Timescales of bubble coalescence, outgassing, and foam collapse in decompressed rhyolitic melts

HAL Id: insu-01102052

https://hal-insu.archives-ouvertes.fr/insu-01102052

Submitted on 21 Nov 2016

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Timescales of bubble coalescence, outgassing, and foam

collapse in decompressed rhyolitic melts

Caroline Martel, Giada Iacono-Marziano

To cite this version:

Timescales

of

bubble

coalescence,

outgassing,

and

foam

collapse

in decompressed

rhyolitic

melts

Caroline Martel

∗

,

Giada Iacono-Marziano

Universitéd’Orléans,CNRS/INSU,ISTO,BRGM,UMR7327,45071Orléans,France

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory:

Received4February2014

Receivedinrevisedform1December2014 Accepted4December2014

Availableonline7January2015 Editor:T.Elliott Keywords: decompression coalescence outgassing foamcollapse permeability rhyoliticmelt

Thetimescaleofdegassingandoutgassinginhydrousrhyoliticmeltsisinvestigatedinawiderangeof conditionsbymeansofdecompressionexperiments.Theevolutionofvesicularity,bubblediameter,and numberdensityischaracterizedasafunctionoftimeeitherofdecompressionorspentatfinalpressure, inordertodeterminetheeffectoffinalpressure,temperature,syn- versuspost-decompressiondegassing, meltcomposition,andmicrolites,onthetimescaleofbubblegrowth,coalescence,andoutgassing. Theresultsuggests thatdifferentbubbleevolutionanddegassing–outgassingtimescalecorrespondingto explosive and effusiveeruptionregimes can be cast inbulk viscosity (melt +bubbles;

η

bulk) versus

decompression time (rather than path) space. The

η

bulk–time relationship defines three domains of

(i) bubble nucleation and growth, restricted to short durations and high

η

bulk (<∼0.03 h for

η

bulk

∼105–6Pa s), (ii) equilibrium degassing with coalescence increasing from negligible (permeability >

10−13_m2_{) toextensive(permeability∼10}−11–12 _m2_{), and (iii)outgassing,restrictedtolongdurations}

and low

η

bulk (>∼10 h for

η

bulk<106Pa s; permeability>10−10m2) that eventuallyleads tofoam

collapse.

ThesefindingsareappliedtothecasestudiesofMtPeléeandMtPinatubotoinferthetransitionfrom pumicetodensepyroclastsinvolcaniceruptionsandthepossibilityofevolvingfromanexplosivePlinian eruptiontoaneffusivedome-growthevent bygivingthevesicularmagmaenoughtimetooutgasand collapse(i.e.hundredstotensofhoursfor

η

bulk∼105to104Pa s,respectively).Wealsoshowthedrastic

effectofmicrolitesonre-arrangingpreexistentbubblesandpotentiallytriggeringalatenucleationevent.

1. Introduction

Explosivevolcaniceruptionshavearousedgreatinterestin un-derstandingthemechanisms ofdegassingandthetransitions be-tween explosive and effusive eruptions of silicic magmas. When silicic magmas rise from depth to the Earthsurface, gas solubil-itydecreases and the oversaturated melt exsolves gases as bub-blesthat growby gasdiffusionandvapor expansion.Either bub-blesremain isolated andtrapped inthemelt or they coalesceto formconnectedgas channelsthat promote gasescape and even-tuallyfoamcollapse(Eichelbergeretal.,1986).Inthecaseof iso-latedbubbles, the magma may be overpressurized withgas and trigger a highly explosive eruption, whereas in the caseof bub-bleinterconnection, themagma mayoutgas anderupt effusively. Thus, investigating the whole degassing process (i.e. the dynam-icsof bubble nucleation, growth,coalescence, and evacuation) is

*

Correspondingauthor.Tel.:+33238255252;fax:+33238494476.

E-mailaddress:caroline.martel@cnrs-orleans.fr(C. Martel).

crucial toour understanding ofthe transitions ineruptive styles. Since the pioneeringstudyof Sparks (1978), degassingand bub-ble dynamics in silicic melts have been extensively investigated throughexperiments(e.g.Navonetal.,1998; Gardneretal.,1999; LarsenandGardner,2000; MartelandBureau,2001; Gondéetal., 2011)andnumericalmodels(e.g.Toramaru,1989,1995;Barclayet al.,1995;ProussevitchandSahagian,1996,1998).

rhyolitic magmas (e.g. Eichelberger et al., 1986; Westrich and Eichelberger,1994;Klug andCashman, 1994, 1996;Bloweretal., 2002; Larsenet al., 2004; Burgisser and Gardner, 2005; Gardner, 2007; Degruyteretal.,2010; Castroetal., 2012a; Takeuchietal., 2009). Up to now, bubble nucleation and growth are the most understood processes, thanks to combined natural, experimental, andnumerical studies.The conditions ofbubble coalescence and foamcollapsearesofarmorepoorlyunderstood,althoughdirectly linked to the ability of the magma to erupt explosively or effu-sively.

In order to assess the respective roles of parameters such as melt composition,pressure,andtemperature, onthemechanisms and timescales of bubble coalescence and outgassing, we con-ducted experiments of decompression-induced degassing in hy-dratedrhyoliticorrhyolite-analoguemelts.Thechoiceofthe com-position is dictated by the fact that rhyolitic melts commonly represent the chemical composition of the matrix that embeds phenocrystsinandesites,dacitesorrhyolites.Decompression sim-ulates magma ascent in a volcanic conduit. These new experi-ments complete aset ofpreviously published decompression ex-periments that were mainly designed for crystallization studies (Martel and Schmidt, 2003; Martel, 2012; Mollard et al., 2012). Thenewdataareacquiredusingthesame(orveryclose)hydrated rhyolitic melts as starting material, were decompressed in simi-lar devices, were processed similarly, so that both the new and previous datacovera largerangeofconditions importantfor ex-amining the degassing–outgassing processes. We determined the time-evolution ofvesicularity, bubble diameterandnumber den-sity,criticaltothecharacterizationofbubblegrowth,coalescence, andfoamcollapseasafunctionoftimeandbulkviscosity(which dependsonmeltcomposition,temperature,H2Ocontent,and bub-ble content). By focusing on long-timescale processes in decom-pressingrhyolite melts, weoffernewinsightsinto bubble coales-cenceandoutgassing,whichare probablythemostelusivestages ofthevesiculationprocess.

2. Experimentalandanalyticalmethods

2.1. Compositionofthestartingmaterials

Threeglass compositionshavebeenused asstarting material: two rhyolites(RHY andSHILL)anda haplotonalite (HTN).RHYis the composition of the rhyolitic matrix glass of the andesite of MtPelée,Martinique(RHYinwt%:75.7SiO2,13.1Al2O3,2.4CaO, 3.6Na2O,1.9K2O,2.5FeO,0.4MgO,0.1MnO,and0.3TiO2).RHY glasspreparationisdescribed inMartel (2012).SHILListhe com-positionof the rhyoliticmatrix glassof the andesite ofSoufriere Hills,Montserrat(SHILLinwt%:75.0SiO2,13.6Al2O3,2.5CaO,4.3 Na2O, 1.7 K2O, 2.0FeO, 0.4MgO, 0.2 MnO, and0.3 TiO2). SHILL glasspreparationisdescribedinMartel andSchmidt (2003).RHY andSHILLareso similarincompositionthatwe consideredthem directlycomparable.HTNisa4-componentsimplifiedcomposition ofarhyolite(HTNinwt%:78.7SiO2,14.1Al2O3,1.8CaO,5.4Na2O). HTN glasspreparationisdescribed inMollard etal. (2012). With respect to RHYandSHILL, HTN allows testing anycompositional differencesdueto eitheran enrichmentin SiO2 (

∼

3 wt%)orthe absenceofferromagnesianoxides.

2.2. Experimentalmethods

Thedecompressionexperimentswereallperformedin externa-lly-heated pressure vessels at temperatures between 850 and 875◦C from an initial pressure ( Pi) of 200 or 150 MPa to final

pressures ( Pf) varyingbetween 10and 60 MPa.The serieswere

decompressed either (i) quasi-continuously following decompres-sionratesfrom0.2to50 000 MPa/hor(ii)rapidly(57–1200 MPa/h)

Table 1

Summaryofthedecompressionseriesandtheirexperimentalconditions. Series name Starting

materiala T (◦C) Pi (MPa) Pf (MPa) Degassing styleb

850SYN50 RHY 850 200 50 SYN

850SYN30 RHY 850 200 30 SYN

850SYN10 RHY 850 200 10 SYN

850POST50 RHY 850 200 50 POST 850POST30 RHY 850 200 30 POST 850POST10 RHY 850 200 10 POST 875POST50 RHY 875 200 50 POST 875POST50HTN HTN 875 200 50 POST 860SYN50 SHILL 860 150 50 SYN

a _{RHYforMtPeléerhyolite,SHILLforSoufriereHillsrhyolite,andHTNfor}

haplo-tonalite(compositionsgivenintext).

b _{SYN for syn-decompressiondegassingand POST for post-decompression}

de-gassing,asdefinedintext.

to Pf followed by a dwell time ranging from4 to 672h at Pf.

The samples were quenched at Pf within

∼

1–2 s. Forease, the

continuous decompressions not followed by a dwell at Pf were

referred as to “syn-decompression” degassing and the rapid de-compressionsfollowedbyadwellat Pf werereferredasto

“post-decompression” degassing although a part of the degassing may havestartedduringdecompression.Adetailedguidetothe experi-mentalproceduresisgivenintheSupplementaryInformation_A,the different decompression series are presented in Table 1,and the experimentalconditionsaregiveninTable 2.

