doi: /j.gca

Transcription

1 doi: /j.gca Geochimica et Cosmochimica Acta, Vol. 69, No. 4, pp , 2005 Copyright 2005 Elsevier Ltd Printed in the USA. All rights reserved /05 $ Experimental constraints on Fe isotope fractionation during magnetite and Fe carbonate formation coupled to dissimilatory hydrous ferric oxide reduction CLARK M. JOHNSON, 1, *ERIC E. RODEN, 2 SUSAN A. WELCH, 1, and BRIAN L. BEARD 1 1 Department of Geology and Geophysics, University of Wisconsin, Madison, WI 53705, USA 2 Department of Biological Sciences, The University of Alabama, Tuscaloosa, AL 35487, USA (Received October 17, 2003; accepted in revised form June 29, 2004) Abstract Iron isotope fractionation between aqueous Fe(II) and biogenic magnetite and Fe carbonates produced during reduction of hydrous ferric oxide (HFO) by Shewanella putrefaciens, Shewanella algae, and Geobacter sulfurreducens in laboratory experiments is a function of Fe(III) reduction rates and pathways by which biogenic minerals are formed. High Fe(III) reduction rates produced 56 Fe/ 54 Fe ratios for Fe(II) aq that are 2 3 lower than the HFO substrate, reflecting a kinetic isotope fractionation that was associated with rapid sorption of Fe(II) to HFO. In long-term experiments at low Fe(III) reduction rates, the Fe(II) aq -magnetite fractionation is 1.3, and this is interpreted to be the equilibrium fractionation factor at 22 C in the biologic reduction systems studied here. In experiments where Fe carbonate was the major ferrous product of HFO reduction, the estimated equilibrium Fe(II) aq -Fe carbonate fractionations were ca. 0.0 for siderite (FeCO 3 ) and ca. 0.9 for Ca-substituted siderite (Ca 0.15 Fe 0.85 CO 3 ) at 22 C. Formation of precursor phases such as amorphous nonmagnetic, noncarbonate Fe(II) solids are important in the pathways to formation of biogenic magnetite or siderite, particularly at high Fe(III) reduction rates, and these solids may have 56 Fe/ 54 Fe ratios that are up to 1 lower than Fe(II) aq. Under low Fe(III) reduction rates, where equilibrium is likely to be attained, it appears that both sorbed Fe(II) and amorphous Fe(II)(s) components have isotopic compositions that are similar to those of Fe(II) aq. The relative order of 56 Fe values for these biogenic minerals and aqueous Fe(II) is: magnetite siderite Fe(II) aq Ca-bearing Fe carbonate, and this is similar to that observed for minerals from natural samples such as Banded Iron Formations (BIFs). Where magnetite from BIFs has 56 Fe 0, the calculated 56 Fe value for aqueous Fe(II) suggests a source from midocean ridge (MOR) hydrothermal fluids. In contrast, magnetite from BIFs that has 56 Fe 0 apparently requires formation from aqueous Fe(II) that had very low 56 Fe values. Based on this experimental study, formation of low- 56 Fe Fe(II) aq in nonsulfidic systems seems most likely to have been produced by dissimilatory reduction of ferric oxides by Fe(III)-reducing bacteria. Copyright 2005 Elsevier Ltd 1. INTRODUCTION It has been recognized for over a decade that bacteria can derive energy through respiration of transition metals bound in natural minerals (Lovley et al., 1987; Myers and Nealson, 1988), and the evidence now seems clear that biologic processing of redox-sensitive metals such as Fe is ubiquitous in surface- and near-surface environments. Microbial Fe(III) reduction may have been associated with some of the earliest respiratory metabolisms on Earth (e.g., Vargas et al., 1998). Mineralogical products of dissimilatory Fe(III) reduction that may be recorded in the rock record include magnetite and Fe carbonates generated during Fe(III) reduction (e.g., Lovley et al., 1987; Roden and Lovley, 1993). Iron may also be incorporated as intracellular minerals, such as magnetite formed by magnetotactic bacteria (Blakemore and Frankel, 1989). In terms of the total Fe that is processed in surface and near-surface environments, dissimilatory Fe(III) reduction (together with Fe(II) oxidation) accounts for many orders of magnitude greater Fe cycling than assimilatory metabolism or * Author to whom correspondence should be addressed (clarkj@ geology.wisc.edu). Present address: Department of Geology and Research School of Earth Sciences, The Australian National University, Canberra, ACT, 0200 Australia. 963 production of intracellular minerals (e.g., Nealson and Stahl, 1997). Metabolic processing of Fe produces significant Fe isotope fractionations (e.g., Johnson et al., 2004). Experimental investigations of Fe isotope fractionation include studies of dissimilatory Fe(III) reduction by bacteria (Beard et al., 1999, 2003a; Icopini et al., 2004), anaerobic photosynthetic Fe(II) oxidation (Croal et al., 2004), ion-exchange chromatography (Anbar et al., 2000; Beard et al., 2003a; Roe et al., 2003), the role of ligands during mineral dissolution (Brantley et al., 2001, 2004), abiotic precipitation of ferric oxides or oxyhydroxides (Bullen et al., 2001; Skulan et al., 2002), sorption of aqueous Fe(II) to goethite (Icopini et al., 2004), and abiotic exchange between Fe(III) and Fe(II) in solution (Matthews et al., 2001; Johnson et al., 2002; Welch et al., 2003). Significant Fe isotope variations are generally restricted to relatively low-temperature systems, including hydrothermal fluids and chemically precipitated minerals (Beard and Johnson, 1999, 2004a; Beard et al., 1999, 2003b; Zhu et al., 2000; Bullen et al., 2001; Sharma et al., 2001; Johnson et al., 2003; Matthews et al., 2004). Magnetite and siderite are minerals that are closely associated in, for example, Banded Iron Formations (BIFs), and these minerals define one of the largest contrasts in Fe isotope compositions in BIFs (Johnson et al., 2003). The mechanisms by which dissimilatory Fe(III) oxide reduc-

2 964 C. M. Johnson et al. tion occurs appear to vary among common dissimilatory metalreducing bacteria (DMRB) such as Geobacter and Shewanella. For example, Fe reduction rates may be strongly affected where the solid-phase Fe(III) substrates are physically isolated from cells, which led Lloyd et al. (1999) and Nevin and Lovley (2000, 2002a,b) to conclude that direct contact between cell and mineral surfaces may be required for the strict anaerobes such as Geobacter sulfurreducens and Geobacter metallireducens. Other DMRB may reduce Fe(III) minerals without direct contact, and a number of workers have proposed that ferric iron may be solubilized by chelating compounds, or reduced indirectly by electron shuttling compounds such as extracellular quinones or humic substances (e.g., Lovley et al., 1996; Newman and Kolter, 2000; Hernandez and Newman, 2001; Shyu et al., 2002). Metal oxide isolation experiments using Geothrix fermentans (also a strict anaerobe) or Shewanella algae (a facultative anaerobe) showed that direct cell-oxide contact was not required for Fe(III) oxide reduction (Nevin and Lovley, 2002a,b), and, in fact, ca. 10% of the soluble Fe in solution existed as Fe(III) in these experiments. It is therefore possible that different mechanisms involved in Fe(III) oxide reduction by different DMRB may produce distinct Fe isotope fractionations. In this contribution, we present the results from six experiments that constrain the Fe isotope fractionations among aqueous Fe(II), magnetite, and Fe carbonate that are produced during dissimilatory Fe(III) oxide reduction by Geobacter and Shewanella. The isotopic effects of other intermediate solid and sorbed species were also investigated. The experimentally determined isotopic fractionations allow constraints to be placed on the role of DMRB in producing Fe isotope variations in the rock record, including magnetite- and siderite-rich layers in Banded Iron Formations. 2. MAGNETITE AND Fe CARBONATE IN LOW- TEMPERATURE ENVIRONMENTS Formation of magnetite under diagenetic conditions has been well studied (e.g., Tarling and Turner, 1999). In many marine environments, particularly where bacterial sulfate reduction is active, detrital magnetite is partially dissolved and converted to Fe sulfides (e.g., Canfield and Berner, 1987; Karlin, 1990a,b; Leslie et al., 1990). In some instances, however, magnetite is produced by diagenetic reactions during breakdown of smectite (Katz et al., 1998, 2000). In addition to these burial diagenetic reactions, magnetite can be produced at or near the sedimentwater interface by microorganisms. For example, magnetite produced by magnetotactic bacteria has been proposed to occur in rocks spanning ages from the Proterozoic (Gunflint Formation) to modern sediments (e.g., Frankel et al., 1979, 1981; Chang and Kirschvink, 1985; Chang et al., 1989). The rate of magnetite production on a per cell basis by dissimilatory Fe(III)-reducing bacteria is ca times greater than that at which magnetotactic bacteria produce magnetite (Konhauser, 1998). Although magnetite production by dissimilatory Fe(III)- reducing bacteria is common in laboratory experiments, relatively few examples exist in modern environments where magnetite formation can be clearly linked to DMRB (Karlin et al., 1987; Eggar-Gibbs et al., 1999). The most important repository of magnetite formed at low temperature lies in the Banded Iron Formations (BIFs). Wellknown occurrences include the 3.8 Ga Isua, 2.5 Ga Transvaal- Hamersley, Ga Superior-Labrador Trough, and 0.7 Ga Rapitan sequences (e.g., Klein and Beukes, 1992). The average oxidation state of fresh Archean through mid-proterozoic BIFs is Fe 2.4, significantly higher than that of igneous rocks (Klein and Beukes, 1992). In the absence of secondary (supergene) alteration, or significant metamorphism, the most common oxide mineral in BIFs is magnetite, which many workers have argued is either a primary precipitate or early diagenetic recrystallization product of fine-grained magnetite or ferric and ferrous hydroxide precursors (e.g., Klein, 1974, 1983). In the absence of sulfate reduction, magnetite is stable to very low Eh conditions, which probably explains the excellent preservation of magnetite in non- or weakly metamorphosed BIFs (Klein, 1974, 1983). The absence of sulfide minerals in oxide and carbonate zones of well-preserved BIFs such as the Transvaal- Hamersley suites suggests that if magnetite formed in deep, relatively anoxic marine settings (e.g., Beukes et al., 1990), it remained stable during diagenesis. Iron carbonate is a common early diagenetic phase in sedimentary rocks (e.g., Mozley and Burns, 1993; Macquaker et al., 1997; Uysal et al., 2000; Raiswell and Fisher, 2000). In general, Fe carbonate is formed in anoxic diagenetic environments where the rate of Fe(III) reduction is greater than that of sulfate reduction (Pye et al., 1990; Coleman, 1993). Siderite formation has been proposed to commonly require biotic mediation to provide the source of both Fe(II) and carbonate through oxidation of organic carbon coupled to dissimilatory Fe(III) reduction (Coleman, 1993; Coleman et al., 1993). Many authigenic Fe carbonates typically have much higher Mg and Ca contents than can be accommodated if they formed in thermodynamic equilibrium at low temperatures (e.g., Mozley and Carothers, 1992; Laverne, 1993; Baker et al., 1995; Hendry, 2002). In addition, experimental studies using Geobacter metallireducens to produce siderite from fluids that contained variable Ca to Mg ratios and constant Fe concentration show that Ca incorporation is rate dependent but that Mg incorporation is rate independent (Mortimer et al., 1997), suggesting that a wide range of metastable Fe carbonate compositions may be produced by DMRB. 3. METHODS We briefly describe the experimental design and analytical methods below. The reader is referred to electronic annex EA-1. Six experiments were performed, involving runs from several weeks to ca. 500 d. In all cases, the ferric Fe source was ferrihydrite, which is commonly referred to as hydrous ferric oxide (HFO). All experiments were conducted at room temperature, ca. 22 C. A summary of the six experiments, noting both conditions and products, is given in Table Isotopic Data Iron isotope compositions were determined using a multi-collector ICP-MS (GV Instruments IsoProbe) on ppb solutions obtained after Fe separation by ion-exchange chromatography. Instrumental mass bias was corrected using a standard-sample-standard approach. Fe isotope variations are reported in standard d notation, in units of per mil (parts per 1000 or ): 56 Fe ( 56 Fe 54 Fe SAMPLE 56 Fe 54 Fe BULK EARTH 1)10 3 (1) where 56 Fe/ 54 Fe BULK EARTH is defined by a wide variety of terrestrial

3 Fe isotope fractionation in biogenic magnetite and siderite 965 Table 1. Summary of experiments conducted. Fe reduction rates [%Fe(II)/day] Solid products Initial Final Rate constant R 2 Duration (d) Initial cell density (ml 1 ) Growth media Experiment Bacteria 1A Shewanella putrefaciens, strain CN Bicarbonate-buffer, with NTA 485 n/a n/a n/a n/a Siderite (FeCO 3 ) 1B Shewanella putrefaciens, strain CN Bicarbonate-buffer, with NTA, 485 n/a n/a n/a n/a Ca-substituted siderite (ca. 10 mm Ca Ca 0.15 Fe 0.85 CO 3 ) Initial conversion of HFO to lepidocrocite, followed by magnetite as the sole solid phase at end of experiment 2 Geobacter sulfurreducens 10 7 Acetate/fumarate growth medium, with NTA Magnetite NMNC Fe(II)(s) 3 Shewanella algae 10 7 Lactate/fumarate growth medium, with NTA 4 Geobacter sulfurreducens 10 8 Bicarbonate-buffered, with NTA Magnetite tr. siderite NMNC Fe(II)(s) 5 Geobacter sulfurreducens 10 8 Bicarbonate-buffered, with NTA Magnetite siderite NMNC Fe(II)(s) 6 Geobacter sulfurreducens 10 8 Bicarbonate-buffered, without Magnetite siderite NMNC Fe(II)(s) NTA Determination of solid products reflect combined results of ferric and ferrous Fe assays, inorganic carbon contents, SEM and TEM imaging, and slow-scan XRD spectra. Fe reduction rates expressed as % Fe(II)/day, where % Fe(II) is relative to the total Fe(II) produced at the end of the experiment. Rates calculated from regression of total system Fe(II) contents versus time, using a first-order rate law (eq. 3; see text). Reduction rates calculated from eq. 4 (see text). NTA is a metal-chelating compound, nitrilotriacetic acid. NMNC Fe(II)(s) is defined as an amorphous, non-magnetic, non-carbonate Fe(II) solid. and lunar igneous rocks that have 56 Fe (Beard et al., 2003a). 57 Fe values may be defined in an analogous manner, based on 57 Fe/ 54 Fe ratios. When describing Fe isotope fractionations between coexisting phases A and B, we follow the traditional definitions such as: A B 56 Fe A 56 Fe B 10 3 In A B, (2) where A-B is the isotope fractionation factor in terms of 56 Fe/ 54 Fe ratios. The external precision (1 SD) of 56 Fe/ 54 Fe and 57 Fe/ 54 Fe ratios is 0.05 and 0.07, respectively, based on 153 analyses of standards during the study. In addition, 99 repeat measurements were made of the experimental products, including 1) reanalysis of separated Fe, and 2) duplicate processing of samples through ion-exchange chromatography; the average reproducibility obtained for both sets of duplicates was The integrity of the anaerobic aqueous Fe(II) solutions produced from the experiments was evaluated by acidification of the original solutions after 6 months, followed by chemical separation and isotopic analysis. This test was designed to evaluate possible oxidation and precipitation in the solutions, and the average reproducibility for these duplicates was 0.10, indicating that no significant oxidation and precipitation occurred between the time the samples were taken and isotopic analysis. The consistency of the experimental runs was tested using duplicate bottles for experiments 5 and 6, each of which contained identical growth media, cultures, and ferric Fe substrate, and the average reproducibility of 56 Fe/ 54 Fe ratios for aqueous Fe(II) that was sampled at identical times was Overview of Experimental Design Experiment 1 was designed to measure Fe isotope fractionation in a system in which Fe carbonate was the primary solid-phase end product of Fe(III) reduction. Recent studies have demonstrated conversion of HFO to siderite and other Fe carbonate phases by Shewanella putrefaciens strain CN32 (Fredrickson et al., 1998; Parmar et al., 2000; Roden et al., 2002), a dissimilatory Fe(III)-reducing bacterium isolated from a core sample (250 m depth) of a shale-sandstone sequence in northwestern New Mexico. This organism was therefore used to generate biogenic siderite for Fe isotope analysis. In addition, the isotopic effects of Ca substitution into siderite were assessed in HFO reduction cultures amended with 10 mm CaCl 2. Experiments 2 and 3 were similar in initial cell density and were designed to compare dissimilatory Fe(III) reduction by G. sulfurreducens (acetate/fumarate-growth media) and S. algae (lactate/fumarategrowth media), respectively, without a carbonate buffer so that Fe carbonate formation was suppressed (Caccavo et al., 1992). The motivation for comparing G. sulfurreducens and S. algae was based on recent findings (Nevin and Lovley, 2002a) which indicate that the latter organism, in contrast to members of the Geobacteraceae (Nevin and Lovley, 2000), has the potential to produce both Fe(III) chelators and electron-shuttling compounds during growth on solid-phase Fe(III) oxides. Production of such compounds by S. algae, particularly Fe(III) chelating compounds, might alter the temporal pattern and/or extent of Fe isotope fractionation during HFO reduction/biomineralization, as compared to G. sulfurreducens. The last three experiments involved relatively high initial cell densities in a carbonate buffer, and were designed to evaluate the effects of formation of coexisting magnetite and Fe carbonate. Experiment 4 used G. sulfurreducens at an initial cell density that was 10 times greater than that of experiments 2 and 3. Experiments 5 and 6 also used G. sulfurreducens and were additionally designed to compare the effects of the presence or absence of a metal-chelating compound, nitrilotriacetic acid (NTA), which is commonly used in experimental studies of DMRB. The presence of chelators such as EDTA and NTA is known to influence the rate and extent of Fe(III) oxide reduction by DMRB, both through complexation or solubilization of Fe(III) (Arnold et al., 1988; Lovley et al., 1994; Lovley and Woodward, 1996) and complexation of biogenic Fe(II) (Urrutia et al., 1998; Royer et al., 2002). Because such compounds are not typically present in natural sedimentary environments, it was important to determine whether the presence of small quantities of NTA in our standard culture medium imparted

