Available online at www.sciencedirect.com

Earth and Planetary Science Letters 263 (2007) 167 – 179 www.elsevier.com/locate/epsl

The role of Al-defects on the equation of state of Al–(Mg,Fe)SiO3 perovskite D. Andrault a,⁎, N. Bolfan-Casanova a , M.A. Bouhifd b , N. Guignot c , T. Kawamoto a,d a

d

Laboratoire Magmas et Volcans, Université Blaise Pascal, Clermont-Ferrand, France b Department of Earth Sciences, Oxford University, Oxford, England c SOLEIL synchrotron, Gif-sur-Yvette, France Institute for Geothermal Sciences, Graduate School of Science, Kyoto University, Beppu, Japan Received 30 May 2007; received in revised form 7 August 2007; accepted 7 August 2007 Available online 24 August 2007 Editor: R.D. van der Hilst

Abstract We performed compression curves of aluminous silicate perovskite (Al–Pv) synthesized under various conditions of pressure, temperature and MgO or SiO2 activities, using laser-heated diamond anvil cell at the ESRF (Grenoble). We refined bulk moduli (K0) from 235 to 270 GPa, in agreement with the wide range of values reported in the literature. We observe that Al–Pv phase synthesized at high temperature, in the SiO2-rich system, is more compressible than Al–Pv phase synthesized at high pressure, in the MgO-rich system. As suggested by various authors, the resolution of this controversy rests on a better understanding of the crystal chemistry of Al in perovskite, which involves at least two competitive mechanisms, substitution of Si in the octahedral site only, or a coupled substitution on both Mg and Si sites. The vacancy mechanism is expected to reduce the K0 significantly, due to the presence of oxygen vacancies. All compression curves performed in this study can be explained by considering that the vacancy mechanism is favored at high temperatures and that the coupled mechanism is favored at high pressures. These trends agree well with previous reports. For (Mg,Fe)(Si,Al)O3 perovskite compositions relevant to the lower mantle, the two previous reports and our new data set for a MORB-type perovskite phase agree well with each other with higher K0 values between 260 and 270 GPa, compared with K0 = 253 GPa for the pure MgSiO3 phase suggested from previous studies. In these compounds, coupled substitution of Al3+ and Fe3+ cations leads to a well constrained crystal chemistry. Therefore, the low K0 value observed in some of the previous studies for Fe-free Al–Pv is likely to be irrelevant for mantle perovskite. Some questions may remain only for the mantle region just below the 670 km discontinuity, where pressures remain moderate, thus potentially allowing for at most 2% of oxygen vacancies. However, it is clear that using low K0 values to extrapolate to greater depths is unjustified. © 2007 Elsevier B.V. All rights reserved. Keywords: lower mantle; silicate perovskite; elastic properties; high pressure

1. Introduction

⁎ Corresponding author. E-mail address: [email protected] (D. Andrault). 0012-821X/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2007.08.012

Magnesium silicate perovskite makes up most of the Earth's lower mantle, comprised between 670 and 2900 km depths (Ito and Takahashi, 1989). The elasticity of this phase thus plays a crucial role in the density of that region and the comparison between laboratory data and

168

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

geophysical data such as density and sound velocity profiles are used to retrieve the composition and temperature of the deep silicate Earth (Wang et al, 1994; Fiquet et al., 1998; Mattern, 2005). Over the past decades, the compressibility of MgSiO3 has been extensively studied and measured values of the bulk modulus, K0, vary between 251 and 262 GPa (see references in Table 1 from Vanpeteghem et al., 2006a). A recent compression study of single crystal MgSiO3 perovskite up to 10 GPa reports a K0 of 253(1) GPa with K′ = 4 and V0 = 162.51(2) Å 3 (Vanpeteghem et al., 2006a,b) in very good agreement with Brillouin and ultrasonic measurements (Sinogeikin et al., 2004; Li and Zhang, 2005). The bulk modulus of polycrystalline MgSiO3 perovskite measured at room temperature up to 94 GPa by Fiquet et al. (2000) is 253(9) GPa with K′ of 3.9(2) and V0 = 162.27(1) Å3 . In comparison, the iron content of perovskite does not seem to affect the compressibility of (Mg,Fe)SiO3 perovskite (Mao et al., 1991) and it was thought since then that the other elements would have no influence as well. Still, this remains questionable in the deepest mantle where iron could be found in the high-spin state (Badro et al, 2004). The equation of state (EoS) of aluminous perovskite has received a lot of attention since Zhang and Weidner (1999) reported that 5 mol% Al2O3 in MgSiO3 decrease the bulk modulus of perovskite down to 234(4) GPa for K′ = 4. Subsequent studies on the same composition, from which Al content is relevant to a pyrolitic mantle, reported values ranging from 251 and 269 for K′ = 4, thus values that display a larger dispersion than for recent MgSiO3 perovskite measurements (see Fig. 1 and Supplementary table summarizing various reports: Anderson et al., 1989; Daniel et al., 2001, 2004; Jackson et al., 2004; Walter et al., 2006). For 10 mol% Al2O3 in perovskite the reported values also vary between 225(1) and 263(7). Note that most of the EoS for aluminous perovskite are reported for K′ = 4, thus the discrepancy does not depend on the choice of EoS (Birch–Murnaghan

Table 1 Chemical composition (wt.%) of glassy starting materials as determined by electron microprobe Al2O3coupled

SiO2 excess

MgO excess

MORBtype

(25 points)

(35 points)

(15 points)

(25 points)

MgO 37.57 (17) 31.06 (1.43) 42.98 (15) FeO SiO2 57.49 (32) 64.11 (1.58) 52.54 (32) 5.39 (07) 5.09 (16) 4.95 (06) Al2O3 Total (wt.%) 100.47 (35) 100.28 (45) 100.50 (30)

18.00 (10) 23.49 (17) 40.59 (26) 16.78 (18) 98.87 (36)

Fig. 1. a) Review of the available previous experimental reports showing (A) a lower or (B) a higher Al–MgSiO3 perovskite bulk modulus (K0) compared to the pure MgSiO3 phase.

second order or third order). The discrepancy between the different experimental results is significant and would represent a substantial change in lower mantle composition according to the recent modeling based on the comparison between seismological profiles and thermo-elastic data sets (Deschamps and Trampert, 2004; Mattern et al., 2005; Samuel et al., 2006). For example, in order to match the seismological speed of sound and density profiles, a perovskite bulk modulus of 235 GPa would require that the lower mantle is composed of aluminous perovskite and stishovite, which is far from the generally accepted pyrolite model (Ringwood, 1970).

