Quartz is a common phase in high-silica igneous rocks and is resistant to post-eruptive alteration, thus offering a reliable record of magmatic processes in silicic magma systems. Here we employ the 75 ka Toba super-eruption as a case study to show that quartz can resolve late-stage temporal changes in magmatic δ18O values. Overall, Toba quartz crystals exhibit comparatively high δ18O values, up to 10.2‰, due to magma residence within, and assimilation of, local granite basement. However, some 40% of the analysed quartz crystals display a decrease in δ18O values in outermost growth zones compared to their cores, with values as low as 6.7‰ (maximum ∆core−rim = 1.8‰). These lower values are consistent with the limited zircon record available for Toba, and the crystallisation history of Toba quartz traces an influx of a low-δ18O component into the magma reservoir just prior to eruption. Here we argue that this late-stage low-δ18O component is derived from hydrothermally-altered roof material. Our study demonstrates that quartz isotope stratigraphy can resolve magmatic events that may remain undetected by whole-rock or zircon isotope studies, and that assimilation of altered roof material may represent a viable eruption trigger in large Toba-style magmatic systems.
The recent advent of micro-analytical techniques has dramatically increased our capabilities of crystal isotope fingerprinting with notable examples including micro-drilling associated with thermal ionisation mass spectrometry (TIMS), laser ablation multi-collector inductively coupled plasma mass spectrometry (LA-MC-ICPMS), and secondary ionisation mass spectrometry (SIMS) (e.g., ref. 1). These tools permit determination of isotope ratios on a sub-crystal scale, allowing the analysis of small sample volumes and thus production of high spatial resolution data within and between individual crystal growth zones (e.g., refs 2, 3, 4).
Crystal-scale oxygen isotope analysis is now routinely carried out using the SIMS technique2,3,4,5. For example, the oxygen isotopic composition of zircon is a robust source of information on the crystallisation environment3,4,6 and, as zircon also lends itself to U-Pb dating, it can help to identify and place constraints on the timing of geological events (e.g., refs 6 and 7). Quartz is a common mineral phase in high-silica metamorphic and magmatic rocks and, like zircon, is largely resistant to alteration and retains its magmatic δ18O values post-eruption8. Unlike zircon, however, quartz can grow to relatively large crystal sizes (e.g., up to 2 mm across in this study), which permits spatially detailed isotopic profiling with SIMS. This level of detail is more challenging to obtain in zircon as crystal size is typically ≤200 μm in most volcanic and metamorphic rocks, which limits intra-crystal analyses to usually a single core and perhaps one or two rim analyses (e.g., refs 2 and 9). Quartz may therefore be especially useful in certain high-silica magmatic settings, such as the Toba caldera in Indonesia, whose erupted products host plentiful quartz phenocrysts. Moreover, because oxygen diffusion in quartz is >104 times faster than in zircon at magmatic temperatures10,11, quartz can potentially provide insight into shorter timescale magmatic processes compared to zircon. Therefore, isotopic changes preserved between the inner and outer rims of quartz crystals may correspond to relatively short timescale magmatic processes e.g., on the order of hundreds of years prior to eruption (cf. ref. 12). Indeed, Matthews et al. (ref. 12) previously applied Ti-diffusion chronometry to quartz crystals from the Young Toba Tuff (YTT) and suggested that short, episodic magma recharge events on a ca. 100-year timescale were important in priming the large-volume YTT magma reservoir for eruption. In the current paper, we present detailed SIMS oxygen isotope analysis of quartz crystals from the YTT, including several of the quartz crystals analysed by ref. 12 for diffusion chronometry to i) test for magma recharge, mixing, and pre-eruptive crustal recycling in the YTT reservoir as recorded in the crystal-scale oxygen isotope record, and ii) assess the utility of oxygen isotope stratigraphy in quartz as a tracer of pre-eruptive magmatic processes in large-volume rhyolitic systems.