4 966 C. M. Johnson et al. any unique effects on Fe isotope fractionation that would not be expected to occur in natural systems Characterization of Solid Products The solid phase assemblages were characterized through a combination of partial and total dissolution of the solid phase and subsequent analysis of Fe(II) and Fe(III) contents using the Ferrozine method (Stookey, 1970), measurement of inorganic carbon (IC) contents, total Fe and Ca contents (by AA and EDS detection), slow-scan ( deg/min) XRD spectra, and SEM and TEM imaging. All solid products were washed in the same buffer that was used in the growth media to remove interstitial aqueous Fe. Isolation of different solid phase components for isotopic analysis was done through partial dissolution in HCl and HAc, accompanied in most cases by Fe(II) and Fe(III) assays of the dissolved component. HAc was preferentially used where solid products were dominated by easily dissolved carbonates and poorly crystalline Fe(II) solids. Because partial dissolution using acids generally occurs congruently, no isotopic fractionation is expected, as has been shown for partial dissolution of hematite in HCl (Skulan et al., 2002; Beard et al., 2003a; Johnson et al., 2004). We further evaluated the effects of partial dissolution using an igneous magnetite (98LH-7; 56 Fe ; Beard and Johnson, 2004b), which was partially dissolved in 1 M and 3 M HCl over time periods of 2 to 21 h (0.2 to 5.7% dissolution), and these tests produced 56 Fe values of the dissolved component that were indistinguishable (average 56 Fe ; n 6) from that of the bulk magnetite. In addition, Ferrozine assays showed that the % Fe(II) of magnetite (33.3%) were successfully recovered when partial dissolution used 1Mor3MHCl. These tests support the expectation that partial dissolution of oxides, and probably carbonates, under low-ph conditions occurs congruently and successfully recovers the Fe isotope composition of the dissolved component; this is distinct from the significant isotopic fractionations that occur during very small extents ( 0.1%) of incongruent dissolution of silicate minerals in the presence of strong organic ligands (Brantley et al., 2001, 2004). The proportions of ferric and ferrous Fe released during partial dissolution provides important constraints on the solid phases that were dissolved. Where significant Fe(III) was released, there remain ambiguities as to the sources of this Fe(III), which may include unreacted ferric hydroxides, magnetite, or mixed Fe(III)-Fe(II) solid phases such as green rust. The results of partial dissolutions that produced a high proportion of Fe(II) are therefore most straightforward to interpret. Interpretive ambiguities notwithstanding, the results of partial dissolutions of the solid products of experiments that produced complex solid phases provide preliminary insights into the isotopic effects of various components and pathways involved in HFO biomineralization. 4. RESULTS A summary of the experiments, Fe reduction rates, and their products is given in Table 1. Tables 2 through 7 report isotopic data for the six experiments (bulk solids and partial dissolutions), the results of wet-chemical assays for Fe(III), Fe(II), and inorganic carbon contents that were used to constrain the solidphase assemblages, as well as summaries of extensive XRD, SEM, and TEM observations. Representative SEM and TEM images of reaction end-products are available in electronic annex EA-2; these images document changes in morphologies of the solid products, as well as important phase transitions that may affect Fe isotope fractionations Fe(III) Reduction Rates Production of aqueous and total Fe(II) was monitored over time during experiments 2 6. HFO reduction ceased after ca. 100 d in experiments 2 and 3, 50 d in experiment 4, and 10 d in experiments 5 and 6. No time course measurements were collected in experiment 1, where complete or near complete conversion of HFO to Fe carbonate occurred within one month based on visual inspection. The accumulation of total Fe(II) over time in experiments 2 6 could be well-described by the integrated form of a first-order rate law: [Fe(II) Tot ](t) [Fe(II) Tot ] 0 [Fe(II) Tot ] Max [Fe(II) Tot ] 0 1 e kt, (3) where [Fe(II) Tot ] o represents the total Fe(II) concentration (solid liquid) at time zero, [Fe(II) Tot ] Max represents the long-term maximum Fe(II) concentration, and k represents a first-order rate constant. In all cases, R 2 values for nonlinear least-squares regression fits of total Fe(II) vs. time data to Eqn. 3 were 0.95 (Table 1). Instantaneous rates of Fe(III) reduction were calculated from the first-derivative of Eqn. 3: d[fe(ii) Tot ] dt k [Fe(II) Tot ] Max [Fe(II) Tot ] 0 (e kt ). (4) The rates of reduction (expressed as % Fe(II)/d) show that initial reduction rates were highest for experiments 5 and 6, intermediate for experiment 4, and lowest for experiments 2 and3(table 1; Fig. 1). Results from previous experiments with S. algae and HFO (Beard et al., 1999) or hematite (Beard et al., 2003a) substrates are shown for comparison (Fig. 1); these earlier experiments all involved significantly lower initial Fe(III) reduction rates, and lower percent total reductions (10.7% for HFO at 23 d, and 1.6% for hematite at 86 d) Aqueous Fe The rate and magnitude of aqueous Fe(II) accumulation varied substantially among the different experiments. Early rapid production of Fe(II) aq occurred for experiments 4, 5, and 6(Tables 5 7). The relatively low Fe(II) aq contents for the first three time samples of experiment 4 may reflect enhanced precipitation of Fe(II)-bearing phases upon pasteurization (Table 5; see electronic annex EA-1), although this does not explain the low Fe(II) aq contents of the last sample (164 d), which was not pasteurized. No samples from experiments 1, 2, 3, 5, and 6 were pasteurized. Production of Fe(II) aq was slowest in experiments 2 and 3 (Tables 3 and 4), commensurate with its 10-fold lower initial cell density as compared to the other experiments. In all cases, the maximum Fe(II) aq contents comprised a minor component of total Fe in the system, ranging from 5% 10% for experiments 1, 3, and 4, to less than 0.25% for experiment 4. Because Fe(II) aq was a minor Fe reservoir in the reaction systems, relatively large changes in 56 Fe values are expected for this phase based on mass balance. Aqueous Fe(II) and total aqueous Fe concentrations in experiments 2 and 3, which compared HFO reduction by G. sulfurreducens and S. algae, were positively correlated (Tables 3 and 4) with slopes of and , respectively. These results suggest that Fe(III) aq comprised perhaps a few percent of the total Fe aq, considerably smaller than the ca. 10% of total Fe aq observed in the experiments of Nevin and Lovley (2000, 2002a,b) where the HFO substrate was physically isolated from the DMRB Solid-Phase End Products In addition to producing aqueous Fe(II), dissimilatory HFO reduction typically results in formation of magnetite and/or Fe

5 Table 2. Results for Experiment 1. Sample Time (d) Bottle Material Fe(II)aq mm/l %Fe system %Fe dissol. Molar Ca/Fe 56Fe 57 Fe Fe(II)-Solid Solid products Experiment 1A - Fe only E1A-Fe-1 (Uninoc. control) HFO VIS: HFO (no reaction) repeat E1A-Fe Tot Solid XRD and SEM-EDS: Only pure FeCC 3 present repeat A: Part. Diss. (0.1M HCl) - Sol. Frac A: Part. Diss. (0.1M HCl) - Residue Aq Fe repeat repeat repeat [1] repeat [1] repeat [1] repeat [2] Experiment 1B - Fe Ca E1B-Fe Ca-1 (Uninoc. control) HFO VIS: HFO (no reaction) repeat repeat E1B-Fe Ca Tot Solid XRD and SEM-EDS: Casiderite (ca. Ca 0.15 Fe 0.85 CO 3 ) repeat A: Part. Diss. (0.1M HCl) - Sol. Frac repeat A: Part. Diss. (0.1M HCl) - Residue B: Part. Diss. (0.1M HCl) - Sol. Frac repeat repeat B: Part. Diss. (0.1M HCl) - Residue repeat C: Part. Diss. (0.1M HCl) - Sol. Frac repeat C: Part. Diss. (0.1M HCl) - Residue D: Part. Diss. (0.1M HCl) - Sol. Frac D: Partial Diss. (0.1M HCl) - Residue E: Part Diss. (0.1M HCl) - Sol. Frac E: Part Diss. (0.1M HCl) - Residue repeat Aq Fe repeat repeat [1] repeat [1] repeat [2] Fe isotope fractionation in biogenic magnetite and siderite Each sample taken from different bottles (0 3). Partial dissolutions performed at room temperatures on timescales varying from 1 3 minutes using acids indicated. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. repeat [2] acidification of original sample after 6 months, followed by separate processing through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and SEM imaging, as well as chemical assays. HFO hydrous ferric oxide (ferrihydrite). Total solid assay of E1-Fe-2; IC/FeT Only pure siderite present. No detectable NMNC Fe(II)(s). Total solid assay of E1-Fe Ca-2: IC/FeT ; stoichiometry consistent with ca. Ca 0.15 Fe 0.85 CO 3. No detectable NMNC Fe(II)(s). IC inorganic carbon, FeT total Fe. Errors in Fe isotope measurements are 2SE based on in-run statistics. 967

6 Sample Time (d) Bottle Material Tot Fe aq mm/l Table 3. Results for Experiment 2. Fe(II)aq mm/l % Fe(II) % Fe(III) %Fe system % Tot solid 56 Fe 57 Fe Fe(II)-Solid Solid products E Tot Solid VIS: HFO (no reaction) A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: HFO (no reaction) repeat A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: HFO (no reaction) A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: brown; XRD: tr. mt. pks. A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid XRD: tr. lepid. pks.; tr. mt. pks.; SEM: tr. lepid. A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid XRD: incr. lepid. pks.; tr. mt. pks.; SEM; incr. lepid. A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat [1] E Tot Solid XRD: very strong lepid. pks.; tr. mt. pks.; SEM: abund. lepid. A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) Sol. Frac of residue of B D: Tot Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac F: Tot. Digest of residue from E G: Part. Diss. (7M HCl) - Sol. Frac H: Tot. Digest of residue from G Aq Fe repeat repeat [1] E Tot Solid XRD: no lepid. pks.; strong, broad mt. pks.; SEM: v. fine-grained repeat A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat repeat [1] C. M. Johnson et al.

7 E Tot Solid XRD: no lepid. pks.; strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat [1] repeat [1] E Tot Solid XRD: no lepid. pks.; strong, broad mt. pks; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) - Sol. Frac of residue of B D: Tot Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac F: Tot. Digest of residue from E G: Part. Diss. (7M HCl) - Sol. Frac H: Tot. Digest of residue from G Aq Fe repeat [1] E Tot Solid XRD: no lepid. pks.; strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat [1] repeat [2] E Tot Solid XRD: no lepid. pks.; strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) - Sol. Frac of residue of B D: Tot Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac F: Tot. Digest of residue from E Aq Fe E Tot Solid ca. 97 XRD: no lepid. pks.; strong, broad mt. pks.; SEM: v. fine-grained Aq Fe ca. 1.5 ca. 1.5 ca. 3 Fe isotope fractionation in biogenic magnetite and siderite Each sample taken from a different bottle (0 12) at time indicated. Partial dissolutions performed at room temperatures; A, D, and F involved acid treatment for 1 hour, whereas B, C, and E involved acid treatment for ca. 10 min. Aqueous Fe from triplicate measurements; uncertainties given as 1. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. repeat [2] acidification of original sample after 6 months, followed by separate processing through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and SEM imaging, as well as chemical assays. HFO hydrous ferric oxide (ferrihydrite), lepid. lepiocrocite, mt. magnetite. Total solid assay of E2-12: IC/FeT IC inorganic carbon, FeT total Fe. Fe(II)/Fe(III) ratio consistent with only mt. as solid phase. No detectable siderite or NMNC Fe(II)(s). Fe(II) and Fe(III) contents determined by Ferrozine assay. Eerros in Fe isotope measurements are 2SE based on in-run statistics. 969

8 Sample Time (d) Bottle Material Table 4. Results for Experiment 3. Tot Fe Aq mm/l Fe(II)aq mm/l % Fe(II) % Fe(III) %Fe system % Tot solid 56 Fe 57 Fe Fe(II)-Solid Solid products E Tot Solid VIS: HFO (no reaction) repeat A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: HFO (no reaction) repeat [1] A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: HFO (no reaction) A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: brown A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe E Tot Solid VIS: brown-black A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat E Tot Solid XRD: tr. mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat E Tot Solid XRD: incl. mt. pks.; SEM: v. finegrained repeat [1] A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) - Sol. Frac of residue of B D: Tot Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac F: Tot. Digest of residue from E G: Part. Diss. (7M HCl) - Sol. Frac H: Tot. Digest of residue from G Aq Fe repeat [1] E Tot. Solid XRD: strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat [1] E Tot Solid XRD: strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat repeat [1] repeat [1] C. M. Johnson et al.

9 E Tot Solid XRD: strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) Sol. Frac of residue of B D: Tot Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac F: Tot. Digest of residue from E G: Part. Diss. (7M HCl) - Sol. Frac H: Tot. Digest of residue from G Aq Fe repeat repeat [1] E Tot Solid XRD: v. strong, broad mt. pks.; SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac Aq Fe repeat repeat [1] repeat [1] E Tot Solid XRD: v. strong, broad mt. pks.; CHEM; NMNC Fe(II)(s); SEM: v. fine-grained A: Part. Diss. (0.5M HCl) - Sol. Frac B: Part. Diss. (0.1M HCl) - Sol. Frac C: Part. Diss. (0.5M HCl) Sol. Frac of residue of B D: Tot. Digest of residue from C E: Part. Diss. (3M HCl) - Sol. Frac repeat F: Tot. Digest of residue from E Aq Fe E Tot Solid ca. 97 XRD: v. strong, broad mt. pks.; CHEM: NMNC Fe(II)(s); SEM: v. fine-grained Aq Fe ca. 1.5 ca. 1.5 ca. 3 Fe isotope fractionation in biogenic magnetite and siderite Each sample taken from a different bottle (0 12) at time indicated. Partial dissolutions performed at room temperatures; A, D, and F involved acid treatment for 1 hour, whereas B, C, and E involved acid treatment for ca. 10 min. Aqueous Fe from triplicate measurements; uncertainties given as 1. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and SEM imaging, as well as chemical assays ( CHEM ). HFO hydrous ferric oxide (ferrihydrite), mt. magnetite, NMNC Fe(II)(s) amorphous, non-magnetic, non-carbonate Fe(II) solid. Total solid assay of E3-12; IC/FeT , indicating 4.6% siderite. High Fe(II)/Fe(III) ratio, in excess of that of magnetite, additionally indicates presence of NMNC Fe(II)(s) at ca. 2.6%. Percent magnetite calculated at 67.8%. Fe(II) and Fe(III) contents determined by Ferrozine assay. Errors in Fe isotope measurements are 2SE based on in-run statistics. 971

10 Sample Time (d) Bottle Material Table 5. Results for Experiment 4. Fe(II)aq mm/l % Fe(II) % Fe(III) %Fe system %Fe dissol. 56 Fe 57 Fe Fe(II)-Solid Solid products E4-1 (HFO stock HFO VIS: HFO (no reaction) material) repeat repeat E4-2 (Uninoculated 22 0 Tot Solid VIS: HFO (no reaction) control) repeat repeat repeat Aq Fe E Tot Solid VIS: HFO (no reaction) repeat repeat repeat Aq Fe E Tot Solid XRD: strong, sharp mt. pks., tr. sid. pks.; SEM: large and fine-grained xtls. A: Part. Diss. (0.1 M HCl rinse) - Sol. Frac A: Part. Diss. (0.1 M HCl rinse) - Residue C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Sol. Frac repeat C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Residue F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Sol. Frac repeat F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Residue Mag Solid repeat Aq Fe repeat repeat [2] E Tot Solid XRD: strong, sharp mt. pks., tr. sid. pks.; SEM: large and fine-grained xtls A: Part. Diss. (0.1 M HCl rinse) - Sol. Frac A: Part. Diss. (0.1 M HCl rinse) - Residue C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Sol. Frac C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Residue F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Sol. Frac repeat F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Residue Mag Solid repeat Aq Fe repeat repeat [2] C. M. Johnson et al.

