Intercomparison measurements of two 33 S-enriched sulfur isotope standards

Despite widespread applications of sulfur isotope mass-independent fractionation (MIF) signals for probing terrestrial and extra-terrestrial environments, there has been no international sulfur isotope reference material available for normalization of D 33 S and D 36 S data. International reference materials to anchor isotope values are useful for interlaboratory data comparisons and are needed to evaluate, e.g. , whether issues exist associated with blanks and mass spectrometry when using di ﬀ erent analytical approaches. We synthesized two sodium sulfate samples enriched in 33 S with di ﬀ erent magnitudes, and termed them S-MIF-1 and S-MIF-2, respectively. The sulfur isotopic compositions of these two samples were measured in ﬁ ve di ﬀ erent laboratories using two distinct techniques to place them on the V-CDT scale for d 34 S and a provisional V-CDT scale for D 33 S and D 36 S. We obtained average d 34 S values of S-MIF-1 ¼ 10.26 (cid:2) 0.22 & and S-MIF-2 ¼ 21.53 (cid:2) 0.26 & (1 s , versus V-CDT). The average D 33 S and D 36 S values of S-MIF-1 were determined to be 9.54 (cid:2) 0.09 & and (cid:3) 0.11 (cid:2) 0.25 & , respectively, while the average D 33 S and D 36 S values of S-MIF-2 are 11.39 (cid:2) 0.08 & and (cid:3) 0.33 (cid:2) 0.13 & (1 s , versus V-CDT). The lack of variation among the interlaboratory isotopic values suggests su ﬃ cient homogeneity of S-MIF-1 and S-MIF-2, especially for D 33 S. Although additional measurements may be needed to ensure the accuracy of the isotopic compositions of S-MIF-1 and S-MIF-2, they can serve as working standards for routine D 33 S analysis to improve data consistency, and have the potential to serve as secondary sulfur isotope reference materials to address issues such as scale contraction/expansion and for normalization and reporting of D 33 S and D 36 S between laboratories. For the same reasons as listed for sulfur isotopes, the same standards were also arti ﬁ cially enriched in 17 O. The calibration is still in progress but ﬁ rst estimations gave D 17 O ¼ 3.3 (cid:2) 0.3 & with unassigned d 18 O.


Introduction
Sulfur has four stable isotopes 32 S, 33 S, 34 S and 36 S with approximate abundances of 94.99%, 0.75%, 4.25% and 0.01%, respectively. The relative abundances of these isotopes in geological materials (e.g., rocks, atmospheric aerosols, water, ice, meteorites, etc.) are affected by different geological, atmospheric, biological, and hydrological processes. Therefore, variations in the relative abundances of sulfur isotopes in a variety of terrestrial and extra-terrestrial materials have the potential to serve as useful tracers of the source and transformation of sulfur in different environments, as well as provide information about their physical and/or chemical conditions. The relative abundances of sulfur isotopes are typically measured as the ratios of the rare isotopes ( 33 S, 34 S and 36 S) to the most abundant isotope, 32 S, and expressed as the delta notation which describes a deviation from a primary isotope reference material: where R represents x S/ 32 S, and x ¼ 33, 34 or 36. The rst principal reference material was troilite from the Canyon Diablo meteorite (Canyon Diablo Troilite -CDT). However, CDT was found to be variable in a prior study, 1 and thus a V-CDT scale was established later by assigning a d 34 S value of À0.3& relative to V-CDT to an internationally distributed silver sulde reference material IAEA-S-1. 2 The assigned value of À0.3& was based on intercomparison measurements from een individual laboratories.
The second historic aspect of sulfur isotope analyses has been on d 34 S. This occurred because of the difficulty of measuring 33 S and 36 S using standard combustion techniques, and a strong mass-dependent correlation between d 33 S, d 34 S and d 36 S that led to the belief that independent information could not be obtained by measuring the two rarest sulfur isotopes. The recognition of mass-independent processes in meteorite samples, [3][4][5] geological samples, [6][7][8][9] and atmospheric and ice-core samples, [10][11][12][13][14][15] as well as the development of new techniques such as the MC-ICP-MS techniques for 33 S measurement 16,17 has made it of interest to have better calibrations over the full range of stable sulfur isotopes. In addition, mass-dependent fractionation processes in the biogeochemical sulfur cycle have also been measured and are known to produce small abundance deviations for 33 S and 36 S from mass-dependent relationships, [18][19][20] and these variations have been shown to be useful in terms of obtaining additional information on the biogeochemical sulfur cycles in, for example, marine environments. [21][22][23] The deviation from massdependent trends was commonly termed mass-independent fractionation (MIF), although some were strictly related to mass-dependent processes, and expressed as the capital delta notation as follows: These sulfur isotope anomalous signals, D 33 S and D 36 S, serve as unique proxies to track both mass dependent and mass independent fractionation processes.
