Chemical Geology 266 (2009) 194–204

Contents lists available at ScienceDirect

Chemical Geology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h e m g e o

A new procedure for separating and measuring radiogenic isotopes (U, Th, Pa, Ra, Sr,
Nd, Hf) in ice cores
Sarah M. Aciego a,⁎, Bernard Bourdon a, Maarten Lupker b, Joerg Rickli a
a
Institute of Isotope Geochemistry and Mineral Resources, Clausiusstrasse 25, NW, ETH-Zurich, Zurich, Switzerland
b
Centre de Recherches Pétrographiques et Géochimiques, CNRS, BP20, Vandoeuvre-les-Nancy, France

a r t i c l e i n f o a b s t r a c t

Article history: A new method for the radiogenic isotope (U–Th–Pa–Ra, Sr, Nd, Hf) analysis of the soluble and insoluble
Received 14 January 2009 components found within ice cores is presented. Melting experiments with rock standards in the presence of
Received in revised form 29 May 2009 EDTA indicate that carbonates, as well as silicates, can be buffered sufficiently to preclude dissolution. The
Accepted 2 June 2009
use of EDTA allows adsorbing species, such as Th and Hf, to remain in solution during melting thus fully
Editor: J. Fein separating the dust (insoluble) and sea salt (soluble) components of the ice after filtration. A new elemental
separation scheme for low sample masses, less than 5 mg solid material, utilizes 4 primary ion exchange
Keywords: columns and two “clean-up” columns to fully isolate U, Th, Pa, Ra, Sr, Nd, and Hf while maintaining high
U-series yields. Elution schemes measured for USGS rock standards and a Chinese loess are presented to provide a
Hf comparison for variable matrix compositions. Mass spectrometer techniques were modified to measure small
Nd aliquots of the standards, equivalent to the amounts found in ice core samples, 10 ng and less. A MC-ICPMS
Sr was employed for the measurement of U, Th, Pa, Ra, and Hf; results of the experiments show that with ion
Ice cores
yields up to 1%, rock standards have errors for 234U/238U of 1%, 230Th/232Th of 1.5%, [228Ra] of 9%, and 176Hf/
Sea salt 177
Hf of 100 ppm. MC-TIMS measurements of Sr and Nd show similar errors for small sample sizes: 87Sr/86Sr
of 50 ppm and 143Nd/144Nd of 80 ppm. This new analytical method increases the number of possible tracers
measured from a single sample, reducing separation times and sample consumption, as well as providing the
addition of a radiometric clock, U-series, to the traditional suite of isotopic tracers, Sr, Nd, and Hf.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction
The isotopic compositions of the hydrosphere and cryosphere are
primarily controlled by rocks exposed at the Earth's surface. Analyzing
and interpreting the isotopic compositions of uranium-series isotopes
(U, Th, Pa, Ra), strontium (Sr), neodymium (Nd), and hafnium (Hf) has
become a powerful tool to trace surface processes (e.g. climate change,
ocean circulation and chemical weathering) both in terms of fluxes and
timescales. One of the more powerful tools for studying these processes
is the ice core record; ice sheets record temperature, humidity,
atmosphere composition, sea ice extent and dustiness all as a function
of time. The Sr, Nd and Pb isotopic composition of dust within ice can
provide information about source area, transport pathways and source
area climate (Biscaye et al., 1997; Delmonte et al., 2004; Fischer et al.,
2007). Work by Pettke (2002) and Van De Flierdt et al. (2004) indicate
that Hf may be an additional useful tracer for the provenance of dust,
although its concentrations are low. Furthermore, Fireman (1986) and
Goldstein et al. (2004) have hypothesized that U-series disequilibria can
develop in ice due to recoil of daughter products from particulate solids
trapped within the ice. The methodology for separating and measuring
⁎ Corresponding author.
E-mail address: aciego@erdw.ethz.ch (S.M. Aciego).
each of these elements depends on the amount of sample available and
the matrix material. In the case of ice cores, the amount of material is so
low (b 1 mg dust/kg ice) that pre-existing methodologies are insufficient
to measure Sr, Nd, Hf, and U-series in a single sample.
The purpose of this work is threefold. First, provide a method for
clean separation of the soluble (precipitation, sea salt, and adsorbed
elements) and insoluble components (carbonate and silicate dust) for
ice samples. Second, provide a chromatographic method that allows
quantitative separation and recovery of each element of interest (U–
Th–Pa–Ra, Sr, Nd, Hf). Third, show that the former procedures allow
for accurate and precise measurements of the isotopic compositions.
These techniques should be applicable to ice cores, as well as to other
studies of sub-milligram samples such as pore and river waters,
micro-milled speleothems and evaporation rinds.
2. Experimental procedures
2.1. Reagents, lab ware, and spikes
All chemical separation and sample handling were performed
under class 10 laminar flow hoods within a class 10,000 clean
laboratory. Ultra-pure reagents were used throughout the chemistry
steps: HCl, HNO3 and HF were double distilled in a PicoTrace cupola

