Improving the routine analysis of siderite for δ13C and δ18O in environmental change research

Rationale The carbon (δ13C) and oxygen (δ18O) isotope composition of siderite (FeCO3) is used widely to understand and quantify geochemical processes to reconstruct past climate and environmental change. However, few laboratories follow precisely the same protocol for the preparation and analysis of siderite‐bearing materials, which combined with the absence of international reference materials and mineral‐specific acid fractionation factors, leads potentially to significant differences in isotope data generated by individual laboratories. Here we examine procedures for the isotope analysis of siderite and discuss factors potentially contributing to inconsistencies in sample measurement data. Methods Isotope analysis of siderite is first assessed using similar versions of the classical off‐line, sealed vessel acid digestion method by comparing data sets obtained from intercomparison materials measured at two participating laboratories. We then compare data from the classical method against those generated using an automated preparation technique using data produced from an independent set of test materials. Results Measurement of siderite δ13C is generally both repeatable and reproducible, but measurement of δ18O may be subject to large (~1‰), method‐dependent bias for siderite reacted at differing temperatures (70°C and 100°C) under classical and automated CO2 preparation conditions. The potential for poor oxygen isotope measurement reproducibility is amplified by local differences in sample preparation protocols and procedures used to calibrate measurement data to international reference scales. Conclusions We offer suggestions for improving the repeatability and reproducibility of δ13C and δ18O analysis on siderite. The challenge of producing consistent isotope data from siderite can only be resolved by ensuring the availability of siderite reference materials to facilitate identical treatment as a basis for minimising method‐dependent contributions to data inconsistency between laboratories.

Rationale: The carbon (δ 13 C) and oxygen (δ 18 O) isotope composition of siderite (FeCO 3 ) is used widely to understand and quantify geochemical processes to reconstruct past climate and environmental change. However, few laboratories follow precisely the same protocol for the preparation and analysis of sideritebearing materials, which combined with the absence of international reference materials and mineral-specific acid fractionation factors, leads potentially to significant differences in isotope data generated by individual laboratories. Here we examine procedures for the isotope analysis of siderite and discuss factors potentially contributing to inconsistencies in sample measurement data.
Methods: Isotope analysis of siderite is first assessed using similar versions of the classical off-line, sealed vessel acid digestion method by comparing data sets obtained from intercomparison materials measured at two participating laboratories.
We then compare data from the classical method against those generated using an automated preparation technique using data produced from an independent set of test materials.
Results: Measurement of siderite δ 13 C is generally both repeatable and reproducible, but measurement of δ 18 O may be subject to large ($1‰), method-dependent bias for siderite reacted at differing temperatures (70 C and 100 C) under classical and automated CO 2 preparation conditions. The potential for poor oxygen isotope measurement reproducibility is amplified by local differences in sample preparation protocols and procedures used to calibrate measurement data to international reference scales.
Conclusions: We offer suggestions for improving the repeatability and reproducibility of δ 13 C and δ 18 O analysis on siderite. The challenge of producing consistent isotope data from siderite can only be resolved by ensuring the availability of siderite reference materials to facilitate identical treatment as a basis for minimising method-dependent contributions to data inconsistency between laboratories.

