Nature and Evolution of PrimitiveVesuviusMagmas: an Experimental Study

MICHEL PICHAVANT1*, BRUNO SCAILLET1, ANNE POMMIER1,2,GIADA IACONO-MARZIANO1 AND RAFFAELLO CIONI3

1INSTITUT DES SCIENCES DE LA TERRE D’ORLEANS (ISTO), UMR 7327, UNIVERSITE D’ORLEANS, 45071 ORLEANS,

FRANCE AND ISTO, UMR 7327, CNRS, 45071 ORLEANS, FRANCE AND ISTO, UMR 7327, BRGM, BP 36009 ORLEANS,

FRANCE2SCRIPPS INSTITUTION OF OCEANOGRAPHY, UC SAN DIEGO, 9500 GILMAN DRIVE, LA JOLLA, CA 92093, USA3DIPARTIMENTO DI SCIENZE DELLA TERRA, UNIVERSITA DEGLI STUDI FIRENZE, VIA LA PIRA 4, 50121 FIRENZE,

ITALY

RECEIVED DECEMBER 16, 2013; ACCEPTED SEPTEMBER 23, 2014

Two mafic eruptive products fromVesuvius, a tephrite and a trachy-

basalt, have been crystallized in the laboratory to constrain the

nature of primitive Vesuvius magmas and their crustal evolution.

Experiments were performed at high temperatures (from 1000 to

�12008C) and both at 0�1MPa and at high pressures (from 50 to

200MPa) under H2O-bearing fluid-absent and H2O- and CO2-

bearing fluid-present conditions. Experiments started from glass

except for a few that started from glass plus San Carlos olivine crys-

tals to force olivine saturation. Melt H2O concentrations reached a

maximum of 6�0 wt % and experimental fO2 ranged from NNO ^

0�1 to NNOþ 3�4 (where NNO is nickel^nickel oxide buffer).

Clinopyroxene (Mg# up to 93) is the liquidus phase for the two

investigated samples; it is followed by leucite for H2O in melt

53 wt %, and by phlogopite (Mg# up to 81) for H2O in melt

43 wt %. Olivine (Fo85) crystallized spontaneously in only one ex-

perimental charge. Plagioclase was not found. Upon progressive crys-

tallization of clinopyroxene, glass K2O and Al2O3 contents strongly

increase whereas MgO, CaO and CaO/Al2O3 decrease; the residual

melts follow the evolution of Vesuvius whole-rocks from trachybasalt

to tephrite, phonotephrite and to tephriphonolite. Concentrations of

H2O and CO2 in near-liquidus 200MPa glasses and primitive

melt inclusions from the literature overlap.The earliest evolutionary

stage, corresponding to the crystallization of Fo-rich olivine, was re-

constructed by the olivine-added experiments. They show that the

primitive Vesuvius melts are trachybasalts (K2O� 4�5^5�5 wt %,

MgO¼ 8^9 wt %, Mg#¼ 75^80, CaO/Al2O3¼ 0�9^0�95)

that crystallize Fo-rich olivine (90^91) as the liquidus phase

between 1150 and 12008C and from 300 to5200MPa. Primitive

Vesuvius melts are volatile-rich (1�5^4�5 wt % H2O and 600^

4500 ppm CO2 in primitive melt inclusions) and oxidized (from

NNOþ 0�4 to NNOþ1�2). Assimilation of carbonate wall-rocks

by ascending primitive magmas can account for the disappearance

of olivine from crystallization sequences and explains the lack of

rocks representative of olivine-crystallizing magmas. A correlation

between carbonate assimilation and the type of feeding system is pro-

posed: carbonate assimilation is promoted for primitive magma

batches of small volumes. In contrast, for longer-lived, large-volume,

less frequently recharged, hence more evolved, cooler reservoirs,

magma^carbonate interaction is limited. Primitive magmas from

Vesuvius and other Campanian volcanoes have similar redox states.

However, the Cr# of Vesuvius spinels is distinctive and therefore

the peridotitic component in the mantle source of Vesuvius differs

from that of the other Campanian magmas.

KEY WORDS: Vesuvius; experimental petrology; phase equilibria;

primitive magmas; carbonate assimilation; magma reservoirs; potassic

series

I NTRODUCTIONVolcanic eruptions of Mt. Vesuvius (Italy) have beenconveniently divided into two groups, open-conduit andclosed-conduit (e.g. Santacroce et al., 1993; Marianelli

*Corresponding author. E-mail: pichavan@cnrs-orleans.fr

� The Author 2014. Published by Oxford University Press. Allrights reserved. For Permissions, please e-mail: journals.permissions@oup.com

JOURNALOFPETROLOGY VOLUME 55 NUMBER11 PAGES 2281^2310 2014 doi:10.1093/petrology/egu057 at U

niversity of California, San D

iego on Novem

ber 11, 2014http://petrology.oxfordjournals.org/

Dow

nloaded from

et al., 1995). Lava effusion and mixed effusive^explosiveactivity characterize the former group, which is best illu-strated by the 1637^1944 period of activity (Santacroceet al., 1993; Marianelli et al., 1995, 1999, 2005; Scandoneet al., 2008). The latter group involves either Plinian orsub-Plinian eruptions with erupted volumes of a maximumof a few km3, as illustrated by the Pompei (AD 79) and thePollena (AD 472) eruptions (Rosi & Santacroce, 1983;Cioni et al.,1995). Mafic compositions (tephrite to phonote-phrite) typically characterize open-conduit eruptions, andmore evolved compositions (tephriphonolite to phonolite)closed-conduit eruptions.Despite this fundamental subdivision, there is an overall

consensus that volcanic activity at Vesuvius is driven bythe periodic arrival of primitive mafic magma batchesat shallow crustal levels (e.g. Santacroce et al., 1993;Scandone et al., 2008). Mafic magmas are directly emittedtogether with more evolved, generally crystal-rich, prod-ucts, during violent Strombolian activity, as for exampleduring the 1794, 1822, 1872, 1906 and 1944 eruptions(Marianelli et al., 1995, 1999, 2005; Scandone et al., 2008).In contrast, eruption of mafic magma is not observed inPlinian events. However, Plinian magma reservoirs suchas that of the AD 79 eruption are periodically rechargedby primitive magma batches that contribute energy andmass to the reservoir and mix with the resident differenti-ates (Cioni et al., 1995). Therefore, it is critical for ourunderstanding of the volcanic activity of Vesuvius to gaina better knowledge of the physical and chemical propertiesof these primitive magma batches.So far, information on primitive magma batches at

Vesuvius has been elusive. Indeed, the absence of primitiveproducts is typical of the activity of Vesuvius (e.g. Dallaiet al., 2011). Geochemical studies have stressed the import-ance of magma mixing in crustal magma reservoirs; mostVesuvius eruption products represent mixtures betweenmafic and more evolved magmas (e.g. Belkin et al., 1993;Santacroce et al. 1993; Villemant et al., 1993; Marianelliet al., 1999). Until now, only detailed mineralogical studieshave provided clues on Vesuvius primitive magmas.Studies of primitive phenocrysts (Fo-rich olivine anddiopside) and of their melt inclusions have documentedthe occurrence of trachybasaltic to tephritic glasses(MgO¼ 8^10wt %; CaO/Al2O3¼1�0^1�1), proposed tobe representative of primitive Vesuvius melts (Marianelliet al., 1995, 1999, 2005; Cioni et al., 1998; Cioni, 2000).Volatile concentrations in these melt inclusions(H2O¼1�5^4�5wt %, CO2¼600^4500 ppm; Marianelliet al.,1995,1999, 2005; Cioni, 2000) have been used to deter-mine depths of magma crystallization (Marianelli et al.,1999, 2005).Another approach to constrain the nature of Vesuvius

primitive magma is experimental. For Vesuvius, previoushigh-pressure experimental work has concentrated on

evolved phonotephritic to phonolitic rocks (Dolfi &Trigila, 1978; Scaillet et al., 2008). One relatively primitivetephrite sample has been experimentally investigated at0�1MPa (Trigila & De Benedetti, 1993). Here we reportthe results of an experimental study on Vesuvius maficproducts, as an extension of our earlier work on Vesuviusphonolites (Scaillet et al., 2008). The data constrain thenature and physical conditions (major element compos-ition, crystallization pressures and temperatures, meltvolatile concentrations and redox state) of primitiveVesuvius magmas. New arguments for interaction betweenthe primitive magmas and carbonate wall-rocks are pro-vided. A correlation between the extent of carbonate as-similation and the type of feeding system is proposed. Ona broader perspective, this study contributes to the experi-mental characterization of mafic K-rich magmas, whichinclude rock-types such as lamproites, lamprophyres andminettes (Barton & Hamilton, 1978; Esperanc� a &Holloway, 1987; Wallace & Carmichael, 1989; Righter &Carmichael, 1996).

EXPER IMENTAL APPROACHAND SELECTION OF START INGSAMPLESIn this study natural mafic products from Vesuvius wereequilibrated under controlled laboratory conditions. Mostexperiments were carried out at high pressures (between50 and 200MPa), both under hydrous but H2O-under-saturated conditions (hereafter designated as fluid-absent;i.e. the melt H2O content is less than the solubility) andin the presence of an H2O- and CO2-bearing fluid phase(hereafter designated as fluid-present; i.e. addition of CO2

imposes the presence of a fluid phase). In addition, vola-tile-free experiments were performed at 0�1MPa.Experimental products were systematically comparedwith phenocryst assemblages and phenocryst and glassinclusion compositions from the eruptive products. Inthis study, eruptions younger than Avellino (3�9 ka BP)have been considered, in particular Pompei (AD 79),Pollena (AD 472), 1631, the 1637^1944 period and 1906 and1944.One major concern at the beginning of this study was

the choice of appropriate starting materials because, asemphasized above, primitive eruptive products are basic-ally lacking at Vesuvius. Another issue is that few maficVesuvius products actually represent magmatic liquids, be-cause of magma mixing and phenocryst accumulation orremoval (e.g. Marianelli et al., 1999).With these difficultiesin mind, two samples were selected, a tephrite (VES9)from a subplinian medieval (8th century) eruption and atrachybasalt (VS96-54A) from the 1944 eruption (Table 1;Fig. 1). VES9 originates from a lava fountaining episodeand is made of fallout crystal-rich (40% crystals) lapilli.

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2282

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Phenocrysts of clinopyroxene (cpx, Fs5^18En34^49Wo46^50,calculated with total Fe as Fe2þ), leucite [K/(KþNa)¼ 0�903^0�943] and rare plagioclase (plag,Ab20^53An40^80Or1^7) are present, together with a fewmagnetite (TiO2¼11�7wt %) and apatite inclusions(Table 2). VS96-54A comes from the main layer of the1944 lava fountain deposits (Marianelli et al., 1999). Thesample is porphyritic (45% crystals). It hosts two distinctphenocryst assemblages, the first comprising diopside(Fs5En48Wo47) and Fo-rich olivine (ol, Fo up to 90) andthe second salite (Fs9^15En36^43Wo48^50), less Fo-rich oliv-ine (Fo50^73), plagioclase (Ab11^16An80^87Or2^5) and minorleucite [K/(KþNa)¼ 0�907^0�914; Table 2; Marianelliet al., 1999].VES9 (K2O¼ 5�55wt %, Table 1; Fig. 1) is compos-

itionally close to the ‘parental mafic magma batch B’of Santacroce et al. (1993), being slightly more primi-tive than the V36 tephrite investigated by Trigila &De Benedetti (1993). MgO (6�73wt %) and CaO/Al2O3

(0�88) are in the middle of the range of primitive meltinclusions from the 1637^1944 period (Marianelli et al.,

1995, 1999, 2005). Compared with VES9, VS96-54A hashigher MgO (7�97wt %) and CaO/Al2O3 (1�04, Table 1),approaching the most primitive 1637^1944 inclusions(Marianelli et al., 1995, 1999, 2005). However, it is lesspotassic (K2O¼ 4�29wt %) and more calcic(CaO¼13�96wt %) than VES9 and the primitive meltinclusion group, suggesting some clinopyroxene accumula-tion (Marianelli et al., 1999).Other starting products were considered during the

course of the study, but none with clearly superior charac-teristics was found. In particular, a tephrite^phonotephrite(VS97-718) from a peripheral eruption of pre-medievalage (Table 1; Fig. 1) was examined in detail. This porphy-ritic sample contains phenocrysts of diopside (Fs4^5En48^49Wo47^48), salite (Fs10^17En32^44Wo46^51), Fo-rich (Fo up to90) olivine with Cr-rich (Cr# up to 81) spinel inclusions,and leucite [K/(KþNa)¼ 0�916^0�930, Table 2].Geochemically, VS97-718 is close to VES9 (Fig. 1), beingonly slightly more magnesian (Table 1). Mineralogically, itis similar toVS96-54A (coexistence of two distinct pheno-cryst assemblages). Moreover, olivine in VS97-718 is tooFo-rich to be in equilibrium with the bulk-rock. Therefore,it was not subjected to specific experimental investigations.However, olivines, diopsides and Cr-spinels in VS97-718are among the most primitive in Vesuvius products andtheir compositions are given inTable 2.

EXPER IMENTAL METHODSStarting materialsVES9 and VS96-54A were ground in an agate mortar to�50 mm and fused in air at 14008C, 0�1MPa in a Pt cru-cible. For each sample, two cycles of melting of 2^4 h each(with grinding between) were performed, yielding ahomogeneous glass whose composition was checked byelectron microprobe (Table 1). The glass was then crushedto �50 mm and stored in an oven.Most experiments used these two glasses directly as

starting materials. A few experiments were conductedwith glass (either VS96-54A or VES9) plus San Carlosolivine crystals (Fo90�5, sieved to 50^100 mm) to force satur-ation of the melt with a Fo-rich olivine under the specificexperimental conditions.

High-pressure experimentsNoble metal capsules, either Au90Pd10 or Au70Pd30 (morerarely Ag70Pd30) at410508C and Au �10508C, were usedas containers (length 15mm, internal diameter 2�5mm,wall thickness 0�2mm). For hydrous but H2O-undersatur-ated experiments, amounts of distilled water between 0�6and 2 ml and about 30mg of glass powder (i.e. less waterthan expected solubilities) were loaded in the capsule.For hydrous but fluid-saturated experiments performedwith H2O^CO2 mixtures, Ag2C2O4 was introduced inthe capsule as the source of CO2, together with distilled

Table 1: Major element composition of primitive Vesuvius

samples

VES9 VS96-54A VS97-718

whole-rock* glassy whole-rock* glassy whole-rock*

SiO2 (wt%) 48�02 48�8(5) 48�21 47�8(5) 48�91

TiO2 0�96 0�93(6) 1�02 1�05(8) 1�00

Al2O3 14�52 13�8(2) 13�42 12�4(2) 13�84

FeOt 7�52 7�44(34) 7�87 7�31(38) 7�05

MnO 0�14 0�13(8) 0�15 0�14(8) 0�14

MgO 6�73 6�51(34) 7�97 8�01(37) 7�52

CaO 12�77 12�6(2) 13�96 14�5(2) 12�39

Na2O 1�80 1�79(8) 1�54 1�57(9) 1�89

K2O 5�58 5�30(12) 4�29 4�11(13) 5�72

P2O5 0�62 0�87(8) 0�69 0�79(8) 0�76

LOI 0�74 – – – –

Total 99�40 98�2 99�12 97�7 99�22

Na2OþK2O 7�38 7�09 5�83 5�68 7�61

CaO/Al2O3 0�88 0�91 1�04 1�17 0�90

Mg# 61�5 60�9 64�3 66�1 65�5

*Whole-rock data:VES9, ICP-AES analysis (CRPG,Vandoeuvre-les-Nancy, France); VS96-54A, analysis fromMarianelli et al. (1999); VS97-718, analysis from R. Cioni(unpublished).yAverage of five electron microprobe analyses; numbers inparentheses are standard deviations in terms of least unitcited.Mg#¼ 100�molar MgO/(MgOþ FeO) with total Fe asFeO. FeOt, all Fe as FeO.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2283

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

water and the glass powder (about 30mg). H2O andAg2C2O4 were weighed so as to generate charges withvariable XH2Oin. [initial molar H2O/(H2OþCO2)]while the (H2OþCO2)/(H2OþCO2þ glass) mass ratio

was kept constant at �10%. All capsules were sealed byarc welding, keeping them in a liquid nitrogen bath to pre-vent water loss. They were then put in an oven for severalhours and reweighted to check for leaks.

Table 2: Representative electron microprobe analyses of phenocrysts in primitiveVesuvius samples

Sample Phase SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O NiO Total Composition

VES9 Lc 54�8 0�06 22�4 0�00 0�29 0�00 0�00 0�00 0�87 19�8 0�00 98�2 Ne3Ks46Qz51

Cpx 50�4 0�83 3�99 0�10 5�21 0�26 15�5 23�6 0�17 0�02 0�12 100�2 Fs8En44Wo48

Plag 51�5 0�03 29�2 0�09 0�75 0�00 0�00 12�5 3�95 0�5 0�10 98�6 An35Ab62Or3

VS96-54A Lc 55�0 0�00 22�1 0�06 0�38 0�05 0�02 0�00 1�20 19�5 0�00 98�3 Ne4Ks45Qz51

Cpx 51�7 0�38 1�97 0�19 3�61 0�03 17�6 23�8 0�13 0�01 0�00 99�4 Fs5En48Wo47

Cpx 47�2 1�38 6�91 0�00 7�44 0�01 13�6 23�0 0�29 0�00 0�01 99�8 Fs12En40Wo48

Ol 40�1 0�01 0�04 0�00 11�5 0�22 50�4 0�32 0�00 0�00 0�12 102�7 Fo88�6

Ol 37�7 0�00 0�05 0�00 24�4 0�33 37�8 0�41 0�00 0�00 0�16 100�9 Fo73�4

Plag 45�7 0�00 32�3 0�00 0�78 0�07 0�00 17�8 1�25 0�28 0�00 98�2 An11Ab87Or2

VS97-718 Lc 54�7 0�00 22�4 0�19 0�30 0�00 0�01 0�00 1�18 19�7 0�13 98�6 Ne4Ks46Qz50

Cpx 53�7 0�18 1�03 0�44 2�62 0�15 17�9 24�3 0�11 0�01 0�06 100�5 Fs4En49Wo48

Cpx 47�4 1�19 6�96 0�18 7�00 0�08 13�0 23�9 0�34 0�00 0�01 100�1 Fs12En38Wo50

Ol 40�6 0�04 0�09 0�04 9�37 0�32 47�2 0�37 0�01 0�00 0�45 98�5 Fo90

Sp 0�02 0�97 8�77 54�9 22�6 0�39 11�6 0�02 0�02 0�00 0�08 99�4 Cr#81

Lc, leucite; Cpx, clinopyroxene; Plag, plagioclase; Ol, olivine; Sp, Cr–Al spinel; Ne¼ 100� atomic (at.) Na/Si;Ks¼ 100� at. K/Si; Qz¼ 100(Si – Na – K)/Si in leucite; Fs¼ 100� at. Fe/(Mgþ FeþCa); En¼ 100� at. Mg/(Mgþ FeþCa); Wo¼ 100� at. Ca/(Mgþ FeþCa) in pyroxene (total Fe as Fe2þ); Fo¼ 100� at. Mg/(Mgþ Fe) in olivine;An¼ 100� at. Ca/(CaþNaþK), Ab¼ 100� at. Na/(CaþNaþK); Or¼ 100� at. K/(CaþNaþK) in plagioclase;Cr#¼ 100� at. Cr/(CrþAl) in spinel.

