The molecular structure of melts along the carbonatite–kimberlite–basalt compositional joint: CO2and polymerisation

(1)

HAL Id: insu-01250952

https://hal-insu.archives-ouvertes.fr/insu-01250952

Submitted on 8 Dec 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.

The molecular structure of melts along the carbonatite–kimberlite–basalt compositional joint:

CO2and polymerisation

Yves Moussallam, Pierre Florian, Dario Corradini, Yann Morizet, Nicolas Sator, Rodolphe Vuilleumier, Bertrand Guillot, Giada Iacono-Marziano,

Burkhard C. Schmidt, Fabrice Gaillard

To cite this version:

Yves Moussallam, Pierre Florian, Dario Corradini, Yann Morizet, Nicolas Sator, et al..

The molecular structure of melts along the carbonatite–kimberlite–basalt compositional joint:

CO2and polymerisation. Earth and Planetary Science Letters, Elsevier, 2016, 434, pp.129-140.

�10.1016/j.epsl.2015.11.025�. �insu-01250952�

(2)

compositional joint: CO

²

and polymerisation

Yves Moussallam, Pierre Florian, Dario Corradini, Yann Morizet, Nicolas Sator, Vuilleumier, Bertrand Guillot, Giada Iacono-Marziano, Burkhard C. Schmidt, Fabrice Gaillard

Abstract

Transitional melts, intermediate in composition between silicate and carbonate melts, form by low degree partial melting of mantle peridotite and might be the most abundant type of melt in the asthenosphere. Their role in the transport of volatile elements and in metasomatic processes at the planetary scale might be significant yet they have remained largely unstudied. Their molecular structure has remained elusive in part because these melts are difficult to quench to glass. Here we use FTIR, Raman,

¹³

C and

²⁹

Si NMR spectroscopy together with First Principle Molecular

Dynamic (FPMD) simulations to investigate the molecular structure of transitional melts and in particular to assess the effect of CO

²

on their structure. We found that carbon in these glasses forms free ionic carbonate groups attracting cations away from their usual ‘depolymerising’ role in breaking up the covalent silicate network. Solution of CO

2

in these melts strongly modifies their structure resulting in a significant polymerisation of the aluminosilicate network with a decrease in NBO/Si of about 0.2 for every 5 mol% CO

2

dissolved.

This polymerisation effect is expected to influence the physical and transport properties of transitional melts. An increase in viscosity is expected with increasing CO

²

content, potentially leading to melt ponding at certain levels in the mantle such as at the lithosphere–asthenosphere boundary. Conversely an ascending and degassing transitional melt such as a kimberlite would become increasingly fluid during ascent hence potentially accelerate. Carbon-rich transitional melts are effectively composed of two sub-networks: a carbonate and a silicate one leading to peculiar physical and transport properties.

Keywords : CO

²

; polymerisation; speciation; transitional melt; kimberlite; glass

1. Introduction

Carbon dioxide (CO

2

) is typically the second-most abundant volatile in terrestrial melts. In common silicate melts, found near the Earth's surface and in its crust, its concentration is typically greatly inferior to that of water and its influence on the melt physical properties secondary. In the Earth's upper mantle however, carbon-rich and typically silica-poor melts, generated by low degree partial melting of the mantle are probably widespread (e.g. Wyllie and Huang, 1976, Eggler, 1978, Dalton and Presnall, 1998, Gudfinnsson and Presnall, 2005, Dasgupta and Hirschmann, 2006 and Dasgupta et al., 2013; Massuyeau et al., in press). Very low degree partial melting of carbonated peridotite at ∼250 km depth produces carbonatite liquid (e.g.

Dasgupta and Hirschmann, 2006). With either increasing temperature, decreasing pressure or

with

(3)

Table 1

Startingoxideandnaturalrockpowdermixcompositionsusedforexperiments(inwt%).ThecompositionofanaturallamproitefromTorreAlﬁna,Italy(Peccerilloetal., 1988),fusedtwiceat1400^◦Candusedtopreparetheoxide-mixcompositionsisreportedatthetopofthetable.

