GEOPHYSICAL RESEARCH LETTERS, VOL. 11, NO. 7, PAGES 637-640, JULY 1984

MINERALOGY AND COMPOSITION OF THE UPPER MANTLE

Don L. Anderson and Jay D. Bass

Seismological Laboratory, California Institute of Technology, Pasadena CA 91125

Abstract. Seismic velocities are calculated picrites and eclogites are used. We compare the for two petrological models of the upper mantle, results for these two rock types with recent an olivine-rich assemblage, pyrolite, and a seismic velocity distributions for the upper man­ garnet-clinopyroxene rich, olivine bearing, as­ tle beneath several distinct tectonic regions in­ semblage, piclogite. These are compared with re­ cluding shield, rise-tectonic and oceanic areas. cent seismic profiles for various tectonic prov­ inces. The .shield data is most consistent with a Procedure cold olivine and orthopyroxene-rich LID (the seismic lithosphere) extending to 150 km followed The elastic constant data and the mineralogy by a high temperature gradient and/or a change in of pyrolite and piclogite assemblages are summa­ mineralogy that serves to decrease the velocity. rized in Table 1, The P = 0 densities and velo­ From 200 to 400 km the velocities follow a 1400°C cities are calculated at various temperatures as­ adiabat. The rise-tectonic mantle is much slow­ suming a linear temperature dependence of bulk er, presumably hotter and is likely to be above modulus, K, rigidity,~, and thermal expansivity, the solidus to depths of at least 300 km. The a, These high temperature values are then used high Vp/Vs ratio of the lower oceanic lithosphere in the finite strain equations of Sammis et al. in the western Pacific is most consistent with [1970] to calculate V ,V ,p,V /V and(= K/P = eclogite. V~ - (4/3)V~ along v~ri~us agiatats. Figure 1 g~ves the results for compressional velocity, Vp, Introduction and shear velocity, Vs for the olivine-rich (pyrolite), and garnet-clinopyroxene rich Large lateral variations in seismic velocities (piclogite) assemblages. We assume that the min­ and the effects of high temperature gradients, eralogy does not change with depth or temperature. anisotropy and partial melting have made it dif­ The effects of pyroxene-garnet reactions will be ficult to infer the composition and mineralogy of discussed later. the upper 400 km of the mantle. The important At depths shallower than 200 km, V is lower minerals in this region of the mantle are un­ for pyrolite but pyrolite and piclogi~e have doubtedly olivine, orthopyroxene, clinopyroxene similar Vs' At greater depth the compressional and garnet and the dominant rock types are pre­ velocities converge but pyrolite Vs is greater sumably harzburgites, lherzolites, pyroxenites than the predicted Vs for piclogite. The dif­ and eclogites. The composition and mineralogy ference in trend between the two assemblages is may change both vertically and laterally. At caused mainly by the different bulk moduli and depths between 200 km and the 400 km discontinu­ pressure derivatives. These curves suggest that ity the average velocities are generally consis­ v is the better mineralogical discriminant above tent with either peridotite or eclogite while 28 0 km, whereas Vs is more strongly affected by those between 400 and 670 km are more consistent temperature and is insensitive to mineralogy. with an olivine eclogite, or piclogite, assem­ These roles are reversed below 200 km. blage [Bass and Anderson, 1984]. The large The parameter Vp/Vs, related to Poisson's lateral variations in velocity above 400 km ratio, is a very weak function of temperature. [Nataf et al., 1984; Anderson, 1984] may be due Figure 2 shows Vp/Vs which, for the parameters to differences in petrology, temperature or crys­ chosen, is nearly independent of temperature for tal orientation. piclogite and weakly dependent on temperature In this paper we compute seismic velocities as for pyrolite. Vp/Vs is, therefore, a good min­ a function of temperature and pressure for two eralogical discr~minant. It is less than 1.71 representative mineral assemblages using the for forsterite, jadeite and orthopyroxene and elastic constant data summarized by Bass and greater than 1.76 for clinopyroxene and garnet. Anderson [1984 and paper in preparation]. Eclogites and piclogite therefore have relative­ Pyrolite is a hypothetical olivine-rich rock ly high Vp/Vs compared to peridotites. which can yield tholeiitic magmas by about 25% The se~smic parameter (=K/P) depends both partial melting [Ringwood, 1975]. We define on temperature and composition but because it is piclogite as an olivine eclogite, the high pres­ independent of the rigidity it should be rela­ sure form of picrite. It is a clinopyroxene­ tively insensitive to high temperature effects garnet rich, olivine-bearing rock which may be such as partial melting and dislocation relaxa­ complementary to or derivative from more ultra­ tion. Both K and P decrease slightly upon par­ basic assemblages. It is not meant to refer to a tial melting. specific mineralogy or composition but is used in Figure 1 also shows an estimate of the solidus the same general sense as the terms basalts, temperature of the mantle [Wyllie, 1984]. The 1400° and 1800°C adiabats, for example, are in the partial melt field at depths shallower than Copyright 1984 by the American Geophysical Union. about 100 and 300 km, respectively. Partial melting reduces the velocity from the values cal­ Paper number 4L6171. culated [Anderson and Sammis 1970, Anderson and 0094-8276/84/0041-6171$03.00 Spetzler, 1970]. Dislocation relaxation effects