Experimental glassesVS96-54A

VES9VES9 + Ol

VS96-54A + Ol

wt% SiO2

wt%

K2O

+ N

a 2O

40 50 60 700

4

8

12

16

trachybasalt

phonolite

tephri-phonolite

phono-tephrite

tephrite

Eruption products194419061637-19441631AD 472AD 79

phonolite

phono-tephrite

trachybasalt

tephri-phonolite

VS96-54AVES9VS97-718

this work

Dolfi & Trigila 1978Trigila & DeBenedetti 1993Scaillet et al.2008

Fig. 1. Total alkalis^silica (TAS) diagram forVesuvius starting materials, experimental glasses and eruption products. Inset shows the compos-ition of the starting materials: trachybasalt (VS96-54A) and tephrites (VES9,VS97-718). Shown for comparison are the V36 tephrite compos-ition of Trigila & De Benedetti (1993), phonotephrite V1 of Dolfi & Trigila (1978) and four phonolites studied by Scaillet et al. (2008). Shadedfields represent whole-rock compositions of the AD 79 (Pompei), AD 472 (Pollena), 1631, 1637^1944 period, 1906 and 1944 eruptions from Rosi &Santacroce (1983), Rosi et al. (1993), Santacroce et al. (1993), Cioni et al. (1995, 1998) and Marianelli et al. (1999, 2005).

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2284

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

High-pressure experiments were all carried out in thesame internally heated pressure vessel, working verticallyand pressurized with Ar^H2 mixtures obtained by sequen-tial loading of H2 and Ar at room temperature (Scailletet al., 1992). Initial H2 pressures ranged from zero (no H2

initially loaded) to 5 bar. This yielded experimental fH2

(measured by Ni^Pd^O sensors; see below), from 0�1 to 12bar (Tables 3 and 4). Total pressure was recorded by atransducer calibrated against a Heise Bourdon tube gauge(uncertainty �20 bar). A double winding Mo furnace wasused, allowing near-isothermal conditions in the 2^3 cmlong hot-spot (gradient 52^38C cm�1). Temperature wasmeasured by three thermocouples (either type S or K)and was recorded continuously (total uncertainty �58C).Run durations were longer for the fluid-absent (14�5^23 h)than for the fluid-present (2^7�5 h) experiments. All runswere drop-quenched, resulting in nearly isobaric quenchrates of �1008C s�1 (Di Carlo et al., 2006).A majority of runs included a Ni^Pd^O sensor capsule,

which served to determine the experimental fH2 (Tayloret al., 1992; Di Carlo et al., 2006; Pichavant et al., 2009).Analysis of the composition of the NiPd alloy after theexperiment allows the fO2 of the sensor capsule to bedetermined (Pownceby & O’Neill, 1994). The fH2 ofthe sensor (and by inference of the experiment, as fH2 isidentical for all capsules) is then obtained from the waterdissociation equilibrium using the fO2 determined above,the dissociation constant of water (Robie et al., 1979) andthe fugacity of pure water at the experimental P and T

(Holloway, 1987). NiPd alloy compositions and the corres-ponding fH2 are listed in Tables 3 and 4. For runsperformed without a sensor, experimental fH2 were esti-mated from the H2 initially loaded into the vessel, the pro-cedure being calibrated from experiments that included asensor. For a given experiment (constant P^T^fH2), thefO2 of each charge varies along with aH2O (or fH2O).The latter was determined for each charge from the H2Ocontent of the quenched glass, using the thermodynamicmodel for H2O solution in multicomponent melts ofBurnham (1979). The oxygen fugacity of each charge isthen calculated from the water dissociation equilibrium,using the fH2 and fH2O determined above, and the dis-sociation constant of water (Robie et al., 1979). Typicaluncertainties on log fO2 are less than 0�25 log units (e.g.Scaillet et al., 1995; Martel et al., 1999; Scaillet & Evans,1999; Costa et al., 2004). In this study, fO2 values are ex-pressed as deviations from the NNO (nickel^nickel oxide)buffer (�NNO values), calculated at the P andTof interest(Tables 3 and 4).At the end of the experiment, the capsules were weighed

to check for leaks and then opened. For each capsule, frag-ments of the run product were mounted in epoxy and pol-ished for scanning electron microscope (SEM)observations and electron microprobe analyses. Glass

chips from selected charges were prepared for Fourier-transform infrared spectrometry (FTIR) measurements.The metallic pellets in the sensor capsule were recovered,mounted in epoxy and then analyzed by electronmicroprobe.

0�1MPa experimentsThese were carried out in two vertical gas-mixing furnacesusing the wire-loop method. Redox conditions were con-trolled with CO^CO2 gas mixtures and the fO2 (fromNNO to NNOþ 0�5) was directly read using a ZrO2 elec-trochemical cell. Temperature was monitored from Pt^PtRh thermocouples. Considerable experimental difficul-ties were encountered, especially with K2O volatilization.To minimize K2O volatilization and, at the same time, Feloss from the charge to the suspension wire, experimentswere repeated under constantT^fO2 conditions during 2^3 cycles of relatively short durations (3 h, each startingwith fresh glass) to progressively saturate the same suspen-sion wire with Fe. Intermediate charges were quenched,then dissolved in HF, and the charge from the last cyclewas retained for detailed study (Pichavant et al., 2009).Both Re and Pt suspension wires were tested but Re wasfound to enhance K2O volatilization and its use was subse-quently avoided.

ANALYT ICAL METHODSAll charges were systematically examined by SEM underback-scattered electron mode (JEOLWINSET JSM 6400instrument, University of Orle¤ ans) to assist in the identifi-cation of the phases and to evaluate the importance ofquench crystallization. Electron microprobe analyses ofnatural and experimental phases were performed with aCameca Camebax, Cameca SX-50 or Cameca SX Five atthe joint BRGM^CNRS^UO facility at Orle¤ ans. Analyseswere carried out under an acceleration voltage of 15 kV,counting times of 10 and 5 s on peak and background re-spectively, and a sample current of 6 nA, except for metal-lic sensor phases, which were analyzed under 20 kVand 20nA. For glasses, a slightly defocused beam of about 5 mmwas used, and for minerals a focused beam of 1^2 mm.Silicate minerals were used as standards. For the variousoxides, the relative analytical errors are 1% (SiO2, Al2O3,CaO), 3% (FeO, MgO, TiO2) and 5% (MnO, Na2O,K2O, P2O5). Phase proportions and FeO and K2O losseswere calculated for each charge using a least-squaresmass-balance routine computed after Albare' de (1995),using the electron microprobe compositions of the glassstarting material and of phases coexisting in the charge.The regression was based on eight major oxides, excludingMnO, P2O5 and H2O.One near-liquidus (charge 5-4) experimental glass was

analyzed for dissolved H2O by Karl-Fischer titration,using equipment and procedures identical to those

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2285

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table 3: H2O-bearing fluid-absent experiments

Charge WBD1 (wt %) H2O melt (wt %) aH2O �NNO2 Phase assemblage3P

R2 �FeO4 (%) KdFe–Mg

cpx–liq KdFe–Mg

ol–liq

Run 3, 214�1MPa, 11508C, 15 h, XNi¼ 0�42, fH2¼ 0�65 MPa

VS96-54A

3-1 3�7 4�1/3�8 0�98 þ1�0 Gl(100) n.d. �16 – –

3-2 3�0 – 0�63 þ0�6 Gl(94), Cpx(6), qu 2�6 �21 0�23 –

3-3 2�5 – 0�49 þ0�3 Gl(93), Cpx(7), qu 2�2 �19 0�21 –

3-4 2�5 – 0�44 þ0�3 Gl(83), Cpx(17), qu 3�4 �25 0�34 –

Run 5, 210�5MPa, 11508C, 10 h, XNi¼ 0�27, fH2¼ 0�28 MPa

VS96-54þOl

5-15 3�4 3�4 0�81 þ1�5 Gl, Ol, Cpx, qu – – 0�24 0�37

5-25 3�3 3�0 0�74 þ1�4 Gl, Ol, Cpx, qu – – 0�25 0�37

VES9þOl

5-45 3�2 3�1/3�1 0�66 þ1�3 Gl, Ol – – – 0�37

Run 1, 210�1MPa, 11008C, 14�5 h, XNi¼ 0�18-0�40, fH2¼ 0�1–0�5 MPa

VS96-54A

1-4 2�6 – 0�40 þ0�9 Gl(69), Cpx(31), qu 0�51 �9 0�19 –

1-5 3�6 – 0�70 þ1�4 Gl(81), Cpx(19), qu 0�18 �3 0�21 –

1-6 2�4 – 0�30 þ0�6 Gl(55), Cpx(45) 4�2 �27 0�30 –

Run 2, 213�5MPa, 11008C, 19 h, XNi¼ 0�14, fH2¼ 0�09 MPa

VS96-54A

2-1 4�2 – 0�81 þ2�5 Gl(75), Cpx(25), qu 0�53 þ1 0�23 –

2-2 4�0 – 0�72 þ2�4 Gl(67), Cpx(33), qu 0�25 þ5 0�31 –

2-3 3�4 – 0�56 þ2�1 Gl(64), Cpx(36), qu 0�24 0 0�24 –

Run 1b, 213�8MPa, 11008C, 17 h, XNi¼ 0�28, fH2¼ 0�28 MPa

VES9

1b-1 4�1 – 0�79 þ1�4 Gl(91), Cpx(9), qu 0�54 �4 0�38 –

1b-2 3�3 – 0�58 þ1�1 Gl(87), Cpx(13), qu 0�35 �7 0�29 –

1b-3 2�2 – 0�31 þ0�6 Gl(78), Cpx(21), qu 0�44 �7 0�33 –

Run 4, 208�1MPa, 10508C, 20�5 h, XNi¼ 0�41, fH2¼ 0�51 MPa

VS96-54A

4-1 5�4 – 0�96 þ1�0 Gl(60), Cpx(36), Phl(4), qu 0�12 �2 0�27 –

4-2 5�9 – 0�94 þ0�9 Gl(49), Cpx(41), Phl(10), qu 0�51 �1 0�18 –

4-3 5�1 – 0�82 þ0�8 Gl(54), Cpx(41), Phl(5), qu 0�13 �4 0�20 –

Run 8, 202�6MPa, 10508C, 21 h, fH2¼ 0�61 MPa

VES9

8-1 6�0 – 1�0 þ0�8 Gl(63), Cpx(29), Phl(8), qu 0�41 �6 0�18 –

Run 9, 206�2MPa, 10008C, 19 h, fH2¼ 6�1 MPa

VES9

9-1 5�7 – 0�95 þ0�7 Gl(58), Cpx(34), Phl(8), qu 0�03 �1 0�24 –

Run 6, 104�5MPa, 11008C, 23 h, XNi¼ 0�26, fH2¼ 0�12 MPa

VS96-54A

6-1 4�1 – 1�0 þ1�7 Gl(75), Cpx(25), qu 0�04 �2 0�25 –

VES9

6-2 3�8 – 1�0 þ1�7 Gl(88), Cpx(12), qu 0�15 �4 0�23 –

Run 10, 101�4MPa, 10508C, 21�5 h, fH2¼ 0�65 MPa

VES9

10-1 4�0 – 0�97 þ0�2 Gl(61), Cpx(33), Phl(6) 2�02 þ6 0�35 –

(continued)

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2286

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

described by Behrens et al. (1996). This glass was also ana-lyzed for H2O by FTIR, together with glasses 3-1, 5-1 and5-2 (fluid-absent experiments), whereas glasses 18-1, 17-2and 5-3 (fluid-present experiments) were analyzed forboth H2O and CO2. To do so, aThermo Fisher FTIR in-strument comprising a Nicolet 6700 spectrometer attachedto a Continuum microscope was used. Spectra wereacquired between 650 and 7000 cm�1 on doubly polishedglass wafers mostly 550 mm thick using an IR lightsource, a KBr beamsplitter and a liquid nitrogen cooledMCT/A detector. Between two and six spots (aperturemostly 50 mm) were analyzed on each sample.Concentrations of H2O and CO2 were determined fromthe Beer^Lambert law. Densities of the experimentalglasses were calculated from the densities of the startingglasses measured at room conditions (VS96-54A:2�769� 0�008; VES9: 2�703�0�007) and using a partialmolar volume of H2O of 12 cm3mol�1 (Richet et al.,2000). H2O concentrations were obtained from theabsorbance of the 3510^3530 cm�1 band, with a linearbaseline drawn between 3800 and 2500 cm�1. An extinc-tion coefficient (e3530) of 67 L mol�1cm�1 was used(Marianelli et al., 1999). For CO2, the absorbance of the1515 cm�1 band was measured on background-subtracted spectra (Dixon et al., 1995). An extinction coeffi-cient (e1515) of 365 L mol�1cm�1 (Marianelli et al., 1999)was used.

Most charges had too many crystals to allow analysisof their glass volatile concentrations by FTIR. Therefore,we had to resort to the ‘by-difference’ method to estimatethe concentration of dissolved volatiles (e.g. Devine et al.,1995). In each electron microprobe session, the differencefrom 100% of glass electron microprobe analyses was cali-brated against the dissolved H2O content, using glassesof known H2O concentrations (those analyzed by Karl-Fischer titration and FTIR) as secondary standards. Forthe fluid-absent (CO2-free) charges, the calibration isstraightforward and glass H2O concentrations with thismethod are estimated to within� 0�5wt %. For the fluid-present (CO2-bearing) charges, glass H2O concentrationsare overestimated, as CO2 dissolved in glass increases thedifference from100%. However, in this study, CO2 concen-trations in glasses are50�2wt %. Therefore, the overesti-mation remains small and the ‘by-difference’ method wasalso applied to estimate glass H2O concentrations inCO2-bearing charges.

EXPER IMENTAL RESULTSThe experiments are divided into three groups, corres-ponding respectively to the fluid-absent, fluid-present and0�1MPa charges. Experimental conditions and results foreach group are detailed in Tables 3^5, and experimentalcompositions are given inTable 6. In total,19 high-pressure

Table 3: Continued

Charge WBD1 (wt %) H2O melt (wt %) aH2O �NNO2 Phase assemblage3P

R2 �FeO4 (%) KdFe–Mg

cpx–liq KdFe–Mg

ol–liq

Run 11, 105�4MPa, 10008C, 22 h, fH2¼ 0�61 MPa

VES9

11-1 3�4 – 0�70 �0�1 Gl(50), Cpx(41), Phl(9), qu 0�06 0 0�22 –

Run 7, 55MPa, 11008C, 16�5 h, XNi¼ 0�62, fH2¼ 0�30 MPa

VS96-54A

7-1 3�0 – 1�0 þ0�4 Gl(68), Ol(1), Cpx(31), qu 0�08 þ3 0�27 0�27

VES9

7-2 2�6 – 0�97 þ0�4 Gl(86), Cpx(14) 0�04 �1 0�19 –

1H2O in glass estimated with the by-difference method. H2O melt in charges 3-1, 5-1 and 5-2 analyzed by FTIR and incharge 5-4 analyzed both by Karl-Fischer titration and FTIR. For charge 3-1 the two numbers given are duplicate FTIRmeasurements, and for charge 5-4, the first is the Karl-Fischer titration and the second the FTIR measurement. aH2Ocalculated from Burnham (1979) with WBD taken as the wt % dissolved H2O and melt composition from Table 6.2�NNO¼ log fO2 – log fO2 of the NNO buffer calculated at experimental P and T (Pownceby & O’Neill, 1994). For run 1,�NNO is the average calculated for fH2¼ 0�1MPa and fH2¼ 0�5MPa.3Phase proportions calculated by mass balance; Gl, glass; Cpx, clinopyroxene; Ol, olivine; Phl, phlogopite; qu, quenchphases identified by SEM.4Apparent loss or gain of FeO (Fe¼ FeOt) calculated as 100(FeOcalc – FeOstarting sample)/FeOstarting sample. FeOcalcand �R2 are obtained from the mass-balance calculations.5Olivine-added charge, mass-balance calculations not performed.Kd

Fe–Mgcpx–liq¼ (Fe/Mg in cpx)/(Fe/Mg in melt) calculated with FeO¼ FeOt in both cpx and melt. Kd

Fe–Mgol–liq¼ (Fe/Mg in

ol)/(Fe/Mg in melt) calculated with FeO¼ FeO in olivine and melt. (See text for the calculation of melt FeO.) XNi, molefraction Ni in the alloy phase of the sensor.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2287

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table 4: H2O-, CO2-bearing fluid-present experiments

Charge XH2Oin1 VBD2

(wt %)

H2O

(wt %)

CO2

(ppm)

aH2O �NNO3 Phase

assemblage4

PR2 (%) �FeO5 Kd

Fe–Mgcpx–liq Kd

Fe–Mgol–liq

Run 18, 207�8MPa, 12008C, 5 h, fH2¼ 0�04 MPa

VES9þOl

18-16 0�72 4�2 – 1302 0�90 þ3�4 Gl, Ol – – – 0�13

Run 17, 213�8MPa, 12008C, 6�5 h, fH2¼ 0�04 MPa

VS96-54þOl

17-26 0�69 3�4 3�0 1861 0�81 þ3�2 Gl, Ol – – – 0�35

Run 5, 210�5MPa, 11508C, 10 h, XNi¼ 0�27, fH2¼ 0�28 MPa

VS96-54þOl

5-36 0�69 2�9 2�6/2�6 1672 0�61 þ1�2 Gl, Ol, Cpx, qu – – 0�22 0�33

Run 13, 200�4MPa, 11508C, 2 h, fH2¼ 0�04 MPa

VS96-54A

13-1 0�10 0�9 – – 0�08 þ1�1 Gl(64), Cpx(36) 0�64 �5 0�35 –

VES9

13-2 0�11 0�7 – – 0�05 þ0�7 Gl(77), Cpx(23) 1�15 �7 0�39 –

Run 15, 98�6MPa, 11508C, 7�5 h, XNi¼ 0�06, fH2¼ 0�01 MPa

VS96-54A

15-1 0�72 1�4 – – 0�27 þ2�5 Gl(73), Cpx(27) 0�70 þ3 0�50 –

15-2 0�50 1�3 – – 0�24 þ 2�4 Gl(69), Cpx(31) 1�12 þ7 0�57 –

15-3 0�29 1�6 – – 0�34 þ2�7 Gl(76), Cpx(24) 0�60 �1 0�31 –

15-4 0�10 1�5 – – 0�30 þ2�6 Gl(69), Cpx(31) 0�97 þ2 0�44 –

VES9

15-5 0�20 1�2 – – 0�21 þ2�3 Gl(86), Cpx(14) 0�71 �4 0�32 –

Run 12, 101MPa, 11008C, 4 h, XNi¼ 0�07, fH2¼ 0�01 MPa

VES9

12-2 0�10 0�7 – – 0�06 þ1�1 Gl(55), Cpx(37), Lc(7) 0�46 �4 0�29 –

12-3 0�00 1�5 – – 0�25 þ2�3 Gl(56), Cpx(37), Lc(7) 0�37 �4 0�26 –

12-5 0�09 2�1 – – 0�41 þ2�7 Gl(60), Cpx(36), Lc(4) 0�98 �2 0�34 –

Run 16, 49�3MPa, 11508C, 3�5 h, fH2¼ 0�04 MPa

VS96-54A

16-1 0�67 1�4 – – 0�47 þ1�4 Gl(81), Cpx(19), qu 0�97 þ6 0�50 –

16-2 0�51 1�4 – – 0�46 þ1�4 Gl(79), Cpx(21), qu 1�32 þ8 0�44 –

16-3 0�25 1�1 – – 0�29 þ1�0 Gl(73), Cpx(27) 0�33 �1 0�37 –

VES9

16-4 0�30 1�1 – – 0�28 þ0�9 Gl(87), Cpx(13) 0�85 �7 0�36 –

1XH2Oin¼ initial molar H2O/(H2OþCO2) in the charge.2Total volatiles (H2OþCO2) dissolved in melt estimated with the by-difference method. H2O and CO2 concentrations inglasses 18-1, 17-2 and 5-3 determined by FTIR. Multiple numbers correspond to duplicate analyses. aH2O calculated fromBurnham (1979) with VBD taken as the wt % dissolved H2O and melt composition from Table 6.3�NNO¼ log fO2 – log fO2 of the NNO buffer calculated at experimental P and T (Pownceby & O’Neill, 1994).4Phase proportions calculated by mass balance; Gl, glass; Cpx, clinopyroxene; Ol, olivine; Lc, leucite; qu, quench phasesidentified by SEM.5Apparent loss or gain of FeO (Fe¼ FeOt) calculated as 100(FeOcalc – FeOstarting sample)/FeOstarting sample. FeOcalcand �R2 are obtained from the mass-balance calculations.6Olivine-added charge, mass-balance calculations not performed.Kd

Fe–Mgcpx–liq¼ (Fe/Mg in cpx)/(Fe/Mg in melt) calculated with FeO¼ FeOt in both cpx and melt. Kd

Fe–Mgol–liq¼ (Fe/Mg in

ol)/(Fe/Mg in melt) calculated with FeO¼ FeO in olivine and melt. (See text for the calculation of melt FeO.) XNi, molefraction Ni in the alloy phase of the sensor.

