Mineralogy and Petrology

, Volume 98, Issue 1, pp 91–110

Iron isotope compositions of carbonatites record melt generation, crystallization, and late-stage volatile-transport processes

Authors

    • Department of Geology and GeophysicsLewis G. Weeks Hall for Geological Sciences
  • Keith Bell
    • Isotope Geochemistry and Geochronology Research Centre2117 Herzberg Laboratories, Carleton University
  • Brian L. Beard
    • Department of Geology and GeophysicsLewis G. Weeks Hall for Geological Sciences
  • Aaron I. Shultis
    • Department of Geology and GeophysicsLewis G. Weeks Hall for Geological Sciences
Original Paper

DOI: 10.1007/s00710-009-0055-4

Cite this article as:
Johnson, C.M., Bell, K., Beard, B.L. et al. Miner Petrol (2010) 98: 91. doi:10.1007/s00710-009-0055-4

Abstract

Carbonatites define the largest range in Fe isotope compositions yet measured for igneous rocks, recording significant isotopic fractionations between carbonate, oxide, and silicate minerals during generation in the mantle and subsequent differentiation. In contrast to the relatively restricted range in δ56Fe values for mantle-derived basaltic magmas (δ56Fe = 0.0 ± 0.1‰), calcite from carbonatites have δ56Fe values between −1.0 and +0.8‰, similar to the range defined by whole-rock samples of carbonatites. Based on expected carbonate-silicate fractionation factors at igneous or mantle temperatures, carbonatite magmas that have modestly negative δ56Fe values of ~ −0.3‰ or lower can be explained by equilibrium with a silicate mantle. More negative δ56Fe values were probably produced by differentiation processes, including crystal fractionation and liquid immiscibility. Positive δ56Fe values for carbonatites are, however, unexpected, and such values seem to likely reflect interaction between low-Fe carbonates and Fe3+-rich fluids at igneous or near-igneous temperatures; the expected δ56Fe values for Fe2+-bearing fluids are too low to produced the observed positive δ56Fe values of some carbonatites, indicating that Fe isotopes may be a valuable tracer of redox conditions in carbonatite complexes. Further evidence for fluid-rock or fluid-magma interactions comes from the common occurrence of Fe isotope disequilibrium among carbonate, oxide, silicate, and sulfide minerals in the majority of the carbonatites studied. The common occurrence of Fe isotope disequilibrium among minerals in carbonatites may also indicate mixing of phenocyrsts from distinct magmas. Expulsion of Fe3+-rich brines into metasomatic aureols that surround carbonatite complexes are expected to produce high-δ56Fe fenites, but this has yet to be tested.

Introduction

Stable and radiogenic isotope studies of carbonatites have been used to monitor the secular evolution of the sub-continental mantle (e.g. Bell and Rukhlov 2004), the evolution of carbonated melts as they migrate from mantle to crustal levels (e.g. Harmer 1999), and sub-solidus cooling and fluid/rock interaction (e.g. Deines 1989). Carbonatites range in age from Archean to present, are found on all continents (Woolley and Kjarsgaard 2008), and have distinct chemical compositions relative to silicate igneous rocks (e.g. Simonetti et al. 1997; Chakmouradian 2006). Although volumetrically small compared to other igneous rocks, carbonatites provide unique probes into the mantle, and, because their ages extend back into the Archean, they can be used to monitor the chemical and isotopic evolution of the mantle over much of Earth’s history. Radiogenic isotope (Sr, Nd, Pb) compositions of carbonatites have shown, unequivocally, that carbonatite magmas are of mantle origin, that many have compositions that are similar to those found in OIBs, and that mixing between isotopically distinct, carbonatitic melts is common. Stable and radiogenic isotope disequilibrium among minerals, even within the same sample, demonstrates the commonly cumulate nature of carbonatites, and is well explained by mixing within magma chambers, as well as the effects of intrusion cooling and alteration (e.g. Simonetti and Bell 1994a; Bizzarro et al. 2003; Haynes et al. 2003). Most young carbonatites (<200 Ma) have isotopic compositions that are typical of sub-oceanic mantle, pointing to sub-lithospheric sources, similar to the HIMU, EM I, and FOZO components defined by oceanic basalts (Bell and Tilton 2001; Bell and Simonetti 2009). A substantial body of isotopic data now exists for carbonatites from East Africa, which suggests mixing between the HIMU and EM I mantle components (e.g. Bell and Dawson 1995; Bell and Simonetti 1996). Noble gas compositions indicate mantle sources (e.g. Marty et al. 1998; Tolstikhin et al. 2002), and some carbonatites have Li, C, and O isotope compositions that are similar to those of oceanic basalts (e.g. Deines 1989; Keller and Hoefs 1995; Halama et al. 2008). Among the models proposed for the sources of carbonated melts, isotopic data generally support one involving mantle upwelling such as plumes/hot spot activity, accompanied in some cases perhaps by interaction with the continental lithosphere (see discussion in Bell and Simonetti 2009, and references within).

Carbonatite complexes commonly contain a wide variety of rock types, and their close spatial association with deep-crustal fracturing and rifting implies an intimate relation between intrusion and tectonism; this is particularly well shown by alkalic-carbonatitic complexes in the East African Rift Valley System (e.g. Bailey 1993) and the Trans-Superior Tectonic and Kapuskasing Structural Zones of the Superior Province, Canada (e.g. Sage 1991). Most carbonatite complexes take the form of circular- or tear-shaped plutons, many associated with silicate rocks of miaskitic affinity. Where related alkalic silicate rocks occur, these form large stocks or ring complexes, and the carbonatites generally occur as plug-like intrusive bodies that have diameters <5 km. Carbonatite complexes may consist solely of carbonatite, usually dolomitic in composition, whereas others are associated with silicate rocks, normally undersaturated with respect to silica. Silicate rocks commonly associated with carbonatites include syenite, nepheline syenite, gabbro, melilitolites, and their volcanic equivalents, as well as pyroxenites (e.g. King and Sutherland 1960; Le Bas 1977). Calciocarbonatites and/or magnesiocarbonatite make up the bulk of the carbonatites within a given complex, but late-stage carbonatites do occur, where these generally comprise only a few percent of the total carbonatite volume. The most common, late-stage carbonatites are composed of ankerite or form ankeritic dolomite-bearing carbonatites, enriched in REEs, fluorite, and incompatible trace elements such as U and Th (Le Bas 1989). Carbonatite complexes are commonly surrounded by fenites, which are metasomatic aureoles produced by expulsion of alkali-rich fluids from the carbonate and/or silicate magmas into the surrounding country rocks.

In this study, we present the first Fe isotope study of carbonatites, including whole-rocks and mineral phases. We show here that the largest range in Fe isotope compositions yet measured in igneous rocks is found in carbonatites. The relatively large Fe isotope fractionations among carbonates, silicates, and oxides at igneous temperatures, coupled with the large contrasts in Fe contents among these mineral groups, makes Fe isotopes a particularly sensitive tracer of processes that are commonly invoked in models for carbonatite genesis and evolution, including magmatic and fluid evolution, crystal fractionation, and liquid immiscibility, (for review, see Lee and Wyllie (1994)). The results of this Fe isotope survey of carbonatites suggest that Fe isotope fractionations among silicates, carbonates, oxides, and sulfides rarely record isotopic equilibrium in carbonatites. Rather, the Fe isotope compositions measured in these minerals record complex differentiation pathways, mixing of phenocrysts from distinct magmas, and late-stage fluid interactions.

Fenitization and the role of fluids in carbonatite evolution

The enormous capacity of carbonated mantle-derived magma to dissolve CO2 and H2O, along with other volatiles such as Cl, F, and S, requires fluid phases to develop and evolve during carbonatite magma differentiation. Fluid compositions can be estimated from fenites, fluid inclusion studies of carbonatites themselves, and mineral chemistry, especially REEs abundances and their distribution patterns. Because CO2 and H2O were the principal volatiles used in melting experiments, the concept of a “carbothermal fluid” of varied CO2 and H2O ratio was introduced in the literature, but it was suggested that other components, especially the halogens, may be just as important as H2O (Gittins 1989).

Fenites, a term coined by Brögger (1921), generally consist of alkali feldspar, sodic pyroxene, and/or alkali amphibole that formed at sub-igneous temperature. Fenites may be zoned relative to a carbonatite intrusion, with an innermost part composed of amphibole and pyroxene, and an outer part rich in biotite (e.g. Le Bas 2008). Metasomatised rocks can be broadly divided into sodic or potassic varieties, although other, more complicated classification schemes have been developed (Morogan 1994). Both sodium- and potassium-rich fenites can occur around a single intrusion, and it has been suggested that these distinctions may be related to depth, where potassium fenitization occurs at the upper levels of a carbonatite complex, and sodic fenitization occurs at greater depths (Le Bas 1989). Fenitization is envisioned to occur by infiltration of fluids from carbonatitic/silicate melts along distinct pathways, producing a network of alteration minerals. In some cases, the presence of disseminated fenitization products implies diffusive processes, where precursor minerals have been completely replaced, reflecting pervasive penetration by fenitizing fluids (e.g. White-Pinilla 1996). Extreme degrees of fenitization has been invoked to explain formation of magmatic silicate rocks, such as syenites, nephelinites and ijolites, by palingenesis of crustal wall rocks after high-grade fenitization (Kramm and Sindern 1998).

