Halogen degassing during ascent and eruption of water-poor basaltic magma

Chemical Geology 263 (2009) 122–130

Contents lists available at ScienceDirect

Chemical Geology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h e m g e o

Halogen degassing during ascent and eruption of water-poor basaltic magma Marie Edmonds a,⁎, Terrence M. Gerlach b, Richard A. Herd c a b c

Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge, CB2 3EQ, United Kingdom Emeritus, United States Geological Survey, Cascades Volcano Observatory, 1300 Cardinal Court #1, Vancouver, WA 98683-9589, USA School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, United Kingdom

a r t i c l e

i n f o

Article history: Accepted 24 September 2008 Keywords: Halogens Basalt Degassing OP FTIR Kilauea

a b s t r a c t A study of volcanic gas composition and matrix glass volatile concentrations has allowed a model for halogen degassing to be formulated for Kīlauea Volcano, Hawai i. Volcanic gases emitted during 2004–2005 were characterised by a molar SO2/HCl of 10–64, with a mean of 33; and a molar HF/HCl of 0–5, with a mean of 1.0 (from approximately 2500 measurements). The HF/HCl ratio was more variable than the SO2/HCl ratio, and the two correlate weakly. Variations in ratio took place over rapid timescales (seconds). Matrix glasses of Pele's tears erupted in 2006 have a mean S, Cl and F content of 67, 85 and 173 ppm respectively, but are associated with a large range in S/F. A model is developed that describes the open system degassing of halogens from parental magmas, using the glass data from this study, previously published results and parameterisation of sulphur degassing from previous work. The results illustrate that halogen degassing takes place at pressures of b 1 MPa, equivalent to b ~ 35 m in the conduit. Fluid–melt partition coefficients for Cl and F are low (b 1.5); F only degasses appreciably at b 0.1 MPa above atmospheric pressure, virtually at the top of the magma column. This model reproduces the volcanic gas data and other observations of volcanic activity well and is consistent with other studies of halogen degassing from basaltic magmas. The model suggests that variation in volcanic gas halogen ratios is caused by exsolution and gas–melt separation at low pressures in the conduit. There is no evidence that either diffusive fractionation or near-vent chemical reactions involving halogens is important in the system, although these processes cannot be ruled out. The fluxes of HCl and HF from Kīlauea during 2004–5 were ~ 25 and 12 t/d respectively. © 2008 Elsevier B.V. All rights reserved.

1. Introduction At Kīlauea and most other volcanoes, halogens are much less abundant than the three primary volatile species H2O, CO2 and S in magmatic vapour and contribute little to magma buoyancy and eruptive processes. Halogens do, however, play an important role in magmatic systems. Chlorine abundance in magma is controlled by fractional crystallisation, H2O degassing, melt-vapour partitioning and vapour separation and loss, making it a useful tracer for magmatic processes (e.g. Sun et al., 2007) and volcanic degassing (e.g. Villemant et al., 2008). When Cl degasses into the atmosphere as a relatively minor constituent of volcanic gases, it has disproportionate atmospheric effects: Cl radicals act to destroy ozone (e.g. Rose et al., 2006). Hydrogen fluoride is much less reactive, but is a strong acid when dissociated in water: it has been linked to fluorosis (Cronin and Sharp, 2002) and will etch the surface of silicate ash particles (Delmelle et al., 2007). At Kīlauea Volcano, F becomes highly enriched in residual melts, and becomes the most significant volatile in slowly-cooled shallow magma bodies and lava lakes (Helz, 1980).

⁎ Corresponding author. E-mail address: [email protected] (M. Edmonds). 0009-2541/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2008.09.022

Halogens typically occur in volcanic gases as hydrogen chloride (HCl) and hydrogen fluoride (HF), with trace amounts of the species hydrogen bromide (HBr) and hydrogen iodide (HI; Aiuppa et al., 2005). In the mantle, halogens are accommodated in the mineral structures of phlogopite, pargasite and apatite, where they occupy hydroxyl ion sites (e.g. Aoki et al., 1981). Like other volatiles, their distribution in the mantle is heterogeneous and related to tectonic setting. Chlorine, in particular, is enriched in subduction settings owing to contributions from subducted sediments and hydrous minerals in altered basalts (e.g. Kent et al., 2002). Major enrichments in halogen concentrations have been documented for ocean island settings such as the Azores and Iceland, relative to Mid-Atlantic Ridge basalts (Schilling et al., 1980; Thordarson and Self, 2003) but the degree of enrichment varies widely between ocean islands (Stroncik and Haase, 2004). Chlorine and fluorine rarely reach saturation in magmas; instead, they either behave as non-volatile, relatively incompatible elements (F and Cl are incorporated into apatite and amphiboles respectively at low pressures; e.g. Brenan, 1993) in water-poor melts, or they partition into a vapour phase (e.g. Webster, 2004). Very little is known about the partitioning behaviour of Cl between melt and vapour in basaltic melts; even less is known about the degassing behavior of F. Fluorine tends to behave as an incompatible, nonvolatile

