Practical and theoretical considerations for the determination of δ13CVPDB values of methylmercury in the environment

Article abstract of DOI:10.1002/rcm.8453

Rationale: Analytical methods that can identify the source and fate of mercury and organomercury compounds are likely to be useful tools to investigate mercury in the environment. Carbon isotope ratio analysis of methylmercury (MeHg) together with mercury isotope ratios may offer a robust tool to study environmental cycling of organomercury compounds within fish tissues and other matrices. Methods: MeHg carbon isotope ratios were determined by gas chromatography/combustion-isotope ratio mass spectrometry (GC/C-IRMS) either directly or following derivatization using sodium tetraethylborate. The effects of a normalization protocol and of derivatization on the measurement uncertainty of the methylmercury δ¹³C_VPDB values were investigated. Results: GC/C-IRMS analysis resulted in a δ¹³C_VPDB value for an in-house MeHg reference material of δ¹³C_VPDB = ?68.3 ± 7.7‰ (combined standard uncertainty, k = 1). This agreed very well with the value obtained by offline flow-injection analysis/chemical oxidation/isotope ratio mass spectrometry of δ¹³C_VPDB = ?68.85 ± 0.17‰ (combined standard uncertainty, k = 1) although the uncertainty was substantially larger. The minimum amount of MeHg required for analysis was determined to be 20 μg. Conclusions: While the δ¹³C_VPDB values of MeHg can be obtained by GC/C-IRMS methods with or without derivatization, the low abundance of MeHg precludes such analyses in fish tissues unless there is substantial MeHg contamination. Environmental samples with sufficient MeHg pollution can be studied using these methods provided that the MeHg can be quantitatively extracted. The more general findings from this study regarding derivatization protocol implementation within an autosampler vial as well as measurement uncertainty associated with derivatization, normalization to reporting scales and integration are applicable to other GC/C-IRMS-based measurements.

Full text of DOI:10.1002/rcm.8453

Dunn Philip (Orcid ID: 0000-0002-3848-6187)
PRACTICAL
AND
THEORETICAL
CONSIDERATIONS
FOR
THE
DETERMINATION OF δ13CVPDB VALUES OF METHYLMERCURY IN THE
ENVIRONMENT
Philip J H Dunn1*, Mine Bilsel2, Adnan Şimşek 2, Ahmet Ceyhan Gören2,3, Murat Tunç2, Nives
Ogrinc4, Milena Horvat4, Heidi Goenaga-Infante1
1National Measurement Laboratory, LGC Limited, Queens Road, Teddington, TW11 0LY,
UK.
2TUBITAK Ulusal Metroloji Enstitüsü (TÜBİTAK UME), PO Box 54, 41470 Gebze, Kocaeli,
Turkey.
3Bezmialem Vakıf University, Faculty of Pharmacy, Department of Analytical Chemistry,
34093 Istanbul-Turkey
4Jožef Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia.
*corresponding author - philip.dunn@lgcgroup.com
This article has been accepted for publication and undergone full peer review but has not been
through the copyediting, typesetting, pagination and proofreading process which may lead to
differences between this version and the Version of Record. Please cite this article as doi:
10.1002/rcm.8453
This article is protected by copyright. All rights reserved.

ABSTRACT
RATIONALE: Analytical methods that can identify the source and fate of mercury and
organomecury compounds are likely to be useful tools to investigate mercury in the
environment. Carbon isotope ratio analysis of methylmercury (MeHg) together with mercury
isotope ratios may offer a robust tool to study environmental cycling of organomercury
compounds within fish tissues and other matrices.
METHODS: MeHg carbon isotope ratios were determined by gas
chromatography/combustion-isotope ratio mass spectrometry (GC/C-IRMS) either directly or
following derivatization using sodium tetraethylborate. The effects of normalisation protocol
and of derivatization on the measurement uncertainty of the methylmercury δ13CVPDB values
were investigated.
RESULTS: GC/C-IRMS analysis resulted in a δ13CVPDB value for an in-house MeHg reference
material of δ13CVPDB = -68.3 ± 7.7 ‰ (combined standard uncertainty, k = 1). This agreed very
well with the value obtained by offline flow-injection analysis-chemical oxidation-isotope ratio
mass spectrometry of δ13CVPDB = -68.85 ± 0.17 ‰ (combined standard uncertainty, k = 1)
although the uncertainty was substantially larger. The minimum amount of MeHg required for
analysis was determined to be 20 μg.
CONCLUSIONS: While the δ13CVPDB values of MeHg can be obtained by GC/C-IRMS
methods with or without derivatization, the low abundance of MeHg precludes such analyses
in fish tissues unless there is substantial MeHg contamination. Environmental samples with
sufficient MeHg pollution can be studied using these methods provided that the MeHg can be
quantitatively extracted. The more general findings from this study regarding derivatization
protocol implementation within an autosampler vial as well as measurement uncertainty
associated with derivatization, normalisation to reporting scales and integration are applicable
to other GC/C-IRMS-based measurements.
This article is protected by copyright. All rights reserved.

INTRODUCTION
Mercury (Hg) is very toxic to humans as well as animals; it can bio-accumulate in terrestrial
and aquatic ecosystems and is therefore a pollutant of particular concern in the environment in
general. There is little doubt that human activities have contributed to the levels of Hg in the
natural environment, particularly during the last century.[1] While there is legislation in place
to limit anthropogenic release of Hg and its use is being phased out where less toxic alternatives
exist, there is still a requirement for measurement of mercury in the environment as well as the
development of methods to trace Hg through ecosystems. These measurements underpin
measures aimed at monitoring mercury pollution, such as the European Water Framework
Directive and the Minamata Convention.[2,3]
While Hg can exist in many chemical forms within the environment such as elemental mercury
2+
(Hg0) and oxidised mercury (Hg2+ and Hg2 ), it is organic forms of mercury such as
methylmercury (MeHg) that are of particular interest when investigating Hg transport within
the environment. Biological activity transforms inorganic Hg (iHg) into organic MeHg and it
is the latter that bio-accumulates and biomagnifies in the food webs. Human exposure to MeHg
largely comes from the consumption of fish and other seafood;[4] however, a better
understanding of the source(s) and fate(s) of MeHg within the environment is required,
particularly within materials such as fish tissues that may enter the human food chain.
For elements with more than one stable isotope, isotope ratio analysis can be used to follow
elements or indeed compounds though biosynthetic pathways. While the environmental
pathways and cycling of bio-element isotope ratios (H, C, N, O and S) have been more
extensively studied,[e.g.5,6] isotope ratio analyses of elements such as Pb, Sr, Mo, Hg and U are
also becoming more common.[e.g. 7,8] Such isotopic studies can measure isotope ratios at natural
abundance or introduce isotopically labelled compounds into the element cycle as tracers that
can then be followed through environmental compartments.
This article is protected by copyright. All rights reserved.