The Toba caldera, north Sumatra (Fig. 1), is thought to originate from subduction of the Investigator Ridge transform fault sited on the Indo-Australian Plate beneath the Sunda Trench, leading to formation of a deep crustal hot zone13,14 that feeds the Toba volcano. The Toba system has produced four major explosive eruptions over the last 1.2 million years15,16, the oldest being the 35 km3 Harrangoal Dacite Tuff (HDT). This event was followed by the quartz-rich 1,000 km3 Old Toba Tuff (OTT) ~0.79 Ma17, and the 60 km3 rhyolitic Middle Toba Tuff (MTT) ~0.5 Ma18. Finally, the cataclysmic 2,800 km3 rhyolitic Youngest Toba Tuff (YTT) erupted ca. 75 ka19. The YTT deposits extend over 20,000 km2 on Sumatra and are recorded as far as 2,000 km away (e.g. in India)20. The Toba super-eruption is the largest and best-known of the late Quaternary, and was speculated to have brought mankind to near extinction21.
High-silica caldera systems like Toba (Indonesia), Yellowstone (USA) and Taupo (New Zealand) produce some of the largest volume eruptions on Earth22. Recent work on magma petrogenesis of the Yellowstone hot spot track, for example, has revealed oxygen isotope diversity in zircon that includes both high- and low-δ18O values relative to ‘normal’ rhyolites9. For instance, the higher δ18O values in Yellowstone zircon are thought to represent assimilation of unaltered crust by ascending magmas, whereas the lower end of the δ18Ozircon spectrum has been attributed to recycling of hydrothermally-altered (low-δ18O) rhyolitic material into the Yellowstone magma system. Similar intra-crystalline isotopic diversity is observed in zircon from other high-silica magmatic systems worldwide, including Toba, where comparable processes are hypothesised to have been in operation (Fig. 2 and ref. 9). To fully resolve the complexity of the Toba crystal record, however, isotopic heterogeneities need to be resolved in more detail. We therefore use Toba as a case-study to assess the utility of quartz as a tracer of pre-eruptive magmatic processes, especially since Toba eruptive rocks are particularly quartz rich23,24. Furthermore, widespread low-δ18O crust is not expected at Toba’s latitude25, meaning that we can also test for the role of late-stage additions of volcanically produced high-temperature hydrothermally-altered crust in driving large silicic eruptions.
Petrography and geochemistry
The YTT samples used in this study are from stratigraphically constrained tuff units exposed along the caldera margin and from outcrops surrounding the Toba caldera. Bulk tuff and pumice samples exhibit no signs of alteration, and have a mineral assemblage comprising quartz, plagioclase, sanidine, biotite, amphibole, and minor pyroxene, zircon and oxides (cf. ref. 23). Bulk tuff and pumice samples have SiO2 contents between 71–73 wt.% (Supplementary Table 1), which fall within the mid-range of pumice compositions previously reported for the YTT (SiO2 = 68–77 wt.%) (e.g., ref. 23).
To provide comparative data for our SIMS approach, we also performed oxygen isotope analyses on YTT whole-rock tuff and pumice clasts by conventional fluorination and on separated crystal phases by bulk crystal laser fluorination (LF) (see Methods for details of analytical procedures). The whole-rock data have δ18O values between 8.2 and 9.9‰ ( = 9.0 ± 0.15‰, 1σ, n = 11; Supplementary Table 2), whereas the older HDT (1.2 My; 61 wt.% SiO2) has a δ18Owhole-rock value of 8.6‰ (n = 1). The YTT bulk crystal analyses by LF give δ18O values of 8.2 to 9.3‰ for quartz ( = 8.9 ± 0.15‰, 1σ, n = 13), 7.6 to 8.1‰ for feldspar ( = 7.8 ± 0.15‰, 1σ, n = 6), 6.8 to 8.9‰ for biotite ( = 7.2 ± 0.15‰, 1σ, n = 8), and 5.5 to 6.4‰ for amphibole ( = 5.9 ± 0.15‰, 1σ, n = 4). In comparison, the δ18Oquartz values from basement granitoids near Toba range from 6.4 to 9.5‰ (±0.15‰, 1σ, n = 5).
