Abstract
The central Taupo Volcanic Zone (TVZ) is an actively rifting continental arc and is well known for its exceptionally high rate of rhyolitic magma generation and frequent caldera-forming eruptions. Two end-member types of rhyolites (R1 and R2) have been previously identified based on differences in their bulk-rock chemistry and mineral assemblage with hydrous phases crystallizing in the R1 type, which are not present or only rare in R2 rhyolites. Here we present new trace element and volatile data from rhyolitic melt inclusions measured in several representative eruptive deposits (R1 and R2 rhyolites) from the central TVZ to examine their volatile concentrations and origin. R1 and R2 show very distinct Cl concentrations, with R2 rhyolites being enriched in Cl by c. 1000 ppm. H2O is slightly higher in the R1 rhyolites, whereas CO2 concentrations are similar between the two end-member types. The origin of these volatile disparities between R1 and R2 melts is assigned to differences in the initial bulk volatile content of the parental magma, possibly associated with distinct input of fluids from the subduction zone. These disparities in bulk volatile concentrations can lead to variations in relative timing of exsolution of volatile phase(s) prior to melt inclusion entrapment.
Supplementary material: Major, trace and volatile composition for the analysed central TVZ rhyolites, and comparison of H2O data between the transmission and reflectance FTIR are available at http://www.geolsoc.org.uk/SUP18767.
Volatiles play an important role in the behaviour of magmas at depth and the style of volcanic eruptions, and they are integral to the development of magmatic-derived hydrothermal systems. The major volatile species in arc volcanic systems are H, C, S and Cl (e.g. Symonds et al. 1994; Giggenbach 1996; Fischer 2008) and they largely originate from the mantle wedge, the subducted altered oceanic crust and its overlying sediments, possibly with some small contribution from the continental crust (e.g. Hilton et al. 2002; Bindeman et al. 2004; Wallace 2005; Zelenski & Taran 2011; Chambefort et al. 2013). These components of subduction zones play an important role in the recycling of aqueous fluids (and fluid-mobile elements) from the surface environment, producing the distinct arc signature associated with subduction zone volcanism (e.g. Kay 1980; Hawkesworth & Ellam 1989; McCulloch & Gamble 1991). The composition and quantity of volatiles transported to the surface by the magmas are however governed by saturation state and timing of exsolution, which will strongly depend on pressure, temperature and composition of the melt and coexisting fluid (e.g. McMillan & Holloway 1987; Webster & Holloway 1988; Métrich & Rutherford 1992; Webster 1992; Burnham 1994; Carroll & Webster 1994; Student & Bodnar 1999; Newman & Lowenstern 2002; Aiuppa et al. 2009).
Magmatic degassing has been documented in many studies, and plays a significant role in eruptive processes and styles. Degassing has been quantified at many arc volcanoes by either monitoring of gas flux at the surface (e.g. Taran et al. 1991; Giggenbach 1996; Aiuppa et al. 2002; Edmonds et al. 2009; Zelenski & Taran 2011) or by assessing volatile concentrations in mineral-hosted melt inclusions and volcanic glass from eruptive deposits (e.g. Anderson et al. 1989; Wallace & Gerlach 1994; Wallace et al. 1995; Métrich & Wallace 2008; Johnson et al. 2010, 2011; Reubi et al. 2013). Crystallization-driven exsolution of fluid phases (i.e. second boiling) has also been investigated in the context of porphyry and epithermal deposits, as the metal partitioning in these exsolved volatile phases likely plays a role in the genesis of ore deposits (e.g. Hedenquist & Lowenstern 1994; Lowenstern 1994b; Dilles et al. 2000; Sillitoe & Hedenquist 2003; Heinrich et al. 2004; Simmons et al. 2005; Williams-Jones & Heinrich 2005; Chambefort et al. 2008, 2013; Tosdal et al. 2009). The heat and mass transfer to hydrothermal systems and hydrothermal alteration will also be highly dependent on volatile release, which drive hydrothermal convection cells around magmatic systems (e.g. Lindgren 1907; Elder 1966; Norton & Knight 1977). For decades, studies have focused on volatile exsolution during crystallization prior to ascent and eruption of magma, and showed that evolved melts in most cases already reached volatile saturation while at low crystallinity before ascent and eruption (e.g. Anderson et al. 1989; Lowenstern et al. 1991; Lowenstern 1993, 1994a; Wallace & Gerlach 1994; Wallace et al. 1995; Johnson et al. 2011). As such, volatile exsolution can play a significant role at different evolutionary stages within a given magmatic system. In order to better understand the role of volatile exsolution in magmatic-hydrothermal systems, questions and/or uncertainties associated with the composition and concentrations of the magmatic fluid phase need to be resolved. For instance, can we assume that most evolved magmatic systems in the same tectonic environment will have similar volatile content and saturation, and exsolve similar fluid compositions? These are important questions that should be considered when assessing heat and mass transfer in active volcanic regions.
The actively rifting central Taupo Volcanic Zone (TVZ) is well known for its exceptionally high rate of rhyolitic magma generation and frequent caldera-forming eruptions (e.g. Wilson 1996), and the extensive petro-chemical research in the TVZ make it an ideal setting to study the diversity of volatile exsolution and fluid compositions. Heterogeneities in bulk-rock geochemistry and petrology have been identified among the erupted rhyolites (e.g. Ewart et al. 1975; Deering et al. 2008), forming a compositional range between two end-member types (rhyolite R1 and R2; Deering et al. 2008, 2010). This diversity in rhyolite types is suggested to be related to a range in volatile compositions within their respective magmatic systems (Deering et al. 2008, 2010); however, little comparative work focusing on characterizing dissolved volatile compositions has been done (cf. Smith et al. 2005; Liu et al. 2006; Johnson et al. 2011). Here, we have directly measured pre-eruptive volatile compositions dissolved in the melts trapped in quartz- and plagioclase-hosted melt inclusions from seven major TVZ ignimbrites and associated fall deposits erupted between c. 320 and 0.7 ka, sourced from four distinct volcanic centres (Fig. 1; Table 1). Volatile and trace element concentrations are compared among the volcanic centres to assess the main factors responsible for their distribution. We particularly focus on addressing the R2 rhyolite type with the Rotorua and Ohakuri volcanic centres, as they are less well studied compared to the R1 type and fewer data exist in the literature. The volatile species, origins, concentrations and saturation are discussed with respect to the depth of magmatic systems and their spatial distribution.