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

The actual discrepancies may arise from different sources. It can either be experimental artifacts (such as errors in the pressure measurement or choice of pressure scale, or also peculiar behavior of metastable Albearing perovskite phases recovered at room pressure). Still, the quality and reproducibility of the different experimental works make it impossible to rule out some studies compared to others, as already pointed out by Yagi et al. (2004). It is very likely that the aluminous perovskite samples synthesized by the different groups indeed show significantly different bulk moduli, whatever technique was used to measure the K0. Alternatively, we should evoke an intrinsic complication related to the crystal chemistry of aluminum in perovskite, and thus potentially find a common explanation for the different results that seem a priori incompatible. Understanding this complication is required in order to define a proper bulk modulus for the relevant lower mantle perovskite phase. From a theoretical point of view, the incorporation of Al in perovskite appears to be simple and two reactions are possible: (1) substitution of Al into the Si site charge balanced by the formation of oxygen vacancies: 2 MgO þ Al2 O3 ¼ 2 MgxMg þ 2 Al′Si þ 5 OxO þ V˙ O˙ ð1Þ Or (2) coupled substitution of Mg and Si by Al: : Al2 O3 ¼ AlMg þ Al′Si þ 3 OxO

ð2Þ

where subscripts indicate the site and superscripts indicate the charge (x) for neutral, (′) for negative charge, (·) for a positive charge (in point defect notation from Kröger and Vink (1956)). XAFS (X-ray absorption fine structure) measurements show that the Al K-edge in perovskite cannot be explained by substitution of Al exclusively in one of the cationic sites (Andrault et al., 1998a,b). High resolution 27 Al NMR spectroscopy also evidences occupation of Al in both sites but in addition this technique enables quantification of the amount of Al in the A or B site of the perovskite structure. In the case of perovskite containing 5 mol% Al2O3 coexisting with stishovite, in the MgSiO3–Al2O3 system, Stebbins et al. (2001) find that the ratio of octahedral to dodecahedral Al3+ is 1, indicating that Al enters via a coupled substitution mechanism (reaction 2). Whereas in the case of perovskite containing 2.5 and 5 mol% Al2O3, in the MgSiO3– MgAlO2.5 system, this ratio becomes 2 (Stebbins et al., 2003, 2006), indicating that Al enters in both sites but preferentially into the Si site. The main difference be-

169

tween former and latter studies is the MgO/SiO2 ratio of the starting material and temperature of the synthesis: in the first case, the MgO/SiO2 ratio is equal to 1; in the second case, the MgO/SiO2 ratio is larger than 1, and the system is undersaturated with respect to silica, which is expected to favor the creation of oxygen vacancies via reaction (1) but coupled substitution still occurs. Navrotsky et al. (2003) also showed that in the system MgSiO3–MgAlO2.5 both incorporation mechanisms are acting, and that more aluminum enters into the Si site than the Mg site. Ab-initio calculations indicate that at low pressures the vacancy mechanism is dominant but becomes unfavored as pressure increases (Brodholt, 2000). These structural ab-initio calculations suggest a significant decrease of K0 in case of the vacancy mechanism, and a negligible effect on K0 in the case of the charge-coupled mechanism (Brodholt, 2000). More recently, Yamamoto et al. (2003) used density functional theory (DFT) and found that the charge-coupled mechanism is taking place at all pressures from 0 to 100 GPa and that aluminous perovskite with 6.25 mol% Al2O3 is 3.4% more compressible than the pure MgSiO3 end member. Finally, Akber-Knutson and Bukowinsky (2004) also found that charge-coupled substitution dominates but that still a few percent of oxygen vacancies are formed. Thus, all theoretical calculations point out to charge-coupled mechanism and to a slight reduction of incompressibility of aluminous perovskite compared to pure perovskite. Up to present, the agreement between theoretical predictions and experimental results was not confirmed, however, in particular because both Al-substitution mechanisms were found to coexist in the same sample (Stebbins et al., 2003) with relative ratio possibly evolving as a function of pressure, temperature, and oxide activities. In this work, we report new compression curves of aluminous perovskites determined for varying pressure and temperature conditions of synthesis, and for different chemical compositions of the starting material, for which we varied the MgO/SiO2 ratio. By using an MgOenriched starting material we expect to favor the Ovacancy mechanism. On the contrary, the SiO2-enriched material is expected to enhance the coupled substitution mechanism. We also used an excess-free starting composition: MgSiO3–Al2O3. We finally investigated pyrope and an Al–(Mg,Fe)SiO3 composition relevant to subducted mid-ocean ridge basalts (MORBs). We purposely restrained our investigation of the compression of aluminous perovskite in the stability field of this phase. Our first reason is that this phase is highly metastable outside of its stability field, to the point that it becomes impossible to recover it at room pressure for

170

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

high Al contents (Andrault et al., 2001). Also, even if the Al–Pv phase is finally recovered and appears relatively well crystalline under the diffractometer, it is not clear how much the local structure around the Alatoms is preserved or if it has undergone peculiar modification (like anomalous expansion) during the ramp of decompression. Finally, we cannot anneal the stresses by laser heating outside of the perovskite stability field and therefore the hydrostaticity may not be enough to insure accurate measurements of volume upon decompression. 2. Experimental procedure 2.1. P–V–T techniques Pressures and temperatures up to 60 GPa and 2300 K were achieved in the diamond anvil cell coupled with YAG or CO2 lasers (LH-DAC). We used DAC mounted with 300 μm-culet diamonds, Re gaskets with hole of less than 80 μm diameter, Ar pressure medium loaded cryogenically, in an experimental procedure similar to that used in our previous study (Andrault et al., 2001). Pressure was first estimated from the shift of the fluorescence of a small ruby chip located on the edge of the sample chamber, so as not to risk any contamination of the sample by corundum. However, to achieve a better reliability of the pressure measurement at 300 K we chose to measure pressure as derived from the EoS of the Ar pressure medium (Jephcoat, 1998), after annealing the sample and Ar with the CO2 laser. When heating with the YAG laser, Pt black was mixed to the sample in order to absorb the radiation, and Pt was also used to retrieve pressure from the EoS of Holmes et al. (1989), at ambient and at high temperature during annealing. Temperatures were determined from the sample thermal emission, using reflective lenses to prevent any chromatic aberration (Schultz et al., 2005). Several temperature measurements were performed during annealing and the stability was of better than 50 degrees within 60 s. The temperature uncertainty is estimated to be within 100 K, as we consider the emissivity factor constant for all investigated pressures. Some of the CO2 laser experiments were performed on a separated optical bench without temperature measurement. Our objective was to produce the lowest heating temperature without any risk of strong heating as it sometimes occurs during the procedure of laser alignment. From optical comparison with the numerous other LH-experiments performed in this work, we are convinced that the sample temperature remained significantly below 2000 K in this case.

Unit cell parameters were determined using angle dispersive X-ray diffraction at the ID-30 beamline of the ESRF at Grenoble (Mezouar et al., 2005). A doubly focused monochromatic X-ray beam of 10 × 15 μm size was used at a wavelength of 0.3738 Å. Wavelength and sample to detector distance were calibrated against iodine absorption and Si standard. Data collection duration was of 60 s. Two dimensional images were integrated using Fit2d (Hammersley, 1996). Full diffractogram refinements (see Fig. 2) were performed at all experimental pressures and temperatures using the GSAS code (Larson and Von Dreele, 1988). Errors in volume determination are dominated by the intrinsic experimental error on ID30 of about 0.2%. Therefore, the error in pressure is intrinsically limited to ∼ 0.5% thanks to the high sensitivity of the unit cell volume of Ar to pressure given its low bulk modulus (V0 = 177.01 Å3; K0 = 1.25 GPa and K0' = 7.89; Guignot et al., in prep, an

Fig. 2. Experimental diffraction patterns (crosses) acquired at ambient conditions in the diamond cell after the laser heating at high pressures of starting materials containing excess SiO2 (top) or excess MgO (bottom). The background due to Compton diffusion of the diamonds was removed for clarity. The quality of the diffraction patterns recorded at high pressure is similar, except for the presence of the Ar diffraction lines that shadow some 2Θ angle regions. The associated Rietveld refinements (continuous lines) show that in both cases the Al–MgSiO3 perovskite is largely dominant, still with clear diffraction lines from stishovite, and overlapped peaks for periclase. The hidden regions in bottom picture correspond to Pt diffraction lines.

