Physics of the Earth and Planetary Interiors 136 (2003) 67–78

Cationic substitution in MgSiO3 perovskite Denis Andrault∗ Laboratoire des Géomatériaux, IPGP, Université Denis Diderot, 4 Place Jussieu, Paris 75252, France Received 24 September 2001; received in revised form 14 May 2002; accepted 28 August 2002

Abstract Various MgSiO3 -based compositions were used to investigate the possible cationic substitutions in Earth’s lower mantle perovskite at about 25 GPa and 1750 K, using a multi-anvil press apparatus. Recovered samples were characterized using X-ray diffraction and electron microprobe. Inserted in the silicate perovskite structure were found (i) high levels of [FeAl], [ScAl], [GaAl], and [YAl], (ii) Na amounts up to 0.9% per formula units, and (iii) various other elements such as Li, N, P, V, and Cr. Strong correlation between contents of pair-substituted cations suggest a major amount of coupled-substitution in the perovskite lattice. The cell parameter ratios a/c and b/c are found to increase with increasing size of the substituted trivalent cation. At an upper limit, a tetragonal lattice (P4/mcc) is observed for a silicate perovskite with [YAl]0.25 (MgSi)0.75 O3 composition. Using a schematic pyrolitic composition for the lower mantle, the possible integration of all types of cations in Mg- and Ca-bearing silicate perovskite phases is discussed. © 2003 Elsevier Science B.V. All rights reserved. Keywords: Silicate perovskite; High pressure; Substitution mechanism

1. Introduction It is well accepted that the high pressure Al-(Mg, Fe)SiO3 perovskite is the major phase in Earth’s lower mantle. This phase and the Ca-bearing constituent (CaSiO3 ) are probably stable down to the core mantle boundary (Andrault, 2001). These perovskite phases co-exist with (Mg, Fe)O magnesiowustite, or perhaps SiO2 stishovite, depending on whether the lower mantle (Ca + Mg + FeII )/Si ratio is above, or below unity, respectively. Experiments performed at high pressure and temperature on pyrolitic material have shown that CaSiO3 perovskite occurs as an almost pure compound (Irifune, 1994). This effect is probably due to the rigidity of its cubic structure, and also because the ionic size of Ca2+ is larger than most of the other important ∗ Tel.: +33-1-4427-4889; fax: +33-1-4427-2487. E-mail address: [email protected] (D. Andrault).

cations. For MgSiO3 perovskite, adopting the Pbnm symmetry, more possibilities for cation substitutions are expected, because the orthorhombic distortion can accommodate different cation sizes. Another dominant issue is how the perovskite structure can accommodate charge deficiency, or excess, after insertion of X+ or X3+ cations in the A2+ B4+ O3 -based stoichiometry. For the insertion of Al3+ in the (Mg, Fe)SiO3 perovskite, three mechanisms could co-exist (Andrault et al., 1998; Brodholt, 2000; Lauterbach et al., 2000; McCammon, 1997; Richmond and Brodholt, 1998; Stebbins et al., 2001; Wood and Rubie, 1996): (1) insertion of Al in octahedral sites with formation of oxygen vacancies, (2) coupled-substitution of Al on both dodecahedral and octahedral sites, (3) coupled-substitution of Al and Fe3+ , after oxidation of some Fe2+ into Fe3+ . The amount of each substitution mechanism actually occurring in lower mantle perovskite is still under

0031-9201/03/$ – see front matter © 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0031-9201(03)00023-2

68

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

discussion, as it probably varies with pressure, temperature, Al content, and/or oxygen fugacity (fO2 ). In a recent study on Al-(Mg, Fe)SiO3 perovskite samples, synthesized in similar experimental conditions as in this study, it is shown that about 20% of trivalent [FeIII Al] cations produce oxygen vacancies, and the remaining 80% enter the perovskite structure via the couple-substitution mechanism (Lauterbach et al., 2000). Because formation of oxygen vacancies is intrinsically related to a decrease of the perovskite molar density, its importance is likely to decrease with increasing pressure (Brodholt, 2000). Also, even if the FeIII /Fe ratio measured in various perovskite compounds seems to vary little in the actual fO2 ranges of experiments (McCammon, 1997), it is, however, possible that the highly reducing conditions that prevail in the deep interior of the Earth favor a mechanism involving less formation of FeIII . In this paper, the accommodation of various cations in MgSiO3 silicate perovskite is investigated. A major goal is to provide new information to the discussion of the “incompatible” character of minor and trace elements in the Earth’s lower mantle. The method followed in this work is to design MgSiO3 -based chemical compositions that possibly favor the insertion of a couple of elements; the starting materials, for example,

contain additions of [CrAlO3 ] or [(Na0.5 Al0.5 )SiO3 ] to MgSiO3 . Some of these coupled-substitutions ([CrAl] for (MgSi), or [NaAl] for 2Mg) may appear to be impossible. It is a major objective of this study to determine the mechanism that applies for the insertion of different cations in the lower mantle perovskite.

2. Experiments Homogeneous glasses or powder mixtures for compositions reported in Table 1 were finely ground in alcohol using an agate mortar. These starting materials were loaded in Pt or Re capsules of about 1.8 mm long and 1.2 mm diameter. All high pressure syntheses were conducted at the Bayerisches GeoInstitut. The high pressure assembly used tungsten carbide cubes with 4 mm truncations, MgO octahedra with 7 mm edges, cylindrical LaCrO3 furnaces, and WRe3 /WRe25 thermocouples (see Rubie, 1999). Sets of cubes were inserted in a split sphere stage to a pressure of about 25 ± 2 GPa. All runs were conducted at about 1750 ◦ C for more than 2 h, according to thermocouple measurements, or using the electrical power–temperature calibration to estimate the temperature for runs in which thermocouple did not survive compression.

