Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Seismic inversion wikipedia , lookup
Schiehallion experiment wikipedia , lookup
History of geology wikipedia , lookup
History of Earth wikipedia , lookup
Supercontinent wikipedia , lookup
Age of the Earth wikipedia , lookup
Oceanic trench wikipedia , lookup
Post-glacial rebound wikipedia , lookup
Future of Earth wikipedia , lookup
Plate tectonics wikipedia , lookup
Geochmica d C’osm&imxn Copyright b 1991 Pergamon 0016-7037/91/$3.00 Acta Vol. 55, pp. 2083-21 IO Press pk. Printed in U.S.A. + .@O INAUGURAL INGERSON LECTURE Phase transformations and their bearing on the constitution and dynamics of the mantle* A. E. RINGWOOD Research School of Earth Sciences, Australian National University, PO Box 4, Canbena, ACT 2605, Australia Abstract-The bulk chemical composition of the Upper Mantle (“pyrolite”) is derived from experimental and petrological studies of the complementary relationships between basaltic magmas and refractory peridotites. The phase transformations which are experienced by pyrolite between depths of 100-800 km are reviewed in some detail, particularly with regard to their capacity to explain the seismic P and S velocity profiles throughout this region. The transition of olivine and pyroxene to &(Mg,Fe)$iO., plus garnet provides a satisfactory explanation of the velocity changes associated with the 400 km discontinuity within the limits of error of the seismic velocity dete~inations. Seismic velocities between 400 and 6.50 km are likewise consistent with this region crystallizing as an assemblage of &y(Mg,Fe)$304 plus garnet. The depth of the 650 km seismic discontinuity corresponds closely to the pressure at which (Mg,Fe)$iOe spine1 disproportionates to MgSi03 perovskite + (Mg,Fe)Q magnesiowtistite. This transformation is completed over a narrow depth interval (~4 km) and is capable of explaining the seismic characteristics of the 650 km discontinuity. The elastic properties and density of the Lower Mantle are readily explained within their observational uncertainties by a pyrolite composition crystallizing as an assemblage of perovskites plus magnesiow~stite. A substantial change in chemical composition (e.g., an increase in Si& and/or a decrease in FeO) at the 650 km discontinuity is not required by available geophysical and petrological data. The near-chondritic ratios of involatile lithophile elements in pyrolite provide important boundary conditions for geochemical Earth models and place severe limitations upon hypotheses which invoke large-scale melting early in the Earth’s history. They also imply that the Mg/Si ratio of the Lower Mantle is similar to that of the Upper Mantle. The geochemical evolution and dynamical behaviour of the mantle are strongly influenced by the petrological differentia~on of pyrolite at mid-ocean spreading centres to form new oceanic lithosphere. The MORB basaltic crust is underlain by a layer of harzburgite. During subduction, these lithologies each respond to sequential phase transformations in a different manner, so that at any given depth, they may differ in density from surrounding pyrolite. Most importantly, between 650 and 750 km, both former basaltic crust and harzburgite are less dense (-0.15 and 0.05 g/cm3, respectively) than pyrolite. Relatively young and thin subducted plates may have attained thermal equilibrium with surrounding mantle by the time they reach the 650 km discontinuity. Because of their buoyancy below 650 km, these plates would not be able to penetrate the 650 km discontinuity and instead are deflected laterally along the discontinuity. This process would eventually produce a layer of former basaltic crust (gametite), buoyantly trapped on top of the 650 km discontinuity, which would partially isolate the convective systems of the Upper and Lower Mantle. In contrast, older and thicker oceanic plates may be sufficiently cool and strong to permit their differentiated upper layers to penetrate the 650 km discontinuity. However, the tips of these plates experience substantial buoyancy stresses at this depth which cause buckling. in consequence, the descending slab piles up and forms a large melange, or “megalith,” of mixed former hanburg&e and former oceanic crust with cross-sectional dimensions amounting to several hundred kilometres. Its integrity is maintained for a Iimited period (- lo8 a) by high viscosity arising from its lower temperature as compared to surrounding mantle. The presence of megaliths may explain a number of geophysical observations including the complex structures present near the intersections of slabs with the 650 km discontinuity, which have recently been imaged by seismic tomography, as well as associated depressions in the depth of this discontinuity. The megahth functions as a “cold finger” in the Lower Mantle and may initiate a sinking convection current. After the cessation of subduction, the megalith ~aduaily warms up, accompanied by reduction in its viscosity, and ultimately becomes entrained in the convective system of the Lower Mantle. Mantle convection is thus envisaged as a hybrid system, with a large degree of independent convection within both the Upper and Lower Mantle, combined with a more limited exchange of material between these regions. This behaviour is enhanced by the high viscosity of the Transition Zone as compared to the Upper and Lower Mantle. The origins of intraplate (hot-spot) magmas are considered in terms of the above model. Partial meiting of the former basaltic crust of the slab at depths of 200-600 km refertihzes overlying depleted peridotite. This refertilized material, entrained by the subducting plate, accumulates in a thin zone immediately overlying the garnetite layer on top of the 650 km * Delivered at the Goldschmidt Conference, Baltimore, MD, May 12, 1988. 2083 A. E. Ringwood 2084 discontinuity. Subsequent reheating of this fertile peridotite causes diapirs to ascend from the 650 km discontinuity accompanied by partial melting within the Upper Mantle to form geochemically enriched magmas which are erupted at hot spots. 1. INTRODUCIION THE DEVELOPMENT OF THE THEORY of plate tectonics has focussed attention on the nature of the “engine” within the mantle which drives plate motions. It is recognized that this engine represents a complex form of convection which is poorly understood. Important boundary conditions on convective processes have been provided by geochemical studies of basalt source regions, some of which have maintained their identities for more than a billion years, despite the mixing induced by mantle convection. It is essential to establish the physical locations of these reservoirs and the processes by which they have formed. The study of these topics has given rise to an interdisciplinary field appropriately named “chemical geodynamics” (ALL~GRE, 1982). Before meaningful progress can be made towards addressing the basic problems of chemical geodynamics, it is essential to possess an adequate understanding of the present physical and chemical constitution of the mantle. Major progress towards achieving this objective has been made during the last few decades. To a large extent, this has arisen from controlled laboratory experimentation on the chemical and physical properties of mantle minerals and rocks over a pressure-temperature regime encompassing a large proportion of the mantle. These topics are reviewed in the first half of the present paper, with particular emphasis on the key role of phase transformations. An attempt is then made to apply knowledge in this field to some topical areas of chemical geodynamics. The mantle is divided into three regions on the basis of seismic velocity distributions (Fig. 1). The Upper Mantle em- braces the region between the Mohorovicic Discontinuity (marking the base of the crust) and a major seismic discontinuity occurring near a depth of 400 km. A second major seismic discontinuity occurs near a depth of 650-670 km and the region between these discontinuities is known as the Transition Zone. The Lower Mantle comprises the large region extending between the “650 km discontinuity” and the Core, which is encountered near a depth of 2900 km. Many earth scientists use the term “upper mantle” to describe the entire region between the Moho and the 650 km discontinuity. This terminology should be discouraged. The Transition Zone possesses an entirely distinct mineralogy to the regions above and below, and this is manifested in important differences in physical properties which are of considerable geodynamic significance. It therefore warrants specific recognition as a major province of the mantle. Although the existence and positions of major seismic discontinuities near 400 and 650 km are well established, there are substantial uncertainties in estimates by various authors in the velocity changes at these discontinuities and also in the seismic velocity gradients between depths of -200 and 800 km, as illustrated in Fig. 1. BENNETT (199 1) has provided an illuminating discussion of the underlying causes of these very real uncertainties, estimates of which are given in Table 1. Accordingly, it is not desirable to select any unique velocity distribution as the basis for detailed interpretations of mantle mineralogy. These interpretations should always include appropriate allowances for uncertainties in the seismic data. In a recent comprehensive seismic investigation, SHEARER (199 I) concluded that the two major intramantle discontinuities were located at depths of 410 f 3 km and 660 t- 8 km. 2. THE UPPER MANTLE (a) Chemical Zoning in the Upper Mantle 41 200 300 400 500 600 700 800 DEPTH km FIG. 1. Compressional and shear velocity profiles derived from seismic observations: PEM (DZIEWONSKI et al., 1975); SHR14 (HELMBERGERand ENGEN, 1974); K8 (GIVEN and HELMBERGER, 1980); PREM (DZIEWONSKI and ANDERSON, 1981); GCA (WALEK, 1984); GH (GRAND and HELMBERGER, 1984). The P-wave velocities of most regions of the Upper Mantle immediately underlying the Mohorovicic Discontinuity (Moho) are in the range 8.1 -t 0.4 km/set. This property, coupled with certain broad petrological and chemical limitations, effectively restricts the mineralogical composition of this region to some combination of olivine, pyroxene(s), and garnet. The principal rock types containing these minerals are peridotite (olivine-pyroxene) and eclogite (garnet-pyroxene). Mineralogical intermediaries between these two rock types are rare. A wide range of evidence reviewed by RINGWOOD (1975) shows that the uppermost mantle or lithosphere is dominantly composed of peridotite, with eclogite widely distributed as local segregations but relatively small in total amount. This interpretation would probably represent a consensus view today; however, the consensus is a relatively recent phenomenon. During the 1950s and 1960s there was an active debate as to whether the bulk composition of the Upper Mantle was dominantly eclogitic or peridotitic. The ill-starred Mohole project was conceived to resolve this debate and to determine 2085 Inaugural Ingerson Lecture Table 1 ESTIMATED PERMISSIBLERANGES OF P-ANOS-WAVE SEISMIC 400 AND 650 KM AND S-WAVE Seismic Velocity Velocity Velocity Ah’D BETWEEN CORRESPONDING RANGES FOR p- 400-650KM (KENNET 1991) P-waves S-waves at 400 km 2.5-5.8 2.8-5.7 at 650 km 3.6-7.3 3.0-7.5 (%) changes discontinuity GRADIENTS parameter changes discontinuity DISCONTINLRTIES VELOCITY CHANGES ATTHE (W) gradients between 400- 1.8-2.9 x 1O-3 2.0-5.0 x 10-3 650 km (km/sec)/km whether or not the Mohorovicic Discontinuity represents a phase change from gabbroic lower crust to eclogitic upper mantle. The debate was effectively resolved via an extensive experimental investigation of the gabbro-eclogite transformation (RINGWOODand GREEN, 1966), which showed that the geophysical characteristics of the Moho in most regions were inconsistent with the proposed phase change. Numerous samples of peridotites from the Upper Mantle have been transported to the surface as xenoliths in kimberlites and alkali basalts, or have been intruded into the crust during erogenic activity. Most of these Upper Mantle peridotites are strongly depleted inlow melting-point components and incompatible elements, so that they would be unable to produce the common types of basaltic magmas if partially melted. Nevertheless, we know that basaltic magmas have been erupted in copious volumes throughout geological time at localities scattered all over the Earth’s surface, in both continental and oceanic settings. It therefore appears that beneath the refractory peridotite layer, there must exist a more primitive source region which has retained a significant basaltic component. RINGWOOD( 1962a,b) and GREEN and RINGWOOD(1963) denoted this primitive source material by the term “pyrolite,” implying a non-specific olivine-pyroxene rock capable of yielding basaltic magmas on partial melting (some workers have preferred to use the term “primitive mantle” in this context). Peridotite is believed to represent the refractory residue remaining after basaltic magma has been extracted from pyrolite. We thus arrive at a chemically zoned model for the Upper Mantle as indicated in Fig. 2. Beneath ocean basins the layer of depleted peridotite is believed to be quite thin (e.g., lo-50 km). The underlying pyrolite provides the source region for the most abundant class of magmas erupted at the Earth’s surface: mid-oceanic ridge basalts (MORBs). The xenolith population in kimberlites which penetrate stable continental cratons implies that the depleted peridotite layer beneath cratons is much thicker, probably in the vicinity of 150-200 km. However, where continental plates are rifted and begin to move apart, MORB basalts are erupted along the newly formed spreading centres (e.g., the Red Sea). This implies that the pyrolite layer is of global extent, continuous beneath ocean basins and continents, although deeper beneath the latter (Fig. 2). (b) Basalt Petrogenesis and the Pyrolite Model The pyrolite model is based upon the complementary relationships between basaltic and komatiitic magmas and their respective peridotitic and dunitic refractory residues. GREEN and RINGWOOD (1964, 1967) showed how the petrogenesis of various classes of basaltic magmas could be interpreted in terms of varying degrees of partial melting of pyrolite at defined depth intervals in the mantle. The methodology employed by these workers during the 1960s was based upon high-pressure-high-temperature experimentation combined with analyses of crystalline and quenched liquid phases via electron-probe techniques. It was this innovation which made it possible to interpret the physico-chemical behaviour of complex, multi-component systems as a function of P and T; and the methodology has been widely adopted by other DEPTH Km OCEAN CONTINENT GREGATIONS FIG.2. Chemically zoned model for the Upper Mantle. 2086 A. E. Ringwood laboratories during the 1970s and subsequently. A second powerful tool for investigating the petrogenesis of basaltic magmas was introduced by GAST (1968), who showed how the partitioning of trace elements could be used to constrain the nature ofthe partial melting and fractional crystallization processes involved in magma genesis. Methods for estimating the chemical composition of pyrolite were discussed extensively by RINGWOOD(1975). Petrogenetic relationships between a primitive MORB and residual harzburgite were employed by GREEN et al. ( 1979) to formulate this parental mantle composition (Table 2). A corresponding estimate was made by SUN (1982) on the basis of complementary relationships between komatiitic magmas and residual dunites (Table 2). A second method is based upon the recognition of types of Iherzolites which have experienced only very small degrees of partial melting and hence are likely to approach the pyrolite composition (e.g., JAGOUTZ et al., 1979; Table 1). These compositions are in close agreement with more recent estimates for the composition of the primitive Upper Mantle obtained by FREY et al. (1985) BONATTI et al. (1986) ZINDLER and HART (1986) and MCDONOUGH ( 1987). The methodology used to derive the major element composition of pyrolite has also been extended to obtain the abundances of minor elements in the primitive Upper Mantle (RINGWOOD, 1966a,b; RINGWOODand KESSON, 1977; SUN, 1982; ZINDLERand HART, 19.86; MCDONOUGH, 1987). The reliability of these estimates has improved successively as the data base has expanded. The estimate of the composition of the primitive Upper Mantle from MCDONOUGH (1987) is given in Table 3. Compositional models of this type have proven to be of considerable utility in several areas of geo- chemistry. For example, the abundance patterns of siderophile and volatile elements in the Earth’s mantle provide important constraints on models of accretion of the Earth and formation of the core (e.g., RINGWOOD, 1966a,b, 1984; W.