While mercury isotope ratios have the potential to help uncover how mercury passes through
the environment,[e.g. 9-12] the isotope ratios of the organic moieties within MeHg and other
organic forms of Hg also have the potential to aid understanding. For example, the study of the
carbon isotope ratios of MeHg within fish tissues or other aquatic organisms, sediments or soils
may reveal the source(s) of carbon within the methyl groups. While there have been some
studies into the carbon isotope analysis of organometallic compounds using hydride
generation,[e.g. 13, 14] the carbon isotope ratio analysis of MeHg or other organomercury
compounds has not been extensively studied in the past, although Masbou and colleagues have
published a comparison of three different approaches for the analysis of MeHg by gas
chromatography/combustion-isotope ratio mass spectrometry (GC/C-IRMS),.[15] Thus far,
however, carbon isotope ratios of MeHg that are fully traceable to the international Vienna
Peedee Belemnite (VPDB) scale have not been reported. Furthermore, only standard deviations
of replicate analyses have been provided in previous studies of MeHg carbon isotope ratios.[15]
In addition, there is as yet no complete uncertainty budget reported for the determination of
MeHg carbon isotope ratios and therefore it is unclear what the analytical limitations of such
methods are.
In this work we have developed methods and derived uncertainty budgets for the fully VPDB-
traceable carbon isotope ratio determination of MeHg with the aim of studying MeHg within
fish tissues. Reference compounds/materials for method development were identified, then
characterised for their bulk 13CVPDB values by flow injection-chemical oxidation-isotope ratio
mass spectrometry (FIA/CO-IRMS). Two methods for the determination of compound-specific
MeHg 13CVPDB values by GC/C-IRMS were developed, with and without derivatization. For
the first time, the performance of the developed methods in terms of measurement uncertainty
as well as normalisation and traceability considerations has been assessed. While the focus of
this work has been on methods that can be applied to the study of MeHg extracted from fish
This article is protected by copyright. All rights reserved.

tissues, the implications for carbon isotope ratio analysis of MeHg within other environmental
samples are also discussed.
MATERIALS and METHODS
Safety considerations
Methylmercury species are toxic and particularly hazardous at the elevated concentrations
described below. Furthermore, derivatives of methylmercury such as methylethylmercury
(MeEtHg) and other doubly substituted organomercury compounds such as dimethylmercury
(Me2Hg) and diethylmercury (Et2Hg) are substantially more hazardous due to their volatility,
and exposure to such compounds should be avoided wherever possible.
Specific procedures to minimise risk arising from mercury species and from other hazards were
applied during this work and included (i) derivatization of Hg species using NaB(Et)4 in
tetrahydrofuran (THF) to avoid the need to prepare aqueous solutions of the derivatization
agent which can result in the release of flammable gases; (ii) carrying out the derivatization
within sealed autosampler vials to remove the need to transfer concentrated solutions of doubly
substituted organomercury compounds between containers prior to instrumental analysis; (iii)
modification of the GC/C-IRMS instrumentation to include a gold trap after the combustion
reactor, with all vent lines being connected to exhaust systems; (iv) sealing of the autosampler
waste vials to avoid release of organomercury compounds into the laboratory atmosphere; and
(v) use of an on-column injector to remove the risk of derivatized MeHg species being vented
during injection (which has the potential to occur with split/splitless injections).
Materials
Methylmercury chloride solids and aqueous solutions were obtained from Alfar Aesar (1000
µg mL-1 solution in H2O, part number 33553, Heysham, UK), VHG (1000 µg mL-1 solution in
This article is protected by copyright. All rights reserved.

H2O, part number VHG-MMC-25, Manchester, NH, USA) and Dr. Ehrenstorfer (solid, part
number DRE-C15100000, Augsburg, Germany). The solid material from Dr. Ehrenstorfer was
dissolved in 18 MΩ cm-1 water (Elga system, Veolia, Marlow, UK) to a concentration of 1000
µg mL-1 thereby resulting in solutions of three different MeHg materials, each at 1000 µg mL-
1
in water being available for analysis. The 1000 µg mL-1 mono-elemental solution of Hg in
dilute nitric acid was obtained from Romil (Waterbeach, UK).
FIA/CO-IRMS analysis
Flow injection analysis/chemical oxidation-Isotope ratio mass spectrometry (FIA/CO-IRMS)
analysis was performed using a Dionex Ultimate 3000 HPLC system (pump and autosampler)
coupled via an LC IsoLink chemical oxidation interface to a Delta V Advantage mass
spectrometer (all from Thermo Scientific, Bremen, Germany). The mobile phase was 18 MΩ
cm-1 water pumped at a flow rate of 500 µL min-1 while the oxidation reagents were phosphoric
acid (1.5 M, prepared from 99 % orthophosphoric acid, Sigma Aldrich, Poole, UK) and sodium
persulphate (1 g mL-1, 99 %, Sigma Aldrich) were each pumped at 30 µl min-1.
Raw isotope delta values for sample gas peaks were obtained using CO2 working gas pulses
introduced to the mass spectrometer via a separate open split within the LC IsoLink at the
beginning (n = 3) and end (n = 3) of each run. These raw delta values were determined outside
the instrumental software to allow use of the International Union of Pure and Applied
Chemistry’s (IUPAC) Commission on Isotopic Abundances and Atomic Weights (CIAAW)-
17
recommended O correction.[16] The raw delta values were corrected for the blank and then
scale calibrated to the VPDB delta scale using reference materials (RMs) analysed at the
beginning and end of each sequence. These RMs were ERM AE672a glycine (13R = 0.010648
± 0.000031; indicative δ13CVPDB = -42.12 ± 0.42 ‰, expanded uncertainties, k = 2, LGC
Standards, Teddington, UK), USGS40 l-glutamic acid (δ13CVPDB = -26.39 ± 0.08 ‰, expanded
This article is protected by copyright. All rights reserved.

uncertainty, k = 2, USGS, Reston, VA, USA), IAEA-CH-6 (δ13CVPDB = -10.45 ± 0.07 ‰,
expanded uncertainty, k = 2, IAEA, Vienna, Austria) and USGS41 l-glutamic acid (δ13CVPDB
=
+37.63 ± 0.10 ‰, expanded uncertainty, k = 2, USGS) that had been prepared as solutions in
18 MΩ cm-1 water at a concentration of approximately 4 mg mL-1 in terms of carbon.
Measurement uncertainties were estimated following the Kragten numerical approach for all
calculations.[17-19] For instrumentally measured terms such as integrated ion current ratios, the
standard uncertainty was taken to be the standard deviation of repeat analyses from the same
vial.
GC/C-IRMS analysis using derivatization
The derivatization protocol and chromatographic separation were based upon the work of Epov
et al.[20,21] MeHg was ethylated using 1 mL of a 0.1 M acetic acid/sodium acetate buffer with
pH 4 prepared from sodium acetate and acetic acid from Sigma Aldrich, added to a 2-mL GC
autosampler vial and cooled in an ice bath. 0.2 mL of aqueous MeHg solution was added,
followed by 20 µL of 10 % NaB(Et)4 in THF (LGC Standards) and 0.5 mL of hexane (LGC
Promochem, Teddington, UK) all of which had been previously cooled in an ice bath. The vial
was sealed and gently shaken for five minutes and then stored in a sealed container at –5 °C
until required for analysis.
GC/C-IRMS analysis was performed using a Trace GC Ultra gas chromatograph with on
column injector and Triplus autosampler coupled via a GC IsoLink combustion interface and
Conflo IV continuous flow interface to a Delta V Advantage isotope ratio mass spectrometer
(all Thermo Scientific). The backflush port of the GC IsoLink and the extraction port of the
Conflo IV were connected to an extraction line. A gold trap (Alfa Aesar gold wires, 0.1 mm
diameter, 99.95% purity, 1 m length, folded in half three times to give 8 strands of 12.5 cm
length inside the gold trap) inside a narrow piece of stainless steel tubing was connected inline,
This article is protected by copyright. All rights reserved.

