Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

ScienceDirect

Geochimica et Cosmochimica Acta 162 (2015) 259–275

Fast grain growth of olivine in liquid Fe–S and the formationof pallasites with rounded olivine grains

Giulio F.D. Solferino a,b,⇑, Gregor J. Golabek a,c, Francis Nimmo d,Max W. Schmidt a

a Department of Earth Sciences, ETH Zurich, 8092 Zurich, Switzerlandb School of Biological, Earth and Environmental Sciences, University College Cork, Cork, Ireland

c Bayerisches Geoinstitut, University of Bayreuth, 95440 Bayreuth, Germanyd Department of Earth and Planetary Sciences, University of California Santa Cruz, 95064 Santa Cruz, CA, USA

Received 4 September 2014; accepted in revised form 10 April 2015; Available online 18 April 2015

Abstract

Despite their relatively simple mineralogical composition (olivine + Fe–Ni metal + FeS ± pyroxene), the origin of pallasitemeteorites remains debated. It has been suggested that catastrophic mixing of olivine fragments with Fe–(Ni)–S followed byvarious degrees of annealing could explain pallasites bearing solely or prevalently fragmented or rounded olivines. In order toverify this hypothesis, and to quantify the grain growth rate of olivine in a liquid metal matrix, we performed a series ofannealing experiments on natural olivine plus synthetic Fe–S mixtures. The best explanation for the observed olivine grainsize distributions (GSD) of the experiments are dominant Ostwald ripening for small grains followed by random grain bound-ary migration for larger grains. Our results indicate that olivine grain growth in molten Fe–S is significantly faster than insolid, sulphur-free metal. We used the experimentally determined grain growth law to model the coarsening of olivine sur-rounded by Fe–S melt in a 100–600 km radius planetesimal. In this model, an impact is responsible for the mixing of olivineand Fe–(Ni)–S. Numerical models suggest that annealing at depths of up to 50 km allow for (i) average grain sizes consistentwith the observed rounded olivine in pallasites, (ii) a remnant magnetisation of Fe–Ni olivine inclusions as measured in nat-ural pallasites and (iii) for the metallographic cooling rates derived from Fe–Ni in pallasites. This conclusion is valid even ifthe impact occurs several millions of years after the differentiation of the target body was completed.� 2015 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

Pallasites are a group of stony-iron meteorites with asimple, yet vexing mineralogical composition. Buseck(1977) found pallasites to contain 64.9 vol% olivine,31.0% metal, 2.3% troilite (FeS), 1.2% schreibersite((Fe,Ni)3P), 0.4% chromite, and 0.2% phosphates on

http://dx.doi.org/10.1016/j.gca.2015.04.020

0016-7037/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: School of Biological, Earth andEnvironmental Sciences, University College Cork, Cork, Ireland.Tel.: +353 21 490 4582.

E-mail addresses: giulio.solferino@ucc.ie (G.F.D. Solferino),gregor.golabek@uni-bayreuth.de (G.J. Golabek), fnimmo@es.ucsc.edu (F. Nimmo), max.schmidt@erdw.ethz.ch (M.W. Schmidt).

average. An occasionally present additional phase is phos-phoran olivine. Texturally, pallasites consist of olivine sur-rounded by a metal matrix with a few additional minorphases. Recent findings provided pallasites bearing up to40 vol% pyroxene, e.g., North West Africa NWA 1911(Bunch et al., 2005), changing the classic definition.Originally separated into a Main Group (MG) and anEagle Station group (ES) (Scott, 1977a), pallasites (theMeteoritical Bulletin website lists 103 pallasite records)are nowadays subdivided into 5–7 groups based on the oxy-gen isotopes signatures, trace element concentrations in theFe–Ni metal, and mineralogical composition (Scott, 2007;Yang et al., 2010; Boesenberg et al., 2012).

http://dx.doi.org/10.1016/j.gca.2015.04.020mailto:giulio.solferino@ucc.iemailto:gregor.golabek@uni-bayreuth.demailto:fnimmo@es.ucsc.edumailto:fnimmo@es.ucsc.edumailto:max.schmidt@erdw.ethz.chhttp://dx.doi.org/10.1016/j.gca.2015.04.020http://crossmark.crossref.org/dialog/?doi=10.1016/j.gca.2015.04.020&domain=pdf

260 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

A further characteristic feature is the morphology of theolivines, which varies between angular and rounded, andconstitutes a key to deciphering the mechanism of pallasiteformation. Macroscopically, there are pallasite specimensbearing solely or prevalently well-rounded, well-sorted oli-vine grains, and samples with almost only angular frag-ments of olivine (fragmental pallasites; Buseck, 1977;Scott, 1977a). The former are commonly attributed to pro-longed annealing, the latter could be the result of a violentmixing or fracturing event. Among pallasites bearing frag-mental olivines, it is possible to distinguish those with aprevalence of large euhedral or subhedral crystals, oftenbroken into several pieces and/or crossed by multiple frac-tures (e.g., the Esquel pallasite), and those with a prepon-derance of shattered fragments in the sub-millimetrerange (Huckitta). Even in highly fragmental pallasites, thesmallest fragments are invariably moderately- towell-rounded. Specimens bearing exclusively one type ofolivine are uncommon, while a combination of two mor-phologies is frequent. The Seymchan pallasite is one ofthe rare samples containing all morphological olivine types.A classification of pallasites based on olivine textures waspresented by Scott (1977b). Nevertheless, there is hardlyany correlation between olivine type and chemical and min-eralogical features, only a tendency that pallasites withmore differentiated metal compositions and higher sulphurcontents (Yang et al., 2010) contain rounded olivine slightlymore frequently than fragmented olivine (Buseck, 1977;Scott, 1977a).

The peculiar characteristics of this meteorite family ledto a number of formation theories. On one hand, the gene-sis of pallasites was ascribed to magmatic differentiationwithin a single chondritic body in a region located nearthe core mantle boundary (CMB) (Ringwood, 1961;Lovering, 1962; Mason, 1963; Anders, 1964; Scott,1977a,b; Wasson and Choi, 2003; Boesenberg et al.,2012). Other authors suggested formation scenarios involv-ing an external source of heat and/or matter related toimpacts of various asteroids into the pallasite parent body(Mittlefehldt, 1980; Malvin et al., 1985; Scott and Taylor,1990; Ulff-Møller et al., 1998; Scott, 2007; Yang et al.,2010).

The composition of the metal fraction in pallasites, inparticular the low Ir content, is consistent with fractionalcrystallization of the liquid in the IIIAB iron meteorite par-ent body after crystallization of 80% of the metal (Scott,1977c; Wasson and Choi, 2003). Nevertheless, there is nounequivocal evidence of a common origin for IIIAB ironmeteorites and pallasites, as indicated by oxygen isotopes(Scott et al., 2009), the concentration of siderophile ele-ments (Yang et al., 2010), and cosmic ray exposures ages(Eugster et al., 2006). Moreover, minor element abundancesin olivine and REE concentrations in phosphates are bestmodelled by low pressure crystallization (Davis andOlsen, 1991; Hsu, 2003), difficult to reconcile with incom-plete silicate-metal separation at the boundary of a growingcore or with mixing of an olivine layer and metal at theCMB of an already differentiated planetesimal in the after-math of a disturbance generated by impacts (Malvin et al.,1985). Similar, the metallographic cooling rates of 2.5–

18 �C/Ma and 13–16 �C/Ma, determined for Main Groupand Eagle Station pallasites, respectively, are not compati-ble with annealing in proximity of the CMB (Yang et al.,2010).

Tarduno et al. (2012) detected a remnant magnetisationin microscopic Fe–Ni inclusions in olivine crystals of twoMain Group pallasites (Bryson et al., 2015). The measuredmagnetic intensities were ascribed to the presence ofdynamo action in their parent body, proposed to be activethroughout their formation and cooling below the Curietemperature of these Fe–Ni particles (�360 �C). Since thetemperature in the proximity of the CMB, while thedynamo is active, is many hundreds of degrees above theCurie temperature of Fe–Ni alloy (Tarduno et al., 2012),the region where pallasite-type material was formed hasto be placed at shallower depths.

In this model a collision event resulted in the fragmenta-tion of olivines in the target body mantle mixing with themetal from the impactor. Fragmentation is followed by oli-vine annealing at relatively shallow depths. In this scenario,the depth range at which the mixed material cools downdetermines the olivine textures, with fragmented or shat-tered olivines being little annealed at the shallowest depthswhile rounded olivines form by longer and hotter annealingdeeper in the planetesimal’s mantle. This hypothesis couldbe tested, if annealing of small fragments in the shallowmantle is shown to result in rounded grains with a grain sizecomparable to that observed in pallasite meteorites beforetemperatures decrease such that the annealing kineticsbecomes insignificant.

Saiki et al. (2003) performed annealing experiments on amixture of olivine and solid Fe–Ni, demonstrating thatrounding of angular olivine fragments in hot solid metalis efficient for the smallest fragments, but inefficient overmeaningful time spans for large fragments >1 mm. Saikiet al. (2003) also pointed out that the absence of sulphurin their experiments might have slowed down the rate oftextural annealing.

A different approach is to test whether grain growth ofolivine fragments in a metal matrix might yield the textureand maximum grain size of olivines observed in pallasiteswith rounded olivine. Guignard et al. (2012) exposed for-sterite plus solid nickel aggregates to high temperatures,finding that forsterite growth in a solid Ni matrix couldyield textures similar to those of pallasites, but growth rateswere too small. A direct comparison of Guignard et al.’s(2012) experiments with actual pallasites is difficult, sincetheir starting material was Fe- and S-free.

Our investigation aims to assess the grain growth kinet-ics of olivine fragments in Fe–S melt. Moreover, we evalu-ate the grain size distribution (GSD) of the experimentallygrown olivines. For comparison, we determine the GSD ofthree pallasites bearing rounded olivine and two with frag-mented olivines in order to discuss their origin and forma-tion conditions. Finally, we elaborate a model of olivinegrowth in a Fe–S matrix in a planetesimal (following theimpact of a differentiated body). In this model, the liquidportion of the impactor core is injected into an olivine layerof the planetesimal’s mantle, leading to the formation ofpallasite meteorites.

