Insights into silicic melt generation using plagioclase, quartz and melt inclusions from the caldera-forming Rotoiti eruption, Taupo volcanic zone, New Zealand

Abstract

The Rotoiti (~120 km³) and Earthquake Flat (~10 km³) eruptions occurred in close succession from the Okataina Volcanic Centre at ~50 ka. While accessory mineral geochronology points to long periods of crystallization prior to eruption (10⁴–10⁵ years) and separate thermal histories for the magmas, little was known about the rates and processes of the final melt production and eruption. Crystal zoning patterns in plagioclase and quartz reveal the thermal and compositional history of the magmatic system leading up to the eruption. The dominant modal phase, plagioclase, displays considerable within-crystal zonation: An_37–74, ~40–227 ppm MgO, 45–227 ppm TiO₂, 416–910 ppm Sr and 168–1164 ppm Ba. Resorption horizons in the crystals are marked by sharp increases (10–30%) in Sr, MgO and X_An that reflect changes in melt composition and are consistent with open system processes. Melt inclusions display further evidence for open system behaviour, some are depleted in Sr and Ba relative to accompanying matrix glass not consistent with crystallization of modal assemblage. MI also display a wide range in XH₂O that is consistent with volatile fluxing. Quartz CL images reveal zoning that is truncated by resorption, and accompanied by abrupt increases in Ti concentration (30–80 ppm) that reflect temperature increases ~50–110°C. Diffusion across these resorption horizons is restricted to zones of <20 μm, suggesting most crystallization within the magma occurred in <2000 years. These episodes are brief compared to the longevity (10⁴–10⁵ year) of the crystal mush zones. All textural and compositional features observed within the quartz and plagioclase crystals are best explained by periodic mafic intrusions repeatedly melting parts of a crystal-rich zone and recharging the system with silicic melt. These periodic influxes of silicic melt would have accumulated to form the large volume of magma that fed the caldera-forming Rotoiti eruption.

Appendices

Appendix: Data repository

Contents

DR1:

Analytical conditions

DR2:

Melt inclusion and matrix glass analyses

DR3:

Feldspar analyses

DR1: Analytical conditions

Glass

Both matrix glasses and MI glasses analysed were microlite free and isotopic showing no significant sign of alteration. The inclusions were optically examined for signs of post-entrapment chemical alteration. Most MI appear to be bubble free, but a few do have small bubbles (<5% of the inclusion volume).

For MI and matrix glass, an accelerating voltage of 15–20 kV, low beam current (2–4 nA) and defocused (10 μm) beam were used for EMP analyses to minimize Na migration. The instrument was calibrated for each set of beam conditions using a suite of appropriate glass and mineral standards. The following count times were used: 20 s for Si, 60 s for Ti, 10 s for Al, 40 s for Fe, 40 s for Mg, 40 s for Mn, 30 s for Ca, 6 s for Na, 20 s for K and 40 s for P. Analytical totals for matrix glasses are in the range 92–97 wt% consistent with secondary meteoric hydration observed in many OVC deposits.

Approximately, 45 Rotoiti and 12 EQF quartz-hosted MI, and 47 Rotoiti and 7 EQF matrix glasses were analysed using SIMS techniques. H₂O, CO₂ and other volatile elements were analysed in approximately 30 Rotoiti and 5 EQF MI. A 2–6 nA, 10 kV (nominal) ¹⁶O⁻ beam was used for trace element analyses, which corresponded to a beam diameter of approximately 7–15 μm at the sample surface. The secondary ions were sampled at 4.5 keV with an offset of 75 eV, ±25 eV energy window and 25-μm image field. Trace elements counts for each analysis were collected for 10 cycles after a 2 min pre-raster to minimize any surface layer contamination. The following elements were analysed using a 6 nA beam, count times for each cycle are in parentheses: ¹H⁺ (50 s), ⁷Li⁺ (30 s), ¹¹B⁺ (50 s), ¹²C⁺ (50 s), ²⁵Mg/2⁺ (50 s), ¹⁹F⁺ (50 s), ²⁶Mg⁺ (50 s), ³⁰Si⁺ (20 s), ³⁵Cl⁺ (50 s), ⁴²Ca⁺ (20 s), ⁵⁶Fe⁺, ²⁸Si⁺ (50 s), ⁸⁵Rb⁺ (50 s), ⁸⁸Sr⁺ (50 s), ⁸⁹Y⁺ (50 s), ⁹⁰Zr⁺ (50 s), ⁹³Nb⁺ (50 s), mass 131 = background (50 s), ¹³⁸Ba⁺ (50 s), ¹³⁹La⁺ (50 s), ¹⁴⁰Ce⁺ (30 s). A dummy mass was added at the start of each cycle to prevent magnet hysteresis. The vesicular matrix glass and some of the small MI were analysed with a smaller 2 nA beam, the volatile elements were not analysed using this set-up but most of the other elements listed above were analysed in addition to ¹⁴¹Pr⁺ (50 s), ¹⁴³Nd⁺ (50 s), ¹⁴⁹Sm⁺ (80 s), ¹⁵¹Eu⁺ (50 s), ¹⁵⁶Gd⁺ (30 s), ¹⁵⁷Gd⁺ (30 s), ¹⁵⁹Tb⁺ (50 s), ¹⁶¹Dy⁺ (50 s), ¹⁶⁵Ho⁺ (50 s). EMP Si contents were used to normalize these SIMS data.