Name SiO2 TiO2 Al2O3 FeO MgO CaO Na2O K2O P2O5 CO2 Total

TA 55.7 1.3 13.1 5.8 9.4 5.4 1.0 7.7 0.5 0.0 100

TA6 16.6 0.2 3.9 0.9 6.9 35.5 0.3 2.3 0.1 33.5 100

TA9 23.1 0.3 5.4 1.5 8.3 30.2 0.4 3.2 0.1 27.5 100

TA10 23.8 0.4 5.5 1.6 8.3 29.7 0.4 3.3 0.1 27.0 100

TA11 30.2 0.6 7.0 2.4 8.5 24.8 0.5 4.1 0.2 21.6 100

TA12 37.7 0.8 8.8 3.4 8.8 19.1 0.6 5.2 0.3 15.2 100

theadditionofCO₂ orH₂O,theliquidproducedbyperidotitepar- tial melting will become increasingly SiO2-rich, forming kimberlite,melilitite andeventually basalt(e.g.WyllieandHuang,1976;

Eggler,1978; Wyllie,1989).Thesemelts–carbonatite, kimberlite andmelilitite –transitionalbetweenpurecarbonatiteandsilicate liquids(∼¹⁰^to ∼⁴⁰^wt% ^SiO2), cancontainup toseveraltens of weightpercent ofCO₂ (Brey andGreen,1976; Moussallametal., 2014). Understanding the effectof CO2 on the structureof these melts/glasses is therefore a cornerstone to predict their physical characteristicsandtransportproperties.

TheeffectofCO₂onmeltpolymerisationhasremainedunclear.

Eggler (1978)suggested that thedissolution ofCO₂ ascarbonate ion(CO²₃⁻)shouldresultinmeltpolymerisationviathereaction:

CO2+2Qⁿ(M^m⁺)CO²₃⁻(M^m⁺)+2Qⁿ⁺¹ (1) WhereQⁿ denotesasilicatetrahedronlinkedbybridgingOatoms to n adjacent tetrahedra (n=0 correspond to an isolated SiO4 tetrahedron, n=4 correspond to a fully connected tetrahedron withfour bridging oxygen), M^m⁺ denotes a metal cation in net- workmodifyingorchargebalancingrole.InEq.(1),theincreasein polymerisation isdenotedbythebuildingofSi–O–Sibonds.

This depiction has gained popularity, being referred to as a largely admitted concept in review articles (e.g. Mysen, 2013, 2012) and books (e.g. Frost and Frost, 2013; Mysen and Richet, 2005). Solid evidenceoftheeffectof CO2 dissolutiononthe aluminosilicate network structure however hasbeen lacking. Mysen andVirgo(1980a,1980b)foundthatCO2depolymerisesalbiteand anortitemeltswhilepolymerisingdiopsideandNaCaAlSi2O7melts.

Konschak (2008)investigatedthealbite –diopsidejointandfound noeffectofCO2 onpolymerisationfromAb50Di50toAb75Di25but found that aslightdepolymerisation isassociated withtheaddi- tionofCO₂ toAb₉₀Di₁₀.AsnotedbyNiandKeppler (2013)these evidenceareallbasedonverysubtledifferencesinRamanspectra, which interpretation can be controversial. All thesestudies have focused ontheincorporation mechanismsofCO2 andits impacts onthemolecular structureofsilicatemelts. Herewe lookatCO₂ intransitionalmelts,typicallyproducedbyverylow-degreepartial mantlemeltingandwithsilicacontentbeingabouthalfofthat of basalts.

The principal question we target in this study is: Does car- bondioxideinﬂuencethedegreeofpolymerisationoftransitional melts? We present results from the ﬁrst Infrared, Raman and Nuclear Magnetic Resonance (NMR) spectroscopy investigation of transitionalglassesandcomparethemwithresultsfromAb-initio FirstPrincipleMolecularDynamicsimulationsappliedtomeltcon- ditions.We explore a compositional rangefrom∼⁴⁴^to ^{23 mol%}

SiO₂ (on a volatile-free basis) with 0 to 26 mol% CO₂. For all compositionsinvestigated,weshowthat (1)CO2 ispresentinthe glassandinthemeltdominantlyascarbonateion(2)thedegree ofpolymerisation oftheglass/melt increaseswithincreasing CO2 content. We concludethat the physicalproperties of transitional meltssuchasviscosity, electricalconductivityandsoundvelocity areexpectedtobegreatlyaffectedbytheir CO₂ contentwithim- plicationsregardingmeltmobilityintheuppermantle.