Prior community efforts have established a consensual value for the V-CDT scale on the basis of the d 34 S for IAEA-S-1 in order to normalize d 34 S measurements of different samples in different laboratories, for data comparability and consistency. Other sulfur reference materials, such as IAEA-S-2 and IAEA-S-3, are also routinely used for the same purpose. Provisional assignments of values for D 33 S and D 36 S have been assigned to V-CDT in various studies, 24,25 but a full assignment has yet to be made. Other researchers have resorted to normalizations to IAEA-S-1 (ref. 19) or CDT for D 33 S and D 36 S. [26][27][28] The number of studies reporting D 33 S and D 36 S of terrestrial and extraterrestrial samples 29 has increased tremendously in the past two decades due to interest in the geological and/or environmental information embedded in D 33 S and D 36 S signals. Such an increase has resulted in a need for the establishment of working materials and interlaboratory comparisons that will lay the groundwork for establishing a consensus for the normalization of D 33 S and D 36 S of V-CDT.
Appropriate data normalization, aside from precise and accurate measurements, is necessary to ensure proper interlaboratory data comparison and to reach consensual conclusions according to D 33 S and D 36 S values measured from the same or similar types of samples. 30 Appropriate data normalization is also important for interpretation of small D 33 S and/or D 36 S values (e.g., 0.03&). The D 33 S and D 36 S values are not directly measured, but calculated from the measured d 34 S, d 33 S and d 36 S values as shown in the above equations. The d 34 S, d 33 S and d 36 S values are typically measured with respect to a laboratory working reference gas (i.e., SF 6 ), and then need to be anchored to the V-CDT scale in order to ensure consistent comparison of data among different laboratories. Although consensual d 34 S V-CDT values of international sulfur isotope reference materials have been established, currently there are no consensual d 33 S V-CDT and d 36 S V-CDT values. Here we use the provisional Wing and Farquhar 24 V-CDT calibration of IAEA-S-1 which assigns D 33 S ¼ 0.094& and D 36 S ¼ À0.700& as the values for IAEA-S-1 on the V-CDT scale. These values correspond to d 33 S and d 36 S of À0.061& and À1.27&, respectively. These values along with the community-dened d 34 S value of IAEA-S-1 (À0.300&) are used to normalize multiple sulfur isotope compositions of particular samples to the V-CDT scale, once the working reference gas is calibrated versus IAEA-S-1, or concurrent measurements of IAEA-S-1 are performed.
Given the small D 33 S and D 36 S values measured in, e.g., stratospheric and tropospheric sulfate aerosols, marine Sbearing materials, meteorites and Proterozoic geological samples, small errors, scale contraction, or dri in one-point scale normalization can lead to signicant differences in the derived D 33 S and D 36 S values for such samples. In addition, the mechanism behind the origin of S-MIF in atmospheric sulfate is still a subject of debate, [31][32][33][34][35][36][37] and observations of small negative D 36 S values in atmospheric sulfate possibly associated with combustion processes 14,15 raise further questions on the photochemical origin of S-MIF. Accurate and precise measurements as well as consistent data normalization are also critical in distinguishing the difference between small non-zero D 33 S and D 36 S values produced by mass-dependent fractionation processes in biogeochemical sulfur cycles and non-zero D 33 S and D 36 S values produced by MIF processes, and in further discerning the contributions of different MIF processes. In principle, data normalization can be considerably improved by using two or more points to provide a normalization spanning a wide delta range. To enable proper data normalization and compatible data reporting from different laboratories, secondary reference materials of D 33 S and D 36 S are necessary in addition to IAEA-S-1. The IAEA-S-1 material is used as a primary reference material to scale or anchor the measured data to the V-CDT scale, rather than a physically real calibration standard.
