ARTICLES
PUBLISHED ONLINE: 16 JANUARY 2017 | DOI: 10.1038/NGEO2878
A key role for green rust in the Precambrian
oceans and the genesis of iron formations
I. Halevy1*†, M. Alesker1†, E. M. Schuster1, R. Popovitz-Biro2 and Y. Feldman2
Iron formations deposited in marine settings during the Precambrian represent large sinks of iron and silica, and have
been used to reconstruct environmental conditions at the time of their formation. However, the observed mineralogy in
iron formations, which consists of iron oxides, silicates, carbonates and sulfides, is generally thought to have arisen from
diagenesis of one or more mineral precursors. Ferric iron hydroxides and ferrous carbonates and silicates have been identified
as prime candidates. Here we investigate the potential role of green rust, a ferrous–ferric hydroxy salt, in the genesis of iron
formations. Our laboratory experiments show that green rust readily forms in early seawater-analogue solutions, as predicted
by thermodynamic calculations, and that it ages into minerals observed in iron formations. Dynamic models of the iron cycle
further indicate that green rust would have precipitated near the iron redoxcline, and it is expected that when the green rust
sank it transformed into stable phases within the water column and sediments. We suggest, therefore, that the precipitation
and transformation of green rust was a key process in the iron cycle, and that the interaction of green rust with various
elements should be included in any consideration of Precambrian biogeochemical cycles.
A
lthough the chemical, mineralogical and isotopic
composition of iron formations (IFs) has been widely
used to constrain the chemistry and oxidation state of the
Precambrian oceans and atmosphere1–3 , many aspects of their
genesis remain poorly understood. It is agreed that even in the
best preserved IFs, observed mineral assemblages reflect diagenetic
alteration of one or more mineral precursors1–5 . Attention has
focused on ferric iron (FeIII ) oxyhydroxides, thought to have
formed by the oxidation of ferrous iron (FeII ) near the ocean
surface, in ‘oases’ of O2 -rich seawater generated by cyanobacteria3 ,
by anoxygenic FeII photosynthesis6 , or by abiotic light-driven FeII
oxidation7 . In contrast, the texture of IFs from various ages supports
primary precipitation of FeII -bearing, nanoparticulate silicates or
carbonates from anoxic waters, and much later oxidation to ferric
oxides8,9 . The identity of the precursors remains uncertain, as do
the pathways for their transformation into the minerals observed
in IFs, and the dependence of the final assemblage on the presence
of organic matter.
A mineral neglected in the study of IFs is the ferrous–
ferric hydroxy salt, green rust (GR). A naturally occurring GR,
fougèrite ([FeII1−x FeIII
Mgy (OH)2+2y ]x + [x/nAn− × mH2 O]x− , with A
x
−
representing OH , Cl− , SO4 2− , CO3 2− ), forms today in waterlogged iron-rich soils, and as a product of steel corrosion
in O2 -poor environments10,11 . Following exposure to air GR
rapidly oxidizes to FeIII oxyhydroxides12,13 . Recently, GR has been
found to form near the iron redoxcline and persist through
the water column and sediments in ferruginous Lake Matano
(Indonesia), an Archaean ocean analogue14 . Additionally, GR has
been inadvertently formed in experiments conducted in Archaeananalogue seawater15,16 . Despite these indications that GR forms
under conditions relevant to Earth’s early oceans, its potential role in
IF genesis is unexplored. Combining thermodynamic calculations,
laboratory experiments and dynamic models of the early iron
cycle, we find that precipitation of GR was probably a major early
sink of seawater iron. Furthermore, following exposure to different
chemical environments, GR transforms into several of the major
minerals observed in IFs, suggesting a role for GR as a precursor
to IF mineral assemblages.
Thermodynamics and experiments suggest a GR precursor
Minerals observed in IFs are thermodynamically stable (Fig. 1a
and Methods), although barriers to nucleation and growth prevent
the direct precipitation of some (for example, haematite), or
require high degrees of supersaturation of others (for example,
siderite). Instead, metastable phases (for example, the hydrous
FeIII oxyhydroxide, ferrihydrite) often precipitate from solution and
transform into the stable minerals. Unless reductants (for example,
organic carbon) are available for reduction of the precursors, iron
may persist in FeIII minerals despite anoxic conditions in the
sediments4 . In O2 -poor seawater-like solutions, thermodynamics
predicts that GR should be a metastable precursor precipitate
(Fig. 1b), and that its transformation to the stable minerals should
depend on solution pH and oxidation state, and on the aqueous
concentrations of dissolved inorganic carbon (DIC), total sulfur
and total silica (Supplementary Fig. 1). Unlike FeIII oxyhydroxide
precursors, approximately two-thirds of the iron in GR is FeII ,
so that organic carbon is not strictly required for its diagenetic
transformation into mixed-valence mineral assemblages. This may
explain the occurrence of FeII /FeIII assemblages in some organic
carbon-poor IFs17 .
To preclude kinetic inhibition of GR precipitation and
constrain its transformation products, we performed precipitation
experiments from Archaean seawater-analogue solutions at
room temperature (∼22 ◦ C), with geologically plausible pH
and concentrations of FeII , DIC, sulfate and silica (Methods).
Initial precipitates were analysed by X-ray diffractometry (XRD),
scanning electron microscopy (SEM) and transmission electron
microscopy (TEM), and reanalysed after ageing periods of
several days to four months (at ∼22 and 50 ◦ C). In all solutions
with pH>∼ 7, irrespective of DIC, sulfate and silica content,
1
Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot 76100, Israel. 2 Department of Chemical Research Support,
Weizmann Institute of Science, Rehovot 76100, Israel. † These authors contributed equally to this work. *e-mail: [email protected]
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
1
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
ARTICLES
a
b
FeOH2+
1.0
1.0
Fe3+
Fe3+
Schwertmannite
Fe(OH)4−
Fe2+
Eh (V)
0.5
Eh (V)
0.5
Haematite
0.0
Fe2+
Ferrihydrite
0.0
FeHCO3+
∑Fe = 10−5 M
∑S = 10−5 M
∑SiO2 = 10−2.7 M
pCO2 = 10−2 atm
T = 25 °C
−0.5
0
2
FeHCO3+
−0.5
Siderite
FeCO30
Greenalite
4
6
8
10
12
14
0
2
4
pH
6
8
GRCO32−
Fe(OH)3−
10
12
14
pH
Figure 1 | Thermodynamic stability of iron-bearing minerals in ‘early seawater’. a, Thermodynamically stable, but kinetically inhibited phases.
b, Metastable, apparently uninhibited phases, which may act as precursors to formation of the stable phases. The grey region represents the range of Eh
and pH encountered in modern marine environments. The Eh of the early, anoxic ocean was lower.
