The dynamics of melting beneath Theistareykir, northern Iceland



[1] U-Th disequilibrium data on a suite of postglacial basalts (<12,000 years old) from Theistareykir, northern Iceland, show a relatively high degree of U series disequilibrium. For 21 out of 27 samples from Theistareykir, 230Th excesses are within 25 ± 5%. Four samples have 230Th excesses >30%. The two geochemically and isotopically most depleted samples (picrites) have excess 230Th of about 15%, similar to enriched basalts from locations adjacent to Theistareykir (Draugarhraun and Asbyrgi). The large range in chemical and isotopic variations of the Theistareykir samples provides an excellent opportunity to investigate the influence of source heterogeneity on the degree of U series disequilibrium. The lack of correlation between U-Th data and radiogenic isotope or chemical parameters indicates that source heterogeneity does not play an important role in creating the excess 230Th. Therefore the Theistareykir basalts provide evidence that the Th systematics result from variations in the timing of the melting and melt extraction process only. The relatively large and constant 230Th excesses for the bulk of the Theistareykir samples (20–30%) suggest that melting and melt extraction must be relatively reproducible, with the melting starting well within the garnet-stability field. Near fractional melting models suggest residual porosity ≤0.1%, upwelling rates ≤1cm/yr (corresponding to melting rates ≤1*10−4 kg/m3yr), and melt ascent velocities ≥3–4m/yr. The lower 230Th excesses of about 15% in both the most depleted lavas from Theistareykir and enriched lavas from locations adjacent to Theistareykir (Draugarhraun and Asbyrgi) indicate some variation in the melting and melt extraction process. The low 230Th excesses of the most depleted melts (picrites) could either be an effect of slow melt migration or be caused by incomplete melt aggregation, with melts created at the very bottom of the melting regime being underrepresented. The low 230Th excesses of the most enriched lavas are most likely due to slow melt migration for melts created at some distance from the center of the melting regime.

1. Introduction

[2] U series isotope systematics in basalts can potentially provide insights into the timescales and dynamics of melting processes in the Earth's mantle [e.g., McKenzie, 1985; Spiegelman and Elliott, 1993]. Unlike other geochemical tracers, the initial 230Th/238U of the mantle source is known because the short-lived decay products of the 238U and 235U decay chain are in secular equilibrium in mantle sources of basalts that have not been melted or infiltrated by melt within the last ∼0.5m.y. During melting, disequilibrium is generated by fractionation of U from its daughter isotopes. The extent of disequilibrium depends on melting parameters such as residual porosity (the amount of melt in contact with the solid residue), melting time (the time period from the onset of melting until the final degree of melting is reached), upwelling rates of both solid and liquid, and the bulk solid-melt partition coefficients (D's) [e.g., McKenzie, 1985; Salters and Longhi, 1999; Spiegelman and Elliott, 1993]. Although recent studies suggest that Th is slightly more incompatible in clinopyroxene (cpx) than U in the spinel stability field [e.g., Landwehr et al., 2001; Salters and Longhi, 1999; Wood et al., 1999], in peridotite garnet remains the only common mantle mineral in which Th is significantly more incompatible than U [Beattie, 1993a, 1993b; Hauri et al., 1994; LaTourrette and Burnett, 1992; LaTourrette et al., 1993; Salters and Longhi, 1999]. Therefore the amount of melting in the garnet-stability field is thought to be an important factor in controlling U-Th systematics, and can so serve as an indicator of the initial depth of melting [e.g., Bourdon et al., 1996a, 1996b].

[3] Consequently, 230Th excesses commonly observed in mid ocean ridge basalts (MORB) have been attributed to variations in the style of melting [Bourdon et al., 1996a, 1996b; Goldstein et al., 1989, 1993, 1994; Lundstrom et al., 1998a, 1999; Newman et al., 1983; Rubin and MacDougall, 1992; Sims et al., 2002]. Some recent studies suggest a potential role for source heterogeneity on the amount of disequilibrium generated during melting and cast some doubt on the suitability of U-Th studies to reliably convey information about the melting dynamics [Bourdon et al., 1996b; Lundstrom et al., 1995, 1998a, 1998b, 1999]. Additional uncertainty in interpreting U-Th systematics in MORB comes from the poor age control of dredged MORB samples, resulting in a potentially significant uncertainty of measured (230Th/238U) (where parentheses denote activity ratios) [Bourdon et al., 1996b], and from contamination with Mn crusts or sediments [e.g., Bourdon et al., 2000; Reinitz and Turekian, 1989]. Moreover, previous U-Th studies of MORB (and OIB) have rarely investigated the effects of source heterogeneity on U-Th systematics using comprehensive data sets with combined U series and several radiogenic isotope ratios (e.g., Sr, Nd, Pb, or Hf isotope ratios), which are the only unambiguous tracer of source heterogeneity. An exception are the recent studies by Stracke et al. [1999] and Sims et al. [1999, 2002]. Therefore deconvolution of the effects of source heterogeneity and depth and/or the time of melting on (230Th/238U) needs to be investigated in a setting that is free of age uncertainty, but also offers the opportunity to distinguish between the effects of source heterogeneity (shown by Sr, Nd, Pb, or Hf isotope ratios) and depth of melting on (230Th/238U).