2.3. Texturalanalysisofthebubbles

Either half the sample-bearing capsule or the biggest pieces of thesample were mountedinepoxy resin foranalysis.Bubbles were investigated usingimagesfromoptical orscanning electron microscopes(SEM).TheimageswereprocessedusingGIMP open-sourcesoftware(GNUImageManipulationProgram)andconverted into binaryimagesforthe determinationoftheporosity (

Φ

)and bubblesizedistributionusingtheSPOsoftware[fabricanalysis us-ing theinterceptmethoddevelopedbyLauneauandRobin (1996)

andLauneauandCruden (1998)].SEMoropticalimages,together withtheir correspondingbinary images,are shownastime-series inSupplementaryInformation_B.Anexampleofonedecompression series is shown in Fig. 1. Our wish was to keep the samples as pristine aspossibletoavoidintroducingbias inthedegassing in-terpretation,butwesometimeshadtomanuallyseparatetouching bubbles where polishingplucking was suspected(e.g. sample D7 inSupplementaryInformation_Be).Wecheckedthat thisprocedure of manual bubble separation did not significantly impact bubble numberdensity(noshiftgreaterthanonelogunit)orbubblesize distribution. For the H2O-saturated samples, some bubbles were trapped in themelt during the hydration procedure.In the sam-ples decompressed in lessthan

∼

0.1 h,thesepre-decompression bubbleswereseveraltimeslargerthanthedecompression-induced ones andwe easily ruled them out fromthe analyses (e.g. sam-ple SH11 from the 860SYN50 series in Supplementary Informa-tion_Bh). However, they were no more distinguishable from the decompression-induced bubbles for longer times and were thus countedtogether.Nevertheless,theseinitial bubbleslikelyhave a negligibleeffectonthewholedegassingprocessbecausetheywere always inproportionbelow1 vol%andinnumberdensityseveral orders ofmagnitudelower thanthe decompression-induced bub-bles (from measurements carried out on the samples held at Pi

andnotsubjectedtodecompression).

Fromthebinaryimagesofthesamples,wedeterminedthe fol-lowingparameters:

–

Φ

,theporositydefinedasthearearatioofbubblestobubbles

Table 2

Experimentalandanalyticalconditions.

Sample Seriesa _{Experimental conditions}b _{Bubble analysis}c _Crystals

Ptime (h) Prate (MPa/h) TimeatPf (h) Φ (vol%) Dmax (μm) log Nv (m−3₎ ψ (vol%) RHY VP2 850SYN50 0.017 9000 0 5(4) 50(10) 13.75 7 VP3 850SYN50 0.25 600 0 24(4) 80(15) 13.88 9 VP14 850SYN50 2 75 0 27(4) 80(20) 14.32 9 VP5 850SYN50 10 15 0 23(4) 130(15) 13.05 11 VP8 850SYN50 48 3.12 0 29(4) 200(15) 12.64 13 VP6 850SYN50 120 1.25 0 27(4) 200(40) 12.20 13 VP10 850POST50 0.25 600 96 29(4) 250(40) 13.20 17 VP18 850SYN10 12.5 14.8 0 66(5) 500(100) 12.70 ∼10 VP13 850SYN10 151.2 1.22 0 61(5) 500(100) 12.42 12 D9 850SYN/POST30 1.5 113 0 47(5) 130(30) 13.51 8 D3 850SYN30 3 56.7 0 49(5)3D _{130(30) 13.71 10} D26 850SYN30 24 7.1 0 56(5)∗ 320(60) 12.14 1 D27 850SYN30 48 3.5 0 51(5)∗ 320(60) 12.00 6 D25 850SYN30 72 2.36 0 57(5)∗ 320(30) 12.08 10 D7 850SYN/POST10 1.5 127 0 80(5)∗ 500(100) 12.93 0 D33 850SYN10 72 2.6 0 60(5)∗3D _{510(100) 11.96 10} D38 850SYN10 144 1.32 0 63(5) 500(100) 12.21 15 D4 850SYN10 333 0.57 0 51(5) 200(100) 12.67 22 D35 850SYN10 960 0.20 0 58(5)∗3D _{320(60) 10.65 16} D20 850POST50 1.5 93 4 26(3) 200(60) 12.97 0 D2 850POST30 3 56.7 96 55(5) 320(60) 12.40 23 D8 850POST30 1.5 113 102 58(5) 320(80) 13.18 20 D34 850POST30 0.25 680 48 51(5)∗3D 400(80) 12.08 23 D39 850POST30 1.5 113 24 54(5)∗3D _{400(60) 12.11 6} D12 850POST10 1.5 127 95 62(5) 500(100) 11.51 27 D6 850POST10 1.5 127 329 43(5) 400(100) 13.02 >30 A1 875POST50 0.125 1200 0 37(5) 127(22) 13.24 0 A2 875POST50 0.125 1200 6 5(2) 51(11) 12.96 0 A3 875POST50 0.125 1200 24 2(2) 80(17) 11.66 0 A4 875POST50 0.125 1200 48 3(2) 80(18) 11.07 1 A5 875POST50 0.125 1200 72 9(2) 75(10) 11.22 1 A6 875POST50 0.125 1200 96 9(2) 75(10) 11.16 1 A7 875POST50 0.125 1200 168 5(2) 75(10) 11.16 3 A8 875POST50 0.125 1200 408 1(1) 50(10) 11.09 6 A9 875POST50 0.125 1200 672 3(2) 50(10) 11.07 12 HTN HTN1 875POST50HTN 0.125 1200 0 44(5) 200(40) 12.66 0 HTN2 875POST50HTN 0.125 1200 6 42(5) 320(70) 11.48 0.1 HTN3 875POST50HTN 0.125 1200 16 28(4) 320(70) 11.91 0.4 HTN4 875POST50HTN 0.125 1200 24 4(2) 80(15) 12.32 1 HTN5 875POST50HTN 0.125 1200 48 3(2) 80(15) 11.65 4

SHILL (fromMartel and Schmidt, 2003)

SH16 860SYN50 0.002 50 000 0 17(2) 6(1) 15.68 0 SH12 860SYN50 0.004 25 000 0 30(4)∗ 10(1) 15.30 2 SH11 860SYN50 0.02 5000 0 30(4)∗ 25(2) 14.93 10 SH1 860SYN50 0.2 500 0 35(4)∗ 100(15) 13.13 7 SH4 860SYN50 10 10 0 32(4)∗ 250(15) 12.05 8 SH3 860SYN50 120 0.8 0 35(4)∗ 400(50) 11.70 18 SH5 860SYN50 360 0.3 0 30(4)∗ 400(50) 11.44 16

a _{DecompressionseriesasdefinedinTable 1(thefirstandlastnumberoftheseriesgives T andP}

f,respectively;SYNandPOSTmeanssyn- orpost-decompression

degassing,respectively).

b _P

timeandPratearethedecompressiondurationanddecompressionrate,respectively.

c _{ΦistheporositymeasuredbySEMorXRCT(}3D_),with(∗_{)denotingthepresenceofpre-decompressionbubbles;Dmaxisthemaximumdiameterofthebubbles;the}

numbersinbracketsgivetheerrorresultingfromrepetitiveanalysesofatleastthreeimagesofthesamesample;Nv isbubblenumberdensitycalculatedafterHiggins

(2000);ψisthepercentageofmicrolitesrecalculatedonabubble-freebasis.

– Dmax,themaximumdiameterofthecontinuousdiameter

dis-tribution of discs having areas equivalent to the real bubble areas (thusnot considering large diameters that are discon-nected/isolatedfromthe histogramofthewhole population). Becausethebubblepopulationisassumedtobemono-modal andshowsmostlysub-sphericalsections,themaximum diam-eterrepresentstheequatorsectionandisassumedtothetrue diameterofthethree-dimensionalbubbles;

– Nv,thestereologicalbubblenumberdensitydeterminedfrom

the2DbubblediameterpopulationusingCSDCorrections soft-ware(Higgins,2000).

Five samples amongst the largest were imaged using XRay-computedtomograph(XRCT;PhoenixNanotom180,ISTO,Orléans, France),inorderto compare2Dwith3D porositiesandto model permeability. Processing any single image from the whole XRCT image stack suggests that 2D

Φ

have a maximum variability of

Fig. 1. SEMandassociatedbinaryimagesfor875POST50HTNforwhichtheHTNstartingmaterialwasrapidlydecompressedat875◦Cfrom200to50 MPa.Fromlefttoright, timespentatPf increasesfrom0.1to48 h(durationgiveninhourinbracketnexttosamplenumberasinTable 2);scalebarof500 μmforallimages.

Table 3 Permeabilitycalculation. Seriesa _Sampleb _Φc (vol%) logκd (m2₎ 850SYN10 D33 60(5) −10.2 850POST30 D39 54(5) −11.5 850POST30 D34 51(5) −12.1 850SYN30 D3 49(5) −13.2 a _{SeriesasinTable 1.} b _{SamplenumberasinTable 2.} c _{SampleporosityasinTable 2.}

d _{Permeability(averageovertheX–Y –Z directions)calculatedafterPalabos}

open-sourcesoftwareusingXRCTimages.

