School of Engineering and Science - Jacobs University

Iron sedimentation and the neodymium isotopic composition of
Archean seawater as inferred from banded iron-formations
by
Brian W. Alexander
A thesis submitted in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
in Geochemistry
Approved, Thesis Committee
_____________________________________
Prof. Dr. Michael Bau, Chair
Jacobs University Bremen
_____________________________________
Prof. Dr. Andrea Koschinsky
Jacobs University Bremen
_____________________________________
Dr. Per Andersson
Swedish Museum of Natural History
_____________________________________
Prof. Dr. Peter Möller
GeoForschungsZentrum Potsdam
_____________________________________
Prof. Dr. Jens Gutzmer
Technische Universität Bergakademie Freiberg
Date of Defense: February 10th, 2009
School of Engineering and Science

ABSTRACT
The neodymium (Nd) isotopic composition of circa three billion-year-old (Ga)
seawater is inferred from studies of banded iron-formations (IF) from South Africa.
Iron-formations from the 2.9 Ga Pongola Supergroup display negative Є Nd values that
are indistinguishable from local shale, suggesting that Nd derived from weathering of
continental crust dominated the Nd isotopic composition of shallow seawater. In
contrast, Є Nd (t) in IF from the 2.95 Ga Pietersburg greenstone belt is clearly
distinguishable from local clastic material, and indicates that where Nd was not sourced
from local crust, seawater could possess more radiogenic Є Nd values of +1. The
Pietersburg IF Є Nd (t) values are therefore interpreted to represent bulk Archean seawater
not affected by local crustal Nd signatures. This Є Nd value of +1 is very similar to
published Є Nd (t) estimates for Archean seawater, when only those IFs which display rare
earth element ratios (REE) common to seawater are considered. Screening the data in
such a manner clearly identifies seawater precipitates, and relatively constant Є Nd (t) of
+1 to +2 for Archean IFs suggest these values are a good average for 2.5-3.8 Ga
seawater.
If bulk Archean seawater typically displayed Є Nd (t) of +1, then hydrothermal
alteration of mafic oceanic crust is considered the likely process for delivering relatively
radiogenic Nd to seawater. Data from this study and previous work suggest significant
high temperature (T) fluid input to seawater between 2.5-3.8 Ga, as supported by
positive Eu anomalies in IFs, and the absence of negative Ce anomalies in any Archean
IFs indicates that seawater was more reducing than modern seawater. Modern high-T
fluids altering mafic oceanic crust possess both positive Є Nd values and Eu anomalies,
and it is possible that these reducing, Fe-rich fluids exerted a greater control on Archean
seawater chemistry than is observed today.
However, the negative Є Nd (t) values for the Pongola IF data indicate that locally,
shallow seawater could be isotopically distinct, and possess an Nd isotopic signature
identical to local crust. If world seawater was isotopically homogenous with respect to
Nd between 2.9-3.0 Ga, these data indicate a significant shift from positive Є Nd values in
seawater during deposition of the Pietersburg IFs, to negative Є Nd values during
deposition of the Pongola IFs. This is considered unlikely, and the Nd isotopic data are
rather interpreted to reflect that, similar to modern oceans, ~3.0 Ga world oceans were
composed of water masses that could be isotopically distinct with respect to Nd.
i

The process(es) by which IFs were deposited in Archean seawater is also
investigated through a study of Fe-rich shales formed in close association with the
Pongola IFs. Major and trace element concentration data for these Fe-rich shales
indicate post-depositional mobility and extreme loss of alkali metals and barium (Ba).
Mobility of these elements in modern clastic sediments has been observed in the
presence of pore water NH + +
4 , and a significant presence of NH 4 in Fe-rich Archean
+
sediments would strongly suggest a biologic origin for IFs. However, a conclusive NH 4
signature is not observed in the Pongola Fe-rich shales, and the behavior of the alkali
metals and Ba is best explained by the presence of significant concentrations of
dissolved Fe 2+ in sediment pore waters and/or bottom seawater. This suggests that
precipitation of Fe(III) mineral phases, either in the water column or at the sediment
water interface, was an unlikely mechanism for the deposition of the primary Fe-rich
sediment.
While studies of seawater prior to ~2.7 Ga are generally constrained to
investigations of IFs, cherts, and rare fluid inclusions, the occurrence of carbonate rocks
is extensive after 2.7 Ga. Many Archean and Paleoproterozoic carbonates are dolomitic
and/or highly silicified, and the retention of primary seawater trace metal distributions
in these rocks has not been conclusively demonstrated. A study of limestones and
silicified dolomites from the ~2.25 Ga Mooidrai carbonates indicates that silicified
dolomites retain primary trace metal signatures consistent with seawater and
indistinguishable from coeval limestones. Therefore, silicified dolomites free of clastic
detritus are considered suitable archives for proxies of ancient seawater.
The accuracy and reproducibility of the trace metal concentration data for 32
elements are evaluated, and the inductively coupled plasma mass spectrometry (ICPMS)
analytical methods used are described. The ICPMS data for Sm and Nd in IF samples
are in excellent agreement with thermal ionization mass spectrometry (TIMS)
measurements, supporting the accuracy of the reported REE data. The data quality is
discussed in the context of repeated analyses of certified reference materials (CRMs)
commonly used in geochemical research. Measured concentrations for IF and carbonate
CRMs match reference data well; however, some elements suffer from major element
interferences in particular rock types. This includes Co in carbonate rocks, and Nb in
Fe-rich rock types. The concentrations of many trace metals in some CRMs are poorly
constrained, and early published datasets appear to frequently overestimate the
abundances of these elements.
ii

ACKNOWLEDGEMENTS
A special thank you to my family and friends for their encouragement and support,
and particularly to Claudia Nitzschmann, as three years became four, and four years
became five. Thanks as well to my advisor, Michael Bau, for introducing me to the
numerous, endlessly fascinating topics regarding the geochemical evolution of the early
Earth. I must confess a compelling attraction to a subject where it is so difficult to prove
my assertions wrong…
A debt of gratitude also to the many people that I came to know and appreciate
within the geochemistry working group at Jacobs, including, but not limited to, faculty
(A. Koschinsky), students (K. Schmidt, S. Kulaksiz), and staff (D. Meißner, J. Mawick,
A. Moje). The various excursions, meetings, and long hours in the lab were much more
enjoyable because of their presence.
from Holland (1972)
iii

This thesis represents original and independently conducted research that has not
been submitted to any other university for the conferral of a degree.
Brian W. Alexander
Bremen, Germany - Jan. 20, 2010
v

TABLE OF CONTENTS
Chapter 1. Introduction..................................................................................................... 1
1.1. Statement of research goal .................................................................................... 1
1.2. Structure of thesis.................................................................................................. 2
1.3. Study area.............................................................................................................. 3
1.4. Research methods.................................................................................................. 5
1.5. Background and review of relevant literature....................................................... 5
1.5.1. Rare earth element geochemistry ................................................................... 6
1.5.2. Rare earth elements in seawater .................................................................... 9
1.5.3. Sm-Nd isotopic signatures in the Earth’s crust and seawater ..................... 20
1.5.4. Iron-formations as seawater archives .......................................................... 24
Chapter 2. Trace element analyses in geological materials using low resolution
inductively coupled plasma mass spectrometry (ICPMS)............................ 39
(published as Jacobs University Technical Report No. 18)
Chapter 3. Continentally-derived solutes in shallow Archean seawater: Rare earth
element and Nd isotope evidence in iron formation from the 2.9 Ga
Pongola Supergroup, South Africa............................................................. 119
(published in Geochimica et Cosmoschmica Acta 72 (2008) 378-394)
Chapter 4. Neodymium isotopes in Archean seawater and implications for the
marine Nd cycle in Earth’s early oceans .................................................... 139
(published in Earth and Planetary Science Letters 283 (2009) 144-155)
Chapter 5. Anoxygenic photoautotrophs and the origin of banded iron-formation ..... 153
Chapter 6. Preservation of primary REE patterns without Ce anomaly during
dolomitization of Mid-Paleoproterozoic limestone and the potential
re-establishment of marine anoxia immediately after the “Great
Oxidation Event” ........................................................................................ 165
(published in South African Journal of Geology (2006) 81-86)
Chapter 7. Concluding remarks.................................................................................... 173
vii

viii

Chapter 1. Introduction
1.1. Statement of research goal
The research presented in this thesis relates to physical and chemical processes
occurring in seawater during the first 2 billion years (2 Ga) of Earth’s history, and
particularly addresses seawater chemistry ca. 3.0 Ga. As samples of seawater from this
time period are not available, except perhaps as microscopic fluid inclusions in some
geologic samples, the research fundamentally rests upon archives of seawater chemistry
that exist in the rock record. For studies of the Archean oceans (ca. 2.5-4.0 Ga) archives
of seawater chemistry are limited due to the scarcity of the rock record. However,
banded iron-formations (IFs) are common and extensive constituents of the Archean
geologic record, and the work described here exclusively deals with IFs and Fe-rich
sediments associated with IF deposition. Iron-formations were formally defined by
James (1954) as “a chemical sediment, typically thin-bedded or laminated, containing
15 per cent or more iron of sedimentary origin, commonly but not necessarily
containing layers of chert”. These chemical sediments are clearly of marine origin (see
Simonson, 2003, for details), and due to their ubiquitous occurrence during the
Archean, IFs are likely one of the best seawater archives prior to ~2.7 Ga.
One question regarding Archean oceans is the extent to which seawater chemistry
was controlled by weathering of oceanic or continental crust. Different conclusions have
been drawn, with some workers supporting weathering of continental crust as a
dominant control on seawater chemistry (e.g., Miller and O’Nions, 1985), while others
have supported hydrothermal alteration of oceanic crust as a primary control (Jacobsen
and Pimentel-Klose, 1988a, 1988b; Veizer et al., 1989; Alibert and McCulloch, 1993;
Bau et al. 1997a). The research presented here attempts to resolve these discrepant
conclusions through new trace metal and isotopic data for ca. 3.0 Ga IFs. A secondary
1

goal of the research is to discern if the process by which IFs formed might possibly be
constrained. Though not the direct focus of this thesis, elucidating the manner in which
IFs precipitated in ancient seawater is attractive, as these sediments remain a
geochemical mystery even after more than 50 years of study.
The general approach followed is to measure the isotopic ratios of the rare earth
elements Sm and Nd, assuming that these measured ratios reflect the ambient seawater
that precipitated the studied IFs. The inferred Sm-Nd isotopic signature of seawater
would reflect different sources of Sm and Nd to ancient oceans, and would constrain the
relative importance of these sources on seawater chemistry. As ancient oceanic and
continental crust possess distinctly different Sm-Nd isotopic signatures, it should be
possible to determine if weathering of oceanic or continental crust was more important
in controlling seawater chemistry.
1.2. Structure of thesis
This thesis is structured as a compilation of independent manuscripts, with the
exception of this, the introductory first chapter, and the final chapter that contains
concluding remarks. Each individual manuscript contains an abstract, background
information, the discussion of results, conclusions, and references. The thesis in total
consists of six chapters that are outlined in the Table of Contents.
This, the first chapter, introduces the research goals and provides background
information relevant to the thesis which is not covered extensively in the individual
manuscripts. The manuscripts that follow are generally written to facilitate publication
in international academic research journals. One exception is Chapter 2, as it describes
the relevant analytical methods employed within the Jacobs University Bremen (Jacobs)
Geochemistry Lab, and which has been published as Jacobs Technical Report No. 18.
2

As a result of this approach, slight differences will exist from chapter to chapter
regarding organization or format. If already published or submitted for publication in an
academic journal, the manuscripts reflect the individual formatting and length
requirements of the respective journal, and where appropriate, the manuscript/chapter is
the printed version of the journal article.
1.3. Study area
The geologic samples used in this thesis are exclusively from South Africa, and
Figure 1 lists the major relevant geologic features of southern Africa. Since the research
goal revolves around the chemistry of ancient seawater, this part of the world is an ideal
area for study. The Kaapvaal craton is one of the oldest and most stable cratons known
in the world (Tankard, 1982), and the oldest rock yet dated on the Kaapvaal craton, and
probably the oldest on the African continent (Poujol et al., 2002), is a 3644 ±4 Ma
tonalitic gneiss from northwest Swaziland (Compston and Kröner, 1988). The Kaapvaal
craton hosts several Archean sequences containing iron-formation deposited in different
tectonic settings, which is advantageous as these represent seawater precipitates from
different marine environments, e.g., island-arc or stable continental shelf. Examples
include IF within the >3.2 Ga Barberton greenstone belt, considered to have formed in
an oceanic plateau/island-arc setting (Lowe, 1994), as well as IF deposited on the stable
Kaapvaal craton during the formation of the ~2.5 Ga Transvaal Supergroup (Beukes,
1983)
The IFs discussed in this thesis were selected as they represent seawater
precipitates from different tectonic settings, yet are of a similar age (ca. 2.9-3.0 Ga).
Samples were obtained from the 2.9 Ga Mozaan Group of the Pongola Supergroup,
which was deposited within a relatively shallow, near-shore continental shelf
3

Figure 1. Map of southern Africa showing locations of major geologic features discussed in thesis. Ironformation
units that provided samples for geochemical analyses were obtained from the Pongola
Supergroup and the Pietersburg greenstone belt.
environment, as well as from the 2.95 Ga Pietersburg greenstone belt, which is
considered to represent a more open ocean, island-arc environment (Fig. 1). The
hypothesis is that Sm-Nd isotopic signatures in IFs of similar age, yet from different
depositional environments, might possess different Nd isotopic signatures that reflect
the respective Nd sources to Archean seawater.
4

1.4. Research methods
The research methods employed for this thesis consist of various techniques for
obtaining geochemical analyses of rock samples. Three analytical techniques are used
for performing elemental and isotopic measurements of the samples discussed. The first
technique is X-ray fluorescence (XRF), which provided major element concentration
data. The second analytical method is inductively coupled plasma mass spectrometry
(ICPMS), which was used for trace metal concentration measurements. The third
analytical method is thermal ionization mass spectrometry (TIMS), which was used for
measuring Sm-Nd isotopic ratios.
The XRF analyses were performed either at the GeoForschungsZentrum (GFZ) in
Potsdam, Germany, or at the University of Johannesburg (South Africa) SpectRAU
facility. The trace metal concentrations determined by ICPMS were provided either by
the GFZ, for the initial research conducted for this thesis, or by the author using
equipment within the Geochemistry Lab at Jacobs University Bremen (JUB). All Sm-
Nd isotopic analyses were performed at the Laboratory for Isotope Geology at the
Swedish Museum of Natural History in Stockholm, Sweden. Details of the XRF
analyses are not discussed, as XRF techniques are well-developed and are only used for
determining major element concentrations (e.g., Si, Al, Fe). A detailed description of
the ICPMS analytical methods is provided in Chapter 2, and is applicable to data
produced at both the GFZ and JUB. Descriptions of the Sm-Nd isotopic analyses are
provided within the individual chapters where relevant.
1.5. Background and review of relevant literature
The principle conclusions of this thesis rest upon the isotopic ratios of Nd in
Archean seawater as inferred from Sm-Nd analyses of iron-formations. Therefore, it is
5

appropriate to briefly discuss: 1) basic rare earth element geochemistry, 2) rare earth
element behavior in seawater, 3) Sm-Nd isotopic signatures in the Earth’s crust and
seawater, and 4) the use of IFs as archives of ancient seawater chemistry.
1.5.1. Rare earth element geochemistry
The rare earth elements include Sc, Y, and the fifteen lanthanide elements which
have atomic masses between 57 and 71 (La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy, Ho,
Er, Tm, Yb, Lu). Hereafter in this thesis, the lanthanides will be exclusively referred to
as the rare earth elements (REE), excluding Sc and Y. However, as described below,
yttrium is useful when grouped with the lanthanides for studying geochemical
processes, and where yttrium data are available discussions of the REE and Y will be
indicated by the acronym REY.
Among the lanthanides promethium (Pm) is a special case, as it has three
radioactive isotopes ( 145 Pm, 146 Pm, 147 Pm), of which 145 Pm has the longest half-life of
~17.7 years. As such Pm is considered an extinct element in natural samples and is not
included in discussions of REY behavior in geological studies. The remaining REY
display similar geochemical behavior, primarily as a result of their +3 valence state,
though Ce and Eu may also exist in the +4 and +2 valence states, respectively. This
geochemically coherent behavior among the REY is also a consequence of very similar
ionic radii between 0.977 Å (Lu) and 1.16 Å (La). The similar ionic radii results from a
progressive filling of the 4f electron orbital that produces a monotonic decrease in the
ionic radii with increasing atomic mass (Figure 2). As a result, lanthanide behavior
tends to vary smoothly and predictably as a function of ionic radii (i.e., atomic mass) in
many geochemical systems.
6

4
Ce
valency
3
Sc
Lu
Y
La
Eu
2
0.8 0.9 1.0 1.1 1.2 1.3
ionic radius (Å)
Figure 2. Valence state and ionic radius (coordination number 8) for rare earth elements. Yttrium has an
ionic radius that falls between Dy and Ho. Cerium and Eu are the only rare earth elements that exist at
valence states other than +3 (ionic radii data from Shannon, 1976).
Rare earth element data are typically normalized for plotting, as REY
concentrations may vary by an order of magnitude (or more) between adjacent elements
in the series, due to the greater abundance of elements with even atomic numbers
relative to elements with odd atomic numbers (Oddo-Harkins rule). Figure 3 shows data
for mid-ocean ridge basalt (MORB) that has been normalized to C1 chondritic meteorite
(Anders and Grevesse, 1989), as well as non-normalized data for comparison, and
where Y has been inserted between Dy and Ho according to its ionic radius (see Figure
2). If none of the REY exhibit anomalous behavior then the normalized REY pattern
will vary smoothly across the series as seen in Figure 3. It should be noted that the C1
chondritic meteorite is often used for normalizing REY data, as it is considered to
possess a REY distribution equivalent to bulk silicate Earth. However, it is sometimes
useful to normalize REY data to other geochemical reservoirs, and another commonly
used REY dataset for normalization purposes (and one used extensively in this thesis) is
upper continental crust, as represented by the Post Archean Average Shale (PAAS) of
McLennan (1989).
7

100
10
MORB
1
0.1
raw data (mg/kg)
chondrite-normalized
La Ce Pr Nd
Sm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 3. Typical rare earth element plot where lanthanide elements are arranged in order of increasing
atomic mass and decreasing ionic radii, with yttrium inserted between Dy and Ho in accordance with its
ionic radius. Normalizing produces smooth REY patterns if none of the REY exhibit anomalous
behavior. Mid-ocean ridge basalt (MORB) data from Niu et al. (1999).
Some of the rare earth elements may be fractionated from neighboring REY as the
result of natural geochemical processes, with Ce and Eu offering the best known
examples. The fractionation of Ce and Eu from the other REY is primarily a
consequence of the different valence states that these elements can possess. Cerium is
readily oxidized from Ce 3+ to Ce 4+ at the Earth’s surface and in the oxic marine
environment, fractionating it from the other REY by forming Ce(IV) compounds of low
solubility. At elevated temperatures (>200 °C) europium can fractionate from the other
REY when Eu is reduced from Eu 3+ to Eu 2+ , e.g., during igneous or hydrothermal
processes, with Eu fractionation occuring due to the different charge and significantly
larger ionic radius of the Eu 2+ ion (Fig. 2). Examples of Ce and Eu fractionation are
presented in Figure 4. The fractionation of Ce from the neighboring elements La and Pr
is shown for average world seawater, and Eu fractionation as observed in high
8

10 -2 360 °C hydrothermal fluid
10 -3
10 -4
sample/PAAS
10 -5
10 -6
10 -7
10 -8
10 -9
average world seawater
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 4. REY distributions in average world seawater and high-T hydrothermal fluid from the mid-
Atlantic ridge normalized to continental crust as represented by PAAS (McLennan, 1989). Seawater
data are the average of 172 worldwide analyses of all depths (average depth 1560 m) for which
complete REY datasets are available (Zhang and Nozaki, 1996; Alibo and Nozaki, 1999; Bau et al.,
1997b; Nozaki et al., 1999; Nozaki and Alibo, 2003). Hydrothermal fluid data for sample BS-07-3/3 of
Bau and Dulski (1999). Note strong negative Ce anomaly in seawater and strong positive Eu anomaly in
high-T hydrothermal fluid.
temperature (T= 360 °C) hydrothermal fluids emanating from the mid-Atlantic oceanic
ridge (MAR).
1.5.2. Rare earth elements in seawater
The abundances of the REY in seawater are extremely low relative to their crustal
averages, with world seawater concentrations ~10 -7
of the levels found in upper
continental crust (see Figure 4). Higher REY abundances are typically found in river
waters and pore waters of marine sediments, though not as high as the concentrations
observed for high-T hydrothermal fluids (Figure 5). It is reasonable that these fluids
(river water, pore water, and hydrothermal fluid) represent the primary dissolved REY
sources for the oceans of the world, and assuming this is true, it is notable that each of
these REY sources contains several times the total REY concentration that exists in
9

10 -2 hydrothermal fluid
10 -3
10 -4
sample/PAAS
10 -5
10 -6
10 -7
10 -8
10 -9
river water
marine pore water
estuaries
world seawater
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 5. Average REE and REY distributions in natural waters. Seawater and hydrothermal fluid data
are the same as in Fig. 4. River water is an average of 155 filtered (0.2-0.45 μm) samples with pH
between 5-9, and represents complete REY datasets for rivers in Venezuela, Australia, USA, and
Sardinia (Tosiani et al., 2004; Lawrence et al., 2006; Bau et al. 2006; Cidu and Biddau, 2007), as well as
complete REE data for rivers in France and Argentina (Tricca et al., 1999; Gammons et al., 2005b; Stille
et al., 2006). The marine pore water REE pattern is the average of 89 filtered (0.2-0.45 μm) samples
compiled from isotope dilution REE data for the eastern and western margins of North America
(Elderfield and Sholkovitz, 1987; German and Elderfield, 1989; Sholkovitz et al., 1989; Sholkovitz et al.,
1992), as well as complete REE datasets for pore waters from the western margins of North and South
America (Haley and Klinkhammer, 2003; Haley et al., 2004). Estuary water is an average of 134 filtered
(0.2-0.45 μm) samples with salinities between 1-30 ‰ (mean= 13.8 ‰, median= 14.2 ‰) and represents
instrument neutron activation analysis data for samples from France (Martin et al., 1976), as well as
isotope dilution REE data from estuaries along the margins of North America (Sholkovitz and Elderfeld,
1988; Goldstein and Jacobsen, 1988; Elderfeld et al., 1990; Sholkovitz et al., 1992), Great Britain
(Elderfeld et al., 1990), South America (Sholkovitz, 1993), and Papua New Guinea (Sholkovitz and
Szymczak, 2000). Additionally, estuarine water average incorporates complete REY data for estuaries
from Australia (Lawrence and Kamber, 2006) and Germany (Kulaksız and Bau, 2007).
modern seawater. Therefore, some process or processes must exist that effectively
remove REY from these source solutions upon mixing with ambient seawater.
The distinctly lower REY concentrations found in estuaries relative to river water
(Fig. 5) indicate that significant REY removal occurs during the mixing of freshwater
and seawater. This phenomenon has been noted in previous experimental studies (Hoyle
et al., 1984), as well studies of natural systems (e.g., Elderfield et al., 1990). The
removal of REY in estuaries is due to the coagulation and ‘salting-out’ of the major
REY-carrying phases present in river water, which are Fe-organic colloids smaller than
10

~0.2 μm (Sholkovitz, 1995). The result of this process is that more than ~70% of the
10 3 times
more REY than ambient seawater, the metal-oxides that precipitate from them
ultimately act as sinks for seawater REY (German et al., 1990; German et al., 1991a).
Pore waters within marine sediments are less well studied than either river or
estuarine waters, and REE data are available for only a handful of sites worldwide (see
Fig. 5 caption for details). Pore waters within marine sediments display strong
variations in both REE concentrations and normalized REE patterns as a function of
depth below the seawater-sediment interface (Elderfield and Sholkovitz, 1987; Haley et
al., 2004). Though data are few, it appears that increased REE concentrations in pore
waters are related to redox cycling of Fe in the upper few centimeters of the sediment
(Elderfield and Sholkovitz, 1987), as well the degradation of organic matter (Haley et
al., 2004). It also appears that pore water fluxes may be a significant source of REE to
seawater; however, the variability in REE patterns and higher concentrations found in
pore waters are generally not observed in seawater sampled a few meters above the
seawater-sediment interface (Elderfield and Sholkovitz, 1987; Haley et al., 2004),
implying that REE sourced from pore waters may be rapidly scavenged from oxic
11

seawater and returned to the sediment column. In consideration of the above
observations, river waters are the dominant source of the REY to modern seawater. It
should be noted that aeolian dust also delivers REY to the oceans, but the magnitude of
this flux to dissolved REY in seawater is debated (e.g., Nakai et al., 1993; Tachikawa et
al., 1999; Greaves et al., 1999; Bayon et al., 2004), and regardless, the REY distribution
in the aeolian flux will generally reflect continental REY sources. It should be noted
that the Pacific Ocean seems atypical in this regard, as volcanic particles derived from
intraplate and circum-Pacific volcanism appear to be a significant REY source, and the
reader is referred to Goldstein and Hemming (2003) for a more thorough discussion.
Within seawater itself, the REY display consistent behavior. Figure 6 shows rare
earth element abundances and distributions from the Atlantic, Pacific, and Indian
oceans. All of the major world oceans have been sampled, with exceptions being the
Southern Ocean (south of 60° latitude) and the central Arctic Ocean. The available rare
earth element dataset for seawater is extensive, and approximately 900 samples
worldwide have been analyzed, with the Pacific Ocean providing the most samples
(47%), followed by the Indian (29%), Atlantic (15%), Arctic (8%), and Mediterranean
(

sample/PAAS
10 -6 10 -7
deep water >2000 m
NW Pacific
NE Indian
SW Pacific
S Atlantic
10 -8
shallow water 2000 m) possesses an
indistinguishable rare earth element pattern and very similar concentrations regardless of the ocean basin
sampled.
and concentrations observed in modern seawater are very similar regardless of location
(Fig. 6).
The consistent REY patterns in modern seawater regardless of location or depth
reflect the interplay between solution complexation effects and particle surface
adsorption processes. The REY are lithophilic and generally insoluble elements, and
precipitation of REY mineral phases in equilibrium with seawater does not control REY
concentrations or distributions, as all of the lanthanides are undersaturated in seawater
(Brookins, 1989). Dissolved REE in seawater do not generally exist as hydrated free
metal ions, but are predominantly complexed by carbonate anions (CO 2- 3 ) as illustrated
in Figure 7 using the data of Cantrell and Byrne (1987).
13

100
MCO + 3 + M(CO 3 )- 2
rare earth element
% total
80
60
40
20
MCO + 3
M(CO 3
) - 2
M 3+
0
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Figure 7. Speciation of the lanthanides in seawater at 25° C and 1 atmosphere pressure with 35.1 ‰
salinity and a total carbonate ion concentration [CO 3 2- ] of 1.39 x 10 -4 moles/kg. Figure adapted from
Cantrell and Byrne (1987), and M refers to individual lanthanide elements. Hydroxide, chloride, fluoride,
and sulfate REE complexes are not shown, as these are minor chemical species affecting REE behavior
in seawater (Cantrell and Byrne, 1987).
The carbonate anions binding to dissolved REY in seawater form either M(CO 3 ) +
or M(CO 3 ) 2
-
complexes, where M refers to the individual element. Monocarbonate
M(CO 3 ) +
complexes are more prevalent for light rare earth elements (LREE), and
dicarbonate M(CO 3 ) - 2 complexes more so for heavy rare earth elements (HREE), with
the relative proportions of these carbonate complexes roughly equal for middle rare
earth elements (MREE) such as Gd and Tb. The calculated carbonate complexation
constants are relatively unaffected by changes in temperature (Cantrell and Byrne,
1987), and it is surmised that the relative lanthanide distributions between mono- and
dicarbonate complexes remains constant throughout the marine water column
(Brookins, 1989).
The solution complexation behavior of the REY strongly affects the process that
controls the distribution of these elements in seawater, which is scavenging by
particulate matter. The REY are readily adsorbed onto particle surfaces in natural
14

waters, as evidenced by the dominant association of the REY with Fe-organic colloids
in river waters (e.g., Sholkovitz, 1995). The particle reactive nature of the rare earths
was well exploited by early workers conducting seawater analyses, in which the REY
were quantitatively removed and concentrated from seawater samples by coprecipitation
with Fe-hydroxides (e.g., Goldberg et al., 1963; Høgdahl et al., 1968;
Elderfield and Greaves, 1982). This scavenging by particulate matter is considered the
primary removal process for the REY in seawater (e.g., de Baar et al., 1985a; Elderfield,
1988), and it appears that particle coatings consisting of organic films and Fe-Mn
oxides are particularly important in this scavenging process (e.g., Byrne and Kim, 1990;
Sholkovitz et al., 1994, and references therein). Experimental work indicates these
scavenging reactions approach steady state within minutes (e.g., Byrne and Kim, 1990;
Koeppenkastrop and De Carlo, 1993; Bau, 1999), suggesting that equilibrium
conditions characterize REY partitioning between seawater and particle surfaces.
The general shape of the REY patterns shown in Figs. 4-6, with a strong
enrichment of the HREE over the LREE in seawater, can be attributed to the
competition between carbonate complexation in solution and reactive particle surfaces.
The exact process by which REY are adsorbed to particle surfaces is still a matter of
study, and Piasecki and Sverjensky (2008) provide a review of proposed sorption
mechanisms in addition to a triple-layer surface complexation model that is consistent
with X-ray studies. However, to a first approximation, particle surface adsorption
constants for individual REY can be described by their first hydrolysis constants (Erel
and Morgan, 1991; Bau et al., 1996). Using such an approach, calculated partition
coefficients between particle surfaces and seawater predict a preferential LREE
enrichment for the adsorbed component (Erel and Stolper; 1993), consistent with LREE
depletion in the solution phase and the observed REY patterns for world seawater. The
15

net effect of carbonate complexation in solution and particle surface adsorption for the
REY is that heavier rare earth elements tend to remain in solution, and conversely, the
LREE partition more easily onto particle surfaces. The result is a generally monotonic
increase in shale-normalized rare earth element concentrations in seawater as a function
of increasing atomic mass. However, as seen in Figs. 4-6, some elements in the REY
series display anomalous behavior relative to their immediate neighbors.
The most strikingly anomalous element among the lanthanides is Ce, which, as
mentioned previously, can exist in the Ce 4+
valence state in oxic marine systems.
Experimental work has demonstrated that when lanthanides are adsorbed onto Mnoxides
and Fe-oxyhydroxides, positive Ce anomalies develop in the solid phases, which
have been attributed to the oxidation of Ce(III) to highly insoluble Ce(IV) on mineral
surfaces (e.g., Koeppenkastrop and De Carlo, 1992; Bau, 1999). Positive Ce anomalies
have also been noted in marine hydrogenetic Fe-Mn crusts (e.g., Bau et al., 1996, and
references therein), consistent with the oxidation of adsorbed Ce(III) to Ce(IV) and the
preferential retention of Ce on metal-oxyhydroxide surfaces in the marine environment.
It should be noted that this fractionation of Ce is not observed under anoxic conditions,
and the REY patterns of seawater sampled from anoxic basins do not display negative
Ce anomalies (e.g., de Baar et al., 1988; German et al., 1991b; Bau et al., 1997b).
The other rare earth element that displays strongly anomalous behavior in
seawater is Y, which is highly enriched relative to neighboring Dy and Ho (Figs. 4-6).
While the ionic radii of Y is similar to Ho, and these elements are considered to be
geochemical twins that exhibit similar behavior in most geological processes, Y is
strongly fractionated from Ho (and the other lanthanides) in seawater. Relative to Ho,
carbonate complexation constants for Y in seawater are similar for moncarbonate
species, and somewhat different for dicarbonate species (Luo and Byrne, 2004).
16

10 2 world seawater (x10 8 )
sample/PAAS
10 1
10 0
hydrogenetic Fe-Mn crust
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 8. Comparison of REY patterns in average world seawater (data same as Fig. 4) and average
(n=12) non-phosphatized hydrogenetic Fe-Mn crusts from the central Pacific Ocean (data from Bau et
al., 1996). Note complementary Ce and Y anomalies between seawater and precipitated Fe-Mn oxide
particles as represented by Fe-Mn crust.
However, the strong positive Y anomaly observed in seawater results from particle
surface complexation effects, as Y adsorbed to particle surfaces is bound less strongly
than the lanthanide elements (e.g., Bau, 1999). The significant fractionation of both Y
and Ce from the other REY in seawater due to processes occurring on particle surfaces
is clearly evident in the REY patterns for hydrogenetic Fe-Mn crusts as shown in Figure
8. This does not imply that hydrogenetic Fe-Mn crusts are solely responsible for these
anomalies, as Ce and Y fractionation has been observed to develop as a function of
increasing salinity in estuaries (Lawrence and Kamber, 2006). It therefore appears that
significant Ce and Y anomalies begin to form during mixing of river waters and
seawater, and may be present in the truly dissolved REY fraction of river waters (Byrne
and Liu, 1998).
Of the remaining REY, La, Gd, and Lu display anomalous behavior in seawater to
varying degrees. Strong positive La anomalies are typical of seawater, though this
17

feature was not recognized in early studies due to the negative anomaly present for
neighboring Ce (e.g., de Baar et al., 1983). Positive Gd and Lu anomalies are also
present in seawater REY patterns, though the magnitude of these anomalies are
significantly smaller (see Fig. 6). The presence of positive La, Gd, and Lu anomalies
was originally postulated to arise from a greater stability of the solution complexes for
these elements, which presumably resulted from the exactly empty, half-filled, and
completely filled 4f electron shell configurations for La, Gd, and Lu, respectively (de
Baar et al., 1985b). Early work by Cantrell and Byrne (1987), as shown in Fig. 7,
predicted no anomalous behavior for these elements, though this study reported
experimental data for only a few REY (e.g., Ce, Eu, Tb), which were then extrapolated
across the entire REY series. More recent experimental studies (Luo and Byrne, 2004)
determined equilibrium formation constants for carbonate complexes of all 15 REY,
and these results are shown in Figure 9. Equilibrium constants for the formation of
monocarbonate REY complexes in high ionic strength solutions mimicking seawater do
not vary smoothly across the REY series, with La, Gd, Y, and Lu showing the greatest
deviations. Predominance diagrams for mono- and dicarbonate REY complexes in
seawater also predict that La, Gd, Y, and Lu should fractionate from their immediate
neighbors (Fig. 9).
It is also likely that the observed positive La and Gd anomalies in seawater are to
some degree generated during adsorption/desorption processes on scavenging particle
surfaces. The authors Byrne and Kim (1990) and Lee and Byrne (1993) suggested that
organic ligands on particle surfaces (e.g., carboxylate) should discriminate against La
and Gd, producing an enrichment of these elements in seawater. This would be
consistent with the conclusions of Bau et al. (1999) regarding REY surface
complexation on hydrous Fe-Mn oxides in seawater. It therefore appears that the La,
18

6.2
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
6.0
5.8
log CO3
β 1
5.6
5.4
5.2
5.0
log CO3
β 1
= [MCO + 3 ][M3+ ] -1 [CO 2-
3 ]-1 TOTAL
9.0
8.5
M(CO 3
) - 2
8.0
7.5
7.0
C M 2
pH
6.5
6.0
5.5
5.0
4.5
MCO + 3
C M 1
M 3+
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 9. Carbonate equilibrium constants and REY speciation for seawater as predicted by Luo and
Byrne (2004). The top diagram represents equilibrium constants (molal concentration) for the formation
of monocarbonate complexes in a 0.7 molal NaClO 4 solution at 25° C. The equilibrium constants for the
REY do not vary smoothly, and local minimums occur for La, Gd, Y, and Lu. The lower diagram
predicts the predominance of REY carbonate complexes in seawater as a function of pH, and the grey
area represents a typical seawater pH range of ~7.7-8.1. The curve C M 1 represents the pH values at which
concentrations of M + and MCO 3 + are equal, and C M 2 is the pH at which MCO 3
+
= M(CO 3 ) 2 .
Gd, and Y anomalies in seawater are due to the interplay of competing solution and
surface complexation effects. Regarding Lu, it is interesting that the predominance
diagram in Fig. 9 rather suggests that Yb is anomalous. As both Yb and Lu are expected
to be primarily complexed by dicarbonate at seawater pH values, the data in Fig. 9
predict that Yb should display a negative anomaly, resulting in an apparent ‘positive’
Lu anomaly. For example, at a seawater pH of 7.5 the data of Luo and Byrne (2004)
predict a greater relative proportion of Yb exists as monocarbonate YbCO + 3 than would
be expected from interpolation between Tm and Lu. This suggests preferential
scavenging of Yb from seawater by reactive particle surfaces; however, there is no
19

sound geochemical basis for anomalous behavior of Yb (e.g., electron configuration,
etc.). Considering that the monocarbonate equilibrium constants for Yb are consistent
with those of Er and Tm, and no discussion of ‘anomalous’ dicarbonate Yb
complexation is provided by Luo and Byrne (2004), it is concluded that the slight
positive Lu anomalies observed in seawater (see Fig. 6) likely represent real excursions
of Lu concentrations from those predicted by extrapolation from Tm and Yb.
1.5.3. Sm-Nd isotopic signatures in the Earth’s crust and seawater
Samarium and Nd each have several isotopes, and 147 Sm radioactively decays to
stable 143 Nd with a half-life of 1.06 × 10 11 years. Even though both Sm and Nd have
similar geochemical behavior, like all the lanthanides, fractionation between Sm and Nd
does occur during differentiation of the Earth’s crust. Figure 10 presents complete REY
patterns for evolved continental crust and MORB that have been normalized to
chondrite (i.e., bulk silicate Earth), showing that crustal differentiation leads to
distinctive Sm/Nd ratios. Upper continental crust is enriched in Nd relative to Sm and
possesses a low Sm/Nd ratio of ~0.17, whereas MORB derived from a depleted mantle
source possesses higher Sm/Nd ratios of ~0.32. As a result of different Sm/Nd ratios
and the subsequent decay of 147 Sm to 143 Nd, Sm-Nd isotopic ratios will evolve
differently in continental and oceanic crust.
The Nd isotopic signature of a geologic sample is reported as the 143 Nd/ 144 Nd
ratio, as 144 Nd is a stable isotope whose abundance does not change with time. The Sm-
Nd isotopic system has been used extensively for dating geologic samples and a
thorough review of the techniques and applications may be found in Faure (1986). The
determination of radiomentric ages relies upon constructing an isochron using measured
147 Sm/ 144 Nd and 143 Nd/ 144 Nd ratios for a suite of cogenetic samples (e.g., minerals).
20

1000
chondrite-normalized
100
10
1
upper continental crust
MORB
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Figure 10. Typical REY distributions in upper continental crust (data from Taylor and McLennan, 1995)
and MORB (Niu et al., 1999) illustrating different Sm/Nd ratios. Light REE are strongly enriched in
continental crust, which possesses negative Eu anomalies due to partitioning of Eu 2+ into plagioclase-rich
residual material during intra-crustal partial melting and fractional crystallization.
From the isochron an initial 143 Nd/ 144 Nd ratio may be determined, which allows the
calculation of a geologic age for the samples.
Alternatively, if the geologic age of a group of samples is already known, then
measured Sm-Nd isotopic ratios in individual samples allow the calculation of initial
143 Nd/ 144 Nd ratios without the need for constructing an isochron. This is the approach
used in this thesis, as the samples investigated are marine sediments, and not igneous
rocks which can often be considered to have crystallized in a geologically instantaneous
time period. The purpose of this research is not to date the sediments studied, but rather
to examine any variations in their initial 143 Nd/ 144 Nd ratios, as these variations may shed
light on sources of Nd to Archean seawater, as well as any isotopic variability in
Archean oceans.
Measured 143 Nd/ 144 Nd ratios in a geologic sample are typically expressed using
the Є Nd (t) notation of DePaolo and Wasserburg (1976). The Є Nd (t) value describes the
21

deviation of the 143 Nd/ 144 Nd ratio measured in a sample relative to the 143 Nd/ 144 Nd ratio
in a chondritic uniform reservoir (CHUR) in parts per 10 4 :
Є Nd (t)
=
⎡(
⎢
⎢⎣
143
Nd/
I
144
t
CHUR
Nd)
i
⎤
− 1⎥
⎥⎦
x 10
4
where ( 143 Nd/ 144 Nd) i is the initial ratio in the sample (i.e., at the time it formed, t), I t CHUR
is the CHUR 143 Nd/ 144 Nd ratio at the time the sample formed, and CHUR is considered
to represent bulk silicate Earth. Since continental crust is characterized by low Sm/Nd
ratios that produce relatively low amounts of radiogenic Nd, it displays negative Є Nd (t)
values. Conversely, mafic oceanic crust derived from a depleted upper mantle is
characterized by positive Є Nd (t) values, as MORB has relatively high Sm/Nd ratios and
excess radiogenic Nd. As the REY in modern seawater are primarily sourced from the
weathering of continental crust, modern seawater possesses negative Є Nd (0) values
between -1 and -20, though extreme values in this range are restricted to shallow waters
(Goldstein and Hemming, 2003). Figure 11 shows average Є Nd (0) values for different
world oceans, as well as simple models for the Nd isotopic evolution in continental
crust and a depleted mantle over the past 4.0 Ga.
The significant Nd isotopic variations between ocean basins indicates that the
marine residence time of Nd (τ Nd ) is less than the mixing time of the oceans (~1000
years, e.g., Broecker and Peng, 1982), and estimates of τ Nd in oxic seawater typically
range from 300-1500 years (Jeandel et al., 1995; Tachikawa et al., 2003). This is
consistent with the particle reactive nature of the REE and the efficient Nd scavenging
processes operating in modern oceans, as REE-rich hydrothermal fluid with highly
positive Є Nd (0) has a negligible effect on 143 Nd/ 144 Nd ratios in world seawater (Fig. 11).
22

15
10
TAG hydrothermal fluid
5
depleted mantle
ε Nd
(t)
0
-5
Pacific Ocean
Indian Ocean
-10
Atlantic Ocean
-15
mafic igneous
-20
felsic igneous
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Ga
continental crust
Figure 11. The range of Nd isotopic ratios (grey bar) as observed in modern seawater (Goldstein and
Hemming, 2003), as well as Nd isotopic evolution in a depleted mantle and average continental crust.
Average ocean basin seawater data from Piepgras and Wasserburg, (1980), and Trans-Atlantic
Geotraverse (TAG) mid-ocean ridge hydrothermal fluid data reported by Mills et al. (2001). Isotopic
evolution lines represent simple linear models where Є Nd (4.56 Ga) = 0, depleted mantle Є Nd (0) = +9
(Salters and Stracke, 2004), and average continental crust Є Nd (0) = -17 (Goldstein et al., 1984). Also
shown are data for mafic and felsic igneous rocks from the Kaapvaal craton, illustrating that ca. 3.0 Ga
crust in the study area displayed a wide range of Є Nd (t) values. Igneous rock data from Hegner et al.
(1984), Wilson and Carlson (1989), Nelson et al. (1992), Kröner and Tegtmeyer (1994), Kröner et al.
(1996), and Chavagnac (2004).
For the early Earth, the isotopic evolution of Nd in continental crust and a
depleted mantle reservoir is debated (cf. Nägler and Kramers, 1998; Patchett and
Samson, 2003), and the isotopic evolution lines in Fig. 11 represent simple linear
models. Regardless of the model favored, however, it appears that by ~3.8 Ga Є Nd values
in the upper mantle were at least +2 (Bennet, 2003), implying that significant crustal
differentiation was occurring in the early Archean. For southern Africa, the range of
observed Є Nd (t) values observed for >2.7 Ga felsic and igneous rocks is large (≥10 Є Nd -
units, Fig. 11), indicating that crustal sources of Nd to Archean seawater in the vicinity
of the Kaapvaal craton were isotopically diverse. Due to the large range in Є Nd (t) values
for Nd sources, and assuming suitable archives for seawater 143 Nd/ 144 Nd ratios are
23

available, it should be possible to distinguish if ca. 3.0 Ga seawater near the Kaapvaal
craton was similar to modern seawater in possessing negative Є Nd (t).
1.5.4. Iron-formations as seawater archives
Reconstructing Nd isotopic ratios in paleoseawater is possible as isotopic
fractionation during incorporation of Nd into pure marine chemical sediments has not
been observed (Frank, 2002). Even marine precipitates possessing strongly fractionated
REY patterns relative to seawater, such as hydrogenetic Fe-Mn crusts (Fig. 8),
accurately record the Nd isotopic composition of ambient seawater (Goldstein and
Hemming, 2003, and references therein). Therefore, sufficiently pure marine chemical
precipitates from the Archean should represent suitable archives for reconstructing
primary seawater Nd isotopic ratios.
Potential seawater archives for the Archean would be carbonate rocks such as
limestones or dolomites; however, the carbonate record prior to ~2.7 Ga is poor,
consisting primarily of thin, discontinuous, and poorly preserved occurrences
(Grotzinger, 1989). Prior to 2.7 Ga the best candidates for archives of seawater
143 Nd/ 144 Nd ratios are iron-formations, as they are ubiquitous in both time and space for
this period of Earth’s history. While the exact process by which they formed remains
unknown, the frequent association of IFs with marine carbonates and sea level rise
(Klein and Beukes, 1989; Simonson and Hassler, 1996) supports the general consensus
that IFs are chemical sediments precipitated from seawater (cf. Simonson, 2003).
The geochemical evidence used in this thesis to screen IF samples for suitability
as seawater archives primarily rests upon measured concentrations of elements such as
Al, Ti, and Th. These refractory elements are good indicators of clastic detritus that may
contaminate pure marine precipitates. Data presented in the following chapters suggests
24

10 0
10 -1
Isua IF (3.8 Ga)
Kuruman IF (2.5 Ga)
10 -4
Campbellrand dolomite (2.5 Ga)
Holocene carbonate reef
La Ce Pr Nd PmSm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
sample/PAAS
10 -2
10 -3
seawater x10 4
Figure 12. Distributions of the REY in average world seawater (same as Fig. 4), and various chemical
sediments that are free from contamination with Al-rich clastic detritus. Data representing the Isua IF
(Greenland) are for the IF-G reference standard analyzed at Jacobs University Bremen as discussed in
Chapter 5. Kuruman (South Africa) IF data represent the average of two analyses as reported in Bau et al.
(1997a). Campbellrand dolomite represents unpublished data from GKP-01 drillcore (Agouron-
Griqualand Paleoproterozoic Drilling Project, administered by University of Johannesburg). Holocene
(≤10,000 years old) carbonate reef data from Webb and Kamber (2000). The REY patterns of IFs and
both ancient and modern carbonates are very similar to each other and modern seawater. These
observations suggest that Archean IFs possessing seawater-like REY distributions are likely to be
excellent archives for the Nd isotopic ratios present in contemporaneous seawater.
that, as a general rule, IF samples containing more than 0.5-0.7% Al 2 O 3 are unsuitable
for reconstructing seawater Nd isotopic ratios. Additionally, the distribution of the REY
is very useful at screening potential seawater archives, as Archean marine precipitates
frequently display REY patterns that are very similar to those observed in modern
seawater.
Figure 12 shows REY patterns observed in Archean IFs and both modern and
ancient marine carbonates. The distribution of the REY is generally indistinguishable
between modern seawater and carbonate reefs, as well as Archean dolomite and IFs.
The general enrichment of the HREE over the LREE and positive La and Y anomalies
found in seawater are common to all marine precipitates shown in Fig. 12 regardless of
age. The primary differences are for Ce, which reflects the different oxidation states of
25

modern and Archean oceans, as well as for Eu, which indicates a greater contribution to
Archean seawater from high-T hydrothermal fluids (e.g., Figs. 4,5). It should be noted
that the small negative Ce anomaly in the shallow water Holocene reef carbonates
relative to world seawater is not inconsistent, as very shallow seawater often displays
smaller negative Ce anomalies compared to bulk seawater (e.g., de Baar et al., 1983).
The fact that Archean IFs commonly have seawater-like REY distributions is not
consistent with REY scavenging by Fe-oxyhydroxides in a manner similar to that
observed in modern oceans. As shown in Fig. 8, adsorption-desorption reactions at
equilibrium strongly fractionate the REY between metal-oxides and seawater. If
Archean IFs precipitated from seawater as Fe(III) minerals, then their observed REY
distributions would require quantitative scavenging of all REY in solution, perhaps in a
manner analogous to the laboratory co-precipitation of REE-Fe-oxide phases from
seawater in early studies (e.g., Goldberg et al., 1963). However, the exact nature of the
primary IF precipitate remains unresolved, though data presented in Chapter 4 suggests
that initial deposition of Fe(II) mineral phases may have been important. Regardless, the
seawater-like REY distributions in Archean IFs, and the lack of other suitable archives,
indicates that IFs offer the best opportunity for discerning the Nd isotopic composition
of >2.7 Ga seawater.
26

References
Alibert, C., McCulloch, M.T., 1993. Rare earth element and Nd isotopic compositions
of the banded iron-formations and associated shales from Hamersley, Western
Australia. Geochim. Cosmochim. Acta 57, 187-204.
Alibo, D.S., Nozaki, Y., 1999. Rare earth elements in seawater: Particle association,
shale normalization, and Ce-oxidation. Geochim. Cosmochim. Acta 63, 363-372.
Anders, E., Grevesse, N., 1989. Abundances of the elements: meteoric and solar.
Geochim. Cosmochim. Acta 53, 197-214.
Bau M. (1999) Scavenging of dissolved yttrium and rare earths by precipitating iron
oxyhydroxide: Experimental evidence for Ce oxidation, Y-Ho fractionation, and
lanthanide tetrad effect. Geochim. Cosmochim. Acta, 63, 67–77.
Bau M. and Dulski P. (1999) Comparing yttrium and rare earths in hydrothermal fluids
from the Mid-Atlantic Ridge: implications for Y and REE behaviour during nearvent
mixing and for the YrHo ratio of Proterozoic seawater. Chem. Geol., 155,
77–90.
Bau M., Koschinsky A., Dulski P., and Hein J.R. (1996) Comparison of the partitioning
behaviours of yttrium, rare earth elements, and titanium between hydrogenetic
marine ferromanganese crusts and seawater. Geochim. Cosmochim. Acta, 60,
1709-1725.
Bau M., Höhndorf A., Dulski P., and Beukes, N.J. (1997a). Sources of rare-earth
elements and iron in Paleoproterozoic iron-formations from the Transvaal
Supergroup, South Africa: Evidence from neodymium isotopes. J. Geol. 105, 121-
129.
27

Bau M., Möller P., and Dulski P. (1997b) Yttrium and lanthanides in eastern
Mediterranean seawater and their fractionation during redox-cycling. Mar. Chem.
56, 123-131.
Bau M., Knappe A., Dulski P. (2006) Anthropogenic gadolinium as a micropollutant in
river waters in Pennsylvania and in Lake Erie, northeastern United States. Chemie
der Erde, (66), 143-152.
Germain Bayon G., German C.R., Burton K.W., Nesbitt R.W., Rogers N. (2004)
Sedimentary Fe–Mn oxyhydroxides as paleoceanographic archives and the role of
aeolian flux in regulating oceanic dissolved REE. Earth Planet. Sci. Lett., 224,
477-492.
Bennet V.C. (2003) Compositional evolution of the mantle, pp. 493-519. In The Mantle
and Core (ed. R.W. Carlson) Vol. 2 Treatise on Geochemistry (eds. H.D. Holland
and K.K. Turekian), Elsevier-Pergamon, Oxford.
Beukes, N.J., 1983. Palaeoenvironmental setting of iron-formations in the depositional
basin of the Transvaal Supergroup, South Africa, in: Trendall, A.F., Morris, R.C.
(Eds.), Iron-formation Facts and Problems, 131-198, Vol. 6 Developments in
Precambrian Geology (Ed. B.F. Windley), Elsevier, Amsterdam.
Brookins, D.G (1989) Aqueous geochemistry of rare earth elements, in: Lipin, B.R.,
McKay, G.A. (Eds.), Geochemistry and Mineralogy of Rare Earth Elements.
Reviews in Mineralogy 21, Mineral. Soc. Amer., 201-225.
Byrne R.H. and Cantrell K.J. (1987) Rare earth element complexation by carbonate and
oxalate ions. Geochim. Cosmochim. Acta, 51, 597-605.
Byrne R.H. and Kim K. (1990) Rare earth element scavenging in seawater. Geochim.
Cosmochim. Acta, 54, 2645-2656.
28

Byrne R.H. and Liu X. (1998) A coupled riverine-marine fractionation model for
dissolved rare earths and yttrium. Aquatic Geochem., 4, 103–121.
Chavagnac V. (2004) A geochemical and Nd isotopic study of Barberton komatiites
(South Africa): implication for the Archean mantle. Lithos, 75, 253– 281.
Cidu R. and Biddau R. (2007) Transport of trace elements under different seasonal
conditions: Effects on the quality of river water in a Mediterranean area. Applied
Geochemistry 22, 2777–2794.
Compston W. and Kröner A. (1988) Multiple zircon growth within early Archaean
tonalitic gneiss from the Ancient Gneiss Complex, Swaziland. Earth Planet. Sci.
Lett., 87, 13-28.
de Baar, H. J. W., M. P. Bacon and P. G. Brewer (1983) Rare earth distributions with a
positive cerium anomaly in the western North Atlantic Ocean. Nature, 301, 324-
327.
de Baar H.J.W., Bacon M.P., Brewer P.G. and Bruland K.W. (1985a). Rare earth
elements in the Atlantic and Pacific Oceans. Geochim. Cosmochim. Acta, 49,
1943-1959.
de Baar H.J.W., Brewer P.G., and Bacon M.P. (1985b) Anomalies in rare earth
distributions in seawater: Gd and Tb. Geochim. Cosmochim. Acta, 49, 1961-1969.
de Baar H.J.W., German C.R., Elderfield H., and van Gaans P. (1988) Rare earth
element distributions in anoxic waters of the Cariaco Trench. Geochim.
Cosmochim. Acta, 52, 1203-1219.
DePaolo, D.J., Wasserburg, G.J., 1976. Nd isotopic variations and petrogenetic models.
Geophys. Res. Lett. 3, 249-252.
Elderfield H. (1988) The oceanic chemistry of the rare-earth elements. Phil. Trans. R.
Soc. Lond., A325, 105-126.
29

