Solubility and fluid-melt partitioning of Cl in hydrous phonolitic melts

PII S0016-7037(00)00386-0

Solubility and fluid-melt partitioning of Cl in hydrous phonolitic melts

S. SIGNORELLI1,* and M. R. CARROLL

2,†

1Department of Earth Sciences, Bristol University, Bristol BS8 1RJ, United Kingdom2Dipartimento di Scienze della Terra, Universita` di Camerino, 62032 Camerino, Italy

(Received October1, 1999;accepted in revised form March7, 2000)

Abstract—Chlorine solubilities and partition coefficients between aqueous chloride solutions and phonoliticmelts were investigated at 860–890°C and pressures of 25 to 250 MPa using both peraluminous (Vesuvius)and peralkaline (Montanã Blanca, Tenerife) phonolitic melt compositions. Depending on the experimentalconditions the phonolites were in equilibrium with either a Cl-bearing aqueous fluid or a subcriticalassemblage of low-Cl aqueous fluid1 Cl-rich brine. The nature of the fluid phase(s) was identified byexamination of fluid inclusions present in run product glasses and the fluid bulk composition was calculatedby mass balance. Chlorine concentrations in the glass increase with increasing Cl molality in the fluid phaseuntil a plateau in Cl concentration is reached when melt coexists with aqueous fluid1 brine. With constantCl molality in the fluid phase, the concentration of Cl in phonolitic melt increases as the pressure decreases.With fluids of similar Cl concentration, higher Cl concentrations are observed in peralkaline phonolitic meltscompared with the peraluminous phonolitic melts; overall the Cl concentrations observed in phonolitic meltsare approximately twice those found in rhyolitic melts under similar conditions. The observed negativepressure dependence of Cl solubility implies that Cl contents of melts may actually increase during magmadecompression if the magma coexists with aqueous fluid and Cl-rich brine.Copyright © 2000 ElsevierScience Ltd

1. INTRODUCTION

Chlorine generally occurs as a minor component in mostmagmas and fumarolic gases. However, it has been observedthat the release of Cl from volcanoes to the atmosphere duringeruptions and quiescent degassing may have significant effectson global climate, contributing to thinning of the ozone layerand the production of acid rain (Johnston, 1980; Devine et al.,1984; Symonds et al., 1988; Albritton, 1989). Studies of Cldegassing behavior may provide insights into subsurface move-ments of magma and eruption likelihood (Love et al., 1998;Francis et al., 1998), but such applications require adequateknowledge of the factors controlling Cl distribution betweenmelt and fluid phases in decompressing magmas. The workdescribed in this paper involves experimental determination ofthe solubility and the fluid/melt partitioning of chlorine inwater-rich, evolved alkaline magmas at shallow crustal depths.The distribution of Cl between a melt and an aqueous phase isgenerally expressed in the literature as Nerst-type partitioncoefficients (DCl

m/aq). However, only components that can varyindependently in a phase or a system have a real distributioncoefficient (Candela and Piccoli, 1995) and this is clearly notstrictly true for Cl (although it would be true for componentssuch as NaCl, KCl, etc.; see Candela and Piccoli, 1995). How-ever, given the lack of knowledge concerning Cl speciation inthe fluid phase we will use a Nernst-style partition coefficient inour discussion of melt-fluid partitioning of Cl. Throughout thispaper we will use fluid as a general term for an aqueous phase,hydrosaline liquid or brine to refer to a Cl-rich aqueous brinephase, and vapor to refer to either a Cl-poor aqueous phase that

may coexist with hydrosaline brine when fluid-phase immisci-bility occurs (e.g., Sourirajan and Kennedy, 1962) or a super-critical fluid; silicate melt will generally be referred to simplyas melt.

Following the pioneering work on chlorine behavior inevolved silicic melts (Holland, 1972; Kilinc and Burnham,1972), numerous studies have investigated the distribution ofchlorine between coexisting melt and fluid phases (see refer-ences below). The experimental results show that Cl partitionsstrongly into aqueous fluids relative to silicate melts by a factorof ;20 to 300. However, most of these studies have beenperformed on rhyolitic melt compositions and at pressuresabove 200 MPa (e.g., Webster, 1992a,b; Kravchuk and Kep-pler, 1994; Student and Bodnar, 1999) where only a single,supercritical fluid phase is present. These data are not directlygermane to the distribution of chlorine between melts and fluidsunder conditions where immiscible vapor and hydrosaline brinecoexist. If we consider the most important binary aqueouselectrolyte system [i.e., (Na, K)Cl–H2O] for interpreting mag-matic fluids, many studies show that this system is character-ized by immiscibility under a wide range ofP–T conditionsencountered in the shallow crust (Sourirajan and Kennedy,1962; Bodnar et al., 1985; Chou, 1987; Anderko and Pitzer,1993a,b). Below the critical point a low salinity aqueous fluidcoexists with a hydrosaline liquid or brine and thus Cl parti-tioning behavior is controlled by the coexistence of threephases (melt, vapor, and brine), neglecting the possible pres-ence of Cl-bearing crystalline phases. The few experimentscarried out in this region for hydrous rhyolitic melts havedocumented that (1) chlorine does not partition strongly intothe water-rich fluid phase and the vapor/melt partition coeffi-cient is;5 at 100 MPa (Webster and Holloway, 1988; Shino-hara et al., 1989; Me´trich and Rutherford, 1992; Williams,1995; Candela and Piccoli, 1995); (2) there is a positive cor-

* ([email protected])† Author to whom correspondence should be addressed:([email protected])

Pergamon

Geochimica et Cosmochimica Acta, Vol. 64, No. 16, pp. 2851–2862, 2000Copyright © 2000 Elsevier Science LtdPrinted in the USA. All rights reserved

0016-7037/00 $20.001 .00

2851

relation between the vapor/melt distribution coefficient andtotal pressure (Shinohara et al., 1989; Malinin et al., 1989); and(3) chlorine solubility, that is the maximum amount of Cl in amelt coexisting with a brine1 vapor assemblage, depends onmelt composition and pressure. The solubility varies from a fewthousand ppm to;1 wt.%, being higher in peralkaline rhyolites(pantellerites) relative to metaluminous compositions (Shino-hara et al., 1989; Me´trich and Rutherford, 1992).

The Cl solubility in water-saturated phonolites was previ-ously investigated (Me´trich and Rutherford, 1992) by a singleexperiment at 850°C and 100 MPa, which showed a melt Clcontent of 62706 170 ppm for a bulk fluid composition of 4.7m Cl (where m indicatesmolality), suggesting considerablyhigher Cl concentrations in phonolitic melts compared withrhyolitic melts formed under similar conditions. The new re-sults presented in this paper provide information concerning thefollowing questions: (1) how does Cl solubility in alkalinephonolites differ from that in the more well-studied rhyoliticmelts; (2) how is Cl solubility in these alkaline compositionsaffected by melt peralkalinity; (3) how do we expect Cl degas-sing behavior to vary at lower pressures where a single aqueousfluid phase may split into coexisting immiscible phases; and (4)how do the experimental data compare with chlorine abun-dance in natural glasses of similar composition. In order toanswer these questions we have conducted water-saturatedexperiments on natural phonolites at 860–890°C and pressuresbetween 25 and 250 MPa, using fluids with different chloridemolalities.

2. EXPERIMENTAL AND ANALYTICAL METHODS

2.1. Starting Materials

Experiments were performed on two natural phonolites with chem-ical compositions reported in Table 1. The K-rich sample is a peralu-minous phonolitic pumice collected from the fallout deposit of the 79A.D. eruption of Vesuvius (southern Italy). The pumice is aphyric toslightly porphyritic, with ,10 vol.% crystals, mainly sanidine withsubordinate biotite, amphibole, opaque minerals, and leucite. The Na-rich sample is a phonolitic peralkaline glass collected from the 2020year-old subunit II A of the Montanã Blanca pumice deposit (CanaryIsland, Spain). The sample consists of 96–99 vol.% fresh, dark-greenglass and 1–4 vol.% euhedral phenocrysts. Feldspar is modally dom-inant (90%), and other phases, in order of decreasing abundance, aresodic clinopyroxene, biotite, magnetite, ilmenite and apatite. The samesample was used previously in a study of H2O solubility in phonoliticmelts (Carroll and Blank, 1997).