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

123

element for a large part of magma evolution, and increases in concentration in the melt with decreasing extents of melting (Dixon et al., 1997) and/or increasing degrees of fractionation (e.g. Byers et al., 1984). A suite of Ko olau (Oahu) melt inclusions analysed by Hauri (2002) show a trend of decreasing F with H2O, suggesting that F is degassed at low pressure. Degassing of halogens has been modeled empirically at Mount Etna, based on field measurements (Aiuppa et al., 2002, 2004), which has confirmed that Cl and F partition into a vapour phase to a greater extent at low pressures. Halogens have only rarely been quantified in volcanic gases emitted from Kīlauea because they are relatively minor components of the gas phase and are thus difficult to measure. There are a few studies that quantify HF in condensates (Murata et al., 1963) and volcanic gases near to the beginning of the current eruption (Gerlach and Graeber, 1985; Greenland, 1988). A filter pack study in 1985 observed large variability in gaseous F/Cl, which was ascribed qualitatively to changes in magma supply and residence time beneath Pu u Ō ō (Miller et al., 1990). Open path Fourier transform infra red (OP FTIR) measurements of halogens in volcanic gases were carried out during 2004–2005 and the concentrations of Cl, F and S were measured in glasses in Pele's tears erupted from Pu u Ō ō in 2006; these data are used to formulate a quantitative model to describe halogen degassing at Kīlauea Volcano. 2. Geological setting Kīlauea Volcano is the youngest and most active subaerial volcano of the Hawaiian Volcanic Chain. Kīlauea makes up the south-eastern portion of the island of Hawai i, abutting the southern flank of Mauna Loa. Kīlauea has been active semi-continuously throughout historical times; the current eruption began in January 1983 on the East Rift Zone. Between 1983 and 1986 the eruption was characterised by regular intervals of lava fountaining activity, interspersed with repose periods averaging 24 days in length (Heliker and Mattox, 2003). Effusion rates reached 390 m3/s and fountains reached heights of N450 m (Heliker and Mattox, 2003). From 1986, the eruption entered a phase of continuous, effusive activity, characterised by eruption rates of around 0.13 km3/year (Heliker and Mattox, 2003). During 2004– 2005, the eruption was characterised by effusion of lava from vents on the south flank of Pu u Ō ō, whilst much of the degassing took place from vents inside the crater (Fig. 1). Lava spattering, low fountaining and gas pistoning activity took place from time to time at the crater and south flank vents during this period (Edmonds and Gerlach, 2007) and Pele's tears and hair were typical eruptive products. Basaltic magma is generated at depths of several tens of kilometres beneath the summit of Kīlauea; degassing is dominated by CO2 during magma ascent and prior to residence in a magma chamber (Dixon et al., 1991). Gerlach and Graeber (1985) proposed that magma degasses in two distinct stages, based on suites of gas samples collected and analysed in 1983 from the East Rift Zone, and from a lava lake at the summit in the early-20th century by T. A. Jaggar. The first stage of degassing occurs in the magma storage area beneath the summit, giving rise to a CO2-rich plume that escapes through the caldera floor; the second stage of degassing occurs as the magma migrates along the East Rift and nears the surface at Pu u Ō ō, where the more soluble gases H2O and SO2 exsolve. This model explains observations of a high CO2 emission rate and high molar C/S at the summit of Kīlauea (type 1 gas) and a high SO2 emission rate and a lower molar C/S at Pu u Ō ō (type 2 gas; Gerlach and Graeber, 1985). Degassing is thought to proceed in equilibrium during lava effusion, with continuous bubble nucleation and growth and minor bubble coalescence, based on vesicle size distributions (VSDs; Mangan et al., 1993; Cashman et al., 1994). Corresponding VSD studies of Pele's tears (erupted during lava spattering or weak strombolian activity) have not been published. Recently, distinct degassing regimes have been recognised, based on gas compositions measured by spectroscopic methods during differ-

Fig. 1. Photographs to show (top) the FTIR spectrometer and telescope, mounted on a tripod, receiving IR from an incandescent vent, with cooler volcanic gases passing in front of the source and absorbing radiation. The spectrometer is connected to a portable acquisition computer system; (bottom) the crater of Pu u Ō ō, showing the degassing vents measured in this study.

ent kinds of eruptive activity at Pu u Ō ō: persistent degassing (equilibrium degassing and limited melt-gas separation), lava spattering (bubble rise from a few tens of metres depth) and gas pistoning (gas slug ascent from N100–500 m depth; Edmonds and Gerlach, 2007). The former is the most common mode of degassing and was prevalent during the measurements presented here. 3. Methods 3.1. Volcanic gas measurements Open path Fourier transform infra red (OP FTIR) spectroscopy is a technique used to measure the composition of gases in open air. The method has been applied successfully to basaltic volcanic gases previously (Mori et al., 1993; Francis et al., 1995). OP FTIR relies on a spectrometer receiving infra red radiation (IR) from a source through a cloud of cooler gas. Under these conditions, IR is absorbed by the gases and concentration-pathlengths can be quantified precisely for various gas species simultaneously. In practice, a variety of IR sources can be utilized; for this study, incandescent vents were used (Fig. 1). In general, the signal-to-noise ratio (and hence the accuracy and precision of the measurements) is inversely proportional to the pathlength and directly proportional to the strength of the IR source and the concentration of gas in the pathlength. Pu u Ō ō provides high

124

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

background spectrum Bj, which is free of, or contains less of, the species i:   Aj = − log Sj =Bj :

Fig. 2. Calibration curves for SO2, HCl and HF, showing concentration-pathlength plotted against absorbance for each of the reference spectra used in the analysis.

concentrations of gases and strong, convenient IR sources. A MIDAC spectrometer with an electrically-cooled indium-antimonide (InSb) detector, which has a field of view of 0.5 m at 80 m, was used for the measurements. The aperture of the spectrometer was aligned with an incandescent vent. Volcanic gases passed in front of the source, absorbing radiation (Fig. 1). For each measurement, 4–8 spectra were averaged, yielding a sampling rate of 4–8 single beam spectra/minute, which was judged to yield the best temporal resolution without sacrificing signal quality. The spectra contain information relating to the wavelength and intensity of radiation between 1800 and 4000 cm− 1 with a resolution of 0.5 cm− 1. For analysis, AutoquantPro software (MIDAC Corp.) was used, based on the principles of Beer's Law:

Aj =

k X

aij LCi j = 1; 2:::n:

ð1Þ

i=1

The observed sample absorbance spectrum, Aj, for each of k species is obtained by normalising the measured spectrum Sj, by a

ð2Þ

In practice, every spectrum contained volcanic gas and hence, the background also contained a small amount. The pathlengthconcentration (ppm m) values are therefore minimum values. Calibration spectra for pure components SO2, HCl and HF at known concentrations are used to calculate the absorptivities αij. Reference spectra for each species were chosen to bracket the expected concentration and thereby provide a more accurate calibration curve from which to calculate concentration-pathlengths (Fig. 2). Fig. 3 shows reference and measured absorbance spectra for sulphur dioxide (SO2), hydrogen chloride (HCl) and hydrogen fluoride (HF) and the fitting windows used for the retrievals. 3.2. Glass geochemistry Samples are Pele's tears collected from Pu u Ō ō in April–May 2006. Tears are typically erupted during lava spattering, which might also be classified as weak strombolian activity, and are glassy droplets up to 1– 2 cm across. The tears were cut and mounted on 30 μm glass slides, polished and carbon-coated. Glass compositions in 8 sections were analysed in the Department of Earth Sciences in Cambridge on a CAMECA SX-100 microprobe. A 10 μm beam was used with a beam current of 60 nA and accelerating voltage of 15 kV for Cl, F, S, P, Cr and Ni, and 4 nA and 15 kV for the other major elements. Detection limits for S, Cl and F were 34–38, 37–40 and 82–87 ppm respectively. Around 170 glass analyses were generated, which were mostly matrix glasses, but with two olivine-hosted melt inclusions (Fig. 4). The matrix was browncoloured glass with sparse plagioclase microlites and scattered phenocrysts of olivine, clinopyroxene and plagioclase (b10 vol.%). Spherical, large bubbles made up ~30 vol.%. Glass immediately surrounding the

Fig. 3. Reference spectra and measured absorbance spectra, showing fitting windows.

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

125

steepening at high concentrations (Fig. 5). Compounds that absorb strongly in the infrared require numerous calibration points because deviation from Beer's law behavior occurs at high concentrations, where the charge distribution from nearby molecules interferes with the absorption of energy. A piece-wise linear regression routine is used here to compensate in part for this non-linear behaviour, although this approach may be subject to error if points lie outside the region of the calibration data. The general effect of attempting to quantify an amount outside of the calibrated region is to underestimate concentration-pathlengths. The SO2/HCl molar ratio in Fig. 5 increases slightly at high concentrations, which indicates that the non-linearity of the HCl calibration curve is more extreme than that for SO2. This is expected, as HCl is a much stronger IR absorber than SO2. This results in an artificially high SO2/HCl for high concentrations. Molar SO2/HCl derived only from the low-concentration SO2 data can therefore be used instead of the high-concentration data in order to negate this calibration effect. This approach yields a slightly lower SO2/HCl for 19 March 2004, 5 March 2005 (by b5%), but does not affect 24 June or 15 July 2005 data significantly. Molar ratios between all three species change rapidly with time and changes in the different ratios correlate weakly (Fig. 5). When plotted on a ternary plot of S/10, Cl and F molar abundances, the gas compositions track between S-rich, F-poor compositions and S-poor, F-rich compositions (Fig. 6). Volcanic gas data from early in the current eruption (1983–1986; Gerlach and Graeber, 1985; Greenland, 1988), collected using Giggenbach sampling techniques (Giggenbach, 1975) and filter pack data collected in 1985 show the same range in HF/HCl as observed in this study (Fig. 7). The range in HF/HCl observed in this study is similar to those measured previously at Kīlauea Volcano. Condensate F/Cl around the 1959–1960 lava flows at Kīlauea ranged from 0.0003 to 0.6 (Murata et al., 1963); and Naughton (1980) found F/Cl in condensates of 0.11–0.82 during the 1974 SW rift zone eruption of Kīlauea. Molar SO2/HCl however, has generally ranged to higher values in previous studies: a mean SO2/HCl of 81 was measured in 1983–1985 during lava fountaining cycles (Gerlach and Graeber, 1985) compared with a mean of 36 in 2004–2005. The mean SO2 flux during 2004–2005 was ~1500 t/d (Elias and Sutton, 2007.). Using the mean SO2/HCl of 33.6 and the mean HF/HCl of 1.0 measured here yields average fluxes of 25 t/d for HCl and 12 t/d HF from Pu u Ō ō (after converting to mass ratios). 4.2. Glass geochemistry Fig. 4. Transmitted light images from an optical microscope built in to the electron microscope, Department of Earth Sciences, Cambridge, to show A: olivine-hosted melt inclusion and B matrix glass with bubble, with measurement profile indicated. Phases labeled Ol (olivine); Pl (plagioclase); Cpx (clinopyroxene–augite); Gl (glass); V (vesicle).

bubbles was typically cracked, with cracks both radial and tangential to the bubble walls (Fig. 4). Matrix glass analyses were carried out in profiles, 10 μm apart, adjacent to bubble walls in order to observe potential diffusion profiles for the volatile species (Fig. 4). 4. Results 4.1. Volcanic gas geochemistry Concentration-pathlengths for each of the three species SO2, HCl and HF are plotted against one another for four separate days of measurements (Fig. 5, Table 1) in order to derive molar ratios between species and assess their variability with time. Abundances of the species range up to 60,000 ppm m SO2, 2300 ppm m HF and 1000 ppm m HCl. SO2 generally correlates with HCl in a linear fashion, particularly at lower end of the range in HCl (Fig. 5). The proportion of HF in the gases is more variable. The SO2–HCl plots show some