A great diversity of compositions has been inferred for fenitizing fluids. Wide variations in the inferred composition of fenitizing fluids has been argued to reflect complex evolution of fluids (e.g. Andersen 1986; Kresten and Morogan 1986; Andersen 1989; Bühn and Rankin 1999; Rankin 2005). Variables that control fenitization by carbonatitic fluids include XCO2 of the fluid, temperature gradients, fO2, FeO/MgO ratio, and activity gradients of SiO2, Al2O3, and CaO (Morogan 1994). At Alnö, for example, two distinct fluids are thought to have been involved in fenitization, derived from carbonatitic and ijolitic magmatic sources (Morogan and Wooley 1988). Based on fenitization products, the carbonatitic-type fluid had \( {\text{X}}_{{{\text{CO}}_2 }} > {\text{X}}_{{{\text{H}}_2 {\text{O}}}} \), high αCaO, possibly high \( \alpha_{{{\text{Na}}_2 {\text{O}}}} \) or \( \alpha_{{{\text{K}}_2 {\text{O}}}} \) (or both), very low \( \alpha_{{{\text{SiO}}_2 }} \), and possibly high F and P contents. Temperatures and fO2 were initially high, but decreased sharply with distance from the source. The occurrence of fluorite, including in some low-grade fenites, suggests interaction with a F-rich fluid (Morogan 1989). Late-stage, selective removal of La and HREE was attributed to passage of a late, highly-oxidizing post-magmatic fluid.

Direct determination of fluid compositions involved in carbonatite/fluid interaction and country rock/fluid interaction have relied primarily on fluid inclusions in minerals such as apatite and fluorite. The complexities of fluid-mineral interaction are well shown by fluid inclusion studies of apatite from the Jacupiranga carbonatite, Brazil (Costanzo et al. 2006), where magmatic evolution is interpreted to have involved crystal fractionation and settling of a carbonatite mineral assemblage in a fluid-stratified magma chamber. Studies of fluid inclusions in apatite from the Fen carbonatite, Norway, identified a magmatic fluid that was rich in CO2 and NaCl, which evolved during magmatic differentiation to a CO2-free, water-dominated system that contained higher salinities and densities as solidification progressed (Andersen 1986). In addition, marked changes in HF fugacity indicated the presence of two or more independent or semi-independent lines of magmatic descent (Andersen and Austrheim 1991). Changes in fluid composition upon late-stage mixing of fenitizing fluids and low-salinity meteoric waters were documented by Williams-Jones and Palmer (2002) based on fluid inclusions in the fenites from the Amba Dongar complex, India. Increases in fO2 during fenitization have been inferred from S isotope compositions of sulfides from ferrocarbonatite from Swartbooisdrift, Namibia (Drueppel et al. 2006).

Pertinent to this survey of Fe isotope compositions of carbonatites, fluid inclusion studies have also demonstrated the importance of Fe-rich chlorides in orthomagmatic carbonatite fluids (Bühn et al. 2002; Rankin 2005). The high Fe3+/Fe2+ ratios measured in many fenites, as well as common occurrence of disseminated iron oxides in fenites, indicates that Fe3+-bearing fluids may be expelled by carbonatite intrusions during crystallization and solidification (e.g. Le Bas 2008). The presence of hematite, especially in the low-grade fenites at Alnö, indicates equilibration with a highly oxidizing fluid (Morogan 1989). Bühn and Rankins’ (1999) fluid inclusion study of a fenite associated with the Kalkfeld carbonatite complex, Namibia, showed that virtually all alkali metals and Cl, and a major proportion F, Th, U, and Ti, were preferentially partitioned into the fluid. In addition, the fenitizing fluid at Kalkfeld was estimated to have contained 3.0 to 4.1 wt. % total Fe.

Iron isotope geochemistry

Iron isotope fractionations among aqueous species and minerals are controlled by redox state and bonding environments (e.g. Beard and Johnson 2004a; Schauble 2004). Geologic substances that contain Fe3+ tend to have higher 56Fe/54Fe ratios relative to Fe2+ substances, with the important exception of pyrite, where Fe is covalently bonded (Polyakov and Mineev 2000). Based on experimentally determined or predicted Fe isotope fractionation factors, the relative order of increasing 56Fe/54Fe ratios is Ca-Mg-Fe carbonate, Fe carbonate, Fe2+ silicate, aqueous Fe2+, magnetite, aqueous Fe3+, hematite, and pyrite, where the relative order for aqueous species may be significantly changed by chloride (Polyakov and Mineev 2000; Welch et al. 2003; Wiesli et al. 2004; Anbar et al. 2005; Polyakov et al. 2007; Hill and Schauble 2008; Shahar et al. 2008).

Although Fe isotope fractionations at igneous or mantle temperatures are relatively small, mantle-derived ultramafic rocks and minerals have a measureable range in Fe isotope compositions, and their average 56Fe/54Fe ratio is about 0.1 per mil (‰) lower than the average of basaltic rocks, possibly reflecting the effects of melting and/or metasomatism (Beard and Johnson 2004b; Poitrasson et al. 2004; Williams et al. 2004; Weyer et al. 2005; Williams et al. 2005; Weyer and Ionov 2007). Within typical analytical uncertainties of ±0.05 to 0.1‰ for 56Fe/54Fe ratios, all basalts appear to be isotopically homogenous. Evolved igneous rocks, however, may, but not always, have relatively elevated 56Fe/54Fe ratios that are ~0.1 to 0.3‰ higher than primitive basaltic rocks, and these isotopic compositions have been interpreted to reflect magmatic differentiation processes, including crystal fractionation, assimilation, and/or fluid exsolution (Poitrasson and Freydier 2005; Schoenberg and Von Blanckenburg 2006; Heimann et al. 2008; Teng et al. 2008).

Samples

We have analyzed carbonatites and their minerals from eleven different countries covering three continents. Most of the 35 samples studied are from Africa. Table 1 summarizes the seventeen complexes studied. Ages range from the Archean to present, and most of the carbonatites are sövites (coarse-grained calcite carbonatite). The majority of the complexes are associated with well-developed zones of fenitization. Almost all of the samples are either plutonic or hypabyssal, other than those from Oldoinyo Lengai; these latter samples are chemically dissimilar to the other carbonatites analyzed in that the Oldoinyo Lengai samples are natrocarbonatites, which have combined alkalis of about 40 wt. %, of which ~31 wt. % is Na2O. In addition to analyses of whole-rock samples, Fe isotope measurements were made of calcite, dolomite, magnetite, spinel, olivine, pyroxene, biotite, phlogopite, melilite, nepheline, perovskite, pyrite, monticellite, and actinolite. Because carbonatites are known to contain a great variety of minerals, this study permitted the first Fe isotope analyses of many of these minerals from samples that formed at igneous temperatures.
Table 1

Carbonatite complexes analyzed for Fe isotope compositions

Name of complex, country

Age (Ma)

Petrologic details

Fenites

AFRICA

Bukusu, Uganda

25 ± 2.5

Sövite, ankerite sovite, and rauhagite. Silicate rocks: melteigite, ijolite, pyroxenite, hornblendite, nepheline syenite.

Yes (K-rich)

Sukulu, Uganda

<40

Sövite. Silicate rocks: rim of syenite, phonolite dikes.

Probably

Tororo, Uganda

40

Sövite. Silicate rocks: Pyroxenite, melteigite, ijolite, nepheline syenite.

Yes

Toror, Uganda

15.5 ± 6

Sövite, ferrocarbonatite, dolomitic sovite. Silicate rocks: trachyte, phonolite, nephelinite melteigite, ijolite, pyroxenite, hornblendite, nepheline syenite.

Yes (K-rich)

Homa Bay, Kenya

12−1.3

Sövite, alvikite and ferrocarbonatite. Silicate rocks: ijolite, phonolite, nephelinite.

Yes

Oldoinyo Lengai, Tanzania

Still active

Natrocarbonatite flows and tuffs. Silicate rocks: phonolitic and nephelinitic tuffs and rare lavas including rare combeite- and melilite-bearing types. Blocks of pyroxenite, ijolite series rocks, nepheline syenite.

Yes (metasomatized blocks)

Panda Hill, Tanzania

113 ± 6

Mainly sövite but areas of ferrocarbonaite and dikes of dolomite carbonatite.

Yes

Sengeri Hill, Tanzania

Same as Panda?

Dolomitic dikes.

Yes

Dicker Willem, Namibia

49 ± 1

Sövite and alvikite. Silicate rocks: Ijolite xenoliths in carbonatite, trachyte dikes.

Yes

NORTH AMERICA

Borden, Canada

1872 ± 13

Sövite, silicocarbonatite, beforsite dikes.

Yes

Oka, Canada

110

Sövite and dolomite carbonatite. Silicate rocks: Okaite-jacupirangite (melilitolite-pyroxenite), and ijolite series rocks; lamprophyre dikes.

Yes

St. Honoré, Canada

ca. 650

Sövite, dolomite and ankerite carbonatite. Silicate rocks: Syenite, nepheline syenite, and ijolite.

Yes

Magnet Cove, USA

ca 100

Sövite. Silicate rocks: Pyroxenite, gabbro, jacupirangite, melteigite, ijolite, syenite, trachyte, phonolite.

Yes

SOUTH AMERICA

Jacupiranga, Brazil

130

Sövites, with a zone of dolomitization, and dikes of alvikite and beforsite. Silicate rocks: peridotite, pyroxenite, jacupirangite, ijolite, nepheline syenite, essexite, tinguaite, monchiquite.

Yes

EUROPE

Kaiserstuhl, Germany

16.0

Sövite, alvikites. Silicate rocks: limburgite, phonolite and leucite tephrite, shonkinite, bergalite, nephelinite, phonolite, monchiquite, essexite and theralite.

Unknown

Kovdor, Russia

365 ± 8

Sövite and dolomite carbonatite. Silicate rocks: Olivinite, pyroxenite, nepheline pyroxenite, melteigite, ijolite, melilite- and monticellite-bearing rocks, nepheline syenite.

