= Chemical Composition of Volcanic Gases W.P. Giggenbach, Abstract “The composition of gases released from voleanoes isa Tancion of deep proceses, suchas vapor. net separation during the generation and rs ofthe magmas, ad shallow preter, active within the oleae strctures themcves. OF the thee major kpes of voleanc systems, those associated ‘ith andesite magmatism ate the most suitable othe application of geochemical surveillance and monitoring techniques, Volatile contests of andesite magmas, largely representing Buds clesed From the subducted slab, are likely tobe high enough te allow a separate, volatle-ch phase to be present during al stages of magni generation and migration, In pte bighly variable solubilities fn magmatic melts, the proportions of volatiles presen inthe vapors discharged fem saeanic fumaroles resemble closely those acquired at depth suggesting thatthe vapor systems have attained steady-state and thatthe overall process ative during the Hie ofthe volatiles selectively nonfractionating ‘The variability in the composition of volcanic gases accessible to sampling, therefore, is lnvgly due to shallow processes, such as reequilibration in response to coling and dilution by meteoric wate, and interaction with fuids of associated hydrothermal systems. The mos i portantfler ofthe redon state of volcanic gasesis that nvaliag HS aed SO, ofthe vapor phase ‘The concentrations of rly responding species, such a 1s apd CO, reflet redox contol by this “gas buffer”, Species with slower kine wespanses, sch as CHa, are largely generated within the slovemoving hydrothermal systems associated wih most acite volcan structure. At shallow’ level, the contents of HS and SO; are affected by the deposition and revaporatie of elemental S and disproporionaion of SO; to HS and sulfate in a hydrothermal environment, Rea “mauris ofthe high!y soluble species HCl and HF are azo aflected by interaction with hyd ‘hermalsolations, The compositions of the inet species, such as COs, Np and the noble pas provide information on the origins of the vapors, from the mantle, the subducted sla, oF boundwater. 1 Introduction Collection of volcanic-gas samples suitable for detailed chemical analysis is generally restricted to volcanoes in comparatively mild, fumarolic states of activity. OF the voleanoes associated with the three major types of maemas, such conditions are most commonly encountered at andesitic voleanoes where comparatively high gas contents, intermediate magma viscosities and suff 1 and Nuclest Scienes, PO Bor 31 Testate of G Lower Hurt, New Zealand22 WH Giggenbch ciently low solidus temperatures combine to protong the degassing process and allow channel ways for the gases rising from deeper levels to remain open. In the case of basaltic magmas, the low viscosities promote rapid separation of volatiles and mek, giving rise to extensive removal of volatiles during effusive cruptions, wich the essentially degassed magma freezing at high temperatures thus impeding steady-state degassing during noneruptive periods. For rhyolitic ‘magmas the situation is reversed; there high viscosities keep high amounts of solatiles dissolved or suspended in the magma to be released, in part, only during violent eruptions. Because of their large numbers, their propensity for comparatively violent eruptions and frequently close association with populated areas, andesitic volcanoes represent also the most important source of volcanic hazard. Be~ cause of this, and often favorable conditions for voleanic-gas sampling, tech- niques based on gas geochemistry are especially valuable in the surveiltance of this type of volcano. ‘Two major sets of processes are likely to affect the composition of gases released from active volcanoes: those active at depth and associated with the generation of the various magma types and their ascent; and those affecting the composition of the gases after separation from their parent magmas, closer to the surface or within the volcanic structure itself. Both are equally important and, therefore, both are discussed in detail, 2 Factors Controlling the Composition of Magmatic Gases 2.1 Magma Type The majority of volcanic-gas samples suitable for detailed analysis has been collected from andesitic volcanoes during comparatively quict states of activ ity. Gases from basaltic magmas are accessible only under exceptional cumstances, such as the occasional formation of stable lava lakes or the rare ‘occurrence of high-temperature fumaroles. Inthe case of rhyolite magmas, the composition of associated vapors can generally be deduced only indirectly. A preliminary evaluation of the composition of gases from geothermal and vol caanie systems associated with non-are basaltic and andesitic magmatism Giggenbach 19922) showed that “basaltic” gases have No/He ratios of 10-200 and CO,/He ratios of 20:000-60 000, much lower than those of “andesitic” {gases of 1000-10 000 and 65 000-900 000 respectively. The excess Np and CO: in the andesitic gases were ascribed to the decompasition of subducted sedi ments with only minor contributions from the mantle. Based on the isotopic composition of high-temperature volcanic water discharges. a >95% sub- ducted marine origin was suggested for the magmatic component in arcstype voleanic, or andesitc fluids (Giggenbach 19926). Chemical Compasion of Volanic Gases 223 By far the major proportion of igneous material produced over the Taupo Voleanie Zone in New Zealand is of rhyolitic composition (Cole 1990). Based oon detailed surveys of the gases there, however, no distinet chemical or isotopic signatures uniquely associated with this highly siliceous type of activity wece detected (Giggenbach and Glover 1992; Giggenbach et al, 1993), except for a generally reduced gas content of vapor discharges from geothermal systems likely to be associated with rhyolitic intrusions as compared to vapors from geothermal and volcanic systems associated with arc-lype, andesitic magma- tism (Giggenbach 1995). The contents of the more soluble constituents, such as, HO and those containing $ and Cl, in gases from rhyolitic centers resemble those of volatiles from andesitic systems; relative contents of the less soluble constituents, He, Ne, Ar, No and CO>, correspond to mixtures of meteoric (air- saturated groundwater) with essentiaily mantle-derived volatiles suggesting that rhyolitic vapors are derived from several sources: the mantle, crustal of subducted sediments, and groundwater. These findings are compatible with recent theories according to which rhyolitic magmas of the Taupo Volcanic Zone may have formed through differentiation of a high-alumina basalt var- ‘ously contaminated by assimilation of crustal sediments (McCulloch et al 1994; Graham et al, 1995). The high-alumina basalt itself is thought to have formed through interaction of material of the mantle wedge with volatiles released from the subducting slab. The differences in the compositions of gases extracted from volcanic rocks may be assumed to reflect, at least to some degree, differences in the compo- sition of the votatiles present in the parent magmas themselves. In Table 1 the volatile contents in melt inclusions or the chilled margins of submarine lavas are given for the three major types of volcanic tock. A major difference among the three sets is the low average H;O content of 2770 mg/kg in non-are basalts as compared to those of 17 000 mg/kg in andesitic rocks and 45 000 mg/kg in rhyolitic rocks. The concentrations of the four remaining species CO;, S, Cl, and F are quite similar in both andesites and rhyolites, but very different from those in the non-arc basalts. COs and S contents in basalts are higher by a factor of about ten, Cl and F contents are much lower. Generation of andesitic and rhyolitic melts from mantle material alone would require addition of large uantities of H;0 and Cl, and removal of much of the original CO> and S. Mantle-derived melts and more siliceous magmas, therefore, appear to have entirely different volatile contents and probably origins (Table 1). One process potentially capable of generating large variations in relative volatile contents of magmas is vapor-melt separation, or partial degassing, ‘which would allow the less soluble species such as the noble gases, N:, and CO. to escape early on, with water and the more soluble S., Cl-, and F-containing species remaining in the residual melts. ei2M WF Gi “Table, Volatile contents of submarine basalts or mat inclusions in vole rocks, in mer. Most values represent averages of bout a oven ii iualanalses Source o ©, cl eh Basals East Pacific Rise 1201630 oo 2” 0 4 Mid-Atlantic Ridge aio 2070 (a a Galapagos Spreading C980 970 10 200 Ca Kitaven, East Rift 4031001080 5 > Kilauea, Summit po 801300 oD} MO Ridge Basal 2100070 - . a, East Rit sri0 983 90 5 4 27 10010 es Andesites Kata, andesite 000 1 10 st Katmai, dacte 0 m9 im 6g Pinatubo asm % wo r Galers 20000 60 mo 0g Whiteland so 100 oh Mean vom = tos 150060 Rhyolis Katmai 00 <6 10 Cater Lake 3ss00 16 mm 0 Bishop sro 220 too Ce i ‘Taupo Volcanic Zone 4000 content of 1 wt % (10-000 mgike) would already start to “boil” at a depth of 80 km, a depth approaching that of the zone of arc-magma generation. In the estimate of the depth of first boiling of the magmas, the effets of supersaturation