Representative matrix glass and melt inclusion compositions obtained in this study are shown in Table 2 and Figs. 8 and 9, with other published data (Hauri, 2002; Swanson and Fabbi, 1973; Harris and Anderson, 1983; Gerlach and Graeber, 1985). Matrix glasses fall within a narrow range of compositions and contain 51.3–51.9 wt.% SiO2 and 6.30–6.75 wt.% MgO. Mean S, Cl and F concentrations are 67, 85 and 173 ppm, with standard deviations of 32, 34 and 80 ppm respectively. On a ternary plot, the matrix glasses in this study plot further from the S apex than those measured previously (Fig. 8), although the matrix glasses of Swanson and Fabbi (1973) were fountain spatter, which suggests the magma may have ascended the conduit rapidly and in disequilibrium with respect to volatile degassing (Namiki and Manga, 2008). Our matrix glasses also display a range of F/Cl ratios, which is consistent with the observation of a range in HF/HCl observed in the volcanic gases. Clear trends relating to crystallisation and degassing cannot be discerned on the plot of Cl and F against S (Fig. 9), although the high end of the range for matrix glasses exceeds most MI Cl contents (Hauri, 2002), suggesting that Cl may be concentrated by some latestage crystallisation prior to degassing. The F contents of glasses in this study are lower than those measured by Hauri (2002) in olivine-

126

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

Fig. 5. Plots to show the composition of volcanic gases at Pu u Ō ō measured by OP FTIR for four days in 2004–5. Left-hand plots show SO2 and HF plotted against HCl (units of ppm m) and right-hand plots show the molar HF/HCl and SO2/HCl ratios plotted against time.

hosted MI from Kīlauea (N350 ppm) and lower than the estimated “residual” (post-degassing and eruption) concentration of F estimated by Gerlach and Graeber (1985). The mass S/Cl and S/F ratio with distance along profiles away from bubble walls are shown in Fig. 10. There is clearly some variability outside the error bars, but no clear trend away from bubble walls on this length-scale. No evidence for diffusive fractionation between S and halogen species during exsolution has therefore survived, but these data do not rule out the possibility that post-quenching diffusion obliterated concentration gradients up to the bubble walls. 5. Discussion We propose that the range and trends in the data are simply related to the last depth of equilibration of the gases and melt in the conduit during open-system degassing. It has been shown that the ascent of large bubbles from 50–100 m in the conduit Pu u Ō ō gives rise to lava spattering activity, with the bubbles containing H2O-rich gases (Edmonds and Gerlach, 2007). It follows that bubbles generated

at shallower depths in such an open-system regime could be enriched in HF compared to deeper-generated larger bubbles, owing to the more rapid depletion of Cl and S in the melt over F. Studies of F degassing in silicic magmas suggest that the vapour–melt partition coefficient is dependent on the total F content of the system and the proportion of H2O in the vapour; in silicic melts containing b1 wt.% F in equilibrium with CO2-bearing fluids, the concentration of F in the fluid increases with decreasing pressure. This behaviour suggests that HF might become enriched in volcanic gases exsolving from a fixed, shallow body of magma over time (Webster and Holloway, 1990). Observations of F-rich segregation veins in olivine tholeiite from Kīlauea provide additional evidence of late-stage F melt enrichment (Helz, 1980; Aoki et al., 1981). Little is known about the partitioning behavior of F at low pressures in mafic melts. It is possible, however, to construct an empirical model based on the glass

Table 1 Mean molar ratios for each of the days of measurement, M: mean, σ: standard deviation, v: coefficient of variability (σ=M) and n: number of spectra (datapoints). Date

Molar ratio

M

σ

V

N

19-Mar-04

HF/HCl SO2/HCl HF/HCl SO2/HCl HF/HCl SO2/HCl HF/HCl SO2/HCl

1.9 40.4 1.2 36.7 1.3 31.2 0.26 26.6

1.6 10.3 0.47 23.6 0.78 7.6 0.16 8.0

84.7 25.5 39.0 64.4 61.6 24.5 63.1 30.2

337

4-Mar-05 24-Jun-05 15-Jul-05

898 197 296

Fig. 6. Ternary plot to show the composition of volcanic gases at Pu u Ō ō on four days in 2004–2005 measured by OP FTIR; axes are S/10, Cl and F (moles).

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

127

Fig. 7. Histograms showing the normalised frequency of HF/HCl (left) and SO2/HCl (right) values from OP FTIR volcanic gas data from Pu u Ō ō for four days during 2004–2005. Previously published data ranges are shown as black bars.

compositions measured in this, and previously published studies, which can be constrained by our knowledge of degassing processes at Kīlauea (Gerlach and Graeber, 1985; Dixon et al., 1991).

Pele's tears as ~10–15%, and this microlite growth was divided into the incremental degassing steps, in proportion to H2O degassing. For closed system degassing, the following mass balance is used to predict Cl concentration in the melt:

5.1. Model for halogen degassing

     X Cl = X0Cl = ð1−χ Þ D X0H2 O −X H2 O + 1 :

The glass geochemistry data, both from this study and those published previously (Figs. 8 and 9) are used to construct a degassing model for halogens, based on partitioning of Cl and F into a vapour phase described by a partition coefficient (taking the case of chlorine, Cl, as an example), DCl fluid-melt, whereby, for open system degassing: X Cl = X0Cl f ðDfluidQmelt −1Þ Cl