Boltzmann method] modified after Degruyter et al. (2010). This model requires the input of reconstructed XRCT images, thresh-oldedforporosity,andsimulatesadownwardsfluidinjectionfrom thetopface ofa cubicsample (whereasthe fourlateralfaces are consideredimpermeable),whichobeysDarcy’slaw.Byrotatingthe sample,permeabilityhasbeencalculatedinthethreespace direc-tions(Table 3).

3. Results

3.1. Syn-decompressiondegassing

Here we address samples with crystallinity

(ψ) <

20 vol%, which degassed in response to quasi-continuous decompression withoutdwelltimeat Pf (Table 2).

3.1.1. Finalpressureof10–15MPa

The bubbles in the 850SYN10 series (Supplementary Informa-tion_Ba)showcomparablesizes,verytortuousoutlines,andstrong features of coalescence with bending and dimpling melt–bubble walls as described by Castro et al. (2012a) (Fig. 2a). Note here thatweinterpretthesemelt–bubblewallcharacteristicsas coales-cenceratherthan quench-inducedbubbleresorptionasdescribed byMcIntoshetal. (2014),sincetheveryfastquenchtimesof1–2 s of our experiments prevent texture evolution during cooling.

Φ

startsdecreasingfrom80_±5 vol% foradecompressionduration of

∼

1 hto50–60_±5 vol% for30 hormore(bluecircles,Fig. 3a).The measured

Φ

of 80_±5 vol% isin agreement within errorwiththe gasfraction(

α

) of84 vol%calculatedforaclosed-system equilib-riumdegassing(Table 4).Duringthisphaseofgeneral

Φ

-decrease,

Dmax remains constant and high (

∼

500 μm) before dropping to

∼

200–300 μm after

∼

200 h (blue circles, Fig. 3b). Meanwhile, log Nv slightlydecreases from

∼

13 to 12 m−3 before drastically

droppingafter

∼

200 hofdecompression (downto1010.5m−3 for

∼

1000 h;bluecircles,Fig. 3c).

The permeability calculated in the sample decompressed in 72 his

∼

10−10.2 _m2_{(D33;Table 3).}

3.1.2. Finalpressureof30MPa

The bubbles in the 850SYN30series are sub-spherical for de-compression durations up to 48 h, butelongate anddisplay tor-tuous outlines when decompressed in 72 h (Supplementary Infor-mation_Bb). Bubble size visibly increases withincreasing decom-pression duration. Thesample decompressed within 1.5 hreveals several bubblecoalescence features withstretching anddimpling melt–bubble walls followingCastro etal.’s (2012a)nomenclature (Fig. 2b).

Φ

of47–49_±5vol% measuredintheH2O-undersaturated runsand

Φ

∼

55vol% measuredintheH2O-saturatedrunsarein agreement with their respective

α

within analytical error (

α

of 54–55 vol% fortheH2O-undersaturated runsand58vol% forthe H2O-saturatedones;Table 4)andthuscompatiblewithan equilib-riumdegassing(greentriangles,Fig. 3a).Fordecompression dura-tionsof1–10 h, Dmax is100–200 μm(greentriangles,Fig. 3b)and

log Nv is

∼

13

.

5 m−3 (greentriangles,Fig. 3c). Fordecompression

duration

>

10 h,

Φ

staysat aplateauvalue of 55 vol%(i.e. equi-librium degassing), while Dmax increases up to 320 μm (plateau

value at

∼

20 h;greentriangles,Fig. 3b)and Nv decreases

drasti-cally(greentriangles,Fig. 3c).

Thepermeabilitycalculatedinthesampledecompressedin3 h is

∼

10−13.2 _m2_{(D3;Table 3).}

3.1.3. Finalpressureof50MPa

The bubbles inthe 850SYN50seriesare sub-spherical and in-creaseinsize withincreasing decompressionduration. Theseries show overallveryfewfeatures ofbubblecoalescence

(Supplemen-taryInformation_Bc).

Φ

reaches 25–30 vol% within lessthan 1-h decompression time andstays at thisplateau value at least dur-ingdecompressiondurationsof100–200 h(reddiamonds,Fig. 3a). The

Φ

plateauvalueof 25–30_±4vol% is closeto

α

=

35 vol% ex-pectedforanequilibriumdegassing(Table 4).Duringtheperiodof

Φ

increase(decompressionduration

<

1 h),Dmax is

∼

80 μm(red

diamonds, Fig. 3b) and a maximum log Nv of 13

.

7–14

.

3 m−3 is

Fig. 2. Featuresofbubblecoalescenceinsamplesdecompressedat850◦Cover1.5hshowingbubble–meltwallthinningbybending,stretching,anddimpling,following

Castroetal.’s (2012a,2012b)nomenclature(a)D7from850SYN10and(b)D9from850SYN30.

Table 4

Gasfractionandviscositycalculations. Starting

glassa

Run T

(◦C)

Before decompression After decompression

Pi (MPa) Cwib (wt%) Psatc (MPa) logηid (Pa s) Pf (MPa) Cw fb (wt%) αc,e _logη meltd (Pa s) Caf _logη bulkg (Pa s) HTN HTN1–HTN5 875 200 5.7 200 4.3 50 2.6 0.42 5.9 1.5 4.2 SHILL SH1–SH16 860 150 4.9∗ 150 4.6 50 2.6 0.34 5.7 1.0 5.5 RHY A1–A9 875 200 5.7 200 4.2 50 2.6 0.42 5.7 1.5 4.1 RHY D2–D3 850 200 5.2 165 4.6 30 2.0 0.54 6.4 1.0 6.2 RHY D9 850 200 5.4 175 4.6 30 2.0 0.55 6.4 1.0 6.2 RHY D4, D6, D38 850 200 5.6 185 4.5 10 1.1 0.84 7.3 2.5 4.9 RHY D20 850 200 5.8∗ 200 4.4 60 2.9 0.34 5.8 1.0 5.9 RHY D25–D27, D34, D39 850 200 5.8∗ 200 4.4 30 2.0 0.58 6.4 1.0 6.2 RHY D7, D12, D33, D35 850 200 5.8∗ 200 4.4 10 1.1 0.85 7.3 2.5 5.1 RHY VP2–VP10, VP14 850 200 5.1 160 4.7 50 2.6 0.35 6.0 1.0 5.8 RHY VP13, VP18 850 200 5.1 160 4.7 15 1.3 0.74 7.0 2.0 4.6 Pel-Pumice 875 240 6.5 240 4.1 25 1.8 0.69 6.3 2.5 5.2 Pel-Surge 875 240 6.5 240 4.1 15 1.3 0.80 6.7 1.5 3.8 Pin-White 780 220 6.0 220 5.0 11 1.2 0.82 8.0 2.5 4.9 Pin-Grey 790 220 6.0 220 4.9 11 1.2 0.82 7.9 2.5 4.8

a _{HTN,SHILL,andRHYasinTable 1;Pel-PumiceandPel-SurgearetheresidualmeltsofMtPeléeP1Plinianpumiceanddome-relatedsurges,respectively;Pin-Whiteand}

Pin-GreyaretheresidualmeltsofMtPinatubo1991whiteandgreypumices,respectively.

b _C

wi isinitialmeltH2Ocontentcalculatedat Pi andT eitherafterLiuetal. (2005)fortheH2O-saturatedorfollowingtheprocedureofMartel (2012)fortheH2

O-undersaturatedsamples;(∗)denotesthepresenceofpre-decompressionbubbles;Cw f isthemeltH2OcontentcalculatedatPf andT afterLiuetal. (2005). c _{SaturationpressurecalculatedafterLiuetal. (2005)forC}

wi.

d _{MeltviscositycalculatedatT afterGiordanoetal. (2008)toaccountforthedifferencesinstartingmeltcompositions(anhydrous),decreasedbytheviscositydifference}

betweentheanhydrousandhydroushaplogranitecalculatedafterHessandDingwell (1996)(ηiforCwiandηmeltforCw f). e _{Gasfractioncalculatedforaclosed-systemequilibriumdegassingafterJaupartandTait (1990)asα= (1+ρ}

w/((Cwi−Cw f)∗ρm))−1,withρmthemeltdensity(taken

as2.2 g/cm3_{;Knocheetal.,1995)andρ}

wwaterdensitycalculatedafterSaulandWagner (1989). f _{Capillarynumbersetto1.0at850}◦_{Cand30–50 MPa(η}

bulk∼ηmelt),1.5at875◦C,and2.5forPf=10 MPa (seetext). g _{BulkviscositycalculatedafterEq.(1)(seetext).}

Dmax starts increasing up to200 μm (reddiamonds, Fig. 3b) and

Nvdecreasesdrastically(reddiamonds,Fig. 3c).

The bubbles in the 860SYN50 series (Supplementary Informa-tion_Bh)aresub-sphericalfordecompressiondurationupto10h,

in-Fig. 3. Timeevolutionofthesyn-decompressiondegassing(filledsymbols;timeisdecompressionduration)andpost-decompressiondegassing(outlinedsymbols;timeis timespentatPf)showing(a)Φ,(b)Dmax,and(c)Nv,for850SYN50,850SYN30,850SYN10,and860SYN50;Themaximumerrorofthemeasurementsisgivenbytheblack

verticallinesontheright-handsideofthediagrams,exceptin(b).In(a),labelsrefertotherunnumbersasinTable 1andthedash–dotlinesshowαasgiveninTable 2; in(c),onlythesampleswithcrystalcontentψ≤16 vol% areshown(seetextforjustification);860SYN50datafromMartelandSchmidt (2003).(Forinterpretationofthe referencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

creasesupto30–35_±4 vol% afteradecompressionofonly12 s(red crosses, Fig. 3a). This

Φ

range is closeto

α

=

34 vol% calculated foran equilibriumdegassing (Table 4). Dmax constantly increases

uptoabout400 μmafterabout120h(redcrosses,Fig. 3b),while

Nv constantly decreases from 1015.7 to 1011.4 m−3 (red crosses, Fig. 3c).