required to initiate bubble formation (Sparks 1978; Bottinga and Javoy 1990; Mangan et al. 1993) were neglected, but so were the combined, potentially significant effects on P, of the other volatiles. According to Fig. 1, COp represents the main pressure-generating species at vesiculacities <1%. At higher vaporimelt Be Chemical Composition of Volcanic Gases 227 weil, Ws a ® 7 : : : K, «| @ 1 ‘oe 7 a 2 : ts 10] oo “EHO a 5 fea : 7 a a {8).8 9 function ofthe vaporimelt volume ratio R.. Vales of P, were calculated by us of Ea. (2) farCOs contents ofthe init met of 10, 1000 and 10-000 age a HO | COs weight ratios of Vand 5 corresponding to H;0/ CO; molevatoe of 0,24 and 12, and forz melt with an inal water content of Sw %, WI Represents conditions assumed for the Ieelydegating magma body Underlying White lands, PR reretens conditions reported for the MAR “popping rock ratios, H3O contributes increasingly more to total gas pressure to become the Fain ‘pressure-generating species for H;0/COz >1 and at Ry *0.1. The pressures generated by a mell containing only water are shown in Fig. 1a for a water content of 5 wt %. A-check as to the validity of the approach may be made by comparing the composition of gases released from the “popping” rock recovered from a depth of 3770 mt along the mid-Atlantic ridge (Javoy and Pineau 1991). Its reported \esicularity of 17% (Ry = 0.2) places it, together with a pressure of 380 bar, in Fig. la at the position of point PR. The total CO; content of the rock had been estimated to be of the order of 14.000 mg/kg at an Hz0/CO> ratio of 1.07 Point PR lies somewhat above the region expected for the reported composi lion, pressure, oF vesicularity, possibly reflecting suppression of full equili- brium vesiculation due to rapid quenching of the lavas. Otherwise the228 WA. Giggenbach agreement between reported degassing conditions and those suggested by Fig. | is excellent. Accepting vapors as discharged from volcanoes during fumarolic and eruptive periods to have started to form already at great depths, degassing is 2 continuous process and the relative proportions of volatiles exsolved and re- maining in the melt continually change with depth. The proportions of a given volatile remaining in the melt as a function of Ry, the steadily increasing vo: ume ratio of vapor to melt, are shown in Fig 1b. In analogy to CO», the solubilities of the nonreactive noble gases can be expected to closely obey Henry's law and to be similar for all three major types of magma, OF che three sets of data on the solubility of mobile gases in basaltic melts considered here, the solubilities reported by Lux (1987) correspond to low values of Q,, those by Hiyagon and Ozima (1986) to highest values. Those proposed by Jambon etal (1986) fall in between and form the basis for the Q, values used in constructing the curves shown in Fig. Ib. They are 140 for He, 310 for Ne, 1300 for Ar, 2600 for Kr and 3600 for Xe. 2.4 Vapor-Melt Distribution of S- and Cl-Cont ining Species ‘The solubility behavior of the S-gases and HCI is greatly complicated by the formation of nonvolatile solids such as sulfides and sulfate (Carroll and Ru- therford 1985; Luh 1990) or metal chlorides (Shinohara et al. 1989). The solubility of § and C_ in silicate melts has been investigated on many occasions, ‘but usually only at high pressures. In the case of the S-gases, the solubility and relative proportions of the major species SO> and HS are a strong function of the composition and redox state of the melt (Haughton et al. 1974; Wallace and Carmichael 1992) rendering derivation of simple, generally applicable dis tribution coefficients difficult. In the case of Cl, by far the predominating species present in high-density fluid phases (Kilinc and Burnham 1972; Shi- nohara etal. 1989; Webster, 1992) are alkali chlorides and not HCI. the species present in the low-pressure vapors released during shallow degassing. According to Eq. (I), and if Cl obeyed Henry's law, the coefficient De, should show a positive pressure dependence, contrary to the findings by Shi- nohara et al. (1989) who ascribe a strong negative pressure dependence to the predominant partitioning of alkali chlorides at high pressures into increasingly dense vapors. Values of the vapor-melt distribution coefficient Qe, may be obtained by use of Eq. (1) and the values of Devas reported by Shinohara etal (1989) for a rhyolitic meit at 810 °C and pressures from 0.6 to 6.0 kbar. The data of Shinohara etal. (1989) correspond, at a pressure of | kbar, to values of of Day of 0.10 oF Ocr of 0.88. At 2 kbar, the value of Dey drops to 0.02 corresponding to an increase in Qcy to 8.8 and in increase in the solubility of alkali chlorides inthe vapor by a factor often. The fractions of Cl remaining in the melt, as calculated by use ofthese values of Qci and Eq. (4) are represented in Fig, 1b by the dashed lines. They suggest that the volatility of C) becomes, at Fe ‘Chemica! Composition of Voleanic Gases 29 high pressures, comparable to that of water and that alkali chlorides become major constituents of any “vapor” phase and, therefore, may contribute con- {GWerably to total vapor pressure and the formation of @ separate fuid phase ‘At low pressures and low densities ofthe vapor phase, partitioning of Clin the form of metal chlorides becomes less important and the behavior of Cl may be described in terms of that of HCI alone. The experiments probably most closely representative of shallow degassing are those of Iwasaki and Katsura (1967) who measured the solubility of HCl in voleanic rock mefts at 1200 °C fend ateospheric pressure. Their data correspond t0 values of Kycusers of 14 500 mg/kg bar in basalt, 4000 mgjkg bar in andesite and 570 mg/kg bar ina liparitic melt, or values of Quci of 0.008, 0.03 and 0. respectively. The values for the more siliceous melts ate of a similar magnitude as those based on the measurements of Shinohara et al. (1989) at lower pressures, but suggest a strong dependence on the composition of the mel. 25 Shallow Degassing For volcanoes in a quiet state of solfataric or fumarolic activity, it was sug- Sted that the gases ave released from the magma at shallow els from Siterancany moe ot Tes fey bing “lve nes” (Gugentach snd Sheppard 1949) Considerable proportions of volales were (ound to be 1 Tas also difsiely all ove avoeancsractre inthe form of COrensiched ‘Si pae, tepoted by Allard eta (1991) for Ena. The process esponsbe io this ype of oll tlese is ikely to correspond 0 subsolidus degasing tom slid bodies of gvs-chargad rock as sugested by Giggenbach eta (1980 for Nevado de Rote nth ase of degassing rom a eey boing ave uta, by far the major proportions ofthe vapors st litely to hae exsolved in persistent equilibrium Tee une auoctated mels during the comparatively stow nse of the bubble wet avo as of vesiulating magma from great depth, festantaneous values Stare largely governed bythe concentrations ofthe two major specis HO Sha COu High depres of turbulence and clos confinement inthe zones of syst! separation of the vapor fom the mel, beneath se sace of che olin underground aca ake, ar kel to promote tense intracion ofthe qeanke eae with the residual rt and therefore maintenance of vapor Salina to quite a shallow deh ln thi ease, the depasing procs i til Sie a be evaluated in terms ofthe continuous, But effectively single-step temoval of gas fom rising batches of magma at oso atmospheric presures ind just above solids temperate JiSjaw of he unertaimues associated with theoreti or even experimental derivation of vaporemet dsibution cosine for S- and Ch-ontining spe ie espe at low pressures, empiri or elective values might be obtained by tse the companion of volts presen in voleaie vapor discharges and ie goat lavas The best results an be exposed for gases and lavas pro200 WF. Gigzenbach duced simultaneously during mild eruptive activity allowing the collection of both representative vapor and rock samples. From 197% onwards, such condi: tions existed occasionally on White Island when fresh lava bombs of basaltic andesite composition (Houghton and Nairn 1991) were ejected from active vents, Water. Cl, and F contents, as measured by NW Dunbar and PR Kyle (pers. comm.), were found to be very similar in separated minerals and lasses, Averages of all 30 analyses are given in Table I. Sulfur contents are from Giggenbach and Sheppard (1984). Unfortunately no CO: analyses are available. The composition of the magmatic gas phase is that derived by Giggenbach and. Sheppard (1989) for samples taken from high-temperature fumaroles ‘The two sets of data, representing the concentrations cm of the species jin the melt and the mol-raction X; in the magmatie vapor phase, may be com- bined to yield values of KP, = cimlM; where P, is the pressure of last equili- brium between exsolved vapor and remaining melt, For HO. the value of K,P, is found to be 6780 (mg/kg). for S: 5555, for Ck: 133 000 and for F: 45 000. Accepting Ky.o to be 30 mgikg bar, as used above, P, becomes 225 bar. corresponding to a magmastatic depth of about 800 m. As suggested above. steady-state magma degassing is a continuous process, possibly commencin, great depth with initial exsolution of the less-soluble gases and increasing contributions from the more-soluble species