ð1Þ

f = 1−MH2 O      MH2 O = X0H2 O −X H2 O + ð1 + χ ÞX H2 O −X H2 O where XCl and XH2O are the concentrations of Cl and H2O in the melt at H2 O a given set of conditions, XCl are the initial concentrations of 0 and X0 H2O and Cl in the primary melt, MH2O is the mass fraction of H2O degassed at each decompression step, f is the fraction of the total H2O remaining in the melt at each decompression step and X is the volume fraction of crystals (microlites) grown at each step as a consequence of H2O degassing. The total proportion of microlites was estimated in the

ð2Þ

The model is based on the observations of volatiles in glasses (Fig. 8), using data from Hauri (2002), this study, Swanson and Fabbi (1973); Harris and Anderson (1983); and Gerlach and Graeber (1985). The S and H2O concentration in the melt with decreasing pressure is predicted using the degassing models of Gerlach (1986) and Dixon et al. (1991). The glass Cl and F concentrations are used to define partition coefficients at each pressure step (over atmospheric pressure), for which glass data is available, for closed and open-system degassing (Fig. 11 A). A negative partition coefficient for the closed-system degassing case reflects an increase in Cl and F in the melt during crystallisation. For open system degassing, the increase in Cl and F during crystallisation is described by a low value of D. Broadly, the distribution coefficient for each is low and approximately constant until pressures of b1 MPa (around 35 m depth), whereupon they increase, reflecting the onset of degassing. The partition coefficient for Cl is higher than that for F in both cases, reflecting the greater tendency for Cl to degas over F. The partition coefficient for Cl increases at a slightly higher pressure than for F, reflecting a slightly earlier onset of degassing for Cl during magma ascent up the conduit.

Table 2 Representative glass compositions in Pele's tears erupted April–May 2006 at Pu u Ō ō, Kīlauea Volcano, Hawai i, in wt.%. 1–5 matrix glasses; 6: melt inclusion in olivine. Point

1

2

3

4

5

6

SiO2 FeO Al2O3 MgO CaO Na2O K2O TiO2 P2O5 Cr2O3 MnO NiO S F Cl Total F/Cl S/Cl

51.5 11.1 13.4 6.70 10.9 2.36 0.430 2.49 0.233 0.032 0.167 0.024 0.0074 0.0205 0.0069 99.38 2.97 2.13

51.5 11.7 13.4 6.67 10.7 2.47 0.419 2.64 0.234 0.033 0.172 0.006 0.0045 0.0165 0.0080 100.02 2.06 1.11

51.8 11.4 13.7 6.68 11.1 2.38 0.408 2.58 0.246 0.034 0.172 0.005 0.0077 0.0148 0.0097 100.48 1.53 1.59

51.6 10.7 13.5 6.53 11.0 2.41 0.378 2.57 0.259 0.039 0.171 0.008 0.0104 0.0180 0.0099 99.27 1.82 2.09

51.9 11.0 13.7 6.48 11.0 2.50 0.452 2.59 0.245 0.028 0.176 0.022 0.0032 0.0185 0.0086 100.14 2.15 0.74

50.7 11.7 13.6 6.35 11.1 2.29 0.422 2.54 0.251 0.029 0.181 0.003 0.0445 0.0205 0.0085 99.31 2.41 10.47

Fig. 8. Glass compositions measured in this study (matrix glasses and olivine -hosted melt inclusions from Pele's tears erupted 2006) and previously published data. Hauri (2002) analysed olivine-hosted melt inclusions; Swanson and Fabbi (1973) matrix glasses of fountain spatter; and Gerlach and Graeber (1985) estimated parental (predegassing), stored (magma equilibrated at summit magma chamber) and residual (post-degassing) concentrations of volatiles from volcanic gas data and glass volatile data from Harris and Anderson (1983).

128

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

Fig. 9. Cl and F concentrations plotted against S concentration in the glass, showing glasses from this study and published previously (legend as for Fig. 8). MG: matrix glass; MI: melt inclusion.

The degassing model is shown in Fig. 11 B and C and Fig. 12. It was found that crystallisation had very little effect on the degassing trends (b3% difference in melt halogen concentrations for each step) and whether degassing was open or closed also had very little effect at these low pressures (Fig. 11 B). Fig. 11 B shows the evolution of melt volatile

Fig. 11. A model to describe halogen degassing at Kīlauea Volcano. A: fluid–melt partition coefficients for Cl and F, calculated as described in the text. B: model evolution of H2O, S, Cl and F melt concentrations with pressure. C: model molar S/Cl and F/Cl in the volcanic gas with pressure.

Fig. 10. Profiles of melt S/Cl and S/F mass ratios, measured by electron microprobe, away from bubble walls in Pele's tears erupted 2006 from Pu u Ō ō. Error bars on the measurements are shown.

Fig. 12. Ternary plot to show the model developed to describe degassing of halogens at Kīlauea Volcano with geochemical data from glasses and volcanic gases also shown (from this study and published previously; see Figs. 6 and 8 for legends).

