Potential Reference Materials for Hematite Oxygen Isotope Analysis

Hematite is a potential mineral for reconstructing the oxygen isotope composition and paleotemperature of paleowater. A highly accurate analysis of oxygen isotopes is essential. However, relative to other oxygenated minerals, we lack hematite reference materials that allow for internationally comparable analyses between different laboratories. To address this issue, we attempted to perform bulk rock oxygen isotope analysis on five hematite reference materials (GBW07223a, GBW07825, YSBC28740-95, YSBC28756-2008, Harvard 92649). Meanwhile, the oxygen isotope ratios of iron oxides (GBW07223a, GBW07825, YSBC28740-95, YSBC28756-2008) were obtained by mass balance involving other oxygen-bearing minerals such as quartz and silicates. In addition, the oxygen isotope ratios of iron oxides in an oolitic hematite (ca. 1.65 billion years ago) are consistent with the results of previous analyses of this class of minerals.

To date, the oxygen isotope of a large number of hematite samples has been analyzed by the laser fluorination method [23], with a few laboratories testing for hematite oxygen isotopes using secondary-ion mass spectrometry (SIMS) [24,25]. However, many laboratories lack an intercomparable reference material (RM) for hematite oxygen isotopes. To address the lack of stable isotope analysis RMs for hematite, we attempted to standardize the oxygen isotope values of a number of hematite RMs procured for principal analysis to assess whether they could be used as potential reference material for comparison between laboratories.

Samples
In this study, we selected five samples of hematite. Two (GBW07223a, GBW07825) were purchased from the China National Standards, and two (YSBC28740-95, YSBC28756-2008) were purchased from NCS Testing Technology Co. They were previously used as reference materials for primary or trace element analysis (see https://www.ncrm.org.cn/Web/Home/Index; http://icloud.ncscrm.com/catalog. aspx). One is a granular reference material of hematite (Harvard 92649), which has previously been used as an RM for electron probe analysis [21]. We also collected a piece of oolitic hematite (Chuan), as shown in Figure 1, from the Chuanlinggou Formation (ca. 1.65 Ga.) in North China, which was crushed into < 63 µm powder. Oxygen isotope composition of hematite has been analyzed on similar such samples by previous researchers [7].

Laser Fluorination Extraction System and Oxygen Isotope Analysis
We designed and built a laser-based fluorination oxygen extraction line ( Figure 2) at the Institute of Geology and Geophysics, Chinese Academy of Sciences. This line was made with a stainless steel tube. BrF 5 was purchased from Tianjin Special Gas Company and placed in Kel-F Teflon distillation traps ( Figure 2). The laser fluorination system utilizes a New Wave Research MIR10-30 laser coupled to a vacuum extraction system. The sample chamber in the laser fluorination system is made of nickel metal and has 36 circular cavities with diameters of 3 mm. The whole system consists of 2 cold fingers, 2 cold traps and one hot trap. Both the cold fingers and the cold traps were immersed in liquid nitrogen during the reaction. KBr was added to the hot trap and heated to 100 • C to react with the F 2 possibly present in BrF 5 and to collect the Br 2 in the later cold trap.
One zircon, Penglai (PL) (IGGCAS; δ 18 O = +5.17% ) [26] and one garnet, 04BXL07 garnet (University of Science and Technology of China; δ 18 O = +3.70% ) [27] were used in the standard set and were ± 0.08% and ± 0.05% , respectively. The measured 18 O/ 16 O ratios were normalised using the value for Standard Mean Ocean Water (SMOW). A ca. 2 mg sample was weighed and placed into the sample chamber. The sample chamber is designed to hold up to 36 samples at a time, with a batch of 10 RMs, including garnet (04BXL07) and zircon (PL) for the analytical test procedure. An excess of BrF 5 was added, and the reaction took approximately 10 minutes to remove adsorbed moisture from the air. The reaction gas was purified and withdrawn from the sample. Usually, two PL zircon samples are analyzed consecutively until the specific value of the PL sample reaches its given value (+5.17% ) before the other samples are analyzed. The fluorination of minerals is usually performed by adjusting the diameter of the CO 2 laser beam to 3 mm and the power of the reaction to between 10 and 20 W. In the presence of BrF 5 , the reaction usually emits a bright white light, and the reaction time is usually 5 minutes, until no white light is emitted and the oxygen of the sample is completely released. Furthermore, the amount of oxygen obtained from the reaction is used to calculate the yield, which is usually > 99%, indicating a complete reaction. The remaining BrF 5 from the reaction is first collected in cold finger 1 and cold trap 1, and trace F 2 gas passes through a hot well containing KBr at 100 • C to produce Br 2 , which is collected in cold trap 2 ( Figure 2). The reference O 2 gas used was purchased high-purity oxygen (99.999%). The oxygen isotope values (δ 18 O) were −9.91% , which ere obtained by calibrating to the international RM quartz (NBS-28). The O 2 gas was analyzed on a MAT 252 mass spectrometer. The mass spectrometer is equipped with three Faraday cups for simultaneous measurement of 32, 33 and 34.