[4] Here, we present U-Th disequilibrium data on postglacial samples from Theistareykir, northern Iceland (age between ≈3000 and 12,000 years) [Slater, 1996; Slater et al., 2001]. Such postglacial flows are easily distinguished from sub-glacial eruptions by their lack of glacial striae and because they flow freely across the landscape and do not form table mountains. Along with the U-Th data presented here, major and trace element concentrations and Sr, Nd, Hf, and Pb isotope data have been reported on the same samples [Slater, 1996; Stracke et al., 2003]. Although Iceland is clearly not a “normal” MORB setting, it is the only locality where a mature ridge system is accessible above sea level. Melting beneath Theistareykir is dominated by melting of depleted mantle [Stracke et al., 2003] and is expected to occur largely as a result of passive mantle upwelling in response to sear-floor spreading with a total spreading rate of 18mm/yr [e.g., Darbyshire et al., 2000a; Maclennan et al., 2002a]. Thus melting beneath Theistareykir should be a good analog to melting beneath “normal” ridges. However, due to the higher temperatures in the sub-Icelandic mantle compared to the adjacent Mid-Atlantic Ridge (MAR, ΔT about 200°C, [Schilling, 1991; Sleep, 1990]), melting at Iceland is expected to start deeper and continue over a greater range in pressure than at the adjacent submerged portions of the MAR. Moreover, the large isotopic variability in combination with the large extents of melting observed in the Theistareykir basalts are ideally suited to investigate the effects of source composition and depth of melting on (230Th/238U), and can potentially offer insight into the dynamics and timescales of melting beneath Iceland.

2. Analytical Techniques

[5] The chemical and mass-spectrometric techniques applied for U and Th have been described in detail by Bourdon [1994] but are described briefly below. Between 80mg and 2g of sample were dissolved in order to yield about 30–40ng of Th for the mass-spectrometric analysis. The samples were spiked with 229Th and 233U and dissolved in a mix of HF-HClO4-HNO3-HCl. Initial separation from Th and U was achieved using 2ml cation exchange columns. The U fraction is purified by processing over 400μl and 100μl anion exchange columns, the Th fraction is purified by processing over 100μl anion exchange columns. Observed total chemistry blanks were ∼5pg for Th and ∼7pg for U, and therefore negligible compared to the analyzed amount of sample.

[6] Th isotope ratios and Th concentrations were determined using the ISOLAB 54 at the NHMFL by SIMS (Secondary Ionization Mass Spectrometry). The instrument and measurement techniques have been described in detail by England et al. [1992] and Bourdon [1994], and similar SIMS techniques with comparable analytical precision have later been applied on other instruments [Layne and Sims, 2000]. Th was loaded as Nitrate with colloidal graphite on pyrolitically coated graphite rods using a semi-automatic loading device as described by Bourdon [1994]. Th was analyzed by successive static measurements of 230Th/232Th and 229Th/232Th. The low abundance isotopes 229Th and 230Th ion beams were measured by a Daly ion counting system after a second stage energy filtering to increase abundance sensitivity, and the 232Th ion beam was collected before the second stage energy filtering in Faraday cups. At the beginning and end of each analysis, the gain calibration factor between the Daly and Faraday collectors was determined by switching the 232Th ion beam between the Daly knob and Faraday collector. The 230Th signal is corrected for interfering 229Th1H by using the measured 232Th1H/232Th ratio (232Th1H/232Th ratios are typically ∼0.8%). A linear interpolation is used to correct 230Th for the 232Th tail by measuring the background on both sides of the 230Th and 229Th peaks. Repeated measurements (n = 33) of about 40ng size loads of the WUN Th standard (WUN = WHOI, UCLA, NHMFL) yielded 230Th/232Th = 4.296*10−6 ± 0.046 (2σ; see Figure 1). 230Th/232Th ratios in the samples reproduced to about 1–2% on the basis of replicate measurements (Table 1).

Figure 1.

Repeated measurements of about 40ng size loads of the WUN Th standard (n = 33). The average value is: 230Th/232Th = 4.296*10−6 ± 0.046 (2σ).

Table 1. U-Th Mass Spectrometry Data for the Theistareykir Samplesa
 Th, ppmU, ppm(238U/232Th)(230Th/232Th)±2(σ)m(230Th/238U)(234U/238U)±2(σ)m
  • a

    Th concentrations and Th isotope ratios are measured by SIMS; U concentrations and isotope ratios are measured by TIMS. Calculations of activity ratios assume λ230 = 9.1577*10−6, λ232 = 4.9475*10−11, λ238 = 1.55125*10*−10, λ234 = 2.8263*10−6 [Cheng et al., 2000]. Errors are estimated on the basis of replicate sample analysis and are <1–3% for U concentrations and <2–3% for Th concentrations. 230Th/232Th and 234U/238U ratios reproduced to about 1–2%. Carbonatite BD144 from Oldoinyo Lengai has also been analyzed. (230Th/232Th) and (238U/232Th) of BD144 are within 1% of the values from Williams et al. [1986] ((230Th/232Th) = 1.007 compared to 1.011 and (238U/232Th) = 11.45 compared to 11.3).