2.2. Experiments

Experimental charges consisted of 30–60 mg of powdered sampleand an aliquot of aqueous solution containing chlorine as (Na, K)Cl,added by micropipette, sealed in a 2.2 mm diameter Ag75Pd25 or Aucapsule. The initial amount of chlorine in aqueous solutions waschecked using a Dyonex high-performance ion chromatograph (HPIC).The initial fluid/powder ratio never exceeded 0.2 by weight, in order tominimize the amount of silicate components dissolved in the fluidphase. To minimize possible cation exchange between the fluid and themelt, the molar Na/K ratio in the fluid phase was adjusted to equal thatof silicate starting materials. Microprobe analyses of run productsverify the success of this method in minimizing changes in meltcomposition (Tables 3 and 4). Capsules containing silicate powder1chloride solution were sealed with an oxy-acetylene torch or a tung-sten-arc welder. They were weighed before and after welding to checkfor loss of aqueous solution and then baked at 120°C overnight to checkfor leaks.

Experiments were conducted in vertical, rapid-quench, water-pres-

surized cold-seal vessels consisting of a Nimonic 105 vessel connectedto a 30 cm long stainless-steel extension by means of a water-cooledcoupling nut. The sample holder consisted of magnetically levitatedstainless-steel filler rod with a magnetic cup at one end and an Inconel600 cup at the other end, where the sample capsules were placed. Thisassemblage produces an oxygen fugacity of NNO1 1 (60.5) aroundthe capsules (verified using NiO–NiPd alloy oxygen fugacity sensors;Taylor et al., 1992) and allowed rapid quenching of the samples bylowering the magnet to the base of the stainless-steel extension. Notethat experiments at different oxidation states might produce resultsdiffering from ours since it has been observed that Cl solubility may beinfluenced by melt iron content, although it is currently not clearwhether ferric or ferrous iron has more effect on melt Cl contents(Metrich and Rutherford, 1992). Quenching was accompanied by apressure rise of;30 MPa as the filler rod in the furnace was replacedby water. The quenching rates are greater than 100°C/s (Carroll andBlank, 1997). Experiments were run for time periods of 48 to 342hours; most runs were.140 h. Temperatures were measured usingsheathed chromel–alumel thermocouples placed in a well at the base ofthe pressure vessel. Thermocouple accuracy was checked against a typeB Pt–Rh thermocouple which was calibrated at the gold melting point.Pressure was measured using Bourdon-tube gauges or an electronicpressure transducer (Nova Swiss). Uncertainties in pressure and tem-perature are estimated as62.5 MPa and65°C. The experimentalprocedure involved first pressurizing the sample to approximately halfof the desired final pressure, then heating the sample and allowingpressure to increase with heating until it reached the desired pressure;at this point pressure was kept constant by manually bleeding thepressure line.

After the experiment the capsules were removed from the pressurevessels, dried and weighed. Any capsules showing anomalous weightchanges were discarded and capsules from successful runs were piercedwith a needle. The punctured capsules were set in an oven at 120°C for5 min and then reweighed to check for total amount of solutionevaporated. Afterwards, capsules were torn open with pliers and frag-ments from each sample were mounted in epoxy resin and prepared aspolished thin sections for petrographic and electron microprobe anal-ysis.

Experiments on the Montanã Blanca (MB) phonolite all producedclear greenish glasses, some of which contained very low (,5 vol.%)

Table 1. Composition (wt.% on water free basis with original ana-lytical totals) of starting phonolites.a

Sample MB Ves

SiO2 59.5 57.3TiO2 0.66 0.19Al2O3 19.0 21.7Fe2O3 3.86 1.28FeO 1.08MnO 0.2 0.13MgO 0.33 0.23CaO 0.79 2.94Na2O 10.09 5.68K2O 5.56 9.53P2O5 0.07 0.04Cl 0.35 0.55F 0.09 0.4L.O.I. 0.45 3.10Total 100.04 99.8

A.I.b 1.2 0.91A/CNKc 0.8 0.87N/NKd 0.73 0.48

a Major elements, Cl and F for MB have been determined by XRF.For Ves analysis for major elements were carried out by XRF and bytraditional wet chemical methods, and Cl and F by potentiometry.

b Agpaitic index5 [(Na2O 1 K2O)/Al2O3] molar ratio.c Molar [(Al2O3/CaO1 Na2O 1 K2O)] ratio.d Molar [Na2O/(Na2O 1 K2O)] ratio.

2852 S. Signorelli and M. R. Carroll

amounts of crystals (Fe–Ti oxides, biotite and apatite). The glassescontain two types of fluid inclusions. Below approximately 2m Cl(fluid composition from mass balance, as discussed below) all theinclusions contain both vapor and liquid phases (type-I inclusion). Thevolume of the vapor phases in these inclusions increases with decreas-ing experimental pressure. Above 2m Cl the inclusions, whatever thepressure and the temperature of the experiment, contain a vapor bubble,liquid and one or more daughter minerals (type-2 inclusions). Theminerals contained in this inclusion type are believed to be alkalichlorides, possibly with other chlorides such as CaCl2 and FeCl2.Generally, experiments from below 200 MPa are found to contain bothtype-I and type-II inclusions (Fig. 1).

Experiments with Vesuvius phonolitic samples (Ves) contain glasswith variable amounts of crystals. Samples produced above 150 MPacontain,5 vol.% of crystals (Fe–Ti oxides, plagioclase, apatite, andbiotite), whereas those from lower pressure experiments show increas-ing amounts of crystals with decreasing pressure. The abundance ofleucite increases markedly as pressure falls from 100 to 50 MPa; noleucite was observed at 150 MPa.

Microprobe analyses of experimental glasses (Tables 1 and 3) indi-cate that the compositions of MB run product glasses do not differsignificantly from the starting material. More substantial compositionalchanges from the bulk starting composition are evident in Ves runproduct glasses (Tables 1 and 4). The main differences are related to theexperiments below 100 MPa, where the crystallization of increasingamounts of leucite with decreasing pressure causes residual melts tobecome more sodic and the ratio of Na2O/(Na2O 1 K2O) increaseswith increasing degree of crystallization. Other than the change in Na/Kand slight variation in peraluminosity the Ves liquids remain broadlyphonolitic in composition (see Table 4).

2.3. Approach to Equilibrium

Figure 2 shows the variation of Cl in phonolitic run-product glassesas a function of experiment duration at 100 MPa and 870°C. For MBphonolite we find no significant change in Cl concentration in the glasswith run durations between 50 and 192 h. Although several previousstudies have suggested that 2–4 day run durations were sufficient toachieve equilibrium Cl solubilities (e.g., Malinin et al., 1989; Shino-hara et al., 1989; Me´trich and Rutherford, 1992), we note that in thecase of the Ves phonolite it appears that the attainment of equilibriumrequires 6 days. The composition of Ves starting material (Na2O/K2O 5 0.60) is almost identical to that (Na2O/K2O 5 0.55) of Metrichand Rutherford (1992), suggesting the results may be comparable.Two-stage experiments (runs ss59 and ss60), carried out by first hold-ing a sample for 48 hours at 150 MPa before lowering the pressure to100 MPa at 5 MPa per hour, yield results in excellent agreement withthe other experiments (Table 2). The agreement between these exper-

iments and those in which the sample was put directly at the desired runconditions supports our claims that the results represent a close ap-proach to equilibrium.