Yes

Siilinjarvi, Finland

2617 ± 10

Sövite, silicocarbonatite. Silicate rocks: Glimmerite, syenite, lamprophyre.

Yes

Geological summary of the complexes arranged by country and continent. Most of the petrological details taken from Woolley and Kjarsgaard (2008). Details of the fenites not given, other than Toror. Some of the analytical uncertainties for the ages are unavailable. Some recent data have been used to assess some of the ages. Included among these are those for Kovdor and Siilinjarvi (Rhuklov and Bell, this volume)

Analytical methods and nomenclature

Mineral separates were obtained by hand picking, with special care given to the carbonate fraction of carbonatite samples. The very low Fe contents of carbonate minerals in carbonatites required mineral separate purity in excess of 99%. Mineral separates were washed in doubly distilled water (2X H2O) prior to dissolution. Iron isotope analysis of carbonatites presents special challenges due to very low ratios of Fe to alkali or alkali-earth elements, as well as the great sensitivity of low-Fe carbonate minerals to small amounts of contamination by Fe-rich silicate or oxide minerals, and the new methods we developed to address these issues are described below.

Partial dissolution experiments and application to carbonate iron isotope analyses

Despite a mineral separate purity of >99%, the very low Fe contents of the carbonate mineral separates raises the possibility that their Fe isotope compositions may have been modified by small amounts of oxide and/or silicate mineral contamination. Total dissolution of several hand-picked, >99% pure calcite mineral separates produced Fe contents that significantly exceeded those determined by electron microprobe analysis by previous studies of the same samples (Haynes et al. 2003), suggesting the presence of small amounts of oxide or silicate contamination that was not visible under a binocular microscope. We therefore developed a partial dissolution protocol that completely dissolved carbonate but did not significantly dissolve oxide or silicate that may have existed in the carbonate mineral separates (Table 2). Partial dissolution experiments of magnetite and silicate (clinopyroxene) minerals involved high (85°C) and low (25°C) temperatures, two sieved grain sizes, and three different molarities of HCl. Small amounts of Fe are dissolved from magnetite using 7 M HCl at high or low temperatures, at 1.6% and 1.1% dissolution at 85°C and 25°C, respectively. Magnetite that was sized to 1 mm and 0.1 mm in diameter responded differently to partial dissolution, where hot 7 M HCl dissolved 1.6% and 1.2%, respectively, and cold 7 M HCl dissolved 1.1% and 0.7%, respectively. At lower HCl molarity the differences in the amount dissolved between grain sizes and temperature was less pronounced. The lowest amount dissolved (0.01%) was accomplished using 1 mm magnetite crystals in cold 1 M HCl.
Table 2

HCl partial dissolution experiments

Total initial Fe (g)

Fe dissolved (µg)

Temperature (°C)

Acid

Grain size (mm)

Mineral

% Fe dissolved

0.01224

199.8

85

7 M HCl

1

MT

1.63

0.02568

308.4

85

7 M HCl

0.1

MT

1.20

0.01464

8.8

85

7 M HCl

0.5

CPX

0.06

0.01385

150.4

25

7 M HCl

1

MT

1.09

0.01397

100.9

25

7 M HCl

0.1

MT

0.72

0.01201

17.6

85

1 M HCl

1

MT

0.15

0.01173

35.5

85

1 M HCl

0.1

MT

0.30

0.00503

1.2

85

1 M HCl

0.5

CPX

0.02

0.02612

2.7

25

1 M HCl

1

MT

0.01

0.01723

11.3

25

1 M HCl

0.1

MT

0.07

0.00935

21.1

85

0.5 M HCl

1

MT

0.23

0.00956

11.6

85

0.5 M HCl

0.1

MT

0.12

0.00295

1.5

85

0.5 M HCl

0.5

CPX

0.05

0.00377

2.2

25

0.5 M HCl

1

MT

0.06

0.01066

3.8

25

0.5 M HCl

0.1

MT

0.04

0.00943

4.3

25

0.5 M HCl

0.5

CPX

0.05

Magnetite (MT) and clinopyroxene (CPX) from samples P10–208 and P2–670, respectively, and initial Fe contents, as determined by electron microprobe analysis, are from Haynes et al. (2003). Dissolved Fe determined by Ferrozine assay

Table 3

Iron isotope data for minerals and whole-rocks from carbonatites

Sample

Mineral

Dissol.

ppm Fe

δ56Fe

δ57Fe

AFRICA

Bukusu, Uganda

BD-1477

Mt

A-1

 

0.01 ± 0.03

0.02 ± 0.04

  

A-2

 

0.01 ± 0.02

0.02 ± 0.03

Sukulu, Uganda

SU-100

WR

  

−0.52 ± 0.06

−0.76 ± 0.04

SU 103

CC

A-1

2996

−0.24 ± 0.03

−0.27 ± 0.05

  

A-2

 

−0.18 ± 0.03

−0.37 ± 0.07

 

WR

  

−0.55 ± 0.05

−0.79 ± 0.03

BD-1499

Mt

A-1

 

0.02 ± 0.03

0.00 ± 0.03

  

A-2

 

0.06 ± 0.06

0.03 ± 0.03

 

CC

A-1

14088

0.25 ± 0.03

0.38 ± 0.03

  

A-2

 

0.34 ± 0.05

0.54 ± 0.07

 

Dol

  

−0.14 ± 0.03

−0.11 ± 0.06

 

WR

  

−0.07 ± 0.03

−0.17 ± 0.04

Tororo, Uganda

BD-1485

Mt

  

0.01 ± 0.02

0.07 ± 0.03

 

CC

 

3175

−0.67 ± 0.03

−0.96 ± 0.03

 

Dol

A-1

 

−0.85 ± 0.07

−1.25 ± 0.03

  

A-2

 

−0.96 ± 0.04

−1.49 ± 0.05

TO-100B

WR

  

−0.37 ± 0.03

−0.58 ± 0.05

Toror, Uganda

DU-365

WR

A-1

 

−0.04 ± 0.06

−0.08 ± 0.03

  

A-2

 

−0.06 ± 0.03

−0.03 ± 0.05

Homa Bay, Kenya

Homa Bay

CC

A-1

4385

−0.72 ± 0.05

−1.07 ± 0.03

  

A-2

 

−0.70 ± 0.04

−1.04 ± 0.05

 

WR

  

−0.45 ± 0.04

−0.57 ± 0.06

Oldoinyo Lengai, Tanzania

OL-2 1993

Mt

  

−0.18 ± 0.05

−0.21 ± 0.04

OL-10 1993

Mt

  

−0.20 ± 0.06

−0.22 ± 0.04

BD-114

WR

A-1

 

−0.49 ± 0.04

−0.78 ± 0.04

  

A-2

 

−0.42 ± 0.04

−0.71 ± 0.07

BD-118

WR

  

−0.62 ± 0.02

−0.99 ± 0.03

Panda Hill, Tanzania

Tan-210

CC

A-1

 

0.38 ± 0.04

0.55 ± 0.04

  

A-2

 

0.30 ± 0.04

0.33 ± 0.03

Tan-212

WR

A-1

 

−0.18±0.09

−0.23 ± 0.06

  

A-2

 

−0.26 ± 0.04

−0.28 ± 0.09

Tan-213

WR

  

−0.30 ± 0.05

−0.42 ± 0.06

BD-724

Mt

A-1

 

0.15 ± 0.05

0.25 ± 0.04

  

A-2

 

0.15 ± 0.04

0.26 ± 0.04

 

CC

 

5053

−0.58 ± 0.03

−0.82 ± 0.03

Sengeri Hill, Tanzania

Sengeri Hill

WR

  

0.80 ± 0.07

1.17 ± 0.04

Dicker Willem, Namibia

5–15

Mt

A-1

 

−0.22 ± 0.02

−0.29 ± 0.04

  

A-2

 

−0.15 ± 0.08

−0.24 ± 0.03

  

A-3

 

−0.10 ± 0.05

−0.19 ± 0.07

 

CC

A-1

12348

−0.36 ± 0.08

−0.57 ± 0.03

  

A-2

 

−0.38 ± 0.05

−0.54 ± 0.03

 

Px

  

0.02 ± 0.07

0.03 ± 0.06

 

Neph

  

0.20 ± 0.03

0.28 ± 0.04

Phalaborwa, South Africa

Phalaborwa

Mt

  

−0.16 ± 0.04

−0.34 ± 0.03

 

CC

A-1

2405

−0.15 ± 0.04

−0.25 ± 0.03

  

A-2

 

−0.15 ± 0.03

−0.20 ± 0.03

NORTH AMERICA

Borden, Canada

BO-203

Mt

A-1

 

−0.12 ± 0.07

−0.16 ± 0.04

  

A-2

 

−0.17 ± 0.03

−0.27 ± 0.03

 

CC

A-1

2232

0.58 ± 0.04

0.89 ± 0.09

  

A-2

 

0.52 ± 0.04

0.66 ± 0.04

 

cpx

A-1

 

0.30 ± 0.07

0.50 ± 0.05

  

A-2

 

0.30 ± 0.03

0.44 ± 0.03

 

Ol

A-1

 

−0.17 ± 0.03

−0.19 ± 0.08

  

A-2

 

−0.11 ± 0.03

−0.15 ± 0.05

 

Phlog

  

0.25 ± 0.03

0.31 ± 0.04

Oka, Canada

P2–670

Mt

  

−0.04 ± 0.03

−0.09 ± 0.04

 

CC

A

596

−0.73 ± 0.04

−1.13 ± 0.04

  

B

 

−0.77 ± 0.08

−1.17 ± 0.04

 

Mica

  

−0.30 ± 0.04

−0.57 ± 0.05

 

cpx

  

