Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa: an experimental approach

Cent. Eur. J. Geosci. • 2(2) • 2010 • 188-198DOI: 10.2478/v10085-010-0008-5

Central European Journal of Geosciences

Trace element behaviour during interaction betweenbasalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

Research Article

Silvio Mollo, Valeria Misiti∗, Piergiorgio Scarlato

Istituto Nazionale di Geofisica e Vulcanologia,Via di Vigna Murata 605, I-00143 Rome

Received 14 February 2010; accepted 9 April 2010

Abstract: We experimentally investigate the major and trace elements behavior during the interaction between two par-tially molten crustal rocks (meta-anorthosite and metapelite) and a basaltic melt at 0.5-0.8 GPa. Results showthat a hybrid melt is formed at the basalt-crust contact, where plagioclase crystallizes. This contact layer isenriched in trace elements which are incompatible with plagioclase crystals. Under these conditions, the traceelement diffusion coefficients are one order of magnitude larger than those expected. Moreover, the HFSEdiffusivity in the hybrid melt is surprisingly higher than the REE one. Such a feature is related to the plagio-clase crystallization that changes the trace elements liquid-liquid partitioning (i.e. diffusivity) over a transientequilibrium that will persist as long as the crystal growth proceeds.These experiments suggests that the behaviour of the trace elements is strongly dependent on the crystalliza-tion at the magma-crust interface. Diffusive processes like those investigated can be invoked to explain someunusual chemical features of contaminated magmatic suites.

Keywords: contamination • trace elements diffusion • basalt-crust interaction

© Versita Warsaw

1. Introduction

The evolution of magmas has been the focal point of petro-logical studies since the first decade of the past century.Even if relatively simple processes have been clearly doc-umented (e.g. fractional crystallization, mixing, mingling),they cannot fully explain the chemical peculiarities ofmany magmas [1–5]. In the case of magma-crust interac-tion, more complex processes must be invoked, (e.g. AFC,∗E-mail: [email protected]

Assimilation Fractional Crystallization; AEC, AssimilationEquilibrium Crystallization; RAFC, Recharge AssimilationFractional Crystallization; RTFA, Recharge Tapping Frac-tionation Assimilation; etc.) to explain geochemical, iso-topic, and, to a minor extent, major elements features notconsistent with simple models [6, 7]. The contamination ofmagmas by crustal material takes place when hot magmasinduce melting of the crust [8–11]. This condition dependson the composition of the crustal materials and on the in-cubation time or rate of magma supply [12]. When crustalmagmas are formed by partial to complete fusion of crustalrocks, crust-basalt melt interaction takes place. In thiscondition the mobile trace elements can be redistributed188

Silvio Mollo, Valeria Misiti, Piergiorgio Scarlato

during interaction and their isotopic compositions can bemodified selectively by diffusion [13–18]. For instance, theNeogene Volcanic Province within the Betic Cordillera hasa volcanism associated to crustal anatexis of metapeliticenclave suites originated at different crustal depths, from20 to 9 km [19]. Crustal anatexis has been explainedby the emplacement of basaltic magmas that, interactingwith the crustal melts, produced the calc-alkaline volcanicactivity ([20] and references therein). Otherwise, meta-anorthosite and associated plutonic rocks constitute animportant component of the older and thicker continentalcrust. However, in some cases the subsequent erosionand tectonic stripping may rapidly uplift these deeplyburied rocks [21] that locally preserve a high pressuremetamorphism [22]. Under this condition, the opportu-nity for physical and chemical magma-crust interactiongenerally increases; rising mafic magmas introduce heatinto the crustal material, causing intracrustal melt gener-ation ([23, 24] and references therein).Here, we experimentally investigate the contaminationof mantle-derived magmas by crustal material. Particu-lar attention has been directed to the behaviour of traceelements in a basaltic melt during its interaction withtwo partially molten crustal rocks (metapelite and meta-anorthosite). The chosen experimental temperatures werekept near the liquidus temperature of the basalt, at 1150and 1200 °C, while pressures were in the range of 0.5-0.8 GPa.

2. Starting materials, experimen-tal procedures and analytical tech-niques

2.1. Starting materials and compositions

A meta-anorthosite (MK72), a metapelite (VA38) and abasalt (PF1) were chosen for the experiments. Analysesof the bulk starting materials are reported in Table 1.MK72 represents the mafic lowermost part of the continen-tal basement from the Pan-African belt (Tanzania), with aP-T stability field in the range of 0.8-1.0 GPa and 900 °Caccording to the literature [25]. Plagioclase is the ma-jor phase (80 vol.%) of this metabasic rock. Granoblasticplagioclase layers are alternated with garnet, hornblende,ilmenite and occasional apatite layers.

Table 1. PF1, VA38, and MK72 whole rock compositions.The ICP analyses were performed by lithium metabo-rate/tetraborate fusion at the Activation Laboratories Ltd(ACTLABS), Ontario, Canada.wt.% PF1 VA38 MK72SiO2 52.23 77.85 52.78TiO2 0.57 0.6 0.84Al2O3 16.74 14.8 25.12FeOtot 9.39 3.57 4.11MnO 0.18 0.1 0.08MgO 7.12 0.82 1.88CaO 10.95 0.17 9.27Na2O 1.83 0.2 5.15K2O 0.86 1.76 0.57P2O5 0.13 0.14 0.18Total 100 100 100Total* 99.95 98.54 99.31ppmCr 120.7 38.02 4.01Rb 23.74 100.38 2.63Sr 433.73 96.82 930.94Y 12.27 20.41 12Zr 33.65 58.16 11.02Nb 2.79 9.56 2.66Cs 0.57 7.88 0.35Ba 243.02 418.27 647.83La 11.15 28.72 7.01Ce 22.22 60.89 16.6Pr 2.7 6.77 2.24Nd 11.68 25.76 11.12Sm 2.7 4.51 2.69Eu 0.9 1.41 1.93Gd 2.74 4.56 2.68Tb 0.36 0.67 0.37Dy 2.5 5.23 2.34Ho 0.5 0.97 0.48Er 1.41 2.38 1.16Tm 0.24 0.33 0.16Yb 1.53 1.89 0.95Lu 0.23 0.3 0.12Hf 1.25 2.35 0.22Ta 0.12 1.06 0.14Pb 6.09 68.27 1.23Th 2.12 7.17 0.07U 0.71 2.14 0.02*original total of the analysis

