Traceelement abundances in midocean ridge basalts and residual peridotites

To understand and model the compositions of melt sources and the partial-melting-fractionation processes, it is essential that the abundances of trace elements with a large range of mineral-melt partition coefficients (Table 24.2) are considered for all matter associated with the process: fertile peridotites, basalts, basalt glasses, melt inclusions in minerals and finally the residual rocks.

First, a representative MORB sample set from the southern mid-Atlantic ridge shows that the normalized concentrations of the most incompatible elements, e.g. Rb, vary by almost two orders of magnitude, from values below the bulk silicate

Table 24.2 Partition coefficients'a for minerals in equilibrium with mafic melts. From Salters and Stracke (2004). Copyright (2004) American Geophysical Union. Modified by permission of the American Geophysical Union

3 GPa 2 GPa

Table 24.2 Partition coefficients'a for minerals in equilibrium with mafic melts. From Salters and Stracke (2004). Copyright (2004) American Geophysical Union. Modified by permission of the American Geophysical Union

3 GPa 2 GPa

Olivine

Opx

Cpx

Garnet

Olivine

Opx

Cpx

Cs

0

0

0

0

0

0

0

Rb

0.0003

0.0002

0.0004

0.0002

0.0003

0.0002

0.0004

Ba

0.000 005

0.000 006

0.0004

0.00007

0.000005

0.000 006

0.0004

Th

0.000 05

0.002

0.00566

0.009

0.00005

0.002

0.0059

Pb

0.003

0.009

0.009

0.005

0.003

0.009

0.012

U

0.000 38

0.002

0.0113

0.028

0.00038

0.002

0.0094

K

0.00002

0.0001

0.001

0.013

0.00002

0.0001

0.001

Ta

0.0005

0.004

0.01

0.015

0.0005

0.004

0.015

Nb

0.0005

0.004

0.01

0.015

0.0005

0.004

0.015

La

0.0005

0.004

0.015

0.0007

0.0005

0.0031

0.03

Ce

0.0005

0.004

0.038

0.017

0.0005

0.004

0.08

Sr

0.000 04

0.0007

0.091

0.0007

0.00004

0.0007

0.091

Nd

0.00042

0.012

0.0884

0.064

0.00042

0.012

0.088

Hf

0.0011

0.024

0.14

0.4

0.0022

0.03

0.2835

Sm

0.0011

0.02

0.1509

0.23

0.0011

0.02

0.299

Ti

0.015

0.086

0.14

0.6

0.015

0.086

0.35

Gd

0.0011

0.065

0.16

1.2

0.0011

0.0065

0.35

Dy

0.0027

0.065

0.17

2

0.0027

0.011

0.4

Er

0.013

0.065

0.18

3

0.013

0.045

0.42

Yb

0.02

0.08

0.25

5.5

0.02

0.08

0.45

Lu

0.02

0.12

0.276

7

0.02

0.12

0.511

a Values for spinel are not listed as they are insignificantly small. Generally the partition coefficients depend on several parameters of the melting processes, such as pressure, temperature, chemical compositions, oxygen fugacity and fluid abundance, and the values given in the table could vary somewhat. For example, Su (2002) presented an opx-melt partition coefficient of 0.007 for Nd that was the same as that used byMcKenzie andO'Nions (1991) in their inversion model, but a factor 1.7 below the value presented in this table. After Salters and Stracke (2004).

Earth (BSE) to 30 times above it (Fig. 24.2). Also, the mean MORB concentrations of these elements are a factor 30 above those obtained for the MORB-source mantle, thus requiring a contribution of highly enriched melts in the bulk MORB magmas. According to Eqn (12.4) such melts must be segregated in (and then removed from) the melting environment at low melt/solid ratios. The REE elements show a smooth pattern and, in accordance with their range in incompatibility, the light rare Earth elements (LREEs) are more variable than the less incompatible heavy rare Earths (HREEs).