3.2. Post-decompressiondegassing

We now address samples with crystallinities

<

20 vol% that mostly degassed during dwell times at Pf after a rapid

decom-pression.Althoughmostly occurringduring dwelltimeat Pf,the

degassing likely startedduring decompression, so that the refer-encetimestartsatthebeginningofthedecompression(Table 2).

3.2.1. Samplesdecompressedat850◦C

These samples belong to the 850POST30 and 850POST50 se-ries.Allsamplesshowsub-roundedbubbles,evenaftermorethan 100 h spent at Pf, andmany bubble coalescence features

(Sup-plementaryInformation_Be andBf). For all samples,

Φ

is close to thecalculated

α

,i.e.54_±5 vol% against

α

=

58 vol% for850POST30 (green outlined triangle, Fig. 3a) and 26–29_±4vol% against

α

=

34–35 vol% for850POST50(redoutlineddiamonds,Fig. 3a).Log Nv

is

∼

12.0and

∼

13

.

1 m−3 _{inthesamplesdecompressedto30 MPa} and50 MPa,respectively(redoutlineddiamonds,Fig. 3c).

The permeabilitycalculatedinthe sample thatstayed 24 hat 30 MPais

∼

10−11.5m2(D39;Table 3).

3.2.2. Samplesdecompressedat875◦C(Pf

=

50 MPa;



P

/

t

=

1200 MPa

/

h)

Bubbles in the 875POST50 series are sub-spherical andshow coalescence featuresaftera decompressiondurationofonly 0.1 h (SupplementaryInformation_Bg). At this duration,

Φ

is 37_±5 vol%, close to

α

=

42 vol% expected at equilibrium (Table 4), Dmax is

∼

125 μm,andlog Nv is13

.

4 m−3 (redfilledcircles,Fig. 4).

How-ever, after

∼

6 h spent at Pf,

Φ

drops below 10 vol%, Dmax is

<

100 μm, and samples are nearly devoid of bubbles (log Nv

∼

11 m−3_{after2daysspentat P}

Fig. 4. Effectofmeltcompositiononthetimeevolutionofthepost-decompressiondegassingshowing(a)Φ,(b)Dmax,and(c)Nv,forRHYandHTNrapidlydecompressed

(P/t=1200 MPa/h)at875◦Cfrom200 MPatoPf=50 MPa (filledsymbols).RHYdecompressedat850◦C(P/t=93–600 MPa/h;outlinedsymbols)isalsoshown

tocomparewithRHYdecompressedat875◦C.Themaximumerroronthemeasurementsisgivenbytheblackverticallinesontheright-handsideofthediagrams,except in(b).In(a),labelsrefertotherunnumbersasinTable 1andthedash–dotlineshowsαasgiveninTable 2.(Forinterpretationofthereferencestocolorinthisfigure legend,thereaderisreferredtothewebversionofthisarticle.)

Bubbles in the 875POST50HTN series are sub-rounded, some-timeselongated,withincreasingsizesfordurationupto16h fol-lowedbyasizedecreasewithincreasingdwelltime (Fig. 1).

Φ

is 42–44vol% for6 h spent at Pf, inagreement with

α

=

42 vol%

expectedat equilibrium (Table 4), before dropping below 5 vol% for longer durations (blue triangles, Fig. 4a). For dwells shorter than24 h,Dmaxincreasesfrom

∼

200to325 μm,beforedecreasing

drasticallytolessthan100μmforlongerdurations(bluetriangles,

Fig. 4b).Log Nvdecreasesfrom12

.

7 m−3 after0.1hto

∼

11

.

5 m−3

after6 hatPf,beforeshowingapeakat12

.

3 m−3for1dayspent

atPf

=

50 MPa (bluetriangles,Fig. 4c).

3.3.Crystallizedsamples

Samplesthatcrystallized morethan20 vol%ofplagioclase mi-crolitesinresponsetolongdwelltimes(

>

2days)werenotplotted inFig. 3 andFig. 4 andwill be discussed separately

(Supplemen-taryInformation_C).Thepermeabilitycalculatedinonecrystallized samplefromthe850POST30seriesis

∼

10−12.1 _m2 _{(D4;Table 3).}

4. Discussion

4.1. Decipheringdegassingandoutgassingfromthetime-evolution curvesofporosity,bubblediameter,andbubblenumberdensity

Thedegassingprocessprimarilystartswithaneventofbubble nucleation.Atthetimeofdegassing,thestartingmelts(i)are crys-talfreeorcontainfewplagioclasemicrolitesthatdidnotrepresent preferentialsitesforbubblenucleation(HurwitzandNavon,1994) and(ii)donotcontainsignificantinitialbubbles(

<

1 vol%). There-fore, we assume a decompression-induced homogeneous bubble nucleation event. Bubble nucleation implies a drastic increase of

Nv,anincreaseof

Φ

becausegasisbeingexsolvedfromthemelt,

whereas Dmax stays nearlyconstant orincreasesslightly(because

bubblesarenotyetsignificantlygrowing).

Once nucleated,bubbles grow by the combinedeffect ofH2O diffusionfromthemelttothebubbleanddecompression-induced gas expansion. Bubble growth impliesincreasing

Φ

and Dmax at

constant Nv becausebubbles stop nucleating anddo not

Fig. 5. Bubblenumberdensitydependenceondecompressionrate(extractedfromFig. 3c,startingfromthehighestNv andshowingonlythesampleswithcrystallinity

<20 vol%thatexperiencedsyn-decompressiondegassing).TheNv−P/t relationshiphasbeendividedintothreelineartrends(boxwidthscorrespondtotheanalytical error)markingtheeffectsof(1)decompressionrate(pink),(2)bubblecoalescence(green),and(3)foamcollapse(blue).The[T06]linesrepresentthetrendcalculatedafter

Toramaru (2006)(seetext).(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

In-situ observations of degassing rhyolitic melts suggest that bubbles are not moving independently of the melt and stay roughly where they nucleated (Martel and Bureau, 2001), which isacombinedconsequenceofthehighviscosityofrhyoliticmelts and short timescale of the experiments. During bubble growth, bubble–meltwallsthin, eventually leading tobubblecoalescence. The bending, stretching, and dimpling of bubble–melt walls (as in Fig. 2) are common features in experimental samples decom-pressedinlessthan

∼

1 hsto Pf

≤

50 MPa andrapidlyquenched

to prevent bubble relaxation and recovery to a spherical shape (Castro et al., 2012a). Bubble coalescence implies an increase of

Dmax andageneraldecreaseofNv,while

Φ

increasesorremains

constant depending on whether H2O is still exsolving from the meltornot.Byextensivecoalescence,wemeanacombineddrastic decreaseofNv andincreaseofDmaxwithtime,whichforinstance

isthecaseforrunD26butnotrunD3(greentriangles,Fig. 3)even ifcoalescencefeaturesareobservedinthissample(Supplementary

Information_B).

Coalescingbubblesmayeventually generategaschannels lead-ing toa pervasive interconnectivity. Atthis pointof high perme-abilitytofluids,the magmaeitherstaysasafoam(e.g.,pumices) orthegasescapesandthefoamcollapses(e.g.obsidiandomesor rhyoliticlava flows;Westrich andEichelberger,1994). Outgassing andfoamcollapseimplyadecreasein

Φ

commonlyaccompanied bydrasticdecreasesinDmax andNv.

4.2. Syn- versuspost-decompressiondegassing

The fewpost-decompression degassingexperimentsshow val-uesof

Φ

, Dmax,and Nv that are comparablewithin error tothe

syn-decompressiondegassingones(outlinedversusfilledsymbols,

Fig. 3), which maybe attributed to a similar degassing behavior over the time period investigated. This similarity could suggest that thedegassing process mainlydependson time available, re-gardlessof whetherdegassing occursduring or after decompres-sion,andonlytoalesserextentondecompressionrate.

Onthe other hand,bubble numberdensityatthe moment of nucleation is predicted to correlate positively with decompres-sionrate(Mourtada-BonnefoiandLaporte, 1999, 2004). Nv

mea-sured in the series potentially recording bubble nucleation (i.e. 860SYN50and850SYN50;red crossesandred diamonds,

respec-tively, in Fig. 3) at a decompression rate of 100 MPa/h suggest log Nv

=

14

.

0±0.5. This value is in good agreement with the es-timationfromin-situdecompressionexperimentsusingarhyolitic meltfree ofbubblecoalescence(Gondéetal.,2011).Applyingthe trendcalculatedbyToramaru (2006),weextrapolatedNv tolower

decompression ratesin order to calculate Nv at the time of

nu-cleationforsamplesaffectedbylatercoalescence[usingmeltH2O diffusion of 2

.

2

×

10−11m

/

s calculated afterZhang andBehrens (2000) andgas/melt interfacialtensionof0.13 N/m calculated af-ter Bagdassarov et al. (1999) at 850◦C and 5.9 wt% melt H2O; note that using a gas/melt interfacial tension of 0.05 N/m that covers the range ofhydrous magmas (Mangan and Sisson, 2005; Gondéetal.,2011) increasesNvby1logunit].