0009-2541/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.chemgeo.2009.06.003
 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204
distillation apparatus and ammonia purchased from Seastar Chemi-
cals. Water was purified using the milli-Q Element Millipore
purification system, resulting in 18.2 MΩ grade water. Ethylenedia-
minetetraacetic acid (EDTA) was purchased as an ultra-pure grade
diammonium salt and dissolved with water to form a 0.5 M solution.
Purification of the EDTA solution can be accomplished by cation
exchange at 0.01 M (Volpe et al., 1991) to a baseline value of 100 ppt U
and Th, and b0.05 fg/g Ra. Purification for Ra using cation exchange is
much more efficient than for U and Th. These elements are retained on
the column resin by adjusting the pH: Ra is retained at pH b 9 while U
and Th are retained at pH b 4 with complete separation at pH b 2 (Fritz
and Umbreit, 1958; Nelson, 1964). However, EDTA must be buffered
above pH 2 otherwise it precipitates within the column; acceptable Th
and U blanks are hard to achieve which may explain why Goldstein et
al. (2004) reported ~ 100 pg blanks with this methodology. In order to
attain better blank levels, we used additional cleaning steps. After one
pass through the cation column at pH 4, to remove the radium and
strontium, further purification was accomplished by repeated pre-
cipitation–dissolution steps. Following the work of Mitchell et al.
(1984), we found that precipitating with clean concentrated HCl,
decanting the supernatant, rinsing with clean 1 M HCl, then re-
dissolving in H2O buffered with ammonia resulted in halving the
blanks. Repeated precipitation–dilution steps result in 0.5 M EDTA
with b10 ppt U, Th, Sr, and Nd, b0.05 fg/g ppt Ra, Pa, and Hf. The pH of
EDTA solutions was adjusted by adding ultra-pure ammonia.
In order to avoid contamination, either from manufacturing or
previous samples, an aggressive cleaning procedure was adopted for
all PFA and PTFE lab ware. Materials were first rinsed in ethanol, then
boiled sequentially in reagent grade fresh aqua-regia, 25% HCl, and
18 MΩ water. Finally, the lab ware was cleaned with a mixture of hot
double distilled HNO3 and HF before rinsing with milli-Q Element
water.
We used a range of rock standards and elemental solutions to
evaluate the success of the melting, filtration, and chemical proce-
dures, as well as mass spectrometer techniques. Three rock powders
are certified USGS standards (AGV-2, BHVO-2, and BCR-2), which have
been characterized for elemental and Sr, Nd, and Hf isotopic
measurements. Two other rock powders are U-series standards
(TML and AThO) that are well characterized for both concentrations
and isotopic compositions. We also used a loess sample (CDL) from
Central China as an internal lab standard with similar characteristics
to dust found within ice cores: diverse mineralogy including clay
minerals and carbonate, a size distribution varying from 0.2 to 10 µm,
and an isotopic composition more suitable for comparison to ice core
dust or source area samples. Dilute Alpha Aesar standards (certified
3% concentration accuracy) were used to test the melting, filtration
and elution procedures. Spikes were calibrated either with our in-
house gravimetric standards (made from Hf, Th, Sr, and Nd solids) or
with international reference standards (226Ra NIST SRM4966, 238U
NIST SRM 3164).
Single element spikes of 229Th, 233U, and 236U were prepared at
dilute concentrations, less than 0.12 ppb, such that weighing errors
could be minimized for ultra-low level samples. The 228Ra spike was
prepared using standard techniques: a ~ 50 year old 50 mg aliquot of a
Th(NO3)4 salt was dissolved and the radium was separated by cation
exchange (Volpe et al., 1991). The initial 228Ra/226Ra ratio of the tracer
was 1.309 ± 0.011, and this ratio was corrected for decay of 226Ra and
228
Ra using λ226 = 0.1205 yr− 1 and λ228 = 4.332 × 10− 4 yr− 1. The
233
Pa spike was prepared regularly every 4–6 months by neutron
irradiation of 232Th in the Paul Scherrer Institut reactor in Villigen,
Switzerland. Solutions of Th metal were dried in pre-cleaned quartz
glass ampoules 4 mm in diameter and 10 cm long, resulting in 1 mg of
Th in nitrate salt form at the bottom of each ampoule. The ampoules
are evacuated to less than 0.01 mbar, then sealed by melting the glass.
After irradiation, the tops of the tubes were cut off with a diamond
saw, the salt dissolved in a concentrated HNO3–dilute HF mixture and
195
removed by pipette. The 233Pa spike was separated from Th and other
elements using the procedures of Bourdon et al. (1999). Spikes of 84Sr,
179
Hf, and 150Nd were used to determine total procedural blanks and
yields.
2.2. Separation of dust from ice core samples
2.2.1. Principles
The components we are attempting to separate from ice cores
consist of three main parts: carbonate and silicate solids, soluble
materials, including elements dissolved in precipitation and sea salt,
and adsorbed materials. The behavior of these components initially
during transport to the ice sheets and finally during sample processing
determines the laboratory procedures we have used to separate them.
Silicates are primarily clays—illite, kaolinite and chlorite (Biscaye
et al., 1997)—with some quartz, minor pyroxenes and olivine (Sala
et al., 2008). The source of the carbonate within the dust fraction is
difficult to determine, as both dust and marine solids could be sources.
However, there is some indication that the marine influence can
dominate the carbonate budget. Banta et al. (2008) found that silicate
dust deposition on Greenland at coastal ice core sites was fairly equal
to the deposition at inland ice core sites, but coastal sites had three
times more carbonate load. Similarly, Sala et al. (2008) showed that
ice from Talos Dome, close to the Ross Sea, had both hydrous and
anhydrous carbonates with the hydrous form dominating the total
abundance, indicating a near marine source. More importantly, while
the hydrous carbonate dissolves rapidly during ice melting (Sala et al.,
2008), the anhydrous form can be buffered and remain insoluble.
The soluble fraction will be some mixture of sea salt, which
contains seawater trace elements, and precipitation. Precipitation in
the form of rain will dissolve some fraction of dust traveling in the
same air mass. Dissolution of dust by snow should be limited in high-
altitude inland ice core locations because of the drying of the air
masses. Conversely, lower altitude, near ocean ice core locations will
have a mix of wet and dry air masses (Fischer et al., 2007), which
could produce precipitation with higher trace element concentrations.
The fraction adsorbed onto the dust grains will contain the particle
reactive U-series daughters, such as 230Th and 231Pa that are ejected
from dust grains during uranium decay, as well as some easily
scavenged trace metals such as Pb. Ideally, in order to extract as much
information as possible about dust and sea salt, we would like to be
able to fully separate these three components. During melting of the
ice core, dissolution must be minimized for both anhydrous carbonates
and silicates while preventing adsorption of elements present in the
soluble fraction or ejected from dust grains. Filtering should fully
separate the soluble and insoluble fraction. This procedure is different
from other ice core procedures designed for mineral dust analyses,
which use a decanting method (Delmonte et al., 2004) or remove all
carbonates (Biscaye et al., 1997).
The biggest hurdle in this procedure is not dissolving the solid dust
grains which would mix the isotopic compositions of the seawater and
continental sources of material. For pure isotopic tracing (e.g. Sr
isotopes) this effect is secondary, but is of the highest importance for
the U-series dating technique. If congruent dissolution occurs, the
outer surface will be affected first, and since this domain in the dust
grains is responsible for the excess of daughter nuclides in the ice,
dissolution will push both the liquid and the solid fractions toward
secular equilibrium and an incorrect, young age. The second source of
uncertainty created by dissolution is that carbonates will dissolve at a
faster rate given the slightly acidic conditions of polar ice (Sala et al.,
2008). Dust-derived carbonate will have an isotopic composition that
could differ from the local marine carbonates and sea salt, thereby
affecting isotopic tracers of sea salt and surface seawater as well as the
initial conditions for U-series dating.
Previous work on U-series dating of ice focused primarily on
adsorption (Fireman, 1986) and minimizing dissolution of silicate
196 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204

material (Goldstein et al., 2004). The EDTA method employed by

Goldstein et al. (2004) on Allan Hills ice was successful at desorbing
the daughter products and preventing dissolution of the silicate
material. But the high initial 234U/238U they calculated could indicate
dissolution of carbonate enriched in 234U, such as the local non-
seawater carbonate (Henderson et al., 2006). EDTA leaching is a
standard protocol used to clean sediment samples, but often in
conjunction with carbonate removal (Di Palma and Mecozzi, 2007).
However, there is evidence that buffering the samples to over pH 10
would minimize dissolution (Rihs et al., 2004), and EDTA concentra-
tions less than 0.0005 M could result in carbonate dissolution at the
same rate as in water (Fredd and Fogler, 1998). In order to test these
assumptions we performed a series of experiments with carbonate
and rocks powders, as described below.

2.2.2. Dissolution with controlled adsorption experiments

Standard powders (BCR-2, CDL, and Merck carbonate) were pre-
filtered with clean ethanol to remove the less than 0.2 µm size fraction
and the powders dried under ambient air conditions. Approximately
10 mg of material was placed in a Teflon container containing 500 mL
of EDTA for 6 h. EDTA concentrations ranged from 0.1 M to 0.00025 M
and pH from 2 to 14. These initial tests immediately showed complete
carbonate powder and moderate CDL (10–50%) dissolution at EDTA
concentrations above 0.01 M and pH below 8. The pure silicate
material (BCR-2) was largely immune to changes in concentration,
and showed minor (1–10%) dissolution at pH b 4. Subsequent
experiments used the optimum lower concentration (0.0005 M),
varying pH and included 229Th and 233U tracers to monitor adsorption.
Fig. 1 shows the results of these experiments for Th. Even at low EDTA
concentration, the pH below 8 causes massive dissolution of CDL,
which is likely due to the loss of carbonate. This difference in silicate
versus carbonate dissolution is shown in Fig. 1b. The Th adsorption
shows a maximum at neutral pH, likely the result of a less aggressive
acid attack and low complexing capability of the EDTA; at high pH, the
NH+ 4 starts to replace exchangeable cations on the grain surfaces
which “helps” the EDTA complexation by blocking the active Fig. 1. Adsorption–dissolution experiments of rock and loess powders in the presence of
adsorbent sites. Uranium adsorption generally followed the same 0.0005 M EDTA. (a) Sequential dissolution of loess sample CDL and adsorption of Th at
pattern as Th, but was in all cases less than 5%. Based on these pH 2 to 12. Complete desorption of Th only occurs at pH above 8; while loess dissolution
is minimized only at pH above 10 (b) Comparison of dissolution of loess CDL and rock
experiments, we found that a 0.0005 M EDTA solution at pH 12
standard BCR-2 at pH between 8 and 13.
resulted in minimal dissolution and maximum desorption.

