USGS44, a new high‐purity calcium carbonate reference material for δ 13C measurements

Rationale The stable carbon isotopic (δ 13C) reference material (RM) LSVEC Li2CO3 has been found to be unsuitable for δ 13C standardization work because its δ 13C value increases with exposure to atmospheric CO2. A new CaCO3 RM, USGS44, has been prepared to alleviate this situation. Methods USGS44 was prepared from 8 kg of Merck high‐purity CaCO3. Two sets of δ 13C values of USGS44 were determined. The first set of values was determined by online combustion, continuous‐flow (CF) isotope‐ratio mass spectrometry (IRMS) of NBS 19 CaCO3 (δ 13CVPDB = +1.95 milliurey (mUr) exactly, where mUr = 0.001 = 1‰), and LSVEC Li2CO3 (δ 13CVPDB = −46.6 mUr exactly), and normalized to the two‐anchor δ 13CVPDB‐LSVEC isotope‐delta scale. The second set of values was obtained by dual‐inlet (DI)‐IRMS of CO2 evolved by reaction of H3PO4 with carbonates, corrected for cross contamination, and normalized to the single‐anchor δ 13CVPDB scale. Results USGS44 is stable and isotopically homogeneous to within 0.02 mUr in 100‐μg amounts. It has a δ 13CVPDB‐LSVEC value of −42.21 ± 0.05 mUr. Single‐anchor δ 13CVPDB values of −42.08 ± 0.01 and −41.99 ± 0.02 mUr were determined by DI‐IRMS with corrections for cross contamination. Conclusions The new high‐purity, well‐homogenized calcium carbonate isotopic reference material USGS44 is stable and has a δ 13CVPDB‐LSVEC value of −42.21 ± 0.05 mUr for both EA/IRMS and DI‐IRMS measurements. As a carbonate relatively depleted in 13C, it is intended for daily use as a secondary isotopic reference material to normalize stable carbon isotope delta measurements to the δ 13CVPDB‐LSVEC scale. It is useful in quantifying drift with time, determining mass‐dependent isotopic fractionation (linearity correction), and adjusting isotope‐ratio‐scale contraction. Due to its fine grain size (smaller than 63 μm), it is not suitable as a δ 18O reference material. A δ 13CVPDB‐LSVEC value of −29.99 ± 0.05 mUr was determined for NBS 22 oil.


| INTRODUCTION
High accuracy measurements of stable carbon isotope ratios (δ 13 C values) in naturally occurring materials are necessary in an increasing number of fields, including oceanography, atmospheric sciences, biology, paleoclimatology, geology, environmental sciences, food and drug authentication, and forensic applications. To achieve high-quality δ 13 C analysis, isotopic reference materials (RMs) are required. In the past several decades, the international isotopic RMs NBS 18, NBS 19, NBS 22, LSVEC, IAEA-CO-1, IAEA-CO-8, IAEA-CO-9, and IAEA-603 have been gradually introduced to the isotope community and used for the determination of δ 13 C values of carbon-bearing materials. [1][2][3] In 1985, the primary recommendation of a Consultants' Group Meeting of the International Atomic Energy Agency (IAEA) 4 was that a new Vienna Peedee Belemnite (VPDB) δ 13 C scale be established with NBS 19 carbonate assigned the value of +1.95 milliurey (mUr) exactly as its single anchor, where 1 mUr = 0.001 = 1‰. 1,5 Implementation of this recommendation improved consistency among δ 13 C measurements. 6 Recognizing that two-point normalization of the δ 2 H and δ 18 O scales substantially improved agreement among laboratories, 7 the IAEA convened a panel in 2004 to review stable carbon isotopic RMs and to recommend a second RM for two-point normalization of the δ 13 C scale. Based on high-precision isotope-ratio mass spectrometry (IRMS), 8,9 a consensus value of −46.6 mUr exactly was assigned to LSVEC lithium carbonate. 10,11 The results (  10,11 The adoption of two-point normalization improved the standard uncertainties of δ 13 C RMs significantly compared with previously assessed uncertainties, as demonstrated in Figure 1 of Coplen et al. 10 Since then, determinations of δ 13 C values of most new secondary RMs for forensic, environmental, paleontological, and atmospheric applications [12][13][14][15][16][17][18][19][20]  Gaithersburg, MD, USA). 21,22 Subsequently, this observation was confirmed by Qi et al. 14 Thus, LSVEC no longer meets minimum requirements for use as a δ 13 C RM, particularly as a scale anchor, and IUPAC has advised against its use as a δ 13 C RM. 23  2 | METHODS

