Chapter 8

Let's take it from the top: the and upper

ZOE: Come and I'll peel off The crust BLOOM: (feeling his occiput dubiously The major divisions of the 's interior- crust, with the unparalleled embarrassment mantle and core - have been known from seis­ of a harassed pedlar gouging the mology for about 80 years. These are based on the symmetry of her peeled pears) reflection and refi·action of P- and S-waves. The Somebody would be dreadfully boundary between the crust and mantle is called the Mohorovicic discontinuity (M-discontinuity jealous if she knew. or Moho for short) after the Croatian seismol­ james joyce, Ulysses ogist who discovered it in 1909. It separates rocks having P-wave velocities of 6-7 kmfs fi·om those having velocities of about 8 kmfs. The term The broad-scale structure of the Earth's interior 'crust' has been used in several ways. It initially is well known from , and lmowledge referred to the brittle outer shell of the Earth of the fine structure is improving continuously. that extended down to the ('weak Seismology not only provides the structure, it layer'); this is now called the ('rocky also provides information about the composi­ layer'). Later it was used to refer to the rocks tion, , dynamics and physical state. A occurring at or near the surface and acquired a 1D seismological model of the Earth is shown petrological connotation. Crustal rocks have dis­ in Figure 8.1. Earth is conventionally divided tinctive physical properties that allow the crust into crust, mantle and core, but each of these to be mapped by a variety of geophysical tech­ has subdivisions that are almost as fundamen­ niques. Strictly spealdng, the crust and the Moho tal (Tables 8.1 and 8.2). Bullen subdivided the are seismological concepts but petrologists speak Earth's interior into shells, from A (the crust) of the 'petrological Moho,' which may actually through G (the inner core). The , occur in the mantle! The Moho may represent a starting at 1000 km depth, is the largest subdi­ transition fi·om to ultramafic rocks or to vision, and therefore it dominates any attempt a high-pressure assemblage composed predomi­ to perform major-element mass balance calcula­ nately of and clinopyroxene. It may thus tions. The crust is the smallest solid subdivision, be a chemical change or a phase change or both, but it has an importance far in excess of its rel­ and differs fi-om place to place. The lower conti­ ative size because we live on it and extract our nental crust can become denser than the mantle, resources frmn it, and, as we shall see, it con­ and may founder and sink into the mantle. tains a large fraction of the terrestrial inventory The present surface crust represents 0.4% of of many elements. the Earth's mass and 0.6% of the silicate Earth. It 92 I LET'S TAKE IT FROM THE TOP: THE CRUST AND

contains a very large proportion of incompatible Table 8.1 Bullen's regions of the Earth's I elements (20-70%), depending on element. These interior include the heat-producing elements and mem­ De pth Range bers of a number of radiogenic-isotope systems Region (km) (Rb-Sr, U-Pb, Sm-Nd, Lu-Hf) that are commonly used in mantle . Thus the conti­ A 0 33 nental crust factors prominently in any mass­ B 33 4 10 upper mantle balance calculation for the Earth as a whole and 220 in estimates of the thermal structure of the Earth c 4 10 1000 transition region (Figure 8.2}. (Rudnic k cru s t). 650 discontinuity The Moho is a sharp seismological boundary D' 1000 2700 lower mantle and in some regions appears to be laminated. 1000 Repetti discontinuity There are three major crustal types- continental, D" transition region 2700 2900 transitional and oceanic. generally E 2900 4980 outer core ranges from 5-15 km in thickness and comprises F 4980 5 120 transition region 60% of the total crust by area and more than G 5 120 6370 inner core 20% by volume. In some areas, most notably near oceanic fracture zones, the oceanic crust is as thin as 3 km. Sometimes the crust is even absent, Table 8.2 I Summary of Earth structure presumably because the underlying mantle is Fraction Fraction cold or infertile, or ascending melts freeze before Depth of Total of Mantle they erupt. Oceanic plateaus and aseismic ridges Region (km) and Crust may have crustal thicknesses greater than 20 km. Some of these appear to represent large volumes Continental crust 0-50 0.00374 0.00554 of material generated at oceanic spreading cen­ Oceanic crust 0-10 0.00099 0.00147 ters or triple junctions, and a few seem to be Upper mantle 10-400 0. 103 0.153 continental fragments. Although these anoma­ Transition (TZ) 400- 650 0075 0.111 lously thick crust regions constitute only about Deep mantle 650- 2890 0.492 0.729 10% of the area of the , they may repre­ Outer core 2890- 5150 0.308 sent more than 25% of the total volume of the Inner core 5150-6370 0.0 17 oceanic crust. They are generally attributed to hot regions of the mantle but they could also represent fertile regions of the mantle or tran­ sient responses to lithospheric extension. , arcs and continental margins are collec­ tively referred to as transitional crust and range from 15-30 km in thickness. Continental crust generally ranges from 30-50 km thick, but thick­ nesses up to 80 km are reported in some active convergence regions. Older regions show a sharp cutoff in crustal thickness at 50 km; this may be the depth of the - phase change and the depth at which over-thickened crust 2000 4000 founders or delaminates. Based on geological and Depth (km) seismic data, the main type in the upper The Preliminary Reference Earth Model (PREM). continental crust is or tonalite in The model is anisotropic in the upper 220 km. Dashed lines composition. The lower crust is probably diorite, are the horizontal components of the seismic velocity (after garnet and amphibolite. The average Dziewonski and Anderson, 1981 ). THE CRUST 93

Shields Platforms 50.------~~~~------~m~e~a~n 80.------~m~e~a~n 41 .85 70 41 .44

>­ u ()' c c Q) :J 40 cr cr~ ~ lL 3:

20 70 80 20 30 70 80 Crustal Thickness (km) Crustal Thickness (km)

Extended Crust Basins 100.------~~~------~m~e~a~n' 20.------~m~e~a~n 30.95 43.68 80 1- 15 >- >­ u g 60 1- c Q) :J ~ 10 cr cr Q) 3: 40 1- u: 5 20 1- rh ~~~~~~~~~rh~~~~~~~~ 0 10 20 30 40 50 60 70 80 70 80 Crustal Thickness (km) Crustal Thickness (km)

Large Igneous Provinces Orogens 10.------~~------m--ea~n~ 60 .------~------=m~e~a~n 35.46 42.62 50

>­ >- 40 u u c c Q) 0 arcs :J ~ 30 33.52 • l l 20 28.66


60 70 80 80 Crustal Thickness (km) Crustal Thickness (km) Crustal thickness histograms for various tectonic rich assemblages at relatively shallow depth. provinces. Note the cutoff near SO km thickness. Thicker The maximum theoretical thickness of mate­ crust exists in some belts but apparently does not rial with crust-like physical properties is about last long (after Mooney et a/., 1998). 50-60 km, although the crust may temporarily achieve somewhat greater thickness because of composition of the continental crust is thought the sluggishness of phase changes at low temper­ to be similar to or diorite. ature and in dry rocks. An Earth model based The terrestrial crust is unusually thin com­ on cosmic abundances of the elements could, in pared with the and and compared principle, have a basaltic crust 200 km thick. lt is with the amount of potential crust in the mantle. usually considered that the Earth's extra crustal This is related to the fact that crustal mate­ material is well mixed into the mantle, or that it rial - on a large body - converts to dense garnet- was never extracted from parts of the mantle. 94 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