Table 1 Starting compositions and phases recovered after experiments at 25 ± 2 GPa and 1750 ◦ C in the multi-anvil press Experiment

Starting composition

Recovered phases

Starting material

H-703 H-1348 1929 H-1333 H-705 H-936 H-702 H-716 2160 2077 1990 1977 H-955 H-709

[YAl]0.25 (MgSi)0.75 O3 [ScAl]0.125 (MgSi)0.875 O3 [ScAl]0.25 (MgSi)0.75 O3 [GaAl]0.125 (MgSi)0.875 O3 [GaAl]0.25 (MgSi)0.75 O3 [CrAl]0.25 (MgSi)0.75 O3 [FeAl]0.25 (MgSi)0.75 O3 [Fe]0.5 (MgSi)0.75 O3 [Mg2 CrP]0.125 (MgSi)0.75 O3 [NaAlSi2 ]0.125 (MgSi)0.75 O3 [NaN]0.25 (MgSi)0.75 O3 [NaP]0.25 (MgSi)0.75 O3 [LiV]0.25 (MgSi)0.75 O3 [CaSi]0.1 [Al]0.4 (MgSi)0.7 O3

Tetragonal Pv Pv Pv, Gt Pv, traces of Gt Pv, Gt, AG-traces Gt, Spi, Sti-traces Pv, Gt, Mw-traces Pv, Spi, Sti Pv, Spi, Sti-traces Pv, Gt, Sti-traces Pv, quenched melt Pv, Sti, Spi, quenched melt Pv, quenched melt Gt

Glass Glass Glass Glass Glass Cr2 O3 , Al2 O3 Glass Glass CrPO4 Al2 O3 , Na2 Si2 O5 NaNO3 Mg3 (PO4 )2 ·8H2 O, Na2 Si2 O5 Li2 O, V2 O5 Glass

Pv, Gt, Mw, Sti, Spi, and AG stand for perovskite, garnet, magnesiowustite, stishovite, spinel-type structure, and (Al2 O3 –Ga2 O3 )-mixed compound, respectively. All silicate perovskite phases are found compatible with the Pbnm space group, except for sample H-703 (see Table 3). The last column describes the starting material preparation: “Glass” stands for the preparation of a homogeneous glass in a Pt crucible. Other samples are powder mixtures of the listed compounds plus MgO and SiO2 (not listed). Sample 1977 was dehydrated in an oven at 400 ◦ C for 2 h.

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

Capsules recovered from high pressure and temperature were cut longitudinally into two pieces using an 80 ␮m diameter diamond wire. Samples were finely polished to perform X-ray diffraction and electron microprobe analyses. Powder X-ray patterns, using Co K␣ and K␤ radiation, and 0.3 or 0.6 mm collimators, were recorded on a Rigaku diffractometer. Pt or Re lines from capsule material were used to check reliability of the 2θ angle measurement. Samples were rotated relative to the X-ray beam during acquisition to improve reliability of the diffraction peaks intensity. The quality of the patterns is enough to estimate the phase contents in each sample, however the grain size is too large to enable Rietveld analyses. Chemical analyses were performed using a Cameca SX50 electron microprobe. The spatial resolution was of a few microns, a size much smaller than that observed for single perovskite phase domains. Possible errors due to overlap of different phases are thus avoided. Depending on the starting material, contents for N, Na, Mg, Al, Si, P, Ca, V, Cr, Fe, Ga, and/or Y elements were analyzed using BN, albite, enstatite, spinel, orthoclase, apatite, wollastonite, vanadite, Cr2 O3 , iron, AsGa, and Y2 O3 as standards, respectively. At least eight different analyses were used to determine chemical compositions and standard deviations for the end-products (Table 2). Note that the Si-analyses of perovskite phases might be affected by some systematic error, because the Si-standard (orthoclase) shows Si in four-fold coordination in contrast with perovskite that presents Si in octahedra. Some of the perovskite compounds showed amorphisation at the electron beam location after the chemical analysis. For this reason, the electron microprobe current was kept low. Difficulties arise in determining the perovskite crystal chemistry from raw numbers of chemical analyses. For samples containing Sc or Li which contents could not be measured in this study due to practical reasons, Sc and Li contents are estimated from the difference between the ideal value of 2 and the sum of all other cations contents. When doing these calculations, the effect of Sc and Li on the K-factors used to correct for the X-ray fluorescence absorption are neglected. For all other samples, the sum of the cation contents was arbitrarily normalized to 2 in order to retrieve the amount of cations found in the MgSiO3 perovskite. Because of the various types of cations

69

Table 2 Perovskite and garnet compositions as determined by electron microprobe analyses Experiment Perovskite H-703 H-1348 1929 H-1333 H-705 H-702 H-716 2160 2077 1990 1977 H-955 Garnet 1929 H-702 H-705 H-936 2077 H-709

Composition Mg0.710(2) Si0.797(2) [Y0.267(1) Al0.226(1) ]O3±ε Mg0.875 Si0.875 [Sc0.125 Al0.125 ]O3 Mg0.712(4) Si0.773(7) [Sc0.286 Al0.229(5) ]O3±ε Mg0.875 Si0.875 [Ga0.125 Al0.125 ]O3 Mg0.763(2) Si0.761(3) [Ga0.250(2) Al0.226(6) ]O3±ε Mg0.718(7) FeII 0.053 Si0.771(5) [FeIII 0.189(5) Al0.269(4) ] O3±ε Mg0.894(18) FeII 0.044 Si0.938(20) [FeIII 0.123(21) ]O3±ε Mg0.972(5) Si1.009(3) [Cr0.019(3) P0.0006(4) ]O3±ε Mg0.976(5) Si0.994(11) [Na0.007(2) Al0.023(4) ]O3±ε Mg0.980(9) Si1.004(3) [Na0.009(1) N0.007(5) ]O3±ε Mg0.987(1) Si1.007(10) [Na0.006(3) P0.0004(2) ]O3±ε Mg0.912(3) Si0.763(5) [Li0.184 V0.141(1) ]O3±ε Mg0.792 [Sc0.050 Al0.391 ]Si0.767 O3±ε Mg0.689(7) FeII 0.123 [FeIII 0.006(12) Al0.370(4) ]Si0.812(8) O3±ε Mg0.780(8) [Ga0.138(2) Al0.302(4) ]Si0.780(9) O3±ε Mg0.780(3) [Cr0.133(13) Al0.277(8) ]Si0.810(6) O3±ε Mg0.646(15) [Na0.117(26) Al0.332(65) ]Si0.905(50) O3±ε Mg0.7 Ca0.1 [Al0.4 ]Si0.8 O3