Storaviti (12,000–10,500 yr BP)
93300.16470.05130.9451.128 1.1941.0160.005
Borgarhraun (10,500–7,000 yr BP)
  0.0227    1.0200.007
  0.0216    1.0310.010
Langavitihraun (10,500–7,000 yr BP)
 0.22750.06930.9251.172 1.2681.0190.009
  0.0706    1.0150.006
  0.0788    1.0030.006
  0.0209    1.0130.013
Theistareykir (3,000–2,700 yr BP)
  0.0573    1.0220.010
Bondholshraun (10,500–7,000 yr BP)
Arnahvammurhraun (12,000–10,500 yr BP)
Hoefuheidharmuli (<12,000 yr BP)
Picrites (12,000–10,500 yr BP)
Samples from Locations Adjacent to Theistareykir Asbyrgi (NE of Theistareykir) (<12,000 yr BP)
  0.0879    1.0240.010
  0.0892    1.0140.009
93230.21030.07121.0271.187 1.1561.0280.007
Draugarhraun (Krafla) (7,000–2,800 yr BP)

[7] U isotope ratios were analyzed on the Finnigan MAT 262 RPQ at the NHMFL (TIMS = Thermal Ionization Mass Spectrometry). U was loaded with H3PO4 and 50μg of colloidal graphite on single Re filaments, and measured by a series of static measurements. 233U, 234U and 235U were measured on an ion counting SEM located behind the RPQ system and 238U was collected on Faraday cups. The gain factor between the ion-counter and the Faraday collectors was determined before and after each measurement by switching the 238U beam between the Faraday collector and the SEM. Due to a software problem, no background correction was employed on the SEM (i.e., for 233U, 234U and 235U; 238U measured on the faraday collectors are background corrected). Typical background values on the SEM are < 1cps. Thus for typical measured intensities of the 233U and 235U beam (5,000–10,000cps), the background correction amounts to <0.05‰, which has a negligible effect on the measured 233U/238U and 235U/238U ratios. U concentrations are calculated using the 233U/238U and 235U/238U ratios. Therefore the lack of background correction on the SEM does not influence the calculated U concentrations. Due to the smaller 234U beam size (about 100cps), however, the lack of a background correction has a significant effect on the measured 234U/238U ratios. Despite this analytical problem, measured (234U/238U) are reported in Table 1, and on the basis of replicate measurements, (234U/238U) reproduced to within 1–2% (Table 1). One has to bear in mind, however, that depending on the 234U count rates relative to the background, actual background corrected (234U/238U) are expected to be up to 1–2% lower than reported values. Thus the Theistareykir samples are considered to be within error of secular equilibrium ((234U/238U) = 1).

[8] Reproducibility of U and Th concentrations is estimated based entirely on replicate measurements of separate powder aliquots, and thus includes instrumental errors and errors originating from sample heterogeneity. On the basis of replicate measurements, Th concentrations reproduced to <3%, while U concentrations generally reproduced to about 1% (differences in some duplicates, however, can be up to 3%), and 238U/232Th reproduced to <3% (Table 1). Reproducibility of the U and Th concentrations is independent of U and Th concentration. Reproducibility for the samples from Draugarhraun based on duplicate measurements (n = 3 for 9366 and n = 4 for 9396, not reported) is <3% for both Th and U concentrations and therefore similar to the reproducibility of samples with lower concentrations (see above).

3. Results

[9] The Theistareykir basalts display a total range of (230Th/238U) from 1.14 to 1.37, with (230Th/232Th) and (238U/232Th) ranging between 1.05 and 1.34 and 0.88 and 1.03, respectively (Figure 2a). 21 out of 27 analyzed samples from Theistareykir have 230Th excesses between 20 and 30%, with (238U/232Th) ranging from 0.86 to 0.99, and (230Th/232Th) between 1.09 and 1.25. Two samples from Bogarhraun and Langavitihraun have 230Th excesses greater than 30%. The two analyzed picrites from Theistareykir have 230Th excesses of about 15% and (238U/232Th) of about 1. Samples from localities adjacent to Theistareykir (Draugarhraun and Asbyrgi) also have (230Th/238U) <1.2, but compared to the picrites (238U/232Th) are lower in the samples from Draugarhraun and higher in the samples from Asbyrgi (Figure 2a).

Figure 2.

Plots of (230Th/232Th) versus (238U/232Th) of the Theistareykir basalts compared to a) other localities in Iceland and basalts from the Mid-Atlantic Ridge (MAR) between 37° and 40°N [Bourdon et al., 1996a] and b) MORB from the Atlantic and Pacific ridges and zero-age basalts from Hawaii. In Figure 2b, only mass spectrometric data are plotted, whereas due to the lack of mass spectrometric U series data on Icelandic volcanics, α-counting data are included in Figure 2a. See text for details. Data sources: Figure 2a: Iceland: α-counting data for the Reykjanes Peninsula, the northern and southeastern volcanic zone (NEVZ, SEVZ) [Hemond et al., 1988; Sigmarsson et al., 1992b], Krafla [Nicholson et al., 1991], Hekla and the Snaefelssness Peninsula [Sigmarsson et al., 1992a, 1992b]. Mass spectrometric data for the Reykjanes Peninsula and Ridge [Peate et al., 2001], and MAR 37–40°N [Bourdon et al., 1996a]. Figure 2b: Mid-Atlantic Ridge (MAR): MAR 37–40°N [Bourdon et al., 1996a], MAR at 33°S [Lundstrom et al., 1998a], MARK area at 23°N [Sturm et al., 2000], MAR 29–30°N [Bourdon et al., 1996b]. East Pacific Rise (EPR): Tamayo region [Bourdon et al., 1993], EPR 9–10°N [Goldstein et al., 1989, 1991, 1993, 1994], Siqueiros Transform and Lamont seamounts [Lundstrom et al., 1999]. Hawaiian basalts: [Cohen and O'Nions, 1993; Cohen et al., 1996; Sims et al., 1995, 1999]. Australian Antarctic Discordance (AAD) [Bourdon et al., 1996b]. Reykjanes Ridge and Reykjanes Peninsula [Peate et al., 2001].