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 261

2. METHODS

2.1. Experimental technique

The starting materials were mixtures of San Carlos oli-vine and Fe–S powders in proportions of 60:40–90:10 vol%. Three runs with Fe–S served to test for carbondiffusion from the graphite capsule into the molten Fe–S.Olivine powder was obtained through crushing and subse-quent milling for a few hours. The final grain size was mea-sured through laser diffractometry on dispersed solutions(“Mastersizer 2000” instrument, Malvern InstrumentsLtd.) resulting in a mean of 1.8(5) lm. The Fe–S mixturecorresponds to the composition of the eutectic in theiron–sulphur binary at 1 GPa, i.e. 70.5 wt% Fe and29.5 wt% S (Teutectic Fe–S = 990 �C; Brett and Bell, 1969;Usselman, 1975; Fei et al., 1997), referred to hereafter asFe–S. The desired Fe–S composition was obtained by mix-ing sulphur pieces (99.999% purity, CeramTM Inc.) andiron powder (99.9% purity, mesh 5 lm CeramTM Inc.).

Olivine and the Fe–S powder were mixed in the desiredproportion, stored in vacuum at 50 �C for several days, andloaded into graphite capsules. The graphite capsules of4.0 mm height and 2.0 mm diameter were fitted into plat-inum capsules that were sealed by arc welding. All experi-ments were performed in an end-loaded piston cylinderpress with a 14 mm bore. Piston cylinder assemblies con-sisted of external NaCl- or Talc-Pyrex sleeves, mostly astraight graphite furnace of 36 mm length and 1 mm wallthickness, and inner crushable MgO cylinders containingthe capsule and a mullite thermocouple ceramics. Onlyfor runs OFS-nf-1 and -14 (Table 1) the graphite heaterswere stepped, with the top and bottom thirds having areduced thickness of 0.5 mm. B-type Pt–Rh thermocoupleswere employed to control the temperature with an accuracy

Table 1Experimental conditions and melt abundance after image analyses.

Run T Duration Metal fractionb

(�C) (h) (vol%)

Runs with olivine plus Fe70S30 mixtures

OFS-1 1100 72 19.5 (6)OFS-2 1200 72 19.4 (7)OFS-3 1300 72 19.3 (6)OFS-4 1300 72 9.4 (4)OFS-5 1300 72 14.9 (6)OFS-nf-1a 1100 150 19.9 (5)OFS-9 1300 39 20.2 (8)OFS-10 1300 145 19.5 (7)OFS-11 1100 300 20.3 (8)OFS-nf-14a 1430 8 19.7 (4)OFS-15 1300 72 45.2 (26)

Runs with iron-sulphide

FS-1 1100 3FS-2 1300 22FS-3 1200 25

a Assembly with a stepped furnace; nf = new furnace.b Metal abundance from digital image analyses of run products.

The error from repeated measurements is reported betweenbrackets.

of ±5 �C. The thermocouple was separated from thePt-capsule by a 0.6 mm thick corundum disk. Run pressurewas calibrated against fayalite + quartz = orthoferrosilite(Bohlen et al., 1980) and the quartz–coesite transition(Bose and Ganguly, 1995). Heating was terminated by shut-ting off power, resulting in a cooling rate of ca. 50 �C/sdown to

Fig. 1. Back Scattered Electrons (BSE) images of experimental samples (a–d) and after processing for digital analyses (e and f). Dark grey/black is olivine, light grey/white is quenched Fe–S melt. Exsolution features in the Fe–S matrix are not visible due to their low BSE contrast incomparison to olivine. (a) run OFS-1 (1100 �C, 72 h, 20 vol% Fe–S); (b) run OFS-11 (1100 �C, 300 h, 20 vol% Fe–S); (c) run OFS-3 (1300 �C,72 h, 20 vol% Fe–S); (d) run OFS-4 (1300 �C, 72 h, 10 vol% Fe–S). (e) binary image of OFS-3; (f) binary image of OFS-3 with touching olivinegrains separated manually by thin white lines.

262 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

(b) The number of olivine grains in the sections had to be>100–150, to obtain a statistically significant GSD; (c)Olivine, FeS and Fe–Ni should represent >90% of the sec-tion’s area; (d) Olivine abundance should not exceed 90%and Fe + FeS abundance should not exceed 50%. Highquality images of pallasites satisfying these criteria wereavailable for the meteorites Brenham, Seymchan, andSpringwater. We also analysed the size distribution of frag-mental pallasite Brahin and a different portion ofSeymchan. For these slabs, image analyses started fromphotographs instead of BSE images. Grain size was equatedto the diameter of each object’s circle-equivalent area. 2Dapparent grain sizes were corrected for sectioning effectsof a sphere with the formula: d3D = d2D/(p/4) (Konget al., 2005).

Roundness and the ratio between perimeter and circleequivalent diameter are used to quantify the shape of the

grains in experimental samples and rounded pallasite slabs.Note that roundness is a measure for equi-dimensionalityand not for smoothness of the surface. The quantitative for-mulation of roundness follows Waddell (1932): 4pA/MA2,where A is the area of the grain and MA is the length ofthe major axis. The circle equivalent diameter is: 2 (A/p)1/2.

2.4. Some background on grain growth

In an aggregate of grains a decrease of grain surfacearea via grain growth leads to a reduction of the surfaceto volume ratio and stabilizes the system by reducing itsGibbs free energy. This is a direct consequence of thegrain boundary energy per unit volume E being inverselyproportional to the grain radius r: E / c/r, where c is thespecific surface energy of the coarsening phase (Evanset al., 2001).

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 263

Grain coarsening could be interpreted as a consequenceof normal grain growth, which is expressed by the mergingof multiple grains to form larger ones (Atkinson, 1988).Normal grain growth theory is usually employed for thestudy of mono-phase aggregates. Ostwald ripening theoryexplains the growth of grains dispersed in a solid or liquidmatrix (Lifshitz and Slyozov, 1961; Wagner, 1961). Normalgrain growth in the presence of a second phase is not influ-enced by the volume fraction of the coarsening phase,whereas Ostwald ripening depends on the ratio of the coars-ening phase and matrix (Yoshino and Watson, 2005; Fauland Scott, 2006; Guignard et al., 2012).

Whatever the governing mechanism, the increase ofaverage grain size (d) with time (t) can be described bythe equation:

dn � dn0 ¼ k0 expð�Ea=RT Þt ð1Þ

where d0 is the starting grain size, d the grain size at time t, nthe growth exponent, k0 a characteristic constant, Ea activa-tion energy, R the gas constant, and T absolutetemperature.

For normal grain growth of a single, pure phase, agrowth exponent of n = 2 is predicted, while chemicalexchange via other phases yields n > 2 (Atkinson, 1988).Ostwald ripening of solid particles may be rate-limited bydiffusion of material from one grain to another throughthe matrix or by removal and precipitation of material fromthe crystal surface, defined as diffusion controlled and sur-face reaction controlled ripening (Martin et al., 1997). Thegrowth exponent predicted for diffusion controlled Ostwaldripening is n = 3, while for surface reaction controlledOstwald ripening it is n = 2.

The distribution of the relative size of grains forming anaggregate that underwent normalised grain growth orOstwald ripening is time-invariant, though the averagegrain size increases with time. Hence, grain size distribu-tions normalised by dividing the value of each size bin bythe mean grain size dm are characteristic for a specificgrowth process. The position of the distribution’s maxi-mum frequency and the width of the grain size distributionscatter with respect to the mean grain size are potentialdiagnostic characteristics (Lifshitz and Slyozov, 1961;Hillert, 1965).

For grain growth controlled by diffusion-limitedOstwald ripening, Hillert (1965) defined a normalised grainsize distribution having a cut-off at 1.5 dm and the peakpositioned at d > dm. First-order surface reaction controlledOstwald ripening, governed by the movement of discretegrowth units from one position to another along grains sur-faces (Hanitzsch and Kalhweit, 1969), has also a peak atd > dm and a maximum grain size dmax of 3.0 dm. Rayleigh and Log-normal distributionscan be adjusted by varying the parameters in their definingequations (Eqs. 4 and 5 in Faul and Scott, 2006).

3. RESULTS

3.1. Composition of olivine and Fe–S melt

Olivine is compositionally homogeneous in each run,indicating that chemical equilibration is reached even forshort experiments and relatively low temperatures(Table 1). With respect to the starting material, olivineFe-contents increase in all experiments by about 2 wt%except for OFS-15 (Table 2), where annealing of a mixturecontaining 40 vol% of Fe–S yielded an olivine shiftedtowards a more Mg-rich composition. OlivineNiO-contents dropped to 60.018 wt% (Table 2), startingfrom an average of 0.44 wt% (Galoisy et al., 1995). Themelting temperature of the resulting Fe–S compositions islower than run temperature in all experiments (Usselman,1975; Fei et al., 1997), the presence of 1–2 wt% Ni in theiron-sulphide may have further reduced the melting temper-ature of the alloy. The texture of the Fe–S portion is indica-tive of exsolution during quench. In run OFS-15 featuresthat might be related to the stability of two coexistingmetallic phases are observed: one with a composition simi-lar to FeS (with clear quench features) and one that isalmost pure iron. The latter has triangular orpseudo-circular shapes (Fig. 2). Precise analysis of theFe–S alloy is difficult, because melt pools are usually small(4–5 lm) but a beam size of �10–15 lm would be needed toaverage across the quench heterogeneities. Nevertheless, theFe–S matrix shows a decrease in iron content (by 5–9 wt%)with a concomitant increase of oxygen (0.2–1.2 wt%) andsulphur (4–9 wt%, Table 2). Analysis of the two quenchphases with a focused beam (beam size 1 lm) revealed thatthe brighter phase has a composition with >80–90 wt% ironand contains up to 2.5 wt% Ni (run OFS-1), whereas thelight grey phase composition is almost stoichiometric FeS(63.5 wt% Fe). The minute size of the Fe–Ni alloy (

Table 2Composition of olivines and iron-sulphide alloys after annealing.