Ourexperimentaldatathereforeshowthreedifferenttrends: (i) A drastic decrease of Nv while Dmax increases (pink trend

in Fig. 5), which is in good agreement with the relation-ship calculatedafter Toramaru (2006) starting withlog Nv

=

15

.

5 for a decompression rate of 50 000 MPa/h (860SYN50) and log Nv

=

14

.

5 for a decompression rate of 100 MPa/h

(850SYN50).Thistrend isthusattributedtotheeffectof de-compressionrateon Nv wherebubblecoalescenceisnot

sig-nificant;

(ii) A slightdecrease of Nv while Dmax isconstant orincreases

where bubbles start to coalesce extensively (green trend in

Fig. 5); atthispoint of extensivecoalescence, the calculated

Nv

− 

P

/

t relationshipsnolongerhold.Theobservedtrend

thus reflects the evolution of Nv with bubble coalescence

quasi-independentlyof



P

/

t (nearlyhorizontaltrends). (iii) A drasticdecreaseofbothNv andDmax(blue trendinFig. 5)

defined by 850SYN10 at a very slow decompression rate (0.2 MPa/h),mostlikelyrevealingbubblelossduetofoam col-lapse.

4.3. Viscositycontrolonthetimeevolutionofbubblegrowth, coalescence,andoutgassing

4.3.1. Meltcomposition

Fig. 4 revealsa strongdifference inthe degassingbehavior of HTNandRHYcompositions.Both seriesstarteddegassingin sim-ilarconditions(comparable Dmax andNv),butthecollapseofthe

bubblyfoam (indicatedby a drasticdropin

Φ

and Dmax)occurs

within the first 6 h for RHY and after 12 h for HTN. Moreover, RHYdoesnotshowevidenceofbubblegrowthorcoalescence (de-creasingDmax withoutNv increase)beforeoutgassing,whileHTN

does(constant

Φ

andincreaseof Dmax whileNv decreases).

Thedifference inviscositybetweenthe two startingmaterials calculatedusingthemodelofGiordanoetal. (2008) isonly

∼

0.2 logunit, mainly dueto their difference of3 wt% SiO2 (Table 4), astheeffectofferromagnesiancomponentsonbothmeltviscosity andbubble–meltsurface tensionisnot known.Wespeculatethat the difference in silica and ferromagnesian content may explain the different outgassing timescales, with the more polymerized, moreviscous, andferromagnesian-free HTN degassing in equilib-riumlonger(moretimeforbubblegrowingandcoalescence)than thesilica-poorer,lessviscous,andferromagnesian-bearingRHY.

4.3.2. Temperature

Ourresults only partially reveal the effect of temperature on thedegassingbehavior, suggestingthat increasingtemperatureby 25◦Cfavorsearlyoutgassingandsubsequentfoamcollapse.Fig. 4

showsthatexperimentsat850◦Cmaintainequilibriumporosities, constant Dmax,andconstant Nv formorethan 4days,differently

tothoseat875◦C,whichexperiencesevereoutgassingandlossof nearlyall thebubblesafter48 h (drastic dropof

Φ

anddecrease ofNv ofmorethan2logunits).

4.3.3. Finalpressure

Finalpressurecontrolstwo crucialparametersthat affectmelt viscosity:melt H2Ocontentandbubblecontent.The H2Ocontent of the melt only varies from 1.1 to 2.9 wt% for the investigated range of Pf (Table 4) andH2O most strongly influences

viscos-ityforcontents

<

1.0wt% (HessandDingwell, 1996). Incontrast, vesicularityvariesovera widerangefrom0to80 vol%(Table 2). Samplesdecompressed to Pf

=

30 and 50 MPa show equilibrium

degassingfollowed by bubblecoalescence without foam collapse fordurations up to 300 h (Fig. 3). On the contrary, samples de-compressedto10 MPastartoutgassingafter1.5 handshow exten-sivebubblecoalescencefollowedbybubblecollapse.Thissuggests that the rhyoliticfoam generated by decompression to pressures

<

10 MPa (with

Φ >

∼

60 vol% and Dmax as large as

∼

500 μm)

cannotsurvivelongerthana coupleofhours inourexperimental conditions.Ourexperimentsthereforeseemtodefinealimiting

Φ

thatcontrolsfoamlifetimebypromotingearlybubblecoalescence, formationofchannelsofgasthateventuallyescapefromthemelt, producingfoamcollapseatlowpressures.

4.3.4. Viscosityofthebubblymelt

Thevariations intemperature,melt composition,H2Ocontent, and bubble content can be gathered by the rheology of a mul-tiphase suspension, making thus likely a relationship between the outgassing timescale andbulk viscosity (melt

+

bubbles) of the sample. The viscosity of the bubbly melt (

ηbulk

) is differ-ent fromthat of the pure melt (

η

melt) because bubbles may act

assolid ordeformable particles that increase ordecrease viscos-ity, respectively, depending on bubble size and shear conditions (Llewellin et al., 2002a; Rust and Manga, 2002). In our experi-mentalconditions,thesamplesareunlikelytoundergoverylarge strain. At low

Φ

,

η

melt may be relevant to account for the

out-gassing timescale. In highly porous samples, however, elongated bubbles(Figs. 1 and 2a)suggest deformation(e.g.bubbles repuls-ing each other to accommodate their volume increase), and the outgassingtimescaleclearly showsinconsistencieswherereported

as a function of

η

melt. For instance, samples decompressed to

10 MPahavethehighest

η

meltduetotheirlowresidualH2O

con-tent(Table 4)butoutgaswellbeforeother samplesdecompressed to 30 or 50 MPa. Moreover, the large difference in outgassing timescales (

∼

90 h in samples decompressed to 50 MPa) cannot simplybeaccountedforbythe0.1–0.3logunitdifferencein

η

melt

between 850 and 875◦C (Table 4). Therefore, other parameters than

η

melt that are affected by temperature (surface tension at

bubble–meltinterface?)maybeconsideredtorendertheobserved degassingtimescale. Besides

η

melt,thekey parameterscontrolling

the viscosity of a bubbly suspension are bubble volume content andthecapillarynumber,Ca,whichdescribestheratioofthe vis-cous stressto therestoringstress: Ca

= (

ηmelt

γ

˙

)/(

σ

/

a

)

, where

γ

˙

isshear strain-rate,

σ

is thesurface tension atthebubble–liquid interface,anda isthenon-deformedbubbleradius.

Following the review andanalysis of the rheology of bubble-bearingmagmasofMaderetal. (2013),weusedLlewellinetal.’s (2002b) equation forsteady, simple-shearing flow giving

η

bulk as

follows:

η

bulk

=

η

melt

×

η

r,∞

+

₁

η

r

_{+ (}

,0

−

_{K Ca}

η

r,

₎

∞m (1)

where

η

r,0

= (

1

−

α

)

−1 is the limiting relative viscosity of the concentrated suspensions at low Ca,

η

r,∞

= (

1

−

α

)

5/3 is the limitingrelative viscosity of theconcentrated suspension athigh

Ca, K

=

6

/

5, and m

=

2 for a monodisperse bubble distribution. We cannot precisely determine Ca for our experiments, because (i) strainrateisnotmeasured,(ii)wewouldneedtheradiusofthe non-deformedbubblepriortocoalescence,butbubblesalready ex-tensivelycoalescedinthesamplesdecompressedto10 MPa.Even approximating Ca by theTaylor deformation parameter(valid for small Ca where bubbles are ellipsoidal and calculated using the length of thesemi-axes of the ellipsoidthat best fits the bubble shape)wouldbeuseless,sincebubbles mayhavehadtimeto re-lax and recover a pseudo-spherical shape before quenching. We thereforesetCa to

∼

1.1fortheexperimentsat850–860◦C,30and 50 MPa,inordertokeep

η

bulk closeto

η

melt(i.e.nosignificant

ef-fectofthebubbles).Ca wassetto1.5forthetwoseriesat875◦C andto 2.5forthe samplesdecompressed to 10 MPa, inorder to obtain

η

bulk

<

η

melt. We assume that both hightemperature and

large bubble size increases Ca by facilitatingbubble deformation and decreasing liquid–vaporsurface tension.The calculated

η

bulk

using(1)arereportedinTable 4.

Reported in

η

bulk-time space, the data allow identification of

different stages of the decompression-induced degassing process (Fig. 6).Thestagewherethebubblesareinaprocessofnucleation andgrowth(

Φ <

α

;nocoalescence)isrestrictedtotimes

<

0.03 h for

η

bulk

∼

105–6Pa s.Forlongerduration,degassingreaches

equi-librium (

Φ

∼

α

), with a threshold of extensive coalescence oc-curring between0.1h for

η

bulk

∼

104Pa s and

∼

10h for

η

bulk

∼

107Pa s. Outgassing starts where

Φ

becomes lower than

α

(re-flectinggasloss).Theoutgassingfieldisrestrictedtolow

η

bulkand

longdurations,i.e.from5 to10hfor

η

bulk

<

104 Pa s to

>

1000h

for

η

bulk

∼

105.5Pa s.Foam collapsedelimitsan area from

∼

10h

for

η

bulk

∼

104Pa s to

∼

1000hfor

η

bulk

∼

105Pa s.

Anotherparameter (beyondthe scopeof thispaper)that may shiftbulk viscosityisthepresenceofcrystals,whichisdiscussed inSupplementaryInformation_C.

4.4. Timerelationshipsbetweenpermeability,bubblecoalescence,and foamcollapse

Thepercolationtheoryforrandomlyplacedmonodisperse bub-bles predicts a strong increase in permeability at

∼

30 vol% (e.g.