Elderfield H. and Greaves M.J. (1982) The rare earth elements in seawater. Nature, 296,
214-219.
Elderfield, H. and E. R. Sholkovitz (1987). Rare earth elements in the pore waters of
reducing nearshore sediments. Earth Planet. Sci. Lett., 82, 280-290.
Elderfeld H., Upstil-Goddard R., and Sholkovitz E.R. (1990) The rare earth elements in
rivers, estuaries and coastal sea waters: processes affecting crustal input of
elements to the ocean and their significance to the composition of seawater.
Geochim. Cosmochim. Acta, 54, 971-991.
Erel Y. and Morgan J.J. (1991) The effect of surface reactions on the relative
abundances of trace metals in deep-ocean water. Geochim. Cosmochim. Acta, 55,
1807-1813.
Erel Y. and Stolper E.M. (1993) Modeling of rare-earth element partitioning between
particles and solution in aquatic environments. Geochim. Cosmochim. Acta, 57,
513-518.
Faure G. (1986) Principles of Isotope Geology. John Wiley and Sons, New York,
Gammons C.H. , Wood S.A.., Pedrozo F., Varekamp J. C., Nelson B.J., Shope C.L., and
Baffico G. (2005b) Hydrogeochemistry and rare earth element behavior in a
volcanically acidified watershed in Patagonia, Argentina. Chem. Geol., 222, 249-
267.
German, C. R. and H. Elderfield (1989). Rare earth elements in Saanich Inlet, British
Columbia, a seasonally anoxic basin. Geochim. Cosmochim. Acta, 53, 2561-2571.
German C. R., Klinkhammer G.P., Edmond J.M., Mitra A., and Elderfield H. (1990)
Hydrothermal scavenging of rare-earth elements in the ocean. Nature, 345, 516-
518.
30

German C. R., Campbell A.C., and Edmond J.M. (1991a) Hydrothermal scavenging at
the Mid-Atlantic Ridge: Modification of trace element dissolved fluxes. Earth
Planet. Sci. Lett., 107, 101-114.
German C.R., Holliday B.P., and Elderfeld H. (1991b) Redox cycling of rare earth elements
in the suboxic zone of the Black Sea. Geochim. Cosmochim. Acta, 55, 3553-3558.
German C.R., Masuzawa T., Greaves M.J., Elderfeld H. and Edmond J.M. (1995)
Dissolved rare earth elements in the Southern Ocean: cerium oxidation and the
influence of hydrography. Geochim. Cosmochim. Acta 59, 1551-1558.
Goldberg E.D., Koide M., Schmitt R.A., and Smith R.H. (1963) Rare-earth distributions
in the marine environment. J. Geophys. Res., 68, 4209-4217.
Goldstein S.L. and Jacobsen S.B. (1988) REE in the Great Whale River estuary,
northwest Quebec. Earth Planet. Sci. Lett., 88, 241-252.
Goldstein S.L. and Hemming S.R. (2003) Long-lived isotopic tracers in oceanography,
paleoceanography, and ice-sheet dynamics, pp. 453-489. In The Oceans and
Marine Geochemistry (ed. H. Elderfield) Vol. 6 Treatise on Geochemistry (eds.
H.D. Holland and K.K. Turekian), Elsevier-Pergamon, Oxford.
Goldstein S.L., O'Nions R.K., and Hamilton P.J. (1984) A Sm-Nd isotopic study of
atmospheric dusts and particulates from major river systems. Earth Planet. Sci.
Lett., 70, 221-236.
Greaves M.J., Elderfield H., and Sholkovitz E.R. (1999) Aeolian sources of rare earth
elements to the Western Pacific Ocean. Marine Chem., 68, 31-38.
Grotzinger, J.P., 1989. Facies and evolution of Precambrian carbonate depositional
systems: emergence of the modern platform archetype. controls on carbonate
platform and basin development, SEPM (Soc. Econ. Paleont. Mineral.) Spec. Pub.
44. 79-106.
31

Haley B.A., Klinkhammer G.P. (2003) Complete separation of rare earth elements from
small volume seawater samples by automated ion chromatography: method
development and application to benthic flux. Mar. Chem., 82, 197–220.
Haley B.A., Klinkhammer G.P. and McManus J. (2004) Rare earth elements in pore
waters of marine sediments. Geochim. Cosmochim. Acta, 68, 1265–1279.
Hegner E., Kroner A., and Hofmann A.W. (1984) Age and isotope geochemistry of the
Archaean Pongola and Usushwana suites in Swaziland, southern Africa: a case for
crustal contamination of mantle-derived magma. Earth Planet. Sci. Lett., 70, 267-
279.
Høgdahl O.T., Melsom S., and Bowen V.T. (1968) Neutron activation analysis of
lanthanide elements in sea water. in Trace Inorganics in Water, pp. 308-325,
Advan. Chem. Series 73.
Holland H.D. (1972) The geologic history of seawater – an attempt to solve the
problem. Geochim. Cosmochim. Acta, 36, 637-651.
Hoyle J., Elderfield H., Gledhill A., and Greaves M. (1984) The behaviour of the rare
earth elements during mixing of river and sea waters. Geochim. Cosmochim.
Acta, 48, 143-149.
Jacobsen, S.B., Pimentel-Klose, M.R., 1988a. A Nd isotopic study of the Hamersley and
Michipicoten banded iron formations: the source of REE and Fe in Archean
oceans. Earth Planet. Sci. Lett. 87, 29-44.
Jacobsen, S.B., Pimentel-Klose, M.R., 1988b. Nd isotopic variations in Precambrian
banded iron formations. Geophys. Res. Lett. 15, 393-396.
James H.L. (1954) Sedimentary facies of iron-formation. Econ. Geol., 49, 235-293.
32

Jeandel, C., Bishop, K., Zindler, A., 1995. Exchange of neodymium and its isotopes
between seawater and small and large particles in the Sargasso Sea. Geochim.
Cosmochim. Acta 59, 535-547.
Klein, C., Beukes, N.J., 1989. Geochemistry and sedimentology of a facies transition
from limestone to iron-formation in the Early Proterozoic Transvaal Supergroup,
South Africa. Econ. Geol. 84, 1733-1774.
Koeppenkastrop D. and De Carlo E.H. (1992) Sorption of rare-earth elements from
seawater onto synthetic mineral particles: An experimental approach. Chem.
Geol., 95, 251-263.
Koeppenkastrop D. and De Carlo E.H. (1993) Uptake of rare earth elements from
solution by metal oxides. Environ. Sci. Technol., 27, 1796-1802.
Kröner A., and Tegtmeyer A. (1994) Gneiss-greenstone relationships in the Ancient
Gneiss Complex of southwestern Swaziland, southern Africa, and implications for
early crustal evolution. Precam. Res., 67, 109-139.
Kröner A., Hegner E., Wendt J.I., and Byerly G.R. (1996) The oldest part of the
Barberton granitoid-greenstone terrain, South Africa: evidence for crust formation
between 3.5 and 3.7 Ga. Precam. Res. 78, 105-124.
Kulaksız S. and Bau M. (2007) Contrasting behaviour of anthropogenic gadolinium and
natural rare earth elements in estuaries and the gadolinium input into the North
Sea. Earth Planet. Sci. Lett., 260, 361-371.
Lawrence M.G., and Kamber B.S. (2006) The behaviour of the rare earth elements
during estuarine mixing-revisited. Marine Chemistry, 100, 147-161.
Lawrence M.G., Greig A., Collerson K.D., and Kamber B.S. (2006) Rare Earth Element
and Yttrium Variability in South East Queensland Waterways. Aqua. Geochem.,
12, 39–72.
33

Lee J.H. and Byrne R.H. (1993) Complexation of trivalent rare earth elements (Ce, Eu,
Gd, Tb, Yb) by carbonate ions. Geochim. Cosmochim. Acta, 57, 295-302.
Liu X. and Byrne R.H. (1998) Comprehensive investigation of yttrium and rare earth
element complexation by carbonate ions using ICP-mass spectrometry. J. Solution
Chem., 27, 803-815.
Lowe D.R. (1994) Accretionary history of the Archean Barberton Greenstone Belt
(3.55-3.22 Ga), southern Africa. Geology, 200, 1099-1102.
Luo Y. and Byrne R.H. (2004) Carbonate complexation of yttrium and the rare earth
elements in natural waters. Geochim. Cosmochim. Acta, 68, 691–699.
Martin J.M., Hogdahl 0., and Phillppot L.C. (1976) Rare earh element supply to the
oceans. J. Geophys. Res., 81, 3119-3124.
McLennan, S.M., 1989. Rare earth elements in sedimentary rocks: influence of
provenance and sedimentary processes, in: Lipin, B.R., McKay, G.A. (Eds.),
Geochemistry and Mineralogy of Rare Earth Elements. Reviews in Mineralogy
21, Mineral. Soc. Amer., 169-200.
Miller, R.G., O'Nions, R.K., 1985. Sources of Precambrian chemical and clastic
sediments. Nature 314, 325-330.
Mills R.A., Wells D.M., Roberts S. (2001) Genesis of ferromanganese crusts from the
TAG hydrothermal field. Chem. Geol., 176, 283-293.
Nägler T.F and Kramers J.D. (1998) Nd isotopic evolution of the upper mantle during
the Precambrian: models, data and the uncertainty of both. Precam. Res., 91, 233-
252.
Nakai S., Halliday A.N., and Rea D.K. (1993) Provenance of dust in the Pacific Ocean.
Earth Planet. Sci. Lett., 119, 143-157.
34

Nelson D.R., Trendall A.F., de Laeter J.R., Grobler N.J., Fletcher I.R. (1992) A
comparative study of the geochemical and isotopic systematics of late Archaean
flood basalts from the Pilbara and Kaapvaal Cratons. Precam. Res., 54, 231-256.
Niu, Y., Collerson, K., Batiza, R., Wendt, J. and Regelous, M. (1999). Origin of
enriched-type mid-ocean ridge basalt at ridges far from mantle plumes: The East
Pacific Rise at 11°20'N. J. Geophys. Res., 104(B4): doi: 10.1029/1998JB900037.
Nozaki, Y., Alibo, D.S., Amakawa, H., Gamo, T., Hasumoto, H., 1999. Dissolved rare
earth elements and hydrography in the Sulu Sea. Geochim. Cosmochim. Acta 63,
2171–2181.
Nozaki Y., Lerche D., Alibo D.S., and Tsutsumi M. (2000a) Dissolved indium and rare
earth elements in three Japanese rivers and Tokyo Bay: Evidence for
anthropogenic Gd and In. Geochim. Cosmochim. Acta, 64, 3975–3982.
Nozaki Y., Lerche D., Alibo D.S., and Snidvongs A. (2000b) The estuarine
geochemistry of rare earth elements and indium in the Chao Phraya River,
Thailand. Geochim. Cosmochim. Acta, 64, 3983–3994.
Nozaki, Y., Alibo, D.S., 2003. Importance of vertical geochemical processes in
controlling the oceanic profiles of dissolved rare earth elements in the
northeastern Indian Ocean. Earth Planet. Sci. Lett., 205, 155-172.
Patchett P.J. and Samson S.D. (2003) Ages and growth of the continental crust from
radiogenic isotopes, pp. 321-348. In The Crust (ed. R.L. Rudnick) Vol. 3 Treatise
on Geochemistry (eds. H.D. Holland and K.K. Turekian), Elsevier-Pergamon,
Oxford.
Piasecki W. And Sverjensky D.A. (2008) Speciation of adsorbed yttrium and rare earth
elements on oxide surfaces. Geochim. Cosmochim. Acta, 72, 3964–3979.
35

Piepgras D.J. and Wasserburg G.J. (1980) Neodymium isotopic variations in seawater.
Earth Planet. Sci. Lett., 50, 128-138.
Poujol M., Robb L.J., Anhaeusser C.R., and Gericke B. (2002) Geochronological
constraints on the evolution of the Kaapvaal craton, South Africa. Econ. Geol.
Res. Institute, Information Circular No. 360, Univ. Witwatersrand, pp. 37.
Shannon R.D. (1976) Revised effective ionic radii and systematic studies of interatomic
distances in halides and chalcogenides. Acta Crystallograph., A32, 751-767.
Salters V., and Stracke A. (2004) Composition of the depleted mantle. Geochemistry,
Geophysics, Geosystems, 5, doi: 10.1029/2003GC000597.
Sholkovitz E.R. (1993) The geochemistry of rare earth elements in the Amazon River
Estuary. Geochim. Cosmochim. Acta, 57, 2181-2190.
Sholkovitz E.R. (1995) The aquatic chemistry of rare earh elements in rivers and
estuaries. Aquatic Geochem. 1, 1-34.
Sholkovitz E.R. (1996) A compilation of the rare earth element composition of rivers,
estuares and the oceans. Woods Hole Oceanog. Inst. Tech. Rept., WHOI-96-13.
pp. 78.
Sholkovitz E.R. and Elderfeld H. (1988) The cycling of dissolved rare earth elements in
Chesapeake Bay. Global Biogeochem. Cycles, 2, 157-176.
Sholkovitz E.R. and Szymczak R. (2000) The estuarine chemistry of rare earth
elements: comparison of the Amazon, Fly, Sepik and the Gulf of Papua systems.
Earth Planet. Sci. Lett., 179, 299-309.
Sholkovitz, E. R, D. J. Piepgras and S. B. Jacobsen (1989). The pore water chemistry of
rare earth elements in Buzzards Bay sediments. Geochim. Cosmochim. Acta, 53,
2847-2856.
36

Sholkovitz E.R., Shaw T.J., and Schneider D.L. (1992) The geochemistry of rare earth
elements in the seasonally anoxic water column and pore waters of Chesapeake
Bay. Geochim. Cosmochim. Acta, 56, 3389-3402.
Sholkovitz E.R., Landing W.M., and Lewis B.L. (1994) Ocean particle chemistry: The
fractionation of rare earth elements between suspended particles and seawater.
Geochim. Cosmochim. Acta, 58, 1567-1579.
Simonson, B.M., 2003. Origin and evolution of large Precambrian iron formations.
Geol. Soc. Amer. Spec. Paper 370, 231-243.
Simonson, B.M., Hassler, S.W., 1996. Was the deposition of large Precambrian iron
formations linked to major marine transgressions?. J. Geol. 104, 665-676.
Stille P., Steinmann M., Pierret M.-C., Gauthier-Lafaye F., Chabaux F., Viville D.,
Pourcelot L., Matera V., Aouad G., and Aubert D. (2006) The impact of
vegetation on REE fractionation in stream waters of a small forested catchment
(the Strengbach case). Geochimica et Cosmochimica Acta 70, 3217–3230.
Tachikawa, K., Jeandel C., and Roy-Barman M. (1999) A new approach to the Nd
residence time in the ocean: the role of atmospheric inputs. Earth Planet. Sci. Lett.,
170, 433-446.
Tachikawa, K., Athias, V., Jeandel, C., 2003. Neodymium budget in the modern ocean
and paleo-oceanographic implications. J. Geophys. Res. 108, 1-13.
Tankard A. J., Jackson M. P. A., Eriksson K. A., Hobday D. K., Hunter D. R., and
Minter W. E. L. (1982) Crustal Evolution of Southern Africa. Springer Verlag,
New York.
Taylor S.R. and McLennan S.M. (1995) The geochemical evolution of the continental
crust. Reviews of Geophysics, 33, 241-265.
37

Tosiani T., Loubet M., Viers J., Valladon M., Tapia J., S. Marrero S., Yanes C.,
Ramirez A., Dupre B. (2004) Major and trace elements in river-borne materials
from the Cuyuni basin (southern Venezuela): evidence for organo-colloidal
control on the dissolved load and element redistribution between the suspended
and dissolved load. Chem. Geol., 211, 305–334.
Tricca A., Stille P., Steinmann M.,Kiefel B., Samuel J., and Eikenberg J. (1999) Rare
earth elements and Sr and Nd isotopic compositions of dissolved and suspended
loads from small river systems in the Vosges mountains France/, the river Rhine
and groundwater. Chemical Geology 160, 139–158.
Veizer, J., Hoefs, J., Lowe, D.R., Thurston, P.C. 1989. Geochemistry of Precambrian
carbonates: II. Archean greenstone belts and Archean sea water. Geochim.
Cosmochim. Acta, 53, 859-871.
Wilson A.H., and Carlson R.W. (1989) A Sm-Nd and Pb isotope study of Archaean
greenstone belts in the southern Kaapvaal Craton, South Africa. Earth Planet. Sci.
Lett., 96, 89-105.
Zhang, J., Nozaki, Y., 1996. Rare earth elements and yttrium in seawater: ICP-MS
determinations in the East Caroline, Coral Sea, and South Fiji basins of the
western South Pacific Ocean. Geochim. Cosmochim. Acta 60, 4631-4644.
38

Chapter 2. Trace element analyses in geological materials using low resolution
inductively coupled plasma mass spectrometry (ICPMS)
39

THIS PAGE INTENTIONALLY LEFT BLANK
40

Brian W. Alexander
Trace element analyses in geological materials
using low resolution inductively coupled plasma
mass spectrometry (ICPMS)
Technical Report No. 18
August 2008
School of Engineering and Science

Abstract
The benefits of inductively coupled plasma mass spectrometry (ICPMS) for
geochemical studies include excellent instrument sensitivity for determining a large
number of elements (10 mg/kg). However, some
elements suffer from interferences due to the major element content of particular rock
types, and are unlikely to be quantifiable when present at low concentrations. This
includes Co in carbonate rocks, and Nb in Fe-rich rock types. The concentrations of
many trace metals in some CRMs are poorly constrained, and early published datasets
appear to frequently overestimate the abundances of these elements.
ii

Table of Contents
List of Figures .......................................................................................................................... iv
List of Tables............................................................................................................................. v
1. Introduction .......................................................................................................................... 1
2. Sample preparation............................................................................................................... 2
3. Sample decomposition ......................................................................................................... 2
4. ICPMS methods ................................................................................................................... 5
4.1. Analyzed elements and mass interferences .................................................................. 5
4.2. External calibration ...................................................................................................... 9
4.3. Data collection............................................................................................................ 13
4.4. Internal standardization .............................................................................................. 14
5. Analytical precision............................................................................................................ 16
6. Analytical recovery ............................................................................................................ 24
7. Analytical accuracy ............................................................................................................ 27
7.1. Reference values and major element interferences .................................................... 27
7.2. High Fe content rocks................................................................................................. 33
7.3. Shales and clastic sediments....................................................................................... 39
7.4. High Si content rocks (cherts) .................................................................................... 42
7.5. Carbonate rocks.......................................................................................................... 44
7.6. Marine ferromanganese nodules and crusts................................................................ 47
7.7. Basalts ........................................................................................................................ 49
7.8. Rare earth element ratios............................................................................................ 51
8. Summary and conclusions.................................................................................................. 53
References ............................................................................................................................... 56
Appendix 1. Analytical data .................................................................................................... 59
Appendix 2. Interferences due to major elements ................................................................... 64
iii

List of Figures
Figure 1. Sample decomposition methods………………………………………………….......3
Figure 2. Preparation of multi-element standards…………..………………………………... 10
Figure 3. Solutions used and analysis order for ICPMS measurements……............................12
Figure 4. ICPMS instrument quantification limits……………...……………………………. 17
Figure 5. Characterization of measures of analytical precision……………………………… 19
Figure 6. Average sample precision for different reference materials……...….…….............. 20
Figure 7. Average run precision for different reference materials…..………………….......... 21
Figure 8. Comparison of different measures of precision…………………............................. 22
Figure 9. Sample and method precision for HNO 3 carbonate decomposition……………….. 24
Figure 10. Analytical recovery for HNO 3 carbonate and HF-HClO 4 decompositions….......... 26
Figure 11. Mg and Ca interferences on 59 Co………………………….…………………….... 30
Figure 12. Analytical results for iron-formation FeR-2…………………………………….... 34
Figure 13. Analytical results for iron-formation FeR-4……………………………….……... 35
Figure 14. REY data for second lot of iron-formation IF-G………………………………..... 37
Figure 15. Analytical results for iron-formation IF-G………………………...……………....38
Figure 16. Analytical results for shale SCo-1………………………………………………... 40
Figure 17. Analytical results for shale SGR-1b……………………………….………............41
Figure 18. Analytical results for chert JCh-1…………………………………….…………... 43
Figure 19. Analytical results for dolomite JDo-1…….………………………………….........45
Figure 20. Comparison of JDo-1 HNO 3 carbonate and HF-HClO 4 decompositions………….46
Figure 21. Analytical results for Fe-Mn nodule JMn-1……………………………….............49
Figure 22. Analytical results for basalt BHVO-2……………………………………..……....50
iv

List of Tables
Table 1. List of certified reference materials (CRMs)…………………………………….........4
Table 2. Isotopes monitored in ICPMS measurements and IQLs….……………………….... ..6
Table 3. Long term stability of HFSE in iron-formation IF-G….............................................. 13
Table 4. Major element interferences for low mass elements (

1. Introduction
The advent of inductively coupled plasma mass spectrometry (ICPMS) has
proven enormously beneficial to geochemical studies. Since commercialization of
ICPMS technology in the early 1980’s, ICPMS has become the method of choice for
geochemical analyses, and is ideal for trace metal determinations and isotopic studies
of many elements. The widespread implementation of ICPMS technology results from
its ability to quickly quantify a large number of elements (>40) in geological samples
on a routine basis. Additionally, ICPMS offers instrument sensitivity that permits
quantification of ng/kg (parts-per-trillion, ppt) concentrations of many elements
regardless of sample matrix, and as such is ideally suited for studies of geochemically
diverse samples. This report describes the ICPMS analytical methods employed
within the Geochemistry Lab at Jacobs University Bremen (JUB) and includes a
critical discussion of the precision and accuracy of these methods.
The Geochemistry Lab at JUB is relatively new, with construction of the lab
completed in the Summer of 2004. In the Fall of 2004 a PerkinElmer DRC-e
quadrupole inductively coupled plasma mass spectrometer (ICPMS) was installed,
which is the principal analytical instrument for determinations of trace metal
concentrations. This report is not intended to provide detailed information regarding
basic ICPMS principles and technology, and the reader is referred to Thomas (2003)
for a thorough review. When samples are decomposed into liquid form, ICP
instruments are ideal for multi-element geochemical analyses, as they allow relatively
fast sample introduction into mass spectrometers, and are readily automated for the
processing of large numbers of samples.
This report describes the steps utilized at JUB to obtain accurate trace metal
concentration data in a variety of rock types, and proceeds from the processing of
hand samples to the decomposition of sample powders, and finally to the methods
employed to ensure the accuracy of the reported data. Sample preparation techniques
are not the primary focus of this paper, and are only briefly discussed. Rather,
considering the relatively short existence of the Geochemistry Lab at JUB and the
necessary development of accurate and routine geochemical analytical techniques,
most of the discussion will focus on a critical assessment of the methods employed in
ICPMS trace metal determinations in a variety of geologic materials, and the relative
precision and accuracy of these measurements.
1

2. Sample preparation
The majority of this report focuses on geochemical analyses of rock reference
standards issued by governmental or research organizations. However, a brief
description is given of the sample preparation methods used for natural rock samples
that are collected as part of research studies at JUB. Rock samples, whether from
outcrops or drillcores, are first processed by hand-trimming to avoid weathering rinds,
quartz/calcite veining, or obvious alteration features. A geologist’s hammer is used to
produce roughly 50 g of small, 1-3 cm pieces that are crushed in a Fritsch Pulverisette
1 jaw crusher to obtain chips approximately 0.5 cm in size. Harder rock types (e.g.,
chert) are processed more finely to produce smaller sizes (≤0.3 cm chips). Sample
chips are rinsed thoroughly with deionized water to remove dust and dried overnight
in a laboratory oven at ~110 °C, after which approximately 20 g of chips are handpicked
to produce homogeneous samples devoid of secondary mineral veins. The
sample chips are then powdered in a Fritsch Pulverisette 6 planetary mill using agate
balls in a sealed agate mortar.
3. Sample decomposition
The ICPMS used in the JUB Geochemistry Lab is configured to accept
samples in liquid form only, though ICP instruments with laser-ablation attachments
for the analyses of solid samples are increasingly common. One advantage of
converting samples to liquid form is that any effects due to heterogeneities within the
sample (e.g., inclusions or minor mineral phases) are removed, and this approach is
typically termed a ‘whole-rock’ analysis. The common method for rocks and minerals
is to decompose the sample in strong mineral acids at elevated temperatures and
pressures. The decomposition procedures used at JUB, and in fact, many of the
ICPMS techniques as well, have been adapted from methods developed by P. Dulski
at the GeoForschungsZentrum (GFZ) in Potsdam, Germany (see Dulski, 1994;
Dulski, 2001).
Two sample decomposition methods are currently employed: the first is a
high-temperature, high-pressure decomposition utilizing concentrated hydrofluoric
(HF) and perchloric (HClO 4 ) acids that is suitable for the total dissolution of silicatebearing
samples (e.g., iron-formations, basalts), while the second method is a lowtemperature
nitric acid (HNO 3 ) decomposition used for dissolving carbonate samples
such as limestones and dolomites. It is important to note that the HF-HClO 4
2

Figure 1. Flow-chart diagram of the two sample decomposition methods employed within the
Geochemistry Lab at JUB for silicate-bearing samples (HF-HClO 4 pressure decomposition) and
carbonate samples such as limestones and dolomites (HNO 3 carbonate decomposition). All reagents
used are super-, ultra-pure grade and dilutions are performed with deionized water (18.2 MΩ). Modified
after Dulski (2001).
decomposition provides complete dissolution of the sample powder, primarily due to
the ability of HF to dissolve silicate minerals, whereas the low-temperature HNO 3
decomposition is considered to dissolve only the carbonate mineral fraction of the
sample powder. The HNO 3 decomposition method, referred to as the ‘carbonate
decomposition’, will not dissolve refractory organic carbon and/or primary silicate
minerals. The carbonate decomposition will attack and leach secondary minerals such
as clays, though the extent of this effect will vary as a function of sample
composition. The two methods are schematically described in Figure 1, and most of
the data and discussion presented here will focus on the HF-HClO 4 decomposition
method. Regardless of the method chosen, every sample decomposition contains at
least one certified reference material (CRM) which is treated analytically as an
unknown sample, and which is considered to provide the best estimate of the accuracy
of the complete decomposition method and ICPMS analysis. The CRMs commonly
used in the Geochemistry Lab are listed in Table 1 and represent a variety of rock
types, and for any individual sample decomposition a CRM is chosen that most
closely resembles the rock type of the samples.
3

Table 1. Certified reference materials (CRM) used as quality assurance standards for routine sample
decomposition and ICPMS measurements within the JUB Geochemistry Lab.
number of sample
CRM
description decompositions
issuing organization
FeR-2 Al-rich iron-formation 3 CCRMP Canadian Certified Reference Materials
Project
FeR-4 iron-formation 2 555 Booth Street Ottawa, Ontario K1A 0G1 Canada
IWG-GIT International Working Group - Groupe
2 first lot International de Travail
IF-G a Isua iron-formation
Service d’Analyse des Roches et des Minéraux,
4 second lot
CPRG-CNRS,
B.P. 20 Vandoeuvre-lés-Nancy Cedex, France
SCo-1 silty marine shale 4
USGS United States Geological Survey
SGR-1b carbon-rich shale 3 U.S. Geological Survey, Box 25046, MS 973
BHVO-2 Hawaiian basalt 7
Denver, CO 80225, USA
JMn-1 manganese nodule 4
JCh-1 chert 3
JDo-1
dolomite
12
(5 HF-HClO 4 ) b
(7 HNO 3 ) b
GSJ Geological Survey of Japan
available from: Seishin Trading Co., 1-2-1 Sanshin
Bulding, Sannomiya, Tyuo-ku, Kobe, 650-0021,
Japan
a original IF-G powder (first lot) no longer available, and newly processed IF-G powder (second lot) available 2006.
b refers to decomposition method used.
Sample decompositions were performed in 30 ml polytetrafluoroethylene
(PTFE) vessels using a PicoTrace DAS acid digestion system (Bovenden, Germany).
As the DAS system contains 16 PTFE vessels, routine sample decompositions consist
of 14 samples, one CRM, and one method blank. Typically, these decompositions will
convert 100 mg of sample powder into 50 g of a clear sample liquid, in a matrix of
either 0.5 molar (moles/l, M) HCl or 0.5 M HNO 3 . This dilution factor of 500
produces a total dissolved solid (TDS) content of 0.2% in the sample liquid (assuming
no loss of volatiles such SiF 4 or CO 2 ), which is generally considered the maximum
TDS content practical for routine ICPMS measurements. The ideal dilution factor for
a particular rock or mineral reflects the balance between two competing effects; the
minimization of matrix effects in the sample by utilizing high dilution factors, versus
the detection and quantification limits imposed by the method and instrumentation.
Typical dilution factors employed within the JUB Geochemistry Lab for ICPMS
measurements range from 250 for very pure, trace metal-poor limestones and cherts,
to ≥5000 for trace metal-rich samples such as marine ferromanganese crusts and
nodules. The use of dilution factors

during decomposition (SiF 4 and CO 2 , respectively), thereby reducing the total
dissolved solid content of the ICPMS sample solution.
Cleaning of the DAS system between sample batches is accomplished by
heating an acid mixture of 5 ml of 15 M HNO 3 and 1 ml 23 M HF in the sealed PTFE
vessels for 12 hours at ~165 °C. Memory effects within the DAS system for the
elements considered in this research have never been observed, and decomposition
method blanks for these elements are essentially controlled by good laboratory
practices and the purity of the reagents used in the decomposition.
4. ICPMS methods
4.1. Analyzed elements and mass interferences
Currently, the concentrations of 32 elements are determined during ICPMS
analyses conducted within the JUB Geochemistry Lab. Initially, and until June 2006,
concentrations of only 24 of these 32 elements were determined during routine
ICPMS analyses, and the current element list and the scanned isotopic masses are
presented in Table 2. Wherever possible, multiple isotopes are monitored for any
given element, as the well-known abundances of natural isotopes provide a method
for checking the quality of the analysis. This is accomplished by normalizing the
signal intensity in cps (counts-per-second) for an isotope by its abundance. If no
isotopic fractionation has occurred in the sample, and the measured intensities are not
affected by interferences (e.g., interferences from isotopes of other elements), then
these normalized intensities should be equivalent for all isotopes of an element.
For example, the rare earth element (REE) Yb is monitored at three isotopic
masses, 171 Yb, 172 Yb, and 174 Yb, which have natural isotopic abundances of 14.3%,
21.9%, and 31.8%, respectively. If the following holds true:
⎛
⎜
⎝
171
Yb
cps
0.143
⎞
⎟
⎠
=
⎛
⎜
⎝
172
Yb
0.219
cps
⎞
⎟
⎠
=
⎛
⎜
⎝
174
Yb
cps
0.318
⎞
⎟
⎠
(1)
then these isotopes of are suitable for quantifying the concentration of Yb in the
analyzed solution. The approach described by Eq. 1 is not suitable if some process,
whether natural or analytical, has fractionated the monitored isotopes for a given
element. Of the 32 elements quantified by ICPMS measurement, only Pb may be
fractionated in such a manner due to natural processes occurring within the sample.
5

Table 2. Elements analyzed in routine ICPMS determinations, specific isotopes measured, interfering
polyatomic species, and instrument quantification limits (IQLs) in HCl and HNO 3 acid matrices. See text for
details. The eight elements in bold type were added to the routine ICPMS analysis in June 2006.
average instrument quantification limit
element
1 Sc
monitored
masses
45 Sc
12 C 16 O 2 1 H + , 13 C 16 O 2
+
observed/corrected interferences
in rock/mineral samples (mg/kg)
0.5 M HCl matrix
(n = 22)*
0.5 M HNO 3 matrix
(n=4)*
0.758 0.233
2 Ti
3 Co
4 Ni
5 Rb
6 Sr
7 Y
8 Zr
9 Nb
10 Mo
47 Ti, 49 Ti
59 Co
60 Ni, 62 Ni
85 Rb
88 Sr
89 Y
90 Zr, 91 Zr
93 Nb
95 Mo, 97 Mo
14 N 16 O 1 2 H + , 12 C 35 Cl + , 14 N 35 Cl + 1.94 0.284
42 Ca 16 O 1 H + , 43 Ca 16 O + , 24 Mg 35 Cl + 0.009 0.020
14 N 16 2 O + 2 , 44 Ca 16 O + , 25 Mg 35 Cl + , 25 Mg 37 Cl + , 27 Al 35 Cl + 0.077 0.168
48 Ca 37 Cl + 0.019 0.009
48 Ca 40 Ar + 0.047 0.280
54 Fe 35 Cl + 0.003 0.004
55 Mn 35 Cl + , 40 Ar 16 O 35 Cl + , 40 Ca 16 O 35 Cl + , 56 Fe 35 Cl + 0.018 0.012
40 Ar 16 O 37 Cl + , 40 Ca 16 O 37 Cl + , 56 Fe 37 Cl + 0.014 0.015
55 Mn 40 Ar + , 44 Ca 16 O 35 Cl + , 44 Ca 16 O 37 Cl + 0.069 0.076
11 Cs
12 Ba
13 La
14 Ce
15 Pr
16 Nd
17 Sm
133 Cs 0.004 0.004
135 Ba, 137 Ba 0.017 0.055
139 La 0.003 0.006
140 Ce 0.003 0.005
141 Pr 0.002 0.004
145 Nd, 146 Nd 0.005 0.007
147 Sm, 149 Sm 0.005 0.005
18 Eu
19 Gd
20 Tb
21 Dy
22 Ho
23 Er
24 Tm
25 Yb
26 Lu
27 Hf
28 Ta
29 W
151 Eu, 153 Eu
158 Gd, 160 Gd
159 Tb
161 Dy, 163 Dy
165 Ho
166 Er, 167 Er
169 Tm
171 Yb, 172 Yb, 174 Yb
175 Lu
178 Hf, 179 Hf
181 Ta
182 W, 183 W
135 Ba 16 O, 134 Ba 17 (OH), 137 Ba 16 O, 136 Ba 17 (OH) 0.002 0.004
142 Nd 16 O, 142 Ce 16 O, 141 Pr 17 (OH), 144 Nd 16 O, 144 Sm 16 O, 160 Dy,
Pr 19 (OH), 143 Nd 17 (OH)
0.003 0.005
141 Pr 18 O, 143 Nd 16 O, 142 Nd 17 (OH), 142 Ce 17 (OH) 0.002 0.004
145 Nd 16 O, 144 Nd 17 (OH), 144 Sm 17 (OH), 147 Sm 16 O, 146 Nd 17 (OH) 0.003 0.004
149 Sm 16 O, 148 Nd 17 (OH), 148 Sm 17 (OH) 0.002 0.004
150 Nd 16 O, 150 Sm 16 O, 149 Sm 17 (OH), 151 Eu 16 O, 150 Nd 17 (OH),
Sm 17 (OH)
0.002 0.006
153 Eu 16 O, 152 Sm 17 (OH), 134 Ba 35 Cl 0.002 0.003
155 Gd 16 O, 154 Gd 17 (OH), 154 Sm 17 (OH), 134,136 Ba 37,35 Cl, 135,137 Ba 37,35 Cl,
Gd 16 O, 155 Gd 17 (OH), 158 Gd 16 O, 157 Gd 17 (OH), 137 Ba 37 Cl
0.002 0.005
158 Gd 17 (OH), 159 Tb 16 O, 138 Ba 37 Cl 0.002 0.004
162 Dy 16 O, 161 Dy 17 (OH), 163 Dy 16 O, 162 Dy 17 (OH) 0.004 0.005
165 Ho 16 O 0.008 0.011
166 Er 16 O, 167 Er 16 O, 166 Er 17 (OH) 0.196 0.067
30 Pb
31 Th
206 Pb, 207 Pb, 208 Pb 0.022 0.035
232 Th 0.003 0.005
32 U
238 U 0.002 0.004
* n equals the number of separate ICPMS runs which contained at least one Certified Reference Material (CRM), with each run consisting
of a separate acid decomposition of a mixed batch of samples and CRMs, and from which the average instrument quantification limits (IQL)
for solid samples were calculated. See text for details regarding the calculation of detection and quantification limits.
6

Fractionation between the three isotopes of Pb monitored during ICPMS
analyses, 206 Pb, 207 Pb, and 208 Pb, will occur due to radioactive decay of U and Th
within the sample. The half-lives of the respective decay series, 238 U → 206 Pb, 235 U →
207 Pb, and 232 Th → 208 Pb, range between ~700 million to ~14 billion years (Faure,
1986). Therefore, changes in the natural modern abundances of the Pb isotopes ( 206 Pb
= 24.1%, 207 Pb = 22.1%, 208 Pb = 52.4%), due to the generation of radiogenic Pb
within the sample may be significant, particularly in the geologically old (>2.5 Ga)
samples commonly analyzed at JUB. Therefore, only in the case of Pb is Eq. 1 not
used to monitor agreement between the concentrations determined from individual
isotopes for a given element. For the remaining elements with more than one atomic
mass available for monitoring, the application of Eq. 1 also provides a method for
identifying interferences that may be affecting isotopes of a particular element.
The issue of ions with masses similar to the isotopes of interest producing
undesirable interferences in ICPMS analyses is common. The elements in geological
materials most suitable for routine ICPMS analyses have masses greater than 80
atomic mass units (amu). This primarily results from the fact that lower mass
elements (e.g., many transition metals) may suffer severe interferences from
polyatomic species generated within the plasma during ionization of the sample.
Many of these polyatomic interferences result from the use of argon as the plasma
source gas in most ICPMS instruments, and the subsequent formation of Ar-oxides
and Ar-hydroxides (ArO + and ArOH + ), though other interfering species may be
significant due to the choice of acid used for decomposing and diluting samples (e.g.,
chloride and nitrogen species).
A common example of the difficulties such polyatomic interferences present
is evident in determinations of Fe, in which the isotope of choice for analysis is the
most abundant one, 56 Fe (91.72%). However, within the plasma large numbers of
40 Ar 16 O + ions are produced which are indistinguishable from 56 Fe in low-resolution
ICPMS analyses. High-resolution ICPMS instruments are capable of distinguishing
between the 40 Ar 16 O + and 56 Fe peaks, but can only achieve this peak resolution at the
expense of ion-beam transmission efficiency and a consequent reduction in
instrument sensitivity. As a result high-resolution ICPMS methods are generally not
suitable for element determinations in the sub-ppb (parts-per-billion, μg/kg) range
necessary for analyses of many trace metals in geological samples, unless intensive
7

sample preparation methods are employed which isolate and concentrate the elements
of interest from undesirable matrix elements.
Table 2 contains interferences commonly observed in ICPMS analyses
performed within the JUB Geochemistry Lab. The inability of the low-resolution
DRC-e ICPMS to resolve interferences from the isotopes of interest does not
significantly inhibit the accurate determination of the 32 elements routinely analyzed,
as correcting for these interferences is possible through careful characterization and
quantification of the interfering polyatomic species. The majority of these corrected
interferences are for the REE (Table 2), as the REE can form significant amounts of
oxide and hydroxide species during ionization. For example, at typical instrument
settings as much as 3% of the Ce present in the sample solution will be ionized to
140 Ce 16 O + , and the
140 Ce 16 O + ion will consequently contribute to any Gd
determinations that utilize the
156 Gd isotope. The isotopes monitored for
quantification of the REE are therefore carefully selected to minimize these effects
and maximize analytical accuracy, though as seen in Table 2 a significant number of
corrections are necessary. While implementing such corrections is not trivial,
uncorrected values may result in reported concentrations for REE elements that are
erroneously high by as much as several percent, as illustrated by the example with
CeO + .
Corrections for polyatomic interferences are performed mathematically,
typically offline using commercially available spreadsheet software such as Microsoft
Excel (the approach followed at JUB). Interferences are identified and quantified by
analyzing solutions that contain relatively high concentrations (50-1000 μg/kg) of a
single element, and surveying the range of masses that might be affected by
interferences produced by oxide, hydroxide, chloride, or nitrogen species of that
particular element. For the REE, the most significant interferences are generated by
REE-oxides or -hydroxides, and therefore are found at 16 and 17 amu above the
interfering element (i.e., REE 16 O + and REE 16 O 1 H + ). The measured intensities of the
interfering element, and the various interferences it produces, allow calculation of
ratios expressing the relative amount of interfering species generated as a function of
the concentration of the interfering element. For example, if as mentioned above, 3%
of 140 Ce in the sample is ionized to 140 Ce 16 O + (i.e., 140 Ce 16 O + / 140 Ce equals 0.03), and if
the Ce concentration in the sample corresponds to a measured intensity of 100,000
8

cps, then 3000 cps must be subtracted from any signal intensity measured at mass 156
(e.g., 156 Gd or 156 Dy). For a detailed treatment of possible REE interfering species and
examples of calculations of the appropriate correction ratios the reader is referred to
Dulski (1994). Considering the number of isotopically different REE-oxyhydroxide
species that may be generated (Table 2), such corrections may seem unwieldy, but
determination of these correction ratios is not necessary with each individual ICPMS
analysis. The approach adopted here, and following the methods of Dulski (1994),
requires periodic determination of these correction ratios at very specific ICPMS
instrument settings which consistently produce similar REEO(H) + /REE + values, and
then ensuring that sample analyses are always performed at identical ICPMS
instrument settings. For the analyses performed within the JUB Geochemistry Lab
this entails tuning and optimization of the ICPMS so that CeO + /Ce + does not deviate
from 0.029 ±0.001.
4.2. External calibration
The determination of trace metal concentrations in sample solutions is
performed using laboratory prepared external calibration standards. This procedure
consists of using commercially available 1000 mg/l single element standards
(Inorganic Ventures, USA) that are combined to form 1 mg/kg (parts-per-million,
ppm) multi-element stock solutions. The 1 mg/kg stock solutions are prepared in acid
matrices that are similar to those present in the 1000 mg/l single element standards,
which are typically ~0.5 M HNO 3 (2.2% v/v). These multi-element stock solutions are
further diluted immediately prior to an individual ICPMS analysis to produce 10 and
20 μg/kg calibration standards, which are analyzed in conjunction with the sample
solutions. Preparation of these external calibration standards is presented
schematically in Figure 2.
Ideally, the determination of elemental concentrations in sample solutions
would proceed using the highly accurate method of standard addition, in which the
sample solutions are split into a minimum of three aliquots, two of which would be
spiked with the elements of interest at different concentrations. By comparing the
ICPMS response for these three solutions it is possible to calculate the concentrations
of the spiked elements in the non-spiked sample solution. The greatest advantage of
the standard addition method is that it minimizes effects caused by matrix differences
9

Figure 2. Diagram showing preparation of multi-element standards used for ICPMS external calibration
and internal standardization. Note that immediately prior to an ICPMS analysis, all solutions receive 10
μg/kg of the Ru, Re, Bi internal standard (see text for details). Trace amounts of HF (0.05 M) are used
to stabilize high field strength elements (Ti, Zr, Nb, Ta, etc.).
between samples and calibration standards, as the standards used for calibration are
incorporated directly into aliquots of the sample solution itself. The greatest drawback
to the standard addition method is that it does not lend itself to multi-element
analyses, as it is most accurate when the spike concentrations for an individual
element are similar to the unknown concentration in the sample. For geological
samples, in which concentrations of trace metals routinely range over three orders of
magnitude, this would necessitate the creation of calibration standards containing 32
potentially different, customized element concentrations, which assumes some preexisting
knowledge of these elemental concentrations in the unknown sample.
Combined with the fact that every sample would require at least triple the ICPMS
analysis time, the standard addition calibration method is generally not recommended
for routine, multi-element ICPMS analyses of geological samples.
However, the external standard calibration method suffers from the
aforementioned ‘matrix effects’. These matrix effects are particularly troublesome for
analyses of geological materials which contain significant amounts of dissolved solids
at typical ICPMS dilution factors. For example, an iron-formation with 90% Fe 2 O 3
that is diluted 1000x would produce a solution containing more than 600 ppm Fe, and
this is generally the maximum dilution suitable for pure iron-formation samples. High
dissolved metal concentrations tend to suppress ICPMS instrument sensitivity in a
10

variety of ways. These range from reductions in ionization efficiency within the
plasma to the deposition of ion-beam inhibiting sample material (e.g., salts or metaloxides)
in the interface region between the plasma source and mass spectrometer.
Therefore, there are two sample dependent matrix effects which need to be addressed
for routine geochemical analyses, one of which may suppress analyte intensities
within an individual sample (e.g., ionization efficiency), and a second effect which
may induce a time-dependent signal drift due to high TDS content in the samples.
Correcting for these sample matrix effects is achieved by using internal
standardization, in which all ICPMS solutions (blanks, external calibration standards,
CRMs, and samples) are spiked with an equivalent amount of an internal standard.
The applicability of this method is critically dependent upon the condition that the
chosen internal standard must not be present in appreciable amounts in any of the
ICPMS solutions (blanks, standards, or samples). For this purpose three elements are
used for internal standardization following the methods of Doherty (1989); Ru, Re,
and Bi. These three elements are ideal as they have low blank values for the reagents
used in the decomposition and ICPMS methods, are present at very low
concentrations in the majority of geological materials, produce no significant
interference effects for the elements of interest, and are themselves free of significant
interferences for their monitored isotopes ( 101 Ru, 187 Re, and 209 Bi).
The default ICPMS analysis consists of running samples in small batches
bracketed by calibration standards and blanks, and is described in Figure 3. A typical
ICPMS analysis consists of four to six of these small sample batches run
consecutively in a single day, and is called an ICPMS run. Prior to analysis,
appropriate sample dilution factors are determined based upon the sample type, and
samples are diluted either in 0.5 M HCl for HF-HClO 4 pressure decompositions, or in
0.5 M HNO 3 for carbonate decompositions. Acid and method blanks, calibration
standards, as well as CRMs are diluted in the same acid matrix as the samples, and all
solutions are spiked with 10 μg/kg of the Ru, Re, Bi internal standard. Matching the
acid matrix of the samples and calibration standards is crucial, as calibration
standards prepared in HNO 3 typically display ~10% greater instrument response than
calibration standards prepared in HCl. While HNO 3 is more commonly used for
preparation of ICPMS solutions as a result of this improved instrument response (as
well as fewer interfering polyatomic species at low masses), it does not produce
11

Figure 3. Diagram depicting the order in which various blanks, calibrations standards, and samples
are analyzed for a single ICPMS sample batch. A typical ICPMS analysis consists of four or more
sample batches run consecutively. The acid blank refers to the 0.5 M HCl or HNO 3 acid matrix used
for diluting all ICPMS solutions for a given analysis. Note that all solutions are spiked with 10 μg/kg
of the internal standard, and that Ru, Re, and Bi in the second 10 μg/kg 32 element calibration
standard within a batch are used for application of internal standard corrections for that batch. The
certified reference material is a commercially available geostandard that is included within every
sample decomposition and run repeatedly with every sample batch. See text for details.
stable solutions with respect to high field strength elements (HFSE) such as Zr, Nb,
Hf, and Ta (Münker, 1998), and therefore these elements are stabilized using trace
amounts of HF in the stock 1 mg/kg calibration solutions. This stability issue for
HFSE in dilute HNO 3 (0.5 M) means that samples processed using the carbonate
decomposition method should be analyzed by ICPMS within 24 hours following
completion of the sample decomposition. For samples processed using the HF-HClO 4
pressure decomposition and diluted in 0.5 M HCl, samples are also analyzed
immediately, though this acid matrix appears sufficient for stabilizing HFSE on
12

Table 3. Change in HFSE concentration in IF-G
iron-formation stored in 0.5 M HCl over a period
of ~11 months.
mg/kg
IF-G
27-Oct-06*
IF-G
16-Sep-07
% difference
2007/2006
Ti 23.9 23.5 -1.88
Zr 0.903 0.865 -4.24
Nb 0.914 0.876 -4.12
Mo 0.562 0.604 7.50
Hf 0.0246 0.0227 -7.57
Ta 0.168 0.157 -6.71
W 251 248 -1.35
* date of ICPMS analysis.
longer time scales, as measured HFSE values in the IF-G geostandard change by less
than 8% over a period of ~11 months (Table 3).
4.3. Data collection
Unknown samples that are analyzed within an ICPMS run (i.e., in a single
day, as part of one or more sample batches that are analyzed consecutively), are
generally measured once, whereas blanks, calibration standards, and CRMs are
measured repeatedly. Prior to beginning any ICPMS measurements, the instrument is
prepared by performing three consecutive analyses of the CRM solution to be used
during that ICPMS run. This has the primary effect of conditioning the cones located
in the interface region between the plasma and the mass spectrometer. These cones
are instrumental in shaping the ion beam and must be periodically cleaned, and this
pre-treatment significantly stabilizes signal intensities. A rinse solution is pumped
through the sample introduction system before the measurement of any solution, be it
a blank, standard, or sample. Nitric acid (0.5 M) is the typical rinsing agent, and a six
minute rinsing step has been found effective for minimizing memory effects. An
ICPMS measurement of any single solution consists of 60 individual scans through
the mass range 45 Sc to 238 U using peak hopping mode and a peak dwell time of 50 ms.
For any individual element, these 60 scans are then averaged to produce the raw
ICPMS data in cps. Scanning time for a single solution is approximately 3.5 minutes,
and when rinsing time, the time required for the autosampler to move to a different
13

solution, etc., are considered, the total amount of time necessary to measure any
single solution is ~10 minutes.
4.4. Internal standardization
After an ICPMS analysis, the raw data in cps are transferred to an Excel
spreadsheet for the application of three separate mathematical corrections. These
corrections are applied to all analytical solutions (blanks, calibration standards,
samples) in the following order: 1) internal standard corrections; 2) corrections for
interfering polyatomic species; and 3) blank corrections. The internal standard (IS)
correction using 101 Ru, 187 Re, and 209 Bi follows the method of Doherty (1989). The
general application of the internal standard correction method is illustrated by the
following equation, where Ru is used to correct the signal intensity of Sr:
I
⎛ ⎞
⎜ I Ru,
std
= ∗ ⎟
, sample I Sr,
sample
(2)
⎜ ⎟
⎝ I Ru,
sample ⎠
corr
Sr
where I is raw intensities in cps, I Ru,std is the measured intensity for the second 10
μg/kg calibration standard within an ICPMS sample batch (see Fig. 3), and I corr is the
corrected intensity. The term in parentheses in (2) is the IS correction factor. This
approach assumes that any matrix or instrument drift effects that suppress or enhance
the signal intensity for Sr proportionally affect the Ru signal intensity, and internal
standard corrections are most effective when the element of interest and the IS are of
similar mass (Thompson and Houk, 1987). Therefore, 101 Ru is used as an internal
standard for all elements with atomic masses lower than 101 (Sc, Ti, Co, Ni, Rb, Sr,
Y, Zr, Nb, and Mo), and 209 Bi is used for elements with masses greater than 209 (Th
and U). For analytes with masses between Ru and Re, the IS correction factor term in
(2) is expanded to a mass-dependent linear function between 101 Ru and 187 Re:
IS correction factor =
⎛
⎜ I
⎜
⎝ I
Re, std
Re, sample
Ru, sample
( M analyte − M Ru ) I Ru, std
+
( M Re − M Ru ) I Ru, sample
Ru, std
− * (3)
I
I
⎞
⎟
⎟
⎠
where M is the atomic mass number for Ru, Re, and the analyte of interest. This
treatment provides an IS correction factor that varies smoothly as a function of mass
between 101 Ru and 187 Re. Internal standard correction factors typically range from
14

0.90, corresponding to an anomalous increase in measured analyte intensities in the
sample, to 1.10, indicating an anomalous decrease in measured analyte intensities.
Experience has shown that analytical accuracy is often compromised when IS
correction factors deviate from unity by more than ~15% (i.e., 1.15), and it
may be necessary to re-analyze the samples using a different dilution factor to
minimize matrix and/or drift effects.
Following internal standard corrections, interference corrections for
polyatomic species are applied to the sample data. These interference corrections are
applied only to elements with atomic masses between 151 (Eu) and 183 (W). A
detailed treatment of the determination and application of these interference
corrections is beyond the scope of this discussion, and the reader is referred to Dulski
(1994). The last correction applied to the sample data before calculating elemental
concentrations involves subtracting the internal standard and interference corrected
blank contribution, which would be due to the reagents used in preparing the samples,
and/or from any laboratory procedures.
4.5. Analytical blanks and quantification limits
Two blank contributions may be defined for the described ICPMS methods.
The first is the acid blank value, which is used for calculating the instrument
detection and quantification limits (IDLs and IQLs). As all ICPMS solutions are
analyzed in a 0.5 M acid matrix (either HCl or HNO 3 ), the acid blank is simply the
purest 0.5 M acid solution that is capable of being produced within the JUB
Geochemistry Lab. Analyses of the acid blank allows the calculation of IDLs and
IQLs for any given element, which represent the best achievable (ideal) ICPMS
sensitivity using the available laboratory reagents. The IDL and IQL equal 3x and 10x
the standard deviation of the acid blank, respectively (MacDougall and Crummett,
1980), and are best characterized through long term, repeated analyses of the acid
blank, as daily fluctuations in instrument sensitivity are common. It is important to
note that IDLs and IQLs calculated from the acid blank are not equivalent to detection
and quantification limits in a solid sample, that is, in a rock or a mineral. As all solid
samples are highly diluted before ICPMS analyses, IDLs and IQLs determined for
solid samples equal the acid blank multiplied by the appropriate dilution factor. Table
2 lists long term average IQLs in solid samples for 0.5 M HCl and 0.5 M HNO 3 acid
matrices.
15

