Earth and Planetary Science Letters 269 (2008) 387–397

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

On the recent bimodal magmatic processes and their rates in theTorfajökull–Veidivötn area, Iceland

G.F. Zellmer a,b,⁎, K.H. Rubin a, K. Grönvold c, Z. Jurado-Chichay a

a Department of Geology and Geophysics, SOEST, University of Hawaii at Manoa, 1680 East West Road, Honolulu, HI 96822, USAb Institute of Earth Sciences, Academia Sinica, 128 Academia Road, Sec. 2, Nankang, Taipei 11529, Taiwan, ROCc Volcanological Institute, University of Iceland, IS-101, Reykjavik, Iceland

a r t i c l e i n f o

⁎ Corresponding author. Institute of Earth Sciences, ARoad, Sec. 2, Nankang, Taipei 11529, Taiwan, ROC. Tel.: +82 2783 9871.

E-mail address: gzellmer@earth.sinica.edu.tw (G.F. Z

0012-821X/$ – see front matter © 2008 Elsevier B.V. Aldoi:10.1016/j.epsl.2008.02.026

a b s t r a c t

Article history:Received 14 November 2007Received in revised form 13 February 2008Accepted 14 February 2008Available online 4 March 2008

Editor: R.W. Carlson

Historical bimodal composition eruptions spanning Torfajökull central volcano and neighbouring Veidivötnfissure swarm in southeastern Iceland provide clues about the nature and timing of basalt and rhyolitepetrogenesis and eruption at an active divergent plate boundary. This study focuses on lavas and tephras of thelast two regional eruptions in 871 and 1477 AD, using samples that approximate mafic and felsic endmembercompositions relative to regionalmixing trends in literature data from the area.Whole rock andmineralU–Th–Raisotopic compositions demonstrate both rapid petrogenetic timescales and limited compositional variation in thebasalt and rhyolite magma sources. Rhyolites display the greatest 230Th excesses (≤17%) at only slightly lower(230Th/232Th) activity ratios than co-eruptive Veidivötn basalts. Both magma types display small but significant226Raexcesses (≤ 10%and ≤ 60% in rhyolites andbasalts, respectively). U-series isotopic and trace element data areconsistent with crustal melting dominated by young (≤60 kyr), K-metasomatized mafic protoliths similar incomposition to the Veidivötn tholeiites as source of the 871 and 1477 rhyolites. Young protolith ages areinconsistentwithpublishedmodels that call onoldmaterials as source rocks for Icelandic rhyolites (e.g., old silicicsegregation lenses or isostatically subsided lavas). Zero-age U–Th and few-ka Ra–Th mineral-whole rockisochrons indicate that crystals formed shortly prior to eruption, consistent with petrographic and compositionalindicators that they are phenocrysts, and suggest that the rhyolite melts are of Late Holocene age.

© 2008 Elsevier B.V. All rights reserved.

Keywords:U-series disequilibriumpetrogenesisrhyolite volcanismrift zonecrustal melting

1. Introduction

1.1. Icelandic bimodal volcanism

The combined effects of divergent plate boundary and hot spotvolcanism have produced a thick basaltic crust in Iceland, contributingto one of few examples of extensive oceanic rhyolite volcanism.Icelandic volcanism tends to be compositionally bimodal, which raisesseveral fundamental questions about how evolved magmas arecreated from their mafic source compositions, the duration of theprocess, and why rhyolitic volcanism is common at some Icelandicvolcanoes yet not at others.

Hypotheses for silicic magma petrogenesis in Iceland include near-liquidus (crystal fractionation ± assimilation) and near-solidus (crustalmelting) processes, and proposed differentiation processes may varyaccording to the local tectonic setting. Multiple authors have favoredextensive (up to 90%) fractional crystallization of basaltic magmas(Carmichael, 1964; Macdonald et al., 1990; Lacasse et al., 2007),sometimes with concurrent crustal assimilation or contamination

cademia Sinica, 128 Academia86 2 2783 9910x602; fax: +886

ellmer).

l rights reserved.

(Nicholson et al., 1991) to form the rhyolites. However, while significantquantities of intermediate composition magmas erupt at some sites(especially atHekla, Sigvaldason,1974), they are generally rare in Iceland(e.g., Walker, 1966; Saemundsson, 1978). Dramatic shifts in magmatic18O in the most felsic Icelandic lavas imply meteorically altered crust intheir petrogenesis (Muehlenbachs et al., 1974). Hence, many previousauthors have suggested that rhyolites are predominantly generated bypartial crustalmelting of unaltered or hydrothermally altered old crustalbasalts (O'Nions and Grönvold, 1973; Sigvaldason, 1974; Hèmond et al.,1988; Sigmarssonet al.,1991; Jónasson,1994; Lacasse et al., 2007;Martinand Sigmarsson, 2007), or their silicic differentiation products (Sigurds-son, 1977; Gunnarsson et al., 1998).

Contemporaneous eruption of basalts and rhyolites has been aprominent manifestation of Icelandic bimodal volcanism duringpostglacial times. Since the 1975–1984 Krafla fires fissure eruption(e.g., Björnsson et al., 1977), lateral magma flow between centralvolcanoes and neighbouring fissures swarms has been considered acommon attribute of eruptions within “volcanic systems” (e.g.,Saemundsson, 1978) along the divergent plate boundary. Examplesof rhyolite-producing central volcanoes with associated basalticfissure eruptions include Askja volcano and Svenagja fissures(Sigurdsson and Sparks, 1978a,b), Grímsvötn volcano and Laki fissures(cf. Bindeman et al., 2006), and Katla volcano and Eldgjá fissures (cf.Lacasse et al., 2007). Basaltic magma injection as a trigger of the 1875

388 G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

Askja plinian rhyolite eruption was explored by Sparks et al. (1977,1981). A similar mechanism was later proposed for the rhyoliticeruptions at Torfajökull central volcano and nearby Veidivötn basalticfissures eruptions, with the twist that basaltic magmas erupted atVeidivötn are thought to be sourced from Bárðarbunga, a secondcentral volcano some 100 km to the north (Blake, 1984; Larsen, 1984;McGarvie, 1984; McGarvie et al., 1990).

Icelandic volcanoes tend to be either tholeiitic or alkalic, dependingon their occurrence within or outside of the primary active rift zones(e.g., Jakobsson, 1979). Postglacial eruptions from Torfajökull haveincluded low and high-K basalts, and high-K rhyolites. The last two

rhyolites (AD 871 and 1477) were erupted simultaneous withVeidivötn tholeiitic basalts along single, continuousfissures, indicatingthat the magma plumbing systems of the Torfajökull and Veidivötnareas are hydraulically and tectonically linked. These eruptionstherefore provide an excellent opportunity to examine the processesand time scales at the heart of bimodal magmatism in Iceland.

1.2. U-series disequilibria

U-series isotopic studies constrain the timing and conditions ofpetrogenesis, making them useful for tests of proposed rhyolitegenerating processes. U-series nuclides are sensitive to the time sincechemical fractionation, so they can be used to make inferences aboutmagma ages upon eruption and to distinguish between recentlymodified and long-lived chemically closed magma sources. Further, thepartitioning behavior of U-series isotopes is strongly dependent on thetypes and proportions of minerals present, and the dynamics of meltinganddifferentiationprocesses. Early studies demonstrated theutilityofU-series isotopes in Icelandic petrogenetic processes (Condomines et al.,1981) and suggested that U-series variations in Icelandic basalts resultedlargely from recent assimilation of partial melts of old hydrothermallyaltered crust (Hèmond et al., 1988). It was later suggested that mixing ofmelts derived from undepleted off-axis mantle and depleted rift zonemantlewas important in off-rift settings (e.g., theSnaefellsnes peninsula,Sigmarsson et al.,1992b). CombinedRa–Thandoxygen isotope data havebeen used to infer that when involved, bulk-digestion of hydratedbasaltic crust can be rapid (b8 ka), such as prior to the 1783–1784 Lakibasaltic fissure eruption (Bindeman et al., 2006). There have been fewprior U-series disequilibria studies of siliceous Icelandicmagmas, limitedto small data sets fromAskja, Krafla andHekla usedwith otherdata typesto argue for a role of crustal melting of hydrothermally altered basalts(Nicholson et al., 1991; Sigmarsson et al., 1991; 1992a).

Finally, the Icelandic plume plays a role in divergent plateboundary volcanism by producing a systematic excess 230Th dropand correlated variations in Pb–Nd–Sr–O isotopes and trace elementratios towards the hot spot center as a result of faster mantleupwelling rates and greater degrees of melting of a more hetero-geneous mantle source (Kokfelt et al., 2003, 2006). The same scenariohas been invoked to explain general isotope variations in Hawaii (Simset al., 1999) as well as temporal fluctuations at Kilauea, wherechanging melting conditions over the past 200 yrs have been detectedby sampling variably wide regions of the upwelling Hawaiian hotspotsource (Pietruszka et al., 2001).