as the gases approach the surface. Less soluble species are therefore more likely to attain vapor-melt equilbrium than the more soluble, late-degassing species. Water has an intermediate so- Iubility, its behavior may, to a first approximation, be considered to be re- presentative of both slightly and highly soluble species. Initially the pressure P, ‘of 225 bar may be taken to estimate values of K; for the other species, That for S of 25 mg/kg bar is considerably lower than the minimum value of about 480 mg/kg bar suggested by the data of Katsura and Nagashima (1974) This discrepancy cannot be ascribed to incomplete degassing as, in this case, equi: librium values of ¢.m and therefore KP, would be even lower. A possible explanation consists of assuming the effective pressure of last vapor-magma ‘equilibrium to have been lower for late-exsolving species than that indicated by the HzO contents. Accepting a value of 480 mg’ke bar for S. the value for P, becomes about 12 bar. suggesting that $ equilibrates to very shallow levels. some 40 to 50 m beneath the lava surface. Similar considerations may apply to Cl where a value of Ke, of about {600 mg/kg bar is suggested by an assumed pressiire of last equilibrium of 225 bar. Taking a pressure of 12 bar, a value of 11 000 mg’ke bar is obtained, well within the range of 4000 to 14 500 mg/kg bar suggested by the data of Iwasaki and Katsura for the solubility of HCl in andesitc to basaltic melts. Accepting different values of Pt for less and highly soluble species. the value of 225bar found for HzO may be used to estimate the concentration of CO; in the equilibrium melt according to eco:m = Keo:PiXcoe: It is only 11.2 mg’ke. a value well below those of 100 to 500 mg/kg reported by Anderson et al, (1989) for the Bishop rhyolite, but within the range of 10 to 25 mg/kg in melt in- clusions at Crater Lake, Oregon (Bacon et al. 1992), 4 (Chemical Composition of Volcanic Gases 231 ‘The assumption of gas release from continually replenished batches of jresh”, or rather bubble-charged magma is fully supported by the short de- gussng times of about two days and the small size of the degassing batch of magma of only some 5*10%m? within the upper parts of the White Island conduit as derived by LeCloarec et al. (1992) on the basis of the distribution of short-lived radioactive isotopes. OF the small instantaneous volume of degas- sing magma, only 28% were found to be actually made up of “deep” magma, The minimum magma volume required to sustain a steady-state discharge rate of some 350 t/d of SOs, over the likely lifetime of the White Island fumarolic system of 10 ka, was estimated to be 250 km? (Giggenbach and Sheppard 1989). Continuous extraction of volatiles from such a large volume, to be released at comparatively steady rates, points to the existence of a highly ef ficient and very deep-reaching convective system beneath White Island and probably many other andesitic volcanoes. 2.6 Effective Values of the Vapor-Melt Volume Ratio ‘According to Eq, (4), the relative proportions of the species remaining in the ‘melt and, therefore, also those exsolved from a given volcano, are largely a function of their initial concentrations c.g and of Ry, the degree of degassing and, therefore, the depth of last equilibrium between exsolved vapor and mek. During an eruption, gases can be expected to exsolve under greatly varying conditions and gas compositions may vary considerably. Unfortunately no sufficiently detailed and reliable data on the composition of eruptive gases are available to check on this conclusion, A commonly invoked “petrological”” technique to obtain information on the behavior of volatiles during an erup- tion is based on the anlysis of melt inclusions, thought to reflect preeruption gas compositions, and of matrix glasses, thought to represent the degassed eruption product. The White Island data and other recent studies, however, show clearly (Westrich and Gerlach 1992; Stix et al. 1993) that the glass in- clusions must have formed in many cases well after most of the gas had already exsolved and that the large excess of volatiles, over that possibly supplied by syneruptive igneous material during recent eruptions from El Chichon, Nevado del Ruiz and Pinatubo, is most likely to be derived from a preexisting free vapor phase, the accumulated vapor as inferred above. Accepting values of K; of 0.56 mg/kg bar for CO, 30 mg/kg bar for H:0, 480 mg/kg bar for S, 11 000 mg/kg bar for HCI and 45 000 mejkg bar for F, the behavior of magmatic volatiles, during shallow degassing, corresponds to that represented by Fig. 1b. There a vertical line is shown representing de- assing at values of R, of 1.0 corresponding to a vesicularity of 50%. In this case, virtually all the less-soluble species are carried in the vapor phase, About 70% of the water, but only about 20% of the initial § and less than 1% of the Cl, assumed to be present as HCl, have transferred to the vapor phase. De- pending on initial CO and H3O contents of the magma, a value of R, of 1.0mW, Gigenbuch would be attained for most magmas having reached depths of less than 2 km. Further expansion of the volume of vapor to a value of R, of 10 would allow Virtually all of the H:O, most of the S, but only 10% of the Cl to reach the vapor phase if assumed to partition in the form of HCI. In view of the comparatively high H;0 contents of most andesitic and rhyolitic lavas, neither steady-state nor eruptive degassing appear to be gov- cerned by high values of R,, Widely varying values of R,, 2nd the occurrence of several stages of magma degassing are suggested by highly variable volatile contents of melt inclusions and matrix glasses at Katmai (Westrich et al. 1991). but not for instance for rhyolitic lavas of the Taupo Voleanic Zone, New Zealand (Dunbar et al. 1989). Generally speaking, close correspondence of ‘melt inclusion aid matrix glass compositions would suggest prolonged coex- istence of magma aad accumulating vapor in two-phase reservoirs well before ‘an eruption, farge differences may indicate rapid rise of a magma and large- seale degassing mainly during the eruption itself. Based on the differing solubility behavior of CO, HO, S, and Cl, and the apparently complete loss of CO», partial [oss of S and high retention of Cl- and. F-containing species in the melt, effective values of R, governing degassing of andesitic magmas are likely to be of the order of unity. Production of sub- ‘marine lavas showing partial degassing of both CO; and S from Kilauea volcano was explained by Dixon et al. (1991) in terms of mixing of a relatively volatile-rich, undegassed component with a subaerially degassed, votatile-de pleted component, with subsequent degassing at the ocean floor allowing CO: contents to adjust to ambient pressures. Comparatively low values of R. also suggest that the usual assumption that low S contents of lavas are due to extensive degassing is not necessarily valid, as recently discussed by Davis etal (1991), but that they are more likely to reflect variations in the original com positions as a function of the tectonic setting of the respective systems 2.7 General Discussion ‘The above findings suggest that magma-degassing processes are likely to lead to only minor variations in the composition of gases released during steady state degassing of a volcano and. therefore, in the composition of gases ac- cessible to sampling. In other words, much of the composition of the gases released from volcanoes is determined or buffered by a, within observational periods, virtually unlimited supply of deep volatiles. The volatile contents of the convecting magma batches may vary with depth, but are likely to reflect attainment of steady-state equilibrium with the rising streams of volatiles, Volatiles present in “quenched” lavas and melt inclusions represent the com: position of “last equilibrium” of the melt with the rising vapors and not that of any source or parent magmas, however defined. Based on the behavior of volatiles, as shown in Fig. 1, and their residual contents in volcanic rocks, as given in Table 1, the main pressure-generacing Chemical Composition of Voleasic Gases 23 species in basaltic magmas is CO>, and that in more siliceous magmas is HO. Applying a fluid-chemical classification, two types of volcanic systems may be distinguished: (1) “CO, voleanoes”, largely associated with basaltic magma- tism; and (2) “water volcanoes” associated with more silicic magmatism, The question which arises is why does the mantle, the main source of basaltic magmas, still contain such high proportions of CO, when it appears to have lost much of the more soluble water? In an earlier investigation (Giggenbach 1992), andestic volcanoes were considered to be “ventholes” allowing excess subducted volatiles to be recycled to the surface, The term “venthole” may be taken to imply the presence and rise of a free volatile phase ali the way down to the zones of arc-magma generation. With increasing depth the nature of this “vapor” can be expected to change considerably and may start to resemble more a hydrous melt, rich in alkali chlorides and even silicates, as suggested by Stolper and Newman (1992). In order for a mantle melt to generate a free vapor phase at a depth of 100 km, at a likely mole