High‐Precision Measurement of Stable Cr Isotopes in Geological Reference Materials by a Double‐Spike TIMS Method

Chromium (Cr) isotopes have been widely used in various fields of Earth and planetary sciences. However, high‐precision measurements of Cr stable isotope ratios are still challenged by difficulties in purifying Cr and organic matter interference from resin using double‐spike thermal ionisation mass spectrometry. In this study, an improved and easily operated two‐column chemical separation procedure using AG50W‐X12 (200–400 mesh) resin is introduced. This resin has a higher cross‐linking density than AG50W‐X8, and this higher density generates better separation efficiency and higher saturation. Organic matter from the resin is a common cause of inhibition of the emission of Cr during analysis by TIMS. Here, perchloric and nitric acids were utilised to eliminate organic matter interference. The Cr isotope ratios of samples with lower Cr contents could be measured precisely by TIMS. The long‐term intermediate measurement precision of δ53/52CrNIST SRM 979 for BHVO‐2 is better than ± 0.031‰ (2s) over one year. Replicated digestions and measurements of geological reference materials (OKUM, MUH‐1, JP‐1, BHVO‐1, BHVO‐2, AGV‐2 and GSP‐2) yield δ53/52CrNIST SRM 979 results ranging from −0.129‰ to −0.032‰. The Cr isotope ratios of geological reference materials are consistent with the δ53/52CrNIST SRM 979 values reported by previous studies, and the measurement uncertainty (± 0.031‰, 2s) is significantly improved.

9.501% and 2.365%, respectively (Meija et al. 2016). Based on the short-lived 53 Mn-53 Cr extinctive radionuclides (3.7 ± 0.4 My), previous studies on Cr isotopes have focused on dating meteorites or identifying the formation process occurring in the solar system during the first 20 My by massindependent Cr isotope variation (Birck and All egre 1988, Lugmair and Shukolyukov 1998, Trinquier et al. 2008, Yamakawa et al. 2009, Qin et al. 2010a). In addition, 54 Cr/ 52 Cr ratios in extra-terrestrial samples that display planetary scale isotopic anomalies have been used to fingerprint different nucleosynthetic provenances (Trinquier et al. 2007, Qin et al. 2010b, Qin and Carlson 2016. Recently, chromium isotopes have been applied as a palaeo-redox proxy in low-temperature environments and especially exploited as a tracer of Earth's atmospheric oxygenation (Frei et al. 2009, Crowe et al. 2013, Planavsky et al. 2014, Cole et al. 2016. Chromium is typically present in nature in three valence states -Cr(II), Cr(III) and Cr(VI). Hexavalent chromium Cr(VI), as either a chromate (alkaline pH) (CrO 4 2-) or a hydrogen chromate (acidic pH) ion (HCrO 4 chromium Cr(III) (Rai et al. 1989, Ellis andBullen 2002). To date, up to~6-7‰ variation of d 53/52 Cr NIST SRM 979 has been found during reduction and oxidation reactions (Schauble et al. 2004, Zink et al. 2010, and a large range from -0.27‰ to 1.23‰ was also found during non-redoxdependent processes (Saad et al. 2017). Accordingly, Cr isotopes have been utilised to trace the natural attenuation of Cr(VI) in groundwater (Blowes 2002) or to constrain the redox state of modern and ancient seawater , Bonnand et al. 2013, Scheiderich et al. 2015. A few studies have been devoted to the inventory of the solid Earth and planetary processes at high temperature, and these studies improve our understanding of the Cr cycle (Schoenberg et al. 2008, Rudge et al. 2009, Moynier et al. 2011, Li et al. 2016, Xia et al. 2017. Regarding the inventory of the solid Earth, Schoenberg et al. (2008) suggested that the d 53/52 Cr NIST SRM 979 value of the igneous silicate Earth is -0.124 ± 0.101‰ by analysing oceanic and continental basalts, mantle xenoliths, ultramafic rocks and cumulates. Farka s et al. (2013) suggested that the d 53/52 Cr NIST SRM 979 value of the bulk silicate Earth (BSE) is -0.079 ± 0.129‰ by analysing mantle-derived chromites.