The second blank contribution to ICPMS analyses is the method blank. The
method blank is a solution that has been processed in exactly the same manner as a
sample during the decomposition method, and represents the contamination added to
a sample as a result of the decomposition method and subsequent dilution in acid.
While significant amounts of concentrated acid are used in the decomposition
methods, these acids (in combination with contamination from labware) contribute
insignificant amounts of trace metals to rocks typically analyzed, even for trace
metal-poor samples like iron-formations and dolomites (Fig. 4). Method blank values
are determined for every ICPMS analysis to monitor possible contamination sources,
but as a result of the above observations, ICPMS data are blank corrected by
subtracting the acid blank intensities. Only in instances where method blank values
are significantly higher than those observed in Fig. 4 are blank corrections performed
using the method blank intensities. The reasoning for this approach is as follows;
since the method blanks are typically below the IQLs determined from the acid
blanks, the method blank is not quantifiable. Only when method blanks exceed the
acid blank IQL are they used for the blank correction. It should be noted that with
regard to the carbonate decomposition method, many refractory elements (e.g., Ti,
Nb, Ta) are not suitable for quantification, as they are expected to be primarily hosted
in silicate-bearing phases that are resistant to dissolution with nitric acid, and in
particular the poor IQLs for Sc mean that it is generally not quantifiable in trace
metal-poor samples regardless of the decomposition method used (Fig. 4).
5. Analytical precision
For the analyses conducted within the Geochemistry Lab at JUB, analytical
precision may be defined in different ways. The highest degree of precision is
expected when considering only the 60 mass scans performed during ICPMS analysis
of a single sample solution. If this sample solution is periodically re-measured with
time during an ICPMS run, i.e., by occasionally returning the autosampler probe to
the vial containing the solution, then analytical precision would be expected to
decrease. Further decreases in analytical precision should occur if one considers
repeated analyses of the same sample solution on different days, or repeated acid
decompositions of the same sample powder, with the decreasing precision reflecting
16

10 2 JDo-1 reference values
10 1
IF-G reference values
IQL HCl (n=21)
method blank HCl (n=19)
10 0
10 -1
mg/kg
10 -2
10 -3
10 -4
10 -5
10 -6
10 -7
10 2 Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
10 1
0.5 M HCl
IQL HNO3 (n=3)
method blank HNO3 (n=3)
10 0
10 -1
mg/kg
10 -2
10 -3
10 -4
10 -5
10 -6
10 -7
0.5 M HNO3
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 4. Long term average instrument quantification limits in mg/kg for solid samples (i.e.,
accounting for dilution factors) as determined from acid blanks, and average method blanks. Data for
acid blank IQLs are also presented in Table 2. The top figure corresponds to a 0.5 M HCl acid matrix,
and includes average method blanks for the HF-HClO 4 pressure decomposition method as well as
reference data for the iron-formation IF-G, which represents a rock type that contains very low trace
metal contents typically dissolved using the HF-HClO 4 method. The bottom figure corresponds to a 0.5
M HNO 3 acid matrix and the carbonate decomposition method, and includes reference data for the
dolomite JDo-1, which represents a very low trace metal content rock typically dissolved using HNO 3 .
Regardless of decomposition method, the method blanks are at least 1 order of magnitude lower than
IQLs for all elements, and generally 3-6 orders of magnitude lower than concentrations observed in the
very trace metal-poor rocks typically analyzed. Method blanks and IQLs are significantly lower in 0.5
M HNO 3 . Note that some elements are likely to be unquantifiable in many trace metal-poor rock
samples due to concentrations close to or below the IQL (e.g., Sc, Nb, Ta).
17

error propagation during the complete sample decomposition and ICPMS analytical
procedure. However, for many elements, in particular Y, Ba, the REE, and U, limits
of precision appear to be controlled by the ICPMS measurement, and not by the
sample decomposition method.
For the discussion of precision, three types of analytical precision are defined:
1) sample precision, calculated from the 60 mass scans of a single solution and
representing raw, uncorrected data; (2) run precision, calculated from element
concentrations determined by repeated analyses of a single solution over the duration
of an ICPMS run (typically 6-12 hours); and (3) method precision, calculated from
element concentrations determined from repeated acid decompositions and ICPMS
analyses of a sample powder. The three types of precision and their relationships are
illustrated in Figure 5. Note that sample precision reflects uncorrected intensities
(cps), whereas run and method precision incorporate internal standard, interference,
and blank corrections. To provide the best estimate of precision for analyses of rocks
and minerals, these three types of precision are calculated using data obtained for the
certified reference materials listed in Table 1. The discussion of sample, run, and
method precision is limited to those analyses where the CRM was run repeatedly as
part of every sample batch (see Fig. 3). Data are discussed as percent relative standard
deviation (%RSD), though it must be stressed that standard deviations themselves
usually vary significantly (±50% relative), so that an average RSD of 2% is expected
to typically range as low as 1% and as high as 3%.
Sample precision in ICPMS determinations for common rock types is
presented in Figure 6, along with the average sample precision for the 10 μg/kg
calibration standard. As the 10 μg/kg calibration standard represents a solution free of
the matrix effects typical of dissolved rock samples, it is expected to display the best
RSD values, which are ~2% for almost all monitored isotopes. However, the 10 μg/kg
calibration standard sample precision is indistinguishable from the sample precision
determined for dissolved rock solutions, particularly for rock types that are not tracemetal
poor, such as basalt and shale. Only for rock types that are very low in trace
metals (e.g., JDo-1 dolomite), are sample precisions worse than those observed in the
10 μg/kg calibration standard, and only then for certain elements (e.g., Rb, Mo, Cs,
Hf, Th, U). The run precision, where measurements of a CRM solution are
periodically repeated over the course of an ICPMS run, is generally ~2%, similar to
18

Figure 5. Diagram depicting various types of precision that may be defined for sample decompositions
and ICPMS analyses. Note that sample precision reflects raw uncorrected intensities (cps), while run
and method precision are calculated from corrected concentration data.
that of the sample precision for many elements, and also varies as function of rock
type (Fig. 7). Shales display slightly poorer run precision (3-5%), though this might
reflect the limited number of analyses where these CRMs were measured repeatedly.
For purposes of long term reproducibility, the method precision is the most suitable
measure of the standard deviation expected for a complete sample decomposition and
ICPMS analyses, as it describes the variability expected in concentration data when a
sample powder is dissolved numerous times over a period of months or years.
Comparisons of the three types of precision discussed here (sample, run, and method)
are presented in Figure 8 for CRMs of four different rock types. Of the 32 elements
analyzed, the method precision RSD is better than 5% for 25 elements in BHVO-2
(basalt), 31 elements in SGR-1b, 26 elements in FeR-2 (iron-formation), and 22
elements in JDo-1 (dolomite). The method precision is also comparable to the sample
and run precision, suggesting that any error associated with the HF-HClO 4 sample
decomposition method (e.g., weighing and/or dilution errors) is small compared to the
error inherent in the ICPMS analysis.
19

sample precision (%RSD)
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
BHVO-2 (n=5)
10 μg/kg calibration standard
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
Sc-45
Ti-47
Ti-49
Co-59
Ni-60
Ni-62
Rb-85
Sr-88
Y-89
Zr-90
Zr-91
Nb-93
Mo-95
Mo-97
Cs-133
Ba-135
Ba-137
La-139
Ce-140
Pr-141
Nd-145
Nd-146
Sm-147
Sm-149
Eu-151
Eu-153
Gd-158
Gd-160
Tb-159
Dy-161
Dy-163
Ho-165
Er-166
Er-167
Tm-169
Yb-171
Yb-172
Yb-174
Lu-175
Hf-178
Hf-179
Ta-181
W-182
W-183
Pb-206
Pb-207
Pb-208
Th-232
U-238
Sc-45
Ti-47
Ti-49
Co-59
Ni-60
Ni-62
Rb-85
Sr-88
Y-89
Zr-90
Zr-91
Nb-93
Mo-95
Mo-97
Cs-133
Ba-135
Ba-137
La-139
Ce-140
Pr-141
Nd-145
Nd-146
Sm-147
Sm-149
Eu-151
Eu-153
Gd-158
Gd-160
Tb-159
Dy-161
Dy-163
Ho-165
Er-166
Er-167
Tm-169
Yb-171
Yb-172
Yb-174
Lu-175
Hf-178
Hf-179
Ta-181
W-182
W-183
Pb-206
Pb-207
Pb-208
Th-232
sample precision (%RSD)
U-238
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
SGR-1b (n=2)
10 μg/kg calibration standard
sample precision (%RSD)
FER-2 (n=3)
10 μg/kg calibration standard
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
Sc-45
Ti-47
Ti-49
Co-59
Ni-60
Ni-62
Rb-85
Sr-88
Y-89
Zr-90
Zr-91
Nb-93
Mo-95
Mo-97
Cs-133
Ba-135
Ba-137
La-139
Ce-140
Pr-141
Nd-145
Nd-146
Sm-147
Sm-149
Eu-151
Eu-153
Gd-158
Gd-160
Tb-159
Dy-161
Dy-163
Ho-165
Er-166
Er-167
Tm-169
Yb-171
Yb-172
Yb-174
Lu-175
Hf-178
Hf-179
Ta-181
W-182
W-183
Pb-206
Pb-207
Pb-208
Th-232
U-238
Sc-45
Ti-47
Ti-49
Co-59
Ni-60
Ni-62
Rb-85
Sr-88
Y-89
Zr-90
Zr-91
Nb-93
Mo-95
Mo-97
Cs-133
Ba-135
Ba-137
La-139
Ce-140
Pr-141
Nd-145
Nd-146
Sm-147
Sm-149
Eu-151
Eu-153
Gd-158
Gd-160
Tb-159
Dy-161
Dy-163
Ho-165
Er-166
Er-167
Tm-169
Yb-171
Yb-172
Yb-174
Lu-175
Hf-178
Hf-179
Ta-181
W-182
W-183
Pb-206
Pb-207
Pb-208
Th-232
U-238
sample precision (%RSD)
JDo-1 (n=3)
10 μg/kg calibration standard 23.0%
Figure 6. Average sample precision (see text for definition) for different rock types dissolved using the
HF-HClO 4 decomposition, as calculated from the number (n) of decompositions where the CRM was
measured repeatedly. The average 10 μg/kg calibration standard is presented for comparison (n=19).
Precision is generally very similar between the rock solutions and the calibration standard, except for
elements that are present at low concentrations in trace-metal poor rocks such as the JDo-1 dolomite
(e.g., Rb, Mo, Cs, Hf, Th, U). This suggests that the standard deviation of ICPMS measurements is
controlled by inherent instrument precision rather than by sample matrix effects.
20

20
18
basalt
BHVO-2 (n=5)
run precision (%RSD)
16
14
12
10
8
6
4
2
0
20
18
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
shales
SCo-1 (n=2)
SGR-1b (n=2)
run precision (%RSD)
16
14
12
10
8
6
4
2
0
20
18
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
iron-formations
FER-2 (n=3)
IF-G (n=3)
run precision (%RSD)
16
14
12
10
8
6
4
2
run precision (%RSD)
0
30
28
26
24
22
20
18
16
14
12
10
8
6
4
2
0
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
dolomite
JDo-1 HCl (n=3)
JDo-1 HNO3 (n=3)
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 7. Average run precision for multiple (n) sample decompositions of certified reference materials
that were analyzed repeatedly within a single ICPMS run (see text for details). All data for HF-HClO 4
sample decompositions, except for JDo-1 dolomite, which includes data for three carbonate (HNO 3 )
decompositions. Note bottom figure (dolomite) uses different scale. Poor RSD values are generally due
to element concentrations approaching the instrument quantification limit (e.g., Sc, see Fig. 4).
21

20
15
BHVO-2 (n=5)
sample precision
run precision
method precision
%RSD
10
5
0
20
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
SGR-1b (n=2)
15
%RSD
10
5
20
0
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
FER-2 (n=3)
70.5% 34.9%
15
%RSD
10
5
0
30
25
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
JDo-1 (n=3)
36.3% 59.3%
%RSD
20
15
10
5
0
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 8. Comparison of average sample, run, and method precision for various CRMs, where n refers
to the number of HF-HClO 4 sample decompositions and ICPMS runs in which the CRM was measured
repeatedly. For most elements method precision is better than 5%, and is similar to, or even lower than,
the sample or run precision. This suggests that for these elements the standard deviation of the data is
primarily controlled by the inherent precision of the ICPMS instrument, and not by the decomposition
method. Poor method precision is also due to low abundances for some elements that approach the IQL,
e.g., Nb, Ta, and W in FeR-2 and JDo-1, as well as Co in JDo-1.
22

The method precision for elements present at low concentrations, particularly
in FeR-2 and JDo-1, is typically poorer (e.g., Nb, Ta, W, as well as Co in JDo-1).
However, some elements that are abundantly present at concentrations ranging from
several mg/kg to several percent also display poor RSD values. The BHVO-2 basalt
contains >1.5% Ti, yet suffers from a method precision of >10%, whereas Ba in JDo-
1 is present at a concentration (6.14 mg/kg) well above the quantification limit and
has an RSD above 20%. In the case of Ti, it is suspected that the poor ionization
efficiency of this metal may mean that the 10 μg/kg calibration standard is not
appropriate for quantification, as it produces a signal intensity that is not sufficiently
greater than that observed for the background acid matrix, and this will be discussed
in greater detail in the section regarding analytical accuracy. The poor method
precision observed for Ba in the JDo-1 dolomite (22.7%) results from an ‘outlier’ for
one of the three decompositions used in the calculation of the RSD value, as
measured Ba in two decompositions is 5.32 and 5.63 mg/kg, while Ba measured in
the third decomposition is 8.61 mg/kg.
To this point the discussion of analytical precision has been limited to the HF-
HClO 4 decomposition method, as much of the research conducted within the JUB
Geochemistry Lab focuses on whole-rock trace metal analyses, which necessitate
complete dissolution of all mineral phases within the sample powder. However, with
regard to analytical precision it is necessary to discuss the HNO 3 carbonate
decomposition as well. As the carbonate decomposition is used for limestone and
dolomite samples, the JDo-1 dolomite is the typical CRM included in decompositions
and analyses of these rock types.
Sample and method precision for JDo-1 using the carbonate decomposition
method are presented in Figure 9, and these measures of precision are comparable for
most elements. This suggests that the carbonate decomposition method does not
significantly increase the error budget for these elements, similar to that observed for
the HF-HClO 4 decomposition. The method precision is better than 5% (RSD) for 25
of the 32 elements analyzed, and several elements displaying poor method precision
(e.g., Rb, Cs, and Hf) are expected to be hosted in refractory mineral phases that
would be resistant to dissolution by nitric acid. Elements at or near the IQLs and
unlikely to be quantifiable in carbonate rocks at concentrations observed in JDo-1
include Sc, Nb, and Ta (see Fig. 4). For many elements the method precision as
23

30
25
JDo-1 (n=3)
sample precision HNO 3
36.3% method precision HNO 3
59.3%
method precision HF-HClO 4
20
%RSD
15
10
5
0
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 9. Average sample and method precision for the HNO 3 carbonate decomposition of the JDo-1
dolomite CRM (for run precision see Fig. 7). Method precision for the carbonate decomposition is
better than 5% for 25 of the 32 elements analyzed, and poorer precision is generally restricted to
elements typically hosted in refractory silicate minerals (Rb, Cs, Hf).
determined for the carbonate decomposition is quite similar to that observed for the
HF-HClO 4 decomposition method, though the HF-HClO 4 decomposition results in
significantly better precision for elements typically hosted by silicate minerals (e.g.,
Rb, Zr, Cs, Hf), due to the complete dissolution of refractory Si-bearing phases.
6. Analytical recovery
The relative precision for most elements analyzed is quite good, but this
provides no information regarding potential loss of these elements as a consequence
of the sample decomposition or ICPMS measurement. Some elements can exist as a
volatile chemical species at some point during the decomposition process, and
therefore may exhibit non-conservative behavior during sample preparation. The best
example of this is Si, which is converted to volatile SiF 4 during HF dissolution of
silicate minerals, and is consequently lost from the sample during evaporation of HF
early in the HF-HClO 4 decomposition method. The ability of the sample treatment
and ICPMS measurements described here to conserve the elements of interest
throughout the decomposition and analytical procedure is termed analytical recovery.
24

Two approaches may be utilized to determine and quantify non-conservative
behavior of the elements of interest. The first involves preparing artificial laboratory
solutions that contain known quantities of the elements of interest, and then simply
treating these artificial solutions as if they were unknown samples by subjecting them
to a full sample decomposition and ICPMS analysis. This characterization of
analytical recovery is discussed here, while the second approach, which involves
determining elemental concentrations in a CRM and comparing these data with the
reference values, is discussed below in the section regarding analytical accuracy.
Analytical recoveries of multi-element spikes (10 μg/kg) for the carbonate and
HF-HClO 4 decomposition methods are presented in Figure 10. Recoveries for most
elements are excellent, particularly for the carbonate decomposition, in which 27 of
32 elements have measured concentrations between 97-103% of the expected values.
However, measured Ni in the carbonate digestion spike solution is ~30% too high,
whereas Ta and W exhibit quite low recoveries of ~24% and ~80%, respectively. The
anomalously high Ni recovery is attributed to contamination, as method blanks for the
carbonate decomposition typically contain 2-5 μg/kg Ni. The significant loss
observed for Ta and W is likely due to the filtration procedure performed for the
carbonate decomposition. During the filtration step, the sample solution (10 ml of 5
M HNO 3 ) is passed over a 0.2 μm cellulose acetate filter (Fig. 1), which is
subsequently rinsed with ~20 ml of 0.5 M HNO 3 . Whereas this rinse step appears
sufficient for most elements, Ta and W (and to a lesser extent Nb) are apparently
retained, and may only be quantitatively rinsed from the filter by adding trace
amounts of HF (0.05 M) to the rinsing solution (K. Schmidt, personal
communication). The addition of HF to the rinsing solution should only be utilized if
information regarding W is critical, as HF can form unstable complexes with other
trace metals, particularly the rare earths, potentially limiting quantitative recovery of
these elements. With regard to Nb and Ta, the use of HF during the filter rinse step is
not particularly informative , as these are immobile elements expected to be hosted in
refractory aluminosilicate phases, which are resistant to dissolution by the carbonate
decomposition method.
For the HF-HClO 4 decomposition, analytical recovery for the 32 elements in
the spike solution is more variable than that observed for the carbonate
decomposition. Low mass elements tend to have slightly low recoveries around 95%,
25

1.35
1.30
1.25
1.20
1.15
HNO 3 carbonate decomp.
HF-HClO 4 decomp.
HF-HClO 4 decomp. pre- June 2006
1.35
1.30
1.25
1.20
1.15
fraction recovered
1.10
1.05
1.00
0.95
0.90
0.85
0.80
0.75
0.30
0.20
0.10
1.10
1.05
1.00
0.95
0.90
0.85
0.80
0.75
0.30
0.20
0.10
Sc Ti Co Ni Rb Sr Y Zr NbMoCs Ba La Ce Pr NdSmEu Gd Tb Dy Ho Er TmYb Lu Hf Ta W Pb Th U
Figure 10. Multi-element analytical recoveries for the carbonate and HF-HClO 4 decomposition
methods, expressed as a ratio of the measured element concentration divided by the spike concentration.
Note break in the y-axis between 0.35-0.75. Spike solutions contained 10 μg/kg of the standard 32
elements analyzed, except for the pre-June 2006 HF-HClO 4 decomposition, which contained 1 μg/kg of
the 24 elements routinely measured prior to June, 2006. Recoveries for most elements are between 95-
105%, particularly for the carbonate and pre-June 2006 decompositions. Notable exceptions are Ni, Ta,
and W for the carbonate digestion, which suggests contamination/interferences (Ni) and loss (Ta and
W) during the decomposition, perhaps as a result of the filtering step. The large spread and general
increase as a function of mass, in the fraction recovered for the 32 element HF-HClO 4 decomposition is
attributed to effects arising from the utilization of the internal standards 101 Ru and 187 Re (see text for
details).
and analytical recovery generally increases with increasing mass, though Ta does not
follow this trend (Fig. 10). However, a test of analytical recovery performed prior to
June 2006, conducted with the 24 elements originally analyzed by ICPMS within the
JUB Geochemistry Lab, resulted in recovery percentages significantly better for many
elements, particularly for elements with atomic masses below 138 (Ba). The exact
reason for the more variable recovery observed for the full 32 element analysis
currently conducted is not clear, though it is likely related to the internal standard
correction. The generally constant analytical recovery of ~95% for elements with
masses below 100, followed by the monotonic increase in recovery between masses
133 (Cs) and 183 (W), is mirrored by the IS correction factor for this analysis, which
was 0.95 for all masses below 133 Cs, before linearly increasing to 0.98 for 183 W, and
then decreasing to 0.96 for Pb, Th, and U.
26

For the more recent, full 32 element test of analytical recovery, the 10 μg/kg
spike solution was analyzed in the middle of an ICPMS run containing Fe-rich
carbonates that had a dilution factor of only 250, and these carbonates displayed
extreme IS correction factors as high as 2.0, reflective of the high TDS content of
these solutions. In contrast, the analytical recovery as determined from the pre-June
2006 analysis was not included within part of an ICPMS run, and was conducted
following routine cleaning and optimization of the ICPMS. It therefore seems that the
discrepancy between the two tests of analytical recovery for the HF-HClO 4
decomposition method, one using a 24 element spike and the other a 32 element
spike, reflects details of the individual tests, including such factors as sample
deposition within the ICPMS interface region affecting signal intensities (i.e., signal
suppression or enhancement).
Based on the few data, it is therefore concluded that for the HF-HClO 4
decomposition analytical recoveries for the majority of elements are typically within
2-3% of their predicted concentrations, similar to that observed for the carbonate
decomposition method. This variability may represent the optimum analytical
recovery possible, as it is comparable to the inherent precision of ~2% observed for
most element measurements (see above discussion regarding precision). However, in
view of the limited data, more tests of analytical recovery are warranted, particularly
using the HF-HClO 4 decomposition method and complete, 32 element spike solutions.
7. Analytical accuracy
7.1. Reference values and major element interferences
The best measure of the suitability of the analytical methods employed within
the JUB Geochemistry Lab is if these methods can accurately (and reproducibly)
quantify the elements of interest in certified reference materials. As mentioned
previously, the CRMs used as routine quality assurance standards are chosen to match
the rock types most commonly analyzed within the JUB Geochemistry Lab. If the
measured CRM concentration data for a given sample decomposition is consistent
with the certified data, then this provides the best measure of conservative element
behavior and overall accuracy for the ICPMS methods employed.
The concentration data for CRMs are generally reported as either
recommended values or preferable/informational values, with recommended values
27

considered more accurate and possessing lower degrees of uncertainty. Concentration
data for most major and some trace elements in CRMs are typically provided by the
issuing organization (Table 1). However, for many trace metals (particularly the REE,
Nb, Mo, Ta, and W), values are either not provided by the issuing organization or are
considered informational or preferred values, indicating that higher degrees of
uncertainty surround these data. Therefore, for a significant number of the 32
elements analyzed at JUB, concentration data in CRMs may only be obtained from
combinations of separate studies and compilations.
One of the most comprehensive and widely referenced compilations of
geoanalytical data was assembled by Govindaraju (1994), which combines data
produced using different analytical methods from a multitude of sources. However,
for the CRMs listed in Table 1 much of the minor and trace metal data in Govindaraju
(1994) is highly variable, particularly for the REE and other metals present at low
concentrations such as Ti, Nb, Ta, Hf, Th, and U. Additionally, for some of the CRMs
data has been published since 1994 using newer analytical methods such as laser
ablation and/or high resolution multi-collector ICPMS. A thorough study of numerous
CRMs, including the ones relevant to this discussion, was conducted by Dulski
(2001), whose data closely match results from this study. Newer data (e.g., Baker et
al., 2002; Bohlar et al., 2004) are consistent with the results presented here and the
values of Dulski (2001), suggesting that the highly variable trace metal concentrations
observed in older compilations of CRM data do not result from sample heterogeneity,
but rather from limitations of older analytical methods.
The reference values for the CRMs used in this study are presented in
Appendix 1, along with average concentration data for these CRMs as measured
within the JUB Geochemistry Lab. The literature data presented are not intended to
reflect an exhaustive or comprehensive survey of all available data for the various
CRMs, and were selected according to three criteria: 1) data published by the CRM
issuing organization, which usually are compilations of data produced by different
laboratories using different methods; 2) studies that provided data for a large number
of the 32 elements considered for this study; and/or 3) work that utilized similar
analytical methods, particularly ICPMS techniques. This approach is intended to
facilitate comparison between the CRM values measured at JUB with worldwide
laboratories using similar analytical methods. Recently, Jochum et al. (2005)
produced a thorough compilation of published major element, trace element, and
28

isotopic data for geologic CRMs, available as an electronic database (GeoReM), and
the reader is referred to this resource for a complete survey of available CRM data.
Analytical accuracy will be discussed individually for each CRM. However,
some general observations can be made regarding the JUB data and the reference
values. The first is that the JUB data typically display greater precision compared to
the average reference values, which is ascribed to the greater number of different
methods and analytical techniques employed in the production of the reference
values. The second is that poor precision and/or accuracy in the JUB data for specific
elements, particularly those with masses

16000
0.5 M HCl matrix
a
1.6
14000
1.4
interference on 59 Co in sample solution (cps)
12000
10000
8000
6000
4000
2000
MgCl +
CaO(H) +
MgCl + interference in JDo-1
CaO(H) + interference in JDo-1
1.2
1.0
0.8
0.6
0.4
0.2
interference on 59 Co in sample powder (mg/kg)
0
0.0
0 10 20 30 40 50 60 70 80 90 100
MgO, CaO in sample powder (wt.%)
16000
0.5 M HNO 3
matrix
b
1.6
14000
1.4
interference on 59 Co in sample solution (cps)
12000
10000
8000
6000
4000
2000
CaO(H) +
CaO(H) + interference in JDo-1
1.2
1.0
0.8
0.6
0.4
0.2
interference on 59 Co in sample powder (mg/kg)
0
MgCl +
0.0
0 10 20 30 40 50 60 70 80 90 100
MgO, CaO in sample powder (wt.%)
Figure 11. Mg and Ca interferences on 59 Co as functions of MgO and CaO content in sample powders.
MgO and CaO are calculated as wt.% assuming the sample powders have been diluted by a factor of
1000, which is a typical dilution factor for Mg- and Ca-rich carbonate rocks (primarily dolomites). Top
figure (a) shows interferences observed in HCl matrix, and the MgCl interference on mass 59 predicted
for JDo-1 (18.4% MgO) is >0.6 mg/kg, significantly greater than literature Co values of ~0.2 mg/kg
(see App. 1). Bottom figure (b) illustrates that MgCl interferences are eliminated when samples are
analyzed in HNO 3 , yet significant CaO(H) + interferences predicted for JDo-1 (34.0% CaO) still
preclude accurate determinations of Co. It appears that reasonable Co determinations may only be
possible for typical carbonate samples (MgO, CaO of 5-40%), when Co contents in the rock are at
least one order of magnitude higher than the predicted interference concentrations (i.e., >10 mg/kg).
30

Table 4. Interferences observed for elements of interest with atomic masses 1000
44 Ca 16 O 37 Cl insignificant interference
* all interfering molecular species presumed to exist in a +1 valence state, and bold text indicates dominant species.
interferences must be inferred, as the total interference at a given mass may represent
combinations of multiple interfering species, and examples would include CO 2,
CO 2 H, NO 2, CCl, and NCl. For the major elements Mg, Ca, Al, Mn, and Fe, high
concentrations in rock samples are not observed to produce significant interferences
for atomic masses >100. The practice of diluting samples with HCl following
decomposition and the subsequent formation of metal-chloride molecular species
(MCl(O) + ) is responsible for many of the interferences listed in Table 4, and these
MCl(O) + interferences are insignificant for samples decomposed using the carbonate
decomposition method. Samples decomposed using the HF-HClO 4 method may be
diluted with 0.5 M HNO 3 to reduce or eliminate MCl(O) + , assuming that all HCl used
during the HF-HClO 4 decomposition is evaporated prior to diluting with HNO 3 .
This suggests that use of HCl should be avoided in order to minimize MCl(O) +
interferences. However, some elements, particularly the HFSE Nb and Ta, are not
stable in low molarity HNO 3 solutions, and require HCl or HF to prevent apparent
31

‘loss’ of these elements in dissolved rock solutions due to adsorption/precipitation
reactions in sample containers (Münker, 1998). For trace concentrations of Nb and
Ta, Münker (1998) suggested diluting dissolved rock samples in 0.3 M HNO 3 and
0.06 M HCl for stabilizing these metals for ≥24 h, and the addition of 0.02 M HF
when Nb and Ta in the diluted solution exceeded approximately 100 μg/kg and 4
μg/kg, respectively. The consistent use of 0.5 M HCl and the formation of MCl(O) +
interferences does not appear to have adversely affected analytical precision or
accuracy for the majority of the ICPMS analyses discussed here. However, for
samples rich in Mg, Mn, and/or Fe that also have very low trace metal concentrations,
it would be useful to more fully investigate mixtures of HNO 3 and HCl in order to
maximize the sensitivity of the ICPMS method.
Some interferences persist regardless of the acid matrix used and result from
analyzing aqueous solutions (e.g., CO 2 on 45 Sc), or from sample ionization in an Ar
plasma. The latter effect is illustrated by presumed 44 Ca + 2 and 48 Ca 40 Ar + interferences
on 88 Sr, which may erroneously increase measured Sr concentrations by 1-3 mg/kg as
CaO contents in sample powders approach 40%. Another example is 55 Mn 40 Ar + ,
which produces a 2-5 mg/kg interference on 95 Mo in sample powders that contain 10-
15% MnO. Figures illustrating the significant interferences observed for Ni, Sr, Y, Zr,
Nb, and Mo are presented in Appendix 2, and permit estimates of the magnitude of
these interferences in various rock samples. It should be stated, however, that
interferences observed in this study may not significantly affect analytical accuracy,
provided the magnitude of the interference is small relative to the concentration of the
trace metal of interest. An example is Nb, which cannot be determined accurately in
Fe-rich, Nb-poor samples (20 mg/kg Nb).
Analytical accuracy, as defined by the agreement between the average
reference value and the measured JUB concentrations (App. 1), is presented in the
following figures as a ratio of (JUB data/reference value), with a ratio of one
indicating perfect agreement between the JUB data and the average CRM reference
value. However, significant uncertainties in reference values exist for some elements,
and some CRMs are more poorly characterized than others. This uncertainty in the
CRM reference values can produce a wide range in the calculated ratio of (JUB
data/reference value). Therefore, the range in calculated ratios due solely to reference
32

value uncertainty (as %RSD) is represented graphically in the following figures by
means of vertical grey bars. A cursory examination indicates this range can be quite
large, as illustrated in Fig. 12 for Mo (0.63–1.33) in FeR-2.
7.2. High Fe content rocks
Banded iron-formations (IFs) are a rock type frequently analyzed within the
JUB Geochemistry Lab, and these samples are typically more than 90% SiO 2 and
Fe 2 O 3 in varying proportions, with NaO, MgO, Al 2 O 3 , and CaO all commonly present
at low concentrations (

1.50 FeR-2 (n=3)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 12. Accuracy estimation for ICPMS analysis of iron-formation FeR-2, where n equals the
number of separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars
represent variability in the calculated ratio that is due solely to uncertainty in the reference average.
Vertical lines represent the variability in the calculated ratio due to the uncertainty of the JUB data, and
all data are presented in App. 1. Precision of JUB data is generally better than that observed for the
average reference values, particularly for elements with atomic masses >100 amu (i.e., heavier than
Mo). Measured JUB data for higher mass elements are generally lower than the reference average,
though this reference average is skewed to higher values by the data of Yu et al. (2001) (see text for
details and App. 1). Data for Nb not included due to probable FeCl interferences and large uncertainty
in JUB data for Nb measurements (70% RSD).
FeR-2 compared to very Al-poor IFs. For FeR-2, the standard deviation of the JUB
data is typically 1-3%, considerably better than the range observed in the literature
data (App. 1). The JUB data tends to be somewhat lower than the reference value
average, with 25 of the 32 elements analyzed displaying concentrations between 85–
100% of the literature data (Fig. 12). However, the average reference values for many
elements are skewed towards higher numbers by the data of a single study (Yu et al.,
2001), and if the data from Yu et al. (2001) are not considered, then measured JUB
concentrations for 24 of 32 elements are within 90-110% of reference values (see
App. 1).
FeR-4
The FeR-4 iron-formation has very similar SiO 2 (50.1%) and Fe 2 O 3 (39.2%)
contents compared to FeR-2, but only one-third as much Al 2 O 3 (1.70%).
34

1.50 FeR-4 (n=2)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 13. Accuracy estimation for ICPMS analysis of iron-formation FeR-4, where n equals the
number of separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars
represent variability in the calculated ratio that is due solely to uncertainty in the reference average.
Vertical lines represent the variability in the calculated ratio due to the uncertainty of the JUB data, and
all data are presented in App. 1. Good agreement is observed for most elements, with exceptions being
Sc, Ti, Co, Ni and U. Reference data for Nb, Mo, and Ta are not shown, and are available only from a
single published study by Yu et al. (2001), who found highly variable results for these elements when
comparing two different sample decomposition methods (HF-HClO 4 versus HF-H 2 SO 4 ). See text for
details.
Consequently, many trace metals have significantly lower concentrations in FeR-4
(e.g., Sc, Zr, Nb, the REE, Hf, Th). The precision of the JUB data is poorly
constrained by the limited number of analyses, though it seems comparable to, if not
significantly better than, the variation observed in literature reference values. The
relative accuracy of the FeR-4 data is presented in Figure 13, and is somewhat better
than that observed for FeR-2, except for elements with the lowest and highest masses
(Sc, Ti, Co, Ni, and U). Poor results for Sc are not unexpected as Sc literature values
for FeR-4 range from 1.1-1.5 mg/kg, which is similar in magnitude to the IQL
observed for Sc (~0.8 mg/kg, Table 2).
However, Ti, Co, and Ni are all present in FeR-4 at concentrations well above
their respective IQLs using the ICPMS methods described here (Fig. 4), so low
signal-to-noise ratios are not expected to adversely affect determinations of these
elements. For Co and Ni, the literature data are reported to only one significant digit
and concentrations of these two elements are apparently low (2 and 8 mg/kg,
35

espectively). Previous workers have observed that many metals with very low
abundances in CRMs are likely present at concentrations lower than initially reported,
and that early data compilations frequently overestimate the true abundance of these
elements (e.g., Dulski, 2001; Yu et al., 2001), perhaps due to older, less sensitive
analytical techniques. It is therefore suggested that the apparent ‘low’ concentrations
for Co and Ni determined at JUB may reflect this observation, though the limited
published data and few JUB analyses of FeR-4 indicate that more work regarding
these elements is necessary.
The Ti concentration determined in this study and the literature reference
value (327 mg/kg and 420 mg/kg, respectively), both seem high enough to suggest
that accurate Ti analyses in FeR-4 are possible. However the Ti reference value is
compiled from analyses utilizing sample fusion and X-ray fluorescence (XRF)
measurements (Abbey et al., 1983), which may suffer from relatively high blanks
from the flux used during sample fusion, and/or poor detection/quantification limits
compared to ICPMS measurements. It is therefore possible that the true Ti abundance
in FeR-4 is lower than 420 mg/kg, and the discrepancy between the JUB data and the
compiled reference average rather reflects the limited sensitivity and precision of
many of the Ti analyses of FeR-4.
IF-G
The IF-G iron-formation is the most Fe-rich and Al-poor IF utilized as a CRM
within the Geochemistry Lab, with 55.9% Fe 2 O 3 , 41.2% SiO 2 , and 0.15% Al 2 O 3 .
Unfortunately, IF-G as prepared by the issuing organization (IWG-GIT, Table 1)
during the original processing run (first lot) is no longer available, and subsequently a
second lot of IF-G was prepared by IWG-GIT. For many of the earliest IF analyses
performed within the Geochemistry Lab, IF-G was the preferred CRM, and as a result
the aliquot of the original first lot IF-G powder was completely consumed. In June
2007 an aliquot of the second lot of IF-G powder was obtained from IWG-GIT, which
subsequently has been analyzed several times. As reported by IWG-GIT, the only
differences between the two IF-G lots regard the concentrations of Cr, Co, and W. As
Co and W are relevant to this discussion, it is noted that Co and W concentrations in
the second IF-G lot are significantly lower than those originally reported for the first
IF-G lot, due to different crushing methods utilized during preparation of the two lots.
36

1
IF-G/shale
0.1
0.01
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
Fig 14. Rare earth and element and yttrium data from four separate HF-HClO 4 decompositions of the
new, second lot of the iron-formation IF-G as prepared and issued by IWG-GIT. Data normalized to
Post Archean Australian Shale (PAAS, McLennan, 1989). The majority of the REE data are consistent,
with exceptions for the light REE (La, Ce, Pr, and Nd).
As a consequence of the above events, most analyses performed at JUB of the
original IF-G lot used the pre-June 2006 ICPMS method, which quantified 24 trace
metals. Only two analyses of this original IF-G lot were performed using the current
32 element ICPMS method, whereas the second IF-G lot has been analyzed several
times using the 32 element method. Unfortunately, results for analyses of the second
IF-G lot are not entirely consistent, and suggest that heterogeneities may exist within
the sample powder. This is illustrated in Figure 14, which is a REE-yttrium (REY)
diagram depicting data from four separate HF-HClO 4 decompositions for the second
lot of IF-G powder. For the majority of the REE (Sm through Lu), the data for the
separate decompositions are quite consistent, whereas La, Ce, Pr, and Nd are much
more variable. This behavior is not well understood at this time. Fractionation across
the REE series for small aliquots of rock samples is often a function of mineralogical
controls on REE distributions within the bulk rock, and the data in Fig. 14 are
preliminarily interpreted to reflect small mineralogical heterogeneities in the second
lot of IF-G.
In consideration of the above observations, the discussion of analytical
accuracy is restricted to the two decompositions of the original lot of IF-G that were
37

1.50 IF-G (n=2)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 15. Accuracy estimation for ICPMS analysis of iron-formation IF-G, where n equals the number
of separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical lines
represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data are
presented in App. 1. Data for Sc not reported due to high IQL, and Nb not reported due to significant
FeCl interference. Reasons for poor observed Ti accuracy are considered to be similar to those proposed
for FeR-4 (see text for details). Lower JUB concentrations for Hf, Ta, Pb, and Th are not considered
anomalous, as it is likely that these elements are present at concentrations lower than originally reported
for IF-G, an observation supported by the data of Dulski et al. (2001) and Bohlar et al. (2004).
performed using the current 32 element ICPMS analysis (see App. 1). Results from
these analyses are presented in Figure 15, and JUB data for 23 of 32 elements are
between 90–100% of the average reference values. Data for Sc and Nb are not shown,
as Sc concentrations in IF-G (~0.3 mg/kg) are below the IQL for Sc (Fig. 4), and
literature Nb concentrations of ~0.1 mg/kg are significantly lower than the predicted
FeCl interferences on 93 Nb of ~0.6 mg/kg (Apps. 1 and 2).
The apparent poor accuracy of the JUB determination for Ti is considered to
arise from reasons similar to those proposed for Ti determinations in FeR-4, i.e., a
combination of low concentration (84 mg/kg) and analytical methodology for
determination of the reference values (e.g., sample fusion and XRF analyses). The
wide discrepancy observed for the Hf, Ta, Pb, and Th data of this study is primarily
due to the inclusion of values from Govindaraju (1994) when calculating the
reference average for these elements. Compared to more recent ICPMS analyses by
Dulski (2001) and Bohlar et al. (2004), JUB data for these elements are quite
38

consistent (App. 1), and similar to FeR-4, it is concluded that older data compilations
frequently overestimate the abundance of many trace elements that are present at very
low concentrations.
7.3. Shales and clastic sediments
Fine-grained clastic rocks and sediments commonly analyzed during
geochemical research at JUB include shales and river sediments. Shales and clastic
sediments may have Al 2 O 3 concentrations as high as ~15%, and the high
aluminosilicate mineral content of these samples produces a relative enrichment in
many trace metals. For example, the REY are incompatible trace metals that behave
similarly to Al in crustal processes, and the REY are good proxies for the Al-rich
continental crust that erodes to form river sediments and shales. In SCo-1, a shale
CRM issued by the USGS, the summed REY concentrations are approximately 170
mg/kg, whereas the IF-G iron-formation discussed above contains ~22 mg/kg of
REY.
This trace metal enrichment facilitates geochemical analyses of shales and
clastic sediments in two ways, with the first being that reference values are better
constrained, as older, less sensitive analytical methods can produce precise and
accurate data for many trace metals. The second advantage arises from the ability to
significantly dilute shale and sediment samples for ICPMS analyses while remaining
well above the IQLs for many trace metals, thereby reducing matrix effects and major
element interferences. This latter effect is illustrated using SCo-1 and IF-G, which, as
mentioned above, have very different REY concentrations. Assuming SCo-1 was
diluted by a factor 7-8 times greater than the dilution factor used for IF-G, then
similar ICPMS signal intensities would be expected for the REY when analyzing
these two CRMs.
SCo-1
The SCo-1 shale CRM (Cody shale) contains 13.7% Al 2 O 3 , 5.14% Fe 2 O 3 ,
approximately 2.6% each MgO and CaO, and only trace amounts of Mn (410 mg/kg).
Figure 16 depicts the analytical accuracy observed for SCo-1. Of the 32 elements
analyzed, 27 are within ±10% of the reference value, with lower masses displaying
the greatest deviation, as the JUB data are significantly higher for these elements (Sc,
Ti, Co, Ni). Some of the discrepancy observed for Sc, Ti, Co, and Ni can be ascribed
39

1.50 SCo-1 (n=4)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 16. Accuracy estimation for ICPMS analysis of shale SCo-1, where n equals the number of
separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical lines
represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data are
presented in App. 1. The triangles for Sc, Ti, Co, and Ni represent average JUB data that excludes a
single analysis that determined anomalously high values for these four elements, and when the
‘anomalous’ data are not considered the JUB averages more closely match the reference values for Sc,
Ti, Co, and Ni.
to a single JUB analysis of SCo-1 that determined much higher concentrations for
these elements, and when these ‘anomalous’ data are not considered (see Fig. 16), the
measured concentrations for 30 of 32 analyzed elements are between 90–110% of the
reference value.
Potential major element interferences on trace metal determinations in SCo-1
would include 27 Al 35 Cl on 62 Ni, and FeCl on 93 Nb (Table 4). The impact of Alchloride
interferences on 62 Ni is significant enough that it is recommended that 62 Ni
not be used for Ni determinations in shales (or sediments) that contain more that a
few percent Al 2 O 3 and that are diluted in HCl (see App. 2). Note that the Ni data
presented in App. 1 and Fig. 16 reflect concentrations for SCo-1 that were determined
solely by monitoring of 60 Ni. As for Nb, the presence of ~5% Fe 2 O 3 in SCo-1 would
be predicted to increase Nb concentrations by roughly 0.1 mg/kg, which is
insignificant when compared to the SCo-1 reference Nb value of 11.9 mg/kg. It is
therefore concluded that Fe 2 O 3 concentrations in shales (or sediments) of 3-7% are
40

1.50 SGR-1b (n=3)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 17. Accuracy estimation for ICPMS analysis of shale SGR-1b, where n equals the number of
separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical lines
represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data are
presented in App. 1. The measured concentrations for Ni, Y, Tb, and Tm are not necessarily considered
anomalous, and likely represent small interferences (Ni) and relatively large uncertainties in the
reference data (Y, Tb, Tm; see text and App. 1).
not likely to adversely affect Nb determinations, assuming Nb is present in the sample
at concentrations of several mg/kg or higher.
SGR-1b
The SGR-1b (Green River) shale is a petroleum and carbonate rich CRM
issued by the USGS. The SGR-1b shale is identical to the original SGR-1 shale (S.A.
Wilson, USGS, personal communication) and contains 6.5% Al 2 O 3 , 4.4% MgO, 8.4%
CaO, 3.0% Fe 2 O 3 , and minor amounts of Mn (267 mg/kg). While the Al content of
SGR-1b is low compared to the SCo-1 shale, it is still high enough to produce
significant Al-chloride interferences on 62 Ni. As a result Ni concentration data
reported for SGR-1b are determined solely from analyses of the 60 Ni isotope. For the
remaining major elements, significantly higher MgO and CaO in SGR-1b compared
to SCo-1 would be expected to erroneously increase measured signal intensities for
60 Ni, due to MgCl and CaO(H) inteferences. The combined magnitude of these Mg
and Ca interferences on 60 Ni is estimated to be ~2 mg/kg (App. 2), and this is
41

consistent with the average Ni concentration for SGR-1b measured at JUB of 30.3
mg/kg, which is slightly higher than the Ni reference average of 28 mg/kg (App. 1).
Figure 17 compares measured JUB trace element data with reference values,
and 28 of the 32 analyzed elements are within ±10% of the average reference value.
Measured concentrations for Y, Tb, and Tm are 10-20% lower than the average
reference value, whereas measured Nb concentrations are >20% higher than the
literature Nb values. The discrepancies for Y, Tb, and Tm may arise from the wide
range observed in reference values, and/or the aforementioned higher trace metal
contents frequently reported by compilations of older data, as concentrations for these
elements as reported by Dulski (2001) are indistinguishable from the JUB data (App.
1). The higher Nb content in SGR-1b as determined at JUB is unlikely to result from
FeCl interferences (see above discussion for SCo-1), and is not necessarily anomalous
considering that literature Nb values solely reflect older data compilations.
7.4. High Si content rocks (cherts)
Occasionally, trace metal analyses are required for rocks which are almost
exclusively composed of quartz (SiO 2 ). Examples of such rocks include microcrystalline
SiO 2 chemical precipitates (cherts), and relatively pure quartz clastic
sediments (e.g., sandstones). To date, little research has been performed at JUB
concerning quartz-dominated clastic rocks, and therefore this section focuses on very
pure SiO 2 chemical precipitates. These cherts are similar to the Si-rich, Fe-poor ironformations
discussed earlier, though for this discussion the term chert refers to a SiO 2
chemical precipitate that contains very little (≤0.5%) Al 2 O 3 , MgO, and CaO.
Concentrations of Fe 2 O 3 may range between 1-5%, and as Fe contents increase
beyond this range the previous discussion regarding iron-formations is considered
more appropriate.
The only chert CRM analyzed within the JUB Geochemistry Lab is JCh-1
(Geological Survey of Japan), which is 98% SiO 2 , 0.73% Al 2 O 3 ,

1.50
JCh-1 (n=3)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 18. Accuracy estimation for ICPMS analysis of chert JCh-1, where n equals the number of
separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical lines
represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data are
presented in App. 1. The JUB data are quite consistent and reproducible, with the exception of Ni. The
reference data for higher mass elements have a wide range of values (i.e., larger grey bars), with JUB
data generally lower. Considering the uncertainty in the reference data, and the very low trace metal
contents, the JUB data have additionally been normalized to the data of Dulski (2001), who has
published the only complete REY dataset for JCh-1 (App. 1). The JUB/Dulski (2001) values are plotted
as white triangles, and the majority of these ratios are close to 1.
with measured concentrations for many metals significantly lower than reference data.
However, the JUB data tend to be much more precise, particularly for higher mass
elements. The exception is Ni, and the large variation in Ni measured at JUB results
from one analysis, which determined a concentration of 14.0 mg/kg. The remaining
two analyses at JUB determined Ni of 7.41 and 8.59 mg/kg, similar to the average
reference value of 8.13 mg/kg (App. 1), and it is concluded that the single, 14.0 mg/kg
Ni analysis is likely anomalous.
The large uncertainty in the average reference values for many elements may
be due to the difficulty in determining the very low trace metal contents in JCh-1, and
few reference data exist for some elements (e.g., the monoisotopic REE Pr, Tb, Ho,
and Tm). One of the most complete trace element datasets yet published for JCh-1 is
from Dulski (2001), whose data compare favorably with the concentrations
determined at JUB, and these data are presented in Fig. 18.
43

7.5. Carbonate rocks
A common type of rock analyzed at JUB are carbonates, which are chemical
precipitates generally containing low trace metal contents, similar to cherts. Carbonate
rocks are dominated by (Ca,Mg)CO 3 minerals, and Ca and Mg may exist in solid
solution within the carbonate mineral structure. As a result, carbonate rock samples
typically analyzed may range from limestones (CaCO 3 ), through dolomites
((Ca,Mg)CO 3 ), to pure magnesites (MgCO 3 ). All of these carbonate rock types are
effectively decomposed using the HNO 3 carbonate decomposition method, though, as
noted previously, any accessory aluminosilicate minerals and refractory organic
carbon are relatively unaffected by dissolution with HNO 3 .
This section discusses the application of both decomposition methods (HF-
HClO 4 and HNO 3 ) to carbonate rocks, and describes the applicability and accuracy of
these methods with respect to specific elements of interest. As the literature reference
values are generally determined by analytical methods capable of thoroughly
characterizing the whole-rock abundance of elements, the inability of the carbonate
decomposition method to dissolve refractory silicate minerals results in ‘poor’
analytical results for elements typically hosted within these minerals (e.g., Zr, Nb, Hf,
Ta, Th, among others). It is therefore necessary to first examine data obtained using
the HF-HClO 4 whole-rock decomposition method, as these data are most comparable
to the average reference value calculated from literature sources.
The most commonly utilized carbonate CRM is the dolomite JDo-1, issued by
the Geological Survey of Japan (GSJ). The JDo-1 dolomite is 34.0% CaO and 18.5%
MgO, with only trace amounts of Al, Mn, or Fe. Therefore, major element
interferences would be expected from the high Ca and Mg contents, and should
primarily affect determinations of Co, Ni, Zr, and Nb. Figure 19 presents comparisons
between concentrations determined at JUB using the HF-HClO 4 decomposition
method with the average reference values. Determinations of Co and Ni are not shown
in Fig. 19, as the elements are severely compromised in HCl acid matrices by
interferences generated from Mg and Ca (Table 4). The JUB measured Co of 1.46
mg/kg is several times higher than the average reference value of 0.234 mg/kg, and
measured Ni is 11.1 mg/kg, also several times higher than the reference value of 2.9
mg/kg.
Of the remaining elements Ti, Sr, Y, the REE, and W show good agreement
with the average reference values, and the reference values themselves are consistent
44

1.50 JDo-1 (n=5)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 19. Accuracy estimation for ICPMS analysis of dolomite JDo-1, where n equals the number of
separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical lines
represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data are
presented in App. 1. Data for Co and Ni not shown due to large MgCl and CaO(H) interferences. The
reference averages for several elements display very large uncertainties (Sc, Rb, Zr, Ba, and Pb, see
App. 1). The JUB data for Sc is close to the IQL (Fig. 4), and Sc measurements at the concentration
levels reported for JDo-1 are unlikely to be reliable.
and calculated from a relatively large number of published studies (App. 1).
Exceptions are Ti and W, as previously published data are only available from Imai et
al. (1996), with Ti reported as 0.00133% TiO 2 . The observed discrepancies for other
elements are variously attributed to either anomalously high reference data from
individual studies that skew the average reference value (Sc, Rb, Zr, and Ba), or to a
lack of published reference data (Nb, Cs, Ta). For Sc, concentrations in JDo-1 are
similar to the IQL in 0.5 M HCl (Table 4), indicating that accurate Sc measurements
in JDo-1 are likely to prove difficult. With respect to Zr and Ba the measured JUB
concentrations agree well with at least two other published studies (see App. 1).
The elements which display the greatest discrepancies, either as uncertainty in
reference values or as deviation of the measured JUB concentration from the average
of published data, are generally incompatible in carbonate minerals. As the JDo-1
dolomite is a very pure marine chemical precipitate, these incompatible elements
(e.g., Sc, Rb, Zr, Nb, Cs, Hf, Ta, Th) are likely to be concentrated in discrete,
45