ratio of H,0/CO> of 2 (I on weight basis), its minimum equilibrium water content, according to Eq, (2), would have to be of the order of 75 wt%, a value not too unreasonable. Water contents of up to 6% were reported by Sisson and Layne (1993) for high-alumina basalts and basaltic- andesite glass inclusions from four subduction-related volcanoes. AS the in- itially formed high-density vapor or hydrous melt rises, it will shed most ofits dissolved solids to resemble more and more the vapor phase actually dis- charged from andesitic volcanoes (Solovova et al. 1992), Most studies into the kinetics of bubble nucleation and growth refer to conditions during eruptions. Due to the long residence times of the melts at great depths, kinetic aspects may be negligible and vapor-liquid distribution processes are likely to approach equilibrium. Accepting the possibility of for mation of a separate vapor phase at depths of 10 km or more, the buoyancy generated by the formation of gas bubbles can be expected to lead (0 con siderable acceleration in the rise of a volatile-charged batch of magma, With decreasing depth and pressure the size of the bubbles grows, ever more ac celerating the process (Bottinga and Javoy 1990) unt the swarms of bubbles or rafts of vesiculating magma reach the top of the conduit. All the while, the \spors are likely 10 reequilibrate with the associated batches of magma with more and more of the more solubie species, especially HO, entering the vapor phase, The rapid expansion and rise of the bubbles over the upper parts ofthe conduit may impede attainment of vapor-melt equilibrium and vapor com- positions may reflect conditions at somewhat deeper levels. If the conduit is ‘pen, the gases escape freely from “boiling” lava lakes or deeper within the roleanic structure from the surfaces of subterranean lava lakes to give rise (0 persistent fumarolic activity. If the conduit is blocked, the vapors can be expected to accumulate to form an ever growing pocket of gas. Assuming the magma column to have been in isostatic equilibrium with its environment, vertical displacement of the column ‘of magma downwards will allow higher and higher pressures from the top of2M WF Giggenbach the depressed magma column to be transmitted, through the vapor pocket, to the base of the blocked conduit. On reaching a sufficient pressure 10 dislodge the obstruction, the voleano erupts with the release of large amounts of stored volatiles. In this scenario volcanic eruptions are in no way caused oF even triggered by the “injection of new miagma”, the most popular explanation invoked for the resumption of voleanic activity, but by the exact opposite. the downward displacement of magma by accumulating gas. The gas accumulation scenario also provides a ready explanation for the vast excess of 8S. as compared to the amounts of lava produced during recent eruptions (Williams et al. 1986; Stix et al. 1993). In techniques for the surveillance of dormant volcanoes, it may therefore be more prositious to look for indications of the rise of rafts of vesiculating magma or the accumulation of gas, rather than for signs of the injection of new magma. Rather than representing an essential vehicle allowing subducted volatiles to return to the surface, as suggested by Giggenbach (1992b), andesitic melts may simply be considered an almost incidental fluid medium facilitating the release of excess volatiles. Independent and free rise of volatiles provides the least contrived explanation for the production of the enormous amounts of yolatiles from many active volcanoes, as compared to the generally puny amounts of syseruptive rock, and also for the occurrence of persistently active volcanoes, such as Stromboli (Francis et al. 1993) and lava lakes. Assuming volcanic systems to have attained steady-state with respect to the supply and discharge of volatiles, the entire vapor-melt sytem is likely to follow a “boiling point versus depth curve”, with the composition of the gases tleased from the “magma geysers” at the surface corresponding closely to that added to the system at mantle depth. Summing up, the vapors released from magmas during steady-state, noneruptive devassing are likely to represent the integrated product of a pro- longed and deep-reaching process affecting large volumes of magma. Due to this large-scale buffering, the compositions of volatiles released from a magma during steady-state degassing are likely to remain comparatively unchanged. at least over conventional observational periods. Most of the variations in the composition of gases collected {rom volcanoes, with time and space. then are probably due to secondary processes, such 2s cooling and reequilibration. and interaction with rock and waters of associated hydrothermal systems. The relative contributions from these secondary, nonmagmatic processes can be expected to vary with the state of activity of a volcano. In order to be able 10 read the messages carried by the gases, itis essential to have a good under- standing of the processes affecting the compositions of the now volcanic vo- latiles in the near-surface environment ‘Chemical Composition of Volcanic Gases 235 3 Factors Affecting the Composition of Volcanic Gases 31 Techniques for the Interpretation of Volcanic Gas Compositions In the evaluation of the composition of volcanic gases, by use of thermo- dynamic techniques, generally two distinctly different approaches are taken, ‘The “global” approach is based on the assumption that the gases, as sampled, have at some stage attained internal equilibrium among one another, and preferably also with rock, and that at least some of the species are still present in proportions reflecting this state of equilibrium. A more step-by-step ap- proach is based on the comparison of simple analytical variables with com- positions expected theoretically for a number of potential reactions and processes. Early investigations using the “global” approach arc those by Ellis (1957) and Matsuo (1960), followed by computer simulations of voleanic-gas compositions by Heald et al. (1963). A further development of this approach Ted to the evaluation of “correspondence temperatures” and “restored com positions” (Gerlach 1979). These techniques are most suitable in the case of very high temperature (> 800 °C) discharges where most of the species ean be assumed to be present in Case to equilibrium amounts and where the effects of secondary processes, such as mixing of various components having equili- brated under a range of conditions, are comparatively unimportant, For low to intermediate temperature discharges, this approach becomes impractical as the effects of a range of processes and contributions from a variety nonmagmatic sources are likely to obscure magmatic signatures. Other global techniques, such as that presented by Symonds and Reed (1993), based on elaborate computer routines, are valuable in the evaluation of specific aspects of volce ric-gas chemistry, such as the deposition of solids, but require large sets of input data, frequently not available, and are susceptible to cross-contamination from spurious data, ‘The other approach allows measured data to “speak for themselves" without any prior assumptions as to the degree of attainment of equilibrium among any of the species and is applicabie 1a only partial analyses (Giggen- bach 1987). Rather than intertwining analytical and thermodynamic. in- formation, the approach keeps measured and theoretical aspects strictly separate and allows attainment of, or deviations from, equitibrium to be judged ona species-by-species, sample-by-sample basis. Both, a more global approach and the species-by-species approach have recently been used successfully by Chiodini et al. (1993) in the evaluation of the reactions governing the com- position of fumarole discharges from Vulcano Island, Italy. The species-by- species approach is largely based on the graphical evaluation of the position of data points with respect to a number of isomolar coexistence lines representing potential reactions governing the most important variable involved in voleanic- 2s reactions, the redox potential. Majer processes possibly controlling the redox potential of volcanic gases are therefore discussed firstft 3.2 Major Processes Controling Redox Potential a Use eens one muerte wal f oe ears S59 enees eal é ‘The elceney of geochemical buer system is determined both by quanti o|gize g298 239 22925 22] 8 tative and thermodynamic factors. In order to form the basis for an effective g|s= % redox buffer, a chemical element is required to occur, over the range of che- | lize eng 2e2 2 tical and physical conditions of interest, in two oxidation states with similar g ziss 7 thermodynamic stabilities and sufciently fast kinetic responses. In order to be : -|28E% E288 249 E2824 BE able to dominate its chemical environment, the components making up a buffer 2/B8o¢ 8528 BORE system have (0 be present and available in quantitatively dominating propor. . See nk oT sions : é ? ‘The most important elements making up a volcanic vapor phase are H, C, > neelareceectecneate © and S. Hydrogen is largely present as H;0 with the other H-eontaining vy] .