The Surtsey Magma Series

The volcanic island of Surtsey (Vestmannaeyjar, Iceland) is the product of a 3.5-year-long eruption that began in November 1963. Observations of magma-water interaction during pyroclastic episodes made Surtsey the type example of shallow-to-emergent phreatomagmatic eruptions. Here, in part to mark the 50th anniversary of this canonical eruption, we present previously unpublished major-element whole-rock compositions, and new major and trace-element compositions of sideromelane glasses in tephra collected by observers and retrieved from the 1979 drill core. Compositions became progressively more primitive as the eruption progressed, with abrupt changes corresponding to shifts between the eruption’s four edifices. Trace-element ratios indicate that the chemical variation is best explained by mixing of different proportions of depleted ridge-like basalt, with ponded, enriched alkalic basalt similar to that of Iceland’s Eastern Volcanic Zone; however, the systematic offset of Surtsey compositions to lower Nb/Zr than other Vestmannaeyjar lavas indicates that these mixing end members are as-yet poorly contained by compositions in the literature. As the southwestern-most volcano in the Vestmannaeyjar, the geochemistry of the Surtsey Magma Series exemplifies processes occurring within ephemeral magma bodies on the extreme leading edge of a propagating off-axis rift in the vicinity of the Iceland plume.

The eruption of Surtsey Volcano, in the Vestmannaeyjar (Vestmann Islands), Iceland (1963-67; Fig. 1), is the canonical example of shallow-to-emergent subaqueous explosive volcanism. It demonstrated the ways in which magma-water interaction drastically affects plume and fragmentation dynamics to create characteristic steam-rich tephra jets and fine-grained deposits. The growth of new islands during the eruption was intensively chronicled in a near-continuous sample and observation record 1-5 that helped it become the type example of pyroclastic activity that is now referred to as Surtseyan Volcanism [6][7][8] . Timing also probably contributed to Surtsey becoming a type locality, because its eruption occurred not long after similar ones (particularly of Capelinhos, Azores, 1957-58 9,10 ) had demonstrated dynamic and depositional 11 similarities, motivating volcanologists to consider that shallow phreatomagmatic eruptions deserved their own unique and unifying classification among eruption styles 12 . Beyond the physical dynamics demonstrated during Surtsey's four-years of activity, the locality that is Surtsey Volcano is significant for having formed in the southwestern-most eruption of the Vestmannaeyjar archipelago, at the propagating tip of Iceland's Pleistocene-Holocene, Eastern Volcanic Zone 13 (EVZ). Its products are therefore key to unravelling petrological characteristics at the leading edge of an off-axis rift in the vicinity of the Iceland plume e.g., 14 .
Here we present the "Surtsey magma series", defined by previously unpublished whole-rock major-element analyses of microcrystalline rocks (bombs and lava) from different vents, and the "Surtsey melt series", defined by new and previously unpublished major and trace-element analyses of sideromelane glasses in tephra collected by observers and retrieved from the 1979 drill core 15 . Samples in both series are linked to their time of eruption and vent of origin to allow a spatiotemporal account of geochemical variability throughout the progression of Surtsey's eruption, from both whole rocks (melt + crystals + /− inclusions) and glass (melt now preserved as glass in tephra, but having been subjected to some fractionation during the eruption) perspectives. Now, in the 50 th anniversary of Surtsey's eruption, and with a recently submitted International Continental Drilling Program (ICDP) proposal to

Progression of the Eruption