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2288

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

(corresponding to 44 charges) and 10 0�1MPa experimentsare reported. All experimental charges were equilibratedunder moderately to fairly oxidizing redox conditions.�NNO range from �0�1 to þ3�4 in the high-pressureexperiments, and the fluid-absent (�NNO¼ ^0�1 to þ2�5)and fluid-present (�NNO¼þ0�7 to þ3�4) charges over-lap. The 0�1MPa charges are more tightly grouped(�NNO¼ 0 to þ0�4).

Fe loss, quench crystallization andevaluation of equilibriumThe importance of Fe loss (owing to Fe alloying withthe metallic capsule) can be evaluated from the mass-balance calculations (Tables 3^5). In the high-pressureexperiments, five charges have 410% relative Fe loss,including three charges with 420% loss; none hasmore than 27% relative Fe loss. At temperatures �11008C,Fe losses or gains are 510% relative, except in charge1-6 (Table 3). At temperatures 411008C, Fe loss reaches21 and 25% relative in charges 3-2 and 3-4 (Table 3).It should be noted, however, that the olivine-addedcharges were not mass-balanced (because the addedolivine crystals were not precisely weighed) and sotheir Fe losses are unknown. In the 0�1MPa experiments,three charges have Fe losses 410% and none has morethan 19% relative loss. Therefore, Fe loss was kept at

relatively low levels in our experiments. No systematicchange in phase assemblage or composition appears be-tween charges with different Fe losses. K volatilizationwas limited to �10% relative in the 0�1MPa experiments(Table 5).Quench phases were commonly detected from SEM

observations and their presence is rather systematic in thehigh-pressure fluid-absent charges (Table 3). In compari-son, in the fluid-present charges, which are on average lessH2O-rich, quench crystals are rare (Table 4). Quench crys-tallization is marked by the appearance of very thin nee-dles (most certainly phlogopite) generally formingovergrowths on pre-existing crystals. Quench phases tendto concentrate in crystal-rich regions, and are typicallyabsent from large glass pools. They are also less frequentin crystal-poor charges. Despite their common occurrence,quench phases are rarely abundant enough for their crys-tallization to have a measurable effect on the melt compos-ition. K2O loss (which measures quench phlogopitecrystallization in high-pressure experiments; Di Carloet al., 2006) is510% relative except in two charges (4-2:12%; 4-3: 18%, Table 3). It is therefore concluded thatquench crystallization is of negligible importance in thisstudy.Most charges (48) were of crystallization type, whether

fluid-absent, fluid-present or volatile-free. In the six

Table 5: 0�1MPa experiments

Charge T (8C) Log fO21 (bar) �NNO Phase assemblage2

PR2 �FeO3 (%) �K2O

4 (%) KdFe–Mg

cpx–liq5

VS96-54A

3 1222 �7�89 þ0�4 Gl(100) n.d. �9 �10 –

4 1212 �7�99 þ0�4 Gl(97), Cpx(3) 1�74 �16 �4 0�27

6 1192 �8�25 þ0�4 Gl(81), Cpx(19) 0�49 �5 �1 0�51

7 1178 �8�37 þ0�4 Gl(80), Cpx(20), Lc(tr) 1�83 �13 �6 0�21

106 1124 �8�70 þ0�1 Gl, Cpx, Lc n.d. n.d. n.d. n.d.

VES-9

9 1209 �7�65 þ0�0 Gl(100) n.d. �6 �5 –

8 1201 �7�77 þ0�1 Gl(100), Cpx(tr) 2�90 �19 �10 0�14

6 1191 �8�00 þ0�2 Gl(100), Cpx(tr) 1�11 �4 �7 0�24

5 1180 �8�18 þ0�2 Gl(77), Cpx(18), Lc(5) 0�43 �5 �1 0�57

76 1123 �8�76 þ0�1 Gl, Cpx, Lc n.d. n.d. n.d. n.d.

1Log fO2 calculated from EMF measured by the zirconia oxygen probe; �NNO¼ log fO2 – log fO2 of the NNO buffercalculated at the experimental T (Pownceby & O’Neill, 1994).2Phase proportions (wt %) calculated by mass balance; Gl, glass; Cpx, clinopyroxene; Lc, leucite; tr, trace (51%).3Apparent loss or gain of FeO (Fe¼ FeOt) calculated as 100(FeOcalc – FeOstarting sample)/FeOstarting sample. FeOcalcand �R2 are obtained from the mass-balance calculations.4Apparent loss or gain of K2O calculated as 100(K2Ocalc – K2Ostarting sample)/K2Ostarting sample. K2Ocalc is obtainedfrom the mass-balance calculations.5Kd

Fe–Mgcpx–liq¼ (Fe/Mg in cpx)/(Fe/Mg in melt) calculated with FeO¼ FeOt in both cpx and melt.

6Mass-balance calculations not performed, crystals too small for analysis.n.d., not determined.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2289

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table6:

Com

positionsofexperimentalproducts(oxidesinwt%)

Charge

Phase

SiO

2TiO

2Al 2O3

FeO

MnO

MgO

CaO

Na 2O

K2O

Cr 2O3

NiO

P2O5

Total

Composition

Startingglasses

VS96-54A

Gl1(5)2

48�9(2)3

1�07(12)

12�7(2)

7�48(44)

0�14(6)

8�19(14)

14�8(2)

1�61(9)

4�20(13)

0�03(3)

0�02(3)

0�81(4)

97�8

Mg89

VES9

Gl(5)

49�7(5)

0�95(9)

14�1(2)

7�58(25)

0�13(8)

6�63(16)

12�8(24)

1�82(11)

5�4(16)

0�00(0)

0�01(3)

0�89(6)

98�1

Mg87

High-pressure

H2O-bearingfluid-absentexperim

ents

VS96-54A

3-1

Gl(5)

49�8(4)

1�02(6)

12�9(1)

6�27(21)

0�11(6)

8�27(12)

15�0(3)

1�54(6)

4�15(6)

0�06(7)

0�00(0)

0�90(5)

93�2

Mg78

3-2

Gl(5)

49�8(4)

1�12(4)

13�8(1)

6�17(31)

0�20(12)

7�61(9)

14�1(1)

1�72(8)

4�54(8)

0�03(6)

0�00(0)

0�93(4)

94�3

Mg76

Cpx(4)

52�0(6)

0�41(6)

2�05(41)

3�08(37)

0�06(6)

16�6(2)

24�5(4)

0�11(2)

0�05(2)

0�07(6)

0�12(14)

n.d.

99�0

En46Wo49

3-3

Gl(5)

49�7(2)

1�05(4)

13�8(2)

6�35(9)

0�16(7)

7�54(17)

14�1(3)

1�69(9)

4�56(4)

0�04(3)

0�02(5)

0�88(9)

95�0

Mg74

Cpx(3)

52�0(1)

0�41(16)

2�19(54)

2�99(40)

0�05(7)

16�9(3)

24�0(4)

0�13(6)

0�17(16)

0�23(2)

0�08(10)

n.d.

99�1

En47Wo48

3-4

Gl(5)

49�7(2)

1�17(7)

15�1(2)

5�94(13)

0�09(8)

6�73(19)

13�2(2)

1�83(12)

5�11(13)

0�07(5)

0�04(4)

1�04(5)

94�9

Mg73

Cpx(5)

50�6(18)

0�59(21)

3�43(154)

4�59(146)

0�07(6)

15�4(14)

23�8(3)

0�17(8)

0�06(4)

0�13(6)

0�04(7)

n.d.

98�9

En44Wo49

5-1(þOl)

Gl(5)

49�1(2)

1�07(6)

12�7(2)

6�77(18)

0�14(9)

9�07(9)

14�52(8)

1�5(10)

4�24(19)

0�01(3)

0�04(5)

0�84(4)

93�7

Mg79

Ol(5)

40�0(5)

0�01(1)

0�03(3)

8�35(36)

0�15(5)

49�1(2)

0�18(11)

0�00(1)

0�01(1)

0�06(3)

0�39(16)

n.d.

98�3

Fo91

Cpx(3)

51�5(6)

0�33(10)

1�74(79)

3�11(41)

0�1(9)

17�1(6)

24�3(2)

0�13(7)

0�02(2)

0�3(14)

0�03(6)

n.d.

98�7

En47Wo48

5-2(þOl)

Gl(5)

49�2(2)

1�09(9)

13�3(2)

6�60(18)

0�15(7)

8�64(6)

13�91(2)

1�76(6)

4�40(9)

0�02(3)

0�01(1)

0�86(4)

93�8

Mg79

Ol(2)

39�9(10)

0�01(2)

0�01(1)

8�70(32)

0�12(18)

49�1(5)

0�18(1)

0�01(1)

0�03(4)

0�00(0)

0�40(6)

n.d.

98�5

Fo91

Cpx(2)

52�4(9)

0�33(12)

1�68(22)

3�19(7)

0�01(1)

16�9(3)

24�5(6)

0�16(0)

0�05(2)

0�26(2)

0�01(1)

n.d.

99�5

En47Wo49

VES9

5-4(þ

Ol)

Gl(6)

49�6(2)

0�97(8)

13�8(1)

6�44(25)

0�14(11)

8�65(19)

12�5(3)

1�7(9)

5�20(16)

0�03(6)

0�05(6)

0�92(5)

94�9

Mg79

Ol(2)

40�0(0)

0�00(0)

0�00(0)

8�81(15)

0�11(4)

49�54(5)

0�14(10)

0�02(2)

0�02(1)

0�04(1)

0�32(7)

n.d.

99�0

Fo91

VS96-54A

1-4

Gl(5)

48�5(3)

1�14(6)

16�9(2)

8�03(19)

0�16(8)

5�14(12)

10�6(1)

2�26(7)

6�1(21)

0�06(8)

0�04(5)

1�05

495�6

Mg62

Cpx(1)

51�7

0�75

3�59

4�32

0�06

14�78

23�89

0�20

0�24

0�107

0�038

n.d.

99�6

En43Wo50

1-5

Gl(5)

48�0(3)

1�10(11)

15�3(3)

8�00(27)

0�17(11)

6�22(8)

12�7(2)

1�95(8)

5�4(10)

0�05(5)

0�03(4)

1�06

494�3

Mg69

Cpx(4)

51�4(9)

0�52(12)

2�69(53)

4�26(50)

0�12(5)

15�9(5)

24�6(4)

0�16(2)

0�10(4)

0�04(8)

0�04(8)

n.d.

99�9

En44Wo49

1-6

Gl(5)

50�4(3)

1�11(10)

18�9(3)

5�30(49)

0�19(7)

3�85(10)

8�81(9)

2�80(10)

7�63(32)

0�01(3)

0�00(0)

1�05

495�2

Mg65

Cpx(4)

49�0(13)

1�02(5)

6�11(92)

5�80(12)

0�07(8)

13�83(7)

23�2(6)

0�32(5)

0�21(12)

0�13(8)

0�06(6)

n.d.

99�8

En41Wo49

2-1

Gl(5)

49�2(7)

1�15(9)

15�6(2)

8�43(44)

0�18(9)

5�55(24)

11�4(3)

2�06(10)

5�27(18)

0�02(3)

0�04(6)

1�08

493�0

Mg69

Cpx(5)

50�5(12)

0�47(9)

2�97(96)

5�23(65)

0�08(6)

15�2(6)

23�5(4)

0�17(4)

0�10(6)

0�03(5)

0�02(5)

n.d.

98�3

En44Wo48

2-2

Gl(6)

49�4(6)

1�18(4)

16�4(2)

8�29(26)

0�21(5)

5�10(22)

10�3(4)

2�27(19)

5�86(14)

0�02(3)

0�02(6)

1�07

493�7

Mg67

Cpx(1)

47�6

0�82

5�07

6�98

0�21

13�7

23�1

0�26

0�16

0�18

0�00

n.d.

98�1

En40Wo49

(continued

)

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2290

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table6:

Continued

Charge

Phase

SiO

2TiO

2Al 2O3

FeO

MnO

MgO

CaO

Na 2O

K2O

Cr 2O3

NiO

P2O5

Total

Composition

2-3

Gl(7)

49�2(2)

1�13(8)

17�0(1)

8�40(41)

0�11(7)

4�86(19)

9�63(33)

2�27(10)

6�27(18)

0�04(4)

0�02(4)

1�06

494�5

Mg65

Cpx(3)

49�2(20)

0�64(17)

4�33(198)

5�90(167)

0�17(6)

14�1(14)

23�2(3)

0�25(9)

0�23(13)

0�04(4)

0�02(3)

n.d.

98�1

En41Wo49

VES9

1b-1

Gl(4)

48�8(7)

1�09(12)

15�5(3)

7�44(19)

0�25(9)

5�75(20)

12�0(2)

2�25(12)

5�79(16)

0�04(7)

0�03(6)

1�08

492�3

Mg69

Cpx(2)

51�0(14)

0�51(25)

2�95(95)

4�67(141)

0�12(6)

15�3(10)

24�1(1)

0�14(9)

0�13(5)

0�04(6)

0�00(0)

n.d.

99�0

En43Wo49

1b-2

Gl(4)

49�2(5)

1�00(7)

16�1(2)

7�50(30)

0�07(6)

5�39(9)

11�1(2)

2�20(15)

6�33(15)

0�06(6)

0�02(4)

1�07

493�5

Mg66

Cpx(2)

51�5(0)

0�43(8)

2�76(64)

4�27(9)

0�04(5)

15�9(3)

23�7(2)

0�16(4)

0�14(1)

0�08(9)

0�04(5)

n.d.

99�1

En45Wo48

1b-3

Gl(4)

49�3(5)

1�01(9)

17�0(1)

7�43(25)

0�13(14)

4�65(10)

9�83(28)

2�38(7)

7�12(9)

0�04(4)

0�11(8)

1�05

495�1

Mg61

Cpx(4)

50�0(18)

0�85(35)

4�06(150)

5�56(126)

0�10(8)

14�9(9)

23�9(7)

0�22(5)

0�13(5)

0�14(6)

0�03(2)

n.d.

99�8

En42Wo49

VS96-54A

4-1

Gl(4)

49�6(2)

1�03(6)

17�5(2)

7�51(38)

0�21(7)

4�09(8)

10�4(9)

2�47(8)

6�10(8)

0�05(7)

0�01(2)

1�14

91�0

Mg59

Cpx(4)

47�3(12)

0�96(24)

4�91(72)

6�65(57)

0�20(14)

13�4(6)

23�4(7)

0�24(6)

0�22(8)

0�05(7)

0�01(3)

n.d.

97�3

En39Wo50

Phl(2)

36�6(4)

2�35(2)

15�7(1)

7�95(1)

0�05(7)

19�6(4)

0�15(13)

0�24(3)

9�85(3)

0�04(5)

0�03(5)

n.d.

92�5

5Phlog81

4-2

Gl(2)

51�7(11)

0�82(4)

19�2(8)

6�49(17)

0�15(5)

1�90(2)

9�55(13)

2�29(100)

6�61(126)

0�10(6)

0�09(11)

1�11

489�6

Mg44

Cpx(1)

46�6

0�89

5�40

7�85

0�17

12�8

24�4

0�18

0�07

0�00

0�00

n.d.

98�3

En37Wo51

Phl(1)

37�0

2�56

15�8

8�50

0�11

18�8

0�47

0�52

9�75

0�00

0�00

n.d.

93�6

Phlog80

4-3

Gl3()

50�4(11)

0�92(7)

18�7(5)

7�31(68)

0�19(4)

2�96(40)

9�12(28)

2�40(81)

6�88(141)

0�03(4)

0�10(7)

1�08

490�7

Mg51

Cpx(2)

48�4(11)

0�82(10)

4�90(55)

6�74(101)

0�27(11)

13�6(5)

23�6(2)

0�30(3)

0�16(6)

0�07(9)

0�07(0)

n.d.

98�9

En40Wo49

Phl(4)

36�4(4)

2�96(18)

15�9(2)

9�32(50)

0�14(3)

18�8(4)

0�21(7)

0�32(5)

9�74(25)

0�05(4)

0�05(9)

n.d.

93�8

Phlog78

VES9

8-1

Gl(3)

52�1(3)

0�98(7)

18�3(3)

6�66(34)

0�09(11)

2�2(31)

8�57(4)

2�72(21)

7�22(33)

0�02(3)

0�05(9)

1�14

91�1

Mg46

Cpx(2)

49�3(8)

0�77(3)

4�55(2)

7�33(23)

0�20(7)

13�4(1)

23�9(0)

0�21(3)

0�05(2)

0�11(1)

0�04(2)

n.d.

99�9

En39Wo50

Phl(2)

36�9(4)

2�56(9)

15�8(0)

10�1(0)

0�07(0)

18�1(3)

0�07(2)

0�24(5)

9�71(17)

0�01(1)

0�00(0)

n.d.

93�5

Phlog76

9-1

Gl(3)

53�4(3)

0�75(6)

18�7(2)

6�10(5)

0�05(4)

1�93(9)

7�43(6)

2�89(12)

7�64(1)

0�01(2)

0�00(0)

1�09

491�5

Mg45

Cpx(1)

45�9

1�02

6�13

8�65

0�30

11�6

24�6

0�34

0�25

0�02

0�13

n.d.

98�9

En34Wo52

Phl(1)

36�6

2�67

16�0

11�7

0�05

17�2

0�33

0�36

10�0

0�00

0�00

n.d.

94�9

Phlog72

VS96-54A

6-1

Gl(5)

48�3(3)

1�15(12)

15�9(2)

8�01(22)

0�17(9)

5�76(25)

11�9(4)

2�02(10)

5�63(19)

0�01(3)

0�04(5)

1�07

493�9

Mg69

Cpx(3)

50�9(6)

0�64(7)

3�14(14)

5�30(46)

0�13(11)

15�3(5)

23�2(4)

0�13(5)

0�10(3)

0�15(11)

0�02(2)

n.d.

99�0

En44Wo48

VES9

6-2

Gl(5)

49�4(5)

1�03(7)

15�5(3)

7�64(30)

0�15(5)

5�39(15)

11�2(2)

2�10(7)

6�47(12)

0�01(1)

0�05(8)

1�06

494�2

Mg69

Cpx(4)

50�7(15)

0�61(31)

3�46(124)

4�85(132)

0�10(9)

14�9(9)

24�1(4)

0�13(1)

0�07(4)

0�21(6)

0�04(5)

n.d.

