Non-volatile vs volatile behaviours of halogens during the AD 79 plinianeruption of Mt. Vesuvius, Italy
H. Balcone-Boissard a,⁎, B. Villemant a,b, G. Boudon a, A. Michel a
a Institut de Physique du Globe de Paris, CNRS, Equipe Géologie des Systèmes Volcaniques, 4 pl. Jussieu, 75005 Paris, Franceb Université P.& M. Curie, Paris, France
Received 20 June 2007; received in revised form 30 January 2008; accepted 1 February 2008
Available online
Editor: R.W. Carlson16 February 2008
Abstract
Pre-eruptive conditions and degassing processes of the AD 79 plinian eruption of Mt. Vesuvius are constrained by systematic F and Clmeasurements in melt inclusions and matrix glass of pumice clasts from a complete sequence of the pumice-fallout deposits. The entire ‘whitepumice’ (WP) magma and the upper part of the ‘grey pumice’ (GP) magma were saturated relative to sub-critical fluids (a Cl-rich H2O vapourphase and a brine), with a Cl melt content buffered at ~5300 ppm, and a mean H2O content of ~5%. The majority of the GP magma was not fluid-saturated. From these results it can be estimated that the WP magma chamber had a low vertical extent (b500 m) and was located at a depth of~7.5 km while the GP magma reservoir was located just beneath the WP one, but its vertical extent cannot be constrained. This is approximatelytwo times deeper than previous estimates. H2O degassing during the WP eruption followed a typical closed-system evolution, whereas GP clastsfollowed a more complex degassing path. Contrary to H2O, Cl was not efficiently degassed during the plinian phase of the eruption.
Keywords: Vesuvius; AD 79 plinian eruption; halogen behaviours; differentiation processes; magma degassing
1. Introduction
The eruptive style of explosive volcanic eruptions is prin-cipally determined by degassing processes, which are primarilycontrolled by the chemical and physical properties of themagma. H2O, SO2, CO2 and halogens (F, Cl) represent the mostimportant volatile components dissolved in magmas. Halogensare of particular interest because they behave as incompatibleelements during melt differentiation and are controlled by theirability to partition into the fluid phase during shallow magma
degassing (Villemant and Boudon, 1998; Villemant et al.,2003). Halogen distribution coefficients between aqueous fluidsand melts vary over a wide range, largely as a function of meltcomposition as shown by experimental results (Kilinc andBurnham, 1972; Bureau et al., 2000; Signorelli and Carroll,2000; Webster and De Vivo, 2002; Webster et al., 2003;Webster, 2004) and studies on natural systems (Villemant andBoudon, 1999). In addition, halogens are generally betterpreserved and easier to analyse than H2O or CO2 in volcanicglass (i.e. matrix glass and melt inclusions). An increasingnumber of high quality measurements of halogens in volcanicglass now exist. These cover a large variety of eruptive stylesand magma compositions which exemplify a large variety of
67H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
behaviours during magma differentiation and degassing, espe-cially for highly silicic magmas. Phonolitic magma eruptionsprovide a particularly suitable material for such studies becausetheir halogen contents are large and highly variable. In this workwe reinvestigate the behaviour of water and halogens during theAD 79 eruption at Vesuvius which emitted tephri-phonolitic tophonolitic magmas during a sustained andmainly plinian activity.This famous eruption has already been subject to numerousstudies; for instance, chronostratigraphy and sedimentology(Sigurdsson et al., 1985; Carey and Sigurdsson, 1987; Cioniet al., 1992), geochemistry and petrology including pre-eruptiveconditions and magma degassing (Mues-Schumacher, 1994;Cioni et al., 1995; Cioni et al., 1998; Cioni et al., 1999; Signorelliet al., 1999; Signorelli and Capaccioni, 1999; Cioni, 2000;Webster et al., 2003), textures and vesiculation conditions(Gurioli et al., 2005), and numerical modelling of eruptiondynamics (Papale and Dobran, 1993; Neri et al., 2002). Here wereport new H2O and halogen compositional data for meltinclusions and matrix glass. These results, combined withprevious geochemical studies, provide new constraints on thepre-eruptive conditions of the AD 79 eruption and also on thefactors controlling halogen behaviour during phonolitic magmadifferentiation and degassing.
2. Geological setting
2.1. Monte Somma–Vesuvius geological history
Mt. Somma–Vesuvius is a complex volcanic system, locatedin the Campanian Plain, Southern Italy, which is bordered by aMesozoic carbonate platform. It is composed of the ancientMonte Somma stratovolcano, the activity of which ceasedfollowing a large caldera collapse, and the Vesuvius volcano(sensu stricto), which is located inside this caldera. The MonteSomma–Vesuvius volcanic system will be referred to asVesuvius from now on. The onset of activity at Vesuviusoccurred ~35 kyr ago. On the basis of chronostratigraphic andarcheomagnetic studies, four distinct periods of activity arerecognized (Raia et al., 2000; Principe et al., 2004) whichdisplay a large spectrum of volcanic activity, from effusive tolarge explosive eruptions. The principal explosive eruptionscomprise four major plinian eruptions (Pomici di Base, ~20000BP; Mercato, ~9000 BP; Avellino, 3500 BP; Pompeii, AD 79)and numerous sub-plinian eruptions such as Pollena (472 AD)or the 1631 eruption. However, other plinian eruptions ordifferent eruption dates are also recognized (e.g. Arno et al.,1987; Dobran, 1993; Rolandi et al., 1998; Cioni et al., 1999;Santacroce et al., 2003). Reported petrological data suggest thatthe magmatic system of Vesuvius is composed of a deepreservoir (10–20 km depth; Nunziata et al., 2006) from whichmagma batches rise and feed shallower magma chambers(typically 3–5 km; DeNatale et al., 1998; Lima et al., 2003;DeNatale et al., 2004; Civetta et al., 2004). Magma seriesranging from silica undersaturated (K-basalt to K-trachyte) tohighly silica undersaturated (K-tephrite to K-phonolite) havebeen recognized (Belkin et al., 1985; Joron et al., 1987; Cioniet al., 1992; Marianelli et al., 1995; Ayuso et al., 1998).
2.2. The AD 79 eruption
The AD 79 eruption of Vesuvius, also known as the“Pompeii eruption”, has been previously described in detail(e.g. Sigurdsson et al., 1985; Cioni et al., 1992; Cioni et al.,1995). The eruption is commonly divided into three phases: aninitial phreatomagmatic phase, followed by a plinian eventwhich produced a thick pumice-fallout deposit and a finalphase that was dominated by numerous collapse events. Thepumice-fallout deposit displays a white to grey colour zoning,which is related to a change in the magma composition fromphonolite to tephri-phonolite (Barberi et al., 1981; Joron et al.,1987). Hereafter “white pumice” (WP) and “grey pumice”(GP) refer to the two main pumice types, which are representedby six fallout units U1, U2, U3, U4 (WP) and U5, U6 (GP)(Fig. 1).
3. Sampling and analytical methods
This study focuses on the pumice-fallout deposits whichrepresent 80% of the erupted magma from this eruption (Cioniet al., 1995). Surges and pyroclastic flow deposits (e.g. theeruptive units EU4 to EU8 described by Cioni et al., 1995) werenot sampled. To achieve a detailed investigation of thedegassing processes, a systematic and statistical sampling of arepresentative section of the pumice-fallout deposits wasundertaken. Six fallout units, U1, U2, U3, U4 (U1234), U5and U6 (U56) were sampled in a proximal 1 m thick section, atthe Terzigno quarry, 6 km southeast of the present crater.Although this site which is located on a bearing of N140° fromthe crater, is not on the main dispersion axis (N160°) it providesan easily accessible and complete section of the AD 79 pumice-fallout deposits. At least 100 pumice clasts were collected fromeach unit sampled by sieving at ϕ 16 mm, to limit bias due towind dispersion and to collect large fragments large enough foranalysis. Each pumice clast was then washed and cut into threepieces for density measurements, chemical analyses and texturalobservations. On the basis of the density distribution histo-grams, at least three pumice clasts were selected from each unitfor chemical analyses: one in the dominant density fraction andthe two others from the high and low density fractions. For unitsU5 and U6, which have large density distributions, 10 and 8clasts were selected, respectively. A piece of each selected clastwas powdered in an agate mortar for chemical analyses and theremaining piece was polished for S.E.M. imaging, pointanalyses (EPMA) or used for mineral and groundmassseparations.
