Generation Of Mid-ocean Ridge Basalts Research Articles

Two-component mantle melting-mixing model for the generation of mid-ocean ridge basalts: Implications for the volatile content of the Pacific upper mantle

We report major, trace, and volatile element (CO2, H2O, F, Cl, S) contents and Sr, Nd, and Pb isotopes of mid-ocean ridge basalt (MORB) glasses from the Northern East Pacific Rise (NEPR) off-axis seamounts, the Quebrada-Discovery-GoFar (QDG) transform fault system, and the Macquarie Island. The incompatible trace element (ITE) contents of the samples range from highly depleted (DMORB, Th/La⩽0.035) to enriched (EMORB, Th/La⩾0.07), and the isotopic composition spans the entire range observed in EPR MORB. Our data suggest that at the time of melt generation, the source that generated the EMORB was essentially peridotitic, and that the composition of NMORB might not represent melting of a single upper mantle source (DMM), but rather mixing of melts from a two-component mantle (depleted and enriched DMM or D-DMM and E-DMM, respectively). After filtering the volatile element data for secondary processes (degassing, sulfide saturation, assimilation of seawater-derived component, and fractional crystallization), we use the volatiles to ITE ratios of our samples and a two-component mantle melting-mixing model to estimate the volatile content of the D-DMM (CO2=22ppm, H2O=59ppm, F=8ppm, Cl=0.4ppm, and S=100ppm) and the E-DMM (CO2=990ppm, H2O=660ppm, F=31ppm, Cl=22ppm, and S=165ppm). Our two-component mantle melting-mixing model reproduces the kernel density estimates (KDE) of Th/La and 143Nd/144Nd ratios for our samples and for EPR axial MORB compiled from the literature. This model suggests that: (1) 78% of the Pacific upper mantle is highly depleted (D-DMM) while 22% is enriched (E-DMM) in volatile and refractory ITE, (2) the melts produced during variable degrees of melting of the E-DMM controls most of the MORB geochemical variation, and (3) a fraction (∼65% to 80%) of the low degree EMORB melts (produced by ∼1.3% melting) may escape melt aggregation by freezing at the base of the oceanic lithosphere, significantly enriching it in volatile and trace element contents. Our results are consistent with previously proposed geodynamical processes acting at mid-ocean ridges and with the generation of the E-DMM. Our observations indicate that the D-DMM and E-DMM have (1) a relatively constant CO2/Cl ratio of ∼57±8, and (2) volatile and ITE element abundance patterns that can be related by a simple melting event, supporting the hypothesis that the E-DMM is a recycled oceanic lithosphere mantle metasomatized by low degree melts. Our calculation and model give rise to a Pacific upper mantle with volatile content of CO2=235ppm, H2O=191ppm, F=13ppm, Cl=5ppm, and S=114ppm.

Mid-ocean ridge basalt generation along the slow-spreading, South Mid-Atlantic Ridge (5–11°S): Inferences from 238U–230Th–226Ra disequilibria

U-series disequilibria have provided important constraints on the physical processes of partial melting that produce basaltic magma beneath mid-ocean ridges. Here we present the first 238U–230Th–226Ra isotope data for a suite of 83 basalts sampled between 5°S and 11°S along the South Mid-Atlantic Ridge. This section of the ridge can be divided into 5 segments (A0–A4) and the depths to the ridge axis span much of the global range, varying from 1429 to 4514m. Previous work has also demonstrated that strong trace element and radiogenic isotope heterogeneity existed in the source regions of these basalts. Accordingly, this area provides an ideal location in which to investigate the effects of both inferred melt column length and recycled materials. 226Ra–230Th disequilibria indicate that the majority of the basalts are less than a few millennia old such that their 230Th values do not require any age correction. The U–Th isotope data span a significant range from secular equilibrium up to 32% 230Th excess, also similar to the global range, and vary from segment to segment. However, the (230Th/238U) ratios are not negatively correlated with axial depth and the samples with the largest 230Th excesses come from the deepest ridge segment (A1). Two sub-parallel and positively sloped arrays (for segments A0–2 and A3 and A4) between (230Th/238U) and Th/U ratios can be modelled in various ways as mixing between melts from peridotite and recycled mafic lithologies. Despite abundant evidence for source heterogeneity, there is no simple correlation between (230Th/238U) and radiogenic isotope ratios suggesting that at least some of the trace element and radiogenic isotope variability may have been imparted to the source regions >350kyr prior to partial melting to produce the basalts. In our preferred model, the two (230Th/238U) versus Th/U arrays can be explained by mixing of melts from one or more recycled mafic lithologies with melts derived from chemically heterogeneous peridotite source regions.

