Fractionation of U and Th during mantle melting: a reprise

CHEMICAL GEOLOGY

ql

ELSEVIER

I]VCLUDING

ISOTOPE OEOSCIENCE

Chemical Geology 139 (1997) 165-183

Fractionation of U and Th during mantle melting: a reprise T i m Elliott 1 Faculteit der Aardwetenschappen, Vrije Universiteit, de Boelelaan 1085, 1081 HV Amsterdam, The Netherlands

Received 31 July 1996; accepted 20 December 1996

Abstract 238U-23°Th disequilibrium in basalts has frequently been interpreted as an index of elemental U-Th fractionation during melting. However, time-dependent melting models can account for 238U-23°Th disequilibrium as a result of 23°Th ingrowth, that is not necessarily related to net elemental fractionations. Such 'ingrowth' models have been key in explaining 238U-23°Th disequilibrium in mid-ocean ridge basalts (MORB), but their importance in accounting for 23°Th-excesses in ocean island basalts (OIB) is less certain. A new compilation of OIB data shows that 238u-E3°Th disequilibrium does not show consistent covariation with commonly used indices of degree of melting. Thus, 23°Th ingrowth must also be important in the melting regimes beneath, at least some, ocean islands. The relative contribution of ingrown 23°Th to total 238u-E3°Zh disequilibrium should depend, most critically, on rates of melting relative to overall degree of melting. Fittingly, in the buoyant Hawaiian plume, there appears to be little time for 23°Th ingrowth, and net U-Th fractionation appears to be most important in generating the largest degrees of 23SU-23°Th disequilibrium. Conversely, in the more sluggish Icelandic plume, 23°Th ingrowth appears to be significant and results in striking disequilibrium even in the largest degree melts. It is apparent that each OIB dataset must be carefully assessed in turn as to whether 238U-23°Th disequilibrium more dominantly reflects degree of melting or rates of melting/mantle upwelling. Certainly, it cannot be taken for granted that OIB (23°Th/232Th) ratios are robust measures of the U/Th ratios of plume sources. © 1997 Elsevier Science B.V. Keywords: uranium disequilibrium; partial melting; ocean island basalts; mathematical models

1. Introduction

Mid-ocean ridge basalts (MORB) and ocean island basalts (OIB) commonly display significant 238U-23°Th disequilibrium, which is generally thought to be caused by melting processes (e.g., Condomines et al., 1988b; Gill and Condomines, 1992; MacDougall, 1995). However, a more detailed interpretation of the unique information provided by U-series data in these mantle-derived lavas is controversial. Two end-member models exist in which t E-mail: [email protected]; FAX: 31-20-646-2457.

U-series disequilibrium has been respectively attributed to net elemental fractionation of parentdaughter pairs in the erupted melt relative to the source (Oversby and Gust, 1968; All~gre and Condomines, 1982; Condomines et al., 1988b) and 'ingrowth' of short-lived daughter nuclides during a finite melting period (McKenzie, 1985a; Williams and Gill, 1989; Spiegelman and Elliott, 1993). 230 226 231 The large Th, Ra and Pa excesses present in many MORB (Condomines et al., 1981a; Newman et al., 1983; Rubin and MacDougall, 1988, 1990; Reinitz and Turekian, 1989; Goldstein et al., 1989, 1991, 1993; Ben Othman and Allbgre, 1990; Volpe

0009-2541/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S0009-2541 (97)00034-X

166

7". Elliott / Chemical Geology 139 (1997) 165-183

and Goldstein, 1993; Lundslxom et al., 1995; Bourdon et al., 1996a,b) are very difficult to explain without invoking daughter nuclide ingrowth (McKenzie, 1985a), but the importance of ingrowth during smaller degrees of melting from actively upwelling plumes is less clear. This has led to diverging interpretations of origin of disequilibria in OIB, even, for example, in near-identical Hawaiian basalts (Cohen and O'Nions, 1993; H6mond et al., 1994b; Sims et al., 1995). A compilation of U-series analyses, together with other geochemical data from a range of OIB, are used in this contribution to attempt to resolve this issue. Beforehand, aspects of U-Th fractionation are reviewed and some of the apparently paradoxical features of the ingrowth models are briefly discussed, and hopefully clarified.

2. U-series decay The principles of U-series decay are fully discussed elsewhere (e.g., Ivanovich and Harmon, 1992), but a brief summary follows. 23SU decays to 2°6pb via a chain of short-lived nuclides including 230Th and 226Ra, whilst 235U decays to 2°Tpb via nuclides including 231Pa. Such nuclide chains, headed by parents which have much longer half-lives than their unstable progeny, evolve to a condition termed secular equilibrium, in which the activity (rate of atomic disintegrations) of each radioactive nuclide in the chain is equal. Thus, in secular equilibrium, the ratio of the activities of any two nuclides in the decay chain is unity, e.g., (23SU/23°Th)= 1, where the parentheses indicate activities. Various processes may induce fractionations of elements in a decay chain and so produce disequilibrium. After a disequilibrium-producing event, individual, intermediate parent-daughter pairs in the chain return to equilibrium over time-scales determined by the half-life of the daughter nuclide. As a reasonable rule of thumb for present analytical capabilities, disequilibrium between parent-daughter pairs can be detected for some five half-lives (of the daughter), after initial perturbation. Apart from 234U, which is not fractionated from 23SU by high-temperature magmatic processes, 230Th is the longest lived of the 238U-series nuclides with a half-life (tl/2) of 75.38 ka, (Meadows et al., 1980). Thus after any

magmatic disturbance, the whole 238U-decay chain will have returned to equilibrium within some 350 ka. The systematics of chain decay result in an important property that makes radioactive disequilibria very useful probes of mantle processes. Most mantle sources are unlikely to have been disturbed for many millions of years prior to melting and can therefore be assumed to be in secular equilibrium. Since, at secular equilibrium, atomic ratios of intermediate nuclides are simply the ratio of their half-lives, the abundance ratios of short-lived nuclides in a source prior to melting are thus constrained. This is a unique feature of short-lived isotope tracers. Any disequilibrium measured in melts can therefore be attributed to the melting process alone. This is in contrast to interpreting stable incompatible element tracers in melts, where the combined effects of melting and the unknown original incompatible element composition of the source must be untangled. Clearly, the utility of U-series nuclide measurements is severely diminished if the mantle source is not in secular equilibrium prior to the major melting event. However, there are few reasonable physical processes that might be invoked to explain why secular equilibrium in MORB and OIB sources might have been recently disturbed. The effect on incompatible elements of small degrees of melting at a 'wet' solidus just prior to major mantle melting at a shallower dry solidus has been discussed by several authors (Schilling et al., 1980; McKenzie, 1985b; Galer and O'Nions, 1986; Plank and Langmuir, 1992). Such a process might be argued to be capable of inducing disequilibrium in an asthenospheric source, but since melting at both wet and dry solidii are part of the same upwelling event, this becomes somewhat semantic. Moreover, if prior wet melting were argued to induce ubiquitous disequilibrium in asthenospheric sources and their melts, it becomes difficult to explain why some samples, including Hawaiian tholeiites (Reinitz and Turekian, 1991; Cohen and O'Nions, 1993; H6mond et al., 1994b; Sims et al., 1995; Cohen et al., 1996) and MORB from deep oceanic ridges (Bourdon et al., 1996b) lie very close to 23°Zh-E38u equilibrium. Furthermore, consideration of the specific geodynamic settings of these samples (Cohen and O'Nions, 1993; H6mond et al., 1994b; Bourdon et al., 1996b) can amply

T. Elliott/ Chemical Geology 139 (1997) 165-183

account for a lack of disequilibrium caused by the melting process. These observations, therefore, lend support to the common assumption of secular equilibrium in the mantle prior to the onset of melting. The 238U-23°Th nuclide pair is the most commonly used in the ,;tudy of magmatic petrogenesis. 238U-23°Th data are conventionally displayed on an equiline plot (Allbgre, 1968), Fig. la. The equiline is the locus of compositions that are in secular equilibrium. The equiline plot is simply an isochron diagram. The parent (238U)is plotted on the x-axis, and the daughter (23°Th) on the y-axis, and both axes are normalised by a stable isotope of the same element

1.5

230a
decay

.~""

.."

/

230Thin-growth ..-'"

:)..."i' V / ,./Elemental..--

~,"" ~""

'"~,"'" /~

fractionalion

" I~IU
:/,,:...,.":...,.,.,." 238u_o,,c.sos -

0.5

-

I

0.5

1.0

1.5 (238U/232Th)

1.6 jj:o°*"**

i

1.41.2-

1.0-

0.8-

0.6

~//

0.6

018

(b) ll.0

as the daughter isotope, (232Th). Strictly speaking,

232Th is, of course, not stable, but for the short time-scales involved in 23SU-23°Th disequilibrium, its activity is effectively constant. Samples out of equilibrium lie to the left (23°Th-excesses) or the right (238U-excesses) of the equiline. Instantaneous fractionation of Th from U leaves the (23°Th/232Th) activity ratio unaffected and produces a horizontal displacement from the equiline (Fig. la). Decay or ingrowth of 23°Th returns samples that are out of equilibrium towards the equiline along a vertical vector (Fig. la). Vertical or horizontal vectors cross contours of constant disequilibrium, that radiate as

/'

Excess230Th ,,," 1.0 -

167

112

114 (238U/232Th)

1.6

Fig. 1. (a) 23Su-23°Zh systematics illustrated on an equiline diagram (All~gre, 1968). Two hypothetical instantaneous melt fractionation events (horizontal solid lines with arrows) are shown together with the time-controlled return of the melts to equilibrium (vertical dashed lines with arrows). Also plotted are the loci of secular equilibrium, the equiline (solid diagonal line), and reference degrees of disequilibrium, at (23°Th/23su)= 1.4 and 1.2, i.e., 40% and 20% 23°Th-excesses and (238U/23°Th)= 1.4 and 1.2, i.e., 40% and 20% 23SU-excesses (dashed diagonal lines). (b) Equiline plot showing 'zero-age' MORB (circles) and OIB (squares) 23Su-23°Th disequilibrium data. Dashed line shows 20% 23°Th-excess. This is not an extensive compilation of all 23Su-23°Th disequilibrium data for MORB and OIB, but serves as a companion plot to Fig, s 7, which shows only samples for which both 87Sr/S6Sr and 2 U 230Th disequilibrium have been analysed. Including additional available 238U-23°Th data would little affect the general form of the plot. No data have been plotted where authors have indicated that secondary processes may have been important. MORB data of Rubin and MacDougall (1988, 1992) are not plotted as some samples have very high (23°Th//232Th) and (238U/232Th) relative to data for similar MORB from a nearby locality (Ben Othman and Allbgre, 1990). It is noticeable that the analyses of MacDonald seamount by Rubin and MacDougall (1989) also have values of (23°Th/232Th) and (23Su/232Th) some 50% higher than those of Hdmond et al. (1994a). Whilst these differences could be simply due to analyses of very different samples, these features are suggestive of a systematic inter-lab bias. In order to present a consistent dataset (in the absence of measurements of the same standard from both labs), analyses of Rubin are not plotted, as a much larger proportion of the OIB dataset comes from the Pads group. Data for MORB: Condomines et al. (1981a), Ben Othman and Allbgre (1990), Goldstein et al. (1991). Data for OIB: Newman et al. (1984a), Condomines et al. (1988a), Hdmond et al. (1988, 1994a), Elliott (1991), Sigmarsson et al. (1992a,b), Williams and Gill (1992), Chabaux and All~gre (1994), Sims et al. (1995), Cohen et al. (1996).