2.4. Mass Balance Calculations

The concentration of chlorine in the fluid phase at run conditions wascalculated from mass balance constraints as follows:

Mass of Cl in the fluid5

CClfl,i p Mfl 1 CCl

sil,i p Msil 2 CClgl p (Msil 1 Malk

gl 1 MH2Ogl 1 MCl

gl 2 MDSfl ),

where

CClfl,i p Mfl 1 CCl

sil,i p Msil 5 total mass of Cl in the experiment

and

CClgl p (Msil 1 Malk

gl 1 MH2Ogl 1 MCl

gl 2 MDSfl ) 5 mass of Cl in final glass

with C 5 concentration; M5 mass, fl5 fluid; i 5 initial; gl 5 glassin run product; alk5 alkalies added to the glass during the run; sil5silicate powder before the run; DS5 silicate dissolved in fluid. Theamount of water in glass after the run was estimated using the exper-imental H2O solubility data of Carroll and Blank (1997). The mass ofalkalies gained by the glass was calculated considering that chlorine isadded to the melt as (Na,K)Cl and that the (Na/K) ratio of melt andfluid remain constant. The mass of silicate glass dissolved into the fluidat run conditions was considered in first approximation negligible. Forwt.% DS of 5.5% (Webster and Holloway, 1988) the calculated Clmolality in the fluid at the end of a run decreases by 5% relative(Metrich and Rutherford, 1992). For experiments conducted in thesubcritical region of the system NaCl–H2O, where two immisciblefluids are present, the calculated molalities represent the chlorine con-tent of both phases (i.e., the sum of Cl in vapor and brine).

The total error (61s) associated with the calculated mass of Cl influid phase is the result of weighing errors for the starting materials(fluid and powder), error in estimation of water content in the runproduct glasses, error on the estimation of Cl in the starting materials(fluid and natural phonolites), and error in the electron microprobeanalysis for Cl in the run product glasses. The errors connected with theweight of starting materials, with the estimation of Cl in the initial fluidphase and the H2O in the product run glasses are very low and theyhave a limited effect on the final result (,5% relative). The largesterrors are associated with the estimation of Cl in phonolitic startingmaterials and in run product glasses. Comparative analyses for Cl byXRF, potentiometry and electron microprobe indicate that the startingmaterials are homogeneous and, for example, the MB phonolite startingmaterial has 35006 100 ppm Cl. For fluids containing less than 1mCl at run conditions the former error affects the calculated fluid com-position by approximately625% relative. Likewise, an uncertainty of6100 ppm Cl in a sample with 5000 ppm Cl in the run product glasscan yield uncertainties in the calculated fluid molality on the order of25%. Taken together, these uncertainties can yield potential errors influid/melt partition coefficients on the order of 50%. However, theseerrors are actually quite small when considered in the context of thewide range in fluid/melt partition coefficients observed in this study(,1 to ;200) and we suggest that the internal precision of the data(e.g., Fig. 3) indicates smaller uncertainties than these worst-caseestimates. For the two experiments conducted under supercritical con-ditions with fluid Cl molalities.15 (runs ss53, ss54; see Table 2) thedilution of water activity in the fluid phase may yield slightly lowerthan expected melt H2O contents. However, the data of Webster(1992a; 1992b) suggest that reductions in water solubility will not bemore that 5–10% relative and this has minimal effect on our conclu-sions.

2.5. Electron Microprobe Analyses

Chlorine and major-element concentrations in glasses and crystalswere determined using JEOL 8600SX and Cameca Camebax electronmicroprobes at the Department of Earth Sciences, Bristol University.Because the run product glasses contained abundant fluid inclusions

Fig. 1. Transmitted light image of type-I and type-II fluid inclusionsin glass quenched from experiment at 100 MPa and 870°C (run ss16,see Table 2). The inclusions consist of vapor bubbles (V), liquid phases(L) and cubic daughter minerals (dm).

2853Solubility and fluid-melt partitioning

and sometimes crystals (e.g., biotite, apatite, leucite and so on), theywere carefully examined by optical and electron microscopy to selectpart of the glasses convenient for the analysis; in all the experiments itwas possible to find multiple large pools of glass suitable for analysis.At least 20 glass analyses were performed on each run product and nodifferences were observed for the same samples analyzed with thedifferent microprobes. Glasses and phenocrysts were analyzed at anaccelerating voltage of 15 kV and 7–10 nA beam current, with a 15315 micron rastered beam. The primary standards were: St. John’s Islandolivine for SiO2 and MgO, andradite for CaO and FeO, albite for Na2O,

Eifel sanidine for K2O, SrTiO3 for TiO2, MnO for MnO, and NaCl forCl. In general, all major elements were counted for 10 seconds on peakand 5 seconds on background; Si was counted for 15 s. Chlorine wascounted for 60 seconds on peak and background to improve analyticalprecision. Sodium migration was minimized by analyzing sodium first,using a low sample current, a wide rastered beam, and a short counttime. Despite these precautions, a sodium loss of 5–10% relativeduring the analysis was detected for water-rich (.5 wt.%) glasses.This was verified by analyzing selected samples with a 2 nAbeamcurrent (Morgan and London, 1996), which invariably gave Na2O

Fig. 2. Chlorine wt.% of (A) MB phonolitic melt and (B) Ves phonolitic melt versus the duration of experiments. Theerror bars on the points show the standard deviation (1s) of the analyses.


values in agreement with that of the starting material. As secondarystandards for chlorine we used KN18 comenditic glass containing3700 ppm Cl (Mosbah et al., 1991), apatite USNM 104021 contain-ing 4100 ppm Cl (Jarosewich et al., 1980) and KE12 pantelleritic

glass containing 3300 ppm Cl (Devine et al., 1984; Me´trich andRutherford, 1992). Analyses of these standards over the period ofthis study (.30 for each standard) give the following results: 0.3560.03 wt.% Cl for KN18, 0.416 0.02 wt.% Cl for the apatite and

Table 2. Experimental run, conditions, and results from starting materials.

No.Duration(hours)

P(MPa)

T(°C)

Mass s.m.(mg)a

Mass fl(mg)b

Cl in glass(wt.% 6 s)