Noble gas isotopic compositions of mantle xenoliths from northwestern Pacific lithosphere

We measured noble gas isotopic compositions of mantle xenoliths and xenocrystic olivines sampled from seamounts—so-called petit-spot volcanoes—on the 135-million-year-old northwestern Pacific Plate. The xenoliths are spinel lherzolites originating from suboceanic lithospheric mantle. The samples' 3He/ 4He ratios are 7.0–8.5 Ra, where Ra signifies atmospheric 3He/ 4He. The 40Ar/ 36Ar ratios are as high as 7000. These observations suggest that the noble gas isotopic compositions of suboceanic lithospheric mantle resemble those of mid-ocean ridge basalt (MORB). A mantle source with a He/U ratio as high as an assumed value for MORB source is necessary to maintain the MORB-like 3He/ 4He over 135 million years, implying that melt extraction at mid-ocean ridges only slightly alters the He/U ratio of the oceanic upper mantle. The 4He/ 40Ar⁎ ratios of the samples described herein are much lower than the theoretical radiogenic production ratio, where an asterisk denotes correction for atmospheric contribution. The low 4He/ 40Ar⁎ is inferred to result from kinetic fractionation in the mantle. When magma migrates through a mantle source, lighter noble gases in the mantle source diffuse selectively into magma channels. The MORB generation does not cause low 4He/ 40Ar⁎, however. If a mantle source is depleted in lighter noble gases during ancient MORB generation, then noble gas isotopic compositions of the mantle source are affected over time by accumulation of radiogenic nuclides. Thereby, the mantle source adopts a radiogenic or nucleogenic noble gas isotopic composition. Recent kinetic fractionation contributes to the low 4He/ 40Ar⁎ in the samples. Based on the diffusive fractionation model, more than 100 years are necessary to produce a residual mantle source with 4He/ 40Ar⁎, as observed in the samples. However, petit-spot volcanoes, found as small knolls, seem to erupt within a short period. The low 4He/ 40Ar⁎ of the samples implies prior volcanism in this region. No recent volcanism has occurred near this region aside from petit-spot volcanism, indicating that petit-spot volcanoes are polygenetic with a long active period. Petit-spot volcanoes are regarded as common magmatism on the subducting oceanic plate. Consequently, the thermal structure and temperature-dependent physical properties of the oceanic plate are, at least partly, affected by remnant heat of the magmatism.

MORB generation beneath the ultraslow spreading Southwest Indian Ridge (9–25°E): Major element chemistry and the importance of process versus source

We report highly variable mid‐ocean ridge basalt (MORB) major element and water concentrations from a single 1050‐km first‐order spreading segment on the ultraslow spreading Southwest Indian Ridge, consisting of two supersegments with strikingly different spreading geometry and ridge morphology. To the east, the 630 km long orthogonal supersegment (<10° obliquity) dominantly erupts normal MORB with progressive K/Ti enrichment from east to west. To the west is the 400 km long oblique supersegment (up to 56° obliquity) with two robust volcanic centers erupting enriched MORB and three intervening amagmatic accretionary segments erupting both N‐MORB and E‐MORB. The systematic nature of the orthogonal supersegments' ridge morphology and MORB composition ends at 16°E, where ridge physiography, lithologic abundance, crustal structure, and basalt chemistry all change dramatically. We attribute this discontinuity and the contrasting characteristics of the supersegments to localized differences in the upper mantle thermal structure brought on by variable spreading geometry. The influence of these differences on the erupted composition of MORB appears to be more significant at ultraslow spreading rates where the overall degree of melting is lower. In contrast to the moderate and rather constant degrees of partial melting along the orthogonal supersegment, suppression of mantle melting on the oblique supersegment due to thickened lithosphere means that the bulk source is not uniformly sampled, as is the former. On the oblique supersegment, more abundant mafic lithologies melt deeper thereby dominating the more enriched aggregate melt composition. While much of the local major element heterogeneity can be explained by polybaric fractional crystallization with variable H2O contents, elevated K2O and K/Ti cannot. On the basis of the chemical and tectonic relationship of these enriched and depleted basalts, their occurrence requires a multilithology mantle source. The diversity and distribution of MORB compositions, especially here at ultraslow spreading rates, is controlled not only by the heterogeneity of the underlying mantle, but also more directly by the local thermal structure of the lithosphere (i.e., spreading geometry) and its influence on melting processes. Thus at ultraslow spreading rates, process rather than source may be the principle determiner of MORB composition.