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

contents with pressure and illustrates the slight increase in both F and Cl concentrations as a consequence of degassing-induced crystallisation and the late onset of degassing compared to H2O and S. Fig. 11 C shows the predicted change in molar gas ratios as a function of pressure, and demonstrates the large changes in gas ratios that can occur as a result of only very small changes in equilibration pressure. This model fits the observed volcanic gas compositions well (Fig. 12). The range in observed gas ratios can therefore be simply explained by gas– melt separation at varying pressures in the upper conduit, up to ~1 MPa above atmospheric pressure. The rapid changes in plume composition may be caused by individual bubbles bursting on the surface of the magma column. The gas composition of the bubbles might reflect their last equilibration depth in the conduit. This model is consistent with other work on halogen degassing from basaltic magmas (Aiuppa et al., 2002, 2004), although at Kīlauea halogen degassing takes place at very low pressures, in contrast to Etna, where it begins at much higher pressures, owing to the higher melt Cl and H2O contents of Etnean magmas. The model can be further tested by applying it to volcanic gases emitted during lava spattering in 2005 (Edmonds and Gerlach, 2007). The gas composition varied throughout cycles of lava spattering, which may be classified as weak strombolian activity. The cycles were interpreted as the bursting of individual large bubbles, consistent with the prevalent view of strombolian eruptive activity (e.g. Houghton and Gonnerman, 2008). The change in gaseous HF/HCl (0.01–2) may be used to calculate the approximate pressure at which these gas compositions are in equilibrium with magma in the conduit. Fig. 13 shows the range in calculated pressures, for both the data presented in this study, and for gases emitted during lava spattering. The gases associated with lava spattering have a significantly “deeper” signature (mode of the histogram 0.28 MPa, compared with 0.03–0.1 MPa for effusive degassing; Fig. 13), consistent with their generation at slightly higher pressures in the conduit. At higher than ~0.3 MPa above atmospheric pressure, the halogen “barometer” becomes relatively insensitive.

129

Our data suggest that diffusive fractionation plays little role in controlling the ratios between S and halogens in the gas phase. Halogen diffusion rates have been measured in basaltic Hawaiitic melt from Mount Etna (Alletti et al., 2007). The study showed that the diffusivities of Cl and F were similar (DF = 1.25 DCl = 4 × 10− 11 m2 s− 1) and about one order of magnitude lower than H2O and a factor of ~ 5 higher than S. In order to generate significant differences in Cl/F in volcanic gases caused by diffusive fractionation, bubble growth rates would have to be high. Changes in Cl/S and F/S, however, could be achieved at lower bubble growth rates. We would therefore expect to see much greater variability in Cl/S and F/S than in Cl/F. In fact, the mean coefficient of variation for HF/HCl is 0.74 and for S/Cl 0.34 (Table 1), which is not consistent with a primary kinetic control. Further, the profiles of S, Cl and F concentrations in glasses adjacent to bubble walls (Fig. 10) show no depletion of S adjacent to bubbles, as you would expect for a slow-diffusing species. The lack of a concentration gradient does not rule out diffusive fractionation however, as post-quenching diffusion may have overprinted it. Another mechanism for inducing variability in halogen concentrations in the gas phase is chemical reactions involving HF. HF may react with silica in plume particles to form silicon tetrafluoride vapour (SiF4). This has been documented at Satsuma-Iwojima (Mori et al., 2002), where SiF4/HF reached values as high as 0.57. Thermodynamic calculations showed that SiF4/HF correlates directly with the total F content of the gases and inversely with temperature. OP FTIR measurements have not detected SiF4 in the Pu u Ō ō gas plume, consistent with the fact that these gases were quenched at a high temperature (N1000 °C). Another possibility is that HF and HCl take part in adsorption reactions in the plume. Water vapour condenses onto the surface of ash particles, which provides a site for dissolution of highly soluble acidic gases such as HF. The fluoride may form soluble salts on the ash surface once the water evaporates, involving Ca2+, Na+ and Mg2+. Cronin and Sharp (2002) found calcium fluosilicate (CaSiF6) and sodium fluoride (NaF) on the surfaces of ash particles at Ambrym and Yasur volcanoes in Vanuatu. Oskarsson (1980) showed that soluble fluoride was transported from Hekla primarily by adsorption onto ash particles after the 1970 eruption. Experiments with ash in the presence of H2O-HF fluids showed that adsorption did not take place at temperatures significantly above 600 °C; the concentration of fluoride adsorbed onto ash particles then increased linearly with decreasing temperature down to around 200 °C. If we take the lack of SiF4 in the gas plume as evidence that the gases were quenched at a high temperature upon exiting the hot vent, then this process is expected to be insignificant. “Etching” has been observed on the surfaces of microspherules collected at Pu u Ō ō and ascribed to the reaction of HF with the silicate surface (Meeker and Hinkley,1993). A study of ash particle surfaces from a range of volcanoes has confirmed that acid-mediated dissolution of the ash's silicate glass and minerals occurs in the presence of fluoride, followed by precipitation at the ash–liquid interface (Delmelle et al., 2007). It is unknown, however, whether these processes can act on timescales of a few seconds, the required timescale for explaining the rapid changes in plume composition observed here. 6. Conclusions

Fig. 13. Histograms to show calculated equilibration pressures, based on HF/HCl in the volcanic gases for gases emitted during effusive activity (this study) and during lava spattering (Edmonds and Gerlach, 2007). These data can be simply interpreted as deeper bubble–melt separation in the lava spattering case, shown in the schematic cartoon to the right.

A study of volcanic gas composition (by OP FTIR) and matrix glass volatile concentration (by electron microprobe) has allowed a model for halogen degassing to be formulated for Kīlauea Volcano, Hawai i. Volcanic gases are characterised by a molar SO2/HCl of 10–64, with a mean of 33; and a molar HF/HCl of 0–5, with a mean of 1.0. The HF/HCl ratio is more variable than the SO2/HCl ratio, and the two correlate weakly. Variations in ratio take place over rapid (seconds) timescales. These measured ranges are similar to halogen ratios measured previously at Kīlauea. Matrix glasses of Pele's tears erupted in 2006 have mean S, Cl and F contents of 67, 85 and 173 ppm respectively, but are associated with a large range in S/F. A model is developed that