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

EoS in good agreement with Jephcoat (1998)), and thanks to laser-annealing of deviatoric stresses. In our experiments, we first synthesized the perovskite samples at a given pressure and temperature, before performing the compression curve at higher pressures. The sample synthesis is performed with care, taking time to heat for sufficiently long time, and checking the change in sample crystallinity using X-ray diffraction. Then, the samples are annealed after each compression steps to resolve the stresses built on pressure increase. We kept annealing temperatures voluntarily moderate (below 2000 K) to disable reequilibration of the point defect chemistry of perovskite, aided by the fact that diffusion of Al in perovskite is extremely sluggish (Miyajima et al., 2001). Our results are all compatible with the fact that the annealing time and temperature remain insufficient to re-equilibrate the point defects concentrations after the first step of the sample synthesis, most probably because the grain size is already too large. The evidence for such slow kinetics of chemical reequilibration is found in the case of the low pressure perovskite sample synthesized in the MgOrich system which did not incorporate all of the 5 mol% Al2O3 during its synthesis and subsequently kept its corundum excess on further compression and annealing steps (see below). Pressures higher than 23–24 GPa are required for the synthesis of the mantle perovskite phase. Therefore, the data coverage between room pressure and 23–24 GPa is intrinsically limited. There are two possibilities to tentatively overcome this problem: (1) one is to perform data acquisition during sample decompression. However, we found these measurements not sufficiently accurate, using Ar pressure medium at least, because the sample stresses cannot be annealed in this pressure range; (2) another is to reload the recovered sample into another cell for a new compression analysis up to ∼ 24 GPa. This technique is questionable, however, because elastic measurements performed outside of the stability field of a given mineral may be irrelevant. Also, the Al-bearing MgSiO3 phase is highly unstable at room pressure, and it is not clear if the local structure of atoms is well preserved or if it undergoes unexpected distortion or partial amorphisation. A consequence is that the EoS cannot be well constrained to high degrees of developments, and, as already done in several previous work, we limited the P–V analyses to the so-called second order Birch–Murnaghan EoS: P = P0 + 3/2 K0 [(V0/V) 7/3 −(V0/V)5/3 ] where the two unknown parameters, K0 and V0, are the room-pressure bulk modulus and volume, respectively. Accurate volume measurements performed in the 25 to 60 GPa pres-

171

sure range, and at room pressure, allow determination of K0 and V0 with reasonable accuracy for this type of material. 2.2. Perovskite compositions We first measured the EoS of pure MgSiO3 perovskite, by performing new LH-DAC experiments up to ∼ 60 GPa. The goal is to obtain the K0 of the MgSiO3 reference compound based on the same pressure medium and the same pressure calibrant (argon) as for the other aluminous perovskites studied here. We fix K'0to 4 and derive K0 = 253(4) GPa and V0 = 162.3(2) Å3 . These values are in perfect agreement with previous reports using the same LH-DAC technique (253(9) GPa (Fiquet et al., 2000)), using single crystal X-ray diffraction (253(1) GPa (Vanpeteghem et al., 2006a)), and using other experimental procedures (Kudoh et al., 1987; Funamori and Yagi, 1993; Wang et al., 1994; Utsumi et al., 1995). Therefore, our experimental procedure is validated. We then investigated five different compositions: three of them consist of aluminous MgSiO3 perovskite (5 mol% Al2O3) containing an excess of MgO, or SiO2, or without any excess; the fourth one has the Mg3Al2Si3O12 pyrope composition and the fifth is MORB-type perovskite as retrieved from Hirose et al. (1999). The starting materials were prepared in the form of glasses in a one atmosphere furnace at temperatures up to 1650 °C. The quenched glasses were ground and re-melted several times. However, given the refractory character of some of them, crystallization could occur during quenching. Only the glassy parts were kept for the experiments and the electron microprobe analysis, reported in Table 1. Analytical conditions for electron microprobe were 15 kV and 15 nA with 10 s counting time. Considering that all Al incorporates into the perovskite phase, the composition of MgO-excess can be described as Mg1.000Si0.858(4)Al0.095(1)O3 − x + 2 mol% of MgO; the SiO2-excess starting material as Mg0.795(32) Si1.000Al0103(3)O3 + x + 5 mol% of SiO2; the MgSiO3– Al2O3 stoichiometric starting material as Mg0.934(3) Si0.959(4)Al0.106(2) without any excess of oxides. The MORB-type perovskite starting material composition can be written as Mg0.502(3)Fe0.368(3)Si0.760(3)Al0.370O3. The effect of the excess of MgO or SiO2 should have a negligible effect on the aluminum content of perovskite for two reasons: 1) the excess of periclase and stishovite is small and 2) the solubility of aluminum in these phases is small. Periclase should dissolve at most 1% Al p.f.u., given that there are no other cations such as Na+ or H+ to charge compensate the

172

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

incorporation of Al3+ in MgO (Kesson et al, 1998; Bolfan-Casanova, 2006). Therefore we can neglect the effect of Al loss from perovskite via incorporation in MgO. The incorporation of aluminum into stishovite is a relatively well known phenomenon under hydrous conditions, and it has been shown that in coexistence with aluminous perovskite stishovite incorporates up to 3% Al p.f.u. at 25 GPa and 1750 °C (Bolfan-Casanova et al., 2003). However, it is likely that under anhydrous conditions the uptake of aluminum by stishovite will be lower as there is no H to charge compensate (by the mechanism Si4+ = Al3+ + H+). Indeed, our starting material is a glass chip which should be less affected by hydration than powders. Therefore, we can consider that in the SiO2-rich starting material, loss of alumina through stishovite is not significant. The present bulk Al content is just below the solubility of Al in perovskite at 22.5 GPa and 1600 °C (Kubo and Akaogi, 2000), thus the system should be undersaturated with respect to Al2O3 and the Al content in perovskite should remain constant at all conditions of this study. The only chemical variable as a function of pressure and temperature is the distribution of Al among the two sites of the perovskite structure. The use of MgO or SiO2 excess is to maximize the presence of oxygen vacancies or coupled substitution, respectively.