99�0

En43Wo50

(continued

)

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2291

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table6:

Continued

Charge

Phase

SiO

2TiO

2Al 2O3

FeO

MnO

MgO

CaO

Na 2O

K2O

Cr 2O3

NiO

P2O5

Total

Composition

10-1

Gl(1)

51�4

0�87

18�4

6�54

0�22

2�51

8�25

2�94

7�8

0�00

0�02

1�07

493�8

Mg48

Cpx(1)

44�4

0�98

8�43

9�82

0�27

10�8

24�0

0�31

0�21

0�24

0�06

n.d.

99�5

En32Wo51

Phl(1)

36�73

3�23

16�37

9�84

0�04

17�51

0�13

0�33

9�76

0�10

0�00

n.d.

94�0

Phlog76

11-1

Gl(2)

54�4(3)

0�60(6)

19�5(2)

5�21(5)

0�12(4)

1�34(9)

5�72(6)

3�29(12)

8�65(1)

0�10(2)

0�07(0)

1�06

494�6

Mg38

Cpx(1)

45�89

1�23

7�30

9�48

0�12

10�9

23�9

0�35

0�29

0�21

0�00

n.d.

99�7

En33Wo51

Phl(1)

36�1

3�1

16�0

11�8

0�3

15�9

0�9

0�3

9�5

0�1

0�0

n.d.

93�9

Phlog71

VS96-54A

7-1

Gl(6)

48�1(3)

1�17(6)

16�1(2)

8�57(32)

0�17(14)

5�45(13)

11�4(2)

2�17(6)

5�72(11)

0�02(2)

0�05(6)

1�05

495�0

Mg61

Ol(7)

40�3(5)

0�03(3)

0�03(3)

13�6(5)

0�28(9)

44�8(4)

0�29(18)

0�02(2)

0�02(2)

0�02(3)

0�19(13)

n.d.

99�5

Fo86

Cpx(2)

49�7(1)

0�70(6)

5�69(19)

5�67(11)

0�20(4)

13�3(4)

22�3(6)

0�40(0)

0�78(11)

0�07(9)

0�03(4)

n.d.

98�7

En41Wo49

VES9

7-2

Gl(5)

49�0(4)

1�10(4)

16�0(2)

7�95(34)

0�18(12)

5�24(15)

11�0(4)

2�15(9)

6�33(11)

0�03(4)

0�00(0)

1�05

495�4

Mg62

Cpx(3)

52�2(5)

0�52(4)

2�29(21)

4�58(30)

0�14(8)

15�9(3)

24�0(2)

0�11(3)

0�10(2)

0�00(0)

0�10(9)

n.d.

99�9

En45Wo48

High-pressure

H2O-,

CO2-bearingfluid-presentexperim

ents

VES9

18-1(þ

Ol)

Gl(10)

49�1(1)

0�88(10)

12�7(1)

8�07(17)

0�16(9)

11�3(3)

11�5(2)

1�46(7)

4�32(15)

0�01(5)

0�02(11)

0�50(12)

95�1

Mg85

Olco

re42�6

0�10

0�00

3�90

0�03

53�6

0�18

0�02

0�00

0�03

0�14

0�00

100�6

Fo96

Olrim

43�1

0�00

0�00

2�32

0�21

54�7

0�31

0�00

0�00

0�00

0�49

0�07

101�1

Fo98

VS96-54A

17-2(þ

Ol)

Gl(9)

48�7(3)

0�96(16)

11�6(2)

8�13(33)

0�15(9)

11�3(2)

13�6(1)

1�43(6)

3�70(9)

0�02(3)

0�00(14)

0�45(13)

95�8

Mg85

Olco

re42�9

0�00

0�04

8�61

0�15

50�3

0�12

0�04

0�05

0�03

0�53

0�04

101�9

Fo91

Olrim

42�7

0�03

0�02

5�78

0�01

51�9

0�24

0�04

0�00

0�01

0�54

0�00

101�2

Fo94

5-3(þ

Ol)

Gl(5)

49�3(4)

1�04(6)

13�8(1)

6�55(10)

0�13(8)

8�11(17)

13�6(1)

1�72(6)

4�68(11)

0�05(6)

0�02(2)

0�97(6)

94�1

Mg77

Ol(1)

40�4

0�00

0 �01

8�57

0�09

49�06

0�1

0�00

0�05

0�09

0�15

n.d.

98�5

Fo91

Cpx(3)

51�4(4)

0�39(13)

2�07(45)

3�06(42)

0�03(4)

17�0(2)

24�2(4)

0�12(4)

0�04(2)

0�21(12)

0�12(12)

n.d.

98�7

En47Wo48

13-1

Gl(5)

48�9(3)

1�04(6)

16�9(1)

7�24(25)

0�16(7)

5�45(11)

10�1(2)

2�33(7)

6�63(19)

0�03(8)

0�01(1)

1�19(7)

97�2

Mg67

Cpx(1)

47�4

0�84

5�79

6�65

0�09

14�5

22�7

0�16

0�03

0�09

0�00

0�00

98�3

En42Wo47

VES9

13-2

Gl(6)

49�6(4)

1�09(7)

17�0(2)

6�78(28)

0�12(9)

5�05(10)

9�60(17)

2�46(12)

7�06(15)

0�03(5)

0�04(6)

1�17(4)

97�5

Mg65

Cpx(1)

47�4

0�80

6�37

7�56

0�00

14�5

22�8

0�25

0�20

0�10

0�05

0�17

100�2

En41Wo47

(continued

)

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2292

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table6:

Continued

Charge

Phase

SiO

2TiO

2Al 2O3

FeO

MnO

MgO

CaO

Na 2O

K2O

Cr 2O3

NiO

P2O5

Total

Composition

VS96-54A

15-1

Gl(5)

48�6(2)

1�17(12)

15�8(2)

7�41(27)

0�12(7)

6�44(17)

11�8(3)

2�02(8)

5�48(16)

0�01(3)

0�05(8)

1�00(4)

96�6

Mg76

Cpx(1)

46�5

0�98

5�72

8�22

0�19

14�4

22�7

0�22

0�15

0�37

0�03

0�00

99�5

En41Wo46

15-2

Gl(5)

48�9(1)

1�16(3)

15�9(2)

7�36(32)

0�16(9)

6�06(7)

11�3(2)

2�17(7)

5�87(7)

0�00(1)

0�01(2)

1�06(9)

96�8

Mg74

Cpx(1)

45�7

1�34

7�34

9�16

0�00

13�3

23�0

0�23

0�11

0�47

0�07

0�11

100�8

En38Wo47

15-3

Gl(5)

48�3(2)

1�12(13)

15�5(1)

7�82(30)

0�13(4)

6�55(13)

11�7(1)

2�42(6)

5�37(12)

0�01(2)

0�03(4)

1�02(5)

96�4

Mg76

Cpx(1)

49�4

0�60

4�47

5�83

0�00

15�9

23�6

0�27

0�16

0�20

0�15

0�14

100�7

En44Wo47

15-4

Gl(5)

48�5(2)

1�04(7)

15�9(2)

7�59(30)

0�11(7)

6�27(12)

11�3(2)

2�29(15)

5�77(12)

0�06(5)

0�04(4)

1�09(9)

96�5

Mg75

Cpx(1)

47�0

0�99

7�50

7�44

0�29

13�8

2 2�5

0�39

0�34

0�45

0�23

0�18

101�1

En40Wo47

VES9

15-5

Gl(7)

49�3(4)

1�02(9)

15�9(2)

7�33(40)

0�11(5)

5�86(15)

10�8(3)

2�32(13)

6�29(15)

0�00(1)

0�02(3)

1�06(9)

96�8

Mg73

Cpx(3)

48�3(2)

0�72(15)

4�42(16)

6�26(33)

0�13(7)

15�4(3)

23�0(3)

0�23(2)

0�14(3)

0�30(4)

0�06(10)

0�09(16)

99�2

En44Wo47

12-2

Gl(4)

50�8(4)

1�01(5)

17�9(4)

7�39(20)

0�25(8)

3�53(5)

7�86(23)

3�17(12)

6�58(21)

0�00(0)

0�06(6)

1�37(4)

97�6

Mg56

Cpx(3)

45�9(9)

0�93(7)

7�56(9)

8�24(44)

0�14(4)

13�6(3)

22�5(2)

0�37(1)

0�19(17)

0�03(2)

0�04(7)

0�40(8)

99�8

En40Wo47

Lc(3)

54�8(2)

0�11(4)

22�0(3)

0�82(10)

0�01(1)

0�05(5)

0�07(10)

0�58(3)

20�1(1)

0�01(2)

0�03(5)

0�03(4)

98�6

Lc97

12-3

Gl(4)

50�5(4)

1�09(3)

17�7(1)

7�83(14)

0�15(13)

3�56(8)

7�95(15)

3�36(13)

6�44(13)

0�04(6)

0�00(0)

1�39(5)

96�5

Mg61

Cpx(2)

46�8(11)

0�94(3)

7�31(81)

7�65(84)

0�11(4)

13�5(4)

21�9(7)

0�38(0)

0�52(15)

0�00(0)

0�01(1)

0 �31(16)

99�5

En40Wo47

Lc(2)

54�6(1)

0�07(2)

21�7(3)

0�73(15)

0�00(0)

0�28(29)

0�55(55)

0�58(2)

19�8(5)

0�07(2)

0�02(3)

0�01(1)

98�4

Lc96

12-5

Gl(5)

50�6(5)

1�07(5)

18�1(2)

7�03(16)

0�13(12)

3�72(11)

7�55(21)

3�03(12)

7�38(12)

0�05(6)

0�00(0)

1�27(11)

95�8

Mg66

Cpx(2)

44�7(4)

1�10(9)

7�83(23)

8�68(28)

0�12(0)

13�6(1)

22�8(6)

0�37(3)

0�19(0)

0�00(0)

0�10(5)

0�45(1)

100�0

En39Wo47

Lc(2)

54�4(2)

0�08(2)

22�0(1)

0�78(1)

0�00(0)

0�03(2)

0�21(23)

0�48(1)

20�3(3)

0�00(0)

0�03(4)

0�17(22)

98�5

Lc97

VS96-54A

16-1

Gl(5)

48�4(3)

1�07(10)

14�7(1)

7�85(29)

0�17(7)

7�32(11)

12�6(2)

1�93(8)

4�97(11)

0�01(1)

0�07(8)

0�91(9)

96�6

Mg73

Cpx(2)

46�4(9)

0�78(13)

6�38(103)

7�79(118)

0�13(14)

14�4(5)

22�7(4)

0�28(2)

0�27(21)

0�40(2)

0�23(29)

0�05(7)

99�8

En41Wo47

16-2

Gl(7)

48�1(3)

1�09(7)

14�9(2)

8�10(37)

0�12(7)

7�11(10)

12�4(2)

2�15(15)

5�05(13)

0�00(0)

0�02(3)

1�01(7)

96�6

Mg72

Cpx(1)

47�4

0�82

6�63

7�49

0 �22

14�84

22�5

0�33

0�48

0�39

0�01

0�09

101�2

En42Wo46

16-3

Gl(5)

48�8(2)

1�18(8)

15�7(1)

7�63(16)

0�17(8)

6�32(17)

11�4(4)

2�10(9)

5�65(20)

0�02(3)

0�04(4)

1�04(11)

96�9

Mg69

Cpx(1)

48�2

0�80

5�26

6�67

0�00

15�0

23�4

0�18

0�18

0�19

0�00

0�00

99�9

En42Wo47

VES9

16-4

Gl(5)

49�3(3)

0�99(8)

15�8(2)

7�13(30)

0�09(7)

6�02(17)

11�0(2)

2�12(10)

6�42(14)

0�08(7)

0�04(4)

1�01(5)

97�0

Mg69

Cpx(1)

48� 1

0�60

4�98

6�20

0�23

14�7

23�5

0�17

0�15

0�43

0�01

0�00

99�0

En42Wo48

(continued

)

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2293

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Table6:

Continued

Charge

Phase

SiO

2TiO

2Al 2O3

FeO

MnO

MgO

CaO

Na 2O

K2O

Cr 2O3

NiO

P2O5

Total

Composition

0�1MPaexperim

ents

VS96-54A

3Gl(7)

49�1(4)

1�03(8)

13�0(2)

6�80(32)

0�19(4)

8�18(12)

15�4(3)

1�58(11)

3�76(12)

0�03(7)

0�06(6)

0�82(10)

98�5

Mg73

4Gl(6)

49�5(3)

1�04(5)

13�6(2)

6�43(31)

0�14(7)

7�83(13)

14�8(3)

1�66(6)

4�20(12)

0�05(6)

0�02(5)

0�83(4)

98�8

Mg73

Cpx(3)

51�2(6)

0�47(5)

3�72(157)

3�84(12)

0�11(10)

17�6(14)

23�8(14)

0�16(24)

0�15(64)

0�46(16)

0�00(6)

0�00(9)

101�5

En48Wo46

6Gl(3)

49�3(2)

1�07(4)

14�7(0)

6�96(10)

0�08(7)

6�77(18)

13�2(1)

1�88(10)

5�08(19)

0�01(2)

0�03(5)

0�91(8)

98�2

Mg69

Cpx(8)

46�3(5)

0�82(10)

6�65(101)

7�75(45)

0�15(10)

14�9(9)

23�0(14)

0�16(20)

0�15(61)

0�09(5)

0�05(6)

0�36(8)

100�4

En42Wo46

7Gl(5)

48�4(3)

1�16(5)

15�7(1)

7�11(19)

0�16(6)

6�10(6)

12�9(3)

2�38(5)

5�07(26)

0�01(3)

0�02(4)

0�96(4)

98�4

Mg66

Cpx(5)

50�2(10)

0�89(26)

4�84(139)

4�00(35)

0�12(12)

16�4(8)

23�6(4)

0�17(10)

0�15(20)

0�13(8)

0�05(6)

0�37(9)

100�9

En46Wo48

Lc(2)

54�6(1)

0�03(8)

22�0(1)

0�00(4)

0 �11(9)

0�10(52)

0�16(78)

0�42(18)

20�2(1)

0�00(0)

0�00(0)

0�00(1)

97�6

Lc97

VES9

9Gl(5)

49�5(3)

0�94(4)

14�4(1)

7�11(18)

0�13(7)

6�61(11)

13�3(2)

1�92(10)

5�14(12)

0�04(5)

0�01(1)

0�89(7)

98�6

Mg69

8Gl(7)

50�3(3)

0�96(6)

14�7(2)

6�18(29)

0�16(9)

6�66(17)

13�3(4)

2�13(5)

4�84(14)

0�02(2)

0�01(3)

0�73(9)

95�8

Mg72

Cpx(1)

52�6

0�51

3�11

2�38

0�00

17�9

22�4

0�24

0�15

0�52

0�00

0�25

100�3

En51Wo46

6Gl(4)

49�2(3)

1�02(4)

14�6(7)

7�24(28)

0�20(7)

6�61(21)

13�4(2)

1�76(9)

4�94(4)

0�01(2)

0�04(7)

0�90(10)

96�3

Mg68

Cpx(1)

48�8

0�81

4�32

4�26

0�11

16�3

22�8

0�19

0�15

0�27

0�16

0�30

98�5

En47Wo47

5Gl(6)

50�1(7)

1�04(7)

15�7(3)

7�04(60)

0�16(11)

5�63(22)

11�5(3)

2�19(9)

5�51(20)

0�08(8)

0�03(7)

1�02(8)

97�6

Mg65

Cpx(3)

43�6(4)

0�89(11)

7�09(12)

9�56(25)

0�20(4)

13�3(15)

22�9(21)

0�26(22)

0�15(92)

0�05(7)

0 �00(1)

0�48(12)

98�5

En38Wo47

Lc(3)

55�0(4)

0�16(0)

21�3(3)

1�43(48)

0�00(0)

0�10(30)

0�26(71)

0�47(8)

19�7(5)

0�04(5)

0�01(3)

0�04(6)

98�5

Lc97

1Glass

analyses

norm

alized

to100%

anhyd

rous,

withallFeas

FeO

.Unnorm

alized

totalis

reported

.2Number

ofmicroprobean

alyses.

3Onestan

darddeviationin

term

sofleastunitcited.

4P2O5notan

alyzed

;P2O5co

ncentrationassumed

equal

to1wt%

before

norm

alization.

5Fco

ncentrationbelow

detection.

Mg¼

Mg#calculatedas

100�molarMgO/(MgOþFeO

).(See

text

forcalculationofFeO

.)En¼100�at.Mg/(MgþFeþCa),Wo¼100�at.Ca/(M

gþFeþCa)

inpyroxene,

calculatedwithFe¼FeO

t.Fs(notgiven

)¼100–En–Wo.Fo¼100�at.Mg/(MgþFe)

inolivine,

calculatedwithFe¼FeO

t.Phlog¼100�at.Mg/

(MgþFe)

inphlogopite,

calculatedwithFe¼FeO

t.Lc¼100�at.K/(NaþK)in

leucite.Gl,glass;Cpx,

clinopyroxene;

Ol,olivine;

Phl,phlogopite;

Lc,

leucite;n.d.,

notdetermined

.

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2294

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

olivine-added charges (three fluid-absent and three fluid-present), olivine dissolution occurred, either alone or sim-ultaneously with clinopyroxene crystallization. Althoughno reversals have been performed, the following lines ofevidence indicate a close approach toward equilibrium inour experiments.

(1) Experimental durations were 2^23 h (13�5 h on aver-age, 5^10 h for the olivine-added charges), sufficientfor equilibrium crystallization of hydrous basalticmelts (Sisson & Grove, 1993; Barclay & Carmichael,2004; Di Carlo et al., 2006; Pichavant & Macdonald,2007). The 0�1MPa experimental cycles each lastedfor relatively short durations (3 h; see above), yetphase assemblages and proportions of crystals aremutually consistent and evolve predictably with tem-perature (Table 5).

(2) Crystal morphologies are euhedral, equant or tabular.The distribution of crystals in both the crystallizationand dissolution^crystallization charges is homoge-neous; no crystal settling was recognized.The difficul-ties encountered in quenching the charges (seeabove) can be taken as an indication for fast crystalnucleation and growth in our experiments.

(3) Sums of residuals from mass-balance calculations(crystallization charges only) range from values50�1to44 (Tables 3^5). Nineteen charges have �R250�5,indicating excellent balance of silicate components be-tween coexisting phases. The highest �R2 values (42)correspond to charges with Fe balance problems, be-cause of either Fe loss (the five charges with Fe loss410% relative) or Fe gain (because of analytical over-estimation; four charges have Fe gains45 and510%relative).

(4) Glasses are chemically homogeneous. In the dissol-ution^crystallization charges, no compositional gradi-ent was detected around the dissolving olivinecrystals.

(5) Compositions of clinopyroxenes are generally homo-geneous. However, in charges 3-4, 2-1, 2-3, 1b-1 and1b-3, standard deviations for clinopyroxene analysesexceed analytical errors for SiO2, Al2O3, FeO andMgO (Table 6). SEM observations reveal core^rimzonation. The dispersion of clinopyroxene^liquid Fe^Mg exchange coefficients reflects these small chemicalheterogeneities although, globally, the dataset suggestscrystal^liquid equilibrium (see below).

(6) Olivine crystallized spontaneously in charge 7-1 ischemically homogeneous with an Fe^Mg exchangecoefficient (Kd) suggestive of a close approach towardcrystal^liquid equilibrium (see below). In compari-son, residual crystals in olivine-added charges didnot completely equilibrate under the imposed experi-mental conditions. At 11508C, residual olivines haveFo and CaO contents in the same range as the starting

San Carlos olivine crystals. For Fe and Mg, oliv-ine^melt equilibration is suggested on the basisof Fe^Mg Kd values (see below) but, for CaO, the re-sidual olivines maintain low concentrations inheritedfrom the starting crystals. At 12008C, both the Fe^Mg Kd values and the CaO contents suggest equilib-rium between residual olivine rims and melt (seebelow).