Cl in fluid(moles/kg)c fl/md DCl

v/m e

Montana Blanca (MB)ss83bisa 338 25 890 34 2.7 0.836 0.03 5–3.93 0.06 17*ss75 198 50 880 41 6 0.346 0.02 0–0.02 0.11 0.2ss64 289 50 880 42 2.93 0.396 0.01 0.2–0.05 0.04 0.4ss66 288 50 880 49.48 2.91 0.536 0.02 1–0.12 0.03 0.8ss65bis 294 50 880 42.03 5.88 0.536 0.02 0.5–0.15 0.11 1ss77 294 50 880 40.19 5.88 0.826 0.03 4–3.82 0.11 17*ss50 301 50 880 49.12 2.92 0.836 0.03 4–2.94 0.03 12*ss67 288 50 880 44.95 3.39 0.856 0.02 4–3.21 0.04 13*ss1 120 100 870 30 3.92 0.296 0.02 0–0.19 0.08 2.4ss5 148 100 870 55 5.00 0.586 0.03 1–0.44 0.04 2.7ss9 48 100 870 60 5.00 0.546 0.02 1–0.7 0.03 4.6ss13 96 100 870 47.67 5.20 0.586 0.01 1–0.6 0.05 3.6ss23 48 100 870 47.89 5.46 0.496 0.03 0.6–0.38 0.06 2.7ss27 148 100 870 48.41 5.61 0.496 0.03 0.6–0.38 0.06 2.8ss25 96 100 870 49.49 5.48 0.486 0.02 0.6–0.36 0.06 2.6ss20 148 100 870 35.3 5.46 0.666 0.03 1–0.57 0.1 3ss14 48 100 870 48.95 6.10 0.816 0.03 5–7.66 0.07 34*ss16 96 100 870 48.58 6.22 0.776 0.02 5–7.75 0.07 36*ss32 148 100 870 48.31 6.39 0.836 0.03 5–7.23 0.07 31*ss59 192 100 870 47.7 5.94 0.86 0.02 4–5.42 0.07 24*ss55 148 150 870 50 6.5 0.266 0.02 0–0.36 0.06 4.9ss56 170 150 870 50.93 6.07 0.356 0.03 0.2–0.41 0.05 4.1ss57 170 150 870 47.27 6.5 0.466 0.01 0.6–0.67 0.06 5.1ss58 170 150 870 46.45 6.31 0.566 0.02 1–1.04 0.06 6.6ss90 175 150 870 37.5 7.90 0.716 0.02 3–3.9 0.13 20*ss45 288 150 870 48.09 6.93 0.686 0.01 4–7.85 0.07 41*ss45bis 170 150 870 49.87 7.09 0.696 0.02 5.5–13.17 0.06 68*ss52 148 200 860 50 6.52 0.226 0.03 0–0.75 0.04 12ss61 168 200 860 45.6 5.87 0.296 0.02 0.2–0.92 0.04 11ss62 168 200 860 44.36 6.14 0.346 0.02 0.5–1.35 0.05 14ss63 168 200 860 45.21 6.52 0.416 0.03 1–2.29 0.05 20ss44 292 200 860 33.18 6.87 0.546 0.03 4–7.62 0.11 49ss44bis 168 200 860 45.25 6.81 0.586 0.02 5–18.2 0.06 111ss81 169 250 860 37.22 8.03 0.136 0.02 0–0.58 0.1 19ss82 169 250 860 37.78 7.52 0.166 0.01 0.2–0.97 0.09 21ss69 168 250 860 38.74 5.16 0.296 0.02 0.5–2.99 0.03 37ss70 170 250 860 37 6.73 0.306 0.02 1–2.66 0.07 32ss68 168 250 860 40.57 6.82 0.396 0.01 2–6.16 0.06 56ss71 170 250 860 41.72 8.17 0.526 0.02 5–15.31 0.08 104ss54 237 250 860 50.99 8.33 0.666 0.03 5–26.03 0.05 140

Vesuvius (Ves)ss78a 342 30 885 42.65 4.08 0.896 0.03 5–5.1 0.07 20*ss49a 291 50 880 48.39 3.34 0.786 0.02 5–8.57 0.04 39*ss15a 48 100 870 42.38 5.94 0.616 0.03 5–9.26 0.08 54*ss17a 96 100 870 46.57 6.35 0.636 0.04 5–9.38 0.08 53*ss31a 144 100 870 41.62 6.05 0.706 0.02 5–8.53 0.09 43*ss48a 240 100 870 45.52 6.14 0.706 0.02 5–9.08 0.08 46*ss48bisa 240 100 870 48.41 6.24 0.716 0.02 5–9.4 0.07 47*ss60a 192 100 870 45.64 6.81 0.686 0.02 5–8.46 0.09 44*ss41a 288 150 870 47.31 6.37 0.576 0.01 5–15.6 0.06 97*ss39a 293 200 870 44.29 7.25 0.476 0.02 5–17.7 0.07 133ss53a 237 250 860 48.41 8.24 0.496 0.03 5–27.59 0.06 200

a Mass of starting material powder.b Mass of starting material fluid.c Initial chlorine molality—chlorine molality of a single fluid phase recalculated as detailed in the text.d Fluid/melt weight ratio during the run.e Vapor/melt distribution coefficient.* These values are for experiments with vapor1 brine and so they are apparent distribution coefficient that would apply to a single-phase fluid

(see text). All the experiments were done in Ag75Pd25 capsules except those marked with (a) done in Au capsules.


0.32 6 0.02 wt.% Cl for KE12. The detection limit for Cl wasestimated to be 0.03 wt.% (2s), whereas the analytical precision wasestimated to be67.5% and 10% (2s) at 8000 ppm and 3000 ppm,respectively.

3. RESULTS

Experimental run conditions, the concentration of Cl in thesilicate melt and that calculated by mass balance for the bulk

fluid are given in Table 2. These data are used to formulateNernst-type partition coefficients for the distribution of Clbetween the fluid and the melt (DCl

v/m). The Cl concentrations ofMB phonolite glasses range from 0.13 to 0.85 wt.% and thoseof Ves glasses range from 0.47 to 0.89 wt.%. The calculated Clconcentrations of the fluid(s) present in the experiments rangefrom approximately 0.02 to 27.6 moles/kg. The apparent fluid/

Fig. 3. Chlorine molality (CClm) of MB phonolitic melts versus Cl molality of a calculated single fluid phase. Experiments

carried out in the two-phase field (melt1 aqueous fluid) are represented as open circles and those in the three-phase field(melt 1 aqueous fluid1 hydrosaline brine) are represented as solid circles. The interpretative curves among the data arebased on Shinohara et al. (1989).CCl

m (moles/kg)' CClm (ppm)/35,450.


melt distribution coefficients vary from 0.2 to 140 for MBphonolite and from 20.3 to 200 for the Ves phonolite. We referto these as apparent distribution coefficients because some ofthe more Cl-rich bulk fluid compositions actually correspond toexperiments in which immiscible vapor and brine coexist withphonolitic melt. Chemical compositions of representative run-product glasses for MB and Ves phonolites are given in Tables3 and 4, respectively.

The variation of Cl in MB phonolite glass with the calculatedCl concentration in a single fluid phase during the runs isshown in Fig. 3. The experiments carried out between 50 and150 MPa display the common behavior of a positive linearcorrelation between Cl in glass and in the fluid phase, when Clmolality in the fluid phase is below 2m Cl, and then a flatdistribution where further increases of Cl in the fluid phase donot increase Cl in the glass. The experiments at 200 and 250

Table 3. Representative analyses of major and trace elements (wt.%6 1s on water-free basis with original analytical totals) from Montanã Blancaexperiments.

Experiment ss83bis ss67 ss16 ss45bis ss63 ss54

T (°C) 890 880 870 870 870 860

P (MPa) 25 50 100 150 200 250

No. analyses 31 25 23 20 22 21

(wt.% 6 s)SiO2 59.376 0.2 59.636 0.40 60.606 0.19 60.796 0.41 60.786 0.51 60.916 0.44TiO2 0.626 0.04 0.606 0.04 0.596 0.06 0.616 0.03 0.576 0.04 0.616 0.05Al2O3 18.876 0.1 19.016 0.19 18.646 0.19 19.066 0.13 19.016 0.17 19.216 0.18FeOtot 3.266 0.14 3.286 0.13 2.886 0.13 2.856 0.17 3.316 0.30 2.896 0.14MnO 0.226 0.06 0.206 0.05 0.136 0.04 0.106 0.04 0.196 0.12 0.106 0.05MgO 0.326 0.03 0.316 0.03 0.346 0.03 0.306 0.02 0.306 0.04 0.226 0.02CaO 0.716 0.03 0.706 0.03 0.796 0.08 0.746 0.05 0.726 0.05 0.756 0.05Na2O 10.026 0.21 9.606 0.01 9.706 0.20 9.246 0.21 8.886 0.25 9.006 0.18K2O 5.766 0.05 5.806 0.08 5.546 0.14 5.576 0.11 5.806 0.14 5.586 0.12Cl 0.856 0.03 0.876 0.02 0.816 0.02 0.746 0.02 0.456 0.03 0.736 0.03Total 97.776 0.48 97.756 0.60 94.856 0.58 93.326 0.75 91.286 0.94 90.906 0.51

Crystalsa O B, O, Ap O B, O B, O B, O, ApA.I. 1.20 1.16 1.18 1.12 1.10 1.10A/CNK 0.79 0.81 0.82 0.87 0.88 0.89N/NK 0.73 0.72 0.73 0.71 0.70 0.71

a Mineralogical assemblage after the run:O 5 Fe–Ti oxides;Ap 5 apatite;B 5 biotite; A.I., A/CNK, and N/NK are defined in Table 1.

