Chapter 24

The nature and cause of heterogeneity

The right to search for truth implies processes internal to the mantle. The fertility also a duty. One must not conceal any and fertility heterogeneity of the are due to of young plates, aseismic part of what one has discovered to ridges and seamount chains, and to delami­ be true. nation of the lower continental , as dis­ Albert Einstein cussed in Part II. TI1ese heterogeneities eventually warm up past the melting point of and become buoyant low-seismic-velocity diapirs that O verview undergo further adiabatic decompression melt­ ing as they encounter thin or spreading regions The clearly controls the location of of the lithosphere. volcanism. The nature and volume of the volcan­ The heat required for the melting of cold ism and the presence of 'melting anomalies' or subducted and delaminated material is extracted 'hotspots,' however, reflect the intrinsic chemi­ from the essentially infinite heat reservoir of cal and petrologic heterogeneity of the upper the mantle, not the core. Melting in the upper mantle. Melting anomalies - shallow regions of mantle does not require an instability of a deep ridges, volcanic chains, flood , radial dike thermal boundary layer or high absolute tem­ swarms- and continental breakup are frequently peratures. Melts from fertile regions of the attributed to the impingement of deep mantle mantle, recycled and subducted thermal plumes on the base of the lithosphere. seamounts, can pond beneath the lithosphere, The heat required for volcanism in the plume particularly beneath basins and suture zones, hypothesis is from the core; plumes from the with locally thin, weak or young lithosphere, or deep mantle create upper mantle heterogeneity. they can erupt. The stress state of the lithophere This violates the dictum of good science: can control whether there is underplating and never go for a deep complex explanation if a shallow sill intrusion, or eruption and dike intrusion. simple one will do. Absolute mantle temperature has little to do with Mantle fertility and melting point variations, this. ponding, focusing and edge effects, i.e. plate tec­ The characteristic scale lengths - 150 km to tonic and near-surface phenomena, may control 600 km- of variations in bathymetry and magma the volumes and rates of . The magni­ chemistry, and the variable productivity of vol­ tude of magmatism may reflect the fertility and canic chains, probably reflect compositional het­ homologous temperature, not the absolute tem­ erogeneity of the asthenosphere, not the scales perature, of the asthenosphere. The chemical and of mantle or the spacing of hot isotopic heterogeneity of the mantle is, in part, plumes. High-frequency seismic waves, scatter­ due to recycling and, in part, due to igneous ing, coda studies and deep reflection profiles are MANTLE HOMOGENEITY: THE OLD PARADIGM 313 needed to detect the kind of chemical het­ on are products of these perceived constraints. erogeneity - blobs and small-scale layering - Absolute temperature, not lithologic diversity, predicted from the recycling and crustal delami­ is the controlling parameter in these models of nation hypotheses. geochemistry and geodynamics, and in the usual interpretations of seismic images and crustal thickness. Mantle homogeneity: the old Th e perception that the mantle is lithologi­ paradigm cally homogenous is based on two assumptions: (1) the bulk of the upper mantle is roughly Global tomography and the geoid characterize isothermal (it has constant potential tempera­ the large-scale features of the mantle. Higher ture) and (2) midocean-ridge basalts are so uni­ frequency and higher resolution techniques are form in composition ('the convecting mantle' required to understand the smaller-scale fea­ is geochemical jargon for what is viewed as tures, and to integrate geophysics with tecton­ 'the h omogenous well-stirred upper mantle') that ics and with mantle petrology and geochemistry. departures from the basic average composition The upper mantle is often regarded as being of basalts along spreading ridges and within extremely homogenous, based on low-resolution plates must come from somewhere else. The tomographic studies and the chemistry of mido­ only way thought of to do this is for nar­ cean ridge basalts. Both of these approaches aver­ row jets of hot, isotopically distinct, mantle to age out the underlying heterogeneity of the man­ arrive from great depths and impinge on the tle. The intrinsic chemical heterogeneity of the plates. shallow mantle, however, is now being recog­ The fact that bathymetry follows the square nized. This heterogeneity contributes to the iso­ root of age relation is an argument that the cool­ tope diversity of magmas and the scattering of ing plate is the main source of density varia­ seismic waves. Melting anomalies themselves - tion in the upper mantle. The scatter of hotspots and swells - reflect lithologic hetero­ depth and heat flow - and many other param­ geneity and variations in fertility and melting eters - as a function of age, however, indicates point of the underlying mantle. The volume of that something else is going on. Plume influence is related more to lithology of the shal­ is the usual, but non-unique, explanation for low mantle than to absolute temperature. Thus, this scatter, and for depth and chemical anoma­ both the locations of volcanism and the volume lies along the ridge. Lithologic (major elements) of volcanism can be attributed to shallow- litho­ and isotope homogeneity of the upper mantle spheric and asthenospheric- processes, processes are two of the linchpins of the plume hypoth­ that are basically athermal and that are intrinsic esis and of current geochemical reservoir mod­ to plate . els. Another is that seismic velocities, crustal Much of mantle geochemistry is based on the thicknesses, ocean depths and eruption rates are assumption of chemical and mineralogical homo­ proxies for mantle potential tempera­ geneity of the shallow mantle, with so-called tures. The asthenosphere, however, is variable normal or depleted m.idocean-ridge basalt in melting temperature and fertility (ability to (NMORB and DMORB) representative of the produce magma) and this is due to recycling homogeneity and depletion of the entire upper of oceanic crust and delaminated continental mantle source (the convecting upper mantle). crust and lithosphere. In addition, seismic veloc­ The entire upper mantle is perceived to be ities are a function of lithology, phase changes a homogenous depleted olivine-rich lithology and melting and are not a proxy for tempera­ approximating pyrolite (pyroxene-olivine-rich ture alone. Some lithologies melt at low tem­ ) in composition; all basalts are formed by perature and have low seismic velocities with­ melting of similar peridotitic lithologies. Venera­ out being hotter than adjacent mantle. These ble concepts such as isolated reservoirs, plumes, can be responsible for melting and tomographic temperature-crustal-thickness relations and so anomalies. 314 THE NATURE AND CAUSE OF MANTLE HETEROGENEITY