Runs with olivine plus iron-sulphide mixtures

Oxide (weight%) OFS-1 OFS-2 OFS-3 OFS-4 OFS-5 OFS-nf-1 OFS-9 OFS-10 OFS-11 OFS-nf-14 OFS-15a S. Carlos Typical s.d.b

SiO2 39.81 39.97 39.79 40.51 39.63 40.28 40.27 40.21 40.17 39.78 42.07 40.81 0.26MgO 48.96 48.86 49.08 48.80 49.49 48.66 49.16 49.31 49.53 49.28 54.29 49.60 0.24FeO 11.29 11.22 11.13 10.77 10.73 11.53 11.24 10.77 10.35 10.98 3.54 8.92 0.66MnO 0.10 0.11 0.11 0.12 0.11 0.11 0.10 0.10 0.09 0.00 0.03 0.23 0.11NiOc 0.012 n/a 0.000 0.005 n/a n/a n/a n/a 0.006 0.000 0.018 0.44 0.008total 100.18 100.15 100.11 100.21 99.96 100.59 100.77 100.40 100.15 100.04 99.95# of analysis 15 16 15 8 30 14 18 15 9 8 9Mg# 89 89 89 89 89 88 89 89 90 89 96 91

Element (weight%)Fe 63.18 64.09 62.78 62.11 62.83 62.70 62.97 62.85 61.98 62.26 61.08 0.65S 35.99 34.74 36.32 35.89 35.96 36.55 35.13 36.77 36.60 36.13 39.34 0.53O 0.53 0.71 0.77 0.75 0.76 0.61 0.37 0.19 0.78 1.17 0.28 0.19Total 99.70 99.54 99.87 98.75 99.55 99.86 98.47 99.81 99.36 99.56 100.71O (atomic%) 1.50 1.94 2.08 2.13 2.11 1.74 1.02 0.52 2.12 3.17 0.83 0.51# of analysis 9 9 8 4 10 9 12 9 7 11 8

Runs with iron-sulphide mix

Element (weight%) FS-1 FS-2 FS-3 s.d.

Fe 66.44 67.96 65.87 0.61S 29.36 30.46 30.12 0.59O 0.51 0.80 0.29 0.17Total 96.31c 99.22 96.28c

# of analysis 18 12 15

Beam size for WDS analysis set to 50 lm.a The iron-sulphide presents evidences of stability of two phases on top of quench exsolution features in the sulphur-rich phase. One phase has composition similar to FeS (see the composition in

the table), and the other is almost pure iron with minor amount of nickel.b s.d. means single standard deviation.c Nickel signal acquisition time for EMPA analyses was set to 120 s in order to enhance the detection limit.

264G

.F.D

.S

olferin

oet

al./G

eoch

imica

etC

osm

och

imica

Acta

162(2015)

259–275

Fig. 2. Back Scattered Electron images of run OFS-15. The imagecontrast was set such that olivine appears black, FeS is dark grey,the white phase is almost pure iron. White circle is almost pure irontexturally unrelated to exsolution during quench. Grey rectanglesare typical quench-induced textural elements.

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 265

3.2. Sample textures

The main textural features of experimental charges donot appear to be influenced by annealing time or tempera-ture (Fig. 1a–d). Roundness of olivine grains is, withinerror, the same for all runs (Table 3). Grains are wellrounded (Waddell, 1932) and show underdeveloped facets,straight grain boundaries and 120� triple junctions atmutual contacts. The ratio between the perimeter and thecircle equivalent diameter of grains increases slightly forruns containing 40–15 vol% of Fe–S, and displays a markedincrease for a Fe–S content of 10 vol% (Table 3). The Fe–Squench is either distributed in pools surrounding small clus-ters of olivine, or located at triple junctions. Changing theolivine/Fe–S ratio from 60/40 to 90/10 caused the disap-pearance of Fe–S melt pools in favour of metal pocketslocated at triple junctions (Fig. 1c and d). In run OFS-4(10 vol% Fe–S) more than 90% of the Fe–S pockets are iso-lated. For samples with 20 vol% Fe–S, melt pools are com-parable in size to olivine grains, larger pools being observedin those samples with a larger average olivine size. Fe–S israrely present along straight grain boundaries (Fig. 1c) andappears as

Table 3Results of digital image and NGSD analyses.

Run name Annealingtime (h)

AnnealingT (�C)

Nominal metalfraction (vol%)

Average grainsize (lm)

Numberof grains

v2 value Roundnessa Perim/CircEqDiamb2nd Order

S. Diff.1st OrderedS. Diff.

Rayleigh Log-normal

OFS-1 72 1100 20 3.2 (8) 731 0.48 0.48 1.33 13.63 0.74 (13) 1.24 (11)OFS-nf-1 150 1100 20 6.1 (8) 1303 0.48 0.59 0.57 12.02 0.73 (12) 1.13 (6)OFS-11 300 1100 20 7.3 (12) 621 0.61 1.44 0.50 35.80 0.74 (12) 1.11 (4)OFS-9 39 1300 20 10.6 (5) 658 0.47 0.68 2.41 18.66 0.76 (11) 1.16 (5)OFS-3 72 1300 20 16.3 (4) 383 0.47 0.51 0.89 22.70 0.78 (11) 1.19 (10)OFS-10 145 1300 20 18.6 (9) 303 0.48 0.49 0.78 47.73 0.76 (10) 1.13 (4)OFS-15c 72 1300 40 9.0 (10) 1308 0.55 0.65 0.78 13.40 0.73 (12) 1.10 (3)OFS-3 72 1300 20 16.3 (4) 383 0.47 0.51 0.89 22.70 0.78 (12) 1.19 (10)OFS-5d 72 1300 15 18.6 (4) 352 0.56 2.60 3.25 9.17 0.75 (12) 1.19 (8)OFS-4e 72 1300 10 21.6 (9) 212 0.56 0.98 0.66 23.38 0.73 (13) 1.51 (22)OFS-2 72 1200 20 7.1 (6) 604 – – – – 0.77 (13) 1.16 (8)OFS-nf-14 8 1430 20 8.9 (5) 554 – – – – 0.75 (12) 1.13 (6)

Pallasite name Average grainsize (mm)

Brenham 4.88 (16) 654 0.61 24.52 0.57 1.08 0.68 (15) 1.10 (8)Seymchan 3.80 (17) 153 1.11 5.71 0.69 0.67 0.57 (16) 1.25 (21)Springwater 6.21 (15) 303 0.75 11.99 4.62 1.04 0.67 (18) 1.15 (10)Brahin 1.24 (13) 277f – –Seymchan breccia 1.96 (19) 599f – –

a Roundness: 4 � [Area]/p � [Major Axis]2; a regular hexagon has a roundness of 0.99, an idiomorphic olivine cross section with an aspect ratio of 1.5 or 2.0 has a roundness of 0.70 or 0.52,respectively.

b Perim/CircEqDiam: {[Perimeter]/2 � [Area/p]1/2}/p.c Olv/Fe–S = 60/40.d Olv/Fe–S = 85/15.e Olv/Fe–S = 90/10.f Number of fragments.

266G

.F.D

.S

olferin

oet

al./G

eoch

imica

etC

osm

och

imica

Acta

162(2015)

259–275

Fig. 3. Grain growth curves. (a) Diamonds: experiments at 1300 �C; squares: experiment at 1200 �C; circles: experiments at 1100 �C. Linesrepresent modelled grain growth curves at each temperature. d0 is set equal to 1.8 lm for all models. Error bars represent 2r values; (b) Graingrowth curve at 1100 �C (grey) with calculated 2r-error envelope (black). Error curves were computed after propagating the experimental andanalytical errors on n, Ea, and k0. Error bars represent 2r values.

0

0.2

0.4

0.6

0.8

1.0

1.2

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Rayleigh

2nd order

1st order

OFS-1

Freq

uenc

y

d / d0

0

0.2

0.4

0.6

0.8

1.0

1.2

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5d / d0

Rayleigh

2nd order

1st order

OFS-11

0

0.2

0.4

0.6

0.8

1.0

1.2

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Rayleigh

2nd order

1st order

OFS-9

Freq

uenc

y

d / d0

0

0.2

0.4

0.6

0.8

1.0

1.2

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5d / d0

Rayleigh

2nd order

1st order

OFS-10

a b

c d

Fig. 4. Normalised grain size distributions of experimental samples and comparison to calculated GSDs. The vertical dashed lines correspondto average grain sizes. The shape of the Rayleigh distribution curves changes from run to run, as each is optimised to fit the observed GSD(Section 3.4), while the shape of 1st and 2nd order surface reaction controlled GSDs is invariable. Log-normal distributions have not beenplotted since their fit with experimental normalised GSDs was very poor. In (a), we also show the measured data and bins, avoided in (b–d) forclarity.

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 267

v2 ¼ 1nb

Xnb

jf c � f mj2

!ð2Þ

where nb is the total number of bins, fc the theoretical nor-malised frequency and fm the measured normalised fre-quency of each bin (Table 3). Rayleigh and Log-normaldistribution fits to each measured GSD were optimised by

adjusting the parameter(s) in their defining equations(Faul and Scott, 2006). The theoretical distributions for dif-fusion and surface reaction controlled Ostwald ripening arenot adjustable. The quality of the fit with Ostwald ripeningand Rayleigh theoretical distributions held similar results,whereas the fit with Log-normal distribution is always poor(Table 3).