11 E Tot Solid XRD: strong, sharp mt. pks., incr. sid. pks.; SEM and TEM: large and finegrained xtls. A: Part. Diss. (0.1 M HCl rinse) - Sol. Frac A: Part. Diss. (0.1 M HCl rinse) - Residue C: Part. Diss. (0.5 M HCl; 1 hr; 1 ml) - Sol. Frac repeat C: Part. Diss. (0.5 M HCl; 1 hr; 1 ml) - Residue F: Part. Diss. (3 M HCl; 2 min; 1 ml) - Sol. Frac F: Part. Diss. (3 M HCl; 2 min; 1 ml) - Residue Mag Solid repeat Aq Fe repeat repeat [2] E Tot Solid XRD: strong, sharp mt. pks., incr. sid. pks.; SEM and TEM: mostly large xtls. A: Part. Diss. (0.1 M HCl rinse) - Sol. Frac A: Part. Diss. (0.1 M HCl rinse) - Residue B: Part. Diss. (0.5 M HCl rinse; 1 ml) - Sol. Frac repeat B: Part. Diss. (0.5 M HCl rinse, 1 ml) - Residue repeat C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Sol. Frac C: Part. Diss. (0.5 M HCl; 1 hr, 1 ml) - Residue D: Part. Diss. (0.5 M HCl; 2 hrs, 1 ml) - Sol Frac. D: Part. Diss. (0.5 M HCl; 2 hrs, 1 ml) - Residue E: Part. Diss. (1 M HCl; 4 hrs, 1 ml) - Sol. Frac E: Part. Diss. (1 M HCl; 4 hrs, 1 ml) - Residue F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Sol. Frac F: Part. Diss. (3 M HCl; 2 min, 1 ml) - Residue Mag Solid repeat [1] Aq Fe repeat [1] Fe isotope fractionation in biogenic magnetite and siderite Each sample taken from a different bottle (0 5) at time indicated. Partial dissolutions performed at room temperatures, following acid treatment and times indicated. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. repeat [2] acidification of original sample after 6 months, followed by separate processing through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and TEM and SEM imaging, as well as chemical assays. HFO hydrous ferric oxide (ferrihydrite), mt. magnetite, sid. siderite. Total solid assay of E4-7 indicates that Fe(II)/Fe(III) ratio is higher than that of magnetite, suggesting solid is a mixture of 3.2% siderite and 96.8% magnetite; the excess Fe(II) phase is inferred to be siderite and not NMNC Fe(II)(s) based on the presence of siderite peaks in the XRD spectra, although the presence of NMNC Fe(II)(s) cannot be excluded. Note that samples E4-3 through E4-6 were pasteurized; no other samples in this study were pasteurized, including sample E4-7. Fe(II) and Fe(III) contents determined by Ferrozine assay. Errors in Fe isotope measurements are 2SE based on in-run statistics. 973

12 Sample Time (d) Bottle Material Fe(II)aq mm/l % Fe(II) % Fe(III) Table 6. Results for Experiment 5. %Fe system %Fe dissol. 56 Fe 57 Fe Fe(II)-Solid IC/FeT % Sid Total solid phase % Non-Sid Fe(II) % Mt % NMNC Fe(II)(s) Solid products E Tot Solid VIS: HFO (no reaction) A: Part. Diss. (1% HAc; 20 min) Sol. Frac. A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue C: Part. Diss. (0.1N HCl; min) - Sol. Frac. repeat C: Part. Diss. (0.1N HCl; min) - Residue Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) Sol. Frac. repeat Aq Fe E Tot Solid VIS: HFO (no reaction) Aq Fe E Tot Solid VIS: brown Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] E Tot Solid VIS: black-brown Mag Solid 7.74 Aq Fe E Tot Solid XRD: sm. broad mt. pks.; SEM: finegrained material A: Part. Diss. (1% HAc; 20 min) Sol. Frac. repeat A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue C: Part. Diss. (0.1N HCl; min) - Sol. Frac. C: Part. Diss. (0.1N HCl; 20 min) - Residue Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] E Tot Solid VIS: black-brown Mag Solid Aq Fe C. M. Johnson et al.

13 E Tot Solid XRD: incr. mt. pks.; sm. sid. pks.; SEM: fine-grained material A: Part. Diss. (1% HAc; 20 min) Sol. Frac. A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue C: Part. Diss. (0.1N HCl; min) - Sol. Frac. repeat C: Part. Diss. (0.1N HCl; 20 min) - Residue Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] repeat [1] E Tot Solid VIS: black-brown Mag Solid Aq Fe E Tot Solid XRD: incr. sid. pks., incl. impure sid. pks., strong mt. pks.; SEM: fine-grained material. A: Part. Diss. (1% HAc; 20 min) Sol. Frac. repeat A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue C: Part. Diss. (0.1N HCl; min) - Sol. Frac. C: Part. Diss. (0.1N HCl; 20 min) - Residue Mag Solid repeat Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] E Tot Solid VIS: black-brown Mag Solid Aq Fe Fe isotope fractionation in biogenic magnetite and siderite 975

14 Sample Time (d) Bottle Material Fe(II)aq mm/l % Fe(II) % Fe(III) %Fe system Table 6. (Continued) %Fe dissol. 56 Fe 57 Fe Fe(II)-Solid IC/FeT % Sid Total solid phase % Non-Sid Fe(II) % Mt % NMNC Fe(II)(s) Solid products E Tot Solid XRD: incr. sid. pks., decreasing impure sid. pks., strong mt. pks.; CHEM: NMNC Fe(II)(s); SEM: finegrained material A: Part. Diss. (1% HAc; 20 min) Sol. Frac. A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue C: Part. Diss. (0.1N HCl; min) - Sol. Frac. C: Part. Diss. (0.1N HCl; 20 min) - Residue Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) Sol. Frac. repeat Aq Fe repeat [1] repeat [2] E Tot Solid VIS: black-brown; CHEM: NMNC Fe(II)(s) Mag Solid Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong sid. pks., minor impure sid. pks.; CHEM: NMNC Fe(II)(s); SEM: finegrained material A: Part. Diss. (1% HAc; 20 min) Sol. Frac. A: Part. Diss. (1% HAc; 20 min) Residue B: Part. Diss. (1% HAc; 12 hrs) Sol. Frac. B: Part. Diss. (1% HAc; 12 hrs) Residue repeat C: Part. Diss. (0.1N HCl; min) - Sol. Frac. repeat C: Part. Diss. (0.1N HCl; min) - Residue Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe C. M. Johnson et al.

15 Fe isotope fractionation in biogenic magnetite and siderite 977 E Tot Solid VIS: black-brown; CHEM: NMNC Fe(II)(s) Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong sid. pks.; SEM: fine-grained material Aq Fe E Tot Solid XRD: strong, broad mt. pks, strong sid. pks., SEM: fine-grained material Mag Solid Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong sid. pks.; CHEM: NMNC Fe(II)(s); SEM; finegrained material Samples taken from three single large bottles in sequence at times indicated; bottles 0 and 1 represent duplicates up through 26.9 days, at which time insufficient sample existed for analysis, and further samples taken from a third bottle (2). Partial dissolutions performed at room temperatures, following acid treatment and times indicated. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. repeat [2] acidification of original sample after 6 months, followed by separate processing through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and SEM imaging, as well as chemical assays ( CHEM ). HFO hydrous ferric oxide (ferrihydrite, mt. magnetite, sid. siderite, NMNC Fe(II)(s) amorphous, non-magnetic, non-carbonate Fe(II) solid. Fe(II) and Fe(III) contents determined by Ferrozine assay. IC inorganic carbon, FeT total Fe. Where fraction of magnetite calculated, all HFO assumed to have been consumed, and all ferric Fe is assigned to magnetite in its stoichiometric proportion. Errors in Fe isotope measurements are 2SE based on in-run statistics. carbonates, which are readily identifiable in XRD spectra and SEM/TEM images of reduction end-products (e.g., Zachara et al., 2002). However, several alternative solid-phase Fe(II)- bearing components (e.g., sorbed/surface-precipitated Fe(II) and/or mixed Fe(II)-Fe(III) hydroxides) may also be produced during HFO reduction, which are difficult to identify by traditional techniques. Although phosphate was present in all of the experiments, vivianite (Fe 3 (PO 4 ) 2 8H 2 O) was not identified in any of the XRD spectra, and could have constituted at most only ca. 3% of the solid mass based on the P content of the culture media Experiment 1 Carbonate minerals were the exclusive end-products of HFO reduction in this experiment. A bimodal distribution of crystallite sizes (ca m and 2 m) was observed. For the Ca-free experiment (experiment 1A), XRD spectra confirmed that pure siderite was the only detectable solid phase, which was also confirmed by EDS spectra obtained on individual crystals during SEM imaging. XRD and EDS spectra obtained on carbonates from the Ca-bearing experiment (experiment 1B) produced consistent mineral stoichiometries of Ca 0.12 Fe 0.88 CO 3 to Ca 0.15 Fe 0.85 CO 3, and this was confirmed for both small and large crystals. A bulk carbonate stoichiometry of ca. Ca 0.15 Fe 0.85 CO 3 was confirmed by the d-spacing shift in XRD spectra. There was no evidence for formation of CaCO 3 minerals from XRD or EDS spectra, although we note that the solutions (including the abiotic controls) were always supersaturated with respect to CaCO 3 (Roden et al., 2002). Determinations of the inorganic carbon (IC) contents of the total solids for the Ca-bearing experiment (experiment 1B) showed an excess of carbonate (molar IC/Fe Total ; Table 2) relative to siderite or the Ca 0.12 Fe 0.88 CO 3 to Ca 0.15 Fe 0.85 CO 3 stoichiometry that was determined for individual crystals by EDS-SEM. A bulk carbonate stoichiometry of Ca 0.15 Fe 0.85 CO 3 would produce a molar IC/Fe Total ratio of 1.18, suggesting the measured molar IC/Fe Total ratio of reflects the presence of 16 14% CaCO 3 in the total solid (1 errors based on triplicate IC analyses). Within the uncertainties, this would not necessarily be detectable in the XRD spectra. The molar IC/Fe Total ratio for the Ca-free experiment (experiment 1A) matched that expected for siderite (molar IC/Fe Total ; Table 2), confirming that siderite was the only solid phase. There was no evidence for additional Fe(II)-bearing solid phases in either the Ca-bearing or Ca-absent experiments, based on EDS spectra obtained during SEM imaging, XRD spectra, or through comparison of the Fe and IC contents of the total solid phase. The 56 Fe values for Fe(II) aq were consistently higher than those of the bulk carbonate for the Ca-absent and Ca-bearing experiments (Fig. 2). Partial dissolution (0.7%) in 0.1 M HCl of siderite produced in experiment 1A yielded a 56 Fe value that was close to that of Fe(II) aq, ca. 1 higher than that of the bulk siderite (Fig. 2). Partial dissolution ( %) of the Cabearing Fe carbonate produced in experiment 1B yielded 56 Fe values that were ca. 1 higher than that of the bulk siderite (Fig. 2). The greatest percent dissolution (2.3%) produced a Ca/Fe ratio of 0.38, which lies most closely to that of the bulk solid. The correlation between 56 Fe, Ca/Fe ratio, and percent dissolution for experiment 1B (Fig. 2) may reflect variable sampling of a minor Ca-rich carbonate, perhaps an Fe(II)-

16 Sample Time (d) Bottle Material Fe(II)aq mm/l % Fe(II) % Fe(III) %Fe system Table 7. Results for Experiment 6. %Fe dissol. 56 Fe 57 Fe Fe(II)-Solid IC/FeT % Sid Total solid phase % Non-Sid Fe(II) % Mt % NMNC Fe(II)(s) Solid products E Tot Solid VIS: HFO (no reaction) Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe E Tot Solid VIS: HFO (no reaction) Aq Fe E Tot Solid VIS: brown Mag Solid repeat Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] E Tot Solid VIS: brown Mag Solid 2.65 Aq Fe E Tot Solid XRD: tr. broad mt. pks.; SEM: fine-grained material Mag Solid repeat Non-Mag Solid Part. Diss. (0.5M HCl; hrs) - Sol. Frac. repeat Aq Fe repeat [1] E Tot Solid VIS: black-brown Mag Solid Aq Fe E Tot Solid XRD: incr. mt. pks., tr. sid. pks.; SEM: finegrained material Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; hrs) - Sol. Frac. repeat Aq Fe repeat repeat [1] E Tot Solid VIS: black-brown Mag Solid Aq Fe C. M. Johnson et al.

17 E Tot Solid XRD: incr. mt. pks., tr. sid. pks.; SEM: finegrained material Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [1] E Tot Solid VIS: black-brown Aq Fe E Tot Solid XRD: strong, broad mt. pks., sid. pks.; SEM: fine-grained material Mag Solid Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat [2] E Tot Solid VIS: black-brown Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong, broad sid. pks.; CHEM: NMNC Fe(II)(s); SEM: fine-grained material Mag Solid repeat Non-Mag Solid Part. Diss. (0.5M HCl; 12 hrs) - Sol. Frac Aq Fe repeat E Tot Solid VIS: black-brown; CHEM: NMNC Fe(II)(s) Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong, broad sid. pks.; SEM: finegrained material Mag Solid Aq Fe E Tot Solid XRD: strong, broad mt. pks., strong, broad sid. pks.; SEM: finegrained material. Mag Solid Aq Fe Fe isotope fractionation in biogenic magnetite and siderite Samples taken from two large bottles in sequence at times indicated, where bottles 0 and 1 represent duplicates. Partial dissolutions performed at room temperatures, following acid treatment and times indicated. repeat re-analysis of separated Fe solution in different mass-spectrometry session. repeat [1] separate processing of original sample through chemistry. repeat [2] acidification of original sample after 6 months, followed by separate processing through chemistry. Solid phases characterized by visual inspection ( VIS ), slow-scan XRD spectra, and SEM imaging, as well as chemical assays ( CHEM ). HFO hydrous ferric oxide (ferrihydrite), mt. magnetite, sid. siderite, NMNC Fe(II)(s) amorphous, non-magnetic, non-carbonate Fe(II) solid. Fe(II) and Fe(III) contents determined by Ferrozine assay. IC inorganic carbon, FeT total Fe. Where fraction of magnetite calculated, all HFO assumed to have been consumed, and all ferric Fe is assigned to magnetite in its stoichiometric proportion. Errors in Fe isotope measurements are 2SE based on in-run statistics. 979

18 980 C. M. Johnson et al. Fig. 1. Comparison of Fe(III) reduction rates for experiments 2 6 (Table 1). Reduction rates are expressed as % Fe(II)/d, where % Fe(II) is relative to total Fe(II) produced at the end of the experiment, and were calculated from regression of total system Fe(II) contents vs. time, assuming a first-order rate law (Eqn. 3 in text); regression results are listed in Table 1. The range in reduction rates for earlier HFO (Beard et al., 1999) and hematite (Beard et al., 2003a) reduction experiments using S. algae are shown for comparison. substituted calcite, consistent with the inferred presence of CaCO 3 based on the measured IC/Fe Total ratio of Based on SEM images, it seems likely that the partial dissolutions preferentially sampled the smaller crystal population ( 2 m), or the surfaces of larger crystals. Fig. 3. Chemical and isotopic data from experiments 2 and 3. Temporal variations in aqueous Fe(II) concentration (A), % Fe(II) in the total solids (B), % Fe(II) in 0.5 M HCl dissolutions (C), and 56 Fe values for aqueous and total solid Fe (D). Horizontal lines in panels B and C show reference line for magnetite (33.3% Fe(II)). The % Fe(II) in the total solids and the 0.5 M HCl partial dissolution reflects the fraction of ferrous Fe relative to the total Fe measured. * notes sample in experiment 2 where solid phase conversion of HFO to lepidocrocite was complete (Table 3). Data from Tables 3 and Experiments 2 and 3 Fig. 2. Fe isotope compositions from experiment 1 relative to molar Ca/Fe ratio of bulk carbonate samples and partial dissolutions of bulk solids. Experiment 1A produced pure siderite (FeCO 3 ), whereas experiment 1B contained Ca in the growth media, and produced a Ca-bearing siderite (ca. Fe 0.85 Ca 0.15 CO 3 ). Percent partial dissolution noted for each sample; 100% denotes total digestions of bulk solids. Data from Table 2. Significant accumulation of Fe(II) aq and conversion of HFO to magnetite did not occur until ca. 10 d in experiments 2 and 3 (Fig. 3). Temporal changes in the ratio of Fe(II) to total Fe in the solids were generally similar for the G. sulfurreducens (experiment 2) and S. algae (experiment 3) cultures, and ap-