3. Results 3.1. Effect of the synthesis pressure on the compression behavior of aluminous perovskite in MgO-rich system The effect of the synthesis pressure on the volume of aluminous perovskite at high pressures was investigated using the MgO-rich starting material. Three different compression curves using this material were obtained, by synthesizing the perovskite phase at three different nominal pressures of 22.5, 34.7, and 46.9 GPa, at temperatures significantly below 2000 K (Table 2), at laser power just enough to allow crystallization of the perovskite grains (as controlled from the sharpness of diffraction peaks). Note that the samples experience a pressure slightly higher than the nominal pressure during the laser heating, because of thermal pressure (Andrault et al., 1998b). The compression of these three samples was performed to 36.7, 46.3, and 55.9 GPa, respectively, the samples being laser annealed after each compression step prior to recording the X-ray diffraction pattern like in our previous study (Andrault et al., 2001). The presence of MgO could hardly be detected in the Xray diffraction patterns because of almost complete peak overlap between perovskite and periclase phases. It appears that the volume of aluminous perovskite in the MgO-rich system depends strongly on the pressure

Table 2 Details of the various experimental runs and EoS results

MgSiO3 Ar-medium MgO-excess Mg_22.5 Mg_34.7 Mg_46.9 SiO2-excess Si_1800 Si_2000 Si_2300 Al2O3-excess Al–Pv_1 Al–Pv_2 Al–Pv_3 Pyrope # Pyr_1 Pyr_2 MORB Morb_1 Morb_2

P-range $ (GPa)

T-synthesis

Laser

V0 (Å3)

K0 (GPa)

Data points

28.1–56.5

–

CO2

162.3 (2)

253 (4)

15

22.5–36.7 34.7–46.3 46.9–55.9

b2000 K b2000 K b2000 K

CO2 CO2 CO2

164.0 (3) 164.5 (3) 163.5 (3)

236 (2) 243 (2) 272 (3)

5 6 6

33.6–52.8 32.7 34.4–36.8

1800 K 2000 K 2300 K

CO2 CO2 YAG

162.7 (4) 163.3 163.4 (2)

272 (3) 253 244 (2)

7 3 6

31.5–48.2 25.6–40.8 26.7–55.3

– – –

CO2 YAG YAG

163.3 (2) 165.3 (2) 164.6 (3)

262 (3) 234 (3) 240 (4)

7 8 14

30. 41.6–57.1

– –

CO2 CO2

164.4 165.2 (2.0)

261 236 (25)

3 6

29.8 34.3–58.7

– –

CO2 CO2

167.5 (3) 167.3 (5)

265 (2) 270 (2)

3 10

$ The reported values are nominal pressures, measured at 300 K after the laser heating. Samples experience higher pressures when heated due to the effect of thermal pressure (Andrault et al., 1998). # For pyrope, the Al–Pv phases could not be recovered at room pressure and the reported V0 are those refined by the EoS.

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

173

cussions. The resulting bulk moduli vary with the pressure of synthesis, and we observe that for aluminous perovskite synthesized at intermediate pressure (34.7 GPa) the bulk modulus is lower (243(2) GPa) than that of MgSiO3 perovskite (257(1) GPa), whereas for aluminous perovskite synthesized at the highest pressure in this study (46.9 GPa), the bulk modulus is higher than that of MgSiO3 perovskite with a value of 272(3) GPa (Table 2). 3.2. Effect of the synthesis temperature

Fig. 3. Experimental compression curves obtained for three different runs using the same Mg-excess starting material. The volume of aluminous perovskite increases with the pressure of synthesis for a constant temperature of annealing below 2000 K (Table 2).

of synthesis (Fig. 3). Comparing the volumes of aluminous perovskite with that of MgSiO3 perovskite, as measured in this study in similar pressure range, we observe that incorporation of aluminum increases the volume of perovskite relative to that of MgSiO3, at high pressure, but also that the difference in volume between Al-bearing perovskite and Al-free perovskite increases with increasing pressure of synthesis. The volumes measured at room pressure after decompression also show a substantial increase of the unit cell volume compared to that of MgSiO3 perovskite, and the higher the pressure of synthesis the lower the V0, an exception being the V0 of the lowest pressure sample, which displays an intermediate V 0 . When synthesized at 46.9 GPa, the V0 is 163.5 Å3, whereas when synthesized at 34.7 GPa, the V0 is 164.5 Å3. The sample synthesized at 22.5 GPa has a V0 of 164.0(3) Å3. One possible explanation is likely the too low pressure of synthesis of this perovskite, which may have not incorporated all the aluminum present in the starting material, because aluminum solubility in perovskite decreases with decreasing pressure (Kubo and Akaogi, 2000; Irifune et al., 1996). This is supported by the diffraction pattern of the recovered sample which exhibits clear Al2O3corundum diffraction lines after the synthesis at a nominal pressure of 22.5 GPa, whereas such lines are clearly absent from the diffractograms of the two other samples synthesized at 34.7 and 46.9 GPa. Thus, this sample is not equivalent to the two other higher pressure samples, and we will not consider it in following dis-

The effect of the synthesis temperature on the volume of aluminous perovskite at 30–35 GPa was investigated using the SiO2-enriched starting material. The use of either CO2- or YAG-lasers provided a large temperature range of investigation. Three different syntheses were performed at nominal pressures (measured at 300 K before and after laser heating) of ∼32.7, ∼ 33.6 and ∼ 34.4 GPa. The former experiment was transformed at 32.7 GPa and at temperature of ∼ 2000 K decompressed without further pressure increase. The two latter samples were further compressed and laser-heated step by step to 52.8 and 36.8 GPa, respectively, before decompression to room pressure. The sample investigated in the 33.6– 52.8 GPa pressure range was synthesized and annealed at a moderate temperature of 1800 K, using the CO2 laser. The sample investigated in the 34.4–36.8 GPa pressure range was synthesized and annealed at a much higher temperature of 2300 K, using the YAG laser.

Fig. 4. Experimental compression curves obtained for three different runs performed at similar pressure with varying synthesis temperature and using the same SiO2-excess starting material. The volume of aluminous perovskite is found to decrease with increasing temperature from 1800, 2000, to 2300 K (Table 1).

174

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

We observe a significant difference in the volume of aluminous perovskite as a function of the synthesis temperature (Fig. 4, Table 2). For all samples, the difference in volume between aluminous perovskite after synthesis and that of MgSiO3 perovskite decreases when the temperature of synthesis increases. Also, we observe a decrease of the bulk modulus from 272(3) to 244(2) as temperature of synthesis increases from 1800 to 2300 K, and the ambient-pressure volumes slightly increase from 162.7(4) to 163.4(2) Å3. 3.3. MgSiO3–Al2O3 system We also performed compression curves for samples in the MgSiO3–Al2O3 system, i.e. stoichiometric starting composition (Fig. 5, Table 2). We observe that the volumes recorded at high pressures are all close to each other, whereas significant variation of the room pressure volumes occurs. As pressure of synthesis increases, the V0 decreases whereas the K0 increases. Such correlations are similar to those observed for the MgO-rich sample, and confirm the effect of synthesis pressure on the cell volume of aluminous perovskite. The results obtained for the pyrope, Mg3Al2Si3O12, composition are very similar to that of the Mg0.95Al0.10Si0.95O3 composition. The data points plot precisely in the same P–V regions between ∼ 25 and ∼ 60 GPa. The perovskite synthesized from pyrope, however, could not be recovered at room pressure, because it amorphises

3.4. P–V EoS of Al–(Mg,Fe)SiO3 Pv-phases relevant to the lower mantle

Fig. 5. Experimental compression curves obtained for three different runs performed for the same MgSiO3–Al2O3 composition at varying pressure and temperature conditions of synthesis. The volumes are found equivalent in the 25–55 GPa pressure range, but the V0 appear significantly different.