1.50
HF-HClO 4
carbonate (HNO 3
)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 20. Comparison of analytical accuracy relative to average reference values for both HF-HClO 4
and carbonate (HNO 3 ) decompositions of the JDo-1 dolomite. Based upon similar results for the two
decomposition methods, Sr, Y, the REE, W, Th, and U are considered to be quantifiable using the
carbonate decomposition method. Additionally, note the excellent results for Ni determined from the
carbonate decomposition method, which were calculated using the 62 Ni isotope. Elements likely to be
hosted in refractory aluminosilicate phases that resist complete dissolution using the carbonate
decomposition include Ti, Rb, Zr, Nb, Cs, Hf, and Ta.
refractory mineral phases. If that is the case, then these refractory minerals may not be
homogenously distributed within the JDo-1 powder, offering a possible explanation
for the wide variation in reported concentrations for incompatible elements.
Compared to the HF-HClO 4 decomposition method the carbonate
decomposition method is not expected to completely dissolve all mineral phases (e.g.,
aluminosilicates), and data for JDo-1 obtained using the two different decomposition
methods are presented in Figure 20. Regarding the carbonate decomposition of JDo-1,
and many other carbonate samples as well, a black residue is commonly filtered out of
the sample solution, and this residue is presumed to be refractory organic carbon. The
black solid retained during the filtering step does not appear to be of significance for
the determination of Ni, Sr, Y, the REE, W, Th, and U, and it is concluded that these
elements are hosted primarily by carbonate minerals, and therefore are quantifiable
using the carbonate decomposition method. The accurate quantification of Ni using
the carbonate method is notable, as this is only possible in an HNO 3 acid matrix and
by monitoring the 62 Ni isotope. Unlike 60 Ni, which is generally not suitable for Ni
46

determinations in carbonate rocks regardless of the acid matrix, concentration
determinations using 62 Ni provide excellent results for samples that are diluted in
HNO 3 , primarily through a reduction in the 25 Mg 37 Cl interference.
Elements that appear to be hosted primarily in refractory mineral phases
resistant to decomposition using the HNO 3 carbonate method include Ti, Rb, Zr, Nb,
Cs, Hf, and Ta, which would be expected for these incompatible elements. Two
elements that may be concentrated in different mineral phases are Ba and Pb (C- and
S-rich minerals?), for which the carbonate decomposition recovers approximately
90% and 80%, respectively, of the concentrations observed for the HF-HClO 4
decomposition method.
As stated above, it appears that the carbonate decomposition provides
satisfactory results for W. However, W is one of three elements apparently ‘lost’
during the analytical recovery test of the carbonate decomposition method, along with
Ta, and to a lesser extent Nb (see Section 6, Fig. 10). The low analytical recovery for
Nb, Ta, and W are attributed to the filtering step of the carbonate decomposition. For
Nb and Ta, the poor analytical recoveries during the carbonate decomposition method
are inconsequential, as regardless, sample dissolution in HNO 3 is not expected to
provide accurate determinations of these elements. In the case of W, the observed loss
during the filtering step is estimated at ~0.002 mg/kg (Fig. 10), which is also
inconsequential when considering the reference concentration of W in the JDo-1
dolomite of ~0.260 mg/kg. It is therefore unlikely that any loss of W during the
filtering step for the carbonate decomposition would significantly affect W
determinations, assuming W is present at concentrations ≥0.050 mg/kg in the sample
powder. This is particularly true when considering the uncertainty in the JUB
measured W data for JDo-1 (~15% RSD).
7.6. Marine ferromanganese nodules and crusts
Samples which comprise a large fraction of the geochemical research
performed with the Geochemistry Lab at JUB are marine ferromanganese nodules and
crusts (referred to collectively as Fe-Mn crusts). These Fe-Mn crusts are typically 13-
15% SiO 2 , 3-5% Al 2 O 3 , ~3% each CaO and MgO, 33-38% MnO, and 8-15% Fe 2 O 3 .
Therefore, the high metal content and interferences from Mn and Fe are expected to
provide the greatest obstacles to accurate trace metal determinations. However, unlike
other metal-rich rocks such as iron-formations, Fe-Mn crusts are highly enriched in
47

many trace metals, and therefore may be diluted significantly following
decomposition. For example, the combined REY concentration within the Fe-Mn
crust JMn-1 is >800 mg/kg, whereas the iron-formation IF-G contains only ~22 mg/kg
REY. While IF samples typically may not be diluted by a factor greater than 1000, Fe-
Mn crust analyses may benefit from higher dilution factors (≥5000).
Several Fe-Mn crust CRMs are available for geochemical studies at JUB, and
include JMn-1 (Table 1), NOD-A-1 and NOD-P-1 (USGS), and GSPN-2 and GSPN-3
(Institute of Rock and Mineral Analysis, People’s Republic of China). For this
discussion only the JMn-1 ferromanganese crust issued by the Geological Survey of
Japan is considered, though additional information regarding analyses at JUB of other
Fe-Mn crust CRMs is available from K. Schmidt.
Figure 21 displays comparisons between measured JUB trace metal
concentrations in JMn-1 with the average reference value. Measured concentrations
are lower than the average reference value for all analyzed elements, with JUB data
for 24 of 32 elements between 80-95% of the reference value. The precision of the
JUB is generally quite good (2-3 % RSD), suggesting that the low measured values
are reproducible and do not result from a single anomalous analysis. As discussed
earlier, at high TDS contents samples may suffer from signal suppression effects
(Section 4.2), which may result in low measured concentrations. However, this does
not appear to be the case with JMn-1, as all of the JUB analyses were performed at
dilution factors of 2500-5000, and the internal standard correction factors for these
analyses were typically between 0.90–1.10.
Rather, it is hypothesized that absorption of water vapor by JMn-1 sample
powder is primarily responsible for the low measured concentrations. For this study,
the JMn-1 powder was not dried prior to analysis, as for most CRMs drying the rock
powder (e.g., overnight at 105-110°C) has not proven to significantly affect the
measured concentrations of the elements of interest. However, Fe-Mn crusts are
microcrystalline marine precipitates composed of poorly ordered mineral phases that,
unlike other rock CRMs, have never undergone mineral recrystallization during
diagenesis and lithification. As a result it is possible that powdered Fe-Mn crusts may
more readily absorb water from ambient laboratory air (even though all CRMs are
stored in dessicators until ready for use), and any increase in sample mass due to this
effect would drive measured concentrations of trace metals towards lower values.
This hypothesis is supported by detailed studies of Fe-Mn crusts by K. Schmidt
48

1.50
JMn-1 (n=4)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 21. Accuracy estimation for ICPMS analysis of Fe-Mn nodule JMn-1, where n equals the
number of separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars
represent variability in the calculated ratio that is due solely to uncertainty in the reference average.
Vertical lines represent the variability in the calculated ratio due to the uncertainty of the JUB data, and
all data are presented in App. 1. The generally low concentrations for many elements as measured at
JUB are hypothesized to result from the adsorption of water vapor by the JMn-1 powder, which would
effectively dilute the sample and lower the measured concentrations (see text for details).
(personal communication), who observed that concentrations measured in non-dried
JMn-1 powders were typically 10-20% lower than concentrations determined for
dried JMn-1 powders.
7.7. Basalts
The last rock type frequently used in geochemical studies at JUB is basalt,
which is representative of oceanic crust. Basaltic rocks are relatively Al-poor, Mgand
Fe-rich igneous rocks that have formed throughout Earth’s history, and have been
the focus of numerous geochemical studies related to crustal differentiation, hot spot
volcanism, seawater-crust interaction, etc. As a result of this research attention
basaltic CRMs are very well characterized, with numerous published studies reporting
data for a large number of trace metals.
One of the best characterized basalt CRMs is BHVO-2, a Hawaiian basalt
issued by the USGS. The BHVO-2 basalt is 13.5% Al 2 O 3 , 7.2% MgO, 11.4% CaO,
0.17% MnO, and 12.3% Fe 2 O 3 . Based on these major element abundances,
49

1.50 BHVO-2 (n=7)
1.25
JUB / reference average
1.00
0.75
0.50
0.25
0.00
Sc Ti Co Ni Rb Sr Y Zr Nb Mo Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta W Pb Th U
Figure 22. Accuracy estimation for ICPMS analysis of basalt BHVO-2, where n equals the number of
separate HF-HClO 4 sample decompositions and ICPMS analyses in 0.5 M HCl. Grey bars represent
variability in the calculated ratio that is due solely to uncertainty in the reference average. Vertical
lines represent the variability in the calculated ratio due to the uncertainty of the JUB data, and all data
are presented in App. 1. Of all the CRMs analyzed at JUB, data for BHVO-2 basalt most closely
matches the average reference values, and does so for the greatest number of elements.
interferences would be expected on Co and Ni (from MgCl, AlCl, and CaO(H)), and
to a lesser extent on Nb from FeCl. However, concentrations of Co, Ni, and Nb in
BHVO-2 are significant (46.0, 119, and 17.5 mg/kg, respectively), and major element
interferences do not appear to affect analytical accuracy for Co, Ni, and Nb (Figure
22).
Of the various CRMs analyzed at JUB, data for BHVO-2 most closely
matches the average reference values, and does so for the greatest number of
elements. Of the 32 elements measured, 30 are within ±10% of the reference value,
and 26 are within ±5% (Fig. 22). The excellent agreement between measured and
reference data is attributed to the fact that BHVO-2 contains significantly higher
concentrations of many elements that display ‘poor’ accuracy for a number of CRMs
(e.g., Co, Ni, Zr, Nb, Hf, Ta, Th), and these higher concentrations facilitate
geochemical analyses. Additionally, the numerous published studies for a large
number of trace metals in BHVO-2 ensures a robust dataset for calculating an average
50

eference value. The result is that for many elements the uncertainty in the average
reference value is similar to that observed for the JUB data (2-5% RSD, App. 1).
The close agreement illustrated for most elements in Fig. 22 is not observed
for Ti in BHVO-2, which has consistently produced low measured values in the JUB
data compared to the compiled literature value (App. 1). The reason for this behavior
is not yet clear, yet it is hypothesized that it may result from a combination of the high
Ti content in BHVO-2 and the relatively low 10 μg/kg Ti calibration standard used for
determination of Ti concentrations. The ionization efficiency of Ti is quite low, and
the 10 μg/kg Ti calibration standard produces an ICPMS signal intensity of only
8000-10,000 cps, relative to rather high blank intensities of 500-800 cps due to CCl
and NO 2 interferences (Table 4). In comparison, the abundance of Ti in BHVO-2
routinely produces 7-9 million cps at dilution factors of 1000. A benefit of ICPMS
instruments is that they frequently offer linear signal responses for many elements at
concentration ranges that exceed six orders of magnitude. However, in the case of Ti,
the combination of blank intensities roughly one order of magnitude less than the
calibration standard intensities, which are themselves approximately three orders of
magnitude less than the sample intensity, may produce a non-linear instrument
response. The fact that good results are observed in those CRMs that contain ~1000-
1500 mg/kg Ti (FeR-2 and SGR-1b), with poorer results at higher Ti abundances
(e.g., >6000 mg/kg as in JMn-1), offers indirect support of this hypothesis. However,
further investigation is warranted, and an approach of higher dilution factors for
BHVO-2 (2500-5000), coupled with an increase in the Ti concentration of the
calibration standard (50 μg/kg?) is recommended.
7.8. Rare earth element ratios
A significant emphasis of much of the research conducted within the Jacobs
University Geochemistry Lab regards the behavior of the rare earth elements and
yttrium in geological samples, including rocks and minerals, as well as natural water
samples such as seawater, hydrothermal fluids, and ground and river waters.
Analytical studies of the REY are generally more focused upon the relative
distribution of these elements within any given sample, rather than their
concentrations, and the REY plot in Fig. 14 is a typical manner of presenting REY
data. Frequently, the shape of REY patterns for interpreting geological samples are
51

Table 5. Select REY ratios and chondrite-normalized REE
anomalies as calculated from reference data and JUB data for
BHVO-2 basalt.
REY ratio avg. ref. data 1 JUB data JUB/ref. avg.
Pr/Sm 0.8885 0.8691 0.978
Pr/Dy 1.017 1.000 0.983
Pr/Yb 2.711 2.726 1.006
Sm/Dy 1.145 1.151 1.005
Sm/Yb 3.051 3.137 1.028
Dy/Yb 2.665 2.726 1.023
Y/Ho 26.44 25.31 0.957
Ce/Ce* 2 1.00 1.01 0.99
Ce/Ce* 3 0.933 0.964 0.97
Eu/Eu* 4 1.00 1.01 0.99
Eu/Eu* 5 1.01 1.02 0.99
1 average reference data (see Appendix 1).
2 calculated as Ce/Ce* = Ce/(0.5La+0.5Pr).
3 calculated as Ce/Ce* = Ce/(2Pr-1Nd) after Bolhar et al. (2004).
4 calculated as Eu/Eu* = Eu/(0.5Sm+0.5Gd).
5 calculated as Eu/Eu* = Eu/(0.67Sm+0.33Tb).
more useful than the absolute REY abundances, because fractionation between REY
elements (and the specific REY patterns that result), may offer clues to the processes
responsible for the formation and subsequent history of many geological samples.
Since the relative distribution of the REY is often more informative for
interpreting geochemical data, it is useful to examine the ability of the described
analytical methods in accurately determining the REY patterns in geological
materials. As this study has focused upon the analyses of CRMs, and the BHVO-2
basalt is considered the best characterized reference material with respect to the REY,
the discussion is limited to relative REY ratios in BHVO-2.
Table 5 lists REY ratios that characterize the relative distributions of the light
rare earth elements (Pr/Sm), the middle rare earth elements (Sm/Dy), and the heavy
rare earth elements (Dy/Yb) for the BHVO-2 basalt. Ratios that include La, Ce, Gd,
Eu, and Lu are not used, as these elements may display anomalous behavior in some
geological samples (e.g., La and Ce in seawater, and Eu in hydrothermal fluids and
plagioclase-rich rocks). Data used for calculating all REY ratios are tabulated in
Appendix 1. No consideration of the relative errors in determining the concentrations
of individual REY elements as reported in Appendix 1 is present in the data of Table
5, as for many of the elements the %RSD is very similar between the average
52

eference value and the measured JUB data (e.g., 3.5% and 2.7% for Yb,
respectively). Relative to the REY ratios calculated from the average reference data,
the JUB ratios are accurate to within 1% for Pr/Yb, 2% for Pr/Dy, 3% for Pr/Sm,
Sm/Yb, and Dy/Yb, and to within 5% for Y/Ho. Considering the well-characterized
nature of the BHVO-2 basalt and the large number of highly precise ICPMS analyses
reported for this CRM, these values are considered the best estimates of the accuracy
of the JUB analytical methods in determining REY ratios.
Also presented in Table 5 are examples of calculated REE anomalies for
chondrite-normalized data of Ce and Eu, two rare earth elements that may be
fractionated from neighboring REE in natural systems (e.g., in seawater for Ce, and in
high-temperature hydrothermal systems for Eu). Normalized REE data that is
anomalous is commonly quantified by means of X/X* ratios, where X is the
normalized measured concentration for a particular rare earth element, and X* is the
predicted normalized concentration for that element. The predicted REE concentration
X* may be calculated in various ways (e.g., Bolhar et al., 2004), and X* is commonly
obtained by interpolating between adjacent REE, as in the calculation of Eu* by
means of (0.5Sm + 0.5Gd). Alternatively, X* may be obtained by extrapolating from
adjacent REE, as in the case of Ce* by means of (2Pr - 1Nd). Different methods for
calculating chondrite-normalized Ce/Ce* and Eu/Eu* are presented in Table 5, and
demonstrate that REE anomalies calculated from measured JUB data accurately
reproduce the anomalies as determined from the average reference data of the BHVO-
2 basalt.
8. Summary and conclusions
The ICPMS analytical methods utilized within the JUB Geochemistry Lab are
capable of reproducibly determining accurate concentrations of many trace metals in
a variety of rock types. The relative precision of the ICPMS measurements is
primarily controlled by the inherent ICPMS instrument precision, and not by sample
preparation and decomposition techniques. On average, for all rock types
decomposed using the HF-HClO 4 decomposition method, the ICPMS instrument
precision (RSD) is better than 3% for eleven elements (Ti, Co, Sr, Y, Zr, Ba, La, Ce,
Pr, Nd, Pb), between 3-5% for sixteen elements (Sc, Ni, Rb, Nb, Sm, Eu, Gd, Tb, Dy,
Ho, Er, Tm, Yb, Lu, Th, U), and between 5-8% for only five elements (Mo, Cs, Hf,
Ta, W). Certainly, the poorer precision observed for some elements is due to low
53

concentrations of these elements in many rock types (e.g., Cs and Hf in JDo-1),
and/or possible inhomogeneities in the CRM sample powders.
The sample decomposition methods employed appear to conservatively retain
all 32 elements analyzed throughout the dissolution procedures, with the only
exceptions being Ni, Ta, and W during the carbonate decomposition. Greater than
expected recoveries of Ni in spike solutions during the carbonate decomposition
suggest a Ni contamination of 2-5 μg/kg occurs, likely during filtration of the
dissolved sample solution. Analytical recoveries of Ta and W are low (24% and 80%,
respectively), and observed to a lesser extent for Nb (92% recovery), and appear to
reflect retention of these elements on the 0.2 μm cellulose acetate filters used during
the carbonate decomposition.
Some analyzed elements are significantly affected by interferences generated
by high major element contents in certain rock types. The affected elements have
atomic masses 10 mg/kg). Determination of low Nb contents (less than ~20 mg/kg) in Ferich
samples is also likely to prove difficult, and Appendix 2 provides a thorough
treatment of observed major element interferences and recommendations regarding
analyses of specific rock types. The potential for reducing chloride-derived
interferences that arise from the use of HCl during sample dilutions should be
investigated more thoroughly. Diluting samples prior to ICPMS analyses in mixtures
of HNO 3 and HCl may satisfactorily mitigate many major element interferences on
low mass trace metals such as Co, Ni, Zr, Nb, and Mo.
The analytical accuracy of the ICPMS measurements is considered excellent,
though accuracy assessments are hampered by the lack of well constrained reference
values for many trace metals in the CRMs discussed. It appears that older data
compilations frequently overestimate the abundances of many trace metals, and for
some elements very few published data exist (e.g., Nb, Mo, Ta, W). The best
constrained CRM is the BHVO-2 basalt, for which the data produced at JUB is in
excellent agreement. Based upon multiple analyses of BHVO-2 it is concluded that
the JUB methods described here provide accurate element determinations for 31 of
54

the 32 analyzed elements, with Ti being the only exception. Element concentrations
in CRMs measured at JUB also agree well with recent published studies that focused
on similar trace metals, and that used similar analytical methods. While the task of
ensuring accuracy in analytical data is a continuous process, it is concluded that the
ICPMS methods described here are suitable for reproducibly determining highly
precise and accurate trace element data in a variety of geologic materials.
55

References
Abbey S., McLeod C.R., and Liang-Guo W. (1983) FeR-1, FeR-2, FeR-3, and FeR-4 four
Canadian iron-formation samples prepared for use as reference materials. Geol. Sur.
Canada paper 83-19.
Baker J.A., Waight T., and Ulfbeck D. (2002) Rapid and highly reproducible analysis of rare
earth elements by multiple collector inductively coupled plasma mass spectrometry.
Geochim. Cosmochim. Acta, 66, 3635-3646.
Barrat J.A., Boulègue J.J., Tiercelin J.-J., and Lesourd M. (2000) Strontium isotopes and rareearth
element geochemistry of hydrothermal carbonate deposits from Lake
Tanganyika, East Africa. Geochim. Cosmochim. Acta, 64, 287-298.
Bédard L.P. and Barnes S.-J. (2002) A comparison of N-type semi-planar and coaxial INAA
detectors for 33 geochemical reference samples. J. Radioanalytical Nuclear Chem.,
254, 485-497.
Bolhar R., Kamber B.S., Moorbath S., Fedo C.M., and Whitehouse M.J. (2004)
Characterisation of early Archaean chemical sediments by trace element signatures.
Earth Planet. Sci. Lett., 222, 43-60.
Dai Kin F., Prudêncio M.I., Gouveia M.A., and Magnusson E. (1999) Determination of rare
earth elements in geological reference materials: A comparative study by INAA and
ICP-MS. Geostand. Newsletter, 23, 47-58.
Doherty W. (1989) An internal standardization procedure for the determination of yttrium and
the rare earth elements in geological materials by inductively coupled plasma-mass
spectrometry. Spectrochim. Acta, 44B, 263-280.
Dulski P. (1994) Interferences of oxide, hydroxide and chloride analyte species in the
determination of rare earth elements in geological samples by inductively coupled
plasma-mass spectrometry. Fresenius J. Anal. Chem., 350, 194-203.
Dulski P. (2001) Reference materials for geochemical studies: new analytical data by ICP-MS
and critical discussion of reference values. Geostand. Newsletter, 25, 87-125.
Eggins S.M. (2003) Laser ablation ICP-MS analysis of geological materials prepared as
lithium borate glasses. Geostand. Newsletter, 27, 147-162.
Faure G. (1986) Principles of Isotope Geology. John Wiley and Sons, New York.
Flem B. and Bédard L.P. (2002) Determination of trace elements in BCS CRM 313/1 (BAS)
and NIST SRM 1830 by inductively coupled plasma-mass spectrometry and
instrumental neutron activation analysis. Geostand. Newsletter, 26, 287-300.
Garbe-Schönberg C.-D. (1993) Simultaneous determination of thirty-seven trace elements in
twenty-eight international rock standards by ICP-MS. Geostand. Newsletter, 17, 81-
97.
Govindaraju K. (1994) 1994 compilation of working values and sample descriptions for 383
geostandards. Geostand. Newsletter, 18, 1-158.
56

Govindaraju K. (1995) 1995 working values with confidence limits for twenty-six CRPG,
ANRT and IWG-GIT geostandards. Geostand. Newsletter, 19 (special), 1-95.
Huang S. and Frey F.A.(2003) Trace element abundances of Mauna Kea basalt from phase 2
of the Hawaii Scientific Drilling Project: Petrogenetic implications of correlations
with major element content and isotopic ratios. Geochem. Geophys. Geosys., 4, 1-43.
Imai N., Terashima S., Itoh S., and Ando A. (1996) 1996 compilation of analytical data on
nine GSJ geochemical reference samples: "Sedimentary Rock Series". Geostand.
Newsletter, 20, 165-216.
Imai N., Terashima S., Itoh S., and Ando A. (1999) 1998 compilation of analytical data for
five GSJ geochemical reference samples: The "Instrumental Analysis Series".
Geostand. Newsletter, 23, 223-250.
Jochum K.P., Nohl U., Herwig K., Lammel E., Stoll B., and Hofmann A.W. (2005) GeoReM:
A New Geochemical Database for Reference Materials and Isotopic Standards.
Geostand. Geoanalytical. Res, 29, 333-338.
Korotev R.L. (1996) A Self-Consistent compilation of elemental concentration data for 93
geochemical reference samples. Geostand. Newsletter, 20, 217-245.
MacDougall D., Crummett W.B., and et al. (1980) Guidelines for Data Acquisition and Data
Quality Evaluation in Environmental Chemistry. Anal. Chem., 52, 2242–2249.
McLennan S. (1989) Rare earth elements in sedimentary rocks: influence of provenance and
sedimentary processes. in Geochmemstry and Mineralogy of the Rare Earth Elements
(eds. B.R. Lipin and G.A. McKay). Reviews in Mineralogy, Vol. 21.
Meisel T., Schöner N., Paliulionyte V. and Kahr E. (2002) Determination of rare earth
elements, Y, Th, Zr, Hf, Nb and Ta in geological reference materials G-2, G-3, SCo-1
and WGB-1 by sodium peroxide sintering and inductively coupled plasma-mass
spectrometry. Geostand. Newsletter, 26, 53-61.
Münker C. (1998) Nb/Ta fractionation in a Cambrian arc/back arc system, New Zealand:
source constraints and application of refined ICPMS techniques. Chem. Geol., 144,
23-45.
Plumlee G. (1998) USGS Certificate of Analysis Basalt, Hawaiian Volcanic Observatory,
BHVO-2, USGS, Boulder CO.
Polat A., Li Jianghai, Fryer B.J., Kusky T., Gagnon J., and Zhang S. (2006) Geochemical
characteristics of the Neoarchean (2800-2700 Ma) Taishan greenstone belt, North
China Craton: Evidence for plume-craton interaction. Chem. Geol., 230, 60-87
Ryder C.H., Gill J., Tepley III F., Ramos F., and Reagan M. K. (2006) Closed- to opensystem-differentiation
at Arenal volcano (1968-2003). J. Volcano. Geotherm. Res.,
157, 75-93.
Schramm B., Devey C.W., Gillis K.M., and Lackschewitz K. (2005) Quantitative assessment
of chemical and mineralogical changes due to progressive low-temperature alteration
of East Pacific Rise basalts from 0 to 9 Ma. Chem. Geol., 218, 281-313.
57

Shafer J.T., Neal C.R., and Regelous M. (2005) Petrogenesis of Hawaiian postshield lavas:
Evidence from Nintoku Seamount, Emperor Seamount Chain. Geochem. Geophys.
Geosys., 6, Q05L09, doi:10.1029/2004GC000875.
Smith D.B. (1995) USGS Certificate of Analysis Cody Shale, SCo-1, United States Geologic
Survey, Boulder, CO.
Terashima S., Usui A., and Imai N. (1995) Two new GSJ geochemical reference samples:
Syenite JSy-1 and manganese nodule JMn-1. Geostand. Newsletter, 19, 221-229.
Thomas R. (2003) Practical Guide to ICPMS. Practical Spectroscopy Series, Taylor and
Francis, London.
Thompson J.J and Houk R.S. (1987) A Study of Internal Standardization in Inductively
Coupled Plasma-Mass Spectrometry. Appl. Spectroscopy, 41, 801-806.
Willbold M. and Jochum K.P. (2005) Multi-element isotope dilution sector field ICP-MS: A
precise technique for the analysis of geological materials and is application to
geological reference materials. Geostand. Geoanalytical. Res., 29, 63-82.
Wilson S.A.(2001) USGS Certificate of Analysis Green River Shale, SGR-1. United States
Geologic Survey, Boulder CO.
Yamamoto Koshi, Itoh N., Matsumoto T., Tanaka Tsuyoshi, and Adachi M. (2004)
Geochemistry of Precambrian carbonate intercalated in pillows and its host basalt:
implications for the REE composition of circa 3.4Ga seawater. Precam. Res., 135,
331-344.
Yu Z., Robinson P., and McGoldrick P. (2001) An evaluation of methods for the chemical
decomposition of geological materials for trace element determination using ICP-MS.
Geostand. Newsletter, 25, 199-217.
58

Appendix 1. Analytical data
Appendix 1 contains the literature data used to calculate the average CRM
reference values discussed in Section 7 regarding analytical accuracy. The average
measured trace metal concentrations as determined within the JUB Geochemistry Lab
for the various CRMs are also presented. As discussed within the main text of this
study, some isotopes with atomic masses

for JMn-1 as described in Section 7.6 preclude definitive statements as to the reason
for this discrepancy, and further studies of this effect in JMn-1 are recommended.
Concentrations of Zr as reported for all CRMs in this study have been
determined solely using 90 Zr. Interferences on the 91 Zr isotope primarily result from
56 M 35 Cl interferences, where M can be Fe, CaO(H), or ArO, and the use of HNO 3 for
diluting samples will reduce these interferences. Good results using 91 Zr in an HCl
acid matrix appear to be limited to samples which contain more than ~40 mg/kg Zr,
and that contain less than ~10% Fe 2 O 3 . The use of 91 Zr for quantifying Zr should be
used with caution, and is likely to prove most useful only for samples diluted in
HNO 3 , or for those samples diluted in HCl for which the Mn content is high enough
to produce significant 55 Mn 35 interferences that preclude the use of 90 Zr for
quantifying Zr (see App. 2).
60

Appendix 1. Literature reference values and measured element concentration data for CRMs analyzed within JUB Geochemistry Lab. Data in mg/kg.
Govindaraju
(1994)
Dulski
(2001)
Yu et al.
(2001)
FeR-2 FeR-4 IF-G 4
Abbey et al.
(1983)
this
study
Govindaraju
(1994)
Dulski
(2001)
Yu et al.
(2001)
Abbey et al.
(1983)
method 1 compiled ICPMS ICPMS compiled ref. avg. 2 %RSD JUB avg. 3
(n=3)
%RSD compiled ICPMS ICPMS compiled ref. avg. %RSD JUB avg.
(n=2)
%RSD compiled ICPMS ICPMS ID MC-
ICPMS
ICPMS ref. avg. %RSD JUB avg.
(n=2)
Sc 6 5.31 6 5.77 5.6 5.34 3.2 1.5 1.09 1.5 1.36 14.2 0.875 0.3 0.2846 0.2923 2.6

Appendix 1 continued. Data in mg/kg.
Smith
(1995)
Korotev
(1996)
Dai Kin
(1999)
SCo-1 SGR-1b JCh-1
Dulski
(2001)
Bédard
and Barnes
(2002)
Meisel
et al.
(2002)
Eggins
(2003)
this
study
Govindaraju Korotev Dai Kin
(1994) (1996) (1999)
Wilson
(2001)
Dulski
(2001)
this
study
Govindaraju
(1994)
Imai
et al.
(1996)
Dulski Flem and
(2001) Bédard
(2002)
method 1 compiled INAA ICPMS ICPMS INAA ICPMS LA- ref. avg. 2 %RSD JUB avg. 3 %RSD compiled INAA ICPMS compiled ICPMS ref. avg. %RSD JUB avg. %RSD compiled compiled ICPMS INAA ref. avg. %RSD JUB avg. %RSD
ICPMS
(n=4)
(n=3)
(n=3)
Sc 11 11.57 11.53 12.7 11.70 5.3 12.8 13.7 4.6 4.94 4.6 4.71 3.4 5.01 2.6 0.85 0.979 1 0.943 7.0 0.931 6.3
Ti 3776 3561 3669 2.9 4080 17.6 1580 1520 1550 1.9 1530 7.9 180 189 185 2.4 174 6.5
Co 11 11.13 11.6 11.24 2.3 12.7 15.1 11.8 11.73 12 11.84 1.0 12.2 3.3 15 15.5 14.4 15.0 3.0 16.5 5.5
Ni 27 24 26 5.9 30.3 22.6 29 26 29 28 5.1 30.3* 5.0 7.5 8.76 8.13 7.7 10.0 28.9
Rb 110 111.5 113 106 111.7 110.4 2.2 113 8.4 83 79 80 81 2.1 82.0 4.5 8.5 8.61 8.6 8.3 8.50 1.5 8.90 0.1
Sr 170 175 169 172.7 171.7 1.4 168 4.3 420 393 420 381 404 4.2 391 3.7 4.6 4.2 4.2 4.3 4.4 4.03 1.6
Y 26 22.8 24.2 26.8 25.0 6.2 22.8 4.4 13 13 9.6 11.9 13.5 9.74 2.2 1.84 1.81 1.8 1.82 0.9 1.77 1.6
Zr 160 166 153 116 165 188.6 158.1 13.8 157 6.6 53 45 53 42 48 10.1 43.7 3.5 11.7 11.5 6.2 9.8 26.0 6.60 3.2
Nb 11 11.6 13.2 11.9 7.8 11.8 4.3 5.2 5.2 5.2 0.0 6.41 1.1 1.7 1.7 0.590 1.5
Mo 1.4 1.4 1.22 5.7 35.1 35 35.1 0.1 34.1 9.2 0.245 10.5
Cs 7.8 7.77 7.7 7.4 7.46 7.63 2.2 7.67 4.1 5.2 5.19 5.2 5.1 5.17 0.8 5.13 2.2 0.3 0.243 0.288 0.29 0.280 7.8 0.277 5.7
Ba 570 559 578 567 592 573 2.0 562 4.3 290 270 290 265 279 4.1 278 1.6 302 302 291 298 1.7 276 4.5
La 30 29.3 28.9 29 30.3 29.7 31.1 29.8 2.4 28.1 3.9 20.3 18.4 18.2 20 18 19.0 5.1 18.2 2.8 1.5 1.52 1.44 1.44 1.48 2.4 1.34 4.8
Ce 62 56.7 57.3 57 60.3 57.5 58.8 58.5 3.1 56.1 3.7 36 34.4 34 36 33.8 34.8 2.8 34.0 3.3 4.72 5.21 4.7 4.66 4.82 4.7 4.42 4.5
Pr 6.6 6.7 7.1 7 6.9 3.0 6.69 3.6 3.9 3.85 4 3.92 1.6 3.91 2.7 0.5 0.37 0.44 14.9 0.335 4.6
Nd 26 25.4 28.4 26 24 27 27.3 26.3 5.0 25.7 3.0 15.5 13.8 13.7 16 13.8 14.6 6.8 14.2 2.4 1.7 2.05 1.41 1.4 1.64 16.2 1.32 4.4
Sm 5.14 5.2 5.0 5.12 5.31 5.52 5.22 3.2 4.96 3.5 2.7 2.6 1.9 2.7 2.42 2.46 12.2 2.52 2.0 0.4 0.359 0.30 0.35 0.352 10.1 0.281 4.0
Eu 1.088 1.3 1.13 1.11 1.17 1.18 1.163 5.9 1.10 3.4 0.56 0.466 0.48 0.56 0.47 0.507 8.5 0.466 2.1 0.08 0.0594 0.060 0.1 0.0749 22.3 0.0484 3.0
Gd 5.3 4.6 4.72 4.8 4.86 5.5 4.53 6.0 2 1.9 2 2.05 1.99 2.7 2.04 1.5 0.304 0.304 0.0 0.286 2.4
Tb 0.675 0.72 0.69 0.69 0.7 0.70 2.1 0.675 3.1 0.36 0.297 0.37 0.3 0.332 10.1 0.295 1.2 0.09 0.0385 0.046 0.04 0.0536 39.5 0.0438 2.5
Dy 4.2 4.0 4.3 4.43 4.23 3.7 4.01 2.7 1.9 1.8 1.9 1.71 1.83 4.3 1.73 0.9 0.4 0.378 0.289 0.356 13.5 0.273 2.4
Ho 0.76 0.8 0.86 0.81 5.1 0.803 2.5 0.38 0.38 0.4 0.34 0.38 5.8 0.347 0.7 0.1 0.060 0.080 25.0 0.0567 2.1
Er 2.2 2.38 2.53 2.68 2.45 7.3 2.35 1.6 1.11 1.2 1.1 0.99 1.10 6.8 1.01 1.0 0.3 0.233 0.174 0.236 21.8 0.167 1.8
Tm 0.4 0.35 0.36 0.37 5.8 0.334 4.8 0.17 0.18 0.17 0.147 0.167 7.3 0.144 1.6 0.025 0.025 0.0240 3.6
Yb 2.27 1.7 2.34 2.37 2.57 2.25 13.0 2.23 1.4 0.94 0.966 1.1 0.94 0.96 0.981 6.1 0.949 1.0 0.1 0.182 0.171 0.16 0.153 20.7 0.162 2.8
Lu 0.341 0.31 0.36 0.35 0.37 0.388 0.353 6.9 0.345 3.3 0.14 0.146 0.14 0.146 0.143 2.1 0.146 1.7 0.04 0.0344 0.026 0.023 0.0309 21.8 0.0269 6.6
Hf 4.75 4.3 4.4 4.8 5.1 4.67 6.2 4.10 5.9 1.39 1.371 1.4 1.32 1.370 2.2 1.30 2.8 0.2 0.195 0.15 0.09 0.159 27.8 0.173 2.0
Ta 0.804 0.9 0.84 0.921 0.866 5.4 0.803 0.5 0.42 0.402 0.411 2.2 0.404 1.6 0.182 0.11 0.146 24.7 0.135 5.8
W 1.4 1.5 1.5 3.4 1.53 1.9 2.57 2.3 2.6 2.49 5.4 2.52 0.3 92.3 92.3 84.7 2.3
Pb 31 30.0 33.7 31.6 5.0 29.7 2.4 38 38 42 39 4.8 40.5 0.3 2 2 2.0 2.0 0.0 1.63 0.9
Th 9.7 9.01 9.5 8.72 9.3 10.23 9.41 5.2 8.92 5.7 4.78 4.48 4.8 4.64 4.68 2.7 4.46 1.6 0.735 0.56 0.63 0.642 11.2 0.555 0.9
U 3 3.1 2.7 3.22 3.01 6.4 3.01 3.8 5.4 5.31 5.4 5.2 5.33 1.5 5.18 2.1 0.736 0.60 0.59 0.642 10.4 0.590 0.3
1 data sources, whether compiled from studies or from single analytical method, and arranged chronologically: INAA = instrumental neutron activation analysis, LA-ICPMS = laser ablation-ICPMS.
2 reference average reported to number of significant digits observed in most precisely measured data from literature sources.
3 JUB average calculated from number of sample decompositions reported in Table 1 and reported to three significant digits only for illustrative purposes (see Section 5 regarding analytical precision).
* values greater than reference average considered to include significant contributions from interfering molecular species, arising from high major element contents (Mg, Al, Ca, Mn, Fe).
this
study
62

Appendix 1 continued. Data in mg/kg.
JDo-1 JMn-1 BHVO-2
Garbe- Govindaraju Imai Dulski Yamamoto
this
this
Terashima Imai Dulski
this
Plumlee Huang Willbold Shafer Schramm Ryder Polat
this
Schönberg (1994) et al. (2001) (2004)
study
study
et al. et al. (2001)
study
(1998) and Frey and et al. et al. (2006) et al.
study
(1993)
(1996)
HF-HClO4 carbonate 4
(1995) (1999)
(2003) Jochum (2005) (2005)
(2006)
(2005)
method 1 ICPMS compiled compiled ICPMS ICPMS ref. avg. 2 %RSD JUB avg. 3 %RSD JUB avg. 3 %RSD ICPMS compiled ICPMS ref. avg. %RSD JUB avg. %RSD compiled ICPMS ICPMS ICPMS ICPMS ICPMS ICPMS ref. avg. %RSD JUB avg. %RSD
HF-HClO4
(n=5)
(n=7)
(n=4)
(n=7)
Sc 0.5 0.14 0.136 0.259 66.0 0.195 23.7 0.265 6.8 13 13 10.1 12.7 32 38 30.1 32 33.0 9.0 30.9 7.4
Ti 7.97 7.97 8.80 10.1 2.10 8.9 6360 6360 5160 10.5 16300 16300 12700 15.3
Co 0.3 0.168 0.234 28.2 1.46* 29.6 0.689* 16.9 1692 1732 1712 1.2 1080 5.9 45 56.2 41 45.7 42 46.0 11.8 44.3 6.0
Ni 2.9 2.9 2.9 0.0 11.1* 2.8 3.04 5.9 12552 12632 12592 0.3 7770 12.1 119 137 107 115 115 119 8.4 124 19.3
Rb 1.5 0.14 0.82 82.9 0.0940 9.1 0.0434 25.0 11.04 10.9 11.6 11.18 2.7 9.99 3.1 9.8 9.48 8.56 10.1 8.71 9.77 8.69 9.30 6.3 9.42 5.6
Sr 108 119 116 117 115 3.6 117 5.7 113 2.6 737 792 800 776 3.6 653 5.5 389 399 394 407 389 396 368 392 2.9 379 4.3
Y 8.9 11.2 10.3 10.4 10.2 8.1 10.6 3.3 11.0 3.0 106.49 111 114 110.50 2.8 95.1 2.5 26 28.3 29 25.8 23.2 28 23 26.2 8.5 24.3 2.6
Zr 0.4 6.21 0.56 2.39 113.1 0.567 1.4 0.237 12.8 350 344 392 362 5.9 318 2.9 172 178 172 174 168 181 158 172 4.0 168 4.9
Nb 0.2 0.2 0.253 40.6 0.0072 12.5 27.42 22.3 24.86 10.3 24.1 10.2 18 19 18.9 19.2 16.6 15.8 15 17.5 9.0 17.5 4.5
Mo 0.4 0.4 0.4 0.0 0.291 7.5 0.282 34.3 316 318 317 0.3 268 2.0 4 3.94 3.97 0.8 4.58 14.9
Cs 0.019 0.019 0.0072 8.8 0.0030 19.8 0.41 0.604 0.37 0.461 22.2 0.324 3.8 0.1 0.11 0.09 0.11 0.09 0.10 8.9 0.0954 4.6
Ba 24 6.14 7 12.4 66.4 6.26 19.1 5.54 2.5 1702 1714 1511 1642 5.7 1450 2.9 130 135 131 142 129 133 125 132 3.8 128 3.4
La 7.1 7.87 7.93 7.7 7.56 7.63 3.9 7.75 3.7 7.84 2.8 124.88 122 121 122.63 1.3 109 2.6 15 15.2 15.3 15.5 14.5 15.3 14.23 15.00 2.9 14.8 2.8
Ce 1.85 2.54 2.49 2.02 2.06 2.19 12.5 2.07 3.4 2.12 2.3 286 277 271 278 2.2 274 19.4 38 38.4 37.6 38.4 35.3 37.5 35.73 37.28 3.1 36.8 3.1
Pr 0.97 0.9 0.956 1.05 0.986 0.972 5.0 1.01 3.9 1.02 2.5 29.72 31.4 32 31.04 3.1 28.3 3.7 5.35 5.57 5.31 5.72 5.1 5.39 4.96 5.34 4.5 5.18 2.9
Nd 4 5.33 5.25 4.09 4.21 4.58 12.8 4.18 3.6 4.25 2.4 127.88 137 123 129.29 4.5 118 3.1 25 24.9 24.5 24.8 23.4 24.6 22.8 24.3 3.2 24.0 3.6
Sm 0.73 0.84 0.788 0.68 0.686 0.745 8.2 0.708 2.9 0.712 2.7 29.25 30.2 27 28.82 4.7 26.7 3.6 6.2 6.16 6.04 5.99 5.9 6.12 5.69 6.01 2.7 5.96 3.1
Eu 0.15 0.19 0.176 0.162 0.154 0.166 8.9 0.159 4.0 0.161 3.0 7.26 7.58 7.53 7.46 1.9 6.60 2.0 2.07 2.03 2.05 2.07 1.96 2.07 1.96 2.03 2.3 2.03 3.2
Gd 0.87 0.87 0.9 0.872 0.878 1.4 0.910 4.1 0.904 3.5 26.05 29.8 32.1 29.32 8.5 27.8 0.5 6.3 6.13 6.23 6.48 5.9 6.32 6.13 6.21 2.7 6.11 2.1
Tb 0.12 0.12 0.116 0.12 0.113 0.118 2.4 0.121 3.9 0.121 3.3 4.42 4.81 4.92 4.72 4.5 4.27 1.6 0.9 0.963 0.933 0.9 0.9 0.959 0.89 0.921 3.1 0.908 2.9
Dy 0.71 1 0.814 0.75 0.747 0.804 12.9 0.754 3.6 0.755 3.4 25.65 28.3 28.6 27.52 4.8 25.3 2.1 5.31 5.3 5.29 5.25 5.2 5.34 5.07 5.25 1.6 5.18 3.0
Ho 0.17 0.2 0.164 0.167 0.164 0.173 7.9 0.168 3.9 0.168 3.6 5.49 5.76 5.58 5.61 2.0 4.93 2.4 1.04 1.01 0.964 1.03 0.95 1.01 0.93 0.991 4.0 0.960 2.5
Er 0.44 0.44 0.46 0.462 0.451 2.3 0.466 3.0 0.465 3.2 13.26 14.6 15.6 14.49 6.6 13.8 1.8 2.54 2.5 2.49 2.49 2.4 2.53 2.39 2.48 2.2 2.49 2.6
Tm 0.06 0.06 0.058 0.056 0.055 0.058 3.5 0.0550 3.8 0.0550 3.9 2.04 2.14 2.21 2.13 3.3 1.94 2.1 0.33 0.35 0.321 0.34 0.32 0.34 0.31 0.330 3.9 0.316 3.0
Yb 0.29 0.36 0.323 0.305 0.303 0.316 7.7 0.301 3.3 0.297 3.5 12.86 13.8 14.4 13.69 4.6 12.5 2.4 2 2.05 1.95 2.01 1.91 2.03 1.84 1.97 3.5 1.90 2.7
Lu 0.05 0.05 0.0494 0.043 0.042 0.0469 7.7 0.0426 4.2 0.0427 3.9 2.06 2.07 2.16 2.10 2.1 1.91 3.1 0.28 0.286 0.269 0.28 0.26 0.278 0.25 0.272 4.4 0.272 3.1
Hf 0.12 0.1 0.0897 0.1032 12.2 0.0148 5.9 0.0053 16.7 6.1 6.23 6.4 6.24 2.0 5.59 4.3 4.1 4.42 4.2 4.58 4.4 4.38 4.35 3.6 4.27 4.7
Ta 0.009 0.009 0.0137 75.1 0.0010 0.0 0.68 0.635 0.658 3.4 0.485 10.6 1.4 1.22 1.08 1.23 1.09 0.686 0.92 1.089 19.8 1.06 6.7
W 0.26 0.26 0.276 13.5 0.248 17.5 37.49 45.3 41.40 9.4 34.7 7.5 0.21 0.29 0.25 16.0 0.228 6.7
Pb 0.5 1 0.19 0.77 0.62 49.2 0.450 4.4 0.352 23.7 444 430 434 436 1.4 360 4.3 1.6 1.53 1.8 1.22 1.58 2.89 1.58 1.74 28.4 1.63 9.2
Th 0.04 0.0429 0.0415 3.5 0.0556 3.3 0.0521 6.1 11.93 11.7 13 12.21 4.6 10.2 3.9 1.2 1.3 1.13 1.32 1.2 1.22 1.53 1.27 9.5 1.23 4.7
U 0.8 0.858 0.88 0.846 4.0 1.03 3.9 0.987 6.1 4.81 5.01 5.3 5.04 4.0 4.32 3.2 0.403 0.446 0.403 0.45 0.42 0.425 0.4 0.421 4.6 0.430 3.4
1 data sources, whether compiled from studies or from single analytical method, and arranged in chronologically.
2 reference average reported to number of significant digits observed in most precisely measured data from literature sources.
3 JUB average calculated from number of sample decompositions reported in Table 1 and reported to three significant digits only for illustrative purposes (see Section 5 regarding analytical precision), with exception of some elements for JDo-1 carbonate decomposition method.
4 data for JDo-1 includes measured concentrations as determined using both the HF-HClO 4 and carbonate (HNO 3 ) sample decomposition methods (see Fig. 1). Note that Ni data for carbonate decomposition solely determined from the 62 Ni isotope.
* values greater than reference average considered to include significant contributions from interfering molecular species, arising from high major element contents (Mg, Al, Ca, Mn, Fe).
63

Appendix 2. Interferences due to major elements
Appendix 2 contains figures illustrating significant molecular interferences for
atomic masses

the concentration of the major element being considered, since as concentrations
increase (e.g., to 500 mg/kg) the high TDS content of the solution begins to suppress
analyte intensities, similar to that observed for rock samples that are only minimally
diluted before analysis.
The quantification of the observed interferences as concentrations in mg/kg
allows predictions to be made regarding the potential impact these interferences might
have upon trace metal determinations. For example, as the CaO content in a carbonate
rock increases from 10% to 40%, then Ni concentrations determined in an HCl acid
matrix would be expected to increase from ~2.5 mg/kg to ~6 mg/kg, due to Ca
interferences on 60 Ni (Fig. A2.1). These calculated interferences in concentration
units of m/kg are obtained from the long term average instrument response as
determined from the 10 μg/kg calibration standard (described in Fig. 2, Section 4).
For example, the 60 Ni isotope has an average instrument response of 2158 cps/μg·kg -1
in 0.5 M HCl, and a sample containing 14% CaO and diluted 1000x would be
expected to generate an CaO(H) interference of ~6700 cps on mass 60. Therefore, Ni
quantified in this sample using measurement of the 60 Ni isotope would be erroneously
overestimated by approximately 3.1 μg/kg (i.e., 2158/6700 = 3.1).
It should be noted that, similar to estimates of interferences based upon raw
data (cps), interference estimates reported as concentrations (mg/kg) will vary with
instrument performance. As the long term average instrument response for the 10
μg/kg calibration standard varies by as much as 30% (RSD), and this is the basis for
the calculation of the interference magnitude in mg/kg, then these calculated
concentrations will vary similarly.
Data for interferences are presented for the following isotopes in the order;
60 Ni, 62 Ni, 88 Sr, 89 Y, 90 Zr, 91 Zr, 93 Nb, and 95 Mo. Regardless of the element of interest,
or the specific interferences that inhibit accurate concentration measurements of these
elements, the key factor is always the relative abundance of the interfering species to
the isotope of interest. In other words, the greatest analytical difficulties arise when
trying to quantify small concentrations of an element in the presence of large
concentrations of interfering elements. As a result, while some interferences may be
large, e.g., several mg/kg in the case of Ca 2 and ArCa on Sr, the effect of these
interferences is minor if the element being measured is abundant in the sample (e.g.,
115 mg/kg Sr in the JDo-1 dolomite).
65

The case of Ni determinations in carbonates has been mentioned previously,
and the conclusion is that accurate Ni determinations in Ca- and Mg-rich rocks are
only possible in HNO 3 acid matrices and by utilizing the 62 Ni isotope. Figures A2.1
through A2.4 illustrate the numerous Ca and Mg interferences on 60 Ni and 62 Ni.
Additionally, Figure A2.5 demonstrates that Al 2 O 3 abundances typical for shales (10-
15%) that are analyzed in 0.5 M HCl are expected to contribute 2-3 mg/kg to the
measured Ni concentration. The data suggest that 62 Ni, when measured in a HNO 3
acid matrix, offers the best isotope for Ni determinations in Mg-,Ca-, and Al-rich
samples. However, 62 Ni is a low abundance isotope of Ni (3.63%), and will produce a
relatively low signal response during ICPMS measurements, and this factor must be
considered during attempts to quantify Ni.
Figure A2.6 illustrates significant interferences on 88 Sr due to various Ca and
presumed Ar species ( 40 Ca 48 Ca, 40 Ar 48 Ca, and 44 Ca 44 Ca). These interferences exist
regardless of the acid matrix used to prepare the samples for ICPMS analysis.
Concentrations of Sr in many rocks are in the range of 50 to several hundred mg/kg
(App. 1), and as Ca and Sr are both alkaline earth metals, their concentrations in
many rocks vary proportionally. As a result, even though high Ca contents may
produce an interference of 1-3 mg/kg on Sr, this frequently is of no great significance
as Sr concentrations themselves may be high in the sample (e.g., ~115 mg/kg in the
case of the JDo-1 dolomite).
Interferences on Y due to high Fe contents in samples are presented in Figure
A2.7 for informative purposes only. The impact of these interferences is likely to be
small (~0.10 mg/kg) relative to typical Y contents in rocks (~10-50 mg/kg, see App.
1), and will not be discussed further.
Figures A2.8 and A2.9 demonstrate that significant interferences due to Mn
and Fe may adversely impact determinations of Zr. The 90 Zr isotope suffers an
interference of several mg/kg in 0.5 M HCl due to 55 Mn 35 Cl (Fig. A2.8). The effects
of high Fe contents on measurements of 91 Zr are much greater due to 56 Fe 35 Cl, and
interferences of 10-20 mg/kg would be expected in Fe-rich samples such as IFs (Fig.
A2.9). Smaller, but still significant effects are noted for 40 Ca 16 O 35 Cl interferences on
91 Zr. These effects would be minimized by the use of HNO 3 as the diluting acid.
Based upon these observations, the 91 Zr isotope is not recommended for most rock
types, unless the use of HCl is avoided, or if Ca- and Fe-poor, Mn-rich samples
require analysis.
66

Potential interferences on monoisotopic 93 Nb are presented in Fig. A2.10, and
are very similar to those described for 91 Zr. This is due to the substitution of the 37 Cl
isotope for 35 Cl in the Fe and Ca molecular species listed above (i.e., 56 Fe 37 Cl and
40 Ca 16 O 37 Cl). Unlike Zr, Nb offers no additional isotope for analysis, and the low
natural abundance of Nb exacerbates the problem of interfering species due to major
elements. Accurate Nb analyses are unlikely in Fe- and Ca-rich rock types, unless the
use of HCl is avoided. However, as discussed in Section 7.1 of the main text, Nb is
unstable in pure HNO 3 acid matrices, and the presence of HCl stabilizes Nb, as well
as other HFSE like Ta, in solution long enough for accurate ICPMS measurements.
As mentioned previously, future work examining the optimum acid matrix for ICPMS
determinations of Nb and Ta is recommended.
The last element that appears to suffer from major element interferences is
Mo, and in particular Mn interferences on the 95 Mo isotope (Fig. A2.11). Similarly to
the Ca-Ar interferences on 88 Sr, Mn affects measured intensities for 95 Mo regardless
of the acid matrix. This is apparently due to the 55 Mn 40 Ar molecular species, which
forms independently of the bulk solution composition. Whereas the use of HCl
appears to suppress 55 Mn 40 Ar formation (Fig. A2.11), the magnitude of the
interference is significant regardless of the acid matrix (e.g., ~3-6 mg/kg at 20-30%
MnO). It is therefore recommended that the 97 Mo isotope be used for Mo
determinations in rock samples that contain more than a few percent MnO, unless Mo
concentrations in the sample are the range of 50 to hundreds of mg/kg.
67

20000
0.5 M HCl
10
9
interference on 60 Ni in sample solution (cps)
15000
10000
5000
CaO(H) +
JDo-1 dolomite (2.9 mg/kg Ni)
MgCl +
8
7
6
5
4
3
2
interference on 60 Ni in sample powder (mg/kg)
1
0
0
0 10 20 30 40 50 60 70 80 90 100
MgO, CaO in sample powder (wt.%)
Figure A2.1. Interferences observed at mass 60 (Ni) in 0.5 M HCl due to sample Mg and Ca content.
JDo-1 contains 18.4% MgO and 34.0% CaO, effectively prohibiting in carbonate rocks
determinations of Ni concentrations that are similar to JDo-1 (2.9 mg/kg Ni).
15000
18
16
interference on 60 Ni in sample solution (cps)
10000
5000
0.5 M HNO 3
JDo-1 dolomite (2.9 mg/kg Ni)
CaO(H) +
14
12
10
8
6
4
interference on 60 Ni in sample powder (mg/kg)
N 2
O 2 + (?)
2
0
0
0 10 20 30 40 50 60 70 80 90 100
MgO, CaO in sample powder (wt.%)
Figure A2.2. Interferences observed at mass 60 (Ni) in 0.5 M HNO 3 due to sample Mg and Ca
content. Use of HNO 3 effectively removes MgCl interference, but does not affect the dominant
CaO(H) interference.
68