63'7 638 Anderson and Bass: Mineralogy and Composition of the Upper Mantle

TABLE 1. Composition and Properties of the northern part of the East Pacific Rise, Gulf Two Mineral Assemblages of California and southern California is drama­ tically different. The temperatures implied for the upper 300 km are above the solidus and part­ Pyrolite Piclogite ial melting is suggested. Vp/Vs is also higher than appropriate for any plausible mineral or mineral assemblage. Vp/Vs is particularly high Olivine (wt.pct.) 57 16 1. 7 3 between 75 and 200 km, suggesting that melting is a maximum in this depth interval. Although Orthopyroxene (wt.pct.) 17 6 1.67 Clinopyroxene (wt.pct.) 12 33 1.74 the velocities are low between 200 and about Garnet (wt.pct.) 14 45 1.80 400 km, the velocity gradient is high. The high P(g/cm3 ) 3.379 3.497 velocity gradient implies a negative temperature gradient, chemical inhomogeneity, mineral and/or Vp(Km/sec) 8.306 8.445 melt reactions or a subsolidus relaxation mecha­ Vs(km/sec) 4.801 4.800 dK/dP 5.3 4.8 nism that is suppressed by pressure. Both d]J/dP 1.9 1.7 pyrolite and piclogite satisfy Vp near 400 km dK/dt(kb/°C) -0.20 -0.22 for the 1400°C adiabat. The average temperature difference between the seismic models appears to d]J/dt&kb/°C) -0.12 -0.12 be about 400°C or less between 300 and 400 km. a(10- j°C~ 2.6 2.6 da/dt(lO- /°C2 ) 1.3 1.3 The inferred temperature for the oceanic pro­ l. 7 30 l. 7 59 files is above the solidus for most of the mantle Vp/Vs above 300 km. The melting curve, of course, is not well known below depths of about 100 km and is an extrapolation from measurements at low pressure. Velocities are still low below 300 km reduce the shear modulus at somewhat lower tem­ and the Vp/Vs ratio is high for the ridge­ peratures [Minster_and Anderson, 1981]. tectonic profile. If the solidus curve is some 300°C lower than shown at 400 km then the low Seismic Models velocities and high gradients between 300 and 400 km could be due to a decreasing amount of Compressional and shear velocities for shields melt with depth. In some tectonic regions high [Given and Heimberger 1980, Grand and Heimberger electrical conductivity occurs at these depths 1984a] are shown in Figure 1. Unfortunately, the and partial melting seems to be required [e.g. Vp and V data are not for the same regions. The Roberts, 1983]. 8 velocities are relatively high to depths of about 150 km and then drop rapidly. The high velocity layer is sometimes called the seismic lithosphere or LID. Below 200 km the velocities fall near the 1400°C adiabat. The trend of Vs between 200 and 400 km parallels the trends of the piclogite adiabats but both pyrolite and piclogite are gen­ erally compatible with the data, particularly con­ sidering the uncertainties in the pressure deri­ ivatives and the possibility of a superadiabatic gradient in this region. The rapid drop in velocity between 140 and 170 km requires a high temperature gradient or a decrease in the olivine content and an increase in the pyroxene content of the mantle. The Vp/Vs ratio for the shield data (Figure 2), is very low in the upper 150 km. Increasing the olivine and orthopyroxene content or decreasing the FeO content of the shield LID, relative to pyrolite, would decrease Vp/Vs. In­ creasing orthopyroxene, however, would decrease V • Therefore, the seismic properties of the sRield lithosphere are best explained by a very olivine-rich, possibly low FeO, peridotite with low temperatures and a low temperature gradient. For a uniform harzburgite composition LID the inferred temperature gradient is only about 2-4°/km, substantially less than the average Fig. 1. Compressional and shear velocities for values of 8-10°/km expected for a conductive geo­ pyrolite and piclogite along various adiabats. therm over the same depth interval. The subse­ The temperatures are for P = 0. The solidus quent rapid drop in velocity, if due to tempera­ curves are from Wyllie [1984]. The portions of ture alone, requires a gradient of at least 10 to the adiabats below the solidus curves are in the 14°C/km. The large increase in Vp/Vs with depth partial melt field. The seismic profiles are for suggests that a change in mineralogy is also in­ shields [Given and Heimberger, 1981; Walck, volved, i.e. an increase in garnet and clinopy­ 1984], tectonic-rise [Grand and Heimberger, roxene relative to olivine and orthopyroxene. 1984a, Walck, 1984], and the North Atlantic The tectonic-rise model, [Grand and [Grand and Heimberger, 1984b] regions. Pacific Heimberger, 1984a; Walck, 1984] appropriate for ocean da.ta are from Shimamura et al. [ 1977]. Anderson and Bass: Mineralogy and Composition of the Upper Mantle 639