Phase equilibriaClinopyroxene, phlogopite, leucite and olivine spontan-eously crystallized in the experiments; apatite was occa-sionally found, but neither plagioclase nor Cr-spinel wasencountered. The most important crystalline phase byfar is clinopyroxene, present in 48 charges out of a totalof 54. The clinopyroxene proportion reaches a maximumof 45wt % in charge 1-6 (Table 3).At 0�1MPa and for an fO2 between NNO and

NNOþ 0�5, phase assemblages and crystallization se-quences are identical for the two compositions investi-gated. Clinopyroxene is the liquidus phase (Table 5).Liquidus temperatures are between 1222^12128C forVS96-54A and 1209^12018C for VES9, significantly lower thanpreviously reported for Vesuvius rocks at 0�1MPa (from�1260 and up to �13308C; Trigila & De Benedetti,1993, fig. 16). In both samples, leucite appears at �11808C,second to clinopyroxene in the crystallization sequence.Despite experiments being performed at temperaturesbelow 11808C and as low as �11258C (Table 5), the appear-ance of the third crystalline phase of the 0�1MPa sequencewas not observed.At high pressures, the phase relations were not com-

pletely explored for each pressure because, in the rangeinvestigated (50^200MPa), the influence of pressure isminor, as detailed below.The 200MPaT^H2O in melt sec-tion for VS96-54A (Fig. 2) illustrates the phase equilibria,which are closely similar forVES9 (not shown). ForVS96-54A, clinopyroxene is the liquidus phase. At 11508C, theliquidus is bracketed by charges 3-1 and 3-2 (Table 3), andthe same liquidus phase assemblage (clinopyroxeneþliquid) persists at 100 and 50MPa (charges 15-1 to 15-4and 16-1 to 16-3 respectively, Table 4). At 11008C, the stablephase assemblage is also clinopyroxeneþ liquid (Fig. 2).Runs 1 and 2, both performed at 11008C, 200MPa(Table 3), show that raising fO2 by 1^2 log units increasesthe proportion of clinopyroxene in the charge but leavesthe stable phase assemblage unchanged. Contours of clino-pyroxene proportions (Fig. 2) exclude the oxidized run 2.The stable phase assemblage is also clinopyroxeneþ liquidat 100MPa (charge 6-1, Table 3) whereas in the 50MPa,11008C charge (7-1, Table 3), olivine occurs togetherwith clinopyroxene. At 10508C for H2O-rich conditions,phlogopite joins clinopyroxene in the crystallization se-quence (Fig. 2). Phlogopite-bearing charges have melt

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2295

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

H2O concentrations between 5�1 and 6�0wt % (run 4,Table 3). For near H2O-saturation, the liquidus tempera-ture for VS96-54A (Fig. 2) is �11208C, about 1008C lowerthan at 0�1MPa.For VES9, clinopyroxene is also the liquidus phase at

200MPa. The 11508C liquidus is half-bracketed by theclinopyroxeneþ liquid charge 13-2 (Table 4), and thesame phase assemblage persists at 100 and 50MPa(charges 15-5 and 16-4 respectively, Table 4). At 11008C,clinopyroxene also crystallizes on the 200MPa liquidus(run 1b, Table 3). The 100 MPa charges demonstrates ashift from clinopyroxeneþ liquid to clinopyroxeneþleuciteþ liquid when decreasing the melt H2O contentbelow �3wt % H2O (6-2, 12-5, 12-3, 12-2, Tables 3 and 4).Clinopyroxeneþ liquid is the stable phase assemblage at50 MPa (charge 7-2, Table 3). At 1050 and 10008C forH2O-rich conditions, phlogopite joins clinopyroxene bothat 200 and 100MPa (charges 8-1, 9-1 and 10-1, 11-1 respect-ively, Table 3).In summary, the high-pressure near-liquidus phase equi-

libria are deceptively simple. Clinopyroxene is the liquidusphase for the two starting samples and clinopyroxeneþliquid is the stable phase assemblage for a wide range ofP^T^H2O melt^fO2 conditions. TheVES9 charges (as nohigh-pressure VS96-54A charge crystallized leucite and,therefore, the location of the leucite saturation curve inFig. 2 is estimated) demonstrate that leucite comes second

in the sequence, crystallizing under H2O-poor concentra-tions (53wt % H2O in melt), consistent with the 0�1MPaphase equilibria. For H2O in melt 43wt %, phlogopitereplaces leucite as the second phase of the sequence. Therespective positions of the liquidus and phlogopite satur-ation curves (Fig. 2) suggest that clinopyroxene persists asthe liquidus phase until H2O-saturation (�6wt % H2O).Therefore, our data do not suggest the possibility thatphlogopite can be stable on the liquidus (Esperanc� a &Holloway, 1987; Righter & Carmichael, 1996). Finally, oliv-ine occurred as a stable phase in only one charge (7-1,Table 3), and no olivine stability field can be delineated.

Composition of crystalline phasesClinopyroxenes are calcic, with 45^52% Wo, 32^51% Enand 4^16% Fs, and Mg# (calculated with total Fe asFe2þ) of 66^93. They contain 1�68^8�43wt % Al2O3, 0�33^1�34wt % TiO2 and 0^0�30wt % Cr2O3 (Table 6). Thereis no systematic difference between cpx crystallized fromeither VS96-54A or VES9. Clinopyroxene chemistrystrongly correlates with the amount of cpx crystals presentin the charge: the higher the amount of cpx, the lower theEn, and the higher the Fs and Al2O3 and TiO2 contents.Upon crystallization, atomic AlþTiþNa increases atthe expense of atomic CaþFeþMg in cpx and so the pro-portion of quadrilateral cpx components decreases.Clinopyroxenes coexisting with olivine (olivine-addedcharges) have the lowest atomic AlþTiþNa and thehighest atomic CaþFeþMg. In detail, compositionsof experimental clinopyroxenes divide into two groups(Fig. 3). In the fluid-absent charges, Wo increases alongwith increasing Al2O3 (trend 1, Fig. 3), whereas in boththe 0�1MPa and the fluid-present charges, Wo decreaseswith increasing Al2O3 (trend 2, Fig. 3). This contrasted be-haviour correlates with systematic differences in meltH2O concentrations, which are higher in the fluid-absentthan in the fluid-present and 0�1MPa charges (see below).Natural clinopyroxenes fromVesuvius follow the H2O-richtrend (Fig. 3). Clinopyroxene contains substantial amountsof atomic Fe3þ (estimated from structural formulae), pro-gressively increasing with atomic AlþTiþFeþNa.However, no correlation appears between atomic Fe3þ incpx and experimental fO2. The Fe^Mg clinopyroxene^liquid exchange coefficient (Kd

Cpx^L) calculated with totalFe as Fe2þ in both phases is 0�30� 0�11 (average of all cpx-bearing charges from this study, n¼ 46). If the most ex-treme values (fluid-present charges 15-1, 15-2, 15-4, 16-1 and16-2, Table 4; 0�1 MPa charges 5, 6 (VS96-54A and 8,Table 5) are filtered out, the average Kd

Cpx^L becomes0�27�0�06, in excellent agreement with experimentaldata for hydrous basaltic compositions (0�26^0�31, Sisson& Grove,1993; 0�29� 0�08, Pichavant &Macdonald, 2007).Olivine crystallized in charge 7-1 is Fo85�5 and it contains

0�29wt % CaO (Fig. 4; Table 6). In the olivine-added ex-periments, compositions of residual crystals divide into

H2O in melt (wt%)

T (

°C)

4 5 630 1 2

1000

1040

1080

1120

1160

1200 VS96-54A200-215 MPa

Cpx+Lc+Liq

Liq

Liq+Vap

Cpx+Phlog+Liq

Cpx

Phlog

Liq

Cpx+Liq

10%

20%35%

Fig. 2. T^H2O in melt phase diagram forVS96-54A showing experi-mental data points (squares), saturation curves (dashed when esti-mated), phase assemblages and stability fields. The diagramcombines experimental results from fluid-absent (runs 1, 2, 3, 4,Table 3) and fluid-present experiments (run 13, Table 4). Total pres-sures range between 200 and 215MPa. Redox conditions range be-tween NNOþ 0�5 and NNOþ1�5, except for run 2 (11008C), whichis more oxidized (Tables 3 and 4). Fine dotted lines are contours ofclinopyroxene proportions (in wt %) constructed excluding datafrom the oxidized run 2. Liq, liquid (silicate melt); Cpx, clinopyrox-ene; Phlog, phlogopite; Lc, leucite; Vap, vapour phase. It should benoted that, in the diagram, the vapour is an H2O-rich, CO2-freephase and thus charge 13-1 (which contains a CO2-rich fluid) liesoutside the LiqþVap field.

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2296

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

two groups. At 11508C, olivines have homogeneous Fo andCaO contents (respectively 90�9^91�3 and 0�10^0�18wt %),close to the San Carlos olivine starting crystals (Fig. 4).The olivine^liquid Fe^Mg exchange coefficient (Kd

Ol^L)calculated by averaging the data for these four chargesand charge 7-1 is 0�34� 0�05, in agreement with literaturevalues for comparable experimental conditions and com-positions (0�27^0�33, Sisson & Grove, 1993; 0�25^0�39,Righter & Carmichael, 1996; 0�32�0�04, Pichavant &Macdonald, 2007). This Kd

Ol^L uses melt FeO contents cal-culated from glass compositions (Kress & Carmichael,1991) for the P,Tand fO2 specific to each charge (Tables 3and 4). At 12008C, residual olivines are chemically zonedin Fe, Mg and Ca. Deviations from the initial olivine com-position are significant. Rims reach Fo of 94�1 and 97�7and CaO of 0�24 and 0�31wt % for 17-2 and 18-1 respect-ively (Fig. 4). The olivine rim in charge 17-2 yields a Kd

Ol^

L of 0�35 whereas the KdOl^L for charge 18-1 (0�13, Table 4)

is clearly lower than the range of equilibrium values. It isworth noting that 18-1 is the most oxidized charge in ourdataset and it is possible that the calibration of Kress &Carmichael (1991) does not perform well for low-FeOt,high Fe3þ/Fe2þ melts such as 18-1 (Table 6).

Leucite compositions were determined in five out of atotal of seven leucite-bearing charges. They are tightlygrouped and close to the leucite end-member (Table 6).There is no clear dependence of leucite composition onthe type of experiment: leucites in high-pressure fluid-pre-sent (run 12, Table 4) and 0�1MPa (Table 5) charges arecompositionally identical. There is also no apparent de-pendence of leucite composition on the composition of thestarting material.Experimental phlogopites have Mg# (calculated with

total Fe as Fe2þ) between 70 and 81, atomic K/(KþNa)between 0�93 and 0�96, and TiO2 contents between 2�35and 3�23wt % (Fig. 5). The F concentration in phlogopitefrom charge 4-1 is below detection. There is no marked dif-ference between micas crystallized from VS96-54A orVES9. Phlogopite becomes more magnesian with increas-ing temperature. AlVI (0�505^0�665, 22 O basis), whichmakes up 9^12% of the octahedral cations, slightly de-creases with increasing temperature. The TiO2 partitioncoefficient between phlogopite and melt (wt % TiO2

in phlogopite/wt % TiO2 in glass; data in Table 6) de-creases from 3�57^5�23 (10008C) to 2�28^3�71 (10508C).If the equation calibrated for lamprophyres is used(Righter & Carmichael, 1996), experimental phlogopitesfrom this study have atomic Fe3þ/�Fe between 0�02 and0�15 (Fe2O3 between 0�2 and 1�7wt %), which leads to aphlogopite^liquid Fe^Mg exchange coefficient (Kd

Phlog^L)of 0�28�0�05 forVesuvius magmas.

VS96-54A VES90.1 MPa

Fluid-absentFluid-present

Vesuvius cpx

+ Ol

H2O < 2 wt%

H2O > 2 wt%

(1)

(2)

Wo50 52 54484644

0

2

4

6

8

10

12 A

l 2O

3 (w

t%)

Fig 3. Al2O3 vsWo (calculated with total Fe as Fe2þ) for the experi-mental clinopyroxenes of this study (Table 6). The data are distin-guished by type of experiment (0�1MPa, fluid-absent, fluid-present),starting material (either VS96-54A or VES9) and addition of olivine(þ Ol). (See Tables 3^5 for details of experimental conditions.) Thegrouping of the fluid-absent H2O-rich (H2O in melt42wt %) ex-perimental data points defines a positive correlation between wt %Al2O3 and Wo [trend (1) indicated by an arrow]. In contrast, boththe 0�1MPa and fluid-present (H2O in melt52wt %) experimentaldata define a negative Al2O3^Wo correlation [trend (2) indicated byan arrow]. Clinopyroxenes from the three olivine-added chargeshave the lowest Al2O3 contents of the experimental dataset. Thegrey field marks the compositional domain for naturalVesuvius clino-pyroxenes (both phenocrysts in lavas and cumulus phases in clinopyr-oxenite nodules). Data for the natural clinopyroxenes are from Rosi& Santacroce (1983), Santacroce et al. (1993), Trigila & De Benedetti(1993), Cioni et al. (1995, 1998), Marianelli et al. (1995, 1999, 2005),Cioni (2000), Fulignati et al. (2000, 2004a), Rolandi et al. (2004),Dallai et al. (2011) and this study (Table 2).

Fo

CaO

(w

t%)

VS96-54A

VES9

Fluid-absentFluid-absent + OlFluid-present + Ol

Fluid-absent + OlFluid-present + Ol

LavasCumulates

0

0.4

0.2

0.6

0.8

1.0

50 60 70 80 90 100

rr

cc

Fig. 4. Variation of CaO vs Fo in experimental olivines of this studyand comparison with olivine compositions in Vesuvius lavas and cu-mulates. Experimental data are given in Table 6. Natural compos-itions (both phenocrysts in lavas and cumulus phases in duniticnodules) are from Trigila & De Benedetti (1993), Cioni et al. (1995),Marianelli et al. (1995, 1999, 2005), Fulignati et al. (2000), Dallai et al.(2011) and this study (Table 2). The data are distinguished by type ofexperiment (either fluid-absent or fluid-present), starting material(either VS96-54A or VES9) and addition of olivine. (See Tables 3^5for details of experimental conditions.) For the chemically zoned oliv-ines of the olivine-added charges 17-2 and 18-1, two data points areplotted, one for the core (labelled c) and the other for the rim(labelled r). Core and rim compositions are connected by dashedlines.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2297

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Melt compositionsCompositions of experimental glasses are given inTable 6.It should be noted that, for charges 4-2 and 4-3, glass com-positions have been corrected to account for quenchphlogopite crystallization (see above). This was done byincreasing their glass K2O concentration respectively to6�61 (from 5�4) and 6�88 (from 5�26) wt % (Table 6).Glasses follow well-defined compositional trends, identi-

cal for theVS96-54A andVES9 charges. Upon progressivecrystallization and considering MgO, CaO or CaO/Al2O3 as a differentiation index, K2O and Al2O3 stronglyincrease (up to �9 and �20wt % respectively) whereasMgO, CaO and CaO/Al2O3 decrease (down to52,56wt% and �0�3 respectively, Fig. 6). Compositions of residualmelts follow the evolution of Vesuvius whole-rocks (Fig. 6).In the total alkalis^silica (TAS) plot, glass compositionscover most of the Vesuvius differentiation sequence, pro-gressively evolving from trachybasalt to tephrite, phonote-phrite, tephriphonolite and approaching the phonolite

domain (Fig. 1). The experimental trends do not dependon the phase assemblage (i.e. presence of leucite, phlogo-pite or olivine) and they reflect the progressive crystalliza-tion of cpx, the dominant mineral phase in ourexperimental charges. Glasses from the olivine-addedcharges extend the compositional trends toward higherMgO, on the one hand, and lower K2O and Al2O3, onthe other (Fig. 6), consistent with partial dissolution of oliv-ine in these experiments. Primitive melt (glass) inclusionsfrom the 1637^1944 eruptive products evolve along trendssimilar to those for the whole-rocks and experimentalglasses. However, they extend to MgO concentrationsbeyond theVS96-54A andVES9 (whole-rocks and glasses)

0 2 4 6 8 10 12MgO (wt%)

Al 2

O3

(wt%

)K

2O (

wt%

)

0

10

15

20

5

0

2

4

6

8

10

Experimental glassesVS96-54A

VES9VES9 + Ol

VS96-54A + Ol

Whole-rocks

Glass inclusions

VS96-54A

VS96-54A

VES9

VES9

(a)

(b)

Fig. 6. MgO variations diagrams for experimental glasses from thisstudy and comparison with selected natural compositions fromVesuvius (whole-rocks and glass inclusions). (a) K2O vs MgO.(b) Al2O3 vs MgO. Data for the experimental glasses are given inTable 6. Whole-rock data for the AD 472, 1631, 1637^1944, 1906 and1944 eruptions are from Rosi & Santacroce (1983), Rosi et al. (1993),Santacroce et al. (1993), Cioni et al. (1998) and Marianelli et al. (1999,2005). The glass inclusions are for the 1637^1944 period of activity[recalculated compositions from Marianelli et al. (1995, 1999, 2005)].The experimental data points are distinguished by starting material(either VS96-54A or VES9). Glasses from the olivine-added experi-ments are plotted with a different symbol. (See Tables 3^5 for detailsof experimental conditions.) For clarity, the whole-rock analyses(Table 1) of VS96-54A (Marianelli et al., 1999) and VES9 (this study,Table 1) are plotted with a larger symbol.

Mg#70 72 74 76 78 80 82 84

TiO

2 (w

t%)

2

1

3

4

5

0

0.0

0.2

0.4

0.6

at. A

lVI

Lavas

Fluid-absent

VS96-54A VES9

(b)

(a) Cumulates

Fig. 5. (a) TiO2 and (b) atomic AlVI (calculated on a 22 O basis) vsMg# (calculated with total Fe as Fe2þ) for experimental phlogopitesof this study and phlogopite compositions in natural Vesuvius prod-ucts. Experimental data are given in Table 6. Data for the lavas arefrom Rosi & Santacroce (1983), Cioni et al. (1995), Rolandi et al.(2004) and Marianelli et al. (2005), and for the cumulates (clinopyrox-enites) are from Cundari (1982). All the data plotted correspond tofluid-absent experiments and are distinguished by starting material(either VS96-54A or VES9). (See Tables 3^5 for details of the experi-mental conditions.)

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2298

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

data points, in a range that is attained only by glasses fromthe olivine-added charges (Fig. 6).

Volatile concentrationsGlass H2O concentrations are detailed in Tables 3 and 4.There is a good match between the different analyticalresults for H2O (e.g. fluid-absent glasses 3-1, 5-1, 5-2, 5-4;fluid-present glasses 17-2, 5-3).H2O concentrations in theVS96-54A andVES9 charges

overlap. The fluid-absent charges have glass H2O concen-trations between 2�2 and 6�0wt % with a frequency max-imum at 3^4 wt %. For the fluid-present charges, glassH2O concentrations range from 0�7 to 4�2wt % (frequencymaximum at 1^2 wt %). Therefore, one major differencebetween the fluid-absent and fluid-present charges con-cerns their respective glass H2O concentrations, whichare systematically higher in the former than in the latter(see above; Fig. 3). Several fluid-absent charges haveaH2O of unity or close to unity (e.g. 7-1, Table 3), implyingthat a hydrous (CO2-free) fluid phase should in fact bepresent.Glass CO2 concentrations in three near-liquidus (1150^

12008C) 200MPa fluid-present charges range from 1302to 1861ppm (Table 4). These results are consistent withthe CO2 solubility determined at 200MPa and 12008C forVES9 (Fig. 7).

