Iron isotope fractionation during planetary differentiation

14
Iron isotope fractionation during planetary differentiation Stefan Weyer a, * , Ariel D. Anbar b , Gerhard P. Brey a , Carsten Mu ¨ nker c , Klaus Mezger c , Alan B. Woodland a a Institut fu ¨ r Mineralogie, Universita ¨t Frankfurt, Germany b Department of Geological Sciences and Department of Chemistry and Biochemistry, Arizona State University, Tempe, AZ, USA c Institut fu ¨ r Mineralogie, Universita ¨t Mu ¨nster, Germany Received 31 December 2004; received in revised form 6 July 2005; accepted 19 September 2005 Available online 25 October 2005 Editor: E. Boyle Abstract The Fe isotope composition of samples from the Moon, Mars (SNC meteorites), HED parent body (eucrites), pallasites (metal and silicate) and the Earth’s mantle were measured using high mass resolution MC-ICP-MS. These high precision measurements (d 56 Fe c F0.04x, 2 S.D.) place tight constraints on Fe isotope fractionation during planetary differentiation. Fractionation during planetary core formation is confined to b0.1x for d 56 Fe by the indistinguishable Fe isotope composition of pallasite bulk metal (including sulfides and phosphides) and olivine separates. However, large isotopic variations (c0.5x) were observed among pallasite metal separates, varying systematically with the amounts of troilite, schreibersite, kamacite and taenite. Troilite generally has the lightest (d 56 Fe c 0.25x) and schreibersite the heaviest (d 56 Fe c +0.2x) Fe isotope composition. Taenite is heavier then kamacite. Therefore, these variations probably reflect Fe isotope fractionation during the late stage evolution and differentiation of the S- and P-rich metal melts, and during low-temperature kamacite exsolution, rather than fractionation during silicate–metal separation. Differentiation of the silicate portion of planets also seems to fractionate Fe isotopes. Notably, magmatic rocks (partial melts) are systematically isotopically heavier than their mantle protoliths. This is indicated by the mean of 11 terrestrial peridotite samples from different tectonic settings (d 56 Fe = +0.015 F 0.018x), which is significantly lighter than the mean of terrestrial basalts (d 56 Fe = +0.076 F 0.029x). We consider the peridotite mean to be the best estimate for the Fe isotope composition of the bulk silicate Earth, and probably also of bulk Earth. The terrestrial basaltic mean is in good agreement with the mean of the lunar samples (d 56 Fe = +0.073 F 0.019x), excluding the high-Ti basalts. The high-Ti basalts display the heaviest Fe isotope composition of all rocks measured here (d 56 Fe c +0.2x). This is interpreted as a fingerprint of the lunar magma ocean, which produced a very heterogeneous mantle, including the ilmenite-rich source regions of these basalts. Within uncertainties, samples from Mars (SNC meteorites), HED (eucrites) and the pallasites (average olivine + metal) have the same Fe isotope compositions as the Earth’s mantle. This indicates that the solar system is very homogeneous in Fe isotopes. Its average d 56 Fe is very close to that of the IRMM-014 standard. D 2005 Elsevier B.V. All rights reserved. Keywords: iron isotopes; solar system; terrestrial planets; Moon; core formation; magma ocean 1. Introduction Iron is by far not only the most abundant element in planetary cores, but is also a major element in the silicate 0012-821X/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2005.09.023 * Corresponding author. Fax: +49 69 79828066. E-mail address: [email protected] (S. Weyer). Earth and Planetary Science Letters 240 (2005) 251 – 264 www.elsevier.com/locate/epsl

Transcript of Iron isotope fractionation during planetary differentiation

www.elsevier.com/locate/epsl

Earth and Planetary Science L

Iron isotope fractionation during planetary differentiation

Stefan Weyer a,*, Ariel D. Anbar b, Gerhard P. Brey a, Carsten Munker c,

Klaus Mezger c, Alan B. Woodland a

a Institut fur Mineralogie, Universitat Frankfurt, Germanyb Department of Geological Sciences and Department of Chemistry and Biochemistry, Arizona State University, Tempe, AZ, USA

c Institut fur Mineralogie, Universitat Munster, Germany

Received 31 December 2004; received in revised form 6 July 2005; accepted 19 September 2005

Available online 25 October 2005

Editor: E. Boyle

Abstract

The Fe isotope composition of samples from the Moon, Mars (SNC meteorites), HED parent body (eucrites), pallasites (metal

and silicate) and the Earth’s mantle were measured using high mass resolution MC-ICP-MS. These high precision measurements

(d56FecF0.04x, 2 S.D.) place tight constraints on Fe isotope fractionation during planetary differentiation.

Fractionation during planetary core formation is confined to b0.1x for d56Fe by the indistinguishable Fe isotope composition

of pallasite bulk metal (including sulfides and phosphides) and olivine separates. However, large isotopic variations (c0.5x) were

observed among pallasite metal separates, varying systematically with the amounts of troilite, schreibersite, kamacite and taenite.

Troilite generally has the lightest (d56Fec�0.25x) and schreibersite the heaviest (d56Fec+0.2x) Fe isotope composition.

Taenite is heavier then kamacite. Therefore, these variations probably reflect Fe isotope fractionation during the late stage evolution

and differentiation of the S- and P-rich metal melts, and during low-temperature kamacite exsolution, rather than fractionation

during silicate–metal separation.

Differentiation of the silicate portion of planets also seems to fractionate Fe isotopes. Notably, magmatic rocks (partial melts)

are systematically isotopically heavier than their mantle protoliths. This is indicated by the mean of 11 terrestrial peridotite samples

from different tectonic settings (d56Fe=+0.015F0.018x), which is significantly lighter than the mean of terrestrial basalts

(d56Fe=+0.076F0.029x). We consider the peridotite mean to be the best estimate for the Fe isotope composition of the bulk

silicate Earth, and probably also of bulk Earth. The terrestrial basaltic mean is in good agreement with the mean of the lunar

samples (d56Fe=+0.073F0.019x), excluding the high-Ti basalts. The high-Ti basalts display the heaviest Fe isotope composition

of all rocks measured here (d56Fec+0.2x). This is interpreted as a fingerprint of the lunar magma ocean, which produced a very

heterogeneous mantle, including the ilmenite-rich source regions of these basalts.

Within uncertainties, samples from Mars (SNC meteorites), HED (eucrites) and the pallasites (average olivine+metal) have the

same Fe isotope compositions as the Earth’s mantle. This indicates that the solar system is very homogeneous in Fe isotopes. Its

average d56Fe is very close to that of the IRMM-014 standard.

D 2005 Elsevier B.V. All rights reserved.

Keywords: iron isotopes; solar system; terrestrial planets; Moon; core formation; magma ocean

0012-821X/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.epsl.2005.09.023

* Corresponding author. Fax: +49 69 79828066.

E-mail address: [email protected] (S. Weyer).

1. Introduction

Iron is by far not only the most abundant element in

planetary cores, but is also a major element in the silicate

etters 240 (2005) 251–264

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264252

portions of planets. Therefore, Fe stable isotope fraction-

ation between planetary reservoirs has the potential to

monitor processes associated with (1) accretion history,

(2) core formation and (3) differentiation of the silicate

reservoirs of planets. However, Fe-isotope variations in

high-temperature rocks are small. Most terrestrial igne-

ous rocks exhibit a range in d56Fe of only 0.3x [1]. Such

small variations were not detectable in the first system-

atic studies of high-temperature rocks [2]. These early

studies used thermal ionization mass spectrometry

(TIMS) and an isotope double spike. With multi-collec-

tor inductively coupled plasma source mass spectrome-

try (MC-ICP-MS) precision was significantly improved

to about 0.1x (2 S.D.) for d56Fe or d57Fe [1,3–5]. Using

this technique, recent studies revealed differences be-

tween mineral phases of high-temperature rocks and

between different planetary geochemical reservoirs [5–

11]. These recent studies motivated the present work, in

which precision was further improved by using high

mass resolution MC-ICP-MS and an internal Cu stan-

dard for mass bias correction [4,12].