Chemical Treatment
The fluorination of natural ironstones releases oxygen not only from iron-bearing minerals like hematite but also from other oxygen-bearing minerals like quartz and silicate minerals. Therefore, in order to analyze oxygen isotopes of iron oxides in natural ironstones, it is necessary to analyze the oxygen content and isotope of quartz and silicate minerals present in natural ironstones.
Sedimentary hematite usually contains significant amounts of silicates, including quartz, and different oxygen-bearing silicate minerals have a significant effect on the determination of oxygen isotopes of iron oxides. Methods for the assessment of oxygen isotopes of iron oxides for hematite are numerous [1,7,11]. In this study, we used the 12N HCl dissolution treatment method. About 1.5 g of the powder sample was weighed and placed in a 15 mL centrifuge tube, and 10 mL HCl was added. To avoid the possible exchange of oxygen from the reagent with the silicate as much as possible, the reaction condition was room temperature at 25 • C for 24 hours. However, we found that at room temperature, even for pure ferric oxide (metal basis), complete dissolution was still not achieved (only 83.4% dissolution), whereas at 90 • C, complete dissolution can be achieved. Therefore, the reaction temperature is very critical for the dissolution of hematite. Previous studies have also shown that a temperature range of 60-80 • C ensures the complete dissolution of hematite in the presence of sufficient HCl [28]. After the 24-hour reaction is completed, the tube was centrifuged for 10 minutes at 4000 rpm. After pouring off the upper clear liquid, secondary ionized water was added for repeated washing and centrifugation more than four times. The residue was then dried in a freeze dryer or oven (60 • C). The weight of the residue as a percentage of the whole rock sample was calculated (Table 1).

Measument of CO 3 2−
In natural hematite samples, CO 3 2− -bearing minerals, such as siderite, calcite or dolomite are common. Therefore, it is necessary to measure the oxygen content of these kinds of minerals. In this study, we used GasBench II system (ThermoFisher company) which is commercial equipment to measure carbon and oxygen isotope of carbonate.
In order to measure the total inorganic carbon, approximately 300 µg of pure carbonate RM and 4 mg of hematite samples were reacted with 100% phosphoric acid at 70°C within a 12 mL vial tube filled with He, and generated CO 2 gas was introduced into an isotope ratio mass spectrometer.