—Density and vesicularity: The fragment was first weightedin air (Ma), before impregnation with molten paraffin to fill allopen and connected vesicles. It was then re-weighed in air (M i
a)and water (M i
w). Density (d) and vesicularity are then calculatedas follows:
d ¼ Ma= Mia �Mi
w
� �
Vesicularity kð Þ ¼ 1004 DRE � dð Þ=DRE
Fig. 1. Stratigraphic section of the AD 79 pumice-fallout deposits at Terzigno(6 km east of the vent). U = units numbering in the pumice-fallout depositsreferred to in the text. Grey line denotes the distribution of white and greypumice deposits (redrawn from Carey and Sigurdsson, 1987). S = surge andPF = pyroclastic flow.
68 H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
where the DRE is the Dense Rock Equivalent and is determinedby pycnometry of the bulk-rock powder (mean of fourmeasurements gives DRE=2.45±0.03 g/cm3 for WP and2.50±0.03 g/cm3 for GP).
— Phenocrysts (pyroxene, feldspar, biotite and leucite) andgroundmass were separated by hand-picking in ten of theselected samples. 50–100 mg of the separated material wasanalysed for selected trace elements by INAA. Data are reportedin Table 1.
— Crystallinities are estimated through mass-balancecalculations using major and trace element compositions ofmineral and groundmass separate. For a given element i, thecomposition of crystals j (Cij) and groundmass (Ci
gdm) arerelated to the whole-rock composition (Ci
WR) by the equation:CiWR=α(C
igdm)+∑βjCij, with α+∑βj=1.
α and βj are the weight fraction of groundmass and minerals,respectively. Crystallinity is given by ∑βj (or 1−α). Thesystem, composed of N linear equations (N=total number ofelements) with (α+∑ βj) unknowns, may be solved by a least-squares method, taking analytical errors into account. Calcula-
tions have been performed using 27 major and trace elements.The propagated error on α values is estimated as ~5%.
The ratio between the volume of vesicles and the volume ofmelt (Vg/Vl) of pumice clasts is calculated by:
Vg=Vl ¼ qliq=qV� �� 1
qV¼ a= 1=qp� �� b=qPhð Þ� �
where ρliq is the DRE measured for WP or GP, α and β thegroundmass and total phenocryst weight fractions (β=1−α) andρp and ρPh are the density of bulk pumice and phenocrysts,respectively. Data are reported in Table 1.
— Bulk-rock analyses: The major element compositionshave been determined by ICP-AES at the CRPG (Nancy,France). The trace element compositions were obtained byINAA (C.E.A., Saclay, France). The analyses are reported in thesupplementary data in the Appendix (Tables 1 and 2).
—Water and halogen contents of bulk rocks and groundmass:The water content of the groundmass was measured following themethod described inMichel andVillemant (2003). Residualwaterand halogens are concentrated as dissolved species in the residualmelts. If halogen contents in hydroxylated minerals are ignored,which is consistent with their very low weight fractions in thesemagmas, volatile contents of groundmass (Ci
gdm) may beestimated by correcting the bulk-rock measurement for pheno-crysts weight fraction (β), estimated by mass-balance calculations(Ci
gdm=CiWR/(1− β)). Due to the low phenocryst content
(β≤15%) of the pumice clasts from the AD 79 eruption, thiscorrection is within the same range as the analytical error (Cioniet al., 1995; Gurioli et al., 2005). Data are reported in Table 1.
— Melt inclusion (MI) and matrix glass (MG) compositions:Fragments of three pumice clasts from each unit were gentlycrushed in an agate mortar and crystals separated by hand-pickingunder a binocular microscope. The selected crystals were thenembedded in resin, abraded and polished on one side to exposeMI. For MG analyses, four fragments of different pumice clastschosen at random among the 100 clasts from each unit wereembedded in resin and prepared in the same way as the minerals.
3.1. MI andMGmicroprobe study: F, Cl andH2Omeasurements
The major elements, F and Cl compositions of MI (Table 2)and MG (see supplementary data in the Appendix) weredetermined by EPMA (CAMECA-SX 100, CAMPARIS).Analyses were performed with an acceleration voltage of15 kV and a beam current of 4 nA. The dwell time was 10s,except for Si (5s) and Na (5s). F and Cl were measured with adwell time of 50s, an acceleration voltage of 15 kV and a beamcurrent of 40 nA. Analytical conditions (in particular dwelltimes) have been established to ensure no Na and F diffusionduring EPMA analyses using glass standards (Table 2). Forinter-calibration of EPMA and pyrohydrolysis, Cl and F havebeen repeatedly analysed by both methods in three natural glasssamples (obsidians from Lipari, Eolian Island (LIP), Italy; Little
Table 1Density, crystallinity and volatile and trace element compositions of plinian clasts
Eruptive unit U1 U1 U1 U1 U2 U2 U2 U3 U3 U3 U3 U3 U4 U4 U4 U4 U4 U5 U5 U5 U5 U5 U5 U5 U6 U6 U6
Density: Samples fall in the dominant (m), most dense (h) or least dense (1) fractions. Vg/Vl: volume ratio of gas over groundmass calculated from density. Halogen and H2O contents are whole-rock compositionscorrected for phenocryst contents. Mean analytical errors are 5% for F and Cl and 10% for H2O. Crystallinity: groundmass and phenocrysts weight fractions are estimated through mass-balance calculations (see text fordetails). Selected trace element composition: whole-rock measurements were made by INAA (C.E.A., Saclay, France). Analytical precision is b0.4% for Th and Co, b1.2% for La and Rb and b5% for Sr, Cr and Ba.
69H.Balcone-B
oissardet
al./Earth
andPlanetary
ScienceLetters
269(2008)
66–79
Table 2Selected EPMA compositions of melt inclusions in WP and GP pyroxenes and feldspars
Melt inclusions in white pumice pyroxenes Melt inclusions in white pumice feldspars Standard glasses
H2O contents were determined using the by-difference method (see Section 3).
70H.Balcone-B
oissardet
al./Earth
andPlanetary
ScienceLetters
269(2008)
66–79
Fig. 2. Alkali-silica diagram for the AD 79 products. Data from the literature arefrom Mues-Schumacher (1994) and Cioni et al. (1995). WP = White pumice;GP = Grey pumice; WR =Whole rock; MI = Melt inclusion MG =Matrix glass;Mean analytical errors for whole-rock analyses are within the symbol size andfor glass, they are ~1% for SiO2 and Na2O+K2O.
71H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
Glass Mountain, California (LGM), United States; Corbettivolcano (CO5), Ethiopia). The results show a slight over-estimation of both F and Cl contents by EPMA compared topyrohydrolysis. For inter-calibration of each EPMA session, thesame natural glass samples have been used as internal standards.Furthermore, standards and samples have always been carboncoated together to avoid analytical differences arising due to thecoating thickness.