1500
0.5 M HCl
4.0
interference on 62 Ni in sample solution (cps)
1000
500
JDo-1 dolomite (2.9 mg/kg Ni)
MgCl +
CaO(H) + (Ni in acid blank?)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
interference on 62 Ni in sample powder (mg/kg)
0
0.0
0 10 20 30 40 50 60 70 80 90 100
MgO, CaO in sample powder (wt.%)
Figure A2.3. Interferences observed at mass 62 (Ni) in 0.5 M HCl due to sample Mg and Ca
content. JDo-1 contains 18.4% MgO and 34.0% CaO, effectively prohibiting in carbonate rocks
determinations of Ni concentrations that are similar to JDo-1 (2.9 mg/kg Ni).
600
4.5
500
4.0
interference on 62 Ni in sample solution (cps)
400
300
200
100
0.5 M HNO 3
MgO, CaO in sample powder (wt.%)
CaO(H) + (Ni in acid blank?)
JDo-1 dolomite (2.9 mg/kg Ni)
not MgCl + (Ni in Mg standard?)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
interference on 62 Ni in sample powder (mg/kg)
0
0.0
0 10 20 30 40 50 60 70 80 90 100
Figure A2.4. Interferences observed at mass 62 (Ni) in 0.5 M HNO 3 due to sample Mg and Ca content.
When monitored in HNO 3 matrices, the 62 Ni isotope is likely the most useful for trace Ni
determinations in carbonate rocks, though the low abundance of 62 Ni (3.63%), which generates low
ICPMS signal intensities, must be considered.
69

1600
1400
0.5 M HCl
4.5
4.0
interference on 62 Ni in sample solution (cps)
1200
1000
800
600
400
200
0
5.1% (FeR-2)
AlCl +
0.0
0 5 10 15 20
13.7% (SCo-1)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
interference on 62 Ni in sample powder (mg/kg)
Al 2
O 3
in sample powder (wt.%)
Figure A2.5. Interferences observed at mass 62 (Ni) in 0.5 M HCl due to sample Al content. Shown are
Al contents of two common CRMs that contain similar Ni (23 mg/kg for FeR-2, and 26 mg/kg for SCo-
1). The lower Al/Ni of FeR-2 is expected to allow reasonable Ni determinations using 62 Ni, however the
poor agreement between 60 Ni and 62 Ni argues against this (see App. 1). The greater Al/Ni of SCo-1
significantly contributes to higher than expected Ni determinations in SCo-1 (see App. 1).
interference on 88 Sr in sample solution (cps)
60000
50000
40000
30000
20000
10000
0.5 M HNO 3
0.5 M HCl
5.2 mg/kg
0
0 10 20 30 40 50 60 70 80 90 100
CaO in sample powder (wt.%)
3.2 mg/kg
Figure A2.6. Interferences observed at mass 88 (Sr) in both 0.5 M HCl and HNO 3 as a function of Ca
content. Note right vertical axis denotes two concentration scales, dependent upon acid matrix.
Significant interferences due to probable combinations of 40 Ca 48 Ca, 40 Ar 48 Ca, and 44 Ca 44 Ca exist
regardless of acid matrix. Note that the discrepancy between acid matrices regarding the magnitude of
the interference arises from different ICPMS signal response (sensitivity) in HCl and HNO 3 . Caution
should be exercised when interpreting Sr concentrations below ~75 mg/kg in samples with high Ca
contents.
0.5 M HCl
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.5 M HNO 3
7
6
5
4
3
2
1
interference on 88 Sr in sample powder (mg/kg)
70

4000
0.40
0.20
3500
0.17 mg/kg
0.18
0.35
interference on 89 Y in sample solution (cps)
3000
2500
2000
1500
1000
500
0.5 M HCl
0.5 M HNO 3
0.13 mg/kg
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Fe 2
O 3
in sample powder (wt.%)
0.5 M HCl
0.5 M HNO 3
0.30
0.25
0.20
0.15
0.10
0.05
interference on 89 Y in sample powder (mg/kg)
Figure A2.7. Interferences observed at mass 89 (Y) in both 0.5 M HCl and HNO 3 as a function of Fe
content. Note right vertical axis denotes two concentration scales, dependent upon acid matrix. A
minor 54 Fe 35 Cl interference exists in 0.5 M HCl, and interferences are linear, though small, for both
acid matrices, perhaps reflecting small Y contamination in artificial Fe solution. Concentrations of Y in
all Fe-rich CRMs are 8-12 mg/kg, suggesting that Fe-interferences in these CRMs are not significant.
12000
1.3 mg/kg
2.5
10000
1.2
interference on 90 Zr in sample solution (cps)
8000
6000
4000
2000
0.5 M HCl
0.5 M HNO 3
0.13 mg/kg
0
0 5 10 15 20
MnO in sample powder (wt.%)
0.5 M HCl
1.0
0.8
0.6
0.4
0.2
0.5 M HNO 3
2.0
1.5
1.0
0.5
interference on 90 Zr in sample powder (mg/kg)
Figure A2.8. Interferences observed at mass 90 (Zr) in both 0.5 M HCl and HNO 3 as a function of MnO
content. Note right vertical axis denotes two concentration scales, dependent upon acid matrix. A strong
55 Mn 35 Cl interference exists in 0.5 M HCl. Concentrations of MnO in Fe-Mn crusts (and Mn ores) may
exceed 30%, and extrapolation of the data suggest possible interferences on 90 Zr of several mg/kg.
71

40000
35000
0.5 M HCl
20
18
interference on 91 Zr in sample solution (cps)
30000
25000
20000
15000
10000
5000
FeCl +
CaOCl +
16
14
12
10
8
6
4
2
interference on 91 Zr in sample powder (mg/kg)
0
0
0 10 20 30 40 50 60 70 80 90 100
CaO, Fe 2
O 3
in sample powder (wt.%)
Figure A2.9. Interferences observed at mass 91 (Zr) in 0.5 M HCl as a function of both Ca and Fe
content. Interferences in HNO 3 acid not observed. Significant interferences produced at low
concentrations (

interference on 95 Mo in sample solution (cps)
5000
4000
3000
2000
1000
0.5 M HNO 3
0.5 M HCl
0
0 5 10 15 20
MnO in sample powder (wt.%)
4.3 mg/kg
2.1 mg/kg
0.5 M HCl
2.0
1.5
1.0
0.5
0.5 M HNO 3
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
interference on 95 Mo in sample powder (mg/kg)
Figure A2.11. Interferences observed at mass 95 (Mo) in both 0.5 M HCl and HNO 3 as a function of
MnO content. Note right vertical axis denotes two concentration scales, dependent upon acid matrix.
Significant interferences produced for both HCl and HNO 3 acid matrices. For Mn-bearing (>5%),
relatively Mo-poor (

Chapter 3. Continentally-derived solutes in shallow Archean seawater: Rare earth
element and Nd isotope evidence in iron formation from the 2.9 Ga
Pongola Supergroup, South Africa
119

THIS PAGE INTENTIONALLY LEFT BLANK
120

Available online at www.sciencedirect.com
Geochimica et Cosmochimica Acta 72 (2008) 378–394
www.elsevier.com/locate/gca
Continentally-derived solutes in shallow Archean seawater:
Rare earth element and Nd isotope evidence in iron
formation from the 2.9 Ga Pongola Supergroup, South Africa
Brian W. Alexander a, *, Michael Bau a , Per Andersson b , Peter Dulski c
a Earth and Space Sciences, Campus Ring 1, Research III 102, Jacobs University Bremen, D-28759 Bremen, Germany
b Laboratory for Isotope Geology, Swedish Museum of Natural History, SE-104 05 Stockholm, Sweden
c GeoForschungsZentrum, D-14473 Potsdam, Germany
Received 27 November 2006; accepted in revised form 25 October 2007; available online 17 November 2007
Abstract
The chemical composition of surface water in the photic zone of the Precambrian ocean is almost exclusively known from
studies of stromatolitic carbonates, while banded iron formations (IFs) have provided information on the composition of deeper
waters. Here we discuss the trace element and Nd isotope geochemistry of very shallow-water IF from the Pongola Supergroup,
South Africa, to gain a better understanding of solute sources to Mesoarchean shallow coastal seawater. The Pongola
Supergroup formed on the stable margin of the Kaapvaal craton 2.9 Ga ago and contains banded iron formations (IFs) that
represent the oldest documented Superior-type iron formations. The IFs are near-shore, pure chemical sediments, and shalenormalized
rare earth and yttrium distributions (REY SN ) exhibit positive La SN ,Gd SN , and Y SN anomalies, which are typical
features of marine waters throughout the Archean and Proterozoic. The marine origin of these samples is further supported
by super-chondritic Y/Ho ratios (average Y/Ho = 42). Relative to older Isua IFs (3.7 Ga) from Greenland, and younger
Kuruman IFs (2.5 Ga) also from South Africa, the Pongola IFs are depleted in heavy rare earth elements (HREE), and
appear to record variations in solute fluxes related to sea level rise and fall. Sm–Nd isotopes were used to identify potential
sediment and solute sources within pongola shales and IFs. The Nd (t) for Pongola shales ranges from 2.7 to 4.2, and Nd (t)
values for the coeval iron-formation samples (range 1.9 to 4.3) are generally indistinguishable from those of the shales,
although two IF samples display Nd (t) as low as 8.1 and 10.9. The similarity in Nd isotope signatures between the shale
and iron-formation suggests that mantle-derived REY were not a significant Nd source within the Pongola depositional environment,
though the presence of positive Eu anomalies in the IF samples indicates that high-T hydrothermal input did contribute
to their REY signature. Isotopic mass balance calculations indicate that most (P72%) of the Nd in these seawater
precipitates was derived from continental sources. If previous models of Fe–Nd distributions in Archean IFs are applied, then
the Pongola IFs suggest that continental fluxes of Fe to Archean seawater were significantly greater than are generally
considered.
Ó 2007 Elsevier Ltd. All rights reserved.
1. INTRODUCTION
* Corresponding author.
E-mail address: b.alexander@jacobs-university.de (B.W.
Alexander).
Studies to determine the characteristics of seawater in
the geologic past are dependent upon proxies which reliably
reflect seawater composition. Such studies have
examined sedimentary rocks and minerals which form directly
from seawater solutions, including inorganic precipitates
(e.g., evaporites) or biologically mediated
precipitates (e.g., CaCO 3 ). Attempts to identify reliable
proxies for paleo-seawater rely on chemical tracers, particularly
elements with known and predictable behavior
in modern seawater and modern seawater precipitates.
0016-7037/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gca.2007.10.028

Nd isotopes in 2.9 Ga Archean surface seawater 379
These elemental tracers must also be relatively unaffected
by geologic processes which commonly affect marine precipitates,
such as diagenesis and metamorphism. One
such group of chemical tracers are the rare earth elements
and yttrium (REY), which have been used as seawater
proxies in a variety of sedimentary rocks of known
marine origin, including limestones, phosphorites, and
banded iron formations. The usefulness of the REYs as
seawater proxies has been discussed extensively by Bau
and Dulski (1996), Webb and Kamber (2000), Shields
and Stille (2001), Nothdurft et al. (2004), Shields and
Webb (2004), and Bolhar et al. (2004). These workers
have concluded that REY distributions in marine chemical
sediments can accurately reflect the REY distribution
of the seawater from which they formed, and that this
relationship can be reliably extended to sediments from
throughout the geologic record.
The information provided by the REYs regarding seawater
composition is a function of the processes which control
their distribution in seawater. In modern marine waters
the variation in REY distributions is dominated by varying
degrees of carbonate complexation, with REY principally
sourced from weathering of continents, as suggested by
the isotopic composition of Nd in modern seawater (Piepgras
and Wasserburg, 1980). High-temperature hydrothermal
inputs (e.g., black smoker fluids) are generally
negligible REY sources to modern seawater, though this
may not have been the case in Earth’s early oceans, a scenario
assessed using Eu, which is enriched in high-temperature
fluids emanating from basaltic, mid-ocean ridge
sources (MORB, Michard, 1989; Bau and Dulski, 1999).
This approach, coupled with Nd isotope systematics, has
been extended to Archean banded iron formations (e.g.,
Derry and Jacobsen, 1990; Danielson et al., 1992), in attempts
to constrain continental- versus MORB-derived Fe
inputs to Earth’s early oceans (Miller and O’Nions, 1985;
Jacobsen and Pimentel-Klose, 1988a). Therefore, the
REY are particularly well suited not only for identifying
Archean seawater proxies, but also for offering insight into
the importance of mid-ocean ridge solute fluxes relative to
solutes sourced from continental weathering.
Numerous workers have used Precambrian iron formations
to study ancient seawater indirectly (e.g., Fryer,
1977; Jacobsen and Pimentel-Klose, 1988a; Derry and Jacobsen,
1990; Shimizu et al., 1990; Towe, 1991; Bau and
Möller, 1993; among others). The similarity of REY patterns
between modern seawater and both Archean iron formations
and carbonates has been noted, and these chemical
sediments have been used to constrain Archean seawater
composition (e.g., Bau and Dulski, 1996; Kamber and
Webb, 2001; Bolhar et al., 2004). Though certain features
are common for many Precambrian IFs (e.g., alternating
bands of Si- and Fe-rich oxide layers), variations are observed
in chemical composition and mineralogy and are
generally attributed to distinct differences in tectonic setting
(e.g., Gross, 1983), or facies changes within a single depositional
basin (Klein and Beukes, 1989). As such, Precambrian
IFs formed in different marine environments may
reflect potential variations in seawater composition between
these environments.
Two broad types of IF are generally recognized: the Algoma
type which is often found in greenstone belts, and the
Superior type associated with stable sedimentary basins and
cratonic margins (Gross, 1983). When considering the mass
of sediment deposited, Superior-type IFs contain far more
Fe than Algoma-type IFs (James and Trendall, 1982;
Klemm, 2000), although this may represent a preservation
bias due to their proposed tectonic settings. The association
of Superior-type IFs with shelf deposits such as quartzites,
carbonates, and shales is consistent with their depositional
model, and Superior-type IFs are particularly useful for
constraining trace element distributions in shallow Archean
seawater.
The Kaapvaal craton in South Africa is widely recognized
as one of the earliest stable cratons formed (Tankard
et al., 1982), and hosts the oldest known Superior-type IFs
in the world within the 3.0 Ga Witwatersrand and Pongola
Supergroups. Work by Beukes and Cairncross (1991)
indicates that some units within both the Witwatersrand
and the Pongola are correlative, however, the Pongola differs
most significantly from the Witwatersrand in that it
experienced lower degrees of metamorphism and structural
deformation. The age, depositional environment, and the
relatively pristine preservation of the Pongola Supergroup
provide unique opportunities for the investigation of Archean
marine chemical sediments deposited in relatively
shallow waters. Rare earth and yttrium distributions and
Sm–Nd isotope systematics in Pongola Supergroup IFs
indicate significant continentally-derived solute fluxes to
shallow Archean seawater.
2. GEOLOGIC SETTING
The 2.9–3.0 Ga Pongola Supergroup is located in eastern
South Africa and southwestern Swaziland and crops
out over a 270-km by 100-km area (Weilers, 1990; Fig. 1).
The extent of the Pongola Supergroup is consistent with a
minimum depositional area of 32,500 km 2 (Button et al.,
1981), and the sequence as a whole consists of two stratigraphic
units; the Nsuze Group and the overlying Mozaan
Group. Sedimentary structures within Nsuze siliciclastic
beds such as lenticular/flaser bedding and herringbone
cross-lamination (von Brunn and Hobday, 1976) and stromatolite
bearing Nsuze carbonates (Mason and von Brunn,
1977; Beukes and Lowe, 1989) indicate a near-shore shallow
water depositional environment (see also Matthews,
1967; von Brunn and Mason, 1977; Tankard et al., 1982).
This study focuses on samples from the Sinqeni Formation
of the Mozaan Group, which outcrops within the Wit
Mfolozi inlier (Fig. 1). The Sinqeni Formation is the most
laterally extensive formation within the Mozaan Group
(Nhleko, 2003) and is dominated by quartz arenite and
shale with minor conglomerate and banded iron formation
(Matthews, 1967; Beukes and Cairncross, 1991). Specifically,
the iron formation is hosted within the Ijzermijn
Member, an approximately 15-m-thick unit which has shale
at the base overlying with a gradational contact coarse
sandstone of the older Dipka Member. The Ijzermijn shale
grades upward into 3- to 5-m-thick iron formation intercalated
with shale, followed again by shale which is capped

380 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
30 o 31 o 32 o 27 o
25 o
SWAZILAND
MOZAMBIQUE
REPUBLIC
OF
SOUTHAFRICA
REPUBLIC
OF
SOUTHAFRICA
Mozaan Group
Nsuze Group
0 50 km
30 o 31 o
WHITE MFOLOZI INLIER
INDIAN
OCEAN
32 o
28 o
29 o
LESOTHO
INDIAN
OCEAN
30 o
PONGOLA
30 o SUPER GROUP
LITHOLOGY AND SAMPLE POSITIONS
shale
conglomerate
iron formation
shale
iron formation
shale
conglomerate
19
18
13x
11x
12x
10x
8
7x
6x
5x
4x
# = sample number (see caption for details)
2
Fig. 1. Map of study area showing extent of Pongola Supergroup, location of White Mfolozi inlier, and relative positions of sampling
locations within a simplified sequence stratigraphy. The x following some sample numbers indicates that two samples were collected, and
numbered accordingly (e.g., 4x refers to samples 41 and 42). Thickness of the lithologic units is not to scale.
with a sharp erosional contact formed by supermature
orthoquartzite (Nhleko, 2003). Sedimentary structures
found within the Mozaan Group are similar to those described
in the Nsuze Group, and Beukes and Cairncross
(1991) recognized only two major depositional environments
for the Mozaan: fluvial braidplain and shallow marine
shelf systems. Studies of the Mozaan Group (Watchorn,
1980; Weilers, 1990) describe the IFs as being deposited
within a distal shelf environment, conclusions that are consistent
with the detailed analysis of Beukes and Cairncross
(1991), who characterized the Mozaan IFs as forming on a
shallow starved outer continental shelf during the peak of a
marine transgression.
Age constraints for the Mozaan are not directly available
from units within the group, though it appears that
deposition occurred between 2940 and 2870 Ma (Beukes
and Cairncross, 1991). This is inferred from a U–Pb zircon
age of 2940 ± 22 Ma for rhyolite within the Nsuze Group,
and a 2871 ± 30 Ma Sm–Nd mineral isochron (Hegner
et al., 1984) for pyroxenite from the post-Pongola intrusive
Usushwana Complex. Following deposition, the Pongola
Supergroup was intruded by differentiated gabbroic sills
(the Usushwana Complex), primarily along the basal
unconformity and as a series of sills in the Mozaan Group
(Button et al., 1981). Granitic plutons emplaced at 2.5–
2.7 Ga bound the Pongola Supergroup primarily along its
eastern margins, resulting in narrow contact aureoles with
relatively little deformation of Pongola strata (Button
et al., 1981; Tankard et al., 1982). Mineral assemblages
within the Pongola generally reflect greenschist facies regional
metamorphism (Tankard et al., 1982; Hegner et al.,
1984; Hunter and Wilson, 1988; Beukes and Cairncross,
1991).
3. SAMPLING AND ANALYTICAL METHODS
Four shale samples and 16 iron formation samples were
collected from the Sinqeni Formation in the White Mfolozi
Inlier (Fig. 1). The shale and IF beds are intercalated, and
are bounded at the top and bottom by beds of small pebble
conglomerate. Desiccation cracks are common in the clastic
units immediately above and below the sequence studied,
indicating that IF deposition was preceded and followed
by relatively lower seawater levels. Field sampling avoided
obvious fault zones and alteration features, and fresh
unweathered samples were subsequently crushed and ana-

Nd isotopes in 2.9 Ga Archean surface seawater 381
lysed at the GeoForschungsZentrum, Potsdam, Germany.
Major element concentrations were obtained by X-ray fluorescence
(XRF), and mineral phase determinations were
obtained by X-ray diffraction (XRD). Trace element concentrations
were determined with a Perkin-Elmer/Sciex
Elan Model 5000 inductively coupled plasma mass spectrometer
(ICP-MS) following the procedures of Dulski
(2001). Samarium and Nd concentrations were also determined
using thermal ionization mass spectrometry (TIMS).
Agreement between TIMS and ICPMS data is better than
2% for Nd (except shale WM2, 3.7%), and better than 4%
for Sm (except shale WM2 and IF WM41, 6.1% and
6.0%, respectively). Iron-formation sample WM121 is atypical
is displaying differences of 7.2% and 10.8% in respective
Nd and Sm concentrations as determined by TIMS and
ICPMS, yet Sm/Nd for WM121 differs by only 2.8% between
the two analytical methods, and Sm/Nd comparisons
between TIMS and ICPMS data never exceeds 5.3%, indicating
that consistent REE ratios were determined regardless
of the method used. Accuracy for other ICPMS
analyses is conservatively estimated to be ±10%, and rare
earth element ratios are estimated to be accurate within
±5%. Samples for Sm–Nd isotope determinations were
spiked using a mixed 147 Sm/ 150 Nd spike, and processed at
the Laboratory for Isotope Geology, Swedish Museum of
Natural History. Samarium and Nd were separated using
ion exchange techniques, and isotopic ratios were determined
with a five collector Finnigan MAT 261 TIMS. Total
analytical blank for Nd was approximately 40 pg, and Nd
was analysed as NdO + in multidynamic mode, with the
data being reduced assuming exponential fractionation
and normalized to 146 Nd/ 144 Nd = 0.7219. External precision
was obtained by repeated analyses of the La Jolla standard
with 143 Nd/ 144 Nd = 0.511853 ± 0.000025 (2r, n = 10).
Samarium was analyzed as an oxide in static mode and normalized
assuming 149 Sm/ 152 Sm = 0.51686. Uncertainties in
TIMS determination of Sm and Nd concentrations are estimated
to about 1% Al 2 O 3 and almost 15% MnO and
will be discussed in detail below. Unless otherwise noted,
results refer only to the remaining 15 IF samples. Normative
quartz and iron-oxides combined represent 96–99%
(wt.) of the samples, with Fe ranging from 10% to 81%.
Aluminium, Ca, Mg, Ti, and P are present only in trace
amounts, and Na and K are absent or not measurable.
Other than Si and Fe, only Mn is present in appreciable
amounts of the major elements studied, and averages
1%. The mineralogy of the iron oxide component for all
samples varies between magnetite and hematite, with the
relative proportion of magnetite increasing with increasing
Fe content. The absence of chlorite or secondary minerals
indicates that the samples have experienced little alteration
or weathering. Immobile trace element concentrations in
the IF samples are low (

Table 1
Major element data for iron formation and shales, with mineral phase determinations for IF samples
Iron-formation samples
Shale samples a
WM41 WM42 WM51 WM52 WM61 WM62 WM71 WM72 WM101 WM102 WM111 WM112 WM121 WM122 WM131 WM132 WM2 WM8 WM18 WM19 pelite b
XRF (wt%)
SiO 2 75.4 26.2 75.7 16.8 19.7 20.7 58.8 70.8 80.9 78.4 80.3 81.0 85.8 88.0 84.4 83.6 51.6 47.9 49.1 48.6 58
Al 2 O 3 0.13 1.22 0.18 0.07 0.67 0.39 0.21 0.02 0.52 0.22 nd nd 0.06 0.06 0.34 0.10 23.3 11.9 24.7 26.8 18.8
Fe 2 O 3 22.2 57.0 21.0 81.0 76.4 76.3 37.5 26.1 15.2 17.3 17.7 18.1 12.0 10.4 14.7 15.0 10.5 29.0 9.16 6.55 9.4
MnO 0.92 14.5 1.85 0.34 1.54 2.50 1.93 2.39 2.11 2.57 1.59 1.27 1.14 1.30 0.35 0.63 0.12 0.43 0.13 0.11 0.27
MgO 0.12 0.25 0.26 0.13 0.20 0.22 0.23 0.27 0.33 0.37 0.30 0.26 0.21 0.19 0.13 0.13 3.01 3.57 3.65 3.18 1.68
CaO 0.15 0.18 0.18 0.18 0.11 0.17 0.24 0.17 0.14 0.21 0.16 0.18 0.14 0.12 0.11 0.11

Table 2
Trace element concentrations in iron-formation and shale samples (mg/kg)
Iron-formation samples
Shale samples
WM41 WM42 WM51 WM52 WM61 WM62 WM71 WM72 WM101 WM102 WM111 WM112 WM121 WM122 WM131 WM132 WM2 WM8 WM18 WM19 pelite a
Rb

384 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
considerably lower (average 1.13). The REY patterns are
not correlated with major element chemistry, nor with
immobile trace element content, but the two groups of patterns
can be distinguished on the basis of relative stratigraphic
position; i.e., high (Dy/Yb) WMS corresponds to
samples collected at the base and at the top of the sequence
REY/WMS
REY/WMS
1
0.1
0.01
0.001
1
0.1
0.01
top of sequence
bottom of sequence
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Y Ho Er Tm Yb Lu
middle of sequence
WM42
WM132
WM131
WM62
WM61
WM52
WM51
WM42
WM41
WM122
WM121
WM112
WM111
WM102
WM101
WM72
WM71
studied, and low (Dy/Yb) WMS is found in samples collected
from the middle of the sequence.
The REY patterns of all the IF samples exhibit WMSnormalized
positive La, Eu, and Gd anomalies. Calculated
(La/La * ) WMS and (Gd/Gd * ) WMS are above unity for all 16
IF samples (Fig. 3), and (Eu/Eu * ) WMS ranges from 1.51 to
2.09 (average 1.69). The magnitude of the La, Eu, and Gd
anomalies are indistinguishable on the basis of stratigraphic
position. Y/Ho averages 42 for all samples, and higher Y/
Ho values are observed for samples collected from the middle
of the sequence, though the difference between the two
groups is small when samples WM121 and WM122 (Y/Ho
of 61 and 66, respectively) are not considered.
The REY pattern of sample WM42 is similar to samples
obtained stratigraphically above and below it, possessing
positive La, Gd, and Y anomalies. The RREY content is
the highest of the IF samples, and though most major element
concentrations for sample WM42 are comparable to
the other IF samples, Al 2 O 3 is enriched at 1.2% and MnO
is significantly higher (14.5%). WM42 also contains somewhat
greater concentrations of Rb, and significantly more
Zr, Hf and Th than any of the other 15 IF samples.
Seven iron-formation samples and three shale samples
were analyzed using TIMS, and Nd and Sm isotope data
are presented in Table 3. Six of the samples (one shale
and five IF samples) display similar 147 Sm/ 144 Nd between
0.1164 and 0.1486, values common to other Archean shales
and iron formation (e.g., Miller and O’Nions, 1985; Alibert
and McCulloch, 1993; Jahn and Condie, 1995; Bau et al.,
1997). Of the remaining four samples, two are iron formations
displaying the highest 147 Sm/ 144 Nd observed (>0.17),
whereas the lowest 147 Sm/ 144 Nd values are observed in the
remaining two shale samples (

Nd isotopes in 2.9 Ga Archean surface seawater 385
Table 3
Nd and Sm concentrations (mg/kg) and isotopic data determined by TIMS for 2.9 Ga Mozaan shales and iron-formations
Sample Nd (ppm) Sm (ppm)
143 Nd/ 144 Nd ± 2r a 147 Sm/ 144 Nd ± 2r Nd (0) Nd (2.9 Ga)
Shale
WM2 30.7 6.34 0.511120 ± 25 0.1247 ± 6 29.61 2.7 ± 0.3
WM8 8.19 1.33 0.510543 ± 25 0.0979 ± 5 40.87 4.0 ± 0.6
WM18 64.7 9.70 0.510389 ± 25 0.0905 ± 5 43.87 4.2 ± 0.3
Iron formation
WM41 0.520 0.148 0.511754 ± 25 0.1721 ± 9 17.24 8.1 ± 0.7
WM51 0.457 0.112 0.511587 ± 25 0.1486 ± 7 20.50 2.6 ± 0.8
WM61 1.76 0.362 0.511029 ± 25 0.1242 ± 6 31.39 4.3 ± 0.6
WM72 0.533 0.104 0.511203 ± 25 0.1302 ± 7 27.99 3.2 ± 0.6
WM102 0.659 0.146 0.511322 ± 25 0.1330 ± 7 25.67 1.9 ± 0.7
WM121 0.194 0.037 0.510921 ± 25 0.1164 ± 6 33.49 3.5 ± 0.5
WM131 0.762 0.215 0.511698 ± 25 0.1766 ± 9 18.34 10.9 ± 0.6
WM131r b 0.748 0.218 0.511734 ± 25 0.1763 ± 9 17.63 10.1 ± 0.8
BCR-2 (3 mg) 28.8 6.6 0.512619 ± 25 0.1378 ± 7
a The errors in the 143 Nd/ 144 Nd refer to the last digits and are estimated from repeated analysis of the La Jolla standard which gave
143 Nd/ 144 Nd = 0.511853 ± 0.000025 (2r, n = 10).
b Replicate digestion, column separation, and analysis of sample WM131.
143 Nd/ 144 Nd of the sample, and 143 Nd/ 144 Nd within a chondritic
uniform reservoir (CHUR) at any given time t. For
the shale samples, Nd (t) falls within a narrow range from
2.74 to 4.24, whereas the IF samples display a bimodal
Nd (t) distribution, with most samples exhibiting Nd (t) similar
to the shales, while the two IF samples with
147 Sm/ 144 Nd >0.17 (WM41 and WM131) have significantly
more negative Nd (t) values of 8.1 and 10.9, respectively.
5. DISCUSSION
5.1. Depositional controls on REYs in Mozaan IFs
The Mozaan IFs REY distributions might have been affected
by a number of processes including: (i) post- and syndepositional
processes, (ii) the degree to which the REY
were scavenged by particulate matter in the water column
prior to deposition, and (iii) also by the various inputs to
the seawater from which the IFs precipitated.
Potential post-depositional mechanisms affecting trace
metal and REY distribution in the Mozaan IFs include diagenetic
and metamorphic processes. During diagenesis,
mobility of the REY would tend to average out REY distributions
in banded iron formations, yet Bau and Dulski
(1992) observed highly variable REY ratios in individual
Fe- and quartz-rich bands within the 2.46 Ga Superior-type
Kuruman IF. Such variation is also present in REY data
for IF mesobands from the Dales Gorge Member in the
Hamersley Group of Australia (Morris, 1993). Bau (1993)
studied the impact of post-burial processes on REY signatures
in Precambrian IFs from the Hamersley Basin in Western
Australia, the Kuruman and Penge IFs in South
Africa, and the Broomstock IFs of Zimbabwe, and concluded
that REYs are effectively immobile during diagenesis
and lithification.
The effects of metamorphism on REY mobility is a function
of water/rock ratios, with LREE depletion and negative
Eu anomalies expected in rocks that host significant
amounts of metasomatic fluids during metamorphism
(Grauch, 1989; Bau, 1993). Iron-formation samples from
regions that have experienced high grade metamorphism
(e.g., Isua, Greenland) do not exhibit Eu or LREE depletion,
and in general coeval detritus free IF samples display
similar REY CN patterns regardless of metamorphic grade
(Bau, 1991, 1993). Therefore, the relatively low-grade
greenschist facies metamorphism which affected Mozaan
IF samples would not have had a significant effect on their
REY distribution.
For syn-depositional processes, REY patterns in chemical
precipitates may be fractionated with respect to contemporaneous
seawater due to the presence of detrital
aluminosilicates, or as a result of exchange effects between
REY scavenging particulates and marine waters. The effect
of clastic contamination can be assessed using trace elements
that are essentially immobile in aqueous solutions,
and hence occur only at ultra-trace levels in seawater. The
low abundances in the Mozaan IFs of such elements as
Al, Ti, Zr, Hf, Y, and Th demonstrates that these are very
pure chemical sediments and that clastic contamination is
negligible, which is consistent with a sediment-starved continental
shelf depositional environment for the Mozaan IFs
(e.g., Beukes and Cairncross, 1991).
The influence of solution complexation, fractionation,
and scavenging within the marine environment on REY
distributions in IFs is more difficult to assess. The REY
concentration in modern seawater is controlled primarily
by scavenging particulate matter (e.g., Elderfield, 1988; Erel
and Stolper, 1993), to the extent that seawater abundances
of REYs are extremely low. The association between Ferich
colloids and REYs has led many workers to conclude
that Fe-oxyhydroxide particles dominated REY scavenging
during the formation of ancient metalliferous sediments
(e.g., Derry and Jacobsen, 1990). This scavenging should
represent the balance between two competing effects; the
solution complexation of the REYs and the Fe-oxyhydroxide
surface complexation. This model assumes that the reac-

386 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
tion kinetics of adsorption/desorption are significantly faster
than the particle residence time in an oxic water column.
This assumption is supported by experimental evidence
indicating that particle-surface/solution REY exchange
equilibrium occurs within minutes (Bau, 1999), suggesting
that most Fe-oxyhydroxide particles settling through an
oxic water column (of relatively constant REY distribution)
should be in equilibrium with the solution phase.
Natural systems that provide insight into such equilibrium
exchange processes are available in the form of modern
marine hydrogenetic ferromanganese crusts and
terrestrial spring-water precipitates. Seafloor Fe–Mn crusts
have trace metal distributions that reflect equilibrium
adsorption–desorption processes between REYs and seawater,
processes that strongly fractionate REYs. Very similar
fractionation is observed between Fe-oxyhydroxides
and terrestrial spring water (Bau et al., 1998). The most
striking difference between these precipitates and the
respective fluids which formed them is the lower Y/Ho observed
in the precipitates, which display negative Y anomalies.
This Y–Ho fractionation is due to the preferential
sorption of Ho relative to Y on the scavenging Fe(–Mn)
particles (e.g., Bau et al., 1996, 1998). Preferential adsorption
of Ho over Y on Fe-oxyhydroxides has been demonstrated
experimentally (Bau, 1999) and is completely
consistent with the observed fractionation of Y and Ho
during precipitation of hydrogenetic Fe–Mn crusts from
seawater. We emphasize that such Y–Ho fractionation is
not observed in Archean iron formations.
If redox conditions in Archean seawater permitted the
precipitation of ferric iron, and if scavenging Fe-oxyhydroxide
particles achieved exchange equilibrium with surrounding
seawater, the above observations predict subchondritic
values of Y/Ho in the IFs. The significantly higher
Y/Ho observed in IFs of different ages strongly suggests
that Fe-oxyhydroxide particles that scavenged REYs could
not be at or near exchange equilibria with respect to ambient
seawater (Bau and Dulski, 1996). Although the actual
reason why IFs display the REY seawater pattern has not
been satisfactorily explained and requires further investigation,
the striking similarity between REY patterns in Archean
marine carbonates and IFs suggests that pure Archean
IFs faithfully record the REY distributions of contemporaneous
seawater.
5.2. Archean seawater
5.2.1. High-temperature hydrothermal input
The geologic setting and low detritus content, coupled
with the observed LREE depletions and positive La, Gd,
and Y anomalies (Figs. 2 and 5), strongly suggests the Mozaan
IFs represent 3.0 Ga seawater. Working under the
assumption that these sediments possess trace element distributions
reflective of contemporaneous seawater, some
general constraints may be placed on the composition of
Archean seawater.
The occurrence of the oxide-facies Mozaan IFs suggests
that the redox level of even very shallow seawater permitted
nearly continuous precipitation of significant amounts of
Fe(III)-bearing (hydr)oxides, while the lack of any Ce
anomaly in normalized data indicates that no process operated
during Fe-sedimentation that was capable of fractionating
Ce. This scenario is clearly different from the oxic
modern marine system, in which Ce is oxidized to immobile
Ce(IV) on particle surfaces, resulting in fractionation from
the other REY and producing the strong negative Ce anomaly
observed in modern seawater. The lack of significant
negative-Ce anomalies in many Archean chemical sediments
has been discussed by numerous authors in terms
of the oxidation history of the Earth’s surface (e.g., Fryer,
1977; Derry and Jacobsen, 1990; Kato et al., 1998; among
others). While the exact process that converted large
amounts of Fe(II) to Fe(III) in seawater during sedimentation
of the Mozaan IFs is unknown, the Ce distributions in
these chemical precipitates indicate Archean seawater did
not possess a negative Ce anomaly, and correspondingly,
redox levels were lower than those observed in modern
marine systems.
Unlike Ce, the Mozaan IFs display pronounced positive
Eu anomalies (Fig. 4), which in modern marine environments
are only observed in high-temperature (>250 °C)
hydrothermal fluids, such as those typically found at midocean
ridges and back-arc spreading centers. It is reasonable
that the REY CN patterns of these high-T fluids in
the Early Precambrian were similar to those observed today
and exhibited Eu anomalies >1 and enrichment in LREEs
(Bau and Möller, 1993), features which are characteristic
(Sm/Yb) CN
10.0
Pongola pelites
PAAS
1.0
Pongola IF
Penge IF
Kuruman IF
Isua IF
Pacific seawater
high-T hydrothermal
fluids
0.3
0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 5 10 15 20 25
(Eu/Eu*) CN
Fig. 4. Plot of chondrite-normalized Sm/Yb and Eu/Eu * for
Mozaan IFs, and including data for 3.7 Ga Isua IFs (Bolhar et al.,
2004), 2.5 Ga Kuruman and Penge IFs (Bau and Dulski, 1996),
shallow (350 °C, Bau and Dulski, 1999). Note
break in horizontal axis. Isua samples are those considered by
Bolhar et al. (2004) to reflect contemporaneous seawater. The grey
shaded area represents the range of values for 62 pelites from the
Pongola Supergroup sampled by Wronkiewicz (1989), and the
crossed-square represents Post-Archean Average Shale (PAAS,
McLennan, 1989). The Mozaan IFs exhibit Eu/Eu * between that of
the Isua and Kuruman IFs, yet display Sm/Yb values similar to
continental crust and significantly higher than any of the other IFs
(except for two Isua and one Kuruman sample). All IF samples
have significantly lower Sm/Yb and Eu/Eu * than high-T hydrothermal
fluids.

Nd isotopes in 2.9 Ga Archean surface seawater 387
of the Mozaan IF samples (Fig. 4). The presence of positive
Eu anomalies is clear evidence that REY from high-T
hydrothermal inputs influenced the continental shelf environment
that hosted the Mozaan IFs.
For constraints on ancient hydrothermal fluxes, modern
black smoker fluids offer the best analogs for the trace
element compositions of high-T fluids which contributed
to the REY distribution in Archean iron formations.
For the Mozaan IFs, the magnitude of the positive Eu
anomalies generally compare well with data for other
IFs (Fig. 5). However, the general REY distributions of
the Mozaan IFs are significantly different when compared
to well studied examples such as the 3.7 Ga Isua IF of
Greenland (Bolhar et al., 2004) and the 2.5 Ga Kuruman
IF from South Africa (Bau and Dulski, 1996), though
regardless of age, these iron formations display the seawater
characteristics of marine chemical sediments (Fig. 5).
Conservative mixing calculations indicate that a high-T
hydrothermal fluid contribution of less than 0.1% is adequate
for producing the Eu/Sm ratios observed within
the Pongola and Kuruman IFs, and reasonably accounts
for most of the Eu/Sm ratios observed in the Isua IFs
REY/WMS
10 1
10 0
high-T hydrothermal fluid (x10 5 )
10 -1
10 -2
10 -3
Isua IF (x2)
Kuruman IF
top/bottom IFs
middle IFs

388 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
fluxes of trace elements would be expected to affect the
REY distribution of the bottom water component, and
though such fluxes are at best difficult to characterize, it
can be ruled out that such low-T input would have displayed
positive Eu anomalies.
Y/Ho
Y/Ho
Sm/Yb
100
80
60
40
28
20
100
80
60
40
28
20
30
10
1
seawater
hydrogenetic
Fe-Mn crusts
0.1% hydrothermal
fluid 1% hydrothermal
fluid
5% hydrothermal
fluid
high-T
hydrothermal fluids
0.1 1 10
Eu/Sm
seawater
10
0.2 1 10 20
Sm/Yb
hydrogenetic
Fe-Mn crusts
seawater
hydrogenetic
Fe-Mn crusts
Pongola IF (2.9 Ga)
Kuruman IF (2.5 Ga)
Isua IF (~3.7 Ga)
modern seawater
modern hydrogenetic Fe-Mn crusts
high-T hydrothermal fluids
1% hydrothermal
fluid
0.1% hydrothermal
fluid
5% hydrothermal
fluid
high-T
hydrothermal fluids
high-T
hydrothermal fluids
5% hydrothermal
fluid
1% hydrothermal
fluid
0.3
0.1 1
10
Eu/Sm
5.2.3. Surface seawater input
The major remaining REY input affecting the Mozaan
IFs would be Archean surface seawater. Rare earth element
distributions in modern seawater, while exhibiting
local variability and distinct trends with depth, are generally
consistent with the pattern shown in Fig. 5, and exhibit
a (Sm/Yb) CN ratio of 0.8. A study by Elderfield
et al. (1990) reported REY data for five coastal seas
(salinity > 20‰) that exhibited (Sm/Yb) CN ranging from
0.70 to 1.24, consistent with coastal seawater from the
East Frisian islands (North Sea) that displays (Sm/Yb) CN
between 0.61 and 0.73 (Kulaksiz and Bau, 2007). Elderfield
et al., 1990 also presented REY data for waters
from six estuaries and 15 rivers. The estuarine waters
(salinity < 10‰) displayed (Sm/Yb) CN that ranged from
0.63 to 4.74, while (Sm/Yb) CN of the river waters varied
between 0.96 and 4.74. Much of the REY load transported
by modern rivers is removed in estuaries by the
settling of suspended detrital particles, and by the salinity-induced
coagulation and settling of colloidal particles
rich in the particle reactive REYs (e.g., Sholkovitz,
1994), processes that homogenize the widely varying
REY distributions observed in modern rivers to produce
the REY pattern typical of modern seawater. Elderfield
et al. (1990) deduced that the colloidal fraction carried
by many rivers would be enriched in the MREEs
(shale-normalized) relative to the light and heavy REYs,
similar to patterns observed for the Mozaan IFs sampled
from the bottom and top of the sequence (Fig. 2), which
represent iron formation deposited under the shallowest
sea-level conditions. Studies of the Kalix river in Sweden
demonstrated that colloidal particles dominate the transport
and export of rare earth elements (Ingri et al., 2000),
and these colloids consist of two fractions, one Fe-rich
and another C-rich (Andersson et al., 2006). General
enrichments in MREEs for Fe-rich organic colloids have
been observed in the Hudson and Connecticut rivers
(Sholkovitz, 1994), and the presence of MREE-enriched
river waters has also been documented by Sholkovitz
and Szymczak (2000) and Hannigan and Sholkovitz
(2001). The speculation that riverine-derived colloidal
particles enriched in the MREEs could exert a strong
influence on REY distributions in the Mozaan IFs relies
on similar processes controlling REY distributions in
b
Fig. 6. Elemental ratio plots for data sets presented in Fig. 5, with
two-component conservative mixing lines for Eu/Sm, Sm/Yb, and
Y/Ho (symbols for all plots defined in c): (a) Y/Ho versus Eu/Sm,
showing that a 0.1% high-T hydrothermal (>350 °C, Bau and
Dulski, 1999) fluid contribution to waters with shallow (

Nd isotopes in 2.9 Ga Archean surface seawater 389
both modern and Archean rivers and estuaries. However,
such processes have not yet been supported by strong evidence
from the geologic record, and thus the Mozaan IFs
are a first hint at their presence in the Archean.
10
5
depleted mantle
5.3. Sources of Nd to Archean seawater
5.3.1. Nd in Archean iron formations
To further constrain solute sources supplying the margin
of the Kaapvaal craton 2.9 Ga, three of the shale samples
and seven of the iron-formation samples were selected for
Nd isotopic analyses (Table 3). Neodymium isotopic variations
in Archean IFs have been examined by previous
workers in attempts to distinguish the source of REY and
Fe in these sediments. Miller and O’Nions (1985) concluded
that continental Nd dominated the isotopic signature of
Precambrian IFs. Jacobsen and Pimentel-Klose (1988a) discussed
the likelihood that Archean surface seawater displayed
continental signatures derived from fluvial inputs,
but concluded that these inputs were an insignificant source
of Nd and Fe to Archean IFs and that both Nd and Fe were
almost exclusively sourced from a hydrothermal component
similar to the flux emanating from modern mid-ocean
ridges. Subsequent studies also suggested REY in Archean
IFs were derived from sources with depleted-mantle Nd isotopic
ratios (Shimizu et al., 1990), and Alibert and McCulloch
(1993) proposed that 50% of Nd in 2.5 Ga
Hamersley IFs was derived from Archean mid-ocean ridge
sources with Nd (t) = +4. A similar conclusion was drawn
for the 2.5 Ga Kuruman and Penge IFs deposited on the
Kaapvaal craton (Bau et al., 1997).
Neodymium isotopic data for Precambrian IFs vary,
with samples generally possessing Nd (t) between 5 and
+5 (Fig. 7). Therefore, it seems likely that the seawater
from which these IFs precipitated possessed variable Nd
values, and that Archean–Paleoproterozoic oceans commonly
ranged within ±5 Nd (t)-units relative to CHUR.
The IF samples from this study exhibit Nd (2.9 Ga) values
that range over 9 -units, and the Nd isotope ratios observed
in the Mozaan IFs are different from most mid- to
late-Archean IFs, which tend towards chondritic or positive
Nd (t) values. One highly negative Nd (t) value of 8.1 for
3.45 Ga iron formation from the Fig Tree/Moodies sequence
of the Barberton greenstone belt (BGB) was reported
by Miller and O’Nions (1985). This sample (SL 28
B) was a small fragment from an individual magnetite band
within the whole-rock sample SL 28 A, which alternatively
displayed Nd (t) = +1.1, similar to Nd (t) = +0.3 for a second
whole-rock sample analysed by these authors. Two
additional analyses of Fig Tree Group IF by Jacobsen
and Pimentel-Klose (1988b) reported Nd (t) of +2.9 and
+4.1. If the highly negative value of 8.1 reported by Miller
and O’Nions (1985) for sample SL 28 B is atypical of
Barberton IF, then these data suggest that IFs deposited
prior to 3.0 Ga on or near the Kaapvaal craton would likely
display chondritic or slightly positive Nd (t). This Nd isotopic
behavior is also observed in younger South African IFs,
as samples from the 2.5 Ga Kuruman and Penge IFs of the
Transvaal Supergroup display a narrow range of Nd (2.5)
between 0.2 and +1.9 (Bau et al., 1997).
ε Nd (t)
0
-5
-10
-15
continental crust
Mozaan IFs (this study)
Australia
Southern Africa
North America
Greenland
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0
stratigraphic age (Ga)
Fig. 7. Initial Nd (t) values for Precambrian IFs from literature
data. The grey area in the top portion of the figure represents Nd (t)
values predicted for Nd isotopic evolution in a depleted mantle
region with an intial Nd (4.56 Ga) = 0 and Nd (0) = +10, while the
dashed line in the lower portion represents a simple continental
crust trend for Nd isotopic evolution where Nd (4.56 Ga) = 0 and
Nd (0) = 17. In order to facilitate comparison with the present
study, data have been screened for samples described as oxidefacies
IF by authors, and are from: Miller and O’Nions (1985),
Jacobsen and Pimentel-Klose (1988a,b; recalculated using
146 Nd/ 144 Nd = 0.7219), Shimizu et al. (1990), Alibert and McCulloch
(1993, 12 Marra Mamba IFs), and Bau et al. (1997). Three
analyses of Isua IFs by Shimizu et al. (1990) which display
Nd (3.7) > +10 are not shown. Most Archean oxide facies iron
formations that appear to have retained a primary isotopic Nd
signature cluster within ±5 -units of Nd (t) = 0 (chondritic
evolution), with notable exceptions being samples from Isua, and
two samples from this study (see text). Nd isotopic data from Frei
et al. (1999) for Isua are not included, as they represent additional
analyses of sample 242573 which was originally described by Miller
and O’Nions (1985). Sample 242573 provided all but one Nd
isotope analyses of Isua IFs as reported by Miller and O’Nions
(1985) and Frei et al. (1999), and displays Nd (t) values that range
from 8.6 to +14.8, leading Frei et al. (1999) to suggest that this
sample did not retain a primary Nd isotopic signature.
5.3.2. Nd in Mozaan shales
The shale samples are correlated with respect to their
Sm–Nd isotope systematics (Fig. 8), and these data suggest
that the shales observed relatively closed-system behavior
with respect to their Nd isotopic evolution. The shales have
model ages (T DM ; Jacobsen and Pimentel-Klose, 1988a) of
3.4 ± 0.1 Ga, that agree well with U–Pb SHRIMP dates for
individual zircon grains from the siliciclastic Dipka member
underlying the IFs, which are predominantly >3.2 Ga as reported
by Nhleko (2003). However, previous Nd isotope
studies of Mozaan Group fine-grained clastic sediments
have revealed a wide range of T DM ages of 1.52–4.73 Ga
(Dia et al., 1990) and 2.92–4.63 Ga (Jahn and Condie,
1995), leading Jahn and Condie (1995) to surmise that Pongola
pelites had very diverse provenances or that some samples
were affected by open-system behavior during
diagenesis or low-grade metamorphism. This large range
in Sm–Nd isotopic data was not observed, however, by Stevenson
and Patchett (1990), who sampled three Mozaan
shales from a single locality and reported Nd (t) values be-

390 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
143
Nd/
144
Nd
0.5125
0.5120
0.5115
0.5110
0.5105
iron-formation
shale
T = 2.9 Ga
ε Nd = -1
WM8
WM18
0.5100
-50
0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
147
Sm/ 144 Nd
T = 2.9 Ga
ε Nd = -5
WM41
WM131
Fig. 8. Analytical results for Nd isotope analyses of Mozaan shales
and iron-formation samples, including Nd (0) values. Reference
isochrons for Nd (t) of 1 and 5 bracket all samples from this
study, except for IF samples WM41 and WM131. The similarity in
the Nd isotopic evolution for the majority of the IFs and the three
shale samples suggests that REY sources for the shallow seawater
which precipitated the Mozaan IFs were predominately derived
from continental crust low in radiogenic Nd.
tween 1.9 to 0.9 (Fig. 8). This suggests that high-resolution
sampling (as in this study) of Mozaan fine-grained clastic
sediment provides consistent Nd isotopic data
supporting relatively homogeneous provenances over short
geologic time spans.
Sample WM8 exhibits a similar Si concentration compared
to the other three Mozaan shales, but has a much
higher Fe content (29% compared to 6–10%, respectively),
and a correspondingly lower Al content. Since this alumininous
Fe-rich sample resides between two IF bands, it is
possible that its major element composition reflects dilution
with IF-type precipitates, and the Fe increase occurred at
the expense of aluminium-rich phases. Samples WM8 and
WM18 have Sm and Nd concentrations that differ by more
than a factor of six, yet display very similar 147 Sm/ 144 Nd
and 143 Nd/ 144 Nd.
5.3.3. Nd in Mozaan iron formations
The good agreement in Sm–Nd isotope systematics between
the shales and the majority of the IF samples
(Fig. 8) argues against a significant REY component derived
from a depleted mantle source contributing to local
seawater. However, the IF samples do display the positive
shale-normalized Eu anomalies typical of Archean IFs,
which requires some high-T hydrothermal REY input. It
appears that such a high-T input, however, did not overwhelm
the terrestrial flux in terms of the Nd isotope signature.
The relationship between Eu anomalies and Sm–Nd
isotopic signatures in Archean iron-formations was examined
by Derry and Jacobsen (1990), who considered twocomponent
conservative mixing between East Pacific Rise
hydrothermal fluids and estimated river water. These
authors concluded that Eu fractionation due to scavenging
-10
-20
-30
-40
ε Nd (0)
of Eu 2+ near vent sites renders these anomalies unsuitable
for mass balance calculations in IFs and that more reliable
mass fraction estimates are obtained using the Nd isotopic
mass balance. However, the inability to reconcile Nd and
Eu mass balance calculations from the mixing of hydrothermal
and river/seawater endmembers is expected, as only the
Nd isotopic signature provides information regarding the
relative mass fractions of the REY sources; Eu anomalies
only provide information regarding the temperature of
REY sources. An extreme example of the lack of correlation
between mantle-like Nd isotopic signatures and positive
Eu anomalies in high-T marine hydrothermal fluids
(and hence, IFs) may be found in the data of Piepgras
and Wasserburg (1985). These authors reported Nd isotopic
data for a hydrothermal system developed on a sediment
covered ridge in the Guaymas Basin. This system
vented hydrothermal fluid in excess of 300 °C, which migrated
through a sedimentary package several hundred meters
thick and possessed a positive Eu anomaly coupled
with Nd (0) = 11.4, indicating that marine fluids hot enough
to fractionate Eu from other REY will display Nd isotopic
signatures similar to the host rock (in this case,
continentally-derived sediments). This illustrates that positive
Eu anomalies cannot be used to constrain MORB-derived
hydrothermal inputs, as they provide little
information regarding the source of REYs to high-T fluids,
but rather reflect only the temperature of the hydrothermal
system.
Of the seven IF samples, five appear to have experienced
an Nd isotopic evolution similar to that of the shales, consistent
with these samples (both shales and IFs) having received
Nd from an isotopically similar source. Two of the
Pongola IF samples, WM41 and WM131, are not consistent
with a single isotopic source or closed-system conditions
following deposition. These samples differ primarily
in their high 147 Sm/ 144 Nd (>0.17), compared to a
147 Sm/ 144 Nd average of 0.13 ± 0.01 for the remaining five
IFs. Considering that WM41 and WM131 represent the
first and last IF beds deposited, respectively, it is unlikely
that post-depositional processes (e.g., metamorphism)
would selectively affect only these samples. The
147 Sm/ 144 Nd average of 0.13 ± 0.01 for five of the seven
Mozaan IFs is identical to the average obtained
(0.13 ± 0.03) from a survey of 61 analyses of predominantly
oxide facies Archean and Paleoproterozoic IFs (Miller and
O’Nions, 1985; Jacobsen and Pimentel-Klose, 1988a,b; Shimizu
et al., 1990; Alibert and McCulloch, 1993; Bau et al.,
1997). Of these literature data, only three samples (two
from the Hamersley Basin and one Barberton greenstone
belt sample) have higher
147 Sm/ 144 Nd than samples
WM31 and WM141. Alibert and McCulloch (1993) reported
data for 26 samples from the Dales Gorge and Joffre
IFs, concluding on the basis of Nd isotopes that these IFs
had suffered a metamorphic resetting of their isotopic system
some 500 Ma after deposition, a conclusion supported
by Rb–Sr and Pb–Pb isotopic analyses of the underlying
Fortescue lavas (Nelson et al., 1992). The 147 Sm/ 144 Nd of
these 26 metamorphosed IF samples averages 0.12 ± 0.02.
It therefore seems that metamorphic events in Archean
IFs capable of altering Nd isotope ratios have little effect