168

T. Elliott/ Chemical Geology 139 (1997) 165-183

straight lines from the origin. Fig. lb shows a compilation of MORB and OIB data on an equiline plot, and illustrates that 23°Th-excesses are common in these mantle-derived melts.

mantle source is in secular equilibrium before melting commences, and so: (23°Th) ...... = (238U) .... ce

3. Models to explain 23Su-230 Th disequilibrium in mantle-derived melts

(230Th/232ThLolt

The half-lives of the intermediate U-series nuclides are short enough that their activities may potentially respond to changes on the time-scales of mantle melting. Whether or not the melting process is fast or slow compared to the half-lives of the nuclides is of great importance in interpreting Useries measurements. The two end-member scenarios of fast and slow melting, and their implications for what information is carried by 238U-23°Th disequilibrium, are summarised below. Useful discussions of the effects of different melting models on U-series disequilibrium are also presented by Williams and Gill (1989), Gill and Condomines (1992), Gill (1993) and MacDougall (1995).

3.1. Instantaneous melting models In modelling U-series data, some workers (e.g., All~gre and Condomines, 1982; Sims et al., 1995) have simply used the same equations of equilibrium melting that are commonly used to model stable element behaviour. Thus short-lived radionuclides are simply fractionated according to their relative incompatibilities during melting and net elemental fractionation is the cause of disequilibrium. Net fractionation is used to refer to elemental fractionation in an erupted melt relative to its unmelted source, as distinct from transitory fractionations that may occur during part of the melting process, which are important in later discussion. Application of simple batch or fractional melting models implicitly assumes instantaneous melt generation and extraction, and so this approach is termed 'instantaneous' modelling. A consequence of instantaneous melting is that the (230Th/232Th) activity ratio of a melt should be a robust measure of the T h / U ratio of its source, regardless of any U - T h fractionation during melting. The argument starts from the assumption that the

If mantle and melt are in isotopic equilibrium (Hofmann and Hart, 1978): =

( 230 Th//

232 Th)sourc e

= (238u/23:Th)

. . . . 0e

= 3.034/[Th/U]sourc ~ The value 3.034 converts the activity ratios into a more conventional weight ratio, indicated by square brackets. This is an appealing scenario, as not only will the measured (230Th//232Th ) yield the source T h / U , but the degree of disequilibrium will give the amount of melt-induced T h - U fractionation. Given estimates of appropriate source mineralogy and partition coefficients, this then yields that holy grail of igneous petrology, the degree of melting. Most importantly, unlike other inversion schemes (e.g., Minster and All~gre, 1978; Alber~de and Tamagnan, 1988; McKenzie and O'Nions, 1991), this method requires no additional estimates of any source compositional parameter: only that the source has been undisturbed within the last 350 ka. Instantaneous melting models can only induce 23°Th-excesses if Th is preferentially fractionated into the melt phase relative to U. Thus, Th must be more incompatible than U and the degree of melting must be comparable or smaller than the bulk partition coefficient of U. Given the very small absolute values of the distribution coefficients of Th and U during mantle melting (Benjamin et al., 1980; LaTourette and Burnett, 1992; Beattie, 1993a,b; LaTourette et al., 1993; Hauri et al., 1994; Lundstrom et al., 1994; Salters and Longhi, 1996) very small degrees of melting ( < 1%, Fig. 2a) are required to obtain net U - T h fractionations comparable to commonly observed degrees of disequilibrium (Fig. lb). Additionally, of the two major U and Th bearing mantle phases, only garnet preferentially retains U relative to Th during melting (Beattie, 1993a; LaTourette et al., 1993; Hauri et al., 1994; Salters and Longhi, 1996). Thus melting must take place almost exclusively in the garnet stability field ( > 80 km for a peridotitic source, > 45 km for eclogitic dominated source; Hirschmann and Stolper, 1996, and

T. EUiott/ Chemical Geology 139 (1997) 165-183

refs. therein), if 23°Ih-excesses are to be accounted for by net fractionation of Th from U. 3.2. Ingrowth models Perhaps counter-intuitively, consideration of the duration of the melting process can actually help explain the presence of 238U-23°Th disequilibrium in many mantle-derived magmas. This was first highlighted by McKenzie (1985a), and later reinforced by Williams and Gill (1989), who examined time-dependent solutions of the 'continuous' melting model of Langmuir et al. (1977). The resulting 'dynamic' melting model can explain production of 23°Th-excesses as the result of 23°Th ingrowth within the residual solid in the melting column. A significant

(a)

1.2o.

1.15-

~

1.10-

J ~

1.05-

1.00

.......

0.0001

1

1

1

0.01

100

I

Degreeofmelting(%) (b)

1.00

- ~ ~ c h

0.90.

!

0.800.70-

i AFM

0.60.

i

I 0.50

..... ..1 ...... ~ ..... ..., ......\

0.0001

0.01

...... ~ ...... 1

Dcgree ofmelang(%)

100

169

consequence of 23°Th ingrowth is that an erupted melt need not have a fractionated (232Th//238U) relative to its source to possess a 23°Th-excess. There are several variants of the ingrowth calculations, but all model the steady state melt output of simple melting regimes resulting from adiabatic decompression. The dynamic melting model simulates a one-dimensional melting column, sketched in Fig. 3. Melts produced during upwelling remain with the residue until a critical porosity is reached, after which melt is

Fig. 2. Fractionation of T h / U ratio relative to the initial ratio of the unmelted source during melting of a hypothetical garnet peridotite. Curves are shown for (a) melts and (b) melted residues. Melting curves are shown for equilibrium (batch), solid line, and accumulated fractional melting (AFM) (Shaw, 1970), dashed line. Modal melting is assumed in the calculation, but since the most significant variations occur at low degrees of melting, where consumption of major phases is small, this simplification has little effect when extrapolating to more realistic scenarios. Significant fractionation of T h / U in the melt compared to its source is only achieved at degrees of melting below 1%, and approaches the limit of fractionation, D o / D x h , at very low degrees of melting. As a modal melting calculation, the residue of the batch melt is simply constantly fractionated relative to the melt, and so shows an identical form. Much more significant fractionations of T h / U are created in the residue by fractional melting (the curve continues far below the minimum value displayed on the y-axis scale), but the highly fractionated residue compositions have vanishingly small U and Th concentrations. The garnet peridotite source is assumed to have 10% garnet (gnt) and 10% clinopyroxene (cpx). Partition coefficients of DTh(Cpx) = 0.015, Du(cpx) = 0.01, DTh(gnt) = 0.0015 and Du(gnt) = 0.01 are used as a synthesis of the abundant work on U-Th partitioning (Benjamin et al., 1980; LaTourette and Burnett, 1992; Beattie, 1993a,b; LaTourette et al., 1993; Hanri et al., 1994; Lundstrom et al., 1994; Salters and Longhi, 1996). Other major mineral phases make an insignificant contribution to U-Th partitioning, although values from Beattie (1993b) were in fact used. Bulk partition coefficients of 0.00166 and 0.00197 for Th and U, respectively, are thus calculated. Notably, even in the limit of fractionation, only some 17% 23°Th-excesses can be generated using these bulk partition coefficients, which is less than observed in many mantle-derived melts. However, recent studies have shown a strong compositional dependence of the partition coefficients of U and Th in clinopyroxene (Lundstrom et al., 1994; Salters and Longhi, 1996), and it is likely that appropriate bulk partition coefficients will be refined with further work. Since it is not the aim of this contribution to model any particular melt composition, but to illustrate general principles, there has been no fine tuning of bulk partition coefficients to obtain requisite T h - U fractionations.