|€"82 2228 “2° 8°32 58 species, such as Hs, CH, or NH; present in only minor amounts. Of the C- af t= ae containing species, CO, predominates by far; the other members of possible g) [3.3 883 888 egeage sy buffer pairs, CHy and CO, form over temperatures from 200 to 1000 °C only in 2|* = = “ ‘minor proportions and are therefore unsuitable to act as voleanic-gas buffers. 2] |,58, sea cae ge Of the major species, only the S species SO, and HyS occur in comparable syz] oo ees con os amounts in voleanic gases and therefore represent the mos likely candidates to Helis. 29.99: age Be act as a major redox buffer. | #[2|'5 aie - z The other major component ofa volcanic system is the associated motten | ZJe| 288 e229, -8, Engg ex oF solid rock, There the major potential redox-controlling element is Fe in its \ g[z) VSR RS Dans two oxidation states. Depending on residence times and general conditions of || ea eee ce nanan es fluid-rock interaction, either gas or rock bulfer may affect gas compositions. A. Glo/2°88 8882 °§, 8888 valid decision as to the actually active buffer is only possible by comparing | = Es ‘measured gas compositions with those expected for buffering of redox potential { a|8288 28288 £8: E8!e7s by any of these systems. A species responding quickly to changes in redox | g/ SEG" S258 SR! BER E potential was found to be H: (Giggenbach 1987). In combination with HO to 23 gees ge aasee form the variable Ry = log(fH2/fH20), it represents the most useful probe gl Ese 2882 92. F2S55 into the redox state of voleanic gases. For equilibration in a vapor phase, and | Bye) e* - cen aK pressures elose to atmospheric, the fugacity rato, / fase may be teal 4] | S882 £228 28888 to the measured concentration ratio zu,/x,0 and analytical data may be aig 25 Bo aes acl compared directly with Ry values expected for redox control by @ number of i potential bufers. s|,|g888 2888 2888 In Fig. 2, values of Ru = log(Su./Su1,0) © log(xn,/191.0) are plotted ver- Ei/2/2825 $885 So585 sus farearoe outlet temperatures forthe volcan gas dachargos we lsted ta , gle Table 2, Potential buffer lines were drawn by use of the temperature depen | 3|£)8988 £882 £8, g2282 88 dence of the equilibrium constsnts, as used earlier (Giggenbach 1987), and i gyry= C= z listed in Table 3. For equilibration at atmospheric pressure, the term : 2 . 4 Fw = logo i8 set equal to 2er0, for equilibration at pressures controlled by Be ied z the presenice of saline brines, the expression for Fw as listed in Table 3 is 3 2 E288 233 | z inserted for temperatures up to 600 °C. The temperature dependence of Fe tlelgsze giz2 2208 z represents an adequate compromise and is also close to that reported by &| $2225 £22 523 e Chiodini et al. (1993) for NaC! saturated brines. al? ead eee # In Fig. 2, data points for high-temperature discharges quite obviously a] |sesz S822 vee gaze follow a trend most closely represented by that for Ry control by the Hs8/SO- ee eee rae ~28 WA Gisgenbuch ig 2, Values of Ri = lost fu,/fud lopisis/rw-o} a5 8 fection of tem erature for samples ited in Table 2 together with fe representing poten tht redo buffs For tymbulr a Table 2 gas buffer. For the CH./CO: and Hs$/SO; systems two lines are shown, one representing variations in Ry as a function of temperature at water vapor pressures of | bar, marked “surface”, the other for control of fro BY the presence of increasingly saline brines, marked “brine”. Most of the data points closely follow the surface line suggesting that Ri values adjust very rapidly to {he redox potential generated by the H$/SO3 gas butler at conditions very close to the surface. At very high temperatures, Ry Values appear to be gov- cerned by the fayalite-magnetite-quartz (FMQ) buffer, as evpected (Carmichael and Ghiorso 1986). Conversely, the fact that Ry values remain close to the Hs$ SO: isomolar coexistence line also suggests that the S gases are oniy slightly affected by any of the other butters, includiag those involving components ofthe tock, such as fayalite, magnetite, pyrthotite or pyrite. Some of the lower-temperature dis- charges deviate from the S-gas buffer line to higher values with a tendency to approach the FeO/FeO1.$ rock-buffer line. In contrast to high-temperature volcanic systems, this buffer was found to exert control of redox conditions in lower temperature, such as geothermal, systems (Giggenbach 1993), Deviations to higher Ry values for the lower-temperature voleaaic gases may eelect in teraction of the gases in 2 hrdrothermal-reaction environment or simply quenching of higher-temperature compositions. Summing up. instanteneous redox conditions in high-temperature volcanic ‘gases are likely to be governed by the presence of HS and 80> in comparable (Chemical Composition of Volsnic Gater 238 “Toble3. Temperature dependence of equilbviam constants ied is thi sty, Tin K. Fupasiy ration ae assumed (0 be equal lo concentration rater, ©€ Rasta usin ils une ed on = lotto) = Ran 3K Fe oT Lee =fostlcotnorfeosi,) = Roc ~ Ro 29- n8/t Lew =lel feo H/fomtyg) © “Rc AR Fhe =961-887T Rune =lalseapeie ati fan tvo 190+ 88/7 Ruan =ltl amie /Aeoio -460 -27077 Rasnt = ltl tims onliio 200 Fee __=ealfwot(rom 100 to eo") = 490= 122077 amounts, Only Jower-temperature gases deviate from this tend and appear to show a tendency to approach redox equilibrium with the rock matrix. None of the redox pairs involving the C gases CO, CHs and CO appears to exert any redox control. Because ofthe generally vey low contents of both CH« and CO, these two gases may provide the basis for additional indicators of conditions and processes affecting the Fsing gases. 33 Carbon Gases The predominant C-containing species in volcanic gases is CO>. Of the major species it is the least soluble in magmatic melts and is, according t0 Fig. 1, virtually completely exsolved from its parent magmas by the time they reach the surface. In volcanic systems, there are no major sources and sinks for CO> and COz ean be considered 2 fully “conservative” component, Any variations in the relecive proportions of a species with respect to COp are likely to reflect the occurrence of processes affecting it rather than COs. A check as to the validity of this argument is carried out by use of Fig. 3. Here relative CO>, 1130, and He contents are plowed representing all available data for White Island together with points representing all available analyses for Neauruhoe volcano, New Zealand, and a set of data recently presented by Chiodini et al (1993) for Vuteano island, Ttaly. For all three systems, variations in the position of data points are largely caused by differences in water contents at close to uniform CO>He ratios Uniform CO,/He ratios in magmatic gases are predominantly the result of the very similar solubility behavior of CO; and He in magmatic melts (Fig. 1a), The preservation of uniform ratios in fumarolic gases, however. shows that CO> behaves chemically similar to the inert noble gas He also under more shallow, voleanic conditions. The pronounced variations in relative Hz0 contents may either be due to variations in the degrée of degassing of the| 240 WF. Giggenbuch aban EW Hoe wabes ohm mn Pa Fig. 3 Relative COs, HO. and He contents of all sumpis avaiable for White Islnd ent [Nguuruhoe, New Zealand topeher with those reported by Chon eta, (1993) foe Valcane, tls and tose ofthe samples sted in Table 2 magma batches feeding the fumaroles, of to addition from shallow meteoric sources, At White Island, highest 1;0/CO; ratios are observed for lower- temperature (<200 °C) vents suggesting that most of the excess HO there is due to addition of groundwater. The reverse process, condensation of mag- matic water on concact with cooler groundwater, is likely to be responsible for the very low H0 contents for some other low-temperature discharges, espe- cially those from Ngauruhoe. The scatter is lower for the high-temperature (400 °C) discharges, but with H0/CO; ratios varying from about 8 to 40, it is still considerable, Some of this variability may indeed be due to variations in the vapor/melt ratio R, and, therefore, to the degree of degassing of magma batches feeding the vents. In this case, vapors produced at low values of R,. or early on during the degassing process, would be be rich in CO. and He, “Late vapors can be expected to be enriched in HzO, but even more so in the more soluble species containing $ and Cl, as discussed farther below. The likely composition of the volatiles present in White Island lavas is represented in Fig. 3 by the point marked “Wl-rock.” Its position reflects the high amounts of water observed in erupted lavas (Table 1). Of the two remaining C-containing species, CO was found to reequilib: to changes in redox potential with similar speed as H2. It may theref emicl Compoxitin of Vane Gases 281 combined with CO; to form the isomolar, redox sensitive variable xco/xco. It mnay in turn be combined with Ry t0 provide @ geoindicator allowing both temperature and redox potential prevailing during the equilibration of Hz and CO to be obtained on the basis of purely analytical information, as shown in Fig. 4. The figure is readily constructed by use of the reactions, and their temperature dependences, given in Table 3 Figure 4 closely resembles Fig. 2, but rather than having to rely on mea sured outlet of other temperatures, the position of data points is determined purely by analytical results allowing