Nd isotopes in 2.9 Ga Archean surface seawater 391
on overall 147 Sm/ 144 Nd values, an observation that supports
the view of Bau (1993), who concluded that REY
exhibited little mobility during all but the most severe
grades of metamorphism. This suggests that the high
147 Sm/ 144 Nd observed in samples WM41 and WM131 is
primary in origin. If these samples represent closed-system
behavior, as the other IF and shale samples appear to, then
they require a source with a lower initial 143 Nd/ 144 Nd (characteristic
of continental crust), coupled with a Sm/Nd ratio
typical of more mafic sources, and as such, they remain
ambiguous.
The IF samples that closely follow the Nd systematics of
the shales display the two distinct types of REY distributions
discussed earlier. WM51 and WM61 exhibit the
MREE to HREE depleted patterns typical at the onset of
IF deposition, whereas the remaining three samples possess
REY patterns that are relatively flat from the MREE to
HREE. The isotopic similarity between the IFs and shales
suggests that soluble REY delivered to the iron-formation
originated from contemporaneous continental crust. The
least radiogenic shale sample displays Nd (2.9 Ga) = 4.2,
so it is reasonable that local continental crust possessed a
similar Nd (2.9 Ga) value. Assuming the mid-ocean ridge
hydrothermal flux displayed Nd (2.9 Ga) = +4, and assigning
the continental flux an Nd (2.9 Ga) value of 4.2, then
Nd isotopic mass balance calculations require that between
72% and 100% of the Nd in these IF samples ( Nd (2.9 Ga)
of 1.9 to 4.3) must have originated from a source(s) possessing
an Nd isotopic signature similar to local continental
crust. If this was the case, this source should possess a REY
distribution similar to that of the Mozaan shale average
WMS, and deviations from the WMS pattern within a pure
chemical sediment would originate from processes occurring
solely within the water column (i.e., chemical complexation
and scavenging on Fe-oxyhydroxide particles, etc.).
The geologic evidence for a transgressive–regressive cycle
of deposition, coupled with the evolution of the REY patterns
in the Mozaan IFs with stratigraphic position, suggests
that varying proportions of a continentally-derived
solute source dominated trace element evolution in shallow
seawater near the Kaapvaal craton 3.0 Ga ago. Therefore,
it is possible that the MREE-enriched patterns observed in
modern low salinity coastal seas and estuaries (Elderfield
et al., 1990) represent processes analogous to those that
controlled the REY distributions observed in the Mozaan
IFs. Unlike modern environments, the importance of organic
compounds and organic-based colloidal particles on
REY distributions in Archean marine settings is difficult
to estimate, though the presence of stromatolitic carbonate
in the underlying Nsuze Group (Mason and von Brunn,
1977; Beukes and Lowe, 1989) indicates that the margin
of the Kaapvaal craton was colonized by microbial communities
prior to deposition of the Mozaan IFs. It is therefore
not unreasonable to expect that organic constituents exerted
some effect on local geochemical processes, and the
above observations are consistent with non-redox sensitive
processes similar to modern ones operating in shallow seas
on Archean continental margins.
The negative Nd (2.9 Ga) values for the Mozaan IFs are
not entirely unusual, however, as supported by negative
Nd (t) values for some younger IFs, as well as by recent
investigations of Archean shallow water carbonates from
southern Africa (Bolhar et al., 2002; Kamber et al., 2004).
These studies concluded that near-shore seawater displayed
REY and Nd isotope signatures indicative of local solute
inputs derived from continental weathering, and offer further
support suggesting that shallow seawater during the
mid- to late-Archean was strongly influenced by continentally-derived
solute fluxes.
A final comment regarding the correlation between Sm–
Nd isotope ratios and Fe in Archean IFs is warranted. Previous
workers have calculated the mass fraction of Fe derived
from different sources by conservatively mixing a
hydrothermal endmember possessing a depleted-mantle
Nd isotope ratio with a river/seawater endmember which
possesses a continental Nd isotope signature (e.g., Miller
and O’Nions, 1985; Jacobsen and Pimentel-Klose, 1988a).
While conclusions have varied regarding the dominant Fe
source, these attempts have all relied on assumptions
regarding the Fe/Nd ratio of the endmembers, and crucial
to the models is the requirement that Fe/Nd did not change
significantly in the Archean ocean. However, there existed
processes in the Archean ocean, such as pyrite precipitation
close to hydrothermal vent sites, and IF precipitation itself,
for example, that fractionated Fe and Nd. Moreover, Fe
and Nd behavior in the continentally-derived flux produces
even more ambiguity due to uncertainties in various parameters
such as the pH of river water, local redox conditions,
and possible involvement of biological processes. As a result,
the current state of knowledge precludes definitive
statements regarding the Fe source to Archean IFs based
upon Nd isotope systematics.
6. CONCLUSIONS
All of the Mozaan iron-formation samples exhibit major
and trace element characteristics typical of very pure marine
chemical sediments, including shale-normalized positive
La, Gd, and Y anomalies, which are fully consistent with a
marine depositional setting. Positive Eu anomalies in all IF
samples indicate a high-T hydrothermal source supplied
REYs to the seawater which precipitated the Mozaan IFs.
The magnitude of this anomaly is consistent with the general
trend of decreasing Eu/Eu * with decreasing age for Precambrian
IFs and is within the range for other Archean IFs,
though it cannot be used to quantify the magnitude of the
high-T hydrothermal flux. The lack of Ce anomalies in all
of the IF samples indicates that local, shallow Archean seawater
was too reducing to permit oxidation of Ce(III) to
Ce(IV), implying relatively low redox levels in the Archean
atmosphere compared to present values.
When compared to many other Archean IFs, the Pongola
samples have higher Sm/Yb ratios and appear to record a systematic
variation in water depth on the margin of the Kaapvaal
craton during IF sedimentation. Dominant factors
influencing the REY distribution of coastal seawater in the
Archean would include atmospheric composition (e.g.,
qCO 2 ) and fluxes of solutes from terrestrial sources. Isotope
systematics of Sm–Nd reveal similarity between iron-formation
and contemporaneous shale, arguing against an Arche-

392 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
an mid-ocean ridge hydrothermal system as the dominant
REY source for these IFs. Rare earth elements in both shale
and IFs appear to be derived primarily from a relatively
homogenous cratonic source older than 3.2 Ga, and consistent
Nd isotopic behavior can be observed in Pongola sediments
that sample relatively short geologic time spans.
These data imply that solutes within shallow seawater along
Archean cratonic margins were sourced primarily from
weathering of continental crust. The Nd isotopic data for
the Mozaan shales (T DM of 3.4 ± 0.1 Ga) and the depositional
age of these sediments suggests that local continental
crust was stable on 100 Ma time-scales, and when coupled
with the laterally extensive nature of the precipitated Mozaan
IFs, this implies that by 2.9 Ga ago a significant amount of
continental crust was exposed to weathering.
ACKNOWLEDGMENTS
We appreciate the assistance of Marina Fischerström and Hans
Schöberg with the Sm–Nd isotopic analyses performed at LIG.
Furthermore, this manuscript significantly benefited from the comments
of A. Bekker and one anonymous reviewer, and their efforts
are gratefully acknowledged. We also thank T. Lyons for helpful
suggestions and editorial guidance.
REFERENCES
Alibert C. and McCulloch M. T. (1993) Rare earth element and Nd
isotopic compositions of the banded iron-formations and
associated shales from Hamersley, Western Australia. Geochim.
Cosmochim. Acta 57, 187–204.
Alibo D. S. and Nozaki Y. (1999) Rare earth elements in seawater:
particle association, shale-normalization, and Ce oxidation.
Geochim. Cosmochim. Acta 63, 363–372.
Anders E. and Grevesse N. (1989) Abudances of the elements:
meteoric and solar. Geochim. Cosmochim. Acta 53, 197–214.
Andersson K., Dahlqvist R., Turner D., Stolpe B., Larsson T.,
Ingri J. and Andersson P. (2006) Colloidal rare earth elements
in a boreal river: changing sources and distributions during the
spring flood. Geochim. Cosmochim. Acta 70, 3261–3274.
Barrett T. J., Fralick P. W. and Jarvis I. (1988) Rare-earth-element
geochemistry of some Archean iron formations north of Lake
Superior, Ontario. Can. J. Earth Sci. 25, 570–580.
Bau M. (1991) Rare earth element mobility during hydrothermal
and metamorphic fluid–rock interaction and the significance of
the oxidation state of europium. Chem. Geol. 93, 219–230.
Bau M. (1993) Effects of syn- and post-depositional processes on
the rare-earth element distribution in Precambrian iron-formations.
Eur. J. Mineral. 5, 257–267.
Bau M. (1999) Scavenging of dissolved yttrium and rare earths by
precipitating iron oxyhydroxide: experimental evidence for Ceoxidation,
Y–Ho fractionation, and lanthanide tetrad effect.
Geochim. Cosmochim. Acta 63, 67–77.
Bau M. and Dulski P. (1992) Small-scale variations of the rareearth
element distribution in Precambrian iron formations. Eur.
J. Mineral. 4, 1429–1433.
Bau M. and Dulski P. (1996) Distribution of yttrium and rare-earth
elements in the Penge and Kuruman iron-formations, Transvaal
Supergroup, South Africa. Precamb. Res. 79, 37–55.
Bau M. and Dulski P. (1999) Comparing yttrium and rare earths in
hydrothermal fluids from the Mid-Atlantic Ridge: implications
for Y and REE behaviour during near-vent mixing and for the
Y/Ho ratio of Proterozoic seawater. Chem. Geol. 155, 77–90.
Bau M. and Möller P. (1993) Rare earth element systematics of the
chemically precipitated component in Early Precambrian iron
formations and the evolution of the terrestrial atmosphere–
hydrosphere–lithosphere system. Geochim. Cosmochim. Acta
57, 2239–2249.
Bau M., Koschinsky A., Dulski P. and Hein J. R. (1996)
Comparison of the partitioning behaviours of yttrium, rare
earth elements, and titanium between hydrogenetic marine
ferromanganese crusts and seawater. Geochim. Cosmochim.
Acta 60, 1709–1725.
Bau M., Höhndorf A., Dulski P. and Beukes N. J. (1997) Sources
of rare-earth elements and iron in Paleoproterozoic ironformations
from the Transvaal Supergroup, South Africa:
Evidence from neodymium isotopes. J. Geol. 105, 121–129.
Bau M., Usui A., Pracejus B., Mita N., Kanai Y., Irber W. and
Dulski P. (1998) Geochemistry of low-temperature water–rock
interaction: evidence from natural waters, andesite, and ironoxyhydroxide
precipitates at Nishiki-numa iron-spring, Hokkaido,
Japan. Chem. Geol. 151, 293–307.
Beukes N. J. and Cairncross B. (1991) A lithostratigraphicsedimentological
reference profile for the late Archaean Mozaan
Group, Pongola Sequence: application to sequence stratigraphy
and correlation with the Witwatersrand Supergroup.
S. Afr. J. Geol. 94, 44–69.
Beukes N. J. and Lowe D. R. (1989) Environmental control on
diverse stromatolite morphologies in the 3000 Myr Pongola
Supergroup, South Africa. Sedimentology 36, 383–397.
Bolhar R., Hofmann A., Woodhead J., Hergt J. and Dirks P.
(2002) Pb- and Nd-isotope systematics of stromatolitic limestones
from the 2.7 Ga Ngezi Group of the Belingwe Greenstone
Belt: constraints on timing of deposition and provenance.
Precamb. Res. 114, 277–294.
Bolhar R., Kamber B. S., Moorbath S., Fedo C. M. and
Whitehouse M. J. (2004) Characterisation of early Archaean
chemical sediments by trace element signatures. Earth Planet.
Sci. Lett. 222, 43–60.
Button A., Pretorius D. A., Jansen H., Stocklmayer V., Hunter D.
R., Wilson J. F., Wilson A. H., Vermaak C. A., Lee C. A. and
Stagman J. G. (1981) The cratonic environment: the Pongola
Supergroup. In Precambrian of the Southern Hemisphere (ed. D.
R. Hunter). Elsevier Scientific Publications Co., Amsterdam,
pp. 501–510.
Danielson A., Möller P. and Dulski P. (1992) The europium
anomalies in banded iron formations and the thermal history of
the oceanic crust. Chem. Geol. 97, 89–100.
DePaolo D. J. and Wasserburg G. J. (1976) Nd isotopic variations
and petrogenetic models. Geophys. Res. Lett. 3, 249–252.
Derry L. A. and Jacobsen S. B. (1990) The chemical evolution of
Precambrian seawater: evidence from rare earth elements in
banded iron formations. Geochim. Cosmochim. Acta 54, 2965–
2977.
Dia A., Allègre C. J. and Erlank A. J. (1990) The development of
continental crust through geological time: the South African
case. Earth Planet. Sci. Lett. 98, 74–89.
Dulski P. (2001) Reference materials for geochemical studies: new
analytical data by ICP-MS and critical discussion of reference
values. Geostand. Newslett. 25, 87–125.
Elderfield H. (1988) The oceanic chemistry of the rare-earth
elements. Phil. Trans. R. Soc. London Ser. A 325, 105–126.
Elderfield H., Upstill-Goddard R. and Sholkovitz E. R. (1990) The
rare earth elements in rivers, estuaries, and coastal seas and
their significance to the composition of ocean waters. Geochim.
Cosmochim. Acta 54, 971–991.
Erel Y. and Stolper E. M. (1993) Modeling of rare-earth element
partitioning between particles and solution in aquatic environments.
Geochim. Cosmochim. Acta 57, 513–518.

Nd isotopes in 2.9 Ga Archean surface seawater 393
Frei R., Bridgewater D., Rosing M. and Stecher O. (1999)
Controversial Pb–Pb and Sm–Nd isotope results in the early
Archean Isua (West Greenland) oxide iron formation: preservation
of primary signatures versus secondary disturbances.
Geochim. Cosmochim. Acta 63, 473–488.
Fryer B. J. (1977) Rare earth evidence in iron formations for
changing Precambrian oxidation states. Geochim. Cosmochim.
Acta 41, 361–367.
Grauch R. I. (1989) Rare earth elements in metamorphic rocks. In
Geochemistry and Mineralogy of Rare Earth Elements (eds. B.
R. Lipin and G. A. McKay); Rev. Mineral. 21, 147–167,
Mineral. Soc. Amer.
Gross G. A. (1983) Tectonic systems and the deposition of ironformation.
Precamb. Res. 20, 171–187.
Hannigan R. E. and Sholkovitz E. R. (2001) The development of
middle rare earth element enrichments in freshwaters: weathering
of phosphate minerals. Chem. Geol. 175, 495–508.
Hegner E., Kröner A. and Hofmann A. W. (1984) Age and isotope
geochemistry of the Archaean Pongola and Usushwana suites
in Swaziland, southern Africa: a case for crustal contamination
of mantle-derived magma. Earth Planet. Sci. Lett. 70, 267–279.
Hunter D. R. and Wilson A. H. (1988) A continuous record of
Archaean evolution from 3.5 Ga to 2.6 Ga in Swaziland and
northern Natal. S. Afr. J. Geol. 91, 57–74.
Ingri J., Widerlund A., Land M., Gustafsson O., Andersson P. and
Ohlander B. (2000) Temporal variations in the fractionation of
the rare earth elements in a boreal river; the role of colloidal
particles. Chem. Geol. 166, 23–45.
Jacobsen S. B. and Pimentel-Klose M. R. (1988a) A Nd isotopic
study of the Hammersley and Michipicoten banded iron
formations: The source of REE and Fe in Archean oceans.
Earth Planet. Sci. Lett. 87, 29–44.
Jacobsen S. B. and Pimentel-Klose M. R. (1988b) Nd isotopic
variations in Precambrian banded iron formations. Geophys.
Res. Lett. 15, 393–396.
Jahn B. and Condie K. C. (1995) Evolution of the Kaapvaal Craton
as viewed from geochemical and Sm–Nd isotopic analyses of
intracratonic pelites. Geochim. Cosmochim. Acta 59, 2239–2258.
James H. L. and Trendall A. F. (1982) Banded iron formation:
distribution in time and paleoenvironmental significance. In
Mineral Deposits and Evolution of the Biosphere (eds. H. D.
Holland and M. Schidlowski). Springer, Berlin, pp. 199–217.
Kamber B. S. and Webb G. E. (2001) The geochemistry of late
Archaean microbial carbonate: Implications for ocean chemistry
and continental erosion history. Geochim. Cosmochim. Acta
65, 2509–2525.
Kamber B. S., Bohlar R. and Webb G. E. (2004) Geochemistry of
late Archaean stromatolites from Zimbabwe: evidence for
microbial life in restricted epicontinental seas. Precamb. Res.
132, 379–399.
Kato Y., Ohta I., Tsunematsu T., Watanabe Y., Isozaki Y.,
Maruyama S. and Imai N. (1998) Rare earth element variations
in mid-Archean banded iron formations: implications for the
chemistry of ocean and continent and plate tectonics. Geochim.
Cosmochim. Acta 62, 3475–3497.
Khan R. M. K., Das Sharma S., Patil D. J. and Naqvi S. M. (1996)
Trace, rare-earth element, and oxygen isotopic systematics for
the genesis of banded iron-formations: evidence from the
Kushtagi schist belt, Archaean Dharwar Craton, India. Geochim.
Cosmochim. Acta 60, 3285–3294.
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, 1733–1774.
Klemm D. D. (2000) The formation of Palaeoproterozoic banded
iron formations and their associated Fe and Mn deposits, with
reference to the Griqualand West deposits, South Africa. J. Afr.
Earth Sci. 30, 1–24.
Kulaksiz S. and Bau M. (2007) Contrasting behaviour of anthropogenic
gadolinium and natural rare earth elements in estuaries
and the gadolinium input into the North Sea. Earth Planet. Sci.
Lett. 260, 361–371.
Mason T. R. and von Brunn V. (1977) 3-Gyr-old stromatolites
from South Africa. Nature 266, 47–49.
Matthews P. E. (1967) The pre-Karroo formations of the White
Umfolozi inlier, northern Natal. Trans. Geol. Soc. S. Afr. 70,
39–63.
McLennan S. M. (1989) Rare earth elements in sedimentary rocks:
influence of provenance and sedimentary processes. In Geochemistry
and Mineralogy of Rare Earth Elements (eds. B.R.
Lipin and G.A. McKay); Rev. Mineral. 21, 169–200, Mineral.
Soc. Amer.
McLennan S. M., Taylor S. R. and Kröner A. (1983) Geochemical
evolution of Archean shales from South Africa. I. The
Swaziland and Pongola Supergroups. Precamb. Res. 22, 93–
124.
Michard A. (1989) Rare earth element systematics in hydrothermal
fluids. Geochim. Cosmochim. Acta 53, 745–750.
Miller R. G. and O’Nions R. K. (1985) Sources of Precambrian
chemical and clastic sediments. Nature 314, 325–330.
Morris R. C. (1993) Genetic moelling for banded iron-formation of
the Hamersley Group, Pilbara Craton, Western Australia.
Precamb. Res. 60, 243–286.
Nelson D. R., Trendall A. F., de Laeter J. R., Grobler N. J.
and Fletcher I. R. (1992) A comparative study of the
geochemical and isotopic systematics of late Archean flood
basalts from the Pilbara and Kaapvaal cratons. Precamb.
Res. 54, 231–256.
Nhleko N. (2003) The Pongola Supergroup in Swaziland, Ph.D.
dissertation Rand Afrikaans University.
Nothdurft L. D., Webb G. E. and Kamber B. S. (2004) Rare earth
element geochemistry of Late Devonian reefal carbonates,
Canning Basin, Western Australia: confirmation of a seawater
REE proxy in ancient limestones. Geochim. Cosmochim. Acta
68, 263–283.
Pichler T., Veizer J. and Hall G. E. M. (1999) The chemical
composition of shallow-water hydrothermal fluids in Tutum
Bay, Ambitle Island, Papua New Guinea and their effect on
ambient seawater. Mar. Chem. 64, 229–252.
Piepgras D. J. and Wasserburg G. J. (1980) Neodymium isotopic
variations in seawater. Earth Planet. Sci. Lett. 50, 128–138.
Piepgras D. J. and Wasserburg G. J. (1985) Strontium and
neodymium isotopes in hot springs on the East Pacific rise
and Guaymas basin. Earth Planet. Sci. Lett. 72, 341–356.
Shields G. and Stille P. (2001) Diagenetic constraints on the use of
cerium anomalies as palaeoseawater redox proxies: an isotopic
and REE study of Cambrian phosphorites. Chem. Geol. 175,
29–48.
Shields G. A. and Webb G. E. (2004) Has the REE composition of
seawater changed over geological time? Chem. Geol. 204, 103–
107.
Shimizu H., Umemoto N., Masuda A. and Appel P. W. U. (1990)
Sources of iron-formations in the Archean Isua and Malene
supracrustals, West Greenland: evidence from La–Ce and Sm–
Nd isotopic data and REE. Geochim. Cosmochim. Acta 54,
1147–1154.
Sholkovitz E. R. (1994) The aquatic chemistry of rare earth
elements in rivers and estuaries. Aqua. Geochem. 1, 3–34.
Sholkovitz E. and Szymczak R. (2000) The estuarine chemistry of
rare earth elements: comparison of the Amazon, Fly, Sepik and
the Gulf of Papua systems. Earth Planet. Sci. Lett. 179, 299–
309.

394 B.W. Alexander et al. / Geochimica et Cosmochimica Acta 72 (2008) 378–394
Stevenson R. K. and Patchett P. J. (1990) Implications for the
evolution of continental crust from Hf isotope systematics of
Archean detrital zircons. Geochim. Cosmochim. Acta 54, 1683–
1697.
Tankard A. J., Jackson M. P. A., Eriksson K. A., Hobday D. K.,
Hunter D. R. and Minter W. E. L. (1982) Crustal Evolution of
Southern Africa. Springer Verlag, New York.
Taylor S. R. and McLennan S. M. (1985) The continental crust: its
composition and evolution. Blackwell Scientific, Oxford.
Towe K. M. (1991) Aerobic carbon cycling and cerium oxidation:
significance for Archean oxygen levels and banded ironformation
deposition. Palaeogeo. Palaeoclim. Palaeoecol. 97,
113–123.
von Brunn V. and Hobday D. K. (1976) Early Precambrian tidal
sedimentation in the Pongola Supergroup of South Africa. J.
Sediment. Petrol. 46, 670–679.
von Brunn V. and Mason T. R. (1977) Siliciclastic-carbonate tidal
deposits from the 3000 M.Y. Pongola Supergroup, South
Africa. Sediment. Geol. 18, 245–255.
Watchorn M. B. (1980) Fluvial and tidal sedimentation in the
3000 Ma Mozaan basin, South Africa. Precamb. Res. 13, 27–42.
Webb G. E. and Kamber B. S. (2000) Rare earth elements in
Holocene reefal microbialites: a new shallow seawater proxy.
Geochim. Cosmochim. Acta 64, 1557–1565.
Weilers B. F. (1990) A review of the Pongola Supergroup and its
setting on the Kaapvaal craton. Econ. Geol. Res. Unit, Info.
Cir. No. 228, Univ. Witwatersrand.
Wheat C. G., Mottl M. J. and Rudnicki M. (2002) Trace element and
REE composition of a low-Temperature ridge-flank hydrothermal
spring. Geochim. Cosmochim. Acta 66, 3693–3705.
Wronkiewicz, D. J. (1989) Geochemistry and mineralogy of mid-
Archean to mid-Proterozoic sediments from the Kaapvaal
craton, southern Africa: source area provenance, weathering
tectonic setting and crustal evolution. Ph.D. dissertation New
Mexico Inst. Mining Tech.
Associate editor: Timothy W. Lyons

THIS PAGE INTENTIONALLY LEFT BLANK
138

Chapter 4. Neodymium isotopes in Archean seawater and implications for the
marine Nd cycle in Earth’s early oceans
139

THIS PAGE INTENTIONALLY LEFT BLANK
140

Author's personal copy
Earth and Planetary Science Letters 283 (2009) 144–155
Contents lists available at ScienceDirect
Earth and Planetary Science Letters
journal homepage: www.elsevier.com/locate/epsl
Neodymium isotopes in Archean seawater and implications for the
marine Nd cycle in Earth's early oceans
Brian W. Alexander a, ⁎, Michael Bau a , Per Andersson b
a Earth and Space Sciences, Jacobs University Bremen, Campus Ring 8, 28759 Bremen, Germany
b Laboratory for Isotope Geology, Swedish Museum of Natural History, Box 50007, 104 05 Stockholm, Sweden
article
info
abstract
Article history:
Received 31 August 2008
Received in revised form 7 April 2009
Accepted 7 April 2009
Available online 9 May 2009
Editor: M.L. Delaney
Keywords:
Pietersburg
Archean
seawater
neodymium
Isua
Published neodymium (Nd) isotopic data for Archean iron-formations (IF) suggest that, overall, seawater
throughout the Archean typically displayed 143 Nd/ 144 Nd close to bulk Earth values, with Є Nd (t) between
−1.5 and +2.5. Neodymium isotopic ratios in seawater during deposition of the ~3.8 Isua (Greenland) IF
likely displayed positive Є Nd (3.8 Ga) of +2.5, as suggested by IF-G, an Isua reference IF that is considered the
best archive for Early Archean seawater. Seawater 143 Nd/ 144 Nd ratios dominated by radiogenic Nd (positive
Є Nd (t)) seem to have persisted for much of the Archean, as IF from the Pietersburg greenstone belt, South
Africa, suggest seawater Є Nd (2.95 Ga)≥+1. However, similarly aged (~2.9 Ga) IFs from South Africa indicate
that significant variations in seawater 143 Nd/ 144 Nd occurred, and clearly show influences from isotopically
distinct crustal sources. These variations are apparently related to depositional environment, with cratonic
margin, shallow-water IFs possessing a continental Є Nd (t) of−3, while IFs associated with sub-aqueous
mafic volcanics display more radiogenic, positive Є Nd (t) values. Such variation in seawater 143 Nd/ 144 Nd is not
possible in an isotopically well-mixed ocean, and similar to today, it appears that marine Nd cycling in the
Archean produced water masses with distinct Nd isotopic ratios. Since the presence of banded ironformations
requires a reducing Archean ocean capable of transporting Fe, metal-oxide precipitation and
scavenging processes near deep sea hydrothermal vent systems would not have scavenged mantle Nd, i.e., Nd
sourced from alteration of oceanic crust. We propose that bulk anoxic seawater prior to 2.7 Ga possessed
relatively constant positive Є Nd (t) of +1 to +2, whereas local shallow-water masses associated with
exposed evolved crust could possess distinctly different, lower Є Nd (t).
© 2009 Elsevier B.V. All rights reserved.
1. Introduction
Several lines of evidence suggest that the hydrothermal alteration
of oceanic crust and its impact on ocean chemistry in the first half of
Earth's history was significantly greater than today. These include
Archean (N2.5 Ga) marine carbonates with higher manganese (Mn)
and iron (Fe) concentrations and low mantle-like strontium (Sr)
isotopic ratios (e.g., Veizer et al., 1989; Eglington et al., 2003).
Additionally, the presence of positive europium (Eu) anomalies in
Archean marine precipitates (carbonates and iron-formations) similar
to modern high-temperature ‘black-smoker’ fluids has been cited as
evidence of an increased mantle-derived hydrothermal component
within Archean seawater (e.g., Fryer et al., 1979; Appel, 1983; Derry
and Jacobsen, 1990; Danielson et al., 1992; Kamber and Webb, 2001).
These observations would be consistent with models of higher heat
production (e.g., McKenzie and Weiss, 1975) and higher sea-floor
spreading rates (e.g., Abbott and Hoffman, 1984) during the Archean,
which in turn implies increased seawater fluxes through oceanic crust.
⁎ Corresponding author. Tel.: +49 421 200 3154; fax: +49 421 200 3229.
E-mail addresses: b.alexander@jacobs-university.de, m.bau@jacobs-university.de
(B.W. Alexander), per.andersson@nrm.se (P. Andersson).
Attempts to resolve the relative importance of Archean marine
hydrothermal fluxes have been limited by the availability of suitable
seawater archives. Archean carbonates are rare (Veizer and Mackenzie,
2004), and thick well-preserved sequences only appear ca.
~2.7 Ga ago (cf. Grotzinger, 1989). Therefore, studies regarding hydrothermal
impacts upon Archean–Paleoproterozoic seawater chemistry
have often used banded iron-formations (IFs) as seawater archives
(e.g., Fryer et al., 1979; Jacobsen and Pimentel-Klose, 1988a; Derry and
Jacobsen, 1990; Alibert and McCulloch, 1993). Iron-formations are
chemical precipitates whose ubiquitous presence throughout the
Archean suggests an increased hydrothermal fluid flux, as modern
vent fluids at oceanic spreading ridges are reducing and Fe-rich, which
are necessary properties of the fluid which ultimately produced IFs.
The conclusion that most IFs precipitated from seawater is generally
accepted and indicated from different lines of evidence (cf. Simonson,
2003). These include the frequent association of IF with marine
carbonates and marine transgressions (Klein and Beukes, 1989;
Simonson and Hassler, 1996), and geochemical evidence such as
rare earth element (REE) distributions in many IF (Bau and Dulski,
1996; Bolhar et al., 2004) that are remarkably similar to both
Archean–Paleoproterozoic marine carbonates (Kamber and Webb,
2001; Bau and Alexander, 2006) and modern seawater.
0012-821X/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsl.2009.04.004

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
145
However, the exact process of IF deposition is unclear, as it implies
contradictory mechanisms in which seawater is reducing enough to
transport large amounts of soluble Fe(II), yet oxidizing enough to
episodically precipitate sediments that may contain N90% Fe (measured
as Fe 2 O 3 ). Therefore, at least with respect to the marine Fe cycle,
Archean seawater during IF deposition was clearly distinct from
modern seawater, and several studies have attempted to quantify the
importance of an Fe-rich hydrothermal flux by examining samarium–
neodymium (Sm–Nd) isotopes in IFs (e.g., Miller and O'Nions, 1985;
Jacobsen and Pimentel-Klose; 1988a, b; Derry and Jacobsen, 1990;
Alibert and McCulloch; 1993; Bau et al., 1997a; Alexander et al., 2008).
This application of Sm–Nd isotopes is possible because fractionation
during crustal differentiation produces different Sm/Nd ratios in
oceanic and continental crust, resulting in unique 143 Nd/ 144 Nd ratios
in these reservoirs due to the subsequent decay of 147 Sm to 143 Nd.
Measured 143 Nd/ 144 Nd ratios are typically expressed using the Є Nd (t)
notation (DePaolo and Wasserburg, 1976), which describes the deviation
of 143 Nd/ 144 Nd in parts per 10 4 from CHUR (chondritic uniform
reservoir, i.e., bulk silicate Earth) at time t. Continental crust, with a
low Sm/Nd ratio and little radiogenic 143 Nd, displays negative Є Nd (t),
whereas oceanic crust derived from a depleted mantle has a higher
Sm/Nd ratio, more radiogenic 143 Nd, and positive Є Nd (t) values.
Therefore, Sm–Nd isotope ratios in a sample of known age can distinguish
whether Nd has been primarily derived from continental or
oceanic crust.
The purpose of this study is to screen IFs for seawater provenance
on the basis of REE patterns, and then re-evaluate the isotopic evolution
of Nd in Archean seawater. Though data between 3.0 and 3.5 Ga
are few, it seems that seawater possessed relatively constant Є Nd (t)of
approximately +1 to +2 from 3.8 Ga until ~2.6 Ga. New Nd isotopic
results for 2.95 Ga IF from the Pietersburg greenstone belt (PGB) in
South Africa support this conclusion, and the PGB samples show strong
evidence for mixing between a detrital sediment source with slightly
negative Є Nd (t) and ambient seawater with positive Є Nd (t) equal to or
greater than +1. It is concluded that, unlike modern seawater, the
great majority of Nd in bulk seawater prior to 2.7 Ga originated from
mantle-derived mafic source rocks. However, comparison of the
Pietersburg IF to similarly aged IF from South Africa indicates that
where evolved local continental crust was present this mantle Nd
signal was not discernible in shallow Archean seawater (Alexander
et al., 2008), implying that like modern oceans, the Archean ocean was
not well-mixed with respect to its Nd isotopic composition.
2. Nd in seawater and previous IF studies
Seawater in modern oceans possesses negative Є Nd (0) between
−1 and −20, though extreme values in this range are restricted to
shallow waters (Frank, 2002; Goldstein and Hemming, 2003). There is
no evidence that radiogenic Nd derived from hydrothermal alteration
of mid-ocean ridge basalt contributes significantly to modern seawater,
and in fact, efficient scavenging of REE by precipitating metal
oxides near hydrothermal vent sites is considered to provide an
ultimate sink for REE in seawater (German et al., 1990). As a result,
Nd derived from continental crust dominates world seawater (e.g.,
Piepgras and Wasserburg, 1980; Andersson et al., 2008, and references
therein). The persistence of a ~20 Є Nd -unit variation between ocean
basins indicates that the marine residence time of Nd (τ Nd ) is less than
the mixing time of the oceans (~1500 yr, e.g., Broecker and Peng,
1982), and estimates of τ Nd in modern oxic seawater based on isotopic
studies typically range from ~500 yr (Tachikawa et al., 2003) to 1000–
2000 yr (Jeandel et al., 1995;). Temporal variability in seawater Є Nd (t)
has also been noted, and Recent (b20 Ka) seawater Nd ratios appear to
have fluctuated by 2–3 Є Nd -units on 1000–5000 yr time scales (e.g.,
Piotrowski et al., 2005; Gutjahr et al., 2008).
Early Nd isotopic studies of IFs attempted to constrain the source of
Fe to coeval seawater, based upon the premise that initial 143 Nd/ 144 Nd
ratios in IFs would indicate if Fe was derived from hydrothermal fluids
originating in ancient oceanic crust, or if the Fe was sourced from
weathering of contemporaneous continental crust. However, this
approach is problematic as recently discussed by Alexander et al.
(2008), as the different geochemical behaviors of Fe and Nd do not
allow definitive statements to be made regarding the Fe source to IF
based upon Nd isotopes.
Regardless of the Fe source to IFs, the use of Sm–Nd isotopes for
discerning the relative impact of oceanic crust-derived hydrothermal
fluxes to Archean seawater should be possible. Neodymium isotopic
data for IFs have varied, with Miller and O'Nions (1985) concluding
that continental crust supplied much of the Nd to Archean oceans.
However, later studies by Jacobsen and Pimentel-Klose (1988a, b),
Alibert and McCulloch (1993), and Bau et al. (1997a) concluded that
contemporaneous seawater Nd was primarily derived from hydrothermal
alteration of mafic oceanic crust. A recent study of 2.9 Ga IF by
Alexander et al. (2008) indicates that shallow Archean seawater along
cratonic margins was dominated by continentally-derived Nd, similar
to coastal seawater in modern oceans. Whereas early studies have
extrapolated data from relatively small numbers of samples and locations
to be representative of the entire Archean ocean (e.g., Miller and
O'Nions, 1985; Jacobsen and Pimentel-Klose, 1988b), the conflicting
interpretations could reflect spatial and temporal heterogeneity with
respect to 143 Nd/ 144 Nd ratios in Earth's early oceans. This study
addresses the possibility that Nd isotopic heterogeneity existed in the
Archean ocean, similar to what is observed in modern oceans.
It is necessary to examine the possibility that primary Nd isotopic
signals recorded in IFs may be reset by post-depositional processes;
however, the Sm–Nd isotopic system is resistant to metamorphism
and consistent Nd isotopic results are obtained for IFs that have experienced
a wide range of metamorphic overprint. Bau et al. (1997a)
reported Sm–Nd data for the stratigraphically equivalent Kuruman
and Penge IFs of the Transvaal Supergroup in South Africa (Beukes,
1983). Whereas the Kuruman IFs were weakly metamorphosed
(b200 °C) and the Penge IFs suffered intense metamorphic overprint
(N500 °C), both IFs yield very similar Sm–Nd isotopic ratios, and REE
patterns are indistinguishable between the two. Moreover, 2.5–3.7 Ga
IFs from a number of locations and with very different metamorphic
histories display remarkably similar, seawater-like REE distributions
(e.g., Alibert and McCulloch, 1993; Bolhar et al., 2004), and in a
detailed study of worldwide IFs Bau (1993) concluded that postdepositional
REE mobility in IFs is negligible.
3. Geologic setting
The Pietersburg greenstone belt outcrops over a 150 km long by
70 km wide area located in northeast South Africa (Fig. 1). The
northeastern portion of the PGB consists of schists, is bounded by the
Limpopo metamorphic province, and experienced the highest degrees
of deformation and lower to upper amphibolite grade metamorphism
(De Wit et al., 1992). The southwestern portion of the belt is
surrounded by igneous intrusives and younger cover of the Traansvaal
Supergroup, and experienced less tectonic deformation and lower
degrees of metamorphism (De Wit et al., 1992). The PGB contains two
major tectonic units separated by a distinct unconformity, above
which lies the sedimentary Uitkyk Formation (De Wit et al., 1992).
Below the unconformity, the South African Committee for Stratigraphy
(SACS, 1980) recognized five formations, from oldest to youngest,
the Mothiba, Ysterberg, Eersteling, Zandrivierspoort, and the Vrischgewaagd.
Only the lowermost Mothiba and Ysterberg Formations
contain IF. However, De Wit (1991) observed significant structural
repetition in the southwestern portion of the PGB, where the IF
samples for this study originated, and concluded that a strict
SACS stratigraphy was not applicable. Therefore, we adopt the stratigraphy
of De Wit (1991) for this area, which labels all units below
the unconformity as simatic basement, due to the silica-magnesia

Author's personal copy
146 B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
Fig. 1. Map showing the location of 2.9–3.5 Ga greenstone belts associated with the eastern margin of the Kaapvaal craton, including the Pietersburg greenstone belt (PGB), its
distribution, and a geologic map of the southwestern region of the PGB (modified after De Wit et al., 1992). The oldest epicratonic sediments in South Africa are located in the
~2.90 Ga Pongola Supergroup, which consists of the older Nsuze Group and the younger, IF-bearing Mozaan Group. Mozaan IF discussed in the text was sampled from the Wit Mfolozi
inlier. Also shown on the detailed map are the type localities for the Pietersburg geologic formations.
lithologies that dominate. The simatic basement primarily consists of
mafic volcanic rocks including metagabbros, serpentinized peridotites,
and massive pillowed metatholeiites, with accessory interlayered
IF, Fe-rich shales, and minor cherts and carbonates (De Wit,
1991).
Few radiometric age data exist for the PGB. The maximum age of
the Uitkyk Formation is interpreted as ~2.90 Ga, as determined
by three detrital zircon uranium–lead (U–Pb) ages that range from
2901±2 Ma to 2957±8 Ma (De Wit et al., 1993). A minimum age of
2687 ±2 Ma for the PGB is provided by a U–Pb single zircon analysis of
the Uitloop granite (Fig. 1), which cross-cuts all units of the PGB
(De Wit et al., 1993). For the simatic basement below the unconformity
a Pb–Pb date for mafic metavolcanics from the Eersteling area
of 3455±128 Ma was reported by Byron and Barton (1990), but this
age was interpreted as reflecting a secondary metamorphic overprint.
The only other data for the simatic basement consists of two zircon
analyses for a metaquartz porphyry associated with IF in the Ysterberg
area. These analyses yielded a Pb–Pb evaporation age of 2939±26 Ma,
and a precise U–Pb zircon age of 2949±0.2 Ma (Kröner et al., 2000). A
granitoid located 15 km west of the Ysterberg area that intrudes into,
and is deformed with, the lower simatic basement yielded a U–Pb
zircon age of 2958±2 Ma (De Wit et al., 1993). These data indicate
that the oldest sections of the PGB are greater than 2.96 Ga in age, and
the fact that iron-formation is described only in the oldest portions
of the simatic basement (Mothiba and Ysterberg Formations) suggests
that IF deposition occurred no later than 2.95 Ga. Therefore, considering
the precision of the U–Pb date (2949±0.2 Ma) reported by
Kröner et al. (2000) for the IF-bearing sequence in the Ysterberg
Formation, we assign the IF in the southwest portion of the PGB a
minimum age of 2.95 Ga for the purposes of interpreting Nd isotopic
data.
For comparison, the Pietersburg samples are discussed in relation
to IF from the ~2.9 Ga Mozaan Group, which is part of the Pongola
Supergroup located southeast of the PGB (Fig. 1). The Mozaan IF
were part of the first sediments unconformably deposited on the older
Nsuze Group, and are considered ideal archives for shallow Archean
seawater (Alexander et al., 2008). Few age constraints exist for
the Mozaan Group. A rhyolite occurring in the upper third of the
older Nsuze Group yielded a SHRIMP U–Pb date of 2977±5 Ma for
single zircons (Nhleko, 2003), and is the best constraint for the upper
age limit of the Mozaan Group (see also Hegner et al., 1994). The
youngest depositional age for the Mozaan Group has been interpreted
as 2837±5 Ma from a SHRIMP U–Pb zircon age obtained from a
quartz-porphyry sill (Gutzmer et al., 1999). Based on the limited data,
it appears that the oldest parts of the Mozaan Group (the IF) may be as
old as ~2.95 Ga, and these sediments are unlikely to be younger than
~2.84 Ga.
4. Sampling and analytical methods
The Pietersburg IF was sampled from an approximately 50 cm long
continuous section of drill core that was subdivided into seven pieces

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
147
varying between 5 and 10 cm in length and labelled as IF1 to IF7. The
entire length of sampled drill core is highly magnetic. The IF is finely
laminated on millimeter scales and displays thicker banding on
centimeter scales. The longest individual pieces of drill core were
subdivided two or three times and these samples are identified as
such, e.g., IF6, IF6a, and IF6b are individual samples from the ~10 cm
long IF6 piece of drill core. This resulted in a total of fourteen samples,
each representing relatively homogenous ~1 cm thick bands that
contained varying amounts of Fe. Samples were crushed with a steel
jaw crusher and powdered with an agate ball mill. Major element
concentrations were determined by X-ray fluorescence (XRF) at the
SpectRAU laboratory of the University of Johannesburg, South Africa.
Trace element concentrations were determined at Jacobs University
Bremen using inductively-coupled plasma mass spectrometry
(ICPMS) following the methods described by Dulski (2001). Briefly,
for both trace metal and Sm–Nd isotopic analyses, powdered samples
were digested under pressure at high temperature (TN160 °C) in
Teflon vessels using concentrated ultrapure perchloric (HClO 4 ) and
hydrofluoric (HF) acids. Following HClO 4 –HF evaporation the samples
were re-dissolved in hydrochloric acid prior to ICPMS analysis or REE
separation. Analytical quality for major and trace element determinations
was monitored using the iron-formation standard FeR-2
(Geological Survey of Canada), analysis of which produced good
agreement with reference values (Table 1).
Samples for Sm–Nd isotope determinations were spiked prior
to dissolution with a mixed 147 Sm/ 150 Nd spike, and analyzed at the
Laboratory for Isotope Geology (LIG), Swedish Museum of Natural
History. Samarium and Nd were separated using ion exchange chromatography
techniques described in Andersson et al. (2008). The
isotopic ratios were determined with a Thermo Scientific TritonTIMS
(thermal ionization mass spectrometer) using multicollector static
mode. Samarium and Nd were loaded on double rhenium filaments. The
total Nd blank averaged 43 pg for three separate digestion batches and is
an insignificant contribution to the sample isotopic composition. Measured
isotope ratios were reduced assuming exponential fractionation.
Samarium ratios were normalized to 149 Sm/ 152 Sm = 0.516747. Neodymium
was run in static mode using rotating gain compensation, and
calculated ratios were normalized to 146 Nd/ 144 Nd=0.7219.
The La Jolla standard yielded 143 Nd/ 144 Nd=0.511848 ±0.0000047
(2σ, n=32). No corrections were applied to the measured ratios and
Table 1
Major and trace element data for Pietersburg IF.
Sample IF1 IF1a IF2 IF2a IF2b IF3 IF4 IF5 IF5a IF6 IF6a IF6b IF7 IF7a FeR-2 Ref a
(wt.%)
SiO 2 49.3 33.7 61.3 44.7 38.4 43.5 38.8 28.6 38.4 49.1 44.7 61.8 44.8 nd 50.4 49.21
TiO 2 0.04 0.05 0.02 0.07 0.07 0.09 0.07 0.04 0.01 0.02 0.03 0.02 0.25 nd 0.19 0.18
Al 2 O 3 0.73 0.72 0.32 1.06 0.91 1.50 1.53 0.33 0.13 0.26 0.54 0.30 5.30 nd 5.19 5.16
Fe 2 O 3 45.6 60.5 36.8 43.7 56.4 48.3 52.5 68.7 58.8 48.7 49.8 37.0 42.1 nd 39.67 39.21
MnO 0.20 0.28 0.05 0.13 0.11 0.21 0.53 0.79 0.64 0.08 0.08 0.09 0.12 nd 0.12 0.12
MgO 3.23 0.07 1.65 4.25 3.55 3.90 3.19 2.67 1.42 1.84 2.17 2.22 5.15 nd 2.23 2.10
CaO 1.75 1.74 0.89 3.45 1.28 2.13 2.77 1.18 1.43 1.07 1.64 2.79 1.33 nd 2.13 2.17
Na 2 O bd bd bd bd bd bd bd bd bd bd bd bd bd nd 0.39 0.51
K 2 O bd 0.02 bd 0.04 0.03 0.02 0.01 bd 0.01 bd 0.01 0.01 0.09 nd 1.34 1.33
P 2 O 5 0.19 0.23 0.11 0.20 0.24 0.30 0.22 0.21 0.10 0.14 0.21 0.16 0.16 nd 0.27 0.27
S bd bd 0.56 0.24 0.12 0.03 0.11 bd bd 0.57 0.17 0.22 0.54 nd 0.33 0.17
Total 101.04 97.31 101.70 98.84 101.11 99.98 99.73 102.52 100.94 101.78 99.35 104.61 99.84 nd 102.27
mg/kg
Sc 1.81 2.06 b1.30 2.86 3.03 2.83 2.85 b1.30 b1.30 b1.30 b1.30 b1.30 7.20 b1.0 4.94 6
Ti 195 306 101 438 488 384 352 172 47.4 80.7 161 91.9 1284 151 1045 1079
Co 4.11 6.00 1.94 7.73 6.94 6.02 6.24 2.61 1.20 1.58 2.31 1.67 14.3 2.29 6.50 7
Ni 26.2 22.3 12.3 57.6 47.2 33.3 31.0 13.9 5.38 8.94 15.1 11.2 128 15.2 22.0 21
Rb 1.17 1.13 0.703 2.46 2.05 2.00 2.07 1.17 0.368 0.52 0.657 0.498 5.60 0.837 62.6 66
Sr 29.5 28.5 11.0 51.2 15.8 33.8 41.0 24.2 29.8 14.8 25.0 38.4 17.9 21.7 60.8 58
Y 10.3 12.6 4.98 10.5 8.96 15.0 18.3 10.8 5.91 6.01 9.68 7.54 12.7 8.74 12.4 15
Zr 9.88 10.0 4.81 14.4 11.7 16.6 15.6 8.20 1.67 3.87 5.15 2.90 61.4 4.90 36.4 39
Cs 0.366 0.417 0.160 0.633 0.60 0.681 0.661 0.467 0.166 0.144 0.265 0.155 1.45 0.273 4.55 5
Ba 2.13 4.26 1.41 6.13 4.33 3.83 3.03 5.77 1.62 2.07 1.31 1.07 6.26 1.39 223 240
La 4.62 4.96 1.96 4.45 5.28 7.69 9.06 4.47 1.43 2.22 3.42 2.44 8.49 2.60 11.8 14
Ce 8.43 9.14 3.53 8.34 9.56 14.0 18.0 7.87 1.88 3.79 5.56 3.87 17.1 4.38 24.1 25
Pr 1.01 1.12 0.425 1.02 1.11 1.65 2.12 0.942 0.204 0.439 0.633 0.454 2.01 0.505 2.83 3
Nd 4.24 4.89 1.80 4.38 4.68 6.84 8.84 3.91 0.894 1.90 2.76 1.95 8.01 2.20 11.5 12
Sm 0.983 1.16 0.393 0.990 1.08 1.55 2.02 0.890 0.183 0.410 0.592 0.431 1.68 0.504 2.44 2.6
Eu 0.418 0.561 0.190 0.473 0.461 0.688 0.973 0.544 0.121 0.241 0.306 0.276 0.601 0.266 1.20 1.25
Gd 1.27 1.55 0.539 1.31 1.39 1.96 2.55 1.24 0.326 0.594 0.879 0.676 1.87 0.743 2.25 2
Tb 0.192 0.236 0.079 0.211 0.201 0.308 0.417 0.191 0.048 0.090 0.140 0.101 0.294 0.118 0.332 0.32
Dy 1.30 1.59 0.560 1.36 1.25 2.07 2.82 1.31 0.344 0.644 0.960 0.708 1.91 0.809 2.10 2
Ho 0.298 0.357 0.128 0.304 0.272 0.463 0.621 0.300 0.093 0.153 0.237 0.180 0.412 0.195 0.436 0.6
Er 0.910 1.12 0.406 0.957 0.840 1.43 1.92 0.945 0.304 0.489 0.751 0.550 1.30 0.618 1.31 1.5
Tm 0.132 0.158 0.061 0.135 0.118 0.199 0.271 0.136 0.045 0.070 0.107 0.081 0.193 0.089 0.187 0.2
Yb 0.870 1.01 0.368 0.881 0.755 1.26 1.73 0.884 0.276 0.462 0.686 0.515 1.27 0.583 1.23 1.3
Lu 0.137 0.166 0.062 0.143 0.123 0.203 0.268 0.141 0.047 0.076 0.114 0.084 0.200 0.096 0.191 0.2
Hf 0.248 0.258 0.118 0.370 0.291 0.438 0.373 0.220 0.037 0.102 0.135 0.073 1.68 0.114 0.955 1
Ta 0.040 0.044 0.017 0.052 0.046 0.074 0.064 0.036 0.007 0.015 0.022 0.013 0.264 0.020 0.147 0.2
W 50.9 140 21.1 43.4 27.9 61.5 134 31.3 59.9 52.2 74.1 39.2 55.9 86.7 2.36
Pb 1.80 2.09 1.29 2.94 1.94 2.25 2.71 2.34 1.41 1.41 1.37 1.69 2.21 1.27 8.43 11
Th 0.330 0.380 0.155 0.473 0.430 0.523 0.506 0.280 0.066 0.174 0.208 0.124 1.46 0.174 2.44 3
U 0.083 0.102 0.041 0.113 0.108 0.146 0.146 0.079 0.020 0.036 0.053 0.031 0.500 0.050 1.33 1.2
bd=below determination limit.
nd=not determined.
a Ref =values for FeR-2 iron-formation standard as provided by issuing agency (Canada Centre for Mineral and Energy Technology-Mining and Mineral Sciences Laboratories, Abbey
et al., 1983), except for italicized data which are obtained from the GeoReM database (Jochum et al., 2005).