Chemical formulae are determined from atomic contents according to the simple rules described in the experimental section. Errors are standard deviations using at least eight different analyses. Elements substituted for (Mg, FeII )SiO3 perovskite are reported in brackets. Chemical compositions of single phase recovered from runs H-1348, H-1333, and H-709 were not measured, and the reported values are those of the starting materials.

found in the perovskite phases, the sum of the cation charges may slightly differ from 6+. Also, Fe adopts FeII and FeIII valences in unknown relative amounts. Here, it is proposed to estimate the FeIII /Fe ratio from microprobe analysis, by constraining the total numbers of occupied octahedral and dodecahedral sites to be equal. For example, element contents measured as (Mg0.718 Si0.771 Al0.269 Fe0.242 ) corresponds to a perovskite compound with chemical composition ( Mg0.718 FeII 0.053 Si0.771 )[ FeIII 0.189 Al0.269 ]O3±ε , in which the bracketed numbers correspond to trivalent cations that are supposed to be equally partitioned between both perovskite sites. For other type of cations, as for (Mg0.646 Si0.905 )[Na0.117 Al0.332 ]O3±ε , the bracketed numbers regroup all cations with a valence different from Mg2+ or Si4+ . Note that this way of writing the perovskite chemical compositions can

70

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

give a distorted view of perovskite crystal chemistry, in particular with a reduced amount of structural defects (for example oxygen vacancies). This problem is discussed in more details below. For the other phases observed in the samples, garnets or spinel-like structures, cation contents were normalized to 8 or 3, respectively, and the same procedure as the one describe above for the perovskite phases was used to derive the chemical compositions.

3. Description of recovered samples From electron microprobe chemical analyses (Fig. 1) and structural refinements using powder X-ray patterns (Fig. 2), it is found that the recovered samples contain either a single phase of perovskite or garnet, or a perovskite in equilibrium with secondary phases. The latter can be garnet, a mixture of stishovite and a spinel-type structure, or a quenched melt (Table 1). Similar type of spinel-type phases than the ones observed in this study were previously reported (Andrault and Bolfan-Casanova, 2001; Fei et al., 1999; Ohtaka et al., 1997; Woodland et al., 1999). Quenched melts are evidenced by the microstructure typical of rapid crystallization. The first important consequence from the fact that perovskite is often observed in contact with another phase(s) is that the chemical activity of each cations is not controlled. Therefore, the thermodynamics of the substitution mechanism cannot be quantified, but the experiments probe instead the accommodation behavior of the silicate perovskite in various chemical conditions. 3.1. Perovskite compounds with X3+ AlO3 substitutions Recovered samples appear to be one perovskite phase for [ScAlO3 ]0.125 , [GaAlO3 ]0.125 , and [YAlO3 ]0.25 additions to MgSiO3 . For one sample containing [GaAlO3 ]0.125 , traces of garnet are also found on the diffraction spectra. Diffraction patterns of [ScAlO3 ]0.125 and [GaAlO3 ]0.125 samples show a perovskite with Pbnm symmetry, as is found for MgSiO3 perovskite. For a sample containing [YAlO3 ]0.25 , the X-ray pattern shows important peaks at 2.857, 2.487 and 2.224 Å, which cannot be interpreted by

Fig. 1. Electron microphotograph of sample H-702 with [FeIII Al]0.25 (MgSi)0.75 O3 starting composition. Below are reported Si, Al, Fe and Mg chemical profiles obtained at the location of the horizontal white line on the image. Three different regions of perovskite, garnet, and magnesiowustite (Mw) are observed. Note the chemical homogeneity of each phase. Profiles show evidence of higher Al or Fe contents in garnet or perovskite phases, respectively. These profiles are only used to compare compositions of co-existing phases. Point analyses are performed for accurate chemical analyses.

the Pbnm symmetry. This set of lines also does not correspond to any MgSiO3 polymorph in this region of the P–T diagram. Instead, all diffraction peaks are explained by the occurrence of a single tetragonal perovskite lattice (Table 3). The presence of 1 0 2, 2 1 0, and 2 1 1, and the absence of 0 1 1, 0 0 3, and 1 1 1 d(h k l) Bragg lines is indicative of the (P4/mcc) space group.

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

71

Fig. 2. Diffraction patterns of samples with [GaAl]0.25 (MgSi)0.75 O3 (a) and [FeIII ]0.5 (MgSi)0.75 O3 (b) starting compositions. Position and intensity of vertical lines correspond to a d(h k l) value, and the intensity of expected diffraction lines for each phase, respectively. (a) A mixture of perovskite and garnet is observed. Small additional peaks, with a major peak located at about 2.60 Å, show evidence for a trace amount of another phase of [Al2 O3 –Ga2 O3 ]-type chemistry. (b) A mixture of perovskite, stishovite and a spinel-type phase. Main spinel peaks are located at 2.53, 2.63, and 2.70 Å.