| Preparation of USGS44
Sixteen bottles of high-purity calcium carbonate powder with a total mass of 8 kg were purchased from Merck (Darmstadt, Germany). To ensure isotopic homogeneity of the RM, the following steps were carried out (as shown in Figure 1). First, approximately 20 g of material was removed from each of these sixteen 500-g bottles, combined, and passed through a 170-mesh (88 μm) stainless steel sieve with an AS200 sieve shaker (Retsch, Newtown, PA, USA) to homogenize the material.
The very small amount of material larger than 88 μm was discarded.
The sieved material was divided and collected in four 4-L glass containers. The same steps were repeated until all materials from the original 16 bottles were combined and either passed through the 88-μm sieve or were discarded after not passing through the sieve.  14 We also noticed that δ 13 C values of LSVEC can be significantly more negative when analyzed using the classical acid digestion method than the values obtained using the elemental analyzer (EA) technique. The observation has not found a satisfactory explanation so far and warrants further experimental investigations. NBS 22 oil, which was anchored to LSVEC, is used as an anchor in this study. 10,11 The use of NBS 22 oil, which was crimp-sealed in silver tubes, 28 made it possible to measure δ 13 C values of USGS44 directly on an EA connected to an isotope-ratio mass spectrometer following the principle of identical treatment. 29 We are aware that IAEA-603 exhibits inhomogeneities at microgram analysis 30 ; however, this problem does not affect the current study due to the relatively large sample amounts (about 0.2-40 mg) used.

| Online combustion continuous-flow IRMS
At the RSIL, the methods used for online δ 13 C analysis are similar to the procedures and techniques used previously for determination of δ 13 C values of secondary δ 13 C RMs. 14

| GasBench and MultiCarb
At the RSIL, a GasBench II gas preparation and introduction system "dry" CO 2 was then heated at −60 C and "focused" in a second cold finger at −160 C for 5 min. The resulting gas was released in a fixed volume and the pressure of the "monitoring gas" was equilibrated with that of the sample. The "monitoring gas" is a Jackson Dome CO 2 with an approximate δ 13 , was fitted to all datasets with the value of y(0) being taken as the true delta value "δ T ". Costech EA, as specified above, was used. One aliquot containing 12 μg carbon (0.10 mg of calcium carbonate) from each of 27 fractions was analyzed. In this sequence, NBS 19 was analyzed at the beginning, middle, and end as a control sample, and IAEA-603 was also analyzed. The measured δ 13 C values of the 0.10-mg samples of USGS44 are presented in Table 3 and they demonstrate that the data quality is comparable with that of NBS 19 and IAEA-603. The overall standard deviation of 0.07 mUr from the nine bottles with 26 analyses at 0.10-mg of USGS44 is slightly higher than 0.05 mUr from 0.70-mg analyses of USGS44, and this is thought to be caused by a variable carbon blank from the tin capsules. The average δ 13 C value of −40.08 mUr from USGS44 in Table 3 is the result of singlepoint normalization against NBS 19.
Considering the need for high-precision δ 13 C analysis to normalize small samples, such as in the analysis of foraminifera 39,40 and basalt, 41 carbonate RMs must be well homogenized. 30 Several studies have demonstrated that using an isotope-ratio mass spectrometer with a GasBench can achieve high-precision δ 13 C analyses, 38,[40][41][42][43][44][45] and methods for analyzing microgram masses of carbonate have been in use for decades. 38,41,43 For this reason, we carried out homogeneity tests with a GasBench as specified above at the RSIL and a MultiCarb system at Geotop. At the RSIL, one sample from the middle of each of the nine bottles (total of 9 vials) of USGS44 was analyzed at 0.1-mg mass level along with NBS 19 and IAEA-603. A 0.02-mUr reproducibility (Table 4) of USGS44 indicates that this material is well homogenized. A 0.05-mUr reproducibility obtained by the MultiCarb system (Table 4) from a randomly selected USGS44 vial with a sample mass of 0.1 to 0.2 mg also confirms that USGS44 is isotopically homogeneous. The average δ 13 C value of −41.94 mUr from USGS44 in Table 4 also is the result of single-point normalization with NBS 19.
The isotopic homogeneity of USGS44 was also evaluated in comparison with NBS 19 using the data from routine sample analysis with the GasBench at the RSIL. Figure 3 shows the data quality from  (Table 3) at the 0.10-mg level, the higher uncertainty of USGS44 obtained with the GasBench probably does not reflect the true material homogeneity, but rather the uncertainty of the analytical method. We suspect that small variable amounts of atmospheric CO 2 (δ 13 C VPDB value $ −8 mUr) were introduced into sample vials when purging the samples, which affects the δ 13 C value of USGS44 (δ 13 C = −42.21 mUr) more than that of NBS 19 (+1.95 mUr).