The water contents of several MI were estimated using EPMA(CAMECA SX50, CAMPARIS) with an acceleration voltage of30 kV, a beam current of 4 nA, and a dwell time of 5s for all majorelements. H2O contents were then estimated using the ‘by-difference method’ calibrated on three natural glass samples ofknown H2O and CO2 contents (Lipari obsidian, Pena Blanca(Mexico)) obsidian and melt inclusions in quartz from Mont Dore(Clochiatti, 1975).
The volume excited by the electron beam is ~1 μm3 and isdependant on the beam energy and the sample chemistry.Working conditions of the electron probe do not provide precisetextural information on this analytical volume. For example MGanalyses may be biased by the presence of undetected microliteswithin the analytical volume. Analytical data are, therefore,selected using major element diagrams combining the spotanalyses of glass and mineral compositions; analyses displayingevidences for contribution from mineral phases are discarded.MI analyses are affected by the same issue: the exposed MIhave a mean diameter of ~25 μm and a variable thickness due totheir shape and the depth of polishing. The composition of thehost mineral was systematically measured close to the inclusionand the coupled inclusion-host mineral analyses were then usedin the same way as described for MG analyses to eliminateaffected samples. Wherever possible, several analyses wereperformed on large inclusions (diameter N40 μm). Potentialboundary layer effects on halogen contents in glass close tophenocrysts due to low melt diffusivities have also beenchecked. Glass wetting phenocrysts and glass remote fromcrystals, analysed in the same fragments have very similarcompositions. This excludes significant diffusion effects, whichis in agreement with experimental results (Baker et al., 2005).
4. Results
4.1. Bulk-rock chemistry and mineralogy
Detailed mineralogical studies of AD 79 eruption products aregiven byBarberi and Leoni (1980), Barberi et al. (1981) andCioniet al. (1995). Feldspar, pyroxene, mica and leucite are theprincipal mineral phases; amphibole, garnet, nepheline andolivine are minor components. The calculated phenocryst weightfractions (Table 1) give the following mean modal composition:feldspar (74%), clinopyroxene (21%), brownish mica (0.4%) andleucite (4.5%) for WP and feldspar (32%), clinopyroxene (36%),brownish mica (32%) and leucite (~0.1%) for GP. The WP andGPmainly differ by a higher feldspar weight fraction inWP and ahigher pyroxene+mica weight fraction in GP. The WP bulk-rockcomposition is phonolitic (SiO2 [55–56%] and (Na2O+K2O)[15.2–15.8%]). The GP correspond to less differentiated mag-
mas: phonolites to tephri-phonolites (SiO2 [53.3–54.2%] and(Na2O+K2O) [12.5–13.5%], Fig. 2). These composition rangesare consistent with previous determinations (Mues-Schumacher,1994; Cioni et al., 1995). The WP whole rocks have slightlyhigher incompatible element contents (e.g. La, Th, Rb) and lowercompatible element contents (e.g. Cr, Co, Ba, Sr) than GP wholerocks (Table 1). In the alkali-silica diagram (Fig. 2), WP and GPare clearly distinct: the existence of intermediate compositions, assuggested by other studies (e.g. Mues-Schumacher, 1994; Cioniet al., 1995) is not confirmed by our data.
4.2. Matrix glass and melt inclusion compositions
The matrix glass (MG) of all pumice clasts is phonolitic(Fig. 2). Groundmass trace element compositions are close towhole-rock compositions due to the low phenocryst weightfraction (β≤15%). Major element compositions of MG fromWP and GP define two groups that show no overlap, similar tothat observed for whole rocks (Figs. 2 and 3). WP MGrepresents the most differentiated melts, with higher total alkalicontents (Na2O+K2O) and Na2O/K2O ratio (~15% and ~1,respectively) than GP MG (~14% and ~0.8, respectively) forsimilar silica contents (SiO2 [55.2–56%]) (Fig. 2). WP MGdisplay higher Al2O3 and Na2O and lower CaO, MgO, Fe2O3,K2O and P2O5 contents than GP MG (Fig. 3). MI analysed inpyroxenes and feldspars are typically phonolitic in composition,except some MI trapped in pyroxenes from GP. WP and GP MIalso define two wholly distinct regions (Fig. 2). MI trapped in
Fig. 3. Major element compositions of whole rocks (WR), matrix glass (MG) and melt inclusions (MI). Comparison with basic MI (crosses) in diopsides from maficmagma of Vesuvius activity (Marianelli et al., 1995).
Fig. 4. Melt inclusion compositions. Histogram of H2O content of melt inclusionsfrom the AD 79 eruption and less differentiated Vesuvius magmas. Data from theliterature are fromMarianelli et al. (1995) (M., 1995), Cioni et al. (1995) (C., 1995),Cioni (2000) (C., 2000) and Sigurdsson et al. (1990) (S., 1990).
72 H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
WP pyroxenes and feldspars have a narrow compositionalrange; they represent the most evolved melts trapped in crystalswith the highest SiO2, Al2O3, Na2O and K2O and the lowestCaO, MgO and Fe2O3 contents. The MI compositional range inGP pyroxenes is greater; no accurate MI analyses in GPfeldspars have been obtained because of their low size and highcrystallinity (see Analytical methods). MI compositions in WPfeldspars are similar to those reported by Cioni (2000). Thecompositions of MI in WP and GP pyroxenes reported heredefine a wider range than those reported by Cioni et al. (1995)and Marianelli et al. (1995).
MI and MG compositions of WP and GP define two distinct,continuous differentiation trends (Fig. 3). They are character-ized by enrichment in Na2O and CaO, and depletion in K2O atan approximately constant Al2O3 content. Evolution trends inSiO2, Fe2O3, and MgO vs Na2O for WP and GP are divergent.These data confirm the existence of a slight but significantcompositional gap between WP and GP melts although theexistence of intermediate compositions has been suggestedpreviously (“grey–white” transition of EU2/3 or “boundarypumices”; Mues-Schumacher, 1994; Cioni et al., 1995). Sinceour analyses have been performed on a systematically andstatistically sampled large population of pumice fragments froma complete pumice-fallout sequence, it is likely that inter-mediate melt compositions, if they exist, would represent syn-eruptive magma mingling between WP and GP magmas duringascent in the feeder conduit and not in the magma chamber. Thisinterpretation is consistent with the rare occurrence of bandedpumices and statistical analysis on bulk-rock compositions,which shows a clear bimodal distribution corresponding to theWP and GP (DeVivo et al., 2003).
4.3. H2O and CO2 contents of glass
Several estimates of H2O contents in MI using the ‘by-difference method’ (Devine and Gardner, 1995) have beenperformed in this work (Table 2). However, a relatively largedata set of H2O measurements in MI from a range of minerals(clinopyroxene (CPx), feldspar (Fp), olivine (Ol) and leucite(Le)) and using various techniques (by-difference, IR or SIMS)
is available for Vesuvius magmas (Sigurdsson et al., 1990;Cioni et al., 1995; Marianelli et al., 1995; Cioni, 2000). Theseresults display a very large range of H2O contents (Fig. 4) whichmay be related to numerous petrological factors highlighted byLowenstern (1995). Examples include post-entrapment degas-sing and/or late melt crystallization which are common in thesematerials and lead to specific analytical difficulties such as aneed for glass homogenisation before analysis and theapplication of large correction factors (e.g. Cioni, 2000).