Of the planetary bodies from which samples are

available, the accretion history of the Moon is of par-

ticular interest because it is thought to have formed as

the result of a giant impact of a Mars size planet onto

the proto Earth [13–17]. In a previous study, Poitrasson

et al. [5] reported a heavier Fe isotope composition for

the bulk Moon and the Earth compared to other plan-

etary bodies, which they interpreted as resulting from

evaporation of iron during the giant impact event.

However, to compare lunar silicate reservoirs with

those of the Earth and other planets it is important to

know whether Fe isotope fractionation occurred during

silicate differentiation on the Moon. To investigate this

question, we analyzed samples from representative

lunar crustal reservoirs.

Another key question is the extent of Fe isotope

fractionation during planetary core formation. This in-

formation is crucial to the interpretation of any Fe

isotope data from the silicate portions of planets,

since most of the total planetary Fe resides in metal

cores that are not accessible for sampling. The silicate

mantles of the planets might have variable Fe isotope

compositions because of the variable core sizes of

planets. Only if the behaviour of Fe isotopes during

core formation is known can d56Fe compositions of

bulk planetary bodies be estimated from silicates. To

address this question, we analyzed silicate and metal

samples from nine pallasites.

The Fe isotope composition of the bulk silicate Earth

is probably best represented by the Fe isotope compo-

sition of samples directly from the mantle because the

mantle contains c99% of the total Fe of the silicate

Earth. Previous studies used both mantle rocks and

differentiated igneous rocks to define the bterrestrialigneous baselineQ [1,5,18]. However, the value for the

silicate Earth determined by these studies is dominated

by rocks generated by partial melting in the mantle

(further referred as magmatic rocks). The values for

the few peridotites measured in these studies agree

within uncertainties with the values for magmatic

rocks. However, the reported uncertainty for individual

or small groups of samples was about 0.1x [1,5],

which is relatively large when compared to the small

differences in Fe isotope composition (b0.2x for

d56Fe) reported between planetary reservoirs [5,8].

Therefore, we decided to use a representative selection

of peridotites from different tectonic settings to estimate

the Fe isotope composition of the silicate Earth and to

determine the extent of modification of Fe isotopes

during partial melting.

2. Samples and analytical techniques

Eight pallasites, 15 lunar rocks, 7 SNC meteorites, 4

eucrites, 11 terrestrial peridotites and 9 international

and in-house standards (including 2 peridotites, 1

komatiite, 4 basalts, 1 hematite and 1 iron meteorite)

were measured for their Fe isotope composition. The

lunar samples were from NASA-Apollo sample return

missions, the SNC meteorites are generally accepted to

originate from the planet Mars [19–21] and the eucrites

are from the HED parent body and may represent

samples from Asteroid 4 Vesta [22,23]. All planetary

samples were aliquots from a N100-mg amount of

powder that was already used for other studies

[17,24]. The terrestrial mantle samples are all bulk

peridotites or olivine separates from peridotites cover-

ing various tectonic settings (xenoliths and orogenic,

garnet and spinel peridotites), including samples from

previous studies [25–29] and also international stan-

dards (JP-1, and PCC-1).

The pallasite samples were provided by different

meteorite collections, namely the Natural History Mu-

seum of London, the Natural History Museum of

Vienna and the Ward collection at the University of

Rochester. From 2 pallasites (Brenham and Mount

Vernon) material from different collections was mea-

sured. Both the silicate and metal separates were ana-

lyzed in this study. The mineral separates consist of a

few (1–10) handpicked clean olivine grains or metal

grains, respectively. The latter also include nonmetallic

phases, such as troilite and schreibersite and have var-

iable amounts of kamacite and taenite.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 253

For Fe isotope analyses, 1–100 mg of each silicate

sample was digested with a HF–HNO3 mixture on a hot

plate. The pallasite metal separates were digested with

conc. HNO3 on a hot plate. All samples were finally

reconstituted in 7.5 M HCl for chemical purification.

Chromatographic separation of Fe from the matrix was

done following the procedure described by [12,30]. The

total amount of Fe loaded on the anion exchange col-

umns (Biorad 2 ml Columns, Biorad AG 1X8 resin)

was 0.2–5 mg.

Iron isotope compositions were measured using MC-

ICP-MS. There are two critical issues limiting the

precision of such analyses: (1) polyatomic ions40Ar14N+, 40Ar16O+ and 40Ar16OH+ interfere with the

isobaric iron isotopes 54Fe+, 56Fe+ and 57Fe+, respec-

tively; and (2) instrument mass bias is much larger

than natural fractionations. To cope with interfer-

ences, most previous studies used desolvating nebu-

lisation, e.g., [3,6,30], collision cell, e.g., [1] or cold

plasma techniques [31] to suppress rather than com-

pletely eliminate the mass interferences. Mass bias

correction was usually done by sample-standard

bracketing (e.g., Belshaw). To improve the analytical

precision relative to these studies, we applied a high

mass resolution technique [4,12,32] to eliminate iso-

baric interferences. Precision was further improved by

addition of a Cu standard solution (NIST 976) to

samples and standards as an internal monitor of

mass bias [12,30].

All measurements, including Fe isotope and Fe, Ni,

P and S concentration measurements on the pallasite

metals, were performed with a Finnigan Neptune MC-

ICP-MS at the Universitat Frankfurt operated in high

mass resolution mode [4,12]. In this mode, the resolv-

ing power on all collectors is sufficient to completely

Table 1

Testing matrix effects on Fe isotope measurements

Sample name Material SSB technique

d56Fe

(x)

Uncertainty

(2 S.D.)

d57Fe U

(2

cpx doped with

IRMM-014

cpx 0.010 0.121 0.007 0

Olivine doped with

IRMM-014

Olivine �0.010 0.135 �0.013 0

Brenham-V Metal

Purified aliquot 1 0.089 0.091 0.116 0

Purified aliquot 2 0.070 0.079 0.104 0

Without purification �0.085 0.159 �0.124 0

Mount-Vernon-V Metal, mainly

Purified aliquot 1 Schreibersite 0.187 0.108 0.267 0

Purified aliquot 2 0.170 0.097 0.242 0

Without purification �0.076 0.045 �0.098 0

eliminate all polyatomic mass interferences (40Ar14N,40Ar16O and 40Ar16OH) and to produce flat top peak

sections that are necessary for high precision measure-

ments. A tandem quartz glass spray chamber (Finni-

gan SIS) combined with an ESI teflon microflow (50

Al) nebuliser was used for sample introduction. All

Faraday collectors were equipped with 1011 V resis-

tors. With this set up a 20–25 V (2–2.5*10�10 A)

signal was usually achieved for a 5-ppm Fe solution

(on 56Fe).

Measurements are reported as d(56Fe/54Fe) (referredto as d56Fe) in x relative to the IRMM-014 standard

(Tables 1 and 2). d(57Fe/54Fe) (referred to as d57Fe) was

measured for quality control (Table 1) and to monitor

for mass independent analytical artifacts. Natural mass

independent fractionation was also of interest for extra-

terrestrial samples. Neither artificial nor natural mass

independent effects were observed.

Instrumental mass bias was corrected using65Cu/63Cu (NIST 976). We assumed 65Cu/63Cu=

0.44563 and the exponential law [33]. It should be

stressed that for d determinations neither assumption

is critical (e.g., [12]). The suitability of Cu to correct for

instrumental mass bias on Fe is demonstrated by the

excellent correlation of the Cu and Fe fractionation

factors (b) in Fig. 1. The regression lines defined by

the standard measurements of each analytical session

and also by replicate sample measurements during end

session (Canon Diablo) have high values for R2 (N0.99)

and very similar slopes of c0.91. The d values of the

mass bias corrected sample measurements were calcu-

lated relative to the mean of the adjacent standard

measurements. With this technique, combining sam-

ple-standard bracketing (further referred as SSB) and

the Cu mass bias correction (further referred as

Cu+SSB technique Number of

analysesncertainty

S.D.)

d56Fe

(x)

Uncertainty

(2 S.D.)

d57Fe Uncertainty

(2 S.D.)