We note thatpolymerisation, inthismanuscriptisstrictly de- fined asthe process bywhich oxygen atomsare sharedbetween silica(oralumina)tetrahedraandisexpressedastheratioofnon- bridging oxygenstotetrahedralcations (NBO/T)orastheratioof non-bridging oxygens per silicon (NBO/Si). We further note that theNBO/Tratioisonlyacalculated‘statisticalaverage’andvarious combinations of Q speciescan give the sameNBO/T. The NBO/T conceptfailspastNBO/T=⁴^but^Q ûnitsât^greater^values^can^still joinupinisolated‘polymerisedunits’.

2. Methods

2.1. Experimentalmethodology 2.1.1. Startingmaterial

Starting materials were produced by mixing powders from a natural lamproitefrom TorreAlfina, Italy (Peccerillo et al., 1988) withvarious amounts ofsynthetic powdersofpure oxides (SiO2, Al₂O₃, MgO, CaCO₃, Na₂CO₃, K₂CO₃) and natural dolomite. The strategy was to producea series ofprogressivelymore silicaand aluminapoorcompositionsadditionallylowinironinorderforthe finalproducttobeanalysablebyNMRspectroscopy.Thesourceof CO₂ inexperimentswas mainlyCaCO₃.TheTorreAlfinarock,was fused twice in air at1400^◦C and quenched to glass in order to ensure homogeneity and remove any volatile present. The composition of all mixtures used as starting materials is reported in Table 1.

2.1.2. HP-HTexperiments

Experiments were performed in the pressure range 0.1 to 1500 MPa. High temperature experiments at 0.1 MPa were performed inahigh-temperaturefurnace. Thesample wascontained in a Pt crucible and heated to 1500^◦C for 30 min and then quenchedtoglassbycompleteimmersionincoldwater.

Experiments athigh pressure were performedby Moussallam et al. (2014) in the pressure range 50 to 350 MPa at relatively constanttemperature(1225to1270^◦C)withnoaddedwaterand underoxidised conditions(log fO2=^FMQ+^3).Experimentswere performedininternally heatedpressurevessels attheISTO-CNRS laboratoryinOrléanswhichcanreachpressuresofupto400MPa (±^{3 MPa)}^under^controlledtemperatureupto1300^◦C(±⁵^◦^C)^(see supplementaryinformation).

Experiments at 1500 MPa and 1350^◦C were performed in a piston-cylinder apparatuswithina 3/4 inch(1.9 cm) assemblies.

Experimentalchargesconsistedofnaturalanhydroussamplepow- der(30 mg)loadedinsealedgold–palladium(Au80Pd20)capsules (1 cm in length, 2.5 mm inner diameter and 2.9 mm outer diameter). The capsules were introduced in a talc–pyrex–graphite furnaceassemblyandsurroundedbyMgO.A B-typethermocouple was locatedat∼¹^mm âtopôf^the^capsule ând^the^run ^tempera- tureshould beconsidered asaminimumvalue. Uncertaintiesare of 10% in relative forpressure andof 15^◦C fortemperature. We used a modified perforated anvil through which cold waterwas circulatedinordertomaximisequenchingrate(>200^◦C s⁻¹).

A pure carbonate glass of composition K₂Mg(CO₃)₂ was syn- thesised in internally heated pressure vessel at 803^◦C and1083

(4)

barunderrelativelyoxidised conditions(logfO2=^FMQ+³⁾ ^with no added water. This composition was shown previously to be quenchable to carbonateglass without crystallisation (Ragone et al.,1966;Gengeetal.,1995).Thesample was enrichedin¹³Cby usingK213CO3(99%¹³C)asstartingmaterial.Dropquenchatthe endoftheexperimentresultedinpurecarbonateglassasindicated bythelackofbirefringenceinpolarisationmicroscopy.