Table 8.3 I Crustal

Range of Crustal Mi neral Composition Abundances (vo l. pet.) Plagiocl ase 31-41 Anorthite Ca(AI2Si 2)0s Albite Na(AI,S i3)0s O rthoclase 9-21 K- K(AI.Si 3)0s Q uartz Si0 2 12-24 Amphi bole NaCa2(Mg. Fe.AI)s [(AI.S i)4 0 11]2(0H)2 0-6 Biotite K( Mg.Fe2+)3(AI .Si3)0 1o(OH.F)2 4-1 1 Muscovite KAI 2(AI.Si3)0 1o(OH) 0-8 Chlorite (Mg.Fe2+)sAI(A I, Si3)0 1o(OH)s 0-3 Hypersthene (Mg.Fe2+ )Si09 0- 11 Augite Ca(Mg.Fe2+)(Si03)2 O li vin e (Mg.Fe2+)2Si0 4 0-3 Oxides ~2 Sphene CaTiSi05 All anite (Ce,Ca.Y)(AI,Fe)3(Si0 4)3(0H) Apatite Cas(P04.C03)3(F.O H,CI) Magnetite FeFe204 Il menite FeTi02

Alternatively, there may be layers or blobs of Table 8.4 Average crustal abundance, den- eclogite in the mantle, representing crust that I sity and seismic velocities of major crustal was previously near the Earth's surface. Such fer­ minerals tile blobs are an alternative to the plume model for anomalous . Volume p vp V, 3 Feldspar (K-feldspar, ) is the most % (g/cm ) (km/s) (km/s) abundant mineral in the crust, followed by and hydrous minerals (such as the micas Quartz 12 2.65 6.05 4.09 and amphiboles) (Table 8.3}. The minerals of the K-feldspar 12 2.57 5.88 3.05 crust and some of their physical properties are Plagioclase 39 2.64 6.30 3.44 given in Table 8.4. A crust composed of these Micas 5 2.8 5.6 2.9 minerals will have an average density of about Amphiboles 5 3.2 7.0 3.8 3 Pyroxene I I 3.3 7.8 4.6 2. 7 gfcm . There is enough difference in the O li vine 3 3.3 8.4 4.9 velocities and Vp/Vs ratios of the more abundant minerals that seismic velocities provide a good mineralogical discriminant. One uncertainty is the amount of serpentinized ultramafic rocks can actually be due to the -eclogite phase in the lower crust since serpentinization decrea­ change, rather than to a chemical change. In ses the velocity of to crustal values. In these cases, crustal thickness is not a good proxy regions of over-thickened crust, the lower por­ for mantle temperature or the amount of m elt tions can have abundant garnet and therefore that has been extracted from the mantle. high density and high seismic velocity, compa­ The floor of the , under the , rable to mantle values. The Moho in these cases is veneered by a layer of tholeiitic basalt that THE CRUST 95

was generated at the midocean ridge systems. The Table 8.5 I Estimates of the chemical compo­ pillow that constitute the upper part of sition of the crust (wt.%) the oceanic crust extend to an average depth of 1-2 km and are underlain by sheeted dikes, gab­ Oceanic Continental Crust bros and olivine-rich cumulate layers. The aver­ Crust age composition of the oceanic crust is much Oxide ( I) (2) (3) more MgO-rich and lower in CaO and Ah03 than the surface MORB. The oceanic crust rests on a Si02 47.8 63.3 58.0 depleted layer of unknown thiclmess. Ti02 0.59 0.6 0.8 In certain models of crustal genesis, the harzbur­ AI20 3 12. 1 16.0 18.0 gite layer is complementary to the crust and is Fe20 3 1.5 9.0 3.5 7.5 therefore about 24 km thick if an average crustal FeO thickness of 6 km and 20% are assumed. MgO 17.8 2.2 3.5 CaO 11.2 4. 1 7.5 Oceanic plateaus and aseismic ridges have thick N a20 1.3 1 3.7 3.5 crust (> 20 km in places), and the corresponding K20 0.03 2.9 1.5 depleted layer would be more than 120 km thick H20 1.0 0.9 if the simple model is taken at face value. Since depleted , including and (1) Elthon (1979). , are less dense than fertile , or (2) Condie (1982). eclogite, several cycles of plate , crust (3) Tayor and McLennan (1985) . generation and would fill up the shal­ low mantle with harzburgite. Oceanic crust and oceanic plateaus may in part be deposited on mafic rocks, by the gabbro-eclogite phase change, ancient, not contemporaneous, buoyant under­ to mantle-like values. The increase in velocity pinnings. from 'crustal' to 'mantle' values in regions of Estimates of the composition of the oceanic thick continental crust may be due, at least and continental crust are given in Table 8.5; in part, to the appearance of garnet as a sta­ another estimate that includes the trace ble phase. The situation is complicated further elements is given in Table 8.6. Note that the by kinetic considerations. Garnet is a common

continental crust is richer in Si02 , Ti02 , A1 2 0 3 , metastable phase in near-surface intrusions such

Na 2 0 and 0 than the oceanic crust. This as pegmatites and metamorphic teranes. On the means that the continental crust is richer in other hand, feldspar-rich rocks may exist at quartz and feldspar and is intrinsically less depths greater than the gabbro-eclogite equilib­ dense than the oceanic crust. The mantle under rium boundary if temperatures are so low, or the stable continental- crust has seismic prop­ rocks so dry, that the reaction is sluggish. erties that suggest that it is less dense than The common assumption that the Moho is a mantle elsewhere. The elevation of chemical boundary is in contrast to the position is controlled primarily by the density and thick­ taken with regard to other mantle discontinu­ ness of the crust and the intrinsic density and ities. It is almost universally assumed that the temperature of the underlying mantle. major mantle discontinuities represent equilib­ It is commonly assumed that the seismic rium solid-solid phase changes in a homoge­ Moho is also the petrological Moho, the bound­ nous mantle. A notable exception to this view ary between sialic or mafic crustal rocks and is the geochemical model that attributes the ultramafic mantle rocks. However, partial melt­ 650 km discontinuity to a boundary separat­ ing, high pore pressure and serpentinization ing the depleted 'convecting upper mantle' from can reduce the velocity of mantle rocks, and the undegassed primitive lower mantle. This is increased abundances of olivine, garnet and strictly an assumption; there is no evidence in pyroxene can increase the velocity of crustal support of this view. It should be kept in mind rocks. High pressure also increases the velocity of that chemical boundaries may occur elsewhere in 96 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