130

M. Edmonds et al. / Chemical Geology 263 (2009) 122–130

describes the open system degassing of halogens from parental magmas, using the glass data presented here, previously published results and parameterisation of sulphur degassing from previous work. The results illustrate that halogen degassing takes place at pressures of b1 MPa, equivalent to b~35 m in the conduit. Partition coefficients for Cl and F are low (b1.5); F only degasses appreciably at pressures of b0.1 MPa above atmospheric pressure, virtually at the top of the magma column. This model reproduces the volcanic gas data and other observations of volcanic activity well and is broadly consistent with other studies of halogen degassing from basaltic magmas. The primary difference between halogen degassing at Kīlauea (which erupts H2O-poor basaltic magma) and at H2O-rich basaltic systems is that halogen degassing is transposed to much lower pressures in the conduit at Kīlauea. The model and observations suggest that variation in volcanic gas halogen ratios is caused by exsolution and gas bubble–melt separation at low pressures in the conduit. There is no evidence that near-vent halogen chemistry is important in the system. Diffusive fractionation and disequilibrium degassing cannot be ruled out as playing an important role in determining halogen gas ratios, but there is little surviving evidence in the erupted glasses. The fluxes of HCl and HF from Kīlauea during 2004–5 were ~25 and 12 t/d respectively. Acknowledgements This work was carried out with the support of the USGS Mendenhall Fellowship Program and the Center for the Study of Active Volcanoes (CSAV) at the University of Hawai i, Hilo. Jeff Sutton and Tamar Elias of the Hawaiian Volcano Observatory (USGS) are thanked for their assistance in the field. The study was funded in part by NERC grant NE/G001537/1. Andy Buckley and Chiara Petrone assisted with electron microprobe measurements at the Department of Earth Sciences, University of Cambridge. References Aiuppa, A., Federico, C., Paonita, A., Pecoraino, G., Valenza, M., 2002. S, Cl and F degassing as an indicator of volcanic dynamics: the 2001 eruption of Mount Etna. Geophysical Research Letters 29. doi:10.1029/2002GL015032. Aiuppa, A., Federico, C., Giudice, G., Gurrieri, S., Paonita, A., Valenza, M., 2004. Plume chemistry provides insights into mechanisms of sulfur and halogen degassing in basaltic volcanoes. Earth and Planetary Science Letters 222, 469–483. Aiuppa, A., Federico, C., Franco, A., Giudice, G., Gurrieri, S., Inguaggiato, S., Liuzzo, M., McGonigle, A.J.S., Valenza, M., 2005. Emission of bromine and iodine from Mount Etna volcano. Geochem. Geophys. Geosyst. 6, Q08008. doi:10.1029/2005GC000965. Alletti, M., Baker, D.R., Freda, C., 2007. Halogen diffusion in a basaltic melt. Geochimica et Cosmochimica Acta 71, 3570–3580. Aoki, K., Ishiwaka, K., Kanisawa, S., 1981. Fluorine geochemistry of basaltic rocks from continental and oceanic regions and petrogenetic application. Contributions to Mineralogy and Petrology 76, 53–59. Brenan, J.M., 1993. Partitioning of fluorine and chlorine between apatite and aqueous fluids at high pressure and temperature: implications for the F and C1 content of high P–T fluids. Earth and Planetary Science Letters 117, 251–263. Byers, C.D., Christie, D.M., Muenow, D.W., Sinton, J.M., 1984. Volatile contents and ferric– ferrous ratios of basalt, ferrobasalt, andesite and rhyodacite glasses from the Galapagos 95.5”W propagating rift. Geochimica et Cosmochimica Acta 48, 2239–2245. Cashman, K.V. , Mangan, M.T., Newman, S., 1994. Surface degassing and modifications to vesicle size distributions in active basalt flows. Journal of Volcanology and Geothermal Research 61, 45–68. Cronin, S.J., Sharp, D.S., 2002. Environmental impacts on health from continuous volcanic activity at Yasur (Tanna) and Ambrym, Vanuatu. International Journal of Environmental Health Research 12, 109–123. Delmelle, P., Lambert, M., Dufrêne, Y., Gerin, P., Óskarsson, N., 2007. Gas/aerosol–ash interaction in volcanic plumes: new insights from surface analyses of fine ash particles. Earth and Planetary Science Letters 259, 159–170. Dixon, J.E., Clague, D.A., Stolper, E.M., 1991. Degassing history of water, sulfur and carbon in submarine lavas from Kīlauea volcano, Hawai i. Journal of Geology 99, 371–394. Dixon, J.E., Clague, D.A., Wallace, P.J., Poreda, R., 1997. Volatiles in alkalic basalts from the North Arch Volcanic Field, Hawaii: extensive degassing of deep, submarine-erupted alkalic series lavas. Journal of Petrology 38, 911–939.