CO3 2− -bearing GR precipitated initially. Ferrihydrite formed at
lower pH. Depending on the solution composition, the GR aged to
iron oxyhydroxides, siderite or iron-bearing silicates (Fig. 2). The
silicates, which were too scarce or poorly crystalline to detect by
XRD, were identified by TEM electron diffraction (ED) patterns
consistent with the structures of modulated 2:1 sheet silicates (for
example, stilpnomelane, minnesotaite). Energy-dispersive X-ray
spectroscopy coupled to the ED measurements indicated that the
iron was associated with the sheet silicates. Sample results are in the
Supplementary Information.
A dynamic model predicts a major GR sink for marine iron
The experiments demonstrate that GR precipitates from Archaean
and Proterozoic ocean-analogue solutions, and that it ages to
minerals observed in IFs in the absence of organic matter. However,
the importance of GR as a precursor to IF mineral assemblages
depends on the dynamics of the early iron cycle. For example, if the
rate of FeII oxidation is rapid enough that FeIII generation outstrips
its precipitation in GR, then FeIII concentrations increase until the
next least soluble phase (ferrihydrite) saturates and precipitates.
Below the iron redoxcline, FeII and DIC or silica concentrations
may be high enough that precipitation of FeII -bearing carbonates
or silicates is a dominant iron sink8,9,18 . Thus, the proportion of
GR in the flux to the sediments depends on the relative rates
of FeII supply, FeII oxidation (FeIII production), and precipitation
and sinking of GR, FeIII oxyhydroxides, and FeII minerals, as
affected by the concentrations of FeII , FeIII , DIC and silica, and by
ocean pH.
To quantify the relative contribution of GR to iron sedimentation
fluxes (fGR ), we developed a dynamic model of the early iron
cycle (Methods). Model inputs are pCO2 , pO2 , the hydrothermal and
riverine influxes, ocean mixing rates, and various thermodynamic
and kinetic constants. The model tracks the aqueous concentrations
of FeII and FeIII in the surface, intermediate and deep ocean, the
saturation state of the various precipitates in these environments,
and their proportion in the sedimentary iron flux. On the basis of the
experimental results shown here for GR and in other studies for FeIII
oxyhydroxides, we assume that the primary precipitates transform
into the minerals observed in IFs, but the transformation is not
explicitly modelled.
2
As the values of the required model parameters are highly
uncertain, we first explore the sensitivity to these values, and then
provide a best-guess temporal evolution of the parameter values
to assess the importance of GR over Earth history. Improving
constraints on early CO2 and O2 levels and ocean chemistry will
reduce uncertainty in such model histories. Due to the strong
dependence of GR saturation on pH (Fig. 3a), the results are highly
sensitive to pCO2 , through its effect on ocean pH (Fig. 3b). Increasing
pO2 leads to decreasing fGR due to more rapid oxidation of FeII
(Fig. 3c). Higher hydrothermal iron influxes result in a higher FeII
concentration in the deep ocean, and in lower fGR (Fig. 3d), if
siderite precipitates at saturation (see discussion below). Within a
reasonable range of values, fGR is insensitive to rates of FeII photooxidation (Supplementary Fig. 3).
We explored the effects of assumptions about the sink of
iron to primary FeII carbonates (siderite) or silicates (greenalite).
Siderite precipitation in modern anoxic environments typically
requires up to ∼50 times supersaturation, possibly due to surfacecontrolled kinetics19 . A requirement for 10 times siderite saturation
(sid = 10) results in deep-ocean FeII concentrations controlled
by upwelling of FeII -bearing water, near-surface oxidation of the
upwelled FeII , and precipitation of GR and/or FeIII oxyhydroxides
as the only iron sinks (Fig. 3 solid). When no supersaturation
is required (sid = 1, for example, if precipitation surface preexists) FeII concentrations are co-controlled by deep-ocean siderite
precipitation and oxidation of upwelled FeII . Accordingly, the iron
sink comprises siderite, GR and/or FeIII oxyhydroxides (Fig. 3
dotted). Irrespective of the required siderite supersaturation and
the resulting FeII concentrations, and using an upper limit on
silica activity of 10−2.7 and an experimentally determined apparent
greenalite solubility18 , greenalite remained undersaturated.
With
estimated
time-dependent
parameter
values
(Supplementary Information), GR constitutes a large fraction
of the iron flux to the sediment during the Archaean (Fig. 3e),
irrespective of uncertainty in GR thermochemical properties
(Supplementary Fig. 4). Before the rise of O2 , a hydrothermal
source of FeII ∼ 2–3 times larger than present is balanced by
oxidation of FeII in the photic zone and formation of GR particles,
as well as by deep-ocean siderite precipitation, if sid = 1. Resulting
FeII concentrations are ∼1–10 µM and <1 nM in the deep and
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
ARTICLES
a
b
200 nm
(030)
(300)
(110)
(120)
(210)
(330)
(120)
(110) (300)
(030)
NaCl
Green rust
5 nm−1
GR
10
20
f
NaCl
GR NaCl
GR
30
GR
GR
NaCl
NaCl GR
GR
40
CuKα 2θ (°)
Green rust
50
NaCl
60
Minnesotaite
70
(006)
(131) (003)
(131)
(110)
S
c
(131) (003)
(131)
(006)
2 nm−1
Siderite + magnetite
500 nm
M
S
M
S
S
g
M
S
S
S
20
30
d
40
CuKα 2θ (°)
M
S M
M
50
Stilpnomelane
60
e
2 nm−1
50 nm
200 nm
Siderite
2 nm−1
Magnetite
(220)
(440)
(660)
ED11
(620)
(642)
200 nm
Figure 2 | Green rust and its ageing products. a,b, TEM image and XRD pattern (a), and indexed ED pattern of green rust (b). c–e, XRD pattern of
siderite (S) and magnetite (M) (c), SEM image of siderite (d), and TEM image and ED pattern of magnetite (e), which are some of the ageing products in
the absence of silica. f,g, TEM image and ED patterns of Fe phyllosilicates, which are some of the ageing products in the presence of silica at room
temperature (f) and at 50 ◦ C (g).
surface ocean, respectively. FeIII concentrations are everywhere
<0.01 nM, too low for saturation and precipitation of FeIII
oxyhydroxides. Realistically, heterogeneity is expected to have
existed in the relative importance of GR and FeIII oxyhydroxides,
for example, due to spatio-temporal variability in phototrophic
activity and in deep water upwelling rates. Similarly, variability
in seawater pH and DIC concentrations (for example, due to
generation of alkalinity and DIC during anaerobic remineralization
of organic matter), or in silica concentrations (for example, due to
variable distance from hydrothermal silica sources), would lead to
heterogeneity in the relative importance of primary FeII carbonate
and silicate precipitation relative to GR or FeIII oxyhydroxide
precipitation.