The Vs structure for the northwest Atlantic [Grand and Helmberger, 1984b] lies between the shield and rise models. Partial melting or some other high-temperature relaxation mechanism is implied for the region between 130 and 300 km. The mantle beneath the northwest Atlantic is sim­ ilar to the rise mantle between 250 and 400 km. The inferred high olivine content for shield lithosphere agrees with evidence from kimberlite xenoliths. However, an olivine-rich lithosphere is not necessarily appropriate for other regions. Although data is extremely limited, the high V /V and

transition zone are consistent with the follow­ olivine-rich assemblage is supported on the basis ing scenerio. Material in the transition zone of velocity, anisotropy and direct sampling for becomes buoyant because of high temperature, the shield lithosphere and the immediately partial melting or because it is displaced by sub-Moho oceanic layer. Olivine is buoyant rela­ subducted material, and rises into the shallow tive to undepleted, garnet-rich lherzolite or mantle. Melting is most extensive under ridges eclogite and can therefore be expected to prefer­ at about 100 km depth. Crystal fractionation entially concentrate in the shallow mantle. Den­ yields picritic or tholeiitic melts which emerge sity increases in the order basalt, peridotite, at midocean ridges and underplate the oceanic eclogite which is the sequence suggested above. lithosphere, leaving behind olivine, clinopyrox- The high velocity gradients in the oceanic mantle 640 Anderson and Bass: Mineralogy and Composition of the Upper Mantle