19 Fe isotope fractionation in biogenic magnetite and siderite 981 proached values consistent with magnetite by 100 d (Fig. 3B; Tables 3 and 4). The broad XRD peaks for magnetite in both the G. sulfurreducens and S. algae cultures indicated that magnetite was very fine grained, which is supported by SEM imaging. A carbonate buffer was not used in this experiment, and no siderite peaks were observed in the slow-scan XRD spectra for both experiments. In experiment 2, the fraction of solid phase Fe that dissolved in the 1-h 0.5 M HCl partial dissolution decreased steadily from ca. 100% in the first few days to 46% at d 16. This drop corresponded to conversion of HFO to m-sized lepidocrocite (Table 3; see also electronic annex EA-2 for SEM images), which is less susceptible to dissolution in dilute acid compared to HFO (e.g., Cornell and Schwertmann, 1996). The fraction of solid phase Fe that dissolved in 0.5 M HCl increased abruptly to 70% 80% after 16 d in association with conversion of lepidocrocite to fine-grained magnetite (Table 3). In contrast, in experiment 3 the fraction of total Fe that dissolved in the 0.5 M HCl partial dissolution decreased from ca. 100% during the first few days of incubation and then remained at 70% 80% throughout the remaining time of the experiment (Table 4), consistent with the observation that HFO and fine-grained magnetite was not accompanied by formation of a resistant intermediate ferric hydroxide phase such as the large lepidocrocite crystals that formed in experiment 2. The long-term ratio of Fe(II) to total Fe in solids differed between experiments 2 and 3 (Fig. 3B; Tables 3 and 4). This ratio remained close to that of magnetite (33.3%) after 40 d in experiment 2 (G. sulfurreducens), but exceeded that of magnetite for the 242 and 291 d samples from experiment 3 (S. algae). Because the solid-phase inorganic carbon (IC) contents of experiment 3 permit at most only a few percent siderite formation (Table 4), the anomalously high % Fe(II) at the end of experiment 3 must reflect a nonmagnetic, noncarbonate solid-phase Fe(II) component that was absent in experiment 2. We will refer to such a component as NMNC Fe(II)(s), because its presence may only be inferred from the chemical assays, and a well-crystalline Fe(II)-bearing phase such as Fe(OH) 2 or green rust was not observed in the XRD spectra and was not apparent in SEM images (see electronic annex EA-2). Assuming that all the Fe(III) in the total solid at the end of experiment 3 came from magnetite, mass-balance calculations indicate that NMNC Fe(II)(s) comprised ca. 28% of total Fe, compared to ca. 5% for siderite (Table 4). The solid-phase IC contents of experiment 2 indicate that no siderite formed in this experiment (Table 3). 56 Fe values for the total solid for experiments 2 and 3 remained constant over time and essentially equal to that of the starting HFO (Fig. 3D; Tables 3 and 4), reflecting the fact that virtually all Fe was contained in the solid phase. 56 Fe values for aqueous Fe steadily increased from low values (-2.5 to 2.0) during the first 20 d (Fig. 3D), and then leveled off to a constant value for experiment 2. In contrast, the 56 Fe values for aqueous Fe continued to increase to the end of experiment 3. Detailed partial dissolutions were performed on solid-phase samples from 16, 68, and 242 d from experiments 2 and 3. Brief extraction (several minutes) using 0.1, 0.5, and 3 M HCl generally dissolved between 0.5 and 10% of the total Fe (Tables 3 and 4), and the data appear to lie along mixing trends between magnetite and a pure Fe(II) solid phase in terms of % Fig. 4. Variations in % Fe(II) (A) and 56 Fe values (B) as a function of percent total Fe dissolved during partial dissolution of the solidphase end products from experiments 2 and 3 at 16, 68, and 242 d. % Fe(II) in panel A reflects the proportion of ferrous Fe relative to total Fe in the sample. The large symbols show results from 0.1 M HCl extractions, and are connected by tie-lines to smaller symbols that show results from 0.5 M HCl extraction of the residual solid from the 0.1 M HCl extraction. Small symbols not attached to tie-lines show results from 3 M HCl partial dissolutions. Bulk mixing trends reflecting variable dissolution of magnetite and a pure Fe(II) component are shown, where the total fraction of the pure Fe(II) component varies from (0.5%) to 0.08 (8%). Data from Tables 3 and 4. Fe(II) percent dissolution variations (Fig. 4). Based on these mixing calculations, the Fe(II) end member comprised 0.5 to 8% of the total solid phase (Fig. 4). Because no carbonates formed in experiment 2, and only a minor amount of siderite was formed in experiment 3 (Tables 3 and 4), it seems likely that NMNC Fe(II)(s) and/or surface-associated Fe(II) were the major contributors of Fe(II) that was liberated during the partial dissolutions. In the case of the 16 d samples, which contained only minute magnetite peaks in the XRD spectra, it is likely that the 0.1 M HCl partial dissolutions are dominated by sorbed Fe(II). Assuming that partial dissolution reflects a mixture between a Fe(II)-only phase and magnetite, the 56 Fe values of the dissolved component suggest that the Fe(II)-only phase has a

20 982 C. M. Johnson et al. 56 Fe value of ca. 1.5 (Fig. 4B). The 0.1 M HCl partial dissolution of the 16 and 242 d samples from experiment 2 had high % Fe(II) relative to the total Fe dissolved ( 85%; Fig. 4A), suggesting that their 56 Fe values of 1.0 to 1.5 lie close to that of the Fe(II)-only component. The 0.1 M HCl partial dissolution for the 16, 68, and 242 d samples from experiment 3 also had high % Fe(II) ( 85% of the Fe liberated; Fig. 4A), and the 56 Fe values for these two samples were also ca. 1.5 (Fig. 4B). In contrast, despite the high % Fe(II) of the 0.1 M HCl partial dissolution of the 242 d sample of experiment 3, its 56 Fe value was significantly higher ( 56 Fe ca. 0; Fig. 4B), and this is correlated with a high 56 Fe value of aqueous Fe in experiment 3 (Fig. 3D) Experiment 4 Fe(II) aq contents decreased with time (Fig. 5A), which corresponded with progressive conversion of HFO to magnetite. The ratio of Fe(II) to total Fe in the solids increased over time to 35.4% at 164 d (Fig. 5B; Table 5), a value only slightly in excess of that expected for magnetite (33.3%). SEM images indicated the presence of some large (ca. 2 m diameter) magnetite crystals early in the experiment, suggesting early rapid formation of magnetite. TEM images and electron diffraction patterns indicated that virtually all of the very finegrained solids analyzed (ca. 50 nm diameter; electronic annex EA-2) also consisted of magnetite. By the end of the experiment, a large portion of the fine-grained solids had recrystallized to coarser-grained magnetite. In addition to magnetite, siderite was identified in the XRD spectra (Table 5), and the height of the siderite peak increased with time. Less than 10% of the total Fe content of the solid phase was recovered in the 1-h 0.5 M HCl partial dissolutions (Table 5), which probably reflects the fact that experiment 4 produced very large magnetite crystals that were resistant to attack by 0.5 M HCl. It is possible that early, rapid conversion to wellcrystallized magnetite was promoted by pasteurization of the 4, 11, and 22 d samples (the 164 d sample was not pasteurized). The proportion of Fe(II) relative to total Fe obtained from the 0.5 M HCl partial dissolutions leveled off at 50% 60% toward the end of the experiment (Fig. 5C; Table 5), suggesting that this treatment preferentially dissolved siderite or NMNC Fe(II)(s). The 56 Fe values for the solid phases remained relatively constant over time (Fig. 5D), as observed in experiments 2 and 3. Similarly, the 56 Fe values for aqueous Fe(II) were initially low and increased with time, reaching, and then slightly exceeding, that of the total solid (Fig. 5D). Partial dissolution of the solids in 0.1, 0.5, and 3 M HCl produced extracts that contained 10 to 100% Fe(II) of the total Fe dissolved, indicating variable contributions from Fe(III)-bearing solids and a pure Fe(II) component, where the highest % Fe(II) was generally associated with the smallest degree of partial dissolution (Fig. 6A). Because the % Fe(II) of the total solids was generally less than that expected for magnetite (33.3%) before 22 d (Fig. 6B), the Fe(III) liberated in the partial dissolutions of the 4, 11, and 22 d samples is likely to reflect a mixture of unreacted HFO and a pure Fe(II) phase. The variations between % Fe(II) and 56 Fe with percent dissolution may be explained by a physical mixture of a low 56 Fe (-3 to 2 ) Fig. 5. Chemical and isotopic data from experiment 4. Temporal variations in aqueous Fe(II) concentration (A), % Fe(II) in the total solids (B), % Fe(II) in 0.5 M HCl dissolutions (C), and 56 Fe values for aqueous and total solid Fe (D). Horizontal lines in panels B and C show reference line for magnetite (33.3% Fe(II)). The % Fe(II) in the total solids and the 0.5 M HCl partial dissolution reflects the fraction of ferrous Fe relative to the total Fe measured. Data from Table 5. pure Fe(II) component and unreacted HFO and/or magnetite that has the same 56 Fe value as the bulk system (Fig. 6B); this low 56 Fe component could be siderite (based on its prominence in the XRD spectra), NMNC Fe(II)(s), sorbed Fe(II), or any combination of these Experiments 5 and 6 Temporal changes in aqueous Fe(II) contents were similar in the presence (experiment 5) and absence (experiment 6) of

21 Fe isotope fractionation in biogenic magnetite and siderite 983 Fig. 6. Variations in % Fe(II) (A) and 56 Fe values (B) as a function of percent total Fe dissolved during partial dissolution of the solidphase end products from experiment 4 (data from Table 5). % Fe(II) in panel A reflects the proportion of ferrous Fe relative to total Fe in the sample. A bulk mixing trend reflecting variable dissolution of magnetite and a pure Fe(II) component is shown in panel A. Bulk mixing trends reflecting variable contributions from magnetite and a pure Fe(II) component during partial dissolution is shown in panel B, with a 56 Fe value of 2.0 (upper line) or 3.0 (lower line); both mixing lines calculated assuming the pure Fe(II) component comprises 0.1% of the total solid. Data from Table 5. NTA (Fig. 7A). The ratio Fe(II) to total solid Fe recovered by 3 M HCl (Fig. 7B; Tables 6 and 7), as well as that obtained in the 0.5 M HCl partial dissolutions (Fig. 7C; Tables 6 and 7), exceeded that expected for magnetite within a few days. The reason for the more extensive Fe(III) reduction observed in experiments 5 and 6 compared to experiment 4 is unknown, given their similar initial cell densities. The high rates of reduction in experiments 5 and 6 produced large quantities of siderite, as well as significant quantities of NMNC Fe(II)(s), which accounted for ca. 25 and 40% of total Fe at the end of the experiments, respectively (Tables 6 and 7). Siderite was prominent in XRD spectra (Tables 6 and 7) as well as SEM images. Early siderite (ca. 2 d) consisted of botryoidal carbonate globules, which were largely replaced by rhombohedral carbonates by ca. 7 d (see electronic annex EA-2 for images). Fig. 7. Chemical and isotopic data from experiments 5 and 6. Temporal variations in aqueous Fe(II) contents (A), % Fe(II) in the total solids (B), % Fe(II) in 0.5 M HCl dissolutions (C), and 56 Fe values for aqueous and total solid Fe (D). Horizontal lines in panels B and C show reference line for magnetite (33.3% Fe(II)). The % Fe(II) in the total solids and the 0.5 M HCl partial dissolution reflects the fraction of ferrous Fe relative to the total Fe measured. Curved arrows in panel D mark the fraction of siderite or solid Fe(II) in excess of siderite (scale on right side of panel D), based on the solid phase proportions (curves pass through the averages of the experiments for X Siderite and X Ex Fe(II) ). Data from Tables 6 and 7. The 56 Fe values for Fe(II) aq showed large changes throughout experiment 5 and 6, beginning at low values of ca. 1.4, increasing by ca. 3 to positive values that exceeded those of the bulk solid (Fig. 7D). Longer-term samples from a third bottle in experiment 5 suggest that at long time periods, the 56 Fe values for Fe(II) aq moved back toward that of the bulk solid. The increase in 56 Fe values for aqueous Fe(II) over the first 27 d of the experiment correlated with large increases in the fraction of siderite, as well as the fraction of nonsiderite Fe(II) in the total solid ( excess Fe(II) ; Fig. 7D). The overall

22 984 C. M. Johnson et al. were dissolved in the 20 min 1% HAc dissolution increased with time (Fig. 8B), tracking changes in the 56 Fe values of aqueous Fe(II) (Fig. 7D). Similar trends were observed for the 12 h 1% HAc dissolutions, although the trend was offset to higher percent dissolution and the range in 56 Fe values was slightly smaller (Fig. 8B). The offset of the 12 h HAc treatment to higher percent dissolution relative to the 20 min HAc treatment is expected, and it is also expected that the smaller percent dissolutions are likely to record the greatest range in 56 Fe values if the solids are isotopically heterogeneous. Temporal trends for the 20 min 0.1 M HCl dissolutions were more complicated, beginning at moderately low values, increasing with time, then decreasing toward the end of the experiment (Fig. 8B), suggesting that the 0.1 M HCl treatment likely sampled different proportions of solids than the HAc treatments. 5. DISCUSSION Experiments 1 and 2 produced single-phase solid assemblages in the end products, siderite and magnetite, respectively, and therefore provide the most straightforward constraints on aqueous Fe-mineral fractionations produced by DMRB. More complex solid assemblages were produced in experiments 3 6, which provide insights into the isotopic effects of intermediate solid products. The temporal changes in Fe isotope compositions of aqueous Fe that were observed in experiments 2 6 constrain the isotopic effects of dynamic, evolving systems and reflect the changing proportions and influence of sorbed Fe(II) and solid products Fe(II) aq -Fe Carbonate Fractionations Fig. 8. Variations in % Fe(II) (A) and 56 Fe values (B) as a function of percent total Fe dissolved during partial dissolution of the solidphase end products from experiment 5. % Fe(II) in panel A reflects the proportion of ferrous Fe relative to total Fe in the sample. Curves in panel B show temporal trends (increasing time) in 56 Fe values for the dissolved component of the 20 min and 12 h 1% HAc partial dissolutions. For the 0.1 M HCl dissolutions, numbers indicate sample number (in time sequence). Data from Table 6. temporal trends in 56 Fe values for the experiments with and without NTA were similar (Fig. 7D), following the trends in Fe(II) aq (Fig. 7A). Partial dissolutions for the solid phases produced in experiments 5 and 6 using 1% HAc and 0.1 M HCl produced high-fe(ii) components (Fig. 8A). The extent of partial dissolution ranged from 1 to 50% (Fig. 8A), indicating the presence of a highly labile solid-phase Fe(II) component that was in much greater abundance than in experiments 2, 3, and 4. Such a solid Fe(II) component is likely to be very fine-grained siderite or its precursors, or NMNC Fe(II)(s), based on the chemical assays, SEM images, and XRD spectra (Tables 6 and 7). The 56 Fe values of the labile solid Fe(II) component varied greatly, from 1.2 to 1.3 (Fig. 8B), which contrasts with the consistently low 56 Fe values for the solid Fe(II) components obtained in partial dissolutions from experiments 2, 3, and 4. The 56 Fe values for the solid Fe(II) components that We infer that the Fe(II) aq -Fe carbonate fractionation measured for the bulk carbonate is least likely to reflect an equilibrium Fe isotope fractionation in experiment 1, and instead rely on data obtained from the partial dissolutions, because these are interpreted to have preferentially dissolved the smaller crystals or surfaces of large crystals that were more likely to be in isotopic equilibrium with Fe(II) aq at the end of the 485 d experiment. Using this approach, we estimate that the equilibrium Fe(II) aq -Fe carbonate fractionation is near zero for pure siderite and 1 for Ca-substituted Fe carbonate (Fig. 9). The later fractionation is only generally constrained because of uncertainty in the sensitivity of Fe(II) aq -Fe carbonate fractionations to the extent of Ca (or other cation) substitutions. In contrast, Fe(II) aq -Fe carbonate fractionations obtained using the bulk solid are ca. 1 higher than those estimated to reflect equilibrium fractionations, and this contrast is observed for both pure siderite and Ca-substituted Fe carbonate. It is possible that these large Fe(II) aq -Fe carbonate fractionations reflect kinetic effects, but it is also possible that they simply reflect disequilibrium conditions at the end of the experiments, where the 56 Fe values for Fe(II) aq that coexisted with the earlyformed bulk carbonate when it formed early in the experiment are unknown. A wide range in Fe(II) aq -Fe carbonate fractionations are predicted from spectroscopic and natural data, spanning values from 0.7 to 3.5 (Fig. 10; Polyakov and Mineev, 2000; Schauble et al., 2001; Johnson et al., 2003), and both ap-