Table 4. Representative analyses of major and trace elements (wt.%6 1s on water-free basis with original analytical totals) from Vesuvius (Ves)experiments.

Experiment ss78 ss49 ss60 ss41 ss39 ss53

T (°C) 885 880 870 870 870 860

P (MPa) 30 50 100 150 200 250

No. analyses 20 24 32 27 30 22

(wt.% 6 s)SiO2 55.446 0.42 56.886 0.36 57.126 0.29 57.246 0.25 57.326 0.37 56.926 0.47TiO2 0.296 0.04 0.246 0.05 0.206 0.05 0.236 0.15 0.206 0.05 0.196 0.04Al2O3 22.406 0.21 21.826 0.34 21.196 0.2 21.636 0.15 22.016 0.26 22.356 0.16FeOtot 2.456 0.16 1.996 0.17 1.236 0.13 1.656 0.06 1.476 0.10 1.486 0.11MnO 0.046 0.01 0.076 0.03 0.096 0.04 0.086 0.05 0.086 0.05 0.066 0.05MgO 0.166 0.02 0.166 0.03 0.146 0.04 0.176 0.02 0.166 0.02 0.116 0.02CaO 2.366 0.13 2.846 0.32 2.496 0.22 2.596 0.06 2.736 0.08 2.826 0.08Na2O 8.316 0.24 7.346 0.19 6.186 0.15 5.646 0.15 5.576 0.13 5.576 0.12K2O 7.666 0.20 7.846 0.31 9.646 0.24 10.156 0.13 9.966 0.16 9.976 0.14Cl 0.906 0.03 0.826 0.02 0.726 0.02 0.626 0.01 0.516 0.02 0.546 0.03Total 99.396 0.50 95.476 0.55 93.906 0.40 92.496 0.50 91.986 0.54 90.616 0.49

Crystalsa Lc, A, C, P, S, O, Sph Lc, C, A, O Lc, O O, B, P, Ap O, P O, B, P, ApA.I. 0.98 0.95 0.93 0.93 0.91 0.90A/CNK 0.93 0.94 0.97 0.96 0.98 0.99N/NK 0.62 0.6 0.5 0.46 0.46 0.46

a Mineralogical assemblage after the run:Lc 5 leucite;C 5 clinopyroxene;O 5 Fe–Ti oxides;A 5 amphibole;P 5 plagioclase;S 5 sanidine;Sph5 sphene;Ap 5 apatite;B 5 biotite; A.I., A/CNK, and N/NK are defined in Table 1.


MPa, on the contrary, show a parabolic increase in melt Clcontent with increasing Cl in the fluid (i.e., the rate of increasein melt Cl content slows in experiments with more Cl-richfluids). Figure 4 shows the isobaric trends for all of the datasuperimposed on the same diagram. This comparison clearlyshows that as pressure increases there is a decrease in slope ofthe straight line and a shift of the pivot point between thestraight line and the plateau towards more rich-Cl fluids untilthe sharp break of slope between them turns into a smoothpassage between 200 and 250 MPa.

4. DISCUSSION

4.1. Cl Solubility Behavior

Previous studies involving hydrous rhyolitic melts equili-brated with chlorine-bearing aqueous solutions at low-mediumpressure have pointed out that the behavior of chlorine in themelt is a strong function of the fluid phase assemblage (1- or2-phase fluid; Shinohara et al., 1989; Malinin et al., 1989;Metrich and Rutherford, 1992). For experiments carried out inthe subcritical region of the system NaCl–H2O where twoseparate volatile phases coexist with silicate melt, Gibbs’ phaserule predicts that, at givenT andP, Cl concentration in all thephases (melt–vapor–hydrosaline brine) is fixed. A flat distri-bution or plateau of Cl content in the silicate melt (Fig. 3) is thetypical expression of the melt being saturated with both ahydrosaline liquid and an H2O-rich vapor and it is assumed todefine the solubility limit for Cl in the hydrous silicate melt. Todefine DCl

v/m it is a common approach to use experimentsperformed in the two-phase field (melt–vapor). Various authors(Webster and Holloway, 1988; Malinin et al., 1989; Shinoharaet al., 1989; Me´trich and Rutherford, 1992) have demonstratedthat in the two-phase field the concentration of Cl in the melt isfunction of the concentration of Cl in the fluid phase. Shinohara(1994) has proposed an alternative approach to calculateDCl

v/m

which utilizes the fixed concentration of all of the phases whena mixture of a vapor and a hydrosaline liquid is in equilibriumwith a silicate melt.CCl

v is obtained from the model NaCl–H2Osystem (e.g., Chou, 1987; Anderko and Pitzer, 1993a) andCCl

m

is obtained from measured Cl content of the silicate melt.Following this model,DCl

l/m, the Cl distribution coefficient be-tween a hydrosaline liquid (l) and a silicate melt (m), can bedetermined.

In order to determine the number of volatiles phases thatwere stable during each experiment, the calculated concentra-tion of Cl in a single fluid phase (on an equivalent molecularfraction or wt.% NaCl basis) was plotted in the pressure-XNaCl

diagram of the system NaCl–H2O (Fig. 5).The effect of KCl onaqueous phase immiscibility was not considered since it has notbeen studied at the temperature of interest of this study. Studiesat lower temperature (400–600°C) suggest that the addition ofKCl tends to enlarge the vapor–brine immiscibility field withthe general effect of decreasing Cl in the aqueous vapor andincreasing Cl in the brine (e.g., Anderko and Pitzer, 1993b). Allthe experiments between 50 and 150 MPa involving a#2m Clfluid phase and those at pressure 200–250 MPa fall in thesupercritical field. This suggests that the silicate melt coexistedwith a single aqueous vapor phase and the observed fluidinclusions (type-I inclusions) support this interpretation. Forexperiments at 200 and 250 MPa, involving$2m Cl fluids,only type-II inclusions were observed. The fluid composition ofall the experiments involving a$2m Cl fluid at pressurebetween 20–150 MPa lies within the immiscibility gap in Fig.5. Thus, at theP andT of interest here, there is a three-phaseequilibrium involving silicate melt, aqueous vapor, and hydro-saline brine. Fluid inclusions found within the run productglasses support this interpretation.

Comparing the results of Fig. 3 with those of Fig. 5, it is clearthat the behavior of Cl in hydrous phonolitic melts is affectedby the phase relations in the fluid phase. When the phonoliticmelt coexists with a single supercritical vapor phase the amountof dissolved Cl depends on the concentration of Cl in thecoexisting fluid phase, in agreement with a monovariant systemat givenT andP. This equilibrium has been used to calculatethe averagedDCl

v/m (slope of the linear solubility trends in Fig.4) as a function of pressure for MB phonolite (Fig. 6). Forexperiments at 200 and 250 MPa the inclination of the curvebelow 2m Cl was used. The averaged distribution coefficientsrange from 0.8 at 50 MPa to 25 at 250 MPa. Experiments in thethree-phase field (subcritical) were not taken into account be-cause they contain two volatile phases and thus only give anapparent fluid/melt distribution coefficient. Alternatively, sincemost of the experiments for the Ves phonolite were performedin the three-phase field (melt–vapor–brine), we have estimatedthe vapor composition in these experiments by the modelNaCl–H2O system from Anderko and Pitzer (1993a). Thesedata, along with the solubility of Cl in Ves phonolite experi-ments yieldDCl

v/m which range from 1.7 at 50 MPa, to 5 at 100MPa, to 16 at 150 MPa. In Fig. 6 we can see a strong pressuredependence of the partition coefficient of Cl between vapor andphonolitic melt similar to that observed for rhyolitic melt (opencircles in Fig. 6). Within the error the trends are parallel andthere is not any significant difference inDCl

v/m between phono-lites and rhyolites.