| δ 13 C stability evaluation
To ensure that USGS44 calcium carbonate is a stable material and that its δ 13 C value does not change when the material is exposed to a humid environment, 14,21,22 a CO 2 equilibration test 14 like that carried out with LSVEC was performed with USGS44. Two Merck CaCO 3 samples (with different lot numbers) were selected for this test. One was a 2-g vial of CaCO 3 , and another was a large bottle containing 500 g of CaCO 3 , which was the candidate material for USGS44. Three aliquots of about 1-g of each material were loaded into an 8-L glass desiccator. For comparison, a set of LSVEC samples was also placed in the CO 2 equilibration desiccator along with CaCO 3 . A vial of water was also placed inside the desiccator to ensure a humid environment.
The desiccator was evacuated and approximately 300 μmol of CO 2 (δ 13 C = −4 mUr) was introduced into the desiccator. After 7 days at ambient temperature, the samples were removed from the desiccator and dried in a vacuum oven at 40 C for 5 h. Comparison measurements between original samples and samples that had been equilibrated with CO 2 were made in the same analytical sequence.
The measured δ 13 C values are shown in Table 5, and demonstrate that LSVEC reacted with CO 2 and its δ 13 C value increased by 1.01 mUr. This confirms previous observations that the δ 13 C value of LSVEC is not stable. 14,21,22 There was no evidence of reaction or exchange between CO 2 and the two Merck CaCO 3 materials, which demonstrates that USGS44 is stable and acceptable for use as a δ 13 C RM.

| Evaluation of carbon blanks
At the RSIL, the carbon blanks from both the tin capsule and the glass filter were carefully evaluated against USGS40 L-glutamic acid. Six to eight 5 × 3.5-mm tin capsules were folded together to act as one sample to produce a substantial CO 2 peak so that the carbon blank and the δ 13 C value of the blank could be determined accurately. A δ 13 C VPDB value of −26.0 mUr, normalized to USGS40, was obtained for the blank of the tin capsules. The carbon blank in each capsule was about 1 μg. The carbon blank is thought to be a byproduct of mineral oil used in the production of the tin capsules, causing the blanks to be similar within the same batch of capsules. Using the same method, six 1.5 × 1.5-mm baked glass filters were combined to act as one sample. The CO 2 peak from the glass filters was too small to  Table 7 was performed using the Monte Carlo method 47  between LSVEC and USGS44 from data in Table 7.  Table 8 (Appendix C and Appendix D, supporting information) was performed using the same approach as described above for Table 7.
Further discussion about NBS 22 can be found in section 3.7.
A concern about the impact of incorrectly assigning the δ 18 O value of the reference injection gas in Isodat arose during the project.
Does it make any difference whether the reference injection gas is assigned as 0 mUr or +23 mUr (or −23 mUr)? We confirmed that as long as one normalizes the δ 13 C measurements with two anchors, the impact upon the normalized δ 13 C VPDB value of the assigned δ 18 O of reference injection CO 2 is insignificant (<0.01 mUr).

