PUBLICATIONS
Journal of Geophysical Research: Solid Earth
RESEARCH ARTICLE Thermodynamic and elastic properties of pyrope 10.1002/2016JB013026 at high pressure and high temperature Key Points: by first-principles calculations • First-principles calculations of thermodynamic properties and Yi Hu1,2, Zhongqing Wu1,3, Przemyslaw K. Dera2,4, and Craig R. Bina5 elasticity of pyrope are done over a wide pressure and temperature range 1Laboratory of Seismology and Physics of Earth's Interior, School of Earth and Space Sciences, University of Science and • If pyrope can be preserved in cold 2 subducted slab, both low temperature Technology of China, Hefei, China, Department of Geology and Geophysics, School of Ocean and Earth Science and 3 and dense mineral assemblage Technology, University of Hawaii at Manoa, Honolulu, Hawaii, USA, National Geophysical Observatory, Mengcheng, contribute to the negative buoyancy China, 4Hawaii Institute of Geophysics and Planetology, School of Ocean and Earth Science and Technology, University of • Contrast in the seismic velocities Hawaii at Manoa, Honolulu, Hawaii, USA, 5Department of Earth and Planetary Sciences, Northwestern University, of the subducted slab and the Evanston, Illinois, USA surrounding mantle results from the lower temperature
Abstract Pyrope is an Mg end-member garnet, which constitutes up to ~15% of the upper mantle. Supporting Information: Thermodynamic and thermoelastic properties of garnet at high pressure and high temperature are • Supporting Information S1 important for understanding the composition and structure of the Earth. Here we report the first-principles Correspondence to: calculations of vibrational properties, thermodynamic properties, and elasticity of pyrope over a wide Z. Wu, pressure and temperature range. The calculated results exhibit good consistency with the available [email protected] experimental results and provide values up to the pressure and temperature range that are challenging for experiments to achieve and measure. Pyrope is almost isotropic at high pressure. The elastic moduli, Citation: especially the shear modulus of pyrope, exhibit nonlinear pressure and temperature dependences. Density Hu, Y., Z. Wu, P. K. Dera, and C. R. Bina and seismic velocities of pyrope along with piclogite, pyrolite, and eclogite are compared with seismic (2016), Thermodynamic and elastic properties of pyrope at high pressure models along the normal upper mantle and the cold subducting lithospheric plate geotherms. Pyrope has and high temperature by first-principles the highest density and seismic velocities among the major minerals of the upper mantle along these calculations, J. Geophys. Res. Solid Earth, geotherms. Eclogite with 30% pyrope and 70% clinopyroxene is denser than the surrounding mantle even 121, doi:10.1002/2016JB013026. along the normal upper mantle geotherm and becomes still denser along the cold slab geotherm. Seismic Received 29 MAR 2016 velocities of eclogite with 30% pyrope along the normal upper mantle geotherm are slower than the Accepted 21 AUG 2016 surrounding mantle but along the cold slab geotherm are faster than the surrounding mantle. Accepted article online 23 AUG 2016
1. Introduction
Garnet with chemical composition (Mg,Fe)3Al2Si3O12 is an important rock-forming mineral in both Earth's upper mantle and subducted lithospheric plates. According to the pyrolitic compositional model, Earth's upper mantle is composed of ~65% olivine, ~ 20% pyroxene, and ~15% garnet by volume [Ringwood, 1975; Frost, 2008; Ita and Stixrude, 1992]. Upper mantle garnet is mainly made of its Mg end-member pyrope,