7'. Elliott / Chemical Geology 139 (1997) 165-183

170

continuously extracted to maintain a constant porosity. So-called 'instantaneous' melts are extracted from all depths of the melting column after critical porosity has been reached. These melts are assumed to rise to the surface instantaneously, without chemical interaction (fractional melt transport) until they are thoroughly mixed, immediately prior to eruption, to produce an aggregate melt. For stable elements, the effects of dynamic melting can vary between accumulated fractional melting (Shaw, 1970) and equilibrium (batch) melting, de-

elts extracted, ,'ed and erupted Melted residue leav~ m a l ~ g column

CRITICAL EXTRACTIONPOP,(

SOLlvua

t

Upwelling Mantle Fig. 3. Cartoon to illustrate the steady state, one-dimensional melting regime modelled in dynamic melting calculations (McKenzie, 1985a). Mantle upwells, intersects its solidus and melts due to adiabatic decompression until it exits, horizontally, out of the melting regime. Thick arrows indicate direction of mantle movement. No melt is extracted until a critical porosity (dp) is reached. After this threshold, each new infinitesimal melt increment (instantaneous melt) generated is extracted to keep the porosity constant. This is shown crudely on the diagram as a series of discrete events (horizontal arrows). At any point in the melting column, melt and residue are in local equilibrium, but once extracted, melts are assumed to be transported in chemical isolation (disequilibrium transport). Dynamic melting assumes instantaneous transport of melts to the surface. Instantaneous melts produced from all depths of the column are mixed prior to eruption, to form the aggregate, steady state melt. The melting column is shaded according to depletion in incompatible element contents (the darker the shading, the more depleted). Lighter shading within the melting column indicates the size of the residual porosity, and the inverted triangle outside the column illustrates the accumulated fraction of extracted instantaneous melts with depth.

pending on the size of the residual melt porosity (McKenzie, 1985a). This formulation thus can reproduce the range of behaviour of the instantaneous models, and so does not affect the notion that net fractionation of highly incompatible element ratios, e.g., (ZaaU/232Th), only occurs at very small degrees of melting (Fig. 2a). The novel aspect of the dynamic model is that it also considers the time elements spent in the upwelling solid. For a steady state melting column, this is only of importance to short-lived nuclides. Fractional melting (or near-fractional melting, with a small equilibration porosity) can result in strongly fractionated elemental ratios in the melted residue (Fig. 2b). If, for example, Th is more incompatible than U, the melted residue will evolve to higher U / T h . As a result of this fractionation, Z3°Th ingrows in an attempt to return the residue to secular equilibrium (Fig. 4a). However, as melting continues during upwelling, the ingrown 230Th will itself be preferentially stripped out into the melt phase and so the residue will ingrow yet more 230Th. By this process, the mean aggregated melt acquires excess 23°Th (Fig. 4b). This occurs regardless of whether or not the overall degree of melting is so large that there is no n e t 238u-E32Th fractionation, although it is clearly essential that at some stage in the melting process, there must have been fractionation of these elements. Fig. 5 illustrates the behaviour of 238U, 23°Th and 232Th with depth in a dynamic melting regime designed to approximate to a plume-melting scenario. Upwelling mantle is assumed to start melting as it intersects its solidus at 120 km, and to melt at a constant melting rate up to a maximum of 5%. The compositions of aggregate melts (i.e., pooled instantaneous melts from the whole melting column below) are shown for depths between the solidus and maximum melting (Fig. 5a). The aggregate melts correspond to different degree melts that would be erupted if the mantle had ceased upwelling at greater depth, for example, if there were a thicker lithospheric lid. Fig. 5b shows the variations in (238U) within the residual solid as a function of depth and therefore degree of melting. As a near-fractional melting process, highly fractionated (232Th/23SU) ratios are generated in the residual solid as melting proceeds (Fig. 5a). The

T. Elliott / Chemical Geology 139 (1997) 165-183

more extreme the U--Th fractionation in the residual solid, the greater the relative amount of 230Th ingrown for a given period of time (Fig. 4a). However, this effect competes with the dramatic decrease in the actual 238U content of the melted residue (Fig. 5b). At greater degrees of melting, although relatively more 23°Th is ingrown for a given amount of 238U, the amount of 23°Th ingrown becomes increasingly small. In the example shown in Fig. 5, after only one quarter of the total amount of melting, the concentration of 238U in the residual solid has been reduced by a factor of one thousand. This extreme depletion neverthele:;s occurs over a significant pe-

(a)

1.4

~

1.3-

mcremet 7 ~In-growth beforenext .lst melt D.,,M ~_._.......~.~meltingepisade 1.2tnca~meat /~tial Melted . / Source residue

1.1 1.0

1.1

i

1.2

i

1.3

I

1.4

1.5

(238U/232Th)

(b)

#

1.5 1.41.31.21.11.0 1.0

1'.1

112

113

114

(238U/232Th)

1.5

171

riod of time, ~ 120 ka for the reasonable mantle upwelling rates used in Fig. 5 (4 c m / a ) . Fig. 5 provides a guide to the changing cause of 238U-23°Th disequilibrium in the aggregate melt, as a near-fractional melting process proceeds. In the very first instantaneous melts, 238U-23°Th disequilibrium is entirely due to net U - T h fractionation, as there has been no time for 230Th ingrowth. With further upwelling and melting, ingrown 23°Th becomes a more important contributor to disequilibrium, as the melted residue acquires a highly fractionated (232Th//238U), Fig. 5b. After a little more than 1% melting (115 km depth), all 238U and 232Th have been effectively stripped out of the residual solid, such that the aggregate melt has the same (232Th/EasU) as its unmelted source. From this point onwards, any ingrown 23°Th from the vanishingly small amount of 238U remaining in the melted residue is so insignificant to the total 23°Zh budget that the (23°Th//E38u) also does not change with further melting. However, provided melts are transported to the surface fast enough (assumed to be instantaneous in dynamic melting calculations) the aggregate melt still bears the record of 23°Th ingrowth from fractionation of U and Th deep in the

Fig. 4. (a) The effects of the dynamic melting regime on 238U23°Th systenlatics illustrated using discrete, multiple melting events. Instantaneous melts are shown as squares, and melted residues as small circles. The effects of melt fractionation are shown by horizontal an'owed lines, whilst the passage of time is indicated by dashed vertical arrows that join initial (open circles) and aged (shaded circles) residues. (b) Full calculation of the continuous evolution of residual solid and instantaneous melt compositions with increasing degree of melting (or height) in the dynamic melting column. Arrows indicate direction of compositional change with increasing degree of melting. In this example of only 1% total melting, the melting trajectories extend beyond the edge of the diagram to extreme (230Th/232Th) and (2380//232Th). The instantaneous melt compositions in fact cross the eqniline, which does not represent a limiting boundary. The averaged instantaneous melts of the whole melting column, which form the mean erupted melt, is also shown as a square. The compositions are calculated using a zero residual porosity, a melting rate of 6.6× 10-5 kg m -3 a- 1 (assuming solid density 3300 kg/m 3) and the same bulk distribution coefficients as in Fig. 3. The step sizes in (a) are 0.1% melt increments, and the averaging effect of the finite steps naturally returns less extreme fractionations than the continuous calculation using otherwise the same parameters.

T. Elliott / Chemical Geology 139 (1997) 165-183

172

melting column. Melting in the upper part of the melting column has no effect on the (232Th/238U) or (23°Th/238U) ratios and for degrees of melting > 1% all 238U-23°Th disequilibrium is due to 23°Th ingrowth alone. Spiegelman and Elliott (1993) presented a somewhat different ingrowth model in which melt transport times are calculated and continuous re-equilibration between melt and matrix occurs through the melting column. Nevertheless, the underlying princi- 100

,(a)

-

105

105"

38U)

DEPTH lena

1.20

(230"[]~238U) ~"~i

" 110

110"

(232Th/23sU)

- 115

115-

1.15

r i 120 1.10 1.05 1.00

Melt/Source

L00

(238U) oo.'"

0.10

0.01

0.001

Residue/Source

Fig. 5. 2 3 8 U - 2 3 ° T h systematics with depth in a dynamic melting regime. (a) T h e (23°Th/238U), dashed line, a n d ( 2 3 2 T h / 2 3 8 U ) , solid line, of aggregate dynamic melts relative to their initial source composition. The compositions plotted at any depth correspond to that of aggregated, erupted melts produced by melting from 120 km to that depth. Since (230Th/238U) is assumed to be unity in all mantle sources, the ratio of (23°Th//238U)melt/(23°Th/238U)source is simply the degree o f disequilibrium in the melt. The ratio ( 2 3 2 T h / / 2 3 8 U ) m e l t / ( 2 3 2 T h / 2 3 8 U ) . . . . . . gives how much of the disequilibrium recorded by ( 2 3 0 T h / 2 3 8 U ) is a result of net elemental T h - U fractionation. The difference between the two curves is thus the amount of ingrown 230Th contributing to disequilibrium, which varies from nothing at the onset of melting to 100% after some 1% melting after 5 km of upwelling. (b) The variation of (238U), dashed line, a n d (232Th/238U), solid line, in the melted residue with amount of upwelling, and therefore degree of melt depletion. The amount of (238U) left in the solid decreases greatly (note logarithmic scale) and continues off-scale to even lower concentrations. The calculations are made for upwelling mantle that starts melting at 120 km depth and finishes at 100 kin, approximating to plume melting beneath a thick, old lithosphere. The solid upwells at a velocity of 4 c m / a and melts at a constant rate of 3.4× 10 -4 kg m -3 a - 1 to a total of 5% melting. Residual porosity set to 0.01%, density of melt and solid are taken as 2800 and 3300 k g / m 3, respectively. Bulk distribution coefficients are the same as those used in Fig. 3.

pie of nuclide ingrowth is the same for this equilibrium percolation model as it is for near-fractional dynamic melting, and indeed for other refinements of the ingrowth model (Qin, 1992, 1993; Iwamori, 1994; Richardson and McKenzie, 1994). As stressed by Spiegelman and Elliott (1993), nuclide ingrowth is the result of a difference in residence time of parent and daughter in the melting column. Continued equilibration throughout the melting column tends to create larger differences in residence time between Th and U, and thus larger degrees of disequilibrium relative to fractional models. In more complex scenarios, where the magnitude or sense of U-Th partitioning changes with depth in the melting column, such as in MORB genesis, this is not necessarily always the case. Dynamic melting is sensitive only to the partition coefficients at the very bottom of the melting column and so impervious to changes higher up. However, as is evident from the modelling by Spiegelman and Elliott (1993), 23SU-23°Th disequilibrium produced by equilibrium percolation in a simulation of a MORB melting scenario is also most strongly affected by U-Th partitioning behaviour at the bottom of the melting column and furthermore produces larger 23°Th-excesses than an equivalent fractional model. The equilibrium percolation model also helps account for disequilibria displayed by very short-lived nuclides, such as 226Ra (half-life 1600 a), as disequilibrium can be created throughout the melting column (given a small enough porosity). In contrast, the dynamic model can create 2: 26Ra-23°Th disequilibrium only at the bottom of the melting column and so requires extremely high melt transport velocities to preserve this signature in erupted melts. In both fractional and equilibrium melt transport scenarios, ingrowth of 23°Th results in a vertical component of displacement of the composition of the erupted melt relative to its source in the equiline diagram (e.g., Fig. 4b). Except at very small degrees of melting, there is negligible net 238U-232Th fractionation (e.g., Fig. 2a and Fig. 5a) and the vector of displacement by the melting process is almost purely vertical (Fig. 4b). This is clearly in stark contrast to the systematics of the instantaneous melting models, in which any disequilibrium is inferred to result from horizontal displacement from the equiline (Fig. la). The amount of disequilibrium produced by ingrowth