The crew of fishing vessel Ísleifur II were first to witness activity in the vicinity of what would become the island of Surtsey, at first mistaking the plume for the smoke of a burning ship, in the early morning of November 14, 1963 2 . Following their reports, a consortium of Icelandic geologists began documenting the eruption in great detail, with many aspects of the eruption given in the Surtsey Research Progress Reports (https://www.surtsey.is) and by S. Thorarinsson [1][2][3][4][5] . Detailed timelines of activity at Surtsey and its satellite vents are given by Jakobsson and Moore 15 and Thordarson 17 , and are summarised below and in Fig. 2a. Grey crosses indicate previously published whole-rock analyses 5,13,22,25,28,31 . Glasses plotted as individual grains (Table S2), the averages of which are listed in Tables 3-4. Eruption day for drill core glasses are not known, but for plotting purposes are assumed to have accumulated at a constant rate over 1 month up to the November 14, 1963 emergence, and are plotted as negative days. Actual estimates of time from breaching the seafloor to emergence are much shorter 15  It is estimated that activity began on the seafloor in early November 1963, with a short submarine fissure building within a few days 15 from a relatively flat ocean floor at about 130 metres below sea level, to the surface by 14 November. Pyroclastic activity from this initial vent, Surtur I, built the foundation of the island that remains today. On 28 December 1963, pyroclastic activity at a new submarine vent Surtla was identified ~2 km to the NE of Surtsey. Surtla's phreatomagmatic jets breached the ocean surface from less than a metre's depth (Fig. 1C) 2 , but the edifice was never observed above water, and its activity had ceased by 6 January 1964. At the end of January 1964, pyroclastic activity on Surtsey continued, but shifted to a new Western vent, Surtur II. Pyroclastic eruption of Surtur II continued until early April, 1964, when it entered an effusive phase (Fig. 1B), with a 120 m wide lava pond forming in the crater. Effusive activity continued until mid-May 1965, and included the establishment of various lava ponds and lava tube systems, intermittent surface flow activity, and the effusion of tube-fed submarine pillow lavas.
As effusive activity at Surtur II waned, submarine pyroclastic activity at the new satellite vent Syrtlingur (Fig. 1D) began ~600 m to the NE of Surtsey. Syrtlingur grew above the surface in late May 1965, and continued in intermittent activity until 17 October 1965. A lack of effusive activity at Syrtlingur left it subject to rapid wave erosion, and though it achieved a maximum height of ~70 metres above sea level, it was washed away within one week of activity ending. The same fate befell the satellite vent Jólnir (Fig. 1E), which is estimated to have began its submarine growth ~800 m to the SW of Surtsey when Syrtlingur was waning, and emerged from the sea surface in late December 1965. Activity at Jólnir continued until early August 1966, and was entirely pyroclastic. The tephra island of Jólnir was washed away by late October.
Jólnir's was the last pyroclastic activity at Surtsey Volcano. However, effusive activity resumed on the main island as Jólnir became inactive, from the original Surtur I vent. Mid August 1966 through to early June 1967 saw variable effusive activity in and around Surtur I, and the opening of several satellite vents in its vicinity. The extensive lava flows on the main island of Surtsey ensured its short-term resistance to erosion by wave action, long enough for palagonitization to cement the bulk of the Surtur I and II tephra cones into resistant tuffs 18,19 . In total, the eruption lasted 3.5 years (Fig. 2), and produced a total of 1.1-1.2 km 3 of material (60-70% tephra), with 0.07 km 3 at each of Syrtlingur and Jólnir, and 0.01 km 3 at Surtla 15 .

Previous studies on Surtsey
Surtsey has been legally protected since its emergence, and has become an important site for multidisciplinary science. Intensive studies, mainly in Earth Sciences, Terrestrial Biology and Marine Biology, provided the background for the successful nomination of Surtsey as a UNESCO World Heritage Site. The above headings are of course extremely broad, and a bibliography of works relating to Surtsey is maintained by the Surtsey Research Society (www.surtsey.is).