BRENHAM

SEYMCHAN fragments

SEYMCHAN rounded

0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

BRAHIN

0

0.2

0.4

0.6

0.8

1.0

1.2

0 0.5 1.0 1.5 2.0 2.5 3.0

Freq

uenc

y

d / d0

SEYMCHAN

SPRINGWATER

Log normal Rayleigh2nd order SPRINGWATER

0 0.5 1.0 1.5 2.0 2.5 3.00

0.2

0.4

0.6

0.8

1.0

1.2

d / d0

d / d0

0

0.2

0.4

0.6

0.8

1.0

1.2

Freq

uenc

y

0 0.5 1.0 1.5 2.0 2.5 3.00

0.2

0.4

0.6

0.8

1.0

1.2

d / d0

a b

c d

Fig. 5. Normalised grain size distributions of pallasites and comparison with calculated normalised GSDs for rounded olivines. Verticaldashed lines correspond to average grain size. (a) Normalised GSDs of three pallasites bearing rounded olivine grains; (b) normalised GSD ofSpringwater pallasite compared with theoretical distributions for Rayleigh, Log-normal, and 2nd order surface reaction controlled growth.Rayleigh and Log-normal normalised GSD were optimised for best fit with the observed curve; (c) normalised GSDs of two different areas ofSeymchan pallasite, one bearing mainly rounded olivine, and the other mainly fragmented crystals (Fig. 6); (d) normalised GSD of thefragmental pallasite Brahin, highlighting the extremely long tail characteristic of mechanically induced fragmentation, followed by littleannealing.

268 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

The results for the normalised GSD of Brenham,Springwater, and the rounded olivine-bearing slice ofSeymchan are more difficult to interpret, since maximumgrain size and maximum frequency give contrasting indica-tions (Fig. 5b). The v2 test indicates that first order surfacereaction controlled growth is the least favourable mecha-nism to interpret olivine growth in these samples.

4. DISCUSSION

4.1. Chemical equilibration of olivine and Fe–S

The almost constant olivine composition in experimentsperformed at different temperatures and annealing times atidentical olivine/Fe–S proportions (Tables 1 and 2), indi-cates that chemical equilibration is readily reached at ourexperimental conditions. Attaining the equilibrium olivinecomposition involves redox reactions such as

Mg2SiO4 þ 6FeSþ 4O2 ¼ Fe2SiO4 þ 2MgFe2O4 þ 3S2or

Fe2SiO4 þ 1=2 S2 ¼ FeSþ FeSiO3 þ 1=2 O2

(Gaetani and Grove, 1999). Such reactions necessarilyinvolve a second oxide or silicate phase, i.e. Mg-ferrite orferrosilite, which were not observed in our runs, but couldhave gone unnoticed. An alternative would be the oxidationof graphite to CO2 which could reside in the porosity of thegraphite capsule, but this would constrain oxygen fugacityto the graphite–CO–CO2 (CCO) buffer, which is inconsis-tent with the oxygen fugacities calculated for the experi-ments. Righter et al. (1990) developed an equation torelate fO2 and olivine forsterite content in pallasites. Theforsterite contents observed in our runs (Table 2) implyan equilibrium fO2, expressed as difference to the iron–wüs-tite (IW) oxygen buffer, of DIW � 0.55 to DIW � 0.85 (inlog units). These values are 4–5 log units below the CCObuffer at 1 GPa. This is consistent with Holloway et al.(1992) suggesting that graphite only imposes a maximumfO2 and in fact the fairly reduced Fe–S in our experimentsleads to reduced fO2-values. In principle, Mg-contents ofolivine should increase with Fe–S fraction while oxygenfugacity decreases. This is clear for experiment OFS-14 with40 vol% Fe–S, but less so for experiments with 10–20 vol%Fe–S (Tables 1 and 2).

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 269

The average NiO content of San Carlos olivine is0.44 wt% (Galoisy et al., 1995), but almost no Ni was pre-sent after annealing (Table 2). Analyses of the iron-sulphidephase revealed that Ni partitioned into the metal alloy, asexpected for the reduced fO2 of our experiments, althoughquantification was difficult due to the quench exsolutiontextures of the metal. If Ni migrated entirely into theiron-sulphide melt, Ni contents would result to 0.35–1.56 wt% for olivine volume fractions of 60–90%.

4.2. Grain growth of olivine in molten Fe–S

Grain growth theory predicts grain growth exponentsfor Ostwald ripening or normal grain growth of 2–3. Forolivine n = 2 was measured exclusively in experiments withnominally pore-free pure olivine aggregates (Karato, 1989).Experiments with olivine plus abundant basalt melt gaven = 3 (Cabane et al., 2005; Schmidt et al., 2012), whereasFaul and Scott (2006) obtained n = 4.3 for olivine + 2–4 wt% basalt. In studies involving olivine + pores or olivi-ne + nickel values of n P 3.5–5 were computed (Nicholsand Mackwell, 1991; Guignard et al., 2012). For a normalgrain growth process growth exponents n > 3 are obtainedby modelling the coarsening of the dominant phase forgrain growth limited by grain boundary diffusion in thematrix phase (n = 4, Ardell, 1972) or by diffusion viaone-dimensional pathways (n = 5, Johnson et al., 1999),both implying a low solubility of olivine in the matrixphase. In both cases the fraction of the matrix phase shouldnot effect the growth rate and hence n.

In our experiments an increase of the Fe–S fractionreduces the olivine growth rate (Table 3), and the growthexponent, computed for the experiments with 20 vol%Fe–S, is n < 3. This seems to preclude the possibility thatolivine growth is controlled by normal grain growth in pres-ence of a second phase and favours Ostwald ripening. Onthe other hand, Ostwald ripening is viable only when thesolubility of the growing phase in the surrounding matrixis non-negligible (Lifshitz and Slyozov, 1961; Wagner,1961). Whether the solubility of Mg, Si, and O in theiron-sulphide melt at the experimental conditions satisfiesthis condition cannot be ascertained from our currentexperiments and data available in the literature.

It should be noted that for anisotropic phases the dis-tinction between normal grain growth and Ostwald ripen-ing might not be relevant (Uli Faul, personalcommunication), and a more complex, intermediate graingrowth mechanism might be envisaged. Fan and Chen(1997) modelled the grain growth of a two-phase system,where both phases were anisotropic, and verified that achange of the volume fraction of the second phase and/orof the interfacial (grain boundary) energy does not modifythe growth mechanism leaving, the growth exponentunchanged. Fan and Chen (1997) noted that grain growthof the dominant phase was controlled by volume diffusion(normal grain growth, n = 3), whereas for the second phaselong distance diffusion prevails (Ostwald ripening, n = 2).

The activation energy of 289 kJ computed for our exper-iments is similar to those of oxygen and silicon lattice diffu-sion in olivine given by Houlier et al. (1988, 1990) and

Gerald and Jaoul (1989). This may indicate that the processresponsible for the grain growth requires diffusion or latticescale reorganisation in the olivine, such as for volumereaction-controlled coarsening. The apparent change inthe GSD for the longest runs may be taken as indicationthat steady-state grain growth was not achieved in allexperiments. However, the impracticality of unequivocallydefining the dominant grain growth mechanism precludesfurther discussion.

4.2.1. Comparison with previous studies

Our experiments aim at quantifying the growth processof olivine grains in pallasite-like material. Other studies onannealing of olivine + metal at high temperature usedolivine + Fe–Ni (Saiki et al., 2003) and forsterite + Ni(Guignard et al., 2012) at subsolidus conditions. For anolivine:Ni ratio of 80:20, Guignard et al. (2012) determineda growth exponent n = 3.5 (or 4.8 if the shortest annealingexperiment was not considered), i.e. a much slower coarsen-ing rate than in our experiments. Guignard et al. (2012)compared their result with data on olivine rounding fromSaiki et al. (2003), concluding that annealing in Fe–Ni ren-ders a faster growth process. Iron in the matrix phase wasproposed to promote faster diffusion leading to the conclu-sion that the nature of the matrix phase influences theripening rate of the dominant phase. Our experiments yieldmuch faster growth rates for a metal matrix with highS-contents in line with an enhanced diffusivity or mobilityof chemical species.

Oxygen fugacity may affect olivine annealing in Fe–S.In order to diffuse in the Fe–S melt, Mg would probablybe complexated as S-species, which would be stabilizedby reducing conditions. Oxygen fugacity in the experi-ments of Saiki et al. (2003) was buffered to Ni–NiO(NNO) and in the case of Guignard et al. (2012) wasNNO-3, i.e. 5 and 2 log units above the oxygen fugacityof our samples. Moreover, chemical diffusion in a liquidphase is orders of magnitude faster than in a crystallinemedium (Bhat, 1973; Bakker et al., 1990). The combinedeffect of our lower oxygen fugacity, the S-rich metal com-position and the liquid vs. solid state of the matrix phaseresults in the observed faster growth rate of olivine in thisstudy.

4.2.2. Comparison with natural pallasites containing rounded

olivines

The main difference between the natural and experimen-tal samples is the composition of their metal-sulphide frac-tion. Brenham and Springwater contain 5.8 and 5.3 wt% ofFeS (troilite) and 28.4 and 27.9 wt% Fe–Ni, respectively(Buseck, 1977). The phase proportions of Seymchan arenot reported, but a dominance of Fe–Ni over FeS is appar-ent in the available slices and slabs. In comparison, ourexperimental metal alloy corresponds to a mixture of80 wt% FeS and 20 wt% Fe.

In the pallasite slices FeS amounts to P15% of the areaof Brenham, 10–15% of Seymchan and

Fig. 6. Photographs of pallasite slices. Images (a–f) are slices of the pallasites Springwater, Brenham, Seymchan (rounded portion), Brahin,Seymchan (fragmental), and again Seymchan (fragmental). White and yellow arrows point at agglomerates of FeS. The white rectanglerepresents a scale bar of 10 mm length.

270 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

the contacts between Fe–S and olivine, and olivine grainsbeing almost always in mutual contact, even in 2D sections(Figs. 1, 2 and 6a–c).