2.2.3. Quantitative recovery of ice core components
Quantitative recovery of the components within an ice core was
tested with simulated ice cores using standard materials. 0.2 ng of
236
U and 229Th spikes in 1 M and 3 M HNO3 were put into two liter
Savillex beakers and neutralized with a 1 mL of 5 M NaOH, then the
beakers were half filled with milli-Q water and frozen. Aliquots of pre-
filtered standards BCR-2 and CDL were sprinkled onto two ice cubes,
while a third was left as a blank, before the beakers were filled and re-
frozen.
Ice samples were removed from the freezer and allowed to warm
at room temperature for approximately 10 min. After 10 min, the ice
could be easily removed from the Teflon beakers. The ice cylinders
were placed on specially designed PFA Teflon racks where they were
rinsed once with milli-Q water before being de-contaminated with
PFA chisels. After approximately 0.5 cm was removed from the outer
part of the cylinder, the ice was rinsed once more and placed in a clean
Teflon beaker to melt. Addition of 0.05 M EDTA at 30 minute intervals
maintained concentrations within the melt at approximately
0.0005 M.
Samples were filtered immediately after melting (4–6 h) with a
PTFE filtration unit equipped with PFA filters of 0.2 µm pore size. The
filtrate was collected and dried down in a clean nitrogen atmosphere.
The liquid residue, including EDTA, was treated with 2 mL 7.5 M HNO3
and 0.5 mL concentrated HF in a capped 15 mL beaker. The filter with
standard rock powder was placed in capped 60 mL Teflon beaker with
5 mL 7.5 M HNO3 and 0.5 mL concentrated HF. The filter solution was
decanted and the filter rinsed twice with 7.5 M HNO3 before the
combined rinses were dried down. Based on the measured 232Th/
229
Th and 238U/236U ratios of the liquids, there was zero dissolution of
BCR-2 and less than 0.01% of CDL. Total procedural blanks were
between 10 and 12 pg. Measurements of the 229Th/232Th and 236U/
238
U of the solids indicated approximately 0.5% adsorption of the Th
onto both standards and less than 0.1% adsorption of U. For the
purposes of using the U-series for dating ice core ice, the adsorption
behavior of U, and in particular 234U, is the most important.
2.3. Column chemistry
Splits of standard rock powders were dissolved in 50 mg batches
using a standard HNO3-HF digestion followed by drying in HCl and
dissolution in 9 M HCl with boric acid to complex any remaining
fluorides. Concentrated H2O2 (500 µL) was added to the dissolution of
CDL to oxidize organic material.
Aliquots equivalent to 0.25–5 mg of material were processed
through a TruSpec ion-exchange column (Fig. 2a). The Eichrom
TruSpec resin (50–100 mesh) was loaded in cleaned PFA columns with
a 500 μL reservoir and changed after every use. This first stage column
splits the sample into three cuts for subsequent chemical separation.
The first cut, containing the Sr, Pb and Ra fractions, was eluted in
1.75 mL 9 M HCl, the second cut, containing the High Field Strength
 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204 197

Fig. 2. (a) Histogram showing the elution scheme of the first (TruSpec) column for elemental separation. Columns were made of heat-shrink Teflon formed with 3 mm internal
diameter and 5 cm long with column volumes of 500 µL. Dashed line for Th indicates the difference in elution for the USGS standards; Th in loess samples shown in solid.
(b) Histogram showing the elution scheme of the first LnSpec column for the REE–Th–Hf separation. The columns have the same dimensions as those in (a).