Table 8.6 1 Composition of the bulk continental crust, by weight*

Si02 57.3 pet. Co 29 ppm Ce 33 ppm Ti02 0.9 pet. Ni lOS ppm Pt 3.9 ppm AI20 3 15.9 pet. Cu 75 ppm Nd 16 ppm FeO 9.1 pet. Zn 80 ppm Sm 3.5 ppm MgO 5.3 pet. Ga 18 ppm Eu 1.1 ppm CaO 7.4 pet. Ge 1.6 ppm Gd 33 ppm Na20 3.1 pet. As 1.0 ppm Tb 0.6 ppm K20 1.1 pet. Se 0.05 ppm Dy 3.7 ppm Li 13 ppm Rb 32 ppm Ho 0.78 ppm Be 1.5 ppm Sr 260 ppm Er 2.2 ppm B 10 ppm y 20 ppm Tm 0.32 ppm Na 2.3 pet. Zr 100 ppm Yb 2.2 ppm Mg 3.2 pet. Nb II ppm Lu 0.30 ppm AI 8.41 pet. Mo I ppm Hf 3.0 ppm Si 26.77 pet. Pd I ppb Ta I ppm K 0.91 pet. Ag 80 ppb w I ppm Ca 5.29 pet. Cd 98 ppb Re 0.5 ppm Sc 30 ppm In 50 ppb lr 0.1 ppb Ti 5400 ppm Sn 2.5 ppm Au 3 ppb v 230 ppm Sb 0.2 ppm Tl 360 ppb Cs 185 ppm Cs I ppm Pb 8 ppb Mn 1400 ppm Ba 250 ppm Bi 60 ppb Fe 707 pet. La 16 ppm Th 3.5 ppm u 0.91 ppm

Taylor and McLennan (1985). *Major elements as oxides or elements.

the mantle. It is hard to imagine how the Earth probably erroneously - treated as one or two could have gone through a high-temperature homogenous layers. and differentiation process and main­ tained a homogenous composition throughout. Seismic velocities in the crust and It is probably not a coincidence that the upper mantle maximum crustal thicknesses are close to the Seismic velocities in the crust and upper man­ basalt-eclogite boundary. Eclogite is denser than tle are typically determined by measuring the peridotite, at least in the shallow mantle, and transit time between an or explosion will tend to fall into normal mantle, thereby and an array of seismometers. Crustal compres­ turning a phase boundary (basalt- eclogite) into sional wave velocities in continents, beneath the a chemical boundary (basalt-peridotite). Some sedimentary layers, vary from about 5 kmfs at are less dense than the mantle below shallow depth to about 7 km/s at a depth of 410 km depth and may settle there, again turn­ 30-50 km. The lower velocities reflect the pres­ ing a phase boundary into a chemical boundary. ence of pores and cracks m.ore than the intrin­ Both the crust and the underlying lithosphere sic velocities of the rocks. At greater depths the have intrinsic densities that increase with depth. pressure closes cracks and the remaining pores That is, the outer shells of Earth are chemically are fluid-saturated. These effects cause a consid­ stratified, probably as a result of early differen­ erable increase in velocity. A typical crustal veloc­ tiation processes. This situation is in contrast to ity range at depths greater than 1 km is 6-7 kmfs. the bulk of the mantle, which is usually - and The corresponding range in shear velocity is THE SEISMIC LITHOSPHERE OR LID 97

about 3.5-4.0 kmfs. Shear velocities can be deter­ Table 8.7 1 Density, compressional velocity mined from both body waves and the and shear velocity in rock types found in of short-period surface waves. The top of the man­ sections tle under continents usually has velocities in the range 8.0-8.2 kmfs for compressional waves and p VP v. Poisson's 4.3-4.7 lanfs for shear waves. Rock Type (g/cm3) (km/s) (km/s) Ratio The compressional velocity near the base of the oceanic crust usually falls in the range Metabasalt 2.87 6.20 3.28 0.31 6.5-6.9 lanfs. In some areas a thin layer at the Metadolerite 2.93 6.73 3.78 0.27 base of the crust with velocities as high as Metagabbro 2.95 6.56 3.64 0.28 7.5 lm1/s has been identified. The oceanic upper Gabbro 2.86 6.94 3.69 0.30 mantle has a P-velocity (Pn) that varies from 3.23 7.64 4.43 0.25 Olivine gabbro 3.85 about 7.9 to 8.6 kmfs. The velocity increases with 3.30 7.30 0.32 oceanic age, because of cooling, and varies with Harzburgite 3.30 8.40 4.90 0.24 Durite 8.45 4.90 0.25 azimuth, due to crystal orientation or to dikes 3.30 and sills. The fast direction is generally close Salisbury and Christensen (1978), to the inferred spreading direction. The average Christensen and Smewing (1981). velocity is close to 8.2 la11/s. but young ocean has velocities as low as 7.6 lm1/s. Tectonic regions also Approx. Rock type SHEAR VELOCITY (P ~ 0) have low velocities. The shear velocity increases Depth Vs (km) denSity from about 3.6-3.9 to 4.4-4.7 lanfs from the base (glee) of the crust to the top of the mantle. 2.62 granodiOrite 2.68 Ophiolite sections found at some continental gneiss restite 2.74 2.77 margins are thought to represent upthrust or 13 gne1ss 2.79 d1orite 2.79 obducted slices of the oceanic crust and upper anorthosite 2.80 mantle. These sections grade downward from diorite 2.80 2.81 pillow to sheeted swarms, intrusives, gabbro 2.86 20 metabasalt 2.87 pyroxene and olivine gabbro, layered gabbro and gabbro 2.87 dolerite 2.93 peridotite and, finally, harzburgite and dunite gabbro 2.95 2.99 (Figure 8.3). Laboratory velocities in these rocks gne1 ss rest1te 2.98 amphibolite 3.07 are given in Table 8.7. There is good agree­ granul•te-mafic 3.10

ment between these velocities and those actually amphibole 3.20 observed in the oceanic crust and upper man­ pyroxenite 3.23 60 eclogite 3 24 tle. The sequence of extrusives, intrusives and 3.29 continental 3.29 moho cumulates is consistent with what is expected at 3.30 3.30 a midocean-ridge chamber. 3 31 The velocity contrast between the lower crust and upper mantle is commonly smaller beneath young orogenic areas (0.5-1.5 lanfs) than beneath and shields (1 - 2 kmfs). Continental increasing the thiclmess of the lower layer. Pale­ systems have thin crust (less than 30 km) and ozoic orogenic areas have about the same range low Pn velocities (less than 7.8 lm1/S ). Thinning of of crustal thiclmesses and velocities as platform the crust in these regions appears to take place areas. by thinning of the lower crust. In island arcs the crustal thickness ranges from about 5 km to 3 5 km. In areas of very thick crust such as in the The seismic lithosphere or LID (70 km) and the (80 lm1) , the thickening occurs primarily in the lower crustal The top of the mantle is characterized, in most layers. Oceanic crust also seems to thicken by places, by a thin high-velocity layer, a seismic lid. 98 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