.215 0.005 0.036 �0.005 0.085 15

.220 0.004 0.044 0.008 0.083 15

.120 0.080 0.061 0.115 0.028 5

.083 0.058 0.037 0.109 0.119 6

.226 0.068 0.004 0.114 0.042 4

.193 0.160 0.056 0.244 0.037 5

.144 0.156 0.045 0.215 0.054 6

.075 0.165 0.063 0.277 0.141 4

Table 2

Fe isotope composition of bulk rock samples and mineral separates

Sample name Material Mass

(mg)

d56Fe

(x)

Uncertainty Number of

analyses

Reference

values(2 S.D.) (2 S.E.)

In-house standards

Canon Diablo Iron meteorite 0.191 0.041 0.009 21

Hematitea Hematite 0.530 0.024 0.014 5 0.567a

International basalt standards

BIR-1 (1) Basalt 0.050 0.039 0.022 5 0.053a, 0.068b

BIR-1 (2) 0.060 0.041 0.028 4

BCR-2 Basalt 0.079 0.047 0.043 3 0.073b, 0.00c

BHVO-1 Basalt 0.117 0.030 0.028 3 0.109b

JB-2 Basalt 0.056 0.033 0.030 3

KAL-1 Komatiite 0.071 0.022 0.015 4

0.076 0.051 0.029 5

Earth mantle

JP-1 Bulk peridotite 13.4 0.003 0.050 0.046 3

PCC-1 Bulk peridotite 17.4 0.043 0.025 0.023 3

BM 90-21 (1) Bulk lherzolite,

Balmuccia, Apls

20.5 0.008 0.065 0.032 6

BM 90-21 (2) Bulk lherzolite,

Balmuccia, Apls

17.8 �0.011 0.042 0.039 3

BM 90-30 Bulk harzburgite,

Balmuccia, Apls

11.2 �0.024 0.021 0.019 3

RC-6 Olivine, Baja California,

Lherzloite

nd 0.015 0.045 0.026 5

5-209 Olivine, Beni Boussera 14.7 0.065 0.040 0.028 4

Bulk lherzolite,

Beni Boussera

18.6 0.060 0.038 0.035 3

PY-18 Olivine, Lherz, Apls 7.0 �0.012 0.017 0.016 3

Bulk lherzolite 38.2 0.000 0.015 0.014 3

SC-1 (a) Olivine, San Carlos,

lherzolite-xenolite

40.2 0.017 0.020 0.018 3

SC-1 (b) Olivine 0.016 0.030 0.028 3

1b-8 Olivine, xenolith,

Eifel, Germany

15.8 0.032 0.029 0.027 3

Vitim Olivine 12.9 0.040 0.030 0.028 3

RC-7 Olivine,

Grt-peridotite-xenolith

nd �0.016 0.077 0.071 3

Mantle meand 0.015 0.055 0.018 11

Pallasites

Brenham-L-69 Olivine 15.7 �0.010 0.026 0.015 5

Brenham-L-59 Olivine 10.7 �0.007 0.038 0.026 4

Brenham-V Olivine c10 �0.002 0.033 0.015 7

Brenham-Ro Olivine �0.016 0.064 0.044 4

Brenham-L-69 Metal, mainly

schreibersite

1.8 0.251 0.027 0.013 6

Brenham-L-59 Metal 5.8 0.001 0.082 0.037 7

Brenham-V (a) Metal c100 0.080 0.061 0.035 5

Brenham-V (b) Metal 0.058 0.037 0.018 6

Brenham-V (c) Metal 0.068 0.004 0.003 4

Brenham-Ro Unknown mineral comp. �0.192 0.051 0.023 7

Mount-Vernon-L Olivine 5.2 0.047 0.022 0.015 4

Mount-Vernon-V Olivine c20 0.027 0.062 0.028 7

Mount-Vernon-L Metal 29.6 0.011 0.049 0.028 5

Mount-Vernon-V (a) Metal, mainly schreibersite c25 0.160 0.056 0.032 5

Mount-Vernon-V (b) Metal, mainly schreibersite 0.156 0.045 0.022 6

Mount-Vernon-V (c) Metal, mainly schreibersite 0.165 0.063 0.044 4

Eagle Station Olivine 4.0 �0.032 0.057 0.025 7

Eagle Station Metal 1.7 �0.167 0.047 0.023 6

Admire Olivine 8.9 0.011 0.043 0.025 5

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264254

Sample name Material Mass

(mg)

d56Fe

(x)

Uncertainty Number of

analyses

Reference

values(2 S.D.) (2 S.E.)

Pallasites

Admire Metal, mainly troilite 5.9 �0.298 0.023 0.021 3

Imilac Olivine 3.3 �0.019 0.061 0.035 5

Imilac Metal 2.7 0.147 0.062 0.036 5

Brahin Olivine 3.6 0.022 0.039 0.036 3

Brahin Metal, mainly troilite 1.8 �0.208 0.062 0.043 4

Marjalahti Olivine 23.6 �0.007 0.052 0.030 5

Marjalahti Metal 25.0 0.154 0.024 0.014 5

Molong Olivine 5.9 0.049 0.011 0.017 2

Molong Metal 1.1 �0.003 0.011 0.017 2

Olivine mean 0.005 0.052 0.016 12

Metal mean (including

all mineral phases)

�0.006 0.349 0.110 12

Pure metal 0.030 0.218 0.097 8

Moon

10084 High-Ti soil 3.8 0.226 0.034 0.024 4

15386 KREEP 2.4 0.140 0.058 0.053 3

15445 KREEP-rich

highland breccia

3.3 0.093 0.013 0.012 3

65015 (1) KREEP-rich

highland breccia

9.5 0.066 0.056 0.025 7

65015 (2) KREEP-rich

highland breccia

2.2 0.070 0.025 0.023 3

62235 KREEP-rich

highland breccia

3.6 0.028 0.038 0.035 3

78155 KREEP-poor

highland breccia

5.3 0.048 0.064 0.059 3

14310 Polymict highland

breccia

4.7 0.102 0.028 0.026 3

72153 (1) High-Ti mare basalt 4.0 0.209 0.030 0.015 6

72153 (2) High-Ti mare basalt 4.4 0.195 0.038 0.034 3

79155 High-Ti mare basalt 3.9 0.191 0.040 0.023 5

75035 High-Ti mare basalt 6.0 0.208 0.040 0.028 4

15495 (1) Low-Ti mare basalt 5.6 0.071 0.055 0.021 9

15495 (2) Low-Ti mare basalt 2.1 0.055 0.029 0.027 3

15555 Low-Ti mare basalt 10.1 0.029 0.029 0.017 5

15475 Low-Ti mare basalt 3.9 0.074 0.030 0.012 8

74220 A-17 orange glass 1.7 0.103 0.062 0.028 7

15426 A-15 green glass 9.6 0.058 0.050 0.025 6

Lunar meand 0.105 0.127 0.037 14

Lunar mean excluding

high-Ti basaltsd0.073

Mars

Shergotty Basaltic shergottite 1.5 0.047 0.059 0.029 6

Zagami Basaltic shergottite 4.2 �0.008 0.021 0.019 3

EETA 79001 (1) Basaltic shergottite 9.1 �0.022 0.040 0.028 4

EETA 79001 (2a) Basaltic shergottite 4.2 �0.001 0.013 0.012 3

EETA 79001 (2b) Basaltic shergottite �0.013 0.043 0.039 3

ALHA 77005 Lherzolitic shergottite 7.9 �0.014 0.025 0.017 4

Lafayette Nakhlites

(augite-olivine cummulate)

9.3 �0.066 0.058 0.029 6

Nakkla Nakhlites

(augite-olivine cummulate)

5.0 �0.046 0.035 0.032 3

ALHA 84001 Orthopyroxenite 4.6 0.017 0.035 0.024 4

Marsian meand �0.012 0.066 0.029 7

Table 2 (continued)

(continued on next page)

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 255

Sample name Material Mass

(mg)

d56Fe

(x)

Uncertainty Number of

analyses

Reference

values(2 S.D.) (2 S.E.)