The rise of atmospheric O2 results in dominance of FeIII
oxyhydroxides in the sedimentary iron flux (Fig. 3e), due partly to
more rapid oxidation of FeII , and partly to an assumed decrease in
the concentrations of reduced greenhouse gases (for example, CH4 ,
C2 H6 ), which causes an increase in pCO2 via the silicate weathering
climate feedback20 , and a concomitant decrease in pH21 . As solar
luminosity increases into the Proterozoic, pCO2 decreases and pH
increases, but rapid FeII oxidation prevents GR from contributing
to the near-surface iron sink. However, FeIII oxyhydroxides formed
in the photic zone may be transformed into GR, either biotically22 or
abiotically23 , when they sink below the iron redoxcline into FeII -rich
waters. The sedimentary flux of iron-bearing particles may have
thus contained an appreciable fraction of GR even if photic zone
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
3
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
ARTICLES
a
c
0.8
0.6
Ωsid: 10
1
0.4
1.0
Fe sink fraction
Fe sink fraction
1.0
Siderite
Ferrihydrite
Green rust
0.2
b
6.6
6.8
7.0
pH
7.2
7.4
7.6
7.8
0.8
Fe sink fraction
Fe sink fraction
0.2
10−5
d
1.0
0.6
0.4
0.2
10−4
pO2 (atm)
10−3
10−2
1.0
0.8
0.6
0.4
0.2
0.0
0.0
10−2
0.6
0.4
0.0
0.0
6.4
0.8
10−1
pCO2 (atm)
1
e
2
3
fhyd (× modern)
4
5
Rise of O2
1.0
Fe sink fraction
0.8
0.6
Proterozoic
Archaean
0.4
0.2
0.0
1.5
2.0
2.5
Age (Ga)
3.0
3.5
Figure 3 | Sensitivity of iron sink to model parameters. a–e, Fraction of precipitates out of the iron sink for variable seawater pH (a), atmospheric pCO2 (b),
atmospheric pO2 (c), the hydrothermal circulation enhancement factor (d) and time (e). In a,c,d, parameters other than the varied parameter were at
default values (Supplementary Information). In b, pH depended on pCO2 , and other parameters were at default values. In e, time-dependent estimates of all
parameters were used (Supplementary Information). Results depend on siderite supersaturation required for precipitation (sid = 10 or 1 in solid or dotted,
respectively). Sensitivity to the parameters is explained in the Supplementary Information.
dynamics favoured formation of FeIII oxyhydroxides. We expect that
GR continued to shuttle iron to the sediments until the second,
Neoproterozoic, rise of atmospheric O2 , when oxygenation of much
of the ocean irreversibly decreased seawater FeII concentrations.
Implications and tests of a GR iron shuttle
A quantitatively important GR shuttle from the surface ocean to
the sediments affects the cycles of several other elements. Silica
sorbs strongly to GR and is known to retard GR oxidation24 .
Adsorption of silica onto sinking GR particles may have coupled
the transport of iron and silica to sediments, leading to enrichment
of both elements in IFs. Phosphate sorbs strongly to GR, but
unlike FeIII oxyhydroxides, which release the phosphate following
reductive dissolution, the phosphate sorbed to GR may get
sequestered in the FeII -phosphate vivianite25 . Did the waning
importance of GR at the Archaean–Proterozoic transition lead
to more efficient phosphate recycling, to a positive feedback on
productivity, organic matter burial and O2 release, and to the
protracted Lomagundi positive C isotope excursion26 ? Nickel sorbs
onto GR more effectively than onto FeIII oxyhydroxides14 , with
proposed implications for the evolution of methanogens, which
require Ni. Reductive immobilization of U and Cr by GR and
isomorphic substitution of divalent transition metals (Zn, Cd, Ni,
Co) for FeII in the GR structure are known23,27–29 . Some of these
4
metals are bioessential nutrients, whereas others have been used to
constrain the oxidation state of the early oceans and atmosphere.
The mobility and chemistry of these elements, and perhaps others,
probably changed radically as the importance of the GR iron
shuttle varied with time. Experimental and theoretical work is
required to fully explore the implications for metal cycles, biological
productivity and evolution, and for the use of transition metals as
proxies for early ocean chemistry.
How might a GR-derived mineral assemblage manifest in the
geologic record? The relatively short timescale of GR transformation
makes it difficult to texturally distinguish primary precipitation of
FeII carbonates or silicates8,9 from their early formation from a GR
precursor in the water column or at the sediment/water interface.
Furthermore, transformation by dissolution and reprecipitation
obliterates many of the microtextures associated with GR
precipitation. The low density of GR relative to other iron-bearing
precipitates (Supplementary Table 5), however, suggests density
gradient-driven sedimentary textures as possible indicators of a
GR precursor. The occurrence of centimetre-scale flame structures
from magnetite bands into overlying chert bands in a Mesoarchaean
IF in the Red Lake Greenstone Belt (northwest Ontario) may be an
example30 . Finally, the different affinity and reactivity of GR towards
a variety of elements, as well as the occurrence of both FeII and FeIII
in GR, may yield distinctive chemical and isotopic fingerprints,
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
although it remains to be seen whether these fingerprints survive
early diagenetic transformation of GR into more stable minerals.
With a combination of experimental and modelling work,
we show that GR was probably a major sink of seawater iron
through much of the Precambrian. Depending on the GR formation
environment and on early diagenetic conditions, a final assemblage
containing combinations of iron oxides, carbonates and silicates is
possible. These findings shed new light on IF genesis, and call for a
revision of our understanding of the early iron cycle and the cycles
of several trace and major elements.