23 Fe isotope fractionation in biogenic magnetite and siderite 985 Fig. 9. Comparison of Fe isotope fractionations between aqueous Fe(II) and Fe carbonate relative to molar Ca/Fe ratio, as determined in experiment 1. Numbers next to symbols indicate the percent dissolution (100% notes analyses of the bulk solid products). Because the partial dissolution experiments are interpreted to sample the most exchangeable portions of the solid carbonate products, we interpret the Fe(II)-Carbonate fractionations obtained by this method to most closely reflect those of the equilibrium isotope fractionation factors. Data from Table 2. proaches predict that Fe(II)aq-Fe carbonate should increase with decreasing mol fraction of Fe, from siderite to ankerite. The effect of Ca substitution observed in the current study provides the first experimental confirmation of these predictions. Our results suggest that bonding changes and distortions in the crystal lattice that accompany even small amounts of Ca substitution into siderite produce very large Fe isotope effects. Because natural Fe carbonates commonly contain significant Mn and Mg as well (see Section 2 above), it seems likely that carbonate stoichiometry may exert a substantial control on Fe isotope fractionations between Fe(II) and carbonate, and we anticipate that this will be recorded in both biologic and abiotic systems. The estimated near-zero equilibrium Fe(II) aq -siderite fractionation in our biologic experiment is less than the positive fractionations predicted from spectroscopic data, but overlaps the range estimated from natural siderites (Fig. 10). However, the estimated equilibrium Fe(II) aq -Fe carbonate fractionation in the Ca-bearing biologic experiment, which produced a Ca/Fe ratio comparable to that of natural siderites, is significantly higher than that estimated from natural minerals (Fig. 10). It is thus far unknown whether the influence of X Fe on Fe(II) aq -Fe carbonate fractionations measured in our biologic system is unique relative to abiotic precipitation systems, because experimental determinations of Fe isotope fractionation across the compositional range of Fe-bearing carbonates have not been made Sorption Effects Fig. 10. Comparison of isotopic fractionations determined between Fe(II) aq and Fe carbonates relative to mole fraction of Fe (total solid) from predictions based on spectroscopic data (Polyakov and Mineev, 2000; Schauble et al., 2001), natural samples (Johnson et al., 2003), and the results of this study of biogenic Fe carbonate formation. All fractionations calculated relative to 56 Fe/ 54 Fe ratios. Total cations were normalized to unity, so that end-member siderite corresponds to X Fe 1.0. Error bars reflect reported uncertainties; analytical errors for data reported in this study are smaller than the size of the symbol. Fractionations measured on bulk carbonate from experiment 1 are interpreted to reflect disequilibrium or kinetic isotope fractionations, whereas those estimated from partial dissolutions are interpreted to lie closer to equilibrium values. For the Ca-bearing experiment, where the bulk solid has a composition of approximately Ca 0.15 Fe 0.85 CO 3, high- Ca and low-ca refer to the range measured during partial dissolution (see Fig. 2). Various ferric Fe substrates that can act as the terminal electron acceptor for DMRB, including HFO, goethite, lepidocrocite, and hematite, have large ranges in their capacity to sorb aqueous Fe(II), and sorption is generally correlated with mineral surface area. In addition to surface sorption onto ferric oxide/hydroxide substrates, sorption of aqueous Fe(II) may also occur on the surfaces of DMRB themselves (e.g., Urrutia et al., 1998; Liu et al., 2001; Roden and Urrutia, 2002). The proportion of Fe(II) sorbed to HFO for the first 16 d of experiments 2 and 3, before formation of magnetite, decreased with time from ca to 89.0%, as calculated from aqueous Fe(II) and total Fe(II) contents of the 0.5 M HCl partial dissolutions (Tables 3 and 4). The measured proportion of sorbed Fe(II) agrees well with that calculated for experiments 2 and 3 using common surface areas and sorption capacities for HFO of 600 m 2 /g and 3x10 6 mole/m 2, respectively (e.g., Roden and Zachara, 1996). The Fe(II) aq -solid fractionations for the early part of experiments 2 and 3 were correlated with Fe reduction rates, where the largest magnitude fractionations were associated with the greatest reduction rates (Fig. 11). We interpret these trends to reflect early kinetic isotope fractionation during rapid sorption of Fe(II) to HFO, where the 56 Fe value of sorbed Fe(II) is

24 986 C. M. Johnson et al. Fig. 11. Isotopic fractionation between Fe(II) aq and the total solid in experiments 2 and 3 as a function of Fe reduction rates during the early part of the experiments (see Fig. 1). Decreasing Fe reduction rates (to the right) correspond to increasing time of incubation. Only time points that reflect the period before significant magnetite formation are plotted to the left of the vertical dashed line, which marks completion of solid phase conversion of HFO to lepidocrocite in experiment 2. The decrease in Fe(II)aq-Solid fractionations with decreasing Fe reduction rates suggests that the early, large-magnitude fractionations reflect kinetic effects due to rapid sorption of Fe(II), followed by approach to isotopic equilibrium with time. The large decrease in sorption capacity during conversion of HFO to lepidocrocite that occurs in the G. sulfurreducens culture is accompanied by an increase in the 56 Fe value for Fe(II) aq, producing a decrease in the magnitude of the Fe(II)aq-Solid fractionation; this is interpreted to reflect release of sorbed Fe(II) that had high 56 Fe values during initially rapid (kinetic) uptake. This change is marked ( * ), which is the same datum that is so marked in Figure 3D. For comparison, the fractionation between Fe(II) aq and ferric substrate measured in the experiments of Beard et al. (1999, 2003a), which involved much slower Fe reduction rates (10 1 to 10 3 % Fe(II)/d) and different substrates (HFO and hematite), are shown by the horizontal band. Data from Tables 3 and 4. relatively high. This is consistent with the experiments of Icopini et al. (2004), who inferred (but did not measure directly) a high- 56 Fe sorbed component based on rapid (1 d) abiotic sorption of Fe(II) aq onto goethite. The 1 change in the Fe(II) aq -solid fractionation in the first part of experiments 2 and 3 (Fig. 11) cannot, however, be explained by a constant Fe(II) aq -Fe(II) sorbed fractionation factor, given the small changes in the proportions of sorbed Fe(II) during this time period. The 56 Fe value of sorbed Fe(II) at 16 d may be estimated using the 0.1 M HCl partial dissolution (Tables 3 and 4). In experiment 2, the 16 d 0.1 M HCl partial dissolution sampled 1.2% of the total solid, which is equivalent to 8.9% of the total sorbed Fe(II) pool, and the 56 Fe value of this dissolution is indistinguishable from that of Fe(II) aq within analytical error (Table 3). In experiment 3, the 16 d 0.1 M HCl partial dissolution sampled 8.8% of the total solid, which is equivalent to 54% of the total sorbed Fe(II) pool, and the 56 Fe value of this dissolution is 0.4 heavier that that of Fe(II) aq (Table 4). A possible explanation for the differences between the experiments is that conversion of HFO to lepidocrocite had largely occurred by 16 d in experiment 2, whereas no such conversion occurred in experiment 3. Moreover, experiment 3 ultimately produced large quantities of NMNC Fe(II)(s), and this may have been a component in the 0.1 M HCl partial dissolution. Nevertheless, we find no evidence that the large 2.1 fractionation between sorbed Fe(II) and Fe(II) aq inferred by Icopini et al. (2004) existed over the longer timescales (several weeks) of our experiments, and we suggest that the results of Icopini et al. (2004) probably reflect a kinetic isotope fractionation upon rapid (1 d) sorption of Fe(II), and are therefore not applicable to the long time scales of dissimilatory Fe(III) reduction that generally occur in nature. Support for an early, transient high- 56 Fe Fe(II) sorbed component comes from the anomalously high Fe(II) aq -solid fractionation for the 16 d sample from experiment 2, which is interpreted to reflect release of a high- 56 Fe sorbed Fe(II) component upon complete phase conversion of HFO to lepidocrocite, commensurate with the large decrease in surface area that is expected to accompany such a conversion. That the Fe(II) aq -solid fractionations measured in experiments 2 and 3 move toward those measured by Beard et al. (1999, 2003a), which involved very low reduction rates of 10 3 to 10 1 % Fe/d (Figs. 1 and 11), suggests that the fractionation measured at low Fe reduction rates most likely reflect equilibrium conditions Fe(II) aq -Magnetite Fractionations The fractionation between Fe(II) aq and magnetite is best constrained by samples from 39 d and later from experiment 2, after initial rapid Fe(II) sorption and HFO phase conversion to lepidocrocite occurred (Fig. 12A). Because Fe mass balance always heavily favored the solid phase, measured Fe(II)aq-Total Solid fractionations will be equal to Fe(II)aq-Magnetite when magnetite is the dominant solid-phase product. The relatively constant Fe(II)aq-Total Solid values observed at low Fe reduction rates (Fig. 12A), in addition to the fact that XRD spectra and SEM images indicate that the biogenic magnetite was very fine grained (Table 3; electronic annex EA-2), suggests that the average Fe(II)aq-Total Solid for the last six time points most closely reflects an equilibrium isotope fractionation Fe(II) aq -Magnetite-NMNC Fe(II)(s) Fractionations Production of biogenic magnetite and siderite in experimental studies of DMRB is often accompanied by formation of poorly crystalline Fe(II)-bearing solid phases that are difficult to detect using XRD spectra, but may be identified through partial dissolution of solids using weak acids (e.g., Fredrickson et al., 1998; Roden et al., 2002). Although we have defined the poorly crystalline, nonmagnetic, noncarbonate Fe(II) component as NMNC Fe(II)(s), we recognize that this may represent several species, including sorbed Fe(II). Partial dissolutions of the solid products at 16, 68, and 242 d for experiments 2 and 3 may be broadly interpreted to reflect variable mixtures of a pure-fe(ii) component and a Fe(III)-rich component that may be magnetite and/or unreacted HFO or lepidocrocite, or even a mixed Fe(III)-Fe(II) phase such as green rust (Fig. 4). The pure-fe(ii) component for the 16 d samples most likely reflects sorbed Fe(II) in experiments 2 and 3. We focus on the partial dissolutions that produced 85% Fe(II) because partial dissolutions that have larger proportions of Fe(III) probably reflect

25 Fe isotope fractionation in biogenic magnetite and siderite 987 NMNC Fe(II)(s) (Fig. 12A) reflected continued sequestration of low- 56 Fe NMNC Fe(II)(s). Partial dissolution of the solid product from the 242 d sample in experiment 3 did not recover all NMNC Fe(II)(s) as calculated for the total solid (28%; Table 4), which suggests that this component was not open to isotopic exchange with Fe(II) aq. We therefore interpret the Fe(II) aq - NMNC Fe(II)(s) fractionations of 0.6 to 0.8 to reflect a disequilibrium, possibly kinetic, fractionation Fe(II) aq -Magnetite-Siderite-NMNC Fe(II)(s) Fractionations Fig. 12. Fe(II) aq -solid fractionations as a function of Fe reduction rate and time for experiments 2, 3, 5, and 6. Fractionations inferred from partial dissolutions are for samples that contained 85% Fe(II), so as to isolate the siderite and/or NMNC Fe(II)(s) component(s) from ferric hydroxide or magnetite (see text). For Fe reduction rates less than 0.01, data are plotted in relative order to the right of the scale break. The contrast in Fe(II) aq -solid fractionations between experiments 2 and 3 (panel A) are interpreted to reflect equilibrium conditions at low Fe reduction rates in experiment 2, but kinetic effects throughout experiment 3 due to extensive formation of NMNC Fe(II)(s). The continuously increasing Fe(II) aq -solid fractionations in experiments 5 and 6 (panel B) are interpreted to reflect the increasing proportion of siderite and NMNC Fe(II)(s) with time and large positive Fe(II) aq -siderite and Fe(II) aq -NMNC Fe(II)(s) fractionations under disequilibrium or kinetic conditions (Table 8) in response to the large initial Fe reduction rates in experiments 5 and 6. Data from Tables 2, 4, 6, and 7. dissolution of magnetite and/or unreacted ferric hydroxide, making it difficult to interpret the measured data. In experiment 2, partial dissolution of the 242 d sample satisfies this criterion, and the Fe(II) aq -NMNC Fe(II)(s) fractionation is 0.2 (Table 3; Fig. 12A). It is possible, however, that this fractionation may reflect that between Fe(II) aq and Fe(II) sorbed. The high proportion of Fe(II) and fraction dissolved in the partial dissolution treatments of the 68 and 242 d samples from experiment 3 (Fig. 4) indicates the presence of an Fe(II) component that is greater than that which can be accounted for by sorbed Fe(II) using reasonable sorption capacities for magnetite. Simple mass-balance calculations using the measured Fe(II) and Fe(III) and inorganic carbon contents suggests that the total solid phase at the end of experiment 3 was 28% NMNC Fe(II)(s), 68% magnetite, and 5% siderite (Table 4). The Fe(II) aq -NMNC Fe(II)(s) fractionation at 68 and 242 d is 0.8 and 0.6, respectively (Table 4; Fig. 12A), suggesting that the continued increase in the 56 Fe values of Fe(II) aq with time in experiment 3 (Fig. 3D) and changes in Fe(II)aq- Coprecipitation of magnetite, siderite, and NMNC Fe(II)(s) in experiments 5 and 6 produced large changes in the measured Fe(II)aq-Total Solid fractionations as a function of time and Fe reduction rate (Fig. 12). Similar changes were observed for experiment 4 (Table 5). These changes must reflect in part changes in the proportions of the different solid phases, as highlighted in Figure 7D (right axis) and listed in Tables 6 and 7. In addition, changes in the overall fractionation factor also likely reflect differences in the specific fractionation factors for the different phases, as indicated by variations in the isotopic fractionations measured for different partial dissolutions (Fig. 12B). The fractionations between Fe(II) aq and the solid components released during partial dissolution ( Fe(II)aq-X ) for the 20 min 1% HAc dissolution seem most likely to record those between Fe(II) aq and NMNC Fe(II)(s), because this was the most mild acid treatment and NMNC Fe(II)(s) is expected to be relatively soluble. If this reasoning is valid, the inferred Fe(II) aq -NMNC Fe(II)(s) fractionation varied from 0.3 to 0.5 from high to low Fe reduction rates (Fig. 12B). The Fe(II) aq -X fractionations for the more aggressive 12 h 1% HAc and 0.1 M HCl partial dissolutions most likely reflect a mixture of siderite and NMNC Fe(II)(s), and are therefore more difficult to interpret, although their trends with time are generally consistent with an increasing proportion of siderite and NMNC Fe(II)(s), as suggested by the Fe(II) and IC contents (Table 6; Fig. 7D). Production of positive overall Fe(II) aq -solid fractionations with time (Fig. 12B) requires that the Fe(II) aq -siderite and Fe(II) aq -NMNC Fe(II)(s) fractionations were consistently positive, such as were measured using the bulk carbonate in experiment 1 and that measured in experiment 3, respectively, where these were interpreted to reflect disequilibrium conditions. These observations may support the interpretation that large positive Fe(II) aq -siderite or Fe(II) aq -NMNC Fe(II)(s) fractionations reflect kinetic effects, given the high Fe reduction rates associated with experiments 5 and 6 (Fig. 12B) Comparison of Fe(II) aq -Magnetite Fractionations from Other Studies Calculated and measured Fe(II) aq -magnetite fractionations encompass a range of 4 to0 (Fig. 13), which is relatively broad compared to the few per mil variations documented for chemically precipitated sediments to date (e.g., Beard and Johnson, 2004a). The largest Fe(II)-Magnetite fractionation of 4.2 (Fig. 13) is one calculated from spectroscopic data, using the b factors from Polyakov and Mineev (2000) and Schauble et al. (2001). More modest Fe(II)-Magnetite fractionations of ca. 1.2 to 3.2 (Fig. 13) have been calculated for

26 988 C. M. Johnson et al. Fig. 13. Comparison of isotopic fractionations between Fe in solution (ferrous and ferric species) and magnetite from predictions based on spectroscopic data (Polyakov and Mineev, 2000; Schauble et al., 2001), natural samples (Johnson et al., 2003), and experimental studies of biogenic magnetite formation (Mandernack et al., 1999; this study). All fractionations calculated relative to 56 Fe/ 54 Fe ratios. Error bars shown reflect reported uncertainties; analytical errors for data reported in this study are smaller than the size of the symbol. Fe(II) aq -magnetite fractionations shown on left side of figure, and Fe(III) aq -magnetite fractionations shown on right side of figure. The measured Fe(II) aq - magnetite fractionation in this study also has been converted to Fe(III) aq -magnetite using the Fe(III) aq -Fe(II) aq fractionation reported by Johnson et al. (2002) and Welch et al. (2003), so that these results may be compared to the Fe(III) aq -magnetite fractionations measured for magnetotactic bacteria by Mandernack et al. (1999). natural samples (Johnson et al., 2003) using the hexaquo Fe(II) factor from Schauble et al. (2001). These values more closely approximate the fractionation of 1.3 estimated from experiment 2. The Fe(II) aq -magnetite fractionations measured here are significantly different from those determined for magnetotactic bacteria (Fig. 13; Mandernack et al., 1999). It is unclear whether the Fe(II)-Magnetite fractionations for magnetotactic bacteria (Mandernack et al., 1999) reflect equilibrium conditions because of inconsistencies in the Fe(II)-magnetite and Fe(III)-magnetite fractionations. For example, Mandernack et al. (1999) report similar Fe(II) aq -magnetite and Fe(III) aq -magnetite fractionations, which would not be expected if the experimental conditions reflected equilibrium fractionation, given the 2.8 to 3.0 fractionation between Fe(III) and Fe(II) in solution (Johnson et al., 2002; Welch et al., 2003). Using the Fe(III) aq -Fe(II) aq fractionation measured by Johnson et al. (2002) and Welch et al. (2003), recalculation of the estimated equilibrium Fe(II) aq -magnetite fractionation measured in the current study to Fe(III) aq -magnetite produces a Fe(III)-Magnetite fractionation of 1.7, which stands in marked contrast to that measured for three experiments using ferric chloride reported by Mandernack et al. (1999) (Fig. 13). Calculation of Fe(III)-Magnetite by this method assumes isotopic equilibrium exists between Fe(III) aq and magnetite, independent of the pathways by which isotopic equilibrium may be attained in a system that contains mixed valance states of Fe. Most of the magnetotactic bacteria experiments were done at 28 C, which produced very rapid formation of magnetite, and it seems likely that the measured isotope fractionations instead reflect kinetic isotope effects, where, as has been shown by other Fe isotope studies, very rapid precipitation of minerals produces no isotopic contrast between mineral and solution Fe (Skulan et al., 2002). We have no explanation for why the low-temperature experiment (4 C) reported by Mandernack et al. (1999) produced the same result as the higher-temperature experiments, which is unexpected given the significant temperature dependence of Fe isotope fractionation factors, as calculated from spectroscopic data (Polyakov and Mineev, 2000; Schauble et al., 2001) or measured experimentally (Welch et al., 2003) Summary of Fractionation Factors The Fe(II) aq -X fractionation factors produced during biogenic mineral formation by DMRB are summarized in Table 8. Although the speciation of aqueous Fe(II) is not precisely known in these studies, the majority of Fe(II) is likely to have existed as hexaquo or hydroxyl species. It is premature to surmise if the inferred equilibrium fractionation factors are unique to DMRB systems because relatively few experimental studies have determined equilibrium fractionation factors in equivalent abiotic systems, and the uncertainties in predicted fractionation factors or those estimated from natural minerals remain large. Wiesli et al. (2004) estimated the equilibrium [Fe II (H 2 O) 6 ] 2 -FeCO 3 fractionation factor at 20 C to be based on slow carbonate precipitation experiments, although it is not yet clear if this is significantly different than Table 8. Summary of Fe isotope fractionations produced by DMRB. Disequilibrium/Kinetic Equilibrium Species A B A B Ref. Fe(II) aq - FeCO Fe(II) aq -Ca 0.15 Fe 0.85 CO Fe(II) aq -Fe 3 O Fe(II) aq - Ferric oxide/hydroxide substrate (DIR) Fe(II) aq - NMNC Fe(II)(s) Fe(II) aq - Fe(II) [HFO] Fe 3 O 4 - FeCO Fe 3 O 4 -Ca 0.15 Fe 0.85 CO References: 1. This study. 2. Beard et al. (1999; 2003a); this study. 3. Icopini et al. (2004); this study.