4.2.3. Rounding of olivine fragments

Rounding of olivine fragments at our experimental con-ditions seems to occur during the earliest stages of anneal-ing and does not increase with run duration (Table 3). Thisobservation confirms the suggestion of Saiki et al. (2003)that the presence of sulphur might catalyse the roundingprocess. The olivine fraction seems to affect the degree ofroundness of grains, quantified by the ratio between grainperimeter and circle equivalent diameter, but only at olivinevolume fractions beyond 85% (Table 3). Guignard et al.(2012) observed that olivine grains were more faceted inexperiments with the largest olivine fraction in olivine–nickel mixtures. This effect could be related to differentgrain boundary energies between olivine–olivine and oli-vine–Fe–S (or olivine–Ni) boundaries.

4.3. Olivine grain growth in pallasites

The width of normalised GSDs measured for the threeselected pallasites extends from a minimum normalisedgrain size between 0.0 and 0.5 dm to a maximum sizebetween 2.5 and 3.0 dm, which is fundamentally similar tothe experimentally determined ones.

Unlike the experimental samples, for natural pallasitesmany factors may have reduced the olivine grain growth(and rounding) rate and potentially changed the dominantgrain growth mechanism. One possible factor is the compo-sition and volume proportion of the metal–sulphide alloy.

We observe that an increase in the metal fraction drasticallyreduces the grain growth rate (Table 3). Moreover, theaverage sulphur amount measured in the metal matrix ofpallasites (0.4–1.0 wt%; Buseck, 1977) is much smaller thanin the experiments (36 wt% S). Possibly our experimentallydetermined growth rate for olivine fragments in molten Fe–(Ni)–S is faster than for lower Fe- (i.e. highest Ni) andS-contents. Even if composition and fraction of the metalalloy could influence the growth and rounding rate andmechanism, only a weak correlation was found betweenthe type of olivine present in the pallasite samples and thesulphur and/or nickel amounts (Buseck, 1977; Boesenberget al., 2012).

4.4. A comparison of boundary conditions of the experiments

and pallasite formation

In order to assess the representativeness of the experi-mentally determined grain growth parameters for olivinecoarsening in pallasites, we compare the experimental con-ditions with P, T, fO2, and composition of the interior ofterrestrial planetesimals. The experiments were performedat 1 GPa, equivalent to �35 km depth in the Earth, butin a body with r 6 600 km, 1 GPa exceeds the maximumpressure. The topology of the Fe–FeS system does notchange significantly for pressures of 0–1 GPa (Brett andBell, 1969; Usselman, 1975; Fei et al., 1997), any deductionrelative to the eutectic melting temperature of the alloy andthe fraction of molten Fe–S at a given temperature remainshence valid for planetesimals. Parameters controlling graingrowth, such as surface energy and grain boundary migra-tion rate are not pressure sensitive (Saiki et al., 2003 and

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 271

references therein). Hence, the effects of pressure over therange of interest should be negligible.

Based on olivine compositions, redox conditions in theparent body of pallasites were estimated to range fromDIW �0.5 to +0.5 (Righter et al., 1990), similar to theDIW �0.9 to �0.6 of the experiments. For fO2 > IW, wüs-tite could stabilize as an additional phase, but was notobserved in pallasites. Instead, for moderate to high sul-phur activities, FeS or FeS2 is formed (Naldrett, 1969).Hence, phase stability is not much influenced by the smalldifference in the experimental and natural fO2. On the otherhand, oxygen fugacity plays a key role for interconnectionof the melt as explained below.

Estimated compositions of the metallic portion of terres-trial planets and planetesimals have 2–15 wt% S (Dreibusand Palme, 1996), much less than the Fe70S30 used in theexperiments. Furthermore, the cosmogenic Fe:Ni ratio is�19.5 (Grossmann, 1972) and pallasites have 3–8 wt% Ni(Buseck, 1977), leading to minor amounts of pentlanditeand pyrrhotite, and the formation of kamacite–taeniteand plessite in the crystallized metal fraction. At 1 atm pres-sure Ni depresses the eutectic temperature of the Fe–Ni–Seutectic to 6800 �C (Clark and Kullerud, 1963; Waldnerand Pelton, 2004) while that of the Fe–S binary is 990 �C(Brett and Bell, 1969; Usselman, 1975; Fei et al., 1997).The smaller S-content of planetesimals should reduce thefraction of molten Fe–(Ni)–S at a given temperature withrespect to our experiments, but Ni counterbalances thiseffect. Moreover, it has been suggested that S-loss duringmeteorite formation is a common process affecting ironsand stony-irons (Boesenberg et al., 2012).

Maximum temperature estimates for the pallasite forma-tion region are between 1100 and >1450 �C (Scott, 1977b;Ohtani, 1983; Saiki et al., 2003; Boesenberg et al., 2012).Cooling below the solidus of the Fe–Ni–S system may havetaken 106–108 years depending on the depth of pallasite for-mation and on the radius of the planetesimal (Fig. 7c). Atthe peak temperature to which pallasites were exposed,the Fe–Ni–S alloy may have been between fully molten orsolidified to more than 70% (Clark and Kullerud, 1963;Usselman, 1975; Urakawa et al., 1987; Waldner andPelton, 2004). Our experiments simulate the scenario whereolivine is surrounded by completely molten Fe–(Ni)–S, thusmaximising olivine growth rates. Ohtani (1983) suggestedthat moderate fractions of molten Fe–S would alreadygreatly enhance growth kinetics, since Fe–S melt wouldform a continuous film around olivine grains. In this case,diffusion would not be much enhanced by higher degreeof melting.

Whether the Fe–(Ni)–S melt forms a continuous filmdepends on its surface tension, which in turn determinesthe dihedral angle (DA) at the contacts of two grains andthe liquid. A critical value of DA = 60� separatesnon-wetting behaviour at higher angles from wetting beha-viour (von Bargen and Waff, 1986). For wetting liquids themelt fraction or connectivity threshold above which liquidsforms an interconnected network is almost zero (vonBargen and Waff, 1986). For Fe–(Ni)–S melts plus olivine,the interconnectivity threshold is debated and may liebetween 6 and 17.5 vol% (Yoshino et al., 2003;

Bagdassarov et al., 2009). Rose and Brenan (2001), andTerasaki et al. (2005) measured dihedral angles of Fe–Smelts and olivine >70� for highly reducing conditions,Gaetani and Grove (1999) observed that at 1 atm pressurean addition of �5 at% oxygen reduces the dihedral angleto

Fig. 7. Olivine grain growth and temperature evolution in the pallasite parent body based on 1D finite-difference conduction modelling(Tarduno et al., 2012). The time on the x-axis starts at the onset of cooling of the planetesimal. D is the depth of the olivine + Fe–S aggregatein the body. The starting grain size (d0) is 100 micrometres. The impact time is t = 0 for all curves except in (d) where impact times areindicated in the upper left corner. (a) and (b): Solid curves represent olivine grain size calculated employing Eq. (1), whereas dot-dash-dotcurves are olivine grain sizes calculated employing the grain growth law fit to the data of Guignard et al. (2012). Dashed curves indicate thatthe selected D lies within the megaregolith in the case of olivine grain size computed with Eq. (1) only. (c) Temperature evolution at variousdepths inside a 200 km body based on a 1D finite-difference conduction model; (d) Olivine grain size for a depth of the metal intrusionD = 40 km in a body with a 200 km radius, calculated with Eq. (1) considering different impact times, t.

272 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

and Guignard et al. (2012) calculated the grain size evolu-tion or the degree of rounding of olivine assuming a con-stant temperature, an approach that overestimates olivinegrowth and degree of rounding.

5.1. Model setup

As argued above, olivine growth becomes negligiblewhen temperature decreases to that of the Fe–Ni–S ternaryeutectic, i.e. 820 �C (Clark and Kullerud, 1963), since oli-vine coarsening in a solid metal matrix is orders of magni-tude slower than in molten metal.

Growth parameters employed in Eq. (1) are those com-puted for our experiments with an olivine/Fe–S volumeratio of 80/20, i.e.: n = 2.42, k0 = 9.43�106 lmn s�1, andEa = 289 kJ/mol. We also calculated the grain size evolu-tion of olivine fragments with a set of values computedby fitting the forsterite:Ni = 80:20 data of Guignard et al.(2012) using their reported grain size at t = 0 for the value

of d0. The least square fit to Eq. (1) yields n = 4.45,k0 = 9.10�107 lmn s�1, and Ea = 320 kJ/mol.

The starting grain size in the numerical model, i.e.100 lm, is an approximation derived from the smallestand least rounded olivine fragments observed in highlyfragmented pallasites (Buseck, 1977; Scott, 1977a) believedto represent those annealed for the shortest time. These pal-lasites are thought to originate from the shallowest depth inthe target body mantle preserving a fragment size distribu-tion representative for the onset of the annealing process.

The scenario of pallasite formation encompasses theimpact of a smaller body onto an already differentiatedone, followed by almost instantaneous vigorous mixing ofthe impactor metal with the now fragmented mantle ofthe target body. The mixed olivine + metal cools down tothe local mantle temperature over less than a few years(Tarduno et al., 2012), accompanied by almost no olivinegrowth. Afterwards, the temperature decrease follows thecooling rate of the local mantle. It should be noted that

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 273

in the 1D finite difference model employed for our calcula-tions, the cooling rate of the planetesimal is reduced by thepresence of a megaregolith layer (Tarduno et al., 2012).

In the impact scenario, annealing is effectively onlyachieved due to heat stored in the parent body. The dura-tion of annealing and the cooling rate depend on the bodyradius, the depth where mixed olivine + metal are located,and on the time of impact relative to planetesimal forma-tion (Fig. 7). The timing of the impact cannot be directlyconstrained (Tarduno et al., 2012 and references therein).Nevertheless, the impact happened at a time such that sub-sequent olivine grain growth of the fragments yields a finalgrain size in the range observed in rounded olivine-bearingpallasites, i.e. �1–20 mm (Buseck, 1977; Scott, 1977a andreferences therein), and the impact derived olivine + Fe–Smix must have been located at a depth which is consistentwith the metallographic cooling rates of the Fe–Ni portion.These are 2.5–18 �C/Ma, corresponding to the temperaturewindow of kamacite–taenite exsolution lamellae formation,i.e. �700–500 �C (Yang et al., 2010).