Figure 2.


[10] The Theistareykir basalts lie well within the range of existing U-Th disequilibrium data on Icelandic volcanics, which have been determined almost exclusively by α-counting [Condomines et al., 1981; Hemond et al., 1988; Krishnaswami et al., 1984; Nicholson et al., 1991; Sigmarsson et al., 1991, 1992a, 1992b] (Figure 2a). Only two of the previous α-counting studies on Icelandic lavas include data on relatively primitive Icelandic basalts [Condomines et al., 1981; Hemond et al., 1988], most of which have 230Th excesses similar to those of the Theistareykir basalts (Figure 2a). Recently, Peate et al. [2001] reported three mass-spectrometric analyses of basalts from the Reykjanes Peninsula, two of which are very similar to the picrites from Theistareykir, whereas one of the samples has significantly higher (238U/232Th) and a 230Th excess <10% (Figure 2a).

[11] Comparing the Theistareykir data with other mass spectrometric U series data shows that the Theistareykir data overlap with data from the Mid-Atlantic Ridge (MAR) at 37°–40°N [Bourdon et al., 1996a], whereas samples from the MAR at 33°S have similar (238U/232Th) but lower (230Th/232Th) (and also and higher U and Th concentrations) [Lundstrom et al., 1998a]. MORB from the East Pacific Rise (EPR) generally have (238U/232Th) greater than 1.1 and 230Th excesses <15% [Bourdon et al., 1993; Goldstein et al., 1994; Lundstrom et al., 1999; Sims et al., 2002]. Samples from the MAR, including Iceland, generally have (238U/232Th) less than 1.1 and 230Th excesses ≥15% [Bourdon et al., 1996a; Lundstrom et al., 1998a]. An exception among the MAR samples are the samples from the MARK area at 23°N [Sturm et al., 2000], which have U-Th characteristics similar to the EPR-MORB (Figure 2b).

[12] Compared to Hawaiian basalts [Cohen and O'Nions, 1993; Cohen et al., 1996; Pietruszka et al., 2001; Sims et al., 1995, 1999], the Theistareykir samples have a similar range in (238U/232Th) but generally have higher 230Th excesses; only the alkali-basalts from the rejuvenated stage of volcanism at Hawaii have 230Th excesses greater than 20%, similar to the Theistareykir basalts [Sims et al., 1995, 1999] (Figure 2b).

[13] (230Th/238U), (230Th/232Th), and (238U/232Th) in the Theistareykir basalts as a whole or for individual flows alone show no clear correlations with other chemical or isotopic parameters (e.g., major and trace element concentrations and ratios, and Sr, Nd, Hf, and Pb isotope ratios, Figures 3a–3c, Table 2). The isotopically most depleted (picrites) and enriched samples from adjacent localities (Draugarhraun and Asbyrgi) have similar 230Th excesses; the majority of the samples with (230Th/238U) between 1.2 and 1.3 show no systematic trend with chemical or isotopic enrichment (Table 1, Figures 3a–3c).

Figure 3.

Plots of a) δ18Oolivine, b) 143Nd/144Nd, and c) (La/Sm)N versus (230Th/238U) in the Theistareykir melts (enriched samples from Asbyrgi and Draugarhraun are not plotted). The lack of correlation between δ18Oolivine and (230Th/238U) confirms that the Theistareykir basalts are not affected by interaction with the preexisting Icelandic crust [see Stracke et al., 2003]. (230Th/238U) show no systematic trend with chemical or isotopic enrichment (143Nd/144Nd, (La/Sm)N).

Table 2. Correlation Coefficient r2 for Correlations Between (230Th/232Th), (238U/232Th), and (230Th/238U) and Major and Trace Element Concentrations and Ratiosa
  • a

    For perfect linear correlations, r2 = 1, with the degree of scatter increasing as r2 approaches 0. All r2 are calculated excluding the samples from Draugarhraun, as these belong to the Krafla volcanic system. r2 is given in absolute values, i.e., without a + or − sign indicating a positive or negative correlation.


4. Discussion

4.1. U-Th Disequilibrium as a Tracer of Crustal Assimilation

[14] Th isotope systematics in Icelandic volcanics have been proposed as a tracer for bulk assimilation of the Icelandic crust via assimilation-fractional crystallization processes (AFC) [e.g., Condomines et al., 1981; Nicholson et al., 1991; Sigmarsson et al., 1991, 1992a, 1992b]. These authors argue that near-equilibrium values of (230Th/238U) in highly evolved lavas with low δ18O (<5) result from assimilation of (or mixing with melts derived from) altered Icelandic crust, containing U and Th in secular equilibrium. We showed previously that the trace element and isotopic characteristics of the Theistareykir basalts are not significantly perturbed by bulk assimilation of the preexisting crust (see Stracke et al. [2003] for a detailed discussion). The U-Th systematics further affirm this conclusion as the most enriched and depleted lavas display similar degrees of disequilibrium, and the bulk of the samples show no trend of increasing “enrichment” (isotopic or with respect to major and trace elements) with decreasing 230Th excesses or oxygen isotope composition [Eiler et al., 2000] (Figures 3a–3c). Moreover, (234U/238U) values close to equilibrium (Table 1) provide no clear indication that significant posteruptive alteration or assimilation of hydrothermally altered material has occurred. It follows that the Th isotope systematics in the Theistareykir basalts reflect the melting process and are not affected by shallow-level interaction with the Icelandic crust.