VA38 comes from the uppermost part of the continental189

Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

crust of the Gennargentu Intrusive Complex (Italy). Therock is a coarse-grained schist. Grained quartz is themost abundant phase (70 vol.%) and muscovite, biotite,sillimanite, and magnetite occur in lepidoblastic layers.Its P-T stability field has been estimated at 0.4-0.5 GPaand 550-650 °C [26, 27].PF1 is a calc-alkaline glassy scoria from Panarea Island(Italy) whose composition is at the boundary betweenbasalt and basaltic andesite according to the TAS dia-gram [28].2.2. Experimental procedures

A PF1 clear glass was obtained from the glassy scoriaby loading the fine grained powder in Fe-presaturatedPt-capsules (5 mm outer diameter) and double meltingthe sample at 1400 °C for 20 minutes in an atmosphericDeltech DT-31 vertical rapid-quench furnace at the Di-partimento di Scienze Geologiche Università “Roma Tre”(Rome). Oxygen fugacity (NNO+2) was buffered by aCO-CO2 gas mixture, controlled by a SIRO2 solid zirco-nia electrolyte oxygen sensor located about 5 mm fromthe sample. This melting process was chosen to avoid thepossibility of interferences with microlites belonging tothe natural product. The glass electron microprobe anal-yses overlap the bulk analysis reported in Table 1, and theloss of iron or alkalies was not detected. PF1 glass, andMK72 and VA38 rocks were then crushed in the mortarand sieved to a grain size of 1 µm.The experiments were conducted in a solid-medium 3/3 inchend-load piston cylinder apparatus at the HP-HT labo-ratory of Istituto Nazionale di Geofisica e Vulcanologia(Rome), by half filling a Pt-capsule (3 mm outer diameter;15 mm length) with PF1 powder. The meta-anorthositeor metapelite powder was then loaded and pressed at thetop of the PF1 powder to fill the upper portion of the cap-sule. Each capsule was filled by a similar thickness ofthe two PF1 and MK72 (or VA38) powders. Runs wereperformed at variable time durations (1, 2, 4 and 8 h)and constant P-T conditions for the PF1-MK72 (0.8 GPaand 1200 °C) and PF1-VA38 (0.5 GPa and 1150 °C) cou-pled compositions. The samples were first pressurized us-ing the piston-in technique to nominal pressure within25 MPa of the actual pressure, and then heated at a rateof 200 °C min−1 up to 20 °C below the set point. Thelast 20 °C were reached with a rate of 40 °C min−1. Thetemperature was controlled by aW95Re5–W74Re26 thermo-couple and held within 3 °C of the experimental tempera-ture. The thermocouple was positioned such that its junc-tion was coincident with the cylindrical axis of the furnaceand the midpoint (length-wise) of the capsules, where thefurnace hot-spot is estimated to be approximately 8 mm

length. The run was quenched by turning off the powerto the heater. The quench was isobaric and the rate wasabout 2000 °C min−1. The apparatus was calibrated forpressure and temperature as described by [29, 30]. Thecapsule was put into a 19.1 mm NaCl-crushable alumina-pyrex assembly. Furnace assembly consisted of an outerNaCl cylinder with a high-purity graphite furnace tubesurrounded by a pyrex glass sleeve within. The capsule(15 mm long) was inserted vertically into holes drilled inthe alumina rods. The rods were then loaded into thegraphite furnace tube. Care was taken to ensure thatthe centre of the capsule coincided with the centre of thehot-zone. Oxygen fugacity was of NNO+2. This valuewas estimated, according to the [31] model, by Mossbauerspectrometrical analyses performed on a PF1 test sam-ple. The NNO+2 value agrees with those measured fromprevious studies where the same piston cylinder assemblywas used [32–34].2.3. Analytical techniques

Major element analyses of the experimental productswere performed at the CNR-IGAG (Rome) with a CamecaSX50 (Electron Probe Micro Analysis) equipped with fivewavelength-dispersive spectrometers using 15 kV acceler-ating voltage, 15 nA beam current, 10 µm beam diameter,and 20 s counting time. The following standards wereused: wollastonite (Si and Ca), corundum (Al), diopside(Mg), andradite (Fe), rutile (Ti), orthoclase (K), jadeite(Na), barite (Ba, S), celestine (Sr), F-phlogopite (F), andmetals (Cr and Mn). Na and K were analyzed first inorder to reduce possible volatilization effects.Trace element analyses were conducted at the CNR-IGG–Pavia with a LA-ICP-MS (Laser Ablation Inductively Cou-pled Plasma Mass Spectrometry) apparatus. The lasersource consists of a Q-switched Nd:YAG laser (Brilliant,Quantel), with a fundamental emission in the near-IR re-gion (1064 nm) which is converted into 266 nm by two har-monic generators. Using mirrors, the laser beam (10 µmin diameter) is carried into a petrographic microscope, fo-cused above the sample, and then projected onto it. Op-timum average instrumental operating conditions are: RFpower 800-900 W, cooling gas 12.08 l min−1, sample gas0.9-1.1 l min−1, auxiliary gas 1.00 l min−1 and carrier gas0.9-1.1 l min−1. The total scan-time is about 700 ms, thesettling time is about 340 ms, and hence the acquisitionefficiency is estimated at about 50%. A typical analy-sis consists of acquiring one minute of background andone minute of ablated sample, thus approximately 170sweeps are required. The mean integrated time for acqui-sition is about 0.9 s for each element. The ablated mate-rial was analyzed with a single-collector double-focusing190