T. Elliott / Chemical Geology 139 (1997) 165-183

during melting is sensitive to melting rate and residual porosity (McKenzie, 1985a; Williams and Gill, 1989; Spiegelman and Elliott, 1993) which have no or less effect, respectively, on the concentrations of stable elements in steady state melts. If 238U-23°Th disequilibrium is interpreted as due to 230Th ingrowth, it thus carries important information about the dynamics of melting, but the (230Th/232Th ) can no longer be used to infer the source U / T h (O'Nions and McKenzie, 1993), nor the degree of disequilibrium used to estimate: the overall degree of melting. Before assessing the relative importance of ingrowth and net elemental fractionation in producing 238U230Th disequilibrium observed in mantle-derived melts, it is important to address some frequently brandished misconceptions about ingrowth models. 3.3. Mass balance

There are potentially puzzling features that result from the ingrowth models, which might appear to violate mass-balance. Initially, the apparently extra 230Th in the erupted melt might seem strange. However, 23°Th is contimlously created as 238U decays, no matter what the environment. The melting processes simply alters the spatial distribution of parent 238U and daughter 230Th. For example, during a nominal 100 ka melting period, the fraction of total uranium that has decayed is: 1 -e

~-~238×1°°'°°°) = 1.5 × 10 -5

This tiny change clearly cannot be resolved as a fractional difference on the y-axis of the equiline diagram. However, if all the 23°Th created from this decayed U is efficiently stripped out of the melting column and into the melt, then the total activity of new 23°Th relative to what was initially in the source is:(1 - e (-x238x 100,00o)) × (A230)/(A238) = 0.92In other words there is the potential to increase the 23°Th activity by ~ 90%, which would translate into 90% 23°Th-excess. "Eais is a limiting value, as the calculation tacitly assumes an infinite partition coefficient for U, a zero partition coefficient for Th and no decay of 23°Th, butt it illustrates the potential for producing 23°Th duntng a finite melting interval. Conceptually, the ingrowth models remove 23°Th that is being produce,d in the melting column but leave the 238U in the column to produce more 23°Th.

173

The difference in upwelling rates of melt and solid result in 230Thbeing rapidly transported to the surface before it decays, whilst 238U has time to reside in the solid, producing more 230Th. If there was no melt movement with respect to the residue then there would be no way of separating the 23°Th daughter from the 23Suparent. With differential transport (and of course D R
174

T. Elliott / Chemical Geology 139 (1997) 165-183

However, the decayed 238U, present in the melt as 23°Wh, is very evident on the equiline plot, and this results from the fact that in the equiline diagram, nuclide concentrations are expressed in terms of activities (atomic concentrations multiplied by activity coefficients). Since the activity coefficient of 230Th is some 60,000 times greater than that of 238U (i.e., 230Th has a much shorter half-life), the convers i o n of one atom of 23SU to 230Th is magnified ~ 60,000 fold in the y- relative to x-axis in the equiline diagram (Fig. 6a). It is the structure of the steady state melting column which allows decayed 238U from the input mantle to be transferred to the erupted melt as excess 230Th. Importantly, it is 230Th, ingrown in response to previous melt extraction, which is preferentially

(a) 0.20

0.25

0.30

'

'

'

1.4

f 1.2

[238U/'232Th]atomie 0.35 0.40

'

j

Locus of


[ 7"0x10-6

o ~

/

I

E

7"

1.O Initial melt

l °=+

//'I~fial source

5"0x10"°

0.8-

4.0xtO-6 0.'8

1~0

1~2

1.4

(238U/232Th)

(b) 1.4

I

I

I

/

.

f

i

1.2.

¢q 1.0.

degr~ ~_melt

0.8.

0.6 0.6

:

/ 0~8

~ ~v Initial source

l~O

melt residue

1~2 (238U/232Th)

1.4

'*

#

(~PtrD

0.6 0.6

stripped into the erupted melt to give it 23°Th-excesses. Thus it is perhaps informative to look at the mass-balance involved in actually setting up the steady state melting column. The steady state melting column needs to be established by the extraction of a melt complement to the m e a n column residue. In the simplest one-dimensional melting column, this is a melt corresponding to half the degree of melting represented by the steady state output. Thus the steady state residue and its melt complement may look somewhat like the small degree m e l t - r e s i d u e pair in Fig. 6b. As long as melting continues, the steady state residue composition will be maintained, but the melt complement can only decay back to equilibrium after its eruption. To complete the cycle, melting must cease and the mean residue will itself return to secular equilibrium (Fig. 6b) and from this perspective the mass-balance with its melt complement is clearer. In the mid-ocean spreading case, the melts removed to set up the steady state melting

Fig. 6. (a) Schematic representation of melt (open squares) and residual solid composition (small open circle) with height in the melting column for the equilibrium transport model. The melt remains in chemical equilibrium with the solid as it passes through the melting column, and so the erupted melt composition is in equilibrium with the residual solid at the top of the melting column. This makes the mass-balance issue easier to illustrate using the equilibrium transport model, although the principle is the same for the dynamic melting column. Arrows indicate direction of evolution of melt and residue with increasing degrees of melting. Double-headed arrows link melt residue pairs (i.e., outpu0 for different degrees of melting. Except at very small degrees of melting ( < 1%) the U-Th output is dominated by the melt. The vertical dashed line indicates the locus of averaged melt-residue compositions (output) for smaller degrees of melting. Arbitrary (23Su/232Th) and (230Th/232Th ) values ale marked on the axes to enable comparison of these values expressed as atomic ratios. This has been done to illustrate the hugely different relative effects of the decay of an atom of 238U to 230Th On the respective axes of the equiline plot. (b) Return of steady state melting column output to equilibrium, illustrated for two cases where U-Th budget is dominated by the erupted melt alone ( > 1% melt) and for smaller degree melts where both melt and residue need to be considered. Passage of time is indicated by dashed, arrowed lines, linking young (open symbol) and aged (shaded symbol) manifestations to otherwise identical compositions. There is no mass loss involved in movement along the vertical dashed lines. For example excess 230Th decays via intermediate nuclides to lead and several alpha particles. The mass of these daughter products is not lost from the system but cannot be shown on the equiline plot.

T. Elliott / Chemical Geology 139 (1997) 165-183

column are those erupted at the on-set of rifting, now at the very edge of the ocean basins.

4. Interpretation 2~°Th-23Su disequilibrium in mantle-derived melts From the earlier discussion, it is clearly important to assess the means by which disequilibrium is produced in mantle-derived melts in order to properly interpret the implications of U-series disequilibrium. The following discussion focuses on MORB and OIB, as in arc settings recent additions to the mantle source from the subducted slab can induce large disequilibrium and s~gnificantly complicate the effects of the melting process alone (All~gre and Condomines, 1982; Newman et al., 1984b; Condomines et al., 1988b; Gill and Williams, 1990; McDermott and Hawkesworth, 1991). MORB are thought to represent large (8-15%) degrees of melting (Klein and Langmuir, 1987; McKenzie and Bickle, 1988) and explaining the common 10-30% 23°Th-excesses that are commonly observed in MORB (Condomines et al., 1981a; Newman et al., 1983; Rubin and MacDougall, 1988; Reinitz and Turekian, 1989; Goldstein et al., 1989, 1991, 1993; Ben Othman and All~gre, 1990; Lundstrom et al., 1995; Bourdon et al., 1996a,b) by net elemental fractionation is highly problematic, given the highly incompatible nature of U and Th. Furthermore, MORB genesis involves upwelling of mantle well into the spinel stability field, in which clinopyroxene should dominate the bulk partition coefficients of U and Th and so even the sense of U-Th fractionation is wrong for an instantaneous melting model. The comparatively long melting time-scales beneath ocean ridges help resolve the dilemma of large 23°Th-excesses, and larger still 226Ra and 231pa excesses (Rubin and MacDougall, 1988, 1990; Volpe and Goldstein, 1993; Goldstein et al., 1993; Lundstrom et al., 1995), in MORB if an ingrowth model is invoked. Also, ingrowth of 23°Th is most sensitive to the partitioning behaviour of U and Th at the bottom of the melting column, and so significant melting in the spinel stability field does little affect the ability of ingrowth models to reproduce significant 23°Th-excesses. Indeed it was the observation of 23°Th-excesses in MORB that prompted the original

175

formulation of dynamic melting by McKenzie (1985a). The mean degree of melting represented by MORB is reasonably well established from the combined constraints of experimental petrology and the need to explain ~ 7 km oceanic crust by rifting (Klein and Eangmuir, 1987; McKenzie and Bickle, 1988), but similar arguments are not readily applied to the OIB melting process. The nature of the melting regime beneath islands, melt production rates, composition and melting behaviour of the (volatilerich?) source are also all rather uncertain. The generally elevated incompatible element contents, light rare earth enrichments and silica undersaturation of OIBs have been taken to indicate that they are lower degree melts than most MORB (Green and Ringwood, 1967; Gast, 1968). At smaller degrees of melting, net elemental U-Th fractionation has more potential to affect significantly any (23°Th//238U) disequilibrium. Moreover, the contribution to disequilibrium from nuclide ingrowth is likely to relatively suppressed in OIB relative MORB as melting rates in hot, actively upwelling plumes should be greater than in colder, passively upwelling upper mantle. Thus the cause of disequilibrium in OIB requires closer examination.