This seismic lithosphere is not the same as the olivine-rich aggregates. Eclogites are highly vari­ plate or the thermal boundary layer. It is defined able but can have Vp and V, as high as 8.8 and solely on the basis of seismic velocities. In some 4.9 lm1js in certain directions and as high as places, e.g. Basin and Range, it is absent, perhaps 8.61 and 4.86 km/s as average values. The above due to delamination. The thiclmess of the LID suggests that corrected velocities of at least 8.6 corresponds roughly to the rheological or elastic and 4.8 kmfs, for Vp and V,, respectively, occur lithosphere, e.g. the apparently strong or coher­ in the lower lithosphere; this requires substan­ ent upper layer overlying the asthenosphere. The tial amounts of garnet, about 26%. The density thermal boundary layer is typically twice as thick, of such an assemblage is about 3.4 gjcm3 . The at least in oceanic regions. The high-velocity roots lower lithosphere may therefore be gravitation­ under cratons are chemically buoyant and are ally unstable with respect to the underlying man­ probably olivine-rich and FeO-poor. They extend tle, particularly when it is cold. The upper mantle to 200-300 km depth. They are long-lived because under shield regions, on the other hand, is con­ they are buoyant, cold, dry, high-viscosity and sistent with a very olivine-rich peridotite which is protected from plate-boundary interactions by buoyant and therefore stable relative to 'normal' the surrounding mobile belts. mantle. Uppermost mantle compressional wave veloc­ Anisotropy of the upper mantle is a poten­ ities, Pn. are typically 8.0-8.2 lunfs, and the tially useful petrological constraint, although it spread is about 7.9-8.6 kmfs. Some long refrac­ can also be caused by organized heterogeneity, tion profiles give evidence for a deeper layer such as laminations or parallel dikes and sills in the lithosphere having a velocity of 8.6 kmfs. or aligned partial melt zones, and stress fields. The seismic lithosphere, or LID, appears to con­ Recycled material may also arrange itself so as tain at least two layers. Long refraction profiles to give a fabric to the mantle. The uppermost on continents have been interpreted in terms mantle under oceans exhibits an anisotropy of of a laminated model of the upper 100 km about 7%. The fast direction is in the direction of with high-velocity layers, 8.6-8.7 kmfs or higher, spreading, and the magnitude of the anisotropy embedded in 'normal' material. Corrected to and the high velocities of P-arrivals suggest that normal conditions these velocities would be oriented olivine crystals control the elastic prop­ about 8.9-9.0 kmfs. The P-wave gradients are erties. Pyroxene exhibits a similar anisotropy, often much steeper than can be explained by self­ whereas garnet is more isotropic. The preferred compression. These high velocities require ori­ orientation is presumably due to the emplace­ ented olivine or large amounts of garnet. The ment or freezing mechanism. the temperature detection of 7-8% azimuthal anisotropy for both gradient or to nonhydrostatic stresses. A peri­ continents and oceans suggests that the shallow dotite layer at the top of the oceanic mantle is mantle at least contains oriented olivine or ori­ consistent with the observations. ented cracks, dikes or lens. In some places. the The anisotropy of the upper mantle, aver­ lithosphere may have formed by the stacking of aged over long distances, is much less than the subducted slabs, another mechanism for creating values given above. Shear-wave anisotropies of anisotropy. the upper mantle average 2-4%. Shear velocities Typical values of Vp and V, at 40 lun in the LID vary from 4.26-4.46 km/s. increas­ depth, when corrected to standard conditions, ing with age; the higher values correspond to a are 8.72 km/s and 4.99 kmjs, respectively. Short­ lithosphere 10-50 My old. This can be compared period surface-wave data implies STP - stan­ with shear-wave velocities of 4.3-4.9 kmfs and dard temperature and pressure - velocities of anisotropies of1-5% found in relatively unaltered 4.48-4.55 lm1js and 4.51-4.64 lrmfs for 5-My-old eclogites, at laboratory frequencies. The com­ and 25-My-old oceanic lithosphere. A value for Vp pressional velocity range in unaltered samples of 8.6 lunjs is sometimes observed near 40 lm1 is 7.6-8.7 lu11js, reflecting the large amounts of depth in the oceans. This corresponds to about garnet. 8.87 lunjs at standard conditions. These values Garnet and clinopyroxene may be impor­ can be compared with 8.48 and 4.93 km/s for tant components of the lithosphere and upper THE SEISMIC LITHOSPHERE OR LID 99

mantle. A lithosphere composed primarily of 0 40 80 120 160 200 peridotite does not satisfY the seismic data. The 0 lithosphere, therefore, is not just cold astheno­ sphere or a pure thermal boundary layer. The 20 roots of cratons, however, do seem to be com­ posed mainly of cold olivine but they are buoy­ ant in spite of the cold temperatures. They are 40 therefore unlikely to fall off. They are probably depleted residual after basalt extraction, and are E' 6 60 probably garnet- and FeO-poor. .<:: li It is likely that the upper mantle is lami­ Q) 0 nated, with the and melt products con­ Previous estimates 80 of seismic thickness centrated toward the top. As the lithosphere cools, underplated basaltic material is incorpo­ rated onto the base of the plate, and as the 100 D Seismic LID plate thickens it eventually transforms to eclog­ ite, yielding high velocities and increasing the I Elastic thickness thickness and mean density of the oceanic plate 120 (Figure 8.4). Eventually the lower part of the plate Age of oceanic lithosphere (My) becomes denser than the underlying astheno­ sphere, and conditions become appropriate for The thickness of the lithosphere as determined subduction or delamination. from flexural loading studies and surface waves. The thickness of the seismic lithosphere, or The upper edges of the open boxes gives the thickness high-velocity LID, is about 150-250 km under of the seismic LID (high-velocity layer, or seismic lithosphere). The lower edge gives the thickness of the the older continental shields. A thin low-velocity mantle LID plus the oceanic crust (Regan and Anderson, zone (LVZ) at depth, as found from body-wave 1984). The LID under continental shields is about studies, however, cannot be well resolved with 150-250 km thick. long-period surface waves. The velocity rever­ sal between about 150 and 200 km in shield areas is about the depth inferred for kimber­ (the peridotite assemblage) can have velocities lite genesis, and the two phenomena may be similar to mixtures of garnet-diopside-jadeite related. (the eclogite assemblage). Garnet-rich assem­ There is very little information about the deep blages, however, have velocities higher than oceanic lithosphere from body-wave data. Sur­ orthopyroxene-rich assemblages. face waves have been used to infer a thicken­ The Vp/Vs ratio is greater for the eclogite ing with age of the oceanic lithosphere to depths minerals than for the peridotite minerals. This greater than 100 kn1 (Figure 8.4). However, when ratio plus the anisotropy are useful diagnostics of anisotropy is taken into account, the thickness mantle mineralogy. High velocities alone do not may be only about 50 km for old oceanic litho­ necessarily discriminate between garnet-rich and sphere. This is about the thickness inferred for olivine-rich assemblages. Olivine is very anisotro­ the 'elastic' lithosphere from flexural bending pic, having compressional velocities of 9.89, 8.43 studies around oceanic islands and at trenches. and 7.72 kmfs along the principal crystallogra­ This is not the same as the thickness of the ther­ phic axes. Orthopyroxene has velocities ranging mal boundary layer (TEL) or the thickness of the from 6.92 to 8.25 lanfs, depending on direction. plate. In natural olivine-rich aggregates (Table 8.11), the The seismic velocities of some upper-mantle maximum velocities are about 8.7 and 5.0 lanfs minerals and rocks are given in Tables 8.9 for P-waves and S-waves, respectively. With 50% and 8.10. Garnet and jadeite have the highest orthopyroxene the velocities are reduced to 8.2 velocities, clinopyToxene and orthopyroxene the and 4.85 kmfs, and the composite is nearly lowest. Mixtures of olivine and orthopyToxene isotropic. Eclogites are also nearly isotropic. For a 100 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