Silvio Mollo, Valeria Misiti, Piergiorgio Scarlato

sector-field ICP–MS (Element, Finnigan Mat, Bremen,Germany). Three parallel analytical profiles were per-formed across the basalt-crustal rock interface to collect arepresentative number of analyses. Screening of the datawas conducted to discriminate glass and crystals. Thetrace element analyses reported below refer to the 8 hruns which display more appreciable element variationswith respect to the whole dataset.Image analyses and chemical maps were used to deter-mine the volume percent of crystals and glass from the ex-perimental products. Analyses were performed at the HP-HT laboratory of INGV (Rome) with a Jeol FE-SEM (FieldEmission Scanning Electron Microscope) 6500F equippedwith an EDS (Energy Dispersion microanalysis System).3. Results

3.1. Textural characters

The basalt-crustal rock run conditions and experimentalassemblages for both the PF1-VA38 and PF1-MK72 prod-ucts are summarized in Table 2.The PF1-VA38 products consist of poorly crystallizedbasalt below a partially molten metapelite (Figure 1). Thethickness of the interaction layer is about 2 mm (Fig-ure 1). Clinopyroxene (En9-Di81) and plagioclase (An83)are the mineral phases on the liquidus. Their abundanceincreases from 2 to 8 vol.% with experiment time and ischaracterised by a pyroxene/plagioclase constant ratio of1.5. The VA38 restitic assemblage consists of quartz andsillimanite. The amount of glass produced by the partialmelting increases from 50 to 60 vol.% with the experimentaltime.Table 2. Basalt-crustal rock run conditions and experimental as-

semblages.

Run# T P t starting products(°C) (GPa) (h) materialPC177 1200 0.8 1 PF1-MK72 g - plg+gPC219 1200 0.8 2 PF1-MK72 g - plg+gPC218 1200 0.8 4 PF1-MK72 g - plg+gPC254 1200 0.8 8 PF1-MK72 g - plg+gPC256 1150 0.5 1 PF1-VA38 cpx+plg+g - q+sil+gPC258 1150 0.5 2 PF1-VA38 cpx+plg+g - q+sil+gPC257 1150 0.5 4 PF1-VA38 cpx+plg+g - q+sil+gPC287 1150 0.5 8 PF1-VA38 cpx+plg+g - q+sil+gPF1-MK72 products consist of a clear basaltic glass be-low a partially molten meta-anorthosite (Figure 2). Theamount of melt in the meta-anorthosite increases with time

Figure 1. FE-SEM backscattered electrons image of PF1-VA38 run.

from 45 to 60 vol.%. A sieve-textured plagioclase is theonly restitic phase present. In Figure 3, two backscat-tered SEM images show that plagioclase has a reactionrim (light grey colour) more An-rich (An56) with respect tothe core composition (An44) that strictly matches the orig-inal chemistry of meta-anorthosite plagioclase (Table 3).At the PF1-MK72 boundary a 0.8 mm wide interactionlayer develops, from which new crystals of plagioclasecrystallize (Table 3 and Figure 3). Plagioclase contentincreases from 5 to 20 vol.% with increasing the experi-mental time: acicular (2×20 µm) crystals occur in the 1and 2 h runs, while tabular shapes (10×50 µm) are presentin 4 and 8 h runs (Figure 3).3.2. Glass chemistries

In PF1-VA38 experiments, the PF1 melt composition be-comes more basaltic andesitic with time due to the crys-tallization of small amounts of clinopyroxene and plagio-clase (Table 4). On the contrary, the partial melting ofVA38 produces a rhyolitic melt (Table 4). In Figure 4, thechemical profiles of major elements show that a compo-sitionally monotonic change between basalt and rhyolitetakes place. Notably, there is a slight diffusion of alkalisfrom the 1 to 8 h runs, but this is not observable at thescale of Figure 4; accordingly, their concentration profilesare very limited during the interaction between basalt andrhyolite, two structurally distinct (i.e., degree of polymeri-sation and viscosity) melts [35].Differently, chemical profiles of trace elements clearlyshow a chemical flux from metapelite towards basalt (Fig-ure 5). This diffusion process occurs for LILEs (Large191

Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

Figure 2. FE-SEM backscattered electrons image of PF1-MK72run.

Figure 3. Plagioclases at the contact and within the interactionlayer from the PF1-MK72 runs. Top: a big sieve-texturedresidual crystal from the meta-anorthosite boundary isdisplayed. Bottom: new plagioclases nucleate andgrowth with time from the light hybrid melt.