To date, there is no international sulfur isotope reference material enriched in 33 S and/or 36 S available for the purpose of global calibration. Individual laboratories generally report D 33 S and D 36 S values normalized using concurrent IAEA-S-1 measurements, but consensus values of D 33 S and D 36 S for IAEA-S-1 on the V-CDT scale have not been assigned. In this study, we report the sulfur isotopic compositions of two synthesized sodium sulfate samples articially enriched in 33 S with different magnitudes. The data we report are from separate analyses performed at ve different laboratories. We evaluate the interlaboratory variations in the reported values and use the data to assess the potential for further use of these samples as secondary reference materials for D 33 S data normalization. Concomitantly, these samples are also enriched in 17 O for the same reasons as listed for sulfur. The calibration is still in progress and the preliminary result of D 17 O ¼ 3.3 AE 0.3& (1s) is reported only for information purposes. In the following, we will not elaborate more on D 17 O.

Synthesis of samples enriched in 33 S
Two sodium sulfate (Na 2 SO 4 ) samples, namely, S-MIF-1 and S-MIF-2, enriched in 33 S were prepared in the stable isotope laboratory at the University of California, San Diego. We chose Na 2 SO 4 as it is chemically stable, is nontoxic, does not become hydrated, and is widely available and easy to manufacture. For S-MIF-1, 20 g of sulfur powder and 0.00445 g of sulfur-33 powder were weighed and mixed in an agate mortar. For S-MIF-2, 20 g of sulfur powder and 0.0015 g of sulfur-33 powder were weighed and mixed. The composition of the initial mixtures corresponded to D 33 S V-CDT values of $30& and 10&, respectively. The powder mixture was transferred into a crucible, and ignited with a ame. The crucible was then placed in a 4 L airtight glass vessel followed by purging of the glass vessel with a ow of O 2 at a rate of 50 mL min À1 . As a result, SO 2 was produced and carried by O 2 into a NaOH solution (made of 33 mL 17 O-enriched water (D 17 O ¼ 47&), 33 mL NaOH (50% w/ w) and 133 mL of pure water (18.2 MU)), where SO 2 was trapped as Na 2 SO 3 . The S(IV)-containing the NaOH solution was placed in an ice-water bath during the collection process. Aer the combustion was complete, we suspended the ow of O 2 and slowly added 80 mL 30% H 2 O 2 (due to the exothermicity of the reaction) to the NaOH solution in order to oxidize the trapped SO 3 2À to SO 4

2À
. For S-MIF-1, 33 mL H 2 SO 4 was added to dilute D 33 S to $10&, and then a few drops of NaOH were added to adjust to neutral pH. For S-MIF-2, no dilution was made and only a few drops of H 2 SO 4 were added to adjust to neutral pH. The different treatments of S-MIF-1 and S-MIF-2 in this last step were intended to produce Na 2 SO 4 with similar D 33 S but different D 17 O. Both S-MIF-1 and S-MIF-2 were then dried and ground, and the powder was collected and stored. In the end, we recovered $136 g and $55.6 g Na 2 SO 4 for S-MIF-1 and S-MIF-2, respectively.

Isotopic analysis
Aer preparation, S-MIF-1 and S-MIF-2 were circulated to ve laboratories, including the stable isotope laboratory at the Ecole Normale Superieure (ENS) de Lyon, the stable isotope laboratory at the Institut de Physique du Globe de Paris (IPGP), the stable isotope laboratory at the University of Maryland, College Park (UMD), the stable isotope Geo-biology laboratory at the Massachusetts Institute of Technology (MIT) and the stable isotope laboratory at the Tokyo Institute of Technology (TIT). In these laboratories, the isotopic compositions of S-MIF-1 and S-MIF-2 were characterized individually.