2. Geological background

The geology of the active volcanic rift zone in South CentralIceland, which includes both Torfajökull and Veidivötn, is summarizedhere on basis of previous studies (Jakobsson,1979; Blake,1984; Larsen,1984; Macdonald et al., 1990; Gunnarsson et al., 1998; Sigmarssonet al., 2000). Torfajökull central volcano is situated on the northernedge of the practically non-rifting Southern Flank Zone, southwest of

Fig. 1. Outline geological map of the Torfajökull–Veidivötn area, based on Vilmundardóttiret al. (1988), Jóhannesson et al. (1990) and Ívarsson (1992), and its positionwith respect tothemain rift zones in Iceland. Sampling locations are indicated by stars. Veidivötn samplesT1, T2, T4, T5, and T6were taken, respectively, fromHravnotn, Breidavatn, Vatnaoldur Inri,Vatnakvisl Creek, andVatnaoldur Fremri. Place name abbreviations for Torfajökull samplesfrom this study and the literature are adopted from Macdonald et al. (1990), and are thefollowing: For the 1477 AD eruption: FRO — Frostastadahraun, LGH — Laugahraun, LJO —

Ljótipollur, NMS—Namshraun, and STU— Stútshraunbasaltflow. For the871ADeruption:BHY— Bláhylur basalt explosion crater, and HRN— Hrafntinnuhraun. For older eruptions:DOM(3) — Domadalshraun 3 (~150 AD); DOM(2) — Domadalshraun 2 (~3100 BP); MKD —

Markarfljöt domes (~3500 BP); HAA — Haölduhraun lava flow (~6500 BP); LAF —

Laufahraunmafic flow, LFD— Laufafell domes, and LFL— Laufafell basalt lavas (~6800 BP);DOM(1) — Domadalshraun 1 (~7000 BP); HSK — Hrafntinnusker (~7500 BP); SLE —

Slettahraun lava flow (~8000 BP). SeeMcGarvie (1985) for a discussion of the evidence for,and accuracy of, these ages. Deposits from BLY, LJO, FRO, STU, NMS and LGH are part of thenortheast Torfajökull group, as discussed in the text.

Fig. 2. Selected EM images of the most recent Veidivötn basalts and associatedTorfajökull rhyolites, with 200 μm scale bars. Insets are from the same sample and to thesame scale as the respective main images.

389G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

the Veidivötn fissure swarm. A simplified geological map of the region(Fig. 1) shows that the fissure swarm occupies roughly 85×10 km2,forming the southern part of the actively rifting Eastern Rift Zone. Ithas erupted tholeiitic basalts of a relatively narrow compositionalrange with a 600–800 year recurrence interval during the historicaland post-glacial eras. In the extreme southwest of the Eastern RiftZone, the Veidivötn fissure swarm intersects Torfajökull centralvolcano, a dominantly rhyolitic plateau rising ~600 m above thesurrounding basaltic terrain and representing both the largest volumeof exposed silicic extrusives and the most productive geothermalsystem in Iceland. The bulk of the plateau is composed of last glaciallavas and tephras. An old ring structure has been identified withinthese deposits, suggested by Saemundsson (1972) to mark the locus ofa caldera complex (Fig. 1). Postglacial silicic lavas with alkalineaffinities have erupted bothwithin and outside this ring structure, andare associated with northeast–southwest trending faults and fissures.Mafic magmas have also erupted along these fissures, predominantlyas transitional alkali basalts in the southwest and as tholeiites in theVeidivötn fissure swarm to the northeast. Direct observations, ashdeposition patterns, and geochronological evidence indicate that thelast three eruptions of tholeiitic basalts from Veidivötn and rhyolitesfrom Torfajökull occurred simultaneously (Larsen, 1984). Abundantevidence of pre- and syn-eruptive mafic and felsic magma mixingoccurs at the intersection of the Veidivötn fissures and the Torfajökullring structure, suggesting that some postglacial Torfajökull rhyoliteeruptions have been triggered by influx of mafic melts. The differencebetween basalt compositions erupted within and to the northeast ofthe Torfajökull ring structure is one reason why various authors havesuggested that the Veidivötn magmas are associated with the volcanicsystem of the compositionally similar Bárðarbunga central volcano,situated about 100 km to the northeast.

This study addresses potential petrogenetic relationships betweenbasaltic and silicic magmas erupted in the Torfajökull region, with afocus on the last two historical eruptions. On basis of geochemical andgeographical patterns, deposits from the 871 AD and 1477 eruptions,which with exception of the 871 AD Hrafntinnuhraun flow (HRN inFig. 1) occurred in what Macdonald et al. (1990) refer to as northeastTorfajökull, are distinguished here from recent, undated eruptives ofwhat will be referred to as “west Torfajökull”, outside and west of thering structure (i.e. LAF, LFD, LFL, MKD; cf. Fig. 1 caption for a key tolocality abbreviations), including associated fissures of transitionalalkali basaltic and intermediate magmas.

3. Methods

Samples were collected using eruption deposit maps of Larsen (1984),with theaimtoavoidmixed-magmasamples. Informationabout samplinglocations, mineral separation techniques, sample preparation for geo-chemical analyses, and analytical methodologies (including analyses ofstandards) are given in the supplementary data in the Appendix.

Whole rockswere analyzed formajor and trace element abundancesby XRF, and for additional trace elements by ICP-MS at the University ofHawaii. Electron microprobe analyses of mineral compositions wereconducted at the Instituto de Geofisica, UNAM, in Mexico City.

U-series and radiogenic isotope analysis methods in the SOESTIsotope Lab are summarized in the appendix, with full discussion in(Pietruszka et al., 2001; Rubin, 2001; Rubin et al., 2005). Highabundance sensitivity ion-counting thermal ionization mass spectro-metry of all U-series nuclides used a Sector54-warp.

4. Results

4.1. Petrographic observations

The products of both historical eruptions are petrographically verysimilar at the thin section scale. Most phenocrysts are euhedral or

subhedral, without significant zoning under the microscope or inmicroprobe transects (see Fig. 2 electron backscatter images).Veidivötn basalts typically contain Ca-rich plagioclase and pyroxenephenocrysts plus olivine microlites set in a fine-grained groundmassdominated by feldspar, pyroxene and accessory magnetite. Torfajökullrhyolites have plagioclase and pyroxene phenocrysts plus magnetitemicrolites set in a glassy to fine crystalline groundmass dominated byfeldspar microlites. Titaniferous amphibole microphenocrysts andapatite microlites occur as accessory phases in the 871 AD rhyolites.

4.2. Comparative mineral chemistry

Veidivötn basalts from 871 AD have Mg-rich olivines (Fo74–86).Clinopyroxenes and plagioclases display wider compositional ranges:En42–54Fs6–15Wo36–47 and An52–92 (Fig. S1a in the Appendix).

Fig. 3. Major and trace element variation diagrams of the Veidivötn and Torfajökull volcanic rocks with silica as differentiation index. All data are normalized to 100% totals. Thisstudy: (■) 871 AD eruption; (♦) 1477 eruption; solid black lines represent differentiation trends of the rhyolites. Literature data (Jakobsson, 1979; McGarvie, 1984; Mørk, 1984;McGarvie, 1985; Gunnarsson, 1987; Hansen and Grönvold, 2000): (×) Most recent eruptions of Veidivötn basalts and Torfajökull rhyolites ( ) Hybrids from the most recentTorfajökull eruptions; the oblong dashed field (3a) and dashed linear trend lines through these samples indicate mixing trends. (ο) Postglacial transitional alkali rocks of westTorfajökull; thick grey lines show trends displayed by thewest Torfajökull samples, but should not be interpreted as indicating continuous differentiation processes (cf. Gunnarsson etal., 1998). Open fields indicate the total extent of tholeiites within the Veidivötn and Torfajökull area, including older samples. Stippled arrows indicate the MELTS (Ghiorso and Sack,1995; Smith and Asimow, 2005) trends for 80% fractional crystallisation of basalt T4, and for 20% partial melting of basalt T2, at 1 kbar and with the following initial values: For T4:enriched in water, [H2O]=0.5 wt.%; T=1200°C (cf. Mørk, 1984); fO2

buffer=FMQ-1. For T2: enriched in alkalis, [Na2O]=2.6 wt.%, [K2O]=0.7 wt.%, and water, [H2O]=3.0 wt.%; T=800°C;fO2

buffer=FMQ+3. The T2 residual mineralogy is pyroxene, feldspar, spinel, and apatite. See text for discussion.