5. U - T h fractionation in OIB Much has been made of the (23°Th//232Th)87Sr/86Sr 'correlation' (All~gre and Condomines, 1982; Condomines et al., 1988b) in MORB and OIB, to suggest that the (23°Th/232Th) of erupted lavas is identical to their source (23°Th//232Th). Since s7Sr/s6 Sr ratios of uncontaminated lavas should indeed reflect source composition (Hofmann and Hart, 1978), it has been proposed that the Th-Sr isotopic correlation therefore implicates the (23°Th/232Th) of a lava to likewise represent its source ratio. This in turn has been taken to suggest that melting and melt extraction are near-instantaneous such that (230Th/232Th ) ratios are unaffected by these processes. This is a very unsatisfactory argument as the correlation of Sr isotopes with (23Su/232Th) is no worse than that with (23°Th//232Th), Fig. 7, but merely off-set by the amount reflecting the degree of disequilibrium (Fig. lb). These global arrays simply

T. Elliott / Chemical Geology 139 (1997) 165-183

176 1.6-

(a)

1.4-

1.2-

"%~%

%°°°

1.0-

"g,... []

0.8-

•% 0.6 0.702

0.703

0. 04

"1 0.706

0.705

".%% 0.707

87Srf16Sr ,.4

(b) 1.2.

o []

08

[]

[]

D 0.6 0.702

0.703

0.704

0.705

[] 0.706

0.707

87Sr/86Sr Fig. 7. (a) Measured(23°Th/232Th)vs. 8 7 S r / 8 6 S r for 'zero-age' MORB glasses (circles)and OIB (squares). The 'Th-Sr array' is taken from Condomines et al. (1988b). (b) (238U//232Th) vs. 87Sr/86Sf~for same samples as in (a). 23aU-23°Th and some 87Sr/a6Sr data are fromrefs. in Fig. 2. Additional87Sr/86Srdata are from Eaby et al. (1984), Newsom et al. (1986), Wright and White (1986/1987), HaUiday et al. (1988, 1990), Hart (1988), Prinzhofer et al. (1989), H6mond et al. (1993), Reisberg et al. (1993), and Barling et al. (1994). are not tight enough to determine whether (2~SU/232Th) or (23°Th/232Th) is more disturbed by melting. This is also hardly surprising given the broad nature of most global isotopic correlations, which must result from mixing of a number of different source compositions (White, 1985; Zindler and Hart, 1986). More compelling evidence is required to assess the manner in which disequilibrium is produced in OIB. Examining variations o f (23°Th//238U) with geochemical indices of degree of melting is a more

useful approach, a s (23°Zh//238U) should strongly depend on the degree of melting if n e t 2 3 8 U - 2 3 2 T h fractionation is the main cause of disequilibrium in OIB. However, it is difficult to unequivocally separate the effects of melting from variable source composition in most geochemical tracers. The existing database of lavas analysed for 238U-23°Th disequilibrium displays variations in long-lived radiogenic isotope ratios, even at restricted localities, implying significant heterogeneity in their sources. In an attempt to minimise this problem, parentdaughter fractionations of long-lived radiogenic isotopic systems can be used as measures of melt-induced fractionation. Measured S m / N d and L u / H f ratios can be compared with model source S m / N d and L u / H f calculated from 143Nd//144Nd and 176Hf//177Hf ratios to assess variations in degrees of melting (DePaolo, 1988; Salters and Hart, 1989)• There are little L u - H f data on the same or similar samples analysed for 2 3 8 U - 2 3 ° T h isotopic data, but there are enough Sm-Nd data to allow comparison between the variations of (230Th/238U) in OIB with an index of inferred Sm-Nd fractionation, Otsm/Nd (DePaolo, 1988), Fig. 8. There is no coherent covariation evident in Fig. 8. Disequilibrium varies widely for near-constant CtSm/Nd at both high and low absolute values of Ctsm/Nd, representing high and low degrees of melting, respectively. Comparing similar lava types, Hawaiian tholeiites have very low (23°Th/238U), ~ 1.02 (Newman et ai., 1984a; Reinitz and Turekian, 1991; Cohen and O'Nions, 1993; H6mond et al., 1994b; Sims et al., 1995; Cohen et al., 1996), whilst uncontaminated, Icelandic tholeiites show much larger degrees of disequilibrium, ~ 1.1-1.35 (Condomines et al., 1981b; H6mond et al., 1988; Nicholson et al., 1991). Alkalic melts from a single study in the Pacific have (23°Th/238U) from 1.07 to 1.32 (H6mond et al., 1994a) at similar Ctsm/Nd. Thus on a • • " " " • global scale, variations of 238 U - 230 Th -thsequlhbnum in OIB seem not to be controlled simply by variations in the degree of melting. N e v e r t h e l e s s , s y s t e m a t i c variations o f (23°Zh//238U) with Otsm/Nd are found in a suite of recent lavas from Hawaii that range from tholeiite to basanite (Sims et ai., 1995). From this observation Sims et al. (1995) concluded that 238U-23°Th disequilibrium could be explained by a batch melting

T. Elliott / Chemical Geology 139 (1997) 165-183 1.4

~

1.3-

e~

o

°i

•

.

[] Iceland O S. Pacific

O

O

A Hawaii

l.l-

•~ J ,

,.0

,

0.4

015

0.6

A

.^AA

~ - - - ~ 0.7 0.8 0~SratNd

0.9

Fig. 8. (23°Th/23aU)vs. ~Xsm/Nd for OIBs. OtSm/Nd is calculated assuming evolution of the sample source with chondritic values until 1.7 Ga, at which point S m - N d was fractionated such that the measured 143Nd/14#Nd i~ produced by subsequent evolution to the present day (DePaolo, 1988). The measured sample S m / N d is then divided by this calculated source S m / N d to derive the inferred melt fractionation, a Sm/Nd- Only analyses of both Sm-Nd and U - T h on the same sample are shown, except for averages of Reunion and Mt. Cameroon data, which show only minor variability. More extensive datasel~ from a single region are shown as the same symbol (see legend). Data from Iceland were carefully screened for indices of crustal contamination and reliability of analysis for low Th samples (H6mond et al., 1988). Data sources are Newman et al. (1984a), Newsom et al. (1986), Wright and White (1986/1987), Comlomines et al. (1988a), Halliday et al. (1988), H6mond et al. (1988, 1993, 1994a), Elliott (1991), Sigmarsson et al. (1992a), Williams and Gill (1992), Barling el al. (1994), Chahaux and All~gre (1994), Sims et al. (1995), Cohen et al. (1996). The dashed line shows the path of melt compositions with variable degrees of batch melting, calculated from the bulk partition coefficients infer,red by Sims et al. (1995). The line illustrates the small changes in S m - N d fractionation for much larger U - T h fractionation at small degrees of melting (i.e., at the left of the diagram).

model, which implies that 23°Th ingrowth was not significant. However, there are other plausible interpretations. The samptes showing the highest degrees of disequilibrium in the study by Sims et al. (1995) come from the post-erosional stage of Haleakala. These recent post-erosional volcanics are some 200 km distant from the most active volcanoes at the head of the Hawaiian chain, and are presumably derived from a cooler, less vigorously upwelling, distal portion of the plume. Thus for these sources smaller degrees of melting should also be associated with lower melting rates, which facilitate greater 230Th ingrowth.

177

Important supplementary evidence, not discussed by Sims et al. (1995), is the presence of significant 226Ra-excesses, (226Ra/23°Th) ~ 1.15, in Hawaiian tholeiites of very similar composition to those also analysed by Sims (Reinitz and Turekian, 1991; Cohen and O'Nions, 1993; Hrmond et al., 1994b; Cohen et al., 1996). The partition coefficients and degrees of melting calculated by Sims et al. (1995) in their self-consistent batch melting model of Th and Nd isotopic data cannot yield sufficiently large 226Ra-excesses to model the tholeiites, even if the partition coefficient of Ra is set to zero. The origin of 226Ra-excesses in the Hawaiian tholeiites has been discussed in some detail by Cohen and O'Nions (1993) and Hrmond et al. (1994b), who were readily able to model their data by an ingrowth model for lavas generated from a rapidly upwelling plume. Thus, net elemental fractionation alone cannot account for all the U-series behaviour in the Hawaiian dataset. Although the 226Ra-excesses in the Hawaiian tholeiites seem to require nuclide ingrowth during melting, it should be noted that 226Ra has a significantly shorter half-life (tl/2 = 1600 a) than 23°Th and so can ingrow much more rapidly. Hawaii has much the largest buoyancy flux of any active plume (Sleep, 1990) and is thus presumably rapidly upwelling. Hence, although 226Ra-excesses probably result from ingrowth, upwelling may nevertheless be too rapid to allow significant 230Th ingrowth and 23°Th-23SU disequilibrium could still be largely the result of net elemental fractionation. Indeed, it is striking that in the dataset of Sims et al. (1995), the Haleakala lavas not only have the highest 23°Th-excesses, but also the highest T h / U despite having the lowest 87 S r / 86 Sr. Lavas with lower 87S r / 86 Sr generally have lower T h / U (Fig. 7b), although, as pointed out, there is some scatter. Consequently, the observation that the highest T h / U ratios are found in Hawaiian lavas with the lowest 87Sr/86Sr, implies that there is a net T h / U increase during melting to produce the alkalic lavas. This issue can be further examined using a recent dataset of Loihi lavas (Garcia et al., 1995). Although these samples show significant isotopic heterogeneity, there is no covariation of long-lived radiogenic isotopic ratios with lava type or incompatible element ratios (Fig. 9a). Thus, these lavas present an

T. Elliott / Chemical Geology 139 (1997) 165-183

178 0.51298

(a) A

0.51296-

z 0.51294-

A

A

0.51292-

A

0.51290-

A

0.51288-

A

0.51286 0.20

0[2

0[4

036

0.28

Sm/Nd 4,0-

(b)

A 3.8-

A

~x

A

3.6.

A 3.4.

A

A 3.2-

3.0 0.20

A

A~A A

!