Lithophile Ion Elements), LREEs (Light Rare Earth Ele-ments), HREEs (Heavy Rare Earth Elements) and HFSEs(High Field Strength Element). Notably, because of thepoor crystallization of clinopyroxene and plagioclase, thetrace element concentrations of the basalt overlap thoseof the original PF1 composition at the scale of Figure 5.In PF1-MK72 experiments, the MK72 melt compositionchanges with increasing the partial melting of the crustalrock from the 1 to 8 h runs. Particularly, the progres-sive dissolution of plagioclase increases the Al2O3 andNa2O contents in the melt with increasing experimentaltime (Table 5). Chemical profiles of the major elementsin PF1 and MK72 glasses are shown in Figure 6. At thePF1-MK72 boundary layer, the ongoing crystallization ofnew plagioclases from the hybrid melt causes discontinu-ities in the elemental trends. The oxide chemical profilesdo not monotonically change as a function of the PF1

Table 3. Microprobe analyses of the plagioclases from the PF1-MK72 experiments.

restitic newwt% plagioclase σ (10) plagioclase σ (10)SiO2 57.11 0.55 52.20 0.53Al2O3 27.03 0.25 29.51 0.27Fe2O3 0.20 - 1.03 0.02CaO 9.11 0.08 13.00 0.17Na2O 6.33 0.06 4.00 0.05K2O 0.20 - 0.24 -Total 99.98 99.98Si 2.563 2.377Al 1.430 1.584Fe 0.007 0.035Ca 0.438 0.634Na 0.551 0.353K 0.011 0.014

Figure 4. PF1-VA38 chemical profiles of major elements of glasses.At the scale of the figure, the chemical trends from 1 to8 h run overlap.

and MK72 chemical gradients; the plagioclase crystal-lization produces a rapid decrease of Al2O3, Na2O andCaO and an increase of FeO, MgO and K2O in the oxidechemical profiles. Such a feature agrees with the Al2O3,Na2O and CaO incorporation into the lattice of crystalliz-ing plagioclase. Trace element profiles at the PF1-MK72boundary layer are shown in Figure 7. The chemical dif-fusion of trace elements into the PF1 melt is easily ob-servable. However, the diffusion process is different tothat expected; steep ramps starting from PF1 towards theMK72 side led to anomalous chemical distribution in thehybrid glass. Trace element concentrations in the hybrid192

Silvio Mollo, Valeria Misiti, Piergiorgio Scarlato

Figure 5. PF1-VA38 trace elements chemical profile of glasses.Selected diffusion trends for LILEs , REEs and HFSEsare reported.

glass are higher than those measured in both the PF1 andMK72 interacting melts.4. Discussion

4.1. Trace element behaviour

Trace element diffusion in silicate melts have been mea-sured in different ways in previous studies: chemical, traceand self diffusion [36–42]. However, experiments were de-signed to calculate trace element diffusivities by couplingtwo halves of the same synthetic glass or two halves of dif-ferent synthetic glasses, doped with variables amounts ofthe investigated elements. Thus, in terms of major compo-nents, a single melt chemistry was adopted, and thereforeelement diffusion occurred in the same compositional ma-trix. Even when two different melts were placed in contact(i.e., basalt-rhyolite or dacite-rhyolite), the investigationwas limited to the behaviour of few elements in the pureglasses [43–47].In the present study, we attempt to determine the traceelement diffusivities between two adjacent melts of con-

Table 4. Microprobe analyses of the differentiated basaltic glassand partially molten metapelite from the PF1-VA38 exper-iments. Only the 8 h data are reported, because analysesoverlap, within the experimental errors.

Run# PC254 PC254Glass PF1 VA38time (h) 8 8wt.% σ (10) σ (10)SiO2 52.79 0.38 76.14 0.35TiO2 0.62 0.03 0.81 0.06Al2O3 16.68 0.24 13.65 0.28FeO 10.28 0.11 4.96 0.22MnO 0.19 0.04 0.13 0.08MgO 6.09 0.11 1.14 0.22CaO 10.14 0.07 0.25 0.14Na2O 1.98 0.08 0.29 0.16K2O 1.07 0.05 2.48 0.1P2O5 0.16 0.03 0.15 0.05Total 100 100

Figure 6. PF1-MK72 major elements chemical profile of glasses.Only 1 and 8 h runs chemical trends are reported. Opencircle: 1 h run; Closed circle: 8 h run.

trasting compositions. The simultaneous presence of sev-eral natural chemical elements, gives us the possibility toinvestigate the diffusion process in conditions similar tothose occurring in nature. To calculate the diffusion coef-ficients of the trace elements, their chemical profiles havebeen treated as a binary diffusion model for a semi-infinitevolume by assuming a concentration-independent diffu-sion coefficient. This approach has been used for both thePF1-VA38 and PF1-MK72 chemical gradients. However,in the PF1-MK72 experiments, we consider that the crys-193

Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

Figure 7. PF1-MK72 trace elements chemical profile of glasses.Selected diffusion trends for LILEs, REEs and HFSEsare reported.

tallization of plagioclase from the hybrid layer modifiedthe original PF1-MK72 chemical gradients. According tothis we determine the trace element diffusion coefficientsby using the chemical gradients formed between the hy-brid melt and the basaltic one. In this model, the solutionof [48] second law is:Ci = Co − erf

(x2√Dit

), (1)

where Ci is the element concentration (wt%) at time t (s)and distance x (cm) from the interface; Co is the concentra-tion at the origin (x=0); and Di (cm2 s−1) is the diffusivityof the element. In Table 6, we report D values calculatedfor the trace elements with appreciable chemical gradientsand measurable diffusion profiles. The diffusion data com-puted for the PF1-VA38 experiments cover a range from0.36×10−7 (Hf) to 6.61×10−7 (Pb) cm2 s−1 and their val-ues agree with those observed in previous multi-elementdiffusion studies on basaltic melts [49, 49–51]. As it is

Table 5. Microprobe analyses of the partially molten meta-anorthosite glasses from the PF1-MK72 experiments.