Elements (HFSE) and the Rare Earth Elements (REE), was eluted in
2 mL 3 M HCl and 3.5 mL 0.25 M HCl and the last cut, containing the U
and Pa fractions, was eluted in 2 mL 0.1 M HCl–1 M HF. This separation
is adapted from Goldstein and Stirling (2003).
The first cut collected from the TruSpec elution was dried down
and brought into solution in 0.2 mL 3 M HNO3 before loading on a
SrSpec column. The Eichrom SrSpec resin (50–100 mesh) was loaded
in precleaned PFA columns with a 100 μL reservoir and changed after
every use. Ra was immediately eluted in the loading solution as well as
0.2 mL 3 M HNO3. After rinsing rubidium (Rb) in 2 mL 3 M HNO3 and
barium (Ba) in 3 mL 7.5 M HNO3, Sr was eluted in 3 mL 0.05 M HNO3
and Pb was eluted in 4 mL of 6 M HCl. This separation scheme is a
combination of traditional SrSpec chemistry for Sr (Horwitz et al.,
1992), Ra (Pietruszka et al., 2002), and Pb (Vajda et al., 1997). Yields
were checked by isotope dilution of BCR-2 and found to be N90% for Sr
and Ra and N97% for Pb. The Ra was further purified using a small
(0.4 mL) AG50 X 8 cation column; the Ra fraction was dried, re-
dissolved in 1.25 M HCl and loaded on the column. Ca, the remaining
impurity, was rinsed off in 4 mL of 2.5 M HCl prior to the pure Ra
elution in 5 mL of 6 M HCl.
The second eluted cut from the TruSpec, in ~ 1.1 M HCl, was dried
down to half volume, 3 mL, and treated with ~ 2.5 mL 11 M HCl, to
bring it to ~3 M HCl, then loaded directly on an Eichrom LnSpec
column (Fig. 2b). The Eichrom LnSpec (HDEHP) resin (100–200 mesh)
was loaded in pre-cleaned PFA columns with a 700 μL reservoir and
changed after every use. Based on the work of Munker et al. (2001),
198 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204
we developed a new method to separate the REEs, Th, and Hf. After
eluting the REEs in 8 mL 3 M HCl, ytterbium (Yb), lanthanum (La) and
titanium (Ti) were rinsed off the column in 8 mL of 6 M HCl followed
by a 0.09 M citric acid–0.4 M HNO3–1% H2O2 mixture and Th eluted in
4 mL 3 M HCl–0.1 M HF. Finally, after rinsing the zirconium (Zr), Hf
was eluted in 5 mL 6 M HCl–0.3 M HF. Yields were checked by isotope
dilution of AThO and BCR-2 and found to be N90% for Th and Hf.
The REEs, collected from the previous LnSpec column, were loaded
in 0.05 M HNO3 onto a second LnSpec column. The resin was loaded
onto a tall 2 mL column and compacted to ensure a better separation.
The columns were calibrated individually and thoroughly cleaned
with 6 M and 1.5 M HCl before re-use. After rinsing lanthanum (La),
cerium (Ce) and praseodymium (Pr) with 5 mL 0.25 M HCl, Nd was
eluted in 0.3 M HCl. Subsequently Sm was eluted in 1 M HCl. Final
yields, checked by isotope dilution of BCR-2, was N85% for Nd.
The third cut from the initial TruSpec column, containing the U and
Pa in 3 mL of 0.1 M HCl–1 M HF, was dried down to ~ 1 mL and further
acidified with 1 mL of 11 M HCl to bring the solution to ~ 5 M HCl–
0.5 M HF. This solution was then loaded directly onto a second pre-
cleaned TruSpec column (50–100 mesh, 100 μL). The higher HCl
normality permitted the fixation of U onto the resin, while the HF kept
Pa in solution. Pa was collected in the initial eluant plus 200 μL of 5 M
HCl–0.5 M HF and U collected in 0.1 M HCl–1 M HF. Yields for both U
and Pa, checked by isotope dilution of TML standard, were N95%.
All of the columns were initially calibrated using certified standard
solutions of known concentration containing both the trace elements
under investigation and the major elements that can negatively affect
elution chemistry and sample measurement. The final elution curves
were obtained using BCR-2 and CDL; the final elution curves for the
Ra–Ba and U–Pa separations were generated by doping the standard
materials with a known amount of 231Pa and 226Ra. Fig. 3 is a flow
chart showing the full chemical separation scheme. The purity of all
fractions was checked using a quadrupole ICP-MS to ensure that each
fraction was clean enough for direct analysis and free of high
concentrations of major or trace elements that would interfere with
analysis or optimization of instrument conditions.
Prior to measurement, all elements separated using the ion
exchange columns were treated with several drops of concentrated
HF, HNO3 and 100 µL of H2O2 to remove any organics (e.g. from the
resin) that could inhibit ionization or clog the nebulizer.
2.4. Mass spectrometry
2.4.1. MC-ICPMS
U, Th, Ra, Pa, and Hf were measured on a Nu Instruments MC-ICPMS
equipped with two ion counters and a retardation filter. The cup
configurations for the five elements measured by ICP-MS are shown in
Fig. 4. The small sample sizes coupled with the low concentrations of
trace elements in typical dust samples necessitate low blanks and
backgrounds from the mass spectrometer. A dedicated set of cones and
a 50 µl/min nebulizer for low abundance work were cleaned prior to
every session; cones were soaked in 10% RBS soap solution before
hand-polishing and the nebulizer was rinsed overnight with 10% HCl–
5% HF. Shared lab torches and spray chambers were cleaned overnight
in 3 M HNO3.
Some modifications to the Nu Plasma standard layout allowed us to
increase our sensitivity significantly while maintaining beam stability.
Nu cones have three designs, an original and two experimental, AB
and WA. The cone configuration of AB sampler and WA skimmer
increased the sensitivity by 30–50%. The drawback of this configura-
tion is that the abundance sensitivity increases from 0.1 ppm to 6 ppm,
which is a major drawback for measuring thorium isotopes. As a
consequence, we used the traditional AB–AB cones for Th and Pa. Two
other variables can be adjusted on the Nu Plasma to increase
sensitivity: the introduction system, aspiration and desolvation, and
the alignment of the fixed lenses can be optimized. An increase in the
sensitivity via the introduction system was attained by using a
nebulizer with a 50 µL/min uptake rate; the lower uptake rate
decreases the overall drop size as well as the range in drop sizes,
thereby increasing the transmission efficiency. By controlling the
nebulizer uptake rate with an external high precision gas flow
controller made by Voegtlin of Switzerland, which maintained
nebulizer gas flow rates to better than 0.1%, the beam stability was
optimized with sensitivity. In order to decrease clipping of the beam
and over-steering through the lenses, a series of re-alignment tools
was made. These tools provided mechanical guides for re-assembly of
the front end of the Nu Plasma with perfect aperture alignment from
sampler cone to line of sight valve. After these adjustments, the
steering voltages dropped below 20 V for the transfer lenses, and the
sensitivity increased by another 10–20%. These small incremental
increases resulted in typical sensitivities between 0.35% and 1.2% for
beam sizes between 200 and 450 V/ppm for Th, U, Ra, Pa, and Hf;
higher than for TIMS measurements, with the exception of Ra.
Uranium was measured statically with 234U collected in ion counter
IC1. Each sample was bracketed with a solution of SRM960, diluted to a
similar concentration as the samples (2–5 ppb, equivalent to 1 ng of
sample uranium). Initially, we had planned to measure the ion counter
relative gains by peak jumping between a Faraday cup and the ion
counter with a 235U beam, but found that the difference in the efficiency
between 10− 2 and 10− 5 V beams to be an additional source of
uncertainty; instead the 238U/234U ratio (18,918; Cheng et al., 2000) of
the bracketing SRM960 was used to calculate the ion counter efficiency.
Uncertainties in the linearity and stability of the ion counter over a range
of ion beam intensities on MC-ICPMS instruments have been noted by
other workers (Richter et al., 2001; Hellstrom, 2003; Ball et al., 2008). At
the low count rates (500–1500 cps) found in such low concentration
samples, it is difficult to apply a correction that accurately accounts for the
count rate dependence of the ion counter gain, therefore concentrations
between standards and samples were matched as best as possible. Ratios
were corrected for mass bias by normalizing to 238U/235U =137.88 using
the exponential law; 238U/236U ratios of the samples were corrected for
mass bias using the 238U/235U ratio of the bracketing standard.
Measurements were acquired in 30 cycles of 8 second integrations after
30 s of baseline measurements at the half masses. Baseline measure-
ments perform the primary function of accounting for noise in the
Faraday cups and background signal originating in the source (desolvat-
ing unit and plasma). In the case of U-series measurements, baseline
measurements must also account for the mass tailing of large ion beams
(e.g. 238U, 232Th) into the ion counters, therefore half mass log-weighted
baselines are optimal (Hellstrom, 2003). For U-series results reported
here, on-peak wash solution blanks were run prior to each measurement
and sample measurements were not made unless blank values on all
isotopes were at Faraday/ion counter ESA deflected baseline values.
Thorium was also measured statically, with 229Th in Faraday cup
L5, and 230Th in the ion counter IC1, equipped with a retardation filter.
Samples were bracketed with a mixed Th105-SRM960 solution to
provide the combined mass bias–gain correction for 232Th/230Th and
the 238U–235U mass bias correction for 232Th/229Th ratios. Concentra-
tions of Th105 were equivalent to 2 ng of material (4 ppb solution),
while the SRM960 was more concentrated (10 ppb) for accurate 238U–
235
U measurements for the mass bias correction. Similar to the
uranium measurements, low count rates (100–500 cps) on the ion
counter necessitated matching the concentration of the samples and
standards as close as possible to avoid non-linear variations in the gain
with count rate. Prior to every measurement, a 100 point peak scan
was made from 229.5 to 230.5 and an exponential fit applied to the
mass scan to subtract the 232Th tail (Hubert et al., 2006; Ball et al.,
2008). Baselines were measured prior to measurement at the half
mass, but only used for correcting the 238U/235U and 232Th/229Th
ratios. Data was acquired in 30 cycles with 8 second integration times.
Protactinium was measured dynamically with simultaneous 233Pa–
231
Pa measurement. The 233Pa beam was collected both in IC1 and IC2
 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204 199

Fig. 3. Summary of the chemical separation scheme for clean cuts of Ra, Sr, Pb, Th, Hf, Nd, U, and Pa.

to account for differences in ion counting efficiencies. Protactinium
was doped with SRM960 in order to correct for mass bias using an
internal 238U–235U correction. Prior to data collection, a 20 second
baseline was measured at half masses. Samples were measured with
5 second integration times until the solution was exhausted (15 to
45 cycles). Typical sample sizes for yields and spike calibrations
ranged from 50 fg to 200 fg of 233Pa and 50 fg to 100 fg of 231Pa.
Radium was measured dynamically with simultaneous 228Ra–226Ra
measurement and the 228Ra beam was collected both in IC1 and IC2 to
account for differences in ion counting efficiencies. Radium measure-
ments were bracketed with SRM960 standard to account for mass bias
using an external 238U–235U correction. Prior to data collection, a
20 second baseline was collected at half masses. Samples were
measured with 5 second integration times until the solution was

Fig. 4. Cup configurations for Nu Instruments (Nu008) MC-ICPMS measurements. U, Th and Hf were static measurements, while Ra and Pa were measured dynamically to account for
differences in ion counting efficiencies in IC1 and IC2.
200 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204