with the formula Mg3Al2Si3O12 and a space group Ia3d [Rickwood et al., 1968]. With increasing pressure and temperature, pyrope garnet gradually incorporates clinopyroxene and orthopyroxene and transforms into majoritic garnet near the base of the upper mantle [Ringwood, 1967]. The detailed volume proportions of these minerals are still under debate because of limited knowledge on wave velocities and density profiles of these minerals at temperature and pressure conditions of the upper mantle [Ringwood, 1979; Ita and Stixrude, 1992]. The detailed elastic properties of these minerals are also critical to understanding the aniso- tropic property of the upper mantle, since lattice-preferred orientation and anisotropic elasticity of major mantle minerals are main reasons for the seismic anisotropy observed in the upper mantle [Kawasaki and Kon'no, 1984; Mainprice et al., 2000]. Pyrope is also an important mineral in the subducted lithospheric plate. Eclogite—a rock formed by high- pressure metamorphism of basalt or gabbro at subduction zones—is mainly composed of garnet and clino- pyroxene [Poli, 1993; Ringwood, 1982]. Recent studies show that at temperatures well below the mantle geotherm, pyroxene-garnet transformation is kinetically inhibited at the conditions of subducting slab, so that pyropic garnet may be well preserved down to the transition zone in the subducting slab [Nishi et al., 2008; Van Mierlo et al., 2013; Bina, 2013]. Vibrational and thermodynamic properties (heat capacity, entropy, ©2016. American Geophysical Union. All Rights Reserved. enthalpy etc.) of the minerals are important for determining the stable and metastable regions of these
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 1 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
minerals in the slab. Detailed elastic properties at the conditions of subducting slab are important for comparing petrological models with seismic data. Wave velocities and elastic moduli of pyrope have been extensively studied. However, due to the challenge of measuring elastic properties at simultaneous high pressure and temperature, most of these studies are limited to high pressure but ambient temperature [Chen et al., 1997; Leitner et al., 1980; O'Neill et al., 1991; Sinogeikin and Bass, 2002a] or high temperature but ambient pressure [Sinogeikin and Bass, 2002b]. Ultrasonic measurements in multianvil press provide bulk and shear moduli, instead of individual elastic con- stants at high pressure and temperature [Gwanmesia et al., 2006; Zou et al., 2012a; Chantel et al., 2016]. The Brillouin scattering measurement by Lu et al. [2013] provides detailed elastic constants of iron-bearing pyrope up to 20 GPa and 750 K, which reaches the pressure of the upper mantle but does not reach the tem- perature of the normal upper mantle. Erba et al. [2014] and Mahmoud et al. [2014] investigated the elasticity of pyrope up to 60 GPa at static condition by using first-principles calculations. The vibrational data for pyrope are also limited; only some Raman [Kolesov and Geiger, 1998, 2000; Hofmeister and Chopelas, 1991] and infrared [Hofmeister et al., 1996] data are available. Several studies reported the heat capacity of pyrope [Richet and Fiquet, 1991; Anderson et al., 1991] but are limited to low pressure. The vibrational properties and thermodynamic properties of pyrope at 0 GPa have also been investigated by using a classical potential [Chaplin et al., 1998] or the first-principles calculations with hybrid DFT functional [Baima et al., 2016; Dovesi et al., 2011; Maschio et al., 2013; Pascale et al., 2005]. Here we report the vibrational properties, thermo- dynamic properties, and elasticity of pyrope at high pressure and high temperature by using first-principles calculations. Our results show a good agreement with experimental and calculated data. Nonlinear pressure and temperature dependence of the elastic moduli are observed. Density, compressional, and shear veloci- ties along the normal upper mantle and cold subducting lithospheric plate geotherms are calculated and compared with seismic models.
