Introduction

Mixing between compositionally distinct components is increasingly recognised as an important process in generating the compositional diversity of oceanic magmatism at various length scales, from microscopic melt inclusions in olivine crystals to suites of lavas from 100 km rift segments (e.g. Langmuir et al. 1978; Saal et al. 1998; Lundstrom et al. 1998; Kamber and Collerson 2000; Reiners 2002). In particular, high precision Pb isotope data on individual mid-ocean ridge segments and on individual volcanic systems on oceanic islands such as Iceland often preserve coherent linear trends that have been attributed to local mixing between two compositionally distinct end-members (e.g. Sims et al. 2002; Stracke et al. 2003; Thirlwall et al. 2004; Baker et al. 2004; Blichert-Toft et al. 2005).

Radiogenic isotope data demonstrate significant compositional heterogeneity in the mantle beneath Iceland and the adjacent ridges (e.g. Hart et al. 1973; Sun and Jahn 1975; Zindler et al. 1979; Condomines et al. 1983; Kurz et al. 1985; Sigmarsson et al. 1992; Hémond et al. 1993; Hanan and Schilling 1997; Hardarson et al. 1997; Hilton et al. 1999; Schilling et al. 1999; Stecher et al. 1999; Skovgaard et al. 2001; Stracke et al. 2003; Thirlwall et al. 2004, 2006; Kokfelt et al. 2006; Brandon et al. 2007). Initial models explained the isotopic variations in terms of mixing between Iceland plume mantle (high 87Sr/86Sr, 206Pb/204Pb, 3He/4He, La/SmN and low 143Nd/144Nd) and the surrounding MORB source mantle (e.g. Hart et al. 1973). However, subsequent work has revealed more complexity in the nature and number of mixing end-member compositions that contribute to Icelandic magmatism and the surrounding ridges (Reykjanes Ridge, Kolbeinsey Ridge), with up to six distinct components suggested (e.g. Hanan and Schilling 1997; Murton et al. 2002; Thirlwall et al. 2004; Kokfelt et al. 2006). There has also been considerable debate as to whether the depleted unradiogenic component is in fact ambient North Atlantic MORB source or an intrinsic depleted component within the upwelling plume mantle (e.g. Kerr et al. 1995; Hanan and Schilling 1997; Fitton et al. 1997; Hanan et al. 2000; Kempton et al. 2000; Chauvel and Hémond 2000; Fitton et al. 2003; Stracke et al. 2003; Thirlwall et al. 2004; Kokfelt et al. 2006). High precision Pb isotope data (Thirlwall et al. 2004; Baker et al. 2004) show that in geographically restricted localities within Iceland, the isotopic variations are consistent with the involvement of two components only. How these local end-members relate to the regional-scale end-members recognised for Iceland as a whole can potentially provide information about the distribution and length scale of heterogeneities within the mantle.

In this paper, we focus on a series of very fresh, virtually aphyric basaltic lava flows that erupted in historic times (after settlement in 874 ad) along the Reykjanes Peninsula in south-west Iceland (Fig. 1). We use high-precision Pb isotope data coupled with other compositional data to investigate the nature of mixing processes within this geographically and temporally restricted sample suite. We also evaluate the relationship of these historic lavas both to the earlier phases of magmatism on the Reykjanes Peninsula and to the adjacent offshore Reykjanes Ridge.

Fig. 1
figure 1

a Map of the Reykjanes Peninsula, Iceland, showing the historic (c. 940 to c. 1340 ad) lava flows and other localities that were sampled for this study or discussed in the text. Compiled from Jónsson (1978, 1983) and Einarsson et al. (1991). Sample locations are shown by the whitecircle symbols. Inset map shows location of the Reykjanes Peninsula within Iceland. b Magmatism in the Reykjanes Peninsula has been divided into a series of volcanic systems (fissure swarms: e.g. Jakobsson et al. 1978), and the historic lava flows are assigned to the following volcanic systems based on the locations of their eruptive vents: Reykjanes (flow 1), Svartsengi (or Grindavik: flows 2–4), Krisuvik (or Trölladyngja: flows 5–12), and Brennisteinsfjöll (or Blafjöll: flows 13–20). c Location of the two offshore samples from the northern Reykjanes Ridge (after Johnson and Jakobsson 1985)

Full size image

Geological background and sample details

The Reykjanes Peninsula in south-west Iceland represents the onshore continuation of the Reykjanes Ridge. The exposed basement volcanic rocks are all younger than 700 ka, and a significant proportion of the peninsula is covered by post-glacial (<13 ka) mafic lavas, as lava shields or fissure-fed lavas. Post-glacial magmatism occurs in a series of five en-echelon NE–SW-trending volcanic systems (fissure swarms), whose arrangement is a consequence of the obliquity of the peninsula relative to the plate spreading direction (Jakobsson et al. 1978; Gudmundsson 1986; Gee et al. 2000; Clifton and Kattenhorn 2006). From west to east, these are the Reykjanes volcanic system (partly submerged), the Svartsengi volcanic system, the Krisuvik volcanic system, and the Brennisteinsfjöll volcanic system, with the Hengill volcanic system forming the eastern edge of the Reykjanes Peninsula magmatism (Fig. 1b). The segment of rift represented by the Reykjanes Peninsula is about 65 km in length. The Hengill volcanic system continues to the north-east to the Langjökull region, forming the ~130 km long Western Volcanic Zone. The period of deglaciation (9–13 ka) was associated with greater magma production rates and the eruption of minor volumes of trace-element-depleted picritic lavas (Jakobsson et al. 1978; Gee et al. 1998a; Maclennan et al. 2002). Although many studies have focused on these unique picritic rocks (e.g. Elliott et al. 1991; Gurenko and Chaussidon 1995; Révillon et al. 1999; Chauvel and Hémond 2000; Skovgaard et al. 2001; Brandon et al. 2007), it is important to emphasize that they are not representative of the main volume of post-glacial magmatism erupted on the Reykjanes Peninsula (e.g. Jakobsson et al. 1978).

The historic lavas were erupted as fissure-fed flows within a 400 year interval from c. 940 to c. 1340 ad, based on published age constraints from historical records, tephrochronology, and 14C dating (Table 1, Jónsson 1983; Einarsson et al. 1991; and references therein). Since this time, there have been no further eruptions on the peninsula itself, and magmatic activity in the region is now concentrated in the ridge segments just offshore where there have been several eruptions within the last century (e.g. Jakobsson 1974; Johnson and Jakobsson 1985). One or two samples were taken from most of the main named historical flows (Fig. 1a), although it should be noted that some of the lava flows were probably part of the same eruptive events. Most samples were taken from the more massive flow interiors inside fissures, avoiding any segregation veins or pipe vesicles, although some samples are from the more vesicular flow tops. For comparative purposes, we included some additional samples in this study: (a) two samples of the recent activity offshore (Johnson and Jakobsson 1985) from a submarine hill southwest of Eldeyjarbodi (~63.5°N) and Steinaholl (~63.1°N) on the northern Reykjanes Ridge (Fig. 1c), and (b) samples of older lava flows that represent the compositional end-members on the Reykjanes Peninsula proposed by Gee et al. (1998b) (Lagafell, ~10 ka—‘depleted’; Stapafell, ~20–100 ka—‘enriched’, see “Discussion”). We also provide high-precision Pb isotope data on additional Reykjanes Peninsula samples from Kurz et al. (1985), Skovgaard et al. (2001) and Brandon et al. (2007) that allow us to put the historic lavas in the broader context of all Reykjanes Peninsula magmatism.