2.2.Analyticaltechniques 2.2.1. Microbeamanalyses

Allexperimentalproductswereexaminedbyopticalmicroscope andscanning electronic microscope (SEM) tocheck forthe pres- enceofquenchcrystals.Electronmicroprobeanalyses(EMPA)were performed on a Cameca SXFive at the ISTO-CNRS laboratory in Orléans. Glasses were analysed using an accelerating voltage of 15 kV,a beamcurrent of6 nAand adefocused beamof 10 μm withacount timeof10 s.Nawasanalysedﬁrstinordertomin- imise itslossduring analyses. No Klosswithtime was observed atouroperatingconditions.

2.2.2. Infraredspectroscopy

InfraredspectrawerecollectedusingaNicolet6700FTIRspec- trometerattachedtoaContinuum microscope.Weused aGlobar internalIR sourcewitha KBrbeamsplitterandaliquidnitrogen cooledMCT/Adetector.Thespectralresolutionwas setto4 cm⁻¹, andspectrawere accumulatedfor128scans.Background spectra wereacquiredbyaccumulating256scansandusedto correctfor background. Spectra were obtained directly on small glass frag- ments(unpolished)intransmissionmode.

2.2.3. Micro-Ramanspectroscopy

An Innova 300-5W Argon ion laser (Coherent^©) operating at 514 nmwas usedastheexciting source toproduce Raman scat- tering.Spectrawere collected bya Jobin–YvonLabram spectrometer (focal distance = ^{300 mm)} êquipped ^with â ^grating ^with 2400 grooves/mmandaCCDdetector.Thespectralfrequencypo- sitionwascalibratedusingtheemissionlinesofNe- andHg-lamps with an accuracy within ±^{1 cm}⁻¹^. Ânalyses ^were ^performed ⁱⁿ confocal mode (hole = ^{500 μm,} ^slit = ^{200 μm)} ând ûsing â

×⁵⁰Ôlympus ôbjective^resultingⁱⁿân ânalysed ^volumeôfâ ^few μm³.Spectrawereacquiredinthe700–1300 cm⁻¹spectralregion which corresponds to the vibrational region for aluminosilicate frameworksymmetricstretch(ν1 forQⁿspecies;e.g.Mysenetal., 1982) andalsothe symmetricstretchfor CO²₃⁻ moleculargroups (ν1 forCO²₃⁻) ataround 1080 cm⁻¹.The acquisition timewas of 10×^{60 s and} ^the ^focus ^depth ^was ôptimised ⁱⁿôrder ^to ôbtain thehighestRamansignal.Atleastthreespectrawereacquiredfor eachsample.

2.2.4. NMRspectroscopy

AllNMRexperimentswereperformedona9.4 TAvanceBruker Spectrometer operatingat 79.5 MHz (²⁹Si) and100.6 MHz (¹³C).

Weused a4mm diameterrotor spinningat14kHzandapplied an Hahn–Echoacquisition witharadio-frequency ﬁeld of50kHz (²⁹Si)and35kHz(¹³C)andan inter-pulsedelayof1rotorperiod.

Therelaxationtimearefoundtobearound75msforbothnuclei becauseofthepresenceofiron(1to4 wt%onvolatilefreebasis) inthe sample. The recycle delayswere set to 250ms, while ac- cumulatingbetween300 000 and900 000 scans,dependingonthe amountofsampleavailable. ATeﬂonspacerwas usedtoposition thesampleinthemiddleoftherotorwhenusingsmallamounts.