Map of the central TVZ, New Zealand. Caldera boundaries and structures after Rowland et al. (2010). Upper left inset shows a photomicrograph of a quartz crystal with melt inclusions from the Ohakuri ignimbrite.
Summary of ignimbrites from the central TVZ analysed in this study.
Geological background
The TVZ is a rifted-arc currently subject to a NW–SE extension of 5–15 mm a−1, from SW to NE (Darby & Meertens 1995; Wallace et al. 2004). It is volcanically and structurally segmented, with its northern and southern segments dominated by andesitic cone-building eruptions (Wilson et al. 1995). The 125×60 km central segment has been active for at least 1.8 Ma and is characterized by explosive rhyolitic caldera-forming eruptions and effusive lava dome-building events, with comparatively minor andesites, dacites and basalts erupted (Houghton et al. 1995; Wilson et al. 1995, 2009; Eastwood et al. 2013). More than 6000 km3 of rhyolite magma has been explosively erupted from at least eight different caldera centres (Wilson et al. 2009). Rifting in the central TVZ is believed to have accelerated c. 0.9 Myrs ago (Wilson et al. 1995, 2009) and has led to the thinning of the crust, accompanied by voluminous basaltic magmatism intruding the lower crustal region from c. 16 to 30 km depth (Harrison & White 2006). Inter-relationships between rifting and accumulation and evacuation of large rhyolitic magma bodies are recognized (Rowland et al. 2010), and regional tectonic architecture partially controls the location and geometry of caldera structures (i.e. Okataina Volcanic Centre; Cole et al. 2010; Seebeck et al. 2010), suggesting a close relationship between tectonics, magmatism and volcanism (Wilson et al. 2009).
The origin of silicic magmatism through multiple stages of crustal assimilation and fractional crystallization processes (AFC) from a basaltic parent has been highlighted in many studies (e.g. Bachmann & Bergantz 2004; Annen et al. 2006). Evidence for AFC processes in the central TVZ have been presented (McCulloch et al. 1994; Deering et al. 2008, 2010); even if basalts form <1 vol% of erupted rocks, they must play an important role in the petrogenesis of the rhyolites and their plutonic equivalents could make up a significant volume of the lower to middle crust.
For this study, we focus on some major eruptive deposits dated between c. 320 and 0.7 ka that provide a good representation of rhyolitic volcanism in the central TVZ in time, space and composition. These eruptive deposits are sourced from the Okataina Volcanic Centre (OVC), the Taupo Volcanic Centre (TVC) and the Rotorua and Ohakuri Volcanic Centres (RoOhVC) as outlined in Figure 1. Essential information about the analysed deposits is summarized in Table 1, and detailed information on the individual eruptive centres can be found in Wilson et al. (1995, 2009) and Cole & Spinks (2009), among others.
Methodology
Sample preparation
We cleaned, oven-dried (50 °C) and crushed 64 individual pumices from ten different fall deposits (for Matahina, Ohakuri, Oruanui, Taupo and Kaharoa eruptive deposits) and ignimbrites (Matahina, Ohakuri, Mamaku, Rotoiti and Oruanui eruptive deposits) from the central TVZ, as listed in Table 1. Quartz and plagioclase crystals were handpicked and melt inclusions were identified using immersion oil (refractive index 1.54). We chose only fully enclosed and glassy melt inclusions and most crystals had multiple inclusions. To avoid boundary layer effects, we selected only >50 µm melt inclusions for analysis (Roedder 1984). For electron microprobe microanalysis (EMPA) and secondary ion mass spectrometry (SIMS), we prepared single-polished one-inch epoxy mounts, where multiple crystals (with similar size) were mounted and specific melt inclusions present at the same level were targeted for exposure. Fourier transform infrared (FTIR) spectroscopy was performed on double-polished crystal wafers, which were different samples to those used for EMPA and SIMS analysis. Notes on melt inclusion appearance and on the potential presence of bubbles were taken before exposure of the targeted inclusions, and a clear distinction in the results section is made between melt inclusions hosting a bubble (open symbols) and those that were not (closed symbols). Melt inclusions for most of the Taupo and all of the Kaingaroa samples were plagioclase-hosted, although the Kaingaroa melt inclusions were too small for trace element analysis and only volatile and major elements are presented here. The Whakamaru ignimbrite did not have any glassy melt inclusions and the results were therefore not included in this study.
Electron microprobe microanalysis (EMPA)
Major element compositions of 235 quartz-hosted and 15 plagioclase-hosted melt inclusions (some within the same crystal) were determined by electron microprobe (EPMA) at the University of Washington (UW, JEOL 733 Superprobe) and Oregon State University (OSU, Cameca SX100). We applied standard analytical conditions with an acceleration voltage of 15 kV, a beam current of 5 nA and a defocused beam of 10 µm in diameter, and measured Na first in the routine to minimize Na migration although it can still occur at these conditions (Morgan & London 1996). All displayed results are normalized to anhydrous conditions.
Volatile compositions (Cl, F, and S) were also analysed with the same EPMA, using two distinct techniques. With the JEOL 733 Superprobe at UW, volatiles were analysed using a beam blanking technique described in Witter & Kuehner (2004) with an acceleration voltage of 10 kV, a beam current of c. 100 nA and a defocused beam of 10 µm in diameter. With the Cameca SX100 at OSU, volatiles were measured simultaneously with the major elements. We used some of the unknown samples as an internal standard between the two different electron microprobes, and no discrepancies in volatile content were observed. The two-sigma (2σ) analytical error was very similar between both electron microprobes, with <3% for Cl and c. 10% for F. Using the beam blanking technique on the JEOL 733 Superprobe, the detection limit for Cl and F was lower with an average of 30 ppm for Cl and 70 for F. Detection limits with the Cameca SX100 was on average 80 ppm for Cl and 110 for F. Sulphur contents did not exceed the detection limit, which was c. 35 ppm using the beam blanking technique on the JEOL 733 Superprobe and c. 95 ppm with the Cameca SX100.