Methods
Methods, including statements of data availability and any
associated accession codes and references, are available in the
online version of this paper.
Received 13 October 2016; accepted 12 December 2016;
published online 16 January 2017
References
1. Bekker, A. et al. Iron formation: the sedimentary product of a complex
interplay among mantle, tectonic, and biospheric processes. Econ. Geol. 105,
467–508 (2010).
2. Klein, C. Some Precambrian banded-iron formations (BIFs) from around the
world: their age, geologic setting, mineralogy, metamorphism, geochemistry,
and origin. Am. Mineral. 90, 1473–1499 (2005).
3. Cloud, P. Paleoecological significance of the banded iron formation. Econ. Geol.
68, 1135–1143 (1973).
4. Walker, J. C. G. Suboxic diagenesis in banded iron formations. Nature 309,
340–342 (1984).
5. Fischer, W. W. & Knoll, A. H. An iron shuttle for deepwater silica in Late
Archean and early Proterozoic iron formations. Geol. Soc. Am. Bull. 121,
222–235 (2008).
6. Kappler, A., Pasquero, C., Konhauser, K. O. & Newmanm, D. K. Deposition of
banded iron formations by anoxygenic phototrophic Fe(II)-oxidizing bacteria.
Geology 33, 865–868 (2005).
7. Cairns-Smith, A. G. Precambrian solution photochemistry, inverse segregation,
and banded iron formations. Nature 276, 807–808 (1978).
8. Rasmussen, B., Krapež, B. & Meier, D. B. Replacement origin for hematite in
2.5 Ga banded iron formation: evidence for postdepositional oxidation of
iron-bearing minerals. Geol. Soc. Am. Bull. 126, 438–446 (2014).
9. Rasmussen, B., Krapež, B. & Muhling, J. R. Hematite replacement of
iron-bearing precursor sediments in the 3.46-b. y.-old Marble Bar Chert,
Pilbara craton, Australia. Geol. Soc. Am. Bull. 126, 1245–1258 (2014).
10. Trolard, F., Bourrié, G., Abdelmoula, M., Refait, P. & Feder, F. Fougerite, a new
mineral of the pyroaurite-iowaite group: description and crystal structure.
Clays Clay Miner. 3, 323–334 (2007).
11. Refait, P., Memet, J.-B., Bon, C., Sabot, R. & Génin, J.-M. R. Formation of the
Fe(II)-Fe(III) hydroxysulfate green rust during marine corrosion of steel.
Corros. Sci. 45, 833–845 (2003).
12. Legrand, L., Mazerolles, L. & Chaussé, A. Oxidation of carbonate green rust
into ferric phases: solid-state reaction or transformation via solution. Geochim.
Cosmochim. Acta 68, 3497–3507 (2004).
13. Génin, J.-M. R., Ruby, C., Géhin, A. & Refait, P. Synthesis of green rusts by
oxidation of Fe(OH)2 , their products of oxidation and reduction of ferric
oxyhydroxides: Eh -pH Pourbaix diagrams. C. R. Geosci. 338, 433–446 (2006).
14. Zegeye, A. et al. Green rust formation controls nutrient availability in a
ferruginous water column. Geology 40, 599–602 (2012).
15. Wiesli, R. A., Beard, B. L. & Johnson, C. M. Experimental determination and Fe
isotope fractionation between Fe(II), siderite and ‘‘green rust’’ in abiotic
systems. Chem. Geol. 211, 343–362 (2004).
ARTICLES
16. Halevy, I. & Schrag, D. P. Sulfur dioxide inhibits calcium carbonate
precipitation: implications for Mars and Earth. Geophys. Res. Lett. 36,
L23201 (2009).
17. Beukes, N. J., Klein, C., Kaufman, A. J. & Hayes, J. M. 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. Bull. Soc. Econ. Geol. 85, 663–690 (1990).
18. Tosca, N. J., Guggenheim, S. & Pufahl, P. K. An authigenic origin for
Precambrian greenalite: implications for iron formation and the chemistry of
ancient seawater. Geol. Soc. Am. Bull. 128, 511–530 (2015).
19. Bruno, J., Wersin, P. & Stumm, W. On the influence of carbonate in mineral
dissolution: II. The solubility of FeCO3 (s) at 25 ◦ C and 1 atm total pressure.
Geochim. Cosmochim. Acta 56, 1149–1155 (1992).
20. Walker, J. C. G., Hays, P. B. & Kasting, J. F. A negative feedback mechanism for
the long-term stabilization of Earth’s surface temperature. J. Geophys. Res. 86,
9776–9872 (1981).
21. Grotzinger, J. P. & Kasting, J. F. New constraints on Precambrian ocean
composition. J. Geol. 101, 235–243 (1993).
22. Ona-Nguema, G. et al. Iron(II,III) hydroxycarbonate green rust formation and
stabilization from lepidocrocite bioreduction. Environ. Sci. Tech. 36,
16–20 (2002).
23. Tamaura, Y. Ferrite formation from the intermediate, green rust II, in the
transformation of γ -FeO(OH) in aqueous suspension. Inorg. Chem. 24,
4363–4366 (1985).
24. Kwon, S.-K. et al. Influence of silicate ions on the formation of goethite from
green rust in aqueous solution. Corros. Sci. 49, 2946–2961 (2007).
25. Hansen, C. R. H. & Poulsen, I. F. Interaction of synthetic sulphate ‘‘green rust’’
with phosphate and the crystallization of vivianite. Clays Clay Miner. 47,
312–318 (1999).
26. Bachan, A. & Kump, L. R. The rise of oxygen and siderite oxidation during the
Lomagundi Event. Proc. Natl. Acad. Sci. USA 112, 6562–6567 (2015).
27. O’Loughlin, E. J., Kelly, S. D., Cook, R. E., Csencsits, R. & Kemner, K. M.
Reduction of uranium(VI) by mixed iron(II)/Iron(III) hydroxide (green rust):
formation of UO2 nanoparticles. Environ. Sci. Technol. 37, 721–727 (2003).
28. Loyaux-Lawniczak, S., Refait, P., Ehrhardt, J.-J., Lecomte, P. & Génin, J.-M. R.
Trapping of Cr by the formation of ferrihydrite during the reduction of
chromate ions by Fe(II)—Fe(III) hydroxysalt green rusts. Environ. Sci. Technol.
34, 438–443 (2000).