Table 8.8 I Density and shear velocity of Table 8.1 0 I Densities and elastic-wave veloc- mantle rocks ities of upper-mantle rocks

Rock p vp v, Vp/Vs reflection Rock type SHEAR VELOCITY (P=O) depths density 4 6 Garnet 3.53 8.29 4.83 1.72 ('

Clark (1966), Babuska (1972), Manghnani and Ramananotoandro (1974), Jordan (1979). Table 8.9 I Densities and elastic-wave veloci- ties in upper-mantle minerals given density, eclogites tend to have lower shear p Vp V, velocities than peridotite assemblages. 3 Mineral (g/cm ) (km/s) Vp IV, The 'standard model' for the oceanic litho­ Olivine sphere assumes 24 km of depleted peridotite - Fo 3.2 14 8.57 5.02 1.71 complementary to and forming contemporane­ Fo93 3.311 8.42 4.89 1.72 ously with the basaltic crust- between the crust Fa 4.393 6.64 3.49 1.90 and the presumed fertile peridotite upper man­ tle. There is no direct evidence for this hypo· Pyroxene thetical model. It is a remnant of the En 3.21 8.08 4.87 1.66 model. The lower oceanic lithosphere may be En so 3.354 7.80 4.73 1.65 much more basaltic or eclogitic than in this sim­ Fs 3.99 6.90 3.72 1.85 ple model, and it might not have formed con­ Di 3.29 7.84 4.51 1.74 temporaneously. Buoyant, refractory and olivine­ Jd 3.32 8.76 5.03 1.74 rich lithologies may have accumulated in the Garnet shallow mantle for billions of years (Gyr). Basalts Py 3.559 8.96 5.05 1.77 can pond beneath plates that are under lateral AI 4.32 8.42 4.68 1.80 compression. Gr 3.595 9.31 5.43 1.71 Kn 3.85 8.50 4.79 1.77 An 3.836 8.51 4.85 1.75 The perisphere Uv 3.85 8.60 4.89 1.76

Sumino and Anderson (1984). The cratonic lithosphere is chemically buoyant relative to the underlying m antle, a result of melt depletion. The shallow mantle elsewhere may THE PERISPHERE 101

Table 8.1 I I Anisotropy of upper-mantle rocks

Mineralogy Direction Vp Vs, Vs2 Vp/Vs Peridotites 100 pet. ol I 8.7 5.0 4.85 1.74 1.79 2 8.4 4.95 4.70 1.70 1.79 3 8.2 4.95 4.72 1.66 1.74 70 pet. ol, I 8.4 4.9 4.77 1.71 1.76 30 pet. opx 2 8.2 4.9 4.70 1.67 1.74 3 8.1 4.9 4.72 1.65 1.72 100 pet. opx I 7.8 4.75 4.65 1.64 1.68 2 7.75 4.75 4.65 1.63 1.67 3 7.78 4.75 4.65 1. 67 1.67 Eclogites 51 pet. ga, 8.476 4.70 1.80 23 pet. cpx, 2 8.429 4.65 1.81 24 pet. opx 3 8.375 4.71 1.78 47 pet. ga, I 8.582 4.91 1.75 45 pet. cpx 2 8.379 4.87 1.72 3 8.30 4.79 1.73 46 pet. ga, I 8.31 4.77 1.74 37 pet. cpx 2 8.27 4.77 1.73 3 8.11 4.72 1.72

Manghnani and Ramananotoandro (1974}, Christensen and Lundquist (1982}.

also be enriched in refractory and buoy­ Ta ble 8. 12 1Measured and estimated ant products of mantle differentiation such properties of mantle minerals as olivine and orthopyroxene (the mantle perisphere). The base of this region may be Mineral related to the Lehmann discontinuity. Perisphere is derived from the prefix peri- mean­ Olivine (Fa. ) 3.37 8.31 4.80 1.73 12 ing all around, surrounding, near by. Peri in Persian ,B -Mg2Si04 3.63 941 548 1.72 folklore is a supernatural being descended from y-Mg2Si04 3.72 9.53 5.54 1.72 fallen angels or supernatural fairies and excluded Orthopyroxene 3.31 7.87 4.70 1.67 from paradise until penance is done. The peri­ (Fs 12) sphere is mainly buoyan t refractory peridotite. It Clinopyroxene 3.32 7.71 4.37 1.76 may be enriched in the large-ion lithophile (LIL) (Hd.12) elements, probably as a result of extraction of jadeite 3.32 8.76 5.03 1.74 these elements from slabs. A shallow enriched Garnet 3.68 9.02 5.00 1.80 layer is one alternative to a primordial deep Majorite 3.59 9.05 * 5.06* 1.79* layer. Perovskite 4.15 10.13* 5.69* 1.78* (Mg.19Fe.21 )0 4.10 8.61 5.0 1 1.72 There is evidence that two or three dis­ Stishovite 4.29 11.92 7.16 1.66 tinct lithologies contribute to the petrogenesis Corundum 3.99 10.86 640 1.70 of observed magm as. These probably correspond to recycled components such as oceanic crust, * Estimated. delaminated lower continental crust and peri­ Dufty and Anderson (1988), Weidner (1986). dotite. 102 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

0 4 5 6 01'~--~==~~~~~------i----, "------... ----- E'e. .<::: 160 Ci Q) 0 240 200 8 9 3.5 4 4.5 5 Velocity (kmls)