Secondary ion mass spectrometer (SIMS)
We used the secondary ion mass spectrometer IMS Cameca 6f at Arizona State University (ASU) to analyse trace elements (Li, B, Rb, Sr, Y, Cs, Ba, La, Ce, Pr, Nd, Sm, Th and U) in 142 melt inclusions on gold-coated epoxy mounts. The primary O-beam intensity was set at 10 nA and we used the composition of the standard NIST 610, measured several times during each session, to convert the measured trace element relative to Si ratios in concentrations (ppm). The 2σ analytical error was <5% for most of the trace elements, except for Cs and Sm for which the 2σ error was 10%, and 15% for U.
Fourier transform infrared (FTIR) spectroscopy
Transmission FTIR
Melt inclusions from 22 quartz crystals from 12 pumices from the Mamaku and Ohakuri ignimbrites and Ohakuri fall deposit (RoOhVC) were doubly intersected and measured by transmission FTIR at the University of Oregon. None of the analysed melt inclusions had a vapour bubble.
In most samples, total water concentrations were calculated using the near-IR peaks at 5200 and 4500 cm−1 using calibrations and parameters after Zhang et al. (1997). For some samples we used the total OH peak at 3500 cm−1 to calculate the water concentration, using Beer's Law (c=MA/ρdε where M is the molecular weight, A is the measured absorbance, ρ is the glass density, d is the thickness of the melt inclusion and ε is the molar absorption coefficient; Stolper 1982). CO2 concentrations were calculated using the 2350 cm−1 peak and Beer's law using a molar absorption coefficient of 1214 Lmol−1 cm−1 (Behrens et al. 2004). Melt inclusion thickness was measured by interference fringes after Wysoczanski & Tani (2006). Uncertainties in measuring inclusion thickness and fitting the background are typically <0.3 wt% for H2O and <5% of the amount present for CO2.
Reflectance FTIR
We measured water content in 40 melt inclusions of the Ohakuri and Mamaku eruptive deposits (RoOhVC) by FTIR reflectance spectroscopy at the University of Oregon using the reflectance technique of Hervig et al. (2003). These data were acquired on the same mounts used for EMPA and SIMS analyses. We calculated the amplitude of the c. 3650 cm−1 peak (caused by O–H resonance) by subtracting the measured reflectance with the reflectance at 3200 cm−1 (corresponding to the background), and normalized this value to a reflectance taken at 4000 cm−1 to account for instrumental variability (Hervig et al. 2003). To calculate the water contents we used the calibration of Johnson et al. (2011), which is based on the same rhyolitic glass standards used by Hervig et al. (2003). Only melt inclusions with a diameter >100 µm were analysed to ensure a good signal to noise ratio; with these analytical conditions, the estimated error was ±0.5 wt%. We found good correlation for H2O values between transmission and reflectance FTIR.
Results
The average major element composition of melt inclusions from the analysed eruptive deposits is presented in Table 2 and plotted in Figure 2. Quartz-hosted melt inclusions have a SiO2 content of 75–80 wt%, and plot on the high SiO2 end of the bulk-rock composition of the TVZ rhyolites (after Karhunen 1993; Beresford 1997; Brown et al. 1998; Smith et al. 2002, 2004; Milner et al. 2003; Gravley 2004; Nairn et al. 2004; Schmitz & Smith 2004; Shane et al. 2007; Deering et al. 2008; Fig. 2). The plagioclase-hosted melt inclusions (Kaingaroa and most of the Taupo analyses) include some slightly less-evolved melt relative to those that are quartz-hosted, with SiO2 contents ranging from 73 to 77 wt% (Fig. 2). In general, the major element composition in the melt inclusions of the analysed TVZ rhyolites is fairly homogeneous.
(a–d) Major element composition of analysed rhyolitic melt inclusions from the central TVZ. Open symbols correspond to melt inclusions containing a vapour bubble. Analytical error is smaller than the size of the symbol. Grey field corresponds to bulk-rock composition after Karhunen (1993); Beresford (1997); Brown et al. (1998); Milner et al. (2003); Gravley (2004); Nairn et al. (2004); Schmitz & Smith (2004); Smith et al. (2004); Shane et al. (2007); Deering et al. (2008).
Average major element and volatile composition of melt inclusions for the analysed eruptive units
Rotorua and Ohakuri Volcanic Centres (RoOhVC)
The Rotorua and Ohakuri Volcanic Centres (RoOhVC) produced R2 rhyolites (Deering et al. 2008). Prior to this study there were no volatile data available for these melts, so we focused on analysing melt inclusions from these magmatic systems. The Mamaku and Ohakuri eruptions both produced slightly heterogeneous magma types in their bulk composition (Milner et al. 2003; Gravley et al. 2007); it is therefore useful to identify the volatile concentration with respect to their different magma types. Major and trace element data for these ignimbrites are discussed in Bégué et al. (2014); here we focus on the volatile distribution of the different magma types. The same terminology for the different magma types is used as in Milner et al. (2003) and Gravley (2004) based on the bulk magma composition, with a more evolved rhyolitic Type 1 (74–78 wt% SiO2) and Type 2 (70–75 wt% SiO2) for the Ohakuri and Mamaku ignimbrites, a rhyodacitic Type 3 65–72 wt% SiO2) for the Ohakuri ignimbrite and a rhyolitic fall deposit from the Ohakuri eruption.