2. Method 2.1. Computational Details The calculations were performed based on density functional theory (DFT) [Hohenberg and Kohn, 1964; Kohn and Sham, 1965] with local density approximation (LDA) [Ceperley and Alder, 1980] by using the Quantum Espresso package [Giannozzi et al., 2009]. The pseudopotential for magnesium was generated by the method of von Barth and Car [Karki et al., 2000]. The weights of five valence configurations with a cutoff radii 2.5 Bohr, 3s23p0,3s13p1,3s13p0.53d0.5, and 3s13p0.5 3s13d1, are 1.5, 0.6, 0.3, 0.3, and 0.2, respectively. Those for oxygen and silicon were generated by the method of Troullier and Martins [1991] with valence configuration 2s22p4 and cutoff radii 1.45 Bohr for oxygen and valence configuration 3s23p43d0 and cutoff radii 1.45 Bohr for silicon. The pseudopotential for aluminum was generated by the Vanderbilt method [Vanderbilt, 1990] with valence configuration 3s23p1 and cutoff radii 1.77 Bohr. Plane- wave kinetic energy cutoff applied was 70 Ry. Brillouin zone summations over the electronic states were performed at gamma point. Structural optimizations were achieved by using damped variable cell shape molecular dynamics [Wentzcovitch, 1991]. For each optimized structure, dynamical matrix was calculated by using density functional perturbation theory [Baroni et al., 2001] to obtain vibrational frequencies at gamma point. 2.2. Quasi-Harmonic Approximation With static internal energy and phonon frequencies known, Helmholtz free energy at high temperature can be calculated by using the quasi-harmonic approximation (QHA) [Wallace, 1998], that is, X X 1 ℏωq;jðÞe; V FeðÞ¼; V; T U ðÞþe; V ℏω ; ðÞþe; V K T ln 1 exp ; (1) st q;j q j B q;j 2 KBT
where subscripts j represent normal mode index, q is phonon wave vector, T is the temperature and e is strain tensor. The first, second, and third terms are the static internal, zero point, and vibrational energies, respec- tively. We repeated the above-mentioned calculation for seven different volumes as shown in Table S1. The largest volume explored is 757.1109 Å3, which corresponds to the volume at 700 K and 0 GPa, and the smallest volume explored is 666.7611 Å3, which corresponds to the condition of 2000 K and 34 GPa. The calculated Helmholtz free energies versus volume were fitted by the isothermal third-order finite strain
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 2 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
equation of state, and then the pressure and volume relationship are described by the third-order Birch- Murnaghan equation of state as shown below: "#()"# 7 5 2 3 3 3 3K V V 3 ′ V PVðÞ¼ T0 0 0 1 þ K 4 0 1 ; (2) 2 V V 4 T0 V
where P is pressure. V and V0 represent the volume at high pressure and ambient pressure, respectively. KT0 and KT0′ are isothermal bulk modulus and its pressure derivative at 0 GPa [Prewitt and Downs, 1998]. All thermodynamic properties can be derived from the calculated free energies. Although the thermodynamic properties at 0 GPa and above 700 K are obtained by extrapolation, the data at pressure and temperature conditions relevant to the Earth's interior are from interpolation. 2.3. Elastic Constants Isothermal elastic constants are given by [Barron and Klein, 1965] ∂2 T 1 F 1 δ δ δ δ δ δ : cijkl ¼ þ P 2 ij kl il kj ik jl (3) V ∂eij∂ekl 2
Here eij (i, j =1–3) is infinitesimal strains and F is the Helmholtz free energy, which is expressed in the quasi- harmonic approximation as equation (1). Adiabatic elastic constants are obtained by using the relation ∂ ∂ S T T S S δ δ ; cijkl ¼ cijkl þ ij kl (4) VCV ∂eij ∂ekl