Sanidine represents the late mineral phase to crystallizewithin the magma chamber and consequently displays the mostdifferentiated MI composition. Provided H2O contents in theseMI are well preserved, they represent the best estimate of theminimum pre-eruptive H2O content. Only three significant datapoints for H2O contents in sanidine MI from the WP exist(Sigurdsson et al., 1990; Cioni, 2000). These range between 4.7and 6.5%. In the GP, H2O content in sanidine MI range between3.5 and 5%. However, these MI display systematic post-entrapment trends and H2O loss (Cioni, 2000): therefore theseH2O contents likely represent minimum values. Major element
Fig. 5. H2O and Cl vs Vg/Vl. Whole-rock compositions are corrected forphenocryst content. H2O: curved lines: theoretical closed-system degassing[Villemant and Boudon (1999) and Villemant et al. (1993)]. Temperature effects(data from Cioni et al., 1998) are accounted in H2O solubility and Vg/Vl
calculations. Solid line = WP (initial H2O content=6%, T=1123 K), dashedlines=GP (initial H2O content=5 and 7%, T=1323 K). The horizontal arrowcorresponds to vesicle expansion without exsolution of a water-bearing fluid.Grey region=open-system degassing. Saturation pressures are calculated fromexperimental data (Marziano et al., 2007). Cl: The mean and range of Cl in meltinclusions are shown for WP (open triangles; 20 analyses) and GP (solidtriangles; 30 analyses).
73H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
compositions of MI in pyroxenes from the AD 79 eruption spana wide range, which extends towards a less differentiated meltcomposition (Cioni, 2000; this work, Fig. 4). Excluding onevalue close to 6.5% all other H2O contents in these MI define acontinuous range between 5% and 1% (Cioni, 2000; this work).Finally, leucite MI correspond to late differentiated melts (highNa2O ~10% and low K2O ~5% contents) and display the lowestH2O contents (0.7 to 2.6%). This is consistent with the late, syn-eruptive crystallization of leucite which likely occurs in stronglydegassed magmas. MI in pyroxenes from less differentiatedmagmas erupted at Vesuvius (Cioni et al., 1995; Marianelliet al., 1995) display a low and relatively narrow range in H2Ocontents (1.5–2.5%).
The consistency of these data may be tested assuming H2O isincompatible (i.e. assuming no volatile phase saturation and ne-glecting the contribution of hydrous minerals) during the magmadifferentiation that produced the AD 79 melts. Thus, H2O con-tentsmay be estimated from the enrichment factors of other highlyincompatible elements, such as Th, between the most primitiveandWP or GPmelts (e.g. [H2O]
WP=[H2O]° ⁎ [Th]WP/[Th]). The
most mafic MI have [H2O]° ~1.8–2.4% (Marianelli et al., 1995).Th contents of the most ‘primitive’ melts are [Th]° ~16 ppm,while those ofWP and GPmelts are [Th]WP ~48 ppm and [Th]GP
~33 ppm, respectively. These values lead to H2O content esti-mates ranging between 5.4–7.2% for WP and 3.7–5% for GPassuming fluid phase-undersaturated conditions. These ranges areconsistent with the highest H2O content recorded inMI of the AD79 magmas. This comparison of MI data leads us to propose bestestimates for maximumH2O contents of 6±1% forWPmelts and5±1% for GP melts, prior to eruption.
CO2 measurements in MI (Cioni, 2000) and isotopicmeasurements of skarn fluid inclusions (Fulignati et al., 2005)from the AD 79 eruption suggest the absence of magmatic CO2.However, an exsolved CO2 phase originating from decarbona-tation reactions at magma chamber walls or from magmaticallyderived CO2 cannot be excluded.
4.4. H2O degassing during the AD 79 plinian eruption
H2O degassing paths of magmas during the AD 79 eruptionmay be modelled using simple assumptions on H2O behaviour(Burnham, 1975; Burnham, 1994; Carroll and Blank, 1997;Villemant et al., 2003). It is assumed that H2O is the majorvolatile species and its solubility is governed by the experi-mental law determined in K-phonolitic melts (Marziano et al.,2007). The pre-eruptive H2O content is estimated from MIanalyses. The closed-system degassing path, typical of plinian-type eruptions, may be theoretically calculated and representedin the [H2O]r–Vg/Vl diagram (Fig. 5) (Villemant and Boudon,1998, 1999). During magma ascent from the H2O saturationdepth, H2O exsolves with decreasing pressure and bubbles formand expand in response to both the solubility decrease and gasvolume increase. As pumice clasts represent magma quenchedat different pressures and therefore reflect different stages ofdegassing, their residual H2O content and vesicularity ([H2O]r,Vg/Vl, Table 1) allow reconstruction of the syn-eruptivedegassing histories for the different eruptive units. The
representative points of the AD 79 pumice-fallout clasts plotfar from initial conditions. Most clasts from WP units plot onthe closed-system evolution line, while the majority of clastsfrom the U5 GP unit plot on the open-system degassing trendand almost all clasts of U6 GP unit plot on an ‘expansion-only’trend. These results suggest that all magma fragments (WP andGP) have suffered large H2O degassing before quenching.
5. Halogen contents of glass
The F and Cl contents of glass (MI, MG) are compared todata from the literature (Cioni et al., 1995; Marianelli et al.,1995; Cioni et al., 1998; Cioni, 2000) (Fig. 6a). F contents inanalysed MI measured are similar for pyroxenes and sanidinesand are similar to the range given in the literature. All analysedMI in sanidines and pyroxenes from fragments of the differentWP units have very similar Cl contents (5300±200 ppm)though they represent a large variation in differentiation degree(4bNa2Ob7%). However, there is a systematic shift of~1000 ppm between our data and those of Cioni (2000) forsanidine MI from WP units (6200±400 ppm), suggesting apossible systematic calibration bias. Nonetheless, both data setsdisplay a constant Cl contents for the same large range in
74 H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
differentiation degree. Contrary to WP, Cl contents of MI inpyroxenes from GP units vary strongly relative to a smallvariation in Na2O content (3–5%). In addition, the range of Clcontents in melts trapped in pyroxenes is greater in U6 than in
Fig. 6. F and Cl contents in AD 79 melt inclusions (MI) and matrix glass (MG). (a)inclusion data from Marianelli et al. (1995). (b) F vs Na2O: Stippled line=F content d(~2.5, Marianelli et al. (1995)). Solid line=differentiation trend. (c) Cl vs Na2O: Lifeldspars MI, 4: WP-GP pyroxenes MI. Solid line: differentiation trend for fluid-u(constant pressure). (d) Cl vs F: F and Cl contents measured in matrix glass and calcufrom matrix glass values by correcting for microcryst contents (WP: 35%, GP: 40%;differentiation trends assuming incompatible behaviours for F and Cl during degasaturated melts (at constant pressure).
U5. MI in leucites from WP and GP units display the highest Fand Cl contents. Finally, less differentiated MI (Na2Ob3%) inpyroxenes from the AD 79 eruption(Cioni et al., 1998; Cioni,2000) and other mafic magmas having supplied Vesuvius
Sanidine, pyroxene and leucite data from Cioni (2000) and mafic magma meltomain for MI in diopside in mafic magmas drawn, calculated from the Cl/F ratioterature data domains (Cioni, 2000): 1: leucites MI, 2: GP feldspars MI, 3: WPndersaturated melts; horizontal arrow: evolution trend for fluid-saturated meltslated in pre-eruptive melts (1=GP; 2=WP). Pre-eruptive contents are calculatedGurioli et al. (2005)). The three lines that pass through the origin correspond tossing-induced crystallization. Horizontal arrow=differentiation trend for fluid-
Fig. 7. F, Cl and Th correlation for magmas (whole rocks) of the Vesuviusactivity. Data from the literature from DeVivo et al. (2003). WP and GPcompositions (this work) are mean values (see Table 1). In the AD 79 magmas, Fbehaves as a highly incompatible element. F is little affected by fluid-saturationand magma degassing. The variation in Cl contents in the AD 79 melts cannot beexplained by simple incompatible behaviour. WP and GPAD 79 whole-rock Clcompositions plot on the ‘fluid-saturation trend’ defined by the Mercato–Avellino–Pompeii melts. Whole rocks of the Pollena eruption, however, plot ona ‘fluid-undersaturated trend’, which characterize a pure incompatible behaviourfor Cl, similar to that observed for F.