Table 3 d(h k l) spacing for the Mg0.710 Si0.797 [Y0.267 Al0.226 ]O3 perovskite hkl

d(h k l) (Å)

Intensity

102 1 1 2/0 2 0 210 211 202 212 2 2 0/0 0 4

2.857 2.487 2.225 2.120 2.025 1.876 1.75

11.0 42.5 32.0 11.4 24.5 27.6 36.3

Interplanar spacings (d(h k l)) correspond to a tetragonal unit cell with a = 4.974 and c = 6.981 Å. The space group (P4/mcc) explains h k l-peaks presences and extinctions.

For [FeAlO3 ]0.25 , [ScAlO3 ]0.25 , and [GaAlO3 ]0.25 additions to MgSiO3 , recovered samples contain a mixture of perovskite and garnet indicating that the maximum solubility has been reached for these compositions at this particular pressure and temperature conditions. Chemical analyses show large cation substitution in these perovskite compounds (up to close than 25% on each of the cation sites), but the Al content in the perovskite is significantly lower than 25%pfu. This type of behavior is common in the Al-MgSiO3 system, as majorite garnet is observed to be in equilibrium with silicate perovskite between 23 and 26.5 GPa (Akaogi and Ito, 1999; Irifune et al.,

72

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

1996; Kubo and Akaogi, 2000). For Al-MgSiO3 perovskite, a pressure of 37.5 GPa is required to insert 25%pfu Al in each of the cation sites (corresponding to the pyrope composition; Ito et al., 1998). For samples with [FeAlO3 ]0.25 or [GaAlO3 ]0.25 additions, microprobe analyses also revealed trace amounts of magnesiowustite or a mixed (Al2 O3 –Ga2 O3 ) phase, respectively. Another experiment performed at about 1400 ◦ C (instead of 1750 ◦ C) for [GaAlO3 ]0.25 added to MgSiO3 shows a significantly larger amount of garnet. This observation suggests that more Al can be inserted in the [GaAl]-MgSiO3 perovskite structure at higher temperature. 3.2. Perovskite phases containing Na+ For samples containing [NaAlSi2 O6 ]0.125 added to MgSiO3 , a similar mixture of perovskite and garnet is obtained, with a trace amount of stishovite. Chemical analysis shows a silicate perovskite with Na content of 0.7(2)%pfu. Note a low Al content of 2.4%pfu in this perovskite (Table 2), compared to about 14%pfu Al as previously reported in Al-MgSiO3 perovskite at similar pressure and temperature conditions (Akaogi and Ito, 1999). This observation suggests a strong interaction between Na and Al elements in the perovskite structure, as the presence of Na tends to decrease the Al solubility in the perovskite. For the sample with [NaNO3 ]0.25 additions to MgSiO3 , optical observations show that the sample is composed of two main regions, with one showing characteristic microstructure of a quenched liquid. In the homogeneous region, diffraction patterns mainly show the presence of Pbnm perovskite. A few low intensity additional peaks suggest traces of other phases. Microanalysis indicates a perovskite N content above the microprobe detection limit at 0.7(5)%pfu, for Na at 0.9(1)%pfu. For all analyses on this perovskite compound, a definite correlation between N and Na contents is observed. For the sample with [NaPO3 ]0.25 additions, the Pbnm silicate perovskite is observed, and also additional Bragg lines. The position of these lines and chemical analysis of the sample are both compatible with the presence of some stishovite and of a spineltype compound. The chemical analysis of the latter yields chemical composition Mg1.041 (Si0.022 Na0.378 P0.600 )1.959 O4±ε , consistent with the AB2 O4 stoi-

chiometry. Characteristic microstructure at the capsule border indicates small amount of partial melting. 3.3. Other perovskite samples For the sample with [Mg2 CrPO6 ]0.125 and [Fe2 O3 ]0.5 additions, the recovered samples consist of Pbnm perovskite co-existing with stishovite and spinel-like phases. The chemical compositions of the spinel-like phases yield (Mg0.901 P0.099 )1.980 (Si0.458 Cr0.542 )1.020 O4±ε and (Mg0.585 FeII 0.415 )1 FeIII 2 O4 chemical fomulae. For the sample with [Mg2 CrPO6 ]0.125 addition, the chemical analysis shows a quite low amount of Cr3+ (1.9(3)%pfu) inserted in the perovskite when compared to the amount of Fe usually found. For the sample with [LiVO3 ]0.25 additions, optical observations show the characteristic microstructure of a quenched liquid for a large part of the sample. In another part of the sample, the characteristic diffraction pattern of Pbnm silicate perovskite is observed. Microanalysis indicates the presence of a significant amount of V and Li in the perovskite structure. For this sample, Si content were measured to 0.763(5), which is a much lower value than that measured for Mg (0.912(3)). This suggests a major substitution of V cations on the Si sites. 3.4. Samples free of perovskite For the sample with [CrAlO3 ]0.25 additions, the diffraction pattern shows the presence of mainly a garnet structure, with a trace amount of stishovite, and no Pbnm perovskite. Microanalysis confirms a stoichiometry compatible with the garnet phase. Again, it is noticeable that a similar composition but with Fe3+ instead of Cr3+ (sample H-702) yield a large amount of Fe-rich silicate perovskite. Note, however, that the high pressure behavior of samples with lower amounts of [CrAlO3 ] additions, such as for [AlScO3 ]0.125 and [AlGaO3 ]0.125 was not investigated in this study. It is very likely that in this case Cr can be found inserted in significant amount into the perovskite lattice, as it is observed for sample 2160 with [Mg2 CrPO6 ]0.125 additions. For the sample containing [CaSiO3 ]0.1 [Al2 O3 ]0.2 additions to MgSiO3 , the recovered sample consisted of a homogeneous garnet phase.