directly after the combustion reactor of the GC IsoLink, to trap any mercury that did not remain
inside the reactor. GC separation of mercury species was performed using a MXT-1 capillary
column (cross-bonded 100 % dimethyl polysiloxane, 15 m, 0.25 mm I.D., 0.25 µm film
thickness, Restek, Bellefonte, PA, USA) using the following oven programme: initial
temperature 40 °C, held for 7 min, ramp to 80 °C at 60 °C min-1, hold for 6 min, ramp to 250
°C at 60 °C min-1 and hold for 1 min (total analysis time 17 min 50 s). The helium carrier gas
flow rate was 2 mL min-1. The injection volume was 1 µL taken directly from the upper hexane
phase within the vial in which derivatization was performed, removing the need to transfer the
volatile and highly toxic mercury species. No sample wash step was implemented within the
autosampler method and the waste vials for the autosampler were sealed using crimp-caps with
septa to avoid the volatile mercury species being released into the laboratory atmosphere. The
backflush valve of the GC IsoLink was turned on at the start of each run, turned off at 150 s
and then turned back on at 800 s to the end of each run.
The instrumental software (Isodat v 3.0, Thermo Scientific) was used for data acquisition and
for integration of ion current chromatograms. Raw delta values for sample gas peaks were
obtained using CO2 working gas pulses introduced to the mass spectrometer via a separate open
split within the Conflo IV at the beginning (n = 3) and end (n = 3) of each run. These raw delta
values were determined outside of the instrumental software as before, using the CIAAW-
recommended 17O correction.[16] The raw delta values were then corrected for the presence of
derivative carbon using a simple mass balance approach (as described elsewhere[19, 22-23]) either
using an external standard iHg solution derivatized within the same sequence of analyses, or
using an internal standard iHg solution added to the MeHg solution prior to derivatization, to
determine the carbon isotopic composition of the ethyl derivatization group (following[15]). The
corrected delta values for the MeHg were then scale calibrated using two MeHg external
standard solutions analysed before and after the samples that had been previously calibrated
This article is protected by copyright. All rights reserved.

for carbon isotopic composition by FIA/CO-IRMS against secondary reference materials. All
calculations were performed within a Kragten-type spreadsheet (KS) to allow the estimation of
measurement uncertainty to be carried out.[17-19] For instrumentally measured terms such as
integrated ion current ratios, the uncertainty was taken to be the standard deviation of repeat
analyses from the same vial.
Before each analytical sequence, the backgrounds for m/z 18, 28, 32, 40 and 44 were checked
to ensure that they were below the established thresholds, and pulses of working gas were used
to assess the stability and linearity of the mass spectrometer response to carbon dioxide.
GC/C-IRMS without derivatization
GC/C-IRMS analysis of MeHg without derivatization was carried out using a Trace GC Ultra
gas chromatograph linked to a MAT 253 isotope ratio mass spectrometer using a GC
combustion III interface (all Thermo Scientific, Bremen, Germany). A ZB-1MS column (60 m
x 0.25 mm I.D., 0.25 µm film thickness, Phenomenex) was pre-treated by injecting 1 µL 0.3
mM methanolic HBr ten times at 1 min. intervals. After the column pre-treatment, 1 µL of
MeHg in benzene solution was injected into the instrument using a split/splitless injector in
split mode (split ratio 1:10) held at 220 °C. The carrier gas (helium) flow rate was 10 mL min-
1. An isothermal GC oven programme was employed at 50 °C for 6 min to ensure elution of
the Hg species. Backflush was used to divert the solvent peak. Backflush was on until 600 s
and then off until the end of the analysis.
MeHg was injected in the form of a 1000 µg mL-1 solution in benzene. While the solid MeHgCl
in-house reference materials (e.g. Dr. Ehrenstorfer) could be directly prepared as 1000 µg mL-
1
solutions in benzene, the MeHg in-house reference materials obtained as aqueous solutions
required an extraction step into benzene. This was achieved by the addition of an equal volume
This article is protected by copyright. All rights reserved.

of benzene to the MeHg aqueous solution, vigorous shaking and allowing the two phases to
separate.
Raw data (δ13C values on a scale defined by the working gas) were obtained using Isodat v3.0
and the Santrock Studley and Hayes (SSH) 17O correction algorithm[24]. The working gas was
introduced in pulses directly into the mass spectrometer at the beginning of each run.
RESULTS AND DISCUSSION
Characterization of in-house MeHg reference materials
GC/C-IRMS instrumentation does not provide structural information for the sample
compounds and therefore a separate means of compound identification is required. This can
consist of e.g. an organic mass spectrometer that is interfaced to same gas chromatograph with
the column eluent split to allow simultaneous compound identification of and isotope ratio
measurement.[25] Where such a system is not available, identification of compounds eluting
from the gas chromatograph should be carried out on a separate GC/MS system that uses the
same chromatographic column and oven programme. For simple analyses involving only few
compounds such as the isotopic composition of MeHg from high-purity standard solutions, it
is also possible simply to compare the chromatogram obtained for the material with that for a
procedural blank. Provided that the sample preparation and instrumental analysis approaches
have been well-studied in the past, it is likely that any peaks present in procedural blank are
not the compound(s) of interest. If only one peak remains within the sample chromatogram that
is not in the procedural blank and which is of the expected size based upon the amount of the
high-purity standard compound injected, identification of peaks is straightforward.
In this work, the MeHg materials were of high purity (>95 % according to the manufacturers),
while the derivatization and chromatographic separations had been previously studied and not
found to produce significant amounts of byproducts.[20,21] Comparison of the chromatograms
This article is protected by copyright. All rights reserved.