−0.28 ± 0.05

−0.31 ± 0.03

P2–680

Mt

  

−0.19 ± 0.04

−0.32 ± 0.07

 

CC

 

176

−0.34 ± 0.02

−0.53 ± 0.03

 

Mica

  

−0.05 ± 0.04

0.01 ± 0.04

 

cpx

A-1

 

−0.18 ± 0.03

−0.24 ± 0.04

  

A-2

 

−0.10 ± 0.04

−0.20 ± 0.05

OC 203

mica

  

−0.09 ± 0.06

−0.16 ± 0.06

 

melilite

A-1

 

0.06 ± 0.04

0.01 ± 0.04

  

A-2

 

−0.02 ± 0.04

−0.04 ± 0.04

OC 302

CC

  

−0.10 ± 0.07

−0.03 ± 0.06

 

perovskite

  

−0.05 ± 0.07

0.03 ± 0.04

OC 305

Mt

A-1

 

−0.54 ± 0.05

−0.88 ± 0.04

  

A-2

 

−0.54 ± 0.02

−0.81 ± 0.04

OC 307

Bio

  

−0.11 ± 0.03

−0.13 ± 0.04

OC 310

CC

  

−0.34 ± 0.05

−0.45 ± 0.04

OC 318

px

  

−0.10 ± 0.04

−0.16 ± 0.05

OC 320

CC

  

−0.57 ± 0.10

−0.85 ± 0.06

 

Mica

A-1

 

−0.07 ± 0.04

−0.11 ± 0.05

  

A-2

 

−0.08 ± 0.04

0.04 ± 0.06

St. Honore, Canada

STH-08

Mt

  

0.08 ± 0.05

0.07 ± 0.06

 

CC

A-1

10151

−0.65 ± 0.05

−0.94 ± 0.03

  

A-2

 

−0.72 ± 0.05

−1.08 ± 0.03

 

Mica

A-1

 

−0.12 ± 0.03

−0.25 ± 0.04

  

A-2

 

−0.05 ± 0.03

−0.01 ± 0.03

 

WR

  

−0.16 ± 0.04

−0.16 ± 0.04

Magnet Cove, USA

MC-1

Spinel

  

−0.02 ± 0.06

−0.02 ± 0.03

 

CC

A

100

−0.87 ± 0.04

−1.09 ± 0.03

  

B

 

−1.12 ± 0.03

−1.46 ± 0.05

 

Mica

  

−0.48 ± 0.07

−0.59 ± 0.04

 

Mont

A-1

 

−0.37 ± 0.05

−0.48 ± 0.06

  

A-2

 

−0.38 ± 0.03

−0.55 ± 0.03

SOUTH AMERICA

Jacupiranga, Brazil

P10–202

Mt

A

 

−0.04 ± 0.04

0.07 ± 0.05

  

B

 

0.01 ± 0.03

−0.03 ± 0.04

 

CC

A-1

169

−0.47 ± 0.04

−0.68 ± 0.04

  

A-2

 

−0.49 ± 0.03

−0.60 ± 0.03

  

B

 

−0.53 ± 0.04

−0.75 ± 0.06

 

Mica

A-1

 

0.07 ± 0.03

0.10 ± 0.04

  

A-2

 

0.11 ± 0.03

0.13 ± 0.04

  

B

 

0.12 ± 0.03

0.26 ± 0.03

P10–208

Mt

A

 

0.08 ± 0.03

0.13 ± 0.03

 

CC

A-1

216

−0.35 ± 0.05

−0.52 ± 0.04

  

A-2

 

−0.30 ± 0.03

−0.38 ± 0.04

  

B

 

−0.35 ± 0.07

−0.48 ± 0.10

  

C

 

−0.34 ± 0.06

−0.52 ± 0.04

 

Mica

  

0.06 ± 0.05

0.13 ± 0.05

 

cpx

  

−0.07 ± 0.06

−0.13 ± 0.05

EUROPE-ASIA

Kaiserstuhl, Germany

K-102

CC

A-1

 

−0.65 ± 0.05

−0.94 ± 0.03

  

A-2

 

−0.60 ± 0.05

−0.91 ± 0.03

 

Mica

A-1

 

−0.19 ± 0.04

−0.27 ± 0.05

  

A-2

 

−0.13 ± 0.03

−0.22 ± 0.03

Kovdor, Russia

KO-101

Mt

A-1

 

−0.43 ± 0.05

−0.60 ± 0.03

  

A-2

 

−0.37 ± 0.03

−0.51 ± 0.04

 

CC

 

632

−0.47 ± 0.03

−0.65 ± 0.03

 

Mica

  

0.13 ± 0.04

0.28 ± 0.04

 

Pyrite

A-1

 

0.08 ± 0.04

0.17 ± 0.04

  

A-2

 

0.09 ± 0.03

0.20 ± 0.03

Sillinjarvis, Finland

SIL 106

Mt

A-1

 

−0.35 ± 0.06

−0.61 ± 0.04

  

A-2

 

−0.33 ± 0.04

−0.51 ± 0.04

 

CC

A-1

4583

0.62 ± 0.03

0.88 ± 0.03

  

A-2

 

0.63 ± 0.03

0.97 ± 0.04

 

Pyrite

  

0.34 ± 0.04

0.41 ± 0.04

 

Act

A-1

 

0.14 ± 0.05

0.21 ± 0.04

  

A-2

 

0.14 ± 0.04

0.19 ± 0.03

Dissolutions A, B, C indicate different dissolutions of same sample, including separate processing through ion-exchange columns and mass analysis. Analyses noted with “-1”, “-2”, etc. indicate repeat analysis of same solution obtained during ion-exchange chromatography, but under different days.

Parallel experiments were performed on clinopyroxene. The 85°C 7 M HCl treatment dissolved 0.06% of the Fe from a 0.5 mm clinopyroxene grain, whereas the 25°C 0.5 M HCl treatment dissolved 0.045% of the Fe from a 0.5 mm clinopyroxene grain. Although the clinopyroxene had a larger amount of Fe partially dissolved at 25°C using 0.5 M HCl as compared to magnetite, the amounts of Fe dissolved in both minerals would be too small to affect the Fe isotope composition of the carbonate. Based on these results, dissolution of a 99% pure carbonate mineral separate that contained 1% magnetite or silicate in cold 0.5 M HCl should produce less than 1% Fe contamination from magnetite or silicate in the dissolved Fe component, which would have no effect on the measured Fe isotope composition of the carbonate.

It is important to note that we did not use acetic acid (HAc) for carbonate dissolution because our previous tests showed that magnetite may undergo incongruent dissolution using HAc (Valaas-Hyslop et al. 2008), which may produce spurious Fe isotope compositions if HAc is used to selectively dissolve carbonate from a mixture of carbonate and magnetite. In addition to producing anomalous δ56Fe values during partial dissolution of magnetite using HAc, Valaas-Hyslop et al. (2008) noted that Fe(II)/FeTotal ratios changed in the solutions relative to those of magnetite, reflecting redox changes in solution through interaction with acetate, and likely formation of a new surface phase. These results were challenged by Von Blanckenburg et al. (2008), who observed no anomalous Fe isotope effects when comparing dissolution of natural whole-rock carbonates using HCl and HAc, although we note that their tests did not involve minerals of known isotopic compositions, nor did they measure Fe(II)/FeTotal ratios of solution produced by partial dissolution of magnetite. We contend that proton-promoted carbonate dissolution without the presence of organic ligands such as acetate remains the safest approach for dissolving carbonate components in the presence of minor mineral contaminants for Fe isotope analysis.

Sample processing and mass analysis

Carbonate mineral separates were digested in 0.5 M HCl, where the sample was left at room temperature for ~1 h, followed by ~20 min in an ultrasonic bath, then centrifuged. Visual inspection showed that no residue was present after centrifugation. This approach was used based on the partial dissolution experiments noted above. Silicate minerals were digested in ~5 ml of 29 M HF and ~500 μl of 7 M HNO3 in Teflon containers on hotplates for 24 h. The solution was then dried down and brought up in 5 ml of 8 M HCl. The solutions were heated on hotplates in Teflon containers for at least 12 h, and once more dried down. Magnetite mineral separates were digested in ~5 ml of 8 M HCl in closed Teflon containers on hotplates for 24 h, followed by centrifugation to remove any residual silicate minerals or other oxides that are more refractory, such as rutile.

Whole-rock powders of carbonatites required special dissolution procedures due to their very high Ca contents, which would produce extensive formation of Ca fluorides using traditional HF dissolution methods. Magnetite, if present, was removed from the samples using a hand magnet and set aside. The remaining powder was dissolved in 0.5 M HCl for ~1 h at room temperature, followed by ~20 min in an ultrasonic bath to remove the calcite fraction. Following centrifugation, the dissolved calcite fraction was set aside, and the remaining “silicate” fraction was washed in 2X H2O, followed by digestion in ~5 ml of 29 M HF and ~500 μl of 7 M HNO3 in closed Teflon containers on hotplates for 24 h. The HF-HNO3 solution was then dried down and brought up in 5 ml of 8 M HCl. The solution was heated on hotplates in closed Teflon containers for at least 12 h and once more dried down. The magnetite fraction was digested as noted above, then brought up in 0.5 M HCl. The calcite fraction and the magnetite fraction were then recombined with the “silicate” fraction to provide a whole-rock analysis.