| The δ 13 C VPDB values from dual-inlet measurements
Ideally, an accurate δ 13 C VPDB determination of USGS44 should have been carried out with two-point normalization. However, a second scale anchor that is independent of LSVEC currently does not exist.
To overcome this deficiency, the best method to obtain an accurate  (Table 9 and Figure 5).    49 The codes are given in Appendix A (supporting information). The ± 0.069 (and ± 0.054) are standard uncertainties and can be expanded by multiplication by 2 to obtain 95% uncertainty bands.   Table 9). The individual δ 13 C VPDB measurements are shown in Figure 5, where the significant variation of the mean values over the years can be seen. This variation points to the limitation of the accuracy achieved for the correction for the various scale contraction contributions that are mentioned above.   (Table 9 and Figure 5). At BGC-IsoLab, the δ 13 C VPDB value of the CO 2 prepared by CIO from USGS44 is more negative by 0.016 mUr than their average value of −42.085 ± 0.008 mUr (Table 9). At CIO, the δ 13 C VPDB value of the CO 2 prepared by BGC-IsoLab from USGS44 is more positive by 0.042 mUr than their average value of −41.992 ± 0.022 mUr ( Table 9). The cause of the differences among these values is unknown. We conclude that the δ 13 C VPDB and δ 18 O VPDB differences between CIO and BGC-IsoLab are not caused by the carbonate treatment to generate CO 2 but suggest a scale realization or instrument problem. As a rule of thumb, the more stretched scale, and thus the one producing more negative results for materials like USGS44, is more likely to be the right one, but that is only true for scales prior to correction. Obviously, there is always the possibility that one stretches the scale by too much. In this case, both groups have gone to considerable lengths to try to produce an isotope-delta scale in which a milliurey (‰) truly represents a milliurey, but we are confronted with a difference that we deem beyond our estimated scale realization uncertainty. In Figure 5, it is clear that BGC-IsoLab has been able to perform scale contraction correction with higher precision than CIO over the years, which, however, does not necessarily imply that that correction is more accurate.

| Comparison of δ 13 C VPDB measurements by DI-IRMS and EA/IRMS
In this study, the EA/IRMS measurements of USGS44 give a δ 13    A comparison of the normalized δ 13 C VPDB values of USGS44 obtained by EA (Table 3) and by GasBench/MultiCarb analysis (Table 4) results in an interesting observation. Although the δ 13 C VPDB values in Tables 3 and 4 were normalized to NBS 19, the average δ 13 C VPDB value for USGS44 of −40.08 ± 0.07 mUr from EA measurement is 1.86 mUr more positive than that of −41.94 ± 0.22 mUr from GasBench/ MultiCarb measurements. The value from the GasBench/MultiCarb analysis is nearer the average values reported in Table 9 of −42.085 ± 0.008 and −41.992 ± 0.022 mUr determined by DI-IRMS measurements. This indicates the importance of correcting for the carbon blank and of using two-point normalization for the stable carbon isotope-delta scale when an EA is used to combust samples.
The δ 13 C scale contraction is substantially less when analyzing carbonates with a GasBench or MultiCarb system.

| SUMMARY AND CONCLUSIONS
The new high-purity, well-homogenized calcium carbonate δ 13 C reference material USGS44 is stable and has a δ 13 C VPDB-LSVEC two- this conundrum. Therefore, we recommend continued use of the twoanchor VPDB-LSVEC scale, but with realization of the δ 13 C VPDB-LSVEC scale using RMs other than LSVEC Li 2 CO 3 , whose use is to be deprecated for δ 13