Oxygen Isotope of Bulk Rocks (δ 18 O Bulk )
The oxygen isotopes of bulk rocks tested in this study were mainly in the range of −1.12 to 2.25% ( Table 2). The two RM of GBW07223a and GBW07825 had consistent bulk rock δ 18 O values within the error ranges of 2.20 ± 0.02% and 2.25 ± 0.08% , respectively. Whereas YSBC28740-95 has a slightly lower value of 0.76 ± 0.26% , it is less reproducible relative to the oxygen isotope values of the other samples. This indicates its possible inhomogeneity. There are two extremes of δ 18 O values: one sample (YSBC28756-2008) has a high δ 18 O value (9.51 ± 0.03% ), and the hematite RM (Harvard 92649) has the lowest δ 18 O value (−1.12 ± 0.07% ). With the exception of RM YSBC28740-95, the analytical repeatability of the oxygen isotopes of the other whole rock samples analyzed in this analysis is high, which also indicates that the samples have good homogeneity, at least within the 2 mg sample size range. The bulk rock δ 18 O values for the newly collected samples (Chuan) are −0.57 ± 0.05% . In addition, it is very noteworthy that this sample (Chuan) has the highest oxygen content (34.09 ± 0.76%) compared to the other standard samples, which indicates that the sample contains more silicates or quartz.
We should note that one of the selected samples (GBW07825) contained high CO 3 2− content, with 3.9% (GBW07825) of its oxygen from CO 3 2− . Since this sample contains low Ca and Mg, the mineral could not have been calcite or dolomite, but rather siderite for GBW07825, given that these samples were rich in iron. The oxygen content from CO 3 2− in the other four samples (YSBC28756-2008, GBW07223a, YSBC28740-95, Chuan) was less than 0.45%, and the oxygen content from CO 3 2− in YSBC28740-95 was not detectable at all. Because the oxygen content of the CO 3 2− -bearing minerals could not be detected through faraday cup (32, 33, 34) of MS, the presence of CO 3 2− -bearing minerals also does not affect the bulk oxygen isotope of the sample to be tested [30][31][32]. The oxygen content of iron-bearing mineral obtained by the fluoridation method will underestimate the total oxygen content of the sample. The sum of the oxygen content obtained by fluoridation and the oxygen content of the carbonate minerals is representative of the total oxygen content of the sample.

Oxygen Isotopes of Residues After HCl Dissolution
Dissolution of hematite at room temperature was used to avoid possible oxygen isotope exchange as much as possible. However, our HCl dissolution experiments for pure hematite (metal basis) showed that pure hematite could not be completely dissolved at 25 • C, and about 16.6% could not be dissolved. For samples (GBW07223a, YSBC28740-95, Chuan), the red mineral (hematite) was clearly present in the residue, which indicates that room temperature is not sufficient to completely dissolve hematite. In contrast, the red hematite residue was clearly not present in the residue, and the pure hematite was completely dissolved at 90 • C, indicating that the temperature of the reaction directly determined the degree of dissolution of hematite. Oxygen isotope analysis of the residues at both temperatures also showsed that 90 • C is the more valid temperature for the complete dissolution of iron oxides.
Except for hematite RM (Harvard 92649), we performed HCl dissolution treatment on all other samples (GBW07223a, GBW07825, YSBC28740-95, YSBC28756-2008, Chuan). The four samples (GBW07223a, GBW07825, YSBC28740-95, YSBC28756-2008, Chuan) were treated with HCl under 25 • C and 90 • C conditions. The percentage content of the residue of hematite at 25 • C was significantly higher than that at 90 • C. The oxygen content of the residue nferred from its molar amount of oxygen (Table 1) is significantly higher at 90 • C than that at 25 • C due to the high oxygen content of silicate minerals compared to Fe 2 O 3 (30.06%).
It is also important to note that because we used two temperature conditions (25 • C and 90 • C) for hematite dissolution, δ 18 O values for both residues were obtained for each sample (Table 1, Figure 3). It is noteworthy that the content of residue at 90 • C is lower than that at 25 • C (Table 1), and the content of residue in these two samples (GBW07223a, YSBC28740-95) is significantly lower than that in the other two samples (GBW07825, YSBC28756-2008). Correspondingly, the sample with the largest change in residue amount (GBW07223a, YSBC28740-95) had the largest change in oxygen isotope value ( Figure 3, Table 1). The residue δ 18 O values at 25 • C are much closer to those of the bulk rock samples (Figure 3, Table 1), which also indicates that hematite is not completely dissolved at 25 • C. The oxygen isotope ratios of the residues of the two samples (GBW07825, YSBC28756-2008) with small residue variation are very similar, which indicates that the dissolution of Fe 2 O 3 was nearly achieved in these two samples at 25 • C. However, we can also see that the δ 18 O values of the residues at 25 • C are obviously lower, which also reflects that a small portion of the Fe 2 O 3 exists in the residue. In addition, we noted that the δ 18 O values of the residues of three (GBW07223a, GBW07825, YSBC28756-2008) of the four RMs and one other collected sample (Chuan) were all in the positive range of 10-13% at 90 • C ( Table 1). The δ 18 O values of the residues of one RM (YSBC28740-95) were relatively low (5.59% ) ( Table 1).
The Hematite RM (Harvard 92649) is known from previous testing that this RM is a relatively pure hematite and that total iron and oxygen are essentially the main elemental composition of this sample, so the low and negligible amount of residue can be expected.