Average volatile compositions are presented in Table 2. The Ohakuri ignimbrite melt inclusions have between 3.5 and 5 wt% H2O but CO2 is almost undetectable, with a measured maximum of 7 ppm (Fig. 3a). However, the data from the Ohakuri fall deposit encompass a larger range of H2O (3.5–6 wt%) and CO2 (20–90 ppm) concentrations along a positive linear trend (Fig. 3a). The Mamaku melt inclusions have a similar range in H2O content as Ohakuri, but all measured Mamaku melt inclusions have some CO2 contents ranging between 20 and 70 ppm (Fig. 3a). The data from the two Mamaku magma types plot separately, with the Mamaku Type 1 having a slight higher H2O–CO2 concentration. Cl contents are high in the Mamaku and Ohakuri magmas and range from 0.25 to 0.35 wt% for all types. No trends are recognizable between the H2O and Cl content for the individual magma types (Fig. 3b); however, a weak positive trend exists between Cl and Y (ranging from 24 to 33 ppm) for some of the Mamaku and Ohakuri magma types (Fig. 4).
Volatile variations for the different magma types from the Rotorua and Ohakuri volcanic centres (RoOhVC). (a) CO2 (ppm) and (b) Cl (wt%) as a function of H2O (wt%); vapour saturation isobars for 50, 100, 150 and 200 MPa at 760 °C in (a) were calculated using VolatileCalc (Newman & Lowenstern 2002); isobars of 50 and 200 MPa in (b) are from Wallace (2005), based on experimental data from Webster et al. (1999). None of the selected melt inclusions for analysis from these units have a vapour bubble. Type 1 and 2 for the Mamaku and Ohakuri ignimbrite correspond to different magma batches (Bégué et al. 2014); the terminology for the different magma types is as Milner et al. (2003) and Gravley (2004), based on the bulk magma composition.
Cl (wt%) as a function of Y (ppm) for the different magma types from the Rotorua and Ohakuri volcanic centres (RoOhVC). None of the selected melt inclusions for analysis from these units have a vapour bubble. Type 1 and 2 for the Mamaku and Ohakuri ignimbrite correspond to different magma batches (Bégué et al. 2014); terminology for the different magma types is as in Milner et al. (2003) and Gravley (2004), based on the bulk magma composition. For most of the magma types the data are scattered (grey field), but there is a positive trend between Cl and Y for the Ohakuri Type 1 and 3 and Mamaku Type 1.
Comparison of volatiles and trace elements among the TVZ volcanic centres
Average volatile contents of the analysed melt inclusions are also presented in Table 2. Chlorine contents span a large range (Figs 5 & 6) and form two distinct groups: (1) the Rotorua–Ohakuri Volcanic Centres (RoOhVC), which have the highest Cl contents ranging from 0.25 to 0.35 wt%; and (2) the Okataina Volcanic Centre (OVC) and the Taupo Volcanic Centre (TVC), which cluster within a much smaller range of 0.15–0.22 wt%. The Oruanui and Taupo eruptive deposits, both from the TVC, have a few outliers with Cl contents up to 0.28 wt%. Fluorine contents also span a large range from almost 0 to 0.08 wt% (Figs 5b & 6a). Melt inclusions from the RoOhVC generally have higher F contents with a relatively small range (0.04–0.08 wt%) (Figs 5b & 6a). The water content in the RoOhVC melt inclusions ranges from 3.2 to 5.5 wt% (average of 4.2 wt%). This range of water contents is similar to the OVC and TVC values, ranging from 3 to 6 wt% (average of 4.8 wt% for the OVC and 4.4 wt% for the TVC; Fig. 6b, c; data from Dunbar et al. 1989; Liu et al. 2006; Johnson et al. 2011). TVZ rhyolites in general have a low CO2 content (Fig. 6c); the RoOhVC melt inclusions have between 0 and 90 ppm CO2 (average c. 35 ppm), which is similar to the OVC data ranging between 0 and 150 ppm (average c. 20 ppm, Johnson et al. 2011). The Oruanui magma from the TVC has CO2 contents ranging between 0 and 200 ppm (average c. 100 ppm) and a few outliers have concentrations of c. 300 ppm (Fig. 6c; Liu et al. 2006).
(a) Cl (wt%) as a function of SiO2 (wt%); (b) Cl (wt%) and (c) F (wt%) as a function of Rb/Sr ratio in analysed rhyolitic melt inclusions from the central TVZ. Open symbols correspond to melt inclusions containing a vapour bubble. Analytical errors for Cl and Rb/Sr ratio are smaller than the size of the symbol.
(a) F (wt%) and (b) H2O (wt%) as a function of Cl (wt%); and (c) CO2 (ppm) as a function of H2O (wt%) in analysed rhyolitic melt inclusions from the central TVZ. Open symbols correspond to melt inclusions containing a vapour bubble. Grey fields in (b) and (c) represent melt inclusion data for the OVC and the Taupo eruption (after Dunbar et al. 1989; Johnson et al. 2011). Isobars of 50 and 200 MPa in (b) are from Wallace (2005), based on experimental data from Webster et al. (1999); vapour saturation isobars for 50, 100, 150 and 200 MPa at 760 °C in (c) were calculated using VolatileCalc (Newman & Lowenstern 2002). Analytical errors for Cl are smaller than the size of the symbol.
The average trace element data from the melt inclusions is presented in Table 3. In general, similar to the volatile data, the trace element compositions distinguish the RoOhVC from the OVC and TVC. The Rb/Sr ratio ranges between 3 and 8 for the RoOhVC and is higher than the OVC and TVC melt inclusions (0.5–3; Fig. 5b, c). Figure 7 shows selected ratios of trace elements arranged chronologically (Fig. 7a–d) and spatially (i.e. depth of the Wadati Benioff Zone; Fig. 7e–h). The RoOhVC is characterized by lower Th/Y and Ba/Y ratios and higher Li/Cs and B/Cl ratios compared to the OVC. The TVC data lie in between the RoOhVC and the OVC and the Taupo eruption has, like the RoOhVC, low Th/Y and Ba/Y ratios. In general, Ba content ranges from 400 to 850 ppm and La content from 7 to 28 ppm. Ba and La are positively correlated for individual volcanic centres, with a linear increase of Ba with increasing La (Fig. 8). We observe two distinct trends, almost parallel to each other, with the OVC having the highest Ba/La ratio and the RoOhVC the lowest. The TVC data lie in between the OVC and RoOhVC, but Taupo products again do not fit with the trend of the Oruanui eruptive deposits (Fig. 8).