DISCUSSIONValidation of experimental resultsBefore discussing the evolution of primitive Vesuviusmagmas it is first necessary to test the applicability of ourexperimental results. This is done below by comparingour synthetic charges with the mineralogical and petrolo-gical characteristics of Vesuvius eruption products.In our experiments, mineral phases, assemblages and

compositions typical of Vesuvius magmas have beenobtained. Clinopyroxene crystallized either alone or withleucite and phlogopite and also, but very rarely, with oliv-ine. Clinopyroxene is by far the dominant mineralphase in our charges, in broad agreement with the modalcomposition of Vesuvius magmas, although leucite, as aproduct of groundmass crystallization, may exceed thecpx proportion in lavas (e.g. Trigila & De Benedetti, 1993).The typical Vesuvius cpx trend (increase of Wo withAl2O3) is experimentally reproduced for H2O 42wt %in the melt (Fig. 3; see also below and Fig. 8). In the sameway, the compositions of experimental and natural leuciteand phlogopite are in excellent agreement (Table 6;Fig. 5). However, one important natural phase, plagioclase,is missing in our experimental products. Calcic (An50^90)plagioclase occurs in low amounts in mafic rocks such asthe 1906 tephras (51%, Santacroce et al., 1993; Table 2),but reaches proportions 45% in more evolved compos-itions (Trigila & De Benedetti, 1993). Another difference

between natural and experimental products concerns oliv-ine. Although never very abundant (51^2%, Santacroceet al., 1993; Trigila & De Benedetti, 1993; Cioni et al., 1995),olivine is ubiquitous in Vesuvius products (Marianelliet al., 1995, 1999, 2005; Dallai et al., 2011). Natural olivineshave a wide range of Fo contents, including a well-definedFo-rich group (Fo86^91, CaO¼ 0�2^0�5wt %) as well asFo-poor, CaO-rich compositions (Fig. 4). The single com-position (Fo85�5) that spontaneously crystallized inthe VS96-54A charge is at the Fo-poor end of the main(Fo-rich) natural group. More Fo-rich crystals (Fo91^98)occur in the olivine-added charges. Concerning experi-mental liquids, the chemical trends exhibited by whole-rocks and melt inclusions have been closely reproduced(Figs 1 and 6) and so the experiments confirm a geneticrelationship between the various members of the Vesuviusdifferentiation sequence, from trachybasalt and tephriteto phonolite.We conclude that there is a close correspond-ence between experimental and natural phase assem-blages and compositions. However, we also stress theexistence of significant differences between synthetic andnatural products, especially concerning plagioclase andolivine.

H2O in glass (wt%)

CO

2 in

gla

ss (p

pm)

1000

2000

3000

4000

5000

0

Experimental glassesVS96-54A + OlVES9 + OlVES9 (Lesne et al. 2011)

0 1 2 3 4 5

200 MPaexp.

200 MPacalc.

Glass inclusionslavascumulates

300 MPacalc.

Fig. 7. H2O vs CO2 concentrations in experimental glasses andnatural glass inclusions. The experimental data plotted are from thisstudy (1150 and 12008C 200MPa fluid-present olivine-added charges,Table 4) and from Lesne et al. (2011), the latter being a CO2 solubilitymeasurement for VES9 melt at 12008C, 200MPa. Experimental datafrom this study are distinguished by starting material (either VS96-54A orVES9).The experimental results collectively define the concen-tration of H2O and CO2 in primitive Vesuvius melts in equilibriumwith H2O^CO2 fluids at a pressure of 200MPa. The continuouscurve is the 200MPa isobar drawn from the experimental data.The dashed curves are 200 and 300MPa fluid^melt saturation isobarscalculated with the model of Iacono-Marziano et al. (2012). The glassinclusions include data for the AD 79, 1794, 1822, 1872, 1906 and 1944eruption products [recalculated compositions from Marianelli et al.(1995, 1999, 2005), inclusions mostly trapped in olivine and morerarely in clinopyroxene; and from Cioni (2000), inclusions in clinopyr-oxene] and for cumulates [dunitic nodules, Fulignati et al. (2000),inclusions trapped in olivine].

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2299

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Conditions of plagioclase crystallization in Vesuviusmagmas can be inferred from previous studies. In high-pressure experiments on phonotephrite V1 (Fig. 1),plagioclase crystallizes after cpx and leucite in the low-temperature, very low melt H2O content region (Dolfi &Trigila, 1978). At 0�1MPa, it appears third in tephrites tophonotephrites (Trigila & De Benedetti, 1993). Plagioclasealso crystallizes in Vesuvius phonolites (Scaillet et al.,2008). Therefore, plagioclase is typical of relatively evolvedVesuvius rocks, consistent with its modal distribution inthe eruption products. We attribute the absence of plagio-clase in our experiments to the choice of starting productsmore primitive, and of conditions closer to the liquidus,than in previous studies. In any case, plagioclase plays nocritical role in the early evolution of Vesuvius magmas(e.g. Santacroce et al., 1993).The mismatch between olivine in synthetic and natural

products is most certainly not an experimental artefact, asa range of techniques, conditions and starting compos-itions have been explored, both in this and in previousexperimental studies onVesuvius rocks. Olivine was foundneither in phonotephrite V1 at high pressure (Dolfi &Trigila, 1978) nor in tephrites^phonotephrites at 0�1MPa(Trigila & De Benedetti, 1993). This phase is also lackingin experiments on Vesuvius phonolites (Scaillet et al.,2008). We suggest that the near-absence of olivine in our

experimental charges (and its total absence in previous ex-perimental studies onVesuvius) reflects the lack of eruptionproducts representative of primitive olivine-crystallizingmagmas (e.g. Dallai et al., 2011). As previously emphasized,only melt inclusions approach the compositions ofVesuvius primitive magmas (Fig. 6). The composition ofolivine that would be at equilibrium with our startingrocks is Fo89�5 for VS96-54 A and Fo87 for VES9 (calcula-tions performed at NNOþ1, 12008C, 200MPa with Kd

ol^

L¼ 0�34 and glass compositions from Table 6). This is

slightly less Fo-rich than the main Vesuvius olivine group(Fo90^91, Fig. 4; Table 2). We conclude that the rocks usedin this study are not representative of the most primitiveVesuvius liquids. This implies that our experiments onVS96-54A and VES9 simulate a stage that is relativelyadvanced in the magmatic evolution.

Primitive Vesuvius magmatic liquids: com-position and intensive variablesConditions of the early evolution of Vesuvius magmas canbe reconstructed from the olivine-added experiments.These have been performed at 200MPa, under both fluid-absent and fluid-present conditions, and with both VS96-54A and VES9 glass. Two types of phase assemblages,olþ liquid and olþ cpxþ liquid, were obtained. Residualolivines in the 11508C charges have Fo ranging from 90to 91, in the range of the most Fo-rich natural crystals(Fig. 4). At 12008C, olivine Fo contents exceed the naturalrange, a consequence of the elevated temperature andhighly oxidizing fO2 (Table 4). Clinopyroxenes coexistingwith Fo90^91 olivines at 11508C have Mg# (90^91), Wo(48^49%), Al2O3 (�2wt %), TiO2 (50�5wt %) andCr2O3 (up to 0�3wt %) similar to diopsides in Vesuviuseruption products (Rosi & Santacroce, 1983; Santacroceet al., 1993; Marianelli et al., 1995, 2005; Cioni et al., 1998;Fulignati et al., 2000, 2004a; Dallai et al., 2011; Table 2;Figs 3 and 8). Therefore, the 11508C 200MPa olivine-added experiments closely approach the environment ofcrystallization of natural diopsides. The lack of spinelin our experiments is attributed to the low Cr concentra-tions of our starting materials (VS96-54A: �200 ppm,P. Marianelli, personal communication; VES9: 147 ppm,this study).The three 11508C 200MPa olivineþ clinopyroxene-

bearing charges define the composition of cotectic meltsin MgO vs CaO space (Fig. 9a). Charge 7-1, which spon-taneously crystallized olivine, provides a point on the11008C, 50MPa olivineþ clinopyroxene cotectic. As foundin other studies (Pichavant et al., 2009), the effect ofdecreasing pressure is to slightly increase the CaO contentof cotectic melts for a given MgO. Relative to the200MPa cotectic, both the VS96-54A and VES9 startingglasses plot in the cpx field (Fig. 9a), consistent with thephase relations determined in this study (cpx as the liqui-dus phase, Fig. 2). In comparison, the three glasses from

Al 2

O3

(wt%

)

Mg#

4

3

2

1

0

5

90 92 948886

lavascumulates

VS96-54A VES90.1 MPa

Fluid-absentFluid-present

Ol

Ol

Fig. 8. Comparison between the compositions of diopsidic(Mg#485) clinopyroxenes in Vesuvius lavas and cumulates and inthe experiments. Mg# in clinopyroxene is calculated with totalFe¼FeO. Data for experimental clinopyroxenes are given inTable 6.For natural clinopyroxenes, the source of the data is as in Fig. 3. Theexperimental data points are distinguished by the type of experiment(0�1MPa, fluid-absent, fluid-present), starting material (either VS96-54A orVES9) and addition of olivine (Ol). (SeeTables 3^5 for detailsof the experimental conditions.) Crystals from fluid-absent and fluid-present charges plot along the natural trend (phenocrysts in lavasand cumulus phases in dunitic and wehrlitic nodules). In comparison,the 0�1MPa experimental clinopyroxenes have higher Al2O3.Clinopyroxenes fromVS96-54A charges (some olivine-added) are inthe range typical of the most primitive natural diopsides(Mg#490, Al2O352wt %).

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2300

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

the 1150 and 12008C olivine-bearing charges define controllines that extend from each starting glass toward Fo-richolivine (Fig. 9a). They progressively shift away from the200MPa cotectic, plotting in the olivine primary field.Glass inclusions range from an evolved (MgO¼ 4^5wt%, CaO �10wt %) to a more primitive (MgO¼ 9^10wt%, CaO¼12^13wt %) end-member (Fig. 9b). Inclusions

with the lowest MgO plot either on the extension of the11508C, 200MPa cotectic, or above it. The most primitive(MgO¼ 8^10wt %) glasses depart from the cotecticregion, plotting in the olivine primary field.Their distribu-tion follows the trend for the olivine-only VES9 charges,suggesting that the most primitive glass inclusions corres-pond to liquids whose composition is controlled by oliv-ine^liquid equilibrium. Indeed, these most primitiveinclusions are trapped in olivine (Fig. 9b). We concludethat olivine (� Cr-spinel) is the first phase to crystallizein primitiveVesuvius liquids (e.g. Marianelli et al., 1995).The P^T^melt H2O^fO2 conditions of primitive

Vesuvius liquids can be constrained from our experimentalresults. For the crystallization of Fo90^91 olivines, tempera-tures between 1150 and 12008C are inferred from the loca-tion of the glass data points (Fig. 9a) and the Fo contentsof olivines in the olivine-added charges. For comparison,temperatures412008C were proposed for the crystalliza-tion of olivineþCr-spinel in 1906 magmas (Marianelliet al., 1995). Clinopyroxene would join olivine on the liqui-dus at �11508C from our experiments (at 1200^11308Cfrom glass inclusion homogenization temperatures;Marianelli et al., 1995; Cioni et al., 1998). Concerning vola-tiles, glass H2O and CO2 concentrations in the olivine-added fluid-present experiments (2�9^4�2wt % H2O;1300^1900 ppm CO2, Table 4) are within the range ana-lyzed in glass inclusions (1�5^4�5wt % H2O; 600^4500 ppm CO2, Fig. 7). Inspection of Fig. 7 indicates trap-ping pressures for the glass inclusions from �200 to a max-imum of 300MPa. This is much shallower than initiallyestimated (300^620MPa, Marianelli et al., 1999) usingfluid^melt saturation models not necessarily calibrated forVesuvius liquid compositions (Behrens et al., 2009; Lesneet al., 2011), and narrowing the range (200^400MPa) pro-posed in later studies (Marianelli et al., 2005). Table 7gives representative examples of primitiveVesuvius liquids.They are trachybasaltic with K2O �4�5^5�5wt %, K2O/Na2O between 2�75 and 4, MgO¼ 8^9wt %, Mg# be-tween 75 and 80 and CaO/Al2O3 between 0�9 and 0�95.Their redox state is estimated at NNOþ 0�6 toNNOþ 0�9 from olivine^liquid equilibrium (Fig. 10). Thisfairly oxidizing fO2 range is independently confirmed bythe application of the olivine^spinel oxybarometer(Ballhaus et al., 1991). �NNO values ofþ 0�7^1�2 are ob-tained for the 1944, 1906 (Marianelli et al., 1995) and 1637^1944 eruption products (Marianelli et al., 2005), in excel-lent agreement with values derived from olivine^liquidequilibrium (Fig. 10). A similar fO2 range (�NNO ofþ0�4^0�9) is calculated from olivine and spinel in VS97-718 (Table 2; see also below).

Early magmatic evolutionFrom the foregoing discussion, the question arises as to thetransition from the early olivine-crystallizing stage (i.e. olis the liquidus phase), documented only by the presence of

MgO (wt%)

CaO

(wt%

)

12

14

16

10

8

VS96-54A

VES9

1100°C 50 MPa

1150°C 200 MPa

Ol

Cpx

OlExperimental glasses

VS96-54A

VES9 + OlVS96-54A + Ol

Iacono Marziano et al. (2008)

+CaCO3

CaO

(wt%

)

(a)

Ol

1150°C

1200°C

1200°C

(Cpx)

(Ol)

(Ol)

(Ol)

B23

V36 12

14

10

8

64 6 8 10 12

(b)

Glass inclusionsin Olin Cpx

Fig. 9. CaO vs MgO for (a) experimental glasses (Table 6) and (b)glass inclusions [recalculated compositions from Cioni et al. (1995,1998), Marianelli et al. (1995, 1999, 2005), Cioni (2000), Fulignati et al.(2000, 2004b) and R. Cioni (unpublished data)]. The glass inclusiondata are distinguished by type of host crystal (either Ol or Cpx).Experiments from this study are distinguished by starting material(either VS96-54A or VES9) and addition of olivine. (See Tables 3and 4 for more information about experimental conditions.)Experimental data points correspond either to a cotectic (Ol+Cpx)or to a single phase assemblage (either Ol or Cpx, specified next tothe appropriate data points). The 11508C, 200MPa olivineþ clinopyr-oxene cotectic is defined by the threeVS96-54A olivine-added charges5-1, 5-2 and 5-3. One point on the 50MPa, 11008C cotectic is definedby the VS96-54A charge 7-1. The three olivine-bearing charges 5-4,17-2 and 18-1 are labelled with temperature (either 1150 or 12008C).They plot on olivine control lines that extend from the VS96-54Aand VES9 starting glasses (Table 6) toward Fo-rich olivine (labelledOl). Additional experimental glasses plotted are from three olivine-and CaCO3-added charges run at 11508C, 200MPa with a Strombolibasalt starting material (Iacono-Marziano et al., 2008). The arrow in-dicates the direction of progressive addition of CaCO3

(labelledþCaCO3). Cpx and Ol (bold lettering) indicate the locationof clinopyroxene and olivine primary phase fields. In (b), the compos-ition of experimental glasses from this study, tephrite V36 (Trigila &De Benedetti, 1993) and Vulsini leucite basanite B23 (Conte et al.,2009; Table 7) are plotted for comparison with the glass inclusions.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2301

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Fo-rich olivine, diopside, Cr-spinel and primitive meltinclusions, to the later clinopyroxene-crystallizing stage(i.e. cpx is the liquidus phase) recorded byVesuvius erup-tion products. This transition must take place at hightemperatures (1100^12008C) and early in the magmaticevolution, as primitive products such as VS96-54A andVES9 (with liquidus temperatures in the range above)have cpx as the liquidus phase (Fig. 2). It is worth empha-sizing that the change in liquidus phase is accompaniedby the disappearance of olivine from the crystallization se-quence, as phase diagrams for mafic Vesuvius productslack ol stability fields (Fig. 2; Dolfi & Trigila, 1978; Trigila& De Benedetti, 1993). In the following sections, potentialmechanisms able to explain this transition are discussed.

Phlogopite crystallization

For melt H2O concentrations of 1�5^4�5wt %, tempera-tures between 1150 and 12008C and pressures of 200^300MPa, Fo-rich olivine and diopsidic clinopyroxene are

in a cotectic relationship in primitive K-rich basaltic li-quids (Barton & Hamilton, 1978; Esperanc� a & Holloway,1987; Righter & Carmichael, 1996; Melzer & Foley, 2000).Consequently, phase assemblages composed successively ofolivine and, when the derivative liquid reaches the cotectic,of olivineþ clinopyroxene, are expected, as observed inthe olivine-added experiments. Therefore, upon decreasingtemperature, Vesuvius magmas should show a progressiveevolution from olivine to olivine^clinopyroxene (� Cr-spinel) phenocryst assemblages (e.g. Marianelli et al., 1995;Fig. 9b). Later, leucite and plagioclase, depending onP^T^melt H2O^fO2, would join the phase assemblage, asobserved in the evolved members of theVesuvius suite, butolivine and clinopyroxene should continue to crystallizeuntil phlogopite appears in the crystallization sequence.Indeed, the phlogopite-forming reaction is a possiblemechanism to explain the replacement of olivine as theliquidus phase and its disappearance from the crystalliza-tion sequence (Barton & Hamilton, 1978; Esperanc� a &Holloway, 1987; Righter & Carmichael, 1996; Melzer &Foley, 2000). However, our experiments show that

Table 7: Primitive potassic Italian compositions (oxides in

wt %): magmatic liquids (Vesuvius) and whole-rock

(Vulsini)

System: Vesuvius Vesuvius Vulsini

Sample: 1906 glass inclusion1 exp. glass 5-42 B233

SiO2 50�39 49�6 49�12

TiO2 1�05 0�97 0�76

Al2O3 12�92 13�8 13�58

FeOt 7�35 6�44 7�12

MnO 0�14 0�14 0�13

MgO 8�64 8�65 9�49

CaO 12�20 12�5 12�99

Na2O 1�64 1�70 1�32

K2O 4�82 5�20 4�35

P2O5 0�86 0�92 0�34

Total 100 99�9 99�2

H2O 3�00 3�2 –

CO2 0�18 – –

Na2OþK2O 6�46 6�90 5�67

CaO/Al2O3 0�94 0�91 0�96

Mg# 754 79 784

Fo5 89�9 90�9 –

1From Marianelli et al. (1999).2From this study (data in Tables 4 and 6), olivine-addedexperiment.3Leucite basanite (Conte et al., 2009).4Calculated with the melt FeO determined at 12008C,200MPa, NNOþ 1 with the model of Kress & Carmichael(1991).5Fo content of coexisting olivine from Marianelli et al.(1999) and Table 6.

F

o

ΔNNO-2 -1 0 +1 +2 +3 +4

88

90

92

94

86

1906

1794

18221872

1944

1200°C200 MPaKdFe-Mg = 0.34

NN

O

Fig. 10. Redox state of primitiveVesuvius liquids as constrained fromolivine^liquid equilibrium. Each curve shows the Fo content of olivinein equilibrium with a given trachybasaltic liquid for different�NNO. Five trachybasaltic liquid compositions have been selected.They correspond to the most primitive glass inclusions from five erup-tions within the 1637^1944 period of activity of Vesuvius [recalculatedcompositions from Marianelli et al. (1995, 2005)]. The composition oftheir host olivine (Marianelli et al., 1995, 2005) is given by the fivecrosses. Chemical equilibrium is assumed between each trapped meltand its host olivine crystal, yielding a unique �NNO equilibriumvalue. The narrow range of �NNO (from þ0�6 to þ0�9) for the fiveglass inclusion-host olivine pairs selected should be noted. Despite sig-nificant differences in the composition of glass inclusions and olivinesbetween eruptions, all primitive magma batches appear to have equi-librated under similar redox conditions. All calculations are per-formed at 12008C, 200MPa with an olivine^liquid Kd

Fe^Mg¼ 0�34 as

determined in this study and for �NNO ranging from �1 to þ3.�NNO values (deviations from the NNO oxygen buffer) are calcu-lated at 12008C and 200MPa (log fO2 of the NNO buffer¼ ^7�7,Pownceby & O’Neill, 1994). Liquid FeO contents are calculatedfrom glass inclusion compositions using the model of Kress &Carmichael (1991).