&NKE,1981; SUN, 1982; W~CNKE and DREIBUS,1988). They have also been used widely in modelling magma petrogenesis by partial melting processes and crust/mantle differentiation (see, e.g., HOFMANN, 1988). The data in Table 3 show that many involatile lithophile elements in pyrolite (e.g., Mg, Ca, Al, Ti, Y, SC, heavy and intermediate REEs, Zr, and Hf) are present approximately in chondritic relative abundances (see also NESBI~Tand SUN, 1976; SUN and NESBITT, 1977; ZINDLER and HART, 1986). Moreover, if the entire mantle is assumed to be of pyrolite composition, a simple relationship exists between the composition of the bulk Earth (mantle and core) and the composition of primitive Cl chondritic meteorites, which display similar relative abundances of involatile elements to the Sun (RINGWOOD, 1966a, 1975; ZINDLER and HART, 1986). Essentially, a geochemically self-consistent Earth model can be derived from the CI chondrite composition by processes involving partial reduction of iron and nickel oxides to form a metallic core together with loss of volatiles. It is important to note, however, that about 20% of the total silicon must be removed from the mantle composition in order to preserve this relationship. Processes which could be responsible for this silicon deficit are discussed in Section 5. (c) Structure of the Upper Mantle The seismic velocities of the uppermost 200 km of the mantle vary in a complex manner both laterally and vertically. Table 2 PYROLITEI MODEL COMPOSITIONS 1 (Jagoutz, et al., 1979) 2 (Sun, 1982) 3 (Green, et al., 1979) SiO2 45.13 44.49 Ti02 0.22 0.22 0.17 Al203 3.96 4.30 4.4 Cr203 0.46 0.44 0.45 CaO 3.50 3.50 3.4 MgO 38.30 37.97 38.8 Fe0 7.82 8.36 7.6 NiO 0.27 0.25 0.26 MnO 0.13 0.14 0.11 Na20 0.33 0.39 0.4 lOOMg/(Mg+Fe) 1. Least depleted 2. Komatiite 3. MORB 89.7 ultramafic - dunite - harzburgite 89.0 xenoliths model model 45.0 90.1 Inaugural lngerson Lecture 2087 Table 3 ESTIMATEDCHEMICALCOMPOSITIONOFTHEPRIMITIVEUPPERMANTIE-TYROLITE" (AFTERMCDONOUGH,~~~~) Element CI I’yrolite Li (ppm) 1.57 1.6 Be 0.027 0.08 B 0.98 0.5 250 C 50000 26 F 61 2545 Na 4950 22.45 Mg% 9.30 2.36 Al % 0.827 Si % 10.25 20.93 95 P (ppm) 1080 350 S 62000 30 Cl 700 240 K 545 Ca % 0.902 2.573 SC 5.98 17.34 Ti 440 1280 82 V 56 2935 Cr 2670 1080 Mn 1900 Fe % 18.10 6.53 105 Co 500 Ni 108ccl 1890 Gl 120 30 56 Zn 310 Ga 10.0 3.9 Ge 32.4 1.1 A5 1.93 0.13 Se 18.6 0.05 Br 3.57 0.075 Rb 2.32 0.635 Sr 7.26 21.05 Y 1.57 4.55 Zr 3.87 11.22 Nb 246 713 MO(ppb) 920 65 4.2 Ru 710 1 Rh 134 5 I’d 555 Pyrolite (Normalized to Mg and CI) Element CI Pyrolite 8 40 0.017 0.023 80 13 0.067 1720 175 0.042 160 5 0.013 2260 13 0.002 430 11 0.011 0.073 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.20 1.21 0.09 0.003 0.003 0.003 0.003 0.002 0.010 0.021 0.031 0.009 1.20 1.07 0.42 1.23 Cd 0.21 In 0.02 Sn 0.18 Sb 0.21 Te 1.00 I 1.18 cs 188 33 0.85 Ba 2410 6989 As 200 710 0.036 La 244 0.0023 Ce 1833 0.018 Pr 632 95.7 0.18 Nd 471 1366 1.18 Sm ELI 153 58 444 1.20 1.21 Gd 205 595 0.61 Tb 0.46 254 0.24 “Y Ho 0.15 Er 165 0.087 Tm 0.072 Yb 0.10 0.075 0.16 0.014 37.2 56.3 708 278 168 108 737 163 479 25.5 74 481 LN 166 25.4 Hf 107 309 Ta 14 41 w 95 21 0.028 Re 36 0.0011 OS 520 3.4 0.009 Ir 480 3.3 0.11 rt AU 1010 6.8 140 0.75 Hg Tl 400 1.20 140 7 1.20 I% 2470 185 1.20 1.20 Pyrolite (Normalized to Mg and CI) 73.7 0.28 10 0.029 Bi 110 2.5 0.002 Th 29 84.1 0.003 U 8.1 21 0.004 From Li to Zr element concentrations and Mg, Al, Si, Ca and Fe are in wt%. are given These variations are caused by several factors. Pyrolite is capable of crystallizing in four distinct mineral assemblages which have well-defined fields in P, T, and fHzo space (GREEN and RINGWOOD, 1970), and possess distinctly different physical properties. Additional complications are introduced by small degrees of partial melting and by anisotropy of olivine. These topics have been reviewed by RINGWOOD (1975), GREEN and LIEBERMANN(1976), and LEVEN et al. (198 1). Below a depth of about 70 km, the stable mineral assemblage displayed by pyrolite is olivine, garnet, clinopyroxene rt orthopyroxene. This assemblage remains stable throughout most of the Upper Mantle to a depth of 400 km (AKAOGI and AKIMOTO, 1979). (However, orthopyroxene is eliminated in ppm, Nb to U are given via solid solution in clinopyroxene CALO and GASPARIK, 1990.) in ppb below about 300 km: PA- Because of its relative depletion in Fe, Ca, and Al, the refractory peridotite layer (Fig. 1) is about 0.06 g/cm3 less dense than underlying garnet.pyrolite (RINGWOOD, 1966~). In consequence, this layer is gravitationally stabilized and has therefore been resistant to disruption by convection (JORDAN, 1979, 198 1; O’HARA, 1975). The development of the sub-continental lithosphere as a chemical boundary layer is probably an evolutionary feature (CLARK and RINGWOOD, 1964). This layer attains its greatest thickness, possibly in the vicinity of 200 km, beneath Precambrian shields and has experienced an extremely complex geological and 2088 A. E. Ringwood geochemical history. Although depleted in its major element chemistry, it appears to have been subjected to repeated episodes of “metasomatism” in which local domains enriched in incompatible elements were formed. Several seismic studies of the lithosphere beneath stable continental regions have provided evidence of a discontinuity at a depth of about 200 km. A high-resolution study by HALES et al. ( 1980) indicated an increase in P-wave velocity of about 0.3-0.5 km/set at this depth, followed by a further small decrease about 30 km deeper. Whereas the P- and S-wave velocity distributions in the sub-continental lithosphere are well explained at most depths by peridotite and garnet pyrolite lithologies, this complex seismic feature at 200 km is not so readily interpreted in these terms (LEVEN et al., 198 1). Moreover, it does not appear to be of global extent (SHEARER, 199 1). For example, it was not recognized in the data of the Early Rise seismic profile which traversed North America with a comprehensive coverage of azimuth (WARREN, 1968). LEVEN et al. (1981) suggested that the discontinuity might be caused by preferred orientation of olivine and pyroxene crystals near a depth of 200 km. These minerals are highly anisotropic in their elastic properties. Preferred orientation of olivines and pyroxene may have been caused by shearing near the boundary of the sub-continental lithosphere with the underlying asthenosphere, as continental plates migrated across the Earth’s surface. (d) Large-scale Melting of the Upper Mantle RINGWOOD (1975, pp. 577-579) pointed out that the Upper Mantle would probably have been molten to a depth of 200-400 km immediately after accretion of the Earth. The fate of this primitive terrestrial magma ocean has since become the subject of considerable speculative discussion. Ringwood concluded that the magma ocean would have crystallized rapidly, and that the differentiated cumulates would have been subducted into the mantle and homogenized by convection. It now seems that the process may have been more complicated. Recent evidence implies that the densities of ultrabasic partial melts generated in the mantle at depths of 250-400 km are higher than those of their olivine plus pyroxene residue (STOLPER et al., 1981; OHTANI, 1984; RIGDEN et al., 1984; AGEE and WALKER, 1988a; MILLER et al., 199 1). Thus, an ultrabasic magma ocean could have been gravitationally stable in the 250-400 depth interval and bounded by an overlying layer of olivine and pyroxene crystals. NKBET and WALKER (1982) proposed that such a subterranean magma ocean existed for as long as two billion years and was the source of much of the komatiitic volcanism which occurred during the Archaean. However, ARNDT (1986) has drawn attention to several geochemical and isotopic difficulties attached to this hypothesis. In particular, it is unable to provide an acceptable explanation of the “depleted” geochemical signatures of many Archaean komatiites. MCFARLANE and DRAKE (1990) also showed from studies of the partitioning of Co and Ni that Upper Mantle rocks do not display the signatures that would be expected on the basis of the olivine-flotation model of ACEE and WALKER (1988b). Recent studies (e.g., HERZBERG and O’HARA, 1985; OHTANI, 1985; TAKAHASHI, 1986; TAKAHASHI and ITO, 1987) have demonstrated that the temperature interval between the solidus and liquidus of peridot&e decreases quite markedly with pressure, from 600°C at atmospheric pressure to less than 1OO’C at 16 GPa (see also, JACKSON, 1977). This has led to the proposal that the Upper Mantle has itself been formed by partial melting of materials from the Transition Zone and Lower Mantle (e.g., HERZBERG and O’HARA, 1985). According to this model, the region of melt extraction was located mainly in the Transition Zone between depths of 400 and 670 km where majorite garnet is the liquidus phase on a pyrolite composition (TAKAHASHI, 1986). Presumably, the mechanism would have involved subsolidus convection throughout the Lower Mantle with upwelling plumes experiencing partial melting via adiabatic decompression as the plumes ascended into the Transition Zone. Although the existence of a narrow solidus-liquidus temperature interval may be consistent with the above scenario, it is by no means sufficient to validate it. The small melting interval could equally reflect the simple facts that olivine happens to possess a high melting point at zero pressure combined with a low melting-point gradient with pressure, whilst pyroxenes and garnet have much lower melting points at zero pressure but possess relatively high melting-point gradients. Given this conjunction of properties, it is inevitable that peridotite would possess a wide melting interval at low pressures and a small melting interval at high pressure. These characteristics alone do not justify the petrogenetic interpretation of the Upper Mantle as a product of partial melting. It was pointed out earlier that many involatile lithophile elements (Mg, Ca, Al, Ti, Zr, Hf, SC, Y, and heavy and intermediate REEs) are present in pyrolite in near-chondritic ratios. Experimental results by KATO et al. (1988a; summarized in Table 4) show that majorite garnet on the liquidi of chondritic and komatiitic melts at 16-20 GPa is enriched by factors of 1.5-2.5 over coexisting liquid in Al and SC and is correspondingly depleted in Ca, Ti, Sm, and La by more than a factor of 2. These observations imply that a partial melting process in which majorite was the principal residual phase would have caused much larger fractionations of these elements than are observed in pyrolite. If there had been a primordial magma ocean, all traces of its existence must have been removed by subsequent subsolidus convection. Hafnium isotopic compositions of Archaean komatiites point to the same conclusion (GRUAUet al., 1990). 3. THE TRANSITION ZONE (a) Historical Background In a classic paper on the elasticity of the mantle, BIRCH ( 1952) concluded that the increases of seismic velocities between depths of about 300-900 km were too large to be caused by self-compression of homogeneous material, and that mantle silicate minerals therefore transformed to denser, closer-packed phases in this depth interval. Testing Birch’s hypothesis provided a challenge for experimentalists since the pressures and temperatures in this region were well beyond the capacities of existing high P,T apparatus. However, the plausibility of his interpretation was strengthened by observations that many germanates;isostructural with silicates at 2089 Inaugural lngerson Lecture Table 4 MgSiO3 PARTITION COEFFICIENTS FOR SELECTED ELEMENTS BETWEENLIQUIDUS PEROVSKITE, PYROPERICH GARNETS AND IJLTRABASIC MELTS, AND Casio3 PEROVSKrrEAND A BASALTICMELT. @ROMUT0 MgSiOs(pv)’ Element CaSi03 BETWEEN et al., 1988a) (pv) D mpv/liq Dcpv/liq Dgnt/liq Ca 0.2 0.8 0.6 Al 0.5-0.8 0.6 2.5 Na 0.02 0.4 0.1 Ti 3 3 0.4 Zr 9 5 0.6 Hf 14 6 0.8 U 25 Th 20 Nb -1 1.4 <O.l SC 5 0.2 1.7 Y 3 2.5 1.3 Yb 2 1.5 1.4 Sm 0.2 6 0.2 La <O.l 5 <O.l 10 PO Sr 3 Ba <O.l <O.l K,Rb,Cs <O.l <O.l * This is the preferred set of partition coefficients obtained in the investigations of Kato et al. (1988a, 1988b). It is believed that they are likely to be accurate to +50% and in many cases to +-30%. The set of partition coefficients obtained by Kato et al.(1988b) represents limits only, owing to constraints imposed by the experimental method employed. Nevertheless, they demonstrate that Dmpv/liq (Sc,Zr,Hf) > 3 and Dmpv/liq (Ca,Sm) < 0.2. low pressures, transformed to much denser polymorphs at modest pressures and that germanates tended to serve as highpressure, crystal-chemical models for silicates (e.g., RINGWOOD, 1962c, 1966~). Moreover, studies of phase relationships displayed by germanate-silicate systems enabled quantitative predictions to be made of the pressures required to transform silicates into denser polymorphs. Thus, from a study of the thermodynamics of Mg$i04-NizGe04 solid solutions (at ambient pressure), RINGWOOD (1956, 1958a) calculated that MgzSiOd should transform from the olivine to the spine1 structure at 17.5 t 5.5 GPa, 15OO”C, and that the spine1 polymorph would be 11 + 3% denser than olivine. (These predictions have subsequently been verified.) A similar result was obtained by direct extrapolation of high-pressure phase boundaries in this system (RINGWOOD and SEABROOK, 1962) and is shown in Fig. 3. It was also found that the olivines Fe2Si04, N&SiO+ and Co2Si04 transformed to denser spine1 + The designation + FeO). Mg’ refers to the molar ratio 100 MgO/(MgO structures at 2-7 GPa (RINGWOOD, 1958b, 1962d, 1963) and that Si02 transformed to the rutile structure near 10 GPa (STISHOV and POPOVA, 1961). These results, reviewed by RINGWOOD (1962c, 1966~) and CLARK and RINGWOOD (1964), were widely considered to have established the validity of Birch’s hypothesis. Nevertheless, the challenge to establish the detailed mineralogical nature of the Transition Zone remained. In 1966, Ringwood and Major developed an apparatus capable of achieving - I8 GPa and 12OO”C, and employed it to discover many new phase transformations in mantle silicates, including those of magnesian olivines (Mg’80)t to spine1 and of (Mg’& to a spinel-like (beta) phase (RINGWOOD and MAJOR, 1966). Transformations of several silicate pyroxenes into a new kind of garnet structure containing up to 25% of octahedrally coordinated silicon were also observed (RINGWOOD, 1967; RINCWOOD and MAJOR, 1971). The latter workers also synthesized a perovskite containing -80% CaSi03, and obtained strong evidence that pure Casio3 perovskite had been synthesized in their experiments. This synthesis was later confirmed by LIU and RINGWOOD (1975). Reviews of this 2090 A. E. Ringwood tories using multi-anvil systems. Preeminent in this field has been the laboratory of Akimoto at the University of Tokyo. Some of the more important experimental investigations carried out by this group are described by AKIMOTO (1972, 1987), AKIMOTO and FUJISAWA(1966, 1968) AKIMOTO et al. (1976, 1977), AKAOGI and AKIMOTO (1977, 1979), and AKAOGI et al. ( 1987). More recently, the laboratory of Ito at the University of Okayama has made a wide range of major contributions (discussed later). During the last few years, there has been considerable proliferation of multi-anvil and diamond anvil laboratories and the field is now extraordinarily active. , 20 NizGeOd 40 60 mole % forsterite OLIVINE solid solutions a0 Mg,SiO, FIG. 3. The System Ni,GeO,-Mg2SiOdat 600°C and O-9 GPa. Extrapolation of the phase boundaries indicates that pure Mg2Si04 olivine should transform to the spine1 structure near 16 GPa (after RINGWOOD and SEABROOK, 1962). work, mainly carried out at the Australian National University, were given by RINGWOOD( 1970, 1975). The early 1970s witnessed the introduction of two new techniques of generating high pressures and temperatures which have since played key roles on this field. BASSETTet al. ( 1967) developed an improved diamond anvil cell capable of generating 30 GPa and, in 1972, combined it with laser heating. Using this system, BASSETTand MING (1973) observed the disproportionation of FezSi04 spine1 into Fe0 + stishovite around 25 GPa and 3000°C. LIU (1974, 1975) used the laser-heated diamond anvil cell to discover the key transitions of MgSi03 to the perovskite structure, the disproportionation of Mg,Si04 spine1 to perovskite plus periclase, and the transformation of pyrope garnet to perovskite. (The existence of MgSi03 perovskite had been predicted by RINGWOOD 1962c, 1966c.) A review of the early ANU diamond anvil work is given by LKJ (1979a), whilst a more comprehensive treatment of high-pressure phase transformations is provided by LIU and BASSETT( 1986). The other important advance in high-pressure technology during this period was the development of the two-stage multianvil system by KAWAI and ENDO (1970) which permitted pressures of up to 25 GPa and temperatures of -2000°C to be generated in significant volumes (a few mm3). Using these systems, RAWA et al. (1974) synthesized MgSiO, ilmenite whilst IT0 et al. (1974) synthesized pure MgzSi04 spinel. So far, we have been concerned primarily with the early discoveries of high-pressure phases which are important in mantle mineralogy. However, the systematic exploration and definition of their phase relationships and the characterization of their properties has been an enterprise of equal importance. This work has been carried out mainly in Japanese labora- (b) Transformations in a Pyrolite Mantle In Section 2, a mode1 bulk chemical composition (pyrolite) for the Upper Mantle was formulated (Table 1). We now wish to inquire whether phase transformations occurring in material of pyrolite composition are capable of explaining the variations of known physical properties (mainly seismic velocities and densities) throughout the Transition Zone. The most significant seismic characteristics in this region are the two major velocity discontinuities near depths of 400 and 650 km (Fig. 1). As discussed earlier, substantial uncertainties exist in the detailed velocity profiles, especially the magnitudes of the velocity increases and in the intervening velocity gradients. The phase transformations which are expected to occur with increasing depth in a mantle of pyrolite composition are summarized in Fig. 4. Phase relationships in the system Mg2Si04-Fe2Si04 (Fig. 5) play a key role in determining mantle structure around depths of 400 km and have been successively refined during the last 25 years (RINGWOODand MAJOR, 1966, 1971; AKIMOTO, 1972; KAWADA, 1977; AKIMOTO~~~ FUJISAWA,1966,1968; AKAOG~et al., 1989, 1984; PYROLITE DEPTH km -1 DENSITY g/cm3 100 200 300 t 3.38 OPX OLIVINE (Mg,Fe),SiO, I \ G+PX 3.42 3.59 3.62 3.68 3.71 a00 0 0.2 0.4 VOLUME 0.6 0.8 1.0 FRACTION FIG. 4. Mineral assemblages and (zero-pressure) densities displayed by pyrolite to a depth of 850 km. It is assumed that the temperature at 400 km is near 1400°C and at 650 km is near 1600 “C in accordance with the mantle geotherm of BROWN and SHANKLAND (1981). 