for a procedural blank, derivatized MeHg solution and derivatized iHg solution (Figure 1)
showed that only one peak was present in each of the derivatized solutions, thereby allowing
identification of the peaks associated with the derivatized MeHg species. For the method using
direct injection of MeHg without a derivatisation step, again only one peak was obtained for
the analysis of MeHg (Figure 2) making identification of the MeHg peak straightforward.
To determine whether the carbon isotope ratios of MeHg in-house reference material solutions
could be measured using a bulk isotope ratio analysis approach such as FIA/CO-IRMS, it was
necessary to demonstrate that each of the three solutions predominately contained carbon in
the form of MeHg. Five subsamples of each of the three MeHg in-house reference material
solutions (VHG, Alfa Aesar and Dr. Ehrenstorfer) were derivatized within the same batch and
then each analysed in triplicate by GC/C-IRMS. Procedural blanks (n=3) and an external
standard iHg solution were also derivatized and analysed in triplicate. The chromatograms
obtained for the derivatized MeHg solutions were examined for peaks not present within the
procedural blanks or attributed to MeEtHg. The only peaks thus identified were attributed to
the presence of Et2Hg by comparison of retention times with those in iHg solutions also
derivatized within the same batch (Figure 1).
These Et2Hg peaks were very small, being < 15 % of the peak area of the MeEtHg peaks (often
much smaller) and therefore < 12 % of the mass of MeEtHg represented by those peaks, and
they could arise from the presence either of EtHg or of iHg in the MeHg in-house reference
material solutions as both of these compounds would produce Et2Hg upon derivatization. The
Et2Hg peaks arising from the analysis of MeHg in-house reference material solutions displayed
raw δ13C values = -30.6 ‰ with standard deviation of 6.7 ‰. This poor precision resulted from
the very small nature of the Et2Hg peaks, making run-to-run variability very high. The delta
values were identical within measurement uncertainty to those of the Et2Hg obtained from
derivatization of the iHg solutions prepared and analysed within the same batch (raw δ13C = -
This article is protected by copyright. All rights reserved.

27.34 ± 0.26 ‰). Only if the ethyl carbon isotope ratio of any purported EtHg present in the
MeHg in-house reference material solutions were a close match to that of the derivatization
agent would this be the case. There are therefore two possible causes of the small Et2Hg peaks
observed following derivatization of the MeHg in-house reference material solutions. First,
there could be iHg within the MeHg standard solutions – indeed this has been observed during
Hg isotope analysis of mercury species within the same MeHg standard solutions (Dmitriy
Malinovskiy, personal communication). Secondly, it is possible for derivatization of MeHg in
the presence of solvent to induce degradation artefacts in small amounts that could also be the
cause of the small Et2Hg peaks.
There was therefore no evidence to suggest the presence of other carbon-containing mercury
species within the MeHg in-house reference material solutions. Hence characterisation of these
solutions by a bulk isotope ratio technique such as elemental analyser-IRMS (EA-IRMS) or
FIA/CO-IRMS to determine the carbon isotope ratio of the MeHg would be possible.
Quantitative conversion of MeHg to CO2 during FIA/CO-IRMS
Quantitative conversion of a target compound or material into the appropriate analyte gas for
analysis by IRMS is a prerequisite for determination of accurate isotope ratios. Indeed it is the
study of analyte gas yield that has recently highlighted the potential for incomplete conversion
of hydrogen within nitrogen- and other heteroatom-containing molecules to hydrogen gas
during high temperature conversion using glassy carbon.[26-28] The simplest means to determine
whether the yield of CO2 is quantitative for a particular material is to compare the peak size
obtained in the IRMS chromatogram for a given mass of the element within the target
compound with the peak size for the same mass of the element within a material known to
exhibit quantitative conversion (e.g. a reference material for isotope ratio). During FIA/CO-
IRMS analysis, the peak area obtained for MeHg was compared with that of the RMs used for
This article is protected by copyright. All rights reserved.

normalisation and as quality control materials (two glycines, two L-glutamic acids and a
sucrose) as all materials were analysed in known amounts in terms of carbon. All peak areas
for the MeHg solutions were of similar peak area to those expected based upon the analysis of
the RMs, suggesting that complete conversion of the MeHg to CO2 had been achieved. The
peak area data can be found in Table 1.
Determination of δ13CVPDB values for in-house MeHg reference materials by FIA/CO-
IRMS
The three MeHg in-house reference material solutions (VHG, Alfa Aesar and Dr Ehrenstorfer)
were analysed by FIA/CO-IRMS in the same sequence as four different normalisation
reference materials to determine their δ13CVPDB values. Each material was analysed using five
repeat injections from the same subsample of solution. The results can be found in Table 2,
which also includes the obtained δ13CVPDB values for the normalization reference materials -
these all agreed with the expected values listed above within uncertainty.
The calibration range afforded by the RMs was from δ13CVPDB = -42.12 ‰ to δ13CVPDB = +37.63
‰ and therefore, for two of the MeHg in-house reference materials (VHG and Alfar Aesar),
calibration via extrapolation was required. Given that the RMs covered just under 80 ‰ while
the extrapolation was by up to 30 ‰, there was only a slight increase in measurement
uncertainty compared with the MeHg in-house reference material within the calibration range.
For each of the three MeHg solutions, over 85 % of the uncertainty obtained was due to the
uncertainty in the reference material expected values. The remainder was from the measured
parameters for the normalization RMs as well as for the MeHg solutions.
This article is protected by copyright. All rights reserved.

Consistency of derivatization for GC/C-IRMS analysis
Where derivatization is used prior to GC/C-IRMS analysis, the derivatization reaction must
involve either no kinetic isotope effect (KIE) or a reproducible KIE for the correction for
derivative carbon to be successful.[22,23] If the KIE is not reproducible, the reaction approach is
unsuitable for GC/C-IRMS analysis. It therefore follows that repeated derivatization of
subsamples of the same material using the same batch of derivatization agent should result in
consistent raw delta values.
Five subsamples of same MeHgCl in-house reference material solution (VHG) were
derivatized within a single batch using the same bottle of NaB(Et)4 in THF. Each of the
derivatized subsamples was then analysed in triplicate by GC/C-IRMS and raw delta values
for MeEtHg obtained (Figure 3). One way analysis of variance (ANOVA) demonstrated no
significant difference in the raw delta values between the subsamples (Fstat = 0.668, Fcrit
3.478, p = 0.63), demonstrating that the reaction was suitable.
=
The peak areas of these analyses were also very consistent despite their small size (average =
1.48 mV.s, standard deviation = 0.19 mV.s, n=15), suggesting that conversion of the MeHg
species to CO2 was consistent. To determine if this conversion and the derivatization and
extraction into the hexane phase were also complete, the signal from the CO2 produced from
known amounts of derivatized MeHg species was compared with that obtained from analysis
of known amounts of fatty acid methyl esters (FAMEs) - compounds that exhibit favourable
combustion characteristics within the instrumentation (which was in the same tune state). The
mass of carbon in the form of MeEtHg within each injection for the 15 analyses described
above was approximately 58 ng (assuming that derivatization and extraction into the hexane
were quantitative) with amplitudes of the associated carbon dioxide peaks of approximately
2.2 V (m/z 44, 3x108  amplification). For a range of even-numbered FAMEs from methyl
myristate to methyl behenate, a signal amplitude of approximately 3.2 V was obtained for the
This article is protected by copyright. All rights reserved.