Diffusive fractionation of noble gases and helium isotopes during mantle melting

The large differences in He and Ar diffusivities in silicate minerals could result in fractionation of the He/Ar ratio during melting of the mantle, producing He/Ar ratios in the primary mantle melts that are higher than those of the bulk mantle. Modeling noble gas diffusion out of the bulk mantle into fast diffusion pathways (such as fractures or melt channels) suggests that significant (order of magnitude) He/Ar fractionation will occur if the fast diffusion channels are spaced several meters apart and the noble gas residence in these diffusion channels is of the order days to weeks. In addition, the 15% difference in 3He and 4He diffusivities could also produce isotopic fractionation between the melt and its solid source. Modeling the behavior of He and Ar during melting shows that small increases (few %) in 3He/ 4He should be correlated with larger variations (factor of 5) in 4He/ 40Ar. However, in order to test this hypothesis the effects of subsequent He–Ar fractionation that occur during degassing have to be corrected. I describe a scheme that can separate He/Ar variations in the primary melt from overprinted fractionation during magmatic degassing. Using the degassing-corrected data, there is a correlation between the primary melt’s 4He/ 40Ar and 3He/ 4He in mid-ocean ridge basalts (MORBs). The slope of the correlation is consistent with the models of preferential diffusion of 3He relative to 4He and of 4He relative to 40Ar from the solid mantle into the melt. Diffusive fractionation of noble gases during melting of the mantle can also account for low 4He/ 40Ar ratios commonly found in residual mantle xenoliths: preferential diffusion of He relative to Ar will produce some regions of the mantle with low 4He/ 40Ar, the complement of the high 4He/ 40Ar ratios in basalts. Diffusive fractionation cannot, however, account for differences between the He and Ne isotopic compositions of MORBs compared with ocean island basalts (OIBs); not only are the extremely high 3He/ 4He ratios of OIBs (up to 50 Ra) difficult to produce at reasonable mantle time and lengthscales, but also the Ne isotopic compositions of MORBs and OIBs do not lie on a single mass fractionation line, therefore cannot result from diffusive fractionation of a single mantle Ne source. If preferential diffusion of He from the solid mantle into primary melts is a significant process during generation of MORBs, then it is difficult to constrain the He concentration of the mantle: He concentrations in basalts and the He flux to the ocean essentially result from extraction of He from a larger (and unknown) volume of mantle than that that produced the basalts themselves. The He concentration of the mantle cannot be constrained until more accurate estimates of the diffusion contribution are available.