29. Refait, P., Drissi, H., Marie, Y. & Génin, J.-M. R. The substitution of Fe2+ ions
by Ni2+ ions in green rust one compounds. Hyperfine Interact. 90,
389–394 (1994).
30. Rowe, C. D. & Wing, B. A. Physical sedimentology constrains the primary
precipitates in banded iron formations. Goldschmidt abstract. 2701 (2015).
Acknowledgements
We thank N. Tosca for valuable comments. I.H. acknowledges funding from a European
Research Council Starting Grant 337183, and an Israeli Science Foundation Grant
764/12. I.H. is the incumbent of the Anna and Maurice Boukstein Career Development
Chair at the Weizmann Institute of Science. Electron microscopy was carried out at the
Moskowitz Center for Nano and Bio-Nano Imaging.
Author contributions
I.H. designed and supervised the research, performed the thermodynamic calculations,
developed and analysed the dynamic model, and wrote the paper. M.A., I.H. and E.M.S.
performed the experiments and analysed the products. R.P.-B. and Y.F. assisted
with analyses.
Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to I.H.
Competing financial interests
The authors declare no competing financial interests.
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
5
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
ARTICLES
Methods
Thermodynamics. Thermodynamic calculations were carried out using the
Geochemist’s Workbench31 . As a basis, we used the thermochemical data in the
LLNL database32 , which contains a variety of relevant iron-bearing silicates (for
example, greenalite, stilpnomelane), carbonates (for example, siderite) and sulfides
(for example, pyrite). We used a version of the database to which various iron
sulfate and iron chloride salts were added33 . We added to the database several
ferrous–ferric hydroxides, including four varieties of GR (carbonate, sulfate,
chloride, hydroxide). We initially took compiled thermochemical properties for the
hydroxides (including the GR)34 , and in preliminary calculations found that only
carbonate and sulfate GR came close to saturation under the conditions of interest.
Then, to explore the effects of uncertainty in the thermochemical properties of
carbonate and sulfate GR, which are less well established than most iron-bearing
minerals, we compiled their values from literature sources and tested the sensitivity
of the thermodynamic calculations to the values of the equilibrium constants used.
In calculations showing the thermodynamically stable phases, we excluded
magnetite and pyrite, which are generally accepted to be of diagenetic origin in
IFs1 . We also excluded various high-temperature/pressure minerals (for example,
ferrosilite, fayalite). Inclusion of magnetite and pyrite in calculations of the
thermodynamically stable phases results in stability fields of these minerals below
the haematite field, but does not affect calculations of the metastable precursors,
from which magnetite and pyrite were excluded. In calculations showing the
possible metastable precursors, we excluded thermodynamically stable minerals
(magnetite, pyrite, haematite, siderite, greenalite) and several metastable minerals
thought to be of diagenetic or metamorphic origin (for example, minnesotaite).
All calculations were carried out at a temperature of 25 ◦ C, a pressure of 1 atm
(1.013 bar) and a total Fe concentration of 10−5 M. Phase diagrams in the main text
(Fig. 1) were calculated with parameter values aimed at simulating an Archaean
ocean. CO2 fugacity was 10−2 atm (ref. 35), total S concentration was 10−5 M
(ref. 36), and total SiO2 concentration was 10−2.7 M (ref. 37). We tested the effect of
uncertainty in the values of these parameters on the expected stable and precursor
mineral assemblage (Supplementary Fig. 1). For the stable mineral assemblage, an
increase in pCO2 and in total SiO2 concentrations leads to growth of the stability
fields of siderite and greenalite, respectively, at the expense of haematite
(Supplementary Fig. 1a,e). An increase in total S has no effect (Supplementary
Fig. 1c), as pyrite precipitation was disallowed (see above).
As mentioned above, we tested the sensitivity also to uncertainty in the
thermochemical properties of carbonate and sulfate GR. Drissi et al.38 used Eh/pH
measurements during Fe(OH)2 oxidation to carbonate GR to calculate the
chemical potential of carbonate GR, given the chemical potentials of CO3 2− , H2 O
and Fe(OH)2 . They reported a mean value of −852,870 cal mol−1 for anhydrous
carbonate GR, but no estimate of the error. Using 2σ uncertainty bounds on
their Eh/pH measurements, and the chemical potentials reported in their
study, we calculated the equilibrium constant for the carbonate GR
dissolution reaction,
(OH)12 (CO3 ) + 13H+ = 4Fe2+ + 2Fe3+ + HCO3 − + 12H2 O
FeII4 FeIII
2
The equilibrium constant is logK = 39.1 ± 0.4 kJ mol−1 . The uncertainty of
0.4 kJ mol−1 on the Gibbs free energy of reaction comes from the 2σ uncertainty
bounds on the Eh/pH measurements. This leads to negligible uncertainty in the
stability field of carbonate GR (Supplementary Fig. 2a), and has no discernible
bearing on the outcome of the thermodynamic calculations and dynamic model in
this study. Much larger uncertainty comes not from the Eh/pH measurements, but
from the choice of chemical potentials of the species participating in the Fe(OH)2
oxidation reaction and in GR formation. Using, instead of the potentials in
Drissi et al.38 , the values recommended in the CRC Handbook of Chemistry and
Physics39 yields logK = 33.2 ± 0.4 kJ mol−1 , significantly expanding the stability
field of carbonate GR at the expense of ferrihydrite and aqueous FeII species
(Supplementary Fig. 2c). Using the more updated chemical potentials for Fe2+ and
Fe3+ from Parker and Khodakovskii40 and the rest of the chemical potentials from
Drissi et al.38 yields logK = 45.5 ± 0.4 kJ mol−1 , significantly diminishing the
stability field of carbonate GR (Supplementary Fig. 2b). However, we can rule out
this high value of logK on the basis of our precipitation experiments, which yielded
GR at pH ∼8. This would not be possible if the equilibrium constant were as large
as 1045.5 .
We similarly explored the sensitivity to the thermochemical properties of
sulfate GR. Refait and Génin41 reported a chemical potential of anhydrous sulfate
GR of −904,100 ± 350 cal mol−1 , and Olowe et al.42 reported a value of
−902,890 cal mol−1 . Using the chemical potentials in these studies, logK = 29.4 to
30.1. Using the chemical potentials in the CRC Handbook of Chemistry and
Physics, logK = 23.5 to 24.1. Using the chemical potentials for Fe2+ and Fe3+ from
Parker and Khodakovskii40 and the rest of the chemical potentials from Refait and
Génin41 , logK = 35.8 to 36.4. Given the greater stability of carbonate GR, this large
uncertainty does not affect the results when consistent equilibrium constants are
used (that is, when the equilibrium constant for both for sulfate and carbonate GR
is calculated using the same chemical potentials). In all of these cases, carbonate
rather than sulfate GR forms.