Quartz crystals in the YTT are typically fragmented due to the high explosivity of the eruption26, but a record of multiple dissolution and reprecipitation events can nevertheless be discerned (e.g., ref. 24). To resolve spatial δ18O zonation within individual quartz crystals, we coupled cathodoluminescence (CL) and SIMS oxygen isotope analyses of 15 quartz crystals from pumice (reflecting pre-eruptive magma batches) and from bulk tuff (encapsulating mixed material on eruption). We present a total of 191 individual SIMS spot analyses that define a range of δ18Oquartz from 6.7 to 10.2‰ ( = 8.8 ± 0.4‰, 1σ; Supplementary Table 3).
In order to assess the meaningfulness of our data in a wider Toba context, we furthermore analysed two clinopyroxene samples from the 1.2 Ma HDT for their 3He/4He ratios. We note that the YTT is virtually devoid of pyroxene (see ref. 23) and hence it was unfeasible to extract sufficient crystals for helium isotope analysis. The 3He/4He values obtained (0.67 RA and 1.76 RA, where RA = air 3He/4He) are lower than typical mantle values (8 ± 1 RA) or contemporary geothermal fluid values at Toba (6.6 RA; ref. 27), and instead approach crustal values (<0.5 RA). Moreover, one whole-rock YTT sample was also analysed for Sr-Nd isotopes and yielded an 87Sr/86Sr value of 0.714033 (±13) and a 143Nd/144Nd value of 0.512113 (±13). These radiogenic isotope data, although few in number, are consistent with previous studies that used radiogenic isotopes and which suggested that substantial crustal material was indeed involved in YTT petrogenesis28,29.
Resolving crustal differentiation processes at Toba
As is often the case in high-resolution analytical approaches2,30, our new SIMS data extend the established δ18O range for the YTT crystal data towards both higher and lower values, with an overall range of 3.5‰ for the full set of YTT quartz crystals (Fig. 3). Employing a crystal-melt fractionation factor (Δ18Oquartz-rhyolite) of 0.5‰ (e.g., ref. 31), the δ18O value of the Toba magma(s) can be calculated from the derived crystal data. The average calculated δ18Omagma value is 8.3 ± 0.4‰ (n = 191, 1σ) for SIMS data and 8.4 ± 0.2‰ (n = 13, 1σ) for the LF data. The datasets show significant overlap, thus ruling out standardisation bias. Moreover, they overlap with calculated δ18Omagma values determined using YTT zircon data from the literature9 (average δ18Omagma value = 8.6 ± 0.2‰; n = 50, 2σ) (Fig. 3). We note that the bulk crystal conventional fluorination quartz data (ref. 32; data obtained pre-1990) produce a slightly higher average δ18Omagma value of 9.1 ± 0.2‰ (n = 12, 1σ) compared to our bulk quartz data (Fig. 3). This is likely due to our use of the more modern laser fluorination technique that provides more accurate results. Nevertheless, it is important to note that all YTT δ18O quartz and corresponding magma values are elevated relative to primitive Sunda arc basalt, which is expected to have δ18O values between 5.5 and 6.1‰ (refs 33, 34, 35), and ~6.7‰ for its closed-system rhyolitic differentiates (ref. 36). Therefore, we suggest that values ≥7‰ in the YTT quartz crystals indicate addition of an external high δ18O component.
Turning our attention to the single crystal data available prior to this study, there is evidence for intra-crystal isotopic heterogeneity from a reconnaissance investigation of zircon from the YTT (beyond 1σ; maximum ∆18Ocore−rim = 0.72‰; Fig. 2 and ref. 9) but more data are required to fully resolve the issue of intra- and inter-crystal variability at Toba. Here, we take the opportunity to better resolve isotopic diversity within and between single crystal grains of our extensive sample suite, with the added advantage that our quartz crystals are larger than the zircon grains of the reconnaissance study9, thus permitting more detailed δ18O profiling through the crystals. Further, by adopting the SIMS approach, we overcome the blending effect that is usually associated with analysis of bulk rock or bulk crystals (cf. ref. 32). Indeed, individual YTT quartz crystals demonstrate core to rim variation of up to 1.8‰, and, importantly, ~40% of the analysed quartz crystals show a significant decrease (beyond 1σ) in their δ18O values from their cores to their outer growth zones (Fig. 4). The remaining ~60% of the analysed quartz crystals display limited resolvable core to rim variation in δ18O values along the analysis profiles. The possibility to conduct a large number of intra-crystal analyses in single quartz grains therefore allows us to record isotopic changes over small length- and hence time-scales, and so counters the averaging effect of whole-grain crystal data or spatially-limited intra-crystal traverses, as is often carried out for co-magmatic zircon (cf. refs 37 and 38) (Fig. 2).