4.2. Effect of Source Heterogeneity on (230Th/238U)

[15] Deriving constraints on melting dynamics using simple quantitative models requires a mineralogically homogeneous source [e.g., McKenzie, 1985; Spiegelman and Elliott, 1993; Stracke et al., 1999]. The range of Sr, Nd, Pb, and Hf isotope ratios in the Theistareykir samples require source heterogeneity [Stracke et al., 2003]. Correlations between isotope ratios, and major and trace element abundances and ratios indicate that melting samples an isotopically and compositionally heterogeneous Icelandic mantle. This behavior is most readily explained if the enriched portions of the Icelandic source have a lower solidus temperature than the associated depleted material. Thus the isotopic range in the Theistareykir source might be linked to mineralogical differences between the isotopically different sources (see Stracke et al. [2003] for a detailed discussion). How would melting of mineralogically different components of the source like garnet-pyroxenite or eclogite influence the chemical and isotopic systematics of the erupted melts?

[16] Melting of garnet-pyroxenite or eclogite is not expected to produce large 230Th excesses [Stracke et al., 1999]. This is because U and Th are not strongly fractionated during melting of eclogite or garnet-pyroxenite, owing to both the compositional dependence of the partition coefficients [e.g., Salters and Longhi, 1999; Salters et al., 2002] as well as the dominance of the clinopyroxene partition coefficient on the bulk partition coefficients. This is caused by both the higher absolute value of DU,Th clinopyroxene compared to DTh,U garnet, as well as by the large abundance of clinopyroxene in these rocks. Because bulk DU/DTh ratios in garnet-pyroxenites and eclogites are smaller than those in garnet-peridotites, melting of eclogite or garnet-pyroxenite leads to less excess 230Th for a given set of physical parameters (e.g., porosity, melting rate and melting time). Melting of a mixed eclogite (garnet-pyroxenite) and garnet-peridotite assemblage is expected to produce trends of decreasing (230Th/238U) with increasing Th/U, 87Sr/86Sr, 206Pb/204Pb and decreasing 143Nd/144Nd and 176Hf/177Hf [Stracke et al., 1999] (see Hirschmann and Stolper [1996], Pertermann and Hirschmann [2003], and Stracke et al. [1999] for detailed discussions on the melting characteristics of pyroxenite/eclogite and peridotite). Consequently, the large 230Th excess associated with high Th/U ratios in MORB cannot be explained by melting garnet-pyroxenite or eclogite as has been previously suggested by Bourdon et al. [1996b] and Lundstrom et al. [1999].

[17] The lack of any significant correlation of (230Th/238U), (230Th/232Th), and (238U/232Th) with radiogenic isotopes in the Theistareykir basalts suggests that compositional heterogeneity is not related to mineralogical heterogeneities in the Theistareykir source. Though considerable mineralogical variations appear possible on the basis of the major and trace element, and long-lived radiogenic isotope variations [Stracke et al., 2003]. However, if the enriched component is not very abundant and melts extensively, it appears possible that it has little effect on the U-Th systematics of these lavas, especially in light of the Th and U depletion of the Theistareykir basalts relative to MORB (indicating that the enriched component has similar Th and U contents compared to the MORB source [Stracke et al., 2003]).

[18] Is it likely, however, that only enriched basalts from other localities like Draugarhraun (Krafla) with low 230Th excesses are influenced by melts derived from a mafic component? In this case, because no correlation between radiogenic isotope ratios and (230Th/238U) is observed, it would have to be assumed that the melts from the mafic component only affect the Th characteristics of the most evolved lavas, whereas the Th isotope composition of the slightly less enriched and evolved lavas are not significantly influenced. Considering the continuous linear mixing trends between radiogenic isotope ratios (and also major and trace elements including Th and U) from the Theistareykir to the Draugarhraun and Asbyrgi lavas such a scenario appears unlikely.

4.3. U-Th Disequilibrium as a Tracer of the Dynamics of Melting

[19] All currently available quantitative models of U series systematics rely on a mineralogically homogeneous source. In the preceding discussion, we have demonstrated that because of the lack of correlation between radiogenic isotope ratios (Sr, Nd, Pb, and Hf) and 230Th excesses source heterogeneity is not likely to influence the U-Th systematics in the Theistareykir basalts. The Th systematics in the Theistareykir basalts are therefore likely to originate from variations in the timing of the melting and melt extraction process only and quantitative models will be used in the following to constrain the dynamics of melting beneath Theistareykir.