analysis of approximately 88 ng of carbon of each. The MeEtHg signal therefore represents
approximately 104 % of the signal size expected based upon the FAME results, suggesting that
complete conversion of MeEtHg to carbon dioxide has been achieved and that both
derivatization and extraction into the hexane were quantitative.
Correction for derivative carbon and contribution to uncertainty
The addition of carbon from the NaB(Et)4 to the MeHg must be accounted for in order to
determine the δ13C value of MeHg. This can be done by a simple mass balance approach as
described elsewhere.[19, 22,23] This necessitates determination of the δ13C value of the derivative
carbon (δ13Cd). Where the derivatization agent has been prepared from solid NaB(Et)4, the solid
could be analysed by an offline approach such as EA-IRMS to determine the δ13Cd value. As
NaB(Et)4 in THF was used, this cannot be done due to the carbon present in the THF and
therefore an indirect approach to determination of the δ13Cd value is required. As the
derivatization procedure also converts any iHg present to diethylmercury (Et2Hg), an external
or internal iHg standard can be used to determine the δ13Cd value.[15] The ethyl groups of both
MeEtHg and Et2Hg can be assumed to be identical provided that (i) the same lot of
derivatization agent is used; (ii) derivatization is carried out within the same batch; and (iii) the
derivatization agent is present in great excess.
While the correction for derivative carbon is straightforward, its application will contribute to
the overall measurement uncertainty. The uncertainty associated with the correction can be
estimated by using a KS or by using the traditional means to combine uncertainty components
within the mass balance equation together as described elsewhere.[23] If one assumes that the
isotopic compositions of the derivatized methylmercury compound and the derivative carbon
(from derivatized iHg) can be measured to the same degrees of uncertainty regardless of
isotopic composition (0.2 ‰); and that the isotopic composition of the methyl mercury is
This article is protected by copyright. All rights reserved.

identical for each different derivative option, it is possible to isolate how the number of carbon
atoms within a derivative affects the uncertainty in the corrected delta value for MeHg (Figure
4).
The KS approach displays a linear increase in measurement uncertainty with increasing
numbers of derivative carbon atoms, while the Docherty equation produces a polynomial
increase. Regardless of the uncertainty estimation approach it is clear that increasing the
number of derivative carbon atoms, increases the associated measurement uncertainty. The
selection of derivatization approach (e.g. between the use of ethylation and propylation) is
therefore a balance between increasing the measurement uncertainty from the correction for
derivative carbon and increasing the signal amplitude which results in better precision of
replicate injections. More generally there is also chromatographic performance to consider as
this can be important in some applications of GC/C-IRMS;[e.g. 29] however. this aspect is of less
importance during the study of MeHg.
One further consideration for the correction of derivative carbon is whether to use iHg as an
internal standard or as an external standard for the determination of the isotopic composition
of the derivative carbon. With the internal standard approach, iHg is added to each sample of
MeHg prior to derivatization and the resulting Et2Hg used during the derivative correction. The
external standard approach uses a separate solution of iHg that is derivatized within the same
batch as any MeHg samples. The internal standard approach does have the advantage that the
two compounds, MeEtHg and Et2Hg, pass through the combustion reactor and other parts of
the instrumentation much closer in time, ensuring that instrumental conditions are as similar as
possible. The advantage of the external standard approach is that should the iHg solution
contain any MeHg, this would not contribute to the isotope ratios determined for any sample
MeHg. There was no difference within measurement uncertainty between the values obtained
for the VHG MeHg normalised using the other two MeHg in-house reference material
This article is protected by copyright. All rights reserved.

compounds when using a derivative carbon correction via an internal or external standard
approach. This was because the raw delta value for the Et2Hg derived from the internal
standard (raw δ13C = -27.35 ± 0.45 ‰) and external standard approaches (raw δ13C = -27.34 ±
0.26 ‰) were also identical within measurement uncertainty for analyses within the same
sequence.
Normalisation of GC/C-IRMS δ13C values using in-house MeHg reference material
solutions
Compound-specific carbon isotope ratios should be reported on the VPDB scale. Delta values
obtained by GC/C-IRMS analyses against a working gas, whether or not this gas has been
calibrated against reference materials, are not traceable to VPDB as the sample and working
gases are not treated identically during analysis. Ensuring traceability to VPDB requires
normalisation of raw delta values using at least two materials of known isotopic composition
that are traceable to the reporting scale analysed within the same sequence.[19, 30] These should
be isotopologues of the sample compound and should have isotopic compositions that span the
expected range of the sample compound to allow calibration via interpolation rather than
extrapolation. For the carbon isotope ratio analysis of individual compounds separated within
a complex mixture this is potentially challenging due to the large number of reference materials
that are necessary and therefore more pragmatic approaches have been recommended;[19]
however, in the current case where only one compound (MeHg) is of interest, traceability to
the reporting scale can be easily achieved while adhering to the ideal approach.
Provided that two methylmercury reference compounds can be sourced with differing isotopic
composition, these can be used for normalisation if they are treated identically to the samples
and analysed within the same sequence using the same batch of derivatization agent.
Determination of the carbon isotopic composition of these MeHg reference compounds can be
This article is protected by copyright. All rights reserved.

carried out by some form of bulk carbon isotope ratio analysis such as EA-IRMS or FIA/CO-
IRMS provided that there is very little carbon present in the material which is not in the form
of MeHg. As discussed above, three different MeHg in-house reference materials (VHG, Alfa
Aesar and Dr. Ehrenstorfer) were sourced and characterised for non-MeHg carbon and also for
bulk carbon isotope ratios by FIA/CO-IRMS. Two were suitable to use for normalisation as
they differed in δ13CVPDB values by just over 30 ‰ (Table 3) while the third (VHG) could be
used as a quality control (QC) material as it was within this calibration range in terms of its
δ13CVPDB value.
Normalisation of sample MeHg δ13C values to the VPDB scale can be achieved using the
offline FIA/CO-IRMS and online GC/C-IRMS measured values for the Alfa Aesar and Dr.
Ehrenstorfer MeHg in-house reference material to create a normalisation plot. The maximum
uncertainty introduced by normalisation is < 0.35 ‰ assuming that the GC/C-IRMS
measurements of the normalisation reference materials and of the sample each have an
associated uncertainty of 0.2 ‰ and that the sample can be calibrated via interpolation. Should
sample MeHg δ13C values lie outside the calibration range, there are two options: first. the
calibration range could be extended (by sourcing a suitable MeHg commercially or by synthesis
of 12C- or 13C-enriched MeHg and gravimetric mixing with the current MeHg in-house
reference material to obtain the desired δ13C value) or calibration via extrapolation could be
employed. The latter is far easier and more cost effective; however, the additional uncertainty
introduced by extrapolation must be accounted for.
This additional uncertainty can be estimated using a KS, where the isotopic compositions of
the normalisation RMs as well as all input parameter standard uncertainties are held constant
but the isotopic composition of the sample is varied. Using such an approach, the effect of each
permil increase in distance from the 30 ‰ calibration range can be shown to result in an
increase in the combined uncertainty of the scale-calibrated δ13C value of 0.01 ‰ (Figure 5).
This article is protected by copyright. All rights reserved.