209 1 Inaugural Ingerson Lecture 1600” C 18 SPINEL (y) SOLID SOLUTION 16 8 6 OLIVINE (cc) SOLID SOLUTION 80 Mg&iOd 60 40 mole % forsterite 20 FenSi FIG. 5. Phase relationships in the system Mg$iO.,-FezSi04 at 422 GPa and at 1600°C (after AKAOGIet al., 1989). KATSURA and ITO, 1989). The most recent work (Fig. 5) shows that the olivine component of pyrolite (Mgls9) would transform to the beta-phase at a depth close to 400 km, with a transition interval of 9-17 km and at temperatures in the vicinity of 1400- 1600°C. This is accompanied by a density increase of about 8% (referred to ambient P and 7) and is believed to be primarily responsible for the major seismic discontinuity near 400 km. AKA~GI et al. ( 1989) obtained a gradient of 1.6 MPa/“C for this transition, which is substantially smaller than earlier estimates. Pyroxenes and garnet with MSi03-A1203 stoichiometry (M = Mg,Fe,Ca) constitute the second most abundant class of minerals in the Upper Mantle. At a depth of about 300 km, appreciable amounts of pyroxenes began to dissolve in the preexisting garnet phase to form complex solid solutions M3(mi,A12)Si3012 (RINGWOOD, 1967, 1975; AKA~GI and AKIMOTO, 1977; AKAOGI et al., 1987). Up to one-quarter of the silicon atoms in these high-pressure garnets, named “majorite,” are octahedrally coordinated. For the pyrolite composition, complete conversion of the pyroxene component to garnet is achieved at a depth of 460 km, assuming a temperature of about 1500°C at this depth (IRIFUNE, 1987; IRIFLJNEand RINGWOOD, 1987a). The slope of this transition is relatively insensitive to temperature, with dP/dT - I.5 MPa/“C. The pyroxene-garnet transformation causes marked positive velocity gradients to occur on either side of the 400 km discontinuity in a zone between depths of 300-460 km. The transformation is accompanied by a density increase of - 10% in the pyroxene component of the Upper Mantle. At a somewhat greater depth, /3(Mg,Fe)zSi0, transforms to the spine1 (y) structure (Fig. 5). Assuming a mantle temperature of about 15OO”C, the phase transformation would occur over a depth interval between -500 and 530 km and would be accompanied by a density increase of about 2%. The elastic properties of @and y Mg$i04 are similar, and the transition may be reflected mainly by a small increase in seismic velocity gradients between these depths (WEIDNER, 1985; WEIDNER and ITO, 1987; RICDEN et al., 1991). Evidence for a small seismic discontinuity near 520 km has been presented by SHEARER( 1990). This discontinuity may, however, be spread out over a depth interval of 30-50 km (CUMMINSet al., 1991). The next major phase transformation occurring at greater depths involves the formation of MgSi03 perovskite (LIu, 1974, 1975). The perovskite polymorph of MgSi03 possesses a density of 4.10 g/cm3 which is about 4% greater than that of an isochemical mixture of periclase plus stishovite. LIU (1979a, 1975) and IT0 et al. (1984) showed that at temperatures in the vicinity of 16OO”C, Mg,Si04 spine1 disproportionates to MgSi03 perovskite plus MgO periclase at a pressure near 23 GPa. This matches the depth of the “650 km discontinuity,“* variously placed between 650 and 690 km according to different seismic studies (Fig. 1). The corresponding transformation in the system Mg$i04-FezSi04 has been extensively studied by YAGI et al. (1979), ITO and YAMADA (1982), ITO et al. (1984), and ITO and TAKAHASHI (1987, 1989). The latter results (Fig. 6) demonstrate that the transition pressure is essentially constant at 23 GPa ( 1600°C) through the composition range Mg*,,-Mg’,, and that the transition is remarkably sharp, being completed within a depth interval smaller than 4 km. Accordingly, the transition, which is accompanied by a (zero-pressure) density increase of I l%, would cause a steep seismic discontinuity at a depth close to 650 km. The transition possesses a negative temperature gradient of -4(*2) MPa/“C (IT0 et al., 1990; IT0 and TAKAHASHI, 1989), a characteristic which has important geophysical implications. Majorite garnet is the second most abundant mineral occurring in the Transition Zone (assuming a pyrolite bulk composition). Its stability has been studied experimentally by IRIFUNE and RINGWOOD(1987a) in a composition identical with the majorite phase of pyrolite and also by IT0 and TAKAHASHI(1987) in a simplified composition. The former investigation showed that majorite is stable from 16 to 20 GPa. Above 20 GPa, a CaSi03 perovskite phase is exsolved from majorite garnet and this is joined near 22 GPa by MgSiO, ilmenite. Exsolution of these phases causes the 1600” C ’ 24. PERO”SK,TE MAGNESIkV~STITE 3;w b I lbH”“l It ST,SH&TE 2 a MpSiOd SPINEL 80 Mg $30, *Also referred to as the “670 km discontinuity.” 2 MAGNESlOWkTlTE PEROVSKITE MSiO~ 60 mole % 40 20 FezSiO, FIG. 6. Pseudo-binary phase diagram for the system Mg,Si04Fe2Si04at 21-26 GPa and 1600°C (after ITOand TAKAHASHI, 1989). 2092 A. E. Ringwood residual garnet to become richer in A1203 and MgO, approaching the pyrope composition. Compositions of coexisting 2 perovskites + garnet at 24.5 GPa, 1400°C, obtained by IRIFUNE and RINCWOOD (1987a) are given in Table 5. Work by ITO and TAKAHASHI(1987) on a simplified composition showed that further transformation of garnet at still higher pressures proceeded over a significant pressure interval and that complete transformation into an assemblage of two perovskites plus a new highly aluminous phase was achieved at a pressure near 27 GPa. MADON et al. (1989a) found that the aluminous phase produced during transformation of a naturally occurring pyrope garnet to perovskites possessed a hollandite-like structure, with a composition close to CaMgAl&O,,. Experiments by ITO and TAKAHASHI(1987) and IRIFUNE and RINGWOOD (1987b) show that the substantial amounts of alumina present in the compositions studied cause the garnet-perovskite(s) transformation to be smeared out over a pressure interval of about 3-4 GPa and may not be completed until depths of 700-720 km in the pyrolite composition (Fig. 4). It should be recognised that uncertainties in seismic velocity distributions permit substantial tradeoffs between the magnitudes of the velocity jumps at the 400 and 650 km discontinuities and the seismic velocity gradients between 400 and 650 km (Fig. 1, Table 1). Likewise, uncertainties in the elastic properties of mantle minerals (particularly the temperature derivatives of the elastic moduli) hinder efforts to provide a precise interpretation of the mineralogy appropriate to a given seismic velocity distribution. WEIDNER (1985), IRIFIJNE (1989), AKAOGI et al. (1987), WEIDNER and ITO (1987), BINA and WOOD (1987), BINA (199 1), YEGANEH-HAERIet al. (1989), and GWANMESIAet al. (1990) have calculated the seismic velocity and/or density distributions within the 200-700 km region of the mantle for a pyrolite bulk composition, using the preceding evidence (Section 3a) on mineral stability fields combined with recent elasticity and density data for the individual mineral phases. They conclude that the pyrolite composition is capable of explaining the positions of the major seismic discontinuities near 400 and 650 km, the velocity changes across the discontinuities, and the velocity gradients and density distributions between them, within the (substantial) uncertainties mentioned above. BASS and ANDERSON(1984) and ANDERSONand BASS (1986) argued that the seismic velocity jump across the 400 km discontinuity and the velocity gradients below 400 km are inconsistent with a pyrolite composition. Their conclusion depends rather critically upon the choice of a preferred set of seismic velocity distributions (PREM) in the mantle and a selected set of elastic properties of minerals. ANDERSONand BASS(1986) claim instead that the 400650 km region is composed of “piclogite” which is chemically equivalent to an eclogite containing an additional 15-20s of olivine. The high positive seismic velocity gradients between 400 and 600 km characteristic of the PREM model are attributed to the persistence of clinopyroxene in this region, accompanied by its increasing solubility in garnet. According to their model, clinopyroxene is not consumed until a depth of 600 km. However, this is contradicted by detailed experimental studies of the stability field of clinopyroxene in relevant compositions, which show that clinopyroxene would be eliminated by 460 km via solid solution in garnet (IRIFUNE, 1987; IRIFUNEand RINGWOOD,1987a; IRIFLJNEet al., 1989). Accordingly, high gradients between 460 and 600 km cannot be caused by this mechanism. A further weakness in the Anderson-Bass model was their choice of a bulk modulus for Table 5 COMPOSITIONS OF CO-EXISTING PHASES PRESENT IN A PYROWTE COMPOSITION (COLUMN 2) AT 24.5 GPA AND14OO’C(FROM IRIFUNEAND RINGWOOD, 1987a) Bulk Mg Garnet Perovskite Composition Ca Perovskite SiOz 51.8 46.6 52.3 51.2 TiOz 0.5 0.1 1.8 1.5 Al203 11.4 18.7 5.3* 1.9 Cr.203 0.9 1.8 0.8 0.2 Fe0 3.1 3.0 3.7 0.6 MgO 23.1 25.3 34.8 3.1 CaO 9.5 4.1 1.0 41.3 NqO 0.9 0.9 0.1 0.3 * This value may be too high because of electronprobe beam overlap on garnet Inaugural Ingerson Lecture majorite which is now known to have been far too high (YAGI et al., 1987). DUFFY and ANDERSON (1989) reexamined the issue in the light of subsequent seismic and mineral-physics data and concluded that the S-wave distribution implied an olivine content of -35% whilst the P-wave distribution was consistent with 53% olivine. Their preferred olivine content was 40%. This is not so very far removed from the pyrolite composition of 55-60% olivine (RINGWOOD, 1975). The bulksound (G/p where K = bulk modulus, p = density) velocity distribution across the 400 km discontinuity was determined by BINA (1991) using Duffy and Anderson’s data set. Bina found that the velocity change was best matched by an olivine content of 68 +_ 11%. Recent progress in the determination of accurate pressure derivatives of elastic moduli for Transition Zone minerals promises to clarify the question of chemical homogeneity between 400 and 650 km (GWANMESIAet al., 1990; RIGDEN et al., 199 1). These results suggest that the velocity contrasts across the 400 km discontinuities for P and S waves can be reconciled with a range of compositional models varying in orthosilicate content from about 45 to 65%, each associated with a different but not unreasonable choice for the values of the (as-yet-unmeasured) temperature derivatives of elastic wave velocities. Compressional wave velocity gradients within the Transition Zone are also compatible with such compositional models, which include the pyrolite composition. However, the S-wave gradients calculated from acoustic measurements may be significantly smaller than most of the seismically determined S-wave gradients (RIGDEN et al., 199 1). If confirmed, this would suggest the presence of some degree of chemical inhomogeneity within the Transition Zone. 4. THE LOWER MANTLE (a) Phase Relationships BIRCH (1952) concluded that the variation of seismic velocities and density throughout the Lower Mantle between depths of 900 and 2700 km could be explained by selfcompression of chemically homogeneous material within the Earth’s gravitational field, and that there was no evidence for the occurrence of major phase transformations or for substantial changes of chemical composition in this region. This interpretation has been supported by static high-pressure investigations on the stabilities of MgSi03 perovskite and MgO which show that these phases remain stable throughout the entire Lower Mantle (KNITTLE and JEANLOZ, 1987; MAO and BELL, 1977). The distribution of iron between perovskite and magnesiowiistite has been studied by YAGI et al. (1979), IT0 et al. (1984), GUYOT et al. (1988), and ITO and TAKAHASHI(1989) who demonstrated a strong preferential partition of Fe0 into magnesiowtistite. This effect tends to decrease with depth. The compositions of Mg-perovskite and magnesiowtistite below 1000 km in a mantle of pyrolite composition would be Mg’,,., and Mg*80,respectively (GUYOT et al., 1988). LIU (1977) demonstrated that up to 25 mol% of A&O3 can be accommodated in solid solution in MgSi03 perovskite. In a study of the transformation of a natural pyrope garnet to 2093 perovskite, MADON et al. (1989b) showed that perovskite (approximately Mg’,,) was capable of accommodating up to 20 atom% of Al in the octahedral Si site at a pressure of 50 GPa. (At higher Al contents an additional aluminous phase, CaMgA1&0i6, would be present.) These results suggest that with the lower (-4%) A&O3 contents characteristic of the pyrolite composition, the magnesian perovskite phase would be capable of accepting all of the AllO3 into solid solution throughout most of the Lower Mantle, and that an additional aluminous phase is unlikely to be present (contrary to a suggestion by RINGWOOD, 1982). Calcium silicate (CaSi03) adopts the perovskite structure at high pressures (RINGWOODand MAJOR, 197 1; LIU, 1975). Experimental investigations by IRIFUNE and RINGWOOD (1987a) show that in the Lower Mantle, calcium would be contained in CaSi03 perovskite and that solid solution between this phase and coexisting MgSi03 perovskite is very limited (Table 5). These results also show that the modest sodium content of the Lower Mantle would be accommodated in solid solution within a CaSi03 perovskite mineral. The experimental evidence cited above indicates that the mineral assemblage adopted by pyrolite in the Lower Mantle would comprise MgSiOJ perovskite + CaSi03 perovskite + (Mg,Fe)O magnesiowtistite. This mineral assemblage was recently identified in a natural (fertile) peridotite composition crystallized at 54 GPa and 1900°C (O’NEILL and JEANLOZ, 1990). (b) Composition The nature of the major seismic discontinuity near 650 km has been the subject of considerable debate. Its depth is coincident with a major phase transformation to perovskite which occurs both in orthosilicate (M2Si04) and metasilicate (MSi03) stoichiometries. The question under debate is whether the discontinuity is primarily caused by the isochemical disproportionation of silicate spine1 to perovskite plus magnesiowilstite or whether, in addition, there is a change in chemical composition at this depth. Chemical models proposed for the Lower Mantle range between the pyrolite composition and an essentially pure perovskititic mineralogy which would be consistent with a chondritic bulk MgO/SiO* ratio for the entire mantle. Extensive shock-wave data on olivines and pyroxenes at Lower Mantle pressures and temperatures show that for the same MgO/(MgO + FeO) ratio, the density difference between these two model mantle compositions is smaller than 0.06 g/cm3 (WATT and AHRENS, 1986). Further ambiguity is introduced by uncertainties over the temperature distribution in the Lower Mantle, by the effects of other minor components (CaO, A1203),by the possibility of additional minor (probably second-order) phase changes in the Lower Mantle, and, finally, by experimental uncertainties in the existing mineralphysics data base. It is evident that it would be extremely difficult, if not impossible, to determine whether the Lower Mantle is of perovskititic or pyrolitic composition from density data alone. An identical dilemma is encountered in comparisons of the elastic properties of the Lower Mantle with those estimated for pyrolite and perovskite compositions. Most workers who have discussed the likely composition 2094 A. E. Ringwood of the Lower Mantle in the light of elasticity and density data have concluded that they are consistent with the interpretation that the Lower Mantle possesses a similar chemical composition to the Upper Mantle, within the uncertainties of the seismic and mineral-physics data base (e.g., RINGWOOD, 1975; JACKSON, 1983; ITO et al., 1984; WATT and AHRENS, 1986; WEIDNER and ITO, 1987; WOLF and BUKOWINSKI, 1987; CHOPELASand BOEHLER, 1989; BUKOWINSKIand WOLF, 1990). However, the experimental and observational uncertainties do not preclude the possibility of a change in chemical composition at the 650 km discontinuity. Thus, a perovskititic Lower Mantle may be permitted, particularly if the Fe0 content is increased to offset the effect of increasing SiOZ content (e.g., JACKSON, 1983). In contrast to the above conclusion, some workers have concluded that the geophysical properties of the 650 km discontinuity and the Lower Mantle are inconsistent with a pyrolite composition and that a change in chemical composition, involving an increase in Fe0 content and possibly also of SiOz, is definitely required (e.g., ANDERSON, 1977, 1983; ANDERSONand BASS,1986; LILJ,1979b; KNITTLEet al., 1986; JEANLOZand KNITTLE, 1989). Most of these workers, and also others (e.g., HERZBERG and O’HARA, 1985; OHTANI, 1985; TAKAHASHI, 1986; AGEE and WALKER, 1988b), have preferred a perovskititic mineralogy (metasilicate, MSi03 stoichiometry) for the Lower Mantle. Without exception, the arguments used by these authors to justify this conclusion have proved to be seriously flawed. LIU (1979b) claimed that the depth of the 650 km discontinuity was not consistent with the pressure at which (Mg,Fe)$i04 spine1 transformed to perovskite + magnesiowtistite. However, significant errors are now believed to have occurred in Liu’s pressure calibrations (JEANLOZand THOMPSON,1983). Subsequently, more accurate phase equilibria studies have shown that the pressure of the transformation (ITO and TAKAHASHI, 1989) is consistent with the depth of this discontinuity. LEES et al. (1983) argued that the 650 km discontinuity is too sharp to be caused by the above phase transformation; however, this conclusion is contradicted by the recent experiments of ITO and TAKAHASHI (1989). JACKSON( 1983) demonstrated that the viability of pyrolite versus perovskititic models for the Lower Mantle would be critically dependent on the mean thermal expansion coefficient (5) of MgSi03 perovskite between about 25 and 1600°C. If this value were relatively “high,” exceeding 25 X 10-6/oC, a perovskititic Lower Mantle would indeed be favoured. On the other hand, if (Ywere smaller than 25 X 10-6/oC, a pyrolite composition would be consistent with the observed density of the Lower Mantle. KNITTLE et al. (1986) and ROSS and HAZEN (1989) measured the thermal expansion coefficient of MgSiOJ perovskite at room temperature, both groups obtaining a value of 22 X 10-6/oC. Knittle et al. attempted to measure (Yat higher temperatures and observed an anomalous increase to -40 X 10-6/oC at -600°C. Taking this at face value, KNITTLE et al. (1986) and JEANLOZand KNITTLE (1989) concluded that the “high” thermal expansivity of MgSi03 perovskite implied that the Lower Mantle must be substantially richer in iron than the Upper Mantle (in accordance with the study of JACKSON, 1983). However, the results of HILL and JACKSON(1990) and CHOPELASand BOEHLER(1989) indicate that KNITTLE et al. (1986) overestimated the thermal expansivity of MgSi03 perovskite. CHOPELAS and BOEHLER (1989) showed that for a wide range of materials, the variation of (Y at high pressure follows a relationship (a In al/a P)= = 5.5 ? 