5.2. Model results

Fig. 7 shows the results of our calculations for twohypothetical parent bodies with radii of 200 and 600 km.The calculations start at the time corresponding to theonset of cooling of the parent body. Since olivine growthrates decrease exponentially with temperature, olivine grainsize curves flatten during cooling of the parent body and themaximum grain sizes that can be reached depend on thedepth of the olivine + Fe–S aggregate.

Employing the olivine grain growth law based on thedata of Guignard et al. (2012) yields a maximum grain size

274 G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275

Bhat B. N. (1973) Thermotransport in liquid Al–Cu alloys. NASATechnical Report R-415.

Boesenberg J. S., Delaney J. S. and Hewins R. H. (2012) Apetrological and chemical reexamination of Main Grouppallasite formation. Geochim. Cosmochim. Acta 89, 134–158.

Bohlen S. R., Essene E. J. and Boettcher A. J. (1980)Reinvestigations and applications of olivine–quartz–orthopy-roxene barometry. Earth Planet. Sci. Lett. 47, 1–10.

Bose K. and Ganguly J. (1995) Quartz-coesite revisited: reversedexperimental determinations at 500–1000 �C and retrievedthermochemical properties. Am. Mineral. 80, 231–238.

Brett R. and Bell P. M. (1969) Melting relations in the Fe-richportion of the system Fe–FeS at 30 kb pressure. Earth Planet.Sci. Lett. 6, 479–482.

Bryson J. F. J., Nichols C. I. O., Herrero-Albillos J., Kronast F.,Kasama T., Alimadadi H., van der Laan G., Nimmo F. andHarrison R. J. (2015) Long-lived magnetism from solidifica-tion-driven convection on the pallasite parent body. Nature 517,472–475.

Bunch T. E., Rumble, III, D., Wittke J. H. and Irving A. J. (2005)Pyroxene-rich pallasites Zinder and NWA 1911: not like theothers. Meteorit. Planet. Sci. 40, A26.

Buseck P. R. (1977) Pallasite meteorites – mineralogy, petrologyand geochemistry. Geochim. Cosmochim. Acta 41, 711–740.

Cabane H., Laporte D. and Provost A. (2005) An experimentalstudy of Ostwald ripening of olivine and plagioclase in silicatemelts: implications for the growth and size of crystals inmagma. Contrib. Mineral. Petrol. 150, 37–53.

Clark L. A. and Kullerud G. (1963) The sulphur-rich portion of theFe–Ni–S system. Econ. Geol. 58, 853–885.

Dasgupta R., Buono A., Whelan G. and Walker D. (2009) High-pressure melting relations in Fe–C–S systems: implications forformation, evolution, and structure of metallic cores in plan-etary bodies. Geochim. Cosmochim. Acta 73, 6678–6691.

Davis A. M. and Olsen E. J. (1991) Phosphates in pallasitemeteorites as probes of mantle processes in small planetarybodies. Nature 353, 637–640.

Dreibus G. and Palme H. (1996) Cosmochemical constraints on thesulfur content in the Earth’s core. Geochim. Cosmochim. Acta60, 1125–1130.

Eugster O., Herzog G. F., Marti K. and Caffee M. W. (2006)Irradiation records, cosmic-ray exposure ages, and transfertimes of meteorites. In Meteorites and the Early Solar System II(eds. D. S. Lauretta and , JrH. Y. McSween). University ofArizona Press, Tucson, pp. 829–851.

Evans B., Renner J. and Hirth G. (2001) A few remarks on thekinetics of static grain growth in rocks. Int. J. Earth Sci. 90, 88–103.

Fan D. and Chen L.-Q. (1997) Diffusion-controlled grain growth intwo-phase solids. Acta Mater. 45, 3297–3310.

Faul U. H. and Scott D. (2006) Grain growth in partially moltenolivine aggregates. Contrib. Mineral. Petrol. 151, 101–111.

Fei Y., Bertka C. M. and Finger L. (1997) High-pressure iron-sulfur compound, Fe3S2, and melting relations in the Fe–FeSsystem. Science 275, 1621–1623.

Gaetani G. A. and Grove T. L. (1999) Wetting of mantle olivine bysulfide melt: implications for Re/Os ratios in mantle peridotiteand late-stage core formation. Earth Planet. Sci. Lett. 169, 147–163.

Galoisy L., Calas G. and Brown, Jr., G. (1995) Intracrystallinedistribution of Ni in San Carlos olivine: an EXAFS study. Am.Mineral. 80, 1089–1092.

Gerald O. and Jaoul O. (1989) Oxygen diffusion in San Carlosolivine. J. Geophys. Res. 94, 4119–4128.

Grossmann G. W. (1972) Condensation in the primitive solarnebula. Geochim. Cosmochim. Acta 36, 597–619.

Guignard J., Bystricky M. and Toplis M. J. (2012) Grain growth inforsterite–nickel mixtures: analogues of small parent bodiesduring early accretion. Phys. Earth Planet. Inter. 204–205, 37–51.

Hanitzsch E. and Kalhweit M. (1969) Zur umlösung aufgedampftermetallkristalle. II. Z. Phys. Chem. Neue. Fol. 65, 290–305.

Hillert M. (1965) On the theory of normal and abnormal graingrowth. Acta Metall. 13, 227–238.

Holloway J. R., Pan V. and Gudmundsson G. (1992) High-pressure fluid-absent experiments in the presence of graphite:oxygen fugacity, ferric/ferrous ratio, and dissolved CO2. Eur. J.Mineral. 4, 105–114.

Houlier B., Jaoul O., Abel F. and Liebermann R. C. (1988) Oxygenand silicon self-diffusion in natural olivine at T = 1300 �C.Phys. Earth Planet. Inter. 50, 240–250.

Houlier B., Cheraghmakani M. and Jaoul O. (1990) Silicondiffusion in San Carlos olivine. Phys. Earth Planet. Inter. 62,329–340.

Hsu W. (2003) Minor element zoning and trace element geochem-istry of pallasites. Meteorit. Planet. Sci. 38, 1217–1240.

Johnson C. H., Richter S. K., Hamilton C. H. and Hoyt J. J. (1999)Static grain growth in a microduplex Ti-6Al-4V alloy. ActaMater. 47, 23–29.

Jones A. P., Tielens A. G. G. M. and Hollenbach D. J. (1996)Grain shattering in shocks: the interstellar grain size distribu-tion. Astrophys J 469, 740–764.

Karato S. (1989) Grain growth kinetics in olivine aggregates.Tectonophysics 168, 255–273.

Kong M., Bhattacharya R. N., James C. and Basu A. (2005) Astatistical approach to estimate the 3D size distribution ofspheres from 2D size distributions. Geol. Soc. Am. 117, 244–249.

Lifshitz I. M. and Slyozov V. V. (1961) The kinetics of precipi-tation from supersaturated solid solutions. J. Phys. Chem.Solids 19, 35–50.

Louat N. P. (1974) On the theory of normal grain growth. ActaMetall. 22, 721–724.

Lovering J. F. (1962) The evolution of the meteorites – evidence forthe co-existence of chondritic, achondritic and iron meteoritesin a typical meteorite parent body. In Researches on Meteorites(ed. C. B. Moore). Wiley, New York, pp. 179–197.

Malvin D. J., Wasson J. T., Clayton R. N., Mayeda T. K. andCurvello W. S. (1985) Bocaiuva – a silicate-inclusion bearingiron meteorites related to the Eagle station pallasites.Meteoritics 20, 257–272.

Martin J. W., Doherty R. D. and Cantor B. (1997) Stability ofMicrostructure in Metallic Systems, 2nd ed. CambridgeUniversity Press, Cambridge, England.

Mason B. (1963) The pallasites. Am. Mus. Novitiates, 2163.Mittlefehldt D. W. (1980) The composition of mesosiderite olivine

clasts and the implications for the origins of pallasites. EarthPlanet. Sci. Lett. 51, 29–40.

Naldrett A. J. (1969) A portion of the system Fe–S–O between 900and 1080 �C and its application to sulfide ore magmas. J.Petrol. 10, 171–201.

Nichols S. J. and Mackwell S. J. (1991) Grain growth in porousolivine aggregates. Phys. Chem. Miner. 18, 269–278.

Ohtani E. (1983) Formation of olivine textures in pallasites andthermal history of pallasites in their parent body. Phys. EarthPlanet. Inter. 32, 182–192.