Given the model NaCl–H2O system from Anderko andPitzer (1993a), the composition of a hydrosaline liquid inequilibrium with a Cl-saturated silicate melt can be estimatedand hydrosaline liquid/silicate melt partition coefficient can becalculated as well. This yields the following estimatedDCl

l/m for

Fig. 4. Summary of variation of the chlorine molality in the melt withthe chlorine molality in a single fluid phase. The vertical bars show thetwo-sigma standard deviations on analyses of chlorine in the silicatemelt.


MB (l 5 hydrosaline liquid, m5 silicate melt):;260 at 50MPa,;125 at 100 MPa and;90 at 150 MPa.

When the phonolitic melt coexists with vapor and hydrosa-line liquid, the amount of Cl in the run product glasses does notdepend on the Cl concentration in the fluids and the Cl solu-

bility is fixed at a givenT andP (plateau in Cl concentration inFig. 3 and 4). Cl solubility in melts saturated with aqueousfluid 1 brine decreases significantly as the pressure increasesfrom 25 to 200 MPa (Fig. 7). However, the two startingmaterials display different behavior at the lower pressuresinvestigated. In Ves phonolite, Cl solubility shows a negativelinear correlation with pressure over the entire pressure rangeinvestigated (30–200 MPa), whereas in MB phonolite the Cl

Fig. 5. Pressure–XNaCl diagram of the system NaCl–H2O with 850°C and 900°C isotherm (redrawn after Anderko andPitzer, 1993a). The polythermal projection of the three-phase vapor1 brine 1 halite equilibrium onto the pressure-composition section is from Bishoff (1991) and Anderko and Pizer (1993a). The circles and the squares represent theapproximate wt.% NaCl equivalent calculated concentration (see the text) of a single fluid phase coexisting at run conditionswith MB and Ves phonolitic melts.

Fig. 6. Variation ofDClv/m with the pressure for MB (solid circles) and

Ves (open squares) samples. Data for rhyolites (open circles) fromHolland (1972), Webster and Holloway (1988), Shinohara et al. (1989),Malinin et al. (1989), Me´trich and Rutherford (1992), Williams (1995),Candela and Piccoli (1995), Student and Bodnar (1999).

Fig. 7. Chlorine solubility in phonolitic melts saturated with aqueousfluid 1 brine versus pressure. Solid circles and open squares representMB and Ves phonolitic glasses (61s standard deviation), respectively.


solubility appears to be constant at pressures below 50 MPa.This might be related to increased amounts of HCl° in the fluidphase at lower pressures (e.g., Shinohara et al., 1984; Fournierand Thompson, 1993; Shinohara et al., 1994) or it mightindicate saturation of the brine with a crystalline chloride phase(difficult to verify since many crystalline chlorides form onquenching). On the other hand, the flattening of Cl solubilitycurve for MB phonolite at low pressure could be the result ofincomplete equilibrium. The linear increase of Cl solubility inVes phonolite through the same range inP may be related tochanges in melt composition due to leucite crystallization atlower pressure (see Table 4). As a result of leucite crystalliza-tion, the Ves phonolitic melts become more sodic and lessperaluminous with decreasing pressure, concomitant factorsthat could favor the dissolution of more Cl instead of thepresence of a plateau.

Various previous studies have demonstrated that the solubil-ity of Cl in hydrous rhyolitic melts depends strongly on meltcomposition (e.g., Carroll and Webster, 1994, and referencestherein). The agpaitic index [A.I.5 molar ratio of (Na1K)/Al] is positively correlated with Cl solubility in hydrousrhyolitic melts and as shown in Fig. 8, changing melt compo-sition from metaluminous rhyolite to peralkaline pantelleritemay yield a three-fold increase in Cl solubility (e.g., Me´trichand Rutherford, 1992). Similar behavior is observed for pho-nolitic melts, with the peralkaline MB phonolite (A.I.5 1.2)having higher Cl solubility than the peraluminous Ves phono-lite (A.I. 5 0.93). In comparing Cl solubilities in rhyolitic andphonolitic melts, the data in Fig. 8 show that for comparablevalues of A.I., the solubility of Cl in phonolites is approxi-mately double that in rhyolitic melts. The most evident differ-ence between the two silicate melts is the lower silica activityof phonolitic melts. The anticorrelation between Cl solubilityand silica activity is consistent with prior experimental work(Kotlova et al., 1960; Iwasaki and Katsura, 1967; Webster,1997a; 1997b; Webster and Rebbert, 1998; Webster et al.,1999).

4.2. Volcanological Implications

Previous studies of Cl solubility in hydrous rhyolites in thepresence of subcritical (Na,K)Cl–H2O fluids have shown that

Cl solubility decreases regularly as the pressure rises. Thisresult has been reaffirmed by the results of our experiments onnatural phonolites. The higher Cl solubility in phonolites withrespect to rhyolites, however, suggests that phonolites duringtheir ascent could transport to the surface larger amounts ofchlorine. This reservoir of chlorine may be potentially emittedto the atmosphere if significant Cl is lost during slow cooling,crystallization and degassing of volcanic products.

The composition of high-temperature gases released fromvolcanoes depends on factors such as gas solubility in thesilicate melt and vapor/melt partitioning and separation duringthe generation and rise of magmas. Giggenbach (1996) hasproposed a model for equilibrium degassing (i.e., closed-sys-tem degassing) that take into account these factors. It, however,requires that volatile solubilities approach Henry’s law behav-ior and that the solubilities decrease with decreasing pressure(as observed for H2O, CO2, noble gases). The negative corre-lation of bothDCl

m/v and melt Cl concentration with pressuremeans that Cl solubility behavior cannot be treated as simply asother volatile species. The solubility of Cl is also complicatedby the formation of nonvolatile alkali chlorides and HCl° at lowpressure. Alternatively, various authors have convincinglydemonstrated that the potential distribution of Cl in magmaticfluids depends on the shape of the immiscibility solvus ofNaCl–H2O system at low pressures and magmatic temperatures(e.g., Shinohara, 1994; Hedenquist and Lowenstern, 1994; He-denquist, 1995). During decompression and/or crystallizationof a magma at low (,200 MPa) pressures it may becomesaturated with both a low-density vapor and a high-densitybrine (first and/or second boiling). As described by Hedenquistand Lowenstern (1994), the low density phase will containcomponents such as H2O, CO2, S, HCl and the higher densitybrine phase will contain alkali salts, metals, water, and sulfates.The shape of NaCl–H2O (see Fig. 5) predicts that upon furtherisothermal decompression, the vapor and the brine will pro-gressively become less and more saline, respectively. Theresults of this study support this model.