Variation of (a) Th/Y and (b) Ba/Y in quartz-hosted melt inclusions from the central TVZ v. time (data in chronological order, not to scale) starting with the 320 ka Matahina eruption (OVC) and ending with the 0.7 ka Kaharoa eruption (OVC). Variation of (c) Th/Y and (d) Ba/Y v. depth of the Wadati Benioff Zone (after Reyners et al. 2006); open symbols correspond to melt inclusions containing a vapour bubble. The grey band represents the systems situated on the western border of the central TVZ (i.e. RoOhVC) compared to the other systems situated on the rift axis (i.e. OVC and TVC). Analytical errors of these elements and ratios are smaller than the size of the symbol.
Ba–La variation in analysed rhyolitic melt inclusions. Open symbols correspond to melt inclusions containing a vapour bubble. Black fields represent bulk data on basalts from the central TVZ, erupted in proximity of the OVC or the TVC and Ohakuri volcano (Rooney & Deering 2014). Analytical error is smaller than the size of the symbol.
Average trace element composition of melt inclusions for the analysed eruptive units.
Discussion
Disparities in volatile exsolution among the TVZ rhyolites
The TVZ rhyolites form a continuum in compositions between R1 and R2 rhyolites. The major differences between these end-member rhyolite types are their mineral assemblage, with hydrous mineral phases (hornblende±cummingtonite±biotite) only present in R1 rhyolites, their crystal content and bulk-rock trace-element composition. Deering et al. (2008, 2010) have interpreted these differences between R1 and R2 to be related to changes in fH2O, fO2 and P–T conditions in the lower to middle crust (15–30 km), where the evolved magmas are generated through fractionation from repeated basalt intrusions. This led the authors of these studies to refer to the R1 and R2 types as ‘cold-wet-oxidizing’ and ‘hot-dry-reducing’ rhyolites, respectively (Deering et al. 2010). Eruption temperatures and reservoir pressures of the rhyolites between R1 and R2 are however very similar (within estimated errors of the used thermometers and geobarometers). Temperatures mostly range between 750 and 800 °C (e.g. Schmitz & Smith 2004; Wilson et al. 2006; Shane et al. 2007; Allan et al. 2012; Bégué et al. 2014) and R1 and R2 rhyolites are stored at similarly shallow depths (c. 100 MPa; e.g. Liu et al. 2006; Johnson et al. 2011; Allan et al. 2012; Bégué et al. 2014). These pressures have been obtained with H2O–CO2 solubility models and with other geobarometers independent of volatile saturation (i.e. amphibole geobarometry, e.g. Ridolfi et al. 2010; rhyolite-MELTS geobarometer, Gualda & Ghiorso 2014).
The volatile and trace element data presented here separate the two rhyolite types. We observe large disparities in chlorine content (difference of c. 1000 ppm; Figs 5 & 6) and smaller differences in F contents (Figs 5b & 6a). Carbon dioxide concentrations are fairly similar (Fig. 6c) however, and water concentrations are on average slightly higher in the R1 magmas (Fig. 6b). These results show that there is a good correlation between the volatile concentrations of R1 and R2 rhyolites and the ‘wet-oxidizing’ and ‘dry-reducing’ types as defined by Deering et al. (2008, 2010). However, it is counter-intuitive that the ‘wet’ R1 magmas have a lower volatile concentration (i.e. Cl and F) compared to the ‘dry’ R2 magmas. These observations therefore lead to several questions. What is the origin of the volatile (especially chlorine) disparities? Why do water and CO2 not show the same disparities between the R1 and R2 melt inclusions analysed here as Cl concentrations? Why do R2 magmas have a higher Cl concentration?
When comparing different rhyolites, one has to consider that the measured volatile concentration in the melt inclusions does not necessarily represent the bulk volatile content of the magma (i.e. volatiles dissolved in the melt, hydrous crystals and exsolved volatile phase, if present), which is inherently what will dictate early magma evolution prior to a gas phase forming. It is therefore important to consider the behaviour of these volatiles in these melts and, more importantly, the ‘saturation state’ of these melts during crystallization (i.e. the exsolution of a volatile phase prior to, or during, crystallization in the silicic magma chamber).
Volatile saturation during crystallization
Many studies have established that silicic magmas often become saturated in a volatile phase during shallow ascent and eruption and during the final stages of crystallization. However, a volatile phase might also form during the earlier stages of crystallization within crystal-poor melt bodies prior to eruption (e.g. Anderson et al. 1989; Lowenstern et al. 1991; Lowenstern 1993, 1994a; Wallace & Gerlach 1994; Wallace et al. 1995). Once saturation is reached, volatile species dissolved in the melt will partition to a volatile phase proportional to their relative solubilities and concentrations.
Exsolution of a fluid phase(s) occurs because the solubilities of the volatile components in the melt are finite; these are strongly dependent on composition, density and number of fluid phases present, which vary with temperature, pressure and melt composition in the system. Chlorine solubility is largely dependent on P–T–X conditions, and an added complexity is that Cl saturated melts can exsolve two separate phases below c. 200 MPa: a low-density vapour phase and a high-density hydrosaline fluid (e.g. Webster 2004; Carroll 2005). Experimental studies on Cl behaviour in silicic magmas show that Cl solubility decreases with decreasing temperature, but it increases with decreasing pressure and there is a strong dependence of the solubility with the composition on the melt and coexisting volatile phase (if present) (e.g. Webster & Holloway 1988; Carroll & Webster 1994; Webster 1997; Signorelli & Carroll 2000; Carroll 2005). The Cl solubility will therefore be affected by the presence and amount of other volatiles, and water concentration in the melt will have an inversely proportional relationship with the Cl solubility (Webster 1997). To date no experimental data exist on how the CO2 content will affect Cl solubility for rhyolitic melts. From experiments on andesites, it is suggested that added CO2 has only a minor effect on the dissolved Cl concentration in the melt; however, CO2 will increase the likelihood for phase separation to form a separate vapour phase and hydrosaline fluid (Botcharnikov et al. 2007). Carbon dioxide generally has a very different behaviour compared to Cl (e.g. Lowenstern 2001) and has a very low solubility in rhyolitic melts (generally an order of magnitude lower than H2O). In silicic magmas, Cl will partition in favour of co-existing volatile phases (either as Cl2 gas or a Cl− ion). Experimentally determined fluid-melt distribution coefficients (DClfl/melt) range from 2 to 80 for evolved silicic compositions (Webster & Holloway 1988; Métrich & Rutherford 1992; Webster 1992; Student & Bodnar 1999). In contrast to Cl, fluorine is highly soluble in silicate melts and will typically favour the melt phase relative to an exsolved fluid. As a result, F concentrations will increase with fractionation as there is no phenocryst phase with a high and fixed F concentration. In general, its content has to exceed c. 8 wt% (on average) for the partition coefficient (DFfl/melt) to be >1 (Carroll & Webster 1994). For the TVZ rhyolites, fluorine is more likely to partition into a volatile phase at a later stage in the magmatic history of a system (i.e. during cooling and crystallization of an intrusion at depth), and this behaviour of F is considered to be independent of H2O and Cl.