Experimental synthesis and diagenesis of GR. GR synthesis and ageing was
performed in seawater-analogue solutions (see below) with variable
concentrations of dissolved inorganic carbon (DIC, ∼25–75 mM) and total SiO2
(0, 100–250 µM). All syntheses were conducted in 1-l glass bottles, at room
temperature (∼22 ◦ C). Bottles were sealed with caps with PTFE-coated silicon
liners, except during purging or diffusive exchange with the atmosphere, when caps
with a gas inlet and outlet were used. The inlet is fused to a ceramic sinter, which
was lowered into the solution for purging, or raised above the solution during
diffusive exchange with the atmosphere. All solutions were magnetically stirred
during preparation.
Doubly distilled water (18.2 m) was purged with N2 gas (99.99%) for 2 h to
remove dissolved O2 . The bottles were then sealed and transferred to an anaerobic
glovebox (O2 < 1 ppm). Inside the glovebox, salts were added to simulate
Precambrian seawater (Supplementary Table 1). Stable, O2 -inert salts were
introduced first (NaCl, MgCl2 × 6H2 O, MgSO4 × 7H2 O), and the solution was
purged with argon gas (99.996%) for 1 h to remove residual O2 brought into the
glovebox with the distilled water or with the salts. FeCl2 × 4H2 O, pre-purified to
remove oxidation products43 , was then added and the solution purged with argon
for an additional 0.5 h. The ceramic sinter was then raised above the liquid surface,
and NaHCO3 (99.7%, ACS) and Na2 SiO3 × 5H2 O were added to simulate waters
with variable concentrations of DIC (24.2, 48.3 or 72.5 mM), and total SiO2 (0, 100,
150, 200 or 250 µM), respectively. Solution pH was adjusted to 7–8 by addition of
NaOH (99.99%, ACS). We note that when the resulting solutions were left to stand,
no precipitate formed over the course of several hours. Argon was flowed
through the headspace for 5 min and the bottles were then taken out of
the glovebox.
Partial oxidation of the FeII to FeIII was achieved by diffusive replacement of the
argon headspace by air. The bottles were exposed to air for ∼1.5 h, without
agitation or mixing, at which time the solution obtained a pale greenish-blue
colour, but remained clear. The bottles were then returned to the glovebox and
sealed. Precipitation of GR was complete within a few hours, but the first analysis
was performed the following day to allow crystallinity to develop. Identical bottles
were left to age in the glovebox, wrapped in aluminium foil to prevent
photochemical reactions in the solution and solid. All synthesis products were aged
at room temperature (∼22 ◦ C), whereas additional products from repeats of the
SiO2 -bearing experiments were also aged at 50 ◦ C. The pH of the ageing solutions
was not buffered or measured, but the persistence of GR (Supplementary Table 2)
indicates that pH did not drop below ∼7.
Characterization of solids. Solids synthesized in the experiments were separated
from solutions by centrifuge. Their morphology and elemental composition were
determined by scanning electron microscopy (SEM, Leo Supra 55VP with EDS
detector, Oxford INCA), and transmission electron microscopy (TEM, Philips
CM-120 operating at 120 kV with EDS detector, EDAX-Phoenix Microanalyzer).
Mineralogy was determined by X-ray diffraction (XRD, Rigaku TTRAX III and
Bruker D2 Phaser, using Cu Kα radiation), and by electron diffraction (ED) in
the TEM.
For TEM analysis, samples were diluted by suspension in de-aerated
ethanol/H2 O and mounted on carbon/collodion-coated 400 mesh-sized Cu grids.
For SEM analysis, samples were mounted on Al, carbon tape-coated SEM stubs (no
sputter coating). For XRD analysis in the Bruker D2 Phaser, samples were mounted
on airtight sample holders, also by dripping a suspension of the solid in de-aerated
ethanol/H2 O. For analysis in the Rigaku TTRAX III, samples were mixed with
glycerol to prevent oxidation and mounted on sample holders. All sample
preparation was done in the anaerobic glovebox, after which samples were
transferred in an underpressurized desiccator to the analytical facilities.
XRD patterns were automatically compared with the International Center for
Diffraction Data (ICDD) database. Lattice parameters (d-spacing) derived from
ED patterns were manually compared with d-spacings from the ICDD database
and from the RRUFF database44 . The average oxidation state of iron in the samples
was not determined. The results of the analyses are shown in Supplementary
Table 2.
Dynamic iron cycle model. The conceptual model is shown in Supplementary
Fig. 5, and the parameter values used are listed in Supplementary Tables 3 and 4.
Parameter choices for a time-dependent calculation over the Archaean and part of
the Proterozoic are discussed in the Supplementary Information.
The ocean is divided into three boxes representing the surface (0–100 m),
intermediate (100–1,000 m), and deep (1,000–4,000 m) ocean. The volume of each
box is calculated under the assumption that oceans cover 85% of Earth’s surface,
which accounts for a smaller land area in the Archaean and early Proterozoic.
Results are insensitive to the choice of fractional land area.
NATURE GEOSCIENCE | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
The mass budgets of FeII and FeIII in the surface ocean are:
d dt
FeII
s
= JRiv × FeII
×
d dt
FeIII
Riv
II II Fe s − Fe i − 4 × JGR − JSid
s
− kph + kO2 × FeII s − Js−i
ARTICLES
factor between 0.1 and 100 in the sensitivity analysis. The value of this scaling
factor in the time-dependent simulations is discussed in the section describing
these simulations.
Mineral precipitation fluxes JGR , JFH and JSid are related to the degree
of saturation:
J=
= kph + kO2 × FeII s − Js−i
×
III III Fe s − Fe i − 2 × JGR − JFH
where subscripted s and i denote the surface and intermediate depth ocean,
respectively. JRiv is the relative riverine water flux (the actual flux in litres per
second divided by the volume of the surface ocean in litres), and [FeII ]Riv is the
concentration of FeII in rivers. Js−i is the relative exchange of water between the
surface and intermediate ocean, kph and kO2 are the rate constants for oxidation by
light-driven processes and O2 , respectively (in units of s−1 ), and JGR , JSid and JFH are
the precipitation fluxes of GR, siderite and ferrihydrite (in moles of precipitate per
litre), respectively, and are described below.