4.3.1. General Constraints on the Dynamics of Melting

[20] Using quantitative models [e.g., McKenzie, 1985; Spiegelman and Elliott, 1993; Spiegelman, 2000], what are the critical parameters that influence the Th-systematics, and how do the constraints derived from the U-Th systematics compare to the information derived from the chemical and long-lived radiogenic isotope systematics [Stracke et al., 2003]?

[21] The most significant parameter influencing the Th systematics is the melting (M) or upwelling rate (W) of the solid mantle. In general, low melting rates (or slow upwelling rates) promote higher 230Th excesses. In addition, other factors such as melting time (T, the time period from the onset of melting until the final degree of melting is reached [Zou and Zindler, 2000]) can significantly influence the Th isotope signatures; short melting times lead to higher degrees of disequilibrium. Furthermore, the partitioning behavior of U and Th plays a critical role because the relative fractionation behavior between U and Th changes between the garnet- and spinel-stability field [e.g., Beattie, 1993a, 1993b; Hauri et al., 1994; Landwehr et al., 2001; LaTourrette and Burnett, 1992; LaTourrette et al., 1993; Salters and Longhi, 1999; Salters et al., 2002; Wood et al., 1999]. As a consequence, the large (230Th/238U) produced by melting in the garnet-stability field might be reduced by progressive melting in the spinel-stability field because Th behaves less incompatibly relative to U in the spinel field than in the garnet-stability field (provided that a significant amount of the original U and Th is still present in the residual source). Finally, low residual porosity promotes large 230Th excesses, but, in contrast to the parameters discussed above which have no influence on the chemical composition of the melts, lower porosity also increases the stable trace element enrichment in the melts generated for any given F.

4.3.2. Quantitative Modeling of the U-Th Systematics

[22] Melting parameters such as residual porosity and upwelling rate are most often estimated from U-Th disequilibrium data on the basis of quantitative models of the melting and melt extraction process [McKenzie, 1985; Spiegelman and Elliott, 1993]. In using such models, however, one has to be aware that any parameter derived on the basis of these models is only an estimation based on a highly simplified model of an actually much more complex melting and melt extraction process. In essence, two end-member type models are often used and/or compared: the so-called dynamic or near-fractional melting model [McKenzie, 1985] and the equilibrium porous flow model [Spiegelman and Elliott, 1993]. The main difference between these two models (see also discussion in the auxiliary material) [Elliott, 1997; Jull et al., 2002; Sims et al., 1999] is that in the near-fractional melting model, melts are instantaneously removed from the solid residue once a critical threshold porosity is reached and transported to the surface without melt-solid interaction, while in the equilibrium porous flow model, melt and solid constantly equilibrate during melt transport. Although there are indications that both processes, rapid disequilibrium melt extraction and slow porous flow with chemical equilibration, are likely to occur during melting and melt transport beneath ridges and OIB (see Kelemen et al. [1997] and Van Orman et al. [1998] for a recent summary and discussion) the reality is most likely somewhere in between and estimates of melting parameters derived using simple end-member models [McKenzie, 1985; Spiegelman and Elliott, 1993] should be used with caution. Moreover, it has recently been suggested that 226Ra and 230Th excesses are produced by different processes at different depths of the melting region. Consequently, so-called two porosity models have been introduced to model U-Th disequilibrium systematics [e.g., Jull et al., 2002; Kelemen et al., 1997; Lundstrom, 2001; Sims et al., 2002]. There appears to be some agreement, however, that 230Th excesses are created deep in the melting region and require rapid melt transport without continuous melt-solid interaction. Major element modeling and the trace element composition of abyssal peridotites also indicate that melting beneath ridges is best described as a near-fractional melting process [e.g., Johnson et al., 1990; Kinzler and Grove, 1992a, 1992b; Langmuir et al., 1992] and field evidence indicates that a significant portion of the melt is likely to be extracted through high porosity channels [Kelemen et al., 1995].

[23] Therefore a near-fractional “dynamic” melting model is used here to derive some first order estimates of melting parameters and to investigate possible causes for the different (230Th/238U) in the Theistareykir basalts. The model used is described in detail in the electronic supplement and supplied in a Microsoft™ Excel spread sheet. Compared to analytical solutions for dynamic melting [McKenzie, 1985; Zou and Zindler, 2000], this incremental dynamic melting model allows for changing partition coefficients and variable proportions of phases entering the melt as a function of the depth of melting (as indicated by recent experimental studies of melting [e.g., Baker et al., 1995; Kinzler and Grove, 1992a, 1992b; Kinzler, 1997; Longhi, 2002; Salters and Longhi, 1999; Salters et al., 2002]) and takes radioactive decay during melt transport into account. For both simple modal or non-modal dynamic melting of stable trace elements and U-Th systematics, the results of this model are in excellent agreement with the analytic solutions (see electronic supplement [McKenzie, 1985; Zou, 1998; Zou and Zindler, 2000]).