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2302

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

temperatures511008C are required for phlogopite to crys-tallize, which is lower than considered above for the transi-tion between the ol- and cpx-crystallizing stages. Thistemperature range is consistent with the observation thatphlogopites inVesuvius products have Mg# (Fig. 5) lowerthan that needed for equilibrium with Fo-rich olivine ordiopside (taking into account our determined Kd

Phlog^L

and KdOl^L). Thus, phlogopites at Vesuvius crystallize

from relatively evolved liquids. We conclude that thephlogopite-forming reaction takes place at a stage moreadvanced in the differentiation course than the transitionbetween the ol- and cpx-crystallizing stages.

CO2 flushing

AtVesuvius, thermal production of CO2 is the consequenceof the interaction between magma and the carbonate base-ment. A CO2-rich fluid phase is thought to flush throughthe magma, increasing the d18O of Fo-rich Ol and diopsidecrystals (Dallai et al., 2011). Thus, the possibility that thisCO2-rich fluid inhibits olivine crystallization needs to beexplored. Indeed, the phase relations of basaltic magmasdepend on the fluid composition (Holloway & Burnham,1972). As a general mechanism, increasing the proportionof CO2 in the fluid (as with CO2 flushing) at constantpressure and fO2 will expand the stability field of Ol (aswell as of other anhydrous silicates such as Cpx) towardshigher temperature, because aH2O (and thus the meltH2O content) will be lowered (Feig et al., 2006; Pichavantet al., 2009). The consequence will be to promote (and notto inhibit) the crystallization of Ol, and of other anhydroussilicates as well. We conclude that a mechanism of CO2

flushing explains neither the removal of olivine from theliquidus nor its replacement by clinopyroxene as the liqui-dus phase.

Magma mixing

The absence of olivine stability fields in phase diagramsdetermined for mafic Vesuvius rocks (this study; Dolfi &Trigila, 1978; Trigila & De Benedetti, 1993) raises the possi-bility that this could be the consequence of the choice ofexperimental starting materials. Magma mixing, moreprecisely mixing between primitive and phonotephritic totephriphonolitic differentiates, is a common process atVesuvius (Santacroce et al., 1993; Cioni et al., 1995;Marianelli et al., 1999, 2005; Table 2). One clear indicationfor magma mixing is the coexistence in the same rock of amixed population of phenocrysts, each representing differ-ent (primitive and evolved) stages in the differentiation se-quence. This is observed in one of our starting products(VS96-54A, and also in VS97-718; see above). If a mag-matic mixture involving primitive and evolved fractionsalong the same differentiation trend (as for Vesuvius) isused as starting material, orders of crystallization andphase proportions will probably be modified comparedwith the primitive end-member magma. However, phases

on or near the liquidus (ol, cpx) should stay the same (foran illustration, see Walker et al., 1979). The disappearanceof ol from the crystallization sequence would remain unex-plained unless the evolved fraction dominates the mixture,forcing temperature to drop to511008C when phlogopitewould have the possibility to crystallize (see above).However, phlogopite is rare to absent in our starting prod-ucts (see also Marianelli et al., 1999). It is never very abun-dant in Vesuvius lavas (e.g. Santacroce et al., 1993; Trigila& De Benedetti, 1993). In clinopyroxenites, phlogopiteoccurs as a minor phase (Cundari, 1982) and it is con-cluded, therefore, that conditions leading to the crystalliza-tion of large amounts of phlogopite are not met atVesuvius.

Crystal accumulation

In the same way, phenocryst accumulation would makethe experimental starting composition unrepresentativeof a liquid. AtVesuvius, selective accumulation of clinopyr-oxene and leucite is observed in the 1906 and 1944 prod-ucts (e.g. Santacroce et al., 1993; Marianelli et al., 1999).Compositions of 1637^1944 whole-rocks reflect clinopyrox-ene, leucite and plagioclase modal proportions (Trigila &De Benedetti, 1993). As previously mentioned, one of ourstarting rocks (VS96-54A) is probably cpx cumulative(Marianelli et al., 1999). However, the experimental resultsfor VES9, which is clearly not cpx cumulative, are almostidentical to those for VS96-54A. In addition, cpx accumu-lation would not explain the disappearance of olivinefrom crystallization sequences. The order of phase appear-ance would be affected (i.e. cpx would crystallize beforeol), but the near-liquidus assemblage should stay identical(cpx, ol) because ol and cpx are in cotectic relationship inprimitive Vesuvius liquids (see above, Fig. 9a). Therefore,despite the complexities of our starting materials, we con-sider that our conclusions regarding liquidus phaseequilibria and crystallization sequences are robust andapplicable toVesuvius magmas.

Carbonate assimilationAlthough long considered as unconventional, carbonateassimilation has been recently reintroduced as a petrogen-etic process both at Vesuvius (Iacono-Marziano et al.,2008, 2009, and references therein; Dallai et al., 2011) andat other Italian volcanic districts (Iacono-Marziano et al.,2007; Freda et al., 2008; Gaeta et al., 2009). Carbonate as-similation by mafic magma produces CO2 gas. The gener-ated CaO melt component, together with the SiO2 meltcomponent, reacts with the Mg2SiO4 olivine componentto produce a CaMgSi2O6 clinopyroxene component(Gaeta et al., 2006; Iacono-Marziano et al., 2008).Therefore, carbonate assimilation promotes reaction be-tween melt and olivine to crystallize clinopyroxene.Details of this reaction have been experimentally simu-lated: at 11508C, 200MPa, and starting from a magmawith both olivine and clinopyroxene as liquidus phases,

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2303

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

progressive addition of CaCO3 (from 0 to 15wt %) leadsto the disappearance of liquidus olivine (Iacono-Marziano et al., 2008). Glasses from these experiments(Fig. 9a) define a compositional trend that extends fromthe near-cotectic region (two data points correspond tocharges with residual olivine still present) to well withinthe cpx primary phase field (one point corresponds to acharge where olivine has totally disappeared). Therefore,carbonate assimilation has the potential to drive primitiveVesuvius magmatic liquids from the olivine into the cpxprimary phase field. In this process, olivine crystals ini-tially present in the magma are resorbed. In detail, theamount of ol consumed depends on the amount (andCa/Mg) of the carbonate material being assimilated. Weconclude that carbonate assimilation provides an explan-ation for many puzzling characteristics of Vesuvius maficrocks, including the absence of products representative ofearly olivine-crystallizing magmas, the low abundanceof olivine in tephrites and phonotephrites, the occurrenceof cpx (and not ol) as the liquidus phase and the ab-sence of ol from crystallization sequences. Carbonateassimilation most probably starts when ascending mantle-derived magmas encounter the base of the crustal carbon-ate sequence (Iacono-Marziano et al., 2009; Dallai et al.,2011), which is inferred to lie at �11km depth beneathVesuvius (Berrino et al., 1998; Pommier et al., 2010).We stress that the main new argument presented in this

study for carbonate assimilation comes from phase equilib-rium considerations; that is, erupted products have cpxon their liquidus despite the evidence that their parentalliquids initially crystallized olivine. The model is similarto that proposed for Colli Albani, a volcanic district of thepotassic Roman Province also built on a thick carbonatesubstratum (Freda et al., 2008; Gaeta et al., 2009). TheColli Albani products have cpx on the liquidus (Iacono-Marziano et al., 2007; Freda et al., 2008), yet they are alsoderived from primitive parental magmas that crystallizeFo-rich olivine, clinopyroxene and spinel (Freda et al.,2008; Conte et al., 2009; Gaeta et al., 2009). The whole-rock composition of a leucite basanite (B23) from Vulsiniis given inTable 7 (see also Fig. 9b), showing the similaritybetween the potassic Roman Province and Vesuvius.Rocks such as B23 carry olivine phenocrysts; they are con-sidered representative of parental magmas at Vulsini(Conte et al., 2009). Their compositional and mineralogicalcharacteristics are analogous to those inferred forVesuviusparental magmas, the direct observation of which isimpossible.

Implications for magma feeding systemsAt Vesuvius, the erupted magmas have a wide range ofcompositions (Fig. 1). Several explanations have been pro-posed to account for this diversity, observed both within asingle eruption and between eruptions. As shown byIacono-Marziano et al. (2008), a decrease of whole-rock

SiO2 with increasing Na2OþK2O [trend (I) in Fig. 11]is the geochemical signature expected for carbonate as-similation. Magma desilication is the consequence of theconsumption of the melt SiO2 component to form clino-pyroxene (Iacono-Marziano et al., 2008). In the recentperiod of activity, only the 1637^1944 whole-rocks show aslight decrease of SiO2 at increasing Na2OþK2O(Marianelli et al., 2005; Fig. 1). In comparison, the 1906and 1944 eruptions have SiO2 either constant or slightlyincreasing with Na2OþK2O.This trend (II, Fig. 11) is theresult of the combination of various processes, includingmixing between primitive and more evolved magmas, ac-cumulation or removal of clinopyroxene and leucite, crys-tal fractionation (Santacroce et al., 1993; Marianelli et al.,1999) and, we suggest, carbonate assimilation. Trend (III),in which the strong increase in SiO2 contrasts with thetwo others and implies a very minor role for carbonate as-similation, is illustrated by the Pompei eruption (AD 79,Fig. 1). Because our experimental glasses (especially themost evolved) mimic this natural trend, we interprettrend (III) to be mainly controlled by cpx fractionationunder closed-system conditions. Therefore, variations inthe extent of carbonate assimilation between eruptionscan be detected.For a volcanic system emplaced on a carbonate substra-

tum the main factors controlling the extent of carbonateassimilation would be (1) the supply rate of fresh magma,(2) its composition, (3) the structure and aspect ratio ofthe feeding system and (4) the volume of stored magma.Primitive magmas would be the most exposed to carbon-ate interaction as high MgO, presence of olivine pheno-crysts and high temperatures enhance the assimilationprocess (see Iacono-Marziano et al., 2008). Mafic magmabatches intruding the base of the carbonate sequenceas an array of dykes or cracks would develop magma^carbonate interfaces that promote assimilation. Carbonateinteraction would be maximized with multiple highaspect ratio intrusions of small volume (Fig. 11b).Conversely, for a large-volume, low aspect ratio reservoir,less frequently recharged, hence colder, magma^carbonateinteraction would be more limited. Fresh magma wouldintersect the reservoir and be consumed by mixing withthe resident differentiates (Cioni et al., 1995; Fig. 11c).Interaction with carbonate wall-rocks would be preventedby a thick skarn shell if the reservoir is long-lived (Cioniet al., 1998; Fulignati et al., 2004a, 2004b), and also by thelow bulk MgO and low temperature of the residentmagma (Iacono-Marziano et al., 2008). The former casecorresponds to systems that feed open-conduit eruptionssuch as during 1637^1944 and in 1906 and 1944. Thesesmall-volume eruptions would be controlled by the epi-sodic arrival at shallow levels (Scandone et al., 2008) oftephritic magma batches generated at 7^11km depth fol-lowing carbonate assimilation by primitive trachybasalts

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2304

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

(Figs 7 and 11b). Products from those eruptions should besignificantly influenced by carbonate assimilation. Whole-rock trends from negative to slightly positive in the TASdiagram would be expected, and are indeed observed(Figs 1 and 11a). The latter case is illustrated by large‘mature’ magma chambers (Cioni et al., 1998) tapped byPlinian eruptions such as the Pompei eruption.Their prod-ucts should be on the whole little affected by carbonate as-similation, consistent with the marked SiO2 increasefound in AD 79 rocks (Figs 1b and 11a). Thus, a correlationbetween the extent of carbonate assimilation and type offeeding system is proposed. The level of carbonate assimi-lation (which can be estimated from the proportion of sedi-mentary CO2 in the gas plume; Iacono-Marziano et al.,2009) might provide information on the type of feedingsystem at present beneathVesuvius (Scaillet et al., 2008).

Implications for potassic magmatismin central^southern ItalyThe new data presented in this study on primitiveVesuviusmagmas make it possible to make comparisons withother potassic volcanic centers of central^southern Italy.

InVesuvius as in other centers, an early phase assemblagecomposed of olivine, Cr-spinel and diopside is found inthe most mafic products. Mineral compositions (notablyFo �89^90 and Cr# 440) demonstrate a near-primarymantle origin for the magmas (Fig. 12a). Anothercommon feature is the scarcity to total absence of eruptionproducts representative of near-primary magmas, as previ-ously discussed for Vesuvius. At Phlegrean Fields, anotherCampanian volcano, eruptions provide no samples of theprimitive feeding magmas, whose characteristics areinferred from rare lava domes (Melluso et al., 2011), andalso from the nearby Ischia (D’Antonio et al., 2013; Morettiet al., 2013) and Procida (Mormone et al., 2011) volcaniccenters. At Colli Albani, in the Roman Province, the mostprimitive compositions are already fairly evolved (Fredaet al., 2008; Gaeta et al., 2009; and see above). Exceptionsoccur (e.g.Vulsini, Conte et al., 2009) but, in most systems,the primitive mineral phase assemblages and compositionshave little or no whole-rock counterpart. This has led vari-ous researchers to consider crystals from the primitivephase assemblage (e.g. Fo-rich olivines) as xenocrysts(see, e.g. Dallai et al., 2011; Melluso et al., 2011; Moretti

wt% SiO2

wt%

K2O

+ N

a 2O

40 50 60 700

4

8

12

16

(I)(II)

(III)0.5 km

K-basalt

tephrite

phonolite

phono-tephrite

tephri-phonolite

carbonate contactaureole

magma-carbonateinterface

cpxcumulate

(a)

(b)

(c)

0.01 km

trachybasalt

Fig. 11. (a) Total alkalis^silica (TAS) variation diagram showing the composition of the parental magma (black star) and the three mainVesuvius whole-rock evolutionary trends. (I) carbonate assimilation; (II) complex open-system evolution; (III) closed-system crystal fraction-ation. (b, c) Types of Vesuvius feeding systems [modified from Cioni et al. (1998)]. (b) illustrates the location of shallow reservoirs that feedopen-conduit eruptions such as those during 1637^1944 and in 1906 and 1944. Colour shading illustrates the progressive transition from trachy-basalt to tephrite with carbonate assimilation. Products from those eruptions should be the most influenced by carbonate assimilation asshown by slightly negative (I) to slightly positive (II) whole-rock trends in the TAS diagram (a). (c) corresponds to a large ‘mature’ magmachamber tapped by Plinian eruptions such as Pompei AD 79 and whose whole-rock evolutionary trends (III) mainly reflect cpx fractionation,and are little affected by carbonate assimilation. (See text for details.)

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2305

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

et al., 2013). Although technically justified (olivine is tooFo-rich for equilibrium with the bulk-rock), a more unify-ing model is favoured here, which recognizes the primitiveminerals as parental magma phenocrysts and interpretstheir crystallization as an early step in the magmaticevolution.Compositions of Cr-spinel are plotted as a function

of the Fo content of their olivine hosts (Fig. 12a) for primi-tive mineral assemblages from various Campanian volca-noes (Vesuvius, Phlegrean Fields, Ischia, Procida). ForPhlegrean Fields, mineral compositions from theAccademia Dome are used (Melluso et al., 2011), and forIschia and Procida, data from the Vateliero (D’Antonioet al., 2013) and Solchiaro eruptions (D’Antonio & DiGirolamo, 1994) respectively. Redox states (�NNO

values) were calculated using the olivine^spinel oxybarom-eter (Ballhaus et al.,1991; Fig. 12b). �NNO values range be-tween þ0�2 to þ1�5 and no large difference is apparentbetween the four volcanoes considered. In particular,there is no evidence for very oxidized magmas, as pro-posed for Ischia (�NNO up to�þ3; D’Antonio et al.,2013; Moretti et al., 2013). The Procida data have lower fO2

than the three other systems because our Solchiaro data-base lacks Cr-spinel inclusions in Fo488 olivines.In the Cr# vs Fo plot (Fig. 12a), the Vesuvius trend is

rooted in the mantle array (Arai,1994a,1994b), thus under-lining the primitive character of olivines and spinels thatcome from VS97-718 and from products of the 1637^1944period. Comparing with the other Campanian systems,the Vesuvius spinels are distinctive, being more Cr-rich.Although mantle melting and magma generation in cen-tral^southern Italy involve complex mantle sources andseveral metasomatic agents, the Cr# of spinel is generallyconsidered as an indicator of the peridotitic component(e.g. Conticelli et al., 2004; Nikogosian & van Bergen,2010; Moretti et al., 2013). Therefore, the distinctive com-position of spinels would imply different peridotitic com-ponents in the source of Vesuvius and of other Campanianmagmas.

CONCLUSIONS(1) Clinopyroxene is the liquidus phase for the two primi-tive VS96-54A and VES9 Vesuvius products. This conclu-sion does not depend on the experimental conditions(H2O-bearing fluid-absent, H2O- and CO2-bearing fluid-present, 0�1MPa volatile-free).(2) Compositions of experimental and natural clinopyr-

oxene, leucite and phlogopite are in excellent agreement.Chemical trends exhibited by whole-rocks and melt inclu-sions have been experimentally simulated. This underlinesthe vital role of clinopyroxene in controlling crystal frac-tionation of Vesuvius magmas and confirms a genetic rela-tionship between the various members of the naturaldifferentiation sequence.(3) The absence of plagioclase in our experimental prod-

ucts is attributed to the primitive nature of the startingcompositions, and to P^T^melt H2O^fO2 conditionscloser to the liquidus than in previous experimental studieson Vesuvius products. Plagioclase plays no critical role inthe early magmatic evolution. The near-absence of olivinein our experiments reflects the lack of rocks representativeof the most primitive magmas at Vesuvius. ExperimentsonVS96-54A and VES9 simulate a stage that is relativelylate in the magmatic evolution.(4) The olivine-added experiments can be used to recon-

struct the early evolution of Vesuvius magmas.Comparison between glasses from these experiments andfrom melt inclusions shows that the most primitiveVesuvius magmas are trachybasalts (K2O� 4�5^5�5wt

ΔNNO

Cr#

90

80

70

60

50

400 0.5 1 1.5 2

VES ISCCFPRO

(b)

90

80

70

60

50

40

Cr#

Fo94 92 90 88 86 84

VES ISCCFPRO

(a)

OSM

A

Fig. 12. Mineral composition (Cr# in spinel, Fo in olivine) andredox state (expressed as �NNO) for primitive Vesuvius magmascompared with other Campanian volcanoes. (a) Cr# in spinel vs Foin olivine; (b) Cr# in spinel vs �NNO [calculated from the oxyba-rometer of Ballhaus et al. (1991)]. All calculations were carried out at12008C and 200MPa (log fO2 of the NNO buffer¼ ^7�7, Pownceby& O’Neill, 1994). In (a), the grey field is the olivine^spinel mantlearray (OSMA, Arai, 1994a, 1994b). In (b), the grey field is the fO2

range forVesuvius determined from olivine^liquid equilibria (Fig. 10).Data sources: VES (Vesuvius), Marianelli et al. (1995, 2005) and thisstudy; ISC (Ischia, Vateliero eruption), D’Antonio et al. (2013); CF(Phlegrean Fields, Accademia Dome), Melluso et al. (2011); PRO(Procida, Solchiaro eruption), D’Antonio & Di Girolamo (1994).