exhausted (15 to 45 cycles). Sample size for the TML standard as
reported in Table 1 was 10 fg of 226Ra spiked with 80 fg of 228Ra,
although yields were measured on 100 fg of 226Ra to ensure accuracy.
Hafnium was measured statically with the isotopes collected in the
full range of Faraday cups. Monitoring of the 182W, 175Lu, and 172Yb
beams allowed us to correct for interference at masses 174 and 176.
Hafnium was normalized using an internal exponential law correction,
with 179Hf/177Hf = 0.7325. Corrections for Lu and Yb interferences were
made using the natural abundances (175Lu= 0.97416, 176Lu= 0.02584,
172
Yb = 0.2183, 174Yb = 0.3138, 176Yb = 0.1276; Rosman and Taylor,
1998). Interference corrections to 176Hf/177Hf were less than 100 ppm.
Bracketing measurements of the hafnium standard JMC475 were run
between every sample with a 30 cycle, 5 second integration routine. Hf
compositions of the samples were measured using the time resolved
(TR) protocol provided in the Nu Plasma software package for laser
analyses. Using this protocol, the magnet was set using the peak center
of the previously measured standard and background zeroes measured
prior to sample introduction. After an on-peak baseline was measured,
the sample was aspirated and a 1–3 minute continual measurement was
made. This method of analysis lowers the amount of material required,
thereby making measurements on ~1 ng of Hf possible. One of the
sources of error when using the TR protocol for laser ablation can be
large variations in beam intensity and an ineffective tau correction,
which accounts for the signal decay time on the Faraday amplifiers.
When using the TR protocol for solutions, this effect is negligible because
the beam is stable. The precision and reproducibility using the TR
protocol are comparable to the standard protocol for high concentrated
uranium measurements (Andersen et al., 2004); the results presented
here show the precision and reproducibility using the TR protocol for Hf
isotopes.
For Hf measurements, baselines were measured on-peak while the
beam was deflected by ESA. Baseline measurements made in this
manner are an accurate representation of the stability and reprodu-
cibility of the Faraday cup “noise”, and therefore is the default method
for the Nu Plasma (Belshaw et al., 1998). However, other protocols
exist to measure baselines: half peak, as described earlier for U-series
measurements where mass tailing can be significant, and on-peak
without ESA deflection in a wash solution prior to sample aspiration.
The latter protocol is usually implemented to correct for memory of
the desolvating unit (e.g. the blank of the wash solution), or
interferences (e.g. 204Hg for 204Pb measurements; Simonetti et al.,
2005). A combined acid background and baseline Faraday cup noise
value has dubious advantages because memory effects are not
constant and the sample solution may behave differently than the
wash solution due to matrix effects in the plasma. In order to evaluate
the two methods, we measured the USGS standard BCR-2 with and
without the ESA deflection and found no discernible difference.
Therefore, we measured both solution and rock standards using ESA
deflection for the baseline, evaluating our blanks by isotope dilution
and memory effects by the reproducibility and accuracy of the
bracketing standard, JMC475.
2.4.2. Sr and Nd measurements by TIMS
Strontium and neodymium were measured on the ETH Thermo
Triton mass spectrometer. Both elements were loaded onto rhenium
filaments that had been previously outgassed for 2 h at 4 A and 10 min
at 5 A. Solution standards of SRM987 for Sr and JNdi-1 for Nd were
measured during the same TIMS runs as the samples.
The Sr fractions were dissolved in 7.5 M HNO3 to yield a 10 ppm
solution (1–50 µL). 1 µL (10 ng) of sample was mixed with 1 µL TaCl5
activator, the sample and activator were loaded onto a filament while
heating with a current of approximately 600 mA. Once the sample was
dry, the filament was heated up slowly to ~ 2 A for 10 s. For half of our
samples we added an additional step to the loading procedure: after
heating to 2 A, we added 0.1 µL of TaCl5 and dried down at 0.5 A. This
last step helped stabilize the sample onto the filament. Prior to
measurement, while the filament was warming up, a baseline was
measured for 250 s. The filaments were heated slowly by increasing the

Table 1
Compilation of standard data for 0.25–5 mg sized aliquots.

Standards ATHO TML AGV-2 BCR-2 BHVO-2 CDL

U, ppm 2.24 ± 0.02 10.99 ± 0.06 1.72 ± 0.01 2.22 ± 0.06
nt; ns 28; 5 6; 2 3; 2 3; 3
Literature1 2.24 10.52
234
U/238U activity 1.001 ± 0.013 0.997 ± 0.008 1.006 ± 0.006 0.912 ± 0.016
nt; ns 91; 13 12; 2 15; 4 18; 6
Literature2 1.00 1.00
Th, ppm 7.32 ± 0.08 32.07 ± 0.72 5.94 ± 0.03 4.95 ± 0.05
nt; ns 28; 5 12; 3 3; 2 3; 3
Literature1 7.4 29.74
232
Th/230Th activity 183477 ± 2450 176540 ± 1976 209961 ± 3993 165123 ± 3860
nt; ns 64; 13 6; 2 15; 4 18; 6
Literature1 182150 173010
230
Th/238U activity 1.09 ± 0.03 1.01 ± 0.02 1.01 ± 0.02 0.83 ± 0.03
226
Ra, fg/g 3916 ± 342
nt; ns 6; 1
Literature1 3600
226
Ra/238U activity 1.06 ± 0.10
176
Hf/177Hf 0.282995 ± 0.000044 0.282867 ± 0.000043 0.283101 ± 0.000050 0.282630 ± 0.000037
nt; ns 13; 2 27; 6 14; 2 18; 4
εHf 7.99 ± 1.54 3.48 ± 1.51 11.73 ± 1.78 −4.92 ± 1.31
Literature3 7.60 3.57 11.88 0.512099 ± 0.000023
143
Nd/144Nd 0.512770± 0.000022 0.512643 ± 0.000037 0.512996 ± 0.000026 6; 4
nt; ns 2; 2 11; 3 2; 2 − 10.52 ± 0.37
εNd 2.58 ± 0.43 − 0.09 ± 0.81 6.98 ± 0.50 0.714040 ± 0.00338
Literature4,5 2.97 − 0.08 6.79 12; 4
87
Sr/86Sr 0.703986 ± 0.000052 0.705018 ± 0.000027 0.703482 ± 0.000052
nt; ns 6; 2 14; 4 6; 2
Literature4,5 0.703981 0.705013 0.703497

Errors are reported as 2SD of the mean. nt is the total number of measurements while ns is the number of individual solutions (dissolutions). Note, these are secondary standards,
primary standard solutions gave the following values: 5 ppb SRM960, mean = 23,356, 2SD = 0.75%; 5 ppb Th105, mean = 269,542, 2SD = 2.5%; 15 ppb JMC475, mean = 0.282155,
2SD = 0.000029; 1 ng JNdi, mean = 0.512102, 2SD = 0.000042; 10 ng SRM987, original mean = 0.710253, 2SD = 0.000106, with modified procedure mean = 0.710255,
2SD = 0.000032. References for literature values: (1) Pietruska et al., 2002 (2) Ball et al., 2008 (3) Weis et al., 2007 (4) Weis et al., 2006 (5) Weis et al., 2005.
 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204 201

current at 200 mA/min with the analyzer gate closed until the filament
temperature reached 1300 °C. The filament was then heated until a
10 mV 88Sr beam was detected, a first focusing was done and the
strontium peak centered before further heating to obtain a beam of
≈100 mV. Initial measurements acquired data at this low beam
intensity for 10 blocks of 20 cycles with 8 second integrations, resulting
in internal and external errors of N200 ppm. Better internal (10 ppm)
and external (50 ppm) errors were achieved by continuing to heat the
filament to reach 5–6 V 88Sr prior to data acquisition for 10 blocks of
20 cycles with 4 second integrations. The modified loading procedure
and manual heating of the filaments between blocks was necessary to
maintain the higher beam intensity. An internal mass fractionation
correction was applied to 87Sr/86Sr using the exponential law and the
isotope ratios were obtained by normalizing to 88Sr/86Sr = 8.375209.
Rubidium interferences were corrected by monitoring 85Rb in the
center cup. Rubidium beam sizes were less than 10− 13 A, such that
corrections did not exceed 100 ppm.
Neodymium was loaded onto double Re zone refined filaments and
run as a metal. Samples were dried down and 1 M HCl–1 M HNO3
added to yield a 1 ppm solution. 1 µL of sample (1 ng Nd) and 0.1 µL of
0.1 M H3PO4 were mixed and loaded drop wise onto the evaporation
filament while heating with a current of 0.8 A. After the filament was
fully dry, the filament current was increased up to 1.8 A to burn off the
phosphoric acid, then brought up to 2.2 A for 10 s. For measurement,
the ionization filament was warmed up first to 1500 °C at 200 mA/
min, then evaporation filament at 100 mA/min to 1.8 A while bringing
the ionization filament to 1650 °C. While warming, Faraday cup
baselines were measured for 300 s. Once the evaporation filament was
at 1.8 A and the temperature measured by the pyrometer was at
1650 °C, the analyzer valve was opened and manual focus begun. The
filament was slowly warmed up to 2.2 A while focusing the ion beam.
Once 144Nd was at 50 mV, measurements were started, with
additional heating between blocks to obtain a signal of 100 mV on
144
Nd. Data was acquired in 4 second integrations for at least
200 cycles. 143Nd/144Nd was normalized to 146Nd/144Nd to 0.7219
using the exponential law and mass 150 was monitored for Sm
interferences; no Sm was ever detected. One of the drawbacks of
measuring Nd as a metal is that the ionization efficiency is not as high
as with oxide measurements (e.g. 3% instead of 5%; Caro et al., 2008).
However, we found that the inherent uncertainty in oxygen
composition when measuring NdO+ (Thirlwall, 1991; Griselin et al.,
2001) and the difficulties in achieving perfect separations (Pr, La, Ce)
would be more problematic to overcome than the slight difference in
external errors that a lower sensitivity would entail.
3. Isotopic measurements: reproducibility of standards, spike
correction and error analysis
U, Hf, Sr and Nd isotopic results are shown in Fig. 5, with the
compilation of data, including Th and Ra, with external reproducibility
in Table 1. Five rock standards and one internal lab standard of loess
were used to evaluate precision and accuracy. The use of rock