75H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
activity (Marianelli et al., 1995) define a low and narrowcomposition range for Cl (3900–4700 ppm). Other data formafic MI (Cioni et al., 1995) display an extremely large range ofCl values (3000–8200 ppm). The compositional regiondepicted in Fig. 6b corresponds to a high density contouraround the data from the literature and is largely defined by themore recent data, which show the least spread dispersion. The Fcontents of MG of WP and GP units are equal to or higher thanthose of MI and plot on the same Na2O–F correlation (Fig. 6b).Cl contents of all MG are significantly higher than those of MIand display two distinct Cl–F differentiation trends (Fig. 6d).
6. Discussion
6.1. F and Cl behaviour during Vesuvius magma differentiation
The F and Cl contents of AD 79 whole-rock samples arecompared to data from the literature (DeVivo et al., 2003).Routine measurements of F and Cl are typically performed withhigh detection limits; the large analytical errors at lowconcentrations likely explain the relatively large spreadobserved in data from the literature. Only analyses fromdedicated volatile studies and values greater than 0.1% for F and0.05% for Cl are considered. The linear correlation betweenNa2O and Th which passes through the origin for all themagmas produced by Vesuvius indicates that both elementshave similar incompatible behaviour during magma differentia-tion with no accompanied degassing (Fig. 7). Na2O and Th maytherefore be used as a differentiation index. Cl–Th diagramsdisplay relatively more complex behaviour. The majority of thedata define a clear linear trend through the origin. However, forsome plinian eruptions (e.g. Mercato, Avellino and Pompeii),large variations in Th are observed at almost constant Clcontents. This may be interpreted as the result of buffering by aCl-rich phase (mineral or fluid) as differentiation proceeds.However, other eruptions, such the sub-plinian Pollenaeruption, do not show evidence of this buffering effect, wherethe initial Cl–Th correlation is preserved up to a Cl enrichmentof ~10000 ppm in the most evolved magmas (Fig. 7). The Clbuffering effect observed in Mercato–Avellino–Pompeii meltsis not compatible with crystallization of Cl-bearing minerals forat least 3 reasons: (i) such a trend in Th–Cl diagram implies abulk solid/melt partition coefficient of Cl that is close to 1 andconstant, which is unlikely; (ii) minerals such as amphibole ormica that can fractionate Cl are almost absent in WP and rare inGP and have low Cl contents; (iii) Cl-bearing accessoryminerals (e.g. those of the sodalite group) would be required tocrystallize in significant amounts (e.g. ~10% of the total mineralfraction for an accessory mineral containing ~5% Cl) and havenot been observed in these magmas. The experimental results onCl partitioning between phonolitic melts and H2O–(Na,K)Clfluids (Métrich and Rutherford, 1992; Signorelli and Carroll,2000; Webster and De Vivo, 2002; Webster et al., 2003;Webster, 2004) and the thermodynamic properties of hydrosa-line solutions (Signorelli and Capaccioni, 1999), suggest thatthe phonolitic melts of the AD 79 eruption may have evolved atsaturation relative to Cl-bearing fluids in sub-critical conditions
(immiscible vapour+brine). The difference in Cl contentsbetween the Mercato–Avellino–Pompeii melts and Pollenamelts cannot be related to bulk melt compositions ordifferentiation degrees which are very similar in all thesemagmas. We, therefore, suggest that the difference in Clbehaviours is primarily controlled by fluid saturation of thephonolitic melts. An exsolved-volatile phase is necessary toaccount for extraction of Cl from the melt, otherwise, involatile-undersaturated conditions, Cl behaves as a pureincompatible and non-volatile element, even at large melt Clcontents.
6.2. Halogen behaviour during differentiation and degassing ofAD 79 magmas
In the AD 79 magmas, F behaves as a highly incompatibleelement and is not affected by magma degassing processes. Incontrast, the variations of Cl contents in the AD 79 melts cannotbe explained by simple incompatible behaviour. Cl content isalmost constant in WP MI (5300±200 ppm) whereas it displaysa large range in GP MI (5500–10 000 ppm). WP MIcompositions plot on the ‘fluid-saturation trend’ defined bythe Mercato–Avellino–Pompeii melts (Fig. 7). U6 MI trappedin pyroxenes have the highest Cl contents (~10000 ppm) andplot on a ‘fluid-undersaturated trend’ characterizing anincompatible behaviour of Cl similar to that observed for F.U5 MI in pyroxenes represent melts at intermediate stagesbetween the two trends. For the least evolved magmas of
Fig. 8. Cl and H2O solubility laws. Solid line=Experimental data of Cl solubilityin the AD 79 WP melt saturated with aqueous fluid+brine from Signorelli andCarroll (2002). Experimental data of water solubility in silicic melts. Greyline=K-phonolite at 1050 °C (Na2O: 5.2%; AD 79 white pumice; Marzianoet al. (2007)). Dotted line: Na-phonolite (Na2O: 10%; Carroll and Blank (1997)).WP melt inclusion Cl contents (5300±200 ppm) defined by the horizontaldotted lines. The vertical dotted lines correspond to the pressure range deducedfrom Cl solubility in melts saturated with a two-phase fluid. It is consistent witha pre-eruptive H2O content of ~5%. The wholeWPmagma has a phonolitic meltcomposition (see Fig. 2).
76 H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
Vesuvius (DeVivo et al., 2003) magma differentiation underfluid-undersaturated conditions allows increasing Cl contents ofresidual melts from ~2000 ppm up to an observed maximum of~10000 ppm (Fig. 7). This is consistent with a simpledifferentiation model, which predicts an enrichment of highlyincompatible elements (such as Th, Na2O, F and Cl) by a factorof ~3 between basic melts (Th ~10 ppm; Na2O ~1.5%) andtephri-phonolitic melts (‘GP type’: Th ~33 ppm; Na2O ~4%)and by a factor of ~5 between basic melts and the mostdifferentiated phonolitic melts (‘WP type’: Th ~47 ppm; Na2O~6–7%). Considering these enrichment factors, the expectedmaximum Cl contents of melts should be ~9500 ppm in GPmelts and ~13000 ppm in WP melts (Figs. 6 and 7). Our resultson U6 MI (maximum of ~10000 ppm) are similar to thepredicted value. On the other hand, no Cl contents close to thepredicted value have been measured in WP MI suggesting thatfluid-saturation and H2O and Cl exsolution were reached duringWP magma differentiation. Progressive differentiation atpressures where melts are fluid-saturated leads to Cl extraction,which buffers the melt Cl content to a value of ~5300 ppm. Thelarge decrease in Cl contents of U5MI, down to the Cl bufferingvalue at almost constant differentiation degree is interpreted asthe results of pyroxene crystallization at different pressures.This suggests that GP magmas were in different fluid-saturationconditions. Such pre-eruptive Cl buffering also explains thealmost uniform Cl contents measured in WP and GP bulkgroundmass (4000–5000 ppm; Fig. 6d). It is therefore assumedthat fluid-saturation conditions were reached in the magmachamber prior to the onset of the eruption for all WP magmasand only for the upper portion of the GP magma. Lower parts ofthe GP magma were not fluid-saturated and may have hadhigher pre-eruptive Cl contents.