Sampling of Surtsey eruption products was systematic, but geochemical analyses have appeared sporadically in the literature (Fig. 2). The first report of whole-rock composition was for a single tephra erupted on Dec 1, 1964 5,24 . The addition of two additional analyses by Steinþórsson 25 then showed a major-element trend from tephra more evolved at Surtur through to more primitive at Syrtlingur. Jakobsson presented additional analyses when defining the VE-I and VE-II magmas 13,28 ; he noted that the latter was lower in MgO (VE-I: MgO = 9.1-10.1 wt% for those without olivine cumulates, VE-II: MgO = 5.4-8.2 wt%, Vestmannaeyjar wide, as per ref. 13) and was probably the fractionated daughter of the former. In 1981, scuba divers visited the eroded remnants of Surtla to collect three additional whole-rock samples 31 . Kokelaar and Durant 31 found Surtla whole-rock samples to be of the VE-II type, as opposed to previously analysed Surtsey lavas which were all VE-I, and found them to be more evolved than previously reported. This allowed the authors to expand upon the concept that Surtsey erupted progressively more primitive magma during the evacuation of a vertically zoned magma chamber. Other chemical data from Surtsey rocks has included major-element analyses of starting materials used for experimental work on the evolution of alkali basalts 22,30 , and early Sr-isotope and REE analyses of Vestmannaeyjar lavas 32 .
The petrogenetic study of Vestmannaeyjar lavas by Furman et al. 27 (based on a PhD thesis 33 ) had the same aims as the analysis of the Surtsey Magma Series presented here. Furman and colleagues examined the same Surtsey samples for which we present major-element chemistry (Table 1), measuring trace elements and Sr-Nd-and Pb-isotopes on a subset of the lavas. They extensively discussed major-element trends in their assessment of magma evolution, with graphical presentation of selected major-element oxides (see Figs 2-4 of Furman et al. 27 ), and tabulation of the major-element chemistry of samples 9850 and 4638 (see Table 6 of Furman et al. 27 ) -the primitive and evolved end-members used in their fractionation modelling. It is the primary objective of this paper to officially and completely present the full suite of Surtsey major-element data.

Samples in the Surtsey magma and melt series
The samples used to define the Surtsey magma and melt series are housed at the Icelandic Museum of Natural History in Gardabaer, and are listed in Table 1.
Whole-rock major-element analyses are from a series of twelve scoria, bomb, and lava samples erupted from late December 1963 to May 19, 1967. All vents are represented in the whole-rock suite. Bombs and lava from Surtur I and II were collected on the island of Surtsey at various stages of activity. Scoria bombs from Surtla and Syrtlingur were dredged from their respective edifices, and were collected directly from the months-long ephemeral subaerial portion of Jólnir.
Major and trace-element analyses of sideromelane glasses in tephra collected on specific dates during the eruption cover pyroclastic activity from all vents, excluding Surtla. Collection of tephra from Syrtlingur and Jólnir was facilitated in part by their fallout deposition onto the main island of Surtsey 3 . The eruption dates for many of the tephras is known with high precision, some having been collected as they rained out directly from the eruptive plume onto the decks of various ocean-going vessels that were nearby or that were deliberately piloted into fallout zones (anecdotally against the better judgement of some of the skippers). In 1979, a joint effort by the Icelandic Museum of Natural History and the Geothermal Research program of the US Geological Survey achieved the drilling of a 180.6 metre-deep borehole into Surtsey 15,34 . Core recovery on the borehole was very good, and the Surtsey magma series includes major-element analyses of sideromelane glass in tephras extracted from the core from 180.1 to 0.1 m depth. We interpret these glasses to represent the pre-emergent growth of the island of Surtsey in the lead up to the establishment of the emergent Surtur I vent.

Analytical methods
The data presented here represent a combination of historical and new analyses from several laboratories.
Whole-rock major-element analysis (ref. 1 in Table 1) was by X-ray fluorescence (XRF) on fused glass beads at the Rock Geochemical Laboratory of the Geological Survey of Denmark and Greenland (Copenhagen). Fe 2 O 3 /FeO was determined by titration, and loss on ignition (LOI) was determined at 1000 °C 35 . Whole-rock trace-element analyses are given in Furman et al. 27 .