Generation of mid-ocean ridge basalts at pressures from 1 to 7 GPa

We propose a model for the generation of average MORBs based on phase relations in the CaO-MgO-Al 2O 3-SiO 2-CO 2 system at pressures from 3 to 7 GPa and in the CaO-MgO-Al 2O 3-SiO 2-Na 2O-FeO (CMASNF) system at pressures from ∼0.9 to 1.5 GPa. The MELT seismic tomography (Forsyth et al., 2000) across the East Pacific Rise shows the largest amount of melt centered at ∼30-km depth and lesser amounts at greater depths. An average mantle adiabat with a model-system potential temperature (T p) of 1310°C is used that is consistent with this result. In the mantle, additional minor components would lower solidus temperatures ∼50°C, which would lower T p of the adiabat for average MORBs to ∼1260°C. The model involves generation of carbonatitic melts and melts that are transitional between carbonatite and kimberlite at very small melt fractions (<0.2%) in the low-velocity zone at pressures of ∼2.6 to 7 GPa in the CMAS-CO 2 system, roughly the pressure range of the PREM low-velocity zone. These small-volume, low-viscosity melts are mixed with much larger volumes of basaltic melt generated at the plagioclase-spinel lherzolite transition in the pressure range of ∼0.9 to 1.5 GPa. In this model, solidus phase relations in the pressure range of the plagioclase-spinel lherzolite transition strongly, but not totally, control the major-element characteristics of MORBs. Although the plagioclase-spinel lherzolite transition suppresses isentropic decompression melting in the CMAS system, this effect does not occur in the topologically different and petrologically more realistic CMASNF system. On the basis of the absence of plagioclase from most abyssal peridotites, which are the presumed residues of MORB generation, we calculate melt productivity during polybaric fractional melting in the plagioclase-spinel lherzolite transition interval at exhaustion of plagioclase in the residue. In the CMASN system, these calculations indicate that the total melt productivity is ∼24%, which is adequate to produce the oceanic crust. The residual mineral proportions from this calculation closely match those of average abyssal peridotites. Melts generated in the plagioclase-spinel lherzolite transition are compositionally distinct from all MORB glasses, but do not have a significant fractional crystallization trend controlled by olivine alone. They reach the composition field of erupted MORBs mainly by crystallization of both plagioclase and olivine, with initial crystallization of either one of these phases rapidly joined by the other. This is consistent with phenocryst assemblages and experimental studies of the most primitive MORBs, which do not show an olivine-controlled fractionation trend. The model is most robust for the eastern Pacific, where an adiabat with a T p of ∼1260°C is supported by the MELT seismic data and where the global inverse correlation of (FeO) 8 with (Na 2O) 8 is weak. Average MORBs worldwide also are well modeled. A heterogeneous mantle consisting of peridotite of varying degrees of major-element depletion combined with phase-equilibrium controls in the plagioclase-spinel lherzolite transition interval would produce the form of the global correlations at a constant T p, which suggests a modest range of T p along ridges. Phase-composition data for the CMASNF system are presently not adequate for quantitative calculation of (FeO) 8-(Na 2O) 8-(CaO/Al 2O 3) 8 systematics in terms of this model. The near absence of basalts in the central portion of the Gakkel Ridge suggests a lower bound for T p along ridges of ∼1240°C, a potential temperature just low enough to miss the solidus for basalt production at ∼0.9 GPa. An upper bound for T p is poorly constrained, but the complete absence of picritic glasses in Iceland and the global ridge system suggests an upper bound of ∼1400°C. In contrast to some previous models for MORB generation that emphasize large potential temperature variations in a relatively homogeneous peridotitic mantle, our model emphasizes modest potential temperature variations in a peridotitic mantle that shows varying degrees of heterogeneity. Calculations indicate that melt productivity changes from 0 to 24% for a change in T p from 1240 to 1260°C, effectively producing a rapid increase to full crustal thickness or decrease to none as ridges appear and disappear.

Modeling of trace element fractionation during non-modal dynamic melting with linear variations in mineral/melt distribution coefficients

Recently significant advancement has been made in the modeling of trace element fractionation during dynamic melting. Some of the theoretical treatments that have contributed to the advancement took into account the variations in source mineral proportions, but all of them assumed constant mineral/melt distribution coefficients. This study introduces linear variations of mineral/melt distribution coefficients into the dynamic melting model. Analytical solutions are provided for this model with variations not only in source mineral proportions but also in mineral/melt distribution coefficients. Applications of these equations to mantle melting to incorporate variable clinopyroxene/melt distribution coefficients during initial stages of melting can reduce the required involvement of garnet for the generation of mid-ocean ridge basalts. The effects of variable mineral/melt distribution coefficients on trace element fractionation need to be carefully assessed in trace element studies of the petrogenesis of igneous rocks and their melt inclusions.

The generation of mid-ocean ridge basalts from the Hf and Nd isotope perspective

This paper presents new Hf and Nd isotope and trace element data for mid-ocean ridge basalts. The intention of this study was to investigate whether melt parameters derived from the isotopic systematics of MORB ( δ (Lu) (Hf) and δ (Sm) (Nd) ) support the variations in depth and degree of melting as inferred from the major elements and ridge depth; that is, the ‘local’ and ‘global’ trend [1]. Variation on a ‘local’ scale is defined as variation within a ridge segment, while the ‘global’ scale variation is defined as variation between ridge segments. Both on a ‘local’ and a ‘global’ scale, δ (Lu) (Hf) and δ (Sm) (Nd) show consistent variations with the major element melt parameters, Na 8 and Fe 8. In addition, the δ melt parameters vary consistently with ridge depth. Basalts from some of the deepest ridges, Australian-Antarctic Discordance and Mid-Cayman Rise, show evidence that a relatively large proportion of the basalt was generated in the presence of garnet. A melting model is derived using best available estimates for a large number of parameters that influence the chemistry of MORBs. With this melting model, the additional constraints of δ (Lu) (Hf) and δ (Sm) (Nd) further define the melting regime beneath the ocean ridges. The onset of melting for shallow ridges, like Kolbeinsey Ridge, starts deeper than for deep ridges, the aggregated melts from the Kolbeinsey Ridge sample a columnar shaped melting regime. Basalts from deep ridges represent a smaller degree of melting, which was generated in a triangular-shaped melting regime. The translation of δ (Lu) (Hf) and δ (Sm) (Nd) in absolute values for degree and depth of melting for individual ridge segments is hampered by uncertainties in the details of the heterogeneity of the MORB reservior.