Finally, we investigated the compression behavior of a MORB-type perovskite, containing large amounts of Al and Fe, i.e. 23.5 wt.% FeO and 16.8 wt.% Al2O3 (Table 1), up to 59 GPa. The two separate runs yield similar results and K0 is refined to ∼ 270 GPa for a V0 of 167.3 Å3 (Fig. 6), even if the samples were synthesized at different pressures. This observation indicates that in the presence of iron, the synthesis conditions have no substantial effect on the elasticity of aluminous perovskite. This is well explained by the fact that aluminum strongly couples to ferric iron (Lauterbach et al., 2000; Nishio-Hamane et al., 2005), and as long as there is enough iron, Al3+ occupying the B site is charge balanced by ferric iron occupying the A site (Vanpeteghem et al., 2006b). Thus, the point defect chemistry in aluminous perovskite is much less variable and is mostly constrained by the relative amount of aluminum and iron. The bulk modulus that we refine is significantly higher than that of pure MgSiO3 perovskite with K0 = 257 GPa. It is comparable with the value of ∼ 265 GPa previously reported for (AlFe)0.10 (MgSi)0.90O3 perovskite (Andrault et al., 2001) and

Fig. 6. Experimental compression curves measured for two separate runs on MORB-type Al-(Mg,Fe)SiO3 perovskite phases. Data for (AlFe)0.05(MgSi)0.95O3 (Andrault et al., 2001) and pyrolitic-type (Ono et al., 2004) perovskite compositions are also shown. Bulk moduli for the natural compositions vary between 260 and 270 GPa, at values higher than that refined for the MgSiO3 perovskite (Table 2).

upon decompression, as is typical for such Al-rich compositions (Andrault et al., 2001). The refined room pressure volume and bulk modulus are therefore determined with less accuracy (Table 2).

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

175

272 GPa for pyrolite-type perovskite (Ono et al., 2004). The good reproducibility, without any noticeable controversy, observed for the different (Al,Fe)-bearing perovskite leaves little doubt for a slightly higher bulk modulus with an increased amount of (Al,Fe). 3.5. Cell distortion of aluminous perovskite A compilation of previous structural data for aluminous perovskite is found in Fig. 3 by Walter et al. (2004) and Fig. 2 of Kojitani et al. (2007). Most previous studies describe the effect of Al content on perovskite cell parameters under ambient conditions in the MgSiO3– Al2O3 system except that of Navrotsky et al. (2003) and Kojitani et al. (2007) who also used MgSiO3–MgAlO2.5 system. The a-axis slightly decreases as a function of increasing Al content. The b-axis is almost constant and the c-axis increases as a function of aluminum content. Concerning distortion angles, the previous reports are compatible with equal trend for pffiffithe ffi evolution of the theta angle, defined as cos H ¼ 2a=c, whereas the phi angle, defined as cos Φ =a/b, shows two different trends that could be related to perovskite samples having variable amount of Al substituted via the vacancy or the coupled mechanism. The cell parameters can be normalized with respect to the pseudo-cubic perovskite lattice (a⁎ ¼ a  pffiffiffi  pffiffiffi 1=3; 1=3 ⁎ V = 2Þ ,b ¼ b  V = 2Þ , and c⁎ ¼ c  ð2V Þ1=3 ) and plotted as a function of the perovskite unit cell volume (Fig. 7). In this diagram, we use V0 = 162.3 Å3 as a reference for the MgSiO3 perovskite volume for all studies. The result is an increase of the cell volume (from 162.5 to 164.8 Å3) as a function of Al content (up to 25 mol% Al2O3) as well as an increase of distortion compared to MgSiO3 perovskite. The advantage of this representation is that it does not require the knowledge of the Al content in the phase, which is often disabled by difficult quantitative analysis in the very small grains synthesized in the LHDAC. In this representation, the different studies appear quite compatible with each other, except for a few trends (that of the b⁎ parameter for study of Irifune et al. (1996) and those for a⁎ and c⁎ for the DAC-samples of Navrotsky et al. (2003)). Generally, the b⁎ parameter is found almost constant, a⁎ slightly decreases and c⁎ increases as a function of the perovskite volume. We also plotted the normalized (a⁎,b⁎,c⁎) cell parameters as a function of cell volume for our ironfree samples that all have the same Al content, 5 mol% Al2O3 (Fig. 7B). We plotted in this diagram room pressure volumes of all Al–Pv samples synthesized during this study, even if we did not perform the compression analysis for some of them (therefore not all

Fig. 7. Correlation between aluminous perovskite volume pffiffifficell  unit 1=3 and the pnormalized cell parameters, a⁎ ¼ a  V = 2 , b⁎ ¼ ffiffiffi1=3 , and c⁎ ¼ c  ð2V Þ1=3 , where a, b, c, and V are the b  V= 2 refined unit cell parameters and volume of the (Z = 4) orthorhombic perovskite lattice. Departure from a value of 1 denotes unit cell distortion compared to the cubic ideal perovskite symmetry. Values are extracted from (A) previous work (Navrotsky et al., 2003; Irifune et al., 1996; Kojitani et al., 2007; Weng et al., 1982; Kubo et al., 2000) and (B) this study. Cell parameters for MORB-type perovskite phase plot at V = 167.4 Å3 and, a⁎ = 0.98, b⁎ = 1.014, and c⁎ = 1.006.

V0 values are reported in Table 2). Values of the room pressure volume of our samples range from 162.7 to 165.3 Å3. For the perovskite samples synthesized in the MgSiO3–Al2O3 system, we observe that the a⁎ and b⁎ parameters decrease with V0, while c⁎ increases. For these samples, the trends are comparable to those retrieved from previous studies, and point out to a moderate increase of the cell distortion with increase of

176

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

the Al–Pv volume (due to a net decrease of a⁎ and increase of c⁎). In contrast, the MgO-rich and SiO2-rich samples display a different behavior, especially for the c⁎ parameters which appears almost constant and a⁎ which increases with increasing V0. If we consider the Mg-excess perovskite compounds only, the sample synthesized at the highest pressure (supposedly with less oxygen vacancies) are found at lower V0. Similarly a trend is observed for the Si-excess perovskite compounds, which show higher V0 for samples synthesized at higher temperatures (supposedly with more oxygen vacancies). Therefore, Al-substitution via the oxygen vacancy mechanism results in a larger increase of the Al–Pv room pressure volume compared to the coupled mechanism. This trend is rather logical because cationic size increases from Si4+, Al3+to Mg2+, and therefore adding more Al in the octahedral site of the perovskite (mainly occupied by Si4+) can easily result in a larger increase of the unit cell volume. What is surprising is that the Mg-excess perovskite synthesized at the highest pressures (minimum of oxygen vacancies) plots in Fig. 7B in the same region as the Si-excess perovskite synthesized at the highest temperatures (maximum of oxygen vacancies). We do not have a definite explanation to this observation. One possibility could be that the varying MgO and SiO2 activities would promote another type of (minor but still substantial) structural defect in the Al-bearing MgSiO3 perovskite. Concerning the MORB-type perovskite compounds, the normalized cell parameters plot at very similar value than the others in Fig. 7B. We do not report the data points in this figure because of a much higher room pressure volume of ∼ 167.4 Å3 (due to the presence of iron). The normalized parameters a⁎ = 0.98, b⁎ = 1.014, and c⁎ = 1.006 are found close to those for the Fe-free perovskite compounds, evidencing that the presence of Fe in aluminous perovskite does not modify drastically the lattice distortion. 4. Discussion 4.1. Fe-free Al-bearing perovskite phases To summarize, we have seen that the properties of iron-free aluminous perovskite are very peculiar for two reasons: (1) the effect of Al incorporation on the cell volume and on the elastic properties of perovskite is generally larger than the effect of Fe incorporation, for similar concentrations (Mao et al., 1991; Kudoh et al., 1992; Weng et al., 1982); (2) the magnitude of the Aleffects is variable and largely dependent on synthesis