4.3.3. Constraints on the Dynamics of Melting Beneath Theistareykir

[24] For modeling melting beneath Theistareykir, the following constraints apply: in order to create 20% melts such as the picrites [e.g., Klein and Langmuir, 1987; Schilling et al., 1999; Stracke et al., 2003] at a locality with a crustal thickness of 20–30km [Darbyshire et al., 2000b; Staples et al., 1997], and assuming an average melt production rate of 0.003/km [Klein and Langmuir, 1987; Langmuir et al., 1992; McKenzie and Bickle, 1988], melting has to start at depths of at least 90–100km, well within the garnet-stability field. Assuming that the garnet-spinel phase transition occurs at depths of about 75km [e.g., Hirschmann and Stolper, 1996; Kinzler, 1997; Longhi, 2002; Salters and Hart, 1989; Takahashi, 1986, and references therein], up to 7.5% melt can be generated in the presence of residual garnet and the Th-systematics of the Theistareykir basalts are therefore dominated by melting in the garnet-stability field (Figure 4; see above). This explains the relatively large 230Th excesses (average about 25%) as well as the lack of correlation between major and trace element tracers of the extent and depth of melting. Even when only 2% of the total extracted melt is generated in the garnet-stability field, the (230Th/238U) of the extracted melt are large and essentially constant (Figure 4), whereas other major and especially trace element parameters vary significantly with the degree and depth of melting (or the relative amount of melting in the garnet-stability field, e.g., Sm/Nd, Lu/Hf). Thus no correlations between major and trace element parameters of melting and 230Th excesses are expected, unless sources with different mineralogical compositions, and therefore different U and Th partition coefficients are present.

Figure 4.

Total amount of melt generated in the garnet-stability field (initial depth of melting) plotted versus the calculated (230Th/238U) of the extracted melt as a function of melt velocity and porosity (ϕ). The total degree of melting is 20%, the upwelling velocity is 1cm/yr. In case more than 2–3% of the total melt are generated in the garnet-stability field, 230Th excesses are essentially constant. The effects of different melt extraction velocities are shown for ϕ = 0.1%, the effects of different ϕ are shown for instantaneous melt extraction (i.e., melt velocity → ∞). Clinopyroxene and garnet partition coefficients for melting in the garnet-stability field (>75km) are from Salters and Longhi [1999] and from Hauri et al. [1994] for melting in the spinel-stability field. See spreadsheet “dy_melt_model_U-Th.xls” and electronic supplement for further details of the modeling.

[25] On the basis of our near-fractional melting model, residual porosity ≤0.1%, upwelling rates ≤1cm/yr (melting rates ≤9.9*10−5 kg/m3yr), and melt ascent velocities ≥3–4m/yr are suggested in order to explain 230Th excesses of about 25% ± 5% (Figure 4). These estimates can change depending on the type of model and the set of partition coefficients used in the modeling. Here, we use the D's given by Salters and Longhi [1999] for melting in the garnet-stability field (>75km depth), and the D's given by Hauri et al. [1994] for melting in the spinel-stability field (<75km depth). While the D's given by Salters and Longhi [1999] and Salters et al. [2002] are most appropriate for melting in the garnet-stability field, a number of different studies provide experimental D's that are appropriate for melting at shallower depths (see Sims et al. [1999] and Stracke et al. [1999] for a recent compilation). The effects of changing the D's for melting in the spinel field can be evaluated by changing the partition coefficients in the supplied spread sheet (see also Sims et al. [1999] for a discussion of the influence of D's on modeling U-Th disequilibria). The relative uniformity of the bulk of the Theistareykir data (21 out of 27 samples have 230Th excesses of 25 ± 5%) further suggests that melt generation and extraction processes beneath Theistareykir are fairly reproducible. However, some variation in the timing and dynamics of melt production and/or extraction is suggested by the lower 230Th excesses of about 15% in the picrites and enriched samples from Draugarhraun (part of the Krafla volcanic system) and Asbyrgi (a location NE of Theistareykir; see Slater [1996] and Slater et al. [2001] for detailed geographic maps).

[26] The picrites are the most depleted melts with the largest extent of melting and are inferred to represent aggregate melts extracted from virtually the entire melting column [Stracke et al., 2003]. Note that the picrites are among the oldest lavas of each eruption sequence and are the least voluminous (at both the Reykjanes Peninsula and at Theistareykir), i.e., each eruption cycle starts with picrites [Jakobsson, 1972; Slater et al., 2001]. Therefore the relatively small amount of total melt generated at the beginning of each melting cycle could lead to slow melt extraction times, allowing excess 230Th to decay substantially during melt transport. Melt transport rates for the picrites would have to be at least a factor of three lower (about 1m/yr) compared to samples with 230Th excesses of about 25 ± 5% (≥3m/yr) in order to account for the differences in (230Th/238U) (Figure 4). Alternatively, faster mantle upwelling rates could be invoked. Large short-term variations in melt production rates, and therefore mantle upwelling rates, are inferred to take place during ice unloading in response to deglaciation at a time that coincides with the emplacement of the oldest lava flows at Theistareykir (picrites, Storaviti) (12,000–10,500 yr BP [Jull and McKenzie, 1996; Maclennan et al., 2002b]). According to the model of Jull and McKenzie [1996] and Maclennan et al. [2002a], mantle upwelling rates are expected to be significantly faster during generation of the oldest lavas from Theistareykir (12,000–10,5000 yr BP) than during generation of the younger lavas. Thus the oldest lavas from Theistareykir (12,000–10,5000 yr BP) would be expected to have significantly lower (230Th/238U) compared to the younger lavas. Although this is true in case of the picrites, the lavas from Storaviti have higher (230Th/238U) comparable to those of the younger lavas, and overall, no systematic variation between age and (230Th/238U) can be observed in the Theistareykir lavas (Table 1). As a third alternative, the picrites could represent aggregates which somehow manage not to include the melt created at the very bottom of the melting column, which has the highest (230Th/238U) as well as the highest Th and U concentrations. Such a scenario would have little effect on other chemical and isotopic parameters, and could, at least in principle, also account for the different U-Th systematics of the picrites compared to the other samples.