Recently, Xia et al. (2017) suggested that the d 53/52 Cr NIST SRM 979 values of fresh fertile peridotites are -0.14 ± 0.12‰ on average and the variation of d 53/52 Cr NIST SRM 979 values of mantle peridotites is caused by partial melting. The d 53/ 52 Cr NIST SRM 979 value of BSE was given as -0.11 ± 0.11‰ by analysing komatiites (Sossi et al. 2018). For the planetary processes, Moynier et al. (2011) suggested that there could be Cr isotopic fractionation during core segregation, as the d 53/ 52 Cr NIST SRM 979 values of meteorites (-0.2‰ to -0.4‰) are lighter than that of the BSE. Schiller et al. (2014) also found light d 53/52 Cr NIST SRM 979 of meteorites ( -0.30‰). However, Bonnand et al. (2016b) found that meteorites have a BSE-like stable Cr isotopic ratio and suggested that Cr isotopes are fractionated during the magmatic process.
Recently, Bonnand and Halliday (2018) suggested that equilibrium fractionation between iron liquid and sulfides or kinetic fractionation during oxidation of Cr may be responsible for Cr isotopic fractionation during fractional crystallisation of meteorite rocks.
Compared with the very large Cr isotopic fractionation at low temperatures, Cr isotopic fractionation in high-temperature environments is much smaller. Thus, small measurement uncertainties during measurement of natural samples are a prerequisite to improve the application of Cr stable isotopic fractionation in high-temperature geochemistry. High-precision Cr isotopicmeasurementshavebeenobtainedforsampleswith extremely high chromium contents such as meteorite and extraterrestrial samples (Moynier et al. 2011, Schiller et al. 2014, Bonnand et al. 2016a). An improved Cr isotope ratio measurement procedure for terrestrial samples that contain less chromium content is also established in this study.
The small measurement uncertainty of stable Cr isotope ratio measurement results requires not only a high-quality chemical separation process but also appropriate methods for mass bias correction, which occurs during measurement and chemical purification. High-quality Cr purification enables a reduction in the isobaric interference from 54 Fe, 50 V and 50 Ti, and matrix element effects that interfere with the production of Cr + thermal ions (Ball and Bassett 2000). Chromium mainly has two valence states in solution, and the behaviours of Cr(III) and Cr(VI) during ion exchange separation are very different (Schoenberg et al. 2008, Trinquier et al. 2008, Yamakawa et al. 2009, Li et al. 2016. To convert all chromium ions to the same valence state, various oxidising or reducing agents ((NH 4 ) 2 S 2 O 8 , K 2 S 2 O 8 , KMnO 4 and HCl) were used in previous studies in the anion or cation resin stage, and correspondingly, two, three or even four columns were used (Table 2; Lugmair and Shukolyukov 1998, Ball and Bassett 2000, Ellis and Bullen 2002, Johnson and Bullen 2004, Halicz et al. 2008, Trinquier et al. 2008, Schoenberg et al. 2008, Yamakawa et al. 2009, Schiller et al. 2014, Rodler et al. 2015, Li et al. 2016, Schoenberg et al. 2016, Li et al. 2017, Zhu et al. 2018. These methods can obtain satisfactory pure Cr cuts from multiple kinds of samples. Generally, 0.5-1 µg of Cr must be loaded onto the filament for precise measurement results with the TIMS. Thus, the chemical purification process must be guaranteed to be applicable for larger test portions sizes. The residual of the oxidising reagents (SO 4 2-) can severely inhibit the Cr + signal during TIMS measurements, but this residual is difficult to remove (Ball and Bassett 2000). In addition, the high ionisation potential of Cr and the inhibition effect of the Cr + signal from organic matter during instrument measurements are the main reasons impeding the acquisition of highprecision isotope ratio results in compositionally complex geological materials during TIMS measurements (Johnson and Bullen 2004, Yamakawa et al. 2009, Li et al. 2016. Previous studies achieved Cr purification from meteorite and terrestrial samples by a two-step column method with AGW50-X8 (Birck and All egre 1988, Lugmair and Shukolyukov 1998, Trinquier et al. 2008, Qin et al. 2010a, Bonnand et al. 2016b. Bonnand et al. (2011) also obtained the precise Cr isotope ratio of carbonates based on a onestep method, which has the same Cr speciation and resin types. In this study, we report an improved two-step column purification scheme without additional oxidising/reducing agent for high-precision measurement of stable Cr isotope 6 4 8 ratios based on the method mentioned above. The resin is modified to Bio-Rad AG50W-X12 (200-400 mesh) cation exchange resins. The inhibition effects of the isobaric interference, matrix and organic matter are tested and assessed. HClO 4 and HNO 3 acids were utilised to eliminate organic matter interference from the resin and were removed without Cr loss. High-precision Cr isotope ratios of geological reference materials ranging in lithologies from ultramafic rock to granitoid are presented in this study.