Iron was separated by anion-exchange chromatography. Samples were passed through ion-exchange columns 2–4 times, where the greatest number of passes were required for high Ca/Fe ratio samples; previous work has shown that multiple passes do not fractionate Fe isotopes compositions because yields are ~100% (Beard et al. 2003). Isotopic compositions were determined on a Micromass IsoProbe, a single focusing, multi-collector inductively-coupled-plasma mass spectrometer with a magnetic sector mass analyzer, equipped with a Cetac Aridus desolvating micro-concentric nebulizer and Elemental Scientific spray chamber. Sample solutions contained 300 ppb Fe, were aspirated at 50 μL/min for eight minutes, resulting in consumption of 120 ng of Fe per analysis. Instrumental mass bias and drift was corrected using a standard-sample-standard approach; details can be found in Beard et al. (2003) and Albarède and Beard (2004).

The accuracy and precision of the Fe isotope measurements were assessed using analyses of standards, as well as multiple analysis of samples. Separated Fe solutions for 35 samples were analyze 2 or 3 times under different running conditions, and the average difference in the analyses is 0.04‰ in 56Fe/54Fe ratios. Five samples were dissolved two or three times, and the average difference between the multiple dissolutions, which included separate processing through ion-exchange columns, was 0.06‰ in 56Fe/54Fe ratios. These assessments of precision and accuracy match those obtained through analysis of ultra-pure Fe standards, and we estimate the average 2σ reproducibility (~2SD) to be 0.06‰ for 56Fe/54Fe ratios.

Nomenclature

Iron has four naturally occurring stable isotopes, 54, 56, 57, and 58, and isotopic compositions have been generally reported using the three major isotopes, either as 56Fe/54Fe or 57Fe/54Fe ratios. Data are reported here in standard δ notation, where:
$$ \delta^{56} {\text{Fe = }}\left[ {{{\left( {{{^{56} {\text{Fe}}} \mathord{\left/ {\vphantom {{^{56} {\text{Fe}}} {^{54} {\text{Fe}}}}} \right. } {^{54} {\text{Fe}}}}} \right)_{\text{sample}} } \mathord{\left/ {\vphantom {{\left( {{{^{56} {\text{Fe}}} \mathord{\left/ {\vphantom {{^{56} {\text{Fe}}} {^{54} {\text{Fe}}}}} \right. } {^{54} {\text{Fe}}}}} \right)_{\text{sample}} } {\left( {{{^{56} {\text{Fe}}} \mathord{\left/ {\vphantom {{^{56} {\text{Fe}}} {^{54} {\text{Fe}}}}} \right. } {^{54} {\text{Fe}}}}} \right)_{\text{standard}} - 1}}} \right. } {\left( {{{^{56} {\text{Fe}}} \mathord{\left/ {\vphantom {{^{56} {\text{Fe}}} {^{54} {\text{Fe}}}}} \right. } {^{54} {\text{Fe}}}}} \right)_{\text{standard}} - 1}}} \right] \times 10^3 $$
(1)
in units of per mil (‰), where (56Fe/54Fe)standard is taken as the average of terrestrial igneous rocks (Beard et al. 2003). Inter-laboratory comparison of Fe isotope ratios can be made by comparison to the measured iron isotope composition of the certified reference material IRMM-014, which has a δ56Fe value of –0.09‰ on the igneous rock scale. The δ57Fe value may be defined in an analogous manner using the 57Fe/54Fe ratio, and δ57Fe and δ56Fe values should be related in a mass-dependent manner.
We discuss Fe isotope fractionations between two phases, A and B, as the difference in the measured δ56Fe values:
$$ \Delta ^{{56}} {\text{Fe}}_{{{\text{A - B}}}} = \delta ^{{56}} {\text{Fe}}_{{\text{A}}} - \delta ^{{56}} {\text{Fe}}_{{\text{B}}} $$
(2)
This is an approximation to the isotope fractionation factor αA-B, which is related to ∆56FeA-B through:
$$ 10^{3} \ln \alpha _{{{\text{A - B}}}} \approx \Delta {}^{{56}}{\text{Fe}}_{{{\text{A - B}}}} $$
(3)
following standard practice. For isotopic fractionations on the order of 1–3‰, use of ∆56FeA-B introduces negligible error relative to analytical uncertainty. Finally, we compare measured Fe isotope fractionations with those predicted from theory or measured in experiment using a self-consistent set of reduced partition function ratios for 56Fe/54Fe, defined as β56/54, following standard practice. To a very good approximation, these can be related to ∆56FeA-B by:
$$\Delta ^{{56}} {\text{Fe}}_{{{\text{A}} - {\text{B}}}} \approx 10^{3} {\text{ln}}\,\,{\text{ $ \beta $ }}_{{\text{A}}} ^{{{56} \mathord{\left/ {\vphantom {{56} {54}}} \right. \kern-\nulldelimiterspace} {54}}} - 10^{3} {\text{ln $ \beta $ }}_{{\text{B}}} ^{{{56} \mathord{\left/ {\vphantom {{56} {54}}} \right. \kern-\nulldelimiterspace} {54}}} $$
(4)

It is important to note, however, that Eqs. (2) and (4) do not imply that δ56Fei is equal to 103ln βi56/54.

Results

The δ56Fe values for carbonate minerals (calcite and dolomite) from the carbonatites studied here, as well as whole-rock samples, span a significantly greater range than that defined by other samples considered to reflect mantle compositions (Fig. 1). Basaltic rocks have δ56Fe = 0.0 ± 0.05‰ (1SD) (e.g. Beard et al. 2003; Poitrasson et al. 2004; Weyer et al. 2005; Weyer and Ionov 2007). Mantle-derived xenoliths and Alpine peridotites define a slightly larger range in δ56Fe values, although most analyses lie within 0.2‰ of δ56Fe = 0.0 (Fig. 1). In contrast, the average δ56Fe value for carbonatites (carbonate minerals or whole-rocks) is clearly less than the average of basaltic or ultramafic rocks (Fig. 1), and the range in δ56Fe values spans the largest range yet measured for igneous rocks, from δ56Fe = −1.0 to +0.8‰. Remarkably, this range spans that commonly measured in low-temperature aqueous environments (e.g. Johnson et al. 2008).
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig1_HTML.gif
Fig. 1

Histograms of δ56Fe values for (a) carbonate minerals and whole-rocks from this study, (b) whole-rock samples of ultramafic rocks, and (c) minerals from ultramafic rocks. Data for (b) and (c) from Zhu et al. (2002), Beard and Johnson (2004b), Poitrasson et al. (2004), Williams et al. (2004), (2005), Weyer et al. (2005), Shultis (2006), and Weyer and Ionov (2007)

The Fe isotope compositions of different minerals in the carbonatites studied vary greatly (Fig. 2). The low-Fe content minerals such as calcite and dolomite have the greatest range in δ56Fe values, where most samples have negative δ56Fe values, but some are positive. The δ56Fe values for silicate minerals and magnetite cluster about δ56Fe = 0.0‰, although there is considerable spread relative to basaltic magmas and silicate minerals from ultramafic rocks (Figs. 1 and 2). Based on temperatures calculated using magnetite-calcite O isotope thermometry for some of the complexes we have investigated (Haynes et al. 2003), there are no clear correlations between crystallization temperature and δ56Fe values for carbonate, silicates, or magnetite. Moreover, the range in δ56Fe values for carbonates, silicates, and magnetite is larger than the range expected to be in equilibrium with the mantle, based on the Fe isotope fractionations predicted from theory or measured in experiment (Fig. 2). Although the relative order of δ56Fe values for carbonates, silicates, and magnetite that is expected based on isotopic fractionation factors is generally reflected in the average δ56Fe values measured, where δ56FeCarbonate < δ56FeSilicate < δ56FeMagnetite, the range in the measured isotopic compositions is considerably larger and does not correlate with crystallization temperatures (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig2_HTML.gif
Fig. 2

δ56Fe-crystallization temperature variations for carbonatites from this study (Table 3). Crystallization temperatures from O isotope thermometry (Haynes et al. 2003), except those that have error bars; these samples are arbitrarily plotted at a temperature of 600 ± 100°C. Shaded curved fields show predicted δ56Fe values as a function of temperature, using the β56/54 factors from Table 4. Fields defined by two end member Fe isotope compositions for magnetite or silicate δ56Fe = 0, which spans the major Fe repositories in the samples analyzed that may have been in equilibrium with mantle peridotite and hence control the Fe isotope composition of the samples and constituent minerals. Cc=calcite, Ank-ankerite, Fa=fayalite, Mt=magnetite

Discussion

Below we touch on several results of our initial Fe isotope survey of carbonatites. First, Fe isotope disequilibrium among minerals is evaluated relative to predicted or experimentally determined Fe isotope fractionation factors. Second, we compare the measured Fe isotope compositions with Li, C, O, and Sr isotope compositions determined on the same samples. Third, the Fe isotope evidence for expulsion of Fe3+-bearing fluids is discussed. Fourth, we bring the Fe isotope variations determined in this study into a model for carbonatite genesis and evolution that is consistent with current petrogenetic models for carbonatites.