Table 1 Major element data (wt%) and published age estimates for Reykjanes Peninsula lavas
Full size table

Results

Petrography and mineral chemistry

The historic lavas are sparsely phyric, with ≪5% phenocrysts of olivine and/or plagioclase, and occasionally trace amounts of clinopyroxene. Phenocryst compositions were analysed in selected samples on grain mounts of hand-picked crystals by electron microprobe in the Geological Institute, Copenhagen University, following the techniques in Larsen and Pedersen (2000). The complete data set of mineral compositions are provided in the Electronic supplementary Appendix Table A1.

Olivine

The range of olivine compositions is summarised in Fig. 2a, which plots the olivine forsterite content against the Mg# of the host sample (Mg# = 100 × Mg2+/(Mg2++Fe2+) assuming that Fe3+/ΣFe = 0.14, based on Mössbauer analyses of Reykjanes Peninsula glasses from Oskarsson et al. (1994). The historic lavas have olivine compositions of Fo78 to Fo86. Sample 456749 from Stapafell contains some olivines that have higher forsterite contents (Fo82 to Fo88), and the offshore sample from Eldeyjarbodi (NI-3605) has olivines with lower forsterite contents (Fo75 to Fo79). For most samples, the olivines have a limited range in forsterite content and they have compositions broadly expected for crystals in equilibrium with the host melt, using a K D(Fe–Mg) value of 0.32 ± 0.04 for the distribution of Fe2+ and Mg between olivine and melt (Putirka 2005). However, a few samples (in particular 456746 and 456749) contain some crystals with significantly higher forsterite contents than expected for equilibrium compositions, and these represent either relicts from an earlier stage in the differentiation of the magma or xenocrysts from earlier magmatic episodes.

Fig. 2
figure 2

a Olivine (Fo%) versus whole rock Mg-number (Mg#) of the host lava. Mg# [molar 100 × Mg2+/(Mg2+ + Fe2+)] is calculated with Fe3+/ΣFe = 0.14 (Oskarsson et al. 1994). The lines indicate the compositions of olivine in equilibrium with a melt of a given Mg-number, assuming a KD(Fe–Mg) value of 0.32 ± 0.04 for the distribution of Fe2+ and Mg2+ between olivine and melt (Putirka 2005). b Plagioclase (An %) versus whole rock Mg# of the host lava

Full size image

Plagioclase

The range of plagioclase phenocryst compositions is summarised in Fig. 2b, which plots the plagioclase anorthite content against the Mg-number of the host sample. Phenocryst compositions vary between An71 and An91, with most lying between An75 and An88, and this is similar to the range found by Jakobsson et al. (1978) for post-glacial fissure lavas on the Reykjanes Peninsula (An68–An88). Plagioclase phenocrysts show minor normal zoning with rim-core variations in An content of 1–4%.

Clinopyroxene

This phase was only found as a phenocryst in three samples (456714, 456746, 456722), and crystals were <1 mm in size. Clinopyroxenes were only analysed in two samples (456714, 456746). They have broadly similar compositions (En44–52Fs9–12Wo37–43 and TiO2 ~0.5%), although those in sample 456746 have slightly higher Mg-numbers (83.5 ± 1.4, 1 SD) and lower Al2O3 (3.1 ± 0.6 wt%, 1 SD) than those in 456714 (81.7 ± 0.9 and 3.7 ± 0.4 wt%, respectively).

Major and trace element data

Major and trace element data are presented in Tables 1 and 2, with analytical details given in the table footnotes. The historic lavas are tholeiitic basalts, with a relatively restricted range in MgO contents (6.5–9.2 wt%) when compared with the compositional range of all recent (<20 ka) Reykjanes Peninsula lavas (Fig. 3). In most cases, lava flows that belong to the same eruptive episode show remarkable compositional homogeneity that does not greatly exceed the analytical uncertainties. This is highlighted by the average composition listed for the Krisuvik Fires eruptive episode (1151–1188 ad) in Tables 1 and 2, for which the signal to noise ratio (the variability of five separate samples divided by the precision of the analytical method based on replicate analyses) is less than two for all elements except for Rb, La and Ce. Other flows from the Krisuvik volcanic system, including the Eldborgarhraun and Afstapahraun yngra flows with slightly younger inferred ages (c. 1325 ad), have indistinguishable compositions to the Krisuvik Fires flows. For these two flows, the compositional deviations for all elements from the mean composition of the Krisuvik Fires flows are all within the 2 SD analytical precision or the 2 SD measured variability for the Krisuvik Fires flows, except for Sr. Jakobsson et al. (1978) and Gee et al. (2000) have documented other examples of the compositional homogeneity of tholeiitic basaltic flows on the Reykjanes Peninsula. These observations are consistent with previous studies of lava flows elsewhere in Iceland (e.g. Sigmarsson et al. 1991; Maclennan et al. 2003a) that suggest that the more evolved flows (i.e. <8 wt% MgO) tend to have a more restricted compositional range, presumably due to longer residence times in the magmatic plumbing system that allows for more thorough homogenisation of magma batches. For the Reykjanes Fires episode (1211–1240 ad), the Stampahraun flow is compositionally different from the other three units that are all inferred to have been part of this eruptive episode, and it was erupted from a different volcanic system (Fig. 1).

Table 2 Trace element data (ppm) for Reykjanes Peninsula lavas
Full size table
Fig. 3
figure 3

Major element variation diagrams for the historic Reykjanes Peninsula lavas: MgO versus a SiO2, b Na2O, c CaO, d Al2O3, e TiO2, f K2O, g–h Fe2O3(t). Published literature data for other Reykjanes Peninsula lavas are shown for reference (Jakobsson et al. 1978; Wood 1978; Condomines et al. 1983; Elliott et al. 1991; Hémond et al. 1993; Gurenko and Chaussidon 1995; Gee et al. 1998b; Révillon et al. 1999; Skovgaard et al. 2001; Martin and Sigmarsson 2005; Kokfelt et al. 2006). Small black circles are published analyses of the historic ‘less enriched’ group. h MgO versus Fe2O3(t) has an expanded scale to show the full compositional diversity of the Reykjanes Peninsula lavas, notably the trace element depleted basaltic and picritic lavas as exemplified by samples from Haleyjabunga, Lagafell, Burfell, Asar, and Vatnsheidi (Fig. 1a). The ‘enriched’ lava shields (volumetrically dominated by the Sandfellshaed and Thrainsskjöldur shields: Fig. 1a) represent about two-thirds of the total volume of post-glacial magmas erupted on the Reykjanes Peninsula, and they were erupted shortly after the deglaciation interval that produced the ‘depleted’ picritic and basaltic eruptions (Jakobsson et al. 1978). The majority of the plotted Reykjanes Peninsula data are from samples younger than <20 ka, and most are post-glacial (i.e. <13 ka), although data from a few sub-glacial eruptions (most likely ~20–100 ka) from Stapafell and the Hengill volcanic system are also plotted

Full size image

Samples from the Reykjanes Peninsula as a whole span a wide range in incompatible trace element ratios: for example Nb/Zr varies from 0.01 to 0.18 (Fig. 4b) and La/SmN varies from 0.28 to 1.65 (Fig. 4a). Gee et al. (1998b) divided the samples into ‘depleted’ (Nb/Zr <0.08) and ‘enriched’ (Nb/Zr >0.08) groups. All of the historic lavas fall into the ‘enriched’ group, with high Nb/Zr (0.10–0.14) and LREE-enriched compositions. For convenience, we divide the historic lavas into two arbitrary groups based on differences in trace element characteristics, namely a ‘less enriched’ group (Nb/Zr 0.10–0.13; La/SmN 1.21–1.36) and a ‘more enriched’ group (Nb/Zr 0.13–0.14; La/SmN 1.38–1.48). The ‘less enriched’ group show trends of increasing TiO2 and K2O with decreasing MgO, and the ‘more enriched’ group are displaced above these trends to higher MgO, TiO2 and K2O (Fig. 3). The ‘less enriched’ group also has higher SiO2 (48.5–49.2 wt%) than the ‘less enriched’ group (47.2–47.9 wt%).