Vesta

Bouvante Basaltic eucrites 0.031 0.036 0.033 3

Juvinas Basaltic eucrites 0.003 0.011 0.010 3

Millbillillie Basaltic eucrites 0.030 0.051 0.047 3

Stannern Basaltic eucrites 0.012 0.021 0.020 3

Vesta meand 0.019 0.027 0.019 4

Planetary mean

(excluding the moon)

0.004 0.027

Average uncertainties

(2 S.D. and 2 S.E.)

of the sample replicates

0.039 0.027

a Reference values are from Williams et al. [37], d56 Fe is calculated from d57 Fe, the hematite is an in-house standard from ETH Zurich.b Reference values are from Poitrasson et al. [5], d56 Fe is calculated from d57 Fe, BCR-1 was measured instead of BCR-2.c Reference values are from Beard et al. [1].d For planetary mean calculations, only the average of replicate measurements is used; the mantle mean includes the average of the bulk rock and

the olivine values for the samples 5-209 and Py-18. The lunar mean excludes the value for the lunar soil (10,084), since the Fe isotope composition

of lunar soils is probably altered by space weathering processes [38].

Table 2 (continued)

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264256

Cu+SSB technique), a precision of typically 0.03–

0.06x (2 S.D.) for d56Fe (and 0.04–0.1x for d57Fe)

was routinely achieved for replicate measurements

(Table 1, Fig. 2). The average precision of replicate

sample measurements for d56Fe was: 2 S.D.=0.039 and2 S.E.=0.027 (in average n =4.5 replicates). This im-

provement by a factor of about 2 compared to previous

studies was essential to resolve the small isotopic dif-

ferences between geochemical reservoirs fractionated at

high temperatures.

Individual samples were typically measured three to

seven times. Only the mean values of the replicate

measurements are given in Tables 1 and 2. Also given

are the uncertainties as two standard deviations (2 S.D.)

and two standard errors of the mean (2 S.E.). The latter

Fig. 1. Plot of the fractionation factors b (from the exponential law)

of 65Cu/63Cu versus 56Fe/54Fe. The b-values for Cu and Fe are well

correlated for both, the IRMM-014 standard solution and Canon

Diablo, indicating that mass discrimination by the ICP source close-

ly follows the exponential law for both elements. However, the

slope of c0.91 shows that mass discrimination for Fe is slightly

less than for Cu.

were used in this study (applying a Student’s t-distribu-

tion on the 95% confidence level) when comparing the

isotopic compositions of planetary bodies, geochemical

reservoirs and different rock types. This treatment is

valid if we assume a Gaussian distribution of each data

set (where the data sets can be replicate measurements of

a sample or random sampling from a geochemical res-

ervoir). This assumption is not necessarily valid for the

samples from Mars and the Moon, since the measured

samples are probably not representative of the bulk

planets. However, using either 2 S.D. or 2 S.E. does

not essentially effect the implications of the study.

3. Results

3.1. Replicates and standards

To exclude the possibility of Fe isotope fractionation

during sample preparation, replicate analyses were per-

ig. 2. Replicate measurements of a solution from the iron meteorite

F

bCanon DiabloQ over a period of c1 year.

Fig. 3. Fe isotope compositions of pallasite metal and olivine sepa

rates. The metal separates include the nonmetallic phases troilite and

schreibersite. While the olivine separates are all F identical with

IRMM-014, the metal separates cover a large range from �0.298xto 0.251x.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 257

formed for several samples. For some replicates, the

entire sample preparation procedure, including diges-

tion and purification, was repeated. In other cases, only

the purification step was repeated on aliquots from the

same sample digest. In almost every case, the Fe iso-

tope compositions of replicates agreed within analytical

uncertainties (Tables 1 and 2). The only exceptions are

the metal separates of the pallasites, which are miner-

alogically, chemically and isotopically heterogeneous

on a small scale (see below).

Possible effects of small amounts of residual matrix

(after sample purification) on the measurement were

also tested by doping a clinopyroxene (cpx) and an

olivine matrix with IRMM-014, after previous separa-

tion of the natural Fe. After doping, the Fe was sepa-

rated from the matrix and the Fe isotope composition

was measured relative to IRMM-014, which should

result in d56Fe=d57Fe=0 (Table 1). This was the case

within uncertainties for both of the doped minerals.

This test also demonstrates that Fe isotope fractionation

on the columns does not affect the Fe isotope analyses,

presumably because our yields were quantitative, or

nearly so (see supplementary data in the Appendix).

The precision of the Cu+SSB technique is typically a

factor of 2–3 better (c0.03–0.06x for d56Fe, 2 S.D.)

than the precision of the SSB technique without Cu mass

bias correction (Table 1), an observation consistent with

[12]. This is due to a few outlier measurements which are

produced when the mass bias occasionally bjumpsQ be-tween two measurements by N0.1x. The effect of sam-

ple matrix on Fe isotope measurements is shown by

comparing the values of the Cu+SSB and the SSB

techniques from purified and unpurified aliquots of

metal separates from two pallasites (Table 1). While

the Cu+SSB values of all aliquots agree with each

other, the SSB values of the unpurified aliquots system-

atically deviate from the purified aliquots. This demon-

strates that the Cu+SSB technique not only improves the

precision but is also more robust against matrix effects

than the conventional SSB technique.

The accuracy and precision of the method was fur-

ther tested by measuring several international and in-

house standards, such as BIR-1, BCR-2, BHVO-1, JB-

2, JP-1, PCC-1, a hematite standard and a purified

solution, prepared from the iron meteorite Canon Dia-

blo (Table 2). Our results agree with those of previous

studies within uncertainties [1,5,9].

3.2. Pallasites

The Fe isotope composition of almost all olivine

separates from pallasites agree with the IRMM-014

standard, within uncertainties (the total range for

d56Fe is �0.032 to +0.049x, Table 2, Fig. 3). In

contrast, the metal separates (which may also include

troilite and schreibersite) vary between �0.30 and

+0.25x, a variation of about 10 times the analytical

precision (Table 2, Fig. 3). However, the metal fraction

in pallasites is quite inhomogeneous and consists of

several mineral phases, such as troilite (FeS), schrei-

bersite ((Fe, Ni)3P), graphite and exsolutions of kama-

cite (a-Fe,Ni, with Ni b6%) [34]. These metallic and

nonmetallic phases could not be separated for this

study. Therefore, the analyzed metal samples always

represent mixtures of these mineral phases. Systematic

variations in the Fe isotope composition are correlated

with the modal mineral composition, which was esti-

mated from the chemical composition of the metal

samples (Table 3). While all troilite-rich samples have

negative d56Fe-values, those of schreibersite-rich sam-

ples are always positive.