Run# PC177 PC219 PC218 PC254time (h) 1 2 4 8wt.% σ (10) σ (10) σ (10) σ (10)SiO2 47.48 1.17 48.44 0.96 49.23 0.53 49.89 0.49TiO2 1.87 0.05 1.68 0.04 1.53 0.04 1.40 0.02Al2O3 22.78 0.59 23.21 0.48 23.55 0.28 23.84 0.24FeO 8.89 0.18 8.02 0.13 7.31 0.10 6.72 0.06MnO 0.18 - 0.16 - 0.15 - 0.13 -MgO 4.18 0.07 3.76 0.05 3.42 0.03 3.13 0.02CaO 9.46 0.27 9.43 0.21 9.40 0.14 9.38 0.10Na2O 3.71 0.13 3.97 0.10 4.18 0.05 4.36 0.05K2O 1.02 0.02 0.94 0.01 0.87 0.02 0.82 0.01P2O5 0.40 0.02 0.36 0.02 0.33 0.01 0.30 0.01Total 100.00 100.00 100.00 100.00

reported in Figure 8, LREEs (La, Ce, Pr, Nd, Sm, Eu andGd) are faster with respect to HREEs (Tb, Dy, Ho, Er,Tm and Yb, Lu), whilst HFSEs (Hf, Th and U) are theslowest ones. These diffusion patterns are consistent withvariations of diffusion coefficients as a function of the ionicradius and charge of the cations. In general, the diffusivitydecreases from LILEs, to LREEs, to HREEs and to HFSE(Figure 8), a trend that is correlated to the increase ofcharge in the same order; for elements of a given charge,diffusion coefficients generally decrease with increasingionic radius [36, 41, 42, 52]. The diffusion coefficients cal-culated for the PF1-MK72 experiments are one order ofmagnitude higher than those computed for the PF1-VA38runs (Table 6 and Figure 8). Figure 8 illustrates how thediffusion coefficients for HFSEs (Th, Zr, Hf, Nb) are com-prised within a range from 10.02×10−7 (La) to 70.12×10−7(Hf) cm2 s−1 in which both LILEs (Rb and Cs) and LREEs(La and Dy) values are included. These variations do notfollow the ionic properties (i.e., radius, charge and fieldstrength) and chemical grouping (LILEs, LREEs, HREEsand HFSEs) of the cations. Such a feature is ascribed tothe crystallization of new plagioclase in the hybrid meltthat occurs at the PF1-MK72 interaction layer. Trace el-ements not compatible with the plagioclase crystal latticeconcentrate in the hybrid melt, thus modifying the origi-nal PF1-MK72 chemical gradients. The occurring diffu-sive patterns agree with the “transient equilibrium” modelproposed by [35]. When two structurally different meltsare brought into contact and begin to mix by diffusion, themajor and trace elements of the melts show a preferencefor one melt relative to the other, with an attempt to estab-lish a distribution between the two melts that resemblesa two-liquid equilibrium. However, the melt structure is194

Silvio Mollo, Valeria Misiti, Piergiorgio Scarlato

largely determined by the proportion of network-formersspecies (principally SiO2 and Al2O3) that are slow-movingrelative to the other components that partition themselvesbetween the two interacting melts. Thus, the behaviourof SiO2 and Al2O3 controls the rate at which concentra-tion gradients of all the others species are eliminated. Inour experiments, the crystallization of plagioclase at thePF1-MK72 interaction layer affects the chemical gradientof SiO2 and Al2O3 (Figure 6). Consequently, their concen-trations do not vary monotonically, driven by a constantchemical gradient as occurs for the PF1-VA38 trends (Fig-ure 4). The trace element concentration increases in thehybrid melt, thus changing the liquid-liquid trace elementpartitioning between the two interacting starting compo-sitions. Furthermore, it has been suggested [50, 53, 54]that the diffusion of cations is influenced by the struc-tural reorganisation of the melt network. When the rate ofstructural reorganization of the melt network is compara-ble to, or greater than the mobility of the individual cations(“intrinsic diffusivity”), network reorganization dominatesthe mobility of all species (“extrinsic diffusivity”), result-ing in a clustering of diffusion coefficients. According tothis, diffusion coefficients for the PF1-MK72 runs showno dependence upon the ionic properties or geochemicalgrouping of the cations. We infer that this behaviour isdue to melt structural reorganization at the PF1-MK72interaction layer that completely overwhelms the intrinsicdiffusivity of the trace elements.

Figure 8. Trace element diffusion coefficients for both PF1-VA38and PF1-MK72 glasses plotted as a function of theirgeochemical grouping: LILEs (Large Lithophile Ion El-ements), LREEs (Light Rare Earth Elements), HREEs(Heavy Rare Earth Elements) and HFSEs (High FieldStrength Element).

4.2. Implications for magmatic processes

Results of previous studies on diffusion in silicate melts,suggest that diffusion is relatively slow, and that on a ge-ologic time scale diffusion is unimportant in large-scale

Table 6. Calculated trace element diffusion coefficients from boththe PF1-VA38 and PF1-MK72 experiments.