During magma transfer to the surface, both F and Cl contentincreases in matrix glass compared to the pre-eruptive melts(Fig. 6d), which is largely attributed to degassing-induced meltcrystallization. During this late and rapid differentiation stage,both F and Cl behave as incompatible elements and are notaffected by fluid saturation. This could be explained by kineticeffects, i.e. the lower F and Cl diffusivities in melts relative toH2O (Gardner et al., 1996; Baker et al., 2005) prevent theirextraction into the exsolved vapour phase during rapid magmaascent. In addition, experimental results show that in silicicmelts, Cl partition coefficients between hydrosaline fluid andmelt strongly decrease with decreasing pressure (Signorelli andCarroll, 2000). This may also contribute to the lack of significantvariations in the Cl content of groundmass in the differentpumice clasts as their vesicularity increases (0.7bVg/Vlb6.5;Fig. 5). Finally, the highest F and Cl contents (5000–7500 ppmand 6000–9000 ppm, respectively; Cioni, 2000) of WP and GPleucite MI may be interpreted as reflecting a late crystallizationstage in the most differentiated melts.
6.3. Geochemical constraints on pre-eruptive conditions
6.3.1. Cl solubility constraintsPressure and depth of the pre-eruptive magma chamber of
the AD 79 eruption has been estimated previously on the basis
of structural and petrological considerations. Pressure estimatesrange between 100 and 150MPa, which correspond to depths of3 to 4.5 km (Barberi and Leoni, 1980; Barberi et al., 1981;Belkin et al., 1985; Sigurdsson et al., 1990). Our resultscombining both H2O and Cl behaviour give new, strongindependent constraints on pre-eruptive magma conditions,namely pressure, depth and the degree of fluid saturation. Theessentially constant Cl contents of most of the pre-eruptivemelts for the AD 79 WP and part of the GP magmas suggeststhat melts, as discussed above, were fluid-saturated prior to theeruption and in equilibrium with sub-critical hydrosaline fluids.Evidence for the presence of such hydrosaline fluids in theVesuvius volcanic system have also been reported by numerousstudies of skarn xenoliths (Belkin et al., 1985; Fulignati et al.,2001; Gilg et al., 2001; Fulignati et al., 2005). Hydrosalinesolutions ((Na,K)Cl–H2O) separate into a low density Cl-poor,a H2O-rich vapour phase and a dense brine when P and T reacha critical point (Anderko and Pitzer, 1993a,b). Therefore, atfixed P and T, the variance is equal to 0 and the composition ofthe multiphase assemblage is fixed, particularly the melt Clcontent (Williams et al., 1997). Considering the experimentaldata on Cl solubility in Vesuvius phonolitic melts saturated withaqueous fluid+brine (Holtz et al., 1995), a Cl buffering value of~5300 ppm corresponds to a pressure of ~185 MPa (depth~7.5 km, assuming lithostatic equilibrium). Therefore, takingthe constant Cl value of ~6200 ppm of Cioni (2000), theestimated pressure is ~35 MPa lower (giving a difference indepth of ~1 km, assuming lithostatic equilibrium). For similar Pand T conditions, other experimental data sets (e.g. Métrich andRutherford, 1992; Signorelli and Carroll, 2000, 2002; Webster
77H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
and De Vivo, 2002; Webster et al., 2003; Webster, 2004) givehigher Cl contents for phonolitic melts in equilibrium with sub-critical hydrosaline fluids. These differences are likely due tothe higher Na contents of the starting material which is knownto enhance Cl and H2O solubility (Holtz et al., 1995; Carroll andBlank, 1997; Webster and De Vivo, 2002).
The compositions of the AD 79 melts and of the Mercato–Avellino–Pompeii bulk rocks show that Cl contents are notstrictly constant but increase slightly with melt differentiation(Fig. 7). This increase, if significant, may be the result of adecrease in differentiation pressure conditions (in equilibriumwith a two-phase hydrosaline fluid) from the least (GP) to themost differentiated phonolites (Avellino). This variation may beestimated as ~50MPa using experimental results (Signorelli andCarroll, 2000). Using the same pressure constraints, the pressurevariation in the WP magma body should not exceed 30 MPa,which corresponds to a vertical extent, b500m. The existence ofboth Cl-rich and Cl-poor MI in U5 shows that the upper part ofthe GP reservoir contained both fluid-undersaturated and fluid-saturated melts and was also at a mean pressure of ~185 MPa.Conversely, U6 MI are more homogeneous and Cl-richindicating fluid-undersaturation and that the GP magma bodyextended to higher pressures (PN185 MPa).
6.3.2. Water solubility constraintsA compilation of data from the literature and this work
suggests that maximum H2O contents were ~6±1% for WPmagmas and ~5±1% for GP magmas. Corresponding H2Osaturation pressures may be derived using experimental H2Osolubilities in phonolitic melts. WP and GP magmas havedifferent alkali contents, which has a significant effect onvolatile solubility, particularly Na, which strongly favours H2Osolubility (Holtz et al., 1995; Carroll and Blank, 1997). H2Osolubility is generally considered not to be strongly dependenton temperature (Holtz et al., 1995). Compositional effects maybe estimated using H2O solubilities determined in Na phonoliticmelts (Carroll and Blank, 1997), K-phonolitic melts (Marzianoet al., 2007) and rhyolitic melts (Burnham, 1975; Burnham,1994; Holtz et al., 1995) and the work of Webster (1997) on thedissolution of Cl and H2O in silicate liquids. Here we assumethat H2O solubility is not significantly modified by Cl addition(Signorelli and Carroll, 2000, 2002). K-phonolitic melts (Na2O~7%), with 6% H2O, become saturated at a pressureapproximately 100 MPa higher than Na-phonolitic melts(Na2O ~12%) (Fig. 8). H2O solubility curves for K-phonolitesand rhyolites (Na2O ~4%) are very similar and differencesin the Na content of WP and GP do not lead to signif-icant differences in H2O solubility. We therefore estimate thatK-phonolites of the AD 79 eruption are saturated at pressures of~250 MPa for 6% H2O and ~185 MPa for 5% H2O. Pressureestimates for WP melts are consistent with those derived fromCl solubilities assuming pre-eruptive fluid-saturated WP meltscontaining 5% H2O.
The above results clearly show that at fluid-undersaturatedconditions H2O, Cl and F behave as incompatible elementsin phonolitic melts. The WP melt differentiation may havebeen enriched in H2O by differentiation processes (fluid-
undersaturated conditions) up to at least ~6% H2O. However,Cl contents of WP MI and MG clearly indicate that, beforeeruption, these melts were saturated with a Cl-rich hydrousfluid+brine with the following conditions: H2O=5%, Cl=5300 ppm. WP magmas were fluid-saturated prior to eruptionand therefore a fraction of the initial WP H2O content wasexsolved in the magma chamber (~20%, assuming a maximuminitial H2O content of ~6%). Contrary to the WP melt, GP meltswere only partially saturated in H2O- and Cl-bearing fluids.However, the MI data do not indicate that maximum pre-eruptive H2O contents of the GP magmas may have been muchgreater than 5%.
7. Conclusions
Fluid-saturation conditions in the AD 79 magmas constrainon the pre-eruptive conditions. Magmas stored in the upper partof the AD 79 magma chamber were saturated with a Cl-poorvapour and brine indicating that the pressure prior eruption was≤185 MPa and that the pressure variation over the WP magmawas less than 30 MPa. This implies that the WP magmachamber had a small vertical extent (b500 m) and was emplacedat a depth of ~7.5 km. The GP magma was only partiallysaturated with such fluids, indicating that the GP reservoirextended from just below the WP reservoir down to a depth thatcannot be constrained by volatile contents. These depths aresignificantly greater than those proposed by previous geochem-ical studies (Barberi and Leoni, 1980) but are consistent withtomographic results that show the existence of a present daymagma reservoir at a depth greater than 8 km (Nunziata et al.,2006).