Plagioclase

Major element concentrations in plagioclase were determined using an EMP, with an accelerating voltage of 20 kV, 40 nA and a ±2 μm beam. The instrument was calibrated for each set of beam conditions using a suite of appropriate standards. More than 550 major element analyses were performed, with most of these analyses along rim to core profiles across the zones observed on BSE images. A high (40 nA) beam current was used, along with the following count times: 10 s for Si, 15 s for Al, 120 s for Fe, 240 s for Mg, 10 s for Ca, 12 s for Na, 30 s for K. The high beam currents and long counting times for Mg (240 s) and Ti (120 s) were used to get higher precision and lower detection limits on these low-abundance elements. A lower beam current (20 nA) was used for some EMP analyses along profiles parallel to SIMS analyses, as only major elements were required to normalize SIMS data.

SIMS trace element analyses were carried out using a 5 nA, 10 kV (nominal) ¹⁶O⁻ beam, which corresponded to a beam diameter of approximately 10 μm. Secondary ions were sampled at 4.5 keV with an offset of 75 eV and 20 eV energy window. Ion yields were calibrated using NIST 610 glass and were corrected for differences in ion yield between glass and feldspar based on calibration against mineral standards of known trace element content. A mechanical stage was used for the profiles, with counts collected every ~25–50 μm along the profiles. At each point, ⁷Li⁺, ²⁶Mg⁺, ³⁰Si⁺, ⁴⁷Ti⁺, ⁸⁸Sr⁺, ¹³⁸Ba⁺ and ¹⁴⁰Ce⁺ counts were collected for 150 s over 3 cycles (n = 185 Rotoiti and n = 59 EQF). Si from EMP analyses performed parallel to the profiles was used to normalize the SIMS data.

Quartz

Ti in quartz was analysed using an EMP with an accelerating voltage of 20 kV, current of 200 nA and ~15 μm beam. Ti counts were simultaneously collected on 4 spectrometers, using 3 LPET and 1 PET crystal, for 240 s. Merging the Ti counts from the spectrometers resulted in low detection limits (<10 ppm), and errors (±10 ppm). Analyses (n = 159 for Rotoiti and n = 270 for EQF) were typically spaced ±30–50 μm apart along core to rim profiles.

The temperatures of quartz crystallization were determined from Ti concentrations using the TitaniQ equation of Wark and Watson (2006): T = −3765/[log (Ti_qtz/X_Ti) − 5.69], where T = temperature in Kelvin and X_Ti = activity of Ti in the quartz. The X_Ti was calculated by dividing the theoretical Ti present in rutile-saturated melt (Ti_rut–melt) by SIMS matrix glass Ti. Ti_rut–melt (in ppm) was calculated using the equation of Hayden et al. (2005), log (Ti_rut–melt) = 7.95 − (5305/T) + (0.124 9 FM), where FM is a melt composition parameter calculated from matrix glass analyses using the equation of Ryerson and Watson (1987).

Other phases

Fe–Ti oxide analyses were performed using a 20 kV, a 20 nA beam current, and a focused beam. Counts of major elements were collected for 10 s, while low abundance elements such as Mn, V, Cr, Ni, Zn, Nb were collected for 30 s. Temperature and oxygen fugacity were determined for Fe–Ti oxide pairs in contact or those within the same phenocryst, using the revised Ghiorso and Evans (2008) formulation. This revised geothermometer shows good agreement with experimental data, suggesting it provides the most reliable temperature and oxygen fugacity estimates (Blundy et al. 2009). Temperatures can either be calculated using Fe⁺²Ti ⇒ (Fe⁺³)₂ exchange or Fe⁺²⇒Mg exchange between the pairs. The Mg exchange geothermometer is much more sensitive to composition, and temperature estimates span a wider range (~600–1150°C). Rotoiti and EQF temperatures and oxygen fugacities determined by the Ghiorso and Evans (2008) method typically display similar ranges (e.g. 698–786°C and .35 to 0.94 NNO for Rotoiti T1; 722–763°C and −0.07 to 0.30 NNO for EQF) to those calculated using the Ghiorso and Sack (1991) 716–803°C and −0.42 to 0.71 NNO for Rotoiti T1; 736–767°C and −0.17 to 0.19 NNO for EQF) method which has been used in previous studies (Shane et al. 2005).

DR2: Melt inclusion and matrix glass analyses (Supplementary material)

DR3: Feldspar analyses

These compositions are similar to those reported in Shane et al. (2005). There is a mistake in Shane et al. (2005): X_An and X_Ab are transposed (Supplementary material).