0.5, which implies a considerable decrease in Cuat Lower Mantle pressures. Thus, at the core-mantle boundary they found that 01is reduced to only 5 X 10d6. CHOPELASand BOEHLER(1989) calculated the densities of perovskite and magnesiowiistite (Mg’**)appropriate to the pyrolite composition using standard procedures but including the effect of pressure on (Y.The calculated density for the pyrolite model throughout the Lower Mantle was found to agree closely with the seismically determined density throughout this region, showing clearly that the pyrolite composition remains viable. Chopelas and Boehler’s estimate of the thermal expansion coefficient of MgSiO, perovskite at high pressures has recently been confirmed by direct measurements (WANG et al., 1991). I conclude that the density of the Lower Mantle is consistent with a pyrolite composition, but that uncertainties in the data base do not preclude the possibility that the Lower Mantle may possess a significantly different composition. A related constraint on Lower Mantle compositional models can be obtained from the observed density distribution throughout this region. Assuming that the Lower Mantle is chemically homogeneous, the observed p versus P relationship can be extrapolated to zero pressure using an appropriate equation of state (CLARK and RINGWOOD, 1964). The decompressed density thereby obtained can be compared with different mineralogical models with the aid of known zero pressure densities of individual minerals and their corresponding thermal expansion coefficients. The most rigorous investigation of this type yet carried out was reported by BUKOWINSKI and WOLF (1990). They demonstrated that a Lower Mantle of pyrolite composition was consistent with the Lower Mantle density distribution and that perovskititic models were “far less satisfactory than pyrolite-like models” although they could not be conclusively eliminated. On the other hand, they found that a class of models possessing higher levels both of SiOz and Fe0 than would be present in a perovskititic stoichiometry were permitted. However, it is difficult to justify these latter models on geochemical grounds. (c) The Lowermost (D”) Region of the Mantle It has long been known that the lowermost region of the mantle immediately overlying the core displays anomalous seismic velocities (e.g., JEFFREY& 1939; BULLEN, 1949; GuTENBERG,1958, 1960). The thickness of this anomalous region (defined by Bullen as the D” layer) is believed to be about 200 km. Seismic P- and S-wave velocity gradients are smaller than elsewhere in the Lower Mantle and may even be negative. Moreover, the seismic properties of the D” layer imply the presence of a high degree of lateral heterogeneity (e.g., YOUNG and LAY, 1987; DZIEWONSKIand WOODHOUSE, 1987). It has been suggested that the anomalous seismic properties could be caused by the presence of a chemical boundary layer 2095 Inaugural Ingerson Lecture (e.g., RINGWOOD,1959, 1961) or of a thermal boundary layer (e.g., STACEYand LOPER, 1983) or perhaps by a combination ofthese effects (LAY, 1989). Interpretation of D” as a thermal boundary layer would imply that a superadiabatic temperature gradient of a few degrees per kilometer exists in this region (STACEY and LOPER, 1983). This interpretation seemed to be supported by measurements of the effect of high pressure on the melting point of iron by WILLIAMSet al. (1987) which would imply that the Outer Core (near the core-mantle boundary) was about 1000°C hotter than the Lower Mantle (at depths of 2500-2700 km). However, a subsequent careful redetermination of the melting point of iron at high pressure by BOEHLERet al. (1990) indicates that the core is probably much cooler than was inferred by WILLIAMSet al. (1987), thereby casting doubt on the need for a thermal boundary layer in the D” region. It is well known that the Outer Core contains substantial amounts of one or more light elements in solution in molten iron. RINGWOOD(1977, 1984) advanced several arguments which suggested that oxygen is the principal light element in the core. Experimental investigations of the system Fe-Fe0 at high pressures by KATO and RINGWOOD(I 989) and RINGWOOD and HIBBERSON(1990) provided strong support for this hypothesis and showed that the solution of oxygen would cause a large depression of the melting point of iron, thereby causing the metal phase to melt at substantially lower temperatures than mantle pyrolite. Their results, in conjunction with those of BOEHLERet al. (1990) confirm that the case for a strong thermal boundary layer involving a superadiabatic temperature change of 500-1000°C across the D” region is decidedly weak. A milder superadiabatic temperature gradient cannot be excluded, but this would not have a large effect upon seismic velocity gradients in the D” region. RINGWOOD (1959, 1961) pointed out that the process of core-formation within the Earth would be likely to lead to a condition of chemical disequilibrium between core and mantle at their mutual interface. He suggested that chemical reactions consequently occurred near the core-mantle boundary (CMB), leading to transport of components (Si, Fe, and Ni) across the boundary, and that the anomalous seismic properties of the lowermost 200 km (D”) zone were caused by these processes. Ringwood estimated that the chemical disequilibrium would generate an EMF in the vicinity of I volt across the CMB and that under certain conditions, charge separation might occur with electrons and ions following different paths, leading to the generation of electrical currents across the CMB. (These effects might be enhanced if perovskite in the Lower Mantle should indeed be in a superionic conductive state as argued by PRICE et al., 1989.) RINGWOOD(1959) suggested that the electrical currents thereby produced might play a significant role, in conjunction with the core dynamo, in generating the Earth’s magnetic field. Some of these ideas have recently been exhumed by KNITTLE and JEANLOZ(1989) and JEANLOZ(1990) who also propose that the core is not in equilibrium with the mantle and that chemical reactions accordingly occur at and near the CMB, ultimately producing the seismic heterogeneities observed in the D” layer. JEANLOZ(I 990) maintains that this is the most dynamically and chemically active region in the mantle. This interpretation was based on observations by KNITTLE and JEANLOZ (1989) that metallic iron and (Mg,Fe)SiOs perovskite chemically react with one another at very high pressure (270 GPa) and temperature (23000°C). Similar observations have been reported by GUYOT et al. (1988). The detailed nature of these reactions was studied by RINGWOODand HIBBERSON(199 1) who showed that several common metallic oxides are soluble in molten iron at high P and T, and at oxygen fugacities near the iron-wiistite buffer. Individual oxides dissolved quasi-congruently, but there are large differences in the solubilities of individual oxides. Fe0 is much more soluble than SiOz, which in turn is much more soluble than MgO. Thus, at the CMB, molten iron would selectively dissolve firstly Fe0 and then SiOZ, leaving a layer enriched in periclase. The rate of this process would depend on the ionic conductivity of the lower-most mantle and would be accelerated if this region were in the superionic conductive state (PRICE et al., 1989). The periclase-enriched chemical boundary layer would possess lower density and seismic velocities than the overlying region. Because of the resultant gravitational instability of this layer, diapirs of periclase-enriched material would ascend from the CMB. The seismic heterogeneity of the D” region might be caused by the presence of large numbers of these diapirs. Ultimately, it is expected that convective mixing in the Lower Mantle would cause the diapirs to be dispersed, thereby producing a relatively high level of chemical homogeneity at depths above 2700 km. 5. GROSS MANTLE DIFFERENTIATION AND THE Mg/Si RATIO We have seen that geophysical evidence permits the Lower Mantle to be either of pyrolitic or perovskititic composition. The former alternative might be favoured on the grounds of economy of hypotheses. Moreover, we should note that it would be coincidental for the depth of a major change in mantle chemical composition to be identical with the depth at which (Mg,FehSiOd spine1 happens to transform to perovskite plus magnesiowtistite. Nevertheless, as noted previously, many scientists have continued to prefer a perovskititic Lower Mantle with metasilicate stoichiometry. In many cases, the primary reason for this preference is that the perovskititic interpretation would permit the bulk mantle to possess a chondritic Mg/Si ratio. A mantle of pyrolite composition possesses an Mg/Si ratio of 1.27 (atomic) whereas the chondritic ratio is 1.05. The pyrolite model thus requires an explanation for the apparent deficiency of SiOZ in the mantle. Advocates of a chemically zoned mantle comprising a pyrolite outer region and a perovskititic Lower Mantle have called upon the processes of crystallization differentiation from the molten state or of extensive partial melting of the mantle to achieve this state (e.g., KUMAZAWA,198 1; OHTAN], 1985; HERZBERGand O’HARA, 1985; ANDERSONand BASS, 1986; TAKAHASHI, 1986; AGEE and WALKER, 1988b). This hypothesis can be tested directly by appropriate experiments. As discussed in Section 2, a wide range of involatile lithophile elements is present in the Upper Mantle in near-chondritic relative abundances (Table 3). This observation provides a strong constraint on proposed global differentiation processes. A, E. 2096 Consider the crystallization of a completely molten mantle. MgSi03 perovskite is found to be the liquidus phase on chondritic and pyrolite compositions below about 700 km (IT0 and TAKAHASHI, 1987), and because of its high relative abundance (-80% of a chondritic Lower Mantle) it is expected to display a broad crystallization field before being joined by magnesiowiisite (- 10%). Casio3 perovskite is not expected to crystallize as a liquidus phase because of its low abundance (- 5%). The partition coefficients of many minor elements between silicate perovskites, garnets, and a range of ultrabasic liquids were determined by KATO et al. (1988a). Some of their results are shown in Table 4. Note that the partition coefficients of Zr, Hf, and SC (Group A) between MgSiOs perovskite and liquid are much higher than unity, whilst the corresponding partition coefficients of the light and intermediate rare earths (Group B elements) are much smaller than unity. These results show that quite modest degrees of fractionation of MgSiOj perovskite, whether caused by crystallization from the melt or by partial melting processes, would result in rather drastic fractionations between Group A and B elements in the residual liquid. An example is given in Fig. 7. It is seen that a large degree of MgSi03 perovskite fractionation is necessary in order to cause substantial changes in the Si/Mg ratio during crystallization of a chondritic mantle from the molten state. However, the separation of only 5-10% of MgSi03 perovskite would have driven the Sm/Hf and Sc/Sm ratios of the residual melts well outside the near-chondritic ratios which are presently observed in the extensive Upper Mantle source regions of midocean ridge basalts. The results of KATO et al. ( 1988a) imply rather strongly that the present mantle does not reflect the operation of comprehensive perovskite-controlled fractionation postulated by the authors cited above, whether or not these arise from global fractional crystallization or partial melting processes. They strongly suggest that the bulk composition of the Lower Mantle is similar to that of the Upper Mantle. 0.8 0.6 0 5 Mg-perovsklte 10 fractionation, % FIG. 7. Variation of Sm/Hf perovskite fractionation from and Si/Mg ratios as a function of Mga chondritic mantle composition, calculated from partition coefficientsgiven in Table 4. The Si/Mg ratio of the present Upper Mantle is also indicated. Shaded areas indicate uncertainties which would be caused by errors of f2 in the Sm/Hf and Sc/Sm ratios used in the calculation (from UT0 et al., 1988a). Ringwood It appears that either (i) the mantle was initially formed in an approximately homogeneous state, or (ii) ifglobal melting and differentiation indeed occurred during, or soon after, formation of the Earth, the record of these processes has been destroyed by subsequent subsolidus convection, which rehomogenised the mantle. RINGWOOD(1990a) has advanced several arguments which support the first, and are unfavourable to the second, of the two alternatives. However, these are beyond the scope of the present review. The primary conclusion is that the Mg/Si ratio of the Upper Mantle is similar to that of the Lower Mantle and is substantially higher than the chondritic ratio. Possible causes of this deviation from the chondritic Mg/Si ratio are considered below. RINGWOOD (1989a) showed that the similar A&O3contents of terrestrial and Venusian basalts imply that the ratio of normative olivine to pyroxene in the Venusian mantle resembled that of the terrestrial mantle and that the Venusian Mg/Si ratio was therefore terrestrial rather than chondritic. He also showed that geochemical relationships among pallasites, howardites, diogenites, and eucrites indicate that the Mg/Si ratio of the eucrite parent body was closer to the terrestrial than to the chondritic value. Moreover, he noted that the spectra of S-type asteroids which predominate in the inner asteroid belt imply high olivine/pyroxene ratios combined with high metal contents. These relationships suggest that Sasteroids possess terrestrial rather than chondritic Mg/Si ratios. This array of evidence suggests that the terrestrial mantle Mg/Si ratio is not unique to the Earth but may be representative of the inner region of the solar system between Venus and the inner asteroid belt. RINGWOOD(1989b) suggested that, rather than being depleted in Si, the Earth’s mantle and the inner solar system may actually possess the primordial solar Mg/Si ratio and that CI chondrites may have been enriched in Si via a cosmochemical process. (Nevertheless, the Mg/Si ratios of the Earth’s mantle and CI chondrites are both consistent with the Mg/Si ratios of the solar photosphere within the errors of measurement of the latter; ANDERSand GREVESSE,1989.) RIETMEIJER(1987, 1988) showed that both chondritic interplanetary dust particles and Halley’s Comet dust particles from the outer parts of the solar system are substantially enriched in relatively volatile elements (Mn, Cu, K, Na, S, Zn, Bi) and also in silicon relative to CI chondrites (outer asteroid belt). He suggested that there had been a radial chemical differentiation of volatile elements (including Si) between the two source regions in the inner and outer solar system and that the existence of this chemical zonation challenges the assumption that the CI chondrites possess primordial compositions. An alternative possibility suggested by RINGWOOD (1979) is that CI chondrites indeed possess the primordial (solar) Mg/Si ratio, but that there has been a significant depletion of silicon (accompanied by more volatile elements) from the inner solar system via high-temperature processing connected with the early evolution of the protosun and solar nebula. 6. PHASE TRANSFORMATIONS AND DENSITY RELATIONSHIPS IN SUBDUCTED OCEANIC LITHOSPHERE So far, this review has focussed on the characteristics of a mantle composed of pyrolite. However, it should be recog- 2097 Inaugural lngerson Lecture MATURE PHANEROZOIC ASEISMIC RIDGES ARCHEAN OCEANIC OCEANIC LllHOSPHERE OCEANIC PLATEAUS LlTHtXPHERE ................................................... ................................................... ................................................... ................................................... ................................................... ................................................... .................................................. ................................................... ................................................... ................................................... ................................................... ................................................... ::f::$::;::;::RESIDUAL :::j:::::::::::::::::i::::::::~ Fo DUNlTE :;::;:::;::;:;::: ................. ~:.~~.~.~~.~.-.~_-.~:,~:.~ 92-94 ................... ................................................... .......................... ................................................... SLIGHTLY DEPLETED PYRCMTE 1 (minus0.1-l% alkallc liquid 62% 01 36% Opx, Cpx, Gnt 60 c La sn ml (b) (a) (c) FIG. 8. Idealized structure of oceanic lithosphere showing chemical and petrological zoning developed during partial melting and differentiation: (a) current mid-oceanic spreading centrcs; (b) oceanic plateaus and aseismic ridges developed over hot spots; and (c) postulated structure of the Archaean oceanic lithosphere. nised that the formation of new lithosphere at mid-oceanic ridges is accompanied by melting and differentiation, leading to a petrologically zoned structure. Oceanic lithosphere displays marked variations in its vertical stratification. The structure of normal lithosphere formed at mid-oceanic spreading centres during most of the Phanerozoic is shown in Fig. 8a. The basaltic crust, -7 km thick, overlies a layer of harzburgite 5-20 km thick which is underlain by lherzolite and pyrolite which have experienced successively smaller degrees of partial melting (e.g., RINGWOOD, 1982). Typical compositions of these lithologies are given in Table 6. There are also extensive oceanic regions where the basaltic crust is Table 6 CH~ICAL COMPOSITIONS Harzburgitecb) 43.6 MORB(C) 49.7 SiO2 44.6 Al203 4.3 0.6 16.4 cao 3.5 0.5 13.1 MgO 38.1 46.5 10.1 FeO 9.1 8.8 8.0 Na20 0.4 2.0 Ti02 0.7 88.2 (a) iW (c) 90.5 69.2 (Sun, 1982) (Michael and Bonatti, 1985) (corrected for NiO, MnO and CrO) (Green, et al., 1979) (primitive MORB) FeO* = Fe0 (total) + NiO + MnO = CrO 2098 A. E. Ringwood they are accompanied by substantial density contrasts which may be of considerable geodynamic significance. The sequence of phase transformations displayed by pyrolite with increasing depth is shown in Fig. 4. The zeropressure densities of the various mineral assemblages can be accurately calculated from those of their individual components and are also shown in Fig. 4. Mineral assemblages and zero-pressure densities for subducted basaltic oceanic crust are shown in Fig. 9 which is based on experiments by IRIFUNE and RINCWOOD (1987a). These authors showed that subducted oceanic crust of MORB composition would remain about 0.1-0.2 g/cm3 denser than pyrolite down to 650 km and below about 750 km. However, between about 650 and 780 km, a large amount ofgamet remains stable in the MORB composition, whereas pyrolite has essentially transformed to the denser perovskite-magnesiowiistite assemblage. Because of the persistence of garnet, the MORB composition is actually about 0.1 g/cm3 less dense than pyrolite over this depth interval. RINGWOOD(1990b) noted that oceanic crust is probably subjected to partial melting during subduction, resulting in the elimination of excess SiOz (stishovite) and that it would be more realistic to calculate the density of subducted oceanic crust at depths below about 500 km on the basis of stishovitefree mineral assemblages. The densities of this modified oceanic crust and pyrolite are shown as a function of depth much thicker than normal and is believed to be underlain by thicker sections of harzburgite. These include oceanic hot spots, oceanic plateaus, and passive continental margins (MCKENZIE and BICKLE, 1988). In these regions, the basaltic oceanic crust may be as much as 20 km thick (Fig. 8b). Mantle temperatures are believed to have been 200-300°C higher in the Archaean than at present (MCKENZIE and WEISS, 1975; BICKLE, 1978, 1986). Higher mantle temperatures would be expected to have produced more advanced degrees of partial