Glass major-element compositions by electron probe microanalysis (EPMA) are reported from several laboratories, and historic and new analyses are both presented where appropriate. Original analysis of drill core glasses (ref. 3 of Table 1) was with an ARL-EMX EPMA at the U.S. Geological Survey, Menlo Park, 1981, using 15 kV potential, 25 nA current and 10 μ m rastered beam 19 . Original analysis of time series tephras (ref. 4 in Table 1) was with a JEOL 733 superprobe at the Massachusetts Institute of Technology, in 1987, using 15 kV, 10 nA, and 10 μ m rastered beam. New analyses of drill core glasses (ref. 5 in Table 1 Trace-element analysis of sideromelane glasses (ref. 7 of Table 1) was by laser ablation inductively coupled mass spectrometry (LA-ICPMS) with a Resonetics RESOlution M-50-LR 193 nm excimer laser ablation system coupled to an Agilent 7500 cs/ce Quadrupole ICP mass spectrometer at the University of Otago. Data were acquired with a spot size of 25-35 μ m, ablating for 60 s at 5 Hz, using He as a carrier gas. Data was reduced offline using Si (from EPMA) and multiple replicate analyses of reference standards BHVO-2G and BCR-2G.
Modelling of major elements uses average olivine phenocryst and plagioclase microphenocryst compositions from Surtsey (supplementary Table S1); average clinopyroxene compositions from experimental phase equilibria studies 22 , and oxide (magnetite + ilmenite) compositions from SteinÞórsson 26 . Trace elements are modelled using standard basalt partition coefficients as compiled in Rollinson 36 , except where noted otherwise.

The Surtsey magma and melt series
The petrography of the Surtsey Magma Series samples is consistent with the general descriptions for VE-I and VE-II types 13,28 . Mineral chemistry is given in Supplementary Table S1. Phenocryst compositions show cr-spinel ranging from Cr# 26-42 (average 32; Cr# = Cr/Cr + Al), olivine ranging from Fo 75 -Fo 87 (average Fo 82 ), and plagioclase ranging from An 53 -An 80 (average An 70 ). None show systematic zoning, and while olivine compositions do not vary systematically through time, plagioclase shows a very slight trend to more anorthite rich compositions. Except for two diopside crystals in one sample of Jólnir tephra (sample 9843), clinopyroxene is not observed as a phenocryst or microphenocryst phase in samples of the magma series, and is present only in the groundmass. Diopside is present along with olivine and plagioclase in rare, small (2-4 cm) gabbroic nodules found in Surtsey tephra and other lavas of Vestmannaeyjar 37 .
Major-element compositions of whole rocks ( Table 2) varied dramatically and systematically during the course of the eruption (Fig. 2). The earliest-erupted whole-rock sample reported here, a scoria bomb from Surtla (4638), is the most evolved in the suite (MgO = 7.01 wt %, K 2 O = 0.68 wt %). Early activity from Surtur I did not yield appropriate samples for whole-rock analysis, but bomb and lava samples from the extended activity at Surtur II showed a marked evolution toward more primitive compositions Temporal trends in tephra glass chemistry (Table 3) are similar to those in whole-rocks, in that they show a progression toward more primitive compositions over time. Glass from tephra in the 1979 drill core overlaps for all major and trace elements with those from Surtur I+II, and glasses from these tephras are chemically indistinguishable (Table 4; Figs 2,3). The earliest reliably time-stamped tephra samples collected after emergence are from pyroclastic activity at Surtur I, and are the most evolved in the suite, both in terms of major (MgO = 5.36 wt %, K 2 O = 0.86 wt %) and incompatible trace elements (Fig. 3). Tephra from Surtur II overlap with that from Surtur I, but with a trend toward less-evolved compositions. Glasses from Syrtlingur and Jólnir are noteworthy firstly for being nearly identical to each other Scientific RepoRts | 5:11498 | DOi: 10.1038/srep11498 in all major and trace elements (except for a slight tendency toward higher SiO 2 and lower incompatible element concentrations at Jólnir; Figs 2-4), and secondly by being notably and ubiquitously less evolved, and more incompatible-element-depleted than Surtsey glasses (Figs 3,4).