| INTRODUCTION
The stable isotope composition of carbonate sediments and fossils provides one of the most important geochemical proxies for environmental change research, 1,2 and has been used as an indicator of palaeoclimate since the seminal work of McCrea. 3 Whilst the mineral used in the majority of studies is calcite (CaCO 3 ), many carbonate-bearing sediments do not contain calcite (endogenic or authigenic) but instead preserve other carbonates, including siderite.
Siderite is a naturally occurring iron-rich carbonate mineral (FeCO 3 ) found in freshwater and marine sediments, [4][5][6][7][8] hydrothermal and metasomatic iron mineralisation, 9,10 and carbonatite magmatism. 11 In addition, siderite may be an important mineral on Mars, where it is implicated in the decarbonisation of the Martian atmosphere. [12][13][14] Siderite can form during early diagenesis from ferrous iron [Fe] 2+ and carbonate [CO 3 ] 2À ions dissolved in marine, brackish, and fresh porewaters in both marine and terrestrial environments. 15 Dissolved [Fe] 2+ only accumulates to significant levels in anoxic waters, under the absence of O 2 where oxidisation to [Fe] 3+ is arrested, and with negligible sulphate concentrations to prevent the preferential formation of hydrogen sulphide and subsequent precipitation of iron sulphide. 16 Siderite is therefore precipitated typically under strongly reducing conditions with elevated CO 2 partial pressure, potentially accompanied by methane formation, and is generated more readily in non-marine sediments due to the lower sulphate and higher organic matter concentrations of freshwater systems, 17 for example due to stratified water columns in biologically productive lacustrine settings. 7 As for calcite, the stable isotope composition of siderite can be a sensitive recorder of information about the environment of its formation, given the oxygen isotope composition (δ 18 O) of siderite is dependent on the temperature and δ 18 O of the water from which the mineral precipitated. [18][19][20] Based on this relationship, siderites have been used as palaeoclimate and palaeoenvironmental indicators. 8,[21][22][23][24][25] Many laboratories measure the carbon isotope composition (δ 13 C) and δ 18 O of siderite, but methods are generally laboratoryspecific, with preparation and analyses carried out under varying experimental conditions. The classical method for determination of δ 13 C and δ 18 O of carbonate minerals developed originally by McCrea 3 is based on the off-line preparation of CO 2 by reaction with hyperconcentrated (anhydrous) phosphoric acid (H 3 PO 4 ≥ 100%) in sealed vessels and analysis using isotope ratio mass spectrometry. A simplified description of reaction chemistry is summarised in equations 1 and 2. 26,27 where M is the divalent cation.
In the phosphoric acid reaction, only two of the available carbonate oxygen atoms are transferred to product CO 2 for isotope analysis. The partitioning and equilibration of oxygen isotopes between CO 2 , H 2 O, and a diverse range of potential oxygen-bearing phosphate species is accompanied by a mineral-specific and temperature-dependent kinetic isotope effect that is corrected for by applying an appropriate, empirically derived, "acid fractionation factor" α 28-31 : Here we examine potential method-dependent contributions to uncertainty in the analysis of siderite by (1)  On the basis of our results we discuss potential sources of measurement inconsistency and offer recommendations to improve the routine isotope analysis of siderite.

| Geochemical characterisation of intercomparison materials
Four siderite-bearing intercomparison materials were prepared from natural rock samples at the BGS (Table 1 and Appendix S1). The samples were hand crushed using an agate mortar and pestle, and the resultant powders were stored in screw-cap sealed glass vials prior to analysis. An aliquot of each siderite was further ground and split into 10-63 μm and <10 μm grain size fractions by milling with agate spheres and then micronizing, respectively, to determine the major elemental composition and mineralogy of each intercomparison material using X-ray diffraction (XRD) and X-ray fluorescence (XRF) at

| Preparation and analysis of CO 2
The methods used to prepare and measure CO 2  3 | RESULTS

| Mineralogy and geochemistry
A summary of the mineralogical composition of each intercomparison material determined by XRD and a full tabulation of XRF and TC elemental analysis data is provided in Appendix S1. Quantitative estimates of the mineralogical composition of intercomparison materials were determined on the basis of geochemical (XRF, TC) data using a normative procedure in which key elements are assigned to specific minerals identified from XRD patterns (Table 3). Available information on the mineralogical and geochemical composition of ETH test siderites are reported in Fernandez et al. 33 All materials were identified as pure siderite with varying, but elevated (X FeCO3 < 0.8) quantities of Mg, Ca, and Mn substitution.