melting, leading to copious generation of komatiitic magmas and a thickened oceanic crust (BICKLE, 1978, 1986). The refractory residuum remaining after extraction of these magmas would have consisted primarily of dunites or orthopyroxene-poor harzburgite containing ohvine of about Mg*94composition (GREEN, 1975). The structure expected for Archaean oceanic lithosphere is illustrated in Fig. 8c, The formation of chemically differentiated oceanic plates is a major geodynamic process which may have involved a large proportion of the volume of the mantle. As differentiated oceanic lithosphere plates descend into the mantle, they experience a complex series of successive phase transformations with increasing depth. The sequences and characteristics of these phase transformations are considerably influenced by the differing chemical compositions of the basalt, harzburgite, dunite, and depleted pyrolite layers (Fig. 8). In some cases, DENSITY DEPTH g/cm3 BASALTIC OCEAN CRUST km 100 3.50 CLINOPYROXENE 200 \ 300 / 400 \ GARNET ST 500 600 700 LiIF SH- I 600 Mg PEROVSKITE 0.1 0.2 0.3 + 0.4 VOLUME 0.5 0.6 I I, Ca PEROVSKITE 0.7 ITE I 0.8 0.9 1.0 FRACTION FIG. 9. Mineral assemblages and (zero-pressure) densities displayed by subducted basaltic (MORB) oceanic crust to a depth of 850 km. 2099 Inaugural Ingerson Lecture in Fig. 10 and the corresponding density difference between pyrolite and oceanic crust is shown in Fig. 11. It is seen that elimination of stishovite would substantially enhance the buoyancy of subducted oceanic crust (relative to surrounding mantle) between depths of 650 and 780 km. Mineral assemblages and associated zero-pressure densities displayed by harzburgite to a depth of 850 km are shown in Fig. 12. It is assumed that the harzburgite follows the same geotherm as pyrolite in Fig. 4, with temperatures of - 1400°C at 400 km and - 1600°C at 650 km. Stability fields shown for mineral assemblages in Fig. 12 are based mainly on references discussed in Section 3, together with the experimental results of IRIFUNE and RINCWOOD (1987b). The density of harzburgite is compared with that of pyrolite (along the same geotherm) in Fig. 10, and the density difference between harzburgite and pyrolite as a function of depth is shown in Fig. 11. Harzburgite is seen to be 0.05-0.08 g/cm3 less dense than pyrolite to a depth of 500 km, owing to its lower Fe0 content and also to its smaller A&O3 content which is responsible, in turn, for a smaller amount of dense garnet. Harzburgite is also -0.05 g/cm3 less dense than pyrolite below 750 km because of its lower Fe0 and higher M2Si04/ MSi03 ratios (RINGWOOD, 1982). However, between depths of 650 and 700 km, harzburgite is up to 0.08 g/cm3 denser than pyrolite. This is because harzburgite fully transforms to a perovskite-magnesiotistite assemblage at 650 km, whereas the broader stability field of garnet in pyrolite extends the completion of this transformation to about 720 km. The implications of these density relationships in subducted oceanic lithosphere for processes of mantle dynamics are discussed in the following section. 7. PHASE TRANSFORMATIONS AND MANTLE DYNAMICS (a) Viscosity Stratification in the Mantle THERMALLY +0.3 - (a) Basalt c fi E +O.l r (b) HarzburgiteminusPyrolite 0.0 I SLAB I I I I I I : I I I/!\ A : _______ I I I I -0.1 - I -0.2 I I 300 I 500 400 1 I 700 600 I 800 DEPTH km FIG. 11. Density differences (obtained from Fig. IO), between a modified (stishovite-free) MORB composition and pyrolite, and between harzburgite and pyrolite as a function of depth along the BROWN and SHANKLAND (198 1) geotherm. with depth in the mantle will be influenced by the successive phase transformations which occur with increasing pressure. KARATO (1989) demonstrated the existence of systematic relationships between (normalized) temperatures and flow stresses for families of dense oxides possessing crystal structures relevant to the Earth’s mantle. These relationships suggested that intrinsic viscosities of crystal structures decreased in the sequence: garnet > spine1 2 olivine 2 perovskite > rocksalt. Karat0 noted that this sequence is consistent with observations of the relative strengths of these minerals in deformed rocks and high-pressure experimental charges. He drew attention to some geophysical implications of these observations, including a high viscosity for the Transition Zone as compared to the Upper Mantle. MEADE Crystal structure has an important influence on the rheology of solids. Accordingly, it is likely that variation of viscosity I EQUlLlSFiATED minus Pyrolite and JEANLOZ strengths of the principal (1990) determined silicate minerals HARZBURGITE DEPTH the relative of the mantle by IENSITY g/cm3 THERMALLY EQUILIBRATED SLAB 4.6 44 : OLIVINE ; Gi 5 (Mg,Fe),SiO, 42 4.0 3.32 3.8 3.6 3.4 200 - - 3.69 300 400 500 DEPTH 600 700 800 600 3.71 km 4.10 FIG. 10. Density profiles displayed by pyrolite, harzburgite, and modified MORB compositions along the geotherm of BROWN and SHANKLAND(1981) at depths between 200 and 900 km in the mantle. Densities have been calculated from experimentally determined phase assemblages in these compositions, corrected for compressibility and thermal expansion along the above geotherm (based on IRIFUNE and RINGWOOD, 1987b; RINGWOOD and IRIFUNE, 1989;and RINGWOOD, 1990b). Densities of the MORB composition below 500 km have been calculated on a stishovite-free basis as discussed in the text. Mg-PEROVSKITE 0 0.2 0.4 0.8 1.0 VOLUME FRACTION FIG. 12. Mineral assemblages and densities displayed by subducted harzburgite to a depth of 850 km. 2100 A. E. Ringwood direct measurements in a diamond anvil cell at high pressures and ambient temperature. They demonstrated that strength decreased in the order: spine1 > perovskite > olivine. in agreement with Karato’s results. They also cited literature evidence relating to the high strength of garnet, relative to other Upper Mantle minerals. MEADE and JEANLOZ(1990) concluded from these results that the Transition Zone may represent a region of high viscosity in the mantle, as compared to the Upper and Lower Mantle. This conclusion, if subsequently verified, would be of considerable geophysical significance. For example, it could have an important influence on the style of convection in the Earth’s mantle. The methodologies employed by the above workers assume that mantle viscosity is governed mainly by dislocation creep. This assumption seems well founded for the Upper Mantle (KARATO et al., 1986). The possibility that a different mechanism (diffusion-controlled creep) might predominate in the Lower Mantle should nevertheless be entertained. If so, it would imply even smaller viscosities in the Lower Mantle. WALL and PRICE (1989) calculated the self-diffusion characteristics of MgSiOj perovskite and obtained an electrical conductivity in good agreement with observed values for the Lower Mantle. The large magnitude of this conductivity suggested that the Lower Mantle might be in a superionic conductive state, which in turn implies high atomic diffusivities and would favour a relatively low viscosity for this region. A number of authors-including HAGER ( 1984), HAGER and RICHARDS (1989) GURNIS and DAVIES (1986), and DAVIES and RICHARDS (199 I)-have developed models of whole-mantle convection which depend rather critically upon the assumption that the viscosity of the mantle increases by factors of 30-100 on passing from the Transition Zone to the Lower Mantle. The evidence for this assumption is weak. Although not yet definitive, the available evidence from mineral physics suggests that the Lower Mantle may possess a lower viscosity than the Transition Zone. ALL~GRE (pers. comm., 1988) suggested that a high-viscosity Transition Zone would enhance the tendency for layered convection in the mantle. This is a potentially important effect which warrants further study. MCCULLOCH (199 1) pointed out that a high viscosity in the Transition Zone may have caused the Upper Mantle to convect independently of the remainder of the mantle during the Archaean. The Archaean crust would then have been extracted only from a reservoir shallower than 400 km. This could provide an explanation of the relatively strong geochemical “depletion signatures” that are indicated by the neodymium isotopic compositions of the oldest Archaean rocks (23.8 Ga). (b) Collision of Slabs with the 650 km Discontinuity: The Formation of Megaliths GIARDINIand WOODHOUSE(1984) concluded from a study of deep seismicity in the Tongan subduction zone that the descending slab encounters strong resistance at the 650 km discontinuity and is deflected laterally. More generally, earthquake source mechanisms show that the lower sections of slabs (below 300 km) are in a state of compressive stress (ISACKS and MOLNAR, 197 1) which would be consistent with the presence of some kind of barrier to slab penetration at 650 km. HAGER (1984) and HAGER and RICHARDS(1989) suggested that the resistance to subduction below 650 km is caused by the high viscosity of the Lower Mantle. However, as noted in the previous section, this interpretation is not supported by experimental evidence on the viscosities of mantle minerals. Nevertheless, there are two other factors which would cause the tips of slabs to experience buoyant resistance when they encounter the 650 km discontinuity. Firstly, the interiors of old slabs are likely to be a few hundred degrees cooler than surrounding mantle at 650 km (OXBURGH and TURCOTTE, 1970). Because of the negative gradient of the transition of spine1 to perovskite + magnesiowtistite (Section 3b), the spine1 + garnet field (Fig. 4) would extend 30-40 km below the 650 km discontinuity in the interior of a slab which was -400°C cooler than surrounding mantle. The lower density of the spine1 would cause a significant buoyant stress at the tip of the slab. Secondly, it was shown in Section 6 that the former basalt and harzburgite components of the slab are 0.20-0.05 g/cm3 less dense than surrounding mantle in the region between 650 and 750 km (Fig. 11). Because of the combination of these two buoyancy effects, the tip of the slab may be subjected to a resistive stress of -500 bars (0.05 GPa) as it sinks below 650 km. RINGWOOD(1982, 1986, 1989b, 1990b) and RINGWOOD and IRIFUNE (1988) have argued that the combined buoyancy stresses would cause the tip of the slab to buckle when it encounters the 650 km discontinuity. Buckling may also be accompanied by plastic thickening. If the slab were, for example, about 400°C cooler than surrounding mantle, its mean viscosity would intensify the buckling process. After a limited amount of buckling (thickening) had occurred at the tip of a subsiding plate, additional penetration through 650 km would be further impeded by the accumulation of high viscosity material at its base. Continuation of this process over an extended period, e.g., IO* years, would lead to the development of a large “megalith” of relatively cool but deformed former oceanic lithosphere, with a mean cross-sectional diameter of several hundred km (Fig. 13). The shapes of megaliths are expected to be highly variable and would be dependent upon the relative rates of subduction versus trench migration. High subduction velocities and low trench migration velocities would yield vertically elongated megaliths, whereas slow subduction and high trench migration velocities would yield megaliths with relatively large horizontal dimensions. Megaliths are believed to be comprised of a melange of former oceanic crust and harzburgite (Fig. 13). The mean density of this differentiated assemblage is slightly smaller than that of chemically equivalent pyrolite (Fig. 11).However, the megalith is likely to be a few hundred degrees cooler than the surrounding mantle, which would tend to cancel the density deficit, causing it to be neutrally buoyant. Megaliths provide “cold fingers” or heat sinks in the Lower Mantle and may generate sinking convection currents in this region. The integrity of megaliths is maintained, initially. by high viscosity arising from their relatively lower temperature. As the outer regions of megaliths are warmed, their viscosity falls and they are likely to be entrained into the convective system of the Lower Mantle. Ultimately, this process of thermal erosion would cause the entire megalith to be mixed into the Lower Mantle. Individual megaliths are thus expected to have a Inaugural lngerson Lecture island arc Calc-alkaline volcanism LITHOSPHERE Resorbtion of depleted pyrolite into Upper Mantle PYROLITE by melts derived from former basaltic crust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...! 400 km discontinuity Accretion of fertilized peridotite \ 650 km discontinuity Former basaltic crust Former harzburgite *:.:,:.:.:.:.:.:.~ ,,,.,.,.,.,.,.,. ,.,., .,.,.,. :::i:::::::::::::::::::::~::,:,:,~:.~,:, 1 /- Descending 1 convection I current t / I : FIG. 13. Model showing structure of mantle and subduction ofa cool, thick plate of differentiated oceanic lithosphere. Previous subduction during the Archaean and also more recent subduction of thickened oceanic crust (Figs. 8b, 14) have produced a gravitationally stable layer of garnetite overlying the 650 km discontinuity. The tip of a cool, thick plate experiences buoyant resistance when it penetrates this layer and encounters the seismic discontinuity at 650 km. Here, the former oceanic crust and harzburgite layers may buckle and plastically thicken to form a large melange (megalith) situated mainly below the seismic discontinuity. The megalith is a transient feature and ultimately becomes entrained in the convective regime of the Lower Mantle. The lower layer of the descending plate of sub-oceanic lithosphere becomes delaminated and resorbed into the Upper Mantle because of its inability to penetrate the gametite layer (-600-650 km) owing to the buoyancy relationships depicted in Figs. 10 and I I. Partial melting of subducted oceanic crust between 200 and 600 km causes geochemical refertilization of adjacent depleted peridotite which is entrained in the flow and becomes concentrated in a thin layer immediately overlying the gametite. This layer of refertilized peridotite ultimately provides a source region for geochemically enriched intraplate (hot-spot) basaltic magmas. transient existence, comparable with the lifetimes of their parental subduction systems. A number of geophysical observations are consistent with the existence of these structures. DZIEWONSKI and WOODHOUSE (1987) found that the degree of seismic heterogeneity at depths of -650 km is greater than at shallower (~500 km) and deeper (>800 km) depths. RINCWOOD (1982,1989b) suggested that megaliths might explain the existence of large positive gravity anomalies (-50 mgL) which extend for up to 1000 km behind subduction zones in the direction of the arc or continent (e.g., Fig. 12 in PHILLIPS and LAMBECK, 1980). CREAGER and JORDAN (1984, 1986) concluded that anomalously seismic high velocities were present at depths of 650- 1200 km below subducting plates in the Western Pacific. This has frequently been cited as evidence for direct penetration of slabs to great depths in the Lower Mantle. However, more recent seismic studies reveal a more complex 2102 A. E. Ringwood picture. FISHER et al. (1988) found that residual sphere anomalies from deep earthquakes in the Mariana slab were consistent with thickening of the slab below 650 km by a factor of 5 or more. This could be explained by the presence of a megalith. ZHOU and CLAYTON (1990) found that highly complex structures underlie the intersections of slabs with the 650 km discontinuity and that they vary considerably for different slabs. The structures appear to include interfmgering, and sub-horizontal extensions, as well as thickening of the slabs. Zhou and Clayton concluded that penetration of slabs into the Lower Mantle was impeded by a barrier at 650 km. SHEARER (199 1) investigated the depths to the 650 km discontinuity (actually at 660 km) on either side of subducted slabs in the northwest Pacific. He found that the discontinuity was depressed by -30 km, compared to its normal depth, for a distance of - 1000 km in the back-arc direction, and for a distance of -500 km in the opposite direction. The depressions could be explained in terms of the spine1 - perovskite + magnesiowiistite transition, if the mantle in these regions were a few hundred degrees cooler than average. Shearer pointed out that the implied temperature distribution is not predicted by models in which slabs slice cleanly through the discontinuity (e.g., CREAGER and JORDAN, 1986) but, models in which the subducting slab instead, “supports spreads out near 660 km. Such models include the gravitationally trapped megalith hypothesis.” CHRISTENSEN and YUEN (1984) concluded that slabs could penetrate directly into the Lower Mantle without major deformation, despite encountering substantial buoyant resistance at 650 km caused, for example, by a negative phase transition gradient and/or a change in chemical composition. However, their analysis applied only to vertically oriented slabs. Most slabs encounter the 650 km discontinuities at angles which are substantially smaller than 90 degrees. Buoyant stresses exerted on the tips of these inclined slabs seem much more likely to lead to buckling. The above scenario for the formation of megaliths has been based upon the subduction of mature oceanic plates, typically 80-120 km thick and 50-100 Ma old (at the point of subduction). Mature plates of this type possess sufficient thermal inertia to be much cooler than surrounding mantle when they reach 650 km. In contrast, relatively young (<30 Ma) and thin (~50 km) slabs possess comparatively low thermal inertia and may also be subducted more slowly than older, thicker slabs. Accordingly, their internal temperatures may approach thermal equilibrium with surrounding mantle by a depth of 650 km. Such slabs become aseismic and ductile at relatively shallow depths and may behave quite differently from old, thick slabs when they reach 650 km (RINGWOOD and IRIFUNE, 1988). Their fate is considered in the following section. discussed in Section 3). RINGWOOD (1982) initially opposed Anderson’s perched eclogite model on the basis of some reconnaissance experiments by LIU (1980). However, subsequent, more elaborate experiments by IRIFUNE and RINGWOOD (1987a) showed that subducted basalt would indeed be substantially less dense than pyrolite (RINGWOOD, 1990b; RINGWOOD and IRIFUNE, 1988) between 650 and 750 km (Fig. 11) and that gravitational trapping of subducted oceanic crust on the 650 km discontinuity deserved serious consideration. RINGWOOD and IRIFUNE (1988) suggested that the subduction of relatively young and thin oceanic plates which approached thermal