an immediate, internally consistent as fessment to be made of the nature of the main bufer systems controlling gas compositions and of temperatures of last equilibrium, not necessarily those measured at the outlet. Again, the buffer controling volcanic gas compositions within the system HO, COs, SOn, HS, Ha, and CO is that based on the § Species, at close to atmospheric pressure. Minor deviations of data points from the isomolar buffer line ae likely to be due to HyS/SO fugacity ratios differing to some degree from the valve of unity, as represented by the line, but may also reflect some loss of CO due to conversion to formate in the highly alkaline [NaOH solutions used to collect most ofthe samples (Giggenbach and Matsuo 1991) The two data points representing low-temperature (< 100 °C) fumsaroles on Pichincha (PC) and El Ruiz (ER), far from the gas buffer line in Fig. 3, lie now close to it suggesting deeper equilibration temperatures of close to 300 and 400 *C respectively, The preservation of higher-temperature compositions in a # 2 So ES Lo 6. y Fe tha! Fig 4, Pot of Ry = login fino) s8t866 Ree = Hel sco/eo:) for gat samples hated in Table 2 logetier with equibelm isoheems and ier wepresenting potemiat redox burs282 WF Giggentaeh these two low-temperature discharges is likely to be due 10 rapid quenching through cooling on contact with shallow groundwaters. Because of the fast speed with which both H; and CO respond to decreasing temperatures and pressures, Hz and CO gas equilibration temperatures are generally close to measured values. A species equilibrating much more slowly was found to be CHa. Again, an aitempt is made to identify the chemical system controling its concentrations by plotting a measurable, isomolar concentration quotient, xcy,/co., Versus ‘measured furnarolic outlet temperatures (Fig. 5). In contrast to Figs. 2 and 4, there appeass to be no obvious correlation with any of the trends representing ‘the most likely redox buffer systems. For all gas discharges, CHa/CO> ratios ‘are much too high to be compatible with buffering by the Hy5/SO: gas buffer For some discharges (WI, SH, NG, VU) data points lie close to the FeO/FeO, ; coexistence line representing the buffer system found to control redox condi. ‘ions in fully-equilibrated hydrothermal systems (Giggenbach 1993). For other discharges, low CH/COs ratios (PP, PC, ER, SN) may reflect the effects of equilibration in a more oxidised environment or, in the ease of ER and PC, of rapid quenching as suggested already by Fig. 2. A potential buffer, at more reduced Ry values is represented by the line marked “caleite-anhydrite”, It * ® ra 2 a Fig. 5. Plot of Rye = lost, /aco,) vers measured uk : tc = los /sco:) vers ‘outlet temperatures For samples listed in “able 2 together with les for potential rox balers He (Chemical Composition of Volanic Gases 242 corresponds to the coexistence of these two minerals in an environment gov- cerned by the reaction CaCO + HS + CH, + SHO = CaSO, + 200: + 8H; 6 for commonly observed values of CO2/H;§ of 100 (Giggenbach 1993). Another potentially important buffer, possibly controlling redox conditions. under magmatic conditions, is represented by the line marked magma-gas equili- brium, according to the reaction 2FexOu + 3Si02 + CO, + 2HyS = 3FeySi0y + CHy +2502 ” Ie corresponds to simultaneous contsal of redox conditions by the H,S/SO> gas and rock bufer, represented under magmatic canditions by fayalite and magnetite (FMQ). According to reaction (7), CHy/CO, ratios should decrease rapidly with decreasing water vapor pressure and temperature. Assuming the comparatively high CHs/CO, ratios as observed for most discharges to be controlled by reaction (7), they would suggest formation of CH, at very high pressures and temperatures, followed by varying degrees of quenching ‘On the basis of patterns exhibited by a large number of fumarolic dis- charges from White Island, Giggenbach (1987) concluded that CH, is largely formed in the hydrothermal envelope surrounding the fumarolic vent system In this cas, the position of the data points for the three Japanese discharges would have to be explained in terms of “reverse quenching” consisting of entrainment of cooler, hydrothermal vapors into a hotter, rapidly rising vol canie gas stream. Another possibility is complete nonequilibrium formation of CCH through thermal decomposition of erustal organic material. At present it is not possible to resolve this uncertainty in the origin of CH. In contrast to hydrothermal systems, it appears premature to use the concentrations of CH 4 the basis of a peoindicator in voleanic systems ‘A species responding with similar speed (o changes in temperature, pres- sure and redox conditions and, therefore, being largely formed in the hydro- thermal envelope surrounding a voleani€ vent system, was found to be NHy (Giggenbach 1987). In some voleanic-bydrothermal discharges, such as Kete- tahi, New Zealand (Giggenbach, unpubl. results), its concentrations may be Birk enough to neutralise the acidity generated by the sulfur gases and for the NH ion to become the most important cation of associated waters. The analytical determination of NH requites special precautions to avoid loss from the generally highly alkaline samples as collected; itis also determined only sporadically in volcanic gases (Giggenbach and Matsuo 1991). Together with the problem of defining @ suitable isomolar concentration quotient (Giggenbach 1987), NH has not yet found widespread application as a seoindicaror. Generally speaking, however, high NH contents point to ex- tensive interaction of voleanic fluids with local rock under hydrothermal conditions3.4 Sulfur Gases The next most important constituents of voleanic-gus discharges are the two species SO; and HyS {Table 2). As suggested by their ability 1o control redox conditions to low temperacures and pressures (Fig. 4), equilibration bewoen the two species is likely to be rapid. A possible indication of primary processes controlling their contents in volcanic gas discharges may therefore be carried out by treating the two species as rapidly interconvertible manifestations of one component. S In Fig, 6, variations in toxal $ contents, S, are shown as a function of H30 and CO, contents. As pointed out above, CO2 can be considered to be a conservative component. Once released from the magma itis likely to remain unaffected by secondary reactions. similar to a noble gs, such as He (Fig. 2) On the other hand, $ was found to be considerably more soluble in magmatic melts than either CO; or HO (Fig. 1). If, as suggested above, variations in H,O/CO; ratios are due to different degrees of magma degassing, the effects of Fig. 6. Relative COs, 1,0, and $, coments ofall samples available foe White Island and Nevur toe, New Zend, and thse eeportat bs Chodin eta. (1993) fr Valeo, together with hose ot the samples listed ia Table 2 Liner are shown representing the eormpostion of ‘pitas por Formed feom the “rginal magma of the composition of the residull mel and of ‘vapor forming from a residual mel, al ay a fanetion of R,. For smbol ee Fig. 3 ondary ‘Chemical Composition of Volcanic Gases 285 this process should be even more pronounced on relative $ and CO; contents. As discussed above, likely values of R, governing degassing at White Island are in the vicinity of unity. Accepting this value, relative concentrations é9 of HO, CO>, and S in the “original magma” can be calculated from measured gas concentrations ¢,, by use of the relation Gio = Gal + 1/0) 8) which in turn may be used to calculate exsolved vapor compositions as a function of Ry. The resulting line is marked in Fig. 6 as “primary vapor". The line was calculated by use of very specilic conditions; however, independent of assumptions, any such degassing curve would follow a similar crescent-shaped trend linking the corner representing the least soluble (CO3) to that for the ‘most soluble component (S). ‘The observed trend in relative CO, HO and $ contents in no way follows that suggested by any degassing curve and rather suggests that much of the variabilities observed for all three systems shown, White Island, Neauruhoe and Vulcano, are largely due to secondary variations in water content. Figur: 6 also contains a line representing the composition of the “residual melt” as a function of Ry. As expected, relative gas contents of the magma move, with increasing degree of degassing farther into the S corner. Gerlach (1986) proposed a two-step degassing process to explain the composition of gases at Kilauea voleano. Assuming some of the vapors at White Island to form by similar secondary degassing, from a magma having already generated the primary vapor at Ry values of unity, a curve is shown representing the compositions of such “secondary vapors", again asa function of R,. For very low values of R,, their compositions correspond to that in equilibrium withthe residual melt, which of course is identical with that of the primary vapor at a R, value of unity. Even at very low values of R, of 0.01, CO; contents ofthe secondary vapors become already very low and secondary vapors formed at significant values of R, become effectively COs free. Such secondary degassing may have affected some of the White Island samples and may explain the