Iron isotope fractionations at igneous temperatures

The wide range in measured Fe isotope fractionations among the minerals in carbonatites illustrated in Fig. 2 is explored on a sample-by-sample basis in Fig. 3. Data for minerals are cast relative to magnetite in traditional “δ-δ” plots, and isotopic fractionation lines are shown for magnetite-pyrite and magnetite-olivine, using the β56/54 factors from Table 4 and a reference temperature of 600°C. Of the 14 samples analyzed where at least two minerals were analyzed, ten samples have Fe isotope data for two or more mineral pairs that could be used to evaluate internal isotopic equilibrium. The greatest confidence in evaluating equilibrium is placed in the magnetite-olivine Fe isotope fractionation factors, where experimental determinations using the “three-isotope-method” (Shahar et al. 2008) have confirmed predicted fractionations based on theory (Polyakov and Mineev 2000; Polyakov et al. 2007). Based on Fe isotope analyses of minerals from a wide variety of silicate-dominated igneous rocks, there does not appear to be a significant Fe isotope fractionation among the various silicate minerals, at least in the case of minerals where Fe is largely Fe2+ (Polyakov and Mineev 2000; Beard and Johnson 2004b; Polyakov et al. 2007; Heimann et al. 2008), and so the magnetite-olivine fractionation curve of Shahar et al. (2008) is taken to be applicable to all fractionations between magnetite and Fe2+ silicate minerals.
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig3_HTML.gif
Fig. 3

Variations in δ56Fe values for magnetite (Mt) relative to other co-existing minerals (Cc=calcite, Dol=dolomite, Px=pyroxene, Amph=amphibole, Neph=nepheline, Py=pyrite, Ank-ankerite, Ol=olivine, Sid=siderite). Minerals from same sample connected by tie lines. Samples grouped relative to those that have equilibrium magnetite-silicate Fe isotope fractionations (a), those that do not have equilibrium magnetite-silicate fractionations (b), and those that have no independent control on Fe isotope equilibrium (c). Note in (b) that one sample analyzed is spinel (sp), not magnetite. Magnetite-olivine and magnetite-pyrite fractionation lines calculated for a temperature of 600°C, using the β56/54 factors in Table 4. Isotopic data from Table 3

Table 4

Set of self-consistent β56/54 factors

T °C

103ln β56/54

103ln β56/54

103ln β56/54

103ln β56/54

103ln β56/54

103ln β56/54

103ln β56/54

103ln β56/54

Olivinea

Magnetiteb

Sideritec

Ankerited

Calcitee

Pyritef

[FeIICl4]2-g

FeIIICl3h

300

1.15

1.76

1.12

0.72

0.32

3.20

1.09

2.17

400

0.84

1.28

0.81

0.52

0.23

2.33

0.79

1.63

500

0.63

0.97

0.62

0.39

0.16

1.77

0.60

1.29

600

0.50

0.76

0.48

0.31

0.14

1.39

0.47

1.06

700

0.40

0.61

0.39

0.25

0.11

1.12

0.38

0.89

800

0.33

0.50

0.32

0.21

0.10

0.92

0.32

0.77

900

0.28

0.42

0.27

0.17

0.07

0.77

0.27

0.68

1000

0.23

0.36

0.23

0.15

0.07

0.66

0.23

0.61

aPolyakov and Mineev (2000)

bCalculated from Shahar et al. (2008) using β56/54 from a

cPolyakov and Mineev (2000)

dPolyakov and Mineev (2000)

eThis study; calculated using magnetite β56/54 from b and 22 assuming ∆56FeMt-CC =∆56FeMt-Ank +∆56FeSid-Ank, which is equivalent to assuming ∆56FeMt-Carbonate varies linearly with Fe content, and that calcite contains very minor Fe

fPolyakov et al. (2007)

gSchauble et al. (2001)

hHill and Schauble (2008)

Of the ten samples where two or more mineral pairs were analyzed, only three samples have Fe isotope fractionations that indicate magnetite-silicate Fe isotope equilibrium (samples P2-670, STH-08, and P10-208), and these are plotted together in Fig. 3a. Six samples that have been measured for magnetite-carbonate and magnetite-silicate fractionations do not record equilibrium magnetite-silicate fractionations (samples 5–15, BO-203, P2-680, P10-202, KO-101, and SIL 106), and this group is plotted in Fig. 3b. An additional sample is plotted in this group, MC-1, where spinel, calcite, mica, and montecellite were analyzed, based on the fact that spinel-silicate fractionations are predicted to be significantly smaller than magnetite-silicate fractionations (Polyakov and Mineev 2000; Polyakov et al. 2007), a conclusion supported by comparing spinel-olivine fractionations measured in ultramafic rocks (Williams et al. 2004; Williams et al. 2005) with the magnetite-olivine fractionations experimentally measured by Shahar et al. (2008). These considerations suggest that sample MC-1 is not in Fe isotope equilibrium. Four samples analyzed for magnetite-calcite or magnetite-dolomite fractionations were not analyzed for magnetite-silicate fractionations (samples BD-1499, BD-1485, BD-724, and Phalaborwa), and hence do not have an independent check on Fe isotope equilibrium; this group is plotted in Fig. 3c.

Several samples that are not in magnetite-silicate Fe isotope equilibrium (Fig. 3b) appear to have equilibrium silicate-silicate or magnetite-pyrite fractionations. Sample 5–15 has similar magnetite-pyroxene and magnetite-nepheline fractionations, suggesting that pyroxene and nepheline are in Fe isotope equilibrium. Samples P2-680 and BO-203 have similar magnetite-pyroxene and magnetite-mica Fe isotope fractionations, indicating that the two silicates are in isotopic equilibrium. Sample MC-1 has similar spinel-mica and spinel-olivine fractionations, suggesting that although spinel is not in Fe isotope equilibrium with the silicates, the two analyzed silicates are in isotopic equilibrium with each other. Two pyrite-bearing samples analyzed (samples KO-101 and SIL 106) appear to have equilibrium or near-equilibrium magnetite-pyrite fractionations, despite the fact that magnetite, calcite, and mica are not in Fe isotope equilibrium in sample KO-101, and magnetite, calcite, and amphibole are not in Fe isotope equilibrium in sample SIL 106 (Fig. 3c).

The Fe isotope fractionations among minerals are cast relative to temperature in Fig. 4. Crystallization temperatures for some of the complexes are taken from magnetite-calcite O isotope thermometry of Haynes et al. (2003), which produces the highest O isotope temperatures and minimizes the effects of sub-solidus equilibration. It is important to note, however, that for some of the samples studied by Haynes et al. (2003), calcite-biotite O isotope temperatures were up to 270°C lower than magnetite-calcite temperatures, indicating sub-solidus cooling and isotopic exchange did occur. If independent crystallization temperatures were not available, a temperature of 600 ± 100°C was assumed in Fig. 4.
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig4_HTML.gif
Fig. 4

Variations in Fe isotope fractionations between magnetite and co-existing minerals relative to temperature (106/T2). Symbols and abbreviations as in Fig. 3. Sample grouping also follows that of Fig. 3: samples that have equilibrium magnetite-silicate Fe isotope fractionations (a), those that do not have equilibrium magnetite-silicate fractionations (b), and those that have no independent control on Fe isotope equilibrium (c). Note in (b) that one sample analyzed is spinel (sp), not magnetite. Temperatures based on O isotope thermometry determined on the same samples. For samples that do not have independent temperature constraints, temperatures were arbitrarily plotted as 600 ± 100°C. Isotopic data from Table 3. Temperature-dependent isotope fractionation curves calculated using β56/54 factors from Table 4

When viewed relative to calculated or estimated crystallization temperatures, the three samples that have equilibrium magnetite-silicate fractionations have relatively high magnetite-calcite fractionations (Fig. 4a), exceeding predicted magnetite-siderite or magnetite-ankerite fractionations (Table 4). We therefore propose an empirical magnetite-calcite Fe isotope fractionation curve using the β56/54 values in Table 4, which has not been measured in experiments nor explicitly calculated from theory, based on a) the observation that the magnetite-calcite Fe isotope fractionations for samples that are in magnetite-silicate equilibrium are generally higher than the predicted magnetite-ankerite or magnetite-siderite fractionations, and b) the predicted and experimentally measured trend of decreasing β56/54 in carbonates with decreasing Fe content (Polyakov and Mineev 2000; Johnson et al. 2005), which would produce increasing magnetite-carbonate fractionations with decreasing carbonate Fe content. When the isotopic data are viewed in the context of our proposed magnetite-calcite fractionation curve, all but one of the samples (BO-23) that do not have equilibrium magnetite-silicate fractionations also have anomalously low magnetite-calcite fractionations (Fig. 4b) Two of the four samples that do not have independent checks on Fe isotope equilibrium (group in Fig. 4c) appear to have equilibrium magnetite-calcite fractionations (samples BD-1485 and BD-724 in Fig. 4c), whereas two samples do not (BD-1499 and Phalaborwa in Fig. 4c). Finally, we note that the two pyrite-bearing samples, which crystallized at markedly different temperatures, have magnetite-pyrite fractionations that generally lie along those predicted by Polyakov and Mineev (2000) and Polyakov et al. (2007), despite their disequilibrium magnetite-silicate Fe isotope fractionations.