Measurement procedure Reagents
In our experiments, all optima-grade HF, HCl and HNO 3 acids were further purified twice by a Savillex TM DST-1000 sub-boiling distillation system (Minnetonka, USA). High purity water (HPW) with a resistivity of 18.2 MΩ cm was obtained from a Milli-Q (MQ) Element system (Millipore, USA) and used in the whole procedure. PFA Savillex TM beakers were cleaned with HNO 3 (14 mol l -1 ) and/or HCl (11 mol l -1 ), diluted to 50% v/v with water for 24 h on a hot plate at 130°C, followed by cleaning with purified HCl (11 mol l -1 ), and were finally rinsed with HPW. Optimagrade perchloric acid was obtained from Thermo Fisher and purified by Bio-Rad AG50W-X12 resin (200-400 mesh) with the chromatographic method. The GSB-Ti, Fe, V, K, Na, Ca, Mg and Mn ultrapure standard solutions used in this study were from the China Iron and Steel Research Institute.

Sample digestion
Sample powders containing 0.5-1 l g Cr were digested in a combination of concentrated HF (29 mol l -1 ) and HNO 3 (14 mol l -1 , 2:1 by volume) in Teflon TM PFA beakers at 130°C for two or three days until solid particles had disappeared. Next, the solutions were heated to evaporation. Then, aqua regia (HCl: HNO 3 = 3:1) was added until the steamed sample completely dissolved. The sample solutions were evaporated to dryness on a hot plate at a surface temperature of~130°C. The products were finally dissolved in 0.5 ml of concentrated HCl (11 mol l -1 ). The dissolved sample solutions were mixed with 0.4 ml 50 Cr-54 Cr doublespike solution for every~1 lg Cr, and then, the mixtures were heated on a hot plate at 130°C overnight to achieve homogenisation. After evaporation to dryness, the mixtures were re-dissolved in 0.2 ml 6 mol l -1 HCl.

Column chemistry
A modified method for Cr purification was developed by a two-step column procedure (Figure 1b, c, Table 1). In previous studies, cation and anion exchange resins (100-200 mesh/ 200-400 mesh) were usually used in the Cr column separation procedure (Table 2; Lugmair and Shukolyukov 1998, Ball and Bassett 2000, Ellis and Bullen 2002, Johnson and Bullen 2004, Halicz et al. 2008, Trinquier et al. 2008, Schoenberg et al. 2008, Yamakawa et al. 2009, Schiller et al. 2014, Rodler et al. 2015, Li et al. 2016, Schoenberg et al. 2016). AG50W-X8 was applied as a conventional cation exchange resin during Cr purification ( Table 2, Trinquier et al. 2008, Qin et al. 2010a, Bonnand et al. 2011, 2016a. The purification methods applicable to silicate and meteorite materials have a limit on the minimum Cr mass fraction of the samples, and this limit is greater than approximately 45 µg g -1 (Table 2). Additionally, a yield of more than 0.5 µg Cr is normally required after separation for high-precision measurement results by MS. Here, analytical grade Bio-Rad AG50W-X12 (200-400 mesh) cation exchange resins were used, and these resins have higher cross-linkage (12%), smaller wet bead size (53-106 µm) and smaller molecular weight limits (~400) than the Bio-Rad AG50W-X8 (200-400 mesh; cross-linkage: 8%; wet bead size: 63-150 µm; molecular weight limits:~1000) cation exchange resins (http://www.bio-rad.com/en-hk/category/ ion-exchange-resins/). Thus, the AG50W-X12 (200-400 mesh) resin has a high sample load capacity so that this scheme is suitable for geological samples with variable Cr content. Silicate samples with Cr mass fractions greater than 17 µg g -1 are suitable for our column chemistry, which mean the maximum mass for sample powders is about 50 mg. The yield of this method is as good as published purification schemes ( Table 2). The adoption of the AG50W-X12 resin also improves the separation efficiency and selectivity, which result in better separation efficiency than the Bio-Rad AG50W-X8 resin (Figure 1a, b). The retardation of Ti and V ions avoids the overlap of the elution of Cr ions and isobaric interference cations (Figure 1a, b). A better purification effect for Cr is also obtained from matrix cations such as Mg + and Ca + (Figure 1a, b).