Velocity-depth profiles for the average Earth, as determined from surface waves (Regan and Anderson, 1984). 'E 400 From left to right, the graphs show P-wave velocities (vertical e. and horizontal), Swave velocities (vertical and horizontal) and .<:. Ci Q) an anisotropy parameter ( I represents isotropy). 0

600 The low-velocity zone or LVZ

A region of diminished velocity or negative veloc· ity gradient in the upper mantle was proposed by Beno Gutenberg in 1959. Earlier, just after 800 had been established, it had been con· eluded that a weak region underlay the rela· tively strong lithosphere. This has been called the Velocity (kmls) asthenosphere. The discovery of a low-velocity zone strengthened the concept of an asthenosphere. Shear-wave velocity profiles for various tectonic Most models of the velocity distribution in provinces; TNA is tectonic North America, SNA is shield the upper mantle include a region of high gra· North America, ATL is north Atlantic. It is difficult to resolve the small variations below 400 km depth (after Grand and dient between 250 and 350 km depth. Lehmann Heimberger, 1984a) . (1961) interpreted h er results for several regions in terms of a discontinuity at 220 lm1 (some· times called the Lehmann discontinuity), and velocity can decrease with depth. It can and many subsequent studies give high-velocity be shown, however, that a high temperature gradients near this depth. Although the global gradient alone is not an adequate explanation. presence of a discontinuity, or high-gradient, and dislocation relaxation both

region near 220 km has been disputed, there is cause a large decrease in velocity. and C02 now appreciable evidence, from reflected phases, decrease the solidus and the seismic velocities. for its existence. The situation is complicated by For partial melting to be effective the melt must the extreme lateral heterogeneity of the upper occur, microscopically, as thin grain boundary 200 km of the mantle. This region is also low Q films or, macroscopically, as dikes or sills, which (high attenuation) and anisotropic. Some upper also are very small compared with · mantle models are shown in Figures 8.5 and 8.6. lengths. Melting experiments suggest that melt· Various interpretations have been offered for ing occurs at grain corners and is more likely to the upper mantle low-velocity zone. This is occur in interconnected tubes. This also seems undoubtedly a region of high thermal gradient, to be required by electrical conductivity data. the boundary layer between the near surface However, slabs, dikes and sills act macroscopi· where heat is transported by conduction and cally as thin films for long-wavelength seismic the deep interior where heat is transported by waves. . If the temperature gradient is high High attenuation is associated with relaxation enough, the effects of pressure can be overcome processes such as grain boundary relaxation, THE LOW-VELOCITY ZONE OR LVZ 103 including partial melting, and dislocation relax­ Tomographic results show that the lateral ation. Small-scale heterogeneity such as slabs and variations of velocity in the upper mantle are dikes scatter seismic energy and this mimics as pronounced, and abrupt, as the velocity varia­ intrinsic anelasticity. Allowance for anelastic dis­ tions that occur with depth. Thus, it is mislead­ persion increases the inferred high-frequency ing to think of the mantle as a simple layered velocities in the low-velocity zone determined or lD system. Lateral changes are, of course, by free-oscillation and surface-wave techniques, expected in a convecting mantle because of vari­ but partial melting is still required to explain ations in temperature and anisotropy due to crys­ the regions of very low velocity. Allowance for tal orientation. They are also expected from the anisotropy results in a further upward revision operation of and crustal processes. for the velocities in this region, as discussed In particular, lithologic heterogeneity is intro­ below. This plus the recognition that subsolidus duced into the mantle by subduction and lower effects, such as dislocation relaxation, can cause crustal delamination. But phase changes and par­ a substantial decrease in velocity has complicated tial melting are more important than tempera­ the interpretation of seismic velocities in the ture and this is why the major lateral changes shallow mantle because one wants to compare are above about 400 km depth. results with laboratory data. Velocities in tectonic The geophysical data (seismic velocities, atten­ regions and under some oceanic regions, how­ uation, heat flow) are consistent with par­ ever, are so low that partial melting is implied. tial melting in the low-velocity regions in the In most other regions a subsolidus mantle com­ shallow mantle. This explanation, in turn, sug­ posed of oriented olivine-rich aggregates can gests the presence ofvolatiles in order to depress explain the velocities and anisotropies to depths the solidus of mantle materials, or a higher­ of about 200 km. Global tomographic inversions temperature mantle than is usually assumed. The involve a laterally heterogenous velocity and top of the low-velocity zone may mark the cross­ anisotropy structure to depths as great as 400 km. ing of the geotherm with the wet solid us of peri­

Mantle tomography averages the seismic dotite, or the solidus of a peridotite-C02 mix. velocity of the mantle over very long wavelengths Its termination would be due to (1) a crossing and travel distances. Sampling theory tells us in the opposite sense of the geotherm and the that extremes of velocity are averaged out in this solidus, (2) the absence of water, C02 or other procedure. As paths get shorter and shorter the volatiles or (3) the removal of water into high­ variance goes up and sections of the path become pressure hydrous phases or escape of C02 . In much slower or much faster than inferred from all of these cases the boundaries of the low­ global tomography. The minimum shear veloc­ velocity zone would be expected to be sharp. ity found along ridges and backarc basins - and Small amounts of melt or fluid (about 1 %) can probably elsewhere - in high-resolution studies, explain the velocity reduction if the melt occurs is smaller than inferred from tomography. If as thin grain-boundary films. Considering the global shear velocities approach the minimum wavelength of seismic waves, magma-filled dikes shear velocities in solid rocks, then a reasonable and sills, rather than intergranular melt films , variance of 5% placed on top of this probably would also serve to decrease the seismic veloc­ means that the low-velocity regions require some ity by the appropriate amount. Slabs that are melting. thin and hot at the time of subduction may The rapid increase in velocity below 220 km stack up in the shallow mantle. The basaltic parts may be due to chemical or compositional changes will melt since the solidus of basalt is lower (e.g. loss of water or C02 ) or to transition from than the ambient temperature of the mantle. relaxed to unrelaxed moduli. The latter expla­ At seismic wavelengths this will have the same nation will involve an increase in Q, and some effect as intergranular melt films. Anelasticity Qmodels exhibit this characteristic. However, the and anisotropy of the upper mantle may also be resolving power for Q is low, and most of the due to these mega-scale effects rather than due seismic Q data can be satisfied with a constant-Q to crystal physics, dislocations, oriented crystals upper mantle, at least down to 400 km. and so on. 104 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

300-400 lun even if melt-solid separation does not occur until shallower depths. Low-melting

!!!_ point materials may be introduced into the shal­ ~ 8.5 low mantle from above, and heat up by conduc­ tion from ambient mantle. do not ·~g have to initiate in thermal boundary layers. ~ Although a strong case can be made for local­ (ij g 8.0 ized or regional partial melting in the shallow ·u; U) mantle, the average seismic velocity over long Q) 6 Rise-Tectonic a. • No. Atlantic paths, such as are used in most tomographic E o Shield 0 studies, may imply subsolidus conditions. If the (.) +W. Pacific partial melt zones are of the order of tens to hun­ dreds of kilometers in lateral dimensions, and tens of kilometers thick, then they will serve to lower the average velocity, and perhaps to intro­ duce anisotropy, in global tomographic models. There are also compositional effects to be consid­ ered. Eclogite, for example, has lower shear veloc­ ities than peridotite at depths between about 100 and 600 km. Eclogite, however, also has a that is lower than ambient mantle temperature.