Fig. 6. Effectoftheviscosityofthebubblymeltonthedegassingtimescale.ηbulkhasbeencalculatedusingEq.(1)(withηmelt,α,andCa giveninTable 4).Thedegassing

processhasbeendividedintothreemainstages:bubblenucleationand growthforwhichΦ <αatshortdegassingdurations,equilibriumdegassingforwhichΦ ap-proximatesα(includingaphaseofmajorcoalescence),andoutgassingforwhichΦ <α(includingaphaseoffoamcollapse)atlongdegassingdurations.Thefilledstars representthepumiceandsurgesamples(Pel_PumiceandPel_Surge)fromtheP1(650B.P.)eruptionofMtPeléeandtheoutlinedstarsrepresentthewhiteandthegrey pumice(Pin_WhiteandPin_Grey)ofthe1991PlinianpumicefallsofMtPinatubo(seetext).

from30up to 80 vol% (Blower,2001). The permeability–porosity relationships oftenfollow powerlawrelationships withan expo-nentdependingontheporegeometry(e.g.Wrightetal.,2009).

Otherfactorshavebeendemonstratedtocontributetothe vari-ationofthepercolationthresholdinmagmas,suchasmelt viscos-ity,aperturesizebetweenpores(e.g.Bouvet deMaisonneuveetal., 2009),shearstresses(e.g.Okumuraetal.,2009),distinctstructures between expanding and collapsing bubbles (Rust and Cashman, 2004), andbubble size distribution. We could not systematically investigatethepermeabilityevolutionwithporosityinoursamples butwewereabletomeasurepermeabilityof3logunitdifferences for sampleswith comparable porosities of 50–60 vol% (Table 3). Thissuggeststhat inour experiments,time isan additional con-trolling factorthat allows the vesicular melt to coalesce,become permeableandoutgas.

A

κ

threshold of

∼

10−13_m2 _at

_Φ

_∼

_{50 vol% for vesicular} foamjustbeforeextensivecoalescenceisprovidedbyrunD3from 850SYN30 (Fig. 6): this sample has a

Φ

of 49_±5vol% and the bubbles are sub-spherical and uniformly sized. Such a low con-nectivity, in the range of natural vesicular pyroclasts (commonly between 10−14 _{and 10}−11_m2_{; Mueller et al., 2005), associated} to a relatively highporosity, maypartly resultfrom the uniform bubblesizedistributioninthesample(aspreviouslysuggestedfor breadcrustbombs; Muelleret al., 2005), but alsofrom the short experimentalduration(3 hatfinalpressure).

Runs D39 and D34 from 850POST30 have

Φ

of 51_±5 and 54_±5 vol%, respectively, and show strong coalescence features within 2 days.

κ

∼

10−12m2 calculated for both samples gives a typical value forfoamed (

Φ

=

50–60 vol%), permeable, but not yet outgassed, samples(Fig. 6). Similarly atlower

η

bulk,D7 from

850SYN/POST10isahighlyporous(

Φ

=

80 vol%)andlikelyhighly permeable sample (Fig. 2a) that did not collapse within 1.5-h (Fig. 6).

κ

∼

10−10m2 calculatedin D33 from 850SYN10gives a thresholdvalueforanoutgassedsamplethatdidnotfullycollapse in 3days.These resultshighlight that

κ

alone isnot a sufficient criteriontodeciphergaslossandfoamcollapse,andthat

η

bulkand

time are crucial parameters to take into account to understand whetherhighly permeablesilicicfoams can collapseafter tensor hundredsofhoursspentatfinalpressure.

4.5. Implicationsforexplosiveversuseffusiveeruptiondynamics

4.5.1. Experiment-naturescaling

Theapplicationoftheexperimentalresultstonaturalcases re-quires much thought regarding scaling of the process of bubble connectivity development. From a strictly physical point of view andignoringanybordereffects,outgassingandfoamcollapse oc-curwhen the gas connectedness is pervasive. It is reasonable to believe that outgassingand foamcollapse timescalescompare in small (sample capsules) and large volumes (e.g. natural volcanic conduits),sincebubblecoalescenceproceedssimultaneouslyatthe sampleinteriorandedge.Asecondconcernistherole ofthe ex-perimentalcontainerontheabilityofgastoescapethemelt.The difficult question to answer is whether the processes of bubble connection,outgassingandfoamcollapsearesloweddownbythe impermeablecapsule.Oursimulationsareprobablynotunrealistic becauseimpermeablecountry-rocks atconduitwallsalsoexistin nature,i.e.sealedbysilica-phasedeposits(Sparks,1997) ormelted throughviscousheatingduringhighstresses(Hessetal.,2008).

The major difference between experimental and natural con-ditions in a volcanic conduit is the amount of deformation ex-perienced by the bubbly melt. In a volcanic conduit, large de-formation is expected at conduit walls. This would make the capillary number a crucial parameter in the bulk viscosity cal-culationand the degassing process. Indeed, the gaspermeability of a non-deformed magma is a few to several orders of mag-nitude lower than for deformed and flowing magmas, in which outgassingisthereforedrasticallyenhanced(Okumuraetal.,2012; Toramaru,2014).Yet,in-situdeformationexperimentsshowedthat shearlocalizesalongthesampleedges leavingthesampleinterior poorly-shearedandpoorly-outgassed(Okumuraetal.,2013).Thus, ourexperimentsmaybe relevantforsimulatingdecompressionof theinteriorofamagmacolumn.

In nature, however, foam collapse is not the only process to generate effusive silicic volcanism. Recent works on explosive– effusive transitions haveexplored the role ofshear fracturing in causing non-explosive magma degassing (e.g. Gonnermann and Manga,2003;Castroetal.,2012b,2014).

4.5.2. ApplicationtothecaseofMtPelée,Martinique

TheP1eruption(650B.P.)ofMtPeléeproducedasurgedeposit ofdenseclastsfollowedbyclimacticPlinianpumicefallout.P1 pre-eruptionconditionswereestimatedtobe

∼

875◦C,

∼

200 MPa,and 6

.

5_±0.5wt% H2Odissolved in the melt (Martel et al., 1998). The pumiceresidualmelt isrhyoliticandshowstypical H2Ocontents of 1

.

8_±0.3 wt%, suggesting a calculated

α

for an assumed close-system equilibrium degassing of 69_±5 vol%, in good agreement withthe measured

Φ

of 71_±4vol% (Martel et al., 2000). The P1 Plinianresidualmelthasan

η

bulk of105.2 Pa s calculatedafter(1)

foraCa of1.5(875◦C–Pf

=

25 MPa;Table 4)andPlinianevents

commonly last only few hours (Rutherford and Gardner, 2000; Castro and Dingwell, 2009), which make the P1 pumice plot in the“Equilibriumdegassing–Extensivecoalescence”field ofFig. 6

(veryclose toD7 sample that is texturally very similar; Fig. 2a). This is in a good agreement with the fact that pumice samples show strong coalescence features but no significant bubble loss (

Φ

∼

α

).

In contrast, the dense clasts from the surges (1902 May 8th or P1) and block-and-ash flows (1902 and1929), for which the eruption durations have been estimated experimentally to more than 2 days (Martel, 2012), show

Φ

of

∼

35 vol% (Martel et al., 2000). This suggests outgassing giventhat the pre-eruption con-ditionsare similartothose oftheP1Plinian event(Martel etal., 1998).AlthoughH2Ocontentsoftheresidualmeltsaredifficultto estimatebecausethematrixesextensivelycrystallized,contentsup to 1.3 wt% H2Ohave been measured in the glassy matrix ofP1 surgelithics(Marteletal.,2000);thiswouldgivean

α

of80 vol%. The vesiculated P1surge residual melt hasan

η

bulk of 103.8Pa s

calculated after (1) for a Ca of 2.5 (875◦C – Pf

=

15 MPa; Ta-ble 4),andthereforeplotsinthe“Outgassing–Foamcollapse”field ofFig. 6.This corroboratesthe hypothesis ofa combinedcontrol of bulk viscosity and degassing timescale on the transition from pumicestobubble-deficientpyroclastsatMtPelée.

Permeability measurements are significantly different for the P1 Plinian pumices and 1929 dome samples: 10−12 and 10−9.5 to−12_m2_{, respectively (Jouniaux et al., 2000). These} per-meabilitiesareinagreementwithourvaluesobtainedfor(i) sam-ples degassed at equilibrium and showing bubble coalescence (10−11.5 to−12.1 m2) and (ii) outgassed and collapsed samples (

<

10−10.2m2).

Nevertheless,weshowintheSupplementaryInformation_C how

microlitecrystallization mayplay a role indegassing, outgassing, and foam collapse. The 1902 surge magmas of Mt Pelée crys-tallized microlites at very shallow level (Martel and Poussineau, 2007),therefore,theinfluenceofaframeworkofbothphenocrysts andmicrolites(

∼

45and

∼

30 vol%,respectively,forMtPeléesurge pyroclasts)should betakenintoaccount inthecalculation ofthe bulkviscositytobetterassesstheglobaldegassingprocess.