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

4. Crystal chemistry of perovskite phases 4.1. Substitution mechanism in [X3+ Al] perovskite phases High degree of (X3+ AlO3 ) substitution is found in MgSiO3 -based perovskite at 25 GPa and 1750 K. For X = Ga, Sc, and Y, the measured total substitution of [X3+ Al] is between 22.9 and 25% on each of the perovskite sites sometimes co-existing with garnet. Pure perovskite compounds were obtained with substitution of 12.5%pfu Sc, 12.5%pfu Ga, or 24.6%pfu Y, indicating that these are minimum values for coupled-substitution in the silicate perovskite. It reveals a relatively large solid solution between MgSiO3 and ScAlO3 (Bass, 1984) or YAlO3 (Ross, 1996) perovskite phases. For a type of starting material, less FeIII (H-702) and much less CrIII (samples H-936 and 2160) were found inserted in the perovskite structure, even if the sizes of Ga, Sc and Y cations bracket those of FeIII and CrIII . This is evidence for a particular behavior of the transition elements in the perovskite lattice. Note also that the amount of Al-substitution in [AlAl]-MgSiO3 compounds was found limited to 7%pfu in each of the polyhedral sites at 26 GPa and 1600 ◦ C (Kubo and Akaogi, 2000), a value significantly lower than the substitution observed for other compounds in this study. This effect could be due to the size of Al that is too small to fit adequately in the dodecahedral site of the perovskite lattice (see discussions below). For samples with X3+ AlO3 (and Al2 O3 and Fe2 O3 ) additions to MgSiO3 , strong predominance of the coupled-substitution mechanism is evidenced by a good correlation between Mg and Si atomic contents (Fig. 3a). The slope is close to 1, which shows evidence for the simultaneous substitution of (MgSi) by (X3+ Al). However, Mg/Si ratio is found slightly below 1. This could be due to a generally higher cation substitution in Mg than in Si sites. Its is an unusual trend for the X3+ substitution in A2+ B4+ O3 perovskite lattices, in which some of the X3+ cations can be inserted in the octahedral site (the Si4+ site) in correlation with the formation of oxygen vacancies (see Lauterbach et al., 2000). This difference could also be explained by a slight overestimate of Si contents in the microprobe measurements. Note that chemical compositions reported for MgSiO3 ,

73

and for samples with [GaAlO3 ]0.125 (H-1333) and [ScAlO3 ]0.125 (H-1348) additions are not determined using the electron microprobe, but instead correspond to the starting material. A strong correlation is found between Al and X3+ (Fe, Ga, Sc or Y) contents in the perovskites (Fig. 3b). This suggests that it is energetically favored to insert Al and X3+ cations together in the structure. The location of each cation in the perovskite structure is probably not random. A simple consideration of the cationic sizes suggests that Al and larger X3+ cations enter octahedral and dodecahedral sites, respectively. This type of consideration can explain the formation of significant amount of FeIII during the synthesis of Al-(Mg, Fe)SiO3 perovskite, if the insertion of [FeIII Al] requires less energy than that of [AlAl]. This type of site segregation is playing a major role in the occurrence of a tetragonal lattice for the sample with [YAlO3 ] additions, as only Y can be responsible for the relative increase of the dodecahedral size that is required for this symmetry change. 4.2. Substitution mechanisms in other perovskite phases In Fig. 3b are also reported results obtained for samples with [NaAl], [NaN], [CrP], or [LiV] additions to MgSiO3 . The total substitution of Na in these samples remains low, and it is therefore difficult to tell if the substitution involves two elements in a coupled mechanism or not. A coupled-substitution could involve [NaAl] and [NaN] substitution for two Mg cations in the dodecahedral sites. This possibility is reinforced by Mg contents that are found significantly lower than those of Si in these two compounds. For sample with [NaAl] additions is measured an Al content of 2.3%pfu, a value significantly lower than the 14%pfu Al reported for an Al-MgSiO3 composition at similar experimental conditions (Kubo and Akaogi, 2000). It thus seems that the integration of some Na in the perovskite lattice significantly reduced the Al-solubility. For the sample with [LiV] additions, a coupledsubstitution could involve a substitution for the couple of (MgSi). This substitution mechanism is compatible with the chemical analyses showing similar Li and V contents. Note, however, that the Li content in this sample was not measured accurately.

74

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

Fig. 3. (a) Correlation between Mg and Si contents for relevant MgSiO3 -based perovskites synthesized in this study. Data on Al perovskite (see [Al, Al]; Kubo and Akaogi, 2000) are also reported. A correlation with a slope close to 1 indicates a major amount of coupled-substitution of trivalent elements on both sites of the perovskite. (b) Correlation between amounts of cations substituted in the perovskite structure. FeIII , Ga, Sc, Y, and Na contents are reported as a function of Al content for the Al perovskite samples. For other perovskite compounds containing amounts of [NaN], [CrP], or [LiV], the content of the second cation is used for the x-axis. A slope close to 1 indicates that the simultaneous insertion of the two different cations is energetically favored.

For the sample with [CrP] addition, a coupledsubstitution in the perovskite lattice could involve a substitution for two Si cations in the octahedral sites. However, the P content is found very low, and it is unlikely that there is a correlation between the two element contents. Instead, it is probable that the 1.9(5)%pfu Cr measured in the perovskite compound is related to a substitution into the octahedral site, that could be correlated with some oxygen vacancies. The fact that this sample contains much less Cr than the

sample with [Fe] addition (H-716) is possibly due to a higher compatibility of Fe with the octahedral site of the silicate perovskite.

5. Effect of cation substitution on the MgSiO3 perovskite lattice For all perovskites synthesized with only low amounts of substituted-cations (samples 2077, 1990,

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78 Table 4 Lattice parameters for perovskite compounds Experiment

Addition

a (Å)

b (Å)

c (Å)

V (Å3 )

MgSiO3 Al0.5 -Pv H-716 H-705 H-702 1929 H-955 H-703 CaSiO3

[Al2 O3 ] [Fe2 O3 ] [GaAlO3 ] [FeAlO3 ] [ScAlO3 ] [LiVO3 ] [YAlO3 ]

4.775 4.778 4.793 4.812 4.822 4.840 4.853 4.974 5.048

4.929 4.938 4.966 4.970 4.978 4.992 4.980 4.974 5.048

6.897 6.943 6.951 6.988 6.979 6.967 6.990 6.981 7.138

162.35 163.81 166.53 167.11 167.54 168.32 168.95 172.74 181.9

Values for MgSiO3 , Al-MgSiO3 , and CaSiO3 perovskite are extracted from previous work (Irifune et al., 1992; Ross and Hazen, 1989; Wang et al., 1996). Cell parameters for the other perovskite phases recovered from runs 2160, 2077, 1977, and 1990 are found to be similar to those of pure MgSiO3 and are not reported here.