Fig. 5. Reproducibility of rock standards. Individual measurements of (a) 143Nd/144Nd, (b) 176Hf/177Hf, and (c) 87Sr/86Sr of BCR-2, and (d) 234U/238U activity ratio of AthO. Literature
values for standards are marked by solid lines (a. Weis et al., 2005, 2006; b. Weis et al., 2007; c. Weis et al., 2005, 2006; d. Ball et al., 2008), while mean values of this data set are
marked by the dashed lines (see Table 1 for values). For (a,b,c) different symbol shapes indicate individual dissolutions and each data point is an individual measurement. For
(d) light gray symbol denote unspiked samples, dark gray symbols indicate spiked samples with a spike subtraction applied to correct the 238U/234U; each data point and error is
based on the mean and standard deviation of standards for the session (1–3 days).
202 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204
standards serves the dual purpose of evaluating the chemical
separation methods as well as the reproducibility of samples on the
mass spectrometers. Precision of these measurements was estimated
based on the long term running means and standard deviations of the
standards, at the same beam sizes (intensities) applicable to small
(1 mg) dust samples.
We evaluated three rock standards (ATHO, TML, and BCR2) to
examine the U-series reproducibility and the efficacy of our spike
corrections. The 2σ external reproducibility obtained for 234U/238U
with ATHO and BCR were 1.3% and 0.6% on 1 ppb solutions, equivalent
to 1 ng of sample material. The 2σ external reproducibility obtained for
230
Th/232Th with ATHO and BCR is 1.3% and 1% based on 1.5 ppb and
3 ppb solutions, equivalent to 1 and 3 ng of sample material measured
over the course of approximately two years. Standard runs at higher
beam sizes have equivalently lower errors comparable to previously
published work (Hellstrom, 2003; Andersen et al., 2004): 2σ SD of 0.1%
(234U/238U) and 0.45% (230Th/232Th) for 30 ppb solutions.
Our measured values are in good agreement with published values.
The accuracy of the measurements can be directly derived by
considering the raw ratios (234U/238U), and indirectly, by the activity
ratios of 230Th/238U and 226Ra/230Th. U-series activity ratios of the
standard TML are well-known to be in secular equilibrium (Luo et al.,
1997; Rubin, 2001; Pietruszka et al., 2002; Ball et al., 2008). Both 234U/
238
U and 230Th/238U have low errors and for TML show activity ratios of 1,
within error. We have used the same chemical separation with spiked
and unspiked aliquots of ATHO and BCR in order to evaluate the accuracy
of the spike subtraction, these results are also shown in Fig. 5d. There is a
slight difference in the average (0.1%, upper line), but well within 2 SD of
the two means.
The radium and protactinium present a different problem. For the
purpose of this work, we only show the elution curves and yields
using doped samples because the amount of Ra and Pa in 1 mg of dust
is too low to measure using existing techniques (b0.25 fg). However,
Goldstein et al. (2004) found 30–200 mg of material in marginal ice
sheet ice from Antarctica. Therefore, there is the possibility that Ra
and Pa would be high enough to measure in some special samples. In
order to evaluate the errors in measuring Ra and Pa, we chose to
optimize our detection limits and yields on small sample sizes. For Ra,
1 mL solutions of TML running at 10 fg/mL yielded 2σ SD errors of 9.5%
and within error of the literature value (Pietruszka et al., 2002). In the
absence of a second rock standard with a well-known 231Pa
concentration, we were only able to calibrate the spike. Errors in the
spike calibration based on external reproducibility of the spiked TML
are b2% for 1 mL aliquots of 100 fg/mL solutions.
Hafnium isotopic results are shown in Fig. 5b. Recent work (Weis
et al., 2007) provides good characterization of “true” values and all of
our measured standards agree with those published values within
error. All of the rock standard data shown were measured using the
time resolved (TR) method, so are an accurate reflection of running a
limited amount of sample with small beam size. External 2σ SD errors
of the daily JMC475 means are only slightly larger than in run 2σ SE,
typically 0.6 ε units. 2σ SD errors on the rock standards are all
~ 40 ppm for solutions equivalent to 0.5 to 1.5 ng of Hf. These errors are
comparable to the 2σ SD 46 ppm external reproducibility measured by
(Weis et al., 2007).
Comparison of Sr isotopic values of the rock standards to other
published values indicates minor offsets, but they all fall within the 2σ
SD uncertainties (Raczek et al., 2003; Weis et al., 2005, 2006). The in-
run precision for each individual measurement is less than 10 ppm,
while the external reproducibility is larger (50 ppm). The original
loading and running technique limited our beam size to 1–2 V and our
external errors to ~ 200 ppm for BCR-2 and for SRM987, while with a
modified loading procedure to stabilize the sample on the filament the
beams reached 5–6 V and errors were reduced to 50 ppm (Fig. 5c:
measurements 7–14). Loess standard CDL shows a higher variability in
87
Sr/86Sr than the other standards, likely due to the heterogeneity of
the material. This variability is found both in the Sr and U isotopic
composition but not in Hf and Nd, likely indicating the source of the
heterogeneity is varying concentrations of carbonate (e.g. Honda et al.,
2004).
Neodymium results all show 2σ SD errors between 40 and 80 ppm,
and all are in good agreement with the published data (Raczek et al.,
2003; Weis et al., 2005, 2006). The long term 2σ SD of JNdi-1 is the
same (80 ppm) as the most frequently run rock standard, BCR-2.
4. Comparison with other methods
Ice processing procedures vary largely among research groups and
depend on the isotope of interest. Bulk melting results in measure-
ments of Pb, Sr, Nd, and trace element concentrations that are
combinations of dust and sea salt (Basile et al., 1997; Biscaye et al.,
1997; Hinkley and Matsumoto, 2001; Burton et al., 2003, 2007). At a
minimum, bulk measurements of Sr isotopes could provide dubious
interpretations of provenance, since as much as 40% of the Sr could be
of sea salt origin (Hinkley and Matsumoto, 2001; Banta et al., 2008).
Separation by centrifuging and decanting without filtration (Del-
monte et al., 2004) may leave some sea salt components admixed
with the dust fraction. Filtration to 0.4 µm or centrifuging, filtration
and acetic acid leaching without initially buffering the melting ice
would likely partially dissolve carbonate in the sea salt fraction
removal (Bory et al., 2003; Delmonte et al., 2008). In all cases, strongly
adsorbed elements (Th, Hf) would partially remain in the dust
fraction, as we have shown in Section 2.2 for Th. In contrast, Goldstein
et al. (2004), extracted soluble and adsorbed materials with EDTA and
filtration. However, the possibility of incomplete dissolution of
carbonate was not addressed. In this work, we attempted to
distinguish between the total sea salt and non sea salt fractions by
effectively preventing carbonate dissolution prior to filtration.
Most of the individual chromatographic separations have been
described elsewhere; here we have combined differing methods to
allow clean separations of many elements from a single split. This
method is therefore particularly adapted for small sample sizes. This
includes a more complete elution scheme for the LnSpec for HFSE to
separate Th from the REEs and Hf (as well as Zr). Also, we find that