The isotopic homogeneity of NMORB has Earth, composed of materials with different strongly influenced thinking about the presumed intrinsic densities, will tend to stratify itself by homogeneity of the upper mantle and the inter­ density. Plate-tectonic processes introduce het­ pretation of 'anomalous' sections of midocean erogeneities into the mantle, some of which ridges. It is common practice to avoid 'anoma­ can be mapped by geophysical techniques. On lous' sections of the ridge when compiling MORB the other hand, diffusion, chaotic advection properties, and to attribute anomalies to 'plume­ and vigorous unidirectional stirring, are homog­ ridge interactions.' In general, anomalies along enizers. Convection is thought by many geo­ the ridge system - elevation, chemistry, physi­ chemists and modelers to homogenize the man­ cal properties - are part of a continuum and tle. Free convection driven by buoyancy is not the distinction between 'normal' and 'anoma­ the same as stirring by an outside agent. Melt­ lous' ridge segments is arbitrary and model ing of large volumes of the mantle, as at ridges, dependent. however, can homogenize the basalts that are erupted, even if they come from a heterogenous mantle. Mantle heterogeneity: toward There are numerous opportunities for gener­ a new paradigm ating (and removing) heterogeneities associated with (Figure 24.1). The tempera­ It is increasingly clear that the upper mantle is tures and melting temperatures of the mantle heterogenous in all parameters at all scales. The depend on plate-tectonic history and processes evidence includes seismic scattering, anisotropy, such as insulation and subduction cooling. Ther­ mineralogy, major- and trace-element chemistry, mal convection requires horizontal temperature isotopes, melting point and temperature. An gradients; cooling from above and subduction of isothermal homogenous upper mantle, however, plates can be the cause of these temperature gra­ h as been the underlying assumption in much of dients. The mantle would convect even if it were mantle geochemistry for the past 35 years. not heated from below. Radioactive heating from One must distinguish fertility from (trace ele­ within the mantle, secular cooling, density inho­ ment) enrichment, although these properties may mogeneities and the surface thermal-boundary be related. Fertility implies a high basalt-eclogite layer can drive . An additional or plagioclase-garnet content. Enrichment implies important element is the requirement that ridges high contents of incompatible elements and long­ and trenches migrate with respect to the under­ term high Rb/Sr, U/Pb, Nd/Sm etc. ratios. Because lying mantle. Thus, mantle is fertilized, contam­ of buoyancy considerations, the most refractory inated and extracted by migrating boundaries - products of mantle differentiation - harzburgite a more energy-efficient process than moving the and lherzolite- may collect at the top of the man­ mantle to and away from stationary plate bound­ tle and bias our estimates of mantle composition. aries, or porous flow of magma over large dis­ The volume fractions and the dimensions of the tances. Lateral return flow of the asthenosphere, fertile components - basalt, eclogite, pyroxenite, and entrained mantle flow, are important ele­ piclogite - of the mantle are unknown. There ments in plate tectonics. Embedded in these flows is no reason to suppose that the upper mantle can be fertile patches. Even if they are confined to is equally fertile everywhere or that the fertile the asthenosphere these patches will move more patches or veins in hand specimens and outcrops slowly than plates and plate boundaries, giving are representative of the scale of heterogeneity the illusion of fixed hotspots. in the mantle. There are three kinds of heterogeneity of Creation of mantle heterogeneity interest to petrologists and seismologists, radial, Recycling contributes to chemical and iso­ lateral and random, or statistical. Melting and topic heterogeneity of the source regions of gravitational differentiation stratify the man­ basalts but it also contributes to the fertility tle. Given enough time, a petrologically diverse and productivity of the mantle. Temperature FATE OF RECYCLED MATERIAL 315