0,22

i

0.24

A i

0,26 0.28 Sm/Nd

Fig. 9. S m / N d vs, (a) 143Nd/144Nd and (b) T h / U (weight ratio) for mafic Loihi lavas (Garcia et al., 1995). More evolved hawaiite lavas are not plotted.

opportunity to investigate the effects of melting largely independent of systematic differences in source composition, that complicate matters elsewhere in Hawaii (e.g., Chen and Frey, 1983). For the less fractionated Loihi samples (the hawaiites have T h / U clearly affected by crystal fractionation, and are not considered), there is a trend of T h / U with other indices of melting (Fig. 9b). Since the range of compositions represented by these Loihi lavas is similar to that of the lavas analysed by Sims et al. (1995), the full range in T h / U at Loihi (24%), Fig. 9b, can be usefully compared to the range of (238U/23°Th) in the Sims et al. (1995) dataset (28%). Whilst this is far from irrefutable evidence, it is suggestive that at Hawaii, 2 3 8 U - 2 3 ° T h disequilibrium may be largely controlled by net U-Th frac-

tionation, with only a minor component of 23°Th ingrowth, perhaps of the magnitude observed in the large degree tholeiitic melts (Cohen and O'Nions, 1993; H6mond et al., 1994b). Iceland is the only other locality for which there are a large number of 2 3 8 U - 2 3 ° T h disequilibrium measurements from a single plume. In stark contrast to Hawaii, there is no systematic variation in degree of 2 3 8 U - 2 3 ° T h disequilibrium with apparent degree of melting. Alkalic lavas produced at the periphery of the plume show no greater disequilibrium than high degree tholeiitic melts produced along the ridge axis towards the centre of the plume (Fig. 8). As part of the ridge system, the crustal thickness at Iceland enables an estimated to be made of the mean degree of melting represented by the dominant tholeiitic melts. As some 20% melts (Klein and Langmuir, 1987), 238U-23°Th disequilibrium in the tholeiites must reflect 23°Th ingrowth, as in the case of MORB. This in turn implies that Iceland is upwelling more slowly than Hawaii, in keeping with its smaller buoyancy flux (Sleep, 1990). The covariafion of (23°Th/238U) in OIB with the buoyancy flux of the associated plume has previously been noted by Chabaux and All~gre (1994). Comparing the chemistry and degree of disequilibrium for the remaining disparate samples of the global OIB 238U-23°Th disequilibrium database is more problematic. These samples cover a wide range of long-lived radiogenic isotopic compositions implying much larger variations in source compositions relative to samples from the individual Icelandic or Hawaiian plumes. As such, Ctsm/Nd should provide the most robust guide to degree of partial melting, and there is no obvious covariation of OtSm/r~dwith (230Th/238U)" However, it is possible that since Th and U are both highly incompatible, whilst Sm and Nd are less so (e.g., Hauri et al., 1994; Salters and Longhi, 1996), that at the very small degrees of melting necessary to fractionate U and Th, Sm and Nd are already near the limit of their maximum fracfionation (i.e., DNd/Dsm) and so are insensitive to variations at these small absolute degrees of melting (Fig. 8). In this case, L a / Y b should become a more sensitive tracer. La is highly incompatible and so its concentration should respond to changes of degree of partial melting even at very low absolute values,

T. Elliott/ Chemical Geology 139 (1997) 165-183

whereas the Yb concentration of the melt should be largely buffered, especially if garnet is controlling U-Th fractionation (Beattie, 1993a; LaTourette et al., 1993; Hauri et a14 1994; Salters and Longhi, 1996). There are problems, in that L a / Y b may vary significantly from source to source and also during fractional crystallisation, but the large changes expected from variable, low degrees of melting should dominate. La/Yb, however, shows no correlation with (23°Th/238U), Fig. 10a. Again, only the Hawaiian samples show any coherent variation of (23°Th/23SU) with rare earth element (REE) fractionation. Thus it appears that in general 238U-23°Th disequilibrium is not controlled by variable degrees of melting (Figs. 8 and 10a). It is possible that highly variable residual mineralogy in different OIB sources could result in different relative T h / U and REE fractionation from location to location, but this is not readily tested with existing data. Hdmond et al. (1994a) analysed a suite of South Pacific lavas for a wide range of geochemical tracers and suggested that the range of 2 3 8 U - 2 3 ° T h disequilibrium observed in the lavas was due to variable contribution from a carbonatite component. This is unlikely for several reasons. Firstly, historically erupted carbonatites show e x t r e m e 230Th deficits, (23°Th/238U)=0.01-0.5 (Williams et al., 1986; Pyle et al., 1991). This is in keeping with experimental work on carbonatite melt partitioning (Walker and Jones, 1991; Walker et al., 1992) which indicates that U is much more readily partitioned over Th into a carbonatitic melt relative to silicate melts and common mantle phases. Secondly, reconstructed compositions of carbonatite liquids from the oceanic environment infer very distinctive incompatible element signatures, notably very high L a / N b (Hauri et al., 1993). If a putative carbonatitic phase were to significantly affect the 238U-23°Th disequilibrium (supposing for the moment that such a phase may have Th-excesses rather than deficits as found in natural carbonitites and in laboratory experiments), it would also be expected to impart this signature to the melt. However, of the lavas analysed by Hdmond et al. (1994a), those with the highest disequilibrium (from MacDonald s,~amount), and thus attributed with the greatest carbonatite contribution by Hrmond et al. (1994a), have the lowest L a / N b (Fig. 10b). MacDonald lavas have a 'HIMU' isotopic signa-

179

1.4

o (a)

[]

rn kela~

[]

Q

1.3.

E

o

S. Pacific

T

/~ Hawaii

mA 1.2.

1

O OOaers

[] ~A~

ll 1.1. []

0

0 0

1.0- ---i~, 0 5

~b

;5

2'0

2'5

3o

(La/Yb)n 1.4-

(b)

].3- ~

1.2

O

1.1.

1.0 0.6

0.7 '

08i

0.9 '

1.0

La/Nb Fig. 10. (a) Chondrite-normalised (Sun and McDonough, 1989), L a / Y b plotted against OIB 23°Th-excesses. Data sources as in Fig. 8, except for Hawaii, where La-Yb data were not always available and frequently measured on different samples, so averages are plotted for Hualalai (Basaltic Volcanism Study Project, 1981; Sims et al., 1995), and Haleakala (Chen and Frey, 1985; Sims et al., 1995), together with error bars depicting the total range of values. MacDonald seamount samples are circled to distinguish them from other S. Pacific samples analysed by Hrmond et al. (1994a). (b) L a / N b vs. (23°Th/23SU) for the S. Pacific samples of Hrmond et al. (1994a).

ture (Zindler and Hart, 1986), typified by radiogenic lead isotopic compositions, and low L a / N b is another characteristic of HIMU lavas (Weaver, 1991). In contrast, the other lavas in the study of Hrmond et al. (1994a) show 'enriched' (EMIl) affinities, and so perhaps it is a difference in source composition that shapes the difference in 238U-23°Th disequilibrium. Whilst the degree of disequilibrium in an erupted lava should be independent of its s o u r c e 2 3 8 U / / 2 3 2 T h

180

T, EUiott / Chemical Geology 139 (1997) 165-183

(it is only required that the source is in secular equilibrium), the degree of disequilibrium may be influenced by changes in bulk source composition that affect the residual phases and partitioning behaviour of U and Th (e.g., Ben Othman and All~gre, 1990; Lundstrom et al., 1995; Hirschmann and Stolper, 1996; Bourdon et al., 1996b). There have been many proponents of recycled oceanic crust as the origin of the HIMU signature (Chase, 1981; Hofmann and White, 1982; Zindler et al., 1982), and from Os isotopic systematics, a rather large proportion of recycled oceanic crust is inferred to be present in HIMU sources (Hauri and Hart, 1993; Reisberg et al., 1993; Marcantonio et al., 1995). The presence of recycled, eclogitic oceanic crust in the HIMU source should increase the proportion of garnet and could thus have a significant effect in enhancing U-Th partitioning during melting, as modelled in some detail by Hirschmann and Stolper (1996) for the MORB melting scenario. EMII mantle sources have long been argued to reflect the presence of recycled sediment (e.g., Cohen and O'Nions, 1982; White and Hofmann, 1982). Only a small mass fraction of sediment can significantly influence the incompatible element systematics of a mantle source and perhaps also without greatly affecting the bulk mineralogy. Thus in the South Pacific, the variation in 238U-23°Th disequilibrium could reflect, at least in part, variations in the bulk composition of mantle sources. Merely invoking more garnet in the source of the MacDonald lavas is not sufficient to explain their full geochemistry, as these lavas do not show greater net REE fractionation as would be expected in the simplest, instantaneous melting models (Fig. 10a). On the contrary, the MacDonald lavas have some of the lowest L a / Y b of the S. Pacific dataset of Htmond et al. (1994a). Such 'cryptic' involvement of garnet in producing higher ( 230T h / 238U) but not affecting L a / Y b is consistent with 230Th ingrowth rather than control by overall degree of melting. Although instantaneous and ingrowth models have apparently been treated as entirely separate species in the above discussion, this is only a convenient short-hand. Ingrowth models reduce to the behaviour of instantaneous models if melting is rapid enough. Thus for high melting rates, it is simpler to model the melting process without additionally considering

the time dependence. In reality, all melting must occur over a finite time-scale, and the important question is whether or not this time-scale is significant or not with respect to the half-lives of the U-series nuclides. Whilst the 238U-23°Th systematics of Hawaii can be adequately represented by batch melting, its 226Ra-23°Th systematics cannot, and neither can the U-series disequilibria of MORB, Iceland or S. Pacific OIB. This by no means implies that an entirely different melting process is required to explain these different observations, as suggested by Sims et al. (1995). Melting rates affect U-series systematics but not stable element concentrations, and so a range of U-series behaviour can be observed for any given stable element composition using an ingrowth model.

6. Summary A global compilation of 238U-23°Th disequilibrium data of OIB shows no coherent variation that might suggest control simply by variable degree of melting, as would be expected for an instantaneous melting model. Beneath Hawaii, it appears possible that net U-Th fractionation can account for a large part of the 238U - 230Th disequilibrium, but this cannot simultaneously explain 226Ra-excesses, and so at least some ingrowth of the shorter-lived nuclides, and probably a little 230Th also occurs. Hawaii may be anomalous due to its large buoyancy flux. The large 23°Th-excess in Icelandic tholeiites almost certainly requires 230Th ingrowth, and this may be true of other OIB, although this is still somewhat uncertain due to paucity of the data. Variations in major element compositions of OIB sources may be important in influencing the (230Th//2380) of their melts.

Acknowledgements The author is grateful for stimulating discussions about U-series disequilibria with many people, amongst whom Bernard Bourdon and Marc Spiegelman deserve a special mention, although naturally any misapprehensions are the responsibility of the author alone. Thoughtful reviews by Dave Peate, Vincent Salters and Matthew Thirlwall greatly improved the manuscript, and thanks to Gareth Davies,

T. Elliott/ Chemical Geology 139 (1997) 165-183

Tibor Dunai and Simone Tommasini for additional comments. TE is supported by the Vrije Universiteit. This is NSG contribution 961211.