Once information of Cl solubility in silicate melts is avail-able, indirect evidence for immiscibility between silicate meltsand H2O–NaCl fluids is provided by the study of the distribu-tion of Cl in glass inclusions present in magmatic phenocrysts.These inclusions are small samples of silicate melt trappedduring crystal growth and they represent the most accessiblesamples of nondegassed magma (Roedder, 1984; Lowenstern,1995, and references therein). Alternatively, matrix glassesfrom rapidly erupted and cooled pumices (e.g., from Plinianeruptions) can provide useful information on some volatiles(e.g., Gardner et al., 1998; Signorelli and Capaccioni, 1999).Data collected from the literature on Cl contents of phonoliticglass inclusions and matrix glasses are presented in Fig. 9. Thecontent of Cl in phonolitic melts is quite variable, ranging from0.18 to 1.03 wt.%. The highest abundance of Cl occur involcanic products from Vesuvius (1.03 wt.%), whereas thelowest pertain to Laacher See (0.18 wt.%), Montanã Blanca(;0.35 wt.%) and Tambora (;0.2 wt.%). Provided that themagma was water saturated and the magmatic reservoir waslocated at pressures below about 200 MPa, the similarity ofnatural and experimental data for Vesuvius suggests that thevast majority of Vesuvian phonolitic melts were saturated withrespect to a hydrosaline liquid. If these assumptions are also

Fig. 8. Solubility of chlorine at 100 MPa as a function of molar(Na1 K)/Al ratio for natural rhyolites and phonolites saturated with anaqueous fluid1 hydrosaline brine.


valid for Laacher See, Montanã Blanca, and Tambora, thecomparison between natural and experimental data suggeststhat the pre-eruptive volatile phase in these systems consistedof only a low-Cl aqueous vapor (i.e., no hydrosaline brine). Thecomparison of Cl in glass inclusions and degassed matrixglasses, when possible, suggests that Cl appears be retained inthe silicate melts during the Plinian phase of phonolitic erup-tions. This is not the case for the Laacher See and Tamboraeruptions, where chlorine has apparently shown a volatile be-havior, with Cl content in matrix glasses lower than in glassinclusions (Devine et al., 1984; Harms and Schmincke, 1998and 2000). There is a sharp gap between the field observationand the laboratory experiment concerning the behavior of chlo-rine during syn-eruptive decompression. Recently, Gardner etal. (1998) have shown that at fast decompression rates Clcannot diffuse out efficiently, although water can. On thecontrary, when the erupted material cools more slowly Cl islost, presumably as a result of groundmass crystallization.

5. CONCLUSIONS

New experimental results on chlorine solubility in phonoliticmelts and vapor/melt and hydrosaline brine/melt partitioningunder magmatic conditions provide insights concerning thegeochemical behavior of Cl in evolved alkaline melts. Theresults of this study are directly applicable to volcano monitor-ing by allowing better interpretation of the origins of variationsin measured halogen emissions from alkaline magmatic sys-tems such those in Italy and Canary Island. In more generalterms, the new data on Cl solubility as a function of magmacomposition and pressure will provide better constraints on thevolcanic degassing of Cl and help in the identification oferuptive products preserved in the geologic record which mayhave been associated with large Cl emissions. This will aidstudies seeking to correlate the record of climatic variability

with the record of volcanic eruptions and will allow betterestimation of past volcanic input to the atmosphere.

At pressures and temperatures outside the vapor–brine im-miscibility field of the NaCl–H2O system the concentration ofCl in phonolitic melts is function of the activity of Cl in thefluid phase. Under conditions within the vapor–brine immisci-bility field, the concentration of Cl in melt is constant and isbuffered by the coexistence of aqueous vapor, hydrosalineliquid 1 silicate melt. Cl solubility and Cl partitioning betweenhydrosaline liquid and melt exhibit a strong negative depen-dence on the pressure, whereas Cl partitioning between aque-ous vapor and melt shows a positive pressure dependence.These features may be attributed to the shape of immiscibilityfield in the NaCl–H2O system. When compared with experi-mental results on rhyolites, the solubility of Cl in phonolites isapproximately twice as high under similar conditions. Theseresults demonstrate that Cl solubility in hydrous silicic melts isa strong function of melt composition and high solubilities canbe expected in alkali-rich, low silica activity compositions.

Acknowledgments—S. Signorelli acknowledges an E. C. bursary in theEnvironment programme—Climatology and Natural Hazards—Con-tract No. ENV-CT97-5081. Additional support was provided by PRE-ERUPT project, Contract No. ENV4-CT96-0259. We thank U. Mues-Schumacher who provided unpublished data about natural phonolitesfrom his M.S. thesis, S. Kearns for the assistance during the electronmicroprobe analysis and D. Hawkins for the analysis of Cl in the initialfluids by HPIC. Detailed comments and suggestions by P. A. Candela,J. D. Webster, and an anonymous reviewer improved the paper, and aregreatly appreciated.

REFERENCES

Albritton D. L. (1989) Stratospheric ozone depletion: global processes.In Ozone Depletion, Greenhouse Gases, and Climate Change. Wash-ington, D.C., pp. 10–18. National Research Council.

Anderko A. and Pitzer K. S. (1993a) Equation-of-state representationof phase equilibria and volumetric properties of the system NaCl–H2O above 573 K.Geochim. Cosmoschim. Acta57, 1657–1680.

Anderko A. and Pitzer K. S. (1993b) Phase equilibria and volumetricproperties of the systems KCl–H2O and NaCl–KCl–H2O above 573K: Equation of state representation.Geochim. Cosmoschim. Acta57,4885–4897.

Bishoff J. L. (1991) Densities of liquides and vapors in boiling NaCl–H2O solutions: A PVTX summary from 300 to 500°C.Am. J. Sci.291,309–338.

Bodnar R. J., Burnham C. W., and Sterner S. M. (1985) Synthetic fluidinclusions in natural quartz. III. Determination of phase equilibriumproperties in the system H2O–NaCl to 1000°C and 1500 bars.Geochim. Cosmochim. Acta49, 1861–1873.

Candela P. A. and Piccoli P. M. (1995) Model ore-metal partitioningfrom melts into vapor and vapor/brine mixtures. InMagmas, Fluidsand Ore Deposits(ed. J. F. D. Thompson), Vol. 23, pp. 101–127.Mineralogical Association of Canada.

Carroll M. R. and Blank J. G. (1997) The solubility of H2O inphonolitic melts.Am. Mineral.82, 549–556.

Carroll M. R. and Webster J. D. (1994) Solubilities of sulfur, noblegases, nitrogen, chlorine, and fluorine in magmas. InVolatiles inMagmas (eds. M. R. Carroll and J. R. Holloway), Vol. 30, pp.231–279. Mineralogical Society of America.

Chou I. M. (1987) Phase relation in the system NaCl–KCl–H2O. III:Solubilities of halite in vapor-saturated liquids above 445°C andredetermination of phase equilibrium properties in the system NaCl–H2O to 1000°C and 1500 bars.Geochim. Cosmochim. Acta51,1965–1975.

Cioni R., Civetta L., Marianelli P., Me´trich N., Santacroce R., andSbrana A. (1995) Compositional layering and syneruptive mixing of

Fig. 9. Histogram of chlorine content of glass inclusions and matrixglasses from phonolites. The data are from the following volcanicsystems: Vesuvius (Cornell and Sigurdsson, 1987; Mues-Schumacher,1994 and unpublished data; Cioni et al., 1995; Cioni et al., 1998;Signorelli and Capaccioni, 1999; Signorelli et al., 1999), Tambora(Devine et al., 1984), Montanã Blanca (Signorelli, unpublished data),and Laacher See (Harms and Schmincke, 1998; 2000).


a periodically refilled shallow magma chamber: The A.D. 79 Plinianeruption of Vesuvius.J. Petrol.36, 739–776.

Cioni R., Marianelli P., and Santacroce R. (1998) Thermal and com-positional evolution of the shallow magma chambers of Vesuvius:Evidence from pyroxene phenocrysts and melt inclusions.J. Geo-phys. Res.103,18277–18294.

Cornell W. C. and Sigurdsson H. (1987) Composition zoning in Pom-pei 79 A.D. pumice deposit; magma mixing and observed trends.Eos68, 434 (abstr.).

Devine J. D., Sigurdsson H., and Davis A. N. (1984) Estimates of sulfurand chlorine yield to the atmosphere from volcanic eruptions andpotential climatic effects.J. Geophys. Res.89, 6309–6325.

Fournier R. O. and Thompson M. J. (1993) Composition of steam in thesystem NaCl–KCl–H2O quartz at 600°C.Geochim. Cosmochim.Acta 57, 4365–4375.