conditions, which control the incorporation scheme of Al. We observe that in MgO-excess samples when pressure of synthesis increases the volume of aluminous perovskite increases (relative to that of MgSiO3 perovskite), whereas the V0 decrease (even if remaining larger than that of MgSiO3 end member). This results in an increase of K0. In the SiO2-excess system, we observe that as temperature of synthesis increases the volume of aluminous perovskite at high pressure decrease (but remain larger than that of the pure end member), whereas the V0 slightly increases, resulting in a decrease of K0. We stress that we cannot monitor in-situ the point defect concentrations in the perovskite samples as a function of P and T during the X-ray diffraction experiments. However, we can rely on the conclusions of previous studies in order to retrieve the evolution of the dominant substitution mechanism. (1) Theoretical calculations: coupled substitution is stabilized as pressure increases (Brodholt, 2000; Yamamoto et al., 2003); (2) Calorimetry and theoretical calculations: nonstoichiometric perovskites have a higher configurational entropy, indicating that vacancy formation is favored at high temperatures (Navrotsky et al., 2003; AkberKnutson and Bukowinsky, 2004); (3) NMR spectroscopy: in going from Al2O3–MgSiO3 system to MgOrich system the ratio of coupled to vacancy mechanism increases (Stebbins et al., 2006); (4) crystal chemical relationships between type of substitution and cell dimensions (Supplementary table and Fig. 7). (5) Abinitio calculations suggest a significant decrease of K0 in case of the vacancy mechanism, and a negligible effect on K0 in the case of the charge-coupled mechanism (Brodholt, 2000). The MgO-rich samples are most representative of the oxygen vacancy compound as expected from reaction 1 and as evidenced by NMR (Stebbins et al., 2003, 2006). Theoretical calculations predict that the oxygen vacancy mechanism becomes unfavored as pressure increases, thus the increase of V0 with decreasing pressure in this system implies that oxygen vacancies expand the unit cell of perovskite, which is generally the case in other phases. The SiO2-excess starting material favors full occupancy of the Si site and thus forces the coupled substitution mechanism. In this system, increasing the temperature of synthesis increases the V0 for the same Al content. Navrotsky et al. (2003) report that the vacancy mechanism has a configurational entropy which is ∼ 50% higher than that of the coupled mechanism, whereas the latter mechanism has a much higher enthalpy of formation. These results suggest that at low temperature the coupled substitution is favored whereas at high temperature the formation of vacancies

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

is favored. Calculations by Akber-Knutson and Bukovinsky (2004) confirm that the vacancy mechanism has a higher entropy than the charge-coupled mechanism. Based on these observations, we expect that in our SiO2excess samples the concentration of oxygen vacancies increases from a minimum value to higher values as temperature increases. Thus, like in the MgO-rich samples, the increase in V0 in the SiO2-rich samples can be directly correlated to an increase of oxygen vacancy concentration. Thus, the evolution of room pressure volumes is fully consistent with previous reports on the Al crystal chemistry in iron-free aluminous perovskite. Our results also show that when V0 increases (i.e. the oxygen vacancy mechanism increases, either due to a decrease of synthesis pressure or due to an increase of synthesis temperature) K0 decreases. Our conclusions are very consistent with ab-initio calculation predicting that for the oxygen vacancy mechanism the bulk modulus decreases (Kröger and Vink, 1956). The only discrepancy with previous expectations is that the coupled substitution mechanism substantially increases the bulk modulus (Yamamoto et al., 2003 and this study), whereas theoretical calculations predict a mild effect (Kröger and Vink, 1956). At this point, we cannot derive information for the bulk moduli derivative K0', because we did not measure, purposely, the volumes upon decompression outside the stability field of perovskite. Our results point out to distinct point defects concentrations for the aluminous perovskites synthesized at different nominal pressures. An interesting point is that higher unit cell volumes are measured for samples synthesized at higher pressure (Fig. 3), which appears contradictory with the classical increase of density with increasing pressure. But in this case, volume and density do not follow the same trend, because the Al content is far from being negligible, and therefore one needs to consider that the Al incorporation affects the molar mass. Incorporating 5 mol% of Al2O3 into the Si site (as MgAlO2.5 into MgSiO3) decreases the molar mass of 0.45%, while incorporating the same amount of Al2O3 but as a coupled defect increases the molar mass of 0.08% relative to MgSiO3. Thus, changing the Alsubstitution mechanism can thus modify the molar mass up to 0.54%. At constant density, such a molar mass difference is equivalent to a change in volume of 0.9 Å3 (for a MgSiO3 room pressure volume of ∼ 162 Å3). This volume variation is of the same order of magnitude than that observed between the different compounds (see Fig. 3). Thus this study has shown how the elasticity of perovskite containing 5 mol% Al2O3 can be affected by

177

pressure and temperature conditions of synthesis. Most of discrepancies around the bulk modulus of iron-free aluminous perovskite can be explained in this manner, as all studies used different conditions and styles (multianvil or DAC) to synthesize their perovskite samples, therefore having different defect populations and elastic properties. We also note that our interpretation explains very well the intriguing and apparently contradictory trends reported by Yagi et al. (2004). Thus, it is clear that the low bulk moduli observed in previous study for some aluminous perovskites synthesized at relatively low pressure and supposedly at relatively high temperatures cannot be used to extrapolate density and elastic parameters to greater depth in the mantle (Fig. 8). 4.2. (Fe, Al)-bearing perovskite phases In the presence of both Al and Fe, the variability of elastic properties vanishes (Fig. 6) because coupled substitution is largely dominant (McCammon et al., 1992), and the amount of oxygen vacancies is at most 2% (Lauterbach et al., 2000). It has been shown that the ferric iron content in aluminous perovskite is constrained solely by the amount of Al3+ (McCammon et al., 1992; McCammon, 2005). These results have demonstrated that the Fe3+/ΣFe ratio does not depend on the oxygen fugacity condition. With respect to volume, the effect of

Fig. 8. Bulk modulus profile expected for pure and (Fe–Al)-bearing MgSiO3 perovskite phases at lower mantle pressures. Lower (∼ 240 GPa) and higher (∼ 265 GPa) perovskite bulk modulus correspond to higher and lower concentration of oxygen vacancy, respectively. The low K0 value observed for Fe-free aluminous perovskite may affect the mantle elastic properties in a shallow region below the 660 km discontinuity, because the perovskite phase may contain oxygen vacancies up to a few percents. However, higher K0 values are expected at greater depths, where oxygen vacancies are not stable.