27 Fe isotope fractionation in biogenic magnetite and siderite 989 the near-zero fractionation we infer for biogenic siderite formation, given the uncertainties in interpreting the partial dissolution results from experiment 1. The disequilibrium or kinetic fractionation factors are significantly different than those estimated to reflect equilibrium conditions. The Fe(II) aq -Fe(II) sorbed fractionation of 2.1 calculated by Icopini et al. (2004) can explain the low 56 Fe values for Fe(II) aq measured during early periods of rapid Fe reduction, but this fractionation appears to be closer to zero under equilibrium conditions. We interpret the Fe(II) aq -Fe(II) sorbed fractionation measured in experiment 2 to be more reliable than that measured in experiment 3 because of the possibility that NMNC Fe(II)(s) was also present in experiment 3 at early time periods. Alternatively, there may be a difference in the isotopic fractionation of sorption to HFO or lepidocrocite. The Fe(II) aq -NMNC Fe(II)(s) fractionation under disequilibrium conditions appears to be larger than the near-zero fractionation inferred to be appropriate for equilibrium conditions. Because natural environments generally involve more crystalline ferric oxide/hydroxide minerals and are nutrient poor relative to most experimental studies of DMRB (e.g., Glasauer et al., 2003), we infer that the fractionation factors that are most applicable to natural systems are those which are estimated to reflect equilibrium conditions in the experiments at low Fe reduction rates Applications to Fe Cycling during Marine Basin Diagenesis and the Origin of Banded Iron Formations Reaction of aqueous Fe(II) and dissolved carbonate with ferric oxides to form siderite and magnetite has been hypothesized to be an important diagenetic process in deep marine basins during formation of some BIFs (e.g., Klein and Beukes, 1989; Beukes et al., 1990; Kaufman, 1996; Sumner, 1997). The sources of Fe(II) aq for such reactions may have included Mid- Ocean Ridge (MOR) hydrothermal sources (e.g., Klein and Beukes, 1989; Beukes and Klein, 1990), as well as reductive dissolution of ferric oxides by Fe(III)-reducing bacteria (e.g., Walker, 1984; Lovley et al., 1987; Nealson and Myers, 1990). Below we use the predicted ranges in 56 Fe values for biogenic magnetite and siderite, based on the Fe isotope fractionations summarized in Table 8, to infer the likely sources of Fe(II) aq (i.e., derived from either MOR sources or dissimilatory Fe(III) reduction) for generation of magnetite and Fe carbonate-bearing BIFs. We compare our experimental results to the observed Fe isotope variations in the exceptionally well-preserved sections of the 2.5 Ga Kuruman Iron Formation in the Transvaal Craton, which have not been subjected to significant metamorphism and are accepted to be some of the freshest BIFs known (Johnson et al., 2003, and references within). The results from the current study suggest that the equilibrium magnetite-siderite fractionation factor at low temperatures (ca. 22 C) is ca. 1.3 (Table 8), and yet few magnetitesiderite pairs from adjacent layers in BIFs plot near this value (Fig. 14). Instead, the data for BIFs that are available so far suggest that Fe isotope fractionations measured for adjacent magnetite-siderite pairs scatter from the predicted Magnetite- Siderite 1.3 to Magnetite-Siderite 0 (Fig. 14). There is Fig. 14. Iron isotope compositions of magnetite and Fe carbonates ( siderite ) from adjacent bands in the 2.5 Ga Kuruman Iron Formation, Transvaal craton (Johnson et al., 2003), compared to Fe(II) aq sources and Fe pathways inferred from the Fe isotope fractionations measured in this study. In part A, measured 56 Fe values for magnetite and siderite are compared, and the data scatter from the equilibrium magnetite-siderite fractionations of 1.3 (Table 8) to a near-zero fractionation. The 56 Fe values of Fe(II) aq that would be in equilibrium with these minerals are given in the top and right scales, based on the Fe(II) aq -mineral fractionations in Table 8. In part B, measured 56 Fe Mt values are compared to the measured magnetite-fe carbonate fractionations ( Magnetite-Fe Carbonate ), and these variations scatter about a mixing line where the 56 Fe value for Fe carbonate remains constant but where the 56 Fe for magnetite changes in response to different isotopic compositions of Fe(II) aq (mixing line also shown in part A). This model assumes that isotopic equilibrium is maintained between Fe(II) aq and magnetite, but not between Fe(II) aq and siderite. The data are interpreted to reflect mixing between Fe(II) aq that was derived from MOR hydrothermal sources and Fe(II) aq produced by DMRB, where decreasing 56 Fe values may reflect increasing amounts of biologic cycling of Fe. no evidence that the magnetite-siderite fractionations in BIFs are as large as those predicted by Polyakov and Mineev (2000), who calculate a 6.2 fractionation in 56 Fe/ 54 Fe at 25 C. The range in 56 Fe values measured for siderite in BIFs is significant, up to 1, and we suspect that some of this may be due to cation substitution, including Mn, Mg, and Ca, based on the significant isotopic effect of Ca substitution observed in this study, and the fact that natural siderite is rarely end-member FeCO 3. It is unknown if the magnetite-siderite fractionation factor determined here is unique to biologic systems. Distinction between biologic and abiologic Fe cycling, however, may be made based on calculated Fe isotope compositions for Fe(II) aq.

28 990 C. M. Johnson et al. Using the Fe(II) aq -mineral fractionation factors from Table 8, adjacent magnetite and siderite bands in BIFs may have formed from a common Fe-bearing fluid where the 56 Fe value of magnetite is 0.5 and Magnetite-Siderite is ca. 1.3 (Fig. 14). In such cases, the calculated 56 Fe for Fe(II) aq is ca. 0.5, which is strikingly similar to the isotopic compositions measured for modern Mid-Ocean Ridge (MOR) hydrothermal fluids (Sharma et al., 2001; Beard et al., 2003b). It is unknown if the Fe isotope compositions of MOR fluids in the Archean was similar to those measured today, but, in the absence of data to the contrary, we will assume that this is the case. For magnetite that has 56 Fe 0, the calculated 56 Fe values for Fe(II) aq are 1.3 (Fig. 14), which lies in the range expected to be produced by DMRB, and we interpret such isotopic compositions to strongly implicate a role for biology in producing magnetite in BIFs. If magnetite that has 56 Fe 0.5 formed from Fe(II) aq that had very low 56 Fe values, this may reflect kinetic effects due to very rapid Fe(III) reduction rates, or biologic cycling of Fe, which would be expected to compound the isotopic fractionation (e.g., Johnson et al., 2004). For adjacent magnetite and siderite bands in BIFs that have Magnetite-Siderite ca. 0, these minerals could not have formed from a common fluid, and in these cases we suggest that such magnetite was produced by DMRB. The 56 Fe values for all siderite layers seem to be well explained through precipitation from Fe(II) aq that was derived from MOR hydrothermal fluids, although an important avenue of future research is to constrain the effects of minor cation substitution on the isotopic variability observed in siderite from BIFs. If the Fe isotope variations in BIFs reflect mixing between MOR and bacterial sources of Fe(II) aq in magnetite, the measured 56 Fe values should reflect the relative contributions of these components, as well as any ferric hydroxide that may have been a precursor phase. Detrital ferric hydroxide would be expected to have 56 Fe values near zero (Beard et al., 2003b; Beard and Johnson, 2004a), suggesting that detrital ferric hydroxide, combined with MOR-sourced Fe(II) aq, is unlikely to produce magnetite that has 56 Fe 0. In contrast, oxidation of MOR-sourced Fe(II) aq could produce ferric hydroxide that has 56 Fe 0, and, ultimately, magnetite that has 56 Fe 0, given the 0.9 to 1.5 ferric hydroxide-fe(ii) aq fractionations that have been measured during abiotic oxidation by O 2 (Bullen et al., 2001) or anoxygenic photosynthesis (Croal et al., 2004). From a mass-balance perspective, production of magnetite that has 56 Fe 0 is most easily accomplished through reductive dissolution of detrital ferric hydroxide ( 56 Fe ca. 0 ), combined with a significant pool of low- 56 Fe Fe(II) aq that has 56 Fe 1.0 to 1.5, although it is also possible that ferric hydroxide produced by Fe(II) oxidation could also produce low- 56 Fe magnetite with a proportionally larger contribution from low 56 Fe Fe(II) aq. Although abiotic reductive dissolution of ferric hydroxide by sulfide can occur (e.g., Poulton, 2003), the paucity of sulfide minerals in magnetite layers in fresh BIFs makes this mechanism unlikely. Large changes in 56 Fe values for ambient Fe(II) aq over the timescales involved in deposition of alternating cm-thick magnetite- and siderite-rich layers is very unlikely for Archean BIFs, given the expected long residence time for Fe and accompanying resistance to changes in isotopic compositions (Johnson et al., 2003). Instead, the low 56 Fe values inferred for Fe(II) aq from low 56 Fe magnetite may reflect the isotopic compositions of interstitial pore waters and/or bottom waters that were closely associated with DMRB, and not those of the open oceans. 6. CONCLUSIONS Successful application of Fe isotope geochemistry to problems in Earth science rests on detailed experimental determination of Fe isotope fractionation factors for biologic and abiotic systems. Although predicted isotopic fractionations based on spectroscopic data provide useful guides, uncertainties in calculated fractionations remain relatively large compared to the range in Fe isotope compositions thus far observed in nature. New experimental data reported here for biogenic magnetite and Fe-carbonate formation during dissimilatory Fe(III) oxide reduction suggest equilibrium 56 Fe/ 54 Fe fractionation factors of 1.3, 0.0, and 0.9 for Fe(II) aq -magnetite, Fe(II) aq -siderite, and Fe(II) aq - ankerite (Ca-bearing Fe carbonate) fractionations, respectively, at room temperature (22 C). Our results support the inference from spectroscopic data and natural samples that carbonate stoichiometry may exert a major control on Fe isotope fractionation factors. Rapid Fe(III) reduction and unique Fe biomineralization pathways have the potential to produce substantially larger Fe(II) aq - mineral fractionations. Rapid sorption of aqueous Fe(II) to ferric hydroxide surfaces produces anomalously low 56 Fe values for Fe(II) aq, which is interpreted to be a kinetic effect. At equilibrium, however, sorbed Fe(II) appears to have Fe isotope compositions that are similar to those of aqueous Fe(II). During rapid Fe(III) reduction, amorphous nonmagnetic, noncarbonate solid Fe(II) phases that have low 56 Fe values may also significantly affect the Fe isotope compositions of aqueous Fe(II) when the system is not in isotopic equilibrium, in a direction opposite to the effect of Fe(II) sorption. However, at equilibrium, amorphous, nonmagnetic, noncarbonate Fe(II) solids appear to have Fe isotope compositions that are similar to those of aqueous Fe(II). It is not yet clear if the fluid-mineral fractionations measured in this study are distinct from those in abiotic or inorganic systems at isotopic equilibrium. Experimental determination of equilibrium Fe isotope fractionation factors in equivalent inorganic systems will be required before we can confidently conclude that there is an Fe isotope vital effect for the system Fe(II) aq -magnetite-fe carbonate. Our results, however, confirm that in the system Fe(II) aq -ferric oxide/hydroxide, low 56 Fe values for Fe(II) aq appear to be unique to biology in nonsulfidic systems under conditions of low Fe reduction rates that most closely approximate natural conditions. Based on the Fe(II) aq -mineral isotope fractionation factors measured here, the contrast in 56 Fe values for Fe carbonates and magnetite from adjacent bands in BIFs can be explained by one of two reaction pathways: 1) in cases where the 56 Fe of magnetite is positive, formation of magnetite and siderite from an abiotic, MOR-derived hydrothermal source of Fe(II) aq that has a 56 Fe of 0.5 is suggested. 2) Fe biomineralization coupled to dissimilatory Fe(III) oxide reduction provides a plausible mechanism for generation of magnetite that has 56 Fe values 0. Although photosynthetic oxidation of Fe(II) has long been the major focus in models that call upon a role for biology in the genesis of BIFS (e.g., Cloud, 1968; Widdel et al.,