In the laboratories of IPGP, UMD, MIT and TIT, S-MIF-1 and S-MIF-2 were analyzed following the conventional SF 6 method. In each laboratory, the Na 2 SO 4 samples were rst reduced to silver sulde (Ag 2 S) using the STrongly Reducing hydrIodic-hypoPhosphorous-hydrochloric acid (STRIP) method 38 or the Kiba reagent 39 method following the standard procedure described in Forrest and Newman. 40 We note that in practice each laboratory uses slightly different reduction techniques for operational convenience. Briey, the collected sulde aer sulfate reduction was converted to silver sulde (Ag 2 S). Aer purication, Ag 2 S was dried, weighed (1-3 mg) and transferred into a small aluminum boat. The aluminum boats were folded and loaded into externally heated nickel reaction tubes. The reaction tubes were evacuated for 0.5-1 hour at $100 C until the desired vacuum was achieved. F 2 (in excess) was then introduced into the reaction tubes to produce SF 6 . The reaction tubes were heated to $250 C and held at this temperature overnight. The produced SF 6 gas was puried rst using a series of cryogenic techniques and then by gas chromatography using helium as the carrier gas. The puried SF 6 was then trapped with liquid nitrogen and transferred under vacuum to a gassource isotope ratio mass spectrometer (Thermo Fisher MAT 253) where its sulfur isotopic composition was analyzed in dualinlet mode. Briey, in each analysis, the sample and reference gas (SF 6 ) were measured one aer another for masses of 127 ( 32 SF 5 + ), 128 ( 33 SF 5 + ), 129 ( 34 SF 5 + ) and 131 ( 36 SF 5 + ). In the end, the measured d 34 S, d 33 S and d 36 S values of the sample were expressed in the delta notation with respect to the reference SF 6 gas.
In the ENS laboratory, S-MIF-1 and S-MIF-2 were analyzed for sulfur isotopic compositions using multi-collector inductively coupled plasma mass spectrometry (MC-ICP-MS). 16 This method allows the measurement of 32 S, 33 S and 34 S at very low sulfur content (180 nmol, but can be as low as 10 nmol sulfur), with a typical reproducibility of AE0.1& for d 34 S and AE0.15& for d 33 S (2s) based on replicate measurements of the in-house Alfa Aesar (AA) standard solution. However, due to the interference of the Ar-36 isotope, 36 S cannot be measured by this method. The chemistry procedure requires a rst step of isolation and purication of sulfur (sulfate) from the sample matrix. In the ENS lab, aer sulfate was isolated with an anion exchange resin (200-400 mesh AG1-X8, in chloride form) and eluted in dilute HNO 3  In Table 1, we list important technical information about sample preparation and analysis in each laboratory. For the IR-MS method in laboratories other than ENS, IAEA-S-1 was measured repeatedly with S-MIF-1 and S-MIF-2. The measurement uncertainties of d 34 S, D 33 S and D 36 S in each laboratory as indicated by repeated measurements of IAEA-S-1 are listed in Table 2. The results of IAEA-S-1 in Table 2 were also used to anchor the sulfur isotopic composition of S-MIF-1 and S-MIF-2 to the V-CDT scale, given its known V-CDT values of d 34 (1)). The working reference gas (i.e., SF 6 ) used in each laboratory possesses different sulfur isotopic compositions, i.e., x S/ 32 S This is because D 33 S and D 36 S are relative values reecting the deviations from the mass-dependent lines, and are not much affected by the scale differences. But when small differences are focused on, they still need to be on the same scale. The means of D 33 S and D 36 S for S-MIF-1 are 9.50 AE 0.08& (1s, N ¼ 33) and 0.11 AE 0.55& (N ¼ 26), respectively, and those for S-MIF-2 are 11.36 AE 0.08& (N ¼ 37) and À0.03 AE 0.54& (N ¼ 23), respectively.