The effect of calibration range can also be investigated using a similar approach whereby a KS
is used to determine the relationship between the δ13C value range afforded by the
normalisation RMs and the additional measurement uncertainty introduced by normalisation
via extrapolation (Figure 5). It is clear that the smaller the difference in isotopic composition
between the normalisation reference materials, the more rapidly the measurement uncertainty
increases when extrapolation is applied. Extrapolation by 100 % of the calibration range
doubles the uncertainty in the resulting delta value and, therefore, unless the increase in
measurement uncertainty is fit-for-purpose, extrapolation should be avoided.
Use of matrix matched normalisation RMs and the need for other corrections
Application of the principle of identical treatment (PIT) ensures that samples and RMs are
treated in the same way during analyses and are of the same matrix, thereby negating some
biases and cancelling some contributions to measurement uncertainty.[31] Often in isotope
analysis it is possible to use organic RMs for normalisation of organic samples, rather than
exactly matching the sample and RM compound. In the situation where two exactly-matched
compounds are available for normalisation with a further QC material also of the same
compound as the sample, all of which are analysed at the same amount level, the PIT can be
applied in the strictest sense and thereby separate calculation stages for blank correction,
normalisation and correction for derivative carbon need not be applied. Simply creating a
calibration plot of the measured (raw) and expected δ13C values for the normalisation reference
materials and applying this to the samples and QC materials of the same compound run within
the same sequence is sufficient. In situations where the analyte compound is not an exact match
for the compounds used for normalisation, separate calculation stages must be employed.
Applying a normalisation directly to the measured data for the sample MeHg does not result in
a value or measurement uncertainty that is different from that from the stepwise approach
This article is protected by copyright. All rights reserved.

involving separate derivative correction and normalisation stages. It is also the case that the
measurement uncertainty budget remains the same, with the majority of the uncertainty coming
from the raw ion current ratios for the MeEtHg peaks within the chromatograms.
Other measurement uncertainty considerations
When peaks are small, automated integration of ion current signals is typically less reliable
than for larger amplitude signals and this can result in larger variation in peak areas between
replicates even if these are repeated injections of the same solution. It is therefore important to
capture this measurement uncertainty in a realistic way. While it has been previously suggested
that the input parameters to a KS for the calculation of raw delta values from integrated ion
current ratios should use the standard deviation of the mean of independent replicates,[18] this
can lead to overestimation of measurement uncertainty in cases where there is significant
variation in peak areas between replicates of the same sample. In such cases (e.g. small sample
sizes with consequent difficulties with integration, or indeed for bulk measurements on
materials for which it is difficult to weigh out the same amount for replicate analyses), it is
better to calculate the raw delta values for each replicate analysis independently and then
determine the mean later in the calculation process. This is because the raw m/z 45/44 and m/z
46/44 ratios will vary in the same way (they are correlated) and therefore there can be
significant variation in these two ratios but far less variation in the raw delta value that they are
used to calculate. Alternatively, the input data into the KS should not be raw ion current ratios,
but rather the raw integrated ion currents individually, as this again would avoid the
manifestation of correlation between the m/z 45/44 and m/z 46/44 ratios.
To illustrate the effect on measurement uncertainty of this correlation, the same raw
instrumental data can be evaluated in two different ways: first, as described elsewhere taking
the mean of the input ion current ratios,[18] and secondly, treating each replicate separately for
This article is protected by copyright. All rights reserved.

raw delta value calculation, then taking the mean and applying the remaining calculation stages
such as normalisation. Doing this for three replicate injections of the Dr. Ehrenstorfer MeHg
resulted in individual raw δ13C values of -28.40 ± 0.20 ‰, -26.42 ± 0.20 ‰ and -27.67 ± 0.20
‰ (combined uncertainty, k = 1). The mean of these three values with combined uncertainty is
-27.50 ± 0.61 ‰ (combined uncertainty, k = 1), while using the mean ion current ratios as input
data resulted in a raw delta value of -27.50 ± 0.82 ‰. This difference in obtained measurement
uncertainty highlights the need to account for or avoid correlation between input parameters –
particularly when the correlation would result in a reduced measurement uncertainty.
A further instance of correlation between input parameters arises from the use of reference
materials of known delta values within the traceability chain for the MeHg in-house reference
materials. For example, the indicative δ13C value for ERM AE672a applied in this work was
obtained by normalisation using reference materials including USGS40, IAEA-CH-6 and
USGS41 – the other RMs used for FIA/CO-IRMS analyses in this work; the delta values of
these reference materials are therefore correlated. Furthermore, the currently accepted delta
values for USGS40, IAEA-CH-6 and USGS41 were all obtained during a single inter-
laboratory exercise and therefore are also probably correlated.[32,33] The relationships between
the δ13CVPDB values of commercially available RMs are complex and accounting for
correlations between them is challenging. Neglecting to account for correlations in this instance
will lead to a larger uncertainty which is a conservative rather than overly-optimistic result and
therefore acceptable.
A final potential source of uncertainty is the so-called linearity effect. This occurs when a single
material analysed over a range of different masses, produces different raw delta values and can
be seen even for homogenous materials.[19] Where sample and RMs for normalisation and QC
are not analysed in exactly the same mass, there will be an additional dispersion of the raw data
obtained, leading to increased standard deviations of replicate analyses and increased
This article is protected by copyright. All rights reserved.

measurement uncertainty. The extent of the linearity effect can be determined via the analysis
of increasing amplitude pulses of working gas – in this work, the linearity was typically less
than 0.02 ‰ V-1 as measured on the middle Faraday collector monitoring m/z 45.
It is also important to check for linearity of the observed delta values for real samples. For
CSIA of MeHg, the mass of sample injected depends on the concentration of solution as well
as the injection volume. The concentration of solution (i.e. of derivatized MeHg present in the
hexane phase) could be controlled by derivatization of different amounts of the stock 1000 µg
mL-1 MeHg solutions. Figure 6 shows the relationships between the amount of derivatized
MeHg injected on column and the signal size obtained, as well as against the raw delta value.
There is a clear linearity effect present; however, the slopes are approximately 0.1 ‰ ng-1.
Provided that the mass of sample injected is controlled to the same extent as the x-axis error
bars shown in Figure 6, the additional dispersion of raw delta values due to the linearity effect
will be approximately 0.3 ‰.
Overall measurement uncertainty for the derivatization approach
The analysis of the VHG MeHg solution by GC/C-IRMS using the other two in-house MeHg
solutions for calibration resulted in a δ13CVPDB value of -68.3 ± 7.7 ‰ (Table 3). This is a
combined standard uncertainty from 15 measurements of each of the three MeHg solutions
(two used for normalisation, one as a “sample”) consisting of three repeat analyses of each of
5 independently prepared replicates. All MeHg signals were between 1 and 3 V in amplitude
for m/z 44 with 3x108 Ω amplification to determine the performance of the method at the lower
amounts of MeHg. The mass of MeHg injected was controlled to within 3 ng (equivalent to
the x-axis error bars of Figure 6). Data analysis was by simple normalisation using the raw δ13C
values for the MeEtHg of the two in-house normalisation reference materials together with the
expected values determined by offline FIA/CO-IRMS (Table 3). The GC/C-IRMS result agrees
This article is protected by copyright. All rights reserved.