Fig. 4
figure 4

a Chondrite-normalised REE patterns for the historic Reykjanes Peninsula lavas, the historic offshore lavas, and the sub-glacial Stapafell eruption (normalising values from Anders and Grevesse 1989). b Zr versus Nb/Zr, to show the relatively restricted range in Nb/Zr of the historic lavas compared to magmatism of the Reykjanes Peninsula as a whole (Hémond et al. 1993; Gurenko and Chaussidon 1995; Gee et al. 1998b, 1999; Révillon et al. 1999; Chauvel and Hémond 2000; Kempton et al. 2000; Hanan et al. 2000; Skovgaard et al. 2001; Kokfelt et al. 2006). These diagrams illustrate the division of the historic lavas into a ‘less enriched’ group (La/YbN 1.3–2.0; Nb/Zr 0.10–0.13) and a ‘more enriched’ group (La/YbN 2.2–2.6; Nb/Zr 0.13–0.14). Boundary between ‘enriched’ and ‘depleted’ samples for the Reykjanes Peninsula as a whole is taken from Gee et al. (1998b)

Full size image

The ‘more enriched’ group are all found in the east of Reykjanes Peninsula within the Brennisteinsfjöll volcanic system, while the ‘less enriched’ group are found to the west in the Reykjanes, Svartsengi, and Krisuvik volcanic systems. These systematic spatial compositional variations in the historic lavas are shown in Fig. 5. For example, within the ‘more enriched’ group, the flows with the highest Nb/Zr are found in the extreme east of the peninsula, while in the ‘less enriched’ group, the flows with the lowest Nb/Zr are found in the extreme west of the peninsula in the Reykjanes volcanic system, and the historic offshore lavas in the northern Reykjanes Ridge have even lower Nb/Zr.

Fig. 5
figure 5

Spatial compositional variations in the historic Reykjanes Peninsula lavas: a SiO2, b MgO, c TiO2, d 206Pb/204Pb, e Nb/Zr. The horizontal axis is the distance of the eruptive fissure for each sampled flow, projected onto a ENE line from the SW tip of the Reykjanes Peninsula to Hengill. The two offshore samples are plotted at arbitrary distances (see Fig. 1c)

Full size image

Radiogenic isotope data (Pb–Sr–Nd)

High precision Pb isotope data and Sr–Nd isotope data on selected samples are presented in Tables 3 and 4, with complete analytical details provided in the table captions. The historic Reykjanes Peninsula samples show a restricted range in radiogenic isotope compositions compared to the isotopic variation of all post-glacial lavas from the Reykjanes Peninsula (Fig. 6: data sources are listed in figure caption). The variations in 87Sr/86Sr (0.70309–0.70314) and εNd (+7.3 to +7.8) are barely outside of external analytical reproducibility (see Table 4 caption). In contrast, high-precision Pb isotope data, analysed by multi-collector inductively coupled plasma mass spectrometer (MC–ICP-MS) using a 207Pb–204Pb double spike to control mass fractionation (Baker et al. 2004), show significant variations (206Pb/204Pb 18.771–18.881; external reproducibility is <±0.003). Furthermore, the data on the historic lavas preserve coherent linear trends on Pb–Pb isotope plots (Fig. 6b, d). The data plot below but parallel to the NHRL (Northern Hemisphere Reference Line: Hart 1984) for 207Pb/204Pb (i.e. negative ∆207Pb: Fig. 6a) and above but at a shallower slope to the NHRL for 208Pb/204Pb (i.e. positive ∆208Pb: Fig. 6c). Pb isotope data were measured on rock chips after leaching in 6 M HCl for ~2 h. Baker et al. (2004, 2005) have demonstrated the robustness of this leaching procedure to recover magmatic Pb isotope compositions from young basaltic rocks. The external precision of the double-spike method for Pb isotope ratios is <~150 ppm for basaltic samples and <~400 ppm for the low Pb concentration (sub-0.2 ppm) picritic samples (Thirlwall et al. 2004; Baker et al. 2004, 2005) based on replicate analyses of individual leached samples. There is good agreement between our double-spike data and those in Thirlwall et al. (2004) on samples from the same flows: e.g. Ögmundarhraun, Haleyjabunga, Stapafell and Lagafell (Table 3). The difference between the average composition of each flow (where more than one sample was analysed per flow) from both studies is <500 ppm for all Pb isotope ratios.

Table 3 High-precision double-spike Pb isotope data for Reykjanes Peninsula lavas
Full size table
Table 4 Sr–Nd–O isotope data for Reykjanes Peninsula lavas
Full size table
Fig. 6
figure 6

Isotope variations within the historic lavas: a, b 206Pb/204Pb versus 207Pb/204Pb, c, d 206Pb/204Pb versus 208Pb/204Pb, e 87Sr/86Sr versus 143Nd/144Nd, f δ18O(V-SMOW) olivine versus 143Nd/144Nd. Data for other Reykjanes Peninsula lavas are shown for comparison (Sun and Jahn 1975; Condomines et al. 1983; Elliott et al. 1991; Hémond et al. 1993; Hanan and Schilling 1997; Gee et al. 1998b; Kempton et al. 2000; Chauvel and Hémond 2000; Skovgaard et al. 2001; Thirlwall et al. 2004, 2006; Blichert-Toft et al. 2005; Kokfelt et al. 2006; Tables 3, 4). These other Reykjanes Peninsula samples are all post-glacial (<13 ka) except for three samples from Thirlwall et al. (2004, 2006) that are 13–20 ka. Error bars plotted on (b) and (d) are ± 150 ppm 2 SD external reproducibility, and the dotted lines are a least-squares linear regression line through the data. On (e), the black line represents a hypothetical mixing curve between ‘depleted’ and ‘enriched’ melts (samples RP95C and RP100C, respectively: Thirlwall et al. 2004), and the grey shaded area represents the estimated analytical uncertainty (~±20 ppm, 2 SD). Estimate of average composition of Reykjanes Peninsula hydrothermally altered crust (87Sr/86Sr = 0.7036, 143Nd/144Nd = 0.51305, δ18Oolivine = +2‰) from Gee et al. (1998b). RP103A is a sample from Kleifarvatn that Thirlwall et al. (2004) use to define their ID2 component. NHRL is the Northern Hemisphere Reference Line (from Hart 1984). For some Reykjanes Ridge samples in (f), δ18Oolivine was calculated from δ18Oglass assuming an olivine-melt fractionation factor of +0.4‰ (e.g. Eiler 2001). Note that the plotted δ18Oolivine values are measured on ~2–10 grains, which will average out any potential within-sample heterogeneity that has been demonstrated in olivines from some mafic lavas elsewhere in Iceland (e.g. Maclennan et al. 2003b; Bindeman et al. 2006)

Full size image

Stable isotope data (δ18O)