Righter K., Arculus R. J., Delano J. W. and Paslick C. (1990)Electrochemical measurements and thermodynamic calcula-tions of redox equilibria in pallasite meteorites: implications forthe eucrite parent body. Geochim. Cosmochim. Acta 54, 1803–1815.

http://refhub.elsevier.com/S0016-7037(15)00226-4/h0040http://refhub.elsevier.com/S0016-7037(15)00226-4/h0040http://refhub.elsevier.com/S0016-7037(15)00226-4/h0040http://refhub.elsevier.com/S0016-7037(15)00226-4/h0045http://refhub.elsevier.com/S0016-7037(15)00226-4/h0045http://refhub.elsevier.com/S0016-7037(15)00226-4/h0045http://refhub.elsevier.com/S0016-7037(15)00226-4/h0050http://refhub.elsevier.com/S0016-7037(15)00226-4/h0050http://refhub.elsevier.com/S0016-7037(15)00226-4/h0050http://refhub.elsevier.com/S0016-7037(15)00226-4/h0050http://refhub.elsevier.com/S0016-7037(15)00226-4/h0055http://refhub.elsevier.com/S0016-7037(15)00226-4/h0055http://refhub.elsevier.com/S0016-7037(15)00226-4/h0055http://refhub.elsevier.com/S0016-7037(15)00226-4/h0055http://refhub.elsevier.com/S0016-7037(15)00226-4/h0060http://refhub.elsevier.com/S0016-7037(15)00226-4/h0060http://refhub.elsevier.com/S0016-7037(15)00226-4/h0060http://refhub.elsevier.com/S0016-7037(15)00226-4/h0060http://refhub.elsevier.com/S0016-7037(15)00226-4/h0060http://refhub.elsevier.com/S0016-7037(15)00226-4/h0065http://refhub.elsevier.com/S0016-7037(15)00226-4/h0065http://refhub.elsevier.com/S0016-7037(15)00226-4/h0065http://refhub.elsevier.com/S0016-7037(15)00226-4/h0065http://refhub.elsevier.com/S0016-7037(15)00226-4/h0070http://refhub.elsevier.com/S0016-7037(15)00226-4/h0070http://refhub.elsevier.com/S0016-7037(15)00226-4/h0075http://refhub.elsevier.com/S0016-7037(15)00226-4/h0075http://refhub.elsevier.com/S0016-7037(15)00226-4/h0075http://refhub.elsevier.com/S0016-7037(15)00226-4/h0075http://refhub.elsevier.com/S0016-7037(15)00226-4/h0080http://refhub.elsevier.com/S0016-7037(15)00226-4/h0080http://refhub.elsevier.com/S0016-7037(15)00226-4/h0085http://refhub.elsevier.com/S0016-7037(15)00226-4/h0085http://refhub.elsevier.com/S0016-7037(15)00226-4/h0085http://refhub.elsevier.com/S0016-7037(15)00226-4/h0085http://refhub.elsevier.com/S0016-7037(15)00226-4/h0090http://refhub.elsevier.com/S0016-7037(15)00226-4/h0090http://refhub.elsevier.com/S0016-7037(15)00226-4/h0090http://refhub.elsevier.com/S0016-7037(15)00226-4/h0095http://refhub.elsevier.com/S0016-7037(15)00226-4/h0095http://refhub.elsevier.com/S0016-7037(15)00226-4/h0095http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0100http://refhub.elsevier.com/S0016-7037(15)00226-4/h0105http://refhub.elsevier.com/S0016-7037(15)00226-4/h0105http://refhub.elsevier.com/S0016-7037(15)00226-4/h0105http://refhub.elsevier.com/S0016-7037(15)00226-4/h0110http://refhub.elsevier.com/S0016-7037(15)00226-4/h0110http://refhub.elsevier.com/S0016-7037(15)00226-4/h0115http://refhub.elsevier.com/S0016-7037(15)00226-4/h0115http://refhub.elsevier.com/S0016-7037(15)00226-4/h0120http://refhub.elsevier.com/S0016-7037(15)00226-4/h0120http://refhub.elsevier.com/S0016-7037(15)00226-4/h0120http://refhub.elsevier.com/S0016-7037(15)00226-4/h0120http://refhub.elsevier.com/S0016-7037(15)00226-4/h0120http://refhub.elsevier.com/S0016-7037(15)00226-4/h0125http://refhub.elsevier.com/S0016-7037(15)00226-4/h0125http://refhub.elsevier.com/S0016-7037(15)00226-4/h0125http://refhub.elsevier.com/S0016-7037(15)00226-4/h0125http://refhub.elsevier.com/S0016-7037(15)00226-4/h0130http://refhub.elsevier.com/S0016-7037(15)00226-4/h0130http://refhub.elsevier.com/S0016-7037(15)00226-4/h0130http://refhub.elsevier.com/S0016-7037(15)00226-4/h0130http://refhub.elsevier.com/S0016-7037(15)00226-4/h0135http://refhub.elsevier.com/S0016-7037(15)00226-4/h0135http://refhub.elsevier.com/S0016-7037(15)00226-4/h0140http://refhub.elsevier.com/S0016-7037(15)00226-4/h0140http://refhub.elsevier.com/S0016-7037(15)00226-4/h0145http://refhub.elsevier.com/S0016-7037(15)00226-4/h0145http://refhub.elsevier.com/S0016-7037(15)00226-4/h0145http://refhub.elsevier.com/S0016-7037(15)00226-4/h0145http://refhub.elsevier.com/S0016-7037(15)00226-4/h0150http://refhub.elsevier.com/S0016-7037(15)00226-4/h0150http://refhub.elsevier.com/S0016-7037(15)00226-4/h0155http://refhub.elsevier.com/S0016-7037(15)00226-4/h0155http://refhub.elsevier.com/S0016-7037(15)00226-4/h0160http://refhub.elsevier.com/S0016-7037(15)00226-4/h0160http://refhub.elsevier.com/S0016-7037(15)00226-4/h0160http://refhub.elsevier.com/S0016-7037(15)00226-4/h0160http://refhub.elsevier.com/S0016-7037(15)00226-4/h0160http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0165http://refhub.elsevier.com/S0016-7037(15)00226-4/h0170http://refhub.elsevier.com/S0016-7037(15)00226-4/h0170http://refhub.elsevier.com/S0016-7037(15)00226-4/h0170http://refhub.elsevier.com/S0016-7037(15)00226-4/h0175http://refhub.elsevier.com/S0016-7037(15)00226-4/h0175http://refhub.elsevier.com/S0016-7037(15)00226-4/h0180http://refhub.elsevier.com/S0016-7037(15)00226-4/h0180http://refhub.elsevier.com/S0016-7037(15)00226-4/h0180http://refhub.elsevier.com/S0016-7037(15)00226-4/h0185http://refhub.elsevier.com/S0016-7037(15)00226-4/h0185http://refhub.elsevier.com/S0016-7037(15)00226-4/h0185http://refhub.elsevier.com/S0016-7037(15)00226-4/h0190http://refhub.elsevier.com/S0016-7037(15)00226-4/h0190http://refhub.elsevier.com/S0016-7037(15)00226-4/h0195http://refhub.elsevier.com/S0016-7037(15)00226-4/h0195http://refhub.elsevier.com/S0016-7037(15)00226-4/h0195http://refhub.elsevier.com/S0016-7037(15)00226-4/h0195http://refhub.elsevier.com/S0016-7037(15)00226-4/h0200http://refhub.elsevier.com/S0016-7037(15)00226-4/h0200http://refhub.elsevier.com/S0016-7037(15)00226-4/h0200http://refhub.elsevier.com/S0016-7037(15)00226-4/h0205http://refhub.elsevier.com/S0016-7037(15)00226-4/h0205http://refhub.elsevier.com/S0016-7037(15)00226-4/h0210http://refhub.elsevier.com/S0016-7037(15)00226-4/h0210http://refhub.elsevier.com/S0016-7037(15)00226-4/h0210http://refhub.elsevier.com/S0016-7037(15)00226-4/h0210http://refhub.elsevier.com/S0016-7037(15)00226-4/h0215http://refhub.elsevier.com/S0016-7037(15)00226-4/h0215http://refhub.elsevier.com/S0016-7037(15)00226-4/h0215http://refhub.elsevier.com/S0016-7037(15)00226-4/h0215http://refhub.elsevier.com/S0016-7037(15)00226-4/h0220http://refhub.elsevier.com/S0016-7037(15)00226-4/h0220http://refhub.elsevier.com/S0016-7037(15)00226-4/h0220http://refhub.elsevier.com/S0016-7037(15)00226-4/h0225http://refhub.elsevier.com/S0016-7037(15)00226-4/h0230http://refhub.elsevier.com/S0016-7037(15)00226-4/h0230http://refhub.elsevier.com/S0016-7037(15)00226-4/h0230http://refhub.elsevier.com/S0016-7037(15)00226-4/h0235http://refhub.elsevier.com/S0016-7037(15)00226-4/h0235http://refhub.elsevier.com/S0016-7037(15)00226-4/h0235http://refhub.elsevier.com/S0016-7037(15)00226-4/h0235http://refhub.elsevier.com/S0016-7037(15)00226-4/h0240http://refhub.elsevier.com/S0016-7037(15)00226-4/h0240http://refhub.elsevier.com/S0016-7037(15)00226-4/h0245http://refhub.elsevier.com/S0016-7037(15)00226-4/h0245http://refhub.elsevier.com/S0016-7037(15)00226-4/h0245http://refhub.elsevier.com/S0016-7037(15)00226-4/h0250http://refhub.elsevier.com/S0016-7037(15)00226-4/h0250http://refhub.elsevier.com/S0016-7037(15)00226-4/h0250http://refhub.elsevier.com/S0016-7037(15)00226-4/h0250http://refhub.elsevier.com/S0016-7037(15)00226-4/h0250

G.F.D. Solferino et al. / Geochimica et Cosmochimica Acta 162 (2015) 259–275 275

Ringwood A. E. (1961) Chemical and genetic relationships amongmeteorites. Geochim. Cosmochim. Acta 24, 159–197.

Rose L. A. and Brenan J. M. (2001) Wetting properties Fe–Cu–Ni–Co–O–S melts against olivine: implications for sulfide meltmobility. Econ. Geol. 46, 145–157.

Saiki K., Laporte D., Vielzeuf D., Nakashima S. and Boivin P.(2003) Morphological analysis of olivine grains annealed in aniron–nickel matrix: experimental constraints on the origin ofpallasites and on the thermal history of their parent bodies.Meteorit. Planet. Sci. 38, 427–444.

Schmidt M. W., Forien M., Solferino G. and Bagdassarov N.(2012) Settling and compaction of olivine in basaltic magmas:an experimental study on the time scales of cumulate formation.Contrib. Mineral. Petrol. 164, 959–976.

Scott E. R. D. (1977a) Pallasites–metal composition, classificationand relationships with iron meteorites. Geochim. Cosmochim.Acta 41, 349–360.

Scott E. R. D. (1977b) Formation of olivine–metal textures inpallasite meteorites. Geochim. Cosmochim. Acta 41, 693–710.

Scott E. R. D. (1977c) Geochemical relationships between somepallasites and iron meteorites. Mineral. Mag. 41, 265–272.