where CV, V, and S are heat capacity, volume, and entropy, respectively. In general, calculations of the elastic tensors require vibrational density of states for many strained configurations. The computational workload for vibrational density of states is 2 or 3 orders larger than that for static internal energy. Therefore, the com- putation of the elastic tensors at high temperature and high pressure is considerably expensive. This is espe- cially noticeable for a crystal like pyrope, with 80 atoms in the primitive cell, whose static internal-energy calculations are already expensive. Wu and Wentzcovitch [2011] developed a new method to reduce signifi- cantly the computational workload. Using the relation between volume dependence and strain dependence of vibrational frequencies and introducing an approximation, Wu and Wentzcovitch [2011] deduced an analytical formula for the second derivative of the free energy with respect to strain. The volume dependence of the frequency, ω d q; j γ dV ¼ q; j (5) ωq; j V and the strain dependence of the frequency, ω d q; j γ11 γ22 γ33 ; ¼ q; je11 þ q; je22 þ q; je33 (6) ωq; j
should produce the same frequency change for the same case. Here γq, j is the mode Grüneisen parameter γii θ and q; j is the strain Grüneisen parameter. For each mode, we can introduce the azimuthal parameters q,j and ϕq,j with γ11 e11 þ e22 þ e33 γ 2θ 2ϕ ; q; j ¼ q; jcos q; jsin q; j (7a) e11 γ22 e11 þ e22 þ e33 γ 2θ 2ϕ ; q; j ¼ q; jsin q; jsin q; j (7b) e22 γ33 e11 þ e22 þ e33 γ 2θ : q; j ¼ q; jcos q; j (7c) e33
The distribution function f(θ, ϕ), which defines the numbers of the azimuthal parameters between (θ, ϕ) and (θ +dθ, ϕ +dϕ), is a constant for an isotropic material. By assuming an isotropic distribution function γii for a crystal, we can obtain all kinds of the average values of q; j, required for the calculation of the elasticity,
from the Grüneisen parameter γq, j. Thus, the method developed by Wu and Wentzcovitch [2011] only requires knowledge of volume dependence of vibrational frequencies. The vibrational modes are calculated at seven different volumes, which correspond to the pressure from negative pressure to the pressure of the
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 3 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
Table 1. Frequencies and Symmetries of Raman and Infrared Modes of Pyropea Raman IR
This Study γ Mode Exp1 Exp2 Exp3 Calc1 Calc2 This Study γ Mode Exp4 Cal2 Cal3
141.9 3.3 T2g 133.0 135 - 97.9 106.5 134.0 2.4 T1u 134.4 121.1 114.6 176.0 2.4 - - - 170.1 171.9 145.1 4.2 134.4 139.8 137.7 208.3 2.0 209.8 222 208 203.7 204.2 201.5 1.5 194.6 189.3 185.5 - - - - 230 - - 221.3 1.9 220.8 215.8 213.5 275.3 1.3 285.0 - 272 266.9 268.3 260.0 1.3 258.5 259.9 259.1 - - - - 285 - - 340.0 1.1 336.0 334.2 333.4 316.8 1.1 319.0 322 318 319.0 320.0 346.6 1.3 - 348.6 346.5 353.5 1.3 351.4 353 350 350.6 353.3 379.7 1.0 383.0 282.9 380.2 384.0 1.4 381.2 383 379 381.9 383.8 418.1 0.8 422.0 423.0 421.3 486.1 0.9 490.8 492 490 492.6 495.3 456.4 1.0 455.0 459.0 457.0 502.0 0.9 510.4 512 510 513.5 514.1 470.9 1.0 478.0 483.6 480.9 588.9 1.0 - 598 598 605.9 607.2 540.1 1.1 535.0 532.8 530.9 635.1 1.0 647.5 650 648 655.3 655.2 583.3 1.0 581.0 582.9 580.6 876.7 0.9 867.6 871 866 861.0 860.6 660.6 0.9 650.0 674.1 673.3 909.7 1.0 - 902 899 896.7 895.9 878.8 1.0 871.0 864.8 865.1 1066.2 0.8 1063.1 1066 1062 1068.4 1067.2 909.2 1.0 902.0 895.9 896.2 208.1 1.7 Eg - 211 203 209.2 209.2 979.4 0.9 972.0 969.7 970.5 309.6 1.3 - 284 - 308.5 309.3 342.7 1.2 - 344 342 336.5 336.7 381.4 1.2 - 375 365 376.9 378.9 - - - - 439 - - 515.1 1.0 - 525 524 526.6 530.7 616.7 0.6 - 626 626 636.0 635.8 858.9 0.8 - - - 864.4 863.2 - - - - 911 - - 947.8 0.9 - 945 938 937.4 936.3 355.6 1.6 A1g 363.6 364 362 352.5 356.5 553.5 1.0 561.4 563 562 564.8 567.5 920.0 0.9 926.0 928 925 926.0 923.8 aExp1, Kolesov and Geiger [2000]; Exp2, Kolesov and Geiger [1998]; Exp3, Hofmeister and Chopelas [1991]; Exp4, Hofmeister et al. [1996]. Cal1, Maschio et al. [2013]. Cal2, Pascale et al. [2005]. Cal3, Dovesi et al. [2011].