equilibrium with surrounding mantle may facilitate this trapping process. Density relationships depicted in Fig. 11 show that the tip of the plate would be buoyant when it encountered the 650 km discontinuity. Moreover, it would be ductile because of elevated temperatures and transformational softening. The plate is therefore likely to be deflected forward along the 650 km discontinuity, as shown by KERR and LISTER (1987) and leading to the trapping of former basaltic crust immediately overlying the discontinuity. Subducted oceanic plateaus with crustal thicknesses of up to 20 km (Fig. 14) would be particularly susceptible to gravitational trapping because of their enhanced net buoyancy below 650 km. The most favourable conditions for gravitational trapping of subducted lithosphere probably existed during the Archaean. A section through Archaean oceanic lithosphere is displayed in Fig. 8c. Former dunitic lithologies would have been 0.12 g/cm3 less dense than pyrolite below 700 km. The thickened (-20 km) mafic or komatiitic crust would have been even more buoyant. It seems improbable that such buoyant differentiated lithosphere would enter the Lower Mantle. Rather, it would be expected to spread out as density currents along the 650 km discontinuity to form a globally encircling layer (KERR and LISTER, 1987). If the mantle were 200-300°C hotter than today, as suggested earlier, its viscosity TRANSITION ZONE Spine1 + Garnet (c) Gravitational Trapping of Oceanic Crust on the 650 km Discontinuity ANDERSON (1979, 1980) proposed that subducted basaltic crust would be less dense than pyrolite immediately below 650 km and would therefore be trapped during subduction to form a “perched eclogite layer.” His proposal formed part of a more complex petrogenetic and structural model which envisaged a layer of “piclogite” between 250 and 650 km (as Transformational softening LOWER MANTLE Perovskite + Mw AP Pyrolite minus Garnetite = 0.18 g/cm 3 dp Pyrolite minus Harzburgite = 0.06 g/cm 3 FIG. 14. Encounter of a thin, thermally equilibrated plate with the 650 km discontinuity, Ieading to buoyant trapping of former oceanic crust (garnetite) immediately above the discontinuity. Inaugural Ingerson Lecture would be 2 or 3 orders of magnitude lower. This would facilitate the disposition of the dunitic and basaltic components to form a stably stratified structure as shown in Fig. 15, ultimately isolating the convective systems of the Upper and Lower Mantle (RINGWOOD, 1990b; RINGWOOD and IRIFUNE, 1988). Convection currents in the Upper and Lower Mantles would erode this differentiated barrier via entrainment at its upper and lower surfaces, whilst it would be replenished by capture of continually subducting lithosphere. This dynamic equilibrium may have led to a steady-state thickness for the barrier and to convective rehomogenisation of the mantle regions on either side. The layer of garnetite-facies former oceanic crust trapped above 650 km may have had a thickness in the vicinity of 50 km. (d) The Style of Mantle Convection The body forces which drive the plate into the mantle arise from the higher densities of the former oceanic crust and underlying depleted harzburgite layers, as compared to surrounding mantle. The lower lithosphere of depleted pyrolite (Fig. 13) is warmer and more ductile (BODINE et al., 1981; KIRBY, 1980) whilst the density difference between the lower lithosphere and underlying asthenosphere is relatively small. RINGWOOD( 1982) suggested that because of the small density difference and its ductile behaviour, the lowermost zone of the descending lithosphere may not move coherently with the upper lithosphere and may gradually become resorbed into the asthenosphere. Thus, the effective width of the slab decreases as it descends, and, by 650 km, it consists mainly of the former basaltic and harzburgite components. Mid-ocean ridge basalts (MORBs) are believed to be produced by partial melting of pyrolite which has experienced episodic extraction of small amounts of alkalic magmas, highly enriched in incompatible elements, over a period exceeding lo9 years (GAST, 1968). Resorbtion and recycling of the lower oceanic lithosphere into the Upper Mantle during subduction appears to provide a promising means of generating MORB source regions (RINGWOOD, 1982). The chemical and isotopic compositions of basalts erupted ARCHAEAN MANTLE km PYROLITE: SpineCGarnet facies 2103 from the mantle provide a wide range of constraints on the existence and lifetimes of geochemical reservoirs in this region. These topics have been extensively discussed by RINGWOOD ( 1982, 1990b) and KESSON and RINGWOOD (1989) in the context of the model of mantle dynamics depicted in Fig. 13. Actually, this model was formulated with these constraints very much in mind. It proposes a hybrid form of convection in which the Upper and Lower Mantles convect essentially independently, and are separated by a garnetite boundary layer within the Transition Zone. Superimposed on this bimodal regime, limited transfer of material from the Upper Mantle to the Lower Mantle occurs in the form of old, thick oceanic plates that penetrate the 650 km discontinuity, forming megaliths which are ultimately entrained into the Lower Mantle convective system. An equivalent amount of material must be transferred from Lower Mantle to Upper Mantle, perhaps as penetrative diapirs overlying rising convection currents in the Lower Mantle. An increasing array of geochemical evidence appears to be more compatible with a hybrid convective system of this type than with more simplistic models of exclusively layered or exclusively whole-mantle convection. For example, trace element and isotopic mass balances indicate that the continental crust was derived by differentiation from a mantle reservoir possessing a substantially larger volume than the mantle regions above a depth of 650 km, but which was much smaller than the volume of the entire mantle (e.g., ZINDLER and HART, 1986; GALER et al., 1989). These studies suggest that the Upper Mantle possesses a highly depleted geochemical signature, whereas the Lower Mantle is only modestly depleted. O’NIONS (1987) and GALER et al. (1989) concluded that highly incompatible elements such as U, Th, and Pb are efficiently extracted from the Upper Mantle and transferred to the crust on a comparatively short timescale (-600 Ma). Replenishment of these elements occurs by injection of less depleted material from the Lower Mantle into the Upper Mantle accompanied by rapid convective re-homogenization of the Upper Mantle. O’NIONSand OXBURGH(1983) inferred from a comparison of the helium and heat fluxes at midocean ridges that whereas most of the radiogenic heat produced in the Lower Mantle was transferred to the Upper Mantle, there was some kind of barrier to the transport of the helium produced by radioactive decay. They suggested that the imbalance might be explained by the presence of a barrier near 650 km, which impeded the transport of helium, but not of heat. The question of slab-mantle coupling PYROLITE: Perovskitite facies iOOMg/(Mg+Fe)=88 FIG. 15. Possible petrological structure of the mantle during the Archaean. Gametite layer represents subducted former (komatiitic) oceanic crust which has been buoyantly trapped above the 650 km discontinuity. Dunite (Fo& complementary to the komatiite is trapped in the layer immediately below 650 km. This stably stratified structure causes the Upper and Lower Mantle to convect independently. A sinking slab would entrain adjacent mantle material by viscous coupling, causing it to follow the path of the slab. In the model depicted in Fig. 13, it is assumed that the width of the entrained region is quite narrow, no more than 10-20 km. In the mantle convection models of DAVIESand RICHARDS (1991), the entrained region extends several hundred kilometers on either side of the slab. In these models, the slab drives the circulation of a large region of adjacent mantle. RICHARDS and DAVIES ( 1989) pointed out that under these circumstances, the integrated negative thermal buoyancy of the sinking region would completely overwhelm the chemical buoyancy effects (Fig. 11) associated with the narrow layers 2104 A. E. Ringwood of former oceanic crust and harzburgite, and that the segregation of former basaltic crust on the 650 km discontinuity illustrated in Fig. 13 would not be possible. Their argument is correct if the style of mantle convection is as they propose. This, however, is a matter of debate. These authors assume that the rheology of the slab-mantle system can be interpreted primarily in terms of temperature-dependent Newtonian viscosities. However, several observations suggest that the actual rheology of the mantle may be considerably more complicated and that deformation might therefore be highly localised. HELFRICH et al. (1989) showed that the upper boundary of the slab to depths of -350 km is defined by a discontinuity at which seismic waves are reflected and converted. A velocity contrast of 5-10s through a layer - lo-20 km thick is indicated. HELFRICHet al. ( 1989) concluded that preferred orientation of olivine within this layer was one of the factors contributing to the velocity anomaly. ANDO et al. (1989) and OHKURA et al. (1990) identified an analogous seismic discontinuity defined by converted waves at a boundary about 35-40 km below the main Wadati-Benioff zone. This lower discontinuity coincides with a second planar set of earthquake epicentres, analogous to the Wadati-Benioff zone. These parallel seismic discontinuities may define the rigid core of the subducted slab, bounded by narrow zones within which displacements are confined. It is difficult to conceive of these structures existing in a mantle deforming according to the model of DAVIESand RICHARDS(1991). A suite of strongly sheared garnet peridotite xenoliths found in kimberlite pipes in Africa, Siberia, and the USA provides direct evidence of mantle rheology. The xenoliths have been deformed at exceptionally high strain-rates amounting to lofold elongation occurring in minutes or hours (GOETZE, 1975; MERCIER, 1979). Moreover, deformation occurred at remarkably high temperatures, frequently in the vicinity of 1400-16OO”C, and at pressures of 4-7 GPa. During deformation, these garnet peridotites have been invaded by silicate melts causing them to become “refertilized.” USSON and RINGWOOD(1989) concluded that these rocks probably represented actual samples of Wadati-Benioff zones and that the combined high temperatures and high strain rates had been caused by episodic shearing associated with Wadati-Benioff zones. These samples testify to the occurrence of deformation processes deep in the Upper Mantle which obey flow laws differing rather profoundly from those assumed by DAVIES and RICHARDS(199 1). It appears that some essential element is lacking in many current treatments of mantle convection based on computer modelling. The missing element may be the heating caused by viscous dissipation along the boundaries of the sinking slab (SCHMELING,1989). This heating would cause a considerable decrease in viscosity together with a parallel increase in strain-rate, thereby concentrating the heating along thin boundary layers. Schmeling found that these effects caused the formation of low-viscosity layers on either side of the sinking slab, with the result that the motion of the slab became largely decoupled from the neighbouring mantle. These pro- 5Members of the Intraplate BasalticAssociation erupted in oceanic regions are often called “Ocean Island Basalts” or “OIBs.” cesses may produce the seismic discontinuities on either side of the slab, as discussed above, and also the style of deformation displayed by high-temperature peridotites. Moreover, they provide mechanisms for generating narrow slabs which are mechanically decoupled from surrounding mantle, as assumed in Fig. 13. 8. HOT SPOTS, PLUMES, AND MANTLE DYNAMICS An important class of magma is erupted through the lithosphere, both in oceanic and continental regions, and can be referred to as the “Intraplate Basaltic Association” (IBA).” Most IBA suites are characterized by enrichments of incompatible elements, whereas MORBs are depleted in these elements. The IBA are produced by widely varying degrees of partial melting and include picrites, tholeiites, alkali basalts, basanites, and nephelinites. Experimental investigations (e.g., GREEN and RINGWOOD, 1967) show that MORBs and the IBA are both produced by partial melting of peridotitic source regions, and that their depleted/enriched characters are inherited from these source regions. If we compare magmas of both types formed by similar degrees of partial melting (e.g., MORBs and Hawaiian tholeiites), the former are depleted in highly incompatible elements (e.g., Cs, Ba, U) by factors of lo-20 as compared to the latter. It was noted earlier (Section 2d; Table 4) that the formation and fractionation of magmas in the presence of MgSi03 and Casio3 perovskites would cause certain characteristic geochemical signatures in the resultant melts. There is no evidence of these signatures in either MORBs or in the IBA. On the contrary, the geochemical characteristics of these magma series are consistent with source-region lithologies dominated by garnet, pyroxene(s), olivine, and (silicate) spinel. We conclude, therefore, that the processes which produce these geochemical characteristics occurred not in the Lower Mantle, but rather in the Upper Mantle and/or the Transition Zone. There is an emerging consensus that the petrogenesis of the IBA is related to the subduction of oceanic crust, which itself consists predominantly of MORBs (e.g., HOFMANNand WHITE, 1980, 1982; CHASE, 198 1). The processes which caused enrichment of incompatible elements in the IBA source region along with characteristic isotopic signatures are widely debated. According to a model proposed by RINGWOOD (1982, 1990b) and illustrated in Fig. 13, bodies of subducted oceanic crust experienced small degrees of partial melting under hydrous conditions at depths between 200 and 600 km. The resultant melts, enriched in incompatible elements, reacted with the adjacent depleted mantle, thereby transferring their isotopic and trace element signatures into these peridotitic protoliths. A significant proportion of this “refertilized peridotite” may have become trapped immediately above the garnetite boundary layer at 650 km (Fig. 13) where it formed a long-lived (_ 1O9a) reservoir. The subcontinental lithosphere may have provided a second environment in which refertilized peridotite was trapped, as illustrated by RINGWOOD(1990b; Fig. I). This material would possess the capacity to produce alkaline IBA magmas if subjected to further episodes of small-degree partial melting. Isotopic compositions of Nd, Hf, Sr, and Pb in the IBA suggest that long time intervals (e.g., lo’-lo9 a) may have elapsed between initial formation of their refertilized peri- 2105 Inaugural Ingerson Lecture dotitic source regions and the subsequent reactivation processes which caused partial melting and actually produced IBA magmas at the Earth’s surface. However, estimates of the time intervals are uncertain because of the possibility that isotopic evolution (particularly in the Pb and Sr systems) may have been affected by sediment contamination and hydrothermal alteration of oceanic crust (e.g., CHAUVEL et al., 199 1). Despite these chronological uncertainties, the evidence previously cited that IBA magmas are formed by partial melting of (refertilized) peridotite firmly indicates a two-stage petrogenetic process separated by a significant time interval. According to the model shown in Fig. 13, the garnetite layer overlying the 650 km discontinuity provides a thermal boundary layer, which is overlain in turn by a thin layer of fertilized former peridotite, representing entrained material formerly adjacent to subducted oceanic crust. It is envisaged that the garnetite layer was subsequently heated by rising convection currents from the Lower Mantle, causing the overlying fertilized peridotite to become buoyant and to ascend as plumes. During ascent, partial melting occurred, leading to the formation of geochemically enriched magmas which erupted from hot spots. Plumes,fiom the core-mantle boundary? According to a currently popular hypothesis, IBA volcanism is caused by plumes which ascend directly from a thermal boundary layer surrounding the core. The enriched geochemical and isotopic characteristics of hot-spot volcanism have been attributed to relative concentration of dense bodies of former oceanic crust by sinking into and within this boundary layer (e.g., HOFMANN and WHITE, 1982; DAVIES, 1990; DAVIES and RICHARDS, 199 1). This scenario encounters certain difficulties: 9 Oceanic crust subducted l-2 Ga ago would have possessed depleted geochemical signatures. Any viable process of remelting of ancient oceanic crust which had collected near the core-mantle boundary should provide a great preponderance of geochemically depleted magmas. However, tholeiites formed by similar degrees of partial melting to MORBs display lo-20-fold enrichment of highly incompatible elements. DAVIES and RICHARDS ( 199 1) suggested that oceanic crust may contain locally enriched “inclusions” which would be preferentially melted because of lower solidus temperatures. This explanation is not viable since the temperatures of ascending plumes exceed those of surrounding mantle and would therefore lead to extensive degrees of partial melting, exceeding those which occur beneath mid-oceanic ridges. Under these conditions, the former depleted MORB component of plumes should be almost totally consumed during melting, and the compositions of plume tholeiites and pi&es should therefore accurately reflect this depleted character. ii) It was demonstrated earlier that the geochemical characteristics of the IBA were produced by multistage petrogenic processes which occurred exclusively in the Upper Mantle and Transition Zone and which seem to have extended over considerable periods oftime (e.g., IO’-lo9 a). It is difficult to reconcile this history with a model which postulates an ultimate source for the IBA at the core-mantle boundary. According to the latter model this source comprised a heterogeneous mixture of former oceanic crust and peridotite (? sediments) which had been stored as a closed system for - lo9 years at very high temperatures without experiencing significant chemical exchange between its separate components. It appears doubtful that the complex geochemical and isotopic characteristics for the IBA could have been generated from this heterogeneous source material within the few million years (GRIFFITHS and CAMPBELL, 1990) that it took the plumes to ascend from the 650 km discontinuity to the lithosphere. iii) The most widely distributed IBA members belong to the alkaline association, which includes alkali basalts, basanites, and nephelinites. These occur extensively in typical hot-spot environments such as Hawaii, where they are allegedly derived from plumes claimed to be ascending from the core-mantle boundary. However, numerous and widely distributed occurrences of alkaline rocks which are petrologically, geochemically, and isotopically indistinguishable from these major hot-spot occurrences are erupted through lithosphere above currently active subduction zones, or where subduction has recently ceased (RINGWOOD, 1982, 1990b). These latter IBA members are clearly derived from sources situated within the Upper Mantle. Few would argue in favour of ultimate source regions near the core-mantle boundary. It does not seem reasonable to attribute these petrologically and geochemically identical alkaline rock types to two fundamentally different categories of source regions. iv) Recent studies of the melting point of iron at high pressures and of the depression of this melting point by dissolved solutes (e.g., oxygen), as reviewed in Section 4c, do not support the existence of a strong thermal boundary layer overlying the core-mantle boundary, which is believed to be essential for the generation of plumes (e.g., STACEY and LOPER, 1983). 