390 G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

Torfajökull 871 AD rhyolites have more iron-rich clinopyroxenes(En30–39Fs20–32Wo36–44); feldspar compositions range from andesine(An45) to anorthoclase (Or29) in two populations distinguished on aternary feldspar diagram. Geothermometry (Fuhrman and Lindsley,1988)suggests that the two populations formed in overlapping temperatureranges, with a slight extension (b50 °C) toward lower temperatures forthemuch less abundant potassic population (cf. Fig. S1a in the Appendix).

Minerals of the 1477 eruption have very similar compositions atslightly smaller overall range, perhaps reflecting the lower number ofanalyses (cf. Fig. S1b in the Appendix). Basalts have Mg-rich olivines(Fo73–82) and clinopyroxenes (En45–53Fs9–15Wo36–43), and calcicplagioclase feldspars (An60–87). A small population of clinopyroxenesis richer in iron, compositionally similar to those found in co-eruptiverhyolites (En36–40Fs20–25Wo36–42). Rhyolitic feldspar compositionsmimic those of the preceding eruption, ranging from andesine(An38) to anorthoclase (Or29), although the two crystal populationsare less distinct than in the 871 eruptives (cf. Fig. S1b in the Appendix).

It is noteworthy that most individual crystals within the rhyolitesare unzoned and chemically homogeneous, despite the overall rangesin crystal chemistry. Similarly, most crystals from the basalts areunzoned to slightly normally zoned. Crystals of both magma typesdisplay little if any signs of host melt disequilibria, in contrast toprevious studies describing extensive crystal-melt disequilibria in theTorfajökull–Veidivötn area (Blake,1984;Mørk,1984). This difference isdue to the focus of this study on endmember lava compositions.

4.3. Whole-rock major and trace element data

Whole rock major and trace element XRF data and trace and rareearth element (REE) ICPMS data appear in Tables 1 and 2 in theAppendix, respectively. Veidivötn basalts analyzed here show asmall compositional range of about 2 wt.% SiO2, mirrored byrelatively small variations in other elements and oxides except forAl2O3 and Sr (Fig. 3 and Fig. S2 in the Appendix). Previouslypublished data on mixed magmas and on fissure eruption basaltsfrom west Torfajökull have intermediate compositions (McGarvie,1985; Gunnarsson, 1987). Relatively low MgO, Ni and Cr of end-member basalts indicate that none of the samples studied here areprimary mantle melts. Torfajökull rhyolites display a larger compo-sitional range: increasing SiO2 is accompanied by increases in K2Oand Nb, decreases in Na2O, Sr, Zr, Hf and REE, and slight decreases inTiO2 and P2O5.

Chondrite-normalized REE abundances in the 871AD and 1477eruptions (Fig. 4) show LREE enrichment over HREE and negativeEu anomalies, both most pronounced in the rhyolites. These trends,which mimic those in Macdonald et al. (1990), are consistent withfeldspar fractionation of a LREE-enriched source. Basalts and rhyo-lites have very similar middle to heavy REE patterns, pointing tocompositional similarities of their respective source regions. Olderpostglacial eruptions of west Torfajökull likewise display overallsimilar REE patterns to the most recent Torfajökull eruptions studied

Fig. 4. Chondrite-normalized REE diagram for the most recent Veidivötn basalts andTorfajökull rhyolites, compared to previously published data from these eruptions(Macdonald et al., 1990) and to data from west Torfajökull (Gunnarsson et al., 1998).Normalization factors are from Sun and McDonough (1989). See text for discussion.

391G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

here, but develop comparatively high Yb/Lu ratios with increasingdifferentiation (cf. Fig. 3d).

4.4. Isotope data

Sr- and U-series isotope data are given in Table 1. All samples showexcess 230Th when plotted on a U–Th equiline diagram (Fig. 5). Thetholeiites lie within the field of postglacial main-rift basalts of Kokfeltet al. (2003) at 12–14% 230Th excess, while the rhyolites with 14–17%230Th excess are displaced to slightly lower (238U/232Th) and (230Th/232Th) activity ratios (Fig. 5a). All whole-rock samples display small butsignificant 226Ra excesses of 33–45% for the basalts, placing them in theoverall Iceland range of Kokfelt et al. (2003), and 3–10% for the rhyolites(cf. Table 1; note (226Ra/230Th)0 are ratios on the date of eruption). Theseare the first published 226Ra data for Icelandic rhyolites.

Table 1Isotope data

Sample # Unit T2 T2-px T2-plag T2-mt T3

Rock type Basalt Basalt Basalt Basalt Rhyo

Eruption year 1477 1477 1477 1477 147787Sr/86Sr 0.703291 0.703154 0.7032σ 0.000015 0.000014 0.000Th ppm 0.3902 0.01272 0.003702 0.097 15.892σ ppm 0.0004 0.00002 0.000004 0.000 0.02U ppm 0.1216 0.00424 0.00376 0.0304 4.7152σ ppm 0.0005 0.00003 0.00002 0.0001 0.021(238U/232Th) 0.945 1.012 3.083 0.945 0.9002σ 0.004 0.007 0.015 0.005 0.004(230U/232Th) 1.063 1.063 1.070 1.0282σ 0.012 0.010 0.009 0.011(230U/232Th)0 1.063 1.063 1.070 1.0292σ 0.013 0.012 0.010 0.012(230Th) dph/μg 0.1013 0.00330 0.255 3.9932σ dph/μg 0.0011 0.00003 0.002 0.042(230Th)0 dph/μg 0.1013 0.00330 0.255 3.9952σ dph/μg 0.0011 0.00003 0.002 0.042(226Ra) dph/μg 0.1470 0.0086 0.156 4.4242σ dph/μg 0.0009 0.0004 0.016 0.014(226Ra)0 dph/μg 0.1587 0.0100 0.130 4.5352σ dph/μg 0.0009 0.0004 0.014 0.015(226Ra/230Th) 1.451 2.601 0.611 1.1082σ 0.012 0.044 0.104 0.011(226Ra/230Th)0 1.567 3.014 0.511 1.1362σ 0.013 0.044 0.104 0.012Ba ppm 37.87 4.46 5.01 6.66 495.3

Parentheses denote activity ratios, which were calculated from atomic abundance analyses

Mineral separates from selected samples have indistinguishable(230Th/232Th) activity ratios from their respective host rocks and definezero-age U–Th isochrons for the 871 AD rhyolite and the 1477 basalt(Fig. 5b). Whole-rock (WR) and mineral separate data are plotted onBa-normalized Ra–Th equiline diagrams in Fig. 6. Assuming theminerals in each rock formed contemporaneously and rapidlycompared to the 226Ra half life, and that Ba is a proxy stable isotopefor 226Ra, the slope of these mineral isochrons yields the time offormation of themineral assemblage (e.g., Volpe andHammond,1991).This method generates ages of 1.4±0.3 and 2.0 (+0.2/−0.1) ka (2σ) forthe 1477basalt and rhyolite respectively (inyears before eruption), and3.2±0.4 ka (2σ) for the 871 rhyolite. Correcting (226Ra/230Th) in each ofthese whole rocks for decay over this time period gives parentalmagma (226Ra/230Th) of 2.03, 1.33 and 1.18 for the 1477 basalt, 1477rhyolite, and 871 rhyolite, respectively. There are multiple potentialcomplicating factors with this simple age assessment, such as timedelay between melt formation and the onset of crystallization, non-instantaneous (perhaps even sequential) formation of phenocrystsduring crustal residence, and imperfections to the use as Ba as a proxyto Ra in mineral melt partitioning. In the discussion section, we willexamine each of these in turn.

Our new Sr isotope data slightly extend the 87Sr/86Sr ranges ofVeidivötn basalts and Torfajökull rhyolites to 0.70307–0.70328 and0.70322–0.70342, respectively (Fig. S3 in the Appendix). By compar-ison, published 87Sr/86Sr data on mafic west Torfajökull samples rangefrom 0.70328 to 0.70348. An early study pointed to distinctly higher87Sr/86Sr in rhyolites than in basalts of this area (O'Nions andGrönvold, 1973), but subsequently published 87Sr/86Sr data collec-tively display a less clear-cut distinction at Torfajökull, with a total87Sr/86Sr variation amounting to 10 times the analytical uncertainty(Stecher et al., 1999). Although our Veidivötn values are slightlyelevated compared to those in the literature, nothing about theanalyses themselves make them suspect; likewise, contamination byrhyolitic melts, which occurs in Veidivötn basalts at the southern endof the fissure (Mørk, 1984), is inconsistent with near-end-memberincompatible element compositions of our samples. Thus the data