top of the lower mantle. Then these vibrational frequencies are fitted and interpolated to other volumes to get the volume dependence of vibrational frequencies. Therefore, the method avoids the computation of vibrational density of states for the strained configurations and reduces the computational workload to less than a tenth of the usual approach. The method has been applied successfully to MgO [Wu and Wentzcovitch, 2011], diamond [Núñez-Valdez et al., 2012], olivine and its polymorph [Núñez-Valdez et al.,
2013], ferropericlase [Wu et al., 2013], stishovite, the CaCl2-type silica [Yang and Wu, 2014], and bridgmanite [Shukla et al., 2015]. 2.4. Geophysical Modeling In order to better understand the contribution of pyrope to the bulk upper mantle and subducting litho- spheric plate, a geophysical model is built and compared to seismic observations. The geotherm of the bulk upper mantle is adopted from Brown and Shankland [1981]. The geotherm of the plate top boundary is estimated with TEMSPOL [Negredo et al., 2004] with a lithospheric age of 130 Myr, a dip angle of 30°, subduction rate of 9.1 cm/yr, and a lithospheric thickness of 95 km. These conditions are equivalent to the subducting plate under northeast Japan [Zhao et al., 1997; Peacock and Wang, 1999]. Mineral data including olivine [Isaak, 1992; Zha et al., 1998; Liu and Li, 2006], orthopyroxenes [Chai et al., 1997; Jackson et al., 2003, 2007], and clinopyroxene [Zhao et al., 1998; Isaak et al., 2006; Duffy and Anderson, 1989] are used. Both piclogite and pyrolite models are adopted as representative bulk mantle mineral assemblages. The piclogite model has ~43% olivine, ~ 36% clinopyroxene, ~ 6% orthopyroxene, and ~15% pyrope by volume, and pyrolite has ~63% olivine, ~ 16% clino-pyroxene, ~ 6% ortho-pyroxene, and ~15% pyrope by volume [Ita and Stixrude, 1992]. The uppermost 6 km layer of the subducting plate is modeled by eclogite, a high pres- sure, and high temperature metamorphic product of mid-ocean ridge basalt, with ~30 vol % garnet and
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 4 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
~70 vol % clinopyroxene [Ringwood, 1982; Perrillat et al., 2006]. Density and seismic velocities of the aggregate mineral assemblage are calculated by using a Voigt-Reuss-Hill average [Hill, 1952] and are compared with the preli- minary reference Earth model (PREM) [Dziewonski and Anderson, 1981] and ak135 [Kennett et al., 1995] models.
3. Results 3.1. Vibrational and Thermodynamic Properties The 240 modes of pyrope at the Brillouin zone center for a primitive Figure 1. Volume-pressure-temperature relationship of pyrope. The solid cell with 80 atoms are divided by sym- lines represent our calculation results. Scattered points are from experi- metry as mental data with circles from Zou et al. [2012b], crosses from Zhang et al. Γ 3A 5A 16E [1998], and down-pointing triangles from Milani et al. [2015]. The dashed total ¼ 1g þ 2g þ g lines correspond to the conditions where the validity of QHA may be þ 42T 1g þ 42T 2g þ 5A1u questionable. þ 5A2u þ 20Eu þ 54T 1u þ 48T 2u: (8) Mulliken notation for vibrational modes is used here [Mulliken, 1955, 1956]. A, E, and T refer to singly, doubly,
and triply degenerate states, respectively. Among them, T1u are infrared active modes, and T2g, Eg, and A1g are Raman active modes. The remaining modes are optically inactive. The calculated results using LDA show good consistency with experimental results and previous first-principles calculation results using B3LYP func- tional (Table 1). The relative differences between experimentally observed and our calculated frequencies for fi most of the vibrational modes are less than 3%. Only ve modes differ from observations by more than 3%. γ ∂lnω Mode Grüneisen parameter ¼ ∂lnV is also calculated and listed in Table 1. The pressure-temperature-volume relation of pyrope up to 2000 K and 40 GPa is shown in Figure 1. The volume given by static energy calculation is ~2% smaller than experiments at ambient temperature. Zero