In detail, 1 ml and 0.33 ml of Bio-Rad AG50W-X12 (200-400 mesh) cation exchange resins were loaded in columns (4 mm in diameter) with different lengths named column 1 and 2, respectively. For column 1 (Figure 1b, Table 1), the resin was washed with 5 ml 6 mol l -1 HCl and 3 ml HPW, and was pre-conditioned with 5 ml 1 mol l -1 HCl. Prior to launching the Cr column, the samples were diluted with HPW to a total volume of 1.2 ml 1 mol l -1 HCl solution and then loaded carefully onto the resin beds. Chromium in the samples formed as Cr(III)-Cl complexes in concentrated HCl at 130°C and was further eluted with 3.5 ml 1 mol l -1 HCl (Larsen et al. 2016). Then, sub-purified 6 4 9 Cr samples were re-digested with 20 ll concentrated HNO 3 and diluted with 2 ml HPW (~0.16 mol l -1 HNO 3 ) before loading on column 2. The Fe, V and matrix elements were almost completely removed by column 1 (Figure 1b, Table 1).
For column 2 ( Figure 1c, Table 1), the resin was washed with 4 ml 6 mol l -1 HCl and 3 ml HPW. The chromium aliquots from column 1 were loaded onto column 2, which was pre-conditioned with 5 ml 0.16 mol l -1 HNO 3 . After sample loading, Ti and Al ions were removed by 3 ml 0.5 mol l -1 HF, and V and other matrix elements were eluted again by following 10 ml 1 mol l -1 HCl. Chromium was eluted with 4 ml 2 mol l -1 HCl. After column separation, the final Cr cuts were died down in concentrated HNO 3 (14 mol l -1 ) several times and then dissolved with concentrated HNO 3 and HClO 4 acid to eliminate organic matter before isotopic measurement.
By implementing a two-step separation procedure, we obtained a highly purified Cr fraction (Figure 1b, c). The Cr yields after the two-column chemistry were higher than 85%, and the whole-procedure Cr blank was less than 3-4 ng, which was considered to be negligible (lower than 0.5%) relative to total amount of loaded Cr (1000 ng). The Cr yields were tested using BHVO-2 by ICP-MS, and the wholeprocedure Cr blank was also measured by ICP-MS.

Cr double-spike technique
The 50 Cr-54 Cr double spike is usually used to correct the mass dependence fractionation generated during sample processing as well as by instrumental mass bias effects for the Cr isotope ratio of samples (Ellis and Bullen 2002, Schoenberg et al. 2008, Frei et al. 2009  The mass-dependent isotopic fractionation in the laboratory, from the column chemistry and from isotopic fractionation by TIMS, was corrected by a 50 Cr-54 Cr double-spike technique. Two single spikes of 50 Cr and 54 Cr were purchased from Oak Ridge National Laboratory. The double spike was prepared by mixing the single spikes in an appropriate proportion ( 50 Cr: 54 Cr ≈ 1:1) and was calibrated by TIMS at the China University of Geosciences. To achieve high-precision isotopic determinations, the proportions of two single spikes and between double spike and sample must be optimised. Our double-spike correction method assumed an exponential mass fractionation law and calculated by iterative Newton-Raphson procedure (Albar ede and Beard 2004) to solve the double-spike equations described in Rudge et al. (2009).