The genesis of mid-ocean ridge basalt

The tholeiitic volcanics erupted at mid-ocean ridges (mid-ocean ridge basalts or MORB) constitute the dominant volcanic lithology on Earth. Analyses of tachylites from Atlantic, Pacific and Indian Ocean spreading centres range widely in 100 Mg/(Mg + Fe 2+) ratios (= M) and M varies from 70 to 30. Glasses with M = 55−65 are the most common variants and only a small percentage of glass analyses has M approaching 70. The latter defines the M -value of basaltic melts in equilibrium with residual upper-mantle source peridotites with M ∼ 88. The frequency histogram of the M -values of average compositions of MORB glasses at 88 ocean floor localities is similar in analysis distribution to the frequency histograms depicting variation in the M -values of glasses from the various spreading centres. M -values and nickel contents of MORB and the nature and compositions of the near-liquidus phases crystallized experimentally from MORB melts at elevated pressures have been applied to identify primary (unfractionated) melts erupted in a mid-ocean ridge environment. However, Ni abundances and high-pressure phase relationships are not necessarily unique or definitive parameters of primary melts. The latter are generally linked genetically with Mg-rich lherzolitic source rocks of ‘pyrolite’ type ( M ∼ 90. The spectrum of M -values displayed by MORB glasses, with a definite bias towards relatively Fe-rich compositions (average M of approximately 600 MORB glasses is 58.6), suggests that the melts may have evolved either via ferromagnesian fractionation of relatively Mg-rich parental melts ( M = 70−80), or by partial melting of a heterogeneous upper mantle with variable M values, or as a result of magma mixing of already fractionated melts and primitive magma batches. For a number of reasons fractonation models based on the extraction of olivine or one or more of olivine, plagioclase and clinopyroxene, either from picritic melts ( M > 75 or ‘primitive’ basaltic melts with M ∼ 70, are questionable as prime controls of MORB chemistry. These include: (1) the extreme rarity of ‘quenched’ picritic or Mg-basaltic melts in ocean ridge environments; (2) the lack of adequate evidence of the appropriate (of necessity voluminous) complementary cumulates (dunites, allivalites, troctolites, anorthosites) demanded by olivine, plagioclase, or olivine + plagioclase fractionation models; and (3) the aberrent frequencies of glass M -values whereby the assumed derivatives ( M = 55−65 are much more abundant (and presumably much more voluminous) than the alleged parents or transitional derivatives (65 < M < 75). The nature of the trends of Na 2O, CaO and Al 2O 3 in Galapagos Spreading Centre tachylites of extended composition ( M = 65−30) indicates the ‘gabbroic’ fractionation is also unlikely to exert important controls on MORB chemistry. As their M -values increase, mid-ocean ridge basalts increase in Al, Ca, Ni, Co, Cr and decrease in Ti, Mn, Na, K and P. Except for Al and Ca, these trends are similar to those displayed by upper-mantle peridotites increasing in M, i.e., becoming more refractory following one or more partial melting episodes. It is suggested that at least a majority of mid-ocean ridge basalts is intrinsically primary and generated by variable degrees of partial melting of heterogeneous lherzolitic upper mantle (80 < M < 90) with variable abundances of elements such as Ti, Al, Ca and Na and also depleted in large ion lithophile (LIL) elements. Negative europium anomalies in the rare-earth patterns of some oceanridge basalts (ferrobasalts with low M) are ascribed mainly to the persistence of residual plagioclase in relatively Fe-rich plagioclase lherzolite source rocks, following low degrees of partial melting. The partial melting events leading to the generation of mid-ocean ridge basalts took place over a relatively modes pressure range (approximately 8–15 kb) which encompassed the transition of plagioclase lherzolite to spinel lherzolite. This proposal appears consistent with the nature and occurrence of megacrysts (xenocrysts) of tschermakitic Cr-diopside (Ca 43Mg 52Fe 5), olivine ( mg 89–91), plagioclase (An 92-85) and spinel (Fe 2Al 60Cr 38) in some MORB. The megacryst compositions suggest that these phases represent disaggregated plagioclase peridotite or spinel lherzolite acquired by melts during their passage through the oceanic upper mantle.