T3-mt T6 T7 T7-px T7plag T7-mt

lite Rhyolite Basalt Rhyolite Rhyolite Rhyolite Rhyolite

1477 871 AD 871 AD 871 AD 871 AD 871 AD

409 0.703236 0.703433015 0.000012 0.000014

7.677 0.5940 17.84 3.661 0.4662 6.7910.008 0.0007 0.02 0.004 0.0006 0.0072.262 0.1823 5.270 1.059 0.1379 2.1480.010 0.0008 0.023 0.005 0.0006 0.0130.894 0.931 0.897 0.878 0.897 0.9600.004 0.004 0.004 0.004 0.004 0.0061.040 1.061 1.045 1.038 1.049 1.0440.012 0.011 0.011 0.008 0.008 0.0101.040 1.062 1.047 1.039 1.051 1.0450.013 0.012 0.012 0.010 0.010 0.0121.950 0.154 4.555 0.928 0.1195 1.7330.022 0.002 0.050 0.008 0.0010 0.0171.951 0.154 4.562 0.930 0.1197 1.7340.023 0.002 0.050 0.008 0.0010 0.0171.502 0.206 4.681 0.820 1.180 1.5410.035 0.003 0.025 0.016 0.080 0.0131.386 0.238 4.761 0.751 1.855 1.4200.032 0.004 0.025 0.015 0.125 0.0120.770 1.335 1.028 0.883 9.879 0.8900.026 0.019 0.012 0.021 0.068 0.0130.711 1.549 1.045 0.809 15.52 0.8190.026 0.020 0.013 0.022 0.07 0.01454.88 61.04 466.2 23.14 796.0 51.65

using λ238U=1.551×10−10 yr−1, λ230Th=9.195×10−6 yr−1 and λ226Ra=4.332×10−4 yr−1.

Fig. 5. Eruption age corrected U–Th equiline diagrams for the most recent eruptionsfrom the Torfajökull–Veidivötn area, all of which display 226Ra excesses. Symbols:(■) 871AD Veidivötn basalt T6; (□) 871AD Torfajökull rhyolite T7; (♦) 1477 Veidivötnbasalt T2; (◊) 1477 Torfajökull rhyolite T3. (6a) Whole rocks. Fields of Iceland'spostglacial main-rift and off-rift mafic volcanic rocks are based on the data by Kokfelt etal. (2003) and have been drawn to enclose N85% of the respective datasets. Althoughtheir U/Th ratios are systematically offset by 2–4% when compared to data on the sameflows reported by Maclennan et al. (2001), the fields enclose most of the recentlyreported U–Th isotope data of mafic samples from Iceland (Sigmarsson et al., 1991;Sigmarsson et al., 1992a,b). Numbers in kyr model the evolution of Veidivötn basaltswith time. Arrow “H?” indicates the affect of questionable metasomatic addition of U toa basaltic source. See text for discussion. (6b) 230Th–238U isochrons for historicallyerupted rocks studied here follow zero-age trends (see text). “plag” plagioclase, “px”pyroxene, “mt” magnetite; whole rocks are unlabeled.

392 G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

presented here suggest that 87Sr/86Sr of Veidivötn basalts andtransitional west Torfajökull postglacial volcanics are more of acontinuum, with the latter generally elevated compared to the former.

5. Discussion

5.1. Petrogenesis and relationships of erupted magma types

It is important to first evaluate petrogenetic processes in the mag-mas of interest before interpreting the U-series results in terms ofmagmatic processes and timescales. While the intermediate volcanicsin northeast Torfajökull are hybrid magmas resulting from pre-eruptive mixing of Torfajökull rhyolites and Veidivötn basalts (Blake,1984;McGarvie,1984;McGarvie et al., 1990), the newwhole-rock datapresented here do not fall on the geochemical mixing trends (cf. Fig. 3and Fig. S2 in the Appendix), implying that they provide insights intothe primary petrogenetic processes operating in the region. The smallMgO range (6.74 to 7.08) in the end member Veidivötn basaltsprovides almost no information about the mineral assemblage thatgoverned their magmatic evolution. However, low MgO, Cr and Nicontents indicate significant olivine removal. Al2O3 and Sr compat-

ibility (cf. Fig. 3), and small negative Eu anomalies (0.91bEu/Eu⁎b0.94;cf. Fig. 4) suggest that plagioclase feldspar was also part of thefractionating assemblage.Within the Torfajökull rhyolites, Al2O3 and Srcompatibility and strongly negative Eu anomalies indicate plagioclasedominated fractionation. Apparently, significant potassic feldspar andbiotite fractionation from rhyolitic melts did not occur since K2O andNb behave incompatibly. TiO2, P2O5, and – at higher SiO2 – Zrcompatiblity implicate titanomagnetite, apatite, and zircon fractiona-tion, respectively, in some of the rhyolites.

The new data also provide insights into the potential petrogeneticrelationship between Torfajökull rhyolites and mafic rocks erupted inthe area. Torfajökull rhyolites were previously suggested to have beengenerated by extensive fractional crystallization of transitional alkalibasalts through icelandites (Macdonald et al., 1990). However, thesimilarity of Al2O3 content in basalts and rhyolites casts doubt on thisscenario (Fig. 3a). Plagioclase feldspar appears to be one of the majorfractionating phases in both the basalts and the rhyolites, so onewould expect the rhyolites to have significantly lower Al2O3 contentsthan the basalts if fractional crystallization was indeed dominantduring their petrogenesis. This is confirmed using MELTS fractionalcrystallization modelling (Ghiorso and Sack, 1995; Smith and Asimow,2005). Note that on Fig. 3a, variations in P, [H2O], and fO2

change theminimum Al2O3 content prior to significant SiO2 enrichment, but donot change the general shape of the liquid line of decent (cf. Zellmeret al., 2005, who followed a similar line of evidence for arc magmadifferentiation).

Alternatively, melting experiments (Beard and Lofgren, 1989; Thyet al., 1990) suggest that Icelandic rhyolite major element geochem-istry derives from partial melting of greenstones and amphibolite-facies metabasalts (cf. Gunnarsson et al., 1998). Off-axis settings inIceland normally erupt transitional or alkali basalts, and at Torfajökullthis carries through to felsic magmas, which are high-K rhyolites. Inaddition, rocks from west Torfajökull display relatively high Zrconcentrations that remain approximately constant at SiO2 valuesgreater than about 56 wt.%, suggesting early fractionation of zircon.This is supported by increasing Yb/Lu ratios with progressivedifferentiation (Figs. 3d and 4), which are known to be stronglyincreased by zircon fractionation (Henderson, 1985; Fujimaki, 1986).Old west Torfajökull rhyolites, which display the highest Yb/Lu ratiosof the area, hence appear to be petrogenetically related to thetransitional basalts sampled in that area.

However, the most recently erupted rhyolites do not display azircon fractionation signature. Instead, they have much lower Yb/Luratios, making the transitional alkali basalts unsuitable as source rocks(cf. Figs. 3d and 4). While the data suggest zircon fractionation may beoperating in the highest silica rhyolites, Yb/Lu ratios may be bufferedby contemporaneous fractionation of apatite (cf. Fig. S2c in theAppendix), which has a weaker but opposite effect on HREE ratios(Henderson, 1985; Fujimaki, 1986). The Yb–Lu–Zr data thereforeindicate that the youngest Torfajökull rhyolites are related to partialmelting of a Zr poor source, such as has produced the historicalVeidivötn tholeiites.

5.2. Tholeiites as a source of the most recent rhyolites?

Unaltered protoliths of tholeiitic composition cannot generatehigh-K off-axis rhyolites with K2ON3 wt.% (Beard and Lofgren, 1989;Thy et al., 1990). However, metasomatism of low-K tholeiites via hightemperature hydrothermal activity (as observed elsewhere in Iceland;Steinthórsson et al., 1987) may produce a suitable protolith. AtTorfajökull, one of Iceland's most productive geothermal areas, wherethe surrounding crust is of transitional alkaline nature (Imsland,1983), metasomatism is likely to increase the alkali content oftholeiitic protoliths prior to their partial melting. Data on alkalimobility is scarce, although K is about three times more mobile thanRb in analogous marine hydrothermal systems (Von Damm, 1990;

393G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

Rubin, 1997), and K/Rb ratios within Veidivötn tholeiites reach valuescomparable to mafic rocks from west Torfajökull (Fig. S2d in theAppendix). MELTS batch melting of a tholeiitic source with slightenrichment of alkalis can generate silica-rich melts with high Al2O3

and K2O content, although further crystal fractionation includingapatite will be required to generate the rhyolites (cf. Fig. 3 and Fig. S2in the Appendix).