29 Fe isotope fractionation in biogenic magnetite and siderite ; Konhauser et al., 2002), our results provide support for previous suggestions (Walker, 1984; Lovley et al., 1987; Nealson and Myers, 1990; Lovley, 1991) that dissimilatory Fe(III) reduction may have been an important pathway for Fe mineralization (and fractionation) during formation of BIFs. Acknowledgments This work reflects an outgrowth from our initial work on Fe isotope fractionation produced by dissimilatory Fe(III) reduction with Prof. Ken Nealson and his group, and Johnson and Beard are indebted to Ken and his colleagues for extensive discussions on the subject. We thank Rebecca Poulson, Silke Severmann, and Rene Wiesli for their invaluable assistance in several aspects of the analytical work, as well as comments on an early version of the manuscript. Journal reviews were provided by three anonymous reviewers, and these, in addition to the comments by AE Kurt Kyser, led to substantial improvement of the paper. This research was supported by the Packard Foundation under a contract to Dianne Newman, as well as the NASA Astrobiology Institute under contracts to Clark Johnson, Ken Nealson, and Jill Banfield. Associate editor: K. Kyser REFERENCES Albarède F. and Beard B. L. (2004) Analytical methods for nontraditional isotopes. In Geochemistry of Non-Traditional Stable Isotopes (eds. C. M. Johnson, B. L. Beard and F. Albarède), pp Rev. in Mineralogy and Geochemistry 55.. Anbar A. D., Roe J. E., Barling J., and Nealson K. H. (2000) Nonbiological fractionation of iron isotopes. Science 288, Arnold R. G., Di Christina T. J., and Hoffman M. R. (1988) Reductive dissolution of Fe(III) oxides by Psuedomonassp Biotechnol. Bioengin. 32, Baker J. C., Kassan J., and Hamilton P. J. (1995) Early diagenetic siderite as an indicator of depositional environment in the Triassic Rewan Group, southern Bowen Basin, eastern Australia. Sedimentology 43, Beard B. L. and Johnson C. M. (1999) High precision iron isotope measurements of terrestrial and lunar material. Geochim. Cosmochim. Acta 63, Beard B. L., Johnson C. M., Cox L., Sun H., Nealson K. H., and Aguilar C. (1999) Iron isotope biosignatures. Science 285, Beard B. L., Johnson C. M., Skulan J. L., Nealson K. H., Cox L., and Sun H. (2003a) Application of Fe isotopes to tracing the geochemical and biological cycling of Fe. Chem. Geol. 195, Beard B. L., Johnson C. M., Von Damm K. L., and Poulson R. L. (2003b) Iron isotope constraints on Fe cycling and mass balance in oxygenated Earth oceans. Geology 31, Beard B. L. and Johnson C. M. (2004a) Fe isotope variations in the modern and ancient Earth and other planetary bodies. In Geochemistry of Non-Traditional Stable Isotopes (eds. C. M. Johnson, B. L. Beard and F. Albarède), pp Rev. in Mineralogy and Geochemistry 55, Mineralogical Society of America and Geochemical Society, Washington DC. Beard B. L. and Johnson C. M. (2004b) Inter-mineral Fe isotope variations in mantle derived rocks and implications for the Fe geochemical cycle. Geochim. Cosmochim. Acta. 68, Beukes N. J. and Klein C. (1990) Geochemistry and sedimentology of a facies transition from microbanded to granular iron-formation in the early Proterozoic Transvaal Supergroup, South Africa. Precamb. Res. 47, Beukes N. J., Klein C., Kaufman A. J., and Hayes J. M. (1990) Carbonate petrography, kerogen distribution and carbon and oxygen isotope variations in an Early Proterozoic transition from limestone to iron-formation deposition, Transvaal Supergroup, South Africa. Econ. Geol. 85, Blakemore R. P. and Frankel R. B. (1989) Biomineralization by magnetogenic bacteria. In Metal-Microbe Interactions (eds. R. K. Poole and G. M. Gadd), pp , IRL Press, Oxford, UK. Brantley S. L., Liermann L., and Bullen T. D. (2001) Fractionation of Fe isotopes by soil microbes and organic acids. Geology 29, Brantley S. L., Liermann L. J., Guynn R. L., Anbar A., Icopini G. A., and Barling J. (2004) Fe isotopic fractionation during mineral dissolution with and without bacteria. Geochim Cosmochim Acta, 68, Bullen T. D., White A. F., Childs C. W., Vivit D. V., and Schultz M. S. (2001) Demonstration of significant abiotic iron isotope fractionation in nature. Geology 29, Caccavo F., Blakemore R. P., and Lovley D. R. (1992) A hydrogenoxidizing, Fe(III)-reducing microorganism from the Great Bay estuary, New Hampshire. Appl. Environ. Microbiol. 58, Canfield D. E. and Berner R. A. (1987) Dissolution and pyritization of magnetite in anoxic marine sediments. Geochim. Cosmochim. Acta 51, Chang S. B. R. and Kirschvink J. L. (1985) Possible biogenic magnetite fossils from the Late Miocene Potamida clays of Crete. In Magnetite Biomineralization and Magnetoreception in Organisms (eds. J. L. Kirschvink, D. S. Jones and B. J. MacFadden), pp Plenum Press. Chang S. B. R., Stolz J. F., Kirschvink J. L., and Awramik S. M. (1989) Biogenic magnetite in stromatolites II. Occurrences in ancient sedimentary environments. Precamb. Res. 43, Cloud P. (1968) Atmospheric and hydrospheric evolution on the primitive earth. Science 160, Coleman M. L. (1993) Microbial processes Controls on the shape and composition of carbonate concretions. Mar. Geol. 113, Coleman M. L., Hedrick D. B., Lovely D. R., White D. C., and Pye K. (1993) Reduction of Fe(III) in sediments by sulphate-reducing bacteria. Nature 361, Cornell R. M. and Schwertmann U. (1996) The Iron Oxides: Structure, Properties, Reaction, Occurrence and Uses. VCH, Weinheim, Germany. Croal L. R., Johnson C. M., Beard B. L., and Newman D. K. (2003) Iron isotope fractionation by anoxygenic Fe(II)-phototrophic bacteria. Geochim. Cosmochim. Acta 68, Eggar-Gibbs Z. G., Jude B., Dominik J., Loizeau J.-L., and Oldfield F. (1999) Possible evidence for dissimilatory bacterial magnetite dominating the magnetic properties of recent lake sediments. Earth Planet. Sci. Lett. 168, 1 6. Frankel R. B., Blakemore R. P. and Wolfe R. S. (1979) Magnetite in freshwater magnetotactic bacteria. Science 203, Frankel R. B., Blakemore R. P., Torres de Araujo F. F., Esquivel D. M. S., and Danon J. (1981) Magnetotactic bacteria at the geomagnetic equator. Science 212, Fredrickson J. K., Zachara J. M., Kennedy D. W., Dong H., Onstott T. C., Hinman N. W., and Li S. (1998) Biogenic iron mineralization accompanying the dissimilatory reduction of hydrous ferric oxide by a groundwater bacterium. Geochim. Cosmochim. Acta 62, Hendry J. P. (2002) Geochemical trends and paleohydrological significance of shallow burial calcite and ankerite cements in Middle Jurassic strata on the East Midlands Shelf (onshore UK). Sed. Geol. 151, Hernandez M. E. and Newman D. K. (2001) Revie: Extracellular electron transfer. CMLS Cell. Mol. Life Sci. 58, Glasauer S., Weidler P. G., Langley S., and Beveridge T. J. (2003) Controls on Fe reduction and mineral formation by a subsurface bacterium. Geochim. Cosmochim. Acta 67, Icopini G. A., Anbar A. D., Ruebush S. S., Tien M., and Brantley S. L. (2004) Iron isotope fractionation during microbial reduction of iron: The importance of adsorption. Geology 32, Johnson C. M., Skulan J. L., Beard B. L., Sun H., Nealson K. H., and Braterman P. S. (2002) Isotopic fractionation between Fe(III) and Fe(II) in aqueous solutions. Earth Planet. Sci. Lett. 195, Johnson C. M., Beard B. L., Beukes N. J., Klein C., and O Leary J. M. (2003) Ancient geochemical cycling in the Earth as inferred from Fe isotope studies of banded iron formations from the Transvaal Craton. Contrib. Mineral. Petrol. 144, Johnson C. M., Beard B. L., Roden E. E., Newman D. K. and Nealson K. H. (2004) Isotopic constraints on biogeochemical cycling of Fe. In Geochemistry of Non-Traditional Stable Isotopes (eds. C. M.

30 992 C. M. Johnson et al. Johnson, B. L. Beard and F. Albarède), pp Rev. in Mineralogy and Geochemistry 55 Mineralogical Society of America and Geochemical Society, Washington, DC. Karlin R. (1990a) Magnetite diagenesis in marine sediments from the Oregon continental margin. J. Geophys. Res. 95, Karlin R. (1990b) Magnetite mineral diagenesis in suboxic sediments at Bettis site W-N, NE Pacific Ocean. J. Geophys. Res. 95, Karlin R., Lyle M., and Heath C. R. (1987) Authigenic magnetite formation in suboxic marine sediments. Nature 326, Katz B., Elmore R. D., Cogoini M., and Ferry S. (1998) Widespread chemical remagnetization: Orogenic fluids or burial diagenesis of clays. Geology 26, Katz B., Elmore R. D., Cogoini M., Engel M. H., and Ferry S. (2000) Associations between burial diagenesis of smectite, chemical remagnetization and magnetite authigenesis in Vocontian trough, SE France. J. Geophys. Res. 105, Kaufman A. J. (1996) Geochemical and mineralogic effects of contact metamorphism on banded iron-formation: An example from the Transvaal Basin, South Africa. Precamb. Res. 79, Klein C. (1974) Greenalite, stilpnomelane, minnesotaite, crocidolite and carbonates in a very low-grade metamorphic Precambrian iron-formation. Can. Mineral. 12, Klein C. (1983) Diagenesis and metamorphism of Precambrian banded iron-formations. In Iron-formations: Facts and Problems (eds. A.F. Trendall and R.C. Morris), pp , Elsevier, Amsterdam. Klein C. and Beukes N. J. (1989) Geochemistry and sedimentology of a facies transition from limestone to iron-formation deposition in the early Proterozoic Transvaal Supergroup, South Africa. Econ. Geol. 84, Klein C. and Beukes N. J. (1992) Proterozoic iron-formations. In Proterozoic Crustal Evolution (ed. K. C. Condie), pp , Elsevier, Amsterdam. Konhauser K. (1998) Diversity of bacterial iron mineralization. Earth- Sci. Rev. 43, Konhauser K. O., Hamade T., Raiswell R., Morrice R. C., Ferris F. G., Southam G., and Canfield D. E. (2002) Could bacteria have formed the Precambrian banded iron formations? Geology 30, Laverne C. (1993) Occurrence of siderite and ankerite in young basalts from the Galápagos Spreading Center (DSDP Holes 506G and 507B). Chem. Geol. 106, Leslie B. W., Hammond D. E., Berelson W. M., and Lund S. P. (1990) Diagenesis in anoxic sediments from the California continental borderland and its influence on iron, sulfur and magnetite behavior. J. Geophys. Res. 95, Liu C., Zachara J. M., Gorby Y. A., Szecsody J. E., and Brown C. F. (2001) Microbial reduction of Fe(III) and sorption/precipitation of Fe(II) on Shewanella putrefaciens strain CN32. Environ. Sci. Technol. 35, Lovley D. R. (1991) Dissimilatory Fe and Mn reduction. Microbiol. Rev. 55, Lovley D. R., Stolz J. F., Nord G. L. Jr., and Phillips E. J. P. (1987) Anaerobic production of magnetite by a dissimilatory iron-reducing microorganism. Nature 330, Lovley D. R., Woodward J. C., and Chapelle F. H. (1994) Stimulated anoxic biodegradation of aromatic hydrocarbons using Fe(III) ligands. Nature 370, Lovley D. R., Coates J. D., Blunt-Harris E. L., Phillips E. J. P., and Woodward J. C. (1996) Humic substances as electron acceptors for microbial respiration. Nature 382, Lovley D. R. and Woodward J. C. (1996) Mechanisms for chelator stimulation of microbial Fe(III) oxide reduction. Chem. Geol. 132, Lloyd J. R., Blunt-Harris E. L., and Lovley D. R. (1999) The periplasmic 9.6 Kilodalton c-type cytochrome of Geobacter sulfurreducens is not an electron shuttle to Fe(III). J. Bacteriol. 181, Macquaker J. H. S., Curtis C. D., and Coleman M. L. (1997) The role of iron in mudstone diagenesis: Comparison of Kimmeridge clay formation mudstones from onshore and offshore (UKCS) localities. J. Sed. Res. 67, Mandernack K. W., Bazylinski D. A., Shanks W. C. III, and Bullen T. D. (1999) Oxygen and iron isotope studies of magnetite produced by magnetotactic bacteria. Science 285, Matthews A., Zhu X-K., and O Nions K. (2001) Kinetic iron stable isotope fractionation between iron (-II) and (-III) complexes in solution, Earth Planet. Sci. Lett. 192, Matthews A., Morgans-Bell H. S., Emmanuel S., Jenkyns H. C., Erel Y., and Halicz L. (2004) Controls on iron-isotope fractionation in organic-rich sediments (Kimmeridge Clay, Upper Jurassic, southern England). Geochem. Cosmochim. Acta, 68, Mortimer R. J. G., Coleman M. L., and Rae J. E. (1997) Effect of bacteria on the elemental composition of early diagenetic siderite: Implication for paleoenvironmental interpretation. Sedimentology 44, Mozley P. S. and Carothers W. W. (1992) Elemental and isotopic compositions of siderite in the Kuparuk formation, Alaska: Effect of microbial activity and water/sediment interaction on early porewater chemistry. J. Sed. Petrol. 62, Mozley P. S. and Burns S. J. (1993) Oxygen and carbon isotopic composition of marine carbonate concretions: An overview. J. Sed. Petrol. 63, Myers C. R. and Nealson K. H. (1988) Bacterial manganese reduction and growth with manganese oxide as the sole electron acceptor. Science 240, Nealson K. H. and Myers C. R. (1990) Iron reduction by bacteria: A potential role in the genesis of banded iron formations. Am. J. Sci. 290A, Nealson K. H. and Stahl D. A. (1997) Microorganisms and biogeochemical cycles: What can we learn from layered microbial communities? In Geomicrobiology (eds. J. F. Banfield and K. H. Nealson), pp Rev. in Mineralogy and Geochemistry 35 Mineralogical Society of America and Geochemical Society, Washington, DC. Nevin K. P. and Lovley D. R. (2000) Lack of production of electronshuttling compounds or solubilization of Fe(III) during reduction of insoluble Fe(III) oxide by Geobacter metallireducens. Appl. Environ. Microbiol. 66, Nevin K. P. and Lovley D. R. (2002a) Mechanisms for Fe(III) oxide reduction in sedimentary environments. Geomicrobiol. J. 19, Nevin K. P. and Lovley D. R. (2002b) Mechanisms for accessing insoluble Fe(III) oxide during dissimilatory Fe(III) reduction by Geothrix Fermentans. App. Environ. Microbiol. 68, Newman D. K. and Kolter R. (2000) A role for excreted quinones in extracellular electron transfer. Nature 405, Parmar N., Warren L. A., Roden E. E., and Ferris F. G. (2000) Solid phase capture of strontium by the iron reducing bacteria Shewanella alga strain BrY. Chem. Geol. 169, Polyakov V. B. and Mineev S. D. (2000) The use of Mössbauer spectroscopy in stable isotope geochemistry. Geochim. Cosmochim. Acta 64, Poulton S. W. (2003) Sulfide oxidation and iron dissolution kinetics during the reaction of dissolved sulfide with ferrihydrite. Chem. Geol. 202, Pye K., Dickson J. A. D., Schiavon N., Coleman M. L., and Cox M. (1990) Formation of siderite-mg-calcite-iron sulphide concretions in intertidal marsh and sandflat sediments, north Norfolk, England. Sedimentology 37, Raiswell R. and Fisher Q. J. (2000) Mudrock-hosted carbonate concretions: A review of growth mechanisms and their influence on chemical and isotopic composition. J. Geol. Soc. 157, Roden E. E. and Lovley D. R. (1993) Dissimilatory Fe (III) reduction by the marine microorganism Desulfuromonas acetoxidans. Appl. Environ. Microbiol. 59, Roden E. E. and Zachara J. M. (1996) Microbial reduction of crystalline iron(iii) oxides: Influence of oxide surface area and potential for cell growth. Environ. Sci. Technol. 30, Roden E. E., Leonardo M. R., and Ferris F. G. (2002) Immobilization of strontium during iron biomineralization coupled to dissimilatory hydrous ferric oxide reduction. Geochim. Cosmochim. Acta 66, Roden E. E. and Urrutia M. M. (2002) Influence of biogenic Fe(II) on bacterial reduction of crystalline Fe(III) oxides. Geomicrobiol. J. 19,

31 Fe isotope fractionation in biogenic magnetite and siderite 993 Roe J. E., Anbar A. D., and Barling J. (2003) Nonbiological fractionation of Fe isotopes: Evidence of an equilibrium isotope effect. Chem. Geol. 195, Royer R. A., Burgos W. D., Fisher A. S., Unz R. F., and Dempsey B. A. (2002) Enhancement of biological reduction of hematite by electron shuttling and Fe(II) complexation. Environ. Sci. Technol. 36, Schauble E. A., Rossman G. R., and Taylor H. P. (2001) Theoretical estimates of equilibrium Fe-isotope fractionations from vibrational spectroscopy. Geochim. Cosmochim. Acta 65, Sharma M., Polizzotto M., and Anbar A. D. (2001) Iron isotopes in hot springs along the Juan de Fuca Ridge. Earth Planet. Sci. Lett. 194, Shyu J. B. H., Lies D. P., and Newman D. K. (2002) Protective role of tol C in efflux of the electron shuttle anthraquinone-2,6-disulfonate. J. Bacteriol. 184, Skulan J. L., Beard B. L., and Johnson C. M. (2002) Kinetic and equilibrium Fe isotope fractionation between aqueous Fe(III) and hematite. Geochim. Cosmochim. Acta 66, Stookey L. L. (1970) Ferrozine A new spectrophotometric reagent for iron. Anal. Chem. 42, Sumner D. Y. (1997) Carbonate precipitation and oxygen stratification in Late Archean seawater as deduced from facies and stratigraphy of the Gamohaan and Frisco Formations, Transvaal Supergroup, South Africa. Am. J. Sci. 297, Tarling D. H. and Turner P. (1999) Paleomagnetism and Diagenesis in Sediments. Special Publication 151. Geological Society of London. Urrutia M. M., Roden E. E., Fredrickson J. K., and Zachara J. M. (1998) Microbial and geochemical controls on synthetic Fe(III) oxide reduction by Shewanella alga strain BrY. Geomicrobiol. J. 15, Uysal I. T., Golding S. D., and Glikson M. (2000) Petrographic and isotope constraints on the origin and evolution of authigenic carbonate minerals and the associated fluid evolution in Late Permian coal measures, Bowen Basin (Queensland), Australia. Sed. Geol. 136, Vargas M., Kashefi K., Blunt-Harris E. L., and Lovley D. R. (1998) Microbiological evidence for Fe(III) reduction on early Earth. Nature 395, Walker J. C. G. (1984) Suboxic diagenesis in banded iron formations. Nature 309, Welch S. A., Beard B. L., Johnson C. M., and Braterman P. S. (2003) Kinetic and equilibrium Fe isotope fractionation between aqueous Fe(II) and Fe(III). Geochim. Cosmochim. Acta 67, Widdel F., Schnell S., Heising S., Ehrenreich A., Assmus B., and Schink B. (1993) Ferrous iron oxidation by anoxygenic phototrophic bacteria. Nature 362, Wiesli R. A., Beard B. L., and Johnson C. M. (2004) Experimental determination of Fe isotope fractionation between aqueous Fe(II), siderite and green rust in abiotic systems. Chem. Geol. 211, Zachara J. M., Kukkadapu R. K., Fredrickson J. K., Gorby Y. A., and Smith S. C. (2002) Biomineralization of poorly crystalline Fe(III) oxides by dissimilatory metal reducing bacteria (DMRB). Geomicrobiol. J. 19, Zhu X. K., O Nions R. K., Guo Y. L., and Reynolds B. C. (2000) Secular variation of iron isotopes in North Atlantic deep water. Science 287, Zhu X. K., Guo Y., O Nions R. K., Young E. D., and Ash R. D. (2001) Isotopic homogeneity of iron in the early solar nebula. Nature 412, APPENDIX SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at doi: /j.gca