The δ18Omagma values of the YTT previously recorded in zircon, and now also in quartz, are consistent with an open and dynamic magma reservoir that received variable crustal and magmatic inputs during its evolution (Fig. 5a). To explore the early quartz crystallisation history at Toba, we modeled how much assimilation of granitic crust with an average δ18O value of 10‰ would be required to achieve our average magma value of 8.3‰, equivalent to the average measured quartz core value of 8.8‰. We employ the following mass-balance mixing model:
where is the average quartz core δ18Omagma value (8.3‰), is the value of the starting magma, assumed here to have a δ18Omagma value of 7‰, is the oxygen isotope ratio of the new component that enters the system, set here as 10‰ (i.e. a whole-rock value in equilibrium with our measured crustal quartz of 9.5‰), and x is the mass fraction of this component. Assuming a rhyolite magma with a δ18O starting value of 7‰ and a crustal contaminant with 10‰, then approximately 43% assimilation of high δ18O granitic crustal material would be required to produce the magma δ18O value in equilibrium with our average quartz core value. However, the exact amount of assimilation is difficult to constrain due to the wide variation in crustal δ18O values, and because it is likely that the highest quartz core values may represent xenocrystals of crustal origin, found at Toba and elsewhere in the Sunda arc (e.g. refs 24 and 39). Our new Sr-Nd-He isotopic data corroborate the oxygen isotope data and provide further evidence for addition of significant amounts of continental crust to the Toba magmatic system; however, it remains an open question whether this crustal component was added in the lower or upper crust or a combination of both. Irrespective of the exact origin of the high δ18O crustal component, the long-term assembly of Toba magma(s) via mantle and particularly crustal contributions would be expected to result in a diverse crystal cargo (cf. refs 38 and 40) and would be an effective way to explain the high δ18O and 87Sr/86Sr values and low 143Nd/144Nd and He ratios that characterise the Toba Tuffs (e.g., refs 24, 28, 38 and 41).
Origin of the low δ18O component at Toba
If the YTT quartz dataset is considered as a whole, a t-test indicates that there is no significant difference between the cores (mean = 8.547; SD = 0.726) and rims (mean = 8.331; SD = 0.866); t (15) = 1.703, p = 0.109, but in addition to the commonly high δ18O values recorded in Toba quartz interiors, approximately 40% of the quartz crystals analysed by SIMS have lower δ18O values in their outer growth zones (Fig. 4a–f). A comparison of Figs 3 and 4 thus underlines that the Toba system demonstrates both large-scale homogenisation and crystal-scale heterogeneity characteristic of large silicic systems (cf. ref. 42). In order to evaluate rapid magmatic processes in large systems such as Toba, the key geochemical signals to evaluate are therefore subtle intra-crystal variations. As noted above, a percentage of 40% of crystals exist where downward changes in δ18O values are observed with progressive crystal growth (i.e. time), which thus requires the input of a much lower δ18O magma component than what would be inferred from bulk rock data and most bulk crystal analyses. Moreover, many cores and rims – apart from the outermost ones – probably grew at different times in different parts of the magma reservoir due to intra-reservoir convection and remobilisation processes42. The quartz crystals in the YTT have thus inherited a range of δ18O values, which when averaged produce a similar value overall. While we cannot always be sure that we have sampled the very outer rims of a crystal, the fact that intra-crystal changes tend to progress to lower values in some 40% of crystals analysed would imply that this incoming magma was recorded only in certain parts of the gigantic YTT magma reservoir, and thus effectively during the closing stages of pre-eruptive YTT chamber evolution.