Hence, the following order of Fe isotope fraction-

ation is indicated in pallasite metals:

d56Fe(troilite) (c�0.25x) bd56Fe(metal) (c0.03x)

bd56Fe(schreibersite) (c+0.20x)

This trend is consistent with the findings of Williams

et al. [9] who observed isotopically lighter Fe isotope

compositions for troilite compared to the metal in iron

-

Table 3

Chemical and calculated mineralogical composition of the pallasite metal separates

Description Mass

(mg)

d56Fe

(x)

Troilite (FeS)a

(mass%)

schreibersite

(Fe, Ni)3Pa

(mass%)

Ni / (Ni+Fe)

Brenham69 2 unaltered grains 1.8 0.251 0.9 94.7 0.432

Brenham59-met1 3 unaltered grains 5.8 0.001 11.1 2.1 0.014

Brenham-Wien 1 unaltered grain c100 0.069 0.5 0.7 0.088

Mt-Vernon-NHM 1 unaltered grain 29.6 0.011 0.3 1.1 0.099

Mt-Vernon-Wien Several altered grains c25 0.160 10.7 73.1 0.342

Marjalahti-met1 1 unaltered grain 25.0 0.154 0.5 2.3 0.163

Admire met1 2 unaltered grains 5.9 �0.298 60.4 0.1 0.002

Imilac 5 slightly altered grains 2.7 0.147 3.1 0.8 0.136

Brahin 3 unaltered grains 1.8 �0.208 65.7 3.3 0.006

Molong 4 unaltered grain 1.1 �0.003 2.8 5.5 0.015

Eagle Station 1 slightly altered grain 1.7 �0.167 3.3 0.9 0.039

a The abundance of troilite and schreibersite was calculated from the S and P concentration of the metal separates.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264258

meteorites. It is also consistent with experimental find-

ings of light Fe isotope compositions of sulfides

(D56Fe(sulfides)c�0.2–0.3x) in equilibrium with sili-

cate melt (at 800–1000 8C [35]). However, our results

are in stark contrast to the findings of Poitrasson et al.

[11], who determined the opposite order of isotope

fractionation in Pallasite metals. This puzzle remains

to be solved.

There is also a crude correlation between the Ni/

(Fe+Ni) ratio of the pure metal separates (without

significant amounts of S and P) and the Fe isotope

composition. Ni-rich metal separates (up to 16.3%)

have systematically higher d56Fe-values than metal

with a low Ni-content (b2%). This indicates Fe isotopes

are fractionated between kamacite and taenite during

kamacite exsolution at temperatures b800 8C [34],

consistent with the findings of Poitrasson et al. [11].

The mean values of the pure metal (+0.03F0.10x2 S.E., n =7), the metal including schreibersite and

troilite (�0.01F0.11x 2 S.E., n =12) and the olivine

separates (+0.01F0.02x 2 S.E., n =12) are identical

to each other and to the IRMM-014 standard within

uncertainties.

3.3. The Earth’s mantle

Bulk peridotites from orogenic peridotite massifs and

olivine separates mainly from peridotitic xenoliths were

analyzed to estimate the Fe isotope composition of the

bulk silicate Earth (Table 2). For several samples, the Fe

isotope compositions of all major Fe-bearing minerals

(olivine, opx, cpx, spinel or garnet) were measured,

Consistent with [36,37], small but systematic differences

on the order of 0.1x between the minerals were ob-

served. These variations will be discussed in detail in a

separate study [48]. As expected, the Fe isotope compo-

sitions of the olivine separates always agree with the bulk

rock, because olivine contains most of the Fe in the

peridotite samples.

We obtained an average d56Fe value of +0.015xF0.018 (2 S.E., n =11) for the analyzed peridotites,

which we take as an approximation for BSE (Table 2).

This value does not agree within 2 S.E. uncertainties

with the igneous rock value obtained by Beard et al. [1]

and Poitrasson et al. [5]. However, the results of this

and previous studies are consistent if we divide the

samples into petrological groups (peridotites and mag-

matic rocks, Fig. 5). The average of the four basalts and

one komatiite measured in this study (d56Fe=

+0.076F0.029, 2 S.E., n =5) are consistent with the

average igneous rock values of [1] (d56Fe=0.09F0.05

1 S.D.) and of [5] (d57Fe=+0.102F0.032, 2 S.E.,

n =13; this corresponds to d56Fe=+0.068). Additional-

ly, the values for all individual standards such as BIR-1,

BHVO-1, BCR-2 (compared with BCR-1), PCC-1 and

JP-1 agree with those of [1] and [5]. Similarly, the few

peridotites measured by [1] and [5] are also on the low

side of their average mantle value and are consistent

with the average mantle determined in this study.

3.4. Moon, Mars and HED

Of the samples from the Moon, Mars and the HED

parent body only the lunar samples differ significantly

(outside 2 S.E. uncertainties) from the IRMM-014 stan-

dard with a mean value of +0.105xF0.037 (2 S.E.,

n =14; Table 2, Figs. 4 and 5). This mean value

excludes one lunar soil sample, since lunar soils are

known to be systematically shifted towards heavier Fe

isotope compositions by space weathering effects [38].

Fig. 4. Fe isotope compositions of all investigated lunar samples. Not

only the soil, but also the high Ti basalts are systematically higher in

d56Fe than lunar glasses, highland rocks (including KREPP) and low

Ti basalts.

Fig. 5. Fe isotope compositions of planets and planetary reservoirs

from the Moon (lunar glasses, high-Ti and low-Ti mare basalts

highland rocks), Mars (SNC meteorites), HED (eucrites) and palla

sites (olivine, metal, troilite and schreibersite) compared to the Earth’s

mantle. Also included are magmatic rocks, measured by Beard et al

[1], Poitrasson et al. [5] (d56Fe are calculated from the given

d57Fe/54Fe values) and from this study.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 259

The heavy Fe isotope composition of lunar rocks is

consistent with earlier studies [5,38]. However, in this

study we observed systematic differences between var-

ious lunar reservoirs and rock types (Fig. 4). High-Ti

basalts are systematically high in d56Fe with a value of

+0.201xF0.018 (2 S.E., n =3), while most of the

highland rocks, the low-Ti basalts and the lunar glasses

display d56Fe values between +0.03x and +0.10x.

Only one KREEP sample (15,386, the only KREEP

basalt) is slightly higher; d56Fe=+0.140x. Excluding

the high-Ti basalts, the average of the lunar samples,

d56Fe=+0.073x, is identical with the average for the

terrestrial basalt standards (d56Fe=+0.076x).

The mean values of both the SNC meteorites

(d56Fe=�0.012xF0.029 2 S.E., n =7) and the

eucrites (d56Fe=+0.019xF0.019 2 S.E., n =4) are

indistinguishable from those of the Earth’s mantle and

the pallasites. Although the eucrite mean is slightly

higher, all of the individual eucrite samples are identical

with IRMM-014 and with each other within 2 S.E.

uncertainties. The SNC meteorites include different

rock types, i.e. three basalts (Shergotty, Zagami and

EETA 79001), one lherzolite (ALHA 77005), two clin-

opyroxenite (Nakhla and Lafayette) and one orthopyr-

oxenite (ALHA 84001). Of the seven martian

meteorites, only three are different from IRMM-014

outside 2 S.E. uncertainties. The only sample with a

positive d56Fe (+0.047x) is Shergotty (basaltic) and

both samples with a negative d56Fe (�0.046x and

�0.066x) are Nakhlites (Nakhla and Lafayette, clino-

pyroxenites).

4. Discussion

4.1. Fe isotope fractionation during planetary core

formation

An important prerequisite to understand and predict

the Fe isotope variations among planetary reservoirs is

the possibility of Fe isotope fractionation during plan-

etary core formation. Pallasites are highly suitable to

address this question because they sample the core–

mantle interface of a differentiated parent body. The

results of this study reveal no systematic variations

between bulk metal and olivine of the pallasites. This

is in contrast to previous studies [6,11] which reported

Fe isotope fractionation between silicate and metal

phases in pallasites.