Data on 87Sr/86Sr ratios in tholeiites, transitional alkali basalts andrhyolites from Torfajökull and Veidivötn are rare (Fig. S3 in theAppendix). No systematic trends are apparent, although the transi-tional alkali rocks display generally higher 87Sr/86Sr ratios than thetholeiites, and are isotopically comparable to the rhyolites. However,fluid mobility of Sr is also well documented, and hydrothermalalteration of intruded low 87Sr/86Sr tholeiites within an areadominated by high 87Sr/86Sr alkali basaltic crust may measurablyincrease 87Sr/86Sr of the protoliths. It may therefore be argued that theSr isotopic differences between tholeiites and rhyolites (Fig. S3 in theAppendix) do not preclude a petrogenetic link between thesecompositions, although a slight admixture of alkali basalt into thetholeiitic protolith might also explain the small Sr isotopic shift. Thereis insufficient data to test mixing scenarios in detail. Future workwould benefit from detailed studies of fluid immobile elements andtheir isotopes, e.g. neodymium and hafnium.

Finally, it is interesting to note that tholeiitic enclaves occur in therecent Torfajökull rhyolites but not in postglacial ones older than 7 ka,which contain transitional or alkali basaltic enclaves (McGarvie et al.,1990). This change in the composition of mafic enclaves is wholly inline with our suggestion that the magmatic feeding system ofTorfajökull Central Volcano shifted from dominantly alkali basalticto dominantly tholeiitic compositions, consistent with the proposedprogressive south-westwards propagation of the Eastern Rift Zonethrough time (e.g., Hey et al., 1998). Tholeiitic magma influx beneathTorfajökull is also consistent with extremely low shear wave velocitiesof the crust in the surrounding region (Allen et al., 2002) that point tosolidified or partially solidified high-temperature intrusives at depthbeneath Torfajökull.

5.3. Whole-rockage constraints on the protoliths of themost recent rhyolites

It has been argued previously that Torfajökull rhyoliticmagmaswerederived from “relatively recently” intruded mafic to intermediateprotoliths on basis of Sr, Nd and Pb isotopic similarity between rhyolitesand transitional basalts there (Stecher et al., 1999). However, signifi-cantly better constraints on the age of the basaltic protoliths to the 871and 1477 Torfajökull rhyolites are provided by the U-series chemistry ofthese rocks. On the broadest geographic scales, U/Th ratios of postglacialIcelandic basalts are controlled by mixing of enriched and depletedmantle components in proportions depending on the distance from theplume centre (Kokfelt et al., 2003), so mafic plutons within ageographically limited region such as Torfajökull central volcano areexpected to display (238U/232Th) activity ratios that are similar to basaltserupted from the same area. In Fig. 5a, the recent Veidivötn basalts plotwithin the field of young Icelandic main-rift lavas. The similarity of(230Th/232Th) in the most recently erupted rhyolites and basalts studiedhere suggests that the rhyolite sources alsowere young. If the protolithswere old (N350 ka), they would have been in U–Th equilibrium andtherefore sit on the equiline at a (238U/232Th) activity ratio of ≥1.03 toproduce the observed (230Th/232Th) values, which is higher than the(238U/232Th) activity ratios displayedby these andmost other postglacialbasalts from Iceland (cf. Fig. 5a).

The very similar (230Th/232Th) activities of the rhyolites and theircoeruptive tholeiites not only strengthen the notion of a petrogeneticlink between these compositions as outlined in the previous section,but also indicate 230Th–238U disequilibrium in the protolith prior topartial melting. The age of the protoliths that formed the rhyolites canbe tentatively quantified if these protoliths originally had U–Th

isotope compositions similar to the most recently erupted Veidivötnbasalts. Ages of 14 (+17/−15) ka (2σ) and 37 (+25/−20) ka (2σ) areobtained for the protoliths of the 871 AD and 1477 rhyolites,respectively (cf. Fig. 5a). These age constraints may be relaxed toslightly older ages if the U/Th ratios in the protoliths increasedconcomitant with alkali enrichment during metasomatism by hightemperature hydrothermal activity. However, significant uraniumenrichment is unlikely, because in contrast to alkali uptake intofeldspar, uranium enrichment is hindered by a lack of suitable hostphases at mafic compositions. In general, partial melting of young andtherefore still hot igneous protoliths is favoured over melting of older,cooler crustal materials (e.g., Annen and Sparks, 2002).

Although it cannot be precluded that some 226Ra excess in therhyolites may be preserved from a recent partial melting event (e.g.,that formed the rhyolite melt and/or the source rock it came from),processes that occurred subsequent to melting may have beensignificant in forming the whole-rock Ra–Th disequilibria. Theseinclude addition of Ra due to its mobility during metasomaticalteration of the protolith (cf. Sturchio et al., 1989), as well as Thscavenging from the melt by late-stage zircon or apatite fractionation.However, the latter scenario appears to be inconsistent with the smalldifferences in Zr and P contents, as well as Th/U ratio, in the tworhyolites which we have analyzed for Ra. As there are too fewconstraints to quantify the relative importance of these processes toforming Ra excesses in the rhyolitic melts, we focus instead on theimplications of the mineral data.

5.4. Insights from pre-eruptive crystallization timescales of basalts andrhyolites

Euhedral to subhedral morphologies in the crystal assemblage ofthe most recent Veidivötn and Torfajökull co-eruptive products(Fig. 2), combined with well-defined fields for plagioclase andpyroxene mineral chemistry data (Fig. S1 in the Appendix), suggestthat crystal inheritance is limited. Only a small number of rhyoliticfeldspars display slightly lower Na contents and higher equilibrationtemperatures. These may be antecrysts from a previous crystal-lization episode. A few pyroxenes of the 1477 eruption are alsocompositionally different from the main population, and again axenocrystic or antecrystic origin of these cannot be precluded.However, the bulk of the crystal assemblage of the most recentlyerupted Veidivötn basalts and Torfajökull rhyolites appears young.This notion is supported by U-series mineral separate analyses ofthe 871 AD rhyolite and the 1477 basalt, which yield ~zero age Th–Uinternal isochrons (Fig. 5b).

Ra–Th data provide tighter time constraints, with pre-eruptivecrystallization ages of 1.4 (±0.3) and 2.0 (+0.2/−0.1) ka (2σ) for the 1477basalt and rhyolite respectively, and 3.2 (±0.4) ka (2σ) for the 871rhyolite (Fig. 6). The results for the 1477 eruption imply that therhyolite started crystallizing about 700 years earlier than the basalt, i.e.that this particular rhyolite spent ~50% more time evolving beforeeruption than did the coeruptive basalt. The contrast in rhyoliteisochron ages suggest that the 871 rhyolite spent roughly 50% moretime between crystallization and eruption than did 1477 rhyolite. It ispossible to look for sequential crystallization in two of the threesamples with mineral WR data because multiple minerals wereanalyzed (i.e., T2 and T7). However, no individual mineral appears tofall off of the best fit linear regressions on the isochron diagrams,suggesting that crystallization was simultaneous for those phases(within error). Further, the overall mineral assemblage in these rocksis characterised by small and euhedral crystals and thus appears tobe cotectic, implying that individual crystal populations are notsignificantly different from each other in their age. If crystallizationoccurred over a protracted time compared to the 226Ra half-life, itwould have occurred at roughly the same rate for all of the mineralspresent.

Fig. 6. Ra–Th mineral isochron diagrams of the most recent eruptions from theTorfajökull–Veidivötn area (226Ra excesses have been corrected to the date of eruption);“wr” whole rock; mineral abbreviations are defined in the Fig. 5 caption. Analyticaluncertainties are smaller than symbol size, unless indicated otherwise. Crystallizationages are based on the assumptions of contemporaneous crystal growth over a shorttime period relative to the half-life of 226Ra and equality of the Ra and Ba crystal-meltpartition coefficients. Alternate scenarios are discussed in the text. Errors werecalculated using Isoplot (Ludwig, 2003). (7a) 1477 Veidivn basalt. (7b) 1477 Torfajökullrhyolite. (7c) 871AD Torfajökull rhyolite. Whole rock values of other samples(unlabeled) are given in each diagram.