0 100 200 300 400 Absorption and the LVZ Depth (km) Elastic-wave velocities are independent of fre­ quency only for a nondissipative medium. In Compressional and shear velocities for two a real solid dispersion must accompany absorp­ petrological models, pyrolite and piclogite, along various tion. The effect is small when the seismic quality adiabats. The temperatures ( C) are for zero pressure. The portions of the adiabats below the solidus curves are factor Q is large or unimportant if only a small in the partial melt field. The seismic profiles are for two range of frequencies is being considered. Even in shields (Given and Heimberger, 1981 ; Walck, 1984), a these cases, however, the measured velocities or tectonic-rise area (Grand and Heimberger, 1984a; Walck, inferred elastic constants are not the true elas­ 1984), and the North Atlantic region (Grand and tic properties but lie between the high-frequency Heimbe rger, 1984b). and low-frequency limits or the so-called 'unre­ laxed' and 'relaxed' moduli. The melting that is inferred for the lower The magnitude of the effect depends on the velocity regions of the upper mantle may be ini­ of the absorption band and the value of tiated by adiabatic ascent from deeper levels. The Q When comparing data taken over a wide fre­ high compressibility and high content of quency band, the effect of absorption can be con­ melts means that the density difference between siderable, especially considering the accuracy of m.elts and residual crystals decreases with depth. present body-wave and free-oscillation data. The High temperatures and partial melting tend to presence of physical dispersion complicates the decrease the garnet content and thus to lower problem of inferring temperature, chemistry and the density of the mantle. Buoyant from mineralogy by comparing seismic data with high­ depths greater than 200 lun will extensively m.elt frequency ultrasonic data. Anelasticity as well on their way to the shallow mantle. Therefore, as anharmonicity is involved in the temperature partially molten material as well as melts can dependence of seismic velocity. be delivered to the shallow mantle. The ulti­ Figures 8.8 and 8.9 show calculations for seis­ mate source of some basaltic melts may be below mic velocities for different mineral assemblages. THE TRANSITION REGION (TR) 105

50 (MgFe)2Si04 from the olivine to the ,B -spinel - - structure (Table 8.12). 111e best fitting N mineralogy at this depth contains less than SO % ~ E olivine. The velocity jump at 410 km is too small 6 40 ~ to accommodate all the olivine (ol) and ortho­ pyroxene (opx) in a pyrolite mantle converting to spinel and majorite; there must be substantial gt 30 and cpx. " " " " " In some places there is a shear-wave 1.82 velocity drop of about 5 % on top of the 410 km discontinuity. This low-velocity zone has a variable thickness ranging up to per­ haps 90 km. This may be due to a dense partial­ • melt layer, in which the solidus has been reduced '" No. Atlantic .. " Rise - Tectonic by the presence of eclogite or C02 . Eclogite itself, • Shield at depth, has a lower shear velocity than peri­ + w. Pacific dotite and even cold eclogite can be a low-shear 1 .62 l__.J.___.l.___..J__....[__ __L__ __L.._ _J__ _l velocity zone at 400 km. 0 1 00 200 300 400 Depth (km)

Seismic parameters for two petrological models Anisotropy and various seismic models. Symbols and sources are the same as in Figure 8.7. Vp/Vs ratios for various minerals are Rayleigh and Love wave data are often 'incon­ shown in the lower panel. The high Vp/Vs ratio for the sistent' in the sense that they cannot be fit rise-tectonic mantle is consistent with partial melting in the simultaneously using a simple isotropic model. upper mantle under these regions. This has been called the Love-wave-Rayleigh­ wave discrepancy, and attributed to anisotropy. Pyrolite is a garnet peridotite composed mainly There are now many studies of this effect of olivine and orthopyroxene. Piclogite is a which has also been called polarization clinopyroxene- and garnet-rich aggregate with or radial anisotropy, or transverse some olivine. Note the similarity in the isotropy. Independent evidence for anisotropy calculated velocities. Below 200 km the seismic in the upper mantle is now strong (e.g. from velocities under shields lie near the 1400 oc adi­ receiver-functions amplitudes and shear-wave abat. Above 150 km depth the shield lithosphere splitting). Radial anisotropy is very strong in the is most consistent with cool olivine-rich mate­ central Pacific. Anisotropy can be caused by crys­ rial. TI1e lower velocity regions have velocities tal orientation or by a fabric of the mantle caused so low that partial melting or some other high­ by slabs, dikes and sills. Seismic waves have such temperature relaxation mechanism is implied. long wavelengths that it is immaterial whether TI1e adiabats falling below the solidus curves the effect is due to centimeter- or tens-of-km-size are predicted to fall in the partial-melt field. At features. Similarly, anelasticity may be due to km­ depth, eclogite and piclogite have lower shear size scatters, rather than to em-sized dislocations. velocities and lower melting temperatures than peridotites. Low shear-velocity regions may be The transition region (TR) subducted or delaminated crust. The mantle transition region, Bullen's The 410 km discontinuity region C is defined as that part of the man­ tle between 410 km and 1000 km (the Repetti The seismic discontinuity at 410 km depth is discontinuity). The lower mantle was defined generally attributed to the of by Bullen as the man tle below 1000 km depth 106 LET'S TAKE IT FROM THE TOP: THE CRUST AND UPPER MANTLE

but more recently it is often taken as the man­ Thermodynamic considerations have been tle below the 650 lan seismic discontinuity. used to argue that the discontinuities are abrupt This has caused immense confusion regarding phase changes of, mainly, olivine to the spinel whether slabs sink into the lower mantle. The crystal structure, and then to a 'post-spinel' 400 krn discontinuity is mainly due phase, not chemical changes as in the standard to the olivine-spinel phase change , geochemical models, and that the deeper one has considered as an equilibrium phase a negative Clapyron slope. This means that cold boundary in a homogenous mantle. The subducting material of the same composition seismic velocity jump, however, is smaller as the surrounding mantle would depress the than predicted for this phase change. The 650 ian discontinuity, inhibiting vertical motion, orthopyroxene-garnet reaction leading to a gar­ and would change to the denser phase only after net solid solution is also complete near this warming up or being forced to greater depth. depth, possibly contributing to the rapid increase Material of different composition and intrinsic of velocity and density at the top of the transi­ density can be permanently trapped at phase tion region. Some eclogites are less dense than boundaries. Geochemical, and many convection, the beta-form of olivine and may be perched at models assume that the 650 km phase change 400 km depth. For these reasons the 400 km separates the 'depleted convecting upper mantle' discontinuity should not be referred to as the from the 'primordial undegassed lower mantle.' olivine-spinel phase change. If the discontinuity Geodynamic modelers assume that if the 650 lm1 is as small as in recent seismic models, then the discontinuity is not a chemical change then there olivine content of the mantle may be lower than can be no deeper chemical change and the man­ in the shallow mantle. This would make sense tle is chemically homogenous. The TZ thus holds since olivine is a buoyant product of mantle dif­ the key to whether there is whole-mantle or ferentiation and would tend to accumulate at the layered-man tie convection. top of the mantle. In the the stable phases are The TR as described by Bullen is a dif­ garnet solid-solution, {3 - and y-spinel and, possi­ fuse region of high seismic wave-speed gradient bly, jadeite. Garnet solid-solution is composed of