below a depth of 100 to 200 km may be due to de­ beneath southwest Japan : The subducting creasing amounts of melt with depth. Although Philippine Sea Plate, Tectonophysics, ~. 1- there may be peridotite-rich regions between 100 44, 1981. and 400 km under oceans and between 200 and 400 Ito, K., Petrological models of the oceanic km under shields the present results show that lithosphere: geophysical and geochemical the seismic data do not demand such an interpre­ tests, Earth. Planet. Sci. Lett., 21, 169-180, tation. The mineralogy of the deep lithosphere 1974. - and underlying mantle may ultimately be deduced ~1inster, B. and Don L. Anderson, A model of on the basis of the Vp/Vs ratio. dislocation~controlled rheology for the mantle, Phil. Trans. Roy. Lond. A, 299, 319- Acknowledgments. We thank the reviewers of 356, 1981. this paper for their comments. This research Nata£, H. -C., I. Nakanishi and D. L. Anderson, was supported by the National Science Foundation Anisotropy and shear-velocity heterogeneities under Grant No. EAR811-5236. Contribution number in the upper mantle, Geophys. Res. Lett., _!_!_, 4078, Division of Geological and Planetary 109-112, 1984. Sciences, California Institute of Technology, Ringwood, A. E., Composition and petrology of Pasadena, California 91125. the Earth's mantle, McGraw-Hill, New York, 618 pp., 1975. :c,~ferences Roberts, R. G., Electromagnetic ~vidence for lateral inhomogeneities within the Earth's Anderson, Don L., Composition of the mantle and upper mantle, Phys. Earth Planet. Interiors, core, Ann. Rev. Earth Planet. Sci., 2• 179- 33, 198-212, 1983. 202, 1977. Sammis, C., Don L. Anderson and T. Jordan, Anderson, Don L., The Earth as a planet: para­ Application of finite strain theory to ultra­ digms and paradoxes, Science, 223, 347-355 sonic and seismological data, J. Geophys. 1984. Res., 75, 4478-4480, 1970. Anderson, Don_L. and C. Sammis, Partial melting Shimamura,~., T. Asada and M. Kumazawa, High in the upper mantle, Phys. Earth Planet. shear velocity layer in the upper mantle of Inter.,~. 41-50, 1970. the western Pacific, ~~ure, 269, 680-682, Anderson, Don L. and H. Spetzler, Partial melting 1977. and the low-velocity zone, Phys. Earth Plant. Sumino, Y. and 0. Nishizawa, Temperature Int., ~. 62-64, 1970. variation of elastic constants of Bass, Jay D. and Don L. Anderson, Composition of pyrope-almandine garnets, J. Phys. Earth, the upper mantle: geophysical tests of two ~. 239-252, 1978. petrological models, Geophys. Res. Lett.,11, Walck, M. C., The P-wave upper mantle structure No. 3, 237-240, 1984. - beneath an active spreading center: the Gulf Given, J. and D. V. Heimberger, Upper mantle of California, Geophys. J. R. astr. Soc., structure of northwestern Eurasia, J. !...!!._, 697-723, 1984. Geophys. Res., 85, 7183-7194, 1981. Wyllie, P. J., Constraints imposed by experi­ Grand, S. and D. V~Helmberger, Upper mantle mental petrology on possible and impossible shear structure of North America, Geophys. magma sources and products, in press, Phil. Journal R.A.S., !...!!._, 399-438, 1984a. Trans. Roy. Soc., London, 1984. Grand, S. and D. V. Heimberger, Upper mantle shear structure beneath the northwest Atlantic ocean, Submitted to Jour. Geophys. Res., 1984b. Jay D. Bass and Don L. Anderson, Seismological Hirahara, K., A large scale three-dimensional Laboratory, California Institute of Technology, seismic structure under the Japan islands and Pasadena, CA 91125 the Sea of Japan;, J. Phys. Earth, 25, 393- 417, 1977. - (Received May 7, 1984; Hirahara, K., Three-dimensional seismic structure accepted May 17, 1984.)