| Carbon and oxygen isotopic composition
Individual δ 13 C and δ 18 O data sets produced by the BGS and UoL laboratories are reported in Appendix S1. The siderite intercomparison data are illustrated as box plots in Figures 1 and 2, and a statistically reduced summary is provided in Tables 4 and 5  open-vessel reactions at 100 C data sets is associated with a single extreme value (see Appendix S1). In two cases these extreme values were recognised as potential statistical outliers (p < 0.05) on the basis of a two-tailed Grubb's test, 54 but all available data were retained for comparison in this study.

| Siderite reaction rate
The reaction rate of the Ivigtut intercomparison siderite at 100 C was investigated at UoL using a freshly crushed (<63 μm) cleavage fragment and the classical acid digestion method. The results are summarised in Figure 3. Rapid dissolution, in which approximately 90% of the total available sample CO 2 was released, occurred within the first 4 h. The remaining 10% was released more slowly and the reaction appeared to be complete after 16 h.  Table 4).
Two δ 13 C data sets from BGS (MR31556, Ivigtut) are characterised by positively skewed distributions and fail a Shapiro-Wilk normal distribution test (Appendix S1); we attribute this to the relatively small number of replicates (n ≤ 12) in sample populations. All δ 13 C data sets fail a Levene comparison of variance test (H 0 : σ 2 BGS = σ 2 UoL ) at the 0.05 level (Appendix S1), showing that the spread of data differs significantly between laboratories. Allowing for the small sample population and the tendency towards bimodality of the MR31556 and Ivigtut data sets, only MR15423 fails a two-tailed comparison of sample means t-test (H0: μ BGS = μ UoL │σ 2 BGS ≠ σ 2 UoL ) at the 0.05 significance level (Table 4) and the mean δ 13 C values measured in each laboratory can be considered effectively identical.
In contrast to δ 13 C values, differences (BGS À UoL) in mean δ 18 O values exhibit a systematic offset of approximately À0.1‰. (Figure 2 and Table 5). All data sets pass a normal distribution test (Appendix S1) and only the data sets for MR17327 fail a comparison of variance test (H 0 : σ 2 BGS = σ 2 UoL ) at the 0.05 level (Appendix S1). However, all data sets fail a two-tailed comparison of sample means t-test (H0: μ BGS = μ UoL │σ 2 BGS ≠ σ 2 UoL ) at the 0.05 significance level (Table 5), consistent with the occurrence of a small laboratory-dependent measurement bias. Although both laboratories follow similar analytical protocols, and we tested specific aspects of the method, the cause of this bias has not been identified.

| ETH test materials
As sample populations and variances of individual BGS, UoL, and ETH data sets differ widely, it is unrealistic to undertake a quantitative comparative statistical assessment. Consequently, only qualitative comparisons are discussed.
Mean δ 13 C values are in reasonable agreement and essentially independent of analytical method (σ ≤ 0.10‰). Maximum and minimum mean values for each material differ by <0.24‰ ( Figure 6 and Table 6) and in at least two cases much of the discrepancy can be

| Sample preparation and storage
Pre-analysis sample preparation, storage, and organic contamination removal procedures are all recognised factors potentially contributing to discrepancy in interlaboratory carbonate stable isotope measurement. 40 In addition to issues identified for carbonates in general, we draw attention to two aspects that may be specific for samples of siderite.
Qi et al. 55 have shown that the isotope composition of finely crushed carbonate materials is susceptible to modification during prolonged contact with humid air, even when stored in apparently tightly sealed screw-cap vials. In addition, powdered carbonates that contain elevated concentrations of transition metals in a low It is generally recommended that such materials are stored in a desiccator to minimise isotope exchange reactions mediated by water molecules absorbed on particle surfaces. To minimise these potential effects siderite samples should be analysed soon after crushing. For long-term storage powders may need to be sealed in evacuated glass ampoules.  (Figures 3 and 5). Although no grain size distribution data are available for any of these yield experiments, tailing of CO 2 yield curves is attributable to the extended time required to react a residual fraction of coarser particles and methoddependent differences in the desorption rate of CO 2 temporarily (or even permanently) absorbed within the phosphoric acid.
Differences in the rate of siderite dissolution at any specified reaction temperature are predicted to be dependent on a range of sample-specific (particle size, sample mass, carbonate chemistry) and  During the initial stages of reaction, a dynamic, isotope disequilibrium state should persist until the siderite reaction has terminated. However, it is also possible that the entire oxygen isotope exchange system will continue to evolve towards a steady state, dependent on the kinetics of individual exchange reactions, over time periods well beyond the point at which the carbonate reaction has actually ceased. 29,38 The direction and magnitude of these variables on the evolving δ 18 O of CO 2 will ultimately be dependent on the method of CO 2 production and extraction, with different outcomes arising from differences in system variables including the isotope composition of