Tephra glass is related to whole-rock analyses from the same time and/or vent, by olivine control, in keeping with the former representing melt and the latter representing melt + crystals (Fig. 3). Tephra glass on the main island of Surtsey, including at all levels in the 1979 drill core, shows some variability, but is bracketed by olivine control from the Surtla sample (4638, Dec. 1963) and the most-evolved and earliest-erupted whole-rock sample from Surtur II (9023, Feb. 1964). Some of the variability in Surtsey tephra glass can be related to minor plagioclase (Na 2 O, CaO, Sr) and Fe-Ti oxide (TiO 2 , FeO t ) microphenocrysts, which are visible in the grains. Tephra glass from Syrtlingur and Jólnir shows no variation beyond analytical uncertainty, and similarly lies along olivine control from associated whole-rocks from the same vents. In this sense, the analysed glass from each vent's tephra faithfully represents the melt fraction of the associated whole-rocks in the Surtsey Magma Series. The fact that glasses from Syrtlingur and Jólnir are dramatically different from those of Surtsey itself indicates that the more-primitive signature of whole-rocks from the satellite vents and later Surtsey whole-rocks is not simply the product of olivine accumulation.

Discussion
Was Surtsey's magma a fractionation series, or a mixing series? In their modelling of the Surtsey Magma Series, Furman et al. 27 found that compositions could be related by up to ~48% fractionation of the assemblage: cpx (44-50%) > ol (26-27%), > plag (17-24%) > Ti-magnetite (6-7%), with stepwise-calculated mineral compositions and proportions. An issue noted as enigmatic by Furman et al. 27 is that although clinopyroxene is the main phase in the modelled fractionation assemblage, it is not present as a phenocryst phase in Surtsey samples 13 , except in one sample of Jólnir tephra as noted above. A similar situation has been noted for lavas from Heimaey 38,39 , which is ~18 km NE of Surtsey, and is the site of the most recent eruption in the Vestmannaeyjar archipelago (Fig. 1A, inset) 37 . At both Surtsey and Heimaey, authors have embraced the "pyroxene paradox" e.g. 40 , relying on efficient, density-driven, clinopyroxene-melt separation at depth to explain the correlation between CaO/Al 2 O 3 (and also Sc/Y 38 ) versus MgO (Fig. 4).
Although the correlation between CaO/Al 2 O 3 and Sc/Y versus MgO are strong in the Surtsey Magma Series, we note several issues with invoking clinopyroxene-dominated fractionation as the main control on variation among Surtsey's magmas. First, although mechanisms for the efficient separation of clinopyroxene due to density contrast with the melt have been discussed by several authors 41-43 , we question whether such a mechanism can be 100% efficient, given the large amounts of clinopyroxene required 27 .   3  3  3  3  3  3  3  2  3  3  2  3  3  3  3  3  3  3  3  3  3  3  3  3  3 3 3    27 . Grey arrows show direction and magnitude for 5% fractionation of different phases. Major elements modelled using measured Surtsey olivine, plagioclase (Table  S1), and oxide 26 compositions, and compositions of clinopyroxene formed in high-(HP) and low pressure (LP) experiments 22 . Note that for all but CaO, the distinction between HP and LP Cpx is trivial. Trace elements modelled using standard distribution coefficients compiled in Rollinson 36 .
Scientific RepoRts | 5:11498 | DOi: 10.1038/srep11498 Second, in 1 atm melting experiments on early Surtsey lavas 30 olivine crystallized at 1220 °C, plagioclase at 1180 °C and clinopyroxene not until 1155 °C. We note that the temperatures of lavas at the surface of the Surtur II crater (e.g., exposed to the air and undoubtedly cooler than anywhere at depth) in March 1965 were still hot, 1151-1162 °C 44 , but density-driven separation of clinopyroxene would have required that it formed early, at high temperature. Third, there is a discrepancy in the trends that would be generated at high versus low pressure. Furman et al. 27 inferred that evolution of the Surtsey magmas would have taken place at pressures > 10 kbar (~30 km). Phase-equilibrium studies by Thy 22 , however, showed that subcalcic diopside crystallized in high-pressure runs (10-30 kbar) and that calcic diopside only crystallized in 1 atm experiments. For CaO/Al 2 O 3 to vary strongly with MgO, only certain clinopyoxene compositions can be used in fractionation modelling (Fig. 4A). Furthermore, although Sc is moderately compatible in clinopyroxene, its partition coefficient (K D Sc) is dependent both on temperature and on composition e.g, 45 . Experiments 22 showed clinopyroxene to be more magnesian at high pressure (En 40-58 %) than at 1 atm (En 36-41 %), which could mean a K D Sc lower by 1/2 or 2/3 for high-P clinopyroxene 46 .