The Monte Carlo method was used to predict optimised the spike/ (spike + sample) ratios (Q) and ratios of double spike (P). Lehn et al. (2013) further proposed an improved method to simulate TIMS analysis by considering Faraday collector damage and achieved excellent repeatability precision of~0.024‰ (2 SE) for Ca isotope ratio measurement results. Here, according to the mathematical equations presented by Lehn et al. (2013), the Monte Carlo simulation, mainly, takes counting statistics (s cs ) and Johnson noise (s jn ) into consideration to optimise spike/ (spike + sample) ratios (Q) and the ratios of 50 Cr spike and 54 Cr spike (P) of Cr isotopes (Figure 2a). The simulation was implemented by MATLAB, and the results are presented in Figure 2a. The ion beam voltage of each isotope, integration time, cycles of each analysis, temperature of the collectors (298 K) and amplifier resistance are all involved in calculation of the theoretical uncertainty curve, which was consistent with that in the experiment. We mixed two single spikes as a 54 Cr/ Cr total ratio of 0.484. The d 53/52 Cr NIST SRM 979 of NIST SRM 979 mixed with different double-spike proportions was determined by TIMS, and the experimental error matched the theoretical simulation uncertainty curve of d 53/52 Cr NIST SRM 979 (Figure 2b, Table S1). Given that the ratio of 54 Cr/ Cr total in double spike is 0.484, the optimal spike proportion should be 10% to 40% of the total Cr (Cr spike + Cr sample ); hence, we selected a Cr spike /Cr total ratio of 0.25 in our method.

Mass spectrometric measurements
The measurements of Cr isotope ratios were performed on a Thermo Fisher Scientific TRITON Plus TIMS in the Isotope Geochemistry Lab of the China University of Geosciences, Beijing. This instrument was equipped with nine Faraday cups linked to 10 11 Ω amplifiers. Chromium was loaded in 2 ll of 3% m/v HNO 3 onto outgassed Re single filaments under a binocular microscope using a 0.5-10 ll range digital pipette. Narrow parafilm dams were placed on the filaments to facilitate the core formation of sample aliquots, and the remaining organics were removed by a brief heating procedure. Then, a mixture solution of 1.3 ll high purity (99.99%) silica gel (< 100 nm) and 1.3 ll saturated H 3 BO 3 was added to the sample drops. The mixture was first dried by slow heating under low current conditions (~0.5 A) through the filament to form a glass, and then, the current was increased slowly until the filament was dull red (~2.2 A) for 2s. Every filament load standard or sample contained 0.5-1 lg of Cr. One sample was loaded on 1-2 filaments. All Cr isotopic data were acquired in static multi-collection mode by the collector array summarised in Table 3. Ion beam measurements at m/z 49 (Ti + ), 51 (V + ) and 56 (Fe + ) allowed us to correct isobaric interferences from Ti + and V + on Cr + and Fe + on Cr + . During a measurement routine, the filament was heated to 1500 mA in 5 min and then heated to 1800 mA in 3 min. Generally, data acquisition started when the current beam of 52 (Cr + ) was stable and higher than 4 9 10 -11 A and the temperature of the filaments was ca. 1330°C. For each analysis, fourteen blocks of twelve cycles at integration times of 8 s were obtained using amplifier rotation. Peak centre and auto-focus were applied at the beginning and after every seven blocks of data acquisition. Each block included a 60 s baseline measurement. Amplifier gains were calibrated at the start of each day to eliminate all gain calibration errors. The total acquisition time was approximately 60 min for one filament, including a 15 min heating routine. Each filament was monitored three or four times.
Results and discussion

Evaluation of Ti, V and Fe isobaric interference and other matrix interference
Iron is a major element, and Ti and V are trace elements in geological materials. All of these elements have direct 6 5 2 isobaric interference on the ion beam of Cr isotopes. In addition, matrix elements (K, Na, Ca and Mg) suppress the emission of Cr on the filament, and this suppression may affect the uncertainty of measurement results of Cr isotopes. Hence, it is important to evaluate the potential effects of isobaric interference from Ti, V, Fe and other matrix elements, even though our Cr separation method provides good separation of Cr from terrestrial samples. To evaluate the isobaric interferences from Ti, V and Fe, 1 lg of unspiked GSB-Cr calibrator was doped with different amounts of GSB-Ti, V and Fe calibration standard solutions (Figure 3, Table S2). The isobaric elements/Cr concentration ratios ranged from 0.001 to 10. We also doped a mixture of K, Na, Ca and Mg to evaluate the inhibition effects from these matrix elements on the Cr + signal intensity during determination (the ratio of mixture to Cr ranged from 0.01 to 0.1).