very well with the bulk value obtained by FIA/CO-IRMS analysis of δ13CVPDB = -68.85 ± 0.17
‰ although the uncertainty is substantially larger. Examination of the uncertainty related to
each stage of the calculations showed that the δ13CVPDB values for each of the five replicates of
VHG MeHg solution could be obtained with a measurement uncertainty of less than 3 ‰ from
repeated injections of the same solution (which included the uncertainty arising from
normalisation and the assignment of the in-house MeHg reference material δ13CVPDB values).
This measurement uncertainty of less than 3 ‰ for replicate injections of the same solution
was also the case for the two MeHg standard solutions used for normalisation. The variability
in results for the normalisation MeHg standard and MeHg sample solutions between the
independently prepared solutions using different lots of derivatization agent and different
derivatization batches was considerably higher and contributed over 85 % to the combined
uncertainty. Substantial variability is to be expected when small signals of the order of 1 V for
m/z 44 are integrated over a baseline that is not completely stable.
GC/C-IRMS analyses of MeHg without derivatization
Chromatographic separation of organomercury compounds can be achieved without
derivatization.29 While such methods require larger amounts of MeHg to produce the same
signal size during GC/C-IRMS analysis (assuming that the same instrument, tune state and
combustion reactor efficiency are used), correction would not be required for derivative carbon
and therefore lower measurement uncertainty would result. As is shown in Figure 4, correcting
raw delta values obtained with 0.2 ‰ standard uncertainty for derivative carbon results in a
combined uncertainty of at least 0.4 ‰ – i.e. a doubling of the measurement uncertainty as a
minimum. As derivatisation is therefore a significant component of uncertainty, investigation
of instrumental methods for carbon isotope ratio analysis of MeHg that do not require
derivatization is valuable. Furthermore, the use of such a method on the same materials as used
This article is protected by copyright. All rights reserved.

in the derivatization approach would act as independent confirmation of the results – in
particular of the correction for derivative carbon – although both techniques rely on the
FIA/CO-IRMS-derived known values for MeHg materials used for normalisation.
The data handling for this approach used a different 17O correction (SSH as opposed to
IUPAC); the effect of this on the raw delta values obtained is expected to be less than 0.06
‰.[16,19] The SSH approach could be implemented within the instrumental software
automatically, rather than requiring external calculations, which was an advantage. For
normalised delta values traceable to VPDB obtained by GC/C-IRMS, the difference in values
obtained between the 17O algorithms is significantly smaller (<0.001‰) provided that the same
algorithm is applied to samples and normalisation RMs.[19] The contribution of the 17O
correction to the measurement uncertainty for obtaining delta values for MeHg with or without
derivatization was therefore deemed negligible.
Analysis of five separate subsamples of two of the MeHg in-house reference material solutions
directly by GC/C-IRMS resulted in a mean raw δ13C value of -42.92 ‰ with a standard
deviation of 0.06 ‰ (Dr. Ehrenstorfer) and a mean raw δ13C value of -73.47 ‰ with a standard
deviation of 0.34 ‰ (VHG). The normalisation plot obtained using these two results, together
with the expected values obtained by FIA/CO-IRMS (Table 3), has a slope of 1.12 and an
intercept of 3.60. The slope is a function of the mass spectrometer tune state, while the intercept
represents the difference between the actual and assigned δ13C value for the working gas. The
standard deviations are comparable with those obtained by analysis of MeHg derivatives within
a single derivatization batch, but required much larger amounts of MeHg – Figure 2 shows the
peak size obtained from the injection of 1000 ng of MeHg (note that the amplitudes of the
signals in Figures 1 and 2 cannot be directly compared as they were obtained using different
instrumentation as noted in the methods section). As discussed above, the uncertainty
introduced by correcting for derivative carbon is approximately double that for a raw,
This article is protected by copyright. All rights reserved.

uncorrected δ13C value and therefore in instances where sufficient MeHg is present and low
uncertainty is required, the direct analysis approach would be more appropriate.
Amounts of MeHg required for GC/C-IRMS analysis
In GC/C-IRMS analysis, the lower limit for sample mass is typically around 10 ng of carbon
on column.[19] This mass of carbon results in sufficient CO2 ions being produced in the ion
source (following combustion of the carbon within the sample compound to CO2) to give a
signal of 3 nA maximum intensity (equivalent to 1 V with 3x108 Ω amplification) on m/z 44.
The exact mass of carbon required on column will of course vary with mass spectrometer
model, tune state of the instrument, etc.; however, this is a suitable estimate. Signals smaller
than 1 V are typically the subject of poorly reproducible integration (fluctuations in the
background or baseline are more significant) which results in larger variation in isotope ratio
between repeat analyses of the same material.
To determine the minimum amount of MeHg required on column, varying amounts of each of
the three MeHg in-house reference material solutions were derivatized and analysed. A plot of
mass of carbon injected (in the form of MeEtHg) against signal size could then be created
(Figure 6A). Given that each MeHg stock solution was 1000 µg mL-1, the three plots should
all overlap; however, there is a slight offset for the Dr. Ehrenstorfer MeHg. As this material
was dissolved in-house to the nominal concentration of the other two solutions, this small offset
is not unexpected. The minimum mass of carbon within the injection volume (1 μL) required
to produce a 1 V signal was approximately 30 ng on column. Given that the hexane phase was
500 μL, this implies that 15 μg of carbon was within the hexane phase. Of this carbon, only
one third is from MeHg while the rest is derivative carbon and therefore 5 μg of carbon from
MeHg (equivalent to approximately 100 μg of MeHg) would be required.
This article is protected by copyright. All rights reserved.

The volume of the hexane phase could be reduced to 100 μL which would also reduce the
amount of MeHg required proportionately to 20 μg; however. a further reduction in the volume
of hexane would make direct sampling from this phase using the autosampler syringe more
difficult. Use of vial inserts could further reduce the volumes of solution required without
impacting the ease of direct sampling of the hexane phase; however, it would then be more
difficult to ensure thorough mixing during the derivatization reaction, which would increase
the risk of incomplete derivatization and associated fractionation. Introduction of a pre-
concentration step between derivatization and analysis would be possible; however, the risk of
losing some of the volatile and highly toxic mercury species would increase. Other options for
increasing sensitivity and thereby reducing the mass of MeHg required include the use of
propylation rather than ethylation. Propylation should increase the signal amplitude by factor
of 33 % over ethylation but it requires larger correction for derivative carbon which has
consequences in terms of measurement uncertainty as discussed above.
Given the need for a combination of multiple analyses from a single derivatization reaction, as
well as independent derivatization reactions for a number of subsamples of the same MeHg,
20 μg is a pragmatic minimum amount of MeHg required. The combined standard uncertainty
of ± 7.7 ‰ obtained for the VHG MeHg solution reported above represents a practical
minimum uncertainty possible for the analysis of the carbon isotope ratio of MeHg at this
amount level used. A reasonable assumption is that this amount of MeHg will need to be trebled
to 60 μg should the GC/C-IRMS approach without derivatization described above be employed
given the corresponding decrease in the number of carbon atoms. Where more MeHg is
available for analysis, the uncertainty associated to the measurement results using the methods
described will decrease as the signal amplitude can be increased.
This article is protected by copyright. All rights reserved.