Hand-picked olivine and/or plagioclase separates (~1.7 mg aliquots) from selected samples were measured for oxygen isotopes at Royal Holloway University of London using a laser fluorination technique (Mattey and Macpherson 1993; Thirlwall et al. 2006). The data are provided in Table 4, and further analytical details are given in the table footnote. The sparsely phyric nature of many of the historic lavas limited the number of samples that could be analysed. All olivines in the historic lavas have light δ18O (+4.2 to +4.3‰) compared to values expected for olivines in equilibrium with normal mantle (+5.2 ± 0.3‰: Mattey et al. 1994). One sample of a more primitive depleted older Reykjanes Peninsula lava (408672) from Skovgaard et al. (2001) gave a heavier δ18O of +5.15‰, very different from the more extreme value of +5.62‰ determined by Skovgaard et al. (2001), but within error of the revised value for this sample of +5.08‰ measured by Thirlwall et al. (2006). An analysis of olivine from an enriched Stapafell sample gave δ18O of 4.77‰, consistent with the values of +4.77 to +4.83‰ determined on a different sample from Stapafell by Thirlwall et al. (2006). Plagioclase in three of the historic lavas have δ18Opl of +5.03 to +5.13‰, while replicate analyses of another sample (456714) have higher δ18Opl of +5.40 and +5.58‰. These values are consistent with the published range for other Reykjanes Peninsula lavas of +4.51 to +5.66‰ (+5.14 ± 0.29‰, 1 SD, n = 23: Thirlwall et al. 2006). For the three samples where both olivine and plagioclase were analysed, the isotopic fractionation for the olivine–plagioclase pairs (Δ18Opl–ol) are +0.7‰ (456742), +0.9‰ (456733) and ~+1.3‰ (456714). The value for sample 456714 is outside of the range of other Icelandic basalts (Δ18Opl–ol = 0.58 ± 0.24 and 0.65 ± 0.28‰) as determined by Thirlwall et al. (2006) and Eiler et al. (2000), respectively, which indicates disequilibrium between these two phases, most likely due to xenocrystic or antecrystic plagioclase. Plagioclase in these samples have compositions mostly between An75 and An85, whereas sample 456714 is the only sample analysed for oxygen isotopes that has a significant population of crystals with higher An contents of An87 to An91 (Fig. 2b).

Discussion

Evidence for two-component mixing in the historic Reykjanes Peninsula lavas

The virtually aphyric nature of the historic lavas means that their compositions approach a true liquid composition. Compositional data demonstrate that the historic lavas do not form a simple cogenetic series of samples that lie on a single liquid line of descent related by differences in the extent of fractional crystallisation. This is apparent from the range in TiO2 and K2O contents at a given MgO content (Fig. 3) and from the range in incompatible trace element ratios such as Nb/Zr and La/Yb (Fig. 4). Given the limited variation in 87Sr/86Sr (0.70309–0.70314) and εNd (+7.3 to +7.8), it is tempting to explain the range in incompatible trace element ratios simply through differences in the melt generation process involving a homogeneous mantle source, but this can be dismissed because of the significant Pb isotope variations revealed with the high precision (double-spike) data (Fig. 6).

A striking feature of the historic lavas are the coherent linear trends shown by high-precision Pb isotope data. All samples plot within analytical uncertainty (±150 ppm external reproducibility) of a least-squares linear regression line on plots of 206Pb/204Pb versus 207Pb/204Pb and 208Pb/204Pb (Fig. 6b, d), except for sample 456723 which lies about 300 ppm off the 206Pb/204Pb versus 208Pb/204Pb trend line. Similar trends in high-precision Pb isotope data have been reported in other ocean island settings (e.g. Hawaii—Abouchami et al. 2000; Eisele et al. 2003: Pitcairn—Eisele et al. 2002: Samoa—Workman et al. 2004), and in each case these trends have been interpreted as binary mixing arrays between compositionally distinct components in the mantle rather than as mantle isochrons, based on correlations with other isotope and elemental ratios. According to Abouchami et al. (2000), whether or not linear Pb isotope arrays represent isochrons or mixing lines can be evaluated by comparing κ (232Th/238U inferred from Pb isotopes) with κ* (232Th/238U calculated from Th and U contents): if κ and κ* are significantly different, then the Pb isotope arrays are unlikely to be isochrons. For the historic Reykjanes Peninsula lavas, κ and κ* are calculated using the procedure outlined by Abouchami et al. (2000). κ is calculated to be ~2.3 from the slopes of the arrays on the 206Pb/204Pb–207Pb/204Pb and 206Pb/204Pb–208Pb/204Pb plots, while κ* = 3.12 ± 0.02 based on six isotope dilution Th and U concentration measurements (Peate et al. 2001; Kokfelt et al. 2003), and this difference is more consistent with a binary mixing origin.

The observed range in 143Nd/144Nd for historic Reykjanes Peninsula magmas is 0.000017 (0.513014–0.513031: Fig. 5e), which is about two times the typical external precision of the 143Nd/144Nd measurement (2 SD of 0.000010: e.g. Thirlwall et al. 2004). This compares with the range in 206Pb/204Pb of 0.11 (18.771–18.881: Fig. 6a), which is greater than 40 times the typical external precision of the double-spike 206Pb/204Pb measurement (2 SD of <0.003: e.g. Thirlwall et al. 2004; Baker et al. 2004). This highlights the advantages of high-precision Pb isotope data to resolve evidence for fine-scale mixing processes (see also Abouchami et al. 2000).

On a plot of La/YbN versus Dy/YbN (where N indicates chondrite normalised ratio: Fig. 7a), binary mixing will produce a linear array, and the historic lavas all plot within the analytical reproducibility (estimated at <±1.5%, 2 SD) of a linear regression line through the data. This contrasts with the strongly curved arrays expected for melting: for example, Fig. 7d shows the hyperbolic trajectory of instantaneous fractional melt compositions from the melting model of Maclennan et al. (2003a) for rift zone tholeiites in the Northern Volcanic Zone of Iceland. The broad correlation of incompatible element ratios (e.g. La/YbN) with Pb isotope ratios (Fig. 7c) provides additional evidence for binary mixing between a ‘depleted’ end-member with 206Pb/204Pb <18.7 and La/YbN <1.3 and an ‘enriched’ end-member with 206Pb/204Pb > 18.9 and La/YbN > 2.7.

Fig. 7
figure 7

a La/YbN versus Dy/YbN, b La/YbN versus 1/Yb, c La/YbN versus 206Pb/204Pb, for the historic Reykjanes Peninsula lavas. Dotted lines in (a) and (c) are least-squares linear regression lines through the data. The absence of a linear trends shown by the data in (b) indicates that variable extents of crystallisation occurred subsequent to mixing. (d) plot showing the strong curvature of melting models for La/YbN versus Dy/YbN; the solid grey line shows the path of instantaneous fractional melt compositions from the melting model of Maclennan et al. (2003a), and compositions that can be generated by mixing of the fractional melts will lie between the solid and dotted grey lines

Full size image

If the linear array on the La/YbN versus Dy/YbN plot was the result of magma mixing just prior to eruption, then the data should also form a linear array on a plot of La/YbN versus 1/Yb. However, Fig. 7b shows that this is not the case, and in general the variations of incompatible element ratios and Pb isotope ratios do not correlate with elemental abundances in the historic lavas. This implies that the mixing must have occurred prior to variable extents of fractional crystallisation. Additional information as to the timing of mixing relative to crystallisation can potentially be obtained from melt inclusions in olivines. Olivine-hosted melt inclusions in samples from elsewhere in Iceland show a wide compositional range, attesting to the presence of compositionally diverse melts concomitant with olivine growth (e.g. Gurenko and Chaussidon 1995; Maclennan et al. 2003b). Furthermore, Maclennan et al. (2003b) showed that for the Borgarhraun flow in NE Iceland, the variability of REE compositions in the melt inclusions decreased as the Fo content of the host olivine crystal decreased, consistent with progressive mixing taking place concurrently with the early stages of crystallisation.