Oxygen Isotope of Iron Oxides (δ 18 O Oxides )
By measuring the moles of O 2 gas released from oxides during fluorination, we can calculate the oxygen content of the bulk rock from the oxygen production. In addition, the oxygen content of the residue from the oxygen yield was obtained. In this way, without distinguishing the specific mineral composition of the residue, mass balance calculation was expressed as where N Bulk refers to the moles of oxygen in the bulk rock, N Residue refers to the moles of oxygen in the residue and N Oxides refers to the moles of oxygen in iron oxides, which can be calculated from Equation (2). δ 18 O Residue refers to the oxygen isotope of the residue remaining after HCl dissolution of hematite.
With Equation (2) we can obtain the oxygen isotope values for iron oxides by following equation In addition, we also tried to perform a calculation of δ 18 O Oxides by Equation (3) after HCl dissolution treatment at 25 • C (Table 1), although the residue under 25 • C includes part of the iron oxides.
Ironstones in nature are not composed of iron-containing minerals such as hematite and magnetite, but also contain other oxygen-containing minerals such as quartz, calcite and other minerals. Thus, it is impossible to directly get the oxygen isotope values of iron oxides. In contrast, a pure hematite RM (Harvard 92649) is composed mainly of Fe and O. The oxygen isotope analysis of iron oxides shows a good homogeneity (1σ = 0.07% ). It is an ideal RM for inter-laboratory comparisons.
The other samples also had more than 16% of the residue undissolved in HCl, so the oxygen isotopes of the residue and the percentage content of the residue need to be evaluated for calculating the oxygen isotope of iron oxides. In sum, the δ 18 O Oxides values of hematite obtained at 25 • C were all higher than those at 90 • C (Table 1, Figure 4). The difference between the δ 18 O Oxides value (−0.89 ± 0.11% ) and the bulk rock δ 18 O is the largest in GBW07825, an RM, relative to the small deviation between δ 18 O Bulk and δ 18 O Oxides of the other samples. In this study, the δ 18 O Oxides value of hematite sample (Chuan) is −4.71 % (Table 1). It is worth noting that the δ 18 O Oxides values of some samples collected from the same location range from −3.41% to −5.67% , which is close to those of the present study [7].

Conclusions
The hematite sample (Harvard 92649) is an ideal hematite RM for inter-laboratory comparisons, and our study shows that it not only has good primary element homogeneity but also has good oxygen isotope homogeneity.
In the present study, ironstones were subjected to HCl dissolution chemistry. By further analysis of the oxygen content and oxygen isotopes of the insoluble residue, the oxygen isotope ratios of iron oxides can be obtained by mass balance calculations.  Acknowledgments: Thanks to Chaofeng Li for providing the sample GBW07825 and to Lanzhen Guo for his help in the sample analysis.

Conflicts of Interest:
The authors declare no conflict of interest.