combining Ra chemistry with Sr–Pb chemistry is possible. This
scheme could be valuable for other hydrology studies that incorporate
traditional tracers with U-series measurements. Our aggressive
chemical cleaning procedures for the lab ware, use of ultra-pure
reagents, careful decontamination of ice, and use of Teflon filters has
resulted in total procedural blanks that are an order of magnitude
better than other published values in ice samples for U, Th, (10 pg)
(Goldstein et al., 2004) and Nd (1 pg) (Delmonte et al., 2008), similar
to published values for Ra and Pa (0.05 fg) (Goldstein et al., 2004), but
worse than others for Sr (150 pg) (Burton et al., 2002).
The techniques for measuring isotopic systems by MC-ICPMS and
TIMS are similar to those previously published with quantification of
errors on small samples. Not just internal errors, but true accounting
for reproducibility, which is particularly important for making age
calculations. U-series measurements are improved by stabilizing and
maximizing beams such that internal and external errors are
improved. This includes optimizing nebulizer running conditions so
that uptake rates (e.g. aerosol size) are low enough to boost sensitivity
while being high enough that the amount of solute is larger than a
single drop. The use of a high precision gas flow controller for the
nebulizer kept uptake rates and beam intensities stable. Alignment of
the lens apertures also increases sensitivity by increasing the
proportion of the beam that actually reaches the detectors. And for
both U and Hf, a modified cone configuration increased the sensitivity
by 50%. In comparison to other methods based on TIMS, we have
similar errors on sample sizes 1–2 orders of magnitude lower for U
and Th, (Goldstein et al., 2004) and comparable results for Ra and Pa.
For Sr, our initial errors were quite high (~200 ppm), especially
 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204
compared to other studies using the Thermo Triton with similar
amounts (Font et al., 2007). By modifying the loading procedure and
running at higher beam sizes, we brought the standard deviation into
the range given in Font et al. (2007), and other ice core work
(Delmonte et al., 2008). The difference in ion yields is not that
significant for Nd measurements on low level samples as a metal or an
oxide. The decision to measure the Nd as a metal was made because of
the possible variations in oxygen isotope composition and the
additional loss of Nd for complete removal of Pr and Ce. Our external
errors are as good as other published data for similar sample sizes: 2σ
SD for JNdi-1 and BCR are 80 ppm compared to 50 ppm on 1 ng sample
sizes using traditional NdO+ measurements (Delmonte et al., 2008)
and 42 ppm on 2 ng samples sizes using a modified loading scheme for
NdO+ measurements (Harvey and Baxter, 2009). The low errors
reported here for 143Nd/144Nd measured as a metal are in agreement
with the work of Caro et al. (2008) who achieved precision of 20 ppm
2σ SD for 10 ng load sizes on a Thermo Triton.
5. Conclusions
In this work we designed a method for melting and filtering ice cores
that limits the co-mingling of the insoluble and soluble components.
Separating sea salt from non-seawater carbonate and dust would allow
more direct measurements of paleo-seawater and paleo-dust. Measure-
ment of the quantities of material found in dust and sea salt, on the order
of 1 ng for trace elements, requires analytical techniques that can achieve
perfect separation and yields while keeping blanks below 10 pg. The
analytical techniques for sequential separation of trace elements required
to measure the full range of isotope tracers (U-series, Sr, Nd, Hf, Pb) are
described. The sequence of columns minimizes the steps required for
complete separation, thus reducing acid volumes and blanks. The MC-
ICPMS and MC-TIMS methods we have employed for U, Th, Ra, Hf, Sr and
Nd result in errors in the range of 1–2%, 3%, 9.5%, 100 ppm, 50 ppm, and
80 ppm respectively. The low errors for MC-ICPMS work are largely due to
the boost in sensitivity from 0.2% to 1% with optimized running conditions.
Given these errors, it is possible to continue to measure traditional ice core
tracers, Sr and Nd, while adding a suite of new isotopic systems to trace
and quantify timescales of ice dynamics.
Acknowledgements
The authors would like to thank the Swiss National Science
Foundation for providing funding. Felix Oberli and Andreas Stracke
provided helpful discussions and suggestions throughout the project.
We thank two anonymous reviewers, and the editor for their helpful
suggestions.
References
Andersen, M., Stirling, C., Potter, E., Halliday, A., 2004. Toward epsilon levels of
measurement precision on 234U/238U by using MC-ICPMS. International Journal of
Mass Spectrometry 237, 107–118.
Ball, L., Sims, K., Schwieters, J., 2008. Measurement of 234U/238U and 230Th/232Th in
volcanic rocks using the Neptune MC-ICP-MS. Journal of Analytical Atomic
Spectrometry 231, 173–180.
Banta, J., McConnell, J., Edwards, R., Engelbrecht, J., 2008. Delineation of carbonate dust,
aluminous dust, and sea salt deposition in a Greenland glaciochemical array using
positive matrix factorization. Geochemistry Geophysics Geosystems 9.
Basile, I., Grousset, F., Revel, M., Petit, J., Biscaye, P., Barkov, N., 1997. Patagonian origin of
glacial dust deposited in East Antarctica (Vostok and Dome C) during glacial stages
2, 4 and 6. Earth and Planetary Science Letters 146 (3–4), 573–589.
Belshaw, N.S., Freedman, P.A., O'Nions, R.K., Frank, M., Guo, Y., 1998. A new variable
dispersion double-focusing plasma mass spectrometer with performance illustrated
for Pb isotopes. International Journal of Mass Spectrometry 181 (1–3), 51–58.
Biscaye, P., Grousset, F., Revel, M., VanderGaast, S., Zielinski, G., Vaars, A., Kukla, G., 1997.
Asian provenance of glacial dust (stage 2) in the Greenland Ice Sheet Project 2 Ice
Core, Summit, Greenland. Journal of Geophysical Research-Oceans 102 (C12),
26765–26781.
Bory, A., Biscaye, P., Piotrowski, A., Steffensen, J., 2003. Regional variability of ice core
dust composition and provenance in Greenland. Geochemistry Geophysics
Geosystems 4.
203
Bourdon, B., Joron, J., Allegre, C., 1999. A method for Pa-231 analysis by thermal
ionization mass spectrometry in silicate rocks. Chemical Geology 157 (1–2),
147–151.
Burton, G., Morgan, V., Boutron, C., Rosman, K., 2002. High-sensitivity measurements of
strontium isotopes in polar ice. Analytica Chimica Acta 469 (2), 225–233.
Burton, G., Boutron, C., Hong, S., Candelone, J., Rosman, K., 2003. Strontium isotope
measurements in Greenland ice from the last glacial maximum to the early
Holocene. Journal de Physique IV 107, 241–244.
Burton, G., Rosman, K., Candelone, J., Burn, L., Boutron, C., Hong, S., 2007. The impact of
climatic conditions on Pb and Sr isotopic ratios found in Greenland ice, 7–150 ky BP.
Earth and Planetary Science Letters 259 (3–4), 557–566.
Caro, G., Bennett, V., Bourdon, B., Harrison, T., von Quadt, A., Mojzsis, S., Harris, J., 2008.
Application of precise 142Nd/144Nd analysis of small samples to inclusions in
diamonds (Finsch, South Africa) and Hadean zircons (Jack Hills, Western Australia).
Chemical Geology 247 (1–2), 253–265.
Cheng, H., Edwards, R., Hoff, J., Gallup, C., Richards, D., Asmerom, Y., 2000. The half-lives
of uranium-234 and thorium-230. Chemical Geology 169 (1–2), 17–33.
Delmonte, B., Basile-Doelsch, I., Petit, J., Maggi, V., Revel-Rolland, M., Michard, A.,