Figure 7d shows that there is a large compositional diversity of melts potentially available within a melt column from mixing between different instantaneous melts, and yet spatially and temporally restricted suites of basaltic lavas, like the historic Reykjanes Peninsula lavas, often show evidence for binary mixing between two compositionally restricted end-members. Maclennan (2008) has argued that this might be a natural consequence of the melt transport network involving channelised porous flow as envisaged by Spiegelman and Kelemen (2003). Melt travelling in the channel centres will be relatively ‘enriched’ in incompatible trace elements and generated over the full depth range of the melt column, while melts in the channel margins are generated in the shallowest mantle and have strongly ‘depleted’ incompatible trace element contents. Mixing between the channel centre and channel margin melts in magma bodies at the top of the melt column can potentially lead to the binary mixing arrays seen in both the whole rock samples and in olivine-hosted melt inclusions.

Fractional crystallisation within the historic Reykjanes Peninsula lavas

Jakobsson et al. (1978) summarised the main petrographic features of the post-glacial Reykjanes tholeiitic lava shield and fissure lavas. For lavas with FeO*/MgO <~1.4, olivine is the sole phenocryst phase, while lavas with FeO*/MgO >~1.4 have both olivine and plagioclase phenocrysts, with minor amounts of clinopyroxene. The historic lavas conform to this pattern, as all the flows (except the Svinahraun flows) contain both olivine and plagioclase phenocrysts. The Svinahraun flows, which belong to the ‘more enriched’ group from the Brennisteinsfjöll volcanic system, have the lowest FeO*/MgO (~1.3) and contain only olivine phenocrysts. The effects of plagioclase crystallisation are apparent as resolvable negative Eu-anomalies (Fig. 4) in samples with FeO*/MgO >1.6 (equivalent to MgO <7.5 wt%). These evolved samples have Eu/Eu* values of 0.94–0.98, while the more primitive higher MgO and lower FeO*/MgO samples have Eu/Eu* values of 0.98–1.03. Crystallisation temperatures were estimated using the olivine-liquid thermometer of Putirka (1997). Not all of the analysed olivine crystals are in equilibrium with their host melt, based on Fe–Mg partitioning (Fig. 2b), and so only those olivines that had measured K D(Fe–Mg) within the expected range for samples of these lava compositions (0.25–0.36: Putirka 2005) were used for temperature calculations. The data indicate that olivine crystallisation in the historic lavas took place at temperatures between ~1,180°C and ~1,200°C, ~15°C lower than the experimentally determined 1 atm. liquidus (Fisk et al. 1980), which is within the ±31°C calibration error estimated by Putirka (1997) for this thermometer.

Although geobarometers such as Putirka et al. (1996) and Nimis and Ulmer (1998) can potentially provide estimates of the pressure of clinopyroxene crystallisation, their applicability is limited because clinopyroxene is not a common phenocryst phase in the historic Reykjanes Peninsula lavas. Clinopyroxene was only analysed in two samples of the ‘less enriched’ group. The Putirka et al. (1996) model requires equilibrium melt-clinopyroxene pairs, and this can be assessed using Fe–Mg partitioning. The equations in Putirka et al. (1996) were used to calculate expected K D(Fe–Mg) values of 0.33 ± 0.01 for these two melt compositions. Disequilibrium between these clinopyroxenes and their host melts is indicated by the significantly lower measured K D(Fe–Mg) values (0.24 ± 0.02). However, the Nimis and Ulmer (1998) geobarometer estimates pressure from just the clinopyroxene composition: sample 456746 gives pressures of 0.21 ± 0.05 GPa and sample 456714 gives pressures of 0.53 ± 0.10 GPa. These are equivalent to crystallisation depths of ~7 ± 2 and ~18 ± 4 km, respectively.

Pressures of crystallisation can also be estimated by comparing sample compositions with experimentally determined ol–plag–cpx cotectic melt compositions. This requires that sample analyses represent liquid compositions, which is a reasonable assumption given the virtually aphyric nature of the historic flows, and also that the melts were in equilibrium with olivine + plagioclase + clinopyroxene. This latter requirement is more difficult to establish unequivocally, given the limited range in MgO within each volcanic system (Fig. 5). The ‘less enriched’ flows show a general trend of decreasing CaO and Ca/Al with decreasing MgO, consistent with cotectic crystallisation. Although clinopyroxene was only found in trace amounts in flows of the ‘less enriched’ group, the absence of more abundant cpx phenocrysts in these flows may be due to resorption during ascent. The model of Kelley and Barton (2008) was used to estimate pressures of crystallisation (to ±0.1 GPa accuracy) for the historic flows, grouped according to volcanic system (Fig. 1): Reykjanes volcanic system, 0.34–0.39 GPa (n = 2); Svartsengi volcanic system, 0.46–0.50 GPa (n = 3); Krisuvik volcanic system, 0.33–0.37 GPa (n = 7); Brennisteinsfjöll volcanic system (excluding the Svinahraun flows), 0.60–0.72 GPa (n = 7). For the ‘less enriched’ group, flows from the Reykjanes and Krisuvik volcanic systems have similar crystallisation depths of ~11–14 km, while the flows from the Svartsengi volcanic system give deeper calculated crystallisation depths of ~16–18 km. The crustal thickness along the Reykjanes Peninsula is 15–18 km (Weir et al. 2001), and so these data, together with the limited clinopyroxene geobarometry suggest that crystallisation in the historic lavas has taken place over a range of depths within the crust. This is consistent with observations at other locations in the Reykjanes Peninsula and elsewhere is Iceland (see summary in Kelley and Barton 2008). The sub-Moho crystallisation depths (~21–26 km) inferred for the ‘more enriched’ lavas from the Brennisteinsfjöll volcanic system are more likely an artifact of these samples crystallising just ol + plag rather than the cotectic ol–plag–cpx assemblage, as they show a trend of inceasing CaO and Ca/Al with decreasing MgO.

The ‘less enriched’ group (except Stampahraun) has a limited range in incompatible element ratios (e.g. Nb/Zr Figs. 4b, 5e) and Pb isotope ratios (Figs. 5d, 6), and so they might be expected to fall on a simple liquid line of descent as MgO decreases from 7.7 to 6.5 wt%, but this is not the case. The lavas from the Krisuvik Fires episode (1151–1188 ad), together with the Eldborgarhraun and Afstapahraun flows from the same volcanic system (that have slightly younger inferred ages ~1325 ad), have remarkably similar compositions and share some distinctive characteristics that distinguish them from the other flows in the ‘less enriched’ group, such as high Ca/Al. Although samples from the adjacent Svartsengi volcanic system have more evolved compositions, they have similar or higher Cr values (124–175 ppm) compared to the flows in the Krisuvik volcanic system (111–149 ppm), but lower Ni contents (65–74 vs. 81–89 ppm). Furthermore, there are small but significant differences in Ba/Nb, two highly incompatible trace elements (Svartsengi volcanic system 4.6–5.0 vs. Krisuvik volcanic system 5.9–6.5). Therefore, the history of mantle-derived melts of similar composition must vary in detail between the different volcanic systems along the length of the Reykjanes Peninsula, as a result of subtle differences in the melt transport regime, depths of crystallisation within the crust, and in the extent of interaction with crustal materials (assimilation and melt percolation through gabbroic cumulates: e.g. Gurenko and Sobolev 2006).