Scott E. R. D. (2007) Impact origins for pallasites. Lunar Planet.Sci. XXXVIII. Lunar Planet. Inst., Houston. #2284 (abstr.).

Scott E. R. D. and Taylor G. J. (1990) Origins of pallasites at thecore–mantle boundaries of asteroids. Lunar Planet. Sci. XXI.Lunar Planet. Inst., Houston. pp. 1119–1120 (abstr.).

Scott E. R. D., Greenwood R. C., Franchi I. A. and Sanders I. S.(2009) Oxygen isotopic constraints on the origin and parentbodies of eucrites, diogenites, and howardites. Geochim.Cosmochim. Acta 73, 5835–5853.

Tarduno J. A., Cottrell R. D., Nimmo F., Hopkins J., Voronov J.,Erickson A., Blackman E., Scott E. R. D. and McKinley R.(2012) Evidence for a dynamo in the Main Group pallasiteparent body. Science 338, 939–942.

Terasaki H., Frost D. J., Rubie D. C. and Langenhorst F. (2005)The effect of oxygen and sulphur on the dihedral angle betweenFe–O–S melt and silicate minerals at high pressure: implicationsfor Martian core formation. Earth Planet. Sci. Lett. 232, 379–392.

Ulff-Møller F., Choi B.-G., Rubin A. E., Tran J. and Wasson J. T.(1998) Effects of liquid immiscibility on trace element fraction-ation in magmatic iron meteorites: a case study of group IIIAB.Meteorit. Planet. Sci. 33, 207–220.

Urakawa S., Kato M. and Kumazawa M. (1987) Experimentalstudy of the phase relation in the system Fe–Ni–O–S up to15 GPa. In High-Pressure Research in Mineral Physics (eds. M.H. Manghnani and Y. Syono). Terra Sci., Tokyo, pp. 95–111.

Usselman T. M. (1975) Experimental approach to the state of thecore: Part 1. The liquidus relations of the Fe-rich portion of theFe–Ni–S from 30 to 100 kb. Am. J. Sci. 275, 278–290.

von Bargen N. and Waff H. S. (1986) Permeabilities, interfacialareas and curvatures of partially molten systems: results ofnumerical computations of equilibrium microstructures. J.Geophys. Res. 91, 9261–9276.

Waddell H. (1932) Volume, shape, and roundness of rock particles.J. Geol. 40, 443–451.

Wagner C. (1961) Theorie der alterung von niederschlägen durchumlösen (Ostwald-reifung). Z. Elektrochem. 65, 581–591.

Waldner P. and Pelton A. D. (2004) Critical thermodynamicassessment and modelling of the Fe–Ni–S system. Metall.Mater. Trans. 35B, 897–907.

Wang C., Hirama J., Nagasaka T. and Ban-Ya S. (1991) Phaseequilibria of liquid Fe–S–C ternary system. ISIJ Int. 31, 1292–1299.

Wasson J. T. and Choi B.-G. (2003) Main-group pallasites:chemical composition, relationship to IIIAB irons, and origin.Geochim. Cosmochim. Acta 67, 3079–3096.

Yang J., Goldstein J. I. and Scott E. R. D. (2010) Main-grouppallasites: thermal history, relationship to IIIAB irons, andorigin. Geochim. Cosmochim. Acta 74, 4471–4492.

Yoshino T. and Watson E. B. (2005) Growth kinetics of FeS meltin partially molten peridotite: an analog for core-formingprocesses. Earth Planet. Sci. Lett. 235, 453–468.

Yoshino T., Walter M. J. and Katsura T. (2003) Core formation inplanetesimals triggered by permeable flow. Nature 422, 154–157.

Associate editor: Rajdeep Dasgupta

http://refhub.elsevier.com/S0016-7037(15)00226-4/h0255http://refhub.elsevier.com/S0016-7037(15)00226-4/h0255http://refhub.elsevier.com/S0016-7037(15)00226-4/h0260http://refhub.elsevier.com/S0016-7037(15)00226-4/h0260http://refhub.elsevier.com/S0016-7037(15)00226-4/h0260http://refhub.elsevier.com/S0016-7037(15)00226-4/h0265http://refhub.elsevier.com/S0016-7037(15)00226-4/h0265http://refhub.elsevier.com/S0016-7037(15)00226-4/h0265http://refhub.elsevier.com/S0016-7037(15)00226-4/h0265http://refhub.elsevier.com/S0016-7037(15)00226-4/h0265http://refhub.elsevier.com/S0016-7037(15)00226-4/h0270http://refhub.elsevier.com/S0016-7037(15)00226-4/h0270http://refhub.elsevier.com/S0016-7037(15)00226-4/h0270http://refhub.elsevier.com/S0016-7037(15)00226-4/h0270http://refhub.elsevier.com/S0016-7037(15)00226-4/h0275http://refhub.elsevier.com/S0016-7037(15)00226-4/h0275http://refhub.elsevier.com/S0016-7037(15)00226-4/h0275http://refhub.elsevier.com/S0016-7037(15)00226-4/h0280http://refhub.elsevier.com/S0016-7037(15)00226-4/h0280http://refhub.elsevier.com/S0016-7037(15)00226-4/h0285http://refhub.elsevier.com/S0016-7037(15)00226-4/h0285http://refhub.elsevier.com/S0016-7037(15)00226-4/h0300http://refhub.elsevier.com/S0016-7037(15)00226-4/h0300http://refhub.elsevier.com/S0016-7037(15)00226-4/h0300http://refhub.elsevier.com/S0016-7037(15)00226-4/h0300http://refhub.elsevier.com/S0016-7037(15)00226-4/h0305http://refhub.elsevier.com/S0016-7037(15)00226-4/h0305http://refhub.elsevier.com/S0016-7037(15)00226-4/h0305http://refhub.elsevier.com/S0016-7037(15)00226-4/h0305http://refhub.elsevier.com/S0016-7037(15)00226-4/h0310http://refhub.elsevier.com/S0016-7037(15)00226-4/h0310http://refhub.elsevier.com/S0016-7037(15)00226-4/h0310http://refhub.elsevier.com/S0016-7037(15)00226-4/h0310http://refhub.elsevier.com/S0016-7037(15)00226-4/h0310http://refhub.elsevier.com/S0016-7037(15)00226-4/h0315http://refhub.elsevier.com/S0016-7037(15)00226-4/h0315http://refhub.elsevier.com/S0016-7037(15)00226-4/h0315http://refhub.elsevier.com/S0016-7037(15)00226-4/h0315http://refhub.elsevier.com/S0016-7037(15)00226-4/h0320http://refhub.elsevier.com/S0016-7037(15)00226-4/h0320http://refhub.elsevier.com/S0016-7037(15)00226-4/h0320http://refhub.elsevier.com/S0016-7037(15)00226-4/h0320http://refhub.elsevier.com/S0016-7037(15)00226-4/h0320http://refhub.elsevier.com/S0016-7037(15)00226-4/h0325http://refhub.elsevier.com/S0016-7037(15)00226-4/h0325http://refhub.elsevier.com/S0016-7037(15)00226-4/h0325http://refhub.elsevier.com/S0016-7037(15)00226-4/h0325http://refhub.elsevier.com/S0016-7037(15)00226-4/h0330http://refhub.elsevier.com/S0016-7037(15)00226-4/h0330http://refhub.elsevier.com/S0016-7037(15)00226-4/h0330http://refhub.elsevier.com/S0016-7037(15)00226-4/h0330http://refhub.elsevier.com/S0016-7037(15)00226-4/h0335http://refhub.elsevier.com/S0016-7037(15)00226-4/h0335http://refhub.elsevier.com/S0016-7037(15)00226-4/h0340http://refhub.elsevier.com/S0016-7037(15)00226-4/h0340http://refhub.elsevier.com/S0016-7037(15)00226-4/h0345http://refhub.elsevier.com/S0016-7037(15)00226-4/h0345http://refhub.elsevier.com/S0016-7037(15)00226-4/h0345http://refhub.elsevier.com/S0016-7037(15)00226-4/h0350http://refhub.elsevier.com/S0016-7037(15)00226-4/h0350http://refhub.elsevier.com/S0016-7037(15)00226-4/h0350http://refhub.elsevier.com/S0016-7037(15)00226-4/h0355http://refhub.elsevier.com/S0016-7037(15)00226-4/h0355http://refhub.elsevier.com/S0016-7037(15)00226-4/h0355http://refhub.elsevier.com/S0016-7037(15)00226-4/h0360http://refhub.elsevier.com/S0016-7037(15)00226-4/h0360http://refhub.elsevier.com/S0016-7037(15)00226-4/h0360http://refhub.elsevier.com/S0016-7037(15)00226-4/h0365http://refhub.elsevier.com/S0016-7037(15)00226-4/h0365http://refhub.elsevier.com/S0016-7037(15)00226-4/h0365http://refhub.elsevier.com/S0016-7037(15)00226-4/h0370http://refhub.elsevier.com/S0016-7037(15)00226-4/h0370http://refhub.elsevier.com/S0016-7037(15)00226-4/h0370

Fast grain growth of olivine in liquid Fe–S and the formation of pallasites with rounded olivine grains1 Introduction2 Methods2.1 Experimental technique2.2 Compositional analysis of olivine and Fe–S2.3 Image analyses2.4 Some background on grain growth

3 Results3.1 Composition of olivine and Fe–S melt3.2 Sample textures3.3 Olivine grain growth3.4 Normalised grain size distributions (GSD) of experimental and natural samples

4 Discussion4.1 Chemical equilibration of olivine and Fe–S4.2 Grain growth of olivine in molten Fe–S4.2.1 Comparison with previous studies4.2.2 Comparison with natural pallasites containing rounded olivines4.2.3 Rounding of olivine fragments

4.3 Olivine grain growth in pallasites4.4 A comparison of boundary conditions of the experiments and pallasite formation4.5 Formation of mixed-type pallasites

5 Modelling olivine grain growth in a planetesimal interior5.1 Model setup5.2 Model results

6 Final remarksAcknowledgementsReferences