Different approaches have been used to assess the saturation state of rhyolitic melts based on the composition of mineral-hosted melt inclusions, including: (1) H2O–CO2 solubility models to establish the saturation pressure (e.g. Newman & Lowenstern 2002; Papale et al. 2006); (2) the analysis of volatile variations in the melt; and (3) a determination of how they evolve relative to other incompatible elements (e.g. Webster 1997). Previous studies in the central TVZ assessed the volatile behaviour and saturation state for many of the youngest Okataina Volcanic Centre magmas (Johnson et al. 2011) and the Oruanui magma (Taupo Volcanic Centre; Liu et al. 2006). From the application of H2O–CO2 solubility models for the OVC and Oruanui melts (TVC), Johnson et al. (2011) and Liu et al. (2006) deduced that these rhyolites crystallized in the presence of an exsolved volatile phase. Furthermore, Johnson et al. (2011) interpreted the Cl concentration from the OVC magmas, which plateaus at 0.2 wt% (Figs 5 & 6), to be associated with Cl partitioning into an exsolved phase. Hydrous mineral fractionation (e.g. hornblende and biotite, which are present mostly in R1 rhyolites) and trace amounts of apatite are suggested to have a negligible effect on the H2O and Cl concentrations, as these minerals only compose a small proportion of the total crystal population (e.g. Schmitz & Smith 2004; Shane et al. 2007; Johnson et al. 2011). The measured volatile content in melt inclusions of these R1 rhyolites therefore represents melts that have experienced some volatile loss. To date there is no literature on the saturation state of R2 rhyolites (i.e. eruptives from the Rotorua–Ohakuri Volcanic Centres, RoOhVC); a detailed assessment of volatile behaviour for these melts will therefore be addressed here.
Application of H2O–CO2 solubility models on the RoOhVC magmas (R2 rhyolites) result in estimated saturation pressures between 75 and 150 MPa (Fig. 3a; using 760 °C as an average temperature of the TVZ magmas; Newman & Lowenstern 2002). These pressures are consistent with the estimated pressures of c. 100 MPa, determined independently of the volatile data (using rhyolite-MELTS; Bégué et al. 2014). On average, the RoOhVC rhyolites have a slightly lower water content compared to R1 magmas (Fig. 6b) which results in slightly lower saturation pressures. From these solubility models, which are calibrated for melt in equilibrium with H2O–CO2-bearing fluids only, the results suggest that the RoOhVC melts may well be saturated with regard to a volatile phase at the time of melt inclusion entrapment. These results are nonetheless inconsistent with experimental studies for silicic melts in equilibrium with H2O–Cl-bearing fluids (Webster et al. 1999). Based on the results of these solubility experiments, the c. 0.3 wt% Cl in the RoOhVC melts would require them to be over-saturated with respect to the 200 MPa isobaric line (Fig. 3b). These results show that the solubilities and saturation with regard to multicomponent volatile phases are complex and may be difficult to predict without some other independent assessment of saturation (i.e. presence of fluid).
Another approach to assess if a volatile phase exsolved from the melt is to compare volatile variations with other incompatible elements. Cl, H2O and CO2 are interdependent and incompatible during volatile under-saturated crystallization (if no hydrous phases are present), and their concentration will be strongly affected during crystallization in the presence of a volatile phase. If the melt is volatile under-saturated, there should be a positive linear correlation of increasing Cl, H2O and CO2 with incompatible elements (Webster 1997). For the RoOhVC (i.e. Mamaku and Ohakuri eruptions), the volatile variation for each individual magma type is treated separately as they are suggested to represent distinct magma batches (Bégué et al. 2014). Chlorine concentrations increase slightly with the Y contents for Mamaku Type 1 and Ohakuri Type 1 and 3 magmas (Fig. 4), suggesting that Cl was possibly not saturated and/or not lost to a coexisting volatile phase during crystallization. For the H2O v. Cl content, the data are very scattered and no clear trends can be identified (Fig. 3b). For the H2O v. CO2 only the Ohakuri fall deposit plot along a positive trend (Fig. 3a), which may be indicative of crystallization in a volatile under-saturated melt. However, for the Mamaku ignimbrite, CO2 contents cluster around c. 25–75 ppm and the melt inclusions of the Ohakuri ignimbrite have no CO2 at all (Fig. 3a). The lack of clear trends in the volatile record may be explained by post-entrapment CO2 diffusion and, to a lesser extent, H diffusion through the quartz host or formation of a vapour bubble, which have been suggested to affect melt inclusion volatile concentrations (e.g. Bodnar 2003; Severs et al. 2007; Steele-MacInnis et al. 2011; Bucholz et al. 2013). Furthermore, these magmas are crystal-poor (Table 1; magma Type 1 has <1 vol% crystals); the record of changes in crystallization with Y or other trace elements is therefore limited compared to systems where changes in volatiles are tracked along a larger chemical spectrum involving more crystallization. This could also be the reason why no clear positive, linear trends are observed between F and the Rb/Sr ratio (Fig. 5b). In addition, plagioclase and pyroxenes represent only a fraction of the crystal cargo, which further reduces the potential affect on the Y content. In comparison, the R1 magmas of the TVZ are crystal-moderate to -rich and variations of volatile contents with incompatible elements are more easily observed.