For alkali basalts such as those of Iceland's EVZ, a range of 1.31-3.5 is probably realistic 47,48 , with lower K D Sc for higher-T. Thus, the fits of fractionation models to observed Sc/Y trends in Surtsey magma depend strongly on the choice of K D Sc (Fig. 4B). The ambiguities outlined above are not in themselves sufficient to discard cpx fractionation as an important control on relationships among samples of the Surtsey Magma Series. However, trends in some trace-element ratios cannot be explained solely by fractionation. In particular, incompatible trace-element ratios such as chondrite-normalized La/Sm N , Th/Hf, Ba/Sr (Figs 5A-C), K/Rb and Zr/Y (not shown), and to a lesser extent Nb/Zr (Fig. 5D), show systematic variability through time to a degree that cannot have been produced by crystal fractionation of a clinopyroxene-dominated assemblage (scaled arrows, Fig. 5).
We suggest that trends in trace-element ratios can be better explained by magma mixing. End members for this inferred mixing are, however, only approximated by published compositions. The enriched component is an alkalic basalt similar to lavas from the EVZ at Eldgjá 49 , which have been used as a proxy for the main basaltic component on Iceland in investigations of ridge-plume interaction along the Reykjanes Ridge 50 , and/or those from the Vestmannaeyjar at Heimaey 38 . Lava from Surtla (4638) is compositionally matched in many trace elements to lavas from Eldgjá, the earliest lavas from Surtur I are better matched to the young (< 5 ka) eruptives on Heimaey that form a rift-parallel array 51 . An important note, however, is that all Surtsey eruptives have significantly lower Nb/Zr than alkalic lavas of the EVZ and Vestmannaeyjar (both for whole rocks analysed by Furman et al. 27 and for glasses analysed in this study), indicating that they may include an as-yet unidentified component. The depleted component is a ridge-like basalt broadly similar to the tholeiites of the submarine Reykjanes Ridge 50,52 , which compositionally approximates global N-MORB 53 on average, but varies with latitude as the northern segment interacts with Iceland 14,50,52 . Figure 5 shows temporal trends for several incompatible trace-element ratios of the Surtsey Magma Series, compared with the aforementioned end-members. Although all  27 , of ~48% fractionation of the approximate assemblage cpx (47%) > ol (27%) > plag (20%) > mgt (6%), with stepwise-calculated equilibrium mineral compositions. Grey arrows show (in 5% increments) fractionation of the same assemblage but with measured ol + plag (Supplementary Table S1), mgt 26 , and the high-P (> 10 kbar, solid) and 1-atm (dashed) experimental cpx compositions from Thy 22 53 . Black arrows to right of each frame shows the effect of 45% (total length of arrow, marked in 5% increments where appropriate) fractionation of cpx-dominated assemblage described by Furman et al. 27 , using starting composition of sample 9850 and standard basalt partition coefficients as compiled by Rollinson 36 . Note that in all cases, trends are from more Icelandic alkalic basalt end members toward more depleted ridge-like end members over time, and that with the exception of Sc/Y, fractionation cannot explain the degree of variation in trace-element ratios. Degree of mixing is not equivalent for all ratios, suggesting that mixing is heterogeneous, and some ratios (e.g., Nb/Zr) are not well represented by lavas from Eldgjá or Heimaey, suggesting that the off-axis location of Surtsey involves as-yet unconstrained, end members. trace-element ratios trend from more EVZ-like Surtla and Heimaey-like early Surtsey to more ridge-like Syrtlingur, Jólnir, and late-Surtsey, the ratios of mixing between the plotted end-members are not consistent (Fig. 5). This is probably because end-member magmas in the off-axis propagating tip of the EVZ remain poorly identified, not having been studied as extensively as magmas of the on-axis Reykjanes Ridge-Iceland Plume region 14,32,49,50,52,[54][55][56][57][58][59][60][61] . We note, additionally, that although temporal trends in Sc/Y could be explained by crystal fractionation, they can be equally well explained by mixing (Fig. 5E).