Jagoutz, E., Grousset, F., 2004. Comparing the Epica and Vostok dust records during
the last 220,000 years: stratigraphical correlation and provenance in glacial
periods. Earth-Science Reviews 66 (1–2), 63–87.
Delmonte, B., Andersson, P., Hansson, M., Schoberg, H., Petit, J., Basile-Doelsch, I., Maggi, V.,
2008. Aeolian dust in East Antarctica (EPICA-Dome C and Vostok): provenance during
glacial ages over the last 800 kyr. Geophysical Research Letters 35 (7).
Di Palma, L., Mecozzi, R., 2007. Heavy metals mobilization from harbour sediments
using EDTA and citric acid as chelating agents. Journal of Hazardous Materials 147
(3), 768–775.
Fireman, E., 1986. Uranium series dating of Allan Hills Ice. Journal of Geophysical
Research-Solid Earth and Planets 91 (B4), D539–D544.
Fischer, H., Siggaard-Andersen, M., Ruth, U., Rothlisberger, R., Wolff, E., 2007. Glacial/
interglacial changes in mineral dust and sea-salt records in polar ice cores: sources,
transport, and deposition. Reviews of Geophysics 45 (1).
Font, L., Nowell, G., Pearson, D., Ottley, C., Willis, S., 2007. Sr isotope analysis of bird
feathers by TIMS: a tool to trace bird migration paths and breeding sites. Journal of
Analytical Atomic Spectrometry 22 (5), 513–522.
Fredd, C., Fogler, H., 1998. The influence of chelating agents on the kinetics of calcite
dissolution. Journal of Colloids and Interface Science 204 (187–197).
Fritz, J., Umbreit, G., 1958. Ion exchange separation of metal ions. Analytica Chimica Acta
19, 509–516.
Goldstein, S., Stirling, C., 2003. Techniques for measuring uranium-series nuclides.
Uranium-Series Geochemistry 23–57.
Goldstein, S., Murrell, M., Nishiizumi, K., Nunn, A., 2004. Uranium-series chronology and
cosmogenic Be-10-Cl-36 record of Antarctic ice. Chemical Geology 204 (1–2),
125–143.
Griselin, M., van Belle, J.C., Pomiès, C., Vroon, P.Z., van Soest, M.C., Davies, G.R., 2001. An
improved chromatographic separation technique of Nd with application to NdO+
isotope analysis. Chemical Geology 172 (3–4), 347–359.
Harvey, J., Baxter, E.F., 2009. An improved method for TIMS high precision neodymium
isotope analysis of very small aliquots (1–10 ng). Chemical Geology 258 (3–4),
251–257.
Hellstrom, J., 2003. Rapid and accurate U/Th dating using parallel ion-counting multi-
collector ICP-MS. Journal of Analytical Atomic Spectrometry 18, 1346–1351.
Henderson, G., Hall, B., Smith, A., Robinson, L., 2006. Control on (U-234/U-238) in lake
water: a study in the Dry Valleys of Antarctica. Chemical Geology 226 (3–4),
298–308.
Hinkley, T., Matsumoto, A., 2001. Atmospheric regime of dust and salt through
75,000 years of Taylor Dome ice core: refinement by measurement of major, minor,
and trace metal suites. Journal of Geophysical Research-Atmospheres 106 (D16),
18487–18493.
Honda, M., Yabuki, S., Shimizu, H., 2004. Geochemical and isotopic studies of aeolian
sediments in China. Sedimentology 51 (2), 211–230.
Horwitz, E., Chiarizia, R., Dietz, M., 1992. A novel strontium-selective extraction
chromatographic resin. Solvent Extraction and Ion Exchange 10 (2), 313–336.
Hubert, A., Bourdon, B., Pili, E., Meynadier, L., 2006. Transport of radionuclides in an
unconfined chalk aquifer inferred from U-series disequilibria. Geochimica et
Cosmochimica Acta 70 (22), 5437–5454.
Luo, X., Rehkamper, M., Lee, D., Halliday, A., 1997. High precision Th-230/Th-232 and U-
234/U-238 measurements using energy-filtered ICP magnetic sector multiple
collector mass spectrometry. International Journal of Mass Spectrometry 171 (1–3),
105–117.
Mitchell, J., Herring, C., Bylina, E., 1984. Purification of EDTA monitored by inductively
coupled plasma atomic emission spectroscopy (ICP-AES). Applied Spectroscopy 38
(5), 653–661.
Munker, C., Weyer, S., Scherer, E., Mezger, K., 2001. Separation of high field strength
elements (Nb, Ta, Zr, Hf) and Lu from rock samples for MC-ICPMS measurements.
Geochemistry Geophysics Geosystems 2.
Nelson, F., 1964. Ion exchange procedures V. Separation of barium and radium. Journal
of Chromatography 16, 403–406.
Pettke, T., 2002. Radiogenic Hf isotopic compositions of continental eolian dust from
Asia, its variability and its implications for seawater Hf. Earth and Planetary Science
Letters 202 (2), 453–464.
Pietruszka, A., Carlson, R., Hauri, E., 2002. Precise and accurate measurement of Ra-226–
Th-230–U-238 disequilibria in volcanic rocks using plasma ionization multi-
collector mass spectrometry. Chemical Geology 188 (3–4), 171–191.
Raczek, I., Jochum, K., Hofmann, A., 2003. Neodymium and strontium isotope data for
USGS reference materials BCR-1, BCR-2, BHVO-1, BHVO-2, AGV-1, AGV-2F GSP-1,
204 S.M. Aciego et al. / Chemical Geology 266 (2009) 194–204
GSP-2 and eight MPI-DING reference glasses. Geostandards Newsletter—the Journal
of Geostandards and Geoanalysis 27 (2), 173–179.
Richter, S., Goldberg, S.A., Mason, P.B., Traina, A.J., Schwieters, J.B., 2001. Linearity tests
for secondary electron multipliers used in isotope ratio mass spectrometry.
International Journal of Mass Spectrometry 206 (1–2), 105–127.
Rihs, S., Sturchio, N., Orlandini, K., Cheng, L., Teng, H., Fenter, P., Bedzyk, M., 2004.
Interaction of uranyl with calcite in the presence of EDTA. Environmental Science &
Technology 38, 5078–5086.
Rosman, K.J.R., Taylor, P.D.P., 1998. Isotopic compositions of the elements 1997
(technical report). Pure and Applied Chemistry 70 (1), 217–235.
Rubin, K., 2001. Analysis of 232Th/230Th in volcanic rocks: a comparison of thermal
ionization mass spectrometry and other methodologies. Chemical Geology 175,
723–750.
Sala, M., et al., 2008. Evidence of calcium carbonates in coastal (Talos Dome and Ross
Sea area) East Antarctica snow and firn: environmental and climatic implications.
Earth and Planetary Science Letters 271 (1–4), 43–52.
Simonetti, A., Heaman, L.M., Hartlaub, R.P., Creaser, R.A., MacHattie, T.G., Böhm, C., 2005.
U–Pb zircon dating by laser ablation-MC-ICP-MS using a new multiple ion counting
Faraday collector array. Journal of Analytical Atomic Spectrometry 20 (8), 677.
Thirlwall, M.F., 1991. High-precision multicollector isotopic analysis of low levels of Nd
as oxide. Chemical Geology: Isotope Geoscience Section 94 (1), 13–22.
Vajda, N., LaRosa, J., Zeisler, R., Danesi, P., KisBenedek, G., 1997. A novel technique for the
simultaneous determination of Pb-210 and Po-210 using a crown ether. Journal of
Environmental Radioactivity 37 (3), 355–372.
Van De Flierdt, T., Frank, M., Lee, D., Halliday, A., Reynolds, B., Hein, J., 2004. New
constraints on the sources and behavior of neodymium and hafnium in seawater
from Pacific Ocean ferromanganese crusts. Geochimica et Cosmochimica Acta 68
(19), 3827–3843.
Volpe, A., Olivares, J., Murrell, M., 1991. Determination of radium isotope ratios and
abundances in geologic samples by thermal ionization mass-spectrometry.
Analytical Chemistry 63 (9), 913–916.
Weis, D., Kieffer, B., Maerschalk, C., Pretorius, W., Barling, J., 2005. High-precision Pb–
Sr–Nd–Hf isotopic characterization of USGS BHVO-1 and BHVO-2 reference
materials. Geochemistry Geophysics Geosystems 6.
Weis, D., Kieffer, B., Maerschalk, C., 2006. High-precision isotopic characterization of
USGS reference materials by TIMS and MC-ICP-MS. Geochemistry Geophysics
Geosystems 7.
Weis, D., Kieffer, B., Hanano, D., Silva, I., Barling, J., Pretorius, W., Maerschalk, C.,
Mattielli, N., 2007. Hf isotope compositions of US Geological Survey reference
materials. Geochemistry Geophysics Geosystems 8.