Application of developed GC/C-IRMS methods to MeHg from fish tissues and other
environmental samples
Analysis of MeHg from environmental samples such as fish tissues requires extraction of the
MeHg from the matrix. Given the relatively large amount of MeHg required for the methods
developed here it was important to determine whether any MeHg signal could be obtained from
a reasonable amount of sample. Assuming that 20 μg of MeHg is the minimum amount required
for GC/C-IRMS analysis using the ethylation approach, the applicability of the method
described above to environmental samples such as fish tissues can be judged by comparison
with the total amount of MeHg within commercially available reference materials certified for
MeHg content. Table 4 provides a list of such materials together with the number of units
requited to provide the 20 μg of MeHg. It is clear that very few of the fish tissue reference
materials contain sufficient MeHg to allow carbon isotope analysis reliably by GC/C-IRMS
with the methods described herein. The extraction and derivatization of MeHg from BCR-463
were performed using 1 g of material but no MeEtHg peak was seen above the baseline
although the expected peak size was only a few hundred mV.
A number of freshly collected fish samples together with some archived fish tissues obtained
from environmental specimen banks within Germany were also available through the European
Metrology Programme for Innovation and Research (EMPIR) Joint Research Project: ENV51
“Traceability for Mercury Measurements” (MeTra).[34] These included roach and pike from
Lake Stechlin, Germany collected in March 2015 and archived bream collected in 2013. The
total Hg contents for these fish samples were between 60 and 1300 μg kg-1.[34] Even at the upper
end of this range, approximately 15 g of fish tissue would be required for each independent
replicate analysis, substantially more than was available. It is important to note that fish
samples collected from other riverine or marine ecosystems could be significantly more
contaminated by MeHg than these reference materials and fish tissue samples. The methods
This article is protected by copyright. All rights reserved.

discussed herein do therefore have the potential to be useful for the analysis of severely polluted
specimens.
MeHg can also be found in other environmental settings including sediments, soils and river
water, for example in locations close to sites where gold is processed using Hg.[35] Table 4 also
lists a number of non-fish reference materials certified for MeHg content and again the amount
of MeHg present is generally too low for carbon isotope analysis using the methods described
above. As with fish tissues, where levels of MeHg contamination are higher than for the
reference materials listed, the methods described herein can provide useful means to study
MeHg in these types of environmental samples.
Whether samples are fish tissues, sediments, or other environmental materials, important
considerations for the application of the GC/C-IRMS methods described herein are quantitative
extraction of the MeHg present to avoid fractionation and the presence of sufficient MeHg.
Concentration levels of MeHg within environmental samples of 4 μg g-1or greater would
require 5 g (or less) of material for carbon isotope ratio analysis. Any extraction method will
require validation or verification for use with new matrices to ensure that it does not introduce
fractionation.
CONCLUSIONS
Methods for GC/C-IRMS carbon isotope analysis of MeHg have been developed including a
novel approach that applied the derivatization protocol directly within an autosampler vial,
rather than in a separate vessel with subsequent transfer. This provided a simple means to
minimise exposure to highly toxic organomercury species and precluded the loss of volatile
MeHg derivatives and it may be of use more widely for GC/C-IRMS analyses of other
compounds where there are similar considerations. Similarly, while the measurement
uncertainty considerations regarding normalisation and derivatization may be rarely applied
This article is protected by copyright. All rights reserved.

for MeHg analysis within environmental samples that are not heavily contaminated with MeHg
at the μg g-1 level, the general approaches will be useful for all applications of GC/C-IRMS
where normalisation or derivatization are applied. Assuming that 20 μg of MeHg can be
extracted from an environmental sample, it is possible to determine the δ13CVPDB value with a
combined standard uncertainty of around 7 ‰. This uncertainty will decrease the more MeHg
is present for analysis, as run-to-run variability in integrated peak areas will be lower for larger
peaks. Where sufficient MeHg can be extracted to apply the method which does not include
derivatization, the achievable measurement uncertainty will also be lower.
ACKNOWLEDGEMENTS
The work described in this paper was funded in part by the UK government Department for
Business, Energy & Industrial Strategy (BEIS) and by the European Commission under the
European Metrology Programme for Innovation and Research (EMPIR) Joint Research
Project: ENV51 “Traceability for Mercury Measurements” (MeTra).
This article is protected by copyright. All rights reserved.

REFERENCES
[1] Futsaeter G, Wilson S. The UNEP global mercury assessment: sources, emissions and
transport.E3S Web of Conferences. Vol. 1. 2013, EDP Sciences.
[2] Directive 2000/60/EC of the European Parliament and of The Council of 23 October 2000
establishing a framework for Community action in the field of water policy. https://eur-
lex.europa.eu/resource.html?uri=cellar:5c835afb-2ec6-4577-bdf8-
756d3d694eeb.0004.02/DOC_1&format=PDF. Accessed February 26, 2019.
[3] UN Treaty Chapter XXVII 17. Minamata Convention on Mercury.
https://treaties.un.org/doc/Treaties/2013/10/20131010%2011-16%20AM/CTC-XXVII-
17.pdf. Accessed February 26, 2019.
[4] Sunderland EM, Mason RP. Human impacts on open ocean mercury concentrations.
Global Biogeochem Cycles. 2007;21(4). doi.org/10.1029/2006GB002876.
[5] Thode HG, Monster J, Dunford HB. Sulphur isotope geochemistry. Geochim Cosmochim
Acta. 1961;25(3):159-174.
[6] Robinson D. δ15N as an integrator of the nitrogen cycle. Trends Ecol Evol. 2001;16(3)153-
162.
[7] Andersen MB, Elliott T, Freymuth H, Sims KWW, Niu Y, Kelley KA. The terrestrial
uranium isotope cycle. Nature. 2015;517(7534):356-359.
[8] Stevenson EI, Hermoso M, Rickaby REM, et al. Controls on stable strontium isotope
fractionation in coccolithophores with implications for the marine Sr cycle. Geochim
Cosmochim Acta. 2014;128: 225-235.
[9] Blum JD, Bergquist BA. Reporting of variations in the natural isotopic composition of
mercury. Anal Bioanal Chem. 2007;388(2):353-359.
This article is protected by copyright. All rights reserved.