178

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179

the insertion of both Fe and Al is critical on V0, which can be increased by more than 3% compared to the MgSiO3 end member (Table 2). We note that previous reports on Al-(Mg,Fe)SiO3 perovskite with compositions relevant to the pyrolite-type lower mantle (Ono et al., 2004; Andrault et al., 2001) and our new data on MORB-type perovskite agree with each other and indicate a significant increase of K0 with (Al,Fe) contents. The different values fall in the 260–270 GPa range for K′ fixed to 4. Acknowledgments We thank M. Mezouar and E. Schultz for their great contribution to the development of laser-heating experiments at the ID30 beamline at the ESRF, and two anonymous reviewers for their useful comments. This work was supported by CNRS and INSUE. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j. epsl.2007.08.012. References Akber-Knutson, S., Bukowinsky, M.S., 2004. The energetics of aluminum solubility into MgSiO3 perovskite at lower mantle conditions. Earth Planet. Sci. Lett. 220, 317–330. Anderson, O.L., Isaak, D.G., Yamamoto, S., 1989. Anharmonicity and the equation of state for gold. J. Appl. Phys. 65 (4), 1534–1543. Andrault, D., Fiquet, G., Itié, J.P., Richet, P., Gillet, P., Häusermann, D., Hanfland, M., 1998a. Thermal pressure in a laser-heated diamond-anvil cell: an X-ray diffraction study. Eur. J. Mineral. 10, 931–940. Andrault, D., Neuville, D., Flank, A.M., Wang, Y., 1998b. Cation coordination sites in Al–MgSiO3 perovskite. Am. Mineral. 83, 1045–1053. Andrault, D., Bolfan-Casanova, N., Guignot, N., 2001. Equation of state of the lower mantle Al–(Mg,Fe)SiO3 perovskite. Earth Planet. Sci. Lett. 193, 501–508. Badro, J., Rueff, J.P., Vanko, G., Monaco, G., Fiquet, G., Guyot, F., 2004. Electronic transitions in perovskite: possible nonconvecting layers in the lower mantle. Science 305, 383–386. Bolfan-Casanova, N., 2006. Water in the transition zone and lower mantle minerals. Earth's Deep Water Cycle, vol. 168. AGU, pp. 57–68. doi:10.1029/168GM06. Bolfan-Casanova, N., Keppler, H., Rubie, D.C., 2003. Water partitioning at 660 km depth and evidence for very low water solubility in magnesium silicate perovskite. Geophys. Res. Lett. 30. doi:10.1029/2003GRL017182. Brodholt, J.P., 2000. Pressure-induced changes in the compression mechanism of aluminous perovskite in the Earth's mantle. Nature 407, 620–622. Daniel, I., Cardon, H., Fiquet, G., Guyot, F., Mezouar, M., 2001. Equation of state of Al-bearing perovskite to lower mantle pressure conditions. Geophys. Res. Lett. 28, 3789–3792.

Daniel, I., Bass, J.D., Fiquet, G., Cardon, H., Zhang, J., Hanfland, M., 2004. Effect of aluminum on the compressibility of silicate perovskite. Geophys. Res. Lett. 31, L15608. doi:10.1029/2004GL020213. Deschamps, F., Trampert, J., 2004. Towards a lower mantle reference temperature and composition. Earth Planet. Sci. Lett. 222, 161–175. Fiquet, G., Andrault, D., Dewaele, A., Charpin, T., Kunz, M., Häusermann, D., 1998. P–V–T equation of state of MgSiO3 perovskite. Phys. Earth Planet. Inter. 105, 21–31. Fiquet, G., Dewaele, A., Andrault, D., Kunz, M., Le Bihan, T., 2000. Thermoelastic properties and crystal structure of MgSiO3 perovskite at lower mantle pressure and temperature conditions. Geophys. Res. Lett. 27, 21–24. Funamori, N., Yagi, T., 1993. High pressure and high temperature in situ X-ray observation of MgSiO3 perovskite under lower mantle conditions. Geophys. Res. Lett. 20 (5), 387–390. Hammersley, J., 1996. Fit2d User Manual, ESRF. Grenoble, France. Hirose, K., Fei, Y., Ma, Y., Mao, H.K., 1999. The fate of subducted basaltic crust in the Earth's lower mantle. Nature 397, 53–56. Holmes, N.C., Moriarty, J.A., Gathers, G.R., Nellis, W.J., 1989. Equations of state of platinum to 660 GPa (6.6 Mbar). J. Appl. Phys. 66, 2962–2967. Irifune, T., Koizumi, T., Ando, J., 1996. An experimental study of the garnet–perovskite transformation in the system MgSiO3–Mg3Al2Si3O12. Phys. Earth Planet. Inter. 96, 147–157. Ito, E., Takahashi, E., 1989. Post-spinel transformation in the system Mg2SiO4–Fe2SiO4 and some geophysical implications. J. Geophys. Res. 94 (88), 10637–10646. Jackson, J.M., Zhang, J., Bass, J.D., 2004. Sound velocities and elasticity of aluminous MgSiO3 perovskite: implications for aluminum heterogeneity in Earth's lower mantle. Geophys. Res. Lett. 31 (10), L10614.1–L10614.4. Jephcoat, A.P., 1998. Rare-gas solids in the Earth's deep interior. Nature 393, 355–358. Kesson, S.E., Fitz Gerald, J.D., Shelley, J.M., 1998. Mineralogy and dynamics of a pyrolite lower mantle. Nature 393, 252–255. Kojitani, H., Katsura, T., Akaogi, M., 2007. Aluminum substitution mechanisms in perovskite-type MgSiO3: an investigation by Rietveld analysis. Phys. Chem. Miner. 34, 257–267. Kröger, F.A., Vink, H.H., 1956. Relations between the concentrations of imperfections in crystalline solids. Solid State Phys. 307–435. Kubo, A., Akaogi, M., 2000. Post-garnet transitions in the system Mg4Si4O12–Mg3Al2Si3O12 up to 28 GPa: phase relations of garnet, ilmenite and perovskite. Phys. Earth Planet. Inter. 121, 85–102. Kubo, A., Yagi, T., Ono, S., Akaogi, M., 2000. Compressibility of Mg0.9Al0.2Si0.9O3 perovskite. Proceeding of the Japan Academy, vol. 76, pp. 103–107. Kudoh, Y., Ito, E., Takeda, H., 1987. Effect of pressure on the crystal structure of perovskite-type MgSiO3. Phys. Chem. Miner. 14, 350–354. Kudoh, Y., Prewitt, C.T., Finger, L.W., Ito, E., 1992. Ionic radius-bond strength systematics, ionic compressibilities, and an application to (MgFe)SiO3 perovskites. In: Syono, Y., Manghnani, M.H. (Eds.), High-pressure Research: Application to Earth and Planetary Sciences. AGU, Washington, pp. 215–218. Larson, A.C., Von Dreele, R.B., 1988. GSAS Manual. Report LAUR 86–748, Los Alamos National Laboratory. Lauterbach, S., McCammon, C.A., van Aken, P., Langenhorst, F., Seifert, F., 2000. Mössbauer and ELNES spectroscopy of (Mg,Fe) (Si,Al)O3 perovskite: a highly oxidized component of the lower mantle. Contrib. Mineral. Petrol. 138, 17–26.