PF1-VA38 PF1-MK72Element D (10−7 cm2 s−1) σ Element D (10−7 cm2 s−1) σY 1.33 0.09 La 10.02 1.10La 1.98 0.40 Dy 61.23 6.09Ce 1.93 0.39 Hf 70.12 6.92Pr 1.90 0.38 Nb 11.03 1.30Nd 1.83 0.37 Zr 20.62 2.34Sm 1.80 0.36 Th 11.28 1.81Eu 1.72 0.34 Cs 12.21 1.19Gd 1.70 0.34 Rb 13.74 1.13Tb 1.67 0.33Dy 1.50 0.30Ho 1.38 0.28Er 1.30 0.26Tm 1.20 0.24Yb 1.11 0.22Lu 1.10 0.22Hf 0.36 0.03Pb 6.61 0.46Th 0.48 0.03U 0.36 0.03magma mixing [35–38, 45, 46]. This statement, however, isnot valid for short magma residence times [55–57]. Duringthese short periods digestion of crustal material in basalticmagma is extraordinarily rapid, as well as, melts genera-tion and extraction form crustal rocks [9, 12, 55, 58–60]. Itfollows that geochemical diversity within single eruptiveepisodes, hence variable overprinting of mantle geochem-ical signatures, is potentially common during magma in-teraction with crustal material. Moreover, several studiesshow that in the case of a large surface area with a “fin-gered” flow morphology at the contact between basalticand more differentiated or crustal magmas, diffusion inthe melt can be an important mixing process, at least forthe mobile trace elements [13–18]. These studies demon-strate that mobile trace elements can be redistributed ona large scale, during interaction at particular magmaticconditions, and that isotopic compositions can be modifiedselectively by diffusion without significantly changing themajor element composition of melts.Data from experimental studies conducted on both maficand silicic compositions, show that after 10.000 yrs at1200 °C, diffusion profiles in melts are relatively short withlengths (defined as x = (Dt)1/2) from 0.1 to few meters ([41]and references therein). Accordingly, the lengths of diffu-sion profiles range from 0.1 (Hf) to 2.57 (Pb) meters whendiffusion coefficients from PF1-VA38 experiments are used.

195

Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

On the contrary, the lengthening of the profiles is consid-erably higher in the case of PF1-MK72 diffusion coeffi-cients, showing values from 3.16 (La) to 8.37 (Hf) meters.It should be noted that magmatic modelling is highly ide-alized being based only on tracer diffusion data, withoutinclusion of gradients of chemical potentials for compo-nents within the melts. Furthermore, the effects of crys-tallization at the magma-country rock interface have neverbeen considered as a possible cause of altered chemicalgradients. Results from this study show that the forma-tion of new crystals changes the liquid-liquid partitioningof cations; under these conditions, the diffusivity of traceelements becomes surprisingly fast and unpredictable. Itis clear that our experiments are not sustainable for nat-ural long-term processes where heat is controlled by abalance between heat convection in the magmatic bodyand heat conduction into crustal rock. Thermal diffusionwill depress the thermal gradient between the two meltsand consequently the effects produced by a chemical dif-fusion process will be shorter. However, we can considerour results reliable in the simplest case of magma-crusttemperature equilibrium, when magmatic heat cannot beremoved rapidly away from the crustal rocks and a con-stant thermal gradient persists in the magmatic body.5. Conclusions

We have experimentally investigated the trace elementdiffusivities between basalt and two crustal rocks. Theformation of new crystals with the interaction betweencrustal and basaltic melts promotes diffusion processesthat are not controlled by the chemical-physical proper-ties of trace elements. We conclude that the behaviourof trace elements must be revised when crystallizationoccurs during magma-crust interaction. At these condi-tions, strong diffusional processes can be the reason fornoticeable mismatches, when Rayleigh models and simpleassimilation fractional crystallization calculations are per-formed to explain the chemical features of contaminatedmagmatic suites.Acknowledgments

This study was supported by Universitá Roma Tre andINGV of Roma. It was also supported by TRIGS Project“Sixth Framework Programme of the European Commis-sion and to the New and Emerging Science and Technol-ogy Pathfinder” and by Project FIRB MIUR “Developmentof innovative technologies for the environmental protectionfrom natural events”. Authors are greateul to D. Dolfi for

her scientific support. We thank B. Watson, G. Iezzi andG. Ventura for the critical review of the manuscript. Wealso thank M. Tiepolo, A. Cavallo and M. Serracino forthe analytical work.References

[1] Barley M.E., Incompatible-element enrichment inArchean basalts; a consequence of contamination byolder sialic crust rather than mantle heterogeneity,Geology, 1986, 14, 947-950[2] Knesel K.M., Davidson P., Isotopic disequilibriumduring melting of granite and implications for crustalcontamination of magmas. Geology, 1996, 24, 243-246[3] Lassen B., Bridgwater D., Bernsteinc S., RosingM., Assimilation and high-pressure fractional crys-tallization (AFC) recorded by Paleo-proterozoic maficdykes, Southeast Greenland, Lithos, 2004, 72, 1-18[4] Dungan M. A., Davidson J.P., Partial assimilative re-cycling of the mafic plutonic roots of arc volcanoes:An example from the Chilean Andes, Geology, 2004,32, 773-776[5] Tegner C., Wilson J.R., Robins B., Crustal assimila-tion in basalt and jotunite: Constraints from layeredintrusions, Lithos, 2005, 83, 299-316[6] Hildreth W., Moorbath S., Crustal contributions to arcmagmatism in the Andes of Central Chile, Contr. Min-eral. Petrol., 1988, 98, 455-489[7] Spera F.J., Bohrson W.A., Energy-Constrained open-system magmatic Processes I: General model andenergy-constrained assimilation and fractional crys-tallization (EC-AFC) formulation, J. Petrol., 2001, 42,999-1018[8] Hildreth W., Gradients in silicic magma chambers:implications for lithospheric magmatism, J. Geophys.Res., 1981, 86, 10153-10192[9] Huppert H.E., Sparks R.S.J., The generation of graniteby intrusion of basalt into the continental crust, J.Petrol., 1988, 29, 599-624[10] Reiners P.W., Nelson B.K., Ghiorso M.S., Assimilationof felsic crust and its partial melt by basaltic magma:thermal limits and extents of crustal contamination,Geology, 1995, 23, 563-566[11] Beard J.S., Ragland P.C., Crawford M.L., Reactive bulkassimilation: A model for crust-mantle mixing in sili-cic magmas, Geology, 2005, 33, 681-684[12] Annen C., Blundy J.D., Sparks R.S.J., The genesis ofintermediate and silicic magmas in deep crustal hotzones, J. Petrol., 2006, 47, 505-539[13] Grove T.L., Kinzler R.J., Baker M.B., Donelly-NolanJ.M., Lesher C.E., Assimilation of granite by basaltic196