Fig. 24.2 Trace elements in mid-ocean ridge basalt (MORB) samples. The average MORB abundances (using all available MORB data), the model abundances in MORB-source mantle and the average abundances in oceanic island basalts (OIBs) are shown for comparison. The bulk silicate Earth (BSE) composition is used for normalization (Tables 17.1 and 24.1). From Hofmann (2003), © Elsevier Science 2003, reproduced by permission.

Fig. 24.2 Trace elements in mid-ocean ridge basalt (MORB) samples. The average MORB abundances (using all available MORB data), the model abundances in MORB-source mantle and the average abundances in oceanic island basalts (OIBs) are shown for comparison. The bulk silicate Earth (BSE) composition is used for normalization (Tables 17.1 and 24.1). From Hofmann (2003), © Elsevier Science 2003, reproduced by permission.

This variability in incompatible elements, including the LREEs, has led to a traditional subdivision of MORBs into three groups, the predominant "normal" (or depleted) N-MORBs, the less common "enriched" E-MORBs, and the "transitional" T-MORBs (Fig. 24.3). The latter have a quite similar composition to that of "all-MORBs", i.e. an average MORB (Table 24.1). From these data alone, however, it is difficult to identify precisely the causes of the variability, which could be mantle-source heterogeneity or differences in the partial melting and fractionation processes.

Second, to shed more light on these processes, trace-element data on melt inclusions have been shown to be of extreme importance. A small amount of melt can be trapped before extensive mixing occurs and can thus preserve compositions resulting from early stages of partial melting (Sobolev, 1996; Shimizu, 1998; Slater et al., 2001). The most surprising feature of melt inclusions is that their compositions are so variable. For example, several olivine-hosted melt inclusions in a single olivine grain from the Atlantic MORBs show LREE patterns that are more depleted than the whole-rock value and thus reveal the initial heterogeneity of the melts (Fig. 24.4).

Trace-element abundances in mantle peridotites are the third important source of relevant information. These rocks, termed abyssal peridotites, are obtained from mid-ocean ridges by dredging and drilling. In general, the sampling of such rocks

Fig. 24.3 Average trace-element abundances in mid-ocean ridge basalts (MORBs). The bulk silicate Earth (BSE) composition is taken from Table 17.1. From Klein (2003), © Elsevier Science 2003, reproduced by permission.

Fig. 24.4 Rare earth elements in olivine-hosted melt inclusions from the mid-Atlantic ridge. One melt inclusion shows an extreme depletion, in contrast with the host matrix (light shading), while two others are less depleted. This example illustrates the progressive mixing of small melt fractions with the host melt, which is itself a mixture of previous such contributions. The C1 composition is taken from Table 11.2. After Sobolev and Shimizu (1993) and Sobolev (1996).

Fig. 24.4 Rare earth elements in olivine-hosted melt inclusions from the mid-Atlantic ridge. One melt inclusion shows an extreme depletion, in contrast with the host matrix (light shading), while two others are less depleted. This example illustrates the progressive mixing of small melt fractions with the host melt, which is itself a mixture of previous such contributions. The C1 composition is taken from Table 11.2. After Sobolev and Shimizu (1993) and Sobolev (1996).

from modern ridges is rare and limited to the uppermost mantle. However, ophiolites expose large mantle sections as well as the mantle-crust interface. One of the best-preserved sections of oceanic lithosphere is the Semail ophiolite complex in Oman. Like most abyssal peridotites, the rocks in this complex are strongly depleted in LREEs and other highly incompatible elements (Fig. 24.5). The La concentrations are typically within 0.005-0.05 of the chondritic values, and the LREE/HREE (normalized) ratios are low; (La/Yb)N — 1/30. The REE pattern of these rocks is smooth, indicating fractionation of all the elements: the lighter the element, the more severe its depletion. The depletion of the most incompatible elements is truly impressive, e.g. [Th] ~ 1 ppb or — 0.013 times the BSE value. Uranium is slightly less depleted, so that U/Th — 1 (Godard et al., 2000).