1977, and 2160), the cell parameters are found to be similar to those of pure MgSiO3 perovskite within the experimental errors of the X-ray diffraction technique used in this study. For the perovskites found with substantial cation substitution, a significant volume increase is observed for all compositions, with a maximum volume increase of about 6.5% for substitution involving large Y3+ cations. Still, the volume of MgSiO3 -based perovskites remains smaller by more than 5% than that of CaSiO3 perovskite (Table 4). To investigate the possible correlation between perovskite volume and size of substituted cations, let us use cation radii tabulated by Shannon and Prewitt (1969). The procedure is to calculate characteristic volume changes for the insertion of each cation, for examples positive, or negative, for the substitution of Al

75

(R = 0.67 Å) in the Si (R = 0.40 Å) site, or in the Mg (R = 0.86 Å) site, respectively. Then, the perovskite crystal chemistry (Table 2) is used to calculate cationic volume change related to a given amount of substitution in octahedral (Voct ), dodecahedral (Vdod ), and in both (Vtot ) sites of the perovskite (Table 5). This operation requires the assumptions that the cation substitution is limited to coupled-substitution mechanism and, in particular, does not involve the formation a significant amount of oxygen vacancies. This assumption seems reasonable for a majority of the perovskite compounds observed in this study (see discussions concerning Fig. 3). It also assumes that smaller (Al3+ ) and larger cations (for example, Ga3+ ) enter octahedral and dodecahedral sites, respectively. First note that all cations used in this study show larger ionic radii than Si4+ , and only Sc3+ and Y3+ are larger cations than Mg2+ . It results a positive Voct for all, and a negative Vdod for most of the perovskite compounds. From the different perovskite compositions, it is derived that the cation excess volume in octahedra Voct can possibly exceed 75% of the volume of Si4+ , confirming that Si is a very small cation for the octahedral site of an oxide perovskite lattice. The combination of the two effects, positive Voct and a negative Vdod is expected to increase significantly the orthorhombic distortion of the Pbnm perovskite lattice that is directly related to the polyhedral volume ratio. In contrast, a smaller distortion can be expected when Vdod is sufficiently positive, as it clearly observed for the sample with [YAlO3 ] additions that shows a tetragonal perovskite lattice.

Table 5 Volume change in A and B sites due to substitution by trivalent cations in MgSiO3 perovskite Experiment

A site

RX

VX

VDod

B site

RZ

VZ

VOct

VTot

Al-Pv MgSiO3 H-705 H-716 H-702 1929 H-703 CaSiO3

Al0.25 Mg Ga0.250 Fe0.061 Fe0.189 Sc0.286 Y0.267 Ca

0.67 0.86 0.76 0.785 0.785 0.885 1.04 1.14

−1.40 0.0 −0.83 −0.64 −0.64 0.24 2.05 3.54

−0.35 0.0 −0.21 −0.04 −0.12 0.07 0.55 3.54

Al0.25 Si Al0.226 Fe0.061 Al0.269 Al0.229 Al0.226 Si

0.67 0.40 0.67 0.785 0.67 0.67 0.67 0.40

0.99 0.0 0.99 1.76 0.99 0.99 0.99 0.0

0.25 0.0 0.22 0.11 0.27 0.23 0.22 0.0

−0.10 0.0 0.02 0.07 0.15 0.30 0.77 3.54

Volume differences between X3+ and Mg (VX ), and between Z3+ and Si (VZ ) are multiplied by X3+ and Z3+ cationic contents from Table 2, to calculate cation excess volumes due to the cation substitution in Mg (Vdod ), Si (Voct ), and both (Vtot ) sites of the perovskite structure. Atomic radii (RX , RZ ) and volumes are in Å and Å3 , respectively.

76

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

The sum of the cationic volume change (Vtot ) is found positive for all compounds except for the [AlAl]-MgSiO3 perovskite. Still, aluminous perovskites are reported with higher unit cell volumes than pure MgSiO3 perovskite (see Table 4). This effect is probably due to the fact that MgSiO3 perovskite lattice is already very distorted, and the relative increase of the octahedral size (Voct is positive even if Vtot is negative) is not well resolved by an additional increase of the distortion, but rather by a slight increase of the lattice volume. For substitutions of cations larger than Al3+ in the dodecahedral site, cation excess volumes (Voct and Vdod ) better compensate each other to achieve a positive Vtot . The perovskite unit cell volumes are found nicely correlated with Vtot . Other correlations between Vtot and ratio of unit cell parameters indicate that the a/b ratio is not much affected by the substitution, while a/c and b/c are more affected. Larger X3+ cations substituted in the dodecahedral perovskite site produce a relative decrease of the c-axis (Fig. 4). Finally, it is noticeable that a tetragonal lattice is evidenced for the first time in a silicate perovskite compound. Its occurrence is not very surprising as the [YAlO3 ]-rich compound shows a relatively high value of Vdod . The volume of this tetragonal lattice

falls precisely in between those of the orthorhombic and cubic lattices of MgSiO3 and CaSiO3 perovskite, respectively.