D. Andrault et al. / Earth and Planetary Science Letters 263 (2007) 167–179 Li, B., Zhang, J., 2005. Pressure and temperature dependence of elastic wave velocity of MgSiO3 perovskite and the composition of the lower mantle. Phys. Earth Planet. Inter. 151, 143–154. Mao, H.K., Hemley, R.J., Fei, Y., Shu, J.F., Chen, L.C., Jephcoat, A.P., Wu, Y., Basset, W.A., 1991. Effect of pressure, temperature and composition on the lattice parameters and density of three (Fe,Mg) SiO3 perovskites up to 30 GPa. J. Geophys. Res. 96 (B5), 8069–8079. Mattern, E., Composition et température dans le manteau profond: interprétations minéralogiques des observations sismologiques, PhD thesis, pp. 231, 2005. Mattern, E., Matas, J., Ricard, Y., Bass, J.D., 2005. Lower mantle composition and temperature from mineral physics and thermodynamic modelling. Geophys. J. Int. 160, 973–990. McCammon, C.A., 2005. Mantle oxidation state and oxygen fugacity: constraints on mantle chemistry, structure, and dynamics. Earth's Deep Mantle: Structure, Composition, and Evolution. Geophysical Monograph, vol. 160, pp. 219–240. McCammon, C.A., Rubie, D.C., Ross II, C.R., Seifert, F., O'Neill, H.S.C., 1992. Mössbauer spectroscopy of 57Fe0.05Mg0.95SiO3 perovskite at 80 and 298 K. Am. Mineral. 77, 894–897. Mezouar, M., Crichton, W.A., Bauchau, S., Thurel, F., Witsch, H., Torrecillas, F., Blattmann, S., Marion, P., Dabin, Y., Chavanne, J., Hignette, O., Morawe, C., Borel, C., 2005. Development of a new state-of-the-art beamline optimized for monochromatic singlecrystal and powder X-ray diffraction under extreme conditions at the ESRF. J. Synchrotron Radiat. 12, 559–664. Miyajima, N., Langenhorst, F., Frost, D.J., Rubie, D.C., 2001. Aluminium diffusion in silicate perovskite and majoritic garnet. Ber. Miner. Ges. 13, 122. Navrotsky, A., Schoenitz, M., Kojitani, H., Xu, H., Zhang, J., Weidner, D.J., Jeanloz, R., 2003. Aluminum in magnesium silicate perovskite: formation, structure, and energetics of magnesiumrich defect solid solutions. J. Geophys. Res. 108 (B7), 2330. Nishio-Hamane, D., Nagai, T., Fujino, K., Seto, Y., Takafuji, N., 2005. Fe3+ and Al solubilities in MgSiO3 perovskite: implication of the Fe3+ AlO3 substitution in MgSiO3 perovskite at the lower mantle condition. Geophys. Res. Lett. 32, L16306. doi:10.1029/2005GL023529. Ono, S., Kikegawa, T., Iizuka, T., 2004. The equation of state of orthorhombic perovskite in a peridotitic mantle composition to 80 GPa: implications for chemical composition of the lower mantle. Phys. Earth Planet. Inter. 145, 9–17. Ringwood, A.E., 1970. Phase transformations and the constitution of the mantle. Phys. Earth Planet. Inter. 3, 109–155. Samuel, H., Farnetani, C.G., Andrault, D., 2006. Chemically denser material in the lowermost mantle: origin and induced seismic velocity anomalies. In: Bass, J.D., van der Hilst, R.D., Trampert, J. (Eds.), Structure, Evolution and Composition of the Earth’s Mantle, Geophysical Monograph, vol. 160. AGU, Washington DC, pp. 101–116.

179

Schultz, E., Mezouar, M., Crichton, W., Bauchau, S., Blattmann, S., Andrault, D., Fiquet, G., Boehler, R., Rambert, N., Sitaud, B., Loubeyre, P., 2005. Double-sided laser heating system for in situ high pressure–high temperature monochromatic X-ray diffraction at the ESRF. High Press. Res. 25 (1), 71–83. Sinogeikin, S.V., Bass, J.D., Zhang, J., 2004. Elasticity of single crystal and polycrystalline MgSiO3 perovskite by Brillouin spectroscopy. Geophys. Res. Lett. 31, L06620. doi:10.1029/2004GL019559. Stebbins, J.F., Kroeker, S., Andrault, D., 2001. The mechanism of solution of aluminium oxide in mantle perovskite (MgSiO3). Geophys. Res. Lett. 28, 615–619. Stebbins, J.F., Kojitani, H., Akaogi, M., Navrotsky, A., 2003. Aluminum substitution in MgSiO3 perovskite: investigation of multiple mechanisms by 27Al NMR. Am. Mineral. 88, 1161–1164. Stebbins, J.F., Du, L.S., Kimberly, K., Kojitani, H., Akaogi, M., Ono, S., 2006. Aluminum substitution in stishovite and MgSiO3 perovskite: high-resolution 27Al NMR. Am. Mineral. 91, 337–343. Utsumi, W., Funamori, N., Yagi, T., Ito, E., Kikegawa, T., Shimomura, O., 1995. Thermal expansivity of MgSiO3 perovskite under high pressures to 20 GPa. Geophys. Res. Lett. 22, 1005–1008. Vanpeteghem, C.B., Zhao, J., Angel, R.J., Ross, N.L., BolfanCasanova, N., 2006a. Crystal structure and equation of state of MgSiO3 perovskite. Geophys. Res. Lett. 33, L03306. Vanpeteghem, C.B., Angel, R.J., Ross, N.L., Jacobsen, S.D., Dobson, D.P., Litasov, K., Ohtani, E., 2006b. Al, Fe substitution in the MgSiO3 perovskite structure: a single-crystal X-ray diffraction study. Phys. Earth Planet. Inter. 155, 96–103. Walter, M.J., Kubo, A., Yoshino, T., Brodholt, J.P., Koga, K.T., Ohishi, Y., 2004. Phase relations and equation-of-state of aluminous Mgsilicate perovskite and implications for the Earth's lower mantle. Earth Planet. Sci. Lett. 222, 501–516. Walter, M.J., Trönnes, R.G., Armstrong, L.S., Oliver, T.L., Caldwell, W.A., Clark, S.M., 2006. Solidus phase relations and perovskite compressibility in the system MgO–AlO1.5–SiO2 with implications for the Earth's lower mantle. Earth Planet. Sci. Lett. 248, 77–89. Wang, Y., Weidner, D.J., Liebermann, R.C., Zhao, Y., 1994. P–V–T equation of state of (Mg,Fe)SiO3 perovskite: constraints on composition of the lower mantle. Phys. Earth Planet. Inter. 83, 13–40. Weng, K., Mao, H.K., Bell, P.M., 1982. Lattice parameters of the perovskite phase in the system MgSiO3–CaSiO3–Al2O3. Year Book Carnegie Instit. Wash. 81, 273–277. Yagi, T., Okabe, K., Nishiyama, N., Kubo, A., Kikegawa, T., 2004. Complicated effects of aluminum on the compressibility of silicate perovskite. Phys. Earth Planet. Inter. 143–144, 81–91. Yamamoto, T., Yuen, D.A., Ebisuzaki, T., 2003. Substitution mechanism of Al ions in MgSiO3 perovskite under high pressure conditions from first-principles calculations. Earth Planet. Sci. Lett. 206, 617–625. Zhang, J., Weidner, D.J., 1999. Thermal equation of state of aluminum-enriched silicate perovskite. Science 284, 782–784.