Author's personal copy
148 B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
Table 2
Sm–Nd concentrations (mg/kg) and isotope ratios for Pietersburg IF.
Sample Sm Nd
147 Sm/ 144 Nd
143 Nd/ 144 Nd±2σ a Є Nd (0) Є Nd (t) Sm Nd
TIMS TIMS t=2.95 Ga ICPMS %RSD ICPMS %RSD
TIMS/ICPMS
TIMS/ICPMS
IF1 0.954 4.116 0.1401 0.511502±4 −22.2 −0.66 0.983 −3.0 4.24 −2.9
IF2 0.426 1.915 0.1344 0.511466±5 −22.9 0.83 0.393 8.4 1.80 6.4
IF3 1.525 6.694 0.1377 0.511464±4 −22.9 −0.48 1.55 −1.6 6.84 −2.1
IF4 2.029 8.862 0.1384 0.511485±3 −22.5 −0.35 2.02 0.4 8.84 0.2
IF4r b 2.035 8.874 0.1386 0.511486±3 −22.5 −0.39
IF5 0.847 3.725 0.1375 0.511499±3 −22.2 0.29 0.890 −4.8 3.91 −4.7
IF5r b 0.856 3.747 0.1381 0.511504±4 −22.1 0.14
IF6 0.424 1.922 0.1334 0.511455±5 −23.1 0.99 0.410 3.4 1.90 1.2
IF7 1.717 8.041 0.1291 0.511290 ±9 −26.3 −0.61 1.68 2.2 8.01 0.4
t=3.8 Ga
IF-G c 0.39 1.72 0.1371 0.511277 2.73
IF-G 0.383 1.687 0.1373 0.511258±4 2.25
BCR-2 d 6.574 28.64 0.1388 0.512628 ±2
a
b
c
2σ refers to last digit in the measured ratio.
r in sample name denotes a replicate digestion and analysis of the same sample, and these data are not discussed in the text.
Data from Stecher et al. (1986). IF-G distributed by GIT-IWG (Groupe International de Travail – International Working Group), Service d' Analyse des Roches et des Minéraux,
CRPG-CNRS, Vandoeuvre-lès-Nancy, France. Sample powder from the original IF-G processing run is no longer available, though a second batch of Isua IF-G iron-formation is
currently available from GIT-IWG. Data for this study, and presumably from Stecher et al. (1986) as well, are for IF-G powder from the original processing run.
d Average of 2 separate digestions of BCR-2 Columbia River basalt reference material (United States Geologic Survey).
the overall reproducibility (external precision) is 9 ppm as estimated
from the standards. Uncertainties in TIMS determination of Sm and Nd
concentrations are estimated as b1% and 2%, respectively.
Agreement between TIMS and ICPMS concentration data is better
than 4.7% for both Sm and Nd, except for sample IF2 where TIMS
values are respectively 8.4% and 6.4% higher relative to ICPMS data
(Table 2). Rare earth element ratios calculated from TIMS and ICPMS
data are significantly more consistent, with all samples displaying
Sm/Nd that varies by no more than 2.2% between the two analytical
methods. Therefore, we very conservatively estimate the accuracy for
ICPMS analyses as better than ±10%, and accuracy for rare earth
element ratios as better than ±5%.
Fig. 2. Selected trace element concentrations plotted as a function of Ti. Trends are similar regardless of elements affinity for felsic (Rb, Th) or mafic crust (Ni, Co). The excellent
correlation between Rb and more immobile metals such as Al and Ti argues against significant post-depositional mobility of trace metals following IF deposition.

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
149
Fig. 3. REY distributions in IF samples normalized to Al-rich IF7 (5.3% Al 2 O 3 ). Inset
displays REY distribution in Al-rich IF7 normalized to chondritic meteorite (Anders and
Grevesse, 1989) and continental crust as represented by Post Archean Australian Shale
(PAAS, McLennan, 1989). PAAS possesses a negative Eu anomaly compared to bulk
Earth, and PAAS-normalized data will correspondingly display positive Eu anomalies
when no Eu fractionation exists. This is the case for Al-rich IF7, for which no anomalous
behavior of Eu is observed. The remaining samples all display positive Eu anomalies
relative to IF7 that significantly vary in magnitude, with greater Eu enrichments in
samples of lower total REY abundance. Positive La, Gd, and Y anomalies are present in
many of the samples and, similar to Eu, the magnitude of these anomalies increases
with decreasing REY concentration. The effect of detrital contamination is observed in
samples IF3 and IF4, which have the second highest Al 2 O 3 contents (1.5%) after IF7, and
relatively flat REY patterns. However, IF4 also possesses greater total REY concentrations
that IF7, which coupled with the positive Eu anomaly in IF4 indicates that a significant
fraction of its REY budget was supplied by the chemical precipitate which produced the
IF. Sample IF7a illustrates the large range of REY concentrations (and REY fractionation)
observed on cm-scales within the drill core, as it was obtained from a homogenous band
adjacent to sample IF7. The range of concentrations and REY distributions observed
throughout the core argues against the widespread mobility of REY at any point
following deposition.
nickel (Ni), and scandium (Sc)), including relatively mobile rubidium
(Rb) (Fig. 2). These correlations remain even if the single sample with
the highest Ti and Al content (IF7) is not considered (Fig. 2).
The iron-formations possess REE and yttrium (REY) concentrations
that vary by an order of magnitude between the samples, and the
distribution patterns of these elements vary significantly as well. This
may be illustrated by normalizing REY data to IF7, which possesses the
highest Al 2 O 3 content (5.3%), a procedure that effectively removes the
influence of any detrital material that may be present (Fig. 3). The Alrich
IF7 itself is light rare earth element (LREE) enriched when
normalized to chondrite (subscript CN , Anders and Grevesse, 1989)
and shows gadolinium/ytterbium ratios ((Gd/Yb) CN ) of ~1 and no
REY anomalies (Fig. 3). The two samples with the next highest Al 2 O 3
concentrations (1.5%, IF3 and IF4) have higher REY abundances than
the remaining samples, and REY concentrations change significantly
on small scales (1–2 cm) throughout the core as a function of Al
content. This is illustrated by samples IF5 and IF5a, which represent
individual ≤1 cm thick bands sampled within 2–3 cm of each other,
and that contain Al, Ti, and REY contents that vary by a factor of ~3.
However, a significant fraction of the REY in most of the samples
must be derived from the chemical precipitate which produced the IF.
Sample IF4 has one-third the Al of sample IF7, yet IF4 has higher REY
concentrations (Fig. 3). In the remaining samples four of the REY
display anomalous behavior to varying degrees, with strong positive
europium (Eu) and Y anomalies and smaller positive lanthanum (La)
and Gd anomalies. The IFs also possess slight positive lutetium (Lu)
anomalies, and the magnitude of all anomalies generally increases as
Sm–Nd analytical quality was monitored using the basalt reference
standard BCR-2 (United States Geologic Survey), which yielded 143 Nd/
144 Nd=0.512628±0.000002. Additionally, the Isua iron-formation
standard IF-G, issued by IWG-GIT (International Working Group –
Groupe International de Travail) was analyzed to provide reference
data for subsequent Sm–Nd isotopic studies of iron-formations, and
these data match well with the single previously published Sm–Nd
isotopic analysis of IF-G (Stecher et al., 1986).
5. Results
Analytical data are presented in Tables 1 and 2. Excluding samples
IF2a and IF7, all samples are more than 91% silicon- and iron-oxides
(SiO 2 and Fe 2 O 3 ), with SiO 2 between 29 and 62%, and Fe 2 O 3 between
37 and 69%. Sample IF2a is enriched in calcium (Ca) and magnesium
(Mg) compared to most samples, and IF7 possesses significantly more
aluminum (Al) and Mg than any other sample. While Al contents are
typically very low (b0.8%), consistent with very low amounts of
detrital aluminosilicates, IF3 and IF4 each contain 1.5% Al 2 O 3 , and IF7
contains 5.3% Al 2 O 3 . Trace elements that are considered immobile,
such as zirconium (Zr), hafnium (Hf), and thorium (Th), correlate well
with Al and titanium (Ti), as do most trace metals (e.g., cobalt (Co),
Fig. 4. Nd concentrations and Є Nd (t) of Pietersburg IF samples as a function of Al
content. Top figure (a): Nd concentrations generally decrease with decreasing Al
concentrations. However, significant Nd must be derived from the chemical precipitate
that formed the IFs, as IF4 contains less than one-third the Al of IF7, but higher Nd
concentrations. Bottom figure (b): Iron-formation samples with Al 2 O 3 N0.5% possess
very consistent negative Є Nd (t), whereas positive Є Nd (t) values are only observed in
samples with the lowest Al concentrations, indicating that contamination with minor
detrital aluminosilicates drives Є Nd (t) to more negative values. This suggests that preexisting
crust that provided aluminosilicates to the IF depositional environment possessed
Є Nd (t) of approximately −0.5, and ambient Fe-rich seawater possessed Є Nd (t)
equal to or greater than +1. Error bars for Є Nd (t) calculated by assuming average errors
in 147 Sm/ 144 Nd ratios of 0.5%.

Author's personal copy
150 B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
REY abundances decrease. However, when comparing samples of
similar REY content the relative size of the anomalies varies. For
example, IF5 possesses a REY distribution and positive La and Y
anomalies almost identical to samples of similar concentration, yet IF5
displays a significantly larger Eu anomaly.
The possibility of significant alteration of the primary REY distributions
is considered unlikely. Mobility of the REY has not been observed
in Archean IFs (Bau, 1993), and concentrations of REY may vary
considerably between adjacent Fe- and Si-rich bands in Archean–
Paleoproterozoic IFs (Bau and Dulski, 1992). This behavior is also
observed in samples IF5 and IF5a, which were obtained from adjacent
layers in a ~4 cm section of drill core and yet possess distinct REY
patterns and very different REY concentrations (Fig. 3). We emphasize
that the lack of significant diagenetic or metamorphic alteration is
further supported by the excellent correlation of relatively mobile Rb
compared to highly immobile elements such as Ti and Th (Fig. 2).
Similarly to immobile trace elements, Є Nd (2.95 Ga) of the IF
samples varies as a function of Al content (Fig. 4), with samples containing
more than 0.5% Al 2 O 3 displaying relatively uniform negative
Є Nd (2.95 Ga) between −0.35 and −0.66. Samples containing less
than 0.5% Al 2 O 3 are significantly more radiogenic, displaying positive
Є Nd (2.95 Ga) between +0.29 and +0.99. The difference between the
two groups of Є Nd (2.95 Ga) values exceeds the analytical error of the
measurements. The correlation between increasing immobile element
contents (e.g., Al, Th, Zr, Rb, and Hf) and decreasing Є Nd (t) suggests
that crustal sources contributing detritus to the Pietersburg IF depositional
basin possessed Є Nd (2.95 Ga) of approximately −0.5, and
seawater responsible for precipitating the IF was more radiogenic with
Є Nd (2.95 Ga) similar to or greater than +1.
6. Discussion
6.1. Nature of the detrital aluminosilicate source
The higher Al 2 O 3 content (≥1.5%) of some IF samples indicates
contamination with detrital aluminosilicates, and candidates for a
detrital source(s) to the iron-formation include lithologies in the
simatic basement. Published data from the Eersteling area are primarily
for major elements and transition metals, and 80 analyses have
characterized metagabbros, metabasalts, amphibolites, serpentinites,
as well as quartz porphyrys (Saager and Meyer, 1982; Jones, 1990;
Byron and Barton, 1990). Aluminum averaged 9.7% in these samples
and only four contained more than 15% Al 2 O 3 (maximum 16.4%),
suggesting that if the aluminosilicate fraction in the iron-formation
was derived from pre-existing simatic basement then 15% Al 2 O 3 is the
likely upper limit for this material, which agrees well with estimates
for Archean mafic rocks (Condie, 1993). The strong correlations
between immobile trace elements and Al allows reasonable estimates
to be made regarding the trace metal distribution in the detrital source,
and by comparing this estimated distribution with trace element data
from the literature it may be possible to restrict the provenance of the
detrital component. Estimated trace metal data (Fig. 5) for the most Alrich
sample (IF7) at higher Al 2 O 3 contents of 10–15% match reasonably
well with a mixture of Archean felsic volcanics and komatiites possessing
compositions as proposed by Condie (1993), and komatiites in
the PGB have been described previously (Saager and Mayer, 1982).
However, few trace element data exist for PGB mafic rocks, making it
difficult to characterize the nature of this aluminosilicate source,
except to state that physical weathering of serpentinized material is an
unlikely candidate for any detritus present in the Pietersburg IF.
The relatively high Al 2 O 3 content in some samples (e.g., IF3, IF4, and
IF7) renders them unsuitable as archives for the marine fluid that
precipitated the Pietersburg IF, but these samples do provide Nd isotopic
information regarding the clastic detrital source during deposition
of the IF. Samples with Al 2 O 3 greater than ~0.5% possess very
consistent Є Nd (t) close to −0.5 (Fig. 4), which is considered to reflect
the Nd isotopic composition of the detrital aluminosilicate source. Close
examination of the Al–Є Nd (t) relationships (Fig. 4) suggeststhattwocomponent
conservative mixing between a detrital aluminosilicate
source and the pure chemical precipitate that formed the IFs might
adequately model the Nd isotopic data. However, two-component
mixing models that attempt to account for detrital contamination
Fig. 5. Trace element discrimination diagrams for simatic basement samples from the southwest region of the PGB. The grey area in both diagrams represents the range of predicted
element concentrations in sample IF7, and was calculated from linear regressions of the metal contents as a function of Al. The lower border of the grey area represents 10% Al 2 O 3 and
the upper border represents 15% Al 2 O 3 , which is a reasonable range for the detrital source (see Section 6). Diagram a contains literature data for possible source lithologies (Jones,
1990; Byron and Barton, 1990), of which pillow lavas and amphibolites are broadly consistent with the predicted distribution pattern. Diagram b shows the distribution patterns for
model Archean komatiites and felsic volcanics of Condie (1993), a mixture of which could reasonably produce the pattern observed in the detrital fraction of the iron-formation
samples, and komatiites are present within the oldest sections of the PGB (Saager and Meyer, 1982).

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
151
generally fail to accurately describe Є Nd (t) in IFs, as pure, Al-free IFs are
themselves best described as mixtures of two components; a siliceous
Nd-poor component (chert), and a Nd-rich Fe-oxide component.
Therefore, mixing between pure IF and aluminosilicates more realistically
reflects three endmembers; an Nd-rich contaminant with a distinct
143 Nd/ 144 Nd ratio, and two IF endmembers that presumably have
similar initial 143 Nd/ 144 Nd ratios, yet very different Nd concentrations.
The relatively constant trace metal ratios as a function of increasing
aluminosilicate content (Fig. 2), and the strong correlation with
decreasing Є Nd (t) suggests that the detrital component was wellmixed
and that the sediment source did not change during the time of
IF deposition represented by our samples. The lack of any negative
Eu CN anomaly in IF7 (Fig. 3) argues against significant contributions
from a highly evolved felsic source that had experienced fractional
crystallization of feldspars. The (Gd/Yb) CN ratios close to unity, i.e., the
lack of HREE depletion, rules out a dominant tonalite–trondhjemite–
granodiorite (TTG) provenance, or that mafic igneous rocks produced
by small-scale partial melting of a garnet-bearing (mantle) source
controlled the REY CN patterns of the detritus. Compared to data
compiled by Wilson (1989), incompatible trace element ratios in IF7,
such as Th/Yb (1.2), Ta/Yb (0.2), Zr/Ta (233), and Zr/Yb (48) as well
as the REY CN patterns are similar to those of modern back-arc basalts.
Assuming this trace element distribution represents the detrital
source, the aluminosilicates in the IFs appear to have originated from
the Archean equivalent of mafic back-arc basalts derived from a
slightly depleted mantle source.
6.2. IFs as seawater archives
The majority of published Sm–Nd isotopic data for Archean IFs are
for oxide-facies samples like the Pietersburg IF, and the following
discussion is therefore restricted to oxide-facies IF. The range of
literature Є Nd (t) values is large, as illustrated by the ~3.8 Ga Isua IFs.
Shimizu et al. (1990) reported Є Nd (3.8 Ga) for four separate Isua
samples that ranged from −5.0 to +22.4. Another sample (242573
of Miller and O'Nions, 1985) displays Є Nd (t) values from −15.2 to
+14.8, and has provided ten Sm–Nd analyses of Isua IF (see also Frei
et al., 1999). To our knowledge only eight individual samples of Isua IF
have been analyzed for Sm–Nd isotopes, and two of these (242573 of
Miller and O'Nions, 1985, and 491243 of Frei and Polat, 2007) have
produced 25 of the 30 individual Sm–Nd analyses reported in the
literature.
Considerable Є Nd (t) variability (between −4.7 to +1.6) is also
observed in younger IFs (e.g., the 2.5 Ga Hamersley IF of Miller and
O'Nions, 1985). Whereas consistent Sm–Nd results may be obtained
for some IFs (e.g., the Transvaal Supergroup, Bau et al., 1997a), any
attempt at reconstructing the Nd isotopic signature of Archean seawater
is hampered by the wide range of reported and often conflicting
Є Nd (t) values. At least in the case of Isua this may reflect resetting of
the isotopic system (Miller and O'Nions, 1985; Shimizu et al., 1990)or
possible contamination with clastic detritus. The last effect is difficult
to judge, as frequently only Sm–Nd data exist for IFs and major element
concentrations such as Al are not provided (e.g., Miller and
O'Nions, 1985; Jacobsen and Pimentel-Klose, 1988b).
However, consistent Sm–Nd data for Isua IF has recently been
reported by Frei and Polat (2007), who analyzed 15 bands of a single
sample and determined Є Nd (3.7 Ga) between +1.7 and +2.7. These
data are very similar to the only other Sm–Nd data for Isua IF as
reported in this study and Stecher et al. (1986) for the IF-G geostandard
(Table 2), which possesses Є Nd (3.8 Ga) of +2.3 to +2.7. Frei
and Polat (2007) also noted the similarity between REY distributions
in their Isua sample and modern seawater, consistent with similar
observations for Isua IF as reported by Bolhar et al. (2004). Modern
seawater, IF-G, and the Pietersburg IF all display very similar
REY patterns (Fig. 6), suggesting that anomalously wide variations
in Є Nd (t) may not be observed in IF samples that possess seawater-like
Fig. 6. Comparison of REY in modern seawater with eleven Pietersburg iron-formation
samples that contain b1.5% Al 2 O 3 . Samples are normalized to PAAS to facilitate
comparison with literature data. The seawater pattern represents the average of 158
worldwide seawater analyses of all depths (average depth 1560 m) for which complete
REY datasets are available (Zhang and Nozaki, 1996; Alibo and Nozaki, 1999; Bau et al.,
1997b; Nozaki et al., 1999; Nozaki and Alibo, 2003). Dashed lines are five Pietersburg IF
samples containing N170 ppm Ti, and solid lines represent six samples containing
b170 ppm Ti. Modern seawater displays pronounced positive Y and La anomalies, which
are also observed in the Pietersburg IF and 3.8 Ga Isua IF-G data (Dulski, 2001), and the
general shapes of the IF patterns are indistinguishable from seawater. Modern seawater
also displays smaller positive Gd anomalies, and slight positive Lu anomalies, features
which are present in the both the Pietersburg IF and IF-G. The only distinct differences
between modern seawater and the Archean IFs are the strong negative Ce anomaly in
seawater and the strong positive Eu anomaly in IF, which respectively reflect lower
relative redox levels and high-T hydrothermal input.
REY distributions, a conclusion supported by the 2.5 Ga Transvaal
Supergroup data of Bau et al. (1997a).
Therefore, to more clearly understand Nd isotopic ratios in Archean
seawater we have focused on those IFs which display clear seawaterlike
REY patterns. The REY are particularly useful for identifying the
marine origin of chemical precipitates (cf. Nothdurft et al., 2004, and
references therein) because seawater REY distributions are unique
relative to REY sources to the oceans (i.e., oceanic or continental
crust). This characteristic REY pattern in seawater is primarily due to
the different complexation behavior of the light and heavy REY with
regard to various aqueous carbonate species (Cantrell and Byrne,
1987), resulting in preferential scavenging of the light rare earth
elements (LREE) in the marine environment (Byrne and Kim, 1990).
This produces the heavy rare earth element (HREE) enrichment characteristic
of seawater (Fig. 6). Furthermore, pure marine precipitates
such as modern corals (Sholkovitz and Shen, 1995), as well as
Holocene microbialitic limestones (Webb and Kamber, 2000) accurately
record REY distributions in coeval seawater. We therefore
conclude that IF samples possessing seawater-like REY distributions
represent the best archives of coeval Precambrian seawater 143 Nd/
144 Nd, and using such an approach to screen IF samples facilitates
comparisons between studies.
The data available to date suggest that the overall shape of the REY
pattern for seawater has not significantly varied for the past 3.8 Ga,
based upon similar REY patterns observed in the Isua IF-G (see also
Bolhar et al., 2004), as well as Archean–Paleoproterozoic marine
carbonates (Kamber and Webb, 2001; Bau and Alexander, 2006). The
most significant differences between the REY distribution in modern
seawater and Archean seawater are for cerium (Ce) and Eu, the only

Author's personal copy
152 B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
REY that can have valence states other than +3. The oxic nature of
modern oceans allows Ce to be readily oxidized at particle surfaces to
highly insoluble Ce(IV), preferentially removing Ce and producing the
very negative Ce anomaly observed in modern seawater. The fact that
negative Ce anomalies have not been observed in marine chemical
sediments (IF or carbonates) older than ~2.3 Ga indicates relatively
reducing oceans, and an atmosphere–hydrosphere redox state for the
early Earth that was lower than today (cf. Holland, 1984).
Whereas Ce fractionation and removal occurs in the water column
and at the sediment–water interface, Eu fractionation occurs during
water–rock interaction at high temperatures via reduction of Eu 3+ to
Eu 2+ . The alteration products of these water–rock reactions discriminate
against the relatively large Eu 2+ ion and produce a subsequent
enrichment of Eu in the fluid phase. This Eu enrichment is present
only in fluids which have generally exceeded ~200 °C, and has been
observed in numerous vent fluids from hydrothermal systems altering
oceanic crust (e.g., Michard et al., 1983; Schmidt et al., 2007, and
references therein). The large positive Eu anomalies in Archean–
Paleoproterozoic IFs and carbonates possessing seawater-like REY
distributions are fully consistent with significant high-T hydrothermal
fluid fluxes to Earth's early oceans. Unfortunately, Eu anomalies are
not suitable for mass fraction calculations regarding the mixing of
high-T hydrothermal fluids and ambient seawater, and only reveal
information concerning the temperature of the REY source (e.g., Bau
and Möller, 1993; Alexander et al., 2008).
6.3. Nd isotope ratios in Archean seawater
Generally, IFs older than 2.25 Ga that possess seawater-like REY
patterns have Є Nd (t) between −5 and +5 (Fig. 7), with much of the
data describing the ~2.5 Ga Hamersley IFs (Australia). The remaining
data are primarily from South Africa, though Frei et al. (2007) have
significantly expanded the Sm–Nd isotopic analyses of Archean IF
from North America; unfortunately, possible ages for these IF are
poorly constrained. Most Є Nd (t) values for IFs older than 2.25 Ga fall
between −1.5 and +2.5, similar to that observed in the PGB samples.
Data for IFs older than 3.0 Ga are dominated by studies of the
relatively few samples from Isua discussed previously. Considering the
wide range of Є Nd (t) values observed in the Isua IF and the few analyses
for samples clearly possessing seawater-like REY distributions, we
consider the IF-G Є Nd (3.8 Ga) value of approximately +2.5 as the best
estimate for average ~3.8 Ga seawater, similar to the conclusions of Frei
and Polat (2007). The Pietersburg data are similar to older IF, though
tending to more CHUR-like values, and support the interpretation that
Nd was significantly, if not dominantly, provided by ocean ridge
hydrothermal systems possessing a combination of radiogenic Nd
signatures and strong positive Eu anomalies. This trend seems consistent
until ~2.6 Ga, when negative Є Nd (t) values and smaller Eu
anomalies (Bau and Möller, 1993) become common in IFs, though in
only a few instances do Archean–Paleoproterozoic IFs display Є Nd (t)
lower than −1.5.
Focusing on the period near 3.0 Ga, the Pietersburg IF Nd isotopic
evolution (Fig. 8) is very similar to that observed for the 2.9 Ga
Mozaan IFs from South Africa (Alexander et al., 2008). The consistent
Nd isotopic evolution observed in the two groups of IF is expected due
to the similar Sm/Nd values observed in IFs with seawater-like REY
distributions. The reasonable range of potential depositional ages for
both the Pietersburg and the Mozaan IFs would be from ~2.84 to
3.1 Ga, and within this time frame the Nd isotopic ratios remain
distinctly different. As initial 143 Nd/ 144 Nd ratios in the Pietersburg IF
are negatively correlated with increasing aluminosilicate content
(Fig. 4), associated crustal rocks displayed Є Nd (2.95 Ga) of −0.5,
whereas ambient seawater during deposition of these IFs displayed
Є Nd (2.95 Ga) of approximately +1. This value is ~3–5 Є-units higher
than seawater contemporaneous with deposition of the Mozaan
samples IF (Alexander et al., 2008). These different Є Nd (t) values
indicated for seawater likely reflect different depositional environments,
as the previously discussed trace element data suggests that
the rocks associated with the Pietersburg IF were generated in an
Fig. 7. Neodymium isotopic signatures in oxide and silicate facies IFs older than 2.4 Ga. World IFs generally possess Є Nd (t) between −5 and +5, though most of the data are more
positive than −1.5. The dashed lines represent Є Nd (t) values predicted for simple Nd isotopic evolution in depleted mantle and continental crust with initial Є Nd (4.56 Ga)=0, and
possessing Є Nd (0) =+10 and −17, respectively. Also shown is the Nd evolution in a depleted mantle as modeled by Nägler and Kramers (1998). Data are screened to distinguish
samples which display seawater-like REY patterns similar to those shown in Fig. 6, and are from Miller and O'Nions (1985), Jacobsen and Pimentel-Klose (1988a, 1988b), 12 samples
considered pristine enough by Alibert and McCulloch (1993) for calculation of possible Archean seawater pH values, and Bau et al. (1997a). All IF samples of 3.7–3.8 Ga age are from
Isua (Greenland), with stratigraphic ages as reported in original datasets. All data have been calculated using a present day 143 Nd/ 144 Nd of 0.512638 in CHUR and normalized to
146 Nd/ 144 Nd=0.7219. The majority of the data fall between −1.5 and +2.5, and the Є Nd (t) of Archean–Paleoproterozoic seawater as suggested by IFs was generally between +1 and
+2 until ~2.7 Ga. Only the 2.9 Ga Mozaan IFs possess distinctly different Є Nd (t) values, which are similar to igneous rocks (rhyolite, andesite, basalt, and gabbro) from the South
Africa–Swaziland border area (Hegner et al., 1984, 1994).

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
153
Fig. 8. Nd isotopic evolution lines for 2.9–3.0 Ga South African IFs. Mozaan IF data does
not include the two samples shown in Fig. 7 with Є Nd (2.9 Ga) lower than −8.0. The
left vertical axis represents the minimum age for the Mozaan Group of 2837±5 Ma
(Gutzmer et al., 1999); the best estimate of the maximum age for the Mozaan IFs is
2977±5 Ma (Nhleko, 2003). The lower age limit for the Pietersburg IFs is indicated as
2958±2 Ma (De Wit et al., 1993). The geologic evidence and the limited radiometric
age data do not preclude the possibility that deposition of the Pietersburg and Mozaan
IFs overlapped in time. The isotopic evolution lines are generally subparallel for all IF
samples, reflecting the similar 147 Sm/ 144 Nd ratios in these samples. The Pietersburg IFs
are consistently more radiogenic than the Mozaan IFs, and are interpreted to represent
Archean seawater masses with higher 143 Nd/ 144 Nd than the seawater that precipitated
the Mozaan IFs.
environment similar to a modern mafic back-arc setting, whereas
the Mozaan IF were clearly deposited in a marine shelf setting along
a stable cratonic margin (Beukes and Cairncross, 1991). The negative
Є Nd (2.9 Ga) values observed in the Mozaan IF are not surprising
considering the occurrence of ca. 2.9 Ga igneous rocks possessing
negative Є Nd (t) near the South Africa–Swaziland border (Hegner
et al., 1984, 1994). These felsic and mafic rocks (e.g., rhyolite, andesite,
basalt, and gabbro) possess Є Nd (t) very similar to the Mozaan
IFs (Fig. 7), and are consistent with published Mozaan Group
shale Є Nd (t) values(Stevenson and Patchett, 1990; Alexander et al.,
2008).
6.4. Implications of Nd isotopic variations within the Archean ocean
The fact that many Archean IFs with seawater-like REY distributions
possess positive Є Nd (t) implies a significant hydrothermal flux
from mantle-derived source rocks to seawater until at least 2.5 Ga.
Certainly more studies of Mesoarchean IF (3.0–3.5 Ga) are necessary,
but existing Nd isotopic data are consistent with the increased
mantle-derived hydrothermal flux implied from Sr isotopic studies of
Archean carbonates (Veizer and Mackenzie, 2004). With the exception
of the Mozaan IF, the first clear evidence of seawater displaying
negative Є Nd (t) is found in N. American IFs possessing Є Nd (2.65 Ga) of
−1.5 (Frei et al., 2007), which is considerably more radiogenic than
the older Mozaan IF. However, it is possible that negative (and highly
variable) Є Nd (t) values were more prevalent in ca. 2.9 Ga seawater
than is suggested by Fig. 7, as it does not include data from Frei et al.
(2007) for oxide-facies IF from the N. American Nemo iron-formation.
This is because the Nemo IF is poorly constrained in age (2.56–2.9 Ga),
and therefore could display Є Nd (2.9 Ga) between −2.76 and +3.71, or
Є Nd (2.56 Ga) between −5.24 and 0.78.
The clearly distinct Є Nd (t) values between the Pietersburg and
Mozaan IF are considered to arise from their different depositional
environments, as the Mozaan IF are the oldest IF deposited on a stable
cratonic margin, whereas older IF like the Pietersburg are associated
with greenstone belts. If ca. 3.0 Ga seawater was homogenous with
respect to 143 Nd/ 144 Nd, then world oceans would have become
dramatically less radiogenic in the time span between deposition of
the Pietersburg and Mozaan IF. We consider this unlikely, as it would
further require a reversion of world seawater Є Nd (t) back to chondritic
or positive values for the remainder of the Archean. We therefore
interpret the data as indicating that Archean seawater could be
inhomogeneous with respect to 143 Nd/ 144 Nd ratios, primarily dependent
upon the nature of local crust exposed for weathering, a situation
that is similar to that observed in modern oceans.
The degree of Nd isotopic inhomogeneity in Archean seawater is
unclear; it may have been that only small parts of the world ocean
deviated significantly from an ‘average’ 143 Nd/ 144 Nd ratio. Excluding
the Mozaan IF, seawater appears to have displayed relatively constant
Є Nd (t) between 0 and +2.5 from 2.7 to 3.8 Ga. This is remarkable, and
though detailed discussions of mantle evolution and the growth of
continental crust are beyond the scope of this study, crustal sources of
Nd were certainly evolving isotopically during this time period
(Fig. 7). However, regardless of the model employed, there is little
evidence that isotopic evolution within Nd source(s) to world oceans
significantly altered the 143 Nd/ 144 Nd ratio of seawater throughout
most of the Archean. This implies that prior to 2.7 Ga, the great bulk of
seawater at any time t was relatively homogeneous with respect to
143 Nd/ 144 Nd, and water masses that deviated from a bulk seawater
average Є Nd (t) of +1 to +2 (i.e., those that formed the Mozaan IF),
were minor components of the world ocean.
An isotopically homogenous world ocean is best explained by
widespread anoxic conditions in Archean seawater (as suggested by
the presence of IF themselves) that would increase the residence time
of Nd in seawater. Evidence for deep water anoxia is provided by U and
Th behavior in altered Archean basalts (Nakamura and Kato, 2007),
and this environment would prohibit the formation of important REY
sinks, increasing the residence time of Nd in Archean bottom seawater.
Shallow (coastal) seawater that could precipitate oxide-facies IF was
most likely characterized by the same short marine residence time of
Nd as is observed in modern seawater, and was dominated by less
radiogenic ‘local’ continental Nd. However, the relatively constant
positive Є Nd (t) values observed in world IFs between 2.7 and 3.8 Ga in
age that possess seawater-like REY distributions suggests hydrothermal
alteration of oceanic crust was dominating the Nd (and hence Fe?)
flux to bulk seawater throughout most of the Archean.
7. Conclusions
Prior to 2.7 Ga, iron-formations possessing seawater-like REY
distributions are the best archives for reconstructing paleoseawater
143 Nd/ 144 Nd ratios. Ambient seawater at the time of deposition of the
Pietersburg IF possessed positive Є Nd (2.95 Ga) of at least +1, which is
slightly lower than the best estimate for seawater Є Nd (3.8 Ga) as
suggested by Isua IF. Local clastic aluminosilicate material, presumably
derived from an Archean equivalent of a modern mafic back-arc
setting, possessed lower Є Nd (t) of approximately −0.5, and relatively
small contributions of this material (≥0.5% Al 2 O 3 ) were sufficient to
dominate the Nd budget in the Pietersburg IF. Crustal sources of Nd to
seawater near the Kaapvaal craton 2.9–3.0 Ga possessed distinctly
different Є Nd (t), and could locally influence seawater 143 Nd/ 144 Nd
ratios, as shown by the Mozaan IF, for example.
The relatively consistent Nd isotopic ratios observed throughout
most of the Archean for IFs with seawater-like REY distributions
suggests that, on average, bulk Archean seawater displayed positive
Є Nd (t) between +1 and +2 until at least 2.7 Ga. However, the range in
Є Nd (t) values observed between the similarly aged Pietersburg and
Mozaan IFs implies that world oceans were not entirely well-mixed
with respect to 143 Nd/ 144 Nd ratios, but that this variability was
restricted to local water masses, and was likely the consequence of

Author's personal copy
154 B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
different depositional environments. In conjunction with the ubiquitous
positive Eu anomalies observed in IFs, the Nd isotopic data
indicate high-temperature hydrothermal alteration of mantle-derived
oceanic crust was a dominant control on bulk seawater composition
for much of the Archean.
Acknowledgements
This research was funded by Exchange Grant 1327 within the
European Science Foundation ArchEnviron Research Networking
Programme. Mark Le Grange is thanked for his assistance with
sampling. The authors would like to gratefully acknowledge the
contributions of M. Fischerström and H. Schöberg regarding the Sm–
Nd isotopic analyses performed at LIG. This manuscript significantly
benefited from the comments of M. Gutjahr and one anonymous
reviewer, as well as suggestions by M.L. Delaney.
References
Abbey, S., McLeod, C.R., Liang-Guo, W., 1983. FeR-1, FeR-2, FeR-3, and FeR-4 four
Canadian iron-formation samples prepared for use as reference materials. Geol. Sur.
Can. paper 83–119.
Abbott, D.H., Hoffman, S.E., 1984. Archaean plate tectonics revisited 1. Heat flow,
spreading rate, and the age of subducting oceanic lithosphere and their effects on
the origin and evolution of continents. Tectonics 3, 429–448.
Alexander, B.W., Bau, M., Andersson, P., Dulski, P., 2008. Continentally-derived solutes in
shallow Archean seawater: rare earth element and Nd isotope evidence in iron
formation from the 2.9 Ga Pongola Supergroup, South Africa. Geochim. Cosmochim.
Acta 72, 378–394.
Alibert, C., McCulloch, M.T., 1993. Rare earth element and Nd isotopic compositions of
the banded iron-formations and associated shales from Hamersley, Western
Australia. Geochim. Cosmochim. Acta 57, 187–204.
Alibo, D.S., Nozaki, Y., 1999. Rare earth elements in seawater: particle association, shale
normalization, and Ce-oxidation. Geochim. Cosmochim. Acta 63, 363–372.
Anders, E., Grevesse, N., 1989. Abundances of the elements: meteoric and solar.
Geochim. Cosmochim. Acta 53, 197–214.
Andersson, P.S., Porcelli, D., Frank, M., Björk, G., Dahlqvist, R., Gustafsson, Ö., 2008.
Neodymium isotopes in seawater from the Barents Sea and Fram Strait Arctic–
Atlantic gateways. Geochim. Cosmochim. Acta 72, 2854–2867.
Appel, P.W.U., 1983. Rare earth elements in the early Archaean Isua iron-formation,
West Greenland. Precambrian. Res. 20, 243–258.
Bau, M., 1993. Effects of syn- and post-depositional processes on the rare-earth element
distribution in Precambrian iron-formations. Eur. J. Mineral. 5, 257–267.
Bau, M., Dulski, P., 1992. Small-scale variations of the rare-earth element distribution in
Precambrian iron formations. Eur. J. Mineral. 4, 1429–1433.
Bau, M., Möller, P., 1993. Rare earth element systematics of the chemically precipitated
component in Early Precambrian iron formations and the evolution of the
terrestrial atmosphere–hydrosphere–lithosphere system. Geochim. Cosmochim.
Acta 57, 2239–2249.
Bau, M., Dulski, P., 1996. Distribution of yttrium and rare-earth elements in the Penge
and Kuruman iron-formations, Transvaal Supergroup, South Africa. Precamb. Res.
79, 37–55.
Bau, M., Alexander, B., 2006. Preservation of primary REE patterns without Ce anomaly
during dolomitization of Mid-Paleoproterozoic limestone and the potential reestablishment
of marine anoxia immediately after the “Great Oxidation Event”.
S. Afr. J. Geol. 109, 81–86.
Bau, M., Höhndorf, A., Dulski, P., Beukes, N.J., 1997a. Sources of rare-earth elements and
iron in Paleoproterozoic iron-formations from the Transvaal Supergroup, South
Africa: evidence from neodymium isotopes. J. Geol. 105, 121–129.
Bau, M., Möller, P., Dulski, P., 1997b. Yttrium and lanthanides in eastern Mediterranean
seawater and their fractionation during redox-cycling. Mar. Chem. 56, 123–131.
Beukes, N.J., 1983. Palaeoenvironmental setting of iron-formations in the depositional
basin of the Transvaal Supergroup, South Africa, in: Trendall, A.F., Morris, R.C. (Eds.),
Iron-formation Facts and Problems, 131–198, Vol. 6 Developments in Precambrian
Geology (Ed. B.F. Windley), Elsevier, Amsterdam.
Beukes, N.J., Cairncross, B., 1991. A lithostratigraphic sedimentological reference profile
for the late Archaean Mozaan Group, Pongola Sequence: application to sequence
stratigraphy and correlation with the Witwatersrand Supergroup. S. Afr. J. Geol. 94,
44–69.
Bolhar, R., Kamber, B.S., Moorbath, S., Fedo, C.M., Whitehouse, M.J., 2004. Characterisation
of early Archaean chemical sediments by trace element signatures. Earth
Planet. Sci. Lett. 222, 43–60.
Broecker, W.S., Peng, T.H., 1982. Tracers in the Sea. Eldigio Press, New York.
Byrne, R.H., Kim, K., 1990. Rare earth element scavenging in seawater. Geochim.
Cosmochim. Acta 54, 2645–2656.
Byron, C.L., Barton, J.M., 1990. The setting of mineralization in a portion of the Eersteling
goldfield, Pietersburg granite–greenstone terrane, South Africa. S. Afr. J. Geol. 93,
463–472.
Cantrell, K.J., Byrne, R.H., 1987. Rare earth element complexation by carbonate and
oxalate ions. Geochim. Cosmochim. Acta 51, 597–605.
Condie, K.C., 1993. Chemical composition and evolution of the upper continental crust:
contrasting results from surface samples and shales. Chem. Geol. 104, 1–37.
Danielson, A., Möller, P., Dulski, P., 1992. The europium anomalies in banded iron
formations and the thermal history of the oceanic crust. Chem. Geol. 97, 89–100.
DePaolo, D.J., Wasserburg, G.J., 1976. Nd isotopic variations and petrogenetic models.
Geophys. Res. Lett. 3, 249–252.
Derry, L.A., Jacobsen, S.B., 1990. The chemical evolution of Precambrian seawater:
Evidence from rare earth elements in banded iron formations. Geochim.
Cosmochim. Acta 54, 2965–2977.
De Wit, M.J., 1991. Archaean greenstone belt tectonism and basin development: some
insights from the Barberton and Pietersburg greenstone belts, Kaapvaal Craton,
South Africa. J. Afr. Earth Sci. 13, 45–63.
De Wit, M.J., Jones, M.G., Buchanan, D.L., 1992. The geology and tectonic evolution of the
Pietersburg Greenstone Belt, South Africa. Precambrian Res. 55, 123–153.
De Wit, M.J., Armstrong, F.L., Kamo, S.L., Erlank, A.J., 1993. Gold-bearing sediments in the
Pietersburg greenstone belt: age equivalents of the Witwatewrsrand Supergroup
sediments. S. Afr. Econ. Geol. 88, 1242–1252.
Dulski, P., 2001. Reference materials for geochemical studies: new analytical data by
ICP-MS and critical discussion of reference values. Geostand. Newsl. 25, 87–125.
Eglington, B.M., Talma, A.S., Marais, S., Matthews, P.E., Dixon, J.G.P., 2003. Isotopic
composition of Pongola Supergroup limestones from the Buffalo River gorge, South
Africa: constraints on their regional depositional setting. S. African J. Geol.106,1–10.
Frank, M., 2002. Radiogenic isotopes: tracers of past ocean circulation and erosional
input. Rev. Geophys. 40, 1–38.
Frei, R., Polat, A., 2007. Source heterogeneity for the major components of ~3.7 Ga
Banded Iron Formations (Isua Greenstone Belt, Western Greenland): tracing the
nature of interacting water masses in IF formation. Earth Planet. Sci. Lett. 253,
266–281.
Frei, R., Bridgewater, D., Rosing, M., Stecher, O., 1999. Controversial Pb–Pb and Sm–Nd
isotope results in the early Archean Isua (West Greenland) oxide iron formation:
preservation of primary signatures versus secondary disturbances. Geochim.
Cosmochim. Acta 63, 473–488.
Frei, R., Dahl, P.S., Duke, E.F., Frei, K.M., Hansen, T.R., Frandsson, M.M., Jensen, L.A., 2007.
Trace element and isotopic characterization of Neoarchean and Paleoproterozoic
iron formations in the Black Hills (South Dakota, USA): assessment of chemical
change during 2.9–1.9 Ga deposition bracketing the 2.4–2.2 Ga first rise of
atmospheric oxygen. Precam. Res. 162, 441–474.
Fryer, B.J., Fyfe, W.S., Kerrich, R., 1979. Archaean volcanogenic oceans. Chem. Geol. 24,
25–33.
German, C.R., Klinkhammer, G.P., Edmond, J.M., Elderfield, H., Mitra, A., 1990.
Hydrothermal scavenging of rare-earth elements in the ocean. Nature 345, 516–518.
Goldstein, S.L., Hemming, S.R., 2003. In: Elderfield, H. (Ed.), Long-lived Isotopic Tracers
in Oceanography, Paleoceanography and Ice Sheet Dynamics. Treatise on
Geochemistry, Vol. 6. Elsevier-Pergamon, Oxford, pp. 453–489.
Grotzinger, J.P., 1989. Facies and evolution of Precambrian carbonate depositional
systems: emergence of the modern platform archetype. Controls on Carbonate
Platform and Basin Development, SEPM (Soc. Econ. Paleont. Mineral). Spec. Pub.,
vol. 44, pp. 79–106.
Gutjahr, M., Frank, M., Stirling, C.H., Keigwin, L.D., Halliday, A.N., 2008. Tracing the Nd
isotope evolution of North Atlantic Deep and Intermediate Waters in the western
North Atlantic since the Last Glacial Maximum from Blake Ridge sediments. Earth
Planet. Sci. Lett. 266, 61–77.
Gutzmer, J., Nhleko, N., Beukes, N.J., Pickard, A., Barley, M.E., 1999. Geochemistry and ion
microprobe (SHRIMP) age of a quartz porphyry sill in the Mozaan Group of the
Pongola Supergoup: implications for the Pongola and Witwatersrand Supergroups.
S. Afr. J. Geol. 102, 139–146.
Hegner, E., Kröner, A., Hofmann, A.W., 1984. Age and isotope geochemistry of the
Archaean Pongola and Usushwana suites in Swaziland, southern Africa: a case for
crustal contamination of mantle-derived magma. Earth Planet. Sci. Lett. 70,
267–279.
Hegner, E., Kröner, A., Hofmann, A.W., 1994. A precise U–Pb zircon age for the Archaean
Pongola Supergroup volcanics in Swaziland. J. Afr. Earth Sci. 18, 339–341.
Holland, H.H., 1984. The Chemical Evolution of the Atmosphere and the Oceans.
Princeton Univ. Press.
Jacobsen, S.B., Pimentel-Klose, M.R., 1988a. A Nd isotopic study of the Hamersley and
Michipicoten banded iron formations: the source of REE and Fe in Archean oceans.
Earth Planet. Sci. Lett. 87, 29–44.
Jacobsen, S.B., Pimentel-Klose, M.R., 1988b. Nd isotopic variations in Precambrian
banded iron formations. Geophys. Res. Lett. 15, 393–396.
Jeandel, C., Bishop, K., Zindler, A., 1995. Exchange of neodymium and its isotopes
between seawater and small and large particles in the Sargasso Sea. Geochim.
Cosmochim. Acta 59, 535–547.
Jochum, K.P., Nohl, U., Herwig, K., Lammel, E., Stoll, B., Hofmann, A.W., 2005. GeoReM: A
new geochemical database for reference materials and isotopic standards.
Geostand. Geoanal. Res. 29, 333–338.
Jones, M.G., 1990. The geology of the Mt Mare area, Pietersburg greenstone belt, South
Africa. Ph.D. thesis, Imperial College, Univ. London.
Kamber, B.S., Webb, G.E., 2001. The geochemistry of late Archaean microbial carbonate:
Implications for ocean chemistry and continental erosion history. Geochim.
Cosmochim. Acta 65, 2509–2525.
Klein, C., Beukes, N.J., 1989. Geochemistry and sedimentology of a facies transition from
limestone to iron-formation in the Early Proterozoic Transvaal Supergroup, South
Africa. Econ. Geol. 84, 1733–1774.
Kröner, A., Jaeckl, P., Brandl, G., 2000. Single zircon ages for felsic to intermediate rocks
from the Pietersburg and Giyani greenstone belts and bordering granitoid
orthogneisses, northern Kaapvaal Craton, South Africa. J. Afr. Earth Sci. 30, 773–793.

Author's personal copy
B.W. Alexander et al. / Earth and Planetary Science Letters 283 (2009) 144–155
155
McKenzie, D., Weiss, N., 1975. Speculations on the thermal and tectonic history of the
Earth,. Geophys. J. R. Astr. Soc. 42, 131–174.
McLennan, S.M., 1989. Rare earth elements in sedimentary rocks: influence of
provenance and sedimentary processes. In: Lipin, B.R., McKay, G.A. (Eds.),
Geochemistry and Mineralogy of Rare Earth Elements. Rev. Mineral. 21, Mineral.
Soc. Amer., pp. 169–200.
Michard, A., Albarède, F., Michard, G., Minster, J.F., Charlou, J.L., 1983. Rare-earth
elements and uranium in high-temperature solutions from East Pacific Rise
hydrothermal vent field (13°N). Nature 303, 795–797.
Miller, R.G., O'Nions, R.K., 1985. Sources of Precambrian chemical and clastic sediments.
Nature 314, 325–330.
Nägler, T.F., Kramers, J.D., 1998. Nd isotopic evolution of the upper mantle during the
Precambrian: models, data and the uncertainty of both. Precambrian Res. 91,
233–252.
Nakamura, K., Kato, Y., 2007. A new geochemical approach for constraining a marine
redox condition of Early Archean. Earth Planet. Sci. Lett. 261, 296–302.
Nhleko, N., 2003. The Pongola Supergroup in Swaziland, Ph.D. thesis, Univ.
Johannesburg.
Nothdurft, L.D., Webb, G.E., Kamber, B.S., 2004. Rare earth element geochemistry of Late
Devonian reefal carbonates, Canning Basin, Western Australia: confirmation of a
seawater REE proxy in ancient limestones. Geochim. Cosmochim. Acta 68, 263–283.
Nozaki, Y., Alibo, D.S., 2003. Importance of vertical geochemical processes in controlling
the oceanic profiles of dissolved rare earth elements in the northeastern Indian
Ocean. Earth Planet. Sci. Lett. 205, 155–172.
Nozaki, Y., Alibo, D.S., Amakawa, H., Gamo, T., Hasumoto, H., 1999. Dissolved rare earth
elements and hydrography in the Sulu Sea. Geochim. Cosmochim. Acta 63,
2171–2181.
Piepgras, D.J., Wasserburg, G.J., 1980. Neodymium isotopic variations in seawater. Earth
Planet. Sci. Lett. 50, 128–138.
Piotrowski, A.M., Goldstein, S.L., Hemming, S.R., Fairbanks, R.G., 2005. Temporal relationships
of carbon cycling and ocean circulation at glacial boundaries. Science 307,
1933–1938.
Saager, R., Meyer, M., 1982. Gold distribution in supracrustal rocks from the Archean
greenstone belts of Southern Africa and from Paleozoic ultramafic complexes of the
European Alps: metallogenic and geochemical implications. Econ. Geol. 77, 1–24.
SACS - S. Afr. Comm. Strat.,1980. Part 1: lithostratigraphy of the Republic of South Africa,
South West Africa/Namibia and the republics of Bophuthatswana, Transkei and
Venda. In: Kent, L.E. (Ed.), Stratigraphy S. Afr. 8. Geol. Surv. S. Afr, pp. 71–80.
Schmidt, K., Koschinsky, A., Garbe-Schönberg, D., de Carvalho, L.M., Seifert, R., 2007.
Geochemistry of hydrothermal fluids from the ultramafic-hosted Logatchev
hydrothermal field, 15°N on the Mid-Atlantic Ridge: temporal and spatial
investigation. Chem. Geol. 242, 1–21.
Shimizu, H., Umemoto, N., Masuda, A., Appel, P.W.U., 1990. Sources of iron-formations in
the Archean Isua and Malene supracrustals, West Greenland: evidence from La–Ce
and Sm–Nd isotopic data and REE. Geochim. Cosmochim. Acta 54, 1147–1154.
Simonson, B.M., 2003. Origin and evolution of large Precambrian iron formations. Geol.
Soc. Amer. Spec. Paper 370, 231–243.
Simonson, B.M., Hassler, S.W., 1996. Was the deposition of large Precambrian iron
formations linked to major marine transgressions? J. Geol. 104, 665–676.
Sholkovitz, E., Shen, G.T., 1995. The incorporation of rare earth elements in modern
coral. Geochim. Cosmochim. Acta 59, 2749–2756.
Stecher, O., Carlson, R.W., Shirey, S.B., Nielson, T., 1986. Nd isotope evidence for the
evolution of metavolcanic rocks from the Archaean block of Greenland and
Labrador. Terra Cognita 6, 236.
Stevenson, R.K., Patchett, P.J., 1990. Implications for the evolution of continental crust
from Hf isotope systematics of Archean detrital zircons. Geochim. Cosmochim. Acta
54, 1683–1697.
Tachikawa, K., Athias, V., Jeandel, C., 2003. Neodymium budget in the modern ocean and
paleo-oceanographic implications. J. Geophys. Res. 108, 1–13.
Veizer, J., Mackenzie, F.T., 2004. Evolution of sedimentary rocks, in: Mackenzie, F.T.
(Ed.), Sediments, Diagenesis, and Sedimentary Rocks, 369–407, Vol. 7 Treatise on
Geochemistry (Eds. H.D. Holland and K.K. Turekian), Elsevier-Pergamon, Oxford.
Veizer, J., Hoefs, J., Lowe, D.R., Thurston, P.C., 1989. Geochemistry of Precambrian
carbonates: II. Archean greenstone belts and Archean sea water. Geochim.
Cosmochim. Acta 53, 859–871.
Webb, G.E., Kamber, B.S., 2000. Rare earth elements in Holocene reefal microbialites: a
new shallow seawater proxy. Geochim. Cosmochim. Acta 64, 1557–1565.
Wilson, M., 1989. Igneous Petrogenesis. Chapman and Hall, London.
Zhang, J., Nozaki, Y., 1996. Rare earth elements and yttrium in seawater: ICP-MS
determinations in the East Caroline, Coral Sea, and South Fiji basins of the western
South Pacific Ocean. Geochim. Cosmochim. Acta 60, 4631–4644.