32 Electronic Annex EA-1 DETAILS ON ANALYTICAL METHODS AND EXPERIMENTAL DESIGN Experimental constraints on Fe isotope fractionation during magnetite and Fe carbonate formation coupled to dissimilatory hydrous ferric oxide reduction by Clark M. Johnson, Eric E. Roden, Susan A. Welch, and Brian L. Beard Geochimica et Cosmochimica Acta Fe Isotope Analysis Isotopic analysis followed the methods used in Skulan et al. (2002) and Beard et al. (2003a). A standard-sample-standard approach was used to correct for instrumental mass bias. Our standard ionexchange methods provide excellent purification from cations such as Ca (e.g., Beard and Johnson, 2004b), which is essential to assure that the mass bias corrections calculated from standards are applicable to bracketed samples (Albarède and Beard, 2004). Iron isotope compositions were determined using a GV Instruments IsoProbe that utilizes a collision cell to remove Ar interferences, where a mixture of hydrogen and argon removed all ArN and ArO interferences on the Fe mass spectrum, as well as essentially all ArOH species. There was no evidence for polyatomic cation isobars such as 40 Ca 14 N or 54 Cr (isobaric with 54 Fe), 40 Ca 16 O (isobaric with 56 Fe), or 40 Ca 16 OH (isobaric with 57 Fe). Over a one-year period, analysis of three Fe solution standards gave the following values: UW J- M Fe: δ 56 Fe= +0.25±0.05 (n=47), δ 57 Fe= +0.39±0.07 ; UW HPS Fe: δ 56 Fe= +0.49±0.05, δ 57 Fe= +0.74±0.07 (n=52); IRMM-14 Fe: δ 56 Fe= -0.09±0.05, δ 57 Fe= -0.11±0.07 (n=54), where all uncertainties are 1σ external standard deviations. External reproducibility for the data presented here may be assessed from 99 replicate analyses. Sixty analyses were duplicated through analysis of the same Fe solution (after ion-exchange chromatography) in different analytical sessions, and the duplicates have an average reproducibility of ±0.06, where 85 % of the duplicates fell within 0.10 of the first analysis. Thirty-one samples were processed through the entire procedure, including ion-exchange separations, and the average reproducibility for these duplicates was also ±0.06, where 90 % of the duplicates fell within ±0.12 of the first analysis. Eight aqueous Fe(II) samples were re-analyzed 6 months after initial collection to test for stability of the solutions in terms of possible oxidation and precipitation of the anaerobically-collected samples, and the average reproducibility was ±0.10. Comparison between laboratories may be made through the IRMM-14 Fe standard, which has a δ 56 Fe value of -0.09±0.05 (Beard et al., 2003a). Note that some labs report Fe isotope variations in ε notation (parts per 10,000) relative to an assumed ε 56 Fe value of zero for the IRMM-14 standard (e.g., Zhu et al., 2001); the ε 56 Fe values reported by these labs may be converted to equivalent δ 56 Fe values as reported here by dividing by 10 and adding It is important to note that Fe isotope data reported in terms of 57 Fe/ 54 Fe ratios will appear to be 1.5 times larger than if reported in terms of 56 Fe/ 54 Fe ratios, although both approaches describe the same isotopic variability on a /mass basis. Experiment 1: Biogenic ferrous carbonate mineral production by Shewanella putrefaciens S. putrefaciens strain CN32 was cultivated aerobically in tryptic soy broth at 37 C. After 16 hr of aerobic growth, cells were collected by centrifugation (7000 x g, 10 min), washed in anaerobic bicarbonate buffer, and re-suspended to a density of ca cells ml -1. The washed cells were used to inoculate (final cell density ca ml -1 ) bicarbonate-buffered growth medium (30 mm NaHCO 3, 0.5 mm KH 2 PO 4, 10 mm NH 4 Cl, plus vitamins and trace elements) containing 10 mm sodium lactate as an electron donor and 10 mmol L -1 of freshly synthesized HFO. The HFO was produced by adjusting the ph of 0.4M FeCl 3 6H 2 O to 7.0 with 1M NaOH, after which the oxide precipitated was washed by centrifugation with distilled H 2 O until the Cl - concentration was < 1 mm (Lovley and Phillips, 1986). 1

33 The culture medium also contained 100 g L -1 of quartz sand (Sigma Chemicals), which promoted carbonate mineral precipitation and prevented the adherence of mineral precipitates to the bottom of the culture bottles. Experiment 1A did not contain Ca, but in Experiment 1B, 10 mm CaCl 2 2H 2 O was added in order to induce formation of Ca-bearing ferrous carbonate precipitates. Previous studies showed that ankerite and siderite are the dominant carbonate mineral end-products of HFO reduction by strain CN32 in the presence and absence of 10 mm Ca, respectively (Roden et al., 2002). No detectable amounts of CaCO 3 minerals precipitated in either the inoculated experiments or the abiotic controls, as determined by XRD spectra, despite the fact that under the conditions of these experiments, CaCO 3 was always oversaturated (Roden et al., 2002). After 485 d of incubation at 30 C, the solid-phase end-products of HFO reduction in Experiment 1 were collected by centrifugation, washed with sterile anaerobic Pipes buffer, and dried under a stream of O 2 -free N 2. The aqueous phase of the cultures was filtered through a 0.2 µm diameter syringe filter and acidified with HNO 3 (1% final concentration). Solid run products were characterized using XRD (bulk material), as well as SEM imaging. Iron and Ca contents of the total solid products were obtained using Atomic Absorption spectroscopy with a graphite furnace. Chemical compositions were determined for individual crystals using EDS spectra obtained during SEM imaging. Experiments 2 and 3: Biogenic magnetite production by Geobacter sulfurreducens and Shewanella algae S. algae and G. sulfurreducens were cultivated in anaerobic (80% N 2, 20% CO 2 headspace) bicarbonate-buffered growth medium (30 mm NaHCO 3, 4.4 mm KH 2 PO 4, 28 mm NH 4 Cl, plus vitamins and trace elements, as described in Lovley and Phillips, 1988), containing 20 mm sodium lactate (S. algae) or 20 mm sodium acetate (G. sulfurreducens) as the electron donor, and sodium fumarate (20 mm) as the electron acceptor. After 3-4 d of growth at 30 C, cells were harvested by centrifugation (7000 g, 10 min), washed in sterile, anaerobic bicarbonate buffer, and re-suspended to a density of ca cells ml -1. Thirteen replicate 50-mL bottles of Pipes-buffered (10 mm, ph 6.8) HFO (50 mmol L -1 ) medium were inoculated with ca cells ml -1 of washed cells that were initially grown in fumarate. Pipesbuffered rather than bicarbonate-buffered medium was used in this experiment in order to favor magnetite production over siderite precipitation during HFO reduction. The 10-fold lower cell density relative to Experiments 1, 4, 5, and 6 was designed to slow down the overall rate of HFO biotransformation in order to observe more clearly temporal patterns in Fe isotope fraction associated with phase conversions, as well as assess any kinetic effects on Fe isotope fractionation. Single culture bottles for each species were sacrificed (no pasteurization) at various times during a 291-d incubation period. The solid and aqueous phases were separated by centrifugation (7000 g, 10 min), and the pellet was stored frozen as a wet paste under N 2 prior to wet-chemical, isotopic, XRD, and SEM analysis as described above. The total solid was analyzed for each bottle, and partial dissolutions of the total solid in 0.5 M HCl was done for each bottle to monitor HFO conversion. In addition, partial dissolutions in weak HCl were done for three time samples for each species. Experiment 4: Biogenic magnetite and siderite production by Geobacter sulfurreducens G. sulfurreducens was grown as in Experiment 2 and washed cells were used to inoculate (final cell density ca ml -1 ) bicarbonate-buffered growth medium (30 mm NaHCO 3, 1 mm KH 2 PO 4, 10 mm NH 4 Cl, plus vitamins and trace elements) containing 20 mm sodium acetate and 50 mmol L -1 of freshly synthesized HFO (see above). Single 50 ml bottles were sacrificed after 0, 4, 11, 22, or 164 d of incubation. The bottles sacrificed at 0, 4, 11, and 22 d were pasteurized at 80 C in a water bath for 15 min; the bottle taken at 164 d was not pasteurized, nor were samples in the other experiments. SEM imaging of control experiments with and without pasteurization did not show significant differences in crystallinity of the solid products, suggesting that re-crystallization during pasteurization was minimal. 2

34 Visual inspection indicated extensive conversion of HFO to magnetite after ca. 4 d of incubation. Timezero controls and uninoculated cultures showed no HFO phase conversion after 164 d, as shown visually and in ferric and ferrous iron assays by Ferrozine. The magnetic precipitates were separated (inside an anaerobic chamber) from the aqueous phase by decanting the liquid into a beaker while holding the culture bottles up to a strong magnet. Virtually all of the mineral precipitates in the cultures were held in place by the magnet, and there was no visual evidence for the presence of residual HFO in the decanted liquid. However, ferric and ferrous iron contents determined on the magnetic concentrates indicated the presence of significant quantities of unreacted HFO up through 22 d (Table 6). The magnetic precipitates were washed twice with sterile, anaerobic Pipes buffer (10 mm, ph 6.7), here again using a magnet to separate the precipitates from the liquid phase. The washed precipitates were dried under a stream of O 2 - free N 2 and stored anaerobically. SEM and TEM imaging, as well as partial and total HCl extractions were used to asses the heterogeneity of the solid phase components and the extent of reaction of HFO to magnetite. These extractions included rinsing the solid in 0.1 M HCl, partial extraction in 0.5 M HCl, and total extraction in 3 M HCl, followed by determination of ferric/ferrous ratios in the extractable components using Ferrozine assays, following the approach of Fredrickson et al. (1998). The solid run products were further characterized for crystal structure by XRD on bulk products, as well as electron diffraction measurements on individual or groups of crystals using TEM. Chemical compositions for individual crystals were obtained using EDS spectra during SEM analysis. Experiments 5 and 6: Biogenic magnetite and siderite production by Geobacter sulfurreducens The culture medium (bicarbonate-buffered) for these experiments was identical to that employed in Experiment 4 with G. sulfurreducens, with the exception that the NTA/trace element solution was omitted from Experiment 6. Two sets of duplicate 100-mL bottles (0 and 1), with NTA (Experiment 5) and without NTA (Experiment 6), were inoculated with ca cells ml -1 of washed acetate/fumarategrown cells. In contrast to previous experiments in which whole bottles of inoculated culture medium were sacrificed over time, in this experiment the same culture bottles were subsampled periodically over time for wet-chemical, isotopic, XRD, and SEM/TEM analysis as described above. Separate 10-mL subsamples were collected for separation of magnetic vs. nonmagnetic solids, which was achieved by immersing a Teflon-coated magnetic stir bar retriever into the subsample, and then transferring the magnetic solids to a vial containing 5 ml of 6M HCl, in which the solids rapidly dissolved. The stir bar retriever was rinsed thoroughly between separations. The residual non-magnetic solids were dissolved in 0.5M HCl, and the quantity of Fe in magnetic vs. nonmagnetic separates was determined with Ferrozine. Sufficient solid and solution Fe was obtained from the original duplicate bottles (0 and 1) for Experiment 5 (NTA-bearing) through 27 d, and for Experiment 6 (NTA-absent) through the end of the entire run at 90 d; for Experiment 5, a third reserve bottle (2) was required for samples taken at 57, 90, and 113 d, and therefore a discontinuity in the reduction progress and Fe isotope compositions is possible. In addition to analysis of the total and magnetic solid, partial dissolution of solids for Experiment 5 was done using weak HCl and HAc. 3

35 Electronic Annex EA-2 SEM AND TEM IMAGES OF SOLID RUN PRODUCTS Experimental constraints on Fe isotope fractionation during magnetite and Fe carbonate formation coupled to dissimilatory hydrous ferric oxide reduction by Clark M. Johnson, Eric E. Roden, Susan A. Welch, and Brian L. Beard Geochimica et Cosmochimica Acta Fig. EA-2-1. Scanning electron microscope images of Fe carbonate produced in experiments with Shewanella putrefaciens strain CN32 (Experiment 1). Large and detailed views of siderite crystals formed in Ca-free experiment (1A). The rounded crystal form and evidence for surface dissolution features suggest that after the ca. 500 d incubation period, dissolution (and possible re-precipitation) of the crystals occurred. Energy-dispersive spectra show that all crystals are pure siderite, and this is confirmed by XRD spectra on bulk samples. 1

36 Fig. EA-2-2. Scanning electron microscope images of Fe carbonate produced in experiments with Shewanella putrefaciens strain CN32 (Experiment 1). Large and detailed views of siderite crystals formed in Ca-bearing experiment (1B). As in Experiment 1A, the rounded crystal form and evidence for surface dissolution features suggest that after the ca. 500 d incubation period, dissolution (and possible reprecipitation) of the crystals occurred. Energy-dispersive spectra show that all crystals are Ca-bearing siderites, where stoichiometries vary from Ca 0.12 Fe 0.88 CO 3 to Ca 0.15 Fe 0.85 CO 3 ; this is confirmed by XRD spectra and AA on bulk-samples. 2

37 Fig. EA-2-3. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 2. Large and detailed views of sample after 4 d of incubation. Sample appears to be entirely unreacted HFO, which is confirmed by XRD spectra, and ferric:ferrous ratios of the solids. 3

38 Fig. EA-2-4. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 2. Large and detailed views of sample after 8 d of incubation. SEM images show first appearance of tabular lepidocrocite (γfeo*oh), and XRD spectra on this sample confirms the presence of γfeo*oh. Ferric:ferrous ratios of the solids indicate that no significant magnetite exists in this sample. 4

39 Fig. EA-2-5. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 2. Large and detailed views of sample after 16 d of incubation. Sample appears to be dominated by tabular lepidocrocite (γfeo*oh), which is confirmed by XRD spectra. Ferric:ferrous ratios of the solids indicate virtually no conversion to magnetite, and XRD spectra do not show the presence of magnetite. 5

40 Fig. EA-2-6. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 2. Large and detailed views of sample after 21 d of incubation. SEM images show that most of the tabular lepidocrocite (γfeo*oh) has been lost, and XRD spectra on this sample does not show evidence for γfeo*oh. A few remnant tabular lepidocrocite crystals remain (detail). Ferric:ferrous ratios of the solids indicate that over half of the solid has been converted to magnetite, and this is confirmed by strong magnetite XRD spectra in this sample. Therefore, the majority of fine-grained material in the sample is interpreted to be fine-grained magnetite. 6

41 Fig. EA-2-7. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 2. Large and detailed views of sample after 99 d of incubation. Sample appears to be dominated by fine-grained magnetite, which is confirmed by XRD spectra. Ferric:ferrous ratios of the solids indicate that the sample is entirely magnetite. Euhedral magnetite observed in Experiment 4 is not seen; large forms in the images are likely to be clumped fine-grained magnetite. 7

42 Fig. EA-2-8. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Shewanella algae of Experiment 3. Large and detailed views of sample after 8 d of incubation. Sample appears to be entirely unreacted HFO, which is confirmed by XRD spectra, and ferric:ferrous ratios of the solids. 8

43 Fig. EA-2-9. Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Shewanella algae of Experiment 3. Large and detailed views of sample after 16 d of incubation. XRD spectra on this sample indicate the presence of magnetite. Ferric:ferrous ratios of the solids indicate some conversion of HFO to magnetite, although the proportion of magnetite is still small. 9

44 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Shewanella algae of Experiment 3. Large and detailed views of sample after 99 d of incubation. Sample consists of aggregates of fine-grained magnetite, as confirmed by XRD spectra and ferric:ferrous ratios of the solids. Note that strong magnetite peaks in XRD spectra are shown in earlier time samples. Ferric:ferrous ratios of this sample suggest that the solid is 100% magnetite. 10

45 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 4. Sample after 22 d of incubation. Large and detailed views of magnetite crystals illustrating large (ca. 1 µm) euhedral magnetite crystals, in addition to fine-grained material. Energy-dispersive spectra obtained on the large crystals are consistent with magnetite, and this interpretation is supported by XRD data on bulk samples. 11

46 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 4. Large and detailed views of sample after 164 d of incubation. The large magnetite crystals observed are similar in size as those observed earlier in the experiment. The proportion of fine-grained material has markedly decreased in the long-term experiment. 12

47 Fig. EA Transmission electron microscope images of fine-grained solid products observed in SEM images of Experiment 4. Large and detailed views of sample after 4 d of incubation. The morphology of the vast majority of grains in the fine-grained material indicates that it is dominated by magnetite, and electron diffraction patterns support this conclusion. Detail of ca. 20 nm-sized magnetite crystal illustrating lattice fringes shown in inset in lower left. However, longbladed crystals match the morphology for ferric oxyhydroxide minerals such as goethite (see detailed inset in upper left), which is interpreted to reflect partial re-crystallization of HFO early in the experiment. The presence of un-reduced ferric oxyhydroxide in the early parts of the experiment is supported by the ferric:ferrous ratios determined for the bulk solid material from this sample. 13

48 Fig. EA Transmission electron microscope images of fine-grained solid products observed in SEM images of Experiment 4. Large and detailed views of sample after 164 d of incubation. All of the fine-grained material appears to have the morphology of magnetite, and this interpretation is consistent with electron diffraction patterns. None of the long-bladed ferric oxyhydroxide crystals observed in the 4 d sample were found in this sample. 14

49 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 5. Large and detailed views of sample after 2 d of incubation. Early siderite formation consists of globules, which we have observed in rapid precipitation experiments in abiologic systems (Wiesli et al., 2004). 15

50 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 5. Large and detailed views of sample after 7 d of incubation. SEM images show that the carbonate globules that are common in the 2 d sample have extensively re-crystallized to rhombohedral siderite. 16

51 Fig. EA Scanning electron microscope images of solid products produced by dissimilatory reduction of HFO by Geobacter sulfurreducens of Experiment 5. Large and detailed views of sample after 27 d of incubation. Siderite consists exclusively as rhombohedral shapes, with small crystals scattered on the surfaces. Magnetite appears to be largely fine-grained. 17