point motion and room temperature effects increase V0 by ~1.6%. The calculated V0 at ambient condition agrees well with experimental data [Zou et al., 2012b; Zhang et al., 1998; Milani et al., 2015] with a difference within 0.5% (Table 2). The good agreements with experiment also hold at higher pressures with the differ- ence within 0.8% at ~33 GPa. The calculated volumes at high temperature are also in good agreement with Zou et al. [2012b], which suggests that the temperature effect on volume and thermal pressure are well included in this calculation. α 1 ∂V Figure 2a shows the thermal expansion ¼ V ∂T P of pyrope at high pressure and temperature. At low temperature the calculations agree well with experimental results [Anderson et al., 1991; Milani et al., 2015; Zou et al., 2012a]. The calculation results start to noticeably deviate the experimental results at ~500 K. The discrepancy increases with increasing the temperature. This pattern is typical for QHA calculations and was also observed by previous studies using the same method [e.g., Wu and Wentzcovitch, 2007; Karki et al., 2000]. QHA, which assumes that phonon frequencies are independent of temperature, omits intrinsic phonon-phonon interaction (anharmonicity). The intrinsic anharmonicity, which increases with increasing the temperature, becomes noticeable at high temperature and results in deviation of the thermal expansion from experiment at high temperature [Wu, 2010; Wu and Wentzcovitch, 2009]. Among thermodynamic prop- erties, thermal expansion is most sensitive to anharmonicity. The effect of anharmonicity at high pressure is less pronounced than at low pressure and can be negligible up to 1500 K at 10 GPa. As shown in Figure 2a, the difference between the calculated and experimental thermal expansion at 10 and 20 GPa is significantly smaller than the one at 0 GPa. ∂E ∂E Heat capacity at constant volume (isochoric) CV ¼ ∂T V and constant pressure (isobaric) CP ¼ ∂T P, which are related by CP = CV(1 + αγthT), are shown in Figures 2b and 2c. CV agrees with experimental [Richet and
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 5 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
Table 2. Thermodynamic Parameters of Pyrope at Ambient Conditions Calculated Experimental 3 a b V0,Å 747.09 750.215(40) , 751.45(15) Thermal expansion, 10 5 K 1 2.724 2.97(45)a, 2.36c 1 1 h d CV, J mol K 324.63, 325.79 323.5 , 1 1 h d CP, J mol K 329.08, 330.70 326.4 Grüneisen parameter 1.68 1.19e, 1.15f Entropy, J mol 1 K 1 264.46, 273.69h 266.27(70)g aZou et al. [2012b]. bZhang et al. [1998]. cAnderson et al. [1991]. dRichet and Fiquet [1991]. eZou et al. [2012b], q = 1.0 fixed. fZou et al. [2012b], q = 1.5 fixed. gHaselton and Westrum [1980]. hBaima et al. [2016].
Fiquet, 1991] and previous DFT calculation data [Baima et al., 2016] and shows a typical QHA pattern, increasing from ~324.63 J/mol K at 300 K to ~487.2 J/mol K at 2000 K, a value very close to the Dulong-
Petit limit (Table 2). CP is calculated to be 329.08 J/mol K at ambient condition (Table 2). At low tempera- ture, the calculated CP is consistent with experimental measurements. Similar to the thermal expansion, the calculated CP begins to deviate from the experimental data at ~1000 K and ambient pressure; the anharmonicity omitted by QHA becomes important above this temperature. On the other hand, the anhar- monicity becomes less visible with increasing pressures. In general, QHA works well along a mantle geotherm [e.g., Karki et al., 2000; Wentzcovitch et al., 2010]. The thermal Grüneisen parameter is an important thermodynamic parameter used to quantify the relation- ship between the thermal and elastic properties of a solid (Figure 2d). The thermal Grüneisen parameter is
(a) (b)
(c) (d)
Figure 2. (a) Thermal expansion, (b) constant volume heat capacity, (c) constant pressure heat capacity, and (d) thermal Grüneisen parameter of pyrope at various temperatures. The solid lines represent the results of this work. The dashed lines along with the shaded area are from Zou et al. [2012b]. The circles are from Anderson et al. [1991]. The plus-sign lines are from Richet and Fiquet [1991]. The dotted lines are from Baima et al. [2016].
HU ET AL. HIGH-PT ELASTICITY OF PYROPE 6 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013026
(a) (b)
Figure 3. (a) Elastic constants of pyrope at various pressures. The solid lines represent the calculated results of this study. The dashed lines represent the calculated static results by Erba et al. [2014]. Scattered points are from experimental results: Sinogeikin and Bass [2000] (right-pointing triangles), Conrad et al. [1999] (circles), Leitner et al. [1980] (rectangles), O'Neill et al. [1991] (downward-pointing triangles), and Chen et al. [1997] (six-pointed stars). (b) Elastic constants of pyrope at various temperatures. The circles are from Sinogeikin and Bass [2002b].