Chapter 5. Anoxygenic photoautotrophs and the origin of banded iron-formation
153

THIS PAGE INTENTIONALLY LEFT BLANK
154

Anoxygenic photoautotrophs and the origin of banded iron-formation
Brian W. Alexander * and Michael Bau
Jacobs University Bremen, D28759 Bremen, Germany
*corresponding author
Oxide-facies banded iron-formation (BIF) comprised of alternating layers of Fe oxides
and quartz formed as a chemical sediment at the Early Precambrian seafloor. Despite
more than 100 years of research, however, the oxidative precipitation mechanism is still
a matter of debate. Some propose an indirect biogenic origin via oxidation of Fe(II) by
molecular oxygen produced photosynthetically by ancient cyanobacteria 1,2 , while others
suggest an abiotic origin via photo-oxidation of Fe(II) 3,4 . Although hypothesized 35
years ago 5 , it was the discovery of Fe(II) oxidizing photoautotrophs 6 that stipulated
further investigation of the potential role of direct biogenic Fe(II) oxidation by
anoxygenic photosynthesis (photoferrotrophy) in BIF genesis. These studies suggest that
anoxygenic photoautotrophs could theoretically have oxidized enough Fe(II) to form
oxide-facies BIF in the Early Precambrian ocean 7,8 before the onset of oxygenic
photosynthesis. However, evidence from the geological record for Fe(II) oxidation in the
absence of oxygen is missing. We present chemical data for ferruginous shales deposited
with oxide-facies BIF from the 2.9 Ga old Pongola Supergroup, South Africa, that
indicate that oxide-facies Precambrian BIF formed in the absence of photosynthetically
produced oxygen. In contrast to non-ferruginous shales deposited below and above the
Pongola BIF, the ferruginous shales are strongly depleted in K, Rb, Cs, Ba and Na, due
to alkali metal exchange for ferrous iron, suggesting that oxide-facies BIF formed in
anoxic, ferruginous seawater. While this is fully compatible with Fe(II) oxidation by
anoxygenic photoautotrophs, it renders the presence of biogenic molecular oxygen
unlikely and suggests that photoferrotrophy, not oxygenic photosynthesis, was a
prerequisite for the formation of oxide-facies BIF on Early Earth.
Although their abundance appears to have peaked at around 2.5 Ga ago 9 , Fe(III) oxide
bearing chemical sediments such as BIF, jasper, jaspilite or ferruginous chert (here subsumed
under oxide-facies iron-formation, IF) occur in the oldest marine sedimentary sequence on
Earth at Isua, Greenland, in all Archean greenstone belts, and in the oldest supracrustal
successions in the Pongola and Witwatersrand Supergroups, South Africa. Hence, formation
155

and preservation of oxide-facies IF was common throughout the first 2 billion years of
Earth’s geologic history. Oxide-facies IF in Archean marine basins is frequently associated
with Fe-rich shales, with the latter commonly assumed to represent mixtures between detrital
clay minerals and Fe(III) oxide minerals.
The Archean (2.85-2.95 Ga) Pongola Supergroup in southern Africa consists of the
predominantly volcanic Nsuze Group, which is unconformably overlain by the predominantly
sedimentary Mozaan Group. Previous studies have concluded that the Pongola Supergroup
was deposited in near-shore shallow marine waters on a stable cratonic margin 10,11 , and
subsequently experienced lower greenschist facies regional metamorphism 10 .
This study discusses samples from the Sinqeni Formation of the Mozaan Group, that
were obtained from the White Mfolozi inlier where a shallow-water sequence of BIF and
ferruginous shale is bounded top and bottom by non-ferruginous shales 12 . The sample set
includes the non-ferruginous shales deposited immediately below and above the BIF and the
ferruginous shales within the BIF. Contacts between the ferruginous shales and the pure
oxide-facies BIF layers can be sharp, but some BIF beds are very impure chemical sediments
due to the co-deposition of detrital aluminosilicates.
Major element, carbon, nitrogen, and trace element data for the samples are provided in
Table 1. Variation in major elements such as Al, Fe, and K is related to the relative position
of the samples in the stratigraphic profile. The non-ferruginous shales located below and
above the BIF display a major element composition typical for Archean shales (Fig. 1). They
are indistinguishable from fine-grained clastic rocks described in a comprehensive study of
the Pongola Supergroup conducted by Wronkiewicz and Condie 13 (see Supplementary
Table), and represent sediment weathered from the Kaapvaal craton ~2.85-3.0 Ga ago.
In contrast, Fe in the ferruginous shales is inversely correlated with Al content (Fig. 1),
a relationship that reflects mixing between Al-rich clastic sediment and the Fe-rich chemical
precipitate that produced the BIF. On average, the ferruginous shales display half the Al 2 O 3
content and a five-fold increase in Fe 2 O 3 compared to the non-ferruginous shales. Similar to
Al and Fe, the alkali metals (with the exception of Na) display variable concentrations and a
distinct bimodal distribution with strong depletions for K, Rb, and Cs in the ferruginous
shales (Fig. 1). Potassium and Rb average 6.3% and 205 ppm, respectively, in the nonferruginous
shales, whereas the ferruginous shales contain less than 0.05% K 2 O and typically
less than 1 ppm Rb. Barium concentrations, unlike the other alkaline earth elements, are also
distinct between the non-ferruginous and ferruginous shales, averaging 567 ppm and less than
22 ppm, respectively (Fig. 1).
156

Table 1. Major and trace element concentration data for Pongola Supergroup shales.
wt.% 1
bottom
_________________________________________________
sample position in studied stratigraphic section ______________________________________________________
WM1 WM-A WM-B WM-C WM2 WM3 WM8 WM9 WM10 WM11 WM12 WM14 WM15 WM16 WM171 WM172 WM18 WM19 WM20
Fe-poor
shale
Fe-poor
shale
Fe-poor
shale
Fe-poor
shale
Fe-poor
shale Fe-shale Fe-shale Fe-shale Fe-shale Fe-shale Fe-shale Fe-shale Fe-shale Fe-shale
impure
BIF
impure
BIF
Fe-poor
shale
Fe-poor
shale
top
Fe-poor
shale
AVG
(n = 9)
Fe-shale
AVG
(n = 8)
AVG
(n = 50)
Fe-poor
shale pelite 4
SiO 2 56.6 56.1 47.4 65.3 51.6 52.4 47.9 54.2 44.9 49.7 49.2 44.5 46.9 36.3 38.0 32.1 49.1 48.6 71.8 47.3 55.8 59.5
TiO 2 0.94 1.11 0.95 0.73 0.90 0.76 0.72 0.82 0.81 0.58 0.47 0.59 0.48 0.83 0.16 0.08 1.35 1.21 0.60 0.67 0.97 0.90
Al 2 O 3 29.2 29.8 30.1 22.2 23.3 14.2 11.9 11.7 14.9 13.3 10.4 11.5 11.0 17.1 4.46 1.86 24.7 26.8 16.7 12.9 25.4 21.1
Fe 2 O 3 0.65 0.88 7.13 3.87 10.5 21.0 29.0 24.2 31.0 29.8 29.0 28.3 27.4 32.2 53.8 63.5 9.16 6.55 2.52 28.0 5.15 6.07
MnO

TOP
relative stratigraphic position (~10m)
SiO 2
Fe 2
O 3
BOTTOM
0 10 20 30 40 50 60 70 80 0 5 10 15 20 25 30 35 40
wt.%
wt.%
TOP
Al 2
O 3
0 5 10 15 20 25 30 35 40
wt.%
Na 2
O
0.0 0.2 0.4 0.6 0.8 1.0 1.2
wt.%
relative stratigraphic position (~10m)
predicted
concentrations
K 2
O
BOTTOM
0 1 2 3 4 5 6 7 8 9 10
wt.%
Rb
0 50 100 150 200 250 300
mg/kg
Cs
0 1 2 3 4 5 6 7 8 9 10
mg/kg
Ba
0 200 400 600 800 1000
mg/kg
Figure 1. Pongola shale major and trace element concentrations plotted in relative stratigraphic order.
The studied sequence is ~10 m thick, and represents a cycle of typical Pongola shale deposition (bottom),
through ferruginous shale and BIF deposition (middle), before returning to non-ferruginous shale deposition
(top of sequence). The grey bar shows the onset and cessation of pure BIF deposition, and the depositional cycle
is clearly evident in the increase in Fe content of the samples, coupled with a corresponding decrease in Al.
Determinations of Na and K for high Fe samples were frequently below X-ray fluorescence quantification
limits, and for these samples Na and K concentrations are plotted as equal to the analytical quantification limits
(0.11% and 0.05%, respectively). Open circles represent predicted concentrations of K, Rb, Cs, and Ba as
calculated from the average K/Al, Rb/Al, Cs/Al, and Ba/Al ratios of the eight ‘normal’ non-ferruginous shale
samples that reside at the top (three samples) and bottom (five samples) of the studied sequence. Extreme
depletion of alkali metals and Ba is inversely correlated with increasing Fe content of the samples and the
occurrence of BIF deposition.
Incompatible elements (Al, Ti, Zr, Th, Hf) have lower concentrations in the ferruginous
shales compared to the non-ferruginous shales, but ratios of these elements are comparable to
published data for Pongola shales 11,13 . For example, Al 2 O 3 /TiO 2 and Zr/Th ratios in Pongola
shales are similar regardless of Fe content, indicating that this component in the ferruginous
shales is geochemically similar to the local Archean continental crust that produced the nonferruginous
shales. We also emphasize that Th/U ratios in both shale types are similar.
158

The Pongola shales are mineralogically and chemically well characterized 11,13-16 and the
observed depletion in K, Rb, Cs, Ba and Na is common in the Mozaan Group. It is, however,
confined to shales with more than 20 wt.% Fe 2 O 3 , a relationship that is also observed in Ferich
shales interbedded with IF from the ~3.0 Ga Buhwa greenstone belt, Zimbabwe 17 . The
major minerals in non-ferruginous shales from the Mozaan Group are muscovite, chlorite,
and illite, with accessory Na- and K-feldspar 11,13 . In seven ferruginous (≥19.8% Fe 2 O 3 )
Mozaan Group shales, Nhleko 11 described quartz, muscovite, biotite, chlorite, albite,
magnetite, siderite, and stipnomelane as abundant minerals in one or more of the samples.
Therefore, the depletion in K, Rb, Cs, Ba and Na in the ferruginous shales is unlikely to be
related to differences in the mineralogical composition of the source material.
The magnitude of the K, Rb, Cs, Ba and Na depletion in the ferruginous shales is
significant, as typical Archean shales, similar to modern shales, commonly contain several
weight percent K 2 O and hundreds of ppm of Rb and Ba. Whereas the ferruginous shales have
roughly half the Al content of typical Pongola shales, the alkali metals and Ba are depleted by
a factor of approximately 100. Predicted concentrations of alkali metals and Ba in the
ferruginous shales may be calculated based upon immobile Al and average K/Al, Cs/Al,
Rb/Al, and Ba/Al ratios in the non-ferruginous shales. Predicted K, Rb, Cs, and Ba
concentrations (Fig. 1) suggest that K and Rb were depleted by more than 97%, Cs by more
than 94%, and Ba between 82-98% during periods of high Fe sedimentation. The lack of any
unusual enrichment of these elements above or below the ferruginous shales suggests that K,
Cs, Rb and Ba loss occurred either within the water column or near the sediment-water
interface, as the mobilized metals apparently escaped into the water column.
The mineralogy of Mozaan Group shales indicates that sheet silicates (e.g., muscovite,
illite) are the primary hosts for K, Rb, Cs and Ba. For these minerals the alkali metals and Ba
are electrostatically bound between silicate layers 18 , and are susceptible to ion-exchange with
dissolved cations. Adsorption/desorption reactions with clay minerals significantly affect
alkali metal distributions between sediments and pore fluids in modern marine systems, and
Cs and Rb, for example, are removed from sediment pore waters by adsorption processes.
Such reactions are reversible, and alkali metals adsorbed to clay minerals may be returned to
solution in the presence of NH +(19,20) 4 . The use of NH + 4 (or Ba 2+ ) to desorb alkali metals from
interlayer atomic sites in sheet silicates is long-established experimental practice in the
determination of clay mineral cation-exchange capacities 21,22 . However, N concentrations in
the ferruginous shales are consistently below 0.01 %, and are similar to non-ferruginous
shales 16 , suggesting that displacement of alkali metals by ammonium did not occur.
159

The mechanism by which the alkali metals and Ba were mobilized from the clastic
sediments is rather a consequence of high dissolved Fe 2+ concentrations. Ferrous iron may
readily exchange with interlayer Na + in smectite 23 , for example, and such Fe 2+ -alkali metal
exchange is also observed for montmorillonite at high ionic strength and high chloride
activity in experiments intended to model seawater-clay particle interactions 24 . An Fe 2+ -alkali
metal exchange is further supported by the fact that extreme K, Rb, Cs, Ba and Na depletion
is confined to the Fe-rich clastic sediments.
The ferruginous shales are closely associated with oxide-facies BIF. Although the
contacts between ferruginous shale beds and BIF beds are often sharp, there also exist impure
BIF samples that are transitional between ferruginous shales and pure BIF. Samples WM171
and WM172 have Al 2 O 3 concentrations of 4.46% and 1.86%, respectively, and elevated
concentrations of TiO 2 , Zr, Hf, and Th (Supplementary Table) compared to associated pure
BIF. Ratios between these immobile elements are similar to those of the ferruginous shales,
indicating that the deposition of ferruginous shale continued during the deposition of oxidefacies
BIF. We emphasize that the detrital aluminosilicate component in WM171 and
WM172 is characterized by the same lack of K, Rb, Cs and Ba (and Na) as the ferruginous
shales, demonstrating the abundance of dissolved ferrous iron in an environment which
formed chemical sediments that are now oxide-facies BIF. This suggests that Fe(III) oxides
precipitated and were preserved in a macro-environment that was reducing with respect to the
Fe 2+ /Fe 3+ redox couple, i.e., in a macro-environment that was anoxic. This is supported by the
close similarity of the Th/U ratios of the ferruginous and the non-ferruginous shales, as
oxidation of U(IV) in the clay minerals during the exchange reaction would have resulted in
the preferential mobilization of U(VI) over Th(IV) 25 , which is not observed.
The oxidation of part of the available dissolved Fe 2+ and the production of ferric-ironrich
sediment could result from a number of processes. These mechanisms include indirect or
direct biological processes, i.e., oxidation by reaction with photosynthetically produced
molecular oxygen 1,2 , or by anoxygenic photoautotrophs 7,8 , respectively. Abiotic
photochemical oxidation of Fe 2+ has also been suggested 3,4 , but is unlikely in the light of
recent experimental results 26 .
The ferruginous shales associated with the BIF formed when fine-grained clastic
sedimentary material was deposited in ferrous iron rich seawater. Ferrous iron was
incorporated into the clay minerals at the expense of K, Rb, Cs, Ba and Na which escaped
into the water column. The contemporaneous precipitation and preservation of oxide-facies
BIF demonstrates that Fe(II) oxidation occurred in an anoxic macro-environment, which is
160

supported by the lack of Th-U fractionation. This renders it unlikely that molecular oxygen
was responsible, as either molecular oxygen would have been abundant enough to
quantitatively oxidize the available dissolved ferrous iron, preventing the formation of alkali
element depleted ferruginous shales, or the Fe(III) oxides would have been reductively
dissolved (by Fe 2+ , for example) as soon as all the molecular oxygen was consumed. Only the
stabilization and preservation of Fe(III) oxides by photoautotrophs would have allowed for
the simultaneous deposition of oxide-facies BIF and ferruginous shales depleted in K, Rb, Cs,
Ba and Na.
Thus, oxide-facies IF could form via photoferrotrophy throughout the Archean prior to
the onset of oxygenic photosynthesis. The formation of Fe(III) (hydr)oxides, i.e., oxide-facies
IF, would have been common in a ferrous-iron-rich Archean ocean capable of sustaining
postulated anaerobic ecosystems dominated by anoxygenic photoautotrophs 27 , in full
agreement with the ubiquitous presence of oxide-facies BIF, jasper, jaspilite and ferruginous
chert in the Archean geological record.
Acknowledgements This research was supported by the European Science Foundation
research program Archean Environmental Studies: the Habitat of Early Life.
References
1. Cloud, P. Significance of the Gunflint (Precambrian) microflora. Science 148, 27-35
(1965).
2. Cloud, P. Paleoecological significance of the banded iron-formation. Econ. Geol. 68,
1135-1143 (1973).
3. Cairns-Smith, A.G. Precambrian solution photochemistry, inverse segregation, and
banded iron formations. Nature 276, 807-808 (1978).
4. Braterman, P.S., Cairns-Smith, A.G., & Sloper, R.W. Photo-oxidation of hydrated Fe2+ -
significance for banded iron formations. Nature 303, 163-164 (1983).
5. Garrels, R.M., Perry Jr., E.A., & Mackenzie, F.T. Genesis of Precambrian ironformations
and the development of atmospheric oxygen. Econ. Geol. 68, 1173-1179
(1973).
161

6. Widdel, F. et al. Ferrous iron oxidation by anoxygenic phototrophic bacteria. Nature 362,
834-836 (1993).
7. Konhauser, K.O. et al. Could bacteria have formed the Precambrian banded iron
formations? Geology 30, 1079-1082 (2002).
8. Kappler, A., Pasquero, C., Konhauser, K.O., & Newman, D.K. Deposition of banded
iron formations by anoxygenic phototrophic Fe(II)-oxidizing bacteria. Geology 33,
865-868 (2005).
9. Klein, C. Some Precambrian banded iron-formations (BIFs) from around the world:
Their age, geologic setting, mineralogy, metamorphism, geochemistry, and origin.
Amer. Mineral. 90, 1473-1499 (2005).
10. Beukes, N.J. & Cairncross, B. A lithostratigraphic-sedimentological reference profile for
the late Archaean Mozaan Group, Pongola Sequence; application to sequence
stratigraphy and correlation with the Witwatersrand Supergroup. S. Afr. J. Geol. 94,
44-69 (1991)
11. Nhleko, N. The Pongola Supergroup in Swaziland, Ph.D. thesis, Rand Afrikaans Univ.
(2003).
12. Alexander, B.W., Bau, M., Andersson, P., & Dulski, P. Continentally-derived solutes in
shallow Archean seawater: Rare earth element and Nd isotope evidence in iron
formation from the 2.9 Ga Pongola Supergroup, South Africa. Geochim. Cosmochim.
Acta 72, 378-394 (2008).
13. Wronkiewicz, D.J. & Condie, K.C. Geochemistry and provenance of sediments from the
Pongola Supergroup, South Africa: Evidence for a 3.0-Ga-old continental craton
Geochim. Cosmochim. Acta 53, 1537-1549 (1989).
14. McLennan, S. M., Taylor, S. R. & Kröner, A. Geochemical evolution of Archean shales
from South Africa. I. The Swaziland and Pongola Supergroups. Precam. Res. 22, 93-
124 (1983).
15. Jahn, B. & Condie, K.C. Evolution of the Kaapvaal Craton as viewed from geochemical
and Sm-Nd isotopic analyses of intracratonic pelites. Geochim. Cosmochim. Acta 59,
2239-2258 (1995).
162

16. Watanabe Y., Naraoka, H., Wronkiewicz, D.J., Condie, K.C., & Ohmoto, H. Carbon,
nitrogen, and sulfur geochemistry of Archean and Proterozoic shales from the
Kaapvaal Craton, South Africa. Geochim. Cosmochim. Acta 61, 3441-3459 (1997).
17. Fedo, C.M., Eriksson, K.A., & Krogstad, E.J. Geochemistry of shales from the Archean
(~3.0 Ga) Buhwa Greenstone Belt, Zimbabwe: Implications for provenance and
source-area weathering. Geochim. Cosmochim. Acta 60, 1751-1763 (1996).
18. Deer, W.A., Howie, R.A., & Zussman, J. An Introduction to the Rock-Forming Minerals.
(Longman Sci. & Tech., Essex, 1992).
19. Comans, R.N.J. et al. Mobilization of radiocaesium in pore water of lake sediments.
Nature 339, 367-369 (1989).
20. James, R.H. & Palmer, M.R. Marine geochemical cycles of the alkali elements and
boron: The role of sediments. Geochim. Cosmochim. Acta 64, 3111-3122 (2000).
21. Chapman, H. D. in Methods of Soil Analysis Part 2 (ed. Black, C.A.) 891-901 (Am. Inst.
Agron., Madison, Wisconsin, 1965).
22. Hendershot, W.H. & Duquette, M. A simple barium chloride method for determining
cation exchange capacity and exchangeable cations. Soil Sci Soc Am J. 50, 605-608
(1986).
23. Kamei, G., Oda, C., Mitsui, S., Shibata, M., & Shinozaki, T. Fe(II )-Na ion exchange at
interlayers of smectite: adsorption-desorption experiments and a natural analogue.
Eng. Geol. 54, 15-20 (1999).
24. Charlet, L. & Tournassat, C. Fe(II)–Na(I)–Ca(II) cation exchange on montmorillonite in
chloride medium: evidence for preferential clay adsorption of chloride–metal ion pairs
in seawater. Aqua. Geochem. 11, 115-137 (2005).
25. Rosing, M.T. & Frei, R. U-rich Archaean sea-floor sediments from Greenland -
indications of >3700 Ma oxygenic photosynthesis. Earth Planet. Sci. Lett. 217, 237-
244 (2004).
26. Konhauser, K.O. et al. Decoupling photochemical Fe(II) oxidation from shallow-water
IF deposition. Earth Planet. Sci. Lett. 258, 87-100 (2007).
27. Canfield, D.E., Rosing, M.T., & Bjerrum, C. Early anaerobic metabolisms. Phil. Trans.
R. Soc. B 361, 1819-1836 (2006).
163

THIS PAGE INTENTIONALLY LEFT BLANK
164

Chapter 6. Preservation of primary REE patterns without Ce anomaly during
dolomitization of Mid-Paleoproterozoic limestone and the potential reestablishment
of marine anoxia immediately after the“Great Oxidation
Event”
165

THIS PAGE INTENTIONALLY LEFT BLANK
166

MICHAEL BAU AND BRIAN ALEXANDER
81
Preservation of primary REE patterns without Ce anomaly
during dolomitization of Mid-Paleoproterozoic limestone and the
potential re-establishment of marine anoxia immediately
after the“Great Oxidation Event”
Michael Bau and Brian Alexander
Geosciences and Astrophysics Program, School of Engineering and Sciences,
International University Bremen, P.O. Box 750561, D-28725 Bremen, Germany
email: m.bau@iu-bremen.de; b.alexander@iu-bremen.de
© 2006 March Geological Society of South Africa
ABSTRACT
Comparison of shallow-water limestone and silicified dolomite from the Mid-Paleoproterozoic Mooidraai Formation, Transvaal
Supergroup, South Africa, shows that the primary rare earth element (REE) distribution of these pure marine sedimentary carbonates
has been preserved during dolomitization and silicification. Both lithologies display REE (and Y) patterns closely resembling those
of present-day seawater, i.e. they show enrichment of the heavy relative to the light REE, positive anomalies of La, Gd and Lu, and
super-chondritic Y/Ho ratios. However, these shallow-water carbonates lack any Ce anomalies, indicating that in the Mid-
Paleoproterozoic the redox-level of surface water in the Griqualand-West sub-basin of the Kaapvaal Craton did not allow for
oxidation of Ce(III). The absence of Ce anomalies from shallow water Mooidraai carbonates indicates a return to marine anoxia
immediately after large amounts of marine sedimentary Mn oxides had been deposited in a highly oxygenated marine environment
in the underlying Hotazel Formation. This suggests that the “Great Oxidation Event” in the Paleoproterozoic was a transition period
characterized by strong fluctuations of the redox level of the Earth’s surface ocean.
Introduction
Interpretation of trace element distribution and isotope
ratios in Precambrian sedimentary carbonates suffers
from the fact that these rocks may occur as (sometimes
silicified) dolomites. Field evidence, such as
dolomitization fronts, suggests that the dolomite is not a
primary seawater precipitate and that dolomitization
occurred during diagenesis, but in most cases such
evidence is missing. Hence, conclusions based on the
trace element and isotope compositions of dolomites are
often questioned. This is unfortunate, since Neoarchean
and Paleoproterozoic carbonates such as those from the
Transvaal Supergroup in South Africa and from
the Hamersley Group in Australia, for example, should
have recorded potential changes in the seawater
composition before and after the “Great Oxygenation
Event” and during episodes of low-latitude glaciation in
the Paleoproterozoic (for recent summaries of early
Precambrian atmosphere-hydrosphere evolution see,
e.g. Holland, 2004, Canfield, 2005 and Catling and Claire,
2005).
The rare earths and yttrium (REY) hosted in marine
sedimentary carbonates can be used as proxies for the
REY distribution in ambient seawater, since the partition
coefficients between carbonate and seawater do not
show major differences within the REY series (Terakado
and Masuda, 1988; Zhong and Mucchi, 1995; Webb and
Kamber, 2000; Tanaka et al., 2004). A positive Eu
anomaly, for example, may reveal the presence of a
high-temperature hydrothermal REY component even in
shallow seawater and the presence or absence of
a Ce anomaly may indicate an oxygenated or
anoxic atmosphere-hydrosphere system, respectively.
Understanding the impact of dolomitization on the REY
distribution in marine sedimentary carbonates, therefore,
might enable us to significantly improve our knowledge
of the chemical evolution of the atmospherehydrosphere
system across the Archean-Proterozoic
boundary and in the early Proterozoic. The topic of REY
mobility during dolomitization had been addressed,
amongst others, by Banner et al. (1988) and Nothdurft
et al. (2003), for example, who studied sedimentary
carbonates of Phanerozoic age. However, their
conflicting conclusions do not prove or disprove the
suitability of REY in Precambrian dolomites as proxies
for REY in ambient seawater, but highlight the need for
further study.
Here, we compare the REY distribution in marine
limestone and in silicified dolomite from the Paleoproterozoic
Mooidraai Formation in the Postmasburg
Group of the Transvaal Supergroup, South Africa.
We show that dolomitization did not modify the primary
REY distribution of the marine sedimentary carbonates,
suggesting that these dolomites still provide information
on the REY distribution in Precambrian seawater.
We also show that neither the limestones nor the
dolomites from the Mooidraai Formation display
significant Ce anomalies and, hence, do not support the
assumption of the existence of strongly oxygenated
surface seawater during Mid-Paleoproterozoic
“Mooidraai times”.
Geology
The Mooidraai Formation occurs within the uppermost
part of the Neoarchean to Paleoproterozoic Transvaal
Supergroup, Northern Cape Province, South Africa.
SOUTH AFRICAN JOURNAL OF GEOLOGY, 2006,VOLUME 109 PAGE 81-86

82
PRESERVATION OF PRIMARY REE PATTERNS WITHOUT CE ANOMALY
Figure 1. Simplified stratigraphic position of the Mooidraai and
Hotazel formations of the Transvaal Supergroup, South Africa.
Its stratigraphic position is indicated in Figure 1.
Deposition of iron formations and intercalated
manganiferous oxides and carbonates of the Hotazel
Formation followed the extrusion of the Ongeluk
basaltic andesites that define the base of the Voëlwater
Subgroup. The Hotazel Formation is best known as host
to the enormous resource of manganese ores that
constitute the Kalahari Manganese Field. Conformably
above the Hotazel Formation follow shallow marine
limestones and dolomites of the Mooidraai Formation.
Carbonate slump breccias and sideritic carbonate units
are restricted to the lower part of the Mooidraai
Formation, whereas microbiolaminated and stromatolitic
carbonates predominate the upper part. The contact of
the Mooidraai Formation with the discordantly overlying
Mapedi shales and quartzites of the Olifantshoek
Formation is erosional. The Mooidraai carbonates did
neither experience significant metamorphism nor
deformation. A detailed description of the geological
setting of the Mooidraai Formation can be found in
Beukes (1983) and Swart (1999).
The petrography of the limestones and of the
silicified dolomites has been described by Tsikos et al.
(2001) and Bau et al. (1999), respectively. Note that the
dolomite samples examined here originate from the
stromatolitic upper part of the Moodraai Formation (Bau
et al., 1999); the limestone samples, on the other hand,
are thought to represent the complete succession (Tsikos
et al., 2001). The geology and the C-O-Sr isotope
geochemistry of the Hotazel and Mooidraai formations
are discussed by Schneiderhan et al. (2006).
Unfortunately, the respective absolute ages of the
Hotazel and the Mooidraai Formation are not very well
constrained. No radiometric age is available for the
Hotazel Formation and the Pb-Pb carbonate “age” for
the Mooidraai dolomites discussed here is 2394 +/- 26
Ma (Bau et al., 1999). Since the reliability of Pb-Pb
carbonate ages is a matter of concern, the correct
depositional age of the Hotazel and Mooidraai sediments
is still unclear. The currently most widely accepted age
for the Voëlwater Subgroup is bracketed by an illconstrained
2222 Ma age for the underlying Ongeluk
lava (the problems associated with the data on which
the Ongeluk age is based have been addressed by Bau
et al., 1999) and by the 2060 Ma ages of the Bushveld
and Molopo Farms igneous complexes (e.g., Walraven
et al., 1990, Walraven and Hattingh, 1993; Reichardt,
1994; Buick et al., 2001). If the stratigraphic correlation
between the Ongeluk and the Hekpoort extrusives is
correct, a recent Re-Os age of 2316 +/- 7 Ma for pyrite
in a carbonaceous shale from the base of the Timeball
Hill Formation (Hannah et al., 2004) below the
Hekpoort basalt narrows the depositional age of the
Hotazel and Mooidraai formations down to between
~2.32 and ~2.06 Ga. However, the “normal” 13 C values
(Figure 2) of the Mooidraai carbonates (Bau et al., 1999;
Tsikos et al., 2001) demonstrate that they formed
before the Lomagundi Event, i.e. before the
worldwide positive excursion of 13 C values of marine
carbonate deposited between ~2.25 Ga and ~2.07 Ga
ago (Aharon, 2005, and references therein). Hence, the
current best estimate suggests that the Hotazel
iron and manganese formations and the Mooidraai
carbonates formed between ~2.32 and ~2.25 Ga ago.
The Hotazel Formation and the Mooidraai Formation,
therefore, represent a well-preserved Mid-
Paleoproterozoic succession of shallow marine chemical
sediments.
Results
The chemical compositions of limestone and silicified
dolomite from the Mooidraai Formation have been
presented by Tsikos et al. (2001) and Bau et al. (1999),
respectively, and will only be summarized here (see
these publications for data). Additional data can be
found in Swart (1999) and Schneiderhan et al. (2006),
but since REY data are incomplete or not available, we
will not consider these in the present contribution which
focusses on the REY distribution.
The Mooidraai limestone samples (Figure 2) are only
slightly silicified (CaO: 34.07 to 52.56%; MgO: 0.60 to
1.94%; SiO 2 : 1.39 to 13.50%) and show low MnO (0.14
to 1.30%) but high Fe 2 O 3 (2.42 to 23.92%). The most Feenriched
carbonates occur in the bottom and top parts
of the Mooidraai Formation (Tsikos et al., 2001). Low
Al 2 O 3 contents (0.10 to 0.53%) indicate that the amount
of detrital aluminosilicates in the chemical sediment is
negligible. Ba and Sr concentrations range from 9 ppm
to 105 ppm, and from 789 ppm to 1524 ppm,
respectively. Concentrations of individual REY vary
within a factor of six between samples; Nd
concentrations, for example, range from 0.26 ppm to
1.69 ppm. REY SN patterns (‘ SN ’: indicates normalization
to post-Archean Australian Shale, PAAS, from McLennan,
1989; note that Tsikos et al. (2001) do not report Y data)
are strongly HREE-enriched (Figure 3).
The Mooidraai dolomite (Figure 2) discussed here is
strongly silicified (CaO: 15.68 to 17.74%; MgO: 10.04 to
11.59%; SiO 2 : 39.3 to 46.5%). MnO and Fe 2 O 3 range from
0.25 to 0.30% and from 1.26 to 1.44%, respectively,
SOUTH AFRICAN JOURNAL OF GEOLOGY

MICHAEL BAU AND BRIAN ALEXANDER
83
parallel to those of the Mooidraai limestones (Figure 3).
Both limestones and dolomites show similar HREE
enrichment, positive La SN and Gd SN anomalies, and even
small positive Lu SN anomalies (Figure 3), although the
size of the positive La SN anomalies is somewhat larger in
the limestones than it is in the dolomites. The presence
of Eu SN anomalies is hard to verify because of the
anomalous Gd content, but the Eu SN /Eu SN * ratios
calculated as Eu SN /(0.67Sm SN +0.33Tb SN ) overlap, and
range from 1.11 to 1.51 with an average of 1.28 in the
limestones compared to a range from 0.91 to 1.29 with
an average of 1.11 in the dolomites.
This preservation of primary REYSN patterns during
dolomitization of marine sedimentary carbonates
demonstrates that Precambrian dolomites can still
provide valuable information on the REY distribution in
Precambrian seawater.
Figure 2. Major element and stable isotope characteristics of
limestone and dolomite of the Mooidraai Formation, Transvaal
Supergroup, South Africa (data from Tsikos et al., 2001, and Bau et
al., 1999). Note that the Mooidraai carbonates do not show
anomalously positive 13 C values, suggesting that they formed
before the Lomagundi Event, i.e. prior to ~2.25 Ga ago.
Implications for Earth’s Mid-Paleoproterozoic
atmosphere and oceans
The REY distribution in the Mooidraai sedimentary
carbonates indicates that these are marine chemical
sediments, as the combination of HREE enrichment with
positive La SN , Gd SN , Lu SN , and Y SN anomalies (or superchondritic
Y/Ho ratios) cannot be observed in lakes and
rivers but is confined to seawater. These anomalies for
which after compensating for the effects of silification is
similar to what is seen in the Fe-poor samples of
Mooidraai limestone. Concentrations of elements
typically associated with detrital aluminosilicates are
very low (Rb: 0.05 to 0.17 ppm; Cs: 0.014 to 0.023 ppm;
Al 2 O 3 , Zr, Hf, and Th are below the respective lower
limit of determination), attesting to the purity of the
chemical sediment. Ba and Sr concentrations
are significantly lower than those of the limestone
samples and range from 1.1 to 2.5 ppm and from
12.8 to 15.3 ppm, respectively. REY concentrations are
also low (e.g., Nd: 0.102 – 0.125 ppm) and REY SN
patterns show pronounced enrichment of the HREE
(Figure 3).
Discussion
Comparing REY in limestones and dolomites
Compared to the limestones, the Mooidraai dolomites
show significantly lower Sr and Ba concentrations and
reflect the loss of Sr, for example, typically observed
during dolomitization (e.g., Veizer, 1983a; b). However,
dolomitization did not produce elevated Fe and Mn
concentrations as is observed in Phanerozoic carbonates
(e.g., Veizer, 1983a; b), suggesting only minor
remobilization of Mn and Fe. This may have implications
for the interpretation of Fe isotope data of Precambrian
sedimentary carbonates, since primary Fe isotope
signatures may be preserved during dolomitization.
Despite dolomitization and strong silicification, the
Mooidraai dolomites show REY SN patterns that are
Figure 3. Shale-normalized REY patterns of limestones and
silicified dolomites from the Mooidraai Formation and of a sample
of manganese formation (braunite lutite; M. Bau, unpubl. data)
from the underlying Hotazel Formation. The parallel patterns of
the carbonates demonstrate that the primary REY distribution of
the limestone was preserved during dolomitization and
silicification. Heavy REE enrichment and positive anomalies of La,
Gd, Y, and Lu indicate that these carbonates are pure marine
chemical sediments. Except for the lack of a negative Ce anomaly,
these REY patterns are very similar to those of modern seawater
and modern marine sedimentary carbonates. Note that in contrast
to the Mooidraai carbonates, manganese formation from the
Hotazel Formation does display a negative Ce SN anomaly.
SOUTH AFRICAN JOURNAL OF GEOLOGY

84
PRESERVATION OF PRIMARY REE PATTERNS WITHOUT CE ANOMALY
Figure 4. Plot of Ce SN /Ce SN * vs. Pr SN /Pr SN * that allows to quantify
Ce and La anomalies. The Mooidraai carbonates show Pr SN /Pr SN *
ratios close to unity, i.e. similar to those of shales, which
demonstrates the absence of negative Ce anomalies. Ce SN /Ce SN *
ratios below unity indicate the presence of positive La anomalies
in the Mooidraai carbonates, as commonly found in seawater and
marine chemical sediments. The absence of significant negative Ce
anomalies suggests that surface seawater during “Mooidraai times”
was too reducing to allow for oxidation of Ce(III). For further
explanation see text. Data from Tsikos et al., 2001 (1); Bau et al.,
1999 (2); Bau et al., 2003, and M. Bau, unpublished data (3);
Taylor and McLennan, 1985 (4, 5).
the non-redox-sensitive REY together with the lack of
siliciclastic detritus also preclude deposition of the
Mooidraai Formation in coastal waters in close proximity
to a land mass, but rather support the drowned-platform
environment suggested by Beukes (1983).
The lack of clear positive Eu SN anomalies in PAASnormalized
REY distribution patterns demonstrates the
absence of a high-temperature hydrothermal REY
component from shallow seawater during “Mooidraai
times”. This is different from what is seen in Neoarchean
carbonates, such as those from the Lower Transvaal
Supergroup, South Africa (Kamber and Webb, 2001), or
from the Hamersley Supergroup, Australia (M. Bau,
unpublished data), and in marked contrast to what is
observed in iron formation from the lower Transvaal
Supergroup, that give strong evidence for a blacksmoker-type
hydrothermal REY source (Bau and Dulski,
1996; Bau et al., 1997).
We emphasize that there exist no significant Ce SN
anomalies in the Mooidraai limestones or dolomites.
As discussed previously (Bau and Dulski, 1996;
Kamber and Webb, 2001), the commonly used
equation to quantifiy Ce SN anomalies (Ce SN /Ce SN * =
Ce SN /(0.5La SN +0.5Pr SN )) must not be applied in studies
of marine chemical sediments due to the presence of
positive La SN anomalies that suggest Ce SN * values that are
much too high. A plot of Ce SN /Ce SN * vs. Pr SN /Pr SN * (Figure
4; Pr SN * = 0.5Ce SN +0.5Nd SN ; for details on this approach
see, for example, Bau and Dulski, 1996; Kamber and
Webb, 2001) clearly demonstrates that neither the
limestones nor the dolomites from the Mooidraai
Formation are characterized by distinct negative Ce SN
anomalies (Figure 4). The Mooidraai carbonates,
although free from detrital aluminosilicates, show
Pr SN /Pr SN * ratios that are considerably smaller
than those of Lower Carboniferous limestones from
Ireland and England, for example, that display
negative Ce SN anomalies (Bau et al., 2003; M. Bau,
unpublished data). Their Pr SN /Pr SN * ratios are, however,
very similar to those typical of shales (Figure 4). The
absence of Ce SN anomalies from the dolomites does not
result from the dolomitization process but is a primary
feature of the chemical sediment and, therefore, reflects
the absence of any Ce SN anomaly from Mooidraai
seawater.
We emphasize that the Mooidraai REY data are yet
another example of how neglecting the existence of
positive La SN anomalies in seawater and in some
marine chemical sediments can lead to false claims of
the presence of negative Ce SN anomalies. Using a
Pr SN /Pr SN * ratio of >1.1 as the positive criterion for
a negative Ce SN anomaly, it appears that except for a few
individual analyses of samples from iron-formations and
cherts in India (e.g., Kato et al., 2002, and references
therein), there exist no published data from Archean or
Early Paleoproterozoic chemical sediments that show
primary Ce SN anomalies.
The absence of distinct negative Ce SN anomalies from
the Mooidraai carbonates, however, is surprising
considering that the Mooidraai Formation was deposited
immediately above the Hotazel Formation. The Hotazel
Formation contains the first and hence oldest significant
deposits of marine sedimentary Mn oxides in the
geological record. The precipitation of these
sedimentary Mn oxides represents a major event in
Earth’s history as the Mn oxide deposits are widespread
and several meters thick (for a recent discussion, see
Schneiderhan et al., 2006). These Mn oxides are the
prot-ore for the Mn ore in the Kalahari Manganese Field,
the largest deposit of sedimentary Mn oxides on Earth.
The precipitation and, in particular, the preservation of
such a huge amount of marine sedimentary Mn oxides
requires a highly oxygenated depositional environment
(e.g., Kirschvink et al., 2000; Kirschvink and Weiss, 2002;
Kopp et al., 2005). The redox level necessary to stabilize
significant amounts of Mn(IV) oxides, however, is
significantly higher than that needed to oxidize Ce(III)
and to form Ce anomalies. Hence, it comes as no
surprise that, indeed, negative Ce SN anomalies occur in
manganese formation from the Hotazel Formation
(Figure 3).
Their stratigraphic position above the Mn-oxidesbearing
Hotazel Formation and, in particular, above the
SOUTH AFRICAN JOURNAL OF GEOLOGY

MICHAEL BAU AND BRIAN ALEXANDER
85
~2.32 Ga-old Timeball Hill Formation that does not
show mass-independent sulphur isotope fractionation
(Bekker et al., 2004), indicates that the deposition of the
Mooidraai carbonates post-dates the “Great Oxygenation
Event”. The lack of specific Ce depletion in Mooidraai
seawater, therefore, suggests that after the deposition of
the Hotazel Formation the redox-level of seawater in the
Griqualand-West sub-basin had returned to a lower level
insufficient to oxidize Ce(III). Apparently, Mn(II)
oxidation had consumed the emerging oxygen reservoir
and biogenic oxygen production was not yet able to
balance the additional oxygen demand. Thus, local
marine anoxia was re-established during Mid-
Paleoproterozoic Mooidraai times.
It remains to be seen whether this retreat to marine
anoxia in the Mid-Paleoproterozoic was a worldwide
phenomenon or whether it was confined to the
Griqualand-West sub-basin of the Kaapvaal Craton.
Nevertheless, the REY data from the Mooidraai
limestones and dolomites suggest that even the
oxygenation of the surface water of the Earth’s oceans
did not follow a unidirectional trend from generally
anoxic to generally oxic conditions. Rather, oxygenated
surface environments became progressively more
abundant until the previously exotic “oxygen oases” had
grown and become more persistent before they
eventually represented the normal state of the system.
Hence, it might be more apt to refer to the oxygenation
of the Earth’s surface system in the Paleoproterozoic as
the “Great Oxygenation Period”. Such a soft rather than
sharp transition is also more in line with, for example,
the contemporaneous formation of oxic and anoxic
paleosols (Rye and Holland, 1998, Yang and Holland,
2003), the observation of abundant hydrothermal mantle
Os in pyrites that do not show mass-independent
sulphur isotope fractionation (Bekker et al., 2004;
Hannah et al., 2004), and low Mo concentration and
unfractionated Mo isotope ratios in the Transvaal
Supergroup (Siebert et al., 2004).
Conclusions
Comparison of marine shallow-water limestone and
silicified dolomite from the Mid-Paleoproterozoic
Mooidraai Formation, Transvaal Supergroup, South
Africa, demonstrates that the REY distribution of the
marine sedimentary carbonate was preserved during
dolomitization and silicification. With one exception,
both lithologies display all the details of the REY
distibution in present-day seawater, such as positive
anomalies for La, Gd and Lu, and a super-chondritic
Y/Ho ratio. However, the Mooidraai carbonates lack the
negative Ce anomaly that indicates oxidation of Ce(III)
in the Earth’s surface system. It appears that after the
deposition of Mn oxides in the Hotazel Formation,
which is indicative of a highly oxygenated supergene
environment, conditions again became sigificantly less
oxic. This suggests that during the transition period from
a rather reducing to an oxygenated atmospherehydrosphere
system in the Paleoproterozoic (the “Great
Oxygenation Period”) the redox-level of the Earth’s
surface ocean fluctuated between reducing and oxic.
Acknowledgements
We gratefully acknowledge many fruitful discussions
of Precambrian geochemistry and geology with
H.D. Holland, J. Gutzmer, and J. Kasting. The paper also
benefited from constructive reviews by A. Lepland,
G. Shields, and J. Gutzmer. In particular, however, we
want to thank Nic Beukes for his advice, help, and kind
hospitality during many years of collaborative work.
His efforts have without doubt stimulated worldwide
interest in and promoted South African chemical
sediments as a valuable source of information on the
chemical evolution of the Earth’s atmospherehydrosphere
system.
References
Aharon, P. (2005) Redox stratification and anoxia of the early Precambrian
oceans: Implications for carbon isotope excursions and oxidation events.
Precambrian Research, 137, 207-222.
Banner, J.K., Hanson, G.L. and Meyers, W.J. (1988) Rare earth element and
Nd isotopic variations in regionally extensive dolomites from the
Burlington-Keokuk Formation (Mississippian): Implications for REE
mobility during carbonate diagenesis. Journal of Sedimentary Petrology,
58, 415-432.
Bau, M. and Dulski, P. (1996) Distribution of yttrium and rare-earth elements
in the Penge and Kuruman Iron-Formations, Transvaal Supergroup, South
Africa. Precambrian Research, 79, 37-55.
Bau, M., Höhndorf, A., Dulski, P. and Beukes, N.J. (1997) Sources of rareearth
elements and iron in Paleoproterozoic iron-formations from the
Transvaal Supergroup, South Africa: Evidence from neodymium isotopes.
Journal of Geology, 105, 121-129.
Bau, M., Romer, R., Lüders, V. and Beukes, N.J. (1999) Pb, O, and C isotopes
in silicified Mooidraai dolomite (Transvaal Supergroup, South Africa):
Implications for the composition of Paleoproterozoic seawater and “dating”
the increase of oxygen in the Precambrian atmosphere. Earth and
Planetary Science Letters, 174, 43-57.
Bau, M., Romer, R., Lüders, V. and Dulski, P. (2003) Tracing element sources
of hydrothermal mineral deposits: REE and Y distribution and Sr-Nd-Pb
isotopes in fluorite from MVT deposits in the Pennine Orefield, England.
Mineralium Deposita, 38, 992-1008.
Bekker, A., Holland, H.D., Wang, P.-L., Rumble, D. III, Stein, H.J., Hannah ,
J.L., Coetzee, L.L. and Beukes, N.J. (2004) Dating the rise of atmospheric
oxygen. Nature, 427, 117-120.
Beukes, N. J. (1983) Palaeoenvironmental setting of iron-formations in the
depositional basin of the Transvaal Supergroup, South Africa. In: A.F.
Trendall and R.C. Morris (Editors), Iron-formation: Facts and Problems.
Developments in Precambrian Geology, 6, 131-209.
Buick, I.S., Maas, R. and Gibson, R. (2001) Precise U-Pb titanite age
constraints on the emplacement of the Bushveld Complex, South Africa.
Journal of the Geological Society London, 158, 3-6.
Canfield, D.E. (2005) The early history of atmospheric oxygen: Homage to
Robert M. Garrels. Annual Reviews Earth and Planetary Sciences, 33, 1-36.
Catling, D.C. and Claire, M.W. (2005) How Earth’s atmosphere evolved to an
oxic state: A status report. Earth and Planetary Science Letters, 237, 1-20.
Hannah, J.L., Bekker, A., Stein, H.J., Markey, R.J. and Holland, H.D. (2004)
Primitive Os and 2316 Ma age for marine shale: implications for
Paleoproterozoic glacial events and the rise of atmospheric oxygen. Earth
and Planetary Science Letters, 225, 43-52.
Holland, H.D. (2004) The geological history of seawater. In: H.D. Holland
and K.K. Turekian (Editors), Treatise of Geochemistry, Elsevier,
Amsterdam, The Netherlands, 6, 583-625.
Kamber, B.S. and Webb, G.E. (2001) The geochemistry of late Archaean
microbial carbonate: Implications for ocean chemistry and continental
erosion history. Geochimica et Cosmochimica Acta, 65, 2509-2525.
Kato Y., Kano T. and Kunugiza, K. (2002) Negative Ce Anomaly in the Indian
Banded Iron Formations: Evidence for the Emergence of Oxygenated
SOUTH AFRICAN JOURNAL OF GEOLOGY

86
PRESERVATION OF PRIMARY REE PATTERNS WITHOUT CE ANOMALY
Deep-Sea at 2.9~2.7 Ga. Resource Geology, 52, 101-110.
Kirschvink, J.L. and Weiss, B.P. (2002) Mars, panspermia, and the origin of
life: Where did it all begin? Palaeontologia Electronica, 4, 8-15.
Kirschvink, J.L., Gaidos, E.J., Bertani, L.E., Beukes, N.J., Gutzmer, J., Maepa,
L.N. and Steinberger, R.E. (2000) Paleoproterozoic Snowball Earth: Extreme
climatic and geochemical global change and its biological consequences .
Proceedings of the National Acadamy of Sciences, 97, 1400-1405.
Kopp, R.E., Kirschvink, J.L., Hilburn, I.A. and Nash, C.Z. (2005) The
Paleoproterozoic snowball Earth: A climate disaster triggered by the
evolution of oxygenic photosynthesis. Proceedings of the National
Acadamy of Sciences, 102, 11131-11136.
McLennan, S.M. (1989) Rare earth elements in sedimentary rocks: Influence
of provenance and sedimentary processes. In: B.R. Lipin and G.A. McKay
(Editors), Geochemistry and mineralogy of rare earth elements. Reviews in
Mineralogy, 21, 169-200.
Nothdurft, L.D., Webb, G.E. and Kamber, B.S. (2003) Rare earth element
geochemistry of Late Devonian reefal carbonates, Canning basin, Western
Australia: Confirmation of a seawater REE proxy in ancient limestones.
Geochimica et Cosmochimica Acta, 68, 263-283.
Reichardt, F.J. (1994) The Molopo Farms complex, botswana: history,
stratigraphy, petrography, petrochemistry and Ni-Cu-PGE mineralization.
Exploration and Mining Geology, 3, 264-284.
Rye , R. and Holland , H.D. (1998) Paleosols and the evolution of
atmospheric oxygen: A critical review. American Journal of Science,
298, 621–672.
Schneiderhan, E.A., Gutzmer, J., Strauss, H., Mezger, K. and Beukes, N.J.
(2006) The chemostratigraphy of a Paleoproterozoic MnF-BIF-dolostone
sucession – the Voëlwater Subgroup of the Transvaal Supergroup in
Griqualand West, South Africa. South African Journal of Geology,
109, 63-80.
Siebert, C., Kramer, J.D., Meisel, Th., Morel, Ph. and Nägler, Th.F. (2005)
PGE, Re-Os, and Mo isotope systematics in Archean and early Proterozoic
sedimentary systems as proxies for redox conditions on the early Earth.
Geochimica et Cosmochimica Acta, 69, 1787-1801.
Swart, Q.D. (1999) Carbonate rocks of the Paleoproterozoic Pretoria and
Postmasburg Groups, Transvaal Supergroup. Unpublished M.Sc thesis,
Rand Afrikaans University, Johannesburg, South Africa, 126pp.
Tanaka, K., Ohta, A. and Kawabe, I. (2004) Experimental REE partitioning
between calcite and aqueous solution at 25°C and 1 atm: Constraints on
the incorporation of seawater REE into seamount-type limestones.
Geochemical Journal, 38, 19-32.
Taylor, S.R. and McLennan, S.M. (1985) The continental crust: its composition
and evolution. Oxford, Blackwell, United Kingdom, 312pp.
Terakado, Y. and Masuda, A. (1988). The coprecipitation of rare-earth
elements with calcite and aragonite. Chemical Geology, 69, 103-110.
Tsikos, H., Moore, J.M. and Harris, C. (2001) Geochemistry of the
Palaeoproterozoic Mooidraai Formation: Fe-rich limestone as end-member
of iron-formation deposition, Transvaal Supergroup, South Africa. Journal
of African Earth Sciences, 32, 19-27.
Veizer, J. (1983a) Chemical diagenesis of carbonate rocks: Theory and
application of trace element technique. In: M.A. Arthur et al. (Editors),
Stable Isotopes in Sedimentary Geology, SEPM Short Course Notes,
10, III/1 – III/100.
Veizer, J. (1983b) Trace elements and isotopes in sedimentary carbonates.
In: E. Roedder (Editor), Carbonates: Mineralogy and Chemistry, Reviews in
Mineralogy, 11, 265-300.
Walraven, F., Armstrong, R.A. and Kruger, F.J. (1990) A chronostratigraphic
framework for the north-central Kaapvaal craton, the Bushveld Complex
and the Vredefort structure. Tectonophysics, 171, 23-48.
Walraven, F. and Hattingh, E. (1993) Geochronology of the Nebo Granite,
Bushveld Complex. South African Journal of Geology, 96, 31-41.
Webb, G. E. and Kamber, B. S. (2000) Rare earth elements in Holocene reefal
microbialites: a new shallow seawater proxy. Geochimica et Cosmochimica
Acta, 64, 1557-1565.
Yang, W. and Holland, H.D. (2003). The Hekpoort paleosol profile in Strata
1 at Gaborone, Botswana: Soil formation during the Great Oxidation Event.
American Journal of Science, 303, 187-220.
Zhong, S.Y. and Mucci, A. (1995) Partitioning of rare earth elements (REEs)
between calcite and seawater solutions at 25ºC and 1 atm, and high
dissolved REE concentrations. Geochimica et Cosmochimica Acta,
59, 443-453.
Editorial handling: J. Gutzmer
SOUTH AFRICAN JOURNAL OF GEOLOGY

Chapter 7. Concluding remarks
The data presented in this thesis presumably constrain certain geochemical
features of seawater for the rather narrow period of time between 2.9-3.0 Ga. Studies of
seawater chemistry that focus on younger rocks benefit from both the increased
preservation of potential seawater archives as well as the greater occurrences of marine
carbonates. However, in many regards, the geochemistry of Archean seawater is poorly
constrained, particularly for the period of time prior to ~3.0 Ga. Though rocks older
than 3.0 Ga are not particularly rare, seawater archives from this time period are limited.
Further work is necessary to identify seawater archives older than 3.0 Ga, and data not
included in this thesis suggest that suitable material from southern Africa is available
for the time period between 3.2-3.6 Ga. Neodymium isotopic studies of these samples
would help refute or support the conclusion that bulk seawater Є Nd values were generally
positive with values between 0 and +2 from 3.0-3.8 Ga.
For younger rocks, unpublished Nd isotopic data indicates that shallow water
marine carbonates from the Transvaal Supergroup in South Africa possess Є Nd (2.5 Ga)
between -0.3 and -2.2, consistent with the conclusions of the Pongola IF study that
weathering of continental crust significantly influenced seawater chemistry near the
Kaapvaal craton after 2.9 Ga. Furthermore, Mn sediments from the ~2.25 Ga Hotazel
Formation in South Africa display seawater-like REY patterns, and possess Є Nd values
between -2.2 and -4.2. These data are consistent with an increasingly important flux of
solutes derived from weathering of continental crust in younger marine chemical
precipitates. A possible test of this hypothesis would be Nd isotopic studies on late
Archean (2.5-2.9 Ga) seawater archives of well defined age and representing a variety
of depositional environments. Such studies may resolve the question of whether or not
173

ulk, open ocean seawater tended to more positive Є Nd values as suggested by the
Pietersburg IF data.
Finally, the suggestion that iron-formations may have precipitated as Fe(II)
minerals is worthy of future study. Many unanswered questions remain regarding this
hypothesis, ranging from the primary mineralogy of such a precipitate to its conversion
to the Fe(III) mineral phases that dominate oxide facies IFs. One intriguing aspect of
such a model regards the REY patterns of many Archean IFs. As discussed in the
Introduction to this thesis, Fe oxy-hydroxides that precipitate from, and are in
equilibrium with, modern seawater, display REY patterns distinctly different from
ambient seawater. The seawater-like REY patterns observed in many Archean-
Paleoproterozoic IFs have been attributed to quantitative scavenging of the REY from
ancient seawater during the process(es) that produced IFs. If, however, the primary IF
precipitate was predominantly composed of Fe(II) mineral phases, then different
mechanisms for incorporating the REY into IFs may have operated, and further
examination of this potential process is warranted.
174