[27] The chemically and isotopically most enriched basalts from locations adjacent to Theistareykir (the quartz-tholeiites from Draugarhraun and the Asbyrgi basalts) also have 230Th excesses of about 15%, similar to the picrites. On the basis of their chemical and radiogenic isotope characteristics, these lavas are inferred to be melts derived from higher average pressures and lower degrees of melting than less enriched lavas from Theistareykir [Stracke et al., 2003]. Thus they are expected to be aggregated melts with a relatively larger contribution of melts produced deep in the melting regime (within the garnet-stability field), and aggregation of melts excluding the high pressure melts as suggested for the picrites appears to be an unlikely explanation for the U-Th systematics of these melts. In order to explain their low excess 230Th, faster upwelling rates and/or slower melt travel times (as for the picrites) could be invoked. An increase in upwelling rate by about a factor of two is required in order to decrease 230Th excesses from about 25% to 15%. Although short-term variations in mantle upwelling rate are suggested due to deglaciation [Jull and McKenzie, 1996; Maclennan et al., 2002b], only lavas erupted directly at the onset of deglaciation are expected to be influenced by increased upwelling rates. The age of the samples from Asbyrgi is not known precisely, but the samples from Draugarhraun (Krafla) are too young to be influenced by short-term temporal variations in mantle upwelling rates during the onset of deglaciation (see above). The fluid dynamics of melting and melt extraction, however, suggest that melting and upwelling rates also vary spatially within a given melting regime [Spiegelman, 1996]. For passive mantle upwelling, which is inferred to be a reasonable model for Theistareykir [Darbyshire et al., 2000a; Maclennan et al., 2002a], upwelling and melting rates decrease away from the center of the melting regime [Spiegelman, 1996], leading to higher (230Th/238U) in melts produced farther from the center of the melting regime. However, if some coupling between the solid upwelling rate and melting rate is assumed (as done by Spiegelman and Elliott [1993] and Spiegelman [2000]; see, e.g., Spiegelman [2000, equation (13)]), a decrease in the upwelling rate will also lead to a decrease in melt extraction rate. Therefore slower average melt extraction velocities can outweigh the effect of slower upwelling rates and due to decay of 230Th during melt transport, melts created at some distance from the center of the melting regime may have significantly less excess 230Th upon eruption than melts created closer to the center. This scenario is consistent with the chemical and isotopic systematics of these melts, which suggests that they are smaller degree melts created at average higher pressure than less enriched melts from Theistareykir [Stracke et al., 2003].

[28] Because of their evolved nature, one might suspect that pre-eruptive storage in high-level magma chambers can affect the U-Th systematics in the Draugarhraun and Asbyrgi melts. However, residence times on the order of 50,000–70,000 years would be required to lower the 230Th excesses from about 25% to 15%, which appear too long on the basis of recent estimates of crystal fractionation times [see, e.g., Hawkesworth et al., 2000, and references therein]. Thus, although storage in crustal magma chambers alone is insufficient to explain the Th systematics of these evolved melts, it could well be a, albeit small, contributing factor.

[29] Attributing the variations in (230Th/238U) to variations in melt extraction velocity, however, is problematic, as little is known about the fluid dynamics of melt extraction mechanisms beneath ridges and ocean islands. However, somewhat tighter constraints on melt extraction velocities may be derived from combining U-Th with U-Pa isotope systematics, and would be an important test on the hypotheses described above.

5. Conclusions

[30] Source heterogeneity does not appear to be an important factor in creating 230Th excesses at Theistareykir, indicating that the compositional heterogeneity in the Theistareykir source is most likely not coupled to mineralogical heterogeneity. Relatively large (230Th/238U) for the bulk of the samples of about 1.25 ± 5 are explained by a reproducible melting process dominated by melting peridotite starting in the garnet-stability field and a near-fractional melting model suggests residual porosity ≤0.1%, upwelling rates ≤1cm/yr (melting rates ≤9.9*10−5 kg/m3yr), and melt ascent velocities ≥3–4m/yr. Lower 230Th excesses of about 15% in both the most depleted and enriched lavas indicate some variations in the melting and melt extraction process. The low 230Th excesses of the most depleted melts are either an effect of slow melt extraction times, or could, in principle, be caused by incomplete melt aggregation excluding a significant fraction of the instantaneous melts created at the very bottom of the melting regime. The low 230Th excesses of the most enriched lavas are most likely an effect of slow melt extraction velocities for melts created at some distance from the center of the melting column.


[31] Michael Bizimis, Tim Elliott, and Bernard Bourdon thanked for discussions and advice. Ted Zateslo is thanked for keeping the ISOLAB 54 and the Finningan MAT 262 at the NHMFL in excellent condition. J. Maclennan is thanked for helpful comments, and K. Sims and Ken Rubin are thanked for journal reviews. A.S. was in part supported by a HSP-III doctoral fellowship from the German Academic Exchange Service (DAAD). D. McKenzie acknowledges support from the NERC and the Royal Society. This work was supported by the U.S. National Science Foundation (EAR 1323-608-22 to AZ).