A comparison of the REE patterns from the Semail harzburgites and dunites (Fig. 24.5) with those of the depleted MORB-source mantle (DMM) model (Fig. 24.2) reveals differences that are fundamental to understanding the cause of DMM depletion. Not only are the Semail patterns much more strongly depleted, but they also have a different shape: a convex-up shape for LREEs and an almost flat pattern with modest depletion for HREEs are typical of the DMM, while the Semail rock patterns are concave-up, with (Gd/Yb)N down to 0.1 and (La/Gd)N mostly around 0.4. The difference is readily explained: the Semail harzburgites and dunites represent the residue after MORB extraction, whereas the model MORB mantle patterns are those of the source rock before the melt was extracted from it.

This observation allows us to conclude that the incompatible-element depletion in the normal MORB-source mantle is not the result of MORB melt extraction (if it were then the mantle pattern would resemble that of the Semail rocks in shape). This conclusion is of course entirely logical: MORB is recycled into the mantle at subduction zones (Chapter 25), whereas the incompatible-element depletion in the mantle results largely from the long-term extraction of matter that formed the continental crust (Chapter 26). It should also be noted here that the depletion pattern is associated with isotope fingerprints. For instance, the high 147Sm/144Nd and 176Lu/177Hf ratios of the depleted MORB-source mantle have engendered high 143Nd/144Nd and 176Hf/177Hf ratios, which show that the depletion of the mantle has occurred on a time scale of billions of years, in sharp contrast with the short lifetime of MORBs but in agreement with the great average age of the continental crust. This will be discussed further in Chapters 27 and 28.

While the depletion of source rocks is a principal outcome of partial melting, melts ascending through depleted residues can also cause enrichment, or refertil-ization, by solidifying there and thus producing the quasi-"primitive" REE patterns occasionally observed in peridotites. The isotopic data, however, may reveal a far-from-primitive evolution of such samples (Section 27.7).

The Evolution of Matter La Ce Pr Nd Sm EuGd Tb Dy Ho Er Tm Yb Lu

La Ce Pr Nd Sm Eu Gd Tb Dy Ho ErTm Yb Lu

Fig. 24.5 Rare earth elements in ophiolitic peridotites from Oman. Note the strong depletion of LREEs, both absolute and relative to the HREEs, and the aberrant pattern compared with the depleted MORB-source mantle (DMM) model; see Fig. 24.2. The C1 composition is taken from Table 11.2. After Godard et al. (2000), © Elsevier Science 2000, reproduced by permission.

La Ce Pr Nd Sm Eu Gd Tb Dy Ho ErTm Yb Lu

Fig. 24.5 Rare earth elements in ophiolitic peridotites from Oman. Note the strong depletion of LREEs, both absolute and relative to the HREEs, and the aberrant pattern compared with the depleted MORB-source mantle (DMM) model; see Fig. 24.2. The C1 composition is taken from Table 11.2. After Godard et al. (2000), © Elsevier Science 2000, reproduced by permission.

24.4 Mid-ocean ridge magmatism: evidence from radioactive trace elements

Constraints on melt generation and transport from radioactive trace elements: introduction to disequilibrium

While long-lived radioactive isotope systematics are useful for dating and for tracing the long-term histories of mantle sources, processes such as melting and the ascent of magmas occur on time scales comparable with the half-lives of some naturally occurring short-lived nuclides: 234U (half-life x234 = 2.45 x 105 yr), 230Th (t230 = 7.74 x 104 yr), 231Pa (t231 = 3.25 x 104 yr), 226Ra (t226 = 1.6 x 103 yr) and 210Pb (t210 = 22.6 yr).

Short-lived isotope-system studies on mantle melts have contributed greatly to our understanding of melting processes. Figure 24.6 shows the decay chains for three long-lived radioactive isotopes, 238U, 235U and 232Th, ultimately decaying into 206Pb, 207Pb, and 208Pb respectively. In particular the intermediate daughters 234U, 230Th and 226Ra in the chain of 238U, but also others, are used as radioactive clocks embedded in the natural fluxes of melts (McKenzie, 2000), waters (e.g. Chabaux et al., 2003; Tricca et al., 2001) and gases (e.g. Lehmann et al., 1999).