‘Enriched’ and ‘depleted’ mantle components on the Reykjanes Peninsula

Recent (<20 ka) magmatism on the Reykjanes Peninsula shows a far greater compositional diversity compared to the historic lavas (Figs. 3, 4b, 6). This variability is shown in petrography, major and trace elements, and isotopes (e.g. Jakobsson et al. 1978; Zindler et al. 1979; Elliott et al. 1991; Gee et al. 1998b; Thirlwall et al. 2004, 2006; Kokfelt et al. 2006). Several authors have suggested that a dynamic melting model could account for much of the observed compositional diversity of melt compositions, at length scales ranging from whole rock samples (e.g. Wood 1981; Elliott et al. 1991) down to olivine-hosted melt inclusions (Gurenko and Chaussidon 1995). Deep melts derived from the base of the melting column will have high Fe and will be enriched in incompatible trace elements. In contrast, shallow melts from the top of the melting column will have low Fe and they will also be depleted in incompatible trace elements as these elements will have been largely removed from the mantle at deeper levels in the melting column. However, the range in Sr–Nd–Pb isotopic compositions in the lavas requires a compositionally heterogeneous source in which the most incompatible element depleted samples (e.g. Nb/Zr ~0.01; La/YbN ~0.2) are derived from a time-integrated depleted source (high 143Nd/144Nd ~0.51316). Therefore, a model involving polybaric melting of a heterogeneous mantle comprised of enriched streaks set in a refractory depleted matrix is the preferred solution of many workers (e.g. Chauvel and Hémond 2000; Fitton et al. 2003; Stracke et al. 2003; Kokfelt et al. 2006). This is consistent with recent olivine-hosted melt inclusion data that preserve, in a single lava sample, the full Pb isotopic range of all Reykjanes Peninsula lavas (Maclennan 2008).

The isotopic variability of recent (<20 ka) Reykjanes Peninsula lavas can be broadly explained by mixing between a ‘depleted’ component with 87Sr/86Sr ~0.70286, 143Nd/144Nd ~0.51316 and 206Pb/204Pb ~18.25, and an ‘enriched’ component with 87Sr/86Sr ~0.70326, 143Nd/144Nd ~0.51296 and 206Pb/204Pb ~18.95 (e.g. Stecher et al. 1986; Gee et al. 1998b; Thirlwall et al. 2004, 2006; this paper). The incompatible element concentrations of the two end-members differ by about an order of magnitude (e.g. Fig. 4b; Gee et al. 1998b), and this marked difference can explain the lack of variation in the Sr and Nd isotope composition of the historic lavas. Simple binary mixing models (Langmuir et al. 1978), using samples from Thirlwall et al. (2004) as representative of the two end-members (‘depleted’ RP95C; ‘enriched’ RP100C), indicate that the restricted range in the Sr and Nd isotope compositions of historic lavas can in fact accommodate a broad range of mixtures involving between 50 and 90% of the ‘enriched’ end-member. The ‘depleted’ end-member is only found in a relatively undiluted form in the small-volume picrites erupted during the last deglaciation (9–13 ka: e.g. Jakobsson et al. 1978).

Post-glacial magmatism on the Reykjanes Peninsula is actually dominated by lavas with ‘enriched’ compositions (i.e. Nb/Zr >0.08; La/YbN >1.0) similar to the historic lavas. For example, Thirlwall et al. (2006) noted that the average Nb/Zr was 0.130 for their suite of 253 samples. Jakobsson et al. (1978) estimated the volumes of the various eruptive units found in the western half of the Reykjanes Peninsula and showed that the ‘depleted’ picritic lavas are volumetrically negligible (less than 2% of the total volume). The two main lava shields (Sandfellshaed and Thrainsskjoldur: Fig. 1), which were erupted shortly after the deglaciation interval that produced the ‘depleted’ picritic and basaltic eruptions, comprise about two-thirds of total volume of post-glacial magmas erupted on the Reykjanes Peninsula and these lavas have ‘enriched’ compositions (Fig. 3: Jakobsson et al. 1978), with La/YbN >1.3 (Zindler et al. 1979).

An intriguing feature of the historic Reykjanes Peninsula lavas is the systematic geographical compositional variation (Fig. 5). Historic lavas erupted from individual volcanic systems have broadly similar compositions, but differ from those in the adjacent volcanic systems. The main geographical trend is one of decreasing trace element enrichment, best shown by the east to west decrease in Nb/Zr that continues in the offshore historic lavas of the northern Reykjanes Ridge (Fig. 5e). While it is tempting to attribute this to a systematic spatial compositional variation in the mantle source beneath the Reykjanes Peninsula, it is important to remember that the historic lavas represent a relatively short time period of magmatic activity (~400 years). In fact, over the last 9,000 years, each of the volcanic systems has erupted lavas that span the full range in Nb/Zr seen in the historic lavas (e.g. Gee et al. 2000). Therefore, the lateral gradient in Nb/Zr shown by the historic lavas along the Reykjanes Peninsula is more likely to be just a coincidence, with the Nb/Zr variations reflecting a complex interplay between the scale of mantle compositional heterogeneity, the footprint of the melt generation zone, and the progressive tapping of melts from this zone. However, the trend of decreasing SiO2 from west to east is part of a longer wavelength variations that continues offshore to the south-west along the northern Rekjanes Ridge and to the north-east along the Western Volcanic Zone (e.g. Jakobsson et al. 2000). SiO2 contents of primitive mafic melts are influenced by the depth of melting as well as source mineralogy. The gradual lateral change in SiO2 therefore might simply reflect an increase in the average depth of melting in response to elevated mantle temperatures and/or more active plume-driven upwelling closer to the center of Iceland, consistent with U-series disequilibrium data (Peate et al. 2001; Kokfelt et al. 2003).

Pb isotope variations in post-glacial (<13 ka) Reykjanes Peninsula lavas

Thirlwall et al. (2004) presented a comprehensive study of high precision Sr–Nd–Pb isotope variations in volcanic rocks from the main rift zones of Iceland and in the adjacent mid-ocean ridges, and they proposed that these data could be explained by the presence of four major compositionally distinct components within the Icelandic mantle (ID1, ID2, IE1, IE2). They noted that samples from geographically restricted locations showed tight isotopic correlations consistent with binary mixing, and that in each region the four mantle components can be mixed in various proportions to provide these local mixing end-members. In their model, the ‘enriched’ component of the Reykjanes Peninsula is composed of a well-blended mixture of IE1 + IE2 and this mixture then variably mixes with the ‘depleted’ component ID1 to form the observed data array (Fig. 8). A minor influence from the ID2 component was required to account for the rare Reykjanes Peninsula samples with positive ∆207Pb.

Fig. 8
figure 8

Regional comparison of Pb isotope compositions of onshore Reykjanes Peninsula lavas and offshore Reykjanes Ridge lavas (this study; Thirlwall et al. 2004; Blichert-Toft et al. 2005). ∆207Pb and ∆208Pb are the vertical deviations × 100 of samples from the Northern Hemisphere Reference Line (after Hart 1984). Star symbols are the inferred mixing components that can explain the isotopic variations of Icelandic lavas (ID1, ID2, IE1, IE2) and the RRE component from Thirlwall et al. (2004)