The differences between this study and prior studies

are probably explained by the small-scale mineralogi-

cally related isotopic heterogeneities (of c0.5x) with-

in the metal part. For example, the Fe isotope

compositions of four different metal fractions from

the same pallasite (Brenham) with different mineral

compositions cover almost the entire range observed

for all investigated pallasites (Table 2, Fig. 3). In con-

trast, the olivines are much more homogeneous in their

,

-

.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264260

Fe isotope composition and all agree within F0.05x.

Poitrasson et al. [11] measured the Fe isotope compo-

sition of pure metal directly adjacent to olivine grains in

three pallasite samples. They assumed isotopic equilib-

rium between silicate and adjacent metal. However, the

results of this study indicate that the Fe isotopes in

troilite, schreibersite and the metal phases reequili-

brated, while olivine did not reequilibrate with these

phases. Thus, equilibrium between olivine and metal

was probably disturbed by the low-temperature isotopic

reequilibration between the phases of the metal part

during cooling. For example, the exsolution of kama-

cite from high-temperature g-Fe,Ni and the formation

of various taenite phases occurs at temperatures as low

as (c400–800 8C) [34,39]. In contrast, the tempera-

tures of core formation on asteroid-sized parent bodies

are assumed to be N1500 8C [40].

Although, low-temperature processes obscure

metal–silicate Fe isotope fractionation, it is still possi-

ble to define an upper limit on the Fe isotope variations

between bulk silicate and bulk metal which might have

been produced during planetary core formation. A sig-

nificant, systematic fractionation during this process

would have produced a difference between the mean

values of all metal and all olivine fractions. However,

the metal and the olivine mean values are identical

within c0.01x. Given the relatively large uncertainty

of the metal mean (a consequence of mineralogical

heterogeneity), we can only confine the Fe isotope

fractionation between bulk metal and olivine in the

pallasites to be V0.1x (2 S.E. of the metal).

Importantly, the magnitude of metal–silicate frac-

tionation was probably smaller during formation of

Earth’s core as compared to differentiation of the palla-

site parent body because of the effect of temperature on

fractionation (equilibrium isotope fractionation factors

are negative correlated with temperature). Compared to

the pallasite parent bodies, metal silicate equilibration

during core formation in large planets like the Earth

occurred at higher temperatures. Formation of the

Earth’s core also occurred at higher pressure, but frac-

tionation factors typically show very little variation

with pressure.

4.2. Fe isotope fractionation during partial melting

In previous studies, Beard et al. [1] and Poitrasson et

al. [5] defined an igneous rock value which was dom-

inated not only by magmatic rocks but also included a

few peridotites. They assumed that Fe isotopes are not

fractionated during partial melting. In this study, we

investigated this matter by analyzing a larger number of

peridotite samples to obtain more representative statis-

tics. The results demonstrate that peridotites are isoto-

pically lighter than magmatic rocks (this study:

D56Femelt-peridotitec0.06x, relative to the magmatic

rocks of [5]: D56Fe melt-peridotitec0.05x and relative

to the magmatic rocks of [1]: D56Fe melt-peridotitec0.07x). Iron isotope fractionation during partial melt-

ing should also have an effect on the depleted peridotite

residue as proposed by Williams et al. [9,37]. There-

fore, we mainly analyzed primitive peridotites (e.g.,

SC-1 San Carlos). However, the depleted peridotites

(e.g., PCC-1) did not show any detectable difference in

Fe isotope composition compared to the primitive ones,

presumably because most of the Fe (N80%) during

partial melting stays in the mantle residue. Therefore,

the effect of partial melting on the peridotitic source

itself should be negligible.

In the light of the new data reported here, the use of

a bterrestrial baselineQ value as a reference for reportingFe isotope compositions as recommended by Beard et

al. [1] is problematic: (1) it is apparent that magmatic

rocks do not represent the bulk planet at the level of

analytical precision now attainable (Table 2, Fig. 5) and

(2) more precise and detailed studies might reveal some

variations among the magmatic rocks and the igneous

rock baseline might depend on the mix of samples used

to define it. Therefore, we recommend using a well-

defined Fe isotope standard, like the IRMM-014, par-

ticularly if small differences in the order of 0.1xbetween high-temperature reservoirs are investigated.

4.3. Fe isotope reservoirs on the Moon

The Fe isotope compositions of the lunar samples

have (1) the highest mean value and (2) the largest

spread among the investigated planetary reservoirs.

However, the lunar samples are all products of partial

melting. They do not include primitive lunar mantle

material, analogous to the terrestrial peridotites. Unfor-

tunately, such materials are not available from the

Moon. A closer look at the various petrogenetic groups

shows that the high-Ti basalts (together with the soil

sample, which is probably not only affected by space

weathering [38], but also originates from a high-Ti

basalt) have by far the highest d56Fe values: For

high-Ti basalts d56Fec+0.2x, while for other groups

+0.03bd56Feb+0.14x (Fig. 4). The latter values are

very similar to those reported for terrestrial magmatic

rocks (this study, [1,5]), though the highland samples,

including one KREEP basalt, are very heterogeneous in

their chemical and mineralogical composition. The sys-

tematic Fe isotope variations between the petrological

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 261

groups of the lunar samples raise the question as to

whether their average represents the bulk composition

of the Moon (as assumed by [5]). Therefore, compar-

isons of bulk planetary Fe isotope compositions, which

are based on these differentiated samples should be

taken with caution. The differences between the planets

might be simply related to the different differentiation

history of the samples and not to planetary formation

processes. The similar mean value of terrestrial basalts

and most of the lunar crustal rocks moreover imply that

the bulk silicate Moon should have a similar Fe isotope

composition as BSE (and other planets).

The particularly heavy Fe of the high-Ti basalts may

be related to the petrogenesis of these unusual rocks. In

contrast to the low-Ti basalts, which originate from a

lunar mantle similar to the Earth’s mantle [41], it is

generally assumed that the high-Ti basalts were gener-

ated mainly in ilmenite- and FeO-rich mantle domains

which are commonly interpreted as a relict cumulate

from a lunar magma ocean [42]. Fe isotopes could

fractionate differently during partial melting of such

domains than during melting of more typical lunar man-

tle. Alternatively, the high-Ti basalts may have inherited

their heavy Fe isotope signature directly from their FeO-

and ilmentite-rich source, which provided a substantial

amount of the Fe to these melts. This latter model would

be more likely, if FeO or ilmenite were almost quantita-

tively assimilated or consumed by the partial melting

reaction, which generated the high-Ti basalts [43,44].

Additional Fe isotope fractionation might have occurred

during fractional crystallisation of the primitive Ti-rich

magmas. Unfortunately, there are no Fe isotope partition

coefficients yet available for FeO or ilmenite in equilib-

rium with a silicate melt. Such data might help to differ-

entiate between these hypotheses and so should be a

subject of future research.

4.4. The Fe isotope composition and homogeneity of

the solar system

Our results on samples from several planetary bodies

provide constraints on the homogeneity of the solar

system Fe isotope composition on a planetary scale.

The data also provide an estimate for the average Fe-

isotope composition of the solar system. The major

uncertainty of such an estimate, however, is that the

Fe isotope composition on planets was fractionated

during silicate differentiation, as indicated on Earth.

Fractionation during core formation over a range of

b0.1x can also not be excluded.

The samples from the Earth mantle can be consid-

ered to represent the bulk silicate Earth (more restric-

tively, the upper mantle). In contrast to the Earth, the

samples from Mars, Moon and the HED parent body

could not be selected to represent the bulk silicate

planet, since only a limited number of samples are

available. As discussed above, the Fe isotope compo-

sition of the lunar samples most likely do not represent

that of the bulk moon. These samples are therefore

excluded from bulk solar system considerations. All

of the samples from the HED parent body investigated

in this study were basaltic eucrites, which are also

differentiated rocks. The martian samples (SNC meteor-

ites) include a mixture of different rock types, i.e. one

lherzolite, 3 basalts and 3 pyroxenites. The petrological

diversity probably also explains the slightly larger var-

iations in the Fe isotope composition compared to the

eucrites. The negative d56Fe values of both nakhlites

(c�0.05x), might indicate small fractionation effects

during magmatic differentiation. If we assume a similar

Fe isotope fractionation during partial melting on Mars

and on the HED parent body as we observed on Earth,

the d56Fe for both planets might be well as low as

c�0.05x relative to IRMM-014.

