Composition of the Upper Mantle: Geophysical Tests of Two Petrological Models

The elastic properties of candidate mantle phases are used to test the viability of olivine-rich (pyrolitic) and CaO + Al,O,-rich (eclogitic) assemblages for the mantle. High temperature adiabats for each phase of interest are constructed and compared to mantle seismic pro. perties. Both pyrolitic and eclogitic assemblages satisfy the seismic properties between 200 and 400 km. Between 400 and 670 km depth an eclogitic assemblage yields a superior match to velocities and velocity gradients. The 400 km seismic discontinuity may represent a chemical boundary between pyrolite and picritic eclogite ("piclogite") or phase transformations in the olivine + orthopyroxene components of a piclogitic assemblage containing about 16% olivine. High velocity gradients in the transition zone may be explained by the transformation of Ca-rich cpx to majorite garnet. Seismic properties at the top of the lower mantle are consistent with pyrolite, piclogite or perovskite, implying that the 670 km discontinuity may be a chemical boundary. Comparisons of laboratory elasticity data with seismic velocity profiles have been used in many studies to constrain the composition of the Earth's mantle. With few exceptions, it has been assumed a priori that the dominant minerals throughout the mantle are olivine (ol) and orthopyroxene (opx), coexisting with a small quantity of garnet (gt) and clinopyroxene (cpx). However, it remains to be demonstrated that ol-rich assemblages provide either a unique or the most satisfactory fit to the seismic data. Anderson (1976, 1979), Liu (1979), Lees et al. (1983), and Jeanloz and Thompson (1983) have discussed the difficulties of explaining the discontinuity at 670 km and the velocities in the transition region on the basis of phase relations in the ol and opx systems. A conclusion of the above studies is that the 670 km discontinuity may represent a chemical boundary. Differences in the Fe and/or Si (opx) content of the upper and lower mantle have been suggested, and Anderson (1979) proposed that the mantle between 220-670 km depth is dominated by gt + cpx (eclogite) rather than ol + opx. It is clearly desirable to test the viability of the gt-cpx hypothesis by direct comparison of the high pressure-high temperature velocities for such an assemblage with mantle velocity and density profiles. Although the stability fields and elastic properties of CaOand Al,O,-rich high-pressure phases are less well constrained than those of ol and opx, sufficient information exists for silicates and analog compounds to examine the plausibility of an eclogitic region. In this paper we calculate the compressional (Vp) and shear (V,) velocities, and density (p) of olivine, pyroxenes, garnets, and their polymorphs, at elevated temperatures and pressures. The properties of assemblages dominated by gt + cpx or ol + opx are then compared to observed mantle velocities and densities. Copyright 1984 by the American Geophysical Union. Paper number 410072. 0094-8276/84/0041-0072$03.00 Data Base Although elasticity data for silicates and oxides has been steadily accumulating, a number of important phases remain uncharacterized. We have therefore relied on systematics to estimate the properties of some high-pressure phases. Data on analog compounds for all pertinent crystal structures are available and they yield consistent elasticity patterns. Bulk modulus (K) and rigidity (~-t) are estimated from mean atomic weight or molar volume systematics and elastic constants of analog compounds. A complete discussion of our data base including Tand P-derivatives will be presented elsewhere (Bass and Anderson, in preparation). The elastic and thermal properties and densities of ol, opx, gt, majorite (mj), (Mg,Fe)O, perovskite (pv), corundum and stishovite have been summarized previously (0. L. Anderson, et. al., 1968; Jeanloz and Thompson, 1983; and references therein). Elasticity data are also available for jadeite (e.g., Hughes and Nishitake, 1963), (3 and 'Y spinels (Weidner et al., 1984; Sawamoto et al., 1984) and diopside (Levien et al., 1979). For some phases only static compression results are available (e.g. mj, MgSiO,-pv) and in these cases measured values of K were used with velocity systematics to define 1-t for a given compound. In the absence of other constraints, both K and 1-t were estimated by velocity systematics. We note that perovskite-structure compounds exhibit well defined elasticity trends (Liebermann et al., 1977). For garnet-structures K ( -170 GPa) and Vp/V, ( -1. 79) are virtually independent of composition and this has been assumed to apply to majorite (K = 220 GPa). Jadeite (jd), diopside (di), and opx, form majoritegarnet type solid solutions (Ringwood and Major, 1966, 1971). Molar volumes of the (hypothetical) majorite endmembers are estimated to be 101.0 (jd) and 122.6 cc/mole (di) by assuming ideal mixing in solid solution. The postmajorite phase of jd-di solutions is thought to have a perovskite structure with end-member volumes of 32.1 (CaSiO,, Ringwood and Major, 1971) and 24.9 cc/mole (NaAlSi,O,, Reid and Ringwood, 1975). Results and Discussion In Figure 1 we show Vp, V, and p for the low and high pressure phases of each component in our model mineral assemblages. The 1400°C adiabats were constructed using third-order Eularian finite strain theory (Sammis et al., 1970). All Fe-Mg silicates contain 10% Fe2+ component, except garnet (20% almandite). Also shown are Earth model PREM (Dziewonski and Anderson, 1981) and other recent profiles (Grand and Heimberger, 1984; Walck 1984; Given and Heimberger, 1980). A common feature is the high velocity gradient between 400 and about 670 km depth. The transition zone gradients are inconsistent with .adiabatic compression of a homogeneous mineral assemblage and a broad phase transition is implied. The depth and breadth of the inferred phase change is an important constraint on the mineralogy in this region.