A comparison of the half-lives shows that the three parent isotopes of the chains decay orders of magnitude more slowly than all the daughters. Thus on a time scale t ~ 5 Myr the abundances of the chain parents may be considered as constants. Assuming a closed-system evolution of the 238U chain (no losses and gains during t ~ 5 Myr) and applying the equations of radioactive decay (see Eqns 1.1, 1.2 and the related text), then, as before letting the nuclide symbol represent the atomic concentration, A238U/At — 0 and the number of 238U decays per unit time is constant:

Here RE denotes the corresponding radioactive equilibrium constant. Under the above assumptions the evolution of each radioactive isotope within the chain, e.g. 234Th, is given by d234Th/dt = -¿234234Th + ^238238U = -k234234Th + RE, (24.4)

for which the solution is

or, for the concentration 'C(t) of any isotope i within the chain, iC(t) = (RE/k)[1 - exp(-kit)]. (24.6)

In view of the above time scale, t ~ 5 Myr » 1/k,, then exp(-kit) — 0 for all intermediate daughter isotopes including 234U, the longest lived one, and 'C(t) — constant. This state is termed radioactive or secular equilibrium: each nuclide within the chain decays at exactly its rate of production, and its abundance is therefore constant. Correspondingly, the radioactive equilibrium concentration ratio (ECR) of any two nuclides, parent j and daughter i, is also constant:

82 Pb |_Rbj stable

I Pbl

138 d

I Pbl

210 Po

5.0 d

210 Bi

1J7 min

1J7 min

^ a-decay y j3-decay

238 I

124 126

^ a-decay y j3-decay

22.6 yr 27 min

128 130 132 134 136 138 140 142 144 146 Number of neutrons

86 Rn At 84 Po Bi

82 Pb Tl

7.04 x 108yr

Decay of U

22.6 yr 27 min

128 130 132 134 136 138 140 142 144 146 Number of neutrons

7.04 x 108yr

Decay of U

124 126 128

130 132 134 136 138 Number of neutrons

140 142 144

_PbJ

stabl^!071

4.8 min

36.1 min

124 126 128

130 132 134 136 138 Number of neutrons

140 142 144

s n to rot

90 Th Ac 88 Ra Fr

86 Rn At 84 Po Bi

82 Pb Tl

Decay of 232 Th

Decay of 232 Th

124 126 128 130 132 134 136 138 140 142 144 Number of neutrons

Fig. 24.6 (a), (b), (c) Radioactive decay chains. Note the highly variable half-lives of isotopes in the chains, allowing the measurement of different time scales. As 238U, 235U, 232Th and the respective short-life isotopes of the chains decay into stable 206,207,208pb, they emit eight, seven and six a-particles (He-atoms) respectively (Eqn 10.7). After McKenzie (2000), © Elsevier Science 2000, reproduced by permission.

_Pbi stable*208 I 34% Ti

3.1 min

^28 1 Ac

5.76 yr

5.76 yr

124 126 128 130 132 134 136 138 140 142 144 Number of neutrons

138 d

Fig. 24.6 (a), (b), (c) Radioactive decay chains. Note the highly variable half-lives of isotopes in the chains, allowing the measurement of different time scales. As 238U, 235U, 232Th and the respective short-life isotopes of the chains decay into stable 206,207,208pb, they emit eight, seven and six a-particles (He-atoms) respectively (Eqn 10.7). After McKenzie (2000), © Elsevier Science 2000, reproduced by permission.

In a state of secular radioactive equilibrium, as seen in Eqn (24.7a) the number of decays per time unit is equal for all nuclides in the chain. The ratio (XilC)/(XjJC), the activity ratio ECR, is often indicated in the literature simply by the use of parentheses, e.g.