These new constraints on the pre-eruptive and eruptiveconditions of the AD 79 plinian eruption, such as deepening ofthe magma chamber, reduction of the range of initial H2Ocontents (5–6%) and large variations in degassing processes,should lead to significant variation in numerical models of theeruption dynamics (Papale and Dobran, 1993; Neri et al., 2002).
The detailed study of halogen contents in melt inclusions andmatrix glass from the AD 79 eruptive products highlights thecomplex behaviour of halogens during phonolitic magmadifferentiation and degassing strongly controlled by fluid-saturation conditions.
Acknowledgments
We are grateful to V. Alaimo for helping with mineralseparation and polishing, M. Fialin and F. Couffignal formicroprobe measurements, J.L. Joron for trace elementmeasurements and A. Erguy for density measurements. Wealso thank an anonymous reviewer who helped to improve themanuscript and we are grateful to J. Mouatt for the constructivereview. IPGP contribution: 2320.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.epsl.2008.02.003.
78 H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
References
Anderko, A., Pitzer, K.S., 1993a. Equation of state representation of phaseequilibria and volumetric properties of the system NaCl–H2O above 573 K.Geochim. Cosmochim. Acta 57, 1657–1680.
Anderko, A., Pitzer, K.S., 1993b. Phase equilibria and volumetric properties ofthe systems KCl–H2O and NaCl–KCl–H2O above 573 K: equation of staterepresentation. Geochim. Cosmochim. Acta 57, 4885–4897.
Arno, V., Principe, C., Rosi, M., Santacroce, R., Sbrana, A., Sheridan, M.F.,1987. Eruptive history in Somma–Vesuvius. In: Santacroce, R. (Ed.), CNRQuaderni de “La Ricerca Scientifica”, 114, Roma. 8, pp. 53–113.
Ayuso, R.A., Vivo, B.D., Rolandi, G., Paone, A., 1998. Geochemical andisotopic (Nd–Pb–Sr–O) variations bearing on the genesis of volcanic rocksfrom Vesuvius, Italy. J. Volcanol. Geotherm. Res. 82, 53–78.
Baker, D.R., Freda, C., Brooker, R.A., Scarlato, P., 2005. Volatile diffusion insilicate melts and its effects on melt inclusions. Annals of Geophysics 48(N. 4/5,August/October), pp. 699–717.
Barberi, F., Leoni, L., 1980. Metamorphic carbonate ejecta from VesuviusPlinian eruptions: evidence of the occurrence of shallow magma chambers.Bull. Volcanol. 43, 107–120.
Barberi, F., Bizouard, H., Clocchiatti, R., Métrich, N., Santacroce, R., Sbrana,A., 1981. The Somma–Vesuvius magma chamber: a petrological andvolcanological approach. Bull. Volcanol. 44, 295–315.
Belkin, H.E., Vivo, B.D., Roedder, E., Cortini, M., 1985. Fluid inclusion geobaro-metry from ejected Mt. Somma–Vesuvius nodules. Am. Mineral. 70, 288–303.
Bureau, H., Keppler, H., Métrich, N., 2000. Volcanic degassing of bromine andiodine: experimental fluid/melt partitioning data and application to strato-spheric chemistry. Earth Planet. Sci. Lett. 57, 57–70.
Burnham, C.W., 1975. Water and magmas; a mixing model. Geochim.Cosmochim. Acta 39, 1077–1084.
Burnham, C.W., 1994. Development of the Burnham model for prediction of theH2O solubility inmagmas. In: Carroll, M.R., Holloway, J.R. (Eds.), Volatiles inmagmas, Reviews in Mineralogy. Mineral. Soc. Am., vol. 30, pp. 123–129.
Carey, S., Sigurdsson, H., 1987. Temporal variations in column height andmagma discharge rate during the 79 AD eruption of Vesuvius. Geol. Soc.Amer. Bull. 99, 303–314.
Carroll, M.R., Blank, J.G., 1997. The solubility of H2O in phonolitic melts. Am.Mineral. 82, 549–556.
Cioni, R., 2000. Volatile content and degassing processes in the AD 79 magmachamber at Vesuvius (Italy). Contrib. Mineral. Petrol. 140, 40–54.
Cioni, R., Marianelli, P., Sbrana, A., 1992. Dynamics of the AD 79 eruption:stratigraphic, sedimentologic and geochemical data on the successions of theSomma–Vesuvius southern sector. Acta Vulcanol. 2, 109–123.
Cioni, R., Civetta, L., Marianelli, P., Metrich, N., Santacroce, R., Sbrana, A.,1995. Compositional layering and syn-eruptive mixing of a periodicallyrefilled shallow magma chamber: the AD 79 plinian eruption of Vesuvius.J. Petrol 36 (3), 739–776.
Cioni, R., Marianelli, P., Santacroce, R., 1998. Thermal and compositionalevolution of the shallow chambers of Vesuvius: evidence from pyroxenephenocrysts and melt inclusions. J. Geophys. Res. 103 (B8), 18.277–18.294.
Cioni, R., Santacroce, R., Sbrana, A., 1999. Pyroclastic deposits as a guide forreconstructing the multi-stage evolution of the Somma–Vesuvius caldera.Bull. Volcanol. 60, 207–222.
Civetta, L., D'Antonio, M., Lorenzo, S.d., Renzo, V.D., Gasparini, P., 2004.Thermal and geochemical constraints on the “deep” magmatic structure ofMt Vesuvius. J. Volcanol. Geotherm. Res 133, 1–12.
Clochiatti, R., 1975. Les inclusions vitreuses des cristaux de quartz. Etudethermo-optique et chimique. Applications géologiques. Mém. Soc. Géol. Fr.122 (96 pp.).
DeNatale, G., Capuano, P., Troise, C., Zollo, A., 1998. Seismicity at Somma–Vesuvius and its implications for 3D tomography of the volcano. J. Volcanol.Geotherm. Res. 82, 175–197.
DeNatale, G., Troise, C., Trigila, R., Dolfi, D., Chiarabba, C., 2004. Seismicityand 3-D substructure at Somma–Vesuvius volcano: evidence for magmaquenching. Earth Planet. Sci. Lett. 221, 181–196.
Devine, J.D., Gardner, J.E., 1995. Comparison of microanalytical methods forestimating H2O contents of silicic volcanic glasses. Am. Mineral. 80,319–328.
DeVivo, B., Ayuso, R.A., Belkin, H.E., Fedele, L., Lima, A., Rolandi, G.,Somma, R., Webster, J.D., 2003. Chemistry, fluid/melt inclusions andisotopic data of lavas, tephra and nodules from N25 ka to 1944 A.D. of theMonte Somma–Vesuvius volcanic activity. Mt. Somma–Vesuvius Geo-chemical archive. Dipartimento di Geofisica e Vulcanologia, Università dinapoli Federico II Open file report 1–2003, p. 143.
Dobran, F., 1993. Global volcanic simulation of Vesuvius. VSG report No. 93-1,CNR-Gruppo Nazionale per la Vulcanologia, Italy, p. 148.
Fulignati, P., Kamenetsky, V.S., Marianelli, P., 2001. Melt inclusion record ofimmiscibility between silicate, hydrosaline, and carbonate melts: applica-tions to skarn genesis at Mount Vesuvius. Geology 89 (11), 1043–1046.
Fulignati, P., Panichi, C., Sbrana, A., Caliro, S., GioncADa, A., Moro, A.D.,2005. Skarn formation at the walls of the AD 79 magma chamber ofVesuvius (Italy): mineralogical and isotopic constraints. Neues Jahrb.Mineral. 181 (1), 53–66.
Gilg, H.A., Lima, A., Somma, R., Belkin, H.E., Vivo, B.D., Ayuso, R.A., 2001.Isotope geochemistry and fluid inclusion study of skarns from Vesuvius.Mineral. Petrol. 73, 145–176.