6. Pyrolitic composition and lower mantle perovskite composition In order to discuss possible insertion of all minor and trace elements of a pyrolitic composition (Allègre et al., 1995; Ringwood, 1979) in the lower mantle perovskite, cationic size, charge, and electronic structure of the inserted cations must be taken into account. Related to the effects of the electronic structure, the unique information available is a limited compatibility of transition elements. Indeed, the experimental results evidence limited solubility of Sc, V, Fe, and Cr cations, even if their size seems to be compatible with the silicate perovskite structure. In particular, more Y3+ than Sc3+ could be inserted in the lattice, even if the difference in size between Y3+ and Mg2+ is significantly larger. Still, this study and previous reports (Irifune, 1994; Kesson et al., 1995; Malavergne et al., 1997) show that the amounts of Fe, Cr, Mn, Co, Ni, V, and Sc inserted in the silicate perovskite are well above those expected for a pyrolitic composition. As

Fig. 4. Ratio of unit cell parameters of perovskite compounds√as a function of the total cation excess volume (Vtot ) due to substitution in MgSiO3 (see Table 5). a/c and b/c ratio are multiplied by 2 for clarity.

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

magnesiowutite can also accommodate large amount of transition elements, it appears that all transition elements of a pyrolitic composition can be inserted in the main lower mantle phases. Then, it appears clear that X3+ elements can easily be inserted into the [XAl]-MgSiO3 perovskite, even for relatively large X3+ cations (Y3+ : R = 1.04 Å). Because about 4 wt.% Al2 O3 are found in pyrolite, the Al content seems sufficiently large to promote the substitution of all X3+ elements, the most abundant being Cr3+ that is found at levels of about ∼0.4 wt.% Cr2 O3 (corresponding to a Cr content much below that found in the perovskite phase recovered from run 2160 (Table 2)). Note that the simultaneous substitution of various X3+ elements was not probed in this study, but it would be surprising to be limited in this sense due to the low natural abundances of X3+ cations. The question remains about the possible integration of the B3+ cation that is smaller than both Si4+ and Al3+ . According to previous results on Al-MgSiO3 perovskite, a reduced amount of [AlB] coupled-substitution, for which Al and B would be inserted in dodecahedral and octahedral sites, respectively, seems possible. About 0.7±0.2%pfu of Na+ was also found inserted in the perovskite structure, when the starting material contained Al, N, or P. This Na content is significantly higher than the ∼0.4 wt.% Na2 O reported for typical pyrolitic composition. At this point, it is difficult to tell if Al, N and/or P are strictly required to insert Na in the structure, but, in any case, the lower mantle perovskite contains enough Al to achieve the Na/Al ratio observed in this study (as in sample 2077 for example). However, the fact that Na+ contents remain relatively low for the different samples probed in this study is not encouraging for the insertion in MgSiO3 -based perovskites of larger alkali elements, such as K+ . Note that the insertion of such large cations could be facilitated at higher pressures, and probably in the CaSiO3 perovskite that possesses a significantly larger dodecahedral site. Note also that K+ was reported to be easily inserted in another silicate phase found to be stable below the 670 km discontinuity (Miyajima et al., 1999). The compatibility of various B4+ cations with the silicate perovskite structure was not probed in this study. A large solid solution was previously reported between CaSiO3 and CaTiO3 (Kubo et al., 1997), suggesting that all pyrolitic TiO2 content could be integrated in the Ca-bearing perovskite. It is also possible

77

that significant amounts of GeO2 can be integrated into one of the two silicate perovskites at high pressure, but no mixed compounds were yet reported between silicate and germanate perovskite phases. It is also possible that large B4+ cations can integrate the perovskite dodecahedral site (the A2+ site), producing an oxygen vacancy. The possible insertion of C4+ at the Si site was probed in preliminary experiments performed up to ∼100 GPa using a laser-heated diamond anvil cell, they do not reveal any reaction between carbonates and MgSiO3 perovskite (Fiquet et al., unpublished results). In contrast with B4+ cations, it seems unlikely that too large A2+ cations are compatible with the silicate perovskite lattice. Ca seems to be already quite big to be coupled with Si in the silicate perovskite, resulting in a CaSiO3 cubic lattice that is very unstable after quench at room pressure and temperature. Finally, more exotic elements such as Li, N, P, and V were found inserted in the silicate perovskite compounds in various amounts. It is difficult to tell by which mechanism these elements are inserted, but their presence confirms the high compliance of this lattice. Many questions remain about the possible insertion of halogens or S, but it is not the subject of this paper. Still, it can be expected that substitution of these elements on an O site can help resolve charge problems related to the insertion of some cations in octahedral or dodecahedral sites. In summary, these experimental results and analyses of the substitution mechanism suggest that most of cations of a pyrolitic composition could be integrated into lower mantle Ca- or Mg-bearing silicate perovskites. The main elements that can hardly be inserted in silicate perovskite phases are cations bigger than Ca2+ , and most probably cations smaller than Si4+ . However, it is not possible to tell if the lower mantle perovskites actually contain all of the elements that are shown to be compatible with the silicate perovskite structure, because material could have been exposed to strong partitioning of some elements with the formation of any liquid phase at various depths in the Earth’s mantle.

Acknowledgements I thank D. Rubie for his invitations to Bayreuth to perform these experiments, N. Bolfan-Casanova,