Edmonds, M., Gerlach, T.M., 2007. Vapor segregation and loss in basaltic melts. Geology 35 (8), 751–754. Elias, T., Sutton, A.J., 2007. Sulfur dioxide emission rates from Kīlauea Volcano, Hawaii, an update: 2002–2006. United States Geological Survey Open-file report 2007–1114, p. 25. Francis, P.W., Maciejewski, C., Oppenheimer, C., Chaffin, C., Caltabiano, T., 1995. SO2: HCl ratios in the plumes from Mt. Etna and Volcano determined by Fourier transform spectroscopy. Geophysical Research Letters 22, 1717–1720. Gerlach, T.M., Graeber, E., 1985. The volatile budget of Kīlauea Volcano. Nature 313, 273–277. Gerlach, T.M., Terrence, M., 1986. Exsolution of H2O, CO2, and S during eruptive episodes at Kilauea volcano, Hawaii. Journal of Geophysical Research 91 (B12), 12177–12185. Giggenbach, W.F., 1975. A simple method for the collection and analysis of volcanic gas samples. Bulletin Volcanologique 34, 132–145. Greenland, L.P., 1988. In: Wolfe, E.W. (Ed.), Gases From the 1983–84 East-Rift Eruption, in U.S. Geological Survey Professional Paper 1463, pp. 145–153. 145 pls. (folded maps, scale 141:150,000) in pocket. Harris, D.M., Anderson, A.T., 1983. Concentrations, sources, and losses of H2O, CO2, and S in Kīlauean basalt. Geochimica et Cosmochimica Acta 47, 1139–1150. Hauri, E., 2002. SIMS analysis of volatiles in silicate glasses, 2: isotopes and abundances in Hawaiian melt inclusions. Chemical Geology 183, 115–141. Heliker, C.C., Mattox, T.N., 2003. In: Heliker, C.C., et al. (Ed.), The First Two Decades of the Pu u O o-Kupaianaha Eruption: Chronology and Selected Bibliography, in U.S. Geological Survey Professional Paper 1676, pp. 1–27. Helz, R.T., 1980. Crystallisation history of Kīlauea Iki Lava Lake as seen in drill core recovered in 1967–1979. Bulletin of Volcanology 43, 675–701. Houghton, B.F., Gonnerman, H.M., 2008. Basaltic explosive volcanism: constraints from deposits and models. Chemie der Erde 68, 117–140. Kent, A.J.R., Peate, D.W., Newman, S., Stolper, E.M., Pearce, J.A., 2002. Chlorine in submarine glasses from the Lau Basin: seawater contamination and constraints on the composition of slab-derived fluids. Earth and Planetary Science Letters 202 (2), 361–377. Mangan, M., Cashman, K.V., Newman, S., 1993. Vesiculation of basaltic magma during eruption. Geology 21, 157–160. Meeker, G.P., Hinkley, T.K., 1993. The structure and composition of microspheres from the Kīlauea volcano, Hawai i. American Mineralogist 78, 873–876. Miller, T.H., Zoeller, W.H., Crowe, B.M., Finnegan, D., 1990. Variations in trace metal and halogen ratios in magmatic gases through an eruptive cycle of the Pu u Ō ō vent, Kīlauea, Hawaii: July–August 1985. Journal of Geophysical Research 95, 12607–12615. Mori, T., Notsu, K., Tohjima, Y., Wakita, H.,1993. Remote detection of HCl and SO2 in volcanic gas from Unzen volcano, Japan. Geophysical Research Letters 20, 1355–1358. Mori, T., Sato, M., Shimoike, Y., Notsu, K., 2002. High SiF4/HF detected in Satsuma–Iwojima volcano's plume by remote FT-IR observation. Earth Planets Space 54, 249–256. Murata, K.J., Ault, W.U., White, D.E., 1963. Halogen acids in fumarolic gases of Kīlauea volcano. IUGG XIII General Assembly, IAV scientific session. Namiki, A., Manga, M., 2008. Transition between fragmentation and permeable outgassing of low viscosity magmas. Journal of Volcanology and Geothermal Research 169, 48–60. Naughton, J.J., 1980. Composition of some components in gas collected during the 1977 eruption at Kīlauea, Hawaii. Journal of Volcanology and Geothermal Research 7, 319–322. Oskarsson, N., 1980. The interaction between volcanic gases and tephra: fluorine adhering to tephra of the 1970 Hekla eruption. Journal of Volcanology and Geothermal Research 8, 251–266. Rose, W.I., Millard, G.A., Mather, T.A., Hunton, D.E., Anderson, B., Oppenheimer, C., Thornton, B.F., Gerlach, T.M., Viggiano, A.A., Kondo, Y., Miller, T.M., Ballenthin, J.O., 2006. The atmospheric chemistry of a 33–34 hour old volcanic cloud from Hekla Volcano (Iceland): insights from direct sampling and the application of chemical box modeling. Journal of Geophysical Research — Atmospheres 111, D20206. doi:10.1029/ 2005JD006872. Schilling, J.-G., Bergeron, M.B., Evans, R., Smith, J.V., 1980. Halogens in the mantle beneath the North Atlantic. Philosophical Transactions of the Royal Society of London. Series A, Mathematical and Physical Sciences 297, 1431 The Evidence for Chemical Heterogeneity in the Earth's Mantle, pp. 147–178. Stroncik, N.A., Haase, K.M., 2004. Chlorine in oceanic intraplate basalts: constraints on mantle sources and recycling processes. Geology 32, 945–948. doi:10.1130/G21027.1. Sun, W.D., Binns, R.A., Fan, A.C., Kamenetsky, V.S., Wysoczanski, R., Wei, G.J., Hu, Y.H., Arculus, R.J., 2007. Chlorine in submarine volcanic glasses from the eastern Manus basin. Geochimica et Cosmochimica Acta 71, 1542–1552. Swanson, D.A., Fabbi, B.P., 1973. Loss of volatiles during fountaining and flowage of basaltic lava at Kīlauea Volcano, Hawaii. Journal of Research of the U.S. Geological Survey 1, 649–658. Thordarson, T., Self, S., 2003. Atmospheric and environmental effects of the 1783–1784 Laki eruption, Iceland: a review and reassessment. Journal Geophysical Research 103, 27411–27445. doi:10.1029/2001JD002042 108(D1). Villemant, B., Mouatt, J., Villament, A.M., et al., 2008. Andesitic magma degassing investigated through H2O vapour–melt partitioning of halogens at Soufrière Hills Volcano, Montserrat (Lesser Antilles). Earth and Planetary Science Letters 269, 212–229. Webster, J.D., 2004. The exsolution of magmatic hydrosaline chloride liquids. Chemical Geology 210, 33–48. Webster, J.D., Holloway, J.R., 1990. Partitioning of F and Cl between magmatic hydrothermal fluids and highly evolved granitic magmas. Geol. Svc. Amer. Spec. Pap 246, 21–34.