Such a change towards relatively low δ18O values is also consistent with the concept of late-stage compositional diversification proposed previously for Toba12,38,43, and similar high-silica systems, such as Yellowstone or the Snake River Plain7,9. It is important to note that such a decrease in δ18O values cannot result from normal closed-system crystal fractionation. In addition, due to the preserved primary textures and intra-crystal compositional variations, we can exclude any post-crystallisation fluid overprint, which would result in irregular domains that reflect fluid fronts. Therefore, either basaltic replenishment with mantle-like δ18O values could have occurred12 or low-δ18O hydrothermally altered materials present in the shallow crust were added to the crystallising magma during the final stages of magmatic evolution, perhaps in a fashion similar to what was recently suggested for Yellowstone and for the Snake River Plain rhyolites7,9,44. In order to quantify the addition of a low-δ18O component to the YTT system, we again employ the mass-balance mixing model given in equation 1 above. In this model, is the lowest recorded oxygen isotope value in the crystal outer zone (δ18Omagma value = 6.2‰), while is the value of the starting magma, approximated here by the average quartz core δ18Omagma value (8.3‰). As before, is the oxygen isotope ratio of the new component that enters the system, and x is the mass fraction of this component. To provide a first order assessment, we used a relatively normal mantle-type δ18O value of 5.5‰ for basalt replenishment35, and we consider hydrothermally-altered volcanic material from the caldera roof to have δ18O = 0‰, as has been suggested for comparable high silica systems elsewhere7,44.
Our modelling suggests that replenishment by basaltic magma with δ18O = 5.5‰ would be required in a proportion of ~75% basalt to 25% rhyolite (at δ18O = 8.3‰) to shift host magma to a δ18Omagma composition identical to values calculated from the lowest quartz rim values of our study (lowest quartz rim δ18Omagma = 6.2‰; Supplementary Figure 1). However, this proportion of mafic magma is unrealistic because quartz would no longer crystallise from a mixed magma of such an intermediate composition since the SiO2 content would be far too low to form quartz45. Indeed, a melt SiO2 content exceeding 70 wt.% is required to crystallise quartz at the pressure of the YTT magma chamber23,45. Furthermore, mafic replenishment would be expected to produce low-δ18O rims grown directly onto resorption surfaces due to contact with mafic magma12 and this is not observed. More generally, the role of mafic replenishment in driving super-eruptions is debated and indeed they may not affect super-eruption type magma reservoirs significantly because incremental mafic replenishments would not usually over-pressurise or pervasively alter the composition of such a large reservoir (cf. refs 46 and 47). However, mafic replenishments can act as a heat source if there is underplating or injection into the lowermost parts of a silicic magma reservoir42. Crystal zones that formed following underplating could reflect a magma hotter than recorded by previous rims. In other words, the parental magma batches would have been heated to temperatures above zircon and quartz saturation at a given δ18O value, such that this hotter magma would have promoted assimilation of roof material. Subsequent cooling below zircon and quartz saturation would crystallise these minerals with diverse δ18O values, which would explain, at least in part, the degree of the isotopic diversity observed in YTT zircon and quartz. High-temperature hydrothermally-altered roof and wall material surrounding the magma reservoir is an therefore alternative low-δ18O component that could have been added, especially given the observation that widespread low-δ18O crustal country rocks are probably not abundant beyond larger volcanic centres in this part of the globe36. Hydrothermally-altered roof rocks can have δ18O values of ~0‰ or lower due to prolonged high-temperature interaction with meteoric waters7,9,44,48. In the case of Toba, the reservoir roof and wall rocks are either composed of silica-rich magmatic rocks or pre-existing granitoids and are probably heavily fractured due to previous caldera collapses15. These rock types are fusible over short timescales, especially when hydrated and hydrothermally overprinted43,48,49,50. Our model suggests that admixing ~25% of a rhyolitic melt derived from a high-temperature altered roof rock with a δ18O value of 0‰ can reproduce the lowest δ18O rim values of the analysed YTT quartz (Fig. 5b). This amount of assimilation is energetically feasible and in line with estimates from other large caldera systems (e.g., ref. 44). For these reasons, we favour pre-eruptive addition of low-δ18O hydrothermally altered crustal melts to explain our new SIMS oxygen data.