4.5.3. ApplicationtothecaseofMtPinatubo,Philippines

The climactic event of Mt Pinatubo, 15 June 1991, lasted for 9 h (Wolfe and Hoblitt, 1996) and produced two varieties of Plinian pumice: a white phenocryst-rich and a grey phenocryst-poor pumice (Hoblitt et al., 1996). The pre-eruptive conditions wereestimatedtobe

∼

780◦C,200_±20MPa,andamelt H2O con-tent

∼

6.0 wt% (Scailletand Evans, 1999). H2Omeasurements in the rhyolitic residual glass of the white pumice give an average of 1.2 wt% (Borisova et al., 2006), indicating efficient degassing, corroboratedbyhighvesicularitiesof

∼

80vol%and

∼

70 vol% mea-suredinwhiteandgreypumiceclasts,respectively(Polacci etal., 2001). The textural differences of the matrix of white and grey pumice (content and number density of bubbles and microlites) was interpretedasresultingfromdifferentflowconditions inthe volcanic conduit (Polacci et al., 2001): strong viscous dissipation at the conduit wallswas proposed to generate the grey pumice by reheating(Polacci etal., 2001). The intensityofthe reheating isnotclearlyquantified,butrheologicalexperimentsina viscome-tersuggesttemperatureincrease

<

10◦Cforappliedstrainrateson theorderof10−3_s−1_{(Hessetal.,2008).}

Thevesicularmeltfromthewhiteandgreypumicehasa

η

bulk

of104.9 _and104.8 _{Pa s,respectively,calculatedafter(1)foraCa of} 2.5( Pf

=

11 MPa)with(i) their respectiveresidualmelt

compo-sitions (Polaccietal., 2001), (ii)780and790◦C,respectively,and (iii)

α

of0.82calculatedforaresidualmeltH2Ocontentof1.2 wt% (Table 4). Both the white andgrey samples plotin the “Equilib-riumdegassing–Extensivecoalescence”fieldofFig. 6(closetothe outgassingfield),inagreementwithpumiceoussamples.This sug-geststhat 1991MtPinatubomagmas didnotsignificantly outgas dueto their rather highviscosity(and low temperature), despite theirhighvesicularities.

Acknowledgements

Appendix A. Supplementarymaterial

Supplementarymaterialrelatedtothisarticlecanbefound on-lineathttp://dx.doi.org/10.1016/j.epsl.2014.12.010.

References

Bagdassarov,N.S.,Dorfman,A.,Dingwell,D.B.,1999.Effectofalkalis, phosphorus andwateronthesurfacetensionofhaplogranitemelt.Am.Mineral. 85,33–40. Barclay,J.,Riley,D.S.,Sparks,R.S.J.,1995.Analyticalmodelsforbubblegrowth

dur-ingdecompressionofhighviscositymagmas.Bull.Volcanol. 57,422–431. Blower,J.D.,2001.Factorscontrollingpermeability-porosityrelationshipsinmagma.

Bull.Volcanol. 63,497–504.

Blower,J.D.,Keating,J.P.,Mader,H.M.,Phillips,J.C.,2002.Theevolutionofbubble sizedistributionsinvolcaniceruptions.J.Volcanol.Geotherm.Res. 120,1–23. Borisova,A.Y.,Pichavant,M.,Polvé,M.,Wiedenbeck,M.,Freydier,R.,Candaudap,F.,

2006.Traceelementgeochemistryofthe1991Mt.Pinatubosilicicmelts, Philip-pines:implicationsforore-formingpotentialofadakiticmagmatism.Geochim. Cosmochim.Acta 70,3702–3716.

BouvetdeMaisonneuve,C.,Bachmann,O.,Burgisser,A.,2009.Characterizationof juvenilepyroclastsfrom theKosPlateau Tuff(Aegean Arc):insights intothe eruptivedynamicsofalargerhyoliticeruption.Bull.Volcanol. 71,643–658. Burgisser,A.,Gardner,J.E.,2005.Experimentalconstraintsondegassingand

perme-abilityinvolcanicconduitflow.Bull.Volcanol. 67,42–56.

Castro,J.M.,Bindeman,I.N.,Tuffen,H.,Schipper,C.I.,2014.Explosiveoriginofsilicic lava:texturaland δD–H2Oevidencefor pyroclasticdegassingduring rhyolite

effusion.EarthPlanet.Sci.Lett. 405,52–61.

Castro,J.M.,Burgisser,A.,Schipper,C.I.,Mancini,S.,2012a.Mechanismsofbubble coalescenceinsilicicmagmas.Bull.Volcanol. 74,2339–2352.

Castro,J.M.,Cordonnier,B.,Tuffen,H.,Tobin,M.J.,Puskar,L.,Martin,M.C.,Bechtel, H.A.,2012b.Theroleofmelt-fracturedegassingindefusingexplosiverhyolite eruptionsatvolcanChaiten.EarthPlanet.Sci.Lett. 333–334,63–69.

Castro,J.M.,Dingwell,D.B.,2009.RapidascentofrhyoliticmagmaatChaiten vol-cano,Chile.Nature 461,780–783.

Degruyter,W.,Bachmann,O.,Burgisser,A.,2010.Controlsonmagmapermeabilityin thevolcanicconduitduringtheclimacticphaseoftheKosPlateauTufferuption (AegeanArc).Bull.Volcanol. 72,63–74.

Eichelberger,J.C.,Carrigan,C.R.,Westrich,H.R.,Price,R.H.,1986.Non-explosive sili-cicvolcanism.Nature 323,598–602.

Gardner,J.E.,2007.Bubblecoalescenceinrhyoliticmeltsduringdecompressionfrom highpressure.J.Volcanol.Geotherm.Res. 166,161–176.

Gardner,J.E.,Hilton,M.,Carroll,M.R.,1999.Experimentalconstraintsondegassing ofmagma:isothermalbubblegrowthduringcontinuousdecompressionfrom highpressure.EarthPlanet.Sci.Lett. 168,201–218.

Giordano,D., Russell,J.K.,Dingwell,D.B., 2008. Viscosityofmagmaticliquids: a model.EarthPlanet.Sci.Lett. 271,123–134.

Gondé,C.,Martel,C.,Pichavant,M.,Bureau,H.,2011.Insitububblevesiculationin silicicmagmas.Am.Mineral. 96,111–124.

Gonnermann,H.M.,Manga,M.,2003.Explosivevolcanismmaynotbeaninevitable consequenceofmagmafragmentation.Nature 426,432–435.

Hamada,M.,Laporte,D.,Cluzel,N.,Koga,K.T.,Kawamoto,T.,2010.Simulating bub-ble numberdensity ofrhyolitic pumices from Plinian eruptions: constraints fromfastdecompressionexperiments.Bull.Volcanol. 72,735–746.

Hess,K.-U.,Cordonnier,B.,Lavallee,Y.,Dingwell,D.B.,2008.Viscousheatingin rhyo-lite;aninsituexperimentaldetermination.EarthPlanet.Sci.Lett. 275,121–126. Hess,K.-U.,Dingwell,D.B.,1996.Viscositiesofhydrousleucograniticmelts:

non-Ar-rhenian model.Am.Mineral. 81,1297–1300.

Higgins, M.D.,2000. Measurementofcrystal sizedistributions. Am.Mineral. 85, 1105–1116.

Hoblitt,R.P.,Wolfe,E.W.,Scott,W.E.,Couchman,M.R.,Pallister,J.S.,Javier,D.,1996. ThepreclimacticeruptionsofMtPinatubo,June1991.In:Newhall,C.G., Punong-bayan,R.S.(Eds.),FireandMud:EruptionsandLaharsofMountPinatubo. Uni-versityofWashingtonPress,Seattle,pp. 457–512.

Hurwitz, S., Navon, O., 1994. Bubble nucleation in rhyolitic melts:experiments athighpressure,temperature,andwatercontent.EarthPlanet.Sci.Lett. 122, 267–280.

Jaupart,C.,Tait,S.,1990.Dynamicsoferuptivephenomena.In:Nicholls,J.,Russell, J.K.(Eds.),ModernMethodsofIgneousPetrology:UnderstandingMagmatic Pro-cesses.In:Mineral.Soc.Am.Rev.,vol. 24,pp. 213–238.

Jouniaux,L.,Bernard,M.-L., Zamora,M., Pozzi,J.-P.,2000. Streamingpotential in volcanicrocksfromMountPelée.J.Geophys.Res. 105(B4),8391–8401. Klug, C., Cashman, K.V., 1994. Vesiculation ofMay 18, 1980, Mount St.Helens

magma.Geology 22,468–472.

Klug,C.,Cashman,K.V.,1996.Permeabilitydevelopmentinvesiculatingmagmas: implicationsforfragmentation.Bull.Volcanol. 58,87–100.

Knoche,R.,Dingwell,D.B.,Webb,S.L.,1995.Melt densitiesforleucogranitesand graniticpegmatites: partialmolarvolumesfor SiO2,Al2O3, Na2O, K2O, Li2O,

Rb2O,Cs2O,MgO,CaO,SrO,BaO,B2O3,P2O5,F2O−1,TiO2,Nb2O5,Ta2O5,and

WO3.Geochim.Cosmochim.Acta 59(22),4645–4652.

Larsen,J.F.,Denis,M.-H.,Gardner,J.E.,2004.Experimentalstudyofbubble coales-cenceinrhyoliticandphonoliticmelts.Geochim.Cosmochim.Acta 68,333–344. Larsen, J.F., Gardner, J.E., 2000. Experimental constraints on bubble interactions inrhyolitemelts:implications forvesiclesizedistributions.EarthPlanet.Sci. Lett. 180,201–214.

Launeau,P.,Cruden,A.R.,1998.Magmaticfabricacquisitionmechanismsina syen-ite:resultsofacombinedanisotropyofmagneticsusceptibilityandimage anal-ysisstudy.J.Geophys.Res. 103,5067–5089.