394 G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

Another issuewith the 226Ramagmaageestimates involves theuseofBa as a proxy for Ra duringmineralmelt partitioning. Some authors havefound reasonable colinear relationships for minerals and melt, implyingthat this conditionholds to afirst order (e.g., Volpe andHammond,1991),whereas others have found non-linearity in some volcanic rocks (e.g.,Cooper et al., 2001). Theoretical treatments of the melt-mineralpartitioning suggest that even small differences in ionic radii can affectmineral melt distribution coefficients (D) of like-charged ions because ofmineral lattice strain. Parameterization of experimental data setsprovides a method to predict differences in, for instance, DBa and DRa

(e.g., Blundy and Wood, 2003). Importantly, each mineral respondsdifferently to temperature andmelt composition. It is extremely unlikely

that a predicted ratio such as DRa/DBa would be the same for multipleminerals in a singlemelt, so Ba normalizedmineral-melt data (i.e., Fig. 6)would not be colinear in young lavas even if the minerals formed at thesame time. Cooper et al. (2001) observed such non-linearity in a crystalrich Kilauean basalt and, using strain rate parameterization to estimateDRa for plagioclase, suggest a crystallization age of 1 kyr before eruptionversus 3.4 kyr if they had assumedDRa=DBa. The difficulty with applyingsuch an estimate here is that minerals are observed to be stronglycolinear on a Ba-normalized diagram such as Fig. 6, such that two pointisochrons (i.e., magnetite-melt, plagioclase-melt, clinopyroxene-melt)yield the same result as a multi-point isochron assuming DRa=DBa.Evaluating each mineral-melt pair separately and using DRa/DBa

estimated from experimental parameterizations produces disparateages for the multiple pairs in a single rock, which seems an unlikelyresult. Further, Blundy and Wood (2003) indicate that fleetingly smallvalues ofDRa andDBa in magnetite imply that these elements will not befractionated fromeachother by this phase. Becausemagnetite-melt pairsyield the same ages as other mineral-melt pairs in T2 and T7 assumingDRa=DBa, this is another reason to suspect that mineral strainparameterized DRa≠DBa are not applicable. As in all studies that usenaturalmineral separates, it is always possibly thatmicroinclusions haveaffected the Th–U–Ra–Ba systematics measured in some of our phases,although it seems rather unlikely that they would have participated injust the right proportion to changemineral specificDRa/DBa to a ratio of 1.

Another way to assess time in this system is to assume anisochronal relationship between coeruptive magmas, such as mightoccur if a basalt reservoir deep in the crust released pulses of magmaat different times that generated basaltic eruptions and rhyolitemelting/eruptions from the same closed system melt source. Suchbasalt–rhyolite pseudo-isochrons for the 1477 and 871 AD eruptionsyield ~5.8 kyr and ~4.7 kyr of pre-eruptive magma aging, respectively,both of which are significantly longer than the internal isochronestimates. Such a scenario implies considerable magma aging withoutcrystallization (or loss of early crystallizing phases). This added timeinterval would also imply parental magma (226Ra/230Th) of roughly 5and 8 for the 1477 and 871 basalts. Such values are unrealistically highrelative to melt model predictions for typical source mineralogy andupwelling conditions for dry mantle melting to produce basaltic melts(Spiegelman and Elliott, 1993). Furthermore, the basalt-rhyolitepseudo-isochron relationship assumes the same Ba/Th for the rhyoliteand basalts before low pressure differentiation, which is difficult toenvision given the large differences between the erupted lavas and theminerals likely to be involved. For these reasons the pseudo-isochronslikely yield extreme overestimates of magma residence times in thecrust, implying that magmas did not reside significantly longer in thecrust than the internal mineral isochron age estimates.

In summary, the U–Th and Ra–Th data presented here indicate thatmelting and preeruptive crystallization of the most recent volcanicdeposits of the Torfajökull–Veidivötn area occurred within a fewthousand years prior to eruption, within the Late Holocene. This isconsistent with the occurrence of tholeiitic enclaves in Torfajökullrhyolites from the early Late Holocene onwards (McGarvie et al.,1990).Interestingly, pre-eruptive crustal magma residence times exceed theeruption recurrence interval for this system by factors of 2 (basalts) toabout 4 (rhyolites), indicating that multiple batches of magma hadbeen accumulating and evolving within the crust several thousandyears prior to these eruptions.

5.5. Petrogenesis of silicic lavas from Torfajökull revisited

Gunnarsson et al. (1998) suggested two principal mechanismsinvolving melting of previously solidified magmas for the petrogen-esis of silicic extrusives in Iceland: (1) melting of small, highlydifferentiated segregation lenses formed in the upper part of priorgabbroic intrusions, followed by subsequent extraction, ascent andpre-eruptive melt accumulation (cf. their Fig. 18); and (2) partial

Fig. 7. Cartoon depicting the petrogenetic model for the most recent eruptions from the Torfajökull–Veidivötn volcanic system. (9a)Within the last ≤37 (+25/−20) kyr, small tholeiiticprotoliths were intruded into the mid crust where they suffer hydrothermal alteration (“H”) and K-enrichment. (9b) By ~4.3 ka b.p., influx of hot basalt leads to partial melting (“p.m.”) of altered tholeiitic protoliths and accummulation of rhyolitic melts in an upper crustal reservoir beneath Torfajökull, where they start to crystallize. (9c) By ~2.5 ka b.p., anotherthermal spike leads to further melting of altered protoliths (“p.m.”) and establishment of a second crystallizing upper crustal rhyolite reservoir. (9d) At 871 AD, tholeiitic magmainflux into the upper crust results in magma mixing (“M”), and triggers eruption of rhyolite from the first rhyolitic reservoir (“ER”), basalt from the Veidivoetn fissures (“EB”), andsome mixed magmas (“EM”). (9e) A similar event occurs in 1477, triggering eruption from the second rhyolitic reservoir. Question marks indicate uncertainty in dike propagationdirection, favoured by some authors to be from the northeast (e.g., Blake, 1984; Larsen, 1984; McGarvie, 1984), although vertical diking from below cannot be precluded. In summary,the system is characterized by small volume magma batches that are transported through the plumbing systemwith relatively high frequency, as opposed to less frequent transportof larger magma volumes that would have to be stored in larger crustal reservoirs for longer periods of time (cf. Zellmer and Annen, in press).

395G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

melting of old, hydrothermally altered extrusives and intrusives,which were brought into the melting region through isostaticsubsidence beneath large, long-lived central volcanoes (see their Fig.19). Such processes require large magma reservoirs. However, seismicstudies have not been successful in imaging such chambers (e.g., Allenet al., 2002). Further, large magma reservoirs would be thermallybuffered, and rapid petrogenetic processes as those evidenced by thisstudy would not be anticipated to operate in such environments.

The U-series data presented here points to a very differentpetrogenetic scenario for the most recently erupted Torfajökullrhyolites (as depicted in Fig. 7), based on the similarity of (230Th/232Th) of coreruptive basalts and rhyolites in both eruptions, andwidespread Ra–Th disequilibria in both whole-rocks and minerals.Within the last ≤37 (+25/−20) kyr, we propose that tholeiitic magmasof composition similar to those of the Veidivötn basalts began tointrude beneath Torfajökull central volcano, where they formedprotoliths that were subsequently metasomatised through high-temperature hydrothermal activity. During the Late Holocene, furtherinflux of hot, mafic magma lead to repeated small degree (F≈0.2, cf.Fig. 3 and Fig. S2 in the Appendix) partial melting of a large volume ofthese tholeiitic protoliths. The evolved partial melts accumulated in

upper crustal magma reservoirs and began to crystallize (and furtherdifferentiate). By 2.5 ka prior to eruption, at least two such reservoirsof rhyolitic composition had formed. Subsequent intersection of therhyolites by new basaltic melt in 871 AD and 1477 resulted in magmamixing and bimodal eruption from fissures of the Torfajökull–Veidivötn area. Given that at least 10 rhyolite eruptions have occurredat Torfajökull during the Late Holocene, of which at least 5 happenedduring the last 4 kyr (Macdonald et al., 1990), it is likely that the uppercrust beneath Torfajökull was and potentially still is host to smallrhyolitic reservoirs formed during the prolonged postglacial history ofintrusion and extrusion at the volcano. The thermal state of the crustin this area is therefore elevated, and the next intrusion of maficmagma is not unlikely to trigger further partial melting at depth and/or mobilization of an existing rhyolite melt body, potentially resultingin future co-eruptive volcanic events at Torfajökull central volcano andthe Veidivötn fissure swarm.

6. Conclusions

1. Postglacial rhyolites erupted from Torfajökull central volcano arethe products of differentiation from recent small degree (F≈0.2)

396 G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

partial melts of previously intruded and metasomatised maficprotoliths.

2. Significant 230Th excesses of the most recent rhyolites, andsimilarity of rhyolitic (230Th/232Th) activity ratios to those of co-eruptive Veidivötn basalts, constrain the age of the protoliths ofthe most recently erupted rhyolites to ≤37 (+25/−20) kyr. Previousmelting models invoking melt extraction from old silicic segrega-tion lenses or rocks with long isostatic subsidence histories are notconsistent with the new data.

3. Ra–Th mineral isochrons yield pre-eruptive crystallization ages ofup to 3.2 ka, suggesting that the crystal load carried within therhyolites is also young and that rhyolite melts were generated inthe Late Holocene. Crystals display simple growth historiesconsistent with their being magmatic phenocrysts formed duringrhyolite cooling prior to eruption. Crystal-melt disequilibria do notoccur in our sample set, and appear to be restricted to the mixedmagma samples considered by previous studies.