Full size image

The new high-precision Pb isotope data (Table 3) for the Reykjanes Peninsula lavas are shown on Fig. 8, together with the high-precision Pb isotope data of Thirlwall et al. (2004), and the compositions of their four inferred mantle components are marked with star symbols. These figures highlight the good agreement between these two independent data sets. On the 208Pb/204Pb versus 206Pb/204Pb plot (Fig. 8b), the Reykjanes Peninsula data form a good linear trend that is consistent with an extrapolation to lower 206Pb/204Pb values of the trend shown by the historic lavas. On the 207Pb/204Pb versus 206Pb/204Pb plot (Fig. 8a), the Reykjanes Peninsula data trend becomes increasingly more scattered at low 206Pb/204Pb values in the trace-element-depleted picritic lavas. Despite the very low Pb contents in these samples (<0.1 ppm: Chauvel and Hémond 2000; Kokfelt et al. 2006; Nielsen et al. 2007), we do not consider the data to have been compromised by analytical issues. We have previously demonstrated the effectiveness of our leaching and analytical procedures to obtain reproducible Pb isotope data on samples with sub-0.2 ppm Pb contents (Baker et al. 2005). For the ‘depleted’ Reykjanes picrites, the compositional extremes are represented by Haleyjabunga and Lagafell. In each location, replicate whole rock samples of the same units give similar Pb isotope compositions, illustrating the robustness of these observations. In detail, the Haleyjabunga sample have significantly higher 207Pb/204Pb (~15.479) than the Lagafell samples (~15.467), despite their lower 206Pb/204Pb (18.30 vs. 18.49). This difference is highlighted on the 206Pb/204Pb versus ∆207Pb plot (Fig. 8c), where ∆207Pb represents the vertical deviation (×100) of a sample from the Northern Hemisphere Reference Line (Hart 1984). The Lagafell samples have negative ∆207Pb values similar to the historic lavas (−2 to −3), while the Haleyjabunga samples straddle the NHRL with ∆207Pb values of −1 to +1. Figure 8d shows that the Haleyjabunga samples and the Lagafell samples are also not colinear with the historic lavas on a 206Pb/204Pb versus ∆208Pb plot.

These isotopic differences are unlikely to be related to alteration and/or assimilation of crustal rocks as the Reykjanes crustal rocks are expected to have high 206Pb/204Pb (~18.8) and fall within the main array of the lavas (see next section). Instead, these data result from isotopic heterogeneity in the mantle and require the presence of at least two components with low 206Pb/204Pb, consistent with the conclusions of Thirlwall et al. (2004). Thirlwall et al. (2004) referred to the component with positive ∆207Pb and high ∆208Pb as ID2, and its inferred composition was based on replicate analyses of a sample (RP103A) from Kleifarvatn that had an unusual Pb isotope composition (low 206Pb/204Pb ~18.16, ∆207Pb ~3.4, ∆208Pb ~62) and enriched Sr–Nd isotope values (Fig. 6e). Based on Fig. 8, it seems that this component has a greater influence on the depleted Reykjanes Peninsula samples than has been previously recognised. The elevated 87Sr/86Sr value inferred for the ID2 component (>0.70345) might also explain the slightly higher Sr isotope composition of Haleyjabunga compared to Lagafell (0.70302 vs. 0.70289: Hémond et al. 1993; Kokfelt et al. 2006; Thirlwall et al. 2006), despite the lower 206Pb/204Pb values (see also Thirlwall et al. 2004).

A role for crustal assimilation?

An important issue to resolve is the extent to which shallow-level interaction with the Icelandic crust might have modified the composition of the lavas. Geothermal fluids and hydrothermally altered rocks from the Reykjanes Peninsula have higher 87Sr/86Sr (0.7037–0.7042: Elderfield and Greaves 1981) than unaltered lavas (<0.7033: Fig. 6e) from some seawater input, but 143Nd/144Nd values are likely to be similar to the average lava composition (e.g. 0.51303–0.51305: Hémond et al. 1993; Gee et al. 1998b) given the limited solubility of Nd in fluids. Therefore, assimilation of hydrothermally altered rocks should produce a shift to elevated 87Sr/86Sr values (Hémond et al. 1993; Gee et al. 1998b). The historic Reykjanes Peninsula lavas have Sr–Nd isotope compositions that plot within analytical error of a mixing curve between the ‘depleted’ and ‘enriched’ mantle components inferred by Stecher et al. (1986) and Gee et al. (1998b), and thus there is no evidence that assimilation of hydrothermally altered crust with elevated 87Sr/86Sr has noticeably affected the composition of the historic lavas (Fig. 6e). The scatter of published data for other earlier Reykjanes Peninsula lavas can be explained by the assimilation of 5–20% of hydrothermally altered crust by some samples (e.g. Gee et al. 1998b), although some scatter can result from incomplete sample leaching procedures (e.g. Thirlwall et al. 2004) or to the influence of the ID2 mantle component (Fig. 6e).

Crustal assimilation is expected to have a minimal effect on the Pb isotope compositions of the erupted lavas. The average 206Pb/204Pb ratio of the Reykanes Peninsula crust can be estimated as ~18.8 ± 0.1 from the broad correlation between trace element enrichment/depletion (e.g. Nb/Zr) and 206Pb/204Pb (Fig. 9), and the average Nb/Zr value of Reykjanes Peninsula lavas of 0.130 (Thirlwall et al. 2004). Its Pb isotope composition should lie within the observed data array of the lavas, because any hydrothermal systems will largely redistribute crustal Pb and the crust is too young for the Pb isotope composition to have significantly changed through radioactive decay. Any assimilation of hydrothermally altered rocks would move Pb isotope compositions along the main array and not perpendicular to it, and therefore could not explain the compositional differences between the different depleted picritic units (e.g. Lagafell vs. Haleyjabunga).

Fig. 9
figure 9

206Pb/204Pb versus Nb/Zr for Reykjanes Peninsula lavas. Data sources: Gee et al. (1998a); Skovgaard et al. (2001); Thirlwall et al. (2004), Nielsen et al. (2007); this study

Full size image

Iceland is notable for the abundance of lavas with δ18O values lighter than expected to be in equilibrium with the mantle (e.g. Muehlenbachs et al. 1974; Condomines et al. 1983; Sigmarsson et al. 1992; Hémond et al. 1993; Eiler et al. 2000; Thirlwall et al. 2006), and the historic Reykjanes Peninsula lavas are no exception (Fig. 6f). These lavas have δ18Oolivine values of +4.2 to +4.3‰, which are significantly lighter than mean lithospheric mantle olivine compositions (5.2 ± 0.3‰: Mattey et al. 1994). The origin of this 18O-depleted signature has been the subject of much debate in the literature over whether it is a feature of the mantle source or acquired through crustal assimilation. Meteoric waters on Iceland have very light δ18O due to the high-latitude location of the island, and the extensive hydrothermal systems in the rift zones means that the altered crustal rocks acquire a light δ18O signature: strongly altered Reykjanes Peninsula crustal rocks are estimated to have δ18O of +2‰ (e.g. Gee et al. 1998a). Bindeman et al. (2006) have suggested that assimilation of palagonitised hyaloclastic volcanic rocks altered by glacial meltwaters might be a suitable mechanism to account for the light δ18O values of many Icelandic basalts, such as the 1783 ad Laki eruption, and Pleistocene hyaloclastites near the Laki eruption site had δ18O of +2.8 to +3.3‰.

The majority of the Reykjanes Peninsula samples show broad correlations between radiogenic isotopes and oxygen isotopes (e.g. Fig. 6f). The ‘depleted’ flows with high 143Nd/144Nd, low 206Pb/204Pb and δ18O values similar to normal mantle (δ18Oolivine = +4.9‰ to +5.1‰) plot at one end, and the ‘enriched’ flows with low 143Nd/144Nd, high 206Pb/204Pb and lighter δ18O values plot at the other (Thirlwall et al. 2006). The historic flows anchor the ‘enriched’ end of the trend with δ18Oolivine of +4.2 to +4.3‰. Several studies have noted that olivines within individual samples of Icelandic basalts can have heterogeneous oxygen isotope compositions (e.g. Maclennan et al. 2003b; Bindeman et al. 2006). The Reykjanes Peninsula δ18Oolivine data plotted in Fig. 6b are analyses on ~2–10 grains, which will average out any potential within-sample heterogeneity. However, the broad trend shown between these averaged δ18Oolivine values and whole-rock 143Nd/144Nd data require that olivines with light δ18O are volumetrically more significant in the ‘enriched’ low 143Nd/144Nd flows. Two Reykjanes Peninsula samples have anomalously light δ18O (+3.4 to +3.8‰: Fig. 6f) and these samples also have elevated 87Sr/86Sr, consistent with some assimilation of altered crust (e.g. Gee et al. 1998b; Thirlwall et al. 2006).