9. CONCLUDING REMARKS An important objective of the Earth Sciences is to explain the radial variation of physical properties throughout the mantle in terms of the chemical compositions and mineral assemblages of its various domains. Major progress towards achieving this objective has been made in recent years and the principal results are reviewed in the first part of this paper. The stage has now been set for progress towards achieving a second major objective-an understanding of the dynamical behaviour of the mantle, including the convective system which drives plate motions. Currently, this objective is impeded somewhat by the absence of an adequate understanding of mantle rheology. However, geochemical studies have been of considerable importance in defining the existence of long-lived, chemically distinct reservoirs in the mantle which place powerful empirical constraints on the style of mantle convection. The model of mantle dynamics proposed in this paper represents an attempt to reconcile geochemical evidence on mantle reservoirs with recent advances in understanding the physical and mineralogical constitution of the mantle. It makes a number of assumptions about possible rheologjcal behaviour of the mantle which have yet to be tested. A basic A. E. Ringwood 2106 issue in mantle dynamics is the extent to which mantle convection homogenises pre-existing chemical heterogeneities (e.g., KELLOGG and TURCOTTE, 1986) or, alternatively, the extent to which chemical heterogeneities themselves modulate or control the convective circulation of the mantle. It has been shown in this and earlier papers (e.g. RINGWOOD, 1982) that large-scale chemical heterogeneities caused by petrological differentiation and phase charges in the mantle can give rise to density contrasts at least as large as those which are believed by many geophysicists to drive mantle convection. Currently, it does not seem possible to decide, CIpriori, whether thermally driven convection would overwhelm the effect of chemical differentiation or vice versa. These two scenarios, however, have quite different implications with regard to the formation and maintenance of geochemical reservoirs. There are grounds for optimism that progress in understanding these latter topics may soon provide firm constraints which govern the style of mantle convection. Acknowledgments-The author is greatly indebted to Dr. I. Jackson, Dr. S. E. Kesson, Dr. M. McCulloch, Dr. M. Drury, Dr. S. Cox, and Dr. B. Kennett who read the manuscript carefully and offered numerous critical and constructive suggestions. Editorial handling: G. Faure REFERENCES AGEEC. B. and WALKERD. (1988a) Static compression and olivine flotation in ultrabasic silicate liquid. J. Geophys. Rex 93, 34373449. AGEEC. B. and WALKERD. (1988b) Mass Balance and phase density constraints on early differentiation of chondritic mantle. Earth Planet. Sci. Lett. 90, 144-156. AKAOGIM. and AKIMOTOS. (1977) Pyroxene-garnet solid-solution equilibria in the systems M&Si4012-Mg3Al&~0,Z and Fe4Si4012Fe3Al$i3012 at high pressures and temperatures. Phys. Earth Planet. Intl. 15, 90- 106. AKA~GI M. and AKIMOTOS. (1979) High-pressure phase equilibria in a garnet Iherzolite, with special reference to Mg*+-Fe*+ partitioning among constituent minerals. Phys. Earth Planet. Intl. 19, 31-51. AKA~GI M., Ross N. L., MCMILLANP., and NAVROTSKYA. (1984) The Mg,Si04 polymorphs (olivine, modified spine1 and spinel)thermodynamic properties from oxide melt solution calorimetry, phase relations, and models of lattice vibrations. Amer. Mineral. 69,499-5 12. A~~ccr M., NAVROTSKYA.. YAGI T., and AKIMOTOS. (1987) Pyroxene-garnet transformation: Thermochemistry and elasticity of garnet solid solutions, and application to a pyrolite mantle. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANI and Y. SYONO), pp. 251-260. American Geophysical Union. Washington, DC. AKA~CI M., ITO E., and NAVROTSKYA. (1989) Olivine-modified spinel-spine1 transitions in the system Mg$iO,-Fe2Si04: Calorimetric measurements, thermochemical calculation, and geophysical application. J. Geophys. Res. 94, 15,671-15,685. AKIMOTOS. (1972) The system MgO-FeO-Si02 at high pressures and temperatures-phase equilibria and elastic properties. Tectonophysics 13, 161-187. AIUMOTOS. (I 987) High-pressure research in geophysics: Past, present and future. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANIand Y. SYONO),pp. l-l 3. American Geophysical Union, Washington, DC. AKIMOTOS. and FUJISAWAH. (1966) Olivine-spine1 transitions in system Mg,Si04-Fe2Si04 at 800°C. Earth Planet. Sci. Lett. 1,237240. AKIMOTOS. and FUJISAWAH. (1968) Olivine-spine1 solid solution equilibria in the system Mg,Si04-Fe$iO.+ J. Geophys. Res. 73, 1467-1479, AKIMOTOS., MATSUIY., and SYONOY. (1976) High-pressure crystal chemistry of orthosilicates and the formation of the mantle transition zone. In Physics and Chemistry of Minerals and Rocks (ed. R. G. J. STRENS),pp. 327-363. J. Wiley, London. AKIMOTOS., YAGI T., and INOUEK. (1977) High temperature-pressure phase boundaries in silicate systems using in-situ X-ray diffraction. In High-Pressure Research: Applications in Geophysics (eds. M. MANGHNANIand S. AKIMOTO),pp. 582-602. Academic Press, New York. ALLBGREC. J. (1982) Chemical Geodynamics. Tectonophysics 81, 109-132. ANDERSE. and GREVESSEN. (1989) Abundances of the elements: meteoritic and solar. Geochim. Cosmochim. Acta 53, 197-214. ANDERSOND. L. (1977) Composition of the mantle and core. Ann. Rev. Earth Planet. Sci. 5, 179-202. ANDERSOND. L. (1979) The upper mantle: Eclogite? Geophys. Res. Lett. 6, 433-436. ANDERSOND. L. (1980) Chemical stratification of the mantle. J Geophys. Rex 84, 6297-6298. ANDERSOND. L. (1983) Chemical composition of the mantle. J. Geophys. Rex (suppl.) 88, B4 I-B52. ANDERSOND. L. and BASSJ. D. (1986) Transition region of the Earth’s Upper Mantle. Nature 320, 32 l-328. ANDO M., KANESHIMAS., OHTAKIT., and OHKURAT. (1989) S to P converted phase from the lower plane of the double-planed deepseismic zone. Eos 70, 1228. ARNDT N. T. (1986) Komatiites: A dirty window to the Archaean mantle. Terra Cognita 6, 59-66. BASS J. D. and ANDERSOND. L. (1984) Composition of the Upper Mantle: Geophysical tests of two petrological models. Geophys. Rex Lett. 11, 237-230. BASSETTW. A. and MING L. (1973) Disproportionation of Fe$iO, to 2Fe + SiOZ at pressures up to 250 kilobars and temperatures up to 3000°C. Phys. Earth Planet. Intl. 6, 154-160. BASSETTW. A., TAKAHASHIT., and STOOKP. (1967) X-ray diffraction and optical observations of crystalline solids up to 300 kb. Rev. Sci. Instrum. 38, 37-42. BICKLEM. J. (1978) Heat loss from the Earth: a constraint on Archaean tectonics from the relation between geothermal gradients and the rate of .elate Droduction. Earth Planet. Sci. Lett. 40. 3Ol. 315. BICKLEM. J. (1986) Implications of melting for stabilisation of the lithosphere and heat loss in the Archaean. Earth Planet. Sci. Lett. SO,3 14-324. BINAC. (199 1) Internally consistent mineralogical interpretation of the 400 km discontinuity. J. Geophys. Res. (in press). BINA C. and WOOD B. J. (1987) The olivine-spine1 transition: Experimental and thermodynamic constraints and implications for the nature ofthe 400 km discontinuitv. J. Geoohvs. - _ Res. 92.48534866. BIRCH F. (1952) Elasticity and constitution of the Earth’s interior. J. Geophys. Rex 57, 227-286. BODINE J. H., STECKLER M. S., and WATTS A. B. (198 1) Observations of flexure and the rheology of the oceanic lithosphere. J. Geophys. Res. 86,3695-3707. BOEHLERR., VON BARGENN., and CHOPELASA. (1990) Melting, thermal expansion, and phase transitions of iron at high pressures. J. Geophys. Res. 95, 21,731-21,736. BONATTIE., OTTONELLOG., and HAMLYNP. (1986) Peridotites from the island of Zarbargad (St. John), Red Sea: Petrology and geochemistry. J. Geophys. Res. 91, 599-631. BROWNJ. M. and SHANKLANDT. J. (198 1) Thermodynamic properties in the Earth and determined from seismic profiles. Geophys. J. Roy. Astron. Sot. 66, 579-596. BUKOWINSKI M. and WOLFG. (1990) Thermodynamically consistent decompression: Implications for lower mantle composition. J. Geophys. Res. 95, 12,583-12,593. BULLENK. E. (1949) Compressibility-pressure hypotheses and the Earth’s interior. Mon. Not. Roy. .4stron. Sot. Geophys. Suppl. 5, 355-368. CHASEC. G. (198 1) Oceanic island Pb: Two-stage histories and mantle evolution. Earth Planet. Sci. Lett. 52, 277-284. Inaugural Ingerson Lecture CHAUVEL C., HOFMANNA. W., and VIDAL P. (1991) HIMU-EM, the French Polynesian connection. Earth Planet. Sci. Lett. (in press). CHOPELASA. and BOEHLERR. (1989) Thermal expansion measurements at very high pressure, systematics, and a case for a chemically homogeneous mantle. Geophys. Res. Lett. 16, 1347-l 350. CHRISTENSEN U. and YUEND. (1984) The interaction of a subducting lithospheric slab with a chemical or phase boundary. J. Geophys. Res. 89,4389-4402. CLARK S. P. and RINGWOODA. E. (1964) Density distribution and constitution of the mantle. Rev. Geophys. 2, 35-88. CREAGERK. C. and JORDANT. H. (1984) Slab penetration into the lower mantle. J. Geophys. Res. 89, 303 l-3049. CREAGERK. C. and JORDANT. H. (1986) Slab penetration into the lower mantle beneath the Mariana and other island arcs of the northwest Pacific. J. Geoohvs. Res. 91. 3573-3589. CUMMINSP., KENNETTB.,*BGwANJ., and BOSTOCKM. (1991) The 530 km discontinuity? Geophys. J. (in press). DAVIESG. ( 1990) Mantle plumes, mantle stirring and hotspot chemistry. Earth Planet. Sci. Lett. 99, 94-109. DAVIESG. and RICHARDSM. (1991) Mantle convection. J. Geol. (in press). DUFFVT. S. and ANDERSOND. L. (1989)Seismic velocities in mantle minerals and the mineralogy of the upper mantle. J. Geophys. Res. 94, 1895-1912. DZIEWONSKIA. M. and ANDERSOND. L. (1981) Preliminary reference Earth model. Phys. Earth Planet. Intl. 25, 297-356. DZIEWONSKIA. and WOODHOUSEJ. (1987) Global images of the Earth’s interior. Science 236, 37-48. DZIEWONSKIA. M., HALESA. L., and LAPWOODE. R. (1975) Parametrically simple Earth models consistent with geophysical data. Phys. Earth Planet. Ml. 10, 12-48. FISHERK., JORDANT., and CREAGERK. (1988) Seismic constraints on the morphology ofthick slabs. J. Geophys. Res. 93,4773-4783. FREY F. A., SVEN J., and STOCKMANH. (1985) The Ronda high temperature peridotite: Geochemistry and petrogenesis. Geochim. Cosmochim. Acta 49, 2469-249 1. GALERS., GOLDSTEINS., and O’NIONSK. (1989) Limits on chemical and convective isolation in the Earth’s interior. Chem. Geol. 75, 257-290. GAST P. W. (1968) Trace element fractionation and the origin of tholeiitic and alkaline magma types. Geochim. Cosmochim. Acta 32, 1057-1086. GIARDINID. and WOODHOUSEJ. W. (1984) Deep seismicity and modes of deformation in Tonga subduction zone. Nature 307, 505-509. GIVENJ. W. and HELMBERGERD. V. (1980) Upper mantle structure of northwestern Eurasia. J. Geophys. Res. 85, 7 183-7 194. GOETZEC. (1975) Sheared lherzolites from the point of view of rock mechanics. Geology 3, 172-I 73. GRAND S. and HELMBERGERD. V. (1984) Upper mantle shear structure of North America. Geophys. J. Roy. Astron. Sot. 76, 399-438. GREEN D. H. (1975) Genesis of Archaean peridotitic magmas and constraints on Archaean geothermal gradients and tectonics. Geology3, 15-18. GREEN D. H. and LIEBERMANNR. C. (1976) Phase equilibria and elastic properties of a pyrolite model for the oceanic upper mantle. Tectonophysics 32, 6 I-92. GREEN D. H. and RINGWOODA. E. (I 963) Mineral assemblages in a model mantle composition. J. Geophys. Res. 68,937-945. GREEN D. H. and RINGWOODA. E. (1964) Fractionation of basalt magmas at high pressures. Nature 201, 1276-1279. GREEN D. H. and RINGWOODA. E. (1967) The genesis of basaltic magmas. Contrib. Mineral. Petrol. 15, 103-190. GREEN D. H. and RINGWOODA. E. (1970) Mineralogy of peridotitic compositions under upper mantle conditions. Phys. Earth Planet. Intl. 3, 359-37 I. GREEN D. H., HIBBERSONW. O., and JAQUESA. L. (1979) Petrogenesis of mid-ocean ridge basalts. In The Earth: Its Origin, Structure and Evolution (ed. M. W. MCELHINNEY),pp. 265-299. Academic Press, London. GRIFFITHS R. and CAMPBELLI. (1990) Stirring and structure in mantle plumes. Earth Planet. Sci. Lett. 99, 66-78. 2107 GRUAU G., CHAUVELC., ARNDT N. T., and CORNICHETJ. (1990) Aluminium depletion in komatiites and garnet fractionation in the early Archaean mantle: Hafnium isotopic constraints. Geochim. Cosmochim. Acta 54, 3095-3 10 1. GURNIS M. and DAVIESG. (1986)Numerical models of high Rayleigh number convection in a medium with depth-dependent viscosity. Geophys. J. Roy. Astron. Sot. 85, 523-541. GUTENBERGB. (1958) Velocity of earthquake waves in the Earth’s mantle. Trans. Amer. Geophys. Union 39, 486-489. GUTENBERGB. (1960) The shadow of the Earth’s core. J. Geophys. Res. 65, 1013-1020. GUYOT F., MAD~N M., PEYRONNEAUJ., and POIRIERJ. P. (1988) X-ray microanalysis of high-pressure/high-temperature phases synthesized from natural olivine in a diamond-anvil cell. Earth Planet. Sci. Lett. 90, 52-64. GUYOT T., PEYRONNEAUJ., and POIRIERJ. P. (1988) TEM study of high-pressure reactions between iron and silicate perovskites (abstr.). Chem. Geol. 70, 61. GWANMESIAG., RIGDEN S. M., JACKSONI., and LIEBERMANN R. C. (1990) Pressure dependence of elastic wave velocities for pMg,SiO, and the composition of the Earth’s mantle. Science 250, 794-797. HAGERB. ( 1984) Subducted slabs and the geoid: constraints on mantle rheology and flow. J. Geophys. Res. 89,6003-60 15. HAGERB. and RICHARDSM. ( 1989) Long-wavelength variations in Earth’s geoid physical models and dynamical implications. Phil. Trans. Roy. Sot. London A328, 309-327. HALESA. L., MUIRHEADK., and RYNN J. (1980) A compressional velocity distribution for the upper mantle. Tectonophysics 63, 309348. HELFRICHG., STEINS., and WOODB. (1989) Subduction zone thermal structure and mineralogy and their relationship to seismic wave reflections and conversions at the slab-mantle interface. J. Geophys. Res. 94, 753-764. HELMBERGERD. V. and ENGEN G. R. (1974) Upper mantle shear structure. J. Geophys. Res. 79, 4017-4028. HERZBERGC. T. and O’HARAM. J. (1985) Origin of mantle peridotite and komatiite by partial melting. Geophys. Res. Lett. 12, 541544. HILL R. and JACKSONI. (1990) The thermal expansion of ScAlOra close silicate perovskite analogue. Phvs. Chem. Mineral. 17, 8996. HOFMANNA. W. (1988) Chemical differentiation of the Earth: The relationship between mantle, continental crust and oceanic crust. Earth Planet. Sci. Lett. 90, 291-3 14. HOFMANNA. W. and WHITE W. M. (1980) The role of subducted oceanic crust in mantle evolution. Carnegie Inst. Washington Yearb. 79,477-543. HOFMANNA. W. and WHITE W. M. (1982) Mantle plumes from ancient oceanic crust. Earth Planet. Sci. Lett. 57, 42 l-436. IRIFUNET. (1987) An experimental investigation of the pyroxenegarnet transformation in a pyrolite composition and its bearing on the constitution of the mantle. Phys. Earth Planet. Ml. 45, 324-336. IRIFUNET. and RINGWWD A. E. (1987a) Phase transformations in primitive MORB and pyrolite compositions to 25 GPa and some geophysical implications. In High Pressure Research in Geophysics (eds. M. MANGHNANIand Y. SYONO),pp. 231-242. American Geophysical Union, Washington, DC. IRIFUNET. and RINCWOODA. E. (1987b) Phase transformations in a harzburgite composition to 26 GPa: Implications for dynamical behaviour of the subducting slab. Earth Planet. Sci. Lett. 86, 365316. IRIFUNET., HIBBERSON W. A., and RINGWWD A. E. (1989) Eclogitegarnetite transformation at high pressure and its bearing on the occurrence of garnet inclusions in diamond. In Kimberlites and Related Rocks Vol. 2: Their Mantle/Crust Setting, Diamonds and DiamondExploration (eds. J. ROSSet al.), pp. 877-882. Geol. Sot. Australia Special Pub]. Blackwell Scientific Publishers, Clayton, Vie. ISACKSB. and MOLNAR P. (1971) Distribution of stresses in the descending lithosphere from a global survey of focal mechanism solutions of mantle earthquakes. Rev. Geophvs. Space Phys. 9, 103-174. 2108 A. E. Ringwood ITO E. and TAKAHASHIE. (1987) Ultrahigh pressure phase transformations and the constitution of the deep mantle. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANIand Y. SYONO), pp. 22 l-229. American Geophysical Union, Washington, DC. ITO E. and TAKAHASHIE. (1989) Post-spine1 transformations in the system Mg,SiO.,-Fe&O, and some geophysical implications. J. Geophys. Res. 94, 10,637-10,646. ITO E. and YAMADAH. (1982) Stability relations of silicate spinels, ilmenites and perovskites. In High-Pressure Research in Geophysics (eds. S. AKIMOTOand M. H. MANGHNANI),pp. 405-419. Terrapubl. Tokyo. ITO E., MA~SUIK., SUITOK., and KAWAIN. (1974) Synthesis of yMg,SiO,. Phys. Earth Planet. Intl. 8, 342-344. ITO E., TAKAHASHIE., and MATSLJIY. (1984) The mineralogy and chemistry of the lower mantle: an implication of the ultrahighpressure phase relations in the system MgO-FeO-Si02. Earth Planet. Sci. Lett. 67, 238-248. ITO E., AKAOGIM., TOPORL., and NAVROTSKYA. (1990) Negative pressure-temperature slopes for reactions forming MgSiOj perovskite from calorimetry. Science 249, 1275-1278. JACKSON1. (1977). Melting of some alkaline-earth and transitionmetal fluorides and alkali fluoroberyllates at elevated pressures: A search for melting systematics. Phys. Earth Planet. Intl. 14, 143164. JACKSONI. (I 983) Some geophysical constraints on the chemical composition of the Earth’s lower mantle. Earth Planet. Sci. Left. 62,91-103. JAGOUTZE., PALMEH., BADENHAUSENH., BLUM K., CENDALES M., DREIBUSG., SPETTELG., LORENZV., and WXNKEH. (1979) The abundances of major, minor and trace elements in the Earth’s mantle as derived from primitive ultramafic nodules. Proc. 10th Lunar Sri. Conf: 2,2031-2050. JEANLOZR. (1990)The nature of the Earth’s core. Ann. Rev. Earth Planet. Sci. 18, 357-386. JEANLOZR. and KNITTLEE. (1989) Density and composition of the lower mantle. Phil. Trans. Roy. Sot. London A328, 377-389. JEANLOZR. and THOMPSONA. (1983) Phase transitions and mantle discontinuities. Rev. Geophys. Space Phys. 21, 51-74. JEFFREYSH. (1939)The times of P, S and SKS and the velocities of P and S. Mon. Not. Roy. Astron. Sot. Geophys. Suppl. 