seamount ophiolites pluton volcanic arc batholiths

basalt

Asthenosphere

flat subduction

Illustration of one end game of the plate-tectonic more than 10% of the seafloor area is capped cycle- the closure of ocean basins. The other end of the by seamounts and plateaus. The delamination cycle is continental breakup, when this diverse material can of over-thickened also intro­ be involved in breakup magmatism. The many oceanic duces fertile material into the asthenosphere; plateaus in the newly opened Atlantic and Indian ocean basins this is warmer and perhaps thicker than sub­ may be, in part, this reactivated material. ducted oceanic crust, and will equilibrate and melt sooner. These warm delaminates are poten­ tial fertile spots and can create melting anoma­ variations are long wavelength while chemical lies. They may account for 5% of all recycled heterogeneity can be of the scale of slabs and material. the source regions of volcanoes. Melting anoma­ The rate at which young oceanic crust and lies may be primarily due to high homologous, delaminated lower continental crust enters the not high absolute, temperature. mantle is comparable to the global rate of Basalts, mafic and ultramafic cumulates and 'hotspot' volcanism, ~ 2 km3 fyr. Melting anomalies depleted harzburgitic rock are constantly formed may therefore be due to fertile blobs. The fates along the 60 000-lan-long mid-ocean spreading­ of older plates with thin crust, and deeper slabs ridge system. The mantle underlying diverging are different. and converging plate boundaries undergoes par­ tial melting down to depths of order 50-200 Ian in regions up to several hundred kilometers Fate of recycled material wide, the processing zone for the formation of magmas - MORB, backarc-basin basalts (BABB) Large-scale chemical heterogeneity of basalts and island-arc basalts (lAB). Midplate volcanoes sampled along midocean-ridge systems occur on and off-axis seamounts process a much smaller length scales of 150 to 1400 Ian. This hetero­ volume of mantle, and the resulting basalts are geneity exists in the mantle whether a migrating therefore - as a consequence of the central limit ridge is sampling it or not. Fertile patches, how­ theorem- much more heterogenous. ever, are most easily sampled at ridges and may Before the oceanic plate is returned to the explain the enigmatic relations between physi­ upper mantle in a subduction zone, it accu­ cal and chemical properties along ridges. Nor­ mulates sediments and the harzburgites become mal oceanic crust may start to melt after being serpentinized; this material enters the man­ in the mantle for ~ 60 million years (Myr), while tle (Figure 24.1). Plateaus, aseismic ridges and delaminated crust may melt after only 20-40 Myr seamount chains also head toward trenches. because it starts out warmer (Figure 5.2). About 15% of the current surface area of Tomographic correlations suggest that is composed of young ( < 20 My) lithosphere and middle-aged plates reside mainly in the bottom 316 I THE N ATUR E AND CAUSE OF MANTLE H ET ERO GENEITY