Isotopic ratios are often good indicators of magma mixing because they are unaffected by crystal fractionation. Surtsey lavas analysed by Furman et al. 27 (not shown), which generally supports a mixing hypothesis. However, because there is significant overlap in the isotopic signatures of these end members, and because isotopic ratios of Surtsey lavas did not vary (beyond analytical uncertainty) during the course of the eruption 27 , isotopic ratios are non-discriminating of magma mixing processes at Surtsey.
Magma mixing as an alternative to crystal fractionation at Surtsey is a simple but attractive hypothesis for several reasons. In addition to explaining much of the variability in compositions listed above, it also explains major-element trends. Specifically, although Vestmannaeyjar is the SW extension of the alkali-olivine basalt region of Iceland 13,67 , only Surtla whole-rock samples plot unambiguously above the alkalic-tholeiitic divide in total alkali-silica space (not shown). The rest of the samples show a temporal trend through transitional space toward tholeiitic compositions. Petrographically, mixing does not require the crystallization of large amounts of clinopyroxene, which is absent except as a groundmass phase and in rare xenoliths; we consider it preferable to avoid drawing upon unseen processes, particularly when doing so results in a paradox. Physically, the injection of new, hot, primitive magma into a pre-existing body of more evolved magma is a well-known mechanism for triggering an eruption e.g, 68 and for sustaining that eruption so long as the supply of new magma persists.
Perhaps the most important argument against fractionation of clinopyroxene in the Surtsey Magma Series involves the small size of the edifice, and the short duration of the eruption. The pyroxene paradox determined for MORB glasses has been defined over long ridge segments, on dredged materials, for which eruptive relationships cannot be accurately known [40][41][42][43]69 . In the Vestmannaeyjar, the lavas of Heimaey that are apparently related by clinopyroxene-dominated fractional crystallization are from ten volcanic centres erupted over tens of thousands of years throughout the Holocene, and into the Upper Pleistocene 38,39,51 . Surtsey's eruption was an essentially continuous event lasting less than four years. Invocation of disparate crystallization histories for pulses of magma erupted only weeks apart from vents separated by only 100's of metres indicates a physically intractable conception of rates and scales of fractionation. In contrast, successive injections of ridge-like magma into more shallowly stored alkalic magma (and other minor components) might be expected to continue through the brief duration of a single eruption.
Regional perspectives on magma mixing at Surtsey. Magma mixing has been established as an important process where the MAR and a mantle plume interact in the Iceland region. Recognition of northward increasing chondrite-normalized La/Sm N by Schilling 14,54 was instrumental in drawing attention to increased ridge-plume interaction along the Reykjanes Ridge toward Iceland. Following this key finding, models of ridge-plume interaction, mantle heterogeneity, and geochemical structure of the Iceland plume have been greatly refined through intensive petrological work 14,32,49,50,52,[54][55][56][57][58][59][60][61]65 . That the end-members for our mixing hypothesis at Surtsey are not rigorously matched by compositions from these studies is not surprising given the complexity of Iceland's magmatic systems. Murton et al. 50 , for example, identified up to six mixing end members along the Reykjanes Ridge, and these interact with the Iceland plume that is itself heterogeneous 66,70 .
Evolution of propagating rifts involves magma that resides in small ephemeral magma chambers becoming progressively more connected as the rift propagates and magma supply increases, until full connectivity is achieved 71,72 . This model has been applied to explain the petrology of lavas at Heimaey 38 , where magma bodies that are still relatively isolated are envisaged to be undergoing polybaric fractionation. We suggest that Surtsey, which is further along the SW propagator than is Heimaey, represents an even more embryonic manifestation of magmatic diversity at a propagating tip. Although the more evolved lavas of Surtsey are geochemically similar to those from Heimaey, there are significant differences in some trace elements, including an offset to lower Nb/Zr (Fig. 5D), that suggest limited communication between the magma bodies of Surtsey and those of the Vestmannaeyjaar as a whole 73 . Furthermore, the irregularity of mixing proportions from one Surtsey vent to another (e.g., Th/Hf in Fig. 5C), suggests limited physical communication among the vents' respective feeder systems.