In the case of partial melting the equilibrium of the melt and residue will be disturbed if the parent and daughter of a decay pair are different elements and thus have different mineral-melt partition coefficients. When the melt (fraction F) and solid (fraction 1 — F) are in equilibrium in the source, combining the mass balance in Eqn (12.1) with Eqn (24.7a) gives for the parent and daughter isotopes in the melt (liquid silicate, LIS) and the solid (SOL)

Xj [F clis + (1 — F) Csol] = A, [F Clis + (1 — F) Csol], (24.8)

because in radioactive and chemical equilibrium the total activities in melt + solid must be the same. Taking into account that CSOL/CLIS = D, the partition coefficient (Section 12.2), Eqn (24.8) then gives the activity ratio in the melt:

If the magma separates from its source region and ascends, erupts and solidifies on a short time scale compared with the half-lives of daughter nuclides, and is rapidly sampled and analysed, the melt fraction can be derived using Eqn (24.9), provided that both partition coefficients are known (Table 24.2).

After a fractionation event such as partial melting, portions of the chain return to equilibrium on time scales ~ 1/Xi for the various daughter isotopes i in the volcanic rock that has been produced. If equilibrium has not been achieved for any pair j, i measured in a volcanic rock, the time scale of magma fractionation or the parameters governing melting, melt segregation and uplift can be estimated from the activity ratios. The 238U chain has a number of parent-daughter pairs with highly differing daughter half-lives that are useful for this (Fig. 24.6); such a pair can also be constituted by two isotopes that are separated by a chain of very shortlived ones, e.g. 210Pb and 226Ra constitute a pair, applicable on time scales up to -100 yr. The others are 226Ra, 230Th (up to - 6000 yr) and 230Th, 234U (up to -300 000 yr). The longest lived pair, 234U, 238U, cannot be used for magmatic systems as it is not fractionated (in surface processes and sedimentology, however, there is fractionation). The 235U chain has one useful pair, 231Pa, 235U (up to -120 000 yr). In the 232Th chain, the pair 228Ra, 232Th is useful for extremely rapid processes up to - 30 yr.

The degree of initial disequilibrium of a magmatic rock depends on several factors, such as the source composition, the mechanisms and rates of melt segregation and ascent, the degree of fractionation in subsurface magma chambers and the contamination of the magmas. A detailed discussion of these processes and the corresponding models can be found in McKenzie (1985, 2000); Spiegelman and Elliott (1993); Iwamori (1994); Bourdon et al. (1996a); Sims et al. (2002); Jull et al. (2002); Rubin et al. (2005).

Constraints on melt generation and transport from the activity ratios in MORBs

Rubin et al. (2005) found (210Pb/226Ra) activity ratios down to 0.85 in unaltered N-MORB samples collected from the Juan de Fuca and East Pacific Rises (Fig. 24.7(a), (b)). As Ra is much more incompatible than Pb in mantle melting, a lower initial (210Pb/226Ra) value (equal to 0.2-0.4, depending on the melting rate) is expected for the initial melt. Therefore the time scale for magma ascent indicated by samples with (210Pb/226Ra) ~ 0.85 is just over two half-lives of 210Pb, ~ 50 years. Assuming some average depth of partial melting beneath ridges, at ~ 50 km (McKenzie and O'Nions, 1991), the rate of melt ascent appears to be as high as ~ 1000 m yr-1.

In the light of this extremely short time scale of melt transport, any disequilibria found in MORBs for pairs with longer-lived daughter isotopes in the chain, e.g. 226Ra, 230Th and 230Th, 234U, have no age significance but reflect the conditions of melting and fractional crystallization. Thus, Th is more incompatible than U for melting at great depths (P > ~ 2 GPa, or depths > 60 km), in the garnet peridotite stability field, whereas at shallower depths the U and Th partition coefficients are similar. Indeed, a (230Th/234U) value up to 1.4 was found in MORBs: a high-230Th signal indicates first that fractionation happened at a great depth and second that the melt ascended without significant equilibration with the wall rocks (Bourdon etal., 1996b).

This interpretation allows the melt fraction in the deepest part of the melt column to be estimated. Assuming that D(Th) is ~ 3 times lower than D(U) (which is - 0.001; Iwamori, 1994) and a typical activity ratio (230Th/238U) = 1.15, then Eqn (24.9) gives F = 0.004, indicating a very low degree of melting.

0 0

Post a comment