comparatively low CO2 contents of gases from Satsuma Iwojima (SI), Mutnovsky (MT) and Klyuchevskoy (KL), but appears to be of ‘only minor importance generally ‘On the basis of Fig. 6, most ofthe variations in relative HO, CO, and $ contents have to be ascrited to secondary, nonmagmatic processes. Because of their quite different chemical properties, SO, and HaS can be expected to respond to these secondary processes in different ways causing distint shifts in their relative concentrations. Using CO; as a reference, potential processes affecting SO; and HS content are evaluated in Fig. 7. It should be kept in rind that the SO and HS contents for the White Island samples are those evaluated indirectly, by use of measured total sulfur contents and the average ‘oxidation state of sulfur as described by Giggenbach (1975). In order to ke discussions simple, data points are shown only for samples from the highe285 W.F. Giggeauch temperature fumarole on Donald Mound, White Island, and for a limited period of rapid change in both outlet temperatues and discharge chemistry from 1974 to 1978. in an earlier account (Giggenbach 1975), these changes were ascribed largely to absorption and desorption process of reactiv into hydrothermal brines surrounding the vent system. Several major processes may be identified in Fig. 7. Deposition and re- volatilisation of elemental sulfur is represented by (ines radiating trom point (0), as shown by a dashed line passing through sample point 3 and marked “loss-eain of elemental sulfur”. The effects of disproportionation of SO; ac- cording to the reactions 4502 + 4:0 = 3H.S0, + HS 9) and 3802 + 2H,0 = 2H2$0,+S(0) , (10) 885 followed by dissolution of the resulting sulfuric acid into associated brines may be evaluated by use of fines radiating from a point outside the triangle of Fig. 7. Relative CO, SOs. td HS contents of sisted White Island and Vues “The diagram allows the evaluation of mixing proceste ivolsng S Spe dation states from 2 to +6 28 discharges, Chenieat Composition of Volcanic Gases 247 Fig. 7, marked " H,SO,, and shown for sample 2 by the line marked “gain-loss of sufate”. Oxidation of H,S to SOa, e.g, by air, and reduction of SO; to HaS, e.g, by excess, higher temperature Hz at a constant C/S ratio, is indicated by horizontal trajectories, Starting with sample 1, the shift to sample 2 corresponds closely to the addition of earlier deposited elemental S, remobilised due to heating from 122 to 400°C. Further heating to 560 °C moves compositions back to that typical of the primary magmatic gas (3), as derived on the basis of later, more stable discharges (Giggenbach and Sheppard 999). Subsequent cooling to 105 °C is accompanied by renewed deposition of elemental S as samples 4 and 5 move away from the position representing elemental S (S(0)). The start of a further heating cycle is marked by @ move of gas compositions to lower C/S ratios, but this time towards the H,S corner. Formation of H2S according to reactions 9} and (10) would lead to an increase in the C/S ratio. The observed decrease in C} 'S may be explained in terms of the addition of HS generated through inter action of high-temperature fluids with earlier deposited elemental S according, (o the reaction 4S(0) + 41,0 = HySOs + 3H, (qu) with the sulfuric acid being absorbed into brines. A further increase in tem- perature to 540 °C brings compositions back to those typical of the primary as, For Vulcano, a number of data points are plotted representing changes in ‘gas compositions from 1982 to 1991 (Giggenbach and Matsuo 1991; Tedesco etal, 1991; Chiodini et al. 1993), The major trend corresponds to the addition of @ highiy reduced vapor phase reaching its peak in 1989. According to Chiodini et al. (1993), much of the variability in the composition of Vuleano 228¢s is due to ming of an oxidised, high-C/S magmatic and a more reduce, low-C/S hydrothermal endmember component. The close position of sample 6 from White Island and the 1989 point for Vuleano suggest that both are af fected by the addition of components from an associated hydrothermal system. Figure 7 shows, that by plotting chronological sequences of relative COs, $0; atid HaS contents, vaiuable information can be extracted on likely processes affecting a rising volcanic gas and, therefore, on changes in the state of activity of a volcano. 35 Halogens The remaining major reactive species in volcanic gases are the halogens, largely present in these low-pressure discharges in the form of hydrogen halides. OF these, HCI and HF predominate, The behavior of HCI in voleanic-vapor dis- charges is investigated in Fig. 8 on the basis of relative CO>, S, and Cl contents, CO; represents again the magmatic reference gas. Any variations in HCCO3May WP, Giggenbuch, ospengs Ovens 200° © m0-mare vents e008 Fig. 8. Relative COs, §, and HCI contents of all available samples from White Island and [Ngauruhos, New Zealand, and of samples reported by Chiodin eta (1993) for Vuleano aly. A line is given representing the compostion of "prinaty vapors’ from the “erignal White Inland ‘magma’ asa function of R,, gether with date pons forthe samples listed in Table 2 ratios, therefore, may again be due to different degrees of degassing of the paren’ magma of to secondary processes such as absorption in, and desorption from associated brines As discussed above, the solubility of HCI in magmatic melts is very high In order to check on possible solublity-induced variations, Fig. 8 contains @ trend line again representing degassing from the “original White Island mae mma", asa function of R,. The composition of this parent magma was asain calculated by use of Eq. (8) and the typical composition of the White Island magmatic-gas discharge (Giggenbach and Sheppard 1989), and assuring a value of Ry of unity, The degassing line again links the least (COs) and most (HC) soluble species, but agsin it, and any other degassing curve do not reproduce trends observed for White Island. The changes in relative COs. S, and HCI contents of high-temperature (> 400 °C) discharges correspond lar ely tothe addition and removal of HCI at comparatively uniform ratios ofS, COs of 0.1 t0 0.5. For low-temperature discharges (<200 °C). HCI contents Fe ‘Chemical Composition of Volcanic Gases 249. are generally low, except for some discharges affected by rapid heating (Gig- genbach 1975). During prolonged cooling periods, even 3 contents decrease due to deposition of elemental sulfur or absorption of H13S0,, together with HCl into invading brines leading to a distinct L-shape in the pattern of relative CO», Sand HCI contents ‘The L-shape is repeated in the data points for Ngauruhoe, from high- temperature, hish-HCl discharges over cooler low-HCl, high-S discharges to the present low-temperature, low-S,, low-HCI vapors. The overall trend for the ‘Vulcano vapors could be interpreted in terms of varying degrees of magma degassing. As pointed out by Chiodini et al, (1993), however, much of the variability there is likely to be due to mixing of COp-rich, magmatic and comparatively Cl+ich, hydrothermal components. The next most important volcanic halogen is F, present in the gas dis- charges largely as HF but complexing to significant degrees with other elements such as Al, C, and Si Symonds et al, 1988; Symonds and Reed 1993), Ratios of F/C1 may provide a useful indicator to monitor volcanie activity; the major distinguishing property of HF is its high solubility in water and its ready adsorption onto solids, such as ash particles or vegetation. The latter, ad- sorption onto vegetation, may give rise to serious problems in agricultural areas affected by prolonged exposure to volcanic fumes (Oskarsson_ 1980; Giggenbach, 1988); the former, adsorption onto ash particles, may provide the basis for a technique to monitor otherwise inataessible volcanic vapors during explosive eruptions. The technique is based on leaching fresh volcanic-ash samples with distilled water. The three components generally present and readily analysed are sulfate, Cl, and F. Figure 9 shows relative SOs, Ci, and F contents in ash leachates from a number of erupting volcanoes. They appear to fall into two groups: comparatively low F leachates from Mt. St Helens (Nehring and Johnston 1981) and Central America (Taylor and Stoiber 1973), and high F/Cl samples from Hekla (Thorainsson and Sigvaldason 1972), Lonquimay (Giggenbach and Naranjo 1989) and Yasur (Giggenbach, 1988). ‘A possible explanation of the observed variations is provided by the widely differing solubilities of the three gases in magmatic melts (Fig. 1). As pre- viously, @ line is shown representing the composition of gases exsolved from ‘typical andesitic magma (M) as a function of R. At low values of Ry, the easolved gases are comparatively which, after atmospheric oxidation, may give rise to the formation of sulfate-rich vapors possibly represented by the Mt. St Helens and Central American leachates. The calculated trend, hhowever, is not able to explain the composition of the extremely F-rich sam- piles. Fig. 9 also contains data points for the average compositions of basaltic (B) and andesitic (A) rocks as given in Table 1. Degassing of basaltic rocks should always give rise to comparatively F-poor vapors, a finding not entirely compatible with