Finally, the appearance of small proportions of a ridge-like mixing end member at Surtsey is consistent with the current tectonic environment of SW Iceland. The WVZ is now receding, the EVZ propagating, and the South Iceland Seismic Zone is accommodating an active ridge jump from the WVZ to the EVZ by transverse faulting 74,75 . This may even lead to the EVZ linking up with the submarine Reykjanes Ridge 38 as it replaces the WVZ as the main rift in Southern Iceland.
Local perspectives on magma mixing at Surtsey. The Surtsey Magma Series documents a temporal progression from more evolved to more primitive compositions. This could suggest evacuation of a vertically zoned magma chamber(s) 31 , but does not make explicit the origin of this zonation. We suggest that the injection of a primitive, ridge-like component into pre-existing, poorly connected volumes of broadly EVZ/Vestmannaeyjar-like alkali basaltic magma caused overpressure to develop in the stored magma system, and caused the eruption that produced Surtsey and its siblings.
Early eruptive products from Surtla (and to a lesser extent Surtur I+II) are the most EVZ-like in the suite (although lower in Nb/Zr), suggesting that they partially represent magma delivered by along-rift transport from the more established EVZ to the NE. However, although Surtla erupted the most evolved magma in the suite, but with 7.01 wt% MgO it is still relatively primitive, as are early lavas on Surtsey. This indicates a relatively short residence time of this magma in shallow storage below what would become the site of Surtsey Volcano -short enough to preclude significant closed-system crystallization and/or differentiation. The regional along-rift disturbance and deformation that would have accompanied early injection of EVZ-like magma into the Surtsey region was probably instrumental in initiating dike propagation from the ridge-like source along axis, into the same area. Injection of this depleted, ridge-like magma caused it to partially mix with ponded EVZ-like magma under Surtsey itself, and this is inferred to have been the most volumetrically significant stored sub-volume. The fact that Syrtlingur and Jólnir saw the eruption of very primitive magma probably indicates that the pre-eruption reservoirs of EVZ-type magma were minimal in these areas. The eruption of similarly primitive lava once activity returned to Surtsey, marked the impending exhaustion of all of the original EVZ-type magma.
Feeder dikes are common on the island of Surtsey, and thin dikes are present at two intervals in the 1979 drill core 15 . Also, gravity surveys on Surtsey have shown separate planar nonfragmental basaltic bodies (e.g, dikes) feeding the Surtur vents 76 . Injection of ridge-like magma into the subsurface at Surtsey would also have produced sub-vertical dikes, given the rift setting and fissure-like beginnings of the eruption 1,2,5,15 . If these were of similar small size as the feeder dikes (10 s of centimetres thickness), with short residence time available for mixing during an active eruption, it is not surprising that mixing should vary through the different times and vents of the Surtsey eruption.

Conclusions
The 1963-67 eruption of Surtsey (Iceland) is the canonical example of shallow-to-emergent phreatomagmatic eruptions. Previously unpublished and new analyses of whole-rocks and sideromelane glass in tephra from time-and space-located samples throughout Surtsey's eruption respectively define a Surtsey Magma Series and Surtsey Melt Series. Progressive changes in the geochemistry of eruptive products can be explained by mixing between ephemeral and poorly connected volumes of broadly EVZ-like alkalic basalt typical of Iceland's Vestmannaeyjar, with primitive ridge-like basalt formed at the propagating ridge tip. Surtsey not only demonstrated the dynamics of magma-water interaction in subaqueous eruptions, but also exemplifies processes occurring at the extreme propagating tip of an off-axis ridge in the vicinity of the Iceland plume.