Absolute and relative abundances of K, Rb, Sr and Ba in circum-Pacific island-arc magmas, with special reference to the Marianas

Absolute and relative abundances of K, Rb, Sr and Ba in 22 specimens representing lavas from six of the ten active and dormant subaerial volcanoes from the northern Mariana island arc are presented. Average concentration of K, Rb, Sr and Ba is about 7000, 15, 300 and 219 ppm, respectively. K, Rb and Ba covary strongly with silica while Sr is relatively constant. Ratios of K/Rb and K/Ba average about 500 and 31, respectively. The Mariana lavas are enriched in K, Rb, Sr and Ba, and their elemental ratios are markedly different from those of the “island-arc tholeiites” and are similar to values typical of the calc-alkaline rock series. Ranges of elemental ratios among different island arcs, and among islands of the Mariana arc, are often distinctive. In many arc lavas, K/Ba is more distinctive than K/Rb in distinguishing these rocks from those of mid-ocean ridge basalts. Consideration of incompatible element ratios shows that magmas of the Mariana arc occupy a range intermediate between the two contrasting oceanic magma-types of mid-ocean ridge basalt and alkali basalt. These data, together with isotopic constraints, support the conclusion that the lavas of the Mariana arc are derived by fractional fusion of a mantle source intermediate in incompatible element and isotopic characteristics between that responsible for the generation of mid-ocean ridge basalts and oceanic alkalic rocks.

Chemical heterogeneity of the Archaean mantle, composition of the earth and mantle evolution

New rare earth element (REE) data for Archaean basalts and spinifex-textured peridotites (STP) show a range of La/Sm ratios (chondrite-normalized) from 0.36 to 3.5, with the bulk of the data in the range 0.7–1.3. This supports the hypothesis, based on Sr isotope initial ratios, that the Archaean mantle was chemically heterogeneous. We suggest that the bulk mantle source for Archaean basaltic magmas was close to an undepleted earth material. An average chemical composition of the Archaean mantle is estimated using chemical regularities observed in Archaean STP and high-magnesian basalts. TiO 2 and MgO data show an inverse correlation which intersects the MgO axis at about 50% MgO (Fo 92). TiO 2 abundance in the mantle source is measured on this plot by assigning an MgO= 38% for the mantle. Concentrations of other elements are also estimated and these data are then used to obtain a composition for the bulk earth. We suggest an earth model with about 1.35 times ordinary chondrite abundances of refractory lithophile elements and about 0.2 times carbonaceous type 1 chondrite abundances of moderately volatile elements (such as Na, Rb, K, Mn). P shows severe depletion in the model earth relative to carbonaceous chondrites, a feature either due to volatilization or core formation (preferred). Our data support the hypothesis of Ringwood that the source material for the earth is a carbonaceous chondrite-like material. The generation of mid-ocean ridge basalts (MORB) is examined in the light of the model earth composition and Al 2O 3/TiO 2, CaO/TiO 2 ratios. It is suggested that for primitive basalts, these values can be used to predict the residual phases in their source. Comparison of chemical characteristics of inferred sources for 2.7-b.y. Archaean basalts and modern “normal” MORB indicates that the MORB source is severely depleted in highly incompatible elements such as Cs, Ba, Rb, U, Th, K, La and Nb, but has comparable abundances of less incompatible elements such as Ti, Zr, Y, Yb. The cause of the depletion in the MORB source is examined in terms of crust formation and extraction of silica-undersaturated melts. The latter seems to be a more likely explanation, since the degree of enrichment of highly incompatible elements in the crust only accounts for up to 40% of their abundances in the bulk earth and cannot match the depletion pattern in normal MORB. A large volume of material, less depleted than the source for normal MORB must therefore exist in the mantle and can serve as the source for the ocean island basalts and “normal” MORB. Three different mantle evolution models are examined and each suggests that the mantle is stratified with respect to abundances of incompatible trace elements. We suggest that no satisfactory model is available to fully explain the spectrum of geochemical and geophysical data. In particular the Pb and Sr isotope data on oceanic basalts, the depletion patterns of MORB and the balance between lithophile abundances in the crust and mantle, are important geochemical constraints to mantle models. Further modelling of the mantle evolution will be dependent on firmer information on the role of subduction, mantle convection pattern, and basalt production through geologic time together with a better understanding of the nature of Archaean crustal genesis.