Francis P., Burton M. R., and Oppenheimer C. (1998) Remote mea-surements of volcanic gas compositions by solar occulation spec-troscopy.Nature396,567–570.

Gardner J. E., Rutherford M., and Hort M. (1998) Degassing of tracegases during volcanic eruptions.Eos79, F936 (abstr.).

Giggenbach W. F. (1996) Chemical composition in volcanic gases. InMonitoring and Mitigation of Volcanic Hazard(eds. R. Scarpa andR. I. Tilling), pp. 221–256. Springer–Verlag.

Harms E. and Schmincke H. U. (1998) Volatile composition of theLaacher See phonolite magma (East Eifel, Germany, 12900 BP):Implications for syneruptive S, F, Cl and H2O degassing.Eos 79,F936 (abstr.).

Harms E. and Schmincke H. U. (2000) Volatile composition of thephonolitic Laacher See magma (12,900 yr BP): Implications forsyneruptive degassing of S, F, Cl and H2O. Contrib. Mineral. Petrol.138,84–98.

Hedenquist J. W. (1995) The ascent of magmatic fluid: Dischargeversus mineralization. InMagmas, Fluids and Ore Deposits(ed.J. F. D. Thompson), Vol. 23, pp 263–289. Mineralogical Associationof Canada.

Hedenquist J. W. and Lowenstern J. B. (1994) The role of magmas inthe formation of hydrothermal ore deposits.Nature370,519–527.

Holland H. D. (1972) Granite, solutions, and base metal deposits.Econ.Geol.67, 281–301.

Iwasaki B. and Katsura T. (1967) The solubility of hydrogen chloridein volcanic rock melts at a total pressure of one atmosphere and attemperatures of 1200°C under anhydrous conditions.Bull. Chem.Soc. Jpn.40, 554–561.

Jarosewich E., Nelen J. A., and Norberg J. A. (1980) Referencesamplesfor electron microprobe analysis.Geostand. News.4, 43–47.

Johnston D. (1980) Volcanic contribution of chlorine to the strato-sphere: More significant to ozone than previously estimated?Science209,491–492.

Kilinc I. A. and Burnham C. W. (1972) Partitioning of chloride be-tween a silicate melt and coexisting aqueous phase from 2 to 8 kb.Econ. Geol.67, 231–235.

Kotlova A. G., Ol’shanskii Y. I., and Tsvetkov A. I. (1960) Sometrends in immiscibility effects in binary silicate and borate systems.Mineral. Geokhim. Tr. Inst. Geol. Rudn. Mest. AN SSSR42, 3–9.

Kravchuk I. F. and Keppler H. (1994) Distribution of chloride betweenaqueous fluids and felsic melts at 2 kbar and 800°C.Euro. J.Mineral. 6, 913–923.

Love S. P., Goff F., Counce D., Siebe C., and Delgado H. (1998)Passive infrared spectroscopy of the eruption plume at Popocate´petlvolcano, Mexico.Nature396,563–567.

Lowenstern J. B. (1995) Applications of silicate melt inclusions to thestudy of magmatic volatiles. InMagmas, Fluids and Ore Deposits(ed. J. F. D. Thompson), Vol. 23, pp. 71–99. Mineralogical Associ-ation of Canada.

Malinin S. D., Kravchuk I. F., and Delbove F. (1989) Chloride distri-bution between phases in hydrated and dry chloride–alumosilicate

melt systems as a function of phase composition.Geochem. Intern.26, 32–38.

Metrich N. and Rutherford M. J. (1992) Experimental study of chlorinebehavior in hydrous silicic melts.Geochim. Cosmochim. Acta56,607–616.

Morgan G. B., VI and London D. (1996) Optimizing the electronmicroprobe analysis of hydrous alkali aluminosilicate glasses.Am.Mineral. 81, 1176–1185.

Mosbah M., Metrich N., and Massiot P. (1991) PIGME fluorine deter-mination using a nuclear microprobe with application to glass inclu-sions.Nucl. Instrum. Methods Phys. Res.58, 227–231.

Mues-Schumacher U. (1994) Chemical variations of the 79 AD pumicedeposits of Vesuvius.Eur. J. Mineral.6, 387–395.

Roedder E. (1984)Fluid Inclusions; Minerallogical Society of Amer-ica. Rev. Mineral.12, 644 pp.

Shinohara H. (1994) Exsolution of immiscible vapor and liquid phasesfrom a crystallizing silicate melt: Implications for chlorine and metaltransport.Geochim. Cosmochim. Acta58, 5215–5222.

Shinohara H., Iiyama J. T., and Matsuo S. (1984) Comportment duchlore dans les syste`ms magma granitique-eau.C.R. Acad. Sci. Paris298,741–743.

Shinohara H., Iiyama J. T., and Matsuo S. (1989) Partition of chlorinecompounds between silicate melt and hydrothermal solutions.Geochim. Cosmochim. Acta53, 2617–2630.

Signorelli S. and Capaccioni B. (1999) Behaviour of chlorine prior andduring the 79 A.D. Plinian eruption of Vesuvius (southern Italy) asinferred from the present distribution in glassy mesostases andwhole-pumices.Lithos 46, 715–730.

Signorelli S., Vaggelli G., and Romano C. (1999) Pre-eruptive volatile(H2O, F, Cl and S) contents of phonolitic magmas feeding the3550-year-old Avellino eruption from Vesuvius, southern Italy.J.Volcanol. Geotherm. Res.93, 237–256.

Sourirajan S. and Kennedy G. C. (1962) The system H2O–NaCl atelevated temperatures and pressures.Am. J. Sci.260,115–141.

Student J. J. and Bodnar R. J. (1999) Synthetic fluid inclusions XIV:coexisting silicate melt and aqueous fluid inclusions in the hap-logranite–H2O–NaCl–KCl system.J. Petrol.40, 1509–1525.

Symonds R. B., Rose W. I., and Reed M. H. (1988) Contribution of Cl-and F-bearing gases to the atmosphere by volcanoes.Nature 334,415–418.

Taylor J. R., Wall V. J., and Pownceby M. I. (1992) The calibration andapplication of accurate redox sensors.Am. Mineral.77, 284–295.

Webster J. D. (1992a) Fluid-melt interactions involving Cl-rich gran-ites: Experimental study from 2 to 8 kbar.Geochim. Cosmochim.Acta 56, 659–678.

Webster J. D. (1992b) Water solubility and chlorine partitioning inCl-rich granitic systems: Effects of melt composition at 2 kbar and800°C.Geochim. Cosmochim. Acta56, 679–687.

Webster J. D. (1997a) Exsolution of Cl-bearing fluids from chlorine-enriched mineralizing granitic magmas and implications for oremetal transport.Geochim. Cosmochim. Acta61, 1017–1030.

Webster J. D. (1997b) Chloride solubility in felsic melts and the role ofchloride in magmatic degassing.J. Petrol.38, 1793–1807.

Webster J. D. and Holloway J. R. (1988) Experimental constraints onthe partitioning of Cl between topaz rhyolite melt and H2O andH2O 1 CO2 fluids: New implications for granitic differentiation andore deposition.Geochim. Cosmochim. Acta52, 2091–2105.

Webster J. D. and Rebbert C. R. (1998) Geochemical evidence of fluidsaturation in felsic magma determined through experimental inves-tigation of H2O and Cl solubilities in F-enriched rhyolites melts.Contrib. Mineral. Petrol.132,198–207.

Webster J. D., Kinzler R. J., and Mathez E. A. (1999) Chloride andwater solubility in basalt and andesitic melts and implications formagmatic degassing.Geochim. Cosmochim. Acta63, 729–738.

Williams T. J. (1995) Copper and HCl in felsic magmatic systems.Ph.D. dissertation, University of Maryland.


Documents

Solubility and fluid-melt partitioning of Cl in hydrous phonolitic melts