extending from 410 to 1000 km. In Bullen's ordinary garnet and Si02-rich garnet (majorite). nomenclature the lower mantle (Region D) The extrapolated elastic properties of the spinel started at 1000 lm1. Birch suggested that the forms of olivine are higher than those observed Repetti discontinuity near 1000 km marked the (Figures 8.9 and 8.10). The high velocity gradients top of the lower mantle and that high seismic throughout the transition zone imply a continu­ wave-speed gradients are caused by polymorphic ous change in chemistry or phase, or in lithol­ phase changes. The early models of Jeffreys and ogy (eclogite vs . peridotite). Appreciable garnet Gutenberg were smooth and had high wave-speed is implied in order to match the velocities. A gradients without abrupt discontinuities, but in spread-out phase change involving clinopyroxene the 1960s it was discovered that there are abrupt (diopside(di) plus jadeite(jd)) transforming to Ca­ jumps in seismic velocity at depths of approx­ rich majorite(mj) can explain the high velocity imately 400 and 650 lrm. During that decade, gradients. various investigations, including detailed stud­ Unusually low temperatures, as expected in ies of the travel times of both first- and later­ the vicinity of a downgoing slab, will warp the arriving body waves, seismic array measurements 410 km discontinuity up by about 8 km per 100 K, of apparent velocities, observations of reflected and the 650 lm1 discontinuity down by about waves such as precursors to the core phase P'P', 5 km per 100 K. The Clapy.ron slopes are uncer­ and analysis of the dispersion of fundamental tain but most estimates for the total thickening/ and higher-mode surface waves, all confirmed the thinning of the TZ lie in the range 12-17 lrm per existence of the discontinuities, which define the 100 degrees. This assumes a purely olivine min­ transition Z are, TZ. eralogy, which is unrealistic. THE TRANSITION REGION (TR) 107

Calculated compressional velocity versus depth for various mantle minerals along a I 400 C adiabat. 11

~ ~ -5 10 0 ~ o; c: ·0u;

200 400 600 800 1000 Depth (km)

upwellings. However, the depths of the 410 and 650 km discontinuities are uncorrelated on a global scale. They also sometim.es show steps, which makes interpretation in terms of tempera­ ture not straightforward. Some places show rapid lateral changes in TZ thickness that may indicate non-thermal effects. It must therefore be borne in mind that temperature may not be the only control. The expected effects of temperature on the depths of the discontinuities are also based on uncertain laboratory calibrations, may be in error by a factor of two, and chemical effects may be stronger than generally supposed. In spite of Depth (km ) these complications, TZ thickness may still prove Calculated shear velocity versus depth for various to be a useful thermometer and an important mantle minerals along a 1400 C adiabat. 'Majorite' (mj), part of any plan to map lateral variations of tem­ 'perovskite' (pv) and 'ilmenite' (il) are structural, not perature in the mantle. mineralogical terms. The dashed lines are two representative There is no global correlation between TZ seismic profiles (after Duffy and Anderson, 1989). thickness and the locations of surface hotspots and the large lower-mantle low-velocity regions. Transition zone thickness is normal beneath The average thickness of the TZ is ~ 242 km; southern Africa (245 km) and the East African uncertainties are typically 3%. Typical thick­ Rift and Afar (244 km), which are underlain nesses beneath high-heat-flow areas are 220- by the postulated 'superplumes' and the pos­ 230 km. The topology of the relevant phase tulated Afar plume. Transition-zone thickness diagrams predicts antisymmetry in the direc­ beneath hotspots is generally within the range tions of deflection of the discontinuities for for normal oceans and often close to the global the cases of both cold downwellings and hot average. 108 LET'S TAKE IT FROM THE T O P: THE C RU ST AND U PPER MANTLE

The observed topography on the discontinu­ penetrate the phase-change region they will sink ities does not seem to be explicable by thermal to the core-mantle boundary. The transition effects to the extent expected. TI1e observations zon e of the upper mantle therefore contin­ are consistent with the decorrelation of seismic ues to be a critical region for investigation. anomalies between the upper and lower man­ tles observed both in tomographic images and The Repetti discontinuity revealed by matched filtering using plate/slab A layered convection model with a chemical reconstructions. A few anomalies appear to interface near 900 km at the base of Bullen's extend from the surface through the TZ and Region C explains the geoid and dynamic topog­ into the deep mantle; it would be interesting to raphy. The evidence for stratification includes calculate if they are more numerous than would the mismatch between tomographic patterns and be expected by chance. spectra between various depth regions, and evi­ Detailed study of some specific regions have dence for slab flattening. TI1ere is a good correla­ yielded surprises. The thinnest TZ region, 181 km tion between subducted slabs and seismic tomog­ thick, is found in Sumatra, where a thick accu­ raphy in the 900-1100 km depth range. The man­ mulation of cold slabs is thought to exist, which tle does not become radially homogenous and would thicken the TZ. In western USA, the thick­ adiabatic until about 800 km depth. A variety ness of the TZ varies from 220 to 270 Jan, with of evidence suggests that there might be an 20 to 30 km relief on each discontinuity, and no important geodynamic boundary, possi­ correlation with surface geology, topography or bly a barrier to convection, and ath er­ between the discontinuities [mantleplumes]. mal boundary at a depth of about 900- There is evidence from scattering of seis­ 1000 km, the bottom of the transition region. mic waves and plate-tectonic-tomographic corre­ Chemical boundaries, in contrast to most lations that there may be a chemical boundary phase-change boundaries, will not be flat, as near 1000-km depth. Much of modern mantle assumed in some layered convection models, and geochemistry is based on the conjecture that the will have little impedance contrast. The latter 650-km phase change is also a major chemical inference is based on plausible compositional change, and that this is the boundary between differences between various lower-mantle assem­ the upper and lower mantles. Geodynamic blages. Complications between 650 and 1300 km models assume that below 650 km depth, the depth in the mantle are perhaps related to slab mantle radioactivity is high. Mantle trapping or thermal coupling and undulations in is also based on the assumption that if slabs can the Repetti discontinuity, the top of Region D.