4. In contrast to old rhyolites from west Torfajökull, which weresourced from the transitional alkali basalts that are abundant in theSouthern Flank Zone, the data suggests that the protoliths of thetwo most recently erupted rhyolites in the area are dominated bytholeiites similar in composition to the co-eruptive Veidivötntholeiites. Transition from alkaline to tholeiitic source protoliths ismirrored by the same transition observed in the composition ofmafic enclaves within postglacial mixed eruption products.Volcanic activity of Torfajökull appears to be sustained by heatprovided to the core of this central volcano through frequent influxof mafic magma of progessively more tholeiitic composition,consistent with the progressive south-westward propagation ofthe Eastern Rift Zone with time.

Acknowledgements

We thank Carlos Linares-López and Teodoro Hernandez-Treviño ofUNAM, Mexico City, for assistance with electron microprobe analysesand mineral separations, respectively, and Doug Pyle for ICP-MSanalyses. Discussions with Dave McGarvie, Olgeir Sigmarsson, andThorvaldur Thordarson, and constructive reviews of Thomas Kokfeltand David Peate improved this paper. Research funding, includingpostdoctoral stipends for GFZ and ZJC during the time when the fieldand analytical work were conducted, was provided by NSF (EAR-0106463 to KHR), the Nordic Volcanological Institute (now part of theUniversity of Iceland) to KG for fieldwork expenses, and by AcademiaSinica and the National Science Council, Taiwan, to GFZ duringcompletion of this work. GFZ thanks Dima Kamenetzky and JörgErzinger for support during productive sabbaticals in Australia andGermany. KHR thanks Jaime Urrutia and the Instituto de Geofisica,UNAM, for support during a productive sabbatical in Mexico.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.epsl.2008.02.026.

References

Allen, R.M., Nolet, G., Morgan, J., Vogfjörd, K., Nettles, M., Ekström, G., Bergsson, B.H.,Erlendsson, P., Foulger, G.R., Jakobsdóttir, S., Julian, B.R., Pritchard, M., Ragnarsson,S., Stefánsson, R., 2002. Plume-driven plumbing and crustal formation in Iceland.J. Geophys. Res. 107 (B8). doi:10.1029/2001JB000584.

Annen, C., Sparks, R.S.J., 2002. Effects of repetitive emplacement of basaltic intrusionson thermal evolution and melt generation in the crust. Earth Planet. Sci. Lett. 203,937–955.

Beard, J.S., Lofgren, G.E., 1989. Effect of water on the composition of partial melts ofgreenstone and amphibolite. Science 244, 195–197.

Bindeman, I.N., Sigmarsson, O., Eiler, J., 2006. Time constraints on the origin of large volumebasalts derived from O-isotope and trace element zoning and U-series disequilibria inthe Laki and Grímsvötn volcanic system. Earth Planet. Sci. Lett. 245, 245–259.

Björnsson, A., Saemundsson, K., Einarsson, P., Tryggvason, E., Grönvold, K., 1977. Currentrifting episode in north Iceland. Nature 266, 1977.

Blake, S., 1984. Magma mixing and hybridization processes at the alkalic, silicic,Torfajökull Central Volcano triggered by tholeiitic Veidivötn fissuring, southIceland. J. Volcanol. Geotherm. Res. 22, 1–31.

Blundy, J., Wood, B., 2003. Mineral-melt partitioning of uranium, thorium and theirdaughters. In: Bourdon, B., Henderson, G.M., Lundstrom, C.C., Turner, S.P. (Eds.),Uranium-series geochemistry. Reviews in Mineralogy and Geochemistry.Mineralogical Society of America and Geochemical Society, Washington, DC,pp. 59–123.

Carmichael, I.S.E., 1964. The petrology of Thingmuli, a Tertiary volcano in easternIceland. J. Petrol. 5, 435–460.

Condomines, M., Morand, P., Allègre, C.J., Sigvaldason, G., 1981. 230Th–238U disequilibriain historical lavas from Iceland. Earth Planet. Sci. Lett. 55, 393–406.

Cooper, K.M., Reid, M.R., Murrell, M.T., Clague, D.A., 2001. Crystal and magma residenceat Kilauea volcano, Hawaii: 230Th–226Ra dating of the 1955 East Rift Eruption. EarthPlanet. Sci. Lett. 184, 703–718.

Fuhrman, M.L., Lindsley, D.H., 1988. Ternary-feldspar modeling and thermometry. Am.Mineral. 73, 201–215.

Fujimaki, H., 1986. Partition coefficients of Hf, Zr, and REE between zircon, apatite, andliquid. Contrib. Mineral. Petrol. 94, 42–45.

Ghiorso, M.S., Sack, R.O., 1995. Chemical mass transfer in magmatic processes IV. Arevised and internally consistent thermodynamic model for the interpolation andextrapolation of liquid–solid equilibria in magmatic systems at elevated tempera-tures and pressures. Contrib. Mineral. Petrol. 119, 197–212.

Gunnarsson, B., 1987. Petrology and petrogenesis of silicic and intermediate lavas on apropagating oceanic rift. The Torfajökull and Hekla central volcanoes, south-centralIceland. PhD Thesis, Johns Hopkins University, Baltimore, Maryland, 432 pp.

Gunnarsson, B., Marsh, B.D., Taylor Jr., H.P., 1998. Generation of Icelandic rhyolites: siliciclavas from the Torfajökull central volcano. J. Volcanol. Geotherm. Res. 83, 1–45.

Hansen, H., Grönvold, K., 2000. Plagioclase ultraphyric basalts in Iceland: the mush ofthe rift. J. Volcanol. Geotherm. Res. 98, 1–32.

Hèmond, C., Condomines,M., Fourcade, S., Allègre, C.J., Oskarsson,N., Javoy,M.,1988. Thorium,strontium and oxygen isotopic geochemistry in recent tholeiites from Iceland: crustalinfluence on mantle-derived magmas. Earth Planet. Sci. Lett. 87, 273–285.

Henderson, P., 1985. General geochemical properties and abundances of the rare earthelements. In: Henderson, P. (Ed.), Rare Earth Element Geochemistry. Developmentsin Geochemistry, vol. 2. Elsevier, Amsterdam, pp. 1–32.

Hey, R.N., Sinton, J.M., Duennebier, F.K., 1998. Propagating rifts and spreading centers.In: Winterer, E.D., Hussong, D.M., Decker, R.W. (Eds.), The Eastern Pacific Ocean andHawaii. Geological Society of America, Boulder, Colorado, pp. 161–176.

Imsland, P., 1983. Iceland and the ocean floor: comparison of chemical characteristics ofthe magmatic rocks and some volcanic features. Contrib. Mineral. Petrol. 83, 31–37.

Ívarsson, G., 1992. Geology and petrochemistry of the Torfajökull Central Volcano inCentral South Iceland, in association with the Icelandic hot spot and rift zones.University of Hawaii, Honolulu. 332 pp.

Jakobsson, S.P., 1979. Petrology of recent basalts of the Eastern Volcanic Zone, Iceland.Acta Nat. Isl. 26, 1–103.

Jóhannesson, H., Jakobsson, S.P., Saemundsson, K., 1990. Geological map of Iceland,sheet 6, South-Iceland, third edition. Icelandic Museum of Natural History andIceland Geodetic Survey, Reykjavík.

Jónasson, K., 1994. Rhyolite volcanism in the Krafla central volcano, north-east Iceland.Bull. Volcanol. 56, 516–528.

Kokfelt, T.F., Hoernle, K., Hauff, F., 2003. Upwelling and melting of the Iceland plumefrom radial variation of 238U–230Th disequilibria in postglacial volcanic rocks. EarthPlanet. Sci. Lett. 124, 167–186.

Kokfelt, T.F., Hoernle, K., Hauff, F., Fiebig, J., Werner, R., Garbe-Schönberg, D., 2006.Combined trace element and Pb–Nd–Sr–O isotope evidence for recycled oceaniccrust (upper and lower) in the Iceland mantle plume. J. Petrol. 47, 1705–1749.

Lacasse, C., Sigurdsson, H., Carey, S.N., Jóhannesson, H., Thomas, L.E., Rogers, N.W., 2007.Bimodal volcanism at the Katla subglacial caldera, Iceland: insight into thegeochemistry and petrogenesis of rhyolitic magmas. Bull. Volcanol. 69, 373–399.

Larsen, G., 1984. Recent volcanic history of the Veidivötn fissure swarm, southernIceland — an approach to volcanic risk assessment. J. Volcanol. Geotherm. Res. 22,33–58.

Ludwig, K.R., 2003. Isoplot/Ex ver. 3, A Geochronological Toolkit for Microsoft Excel.Berkeley Geochronology Center, Berkeley, California.

Macdonald, K.C., McGarvie, D.W., Pinkerton, H., Smith, R.L., Palacz, Z.A., 1990.Petrogenetic evolution of the Torfajökull Volcanic Complex, Iceland I. Relationshipbetween the magma types. J. Petrol. 31, 429–459.