| Siderite reference materials
The use of globally available, matrix-specific reference materials to improve the reproducibility of isotope data is well established. 36

| Recommendations for isotope analysis of siderite
We make the following suggestions for improving the repeatability and reproducibility of δ 13 C and δ 18 O values. Given the proliferation of automated gas preparation systems (based on multiple combinations of gas preparation conditions) and differing mass spectrometry currently in use it is impossible to establish a standard protocol for routine analysis. The challenge of producing consistent data sets can only be resolved practically by the global availability of siderite reference materials and the reliance on identical treatment as a basis for reducing method-dependent contributions to local measurement bias.
• To minimise reaction times a sample particle size ≤200 μm is necessary. However, fine-grained siderite powders may be subject to rapid oxidation in contact with moist air. Storage in a desiccator is a minimum requirement and analysis soon after crushing is recommended. The use of water-based oxidising agents to remove organic contamination should be avoided. • No attempt has been made to assess the effects of differences in sample mass and phosphoric acid volume. Maintaining a consistent, laboratory-defined ratio between these variables is advisable. It is noted that mixing larger quantities of fine-grained siderite powder with smaller volumes of phosphoric acid can lead to solidification issues potentially arresting the reaction before completion and trapping a portion of the liberated CO 2 .
• Elevated reaction temperatures are required for efficient sample processing. A temperature of 100 C offers a compromise between reaction rate limitations and the potential negative effects from the high vapour pressure of H 2 O in the sealedvessel headspace.
• The integrated effects of temperature, particle size, mineral chemistry, and concentration of phosphoric acid on siderite reaction rate are poorly defined. We recommend a minimum reaction time of 24 h at 100 C for sealed-vessel procedures. The use of a consistent reaction time should minimise effects related to possible time-temperature-sensitive oxygen isotope equilibration between CO 2 -H 2 O-H 3 PO 4 .
• When using a generic α CO2Àsiderite and discounting impurities, this will affect the accuracy of the data for specific samples but not the precision achieved on the data. In the absence of rigorous determination of α CO2Àsiderite , we recommend the continuing use of the fractionation factors of Rosenbaum and Sheppard. 29 Extrapolation to lower temperatures (<<100 C) should be avoided.
Possible method-dependent differences of α CO2Àsiderite between sealed-and open-vessel procedures reported by Fernandez et al. 33 have not been confirmed in this study and the use of methodspecific fractionation factors requires validation.
• The application of the principle of identical treatment, using published oxygen isotope fractionation factors (α CO2Àcarbonate ) for siderite samples and calcite calibration materials, which may not be internally consistent at elevated reaction temperatures, provides the only available procedure for automated determination of the isotope composition of siderite but may result in poorer measurement reproducibility. Normalisation to a defined siderite-specific scale (relative to VPDB), based on reliable reference materials, should mitigate local, methoddependent bias across the range of manual and automated gas preparation and measurement procedures currently in use. However, poorly defined uncertainties associated with existing estimates of α CO2Àsiderite and the effects of siderite chemistry on α CO2Àsiderite remain.