The value of JRiv was chosen to equal the present-day flux
0
(JRiv
= 4 × 1016 l yr−1 )45 , scaled by the fraction of land area relative to the present
0
0
0
= 0.3. [FeII ]Riv was varied between
, where fland
× fland /fland
land fraction, JRiv = JRiv
0 and 1 × 10−6 M. For lack of specific constraints, the exchange flux between the
surface and intermediate ocean was chosen to equal estimates of the present value
(∼1.9 × 1018 l yr−1 ), as constrained by hydrographic data46 . There is uncertainty in
the choice of exchange fluxes, due, for example, to varying continental
configuration on the timescales of interest. However, the physics of a fluid on a
rotating planet, and the temperature and salinity budgets at the ocean’s surface
dictate that ocean fluxes cannot be radically different from the adopted values,
except for geologically brief intervals.
The abiotic rate constant for oxidation by O2 is from ref. 47:
(C − Csat )
τ
× fmin
where C is the concentration (FeII or FeIII ), Csat is the concentration at saturation of
the mineral of interest, τ is a timescale for return to saturation (chosen to be 1 h),
and fmin is a unitless factor, which increases sigmoidally from zero when C < Csat to
unity when C ≥ 2 × Csat . Mineral dissolution is unaccounted for, as fmin is zero
whenever minerals are undersaturated. This does not affect the results at a steady
state on the timescales of interest here. Details of the calculation of fmin and Csat for
the minerals included in the model are in the Supplementary Information.
Similar mass budgets can be constructed for the intermediate ocean:
d II Fe i = Js−i × FeII s − FeII i − Ji−d × FeII i − FeII d − 4 × JGR − JSid
dt
d III Fe i = Js−i × FeIII s − FeIII i − Ji−d × FeIII i − FeIII d − 2 × JGR − JFH
dt
where a subscripted d denotes the deep ocean, Ji−d is the relative exchange of water
between the intermediate and deep ocean (a water flux of ∼1.2 × 1018 l yr−1
(ref. 50), divided by the volume of the intermediate ocean). The mineralization
fluxes are defined identically to those in the surface ocean, except that the
concentrations of FeII and FeIII in the intermediate ocean are used. In the
deep ocean:
d II Fe d = JHyd × FeII Hyd + J
dt
i−d
×
II II Fe i − Fe d − 4 × JGR − JSid
d III Fe d = Ji−d × FeIII i − FeIII d − 2 × JGR − JFH
2
= k × [O2 ] × OH−
kab
O2
dt
log k = log k0 − 3.29 × I 0.5 + 1.52 × I
log k0 = 21.56 − 1,545/T
where I is the ionic strength (0.7, similar to modern seawater), [OH− ] is the
hydroxide concentration (in M), [O2 ] is the concentration of dissolved oxygen
(in M), [O2 ] = KH × pO2 , and
KH = 1.3 × 10−3 × e1,500×( T − 298.15 )
1
1
is Henry’s law constant for O2 (M atm−1 ). We account for microaerophilic
microbial oxidation of FeII , which can be many times faster than abiotic oxidation,
using information from culture experiments48 . The total FeII oxidation rate
constant (abiotic + microbial) is:
kO2 =
kab
O2
1 − fbio
II
where fbio if the fraction of microbial Fe oxidation out of the total oxidation as a
function of pO2 (in atm). At the upper boundary of the experimental results:
fbio = 1.074 × 10−3 × e−3.446×log[pO2 ]
We cap fbio at 0.95, higher than the maximal experimental value (0.88). This choice
and the fit to the upper (rather than lower) boundary of the experimental results,
maximizes fbio and, therefore, the total oxidation rate constant. This conservative
choice leads to underestimation of the importance of GR as a function of
atmospheric pO2 .
The base rate constant for photo-oxidation of FeII was derived from the results
of published photo-oxidation experiments49 , in which a photo-oxidation flux of
200 mg FeII cm−2 yr−1 for a modern solar flux and down to a depth of 100 m in a
water column with 100 µM FeII was calculated. This translates into a rate constant
∼1.1 × 10−7 s−1 . As light-driven oxidation occurs only in the surface ocean box,
this rate constant must be normalized by the depth of the surface ocean,
which is here identical to that in the above study, making for a normalization factor
of unity.
Anoxygenic FeII photosynthesis can be many times more rapid than abiotic
oxidation6 . Conversely, an iron redoxcline located several metres to tens of metres
below the ocean surface can result in less radiation available for light-driven FeII
oxidation, both biological and abiotic. To account for possibly higher and lower
rates of photo-oxidation, we multiplied the base rate constant by a unitless scaling
where mineral precipitation fluxes are defined as above. The default value of the
flux of water out of near-axis hydrothermal vents, JHyd , was taken to be ∼2 times
the present-day flux of 3 × 1013 l yr−1 (ref. 45), divided by the volume of the deep
ocean box. According to the dependence of the hydrothermal flux on the heat flux
out of the oceanic crust through time51 , this is equivalent to the hydrothermal water
II
flux
II∼
2,800 million years ago. The concentration of Fe in hydrothermal fluids,
Fe Hyd , was taken to be 80 mM, as indicated by thermodynamics of sulfate-poor
hydrothermal solutions52 .
We did not consider the fate of the mineral particles after their precipitation
from seawater (that is, sinking, deposition, and early diagenesis), as we aim to
assess the fraction of iron exiting the ocean via a GR precursor. The exact timing of
transformation into the thermodynamically stable minerals, whether still in the
water column or in the sediments, is a separate and important problem, which was
not addressed here. On the basis of the experiments conducted in this study for GR,
and in other studies for FeIII oxyhydroxides, we assume that these primary
precipitates transform into the stable minerals observed in IFs on relatively short
timescales, under the appropriate chemical conditions (for example, presence of
organic matter and FeIII reducers for the transformation of ferrihydrite into
FeII -bearing or mixed-valence minerals).
Overall in the ocean at the model steady state, the hydrothermal and riverine
source of FeII is balanced by oxidation in the surface ocean (by O2 and light-driven
processes), and by precipitation of FeII -bearing minerals (GR or siderite). The FeIII
source (FeII oxidation) is balanced by precipitation of FeIII -bearing minerals (FeIII
oxyhydroxides or GR).