Silvio Mollo, Valeria Misiti, Piergiorgio Scarlato

magma at Burnt Lava flow, Medicine Lake volcano,northern California: Decoupling of heat and masstransfer, Contrib. Mineral. Petrol. 1988, 99, 320-343[14] Baker D.R., Tracer versus trace element diffusion: Dif-fusional decoupling of Sr concentration from Sr iso-tope composition, Geochim. Cosmochim. Acta, 1989,53, 3015-3023[15] Lesher C.E., Decoupling of chemical and isotopic ex-change during magma mixing, Nature, 1990, 344, 235-237[16] Allen C.M., Local equilibrium of mafic enclaves andgranitoids of the Turtle pluton, southeast California:Mineral, chemical, and isotopic evidence, Am. Min-eral., 1991, 76, 574-588[17] Blichert-Toft J., Lesher C.E., Rosing M.T., Selectivelycontaminated magmas of the Tertiary East Greenlandmacrodike complex, Contrib. Mineral. Petrol., 1992,110, 154-172[18] Snyder D., Tait S., The imprint of basalt on the geo-chemistry of silicic magmas. Earth Planet, Sci. Lett.,1998, 160, 433-445[19] Álvarez-Valero A.M., Cesare B., Kriegsman L.M., For-mation of melt-bearing spinel–cordierite–feldsparscoronas after garnet in metapelitic xenoliths. Reac-tion modelling and geodynamic implications, J. Meta-morph. Geol., 2007, 25, 305-320[20] Álvarez-Valero A.M., Kriegsman L.M., Partial crustalmelting beneath the Betic Cordillera (SE Spain): Thecase study of Mar Menor volcanic suite, Lithos, 2008,101, 379-396[21] Cotkin S.J., Igneous and metamorphic petrology ofthe eclogitic Seljeneset Meta-anorthosite and relatedjotunites, Western Gneiss Region, Norway, Lithos,1997, 40, l-30[22] Griffin W.L., Brueckner H.K., REE, Rb-Sr and Sm-Ndstudies of Norwegian eclogites, Chem. Geol., 1985,52, 249-271[23] Laube N., Springer J., Crustal melting by ponding ofmafic magmas: A numerical model, J. Volcan. Geother.Res., 1998, 81, 19-35[24] Annen C., Sparks R.S.J., Effects of repetitive emplace-ment of basaltic intrusions on thermal evolution andmelt generation in the deep crust, Earth Planet. Sci.Lett., 2002, 203, 937-955[25] Appel P., Möller A., Schenk V., High-pressure gran-ulite facies metamorphism in the Pan-African Belt ofeastern Tanzania: P-T-t evidence against granuliteformation by continent collision, J. Met. Geol., 1998,16, 91-509[26] Spear F.S., Cheney J.T., A petrogenetic grid for peliticschists in the system SiO2-Al2O3-FeO-MgO-K2O-H2O, Contr. Mineral. Petrol., 1989, 101, 149-164

[27] Misiti V., Petrology of high-grade thermometamorphicrocks of Gennargentu (Sardinia, Italy), 2003, PhDthesis, University La Sapienza of Rome[28] Le Bas M.J., Le Maitre E.W., Streckeisen A., ZanettinB., A chemical classification of volcanic rocks basedon the total alkali-silica diagram, J. Petrol., 1986, 27,745-750[29] Bohlen S.R., Equilibria for precise pressure cal-ibration and a friction less furnace assembly forthe piston-cylinder apparatus, Neues Jahr. Mineral.Monat., 1984, 9, 404-412[30] Hudon P., Baker D.R., Toft P.B., A high-temperatureassembly for 1.91-cm (3/4-in.) piston-cylinder appa-ratus, Am. Mineral., 1994, 79, 145-147[31] Kress V.C., Carmichael I.S.E., The compressibility ofsilicate liquids containing Fe2O3 and the effect ofcomposition, temperature, oxygen fugacity and pres-sure on their redox states, Contr. Mineral. Petrol.,1991, 108, 89-92[32] Kushiro I., Partial melting of mantle wedge and evo-lution of island arc crust, J. Geophys. Res., 1990, 95,15929-15939[33] Kawamoto T., Hirose K., Au–Pd sample containers ofmelting experiments on iron and water bearing sys-tems, Eur. J. Mineral., 1994, 6, 381-385[34] Freda C., Gaeta M., Misiti V., Mollo S., Dolfi D.,Scarlato P., Magma–carbonate interaction: An exper-imental study on ultrapotassic rocks from Alban Hills(Central Italy), Lithos, 2008, 101, 397-415[35] Watson E.B., Basalt contamination by continentalcrust: Some experiments and models, Contr. Mineral.Petrol., 1982, 80, 73-87[36] Jambon A., Tracer diffusion in granitic melts: experi-mental results for Na, K, Rb, Cs, Ca, Sr, Ba, Ce, Eu to1300 °C and a model of calculation, J. Geophys. Res.,1982, 87, 10797-10810[37] Lowry R.K., Henderson P., Nolan J., Tracer diffu-sion of some alkali, alkaline-earth and transition ele-ment ions in basaltic and andesitic melt, and the im-plications concerning melt structure, Contr. Mineral.Petrol., 1982, 80, 254-261[38] Baker D.R., Watson E.B., Diffusion of major andtrace elements in compositionally-complex Cl- and F-bearing silicate melts, J. Non-Cryst. Sol., 1988, 102,62-70[39] Baker D.R., Tracer diffusion of network formers andmulti-component diffusion in dacitic and rhyoliticmelts, Geochim. Cosmochim. Acta, 1992, 56, 617-631[40] Mungall J.E., Dingwell D.B., Chaussidon M., Chemi-cal diffusivities of 18 trace elements in granitoid melts,Geochim. Cosmochim. Acta, 1999, 63, 2599-2610[41] Koepke J., Behrens H., Trace element diffusion in an-197