References Alber~de, F., Tamagnan, V., 1988. Modelling the recent geochemical evolution of the Piton de la Fournalse volcano, R6uninn Island, 1931-1986. J. Petrol. 29, 997-1030. All~gre, C.J., 1968. 23°Th dating of volcanic rocks. A comment. Earth Planet. Sci. Lett. 3, 338. All~gre, C.J., Condomines, M., 1982. Basalt genesis and mantle structure studied through Th-isotopic geochemistry. Nature 299, 21-24. Barling, J., Goldstein, Si,., Nicholls, I.A., 1994. Geochemistry of Heard Island (southern Indian Ocean): characterisation of an enriched mantle component and implications for enrichment of the sub-Iodian ocean mantle. J. Petrol. 35, 1017-1053. Basaltic Volcanism Stud)' Project, 1981. Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, New York, 1286 pp. Beattie, P., 1993a. Uranium-thorium disequilibria and partitioning on melting of garnet peridotite. Nature 363, 63-65. Beattie, P.D., 1993b. The generation of uranium series disequilibrium by partial melting of spinel peridotite: constraints from partitioning studies. E ~ h Planet. Sci. Lett. 117, 379-391. Benjamin, T.M., Heuser, W.R., Bumett, D.S., 1980. Actinide crystal-liquid partitioning for clinopyroxene and Ca3(PO)4. Geochim. Cosmochim Acta 44, 1251-1264. Ben Othman, D., All~gre, C.J., 1990. U-Th systematics at 13°N east Pacific Ridge segment. Earth Planet. Sci. Lett. 98, 129137. Bourdon, B., Langmuir, C.H., Zindler, A., 1996a. Ridge-hotspot interaction along the lvlid-Atlantic ridge between 37o30' and 40030'N: the U-Th di,,;equilibrium evidence. Earth Planet. Sci. Lett. 142, 175-189. Bourdon, B., Zindler, A., Elliott, T., Langmuir, C.H., 1996b. Constraints on mantle melting at mid-ocean ridges from global 23SU/230Th disequilibrium data. Nature 384, 231-235. Chabanx, F., All~gre, C.J., 1994. 238U-23°Th-226Ra disequilibria in volcanics: A new insight into melting conditions. Earth Planet. Sci. Lett. 126, 61-74. Chase, C.G., 1981. Ocean island Pb: two stage histories and mantle evolution. Earth Planet. Sci. Lett. 52, 277-284. Chen, C.-Y., Prey, F.A., 11983. Origin of.Hawaiian tholeiite and alkalic basalt. Nature 302, 785-789. Chen, C.-Y., Frey, F.A., 1985. Trace element and isotope geochemistry of lavas from Haleakala Volcano, East Maul: implications for the origin of Hawaiian basalts. J. Geophys. Res. 90, 8743-8768. Cohen, A.S., O'Nions, R.K., 1993. Melting rates beneath Hawaii: evidence from uranium series isotopes in recent lavas. Earth Planet. Sci. Lett. 120, 169-175. Cohen, A.S., O'Nions, R.K., Kurz, M.D., 1996. Chemical and isotopic variations in Manna Loa tholeiites. Earth Planet. Sci. Lett. 143, 111-124.

181

Cohen, R.S., O'Nions, R.K., 1982. Identification of recycled material in the mantle from Sr, Nd and lab isotope investigations. Earth Planet. Sci. Lett. 61, 73-84. Condomines, M., Morand, P., All~gre, C.J., 1981a. 2a°Th-238U radioactive disequilibria in tholeiites from the FAMOUS zone (Mid-Atlantic Ridge, 36°50'N): Th and Sr isotopic geochemistry. Earth Planet. Sci. Lett. 55, 247-256. Condomines, M., Morand, P., All~gre, C.J., Sigvaldason, G.E., 1981b. 23°Th-238U disequilibria from historical lavas of Iceland. Earth Planet. Sci. Lett. 66, 393-406. Condomines, M., Bachelery, P., Boucbez, R., Ma, J.L., 1988a. U-Th-Ra disequilibria in Piton de la Foumaise lavas (Rtunion island). Chem. Geol. 70, 126. Condomines, M., Htmond, C., All~gre, C.J., 1988b. U-Th-Ra radioactive disequilibria and magmatic processes. Earth Planet. Sci. Lett. 90, 243-262. DePaolo, D.J., 1988. Neodymium Isotope Geochemistry: an Introduction. Springer Vedag, New York, 187 pp. Eaby, J., Clague, D.A., Delaney, J.R., 1984. Sr isotopic variations along the Juan de Fuca Ridge. J. Geophys. Res. 89, 7883-7890. Elliott, T.R., 1991. Element Fractionation in the Petrogenesis of Ocean Island Basalts. PhD Thesis (unpublished), Open University, Milton Keynes, UK. Galer, S.J.G., O'Nions, R.K., 1986. Magmagenesis and the mapping of chemical and isotopic variations in the mantle. Chem. Geol. 56, 45-61. Garcia, M.O., Foss, D.J.P., West, H.B., Mahoney, JJ., 1995. Geochemical and isotopic evolution of Loihi volcano, Hawaii. J. Petrol. 36, 1647-1674. Gast, P.W., 1968. Trace element fractionation and the origin of tholeiitic and alkaline magma types. Geochim. Cosmochim. Acta 32, 1057-1086. Gill, J.B., 1993. Melts and metasomatic fluids: evidence from U-series disequilibria and Th isotopes. Philos. Trans. R. Soc. London A342, 79-90. Gill, J., Condomines, M., 1992. Short-lived radioactivity and magma genesis. Science 257, 1368-1376. Gill, J.B., Williams, R.W., 1990. Th isotope and U-series studies of subduction-related volcanic rocks. Geochim. Cosmochim. Acta 54, 1427-1442. Goldstein, S.J., Murrel, M.T., Janecky, D.R., 1989. Th and U isotopic systematics of basalts from the Juan de Fuca and Gorda Ridges by mass spectrometry. Earth Planet. Sci. Lett. 96, 134-146. Goldstein, S.J., Murreil, M.T., Janecky, D.R., Delaney, J.R., Clague, D.A., 1991. Geochronology and petrogenesis of MORB from the Juan da Fuca and Gorda ridges by 23gU - 230Th disequilibrium. Earth Planet. Sci. Lett. 107, 25-41. Goldstein, S.J., Murrell, M.T., Williams, R.W., 1993. 231Pa and 23°Th chronology of mid-ocean ridge basalts. Earth Planet. Sci. Lett. 115, 151-159. Green, D.H., Ringwood, A.E., 1967. The genesis of basaltic magmas. Contrib. Mineral. Petrol. 15, 103-190. Halliday, A.N., Dicldn, A.P., Fallick, A.E., Fitton, J.G., 1988. Mantle dynamics: a Nd, Sr, Pb and O isotopic study of the Cameroon Line volcanic chain. J. Petrol. 29, 181-211. Halliday, A.N., Davidson, J.P., Holden, P., DeWolf, C., Lee,

182

T. Elliott / Chemical Geology 139 (1997) 165-183

D.-C., Fitton, J.G., 1990. Trace-element fractionation in plumes and the origin of HIMU mantle beneath the Cameroon line. Nature 347, 523-528. Hart, S.R., 1988. Heterogeneous mantle domains: signatures, genesis and mixing chronologies. Earth Planet. Sci. Lett. 90, 273-296. Hauri, E.H., Shimizu, N., Dieu, J.J., Hart, S.R., 1993. Evidence of hot-spot related carbonatite metasomatism in the oceanic upper mantle. Nature 365, 221-227. Hauri, E.K., Wagner, T.P., Grove, T.L., 1994. Experimental and natural partitioning of Th, U, Pb and other trace elements between garnet, clinopyroxene and basaltic melts. Chem. Geol. 117, 149-166. Hauri, E.R., Hart, S.R., 1993. Re-Os isotope systematics of HIMU and EMII oceanic island basalts from the South Pacific Ocean. Earth Planet. Sci. Lett. 114, 353-371. Htmond, C., Condomines, M., Foucade, S., All~gre, C.J., Oskarsson, N., Javoy, M., 1988. Thorium, strontium and oxygen isotopic geochemistry in recent tholeiites from Iceland: crustal influence on mantle-derived magmas. Earth Planet. Sci. Lett. 87, 273-285. H~mond, C., Amdt, N.T., Lichtenstein, U:, Hofmann, A., 1993. The heterogeneous Iceland Plume: Nd-Sr-O isotopes and trace element constraints. J. Geophys. Res. 98, 16833-16850. H~mond, C., Devey, D.W., Chauvel, C., 1994a. Source compositions and melting processes in the Society and Austral plumes (South Pacific Ocean): element and isotope (Sr, Nd, Pb, Th) geochemistry. Chem. Geol. 115, 7-45. H~mond, C., Hofmann, A.W., Heusser, G., Condomines, M., Raczek, I., Rhodes, J.M., 1994b. U-Th-Ra systematics in Kilauea and Manna Loa basalts, Hawaii. Chem. Geol. 116, 163-180. Hirschmann, M.M., Stolper, E.M., 1996. A possible role for garnet pyroxenite in the origin of the 'garnet signature' in MORB. Contrib. Mineral. Petrol. 124, 185-208. Hofmann, A.W., Hart, S.R., 1978. An assessment of local and regional isotopic equilibrium in the mantle. Earth Planet. Sci. Lett. 38, 44-62. Hofmann, A., White, W.M., 1982. Mantle plumes from ancient oceanic crust. Earth Planet Sci. Lett. 57, 421-436. Ivanovich, M., Harmon, R.S., 1992. Uranium-Series Disequilibrium, Applications to Earth, Marine and Environmental Sciences, 2nd ed. Oxford University Press, New York, 912 pp. Iwamori, H., 1994. 23Su-23°Th-226Ra and 235U-231pa disequilibria produced by mantle melting with porous and channel flows. Earth Planet. Sci. Lett. 125, 1-16. Klein, E.M., Langmuir, C.H., 1987, Global correlations of ocean ridge basalt chemistry and axial depth and crustal thickness. J. Geophys. Res. 92, 8089-8115. Langrnuir, C.L., Bender, J.F., Bence, A.E., Hanson, G.N., Taylor, S.R., 1977. Petrogenesis of basalts from the FAMOUS area: mid-Atlantic ridge. Earth Planet. Sci. Lett. 36, 133-156. LaTourette, T.Z., Burnett, D.S., 1992. Experimental determination of U and Th partitioning between clinopyroxene and natural and synthetic basaltic liquid. Earth Planet. Sci. Lett. 110, 227-244. LaTourette, T.Z., Kennedy, A.K., Wasserburg, G.J., 1993. Ura-