part of the transition region, near and just chemical heterogeneity at the tens of km scale, below 650 km. Plates that were young ( < 30 Myr) the scales of recycled crust and lithosphere. The at the time of subduction (e.g. Farallon slab hundreds of km scales are comparable to the under western North America) and slabs sub­ segmentation of ridges, trenches and fracture ducted in the past 30 Myr may still be in the zones, and the scales of delaminated crust along upper mantle. Old , thic k s labs a ppear island arcs. Chunks of slabs having dimensions t o collect at 7 5 0- 900 km, the probable of tens by hundreds of km are inserted into the base of the layer accessible to surface volca­ mantle at trenches. They are of variable age, noes. The quantitative and statistica l and equilibrate and are sampled over various meth ods of d e t ermini n g the d e pth o f time scales. The lateral dimensions of plates, and s ubducti o n and comparison of t o mographic the separation distances of trenches and aseis­ and geod y namic mo d els are superior to the mic ridges are also likely to show up as scale visual analysis of selected color tomographic lengths in chemical and physical variations along cross-sections - qualitative chromotomography. ridges. A chemically stratified mantle will have some TI1e central limit theorem (CLT) is essential deep high-velocity patches and some will appear in trying to understand the range and variabil­ to correlate with shallower structures. ity of mantle products extracted from a het­ erogenous mantle. Depending on circumstances, small domains - tens to hundreds of km in Scale of mantle heterogeneity extent - can also be isolated for long periods of time until brought to a ridge or across the In the plume model isotopic differences are melting zone. Mineralogy, diffusivity and solu­ attributed to different large (400-2000 km in bility are issues in determining the size of iso­ extent) reservoirs at different depths. In the latable domains. When a multi-component man­ marble cake and plum-pudding models the tle warms up to its solidus - not necessarily the characteristic dimensions of isotopic hetero­ same as the surrounding mantle - the erupted geneities are centimeters to meters. Chemical magmas can be variable or homogenous; this is differences between ridge and nearby seamount controlled by sampling theory, the statistics of and island basalts may be due, in part, to the large numbers and the CLT. Even under a ridge nature of the sampling of a common heteroge­ the melting zone is composed of regions of vari­ nous region of the upper mantle. In order for able melt content. The deeper portions of the this to work there must be substantial chem­ zone, and those regions on the wings. will expe­ ical differences over dimensions comparable to rience small degrees of melting but these will be the volume of mantle processed in order to fuel blended with high-degree melts under the ridge, the volcano in question, e.g. tens to hundreds prior to eruption. Magma cannot be considered ofkm. to be uniform degrees of melting from a chem­ Chemical differences along ridges have char­ ically uniform mantle. Blending of magmas is acteristic scales of 200-400 km. Inter-island dif­ an alternate to the point of view that convec­ ferences in volcanic chains, and seamount chem­ tion is the main homogenizing agent of man­ ical differences, occur over tens of km e.g. the tle basalts. There are also differences from place Loa and Kea trends i n Hawaii. If hetero­ to place and with depth, i.e. large-scale hetero­ geneities were entirely grain-sized or Ian-sized, geneities. then both OIB and MORB would average out the The isolation time of the upper mantle is heterogeneity in the sampling process. If het­ related to the time between visits of a trench or erogeneities were always thousands of lan in a ridge. With current migration rates a domain extent and separation, then OIB and MORB sam­ of the upper mantle can be isolated for as pling differences could not erase this. Tilere­ long as 1 to 2 Gyr. These are typical mantle fore, there must be an important component of isotopic ages and are usually attributed to a COMPOSITION OF OIB SO URCES- ECLOGITE? 317 convective-overturn time. Either interpretation is of renewed interest. The possibility that the shal­ circumstantial. low mantle is lithologically variable, containing materials with higher latent basaltic melt frac­ tions than lherzolite, means that the mantle Composition of OIB sources ­ can be more-or-less isothermal on a local and eclogite? regional scale, yet at given depth closer to the solidi of some of the lithologies than oth ers. Fer­ Subducted or delaminated basalt converts to tile patches can also account for melting anoma­ eclogite at depths greater than about 50 km. lies along the global ridge system. Thu s, if the are not uniform; they includes extru­ upper man tle is s ufficient ly heteroge­ sives, dikes, sills and diverse cumulates and nous , plumes and high absol ute temper­ may be suitable p rotolith s for composition­ atu res a r e not requ ired as an explanation ally distinct and diverse ocean-island basalts. The for melting anomalies. The viability of the ther­ possible roles of garnet pyroxenite and eclog­ mal plume hypothesis boils down to the viability ite in the mantle sources of flood basalts and of the assumption that the upper mantle is chem­ ocean islands have recently become a matter ically homogenous [mantleplu mes].