Trace element behaviour during interaction between basalt and crustal rocks at 0.5-0.8 GPa:an experimental approach

desitic melts: an application of synchrotron X-ray flu-orescence analysis, Geochim. Cosmochim. Acta, 2001,65, 1481-1498[42] Behrens H., Hahn M., Trace element diffusion andviscous flow in potassium-rich trachytic and phonoliticmelts, Chem. Geol., 2009, 259, 63-77[43] Fisk M.R., Basalt magma interaction with harzbur-gites and the formation of high-magnesium andesites,Geophys. Res. Lett., 1986, 13, 467-470[44] Kelemen P.B., Reaction between ultramafic rock andfractionating basaltic magma I. Phase relations, theorigin of calc-alkaline magma series, and the forma-tion of discordant dunite, J. Petrol., 1990, 31, 51-98[45] Chakraborty S., Dingwell D.B., Rubie D.C., Multicom-ponent diffusion in ternary silicate melts in the sys-tem K2O-Al2O3-SiO2: I. Experimental measurements,Geochim. Cosmochim. Acta, 1995a, 59, 255-264[46] Chakraborty S., Dingwell D.B., Rubie D.C., Multi-component diffusion in ternary silicate melts in thesystem K2O-Al2O3-SiO2: II. Mechanisms, systemat-ics and geological applications, Geochim. Cosmochim.Acta, 1995b, 59, 265-277[47] Liang Y., Davis A.M., Energetics of multicomponentdiffusion in molten CaO-Al2O3-SiO2, Geochim. Cos-mochim. Acta, 2002, 66, 635-646[48] Fick A., On liquid diffusion, Philosoph. Mag. J. Sci.,1855, 10, 31-39[49] Margaritz M., Hofmann A.W., Diffusion of Eu and Gdin basalt and obsidian, Geochim. Cosmochim, Acta,1978, 42, 847-858[50] LaTourette T., Wasserburg G.J., Fahey A.J., Self dif-fusion of Mg, Ca, Ba, Nd, Yb, Ti, Zr and U in hap-lobasaltic melt, Geochim. Cosmochim. Acta, 1996, 60,1329-1340[51] LaTourette T., Wasserburg G.J., Self diffusion of eu-ropium, neodymium, thorium, and uranium in hap-lobasaltic melt: The effect of oxygen fugacity and therelationship to melt structure, Geochim. Cosmochim.Acta 1997, 61, 755-764[52] Roselieb K., Jambon A., Tracer diffusion of potassium,rubidium and cesium in a supercooled jadeite melt,Geochim. Cosmochim. Acta 1997, 61, 3101-3110[53] Dingwell D.B., The effects of structural relaxation oncationic tracer diffusion in silicate melts, Chem. Geol.,1990, 182, 209-216[54] Mungall J.E., Dingwell D.B., Actinide diffusion in ahaplogranitic melt: Effects of temperature, water con-tent, and pressure, Geochim. Coscmochim. Acta, 1997,61, 2237-2246[55] Costa F., Dungan M., Short time scales of magmaticassimilation from diffusion modeling of multiple ele-ments in olivine, Geology, 2005, 33, 837-840

[56] Dufek J., Bergantz G.W., Lower Crustal Magma Gen-esis and Preservation: a Stochastic Framework forthe Evaluation of Basalt–Crust Interaction, J. Petrol.,2005, 46, 2167-2195[57] Kuritani T., Kitagawa H., Nakamura E., Assimilationand Fractional Crystallization Controlled by Trans-port Process of Crustal Melt: Implications from an Al-kali Basalt–Dacite Suite from Rishiri Volcano, Japan,J. Petrol., 2005, 46, 1421-1442[58] Koyaguchi T., Kaneko K., A Two-stage Thermal Evo-lution Model of Magmas in Continental Crust, J.Petrol., 1999, 40, 241-254[59] Bryan S.E., Ferrari L., Reiners P.W., Allen C.M.,Petrone C.M., Ramos-Rosique A., Campbell I.H.,New Insights into Crustal Contributions to Large-volume Rhyolite Generation in the Mid-TertiarySierra Madre Occidental Province, Mexico, Revealedby U-Pb Geochronology, J. Petrology, 2008, 49, 47-77[60] Van Der Zander I., Sinton J.M., Mahoney J.J., LateShield-Stage Silicic Magmatism at Wai’anae Vol-cano: Evidence for Hydrous Crustal Melting inHawaiian Volcanoes, J. Petrol., 2010, 51, 671-701

198