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2306

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

%, MgO¼ 8^9wt %, Mg#¼ 75^80, CaO/Al2O3¼0�9^0�95) similar to other potassic magmas from the RomanProvince. They crystallize Fo-rich olivine (90^91) on theirliquidus between 1150 and 12008C and from a maximumof 300 to �200MPa. They contain 1�5^4�5wt % H2O and600^4500 ppm CO2, and are fairly oxidized (fromNNOþ 0�4 to NNOþ1�2).(5) The transition from an early olivine-crystallizing to a

later clinopyroxene-crystallizing stage requires a specificmechanism during differentiation of the Vesuvius mag-matic suite. Assimilation of carbonate wall-rocks by as-cending primitive magmas satisfactorily explains the factthat erupted rocks have cpx on their liquidus despite evi-dence that their parental liquids initially crystallizedolivine. Carbonate assimilation erases the memory of theearly olivine-crystallizing stage inVesuvius magmas.(6) Carbonate assimilation is promoted for primitive

magma batches of relatively small volume. For longer-lived, large-volume, less frequently recharged, hencecolder, magma reservoirs, magma^carbonate interactionis limited. A correlation between the extent of carbonateassimilation and type of feeding system is proposed, withpotential implications for the monitoring of the deep-seated activity of the volcano.(7) Primitive magmas from Vesuvius and other

Campanian volcanoes do not show systematic differencesin their redox states. In contrast, the Cr# of Vesuvius spi-nels is distinctive and implies different peridotitic compo-nents in the mantle source of Vesuvius and of otherCampanian magmas.

ACKNOWLEDGEMENTSI. Arienzo, A. M. Conte, S. Conticelli, M. D’Antonio, P.Marianelli, L. Melluso, R. Moretti and M. Piochi arethanked for discussions, data and samples. This workbenefited from constructive reviews by M. Rutherford, M.Gaeta and an anonymous reviewer.

FUNDINGThis study was supported by the 2004^2006 DPC-INGVproject on Vesuvius and by the MED-SUV project. G.Iacono-Marziano was funded by the ANR projectInterVol.

REFERENCESAlbare' de, F. (1995). Introduction to Geochemical Modeling. Cambridge

University Press, 543 p.Arai, S. (1994a). Characterization of spinel peridotites by olivine^

spinel compositional relationships: review and interpretation.Chemical Geology 113, 191^204.

Arai, S. (1994b). Compositional variation of olivine^chromian spinelin Mg-rich magmas as a guide to their residual spinel peridotites.Journal of Volcanology and Geothermal Research 59, 279^293.

Ballhaus, C., Berry, C. F. & Green, D. H. (1991). High pressure experi-mental calibration of the olivine^orthopyroxene^spinel oxygengeobarometer: implications for the oxidation state of the uppermantle. Contributions to Mineralogy and Petrology 108, 82^92.

Barclay, J. & Carmichael, I. S. E. (2004). A hornblende basalt fromWestern Mexico: water-saturated phase relations constrain a pres-sure^temperature window of eruptibility. Journal of Petrology 45,485^506.

Barton, M. & Hamilton, D. L. (1978). Water-saturated melting rela-tions to 5 kilobars of three Leucite Hills lavas. Contributions to

Mineralogy and Petrology 66, 41^49.Behrens, H., Romano, C., Nowak, M., Holtz, F. & Dingwell, D. B.

(1996). Near-infrared determination of water species in glasses ofthe system MAlSi3O8 (M ¼ Li, Na, K): an interlaboratory study.Chemical Geology 128, 41^63.

Behrens, H., Misiti, V., Freda, C., Vetere, F., Botcharnikov, R. E. &Scarlato, P. (2009). Solubility of H2O and CO2 in ultrapotassicmelts at 1200 and 12508C and pressures from 50 to 500 MPa.American Mineralogist 94, 105^120.

Belkin, H. E., Kilburn, C. R. J. & De Vivo, B. (1993). Sampling andmajor-element chemistry of the recent (AD1631^1944) Vesuvius ac-tivity. Journal of Volcanology and Geothermal Research 58, 273^290.

Berrino, G., Corrado, G. & Riccardi, U. (1998). Sea gravity data in theGulf of Naples: a contribution to delineating the structural patternof the Vesuvian area. Journal of Volcanology and Geothermal Research

82, 139^150.Burnham, C.W. (1979). The importance of volatile constituents. In:

Yoder, H. S., Jr (ed.) The Evolution of Igneous Rocks. PrincetonUniversity Press, pp. 439^482.

Cioni, R. (2000). Volatile content and degassing processes in the AD79 magma chamber at Vesuvius (Italy). Contributions to Mineralogy

and Petrology 140, 40^54.Cioni, R., Civetta, L., Marianelli, P., Metrich, N., Santacroce, R. &

Sbrana, A. (1995). Compositional layering and syn-eruptivemixing of a periodically refilled shallow magma chamber:the AD 79 plinian eruption of Vesuvius. Journal of Petrology 36,739^776.

Cioni, R., Marianelli, P. & Santacroce, R. (1998). Thermal and com-positional evolution of the shallow magma chambers of Vesuvius:evidence from pyroxene phenocrysts and melt inclusions. Journal ofGeophysical Research 103, 18227^18294.

Conte, A. M., Dolfi, D., Gaeta, M., Misiti, V., Mollo, S. &Perinelli, C. (2009). Experimental constraints on evolution of leu-cite-basanite magma at 1 and 10�4 GPa: implications for parentalcompositions of Roman high-potassium magmas. European Journalof Mineralogy 21, 763^782.

Conticelli, S., Melluso, L., Perini, G., Avanzinelli, R. & Boari, E.(2004). Petrologic, geochemical and isotopic characteristics ofpotassic and ultrapotassic magmatism in central^southern Italy:inferences on its genesis and on the nature of mantle sources.Periodico Mineralogia 73, 135^164.

Costa, F., Scaillet, B. & Pichavant, M. (2004). Petrological and experi-mental constraints on the pre-eruption conditions of Holocenedacite from Volca¤ n San Pedro (368S, Chilean Andes) and theimportance of sulphur in silicic subduction-related magmas.Journal of Petrology 45, 855^881.

Cundari, A. (1982). Petrology of clinopyroxenites ejecra from Somma^Vesuvius and their genetic implications. Tschermak Mineralogisches

und Petrographische Mitteilungen 30, 17^35.Dallai, L., Cioni, R., Boschi, C. & D’Oriano, C. (2011). Carbonate-

derived CO2 purging magma at depth: influence on the eruptiveactivity of Somma^Vesuvius, Italy. Earth and Planetary Science Letters310, 84^95.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2307

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

D’Antonio, M. & Di Girolamo, P. (1994). Petrological and geochemicalstudy of mafic shoshonitic volcanics from Procida^Vivara andVentotene islands (Campanian region, south Italy). Acta

Vulcanologica 5, 69^80.D’Antonio, M., Tonarini, S., Arienzo, I., Civetta, L., Dallai, L.,

Moretti, R., Orsi, G., Mariachiara, A. & Trecalli, A. (2013).Mantle and crustal processes in the magmatism of the Campaniaregion: inferences from mineralogy, geochemistry, and Sr^Nd^Oisotopes of young hybrid volcanics of the Ischia island (southItaly). Contributions to Mineralogy and Petrology 165, 1173^1194.

Devine, J. D., Gardner, J. E., Brack, H. P., Layne, G. D. &Rutherford, M. J. (1995). Comparison of microanalytical methodsfor estimating H2O contents of volcanic glasses. American

Mineralogist 80, 319^328.Di Carlo, I., Pichavant, M., Rotolo, S. G. & Scaillet, B. (2006).

Experimental crystallization of a high-K arc basalt: the goldenpumice, Stromboli volcano (Italy). Journal of Petrology 47, 1317^1343.

Dixon, J. E., Stolper, E. M. & Holloway, J. R. (1995). An experimentalstudy of water and carbon dioxide solubilities in mid-ocean ridgebasaltic liquids. Part I: calibration and solubility models. Journal ofPetrology 36, 1607^1631.

Dolfi, D. & Trigila, R. (1978). The role of water in the 1944 Vesuviuseruption. Contributions to Mineralogy and Petrology 67, 297^304.

Esperanc� a, S. & Holloway, J. R. (1987). On the origin of some mica-lamprophyres: experimental evidence from a mafic minette.Contributions to Mineralogy and Petrology 95, 207^216.

Feig, S.T., Koepke, J. & Snow, J. E. (2006). Effect of water on tholeiiticbasalt phase equilibria: an experimental study under oxidizingconditions. Contributions to Mineralogy and Petrology 152, 611^638.

Freda, C., Gaeta, M., Misiti, V., Mollo, S., Dolfi, D. & Scarlato, P.(2008). Magma^carbonate interaction: an experimental study onultrapotassic rocks from Alban Hills (Central Italy). Lithos 101,397^415.

Fulignati, P., Marianelli, P. & Sbrana, A. (2000). The feeding systemof the 1944 eruption of Vesuvius: melt inclusion data from duniticnodules. Neues Jahrbuch fu« r Mineralogie, Monatshefte 9, 419^432.

Fulignati, P., Marianelli, P., Santacroce, R. & Sbrana, A. (2004a).Probing theVesuvius magma chamber^host rock interface throughxenoliths. Geological Magazine 141, 417^428.

Fulignati, P., Marianelli, P., Me¤ trich, N., Santacroce, R. & Sbrana, A.(2004b). Towards a reconstruction of the magmatic feeding systemof the 1944 eruption of Mt Vesuvius. Journal of Volcanology and

Geothermal Research 133, 13^22.Gaeta, M., Freda, C., Christensen, J. N., Dallai, L., Marra, F.,

Karner, D. B. & Scarlato, P. (2006). Time-dependent geochemistryof clinopyroxene from the Alban Hills (Central Italy): clues to thesource and evolution of ultrapotassic magmas. Lithos 86, 330^346.

Gaeta, M., Di Rocco, T. & Freda, C. (2009). Carbonate assimilationin open magmatic systems: the role of melt-bearing skarns andcumulate-forming processes. Journal of Petrology 50, 361^385.

Holloway, J. R. (1987). Igneous fluids. In: Eugster, H. P. &Carmichael, I. S. E. (eds) Thermodynamic Modeling of Geological

Materials: Minerals, Fluids and Melts. Mineralogical Society of America,

Reviews in Mineralogy 17, 211^233.Holloway, J. R. & Burnham, C.W. (1972). Melting relations of basalt

with equilibrium water pressure less than total pressure. Journal ofPetrology 13, 1^29.

Iacono-Marziano, G., Gaillard, F. & Pichavant, M. (2007). Limestoneassimilation and the origin of CO2 emissions at the Alban Hills(Central Italy): constraints from experimental petrology. Journalof Volcanology and Geothermal Research 166, 91^105.

Iacono-Marziano, G., Gaillard, F. & Pichavant, M. (2008). Limestoneassimilation by basaltic magmas: an experimental re-assessment

and application to Italian volcanoes. Contributions to Mineralogy and

Petrology 155, 719^738.Iacono-Marziano, G., Gaillard, F., Scaillet, B., Pichavant, M. &

Chiodini, G. (2009). Role of non-mantle CO2 in the dynamics of vol-cano degassing: the MountVesuvius example. Geology 37, 319^322.

Iacono-Marziano, G., Morizet, Y., Le Trong, E. & Gaillard, F. (2012).New experimental data and semi-empirical parameterization ofH2O^CO2 solubility in mafic melts. Geochimica et Cosmochimica Acta97, 1^23.

Kress, V. C. & Carmichael, I. S. E. (1991). The compressibility ofsilicate liquids containing Fe2O3 and the effect of composition, tem-perature, oxygen fugacity and pressure on their redox states.Contributions to Mineralogy and Petrology 108, 82^92.

Lesne, P., Scaillet, B., Pichavant, M. & Be¤ ny, J.-M. (2011). The carbondioxide solubility in alkali basalts: an experimental study.Contributions to Mineralogy and Petrology 162, 153^168.

Marianelli, P., Metrich, N., Santacroce, R. & Sbrana, A. (1995).Mafic magma batches at Vesuvius: a glass inclusion approach tothe modalities of feeding stratovolcanoes. Contributions to Mineralogy

and Petrology 120, 159^169.Marianelli, P., Metrich, N. & Sbrana, A. (1999). Shallow and deep res-

ervoirs involved in magma supply of the 1944 eruption ofVesuvius. Bulletin of Volcanology 61, 48^63.

Marianelli, P., Sbrana, A., Metrich, N. & Cechetti, A. (2005).The deep feeding system of Vesuvius involved in recent violentstrombolian eruptions. Geophysical Research Letters 32, L2306.

Martel, C., Pichavant, M., Holtz, F., Scaillet, B., Bourdier, J.-L. &Traineau, H. (1999). Effects of fO2 and H2O on andesite phase rela-tions between 2 and 4 kbar. Journal of Geophysical Research 104,29453^29470.

Melluso, L., De’ Gennaro, R., Fedele, L., Franciosi, L. & Morra, V.(2011). Evidence of crystallization in residual, Cl^F-rich, agpaıtic,trachyphonolitic magmas and primitive Mg-rich basalt^trachyphonolite interaction in the lava domes of the PhlegreanFields (Italy). Geological Magazine 49, 532^550.

Melzer, S. & Foley, S. F. (2000). Phase relations and fractionationsequences in potassic magma series modelled in the systemCaMgSi2O6^KAlSiO4^Mg2SiO4^SiO2^F2O^1 at 1 bar to 18 kbar.Contributions to Mineralogy and Petrology 138, 186^197.

Moretti, R., Arienzo, I., Orsi, G., Civetta, L. & D’Antonio, M. (2013).The deep plumbing system of Ischia: a physico-chemical windowon the fluid-saturated and CO2-sustained Neapolitan volcanism(southern Italy). Journal of Petrology 54, 951^984.

Mormone, A., Piochi, M., Bellatrecia, F., De Astis, G., Moretti, R.,DellaVentura, G., Cavallo, A. & Mangiacapra, A. (2011). A CO2-rich magma source beneath the Phlegrean volcanic district (south-ern Italy): evidence from a melt inclusion study. Chemical Geology287, 66^80.

Nikogosian, I. K. & van Bergen, M. J. (2010). Heterogeneous mantlesources of potassium-rich magmas in central^southern Italy: meltinclusion evidence from Roccamonfina and Ernici (Mid LatinaValley). Journal of Volcanology and Geothermal Research 197, 279^302.

Pichavant, M. & Macdonald, R. (2007). Crystallization of primitive bas-altic magmas at crustal pressures and genesis of the calc-alkaline igne-ous suite: experimental evidence from StVincent, Lesser Antilles arc.Contributions to Mineralogy and Petrology 154, 535^558.

Pichavant, M., Di Carlo, I., Le Gac, Y., Rotolo, S. & Scaillet, B.(2009). Experimental constraints on the deep magma feedingsystem at Stromboli volcano, Italy. Journal of Petrology 50, 601^624.

Pommier, A., Tarits, P., Hautot, S., Pichavant, M., Scaillet, B. &Gaillard, F. (2010). A new petrological and geophysical investiga-tion of the present-day plumbing system of Mount Vesuvius.Geochemistry, Geophysics, Geosystems 11, doi:10.1029/2010GC003059.

JOURNAL OF PETROLOGY VOLUME 55 NUMBER 11 NOVEMBER 2014

2308

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from

Pownceby, M. I. & O’Neill, H. St. C. (1994). Thermodynamic datafrom redox reactions at high temperatures. III. Activity^compos-ition relations in Ni^Pd alloys from EMF measurements at850^1250 K and calibration of the NiO þ Ni^Pd assemblage as aredox sensor. Contributions to Mineralogy and Petrology 116, 327^339.

Richet, P., Whittington, A., Holtz, F., Behrens, H., Ohlhorst, S. &Wilke, M. (2000). Water and the density of silicate glasses.Contributions to Mineralogy and Petrology 138, 337^347.

Righter, K. & Carmichael, I. S. E. (1996). Phase equilibria of phlogo-pite lamprophyres from western Mexico: biotite liquid equilibriaand P^Testimates for biotite-bearing igneous rocks. Contributions toMineralogy and Petrology 123, 1^21.

Robie, R. A., Hemingway, B. S. & Fisher, J. R. (1979).Thermodynamic

properties of minerals and related substances at 298�15 K and 1 bar (105

pascals) pressure and at higher temperatures. US Geological Survey Bulletin

1452.Rolandi, G., Munno, R. & Postiglione, I. (2004). The A.D. 472 erup-

tion of the Somma volcano. Journal of Volcanology and Geothermal

Research 129, 291^319.Rosi, M. & Santacroce, R. (1983). The A.D. 72 ‘Pollena’ eruption:

volcanological and petrological data for this poorly-known, pli-nian-type event at Vesuvius. Journal of Volcanology and Geothermal

Research 17, 249^271.Rosi, M., Principe, C. & Vecci, R. (1993). The 1631 Vesuvius eruption.

A reconstruction based on historical and stratigraphical data.Journal of Volcanology and Geothermal Research 58, 151^182.

Santacroce, R., Bertagnini, A., Civetta, L., Landi, P. & Sbrana, A.(1993). Eruptive dynamics and petrogenetic processes in a veryshallow magma reservoir: the 1906 eruption of Vesuvius. Journal ofPetrology 34, 383^425.

Scaillet, B. & Evans, B.W. (1999). The 15 June 1991 eruption of MountPinatubo. I. Phase equilibria and pre-eruption P^T^fO2^fH2Oconditions of the dacite magma. Journal of Petrology 40, 381^411.

Scaillet, B., Pichavant, M., Roux, J., Humbert, G. & Lefe' vre, A.(1992). Improvements of the Shaw membrane technique for meas-urements and control of fH2 at high temperatures and pressures.American Mineralogist 77, 647^655.

Scaillet, B., Pichavant, M. & Roux, J. (1995). Experimental crystal-lization of leucogranite magmas. Journal of Petrology 36, 663^705.

Scaillet, B., Pichavant, M. & Cioni, R. (2008). Upward migration ofVesuvius magma chamber over the past 20,000 years. Nature 455,216^220.

Scandone, R., Giacomelli, L. & Speranza, F. (2008). Persistent activ-ity and violent strombolian eruptions at Vesuvius between 1631 and1944. Journal of Volcanology and Geothermal Research 170, 167^180.

Sisson, T.W. & Grove, T. L. (1993). Experimental investigation of therole of H2O in calc-alkaline differentiation and subduction zonemagmatism. Contributions to Mineralogy and Petrology 113, 143^166.

Taylor, J. R.,Wall,V. J. & Pownceby, M. I. (1992). The calibration andapplication of accurate redox sensors. American Mineralogist 77,284^295.

Trigila, R. & De Benedetti, A. A. (1993). Petrogenesis of Vesuvius his-torical lavas constrained by Pearce element ratios analysis andexperimental phase equilibria. Journal of Volcanology and Geothermal

Research 58, 315^343.Villemant, B., Trigila, R. & De Vivo, B. (1993). Geochemistry of

Vesuvius volcanics during 1631^1944 period. Journal of Volcanologyand Geothermal Research 58, 291^313.

Walker, D., Shibata, T. & DeLong, S. E. (1979). Abyssal tholeiitesfrom the Oceanographer fracture zone. Part II. Phaseequilibria and mixing. Contributions to Mineralogy and Petrology 70,111^125.

Wallace, P. & Carmichael, I. S. E. (1989). Minette lavas and associatedleucitites from the western front of the Mexican volcanic belt: pet-rology, chemistry and origin. Contributions to Mineralogy and Petrology

103, 470^492.

PICHAVANT et al. PRIMITIVEVESUVIUSMAGMAS

2309

at University of C

alifornia, San Diego on N

ovember 11, 2014

http://petrology.oxfordjournals.org/D

ownloaded from