Launeau,P.,Robin,P.-Y.,1996.Fabricanalysisusingtheinterceptmethod. Tectono-physics 267,91–119.

Liu,Y.,Zhang,Y.,2000.Bubblegrowthinrhyoliticmelt.EarthPlanet.Sci.Lett. 181, 251–264.

Liu,Y.,Zhang,Y.,Behrens,H.,2005.SolubilityofH2Oinrhyoliticmeltsatlow

pres-sures anda newempirical modelformixedH2O–CO2 solubility inrhyolitic

melts.J.Volcanol.Geotherm.Res. 143,219–235.

Llewellin,E.W.,Mader,H.M.,Wilson,S.D.R.,2002a.Therheologyofabubblyliquid. Proc.R.Soc.A 458,987–1016.

Llewellin,E.W.,Mader,H.M.,Wilson,S.D.R.,2002b.Theconstitutiveequationand flowdynamicsofbubblymagmas.Geophys.Res.Lett. 29,2170.

Lyakhovsky,V.,Hurwitz,S.,Navon,O.,1996.Bubblegrowthinrhyoliticmelts: ex-perimentalandnumericalinvestigation.Bull.Volcanol. 58,19–32.

Mader,H.M.,Llewellin,E.W.,Mueller,S.P.,2013.Therheologyoftwo-phasemagmas: areviewandanalysis.J.Volcanol.Geotherm.Res. 257,135–158.

Mangan,M.,Sisson,T.,2000.Delayed,disequilibriumdegassinginrhyolitemagma: decompression experiments and implications for explosive volcanism.Earth Planet.Sci.Lett. 183,441–445.

Mangan,M.,Sisson,T.,2005.Evolutionofthemelt–vaporsurfacetensionin sili-cic volcanicsystems: experimentswith hydrous melts.J.Geophys. Res. 110, B01202.http://dx.doi.org/10.1029/2004JB003215.

Martel,C.,2012.Eruptiondynamicsinferredfrommicrolitecrystallization experi-ments:applicationtoPliniananddome-formingeruptionsofMt.Pelée (Mar-tinique,LesserAntilles).J.Petrol. 53,699–725.

Martel, C., Bourdier,J.-L.,Pichavant,M., Traineau, H.,2000.Textures, water con-tentanddegassingofsilicicandesitesfromrecentpliniananddome-forming eruptions at Mt Pelée volcano (Martinique,Lesser Antilles arc). J. Volcanol. Geotherm.Res. 96,191–206.

Martel, C., Bureau,H., 2001. In-situ high-pressure and high-temperature bubble growthinsilicicmelts.EarthPlanet.Sci.Lett. 191,115–127.

Martel, C., Pichavant,M.,Bourdier, J.-L.,Traineau, H.,Holtz,F.,Scaillet,B., 1998. Magmastorageconditionsandcontroloferuptionregimeinsilicicvolcanoes: experimentalevidencefromMt.Pelée.EarthPlanet.Sci.Lett. 156,89–99. Martel,C.,Poussineau,S.,2007.Diversityoferuptivestyleinferredfromthe

micro-litesofMt.Peléeandesite(Martinique,LesserAntilles).J.Volcanol.Geotherm. Res. 166,233–254.

Martel,C.,Schmidt,B.C.,2003.Decompressionexperimentsasaninsightintoascent ratesofsilicicmagmas.Contrib.Mineral.Petrol. 144,397–415.

McIntosh, I.M.,Llewellin, E.W., Humphreys, M.C.S.,Nichols, A.R.L., Burgisser, A., Schipper, C.I., Larsen,J.F.,2014. Distributionofdissolved waterinmagmatic glassrecordsgrowthandresorptionofbubbles.EarthPlanet.Sci.Lett. 401,1–11. Mollard,E.,Martel,C., Bourdier,J.-L.,2012. Decompression-inducedexperimental crystallizationinhydratedsilica-richmelts:empiricalmodelsofplagioclase nu-cleationandgrowthkinetics.J.Petrol. 53,1743–1766.

Mourtada-Bonnefoi,C.,Laporte,D.,1999.Experimentalstudyofhomogeneous bub-blenucleationinrhyoliticmagmas.Geophys.Res.Lett. 26,3505–3508. Mourtada-Bonnefoi, C.,Laporte,D., 2004.Kineticsofbubble nucleationina

rhy-oliticmelt:anexperimentalstudyoftheeffectofascentrate.EarthPlanet.Sci. Lett. 218,521–537.

Mueller,S.,Melnik,O.,Spieler,O.,Scheu,B.,Dingwell,D.B.,2005.Permeabilityand degassingofdomelavasundergoingrapiddecompression:anexperimental de-termination.Bull.Volcanol. 67,526–538.

Navon, O., Chekhmir,A., Lyakhovsky, V.,1998. Bubble growthin highlyviscous melts:theory,experiments,andautoexplosivityofdomelavas.EarthPlanet.Sci. Lett. 160,763–776.

Okumura,S.,Nakamura,M.,Nakano,T.,Uesugi,K.,Tsuchiyama,A.,2012. Experimen-talconstraintsonpermeablegastransportincrystallinesilicicmagmas.Contrib. Mineral.Petrol. 164,493–504.

Okumura, S., Nakamura,M., Takeuchi,S., Tsuchiyama, A.,Nakano, S., Uesugi,K., 2009.Magmadeformationmayinducenon-explosivevolcanismviadegassing throughbubblenetworks.EarthPlanet.Sci.Lett. 281,267–274.

Okumura, S.,Nakamura,M.,Uesugi,K.,Nakano,T.,Fujioka,T., 2013.Coupled ef-fectofmagmadegassingandrheologyonsilicicvolcanism.EarthPlanet.Sci. Lett. 362,163–170.

Polacci,M.,Papale,P.,Rosi,M.,2001.Texturalheterogeneitiesinpumicesfrom cli-macticeruptionofMountPinatubo,15June1991,andimplicationsformagma ascentdynamics.Bull.Volcanol. 63,83–97.

Proussevitch,A.A.,Sahagian,D.L.,1998.Dynamicsandenergeticsofbubblegrowth inmagmas:analyticalformulationandnumericalmodeling.J.Geophys.Res. 103 (B8),18223–18251.

Rust,A.C.,Cashman,K.V.,2004.Permeabilityofvesicularsilicicmagma:inertialand hysteresiseffects.EarthPlanet.Sci.Lett. 228,93–107.

Rust,A.C.,Manga,M.,2002.Effectsofbubbledeformationontheviscosityofdilute suspensions.J.Non-Newton.FluidMech. 104,53–63.

Rutherford,M.J.,Gardner,J.E.,2000.Ratesofmagmaascent.In:Sigurdsson,H.(Ed.), EncyclopediaofVolcanoes.AcademicSanDiego,California,pp. 207–217. Sahimi,M., 1994.ApplicationsofPercolation Theory.TaylorandFrancis,London,

258 pp.

Saul,A.,Wagner,W.,1989.Afundamentalequationforwatercoveringtherange fromthemeltinglineto1273 Katpressuresupto25 000 MPa.J.Phys.Chem. 18 (4),1537–1565.

Scaillet,B.,Evans,B.W.,1999.The15June1991eruptionofMountPinatubo.I.Phase equilibriaandpre-eruption P –T – f O2– f H2Oconditionsofthedacitemagma.

J. Petrol. 40,381–411.

Sparks,R.S.J.,1978.Thedynamicsofbubbleformationandgrowthinmagmas:a reviewandanalysis.J.Volcanol.Geotherm.Res. 3,1–37.

Sparks,R.S.J.,1997.Causesandconsequencesofpressurisationinlavadome erup-tions.EarthPlanet.Sci.Lett. 150,177–189.

Takeuchi,S.,Tomiya,A.,Shinohara,H.,2009.Degassing conditionsforpermeable silicicmagmas: implications from decompression experimentswith constant rates.EarthPlanet.Sci.Lett. 283,101–110.

Toramaru,A.,1989.Vesiculationprocessandbubblesizedistributionsinascending magmaswithconstantvelocities.J.Geophys.Res. 94(B12),17523–17542. Toramaru,A.,1995.Numericalstudyofnucleationandgrowthofbubblesinviscous

magmas.J.Geophys.Res. 100(B2),1913–1931.

Toramaru, A.,2006.BND(bubblenumberdensity)decompression ratemeterfor explosivevolcaniceruptions.J.Volcanol.Geotherm.Res. 154,303–316. Toramaru,A.,2014.Onthesecondnucleationofbubblesinmagmasundersudden

decompression.EarthPlanet.Sci.Lett. 404,190–199.

Westrich,H.R.,Eichelberger,J.C.,1994.Gastransportandbubblecollapseinrhyolitic magma:anexperimentalapproach.Bull.Volcanol. 56,447–458.

Wolfe,E.W.,Hoblitt,R.P.,1996.Overviewoftheeruptions.In:Newhall,C.G., Punong-bayan,R.S.(Eds.),FireandMud:EruptionsandLaharsofMountPinatubo. Uni-versityofWashingtonPress,Seattle,pp. 3–20.

Wright,H.M.N.,Cashman,K.V.,Gottesfeld,E.H.,Roberts,J.J.,2009.Porestructureof volcanicclasts:measurementsofpermeabilityandelectricalconductivity.Earth Planet.Sci.Lett. 280,93–104.

Zhang,Y.,Behrens,H.,2000.H2Odiffusioninrhyoliticmelts andglasses.Chem.