the composition of the Hekla leachates. Relative SO,, Cl and F in these leachates, therefore, are likely to be subject to secondary processes, such as relative degrees of oxidation ofS species to S04, the mode of formation of acid droplets, which in tun would be largely a function of the moisture290 WE. Giggenbech $0, cl OF we aS or MD Fig. 9. Relative SOs Ch, and F contets of ash leachate, o8 weight basi, together with 2 ie representing the composition of primary vapors’ formed ftom an oignalandeitic magma’ (4) 254 function of Ry. F/CL ratios along the CH-F axis ate on molar bass Points A and B represen the Average compositions of andesite und basic lavas as lned in Table I; points M and represent the assumed comperition a typical andesite magma and tha ofthe eidual mek remaining alter Single-sep closedsstem degatting espectvely content and possibly the temperature of the eruption cloud. While results for {individual eruptions appear to be internally consistent, additional information is required to explain the large variations from volcano to volcano, ‘The remaining halogens, Br and I, are usually only analysed for specific purposes, such as the delineation of possible admixture of seawater (Sugiura and Mizutani 1963), A species showing considerable promise as an indicator of the occurrence of high-temperature vapor-liquid separation processes is boron, but little is known ofits volcanological behavior (Martini and Cellini Legittimo 1984), 4 a Seen ESSE SO UESEOUSESTSTSTOOSES 3.6 Nonreactive Gases ‘The only remaining species in Table 2 to be considered are No and Ar; He has already been used earlier in Fig, 3 to check on the behavior of COz and HO. It js again used in Fig. Yas pact of a classification scheme of terrestrial gases into ‘meteoric, magmatic, and crustal components. OF these the meteoric component is represented by two points corresponding to the composition of air and air- saturated water (asw), or groundwater. Volcanic gases generally have much lower relative Ar contents and data points plot along the opposite axis, de- pending of tectonic setting, either close to the He corner or at higher Nz contents, Data points representing the fst case are gases from fumaroles on Sicrra Negra (SN) and Aleedo (AL) voleanoes on the Galapagos hot spot and from Halemaumau crater (HA) on Hawaii (Giggenbach, unpubl. results). PR represents the composition of free gases in “popping” rocks recovered from the mid-Atlantic ridges (Javoy and Pineau 1991). The close positions of data points for sainples from a mid-ocean ridge and from hotspots suggests that gases 05 fa oh Hepes ake on Fig. 10, Relative Ny, He, and Ar contents of samples Usted in Table 2, together with those of potential meteoric. “andes. male and erostaendmember components2 WLP, Giggenbach associated with non-are basaltic magmatism generally have low Nz He and No! Ar ratios (Matsuo et al. 1978; Giggenbach 19922), but high He/Ar ratios close (o the point marked mantle. Another group of gases also plots close to the He corner, those associated with natural-gas discharges or guses from stable continental settings generally In these eases high He contents are duc to the addition of ‘He from radioactive decay of U and Th in the crust, For New Zealand, the radioactive He/Ar production ratio was found to be 0.60.3 (Giggenbach et al. 1993). The corre sponding point is shown in Fig. 10 and marked “crust”. A distinction between the two types of high-He gases is possible on the basis of their Ne isotopic compositions. The "He/‘He ratio of mantle-derived gases, normalised to air, is generally close to 8 Ry in gases associated with mid-ocean spreading centers, but may exceed 20 Ra at some hot-spots. Those of crustal origin generally have ratios <0.1 Ry, For gases from volcanic systems along convergent plate boundaries, re- lative Nz contents are usually vell above those in air and data points plot Within the fikd marked “andesite”. They show considerable variability pos- sibly due to differences in the proportions of marine sediments reaching the zones of arc-magma generation (Kita et al. 1993), Nevertheless, retaive No, He, and Ar contents, together with the isotopic composition of He provide & reliable technique for the classification of most voleanic-gas discharges. This technique, and all the others discussed earlier, rely on the availability of sufficiently complete and accurate sets of data, obtained only on well-col- lected samples, analysed by reliable methods. Over the years. a large number of procedures for the collection and analysis of volcanic-gas samples have been used. A comparatively simple technique was proposed back in 1975 (Giggen- bach 1975) It is based essentially on the use of evacuated flasks (300 ml) containing an alkali hydroxide solution, usually 50 ml of 5 N NaOH, for the absorption of the acid gases, such as CO:, SOs, HsS and HCI. They are ana. lysed by wet chemical techniques. The nonabsorbed gases, such as He, Ho, Ar. Nz, CO, and CH, remain in the gas space of the bottle and are analysed by gas chromatography. The technique is suitable for the callection ofa wide range of discharges from low-temperature bubbling pools, over geothermal wells, 0 high-temperature volceni¢ vents. Over the years it has been tested on many ‘occasions, Possible problems with this technique and ways to overcome them have been discussed by Giggenbach and Matsuo (1991). References Allard P. Carbonele 3, Dulevie D. LeBroec JL, Morel F. Robe MC. Saurenas IM, Falvte Pierre: R, MartinD, Sabrovs JC, Zeitwoog P (1931) Eruptve and difusve emission of CO, Grom Mount ina, Rate 387-391 Chemica! Composition of Volcanic Gases 253 [Anderson AT, Newsman S, Wilms SN, Druitt TH, Sirius C, Stipe (1989) H,0, CO, Clana us it Pinan and ashflow Bishop thyolite. Geology 17: 221-225, Bacon €R, Nexmun S,Stolper £ (1992) Water, CO;.Cl and F in melt Inclusions in phenocrysts from three Holocene explosive eruptions crater Lake, Oregon. Amt Mineral 7: 1021-1030 Blank JG, Stolper EM, Carrll MR (1998) Solubilites of carbon dione avd watt in hyo melt ‘¢ 830 °C and 750 bar, Earth Planet Sei Lett 119: 27-36 Bottings Y,4380¥ M (1990) MORB degassing: Bubble growth and ascent. Chem Geol 8: 285-270 Byers CD, Garcia MO, Muenow DW (1985) Volatiles a basalt lases om the Eas Pacie Rie 112° N: implications for MORB sources and submarine lva fow morphology Earth Planet, Sei Lew 9.9.20 CCannichaetISE, Ghiorso MS (1986) Oridation-redution relations in basic magma: a case for homogeneous equilibria. Earth Planet Sci Let 78: 200-210 CCawoll MR, Rutherford MJ (1985) Sulde and sulfate saturation in hydrous silicate els. J Geopinys Res SNC: 601-612 CChiodini G,CioniR, Marin (1993) Reactions gnerning the chemistry of rater fumaroe from Vuleano Island, tly, and implication for volcanic sureilans®. App Geochem 8:357-37L Cole JW (1999) Structural control and origin of volcanism in the Taupo volcanic zone New Zealand, Bull Voloanol $2: 445-459 D'Amore F, Bolognesi L (1983) Isotopic variation ofthe hydrothermal system on Vulcaro Island, aly. Geochim Cosmochim Acta 57: 2069-2082 Davis AS, Clagve DA, Schulz MS, Hein JR (1991) Low sulfor content in submarine lavas: an ‘unreliable indicator of subarial eruption. Geology 19: 750-753, ‘Delaney JR, Muenow DW, Grabam DG (1978) Abundance and distribution of wate, carbon and ‘sulfur inthe lasy ris of submarine pillow barals, Geochim Cosmochim Acta 4: 581-894 Dixon JE, Clague DA, Stlpec E (1991) Degasing hisory of water, sufur and easbon in submarine Ibvas from Klsues, Hawai. J Geol 99: 371-334 ‘Dunbar NW, Hervig RL- Kyle PR (1985) Determination of pre eruptive H.0, F and Cheastents of slice magmas using tek iaclusons: example from Taupe Voleaic Centee, New Zeal Ball Vleanol Si: 177-184 Elis AJ (1957) Chemical equilibrium io magmatic gases. Am J Sci 285. 416-431 Francs P, Oppenteimer C, Stevenson D (1993) Endogenous growth of persistently active vo noes, Nature 36: 54-857 Gerlach TM (1979) Evaluation and restauration ofthe 1970 voleanic gas analyses from Me Etna Sic. J Voleanal Geotherm Res 6: 165-178 ‘Gestach TM (1986) Exsolusion of H,0, CO;, and § during erepive episodes at Kilavea voleano Hawai. J Geophys Res 9: 12177-12185 GGiegenbach WE (1975) A simple method forthe collection and alysis of volcanic gas samples Bull Voleanol 39: 132-145 Giggenbach WF (1987) Redox procescs governing the chemist offumarole gas discharges fom ‘White Island, New Zealand. Appl Geochem 2: 43-161 Grewenbach WE (1988) Yatur voeano investgntons, Vanuatu, Sept 1988. NZ Geo! Sure C136: 45-60 Giggenbach WF (1992s) The composition of gases in geothermal ant volcinicspsems 25 4 Tanetion of tectonic sting Prac Int Symp Water-Rock Interaction 8: 873-878 Giggenbach WE (19928) Isotopic shifts in waters (tom geothermal and volcanic gysem slong ‘onnergent plate bourdaries and ther origin, Earth Planet Sei Let 113: 495-Si0 Giggentuch WE (1993) Redox conesl of as compositions in Philippine volcane-hyérothermal stems, Geothermis 22: 575-587 Giggeabach WF (1995) Variations in the chemical and isotopic composition of fut Gocharged ‘over the Taupo Voleanc Zane. J Voleanal Geatherm Res 68 89-116, Giggenbach WF. Giower RH (1992) Tectonic regime ard major prosesses governing the cheisry tof wate and gas icharges from the Rotorus geothermal Feld, New Zealand, Geothermics die imta0