The Fe isotope composition of bulk HED would be

of particular interest in comparison to that of the main

group pallasites olivines, since there is a debate as to

whether these meteorites, and the IIIAB irons, originate

from the same parent body. This hypothesis is mainly

based on their indistinguishable oxygen isotope com-

position, which indicates that they come at least from

the same oxygen reservoir [45]. However, if the main

group pallasites and the HED meteorites originate from

the same parent body they must also have the same Fe

isotope composition.

The mean value of the pallasites of 0.00F0.10x(for both, metal and olivine combined) gives an esti-

mate for the Fe isotope composition of the pallasite

parent bodies. There are in total three different groups

of pallasites, representing at least three different parent

bodies, of which two, the main group and the Eagle

station group parent body, were analyzed in this study.

Although the Eagle Station group is only represented

by one sample, it can be stated that the difference

between the olivines of the two groups (representing

the mantle of the parent bodies) is smaller than 0.05x.

Despite possible fractionation processes, the Fe iso-

tope composition of planetary bodies, as represented by

peridotites from the Earth’s mantle, SNC meteorites,

eucrites and pallasites is remarkably homogeneous and

all planetary means agree with the standard IRMM-014

within F0.05x (Table 2, Fig. 5). This value is also

consistent with the average Fe isotope composition of

chondrites [11,31,46]. This implies that iron was well

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264262

mixed in the early inner solar system during accretion

and planet formation.

5. Conclusions

(1) High mass resolution MC-ICP-MS using Cu to

control for variations in mass bias, is a suitable

method to perform high precision Fe isotope mea-

surements. External reproducibilities of c0.03–

0.06x (2 S.D.) for d56Fe are routinely achieved

on replicate sample measurements, which is

sufficient to reveal small differences in the Fe

isotope composition between different high-tem-

perature rock types and geochemical reservoirs of

planetary bodies.

(2) Pallasite silicate and metal samples have identi-

cal mean Fe isotope compositions, implying that

there is little or no Fe isotope fractionation

during planetary core formation (b0.1x). The

large range in the Fe isotope composition of the

investigated metal samples (N0.5x) can be

linked to chemical and petrological sample het-

erogeneities and Fe isotope fractionation be-

tween the various metal (kamacite, taenite) and

nonmetallic (troilite, schreibersite) phases. This

implies that Fe-isotope fractionation occurred

during the differentiation of the metal melt and

during kamacite exsolution at relatively low

temperatures.

(3) Partial melting of the silicate mantle produces

magmatic rocks with a systematically heavier Fe

isotope composition than the starting material.

This conclusion follows from the difference be-

tween the mean of terrestrial basalts (this study,

[1,5]) and the mean of 11 terrestrial peridotite

samples from different tectonic settings studied

here (d56Fe=+0.015F0.018x). The peridotite

mean seems to be most suitable to define the Fe

isotope composition of the bulk silicate Earth.

Combined with conclusion (2), this value proba-

bly represents bulk Earth.

(4) The isotopically heavier Fe isotope composition

of the lunar samples with an average of d56Fe=

+0.102x (for all samples except the soil) prob-

ably reflects fractionation that occurred during

the differentiation history of the Moon rather

than during accretion. High-Ti basalts have

clearly the highest d56Fe of the analyzed lunar

samples (+0.201F0.018x), while low-Ti

basalts, highland rocks and lunar glasses have

an average d56Fe of +0.073F0.022x, identical

with terrestrial basalts.

(5) The mean Fe isotope compositions (d56Fe) of

the SNC meteorites (�0.012F0.029x), the

eucrites (+0.019F0.019x), the pallasite metals

(�0.006F0.110x), pallasite silicates (+0.005F0.016x) and the terrestrial peridotites (+0.015F0.018x) are all identical within uncertainties

implying an isotopically homogeneous solar

system.

Acknowledgements

We thank the Natural History Museum of London,

the Natural History Museum of Vienna and the Ward

collection at the University of Rochester for provid-

ing pallasite samples. Asish Basu (University of

Rochester) is thanked for providing peridotite sam-

ples. We are grateful to Rama Chakrabarti and Gail

Arnold for their support in sample preparation. Helen

Williams is thanked for providing an in-house stan-

dard and for fruitful discussions. Mark Rehkamper

and one anonymous reviewer are thanked for their

constructive and helpful reviews. This study was

supported by the DFG (Deutsche Forschungsge-

meinschaft; Project: WE 2850/1-1).

Appendix A. Supplementary data

Supplementary data associated with this article can

be found, in the online version, at doi:10.1016/j.epsl.

2005.09.023, [47].

References

[1] B.L. Beard, C.M. Johnson, J.L. Skulan, K.H. Nealson, L. Cox,

H. Sun, Application of Fe isotopes to tracing the geochemical

and biological cycling of Fe, Chem. Geol., Special Issue 195

(2003) 87–117.

[2] B.L. Beard, C.M. Johnson, High precision iron isotope measure-

ment of terrestrial and lunar materials, Geochim. Cosmochim.

Acta 63 (1999) 1653–1660.

[3] N.S. Belshaw, X.K. Zhu, Y. Guo, R.K. O’Nions, High precision

measurements of iron isotopes by plasma source mass spectrom-

etry, Int. J. Mass Spectrom. 197 (2000) 191–195.

[4] S. Weyer, J.B. Schwieters, High precision Fe isotope measure-

ments with high mass resolution MC-ICPMS, Int. J. Mass

Spectrom. 226 (2003) 355–368.

[5] F. Poitrasson, A. Halliday, D. Lee, S. Levasseur, N. Teutsch, Iron

isotope differences between Earth, Moon, Mars and Vesta as

possible records of contrasted accretion mechanisms, Earth

Planet. Sci. Lett. 223 (2004) 253–266.

[6] X.K. Zhu, Y. Guo, R.J.P. Williams, R.K. O’Nions, A. Mat-

thewsa, N.S. Belshaw, G.W. Canters, E.C. de Waal, U. Weser,

B.K. Burgess, B. Salvato, Mass fractionation processes of

transition metal isotopes, Earth Planet. Sci. Lett. 220 (2002)

47–62.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264 263

[7] S. Weyer, G. Arnold, R. Chakrabarti, A.D. Anbar, Iron isotope

fractionation at high temperatures, EGS AGU EUG 2003 Joint

Assembly, Abstract Volume, 2003 (EAE03-A-12590).

[8] S. Weyer, A. Woodland, C. Munker, G. Arnold, R.. Anbar, A.

Anbar, Iron isotope variations in the Earth’s mantle and the

terrestrial planets., Geochim. Cosmochim. Acta 68 (2004) A1113.

[9] H. Williams, C. McCammon, A. Peslier, A. Halliday, N.

Teutsch, S. Levasseur, J. Burg, Iron isotope fractionation

and the oxygen fugacity of the mantle, Science 304 (2004)

1656–1659.

[10] H.M. Williams, A.N. Halliday, N. Teutsch, S. Levasseur, Iron

isotope fractionation in iron meteorites: new insights into metal–

sulfide segregation and core crystallization, Eos Trans.-Am.

Geophys. Union 85 (47) (2004) (Fall Meet. Suppl., Abstract

P33A).

[11] F. Poitrasson, S. Levasseur, N. Teutsch, Significance of iron

isotope mineral fractionation in pallasites and iron meteorites

for the core–mantle differentiation of terrestrial planets, Earth

Planet. Sci. Lett. 234 (2005) 151–164.