Thirlwall et al. (2006) showed that some samples from the offshore Reykjanes Ridge also had light δ18Oolivine values (down to +4.5‰) and argued, based on curvature of Nd–O mixing arrays and normal Cl/K ratios, that these values represented mantle compositions. This strengthened their arguments that the light δ18O in samples from the Reykjanes Peninsula with similar Sr–Nd–Pb isotope compositions were also unaffected by crustal assimilation. Burnard and Harrison (2005) argued that glass samples with high 40Ar/36Ar ratios (>2,000) cannot have been affected by surface alteration or assimilation of altered crust as this would have introduced argon with an atmospheric composition (40Ar/36Ar = 296). They showed that two samples from Dagmalafell in the Hengill volcanic system of the Reykjanes Peninsula had light δ18Oglass values of +4.6‰ (equivalent to δ18Oolivine values of ~+4.3‰) coupled with high 40Ar/36Ar ratios (4,900–6,500), which indicates that the mantle source had light δ18Oolivine of ~+4.3‰. Thus, it is reasonable to suggest that the light δ18Oolivine of the historic lavas represent mantle values, although it is difficult to categorically rule out any role for assimilation. An assessment of within-sample δ18O heterogeneity through a detailed study of individual olivine crystals might help to resolve this issue (e.g. Bindeman et al. 2006). Another approach would be to look at melt inclusions in olivine and plagioclase crystals to see if there is any evidence for elevated chlorine abundances that should be an unequivocal tracer for the assimilation of crustal lithologies hydrothermally altered by fluids of seawater origin (e.g. Kent et al. 1999). While hydrothermal systems across Iceland are mostly dominated by meteoric water, the active systems in the south-eastern part of the Reykjanes Peninsula (e.g. Reykjanes, Svartsengi) are dominated by seawater flow that should introduce significant quantities of chlorine into the altered crustal lithologies (e.g. Kadko et al. 2007).

Comparison of onshore Reykjanes Peninsula lavas and offshore Reykjanes Ridge lavas

Figure 8 shows a comparison of the Pb isotopic variations in the Reykjanes Peninsula lavas with those in the offshore Reykjanes Ridge mid-ocean ridge lavas (data from Thirlwall et al. 2004; Blichert-Toft et al. 2005; this paper). Both groups of samples define reasonably linear arrays on a plot of 206Pb/204Pb versus 208Pb/204Pb (Fig. 8b), but the Reykjanes Ridge samples have a steeper slope than the Reykjanes Peninsula lavas. This means that at low 206Pb/204Pb values, the Reykjanes Ridge samples have lower 208Pb/204Pb values than the Reykjanes Peninsula samples (consistent with the observations of Thirlwall et al. 2004). The radiogenic end of the Reykjanes Ridge array intersects the Reykjanes Peninsula array at an intermediate point that Thirlwall et al. (2004) defined as the Reykjanes Ridge Enriched (RRE) component. This is more clearly illustrated on the plot of 206Pb/204Pb versus ∆208Pb (Fig. 8d). The fact that the RRE component lies at a point in the middle of the Reykjanes Peninsula data field led Thirlwall et al. (2004) to argue that the two local mixing end-members responsible for the Reykjanes Peninsula trend must have been well-mixed prior to their advection down the ridge and furthermore that this must have occurred at mantle depths.

Figure 8d shows that ∆208Pb can only be used to clearly distinguish Reykjanes Peninsula and Reykjanes Ridge compositions that have 206Pb/204Pb values less than ~18.6. Moving south down the Reykjanes Ridge, the first samples that have Pb isotope compositions that are resolvable from the Reykjanes Peninsula lavas are found south of 62.4°N, with 206Pb/204Pb <18.6 and Nb/Zr <0.08. The ridge section between 62.4°N and the Iceland coast at 63.8°N, a distance of ~200 km, contains samples with compositions transitional between the historic Reykjanes Peninsula lavas and true Reykjanes Ridge compositions (206Pb/204Pb 18.6–18.7, Nb/Zr 0.08–0.10), although there is one sample (186D2: Thirlwall et al. 2004) that is similar in composition to the onshore historic lavas (206Pb/204Pb 18.83, Nb/Zr 0.124). This poorly sampled region clearly represents an important transition zone.

On the Reykjanes Peninsula, samples with 206Pb/204Pb <18.6 and Pb isotope compositions distinct from the Reykjanes Ridge lavas all have ‘depleted’ compositions (Nb/Zr <0.08: Fig. 9). Gee et al. (1998a) showed that the eruption of lavas with Nb/Zr <0.08 and 206Pb/204Pb <18.6 was restricted to the interval between 9,000 and 13,000 years ago, a period in which the ice sheets that covered the Reykjanes Peninsula during the Weichselian glaciation disappeared. In most of the volcanic systems within Iceland, this deglaciation period has been recognised as a time of increased volcanic activity (e.g. Jakobsson et al. 1978; Gee et al. 1998a; Maclennan et al. 2002). Ice unloading affected both the state of stress within the crust and also the processes of melt generation within the mantle. Jull and McKenzie (1996) modelled how the unloading increased melt production through mantle decompression: the largest increase in melt production occurred at depths of about 60 km where the mantle had already been depleted by ~15% melting. Thus, the deglaciation interval should have been characterised by increased production and release of incompatible-element-depleted melts from the shallow, depleted mantle. Under normal conditions, such melts would have little compositional leverage and would be overwhelmed by incompatible-element-enriched melts derived either from melting of fertile enriched veins or streaks or from initial deep melting of the refractory peridotitic matrix. This is consistent with the modelling of platinum group element abundances in Icelandic lavas by Momme et al. (2003) who concluded that the ‘depleted’ picrites formed when unique tectonic conditions (i.e. glacial rebound) allowed the selective tapping of melts from the shallow, depleted central part of a triangular melting regime without contributions from low degree melts from the flanks and deeper levels of the melt generation zone.

It is worth noting that in the absence of the deglaciation event, ‘depleted’ melts with 206Pb/204Pb <18.6 would not be represented in the Reykjanes Peninsula lavas, except perhaps as melt inclusions trapped in early formed olivines (e.g. Gurenko and Chaussidon 1995; Maclennan 2008). Instead, the RRE component would be considered to be the end-member to both the Reykjanes Peninsula and Reykjanes Ridge compositional trends and thus an important component within the Icelandic mantle, despite its clearly hybrid origin. The eruption of ‘depleted’ picrites during the deglaciation interval represents a good example of how surface processes can influence the sampling of compositional heterogeneities within the mantle.

Summary

Historic eruptions of tholeiitic basalts (MgO 6.5–9.2 wt%) on the Reykjanes Peninsula, Iceland, took place in a short time period between c. 940 and c. 1340 ad. They have a restricted composition variation in terms of incompatible trace element ratios and isotope ratios compared to earlier Reykjanes Peninsula lavas. High-precision double-spike Pb isotope data show coherent linear trends and good correlations with incompatible trace element ratios, indicating the importance of binary mixing processes. Poor correlations of Pb isotopes and incompatible trace element ratios with element concentrations show that mixing occurred prior to extensive fractional crystallisation. Pb isotope data on the older post-glacial Reykjanes Peninsula magmatism extend to much lower 206Pb/204Pb values than in the historic lavas. These earlier trace-element-depleted samples show significant heterogeneity in 207Pb/204Pb and 208Pb/204Pb, which requires the involvement of at least three compositionally distinct components within the local mantle.