78

D. Andrault / Physics of the Earth and Planetary Interiors 136 (2003) 67–78

D. Frost, D. Krausse, D. Neuville, B. Poe, and H. Schultze for great help with experiments, N. Bolfan-Casanova, J.M. Jackson, and two anonymous reviewers for helpful information and comments, and J. Bass for his invitation to participate to this journal special issue. High pressure experiments were performed at the Bayerisches Geoinstitut under the EU “IHP Access to Research Infrastructures” Programme (Contract No. HPRI-1999-CT-00004 to D.C. Rubie). This is a CNRS and IPGP contribution. References Akaogi, M., Ito, E., 1999. Calorimetric study on majorite-perovskite transition in the system Mg4 Si4 O12 –Mg3 Al2 Si3 O12 : transition boundary with positive pressure–temperature slopes. Earth Planet. Sci. Lett. 114, 129–140. Allègre, J.A., Poirier, J.P., Humler, E., Hofmann, A.W., 1995. The chemical composition of the Earth. Earth Planet. Sci. Lett. 134, 515–526. Andrault, D., 2001. Evaluation of (Mg, Fe) partitioning between perovskite and magnesiowustite up to 120 Gpa. J. Geophys. Res. 106, 2079–2087. Andrault, D., Bolfan-Casanova, N., 2001. High pressure phases in MgFe2 SiO4 and Fe2 O3 –MgSiO3 systems. Phys. Chem. Miner. 28, 211–217. Andrault, D., Neuville, D., Flank, A.M., Wang, Y., 1998. Cation coordination sites in Al-MgSiO3 perovskite. Am. Miner. 83, 1045–1053. Bass, J.D., 1984. Elasticity of single-crystal SmAlO3 , GdAlO3 , and ScAlO3 perovskites. Phys. Earth Planet. Inter. 36, 145– 156. Brodholt, J.P., 2000. Pressure-induced changes in the compression mechanism of aluminous perovskite in the Earth’s mantle. Nature 407, 620–622. Fei, Y., Frost, D.J., Mao, H.K., Prewitt, C.T., Häusermann, D., 1999. In situ structure determination of the high-pressure phase of Fe3 O4 . Am. Miner. 84, 203–206. Irifune, T., 1994. Absence of an aluminous phase in the upper part of the Earth’s lower mantle. Nature 370, 131–133. Irifune, T., et al., 1992. A performance test for WC anvils for multianvil apparatus and phase transformation in some aluminous minerals up to 28 GPa. In: Syono, Y., Manghnani, M.H. (Eds.), High-Pressure Research: Application to Earth and Planetary Sciences. AGU, Washington DC, pp. 43–50. Irifune, T., Koizumi, T., Ando, J.I., 1996. An experimental study of the garnet-perovskite transformation in the system MgSiO3 – Mg3 Al2 Si3 O12 . Phys. Earth Planet. Inter. 96, 147–157. Ito, E., Kubo, A., Katsura, T., Akaogi, M., Fujita, T., 1998. Highpressure transformation of pyrope (Mg3 Al2 Si3 O12 ) in a sintered diamond cubic anvil assembly. Geophys. Res. Lett. 25 (6), 821–824.

Kesson, S.E., Fitz Gerald, J.D., Shelley, J.M., Withers, R.L., 1995. Phase relations, structure and crystal chemistry of some aluminous silicate perovskites. Earth Planet. Sci. Lett. 134, 187–201. Kubo, A., Akaogi, M., 2000. Post-garnet transitions in the system Mg4 Si4 O12 –Mg3 Al2 Si3 O12 up to 28 GPa: phase relations of garnet, ilmenite and perovskite. Phys. Earth Planet. Inter. 121, 85–102. Kubo, A., Suzuki, T., Akaogi, M., 1997. High pressure phase equilibria in the system CaTiO3 –CaSiO3 —stability of perovskite solid solutions. Phys. Chem. Miner. 24 (7), 488–494. 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. Miner. Petrol. 138, 17–26. Malavergne, V., Guyot, F., Wang, Y., Martinez, I., 1997. Partitioning of nickel, cobalt and manganese between silicate perovskite and periclase: a test of crystal field theory at high pressure. Earth Planet. Sci. Lett. 146, 499–509. McCammon, C.A., 1997. Perovskite as a possible sink for ferric iron in the lower mantle. Nature 387, 694–696. Miyajima, N., Fujino, K., Funamori, N., Kondo, T., Yagi, T., 1999. Garnet-perovskite transformation under conditions of the Earth’s lower mantle: an analytical transmission electron microscopy study. Phys. Earth Planet. Inter. 116, 117–131. Ohtaka, O., Tobe, H., Yamanaka, T., 1997. Phase equilibria for the Fe2 SiO4 –Fe3 O4 system under high pressure. Phys. Chem. Miner. 24, 555–560. Richmond, N.C., Brodholt, J.P., 1998. Calculated role of aluminum in incorporation of ferric iron into magnesium silicate perovskite. Am. Miner. 83, 947–951. Ringwood, A.E., 1979. Origin of the Earth and Moon. Springer, New York. Ross, N.L., 1996. Distortions in GdFeO3 -type perovskite with pressure: a study of YAlO3 to 5 GPa. Phase Trans. 58, 27–41. Ross, N.L., Hazen, R.M., 1989. Single crystal X-ray diffraction study of MgSiO3 perovskite from 77 to 400 K. Phys. Chem. Miner. 16, 415–420. Rubie, D.C., 1999. Characterising the sample environment in multianvil high-pressure experiments. Phase Trans. 68, 431– 451. Shannon, R.D., Prewitt, C.T., 1969. Effective ionic radii in oxides and fluorides. Acta Crystallogr. B25, 925–946. Stebbins, J.F., Kroeker, S., Andrault, D., 2001. The mechanism of solution of aluminium oxide in mantle perovskite (MgSiO3 ). Geophys. Res. Lett. 28, 4615–4619. Wang, Y., Weidner, D.J., Guyot, F., 1996. Thermal equation of state of CaSiO3 perovskite. J. Geophys. Res. 101, 661–672. Wood, B.J., Rubie, D.C., 1996. The effect of alumina on phase transformations at the 660-kilometer discontinuity from Fe–Mg partitioning experiments. Science 273, 1522–1524. Woodland, A.B., Angel, R.J., Koch, M., Kunz, M., Miletich, R., 1999. Equations of state for Fe2+ 3 Fe2+ 3 Fe3+ 2 Si3 O12 “skiagite” garnet and Fe2 SiO4 –Fe3 O4 spinel solid solutions. J. Geophys. Res. 104, 20049–20058.