Co-variations in Fe, Li, C, O, and Sr isotopes

Our survey of Fe isotopes in carbonatites includes samples (same powders) previously analyzed for Li, C, O, Sr, Nd, and Pb isotopes from the studies of Bell and Tilton (2001), Haynes et al. (2003), and Halama et al. (2008). Calcium carbonate magma in equilibrium with silicate mantle should have a δ56Fe value of ~ −0.3 to −0.8‰, assuming an average silicate mantle δ56Fe value of 0.0 to −0.1‰ (Beard and Johnson 2004b; Williams et al. 2004; Weyer et al. 2005; Williams et al. 2005; Schoenberg and Von Blanckenburg 2006; Weyer and Ionov 2007), and equilibration temperatures of 400 to 1,000°C, using the β56/54 values in Table 4. That the β56/54 values for carbonates are lower than any other mineral (Table 4) indicates that crystal fractionation of non-carbonate, Fe-bearing minerals (silicates, oxides, sulfides) will move a carbonatite magma toward more negative δ56Fe values, lower than the −0.3 to −0.8‰ values that are expected for carbonate magmas in equilibrium with the silicate mantle. In Fig. 5, we compare the Fe, Li, O, and Sr isotope compositions of the carbonatites studied here with those expected for primary carbonate melts, as well as changes that may occur through magmatic differentiation and fluid interaction.
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig5_HTML.gif
Fig. 5

Variations in Sr, O, C, and Li isotope compositions, as available, for samples analyzed in this study for Fe isotope compositions. For most samples, whole-rocks were analyzed, but for some O and C isotope compositions, calcite was analyzed, and these were plotted against δ56Fe values determined on the same calcite mineral separate. Dark gray boxes denote fields for primary carbonate magma compositions. Arrows denote various processes that may change the isotopic compositions, as discussed in the text. Sr, O, C, and Li isotope data from Bell and Tilton (2001), Haynes et al. (2003), and Halama et al. (2008). Fe isotope data from Table 3

Based on Sr, Nd, and Pb isotopes in East African carbonatites, Bell and Tilton (2001) interpreted the isotopic compositions to largely reflect mixing between HIMU and EM I mantle components. Our sample suite includes the carbonatite complexes that lie closest to these end members (EM I: Homa Bay; HIMU: Sukulu), and data for the other complexes analyzed here scatter between these mantle components (Fig. 5a). Similar trends are observed for Nd and Pb isotope variations (not shown). For the carbonatites that have δ56Fe values that reflect equilibrium with the mantle (δ56Fe < −0.3‰), there is no correlation between Fe isotope compositions and Sr, Nd, or Pb isotope ratios, suggesting that Fe isotopes are not a sensitive indictor of EM I or HIMU reservoirs. This conclusion is consistent with the observation that δ56Fe values for basaltic rocks are essentially homogeneous regardless of their radiogenic isotope compositions (e.g. Beard et al. 2003; Weyer and Ionov 2007).

A larger suite of the samples analyzed here has been analyzed for C and O isotopes, including from Oka (Canada), Magnet Cove (USA), and Jacupiranga (Brazil) from Haynes et al. (2003), as well as Sukulu (Uganda), Toror (Uganda), Homa Bay (Kenya), Oldoinyo Lengai (Tanzania), Panda Hill (Tanzania), and Borden (Canada) from Halama et al. (2008). With one exception, all δ18O values for the samples analyzed for Fe isotopes fall within a restricted range that is consistent with derivation from the mantle. Sample DU-365 from Toror has an elevated δ18O value (Fig. 5b), suggesting fluid loss, which will tend to increase δ18O values (e.g. Deines 1989), although it is not clear that fluid loss would produce a sufficient increase in δ18O values. The fact that this sample has a relatively high δ56Fe value is consistent with passage of Fe3+-bearing fluids, as will be discussed below. The δ13C values of the samples analyzed for Fe isotopes vary greatly (Fig. 5c), scattering beyond the range that reflects equilibrium with the mantle. Relatively high δ13C values in carbonatites are generally interpreted to reflect extensive crystal fractionation, whereas loss of CO2 should decrease δ13C values (e.g. Deines 1989). Extensive crystal fractionation, however, should also produce an increase in δ18O values (Deines 1989), which is not generally observed in the samples studied. An alternative explanation for an increase in δ13C values but not δ18O values, is passage of CO2-bearing fluids that produced a net addition of high-δ13C carbon, rather than CO2 loss. As will be discussed below, samples that have the highest δ56Fe values are interpreted to reflect passage of Fe-bearing solutions, and if such solutions were CO2 bearing, this may be the explanation for the tendency of high δ56Fe values to be associated with high δ13C values.

Eight samples analyzed by Halama et al. (2008) for Li isotopes were analyzed in this study, and the data suggest a general negative correlation between δ7Li and δ56Fe values (Fig. 5d), although the data base is somewhat limited. Halama et al. (2008) interpreted decreasing δ7Li values, relative to mantle values, to reflect expulsion of Li-bearing fluids upon extensive crystal fractionation, and they noted a broad negative correlation between δ7Li and δ13C values that supports this interpretation. Because fluids rich in Li and Fe would most likely be chloride bearing, it is anticipated that further studies of carbonatite complexes that experienced fluid loss would have strong correlations between Li and Fe isotopes, as well as fluid inclusion evidence for chloride-bearing brines.

Evidence for interactions with Fe3+-bearing fluids

Calciocarbonatite magmas in equilibrium with the mantle must have negative δ56Fe values, most likely less than −0.3‰, posing a significant interpretive problem for the large number of carbonatites and calcites that have δ56Fe > −0.3‰. Crystal fractionation of silicates, oxides, and sulfides from carbonate magmas cannot explain δ56Fe > −0.3‰, as noted above. Although the modal abundances of silicates, oxides, and sulfides are significantly less than those of carbonate minerals in the samples studied, the majority of Fe in all of the samples lies in the non-carbonate minerals, confirming that crystal fractionation, by itself, should strongly decrease δ56Fe values in the carbonatite complexes studied here. Similarly, generation of an immiscible silicate-carbonate melt (e.g. Kjarsgaard and Hamilton 1989; Brooker 1998; Kjarsgaard 1998; Lee and Wyllie 1998) should produce negative δ56Fe values in the carbonate melt component, particularly if peralkaline silicate magmas were involved; the high alkali contents in such magmas produce high Fe3+/Fe2+ ratios in the melt due to M2+-Fe2+ and M+-Fe3+ charge balance, without significant changes in fO2, and Fe3+-bearing silicates have δ56Fe values ~0.2 to 0.4‰ higher than Fe2+-bearing silicates at igneous temperatures, based on calculated Fe isotope fractionation factors (Polyakov and Mineev 2000), as well as experiments involving Fe3+-rich silicate melts (Schuessler et al. 2007). The general effect of increasing δ56Fe values with increasing Fe3+/Fe2+ ratios in silicate minerals has been observed in igneous complexes (Schoenberg et al. 2009).

The strong negative correlation between magnetite-calcite Fe isotope fractionations and the δ56Fe value of calcite (Fig. 6a) indicates that the non-equilibrium magnetite-calcite fractionations illustrated in Figs. 3 and 4 largely reflect anomalously high δ56Fe values for calcite. In contrast, δ56Fe values for magnetite are relatively constant over the wide range of magnetite-calcite Fe isotope fractionations, although there is a weak tendency for the most disequilibrium magnetite-calcite fractionations to be associated with the lowest δ56Fe values for magnetite (Fig. 6b). We propose that passage of Fe-bearing fluids through carbonate magmas or crystallized minerals is the most likely explanation for the anomalously high δ56Fe values for calcite. The very low Fe contents of calcite makes this mineral particularly sensitive to modification by Fe-bearing fluids, and hence can explain the strong correlation between the δ56Fe values for calcite and the magnetite-calcite fractionations (Fig. 6). Based on calculated β56/54 factors for Fe- and Fe-bearing fluids (Table 4), however, the δ56 Fe values of calcite should decrease upon loss of a Fe-bearing fluid. Although Heimann et al. (2008) documented evidence for increasing the δ56 Fe values of evolved, metaluminous silicate magmas through loss of an Fe-bearing fluid, this model cannot explain the increase in δ56Fe values for calcite in carbonatites because, under equilibrium conditions, the δ56Fe values of calcite are lower than those of Fe2+- or Fe3+-bearing fluids, silicates, or oxides (Table 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig6_HTML.gif
Fig. 6

Variation in magnetite-calcite Fe isotope fractionations (Δ56FeMt-Cc) with δ56Fe values for calcite (a) and magnetite (b). Gray box indicates Δ56FeMt-Cc fractionations and δ56FeCc and δ56FeMt values that would be in equilibrium with the mantle at igneous temperatures. Equilibration temperatures for calcite-magnetite fractionation shown as dashed lines, as calculated using β56/54 factors from Table 4. Note that one sample, spinel (sp) was analyzed instead of magnetite. Symbols indicate groupings of data according to magnetite-silicate Fe isotope equilibrium (samples in Figs. 3a and 4a), magnetite-silicate disequilibrium (samples in Figs. 3b and 4b), or no independent control on Fe isotope equilibrium (samples in Figs. 3c and 4c). The strong negative correlation indicates that the Δ56FeMt-Cc fractionations are controlled by the δ56FeCc values. Three models for fluid-rock interaction shown, where arrows indicate direction of increasing fluid/rock ratios from zero (in gray box) to 10 (end of arrow) for Models A, B, and C of Table 5

A simple fluid-rock mixing model (e.g. Criss 1999) may explain the trend of decreasing magnetite-carbonate fractionations with increasing δ56Fe values for calcite (Fig. 6), reflecting net addition of Fe that had high-δ56Fe values, rather than fluid loss. A mixing model that incorporates a range of Fe contents in calcite and fluid, as well as isotopic compositions (Table 5), can produce the spread in δ56Fe values for calcite. Note that the much higher Fe contents of magnetite, relative to calcite, results in no significant change in the δ56Fe values of magnetite in the fluid-rock mixing model (Fig. 6b). A fluid-rock mixing model that reflects a net addition of Fe through passage of Fe-bearing fluids is consistent with a weak correlation between calcite Fe contents and δ56Fe values (not shown; data in Table 3). The fluid-rock mixing model that most successfully explains the data uses a relatively high δ56Fe value for the Fe-bearing fluid (model C in Fig. 6 and Table 5), which, at igneous temperatures, would most likely be an Fe3+-bearing fluid based on β56/54 values (Table 4). In this model, an Fe2+-bearing fluid cannot explain the wide range in δ56Fe values of calcite, even at very high fluid/rock ratios, given the lower β56/54 factors relative to those of Fe3+-bearing fluids. Invoking an Fe3+-bearing fluid is consistent with the peralkaline nature of carbonatite-silicate systems, where Fe3+/Fe2+ ratios are relatively high as compared to metaluminous silicate magmas, the common occurrence of Fe3+ oxides (magnetite, hematite) in carbonatites, and the abundance of Fe3+-oxides in the fenites that surround intrusive carbonatite complexes.
Table 5