Except for the pH sensitivity analysis, in which ocean pH was prescribed
independently, ocean pH was calculated from charge balance (for all other
calculations), given pCO2 , and the concentration of conservative ions in seawater:
[H+ ] − [OH− ] − [HCO3 − ] − 2 × [CO3 2− ] + ...
... + [Na+ ] + [K+ ] + 2 × [Mg2+ ] + 2 × [Ca2+ ] + 2 × [Fe2+ ] − [Cl− ]
−[Br− ] − 2 × [SO4 2− ] = 0
For lack of better constraints, the concentrations of Na+ , K+ , Mg2+ , Cl− and Br−
were taken to be identical to the present-day ocean (469, 10.2, 52.8, 546 and
0.84 mM, respectively). The concentration of Ca2+ was calculated using the
prescribed value of pCO2 and assuming CaCO3 saturation. The concentration of free
Fe2+ was taken from the dynamic model. Substituting [OH− ] = Kw /[H+ ],
[HCO3 − ] = pCO2 × KH × KA1 /[H+ ], [CO3 2− ] = pCO2 × KH × KA1 × KA2 /[H+ ]2 ,
NATURE GEOSCIENCE | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO2878
ARTICLES
[Ca2+ ] = KSP × [H+ ]2 /(pCO2 × KH × KA1 × KA2 ) into the above charge balance
yields a quartic equation for [H+ ], which has only one real, positive root.
Code availability. The code used to calculate the fractions of green rust,
ferrihydrite and siderite in the flux of iron to the sediments is available on request.
Data availability. Representative electron microscope images, electron diffraction
patterns and X-ray diffraction patterns used to identify the primary precipitates
and ageing products are available in the Supplementary Information.
References
31. Bethke, C. Geochemical and Biogeochemical Reaction Modeling (Cambridge
Press, 2008).
32. Johnson, J. W., Oelkers, E. H. & Hegelson, H. C. SUPCRT92: a software
package for calculating the standard molal thermodynamic properties of
minerals, gases, aqueous species, and reactions from 1 to 5000 bar and 0 to
1000 ◦ C. Comput. Geosci. 18, 899–947 (1992).
33. Tosca, N. J. et al. Geochemical modeling of evaporation processes on Mars:
insight from the sedimentary record at Meridiani Planum. Earth Planet. Sci.
Lett. 240, 122–148 (2005).
34. Rickard, D. & Luther, G. W. III. Chemistry of iron sulfides. Chem. Rev. 107,
514–562 (2007).
35. von Paris, P. et al. Warming the early Earth—CO2 reconsidered. Planet. Space
Sci. 56, 1244–1259 (2008).
36. Crowe, S. A. et al. Sulfate was a trace constituent of Archean seawater. Science
346, 735–739.
37. Siever, R. The silica cycle in the Precambrian. Geochim. Cosmochim. Acta 65,
3265–3273 (1992).
38. Drissi, S. H., Refait, Ph., Abdelmoula, M. & Génin, J. M. R. The preparation
and thermodynamic properties of Fe(II)-Fe(III) hydroxide-carbonate (green
rust 1); Pourbaix diagram of iron in carbonate-containing aqueous media.
Corros. Sci. 37, 2015–2041 (1995).
39. Haynes, W. M. CRC Handbook of Chemistry and Physics
97 edn (CRC Press, 2016).
40. Parker, V. B. & Khodakovskii, I. L. Thermodynamic properties of the aqueous
ions (2+ and 3+) of iron and the key compounds of iron. J. Phys. Chem. Ref.
Data 24, 1699–1745 (1995).
41. Refait, Ph. & Génin, J. M. R. The transformation of chloride-containing green
rust into sulphated green rust two by oxidation in mixed Cl− and SO4 2−
aqueous media. Corros. Sci. 36, 55–65 (1994).
42. Olowe, A. A. & Génin, J. M. R. in Corrosion Science and Engineering:
Proceedings of an International Symposium in honour of Marcel Pourbaix’s 85th
Birthday Vol. 158 (eds Rapp, R. A., Gocken, N. A. & Pourbaix, A.) RT 297
(Rapports Techniques CEBELCOR, 1989).
43. Gayer, K. H. & Wootner, L. The hydrolysis of ferrous chloride at 25◦ . J. Am.
Chem. Soc. 78, 3944–3946 (1956).
44. Downs, R. T. The RRUFF Project: An Integrated Study of the Chemistry,
Crystallography, Raman and Infrared Spectroscopy of Minerals
(19th General Meeting of the International Mineralogical
Association, 2006).
45. Elderfield, H. & Schultz, A. Mid-ocean ridge hydrothermal fluxes and the
chemical composition of the ocean. Annu. Rev. Earth Planet. Sci. 24,
191–224 (1996).
46. Ganachaud, A. & Wunsch, C. Improved estimates of global ocean circulation,
heat transport and mixing from hydrographic data. Nature 408,
453–457 (2000).
47. Millero, F. J., Sotolongo, S. & Izaguirre, M. The oxidation kinetics of Fe(II) in
seawater. Geochim. Cosmochim. Acta 51, 793–801 (1987).
48. Druschel, G. K., Emerson, D., Sutka, R., Suchecki, P. & Luther, G. W. III
Low-oxygen and chemical kinetic constraints on the geochemical niche of
neutrophilic iron(II) oxidizing microorganisms. Geochim. Cosmochim. Acta 72,
3358–3370 (2008).
49. Anbar, A. D. & Holland, H. D. The photochemistry of manganese and the
origin of banded iron formations. Geochim. Cosmochim. Acta 56,
2595–2603 (1992).
50. Hotinski, R. M., Kump, L. R. & Najjar, R. G. Opening Pandora’s Box: the
impact of open system modeling on interpretations of anoxia.
Paleoceanography 15, 267–279 (2000).
51. Sleep, N. H. & Zahnle, K. Carbon dioxide cycling and implications for climate
on ancient Earth. J. Geophys. Res. 106, 1373–1399 (2001).
52. Kump, L. R. Jr & Seyfried, W. E. Hydrothermal Fe fluxes during the
Precambrian: effect of low oceanic sulfate concentrations and low hydrostatic
pressure on the composition of black smokers. Earth Planet. Sci. Lett. 235,
654–662 (2005).
NATURE GEOSCIENCE | www.nature.com/naturegeoscience
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.