nium-thorium fractionation by garnet--evidence for a deep source and rapid rise of oceanic basalts. Science 261,739-742. Lundstrom, C.C., Shaw, H.F., Ryerson, F.J., Phinney, D.L., Gill, J.B., Williams, Q., 1994. Compositional controls on the partitioning of U, Th, Ba, Pb, Sr and Zr between clinopyroxene and haplobasaltic melts: implications for uranium series disequilibria in basalts. Earth Planet. Sci. Lett. 128, 407-423. Lundstrom, C.C., Gill, J., Williams, Q., Perfit, M.R., 1995. Mantle melting and basalt extraction by equilibrium porous flow. Science 270, 1958-1961. MacDougall, J.D., 1995. Using short-lived U and Th series nuclides to investigate volcanic processes. Annu. Rev. Earth Planet. Sci. 23, 143-167. Marcantonio, F., Zindler, A., Elliott, T., Staudigel, H., 1995. Os isotope systematics of La Paima, Canary Islands: evidence for recycled crust in the mantle source of HIMU ocean islands. Earth Planet. Sci. Lett. 133, 397-410. McDermott, F.M., Hawkesworth, C., 1991. Th, Pb, and Sr isotope variations in young island arc volcanics and oceanic sediments. Earth Planet. Sci. Lett. 104, 1-15. McKenzie, D., 1985a. 23°Th-23Su disequilibrium and the melting process beneath ridge axes. Earth Planet. Sci. Left. 72, 149157. McKenzie, D., 1985b. The extraction of magma from the crust and mantle. Earth Planet. Sci. Lett. 74, 81-91. McKenzie, D,, Bickle, M.J., 1988. The volume and composition of melt generated by extension of the lithosphere. J. Petrol. 29, 625-679. McKenzie, D., O'Nions, R.K., 1991. Partial melt distributions from inversion of rare earth element concentrations. J. Petrol. 32, 1021-1091. Meadows, J.W., Armani, R.J., Callis, E.L., Essling, A,M., 1980. Half-life of 230Th. Phys. Rev. C22, 750-754. Minster, J.F., All~gre, C.J., 1978. Systematic use of trace elements in igneous processes, Part III. Inverse problem of batch melting in volcanic suites. Contrib. Mineral. Petrol. 68, 37-52. Newman, S., Finkel, R.C., MacDougall, J.D., 1983. 23°Th-23Su disequilibrium systematics in oceanic tholeiites from 21°N on the East Pacific rise. Earth Planet. Sci. Lett. 65, 17-33. Newman, S., Finkel, R.C., MacDougall, J.D., 1984a. Comparison of 23°Th-23Su disequilibrium systematics in lavas from three hot spot regions: Hawaii, Prince Edward and Samoa. Geochim. Cosmochim. Acta 48, 315-324. Newman, S., MacDougall, J.])., Finkel, F.C., 1984b. 23°Th-23Su disequilibrium in island arcs: evidence from the Aleutians and the Marianas. Nature 308, 268-270. Newsom, H.E., White, W.M., Jochum, K.P., Hofmann, A.W., 1986. Siderophile and chalcophile element abundances in oceanic basalts, Pb isotope evolution and growth of the Earth's core. Earth Planet. Sci. Lett. 80, 299-313. Nicholson, H., Condomines, M., Fitton, J.G., 1991. Geochemical and isotopic evidence for crustal assimilation beneath Krafla, Iceland. J. Petrol. 32, 1005-1050. O'Nions, R.K., McKenzie, D., 1993. Estimates of mantle thorium/uranium ratios from Th, U and Pb isotopic abundances in basaltic melts. Philos. Trans. R. Soc. London A342, 65-77. Oversby, V.M., Gast, P.W., 1968. Lead isotopic compositions and

T. Elliott / Chemical Geology 139 (1997) 165-183 uranium decay series disequilibrium in recent volcanic rocks. Earth Planet. Sci. Lett. 5, 199-206. Plank, T., Langmuir, C.tl., 1992. Effects of the melting regime on the composition of ~the oceanic crest. J. Geophys. Res. 97, 19749-19770. Prinzhofer, A., Lewin, E., All~gre, C.J., 1989. Stochastic melting of the marble cake mantle: evidence from local study of the East Pacific Rise. Earth Planet. Sci. Lett. 92, 189-206. Pyle, D.M., Dawson, J.B., Ivanovich, M., 1991. Short-lived decay series disequilibria in the natrocarbonatite lavas of Oldoinyo Lengai, Tanzania: constraints on the timing of magma genesis. Earth Planet. Sci. LeL~. 105, 378--396. Qin, Z., 1992. Disequilibrium partial melting model and its implications for trace element fractionations during mantle melting. Earth Planet. Sci. Lett. 112, 75-90. Qin, Z,, 1993. Dynamics of melt generation beneath mid-ocean ridge axes: theoretical analysis based on 238U - 230Th- 226Ra and 231Pa- 23~U. Gec,chim. Cosmochim. Acta 57, 1629-1634. Reinitz, I., Turekian, K., 1989. (230Th/238U) and (226Ra/230Th) fractionation in young basaltic glasses from the East Pacific Rise. Earth Planet. Sci. Lett. 94, 199-207. Reinitz, I.M., Turekian, K.K., 1991. The behavior of the uranium decay chain nuclides and thorium during the flank eruptions of Kilanea (Hawaii) between 1983 and 1985. Geochim. Cosmochim. Acta 55, 3735-3740. Reisberg, L., Zindler, A., Mareantonio, F., White, W., Wyman, D.A., Weaver, B., 11993. Os isotope systematics in ocean island basalts. Earth Planet. Sci. Lett. 120, 149-167. Richardson, C., McKen~,ie, D., 1994. Radioactive disequilibria from 2D models of melt generation by plumes and ridges. Earth Planet. Sci. Lett. 128, 425-437. Rubin, K.H., MacDougall, J.D., 1988. 226Ra excesses in mid-ocean ridge basalts and mantle melting. Nature 355, 158-161. Rubin, K.H., MacDoug,'dl, J.D., 1989. Submarine magma degassing and explosive magmatism at MacDonald (Tamaxii) seamount. Nature 341, 50-52. Rubin, K.H., MacDougall, J.D., 1990. Dating of neovolcanic MORB using (226Ra/23°Th) disequilibrium. Earth Planet. Sci. Lett. 101,313-322. Rubin, K.H., MacDougaU, J.D., 1992. Th-Sr isotopic relationships in MORB. Earth Planet. Sci. Lett. 114, 149-157. Salters, V.J.M., Hart, S.R., 1989. The hafnium paradox and the role of garnet in the source of mid-ocean ridge basalts. Nature 342, 420-422. Salters, V.J.M., Longhi, J., 1996. Partitioning of trace elements during primary melting of MORB mantle. J. Conf. Abstr. 1, 529. Schilling, J.-G., Bergeron, M.B., Evans, R., 1980. Halogens in the mantle beneath the N. Atlantic. Philos. Trans. R. Soc. London Ser. A297, 147-178. Shaw, D.M., 1970. Trace element fractionation during anataxis. Geochim. Cosmochirn. Acta 34, 237-243. Sigmarsson, O., Condornines, M., Fourcade, S., 1992a. Mantle and crustal contribution in the genesis of recent basalts from off-rift zones in Iceland: constraints from Th, Sr and O isotopes. Earth Planet Sci. Lett. 110, 149-162.

183

Sigmarsson, O., Condomines, M., Ibarrola, E., 1992b. 238U-23°Zh radioactive disequilibria in historic lavas from the Canary Islands and genetic implications. J. Volcanol. Geotberm. Res. 54, 145-156. Sims, K.W.W., DePaolo, D.J., Murrell, M.T., Baldridge, W.S., Goldstein, S.J., Clague, D.A., 1995. Mechanisms of magma generation beneath Hawaii and mid-ocean ridges: uranium/thorium and samarium/neodymium isotopic evidence. Science 267, 508-512. Sleep, N.H., 1990. Hotspots and mantle plumes: some phenomenology. J. Geophys. Res. 95, 6715-6736. Spiegelman, M., EUiott, T., 1993. Consequences of melt transport for uranium series disequilibrium in young lavas. Earth Planet. Sci. Lett. 118, 1-20. Sun, S.-S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: A.D. Sannders and M.J. Norry (Editors), Magmatism in the Ocean Basins. Geol. Soc. London, Spec. Publ. 42, 313-345. Volpe, A.M., Goldstein, S.J., 1993. 226Ra-E3°Th disequilibrium in axial and off-axis mid-ocean ridge basalts. Geochim. Cosmochim. Acta 57, 1233-1241. Walker, D., Jones, J.H., 1991. Experimental partitioning of Nb, Mo, Ba, Co, Pb, Th and U between carbonate and silicate liquid. EOS, Trans. Am. Gemphys. Union 72, 574. Walker, D., Beattie, P., Jones, J.H., 1992. Partitioning of U - T h - P b and other incompatibles between angite and carbonate liquid at 1200°C and 55 kbar. EOS, Trans. Am. Geophys. Union 73, 616. Weaver, B.L., 1991. The origin of ocean island basalt end member compositions. Earth Planet. Sci. Lett. 104, 381-397. White, W.M., 1985. Sources of oceanic basalts: radiogenic isotope evidence. Geology 13, 115-118. White, W.M., Hofmann, A.W., 1982. Sr and Nd isotope geochemistry of oceanic basalts and mantle evolution. Nature 296, 821-825. Williams, R.W., Gill, J.B., 1989. Effects of partial melting on the uranium decay series. Geochim. Cosmochim. Acta 53, 16071619. Williams, R.W., Gill, J.B., 1992. Th isotope and U-series disequilibria in some alkali basalts. Geophys. Res. Lett. 19, 139-142. Williams, R.W., Gill, J.B., Bruland, K.W., 1986. Ra-Th disequilibria systematics: timescale of carbonatite magma formation at Oldoinyo Lengai volcano, Tanzania. Geochim. Cosmochim. Acta 53, 1607-1619. Wright, E. and White, W.M., 1986/1987. The origin of Samoa: New evidence from St, Nd, and Pb isotopes. Earth Planet. Sci. Lett., 81: 151-162. Zindler, A., Hart, S., 1986. Chemical geodynamics. Annu. Rev. Earth Planet. Sci. 14, 493-571. Zindler, A., Jagoutz, E., Goldstein, S., 1982. Nd, Sr and Pb isotopic systematics in a three-component mantle: a new perspective. Nature 298, 519-523.