From the assessment of saturation state above, some of the Mamaku and Ohakuri magmas from the RoOhVC appear to be under-saturated while others remain inconclusive. Importantly, if we consider Cl variations with regard to other incompatible elements, then the RoOhVC melt inclusions unlikely record Cl-loss to a coexisting phase; some care is therefore required when solely using the H2O–CO2 solubility models, which here would indicate saturation. We suggest that our results for R2 melt inclusions may reflect that these melts were either at, or very close to, saturation in a volatile phase during melt inclusion entrapment. However, regardless of volatile saturation, we can conclude that the R1 magmas experienced a higher amount of degassing prior to and/or during crystallization.
Differences in bulk volatiles
Disparities in volatile saturation between R1 and R2 rhyolites is suggested here, with the OVC and Oruanui magmas (TVC) being saturated prior to quartz crystallization (Liu et al. 2006; Johnson et al. 2011) compared to the RoOhVC melts where saturation, if reached, would have occurred during or after crystallization. Measured volatile concentrations in melt inclusions therefore represent variously degassed melts.
R1 magmas show clear volatile exsolution, which initiated the partitioning of Cl in the co-existing phase, as DClfl/melt is ≫1 (Webster & Holloway 1988; Métrich & Rutherford 1992; Webster 1992; Student & Bodnar 1999), and ‘stabilized’ the Cl content at its solubility value of c. 0.2 wt% (Figs 5 & 6a). Bulk Cl contents in the OVC therefore had to be much higher than the concentration measured in melt inclusions. Equally, most of the H2O partitioned into the coexisting volatile phase (Fig. 6b, c), requiring a higher initial bulk water content. Chlorine solubility is negatively correlated with the water content and will be higher in rhyolites with lower H2O. This may explain the high Cl content in the RoOhVC magmas (Figs 5 & 6), and suggest that R2 magmas have a lower bulk water content.
The exsolution of a volatile phase is highly dependent on the initial water content, among other parameters. Here, similar H2O concentrations in the OVC and RoOhVC magmas (c. 3–6 wt%) can be explained by the similar water solubilities at storage conditions of these melts (similar pressures and temperatures), independent of the initial water content. The variability in initial water content is also in agreement with the suggested petrogenesis of R1 and R2 rhyolites and the ubiquitous presence of hornblende (±biotite) in the R1 type (absent in most or only sparse in R2, Table 1). Variations in temperature, fH2O and fO2 in the lower crust are shown to dictate the resulting mineral assemblage of the rhyolites (Deering et al. 2008), which are generated largely through fractional crystallization (e.g. McCulloch et al. 1994; Deering et al. 2008, 2010).
Bulk Cl concentrations, along with water and other volatiles, are also suggested to be much higher in the R1 compared to the R2 rhyolites. Some rough estimations of the differences of bulk Cl can be made using a DClfl/melt of 10 (for pressures of c. 100 MPa, e.g. Signorelli & Carroll 2000) and by considering the coexistence of c. 20 vol% bubbles with the melt. With these assumptions, the bulk Cl content in R1 rhyolites may be at least twice as high (c. 0.6 wt%) compared to the R2 magmas, which will have non-negligible effects on the amount of mass and heat transfer from magmatic systems.
Distribution of volatiles in the central TVZ
From the melt inclusion data, we suggest that the rhyolites form two groups based on their inferred bulk volatile content: R1 rhyolites have higher bulk volatile concentrations corresponding to the ‘wet-oxidizing’, compared to the ‘dry-reducing’ R2 magmas (after Deering et al. 2008, 2010). The distribution of the melt inclusion data therefore mirrors the variability in bulk-rock geochemistry and mineral assemblage indicated by the different rhyolite types. Deering et al. (2010) suggest that the last 550 ka in TVZ magmatism can be subdivided into two broad cycles each with R1 followed by R2 magmatism. They suggest that the more buoyant R1 magmas, hence volatile-rich, are focused along the central rift axis in comparison to the volatile-poorer R2 magmas. The variability of these rhyolite types has previously been suggested to be associated with subduction zone processes and/or crustal level modulation induced by caldera collapse (Deering et al. 2008, 2010; Rooney & Deering 2014). Our volatiles and trace element data from the melt inclusions therefore appear to be consistent with differences originating from these processes as well.
Influence of slab-derived fluids?
The origin of the major volatiles in primary arc magmas is mainly the mantle wedge, the subducted altered oceanic crust and its overlying sediments (e.g. Wallace 2005; Zelenski & Taran 2011). It has long been established that these components of subduction zones play an important role in the generation of aqueous fluids and the transport of fluid-mobile elements producing, ultimately, the typical arc signature in subduction zone volcanism (e.g. Kay 1980; Hawkesworth & Ellam 1989; McCulloch & Gamble 1991).
In the TVZ, crystal fractionation and assimilation models (AFC) have been used to explain the rhyolite petrogenesis from a parental mafic magma (e.g. McCulloch et al. 1994; Graham et al. 1995; Price et al. 2005; Deering et al. 2011), and a large presence of mantle component in the silicic magmas has been suggested with only minor assimilation (<25%) in the lower crustal region (e.g. McCulloch et al. 1994; Deering et al. 2008). The compositional and petrological variability present in the rhyolites has been suggested to reflect distinct zones of melting and magma generation in the mantle (Deering et al. 2008, 2011), and a recent study that utilized basalt bulk-rock geochemistry has provided a potential link between the ‘wet’ R1 and ‘dry’ R2 rhyolite types and the variability in the mafic parent (Rooney & Deering 2014). The observed range of composition of these basalts suggests different sources for the primitive melt generation, with disparities in slab-derived components from the subduction zone (Rooney & Deering 2014). A stronger petrogenetic connection between the basalts and the rhyolites has been established for the central TVZ (Rooney & Deering 2014), which gives more confidence in using the rhyolite compositions to assess the slab-derived flux. Compared to the primitive basalts, the rhyolites are also much more abundant in the central TVZ, permitting a better evaluation of the fluid flux variability through time and space.