Gurioli, L., Houghton, B.F., Cashman, K., Cioni, R., 2005. Complex changes ineruption dynamics during the AD 79 eruption of Vesuvius. Bull. Volcanol.67, 144–159.
Holtz, F., Behrens, H., Dingwell, D.B., Johannes, W., 1995. H2O solubility inhaplogranitic melts: compositional, pressure, and temperature dependence.Am. Mineral. 80, 94–108.
Joron, J.L., Metrich, N., Rosi, M., Santacroce, R., Sbrana, A., 1987. Chemistryand petrography. In: Santacroce, R. (Ed.), “Somma–Vesuvius”. QuADernide la ricerca scientifica, CNR, Roma 8, pp. 105–171.
Kilinc, I.A., Burnham, C.W., 1972. Partitioning of chloride between silicatemelts and coexisting aqueous phase from 2 to 8 kbars. Earth Planet. Sci. Lett.102, 413–429.
Lima, A., Danyushevsky, L.V., Vivo, B.D., Fedele, L., 2003. A model for theevolution of the Mt. Somma–Vesuvius magmatic system based on fluid andmelt inclusion investigations. In: De Vivo, Bodnar (Eds.), Melt Inclusions inVolcanic Systems: methods, applications, and problems. Developments inVolcanology Series, vol. 5. Elsevier , pp. 227–249.
Lowenstern, J.B., 1995. Applications of silicate melt inclusions to the study ofmagmatic volatiles. In: Thompson, J.F.H. (Ed.), Magmas, fluids, and oredeposits. Mineral Ass. Can., vol. 23, pp. 71–99.
Marianelli, P., Métrich, N., Santacroce, R., Sbrana, A., 1995. Mafic magmabatches at Vesuvius: a glass inclusions approach to the modalities of feedingstratovolcanoes. Contrib. Mineral. Petrol. 120, 159–169.
Marziano, G.I., Schmidt, B., Dolfi, D., 2007. Equilibrium and disequilibriumdegassing of a phonolitic melt (Vesuvius AD 79 “white pumice”) simulatedby decompression experiments. J. Volcanol. Geotherm. Res. 161, 151–164.
Métrich, N., Rutherford, M.J., 1992. Experimental study of chlorine behaviourin hydrous silicic melts. Geochim. Cosmochim. Acta 56, 607–616.
Michel, A., Villemant, B., 2003.Determination of halogens (F, Cl, Br, I), sulfur andwater in seventeen geological reference materials. Geostand. Newsl. 27 (2),163–171.
Mues-Schumacher, U., 1994. Chemical variations of the A.D. 79 pumicedeposits of Vesuvius. Eur. J. Mineral. 6, 387–395.
Neri, A., Papale, P., Seppia, D.D., Santacroce, R., 2002. Coupled conduit andatmospheric dispersal dynamics of the AD 79 Plinian eruption of Vesuvius.J. Volcanol. Geotherm. Res. 120, 141–160.
Nunziata, C., Natale, M., Luongo, G., Panza, G.F., 2006. Magma reservoir atMt. Vesuvius: size of the hot, partially molten, crust material detected deeperthan 8 km. Earth Planet. Sci. Lett. 242, 51–57.
Papale, P., Dobran, F., 1993. Modeling the ascent of magma during theplinian eruption of Vesuvius in AD 79. J. Volcanol. Geotherm. Res. 58,101–132.
Principe, C., Tanguy, J.C., Arrighi, S., Paiotti, A., Goff, M.L., Zoppi, U., 2004.Chronology of Vesuvius activity from AD79 to 1631 based on arche-omagnetism of lavas and historical sources. Bull. Volcanol. 66, 703–724.
Raia, F., Webster, J.D., Vivo, B.D., 2000. Pre-eruptive volatile contents ofVesuvius magmas: constraints on eruptive history and behaviour. I. Themedieval and modern interplinian activities. Eur. J. Mineral. 12, 179–193.
79H. Balcone-Boissard et al. / Earth and Planetary Science Letters 269 (2008) 66–79
Rolandi, G., Petrosino, P., Mc Geehin, J., 1998. The interplinian activity atSomma–Vesuvius in the last 3500 years. J. Volcanol. Geotherm. Res. 82,19–114.
Santacroce, R., Sbrana, A., Andronico, A., Arrighi, S., Benvenuti, E., Cioni, R.,Di Vito, M., Gurioli, L., Luperini, W., Marianelli, P., Sulpizio, R., Zanchetta,G., 2003. Carta geologica del Vesuvio, 1:15000 Progetto CARG, ServizioGeologico Nazionale, CNR. In: Selca (Eds.).
Signorelli, S., Capaccioni, B., 1999. Behaviour of chlorine prior and during the79 A.D. plinian eruption of Vesuvius (southern Italy) as inferred from thepresent distribution in glassy mesostase and whole-pumices. Lithos 46,715–730.
Signorelli, S., Carroll, M.R., 2000. Solubility and fluid-melt partitioning of Cl inhydrous phonolitic melts. Geochim. Cosmochim. Acta 64 (16), 2851–2862.
Signorelli, S., Carroll, M.R., 2002. Experimental study of Cl solubility inhydrous alkaline melts: constraints on the theoretical maximum amount ofCl in trachytic and phonolitic melts. Contrib. Mineral. Petrol. 143, 209–218.
Signorelli, S., Vaggelli, G., Romano, C., 1999. Pre-eruptive volatile (H2O, F, Cland S) contents of phonolitic magmas feeding the 3550-year old Avellinoeruption from Vesuvius, southern Italy. J. Volcanol. Geotherm. Res. 93,237–256.
Sigurdsson, H., Carey, S., Cornell, W., Pescatore, T., 1985. The eruption ofVesuvius in A.D. 79. Natl. Geogr. Res. 1, 332–387.
Sigurdsson, H., Cornell, W., Carey, S., 1990. Influence of magma withdrawal oncompositional gradients during the AD 79 Vesuvius eruption. Nature 345,519–521.
Villemant, B., Boudon, G., 1998. Transition between dome-forming and plinianeruptive style: H2O and Cl degassing behaviour. Nature 392, 65–69.
Villemant, B., Boudon, G., 1999. H2O and halogen (F, Cl, Br) behaviour duringshallow magma degassing processes. Earth Planet. Sci. Lett. 168, 271–286.
Villemant, B., Trigila, R., Vivo, B.D., 1993. Geochemistry of Vesuviusvolcanics during 1631–1944 period. J. Volcanol. Geotherm. Res. 58,291–313.
Villemant, B., Boudon, G., Nougrigat, S., Poteaux, S., Michel, A., 2003. Waterand halogens in volcanic clasts: tracers of degassing processes during plinianand dome-building eruptions. Geol. Soc. Lond. Spec. Publ. 213, 63–79.
Webster, J.D., 1997. Exsolution of magmatic volatile phases from Cl-enrichedmineralizing granitic magmas and implications for ore metal transport.Geochim. Cosmochim. Acta 61 (5), 1017–1029.
Webster, J.D., 2004. The exsolution of magmatic hydrosaline chloride liquids.Chem. Geol. 210, 33–48.
Webster, J.D., De Vivo, B., 2002. Experimental and modeled solubilities ofchlorine in aluminosilicate melts, consequences of magma evolution, andimplications for exsolution of hydrous chloride melt at Mt. Somma–Vesuvius. Am. Mineral. 87, 1046–1061.
Webster, J.D., De Vivo, B., Tappen, C., 2003. Volatiles, magmatic degassing anderuptions of Mt. Somma–Vesuvius: constraints from silicate melt inclusions,Cl and H2O solubility experiments and modeling. Volatiles, magmaticdegassing and eruptions of Mt. Somma–Vesuvius, pp. 207–226.