2.3.FirstprinciplesMDsimulations

First-Principles Molecular Dynamics (FPMD) simulations were performedontwosystems.The ﬁrstone isasimpliﬁedTA10sys-

tem(32 wt%SiO2;expressed ona volatile-freebasis;cf.Table 1) composedofSiO2(x=⁰.30),Al2O3(x=⁰.07),MgO(x=⁰.14)CaO (x=0.44)andK2O(x=0.05)(werexreferstothemolefractionof eachcomponent).Asimulationboxwithasidelength L=¹⁴.7 Å was used. This box consisted of 238 atoms(135 O, 31 Si, 7 Al, 15 Mg, 45Ca and5 K). Thisgave a densityof 2.92 g/cm³.The systemwasrun atT=^{1773 K.}^The^second^systemîsâ^simplified TA10compositioncontaining13.8 wt%ofCO2,whichtranslatesin terms ofmolefractions toCO2 (x=⁰.18), SiO2 (x=⁰.25), Al2O3 (x=⁰.05),MgO (x=⁰.12), CaO(x=⁰.36)andK₂O(x=⁰.04). In thiscase thesimulation box contained304 atoms(179 O, 31 Si, 7 Al, 15Mg, 45Ca, 5 Kand 22C) andits side length measures L=¹⁵.7 Å,correspondingtoadensityof2.83 g/cm³.Forthissys- tem,wefixedthetemperatureatT=^{1498 K.}^Theâdopted^temper- atureconditionsrepresentmeltconditionsalmostidenticaltoglass synthesesconditions.Theexperimentalglassstructurehoweveris thatpreservedattheglasstransitiontemperature,whichislower than the synthesis temperature, while the structure observed by FPMD isthat ofthe melt. The simulations reported required the useofabout215 000 singleCPUhours,usingthehighperformance oftheIDRISsupercomputer.Additionalsimulationdetailsaregiven inthesupplementaryinformation.

Theoretical NMR spectra wereobtained fromcalculationsper- formedonMDboxes,detailsaregiveninthesupplementaryinfor- mation.

3. Results

Five starting compositions were saturated with CO₂, at pres- suresfrom0.1to 1500MPa.Thisallowedustocreatesamplesof similarcompositionswithvariableCO2 content.Allexperimentare reportedinTable 2.Analyticaldatafrompressurebetween100to 350MPaare fromtheexperimentsreportedinMoussallam etal.

(2014).The FPMDab-initio molecular dynamic simulations were performedonasinglecomposition(with32wt%SiO2onavolatile freebasis)equilibratedatasinglepressure,withandwithoutCO₂. 3.1. Infraredspectroscopy

InfraredspectraforCO2-bearingglassesrangingincomposition from24to44 wt%SiO2 (expressedonavolatile-freebasis)arere- portedinFig. 1togetherwithaninfraredspectrumobtainedona purecarbonateglassofcomposition50:50mol%MgCO3–K2CO3.In allcompositionsinvestigatedCO2intheglassisintheformofcar- bonate,incontrasttorecentfindingsinwater-saturatedcarbonate melts (Foustoukos and Mysen, 2015). The absence of absorption peakat∼^{2350 cm}⁻¹ ⁱⁿâll^spectra^shows^theâbsenceôf^molecu- larCO2 inthequenchedglass,althoughitmayhavebeenpresent inthemelt(seesection3.4).Thecarbonatev₃ doublethasasim- ilarpositionacross composition(mid-point at∼^{1460 cm}⁻¹⁾^with a splittingofv3≈^{70 cm}⁻¹^.În ^detail,^small^variations ^between peak positionoccurfromthemostsilica-rich(TA12;44wt%SiO₂ onavolatilefreebasis)tothemostsilica-poorglass(TA6;24wt%

SiO2 on a dry basis) with the mid-point shifting from 1465 to 1455 cm⁻¹ andthe v₃ split shifting from80 to60 cm⁻¹.The spectra obtained on pure carbonate glass shows a mid-point at 1450cm⁻¹andasplittingofv3≈^{60 cm}⁻¹^.

3.2. Ramanspectroscopy

Raman spectra of quenched glasses ranging in composition from 24to 44 wt% SiO₂ (expressed on a volatile-free basis) and containing 0 to 22 wt% CO2 are presented in Fig. 2 in the fre- quencyrange700to1200 cm⁻¹.Allglassescontaining CO2 show a strong peak at ∼^{1080 cm}⁻¹ corresponding to the stretch- ing of C–O vibration in CO²₃⁻ conﬁguration while no evidence