Volatiles and fluid-mobile elements are transported along with aqueous fluids from the subducting slab, and participate in melt generation processes. Thorium and Ba are such elements, representing components of the subduction zone, and Y represents a mantle component. Ba/Y and Th/Y ratios are therefore independent of mantle source, partial melting or fractional crystallization and represent a measure of the subduction input (Pearce & Peate 1995; Pearce & Stern 2006). This input seems to be distinct between the RoOhVC and OVC-TVC magmas (Fig. 7a–d). With increasing distance from the trench (i.e. increasing depth of the Wadati Benioff Zone, after Reyners et al. 2006) there is decreasing input from slab-derived components, indicating that the contribution of slab-derived components were probably not as important for the RoOhVC as for the OVC. Similar patterns have been observed at numerous subduction zone systems on Earth (e.g. Izu-Bonin System, Hochstaedter et al. 2001; Tonga System, Jenner et al. 1987; Bach et al. 1998). These results indicate that trace element variations in the rhyolitic melt inclusions can be linked to the basalt bulk-rock data (Fig. 8), consistent with the findings of Rooney & Deering (2014), showing that the variability in the TVZ rhyolites is a result of distinct sources for melt generation in the mantle region and processes operational in the subduction zone. There is very little published volatile data for TVZ basalts; caution should therefore be applied in trying to establish a direct link in the volatile concentrations between the R1 and R2 rhyolites and the basalt distribution.
In most of the measurements of volatiles and trace elements from this study, the TVC magmas usually have compositions in between the OVC and the RoOhVC magmas (Figs 7 & 8). These results are in agreement with the observations of Deering et al. (2008, 2010), suggesting that the TVC magma composition is less ‘wet and oxidizing’ than the OVC magmas (representing the R1 end-member composition). In Figure 8, the Oruanui rhyolite (TVC) plots along a trend above the OVC magmas and has lower Ba/Y and Th/Y than the OVC, although the depth of the subducting slab beneath the volcanic centres is very similar (Fig. 7c, d). These results may be indicative of disparities in slab-derived fluid flux between the OVC and the TVC, as suggested by Reyners et al. (2006) and Eberhart-Phillips et al. (2008), and is also in agreement with the basalt study of Rooney & Deering (2014), suggesting a decrease in slab-derived contribution from north to south along the rift axis. Compositional differences between the Oruanui and the Taupo rhyolites from the TVC have been suggested to be unrelated to changes in fluid flux from the subduction zone (Deering et al. 2010).
Conclusion
We focused on the volatile and trace element composition of melt inclusions from seven major rhyolitic ignimbrites in order to identify the spatial distribution of the major volatile species at the scale of the entire central TVZ, and to establish the importance of volatile saturation in these silicic systems. H2O–CO2 solubility models are combined with the study of chlorine variations along with incompatible elements to assess the saturation state of some of these rhyolitic melts, and to allow for comparison among the distinct volcanic centres. Some of the major conclusions are:
Rhyolites produced in the central TVZ over the past c. 320 ka have distinct volatile, fluid-mobile and highly incompatible element compositions (e.g. Cl, F, Th, Ba), with the rhyolites from the Okataina Volcanic Centre (OVC) and the single Oruanui eruptive event and the Rotorua-Ohakuri Volcanic Centres (RoOhVC) and Taupo magmas forming two distinct types (R1 and R2 types, respectively).
These volatile concentrations from R1 and R2 rhyolites mirror the variability in bulk-rock geochemistry and petrology of the rhyolitic continuum between ‘wet-oxidizing’ and ‘dry-reducing’ types established by Deering et al. (2008, 2010). However, the ‘wet’ R1 rhyolitic melt inclusions record a lower volatile concentration compared to the ‘dry’ R2 melt inclusions. This emphasizes the importance of considering the saturation state of these melts at the time of melt inclusion entrapment.
The origin of the disparities in volatile concentration between R1 and R2 is suggested to be related to differences in the bulk volatile concentration of the parental magmas, leading to differences in volatile saturation between R1 and R2 systems: R1 rhyolites were clearly saturated during crystallization, whereas R2 rhyolic melt inclusions are suggested to record either less degassing than R1 or no degassing at all.
These results show that silicic systems with similar reservoir depths, eruption temperatures and major element compositions may exsolve a volatile phase at different stages in their magmatic history, and it can not always be assumed that all high-silica melts reached volatile saturation during crystallization. The differences in volatile content will not only have an effect on the relative timing of exsolution of a volatile phase in a magmatic system, but the composition of this exsolved magmatic volatile phase as well.
Acknowledgments
We would like to thank S. Kuehner, F. Tepley, R. Hervig and L. Williams for their help with analytical procedures at University of Washington, Oregon State University and Arizona State University; P. Wallace for his contribution and help with the FTIR at the University of Oregon and for his constructive and insightful review of an early version of the manuscript; and R. Spiers for his help with the sample preparation. This work was supported by the Mighty River Power Ltd. Source to Surface geothermal research programme with the University of Canterbury, a University of Canterbury doctoral scholarship and the Mason Trust Fund. Additional funding to Chambefort came from GNS Science CSA (Core Science Area) Geothermal Research Programme and the Royal Society of New Zealand Marsden Fund. We also acknowledge NSF (EAR0948878) for continued support of the ASU National SIMS Facility. We also thank J. Lowenstern, anonymous reviewers and editor S. Straub for their comments, which significantly improved the manuscript. Finally, we thank the corresponding editor G. Zellmer for the opportunity to contribute to this special publication.
- © 2015 The Geological Society of London
References
Magmatic volatile distribution as recorded by rhyolitic melt inclusions in the Taupo Volcanic Zone, New Zealand