Parameters for fluid-rock interaction model

Model

Initial δ56FeCalcite

Initial Fe concentration calcite (ppm)

Initial δ56FeFeCl3

Initial Fe concentration fluid (ppm)

δ56FeMagnetite

A

−0.75

10,000

0.27

400

0.1

B1

−0.75

100

0.27

400

0.0

C

−0.75

100

0.56

10,000

−0.1

Magnetite Fe concentration assumed to be stoichiometric, at 724,138 ppm. Initial δ56FeFeCl3 = +0.27 reflects predicted composition at 800°C, and initial δ56FeFeCl3 = +0.56 reflects predicted composition at 600°C, using β56/54 factors from Table 4

The slight decrease in δ56Fe values for magnetite for samples that have high δ56Fe values for calcite (Fig. 6) remains a puzzle. This weak trend cannot be explained by a net addition of a high-δ56Fe, Fe3+-bearing fluid. Crystallization of magnetite from a high-δ56Fe carbonate magma, if in isotopic equilibrium with calcite, would produce high δ56Fe values for magnetite. One possible explanation is that magnetite in the high-δ56 Fe calcite samples equilibrated with an Fe3+-rich fluid at low temperatures, which could potentially produce a decrease in δ56Fe values for magnetite, given the relative β56/54 factors (Table 4). In such a model, however, calcite must maintain isotopic disequilibrium with the fluid and magnetite. It is also possible that the calcite and magnetite reflect, in part, physical mixtures of phenocrysts from unrelated magmas. A deeper understanding of the Fe isotope exchange kinetics of carbonate, oxide, and fluid is required to test these models, as well as experimental confirmation of the Fe isotope fractionations between Fe3+- and Fe2+-bearing fluids, oxides, and carbonates at igneous temperatures.

Fe isotope constraints on carbonatite genesis and evolution

We bring the above discussion into a summary model for carbonatite genesis and evolution in terms of Fe isotope variations in Fig. 7. This model is consistent with current petrogenetic models for carbonatites. Generation of carbonate magmas in the mantle, either in a plume or the lithosphere, will produce negative δ56Fe values, certainly < −0.3‰ at mantle temperatures, but probably not lower than −0.8‰ at relatively low temperatures, as discussed above. It is possible that slightly lower δ56Fe values for carbonatite magma may be produced if melting of a previously carbonated peridotite (e.g. Dalton and Presnall 1998; Wyllie and Lee 1998; Yaxley et al. 1998) occurred, assuming such a peridotite had slightly lower δ56Fe values due to addition of low-δ56Fe value carbonate. If mafic silicate magmas were associated with carbonatite genesis, these should have δ56Fe values of ~0.0 ± 0.1‰, given the relatively homogenous nature of basaltic magmas in terms of Fe isotope compositions (e.g. Beard et al. 2003; Schoenberg and Von Blanckenburg 2006; Weyer and Ionov 2007). Crystal fractionation in carbonatite magmas (e.g. King and Sutherland 1960; Le Bas 1989; Simonetti and Bell 1994b) will produce a decrease in δ56Fe values for carbonate, and this is a likely explanation for the samples that have the most negative δ56Fe values. Generation or evolution of carbonatite magmas through liquid immiscibility (e.g. Lee and Wyllie 1998) will also decrease δ56Fe values in carbonatites, given the positive silicate-carbonate fractionation factors at igneous temperatures, and this mechanism will exert the greatest Fe isotope effect at low temperatures, where the fractionation factors are relatively large, or if a peralkaline, Fe3+-bearing silicate magma is involved. In a magmatic system where carbonate and silicate magmas coexist, the Fe isotope compositions of the carbonate magma should be shifted greatest because of its relatively low Fe contents, whereas the δ56Fe values of the silicate magmas would likely remain close to zero to the degree that the majority of the Fe is contained in the silicate component of the system.
https://static-content.springer.com/image/art%3A10.1007%2Fs00710-009-0055-4/MediaObjects/710_2009_55_Fig7_HTML.gif
Fig. 7

Cartoon illustrating our preferred model for Fe isotope variations in carbonatites. Silicate magmas in equilibrium with the mantle will have δ56Fe values near zero, whereas carbonatites in equilibrium with the mantle will have δ56Fe values ≤ −0.3‰; these compositions apply to generation of carbonatites in plume environments or the lithospheric mantle. Crystal fractionation or liquid immiscibility will tend to move carbonatite magmas toward more negative δ56Fe values, reaching values as low as ~ −1.0‰, whereas these processes will not significantly change the δ56Fe values of related silicate magmas. Residence of crystallizing carbonatite magmas in the upper crust will release fluids through the outer portions of the intrusive complexes, as well as the surrounding country rocks, producing fenite zones. Relatively high δ56Fe values for carbonates in carbonatites require addition of a high-δ56Fe fluid, which would most likely be an Fe3+-bearing fluid, given the relatively high β56/54 factors for Fe3+ fluids (Table 4). Based on fluid-rock modeling, and currently available β56/54 factors, the δ56Fe values of calcite-rich carbonatites might be increased up to +0.4‰ by passage of Fe3+-rich fluids

At high levels in the crust, continued crystallization of carbonatite magmas will produce a free fluid phase, which, based on the discussion above, is likely to be chloride- and Fe-rich. At Oldoinyo Lengai, for example, the natrocarbonatites show that Cl, F, and S are probably constituents of carbonatite fluids, and the mineralogy of the fenites indicates that the fluid must have contained Ca, Mg, K, Na, and Fe3+ (e.g. Morogan and Martin 1983; Gittins 1989). An NaCl brine was considered to be the fenitizing fluid at Callander Bay, Ontario, Canada (Currie and Ferguson 1971). The abundance of acmitic pyroxene, riebeckite, and hematite in fenites suggests that fenitization occurred at a depth where the fenitizing fluids had a relatively high f O2, probably close to the hematite-magnetite buffer (Gittins 1989). Fenitization of the surrounding granite terrane at the Iron Hills carbonatite, Colorado, was attributed to an early- and a late-stage fluid. Fluid evolution, based mainly on pyroxene and amphibole, began with high Mg/Fe ratios and high total Fe and Fe 3+, relative to the unaltered country rock, and as the Mg/Fe ratio decreased, high Fe 3+ aegerine-augite and aegerine were formed (White-Pinilla 1996). All fluids from Iron Hills had overlapping temperatures between 510 and 560°C (Lowers 2005).

Loss of Fe-rich fluids provides an additional means for producing low-δ56Fe values in the remaining carbonate magma because Fe-bearing fluids, particularly those that are Fe3+-rich, will have relatively positive δ56Fe values (Table 4). We propose, however, that the high δ56Fe values in calcite were produced not by fluid loss but by net addition of high-δ56Fe iron through passage of Fe3+-rich brines, increasing δ56Fe values in calcite up to +0.4‰ (circular inset in Fig. 7). These brines must have passed through the carbonatites at relatively high temperatures to retain their high O isotope temperatures (Haynes et al. 2003), as there is no correlation between δ56FeMt-CC fractionations and O isotope temperatures (Fig. 4). Assuming a simple intrusion geometry, our proposal predicts a concentrically zoned carbonatite complex, where the inner zones have relatively low δ56Fe values, and the outer parts of the carbonatites, as well as the fenites surrounding the carbonatites, have relatively high δ56Fe values. In our broad survey, there is insufficient sampling at any single complex to test this proposal, nor are Fe isotope data available for fenites. Such studies would be fruitful lines of future research that would provide a full understanding of the Fe isotope mass balance of carbonatite complexes and their associated alteration and mineralization zones.

Conclusions

Previous stable isotope studies of carbonatites, including Li, C, and O, have provided constraints on crystallization temperatures and fluid-rock interactions. Iron isotope variations in carbonatites suggest that this relatively new stable isotope system can also provide a tracer of fluid interactions and cooling history. Many minerals in the carbonatites studied are out of Fe isotope equilibrium at igneous temperatures, and, considering the large contrasts in Fe contents among carbonate, silicate, oxide, and sulfide minerals, isotopic disequilibrium likely reflects the effects of cooling and fluid/rock interaction. Iron isotope disequilibrium may also record mixing of phenocrysts from distinct magmas. Iron isotopes, therefore, provide additional evidence for isotopic disequilibrium in carbonatite magmas, which previous stable (e.g. Haynes et al. 2003) and radiogenic (e.g. Simonetti and Bell 1993) isotope studies have demonstrated. Because Fe isotope fractionations between fluids and minerals/magmas are sensitive to oxidation state, Fe isotopes represent a stable isotope system that is uniquely poised to investigate the redox state of carbonatite magmas and their fluids. Moreover, the very large range in Fe contents of carbonates, oxides, silicates, and sulfides provides a means to investigate differential mass-balance responses to differentiation and fluid-interaction processes. As our understanding of Fe isotope fractionation factors improves, as well as Fe diffusion rates in minerals, the exact mechanisms responsible for producing Fe isotope distributions among carbonatite minerals and whole-rocks will be better understood.

Acknowledgments

C.M.J., B.L.B., and A.I.S. thank the organizers of this special volume in honor of our co-author Keith Bell, including guest editor Antonio Simonetti. This work was supported by the Department of Geology and Geophysics (U.W. Madison), the Geological Society of America, and the National Science Foundation (grant EAR-0525417). In addition to samples in the collection of K.B., samples were provided by J.B. Dawson, D. Moecher, E. Haynes, and M. Spicuzza. Journal reviews were provided by R. Schoenberg and an anonymous reviewer, whose comments helped to improve the manuscript. We thank A. Simonetti for editorial handling of the paper.

Copyright information

© Springer-Verlag 2009