[12] G.L. Arnold, S. Weyer, A.D. Anbar, Fe isotope variation in

natural materials measured using high mass resolution MC-

ICPMS, Anal. Chem. 76 (2004) 322–327.

[13] W. Hartmann, D. Davis, Satellite-sized planetisimals and lunar

origin, Icarus 24 (1975) 505–515.

[14] A. Cameron, W. Benz, Origin of the moon and the single impact

hypothesis, Icarus 92 (1991) 204–216.

[15] R.M. Canup, E. Asphaug, Origin of the Moon in a giant

impact near the end of the Earth’s formation, Nature 412

(2001) 708–712.

[16] T. Kleine, C. Munker, K. Mezger, H. Palme, Rapid accretion and

early core formation on asteroids and the terrestrial planets from

Hf-W chronometry, Nature 418 (2002) 952–955.

[17] C. Munker, J. Pfander, S. Weyer, A. Buchl, T. Kleine, K.

Mezger, Evolution of planetary cores and the Earth–Moon

system from Nb/Ta systematics, Science 301 (2003) 84–87.

[18] L. Beard, C. Johnson, Fe isotope variations in the modern and

ancient Earth and other planetary bodies, Geochemistry of Non-

Traditional Stable Isotopes, Reviews in Mineralogy and Geo-

chemistry, vol. 55, 2004, pp. 319–357.

[19] P. Wasson, G. Wetherhill, Dynamical, chemical, and isotopic

evidence regarding the formation location of asteroids and

meteorites, in: T. Gehrels (Ed.), Asteroids, University of Arizona

Press, Tucson, 1979, pp. 926–974.

[20] D. Bogard, P. Johnson, Martian gases in an antartic meteorite?

Science 221 (1983) 651–654.

[21] R.P. Binzel, S. Xu, Chips of asteroid 4 Vesta: evidence for the

parent body of basaltic achondrite meteorites, Science 260 (1993)

186–191.

[22] T. McCord, J. Adams, T. Johnson, Asteroid Vesta: spectral

reflectivity and compositional implications, Science 168

(1970) 1445–1447.

[23] G. Consolmagno, M. Drake, Composition and evolution of the

eucrite parent body: evidence from rare earth elements, Geo-

chim. Cosmochim. Acta 41 (1977) 1271–1282.

[24] S. Weyer, C. Munker, M. Rehkamper, K. Mezger, Determi-

nation of ultra low Nb, Ta, Zr, and Hf concentrations and

precise Nb/Ta and Zr/Hf ratios by isotope dilution analyses

with multiple collector ICP-MS, Chem. Geol. 187 (2002)

295–313.

[25] A. Basu, Geochemistry of ultramafic xenoliths from San Quin-

tin, Baja California, in: F. Boyd, H. Meyer (Eds.), Second

International Kimberlite Conference, 1979.

[26] I. MacGregor, A. Basu, Thermal structure of the lithosphere: a

petrologic model, Science 185 (1974) 1007–1011.

[27] A. Woodland, J. Kornprobst, B. Wood, Oxygen thermobaro-

metry of orogenic lherzolite massifs, Petrology 33 (1992)

203–230.

[28] S. Weyer, C. Munker, K. Mezger, Nb/Ta, Zr/Hf and REE in the

depleted mantle: implications for the differentiation history of

the crust–mantle system, Earth Planet. Sci. Lett. 205 (2003)

309–324.

[29] H.-M. Seitz, G. Brey, Y. Lahaye, S. Durali, S. Weyer, Lithium

isotopic signatures of peridotite xenoliths and isotopic fraction-

ation at high temperature between olivine and pyroxenes, Chem.

Geol. 212 (2004) 163–177.

[30] A.D. Anbar, J.E. Roe, J. Barling, K.H. Nealson, Nonbio-

logical fractionation of iron isotopes, Science 288 (2000)

126–128.

[31] K. Kehm, E. Hauri, C. Alexander, R. Carlson, High precision

iron isotope measurements of meteoritic material by cold

plasma ICP-MS, Geochim. Cosmochim. Acta 67 (2003)

2879–2891.

[32] D. Malinovsky, A. Stenberg, I. Rodushkin, H. Andren, J. Ingri,

B. Ohlander, D. Baxter, Performance of high resolution MC-

ICP-MS for Fe isotope ratio measurements in sedimentary

geological materials, J. Anal. Atom. Spectrom. 18 (2003)

687–695.

[33] W.A. Russel, D.A. Papanastassiou, T.A. Tombrello, Ca isotope

fractionation on the Earth and other Solar System materials,

Geochim. Cosmochim. Acta 42 (1978) 1075–1090.

[34] D. Mittlefehldt, T. McCoy, C. Goodrich, A. Kracher, Non-

chondritic meteorites from asteroidal bodies, in: J. Papike

(Ed.), Planetary Materials, Reviews in Mineralogy, vol. 36,

1998, pp. 4-1-195.

[35] J. Schuessler, R. Schoenberg, H. Behrens, F. Von Blanckenburg,

Experimental calibration of the Fe isotope fractionation between

pyrrhotite and silicate melt, Goldschmidt Conference Abstracts,

2005, pp. A211.

[36] B.L. Beard, C.M. Johnson, Inter-mineral Fe isotope varia-

tions in mantle-derived rocks and implications for the Fe

geochemical cycle, Geochim. Cosmochim. Acta 68 (2004)

4727–4743.

[37] H. Williams, A. Peslier, C. McCammon, A. Halliday, S.

Levasseur, N. Teutsch, J. Burg, Systematic iron isotope varia-

tions in mantle rocks and minerals: the effect of partial

melting and oxygen fugacity, Earth Planet. Sci. Lett. 235

(2005) 435–452.

[38] R. Wiesle, B. Beard, L. Taylor, C. Johnson, Space weathering

processes on airless bodies: Fe isotope fractionation in the lunar

regolith, Earth Planet. Sci. Lett. 216 (2003) 457–465.

[39] C. Yang, D. Williams, J. Goldstein, Low temperature phase

decomposition in metal from iron, stony-iron and stony meteor-

ites, Geochim. Cosmochim. Acta 61 (1997) 2943–2956.

[40] G.J. Taylor, Core formation in asteroids, J. Geophys. Res. 97

(1992) 14717–14726.

[41] J.H. Jone, H. Palme, Geochemical constraints on the origin of

the Earth and Moon, in: R.M. Canup, K. Righter (Eds.), Origin

of the Earth and Moon, University of Arizona Press, Tucson,

2000, pp. 197–216.

[42] S. Taylor, P. Jakes, The geochemical evolution of the moon, 5th

Lunar and Planetary Science Conference, 1974, pp. 1287–1305.

[43] A. Ringwood, S. Kesson, A dynamic model for mare basalt

petrogenesis, 7th Lunar and Planetary Science Conference,

1976, pp. 1697–1722.

S. Weyer et al. / Earth and Planetary Science Letters 240 (2005) 251–264264

[44] C.K. Shearer, J.J. Papike, Magmatic evolution of the Moon, Am.

Mineral. 84 (1999) 1469–1494.

[45] R. Clayton, T. Mayeda, Oxygen isotope studies of achondrites,

Geochim. Cosmochim. Acta 60 (1996) 1999–2018.

[46] X.K. Zhu, Y. Guo, R.K. O’Nions, E.D. Young, H.D. Ash,

Isotopic heterogeneity of iron in the early solar system, Nature

412 (2001) 311–313.

[47] C.N. Marechal, P. Telouk, F. Albarede, Precise analysis of

copper and zinc isotopic compositions by plasma-source mass

spectrometry, Chem. Geol. 156 (1999) 251–273.

[48] S. Weyer, A.B. Woodland, D. Ionov, G.P. Brey, Evidence for

partial melting, melt percolation and mantle metasomatism from

Fe isotopes. (In preparation).