In order to better compare the data from different laboratories and to evaluate the isotopic variability of S-MIF-1 and S-MIF-2, these values need to be anchored on the same scale (i.e., the V-CDT scale). For measurements done at IPGP, UMD, MIT and TIT, the international reference material IAEA-S-1 was also measured at the same time with S-MIF-1 and S-MIF-2, and the results are reported with respect to laboratory working reference gases and listed in Table 2. IAEA-S-1 has a consensual d 34 S V-CDT value of À0.300&, but its d 33  For measurements done at the ENS, the d 34 S V-CDT value of the working standard solution AA was 4.86 AE 0.14& as calibrated using international reference materials. According to the mass-dependent fractionation law, its d 33 S V-CDT value was calculated to be À2.52&. These values were then used to convert the raw d 34

Characterization of isotopic variability
The isotopic compositions of S-MIF-1 and S-MIF-2 on the V-CDT scale are listed in Tables 5 and 6, respectively. As shown in these tables, the d 34 S, d 33 S and d 36 S V-CDT values of S-MIF-1 (or S-MIF-2) from different laboratories are identical. This is as expected because now they are all on the same scale (i.e., x R ref is the same in eqn (1)), and the small or negligible difference is due to measurement uncertainties. Considering all data from the ve laboratories, the uncertainties of d 34 S V-CDT for S-MIF-1 and S-MIF-2 are AE0.22 and AE0.26& (1s), respectively, comparable to or better than those of international sulfur reference materials (e.g., AE0.2& for IAEA-SO-6 and AE0.4& for NBS-127). Regarding the D 33 S V-CDT and D 36 S V-CDT , it can be noted that they are slightly different from those calculated from the raw d 34 S, d 33 S and d 36 S data (Tables 3 and 4). These small differences may be considered to be negligible regardless of the scale when the capital delta values are large, e.g., in the case of D 33 S of S-MIF-1, it is 9.50 AE 0.08& and 9.54 AE 0.09& before and aer anchoring on the V-CDT scale. But when small capital values are the subject of interest, the difference becomes important, e.g., in the case of D 36 S of S-MIF-1, it is 0.11 AE 0.55& and À0.14 AE 0.25& before and aer anchoring on the V-CDT scale.
Overall, the uncertainties of D 33 S V-CDT for S-MIF-1 and S-MIF-2 are AE0.09 and AE0.08&, respectively. For D 36 S V-CDT , the uncertainties are AE0.25 and AE0.13& for S-MIF-1 and S-MIF-2, respectively. The relatively large uncertainties in D 36 S V-CDT are in part due to the extremely low abundance of 36 S, which makes its accurate measurement more difficult, but maybe more likely come from an isobaric interference on the 131 peak during IRMS measurements probably due to C 3 F 5 + . 44 In addition, the uncertainties of the small delta values are larger than those of the capital delta values as always observed, and the best precision is obtained for D 33 S V-CDT . This is because the uncertainties of the small delta values are in part from sample preparation and conversion processes, but these processes would only induce mass-dependent fractionation and thus won't signicantly affect the capital delta values.
It can be noted that the V-CDT values of S-MIF-1 and S-MIF-2 derived from the measurement (one analysis available) done at MIT are slightly different from those done at other labs, i.e., they are at the low end for D 33 S V-CDT but the high end for D 36 S V-CDT among all data derived. However, this is only one measurement and its involvement in the global means did not affect the results (the means and standard deviations) much. S

Summary
There is a compelling need to have international sulfur and oxygen isotope reference materials enriched in 33  samples. This is important not only in terms of data comparisons within a laboratory and/or among different laboratories, but also regarding the differentiation of small D 33 S and D 36 S values from mass-dependent and mass independent fractionation processes. Currently there is only one international sulfur reference material, IAEA-S-1, with established D 33 S and D 36 S values (0.094 AE 0.004&, and À0.7 AE 0.1&, respectively) reported on the V-CDT scale, 24 but IAEA-S-1 can be regarded more as a primary reference material. There are no reference materials with apparently large anomalies in D 33 S and D 36 S. In this report, we synthesized two sodium sulfate samples, S-MIF-1 and S-MIF-2, articially enriched in 33 S and 17 O. The preliminary assessments of their oxygen isotopic compositions yielded D 17 O ¼ 3.3 AE 0.3&. The sulfur isotopic compositions of these two samples were characterized at ve different laboratories using two distinct methods, the conventional IR-MS method and the newly developed MC-ICP-MS method. 16 Except for one data point from the MIT laboratory, results from the other four laboratories are in good consistency. The good precision of these isotopic values indicates isotopic homogeneity of S-MIF-1 and S-MIF-2. Although further calibration efforts may be needed to improve the accuracy of D 33 S V-CDT assessments of S-MIF-1 and S-MIF-2, their current values can be adopted to calibrate D 33 S measurements. In particular, mixing them with other sulfur reference materials with zero D 33 S such as IAEA-SO-5 and IAEA-SO-6 should generate working standards with different D 33 S values, which can be used to establish a calibration curve spanning a large D 33 S range (e.g., 0 to 11&) for better data normalization. These standards are available for the community and can be requested on demand from Joel Savarino.

Conflicts of interest
There are no conicts to declare.