4,498-533. JORDANT, (1979) Mineralogies, densities and seismic velocities of garnet Iherzolites and their geophysical implications. In TheMantle Sample: Inclusions in Kimberlites and Other Volcanics (eds. F. R. BOYD and H. MEYER), pp. I-14. American Geophysical Union, Washington, DC. JORDANT. (198 1) Continents as a chemical boundary layer. Phil. Trans. Roy. Sot. London A301, 359-373. KARATO S. (1989) Plasticity-crystal structure systematics in dense oxides and its implications for the creep strength of the Earth’s deep interior: A preliminary result. Phys. Earth Planet. In/l. 55, 234-240. KARATO S., PATERSONM., and FITZ GERALDJ. (1986) Rheology of synthetic olivine aggregates: influence of grain size and water. J. Geophys. Res. 91, 8 151-8 176. KATO T. and RINGWOODA. E. (1989) Melting relationships in the system Fe-Fe0 at high pressures: Implications for the composition and formation of the Earth’s core. Phys, Chem. Mineral. 16, 524538. &TO T., RINGWOODA. E., and IRIFUNET. (1988a)Experimental determination of element partitioning between silicate perovskites, garnets and liquids: constraints on early differentiation of the mantle. Earth Planet. Sci. Lett. 89, 123-145. KATO T., RINGWOODA. E., and IRIFUNET. (1988b) Constraints on element partition coefficients between MgSi03 perovskite and liquid determined by direct measurements. Earth Planet. Sci. Lett. 90,65-68. KATSURAT. and ITO E. ( 1989) The system Mg,Si04-FeZSi04 at high pressures and temperatures. Precise determination of stabilities of olivine, modified spine1 and spine]. J. GeophJ!s. Res. 94, 15,66315,670. KAWADAK. (1977) The system MgzSiO,-FezSiO1 at high pressures and high temperatures and the Earth’s interior. Ph.D. dissertation, University of Tokyo. KAWAI N. and ENDOS. (1970) The generation of ultrahigh hydrostatic pressures by a split-sphere apparatus. Rev. Sci. Instrum. 41, 117% 1181. KAWAIN., TACHIMORIM., and ITO E. (1974) A high pressure hexagonal form of MgSi03. Proc. Japanese Acad. 50,378-380. KELLOGG L. and TURCOTTED. (1986) Mantle homogenization. Eos 67, 367. KENNETTB. L. N. (1991) Seismic velocity gradients in the upper mantle. Geophys. Res. Left. (submitted). KERR R. and LISTERJ. (1987) The spread of subducted lithospheric material along the mid-mantle boundary. Earth Planet. Sci. Lett. 85, 24 1-247. KESSON S. E. and RINGWOOD A. E. (1989)Slab-mantle interactions. 1.Sheared and refertilised garnet peridotite xenoliths-samples of Wadati-Benioff zones? Chem. Geol. 78, 83-96. KIRBY S. H. (1980) Tectonic stresses in the lithosphere: Constraints provided by the experimental deformation of rocks. J. Geophys. Res. 85,6353-6363. KNITTLE E. and JEANLOZR. (1987) Synthesis and equation of state of (Mg,Fe)SiO-( perovskite to over 100 gigapascals. Science 235, 668-670. KNI~TLE E. and JEANLOZR. (1989) Simulating the core-mantle boundary: an experimental study of high-pressure reactions between silicates and molten iron. Geophys. Re.r. Lett. 16, 609-6 12. KNI-ITLEE., JEANLOZR., and SMITHG. L. (1986) Thermal expansion of silicate perovskite and stratification of the Earth’s mantle. Nature 319, 214-216. KUMAZAWAM. (1981) Origin of materials in the Earth’s interior and their layered distribution. Japanese J. Petrol. Mineral Econ. Geol. Spec. Iss. 3, 239-247. LAY T. (1989) Structure ofthe core-mantle transition zone: A chemical and thermal boundary layer. Eos 70,49-59. LEESA., BUKOWINSKIM., and JEANLOZR. (1983)Reflection properties of phase transition and compositional change models of the 670 km discontinuity. J. Geophys. Res. 88, 8145-8 159. LEVENJ. H., JACKSONI., and RINGWOOD A. E. (198 1) Upper mantle seismic anisotropy and lithospheric decoupling. Nature 253, 708710. LIU L.-G. (1974) Silicate perovskite from phase transformations of pyrope-garnet at high pressure and temperature. Geophys. Res. Lett. 1, 277-280. LIU L.-G. (I 975) Post-oxide phases of forsterite and enstatite. Geoph.vs. Res. Lett. 2, 417-419. LIU L.-G. (1977) The system enstatite-pyrope at high pressures and temperatures and the mineralogy of the Earth’s mantle. Earth Planet. Sri. Lett. 41, 398-404. Lru L.-G. (1979a) Phase transformations and the constitution of the deep mantle. In The Earth: Its Origin, Structure and Evolution (ed. M. W. MCELHINNY),pp. 177-202. Academic Press, London. LIU L.-G. (1979b) On the 650 km seismic discontinuity. Earth Planet. Sci. Lett. 42, 202-208. LIU L.-G. (1980) The mineralogy of an eclogitic earth mantle. Phys. Earth Planet. Intl. 23, 262-267. LIU L.-G. and BASSETT W. A. (1986) Elements, Oxides and Silicates: High Pressure Phases with Implications ,for the Earth’s Interior. Oxford University Press. LIU L.-G. and RINGWOODA. E. (1975) Synthesis of perovskite-type polymorph of Casio,. Earth Planet. Sci. Lett. 28, 209-2 I 1. MAD~N M., CASTEX J., and PEYRONNEAUJ. (1989a) A new aluminocalcic high-pressure phase as a possible host of calcium and aluminium in the lower mantle. Nature 342, 422-425. MAD~N M., GUYOTF., PEYRONNEAU J., and POIRIERJ. P. (1989b) Electron microscopy of high-pressure phases synthesized from natural olivine in a diamond anvil cell. Phys. Chem. Mineral. 16, 320-330. MAO H. K. and BELL P. (I 977) Pressure-volume equations of state of Me0 and Fe to 1 Mbar. Carnegie Yearb. 76. . Inst. Washinnton ._ 5 19-1522. MCCULLOCHM. (199 I) The role of subducted slabs in an evolving Earth. Earth Planet. Sci. Lett. (in press). MCDONOUGHW. F. (1987) Chemical and isotopic systematics of basalts and peridotite xenoliths: Implications for the composition Inaugural Ingerson Lecture and evolution of the Earth’s mantle. Ph.D. thesis, Australian National University, Canberra. MCFARLANE E. and DRAKEM. (1990) Element partitioning and the early thermal history of the Earth. In Origin of the Earth (eds. H. E. NEW~~M and J. H. JONES),pp. 135-l 50. Oxford Univ. Press. MCKENZIED. and BICKLEM. J. (1988) The volume and composition of melt generated by extension of the lithosphere. J. Petrol. 29, 625-680. MCKENZIED. and WEISSN. (1975) Speculations on the thermal and tectonic history of the Earth. Geophys. J. Roy. Astron. Sot. 42, 131-174. MEADEC. and JEANLOZR. (1990) The strength of mantle silicates at high pressures and room temperatures: Implications for the viscosity profile of the mantle. Nature 348, 533-535. MERCIERJ. C. (1979) Peridotite xenoliths and the dynamics of kimberlite intrusion. In The Mantle Sample: Inclusions in Kimberlites and Volcanic Rocks (eds. F. R. BOYDand H. MEYER), pp. 1972 12. Amer. Geophysical Union, Washington, DC. MICHAELP. J. and BONATTIE. (1985) Peridotite composition from the North Atlantic: Regional and tectonic variations and implications for partial melting. Earth Planet. Sci. Lett. 73, 9 I-104. MILLERG. H., STOLPERE. M., and AHRENST. J. (I 99 1) The equation of state of a molten komatiite, komatiite petrogenesis, and the evolution of the Hadean mantle. J. Geophys. Rex (in press). NESBITTB. and SUN S. (1976)Geochemistry of Archaean spinifextextures peridotites and magnesian and low-magnesian tholeiites. Earth Planet. Sci. Lett. 31, 433-453. NISBETE. G. and WALKERD. (1982) Komatiites and the structure of the Archaean mantle. Earth Planet. Sci. Lett. 60, 105-I 13. O’HARA M. J. (1975) Is there an Icelandic mantle plume? Nature 60, 105-I 13. OHKURA T., KANESHIMA S., and ANDO M. (1990) Two layered structure of a subducting oceanic plate. Eos 71, 555. OHTANIE. (1984) Generation of komatiite magma and gravitational differentiation in the deep upper mantle. Earth Planet. Sci. Lett. 67,261-272. OHTANI E. (1985) The primordial terrestrial magma ocean and its implications for stratification of the mantle. Earth Planet. Sci. Lett. 78,70-80. G’NEILL B. and JEANLOZR. (1990) Experimental Petrology of the lower mantle: A natural peridotite taken to 54 GPa. Geophys. Res. Lett 77, 1477-1480. O’NIONS K. (1987) Relationships between chemical and convective layering in the Earth. J. Geol. Sot. London 144,259-274. O’NIONS K. and OXBURGH E. R. (1983) Heat and helium in the Earth. Nature 306, 429-43 1. OXBURCHE. R. and TURCOTTED. L. (I 970) Thermal structure of island arcs. Bull. Geol. Sot. Amer. 81, 1665-1688. PACALOR. and GASPARIKT. (1990) Reversals of the orthoenstatiteclinoenstatite transition at high pressures and temperatures. J. Geophys. Res. 95, 15,853-15,858. PHILLIPSR. J. and LAMBECKK. (1980)Gravity fields of the terrestrial planets: long wavelength anomalies and tectonics. Rev. Geophys. Space Phys. 18, 27-76. PRICE G., WALL A., and PARKERS. (1989) The properties and behaviour of mantle minerals: A computer simulation approach. Phil. Trans. Roy. Sot. London A328, 39 l-407. RICHARDSM. and DAVIESG. (1989) On the separation of relatively buoyant components from subducted lithosphere. Geophys. Res. Lett. 16, 83 l-834. RIETMEIJERF. (1987) Chondritic interplanetary dust and primitive chondrite matrices: the search for chemically pristine solids in the solar system. Lunar Planet. Sci. 18, 832-834. RIETMEIJERF. (1988) On chemical continuum in early solar system dust at >I.8 A.U. Chem. Geol. 70, 33. RICDEN S. M., AHRENST. J., and STOLPERE. E. (1984) Densities of silicate legends at high pressure. Science 226, 1071-1074. RIGDEN S., GWANMESIAG. D., FITZ GERALDJ., JACKSONI., and LIEBERMANN R. C. (199 1) High pressure elasticity of the Mg$iO, polymorphs and the structure of the transition zone of the Earth’s mantle. Nature (in press). RINGWOODA. E. (1956) The olivine-spine1 transition in the Earth’s mantle. Nature 178, 1303-l 304. RINGW~~D 2109 A. E. (1958a) Constitution of the mantle Part I. Geochim. Cosmochim. Acta 13,303-32 1. RINGWOODA. E. (1958b) The olivine-spine1 transition in fayalite. Bull. Geol. Sot. Amer. 69, 129. RINGWOODA. E. (1959) On the chemical evolution and densities of the planets. Geochim. Cosmochim. Acta 15, 257-283. RINGWO~DA. E. (1961) Silicon in the metal phase of enstatite chondrites and some geochemical implications. Geochim. Cosmochim. Acta 25, 1-13. RINGWOODA. E. (1962a) A model for the upper mantle. J. Geophys. Res. 67, 857-867. RINGWOODA. E. (1962b) A model for the upper mantle 2. J. Geophys. Res. 67,4473-4477. RINGWOODA. E. (1962~) Mineralogical constitution of the deep mantle. J. Geophys. Res. 67,4005-40 10. RINGWOODA. E. (1962d)Prediction and confirmation of olivinespine1 transition in Ni$iO,. Geochim. Cosmochim. Acta 26,457469. RINGWOODA. E. (1963) Olivine-spine1 transformation in cobalt orthosilicate. Nature 198, 79-80. RINGWOODA. E. (1966a) The chemical composition and origin of the Earth. In Advances in Earth Sciences (ed. P. M. HURLEY),pp. 287-356. MIT Press, Cambridge. RINGWOODA. E. (1966b) Chemical evolution of the terrestrial planets. Geochim. Cosmochim. Acta 30,41-104. RINGWOODA. E. (1966~)Mineralogy of the mantle. In Advances in Earth Science (ed. P. HURLEY), pp. 357-398. MIT Press, Cambridge. RINGWOODA. E. (1967) The pyroxene-garnet transformation in the Earth’s mantle. Earth Planet. Sci. Lett. 2, 255-263. RINGWOODA. E. (1970) Phase transformations and the constitution of the mantle: A review. Phys. Earth Planet. Intl. 3, 1099155. RINGWOODA. E. (I 975) Composition and Petrology, of the Earth’s Mantle. McGraw-Hill. New York. RINGWOODA. E. (1977) Composition of the core and implications for origin of the Earth. Geochem. J. 11, 111-135. RINGWOODA. E. (1979) Origin of the Earth and Moon. SpringerVerlag, New York. RINGWOODA. E. (1982) Phase transformations and differentiation in subducted lithosphere: Implications for mantle dynamics, basalt petrogenesis, and crustal evolution. J. Geology 90, 61 l-643. RINGWOODA. E. (1984) The Earth’s core: Its composition, formation and bearing upon the origin of the Earth. Proc. Roy. Sot. London A395, I-46. RINGWOODA. E. (1986) Dynamics of subducted lithosphere and implications for basalt petrogenesis. Terra Cognita 6, 67-77. RINGWWD A. E. (1989a) Significance of the terrestrial Mg/Si ratio. Earth Planet. Sci. Lett. 95, 1-7. RINGWOODA. E. (1989b) Constitution and evolution of the mantle. In Kimberlites and Related Rocks: Vol. I-Their Comoosition, Occurrence, Origin and Emplacement (eds. J. ROSS et’al.), pp. 457-485. Geol. Sot. Australia Special Publ., Blackwell Scientific Publishers, Clayton, Vie. RINGWOODA. E. (1990a) Earliest history of the Earth-Moon system. In Proc. Conf on Origin of the Earth (eds. H. NEWSONand J. JONES),pp. 10l- 134. Oxford University Press, Oxford. RINGWOODA. E. (1990b)Slab-mantle interactions. 3. Petrogenesis of intraplate magmas and structure of the upper mantle. Chem. Geol. 82, 187-207. RINGWOODA. E. and GREEN D. H. (I 966) An experimental investigation of the gabbro-eclogite transformation and some geophysical implications. Tectonouhvsics 3. 383-427. RINGWOODA. E. and HI~BERSONW. (1990) The system Fe-Fe0 revisited. Phys. Chem. Mineral. 17, 313-319. RINGWOODA. E. and HIBBERSONW. ( I99 1) Solubilities of mantle oxides in molten iron at high pressures and temperatures: Implications for the composition and formation of Earths core. Earth Planet. Sci. Lett. (in press). RINGWOOD A. E. and IRIFUNET. (1988) Nature ofthe 650-km seismic discontinuity: Implications for mantle dynamics and differentiation. Nafure 331, 131-136. RINGWOODA. E. and KESSONS. E. (1977) Basaltic magmatism and the bulk composition of the Moon II. Siderophile and volatile 2110 A. E. Ringwood elements in Moon, Earth and chondrites: Implications for lunar origin. The Moon 16,425-4&l. RINGWOODA. E. and MAJORA. ( 1966) Synthesis of MgzSi04-Fe2Si04 spine1 solid solutions. Earth Planet. Sci. Lett. 1, 241-245. RINGWOODA. E. and MAJOR A. (197 I) Synthesis of majorite and other high pressure garnets and perovskites. Earth Planet. Sci. Lett. 12,41 I-418. RINGWOODA. E. and SEABROOKM. P. (1962) Olivine-spine1 equilibria at high pressure in the system Ni,GeO,-Mg$SiO+ J. Geophys. Res. 67, 1975-1985. Ross N. and HAZENR. (1989) Single crystal X-ray diffraction study of MgSiOs perovskite from 77 to 400 K. Phys. Chem. Mineral 16, 415-420. SCHMELINGH. (1989) Compressible convection with constant and variable viscosity: The effect on slab formation, geoid, and topography. J. Geophys. Rex 94, 12,463-12,48 1. SHEARERP. (1990) Seismic imaging of upper mantle structure-new evidence for a 520 km discontinuity. Nature 344, 12 l- 126. SHEARERP. ( I99 1) Constraints on upper-mantle discontinuities from observations of long period reflected and converted phases. J. Geophys. Rex (in press). STACEYF. and LOPERD. (1983)The thermal boundary-layer interpretation of D” and its role as a plume source. Phys. Earth Planet. Intl. 33, 45-55. STISHOVS. M. and POPOVAS. V. (1961) New dense polymorphic modification of silica. Geokhimiya 10, 837-839. STOLPERE. M., WALKERD., HAGER B., and HAYS J. (1981) Melt segregation from partially molten source regions: Importance of melt density and source region size. J. Geophys. Res. 86, 62616271. SUN S.-S. ( 1982) Chemical composition and the origin of the Earth’s primitive mantle. Geochim. Cosmochim. Acta 46, 179-192. SUN S. and NESBITTR. W. (1977) Chemical heterogeneity of the Archaean mantle, composition of the Earth and mantle evolution. Earth Planet. Sci. Lett. 44, 119-l 38. TAKAHASHIE. (1986) Melting of a dry peridotite KLB-1 up to 14 GPa: Implications on the origin of peridotitic Upper Mantle. J. Geophys. Res. 91,9367-9382. TAKAHASHIE. and ITO E. (1987) Mineralogy of mantle peridotite along a model geotherm up to 700 km depth. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANIand Y. SYONO), pp. 427-437. American Geophysical Union, Washington, DC. WALEK M. C. (1984) The P wave upper mantle structure beneath an active spreading centre: the Gulf ofcalifornia. Geophys. J. Roy. Astron. Sot. 76, 697-723. WALL A. and PRICE G. (1989) Electrical conductivity of the lower mantle molecular dynamics simulation of MgSiOSperovskite. Phys. Earth Planet. Int. 58, 192-194. WANG Y., WEIDNERD., LIEBERMANN R., Lru X., Ko J., VAUGHAN M., ZHAO Y., YEGANEH-HAERIA., and PACALOR. (1991) Phase transition and thermal expansion of MgSi03 perovskite. Science 251,410-412. WXNKE H. (1981) Constitution of terrestrial planets. Phil. Trans. Roy. Sot. London A303,287-302. WANKE H. and DREIBUSG. (1988) Chemical composition and accretion history ofterrestrial planets. Phil. Trans. Roy. Sot. London A325,545-557. WARREND. H. (1968) Project Early Rise, travel times and amplitudes. USGS Open-File Rept., Menlo Park, CA. WATT J. P. and AHRENST. J. (I 986) Shock wave equation of state of enstatite. J. Geophys. Res. 91, 7495-7503. WEIDNERD. J. (1985) A mineral physics test of a pyrohte mantle. Geophys. Rex Lett. 12, 4 17-420. WEIDNERD. J. and ITO E. (1987) Mineral physics contraints on a uniform mantle composition. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANIand Y. SYONO),pp. 439-446. American Geophysical Union, Washington, DC. WILLIAMSQ., JEANLOZR., BASSJ., SVENDSENB., and AHRENST. (I 987) The melting curve of iron to 250 gigapascals: A constraint on the temperature at the Earth’s center. Science 236, 18l- 182. WOLFG. and BUKOWINSKIM. (1987) Theoretical study ofthe structural properties and equations of state of MgSiO, and CaSiOs perovskites: Implications for lower mantle composition. In HighPressure Research in Mineral Physics (eds. M. MANGHNANIand Y. SYONO),pp. 3 13-33 I. American Geophysical Union, Washington, DC. YAGI T., BELLP. M., and MAO H. K. (1979) Phase relations in the system MgO-FeO-Si02 between 150 and 700 kbar at 1000°C. Carnegie Inst. Washington Yearb. 78, 6 14-6 18. YAGI T., AKAOGIM., SHIMOURAO., TAMAIH., and AKIMOTOS. (1987) Equations of state of garnet solid solutions determined by high pressure X-ray diffraction using synchrotron radiation. In High-Pressure Research in Mineral Physics (eds. M. MANGHNANI and Y. SYONO), pp. 141-148. American Geophysical Union, Washington, DC. YEGANEH-HAERIA., WEIDNERD. J., and ITO E. (1989) Elasticity of MgSiO, in the perovskite structure. Science 243, 787-789. YOUNG C. and LAYT. (1987) The Core-mantle boundary. Ann. Rev. Earth Planet. Sci. 15, 25-46. ZHOUH. and CLAYTONR. (1990) P and S wave travel time inversions for subducting slab under the island arcs of the northwest Pacific. J. Geophys. Res. 95, 682 l-685 1. ZINDLERA. and HART S. (1986) Chemical geodynamics. Ann. Rev Earth Planet. Sci. 14, 493-57 1.