A corollary of the roof assimilation model is the initiation of a positive feedback loop wherein explosive shattering of H2O-rich silicic rock on contact with magma could accelerate assimilation51. Moreover, after several cycles of caldera-forming activity at Toba, the reservoir roof (i.e. the caldera floor rocks) would be highly fractured and liable to mechanical break-up by piecemeal intra-caldera structures on a variety of scales (e.g., refs 15 and 52). Such a faulted and fractured roof would provide pathways for infiltrating hydrothermal fluids, and would provide an easily mobilised reservoir of low-δ18O material spatially adjacent to the magma reservoir. Indeed, mechanical roof failure and roof subsidence into the magma reservoir is a highly probable eruption trigger at large caldera systems such as Toba52,53, especially as this caldera sits upon the Batak Tumor, an enormous swell or tumescence area around Lake Toba, which is related to emplacement of the large Toba reservoir and probably also to buoyant magma rise15 (Fig. 1). Associated roof uplift from emplacement and rise of silicic magma within this swell will have imparted an extensional regime onto the reservoir roof (e.g., refs 15, 46, 47 and 52), which would thus have been intensely fractured and weakened already prior to the YTT event and hence made susceptible to partial collapse and rapid recycling during caldera unrest.
The YTT whole rock and pumice samples were crushed, sieved, cleaned and picked for pristine quartz, feldspar, biotite and amphibole. For CL characterisation, quartz grains were mounted in epoxy resin and analysed with a CL detector attached to a JEOL microprobe at the Department of Earth Sciences at Bristol University. Whole rock powders were analysed for oxygen isotopes using a conventional line and mineral grains by using a laser fluorination line at the University of Cape Town. For SIMS analysis, quartz crystals were fixed in epoxy mounts and coated with gold. In-situ O-isotope measurements were made using a CAMECA IMS 1280 SIMS instrument at the Swedish Museum of Natural History, Stockholm (NordSIM). A 20 keV Cs+ primary beam of ca. 2.5 nA was used in critically-focussed mode together with a 5 μm raster to sputter a ca. 10 μm sample area. The runs comprised a 90 second pre-sputter period with a raster of 20 μm, and field aperture centering using the 16O signal followed by 64 seconds of data acquisition using two Faraday detectors in the multicollector system operating a common mass resolution of ca. 2500. Corrections for instrumental mass fractionation were determined using international reference material NBS-28 (silica sand), with further details given in refs 60 and 61. Full descriptions of all methods employed, including for Sr-, Nd-, and He-isotope analyses, are provided in the Supplementary Information.
How to cite this article: Budd, D. A. et al. Magma reservoir dynamics at Toba caldera, Indonesia, recorded by oxygen isotope zoning in quartz. Sci. Rep. 7, 40624; doi: 10.1038/srep40624 (2017).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work is dedicated to the memory of Prof. J.P. Davidson. We thank Dr. Ir. A. Djumarma Wirakusumah of the Geological Survey of Indonesia for granting permission to undertake fieldwork in Sumatra. M. Mahjum is thanked for logistical support in the field and S. Kearns, A. Berger, K. Lindén, L. Ilyinsky and B. Dahren are thanked for analytical support. We gratefully acknowledge the Swedish Research Council (VR), the Royal Swedish Academy of Sciences (KVA), Scripps Institution of Oceanography, Istituto Nazionale di Geofisica e Vulcanologia (INGV), the National Research Foundation (NRF) of South Africa, the Centre for Natural Disaster Science (CNDS), and the Otterborgska research fund at Uppsala University for generous financial support. NordSIM is financed and operated under a joint Nordic contract. This is NordSIM publication #486.