Maclennan, J., McKenzie, D., Grönvold, K., 2001. Plume-driven upwelling under centralIceland. Earth Planet. Sci. Lett. 194, 67–82.

Martin, E., Sigmarsson, O., 2007. Crustal thermal state and origin of silicic magma inIceland: the case of Torfajökull, Ljósufjöll and Snæfellsjökull volcanoes. Contrib.Mineral. Petrol. 153, 593–605.

McGarvie, D.W., 1984. Torfajökull: a volcano dominated by magma mixing. Geology 12,685–688.

McGarvie, D.W., 1985. Volcanology and petrology of mixed magmas and rhyolitesfrom the Torfajökull volcano, Iceland. PhD Thesis, University of Lancaster,Lancaster, U.K.

McGarvie, D.W., Macdonald, R., Pinkerton, H., Smith, R.L., 1990. Petrogenetic evolutionof the Torfajökull Volcanic Complex, Iceland II. The role of magma mixing. J. Petrol.31, 461–481.

Mørk, M.B.E., 1984. Magma mixing in the post-glacial Veidivötn fissure eruption,southeast Iceland: a microprobe study of mineral and glass variations. Lithos 17,55–75.

397G.F. Zellmer et al. / Earth and Planetary Science Letters 269 (2008) 387–397

Muehlenbachs, K., Anderson, A.T., Sigvaldason, G.E., 1974. Low 18O basalts from Iceland.Geochim. Cosmochim. Acta 38, 577–588.

Nicholson, H., Condomines, M., Fitton, J.G., Fallick, A.E., Grönvold, K., Rodgers, G., 1991.Geochemical and isotopic evidence for crustal assimilation beneath Krafla, Iceland.J. Petrol. 32, 1005–1020.

O'Nions, R.K., Grönvold, K., 1973. Petrogenetic relationships of acid and basic rocks inIceland: Sr-isotopes and rare-earth elements in late and postglacial volcanics. EarthPlanet. Sci. Lett. 19, 397–409.

Pietruszka, A.J., Rubin, K.H., Garcia, M.O., 2001. 226Ra–230Th–238U disequilibria ofhistorical Kilauea lavas (1790–1982) and the dynamics of mantle melting within theHawaiian plume. Earth Planet. Sci. Lett. 186, 15–31.

Rubin, K.H., 1997. Degassing of metals andmetalloids from erupting seamount andmid-ocean ridge volcanoes: Observations and predictions. Geochim. Cosmochim. Acta61, 3525–3542.

Rubin, K.H., 2001. Analysis of 232Th/230Th in volcanic rocks: a comparison of ThermalIonization Mass Spectrometry and other methodologies. Chem. Geol. 175, 755–782.

Rubin, K.H., van der Zander, I., Smith, M.C., Bergmanis, E.C., 2005. Minimum speed limitfor ocean ridge magmatism from 210Pb–226Ra–230Th disequilibria. Nature 437,534–538.

Saemundsson, K., 1972. Jardfraediglefsur um Torfajökullsvaedid. Naturufraedingurinn42, 81–144.

Saemundsson, K., 1978. Fissure swarms and central volcanoes of the neovolcanic zonesof Iceland. Geol. J. Spec. Issue 10, 415–432.

Sigmarsson, O., Hemond, C., Condomines, M., Fourcade, S., Oskarsson, N., 1991. Origin ofsilicic magma in Iceland revealed by Th isotopes. Geology 19, 621–624.

Sigmarsson, O., Condomines, M., Fourcade, S., 1992a. A detailed Th, Sr and O isotopestudy of Hekla: differentiation processes in an Icelandic volcano. Contrib. Mineral.Petrol. 112, 20–34.

Sigmarsson, O., Condomines, M., Fourcade, S., 1992b. Mantle and crustal contribution inthe genesis of Recent basalts from off-rift zones in Iceland: Constraints from Th, Sr,and O isotopes. Earth Planet. Sci. Lett. 110, 149–162.

Sigmarsson, O., Karlsson, H.R., Larsen, G., 2000. The 1996 and 1998 subglacial eruptionsbeneath the Vatnajökull ice sheet in Iceland: contrasting geochemical andgeophysical inferences on magma migration. Bull. Volcanol. 61, 468–476.

Sigurdsson, H., 1977. Generation of Icelandic rhyolites by melting of plagiogranites inthe oceanic layer. Nature 269, 25–28.

Sigurdsson, H., Sparks, R.S.J., 1978a. Lateral magma flow within rifted Icelandic crust.Nature 274, 126–130.

Sigurdsson, H., Sparks, R.S.J., 1978b. Rifting episode in north Iceland in 1874–1875 andthe eruptions of Askja and Sveinagja. Bull. Volcanol. 41, 149–167.

Sigvaldason, G.E., 1974. The petrology of Hekla and origin of silicic rocks in Iceland. Soc.Sci. Islandica 5, 1–44.

Sims, K.W.W., DePaolo, D.J., Murrell, M.T., Baldridge, W.S., Goldstein, S., Clague, D., Jull,M., 1999. Porosity of the melting zone and variations in the solid mantle upwelling

rate beneath Hawaii: inferences from the 238U–230Th–226Ra and 235U–231Padisequilibria. Geochim. Cosmochim. Acta 63, 4119–4138.

Smith, P.M., Asimow, P.D., 2005. Adiabat_1ph: A new public front–end to the MELTS,pMELTS, and pHMELTS models. Geochem. Geophys. Geosyst. 6. doi:10.1029/2004GC000816.

Sparks, R.S.J., Sigurdsson, H., Wilson, L., 1977. Magma mixing: a mechanism fortriggering acid explosive eruptions. Nature 267, 315–318.

Sparks, R.S.J., Wilson, L., Sigurdsson, H., 1981. The pyroclastic deposits of the 1875eruption of Askja, Iceland. Philos. Trans. R. Soc. Lond., A 299, 241–273.

Spiegelman, M., Elliott, T., 1993. Consequences of melt transport for uranium seriesdisequilibrium in young lavas. Earth Planet. Sci. Lett. 118, 1–20.

Stecher, O., Carlson, R.W., Gunnarsson, B., 1999. Torfajökull: a radiogenic end-member ofthe Iceland Pb-isotopic array. Earth Planet. Sci. Lett. 165, 117–127.

Steinthorsson, S., Hardarson, B.S., Ellam, R.M., Larsen, G., 2000. Petrochemistry of theGjálp-1996 subglacial eruption, Vatnajökull, SE Iceland. J. Volcanol. Geotherm. Res.98, 79–90.

Steinthórsson, S., Óskarsson, N., Arnórsson, S., Gunnlaugsson, E., 1987. Metasomatism inIceland: hydrothermal alteration and remelting of oceanic crust. In: Helgeson, H.C.(Ed.), Chemical transport inmetasomatic processes, Dordrecht, Holland, pp. 355–387.

Sturchio, N.C., Bohlke, J.K., Binz, C.M., 1989. Radium–thorium disequilibrium andzeolite–water ion exchange in a Yellowstone hydrothermal environment. Geochim.Cosmochim. Acta 53, 1025–1034.

Sun, S.S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanicbasalts: implications for mantle composition and processes. In: Saunders, A.D.,Norry, M.J. (Eds.), Magmatism in ocean basins. Geological Society SpecialPublication, vol. 42, pp. 313–345.

Thy, P., Beard, J.S., Lofgren, G.E., 1990. Experimental constraints on the origin of Icelandicrhyolites. J. Geol. 98, 417–421.

Vilmundardóttir, E.G., Snorrason, S.P., Larsen, G., Gudmundsson, Á., 1988. Geologicalmap, Sigalda–Veidivötn, 3340B. National Energy Authority, Hydro Power Divisionand The National Power Company, Reykjavík.

Volpe, A.M., Hammond, P.E., 1991. 238U–230Th–226Ra disequilibria in young Mount St.Helens rocks: time constraint for magma formation and crystallisation. EarthPlanet. Sci. Lett. 107, 475–486.

Von Damm, K.L., 1990. Seafloor hydrothermal activity: black smoker chemistry andchimneys. Ann. Rev. Earth Planet. Sci. 18, 173–204.

Walker, G.P.L., 1966. Acid volcanic rock in Iceland. B. Volcanol. 29, 375–402.Zellmer, G.F., Annen, C., in press. An introduction to magma dynamics. In: Annen, C.,

Zellmer, G.F. (Eds.), Dynamics of crustal magma transfer, storage and differentiation.Geological Society Special Publication 304.

Zellmer, G.F., Annen, C., Charlier, B.L.A., George, R.M.M., Turner, S.P., Hawkesworth, C.J.,2005. Magma evolution and ascent at volcanic arcs: constraining petrogeneticprocesses through rates and chronologies. J. Volcanol. Geotherm. Res. 140, 171–191.