As a result of high ionisation potential, Ti and V are very difficult to ionise with TIMSa finding also proposed by Li et al. (2016) (Figure 2). Nevertheless, we could not obtain a stable Cr + single when the ratio of Ti to Cr is greater than 0.01. The presence of Ti suppresses the signal intensity of Cr + and results in poor precision (Li et al. 2016). We also noticed that the signal of Cr + attenuates rapidly during determination. No signals of 51 (V + ) and 49 (Ti + ) were detected during instrument measurements, as the 49 Ti / 52 Cr and 51 V / 52 Cr ratios of samples doped with GSB-Ti and V were indistinguishable from those of the pure GSB-Cr calibration material. This finding suggests that Ti and V only inhibit the emission of Cr but show no isobaric interferences. The mixture of K, Na, Ca and Mg provides a similar effect as that of Ti and V for Cr isotope measurement (Figure 3).
However, the presence of both Cr + and Fe + beams can be observed in pure GSB-Cr doped with Fe during measurement, and a weak positive relationship was observed between the 53 Cr/ 52 Cr ratios and the amount of   Lehn et al. (2013)  6 5 3 added Fe (Figure 3). The samples doped with less than 1 lg Fe could show bias-free 53 Cr/ 52 Cr ratios depending on our analysis (Figure 3). Iron can be ionised with Cr due to the lower ionisation potential than Ti and V at slightly higher temperature. Excess Fe caused strong 56 (Fe + ) signal interference and inhibited the emission of Cr + for a sample doped with 10 lg Fe (Figure 3). It is noteworthy that the current beam of 56 (Fe + ) increases with temperature, especially when the temperature is higher than 1350°C. Therefore, in our experience, data acquisition should be treated carefully when the temperature of the filaments reaches 1350°C.
In contrast to analysis by MC-ICP-MS, Cr isotopic measurements by TIMS do not have strong isobaric and matrix interferences, as the Ti, V and most of the Fe have low ionisation efficiency or even no ionisation due to the high ionisation potential and different ionisation temperature relative to those of the Cr isotope (Li et al. 2016). The addition of Ti, V and Fe inhibits the ionisation of Cr, and only Fe can be ionised with Cr ( Figure 3).

Elimination of organic matter in the sample after chemical separation
During the measurement of Fe, Cr, Pb and Cu isotopes using TIMS, organic matter is considered a significant interference (Kuritani and Nakamura 2002, Johnson and Bullen 2004, Yamakawa et al. 2009).
The organic matter in samples is most likely derived from the exchange resin or colloidal silica. Previous work used ultraviolet radiation, H 2 O 2 or HClO 4 treatment to eliminate organic residues originating from the column resin (Kuritani and Nakamura 2002, Johnson and Bullen 2004, Yamakawa et al. 2009. During our experiment, the beam current of 52 (Cr + ) increased smoothly and was stable over 10 9 10 -11 A when 0.5 lg of pure Cr isotope reference materials NIST SRM 979 and NIST SRM  6 5 4 3112a were measured. By contrast, the beam of 52 (Cr + ) in NIST SRM 979, NIST SRM 3112a or geological reference materials (~0.5-1 lg Cr) that underwent chemical procedures usually did not exceed 1 or 2 9 10 -11 A and attenuated to zero rapidly. Thus, the elimination of organic matter is required for the measurement of high-precision Cr isotope ratio in complex rock samples.
The organic component in samples is highly persistent after the repeated drying of samples with drops of nitric acid at elevated temperature or adding H 2 O 2 (Kuritani and Nakamura 2002, Johnson and Bullen 2004, Yamakawa et al. 2009. Both H 2 O 2 and HNO 3 were tried many times to eliminate organic matter after Cr purification with different temperature/time conditions (for the repeated digestion of a BHVO-2 sample with 1 µg Cr). As shown in Figure 4, the signal intensity decayed rapidly after drying 4-8 times repeatedly samples with drops of H 2 O 2 , and similar results were obtained even at room temperature/ heating the samples with H 2 O 2 at 70°C for 1-14 days. The signal intensity also decayed after repeatedly drying samples 5-8 times with drops of HNO 3 . HClO 4 has not been commonly used recently, as perchloric acid results in Cr loss by evaporation of Cr (Makishima et al. 2002). However, in this study, an optimised process performed with HClO 4 , which is the most effective medium to eliminate organic interference after trying different processes (HNO 3 /H 2 O 2 with different temperature/time conditions). All samples were heated on a 120°C hot plate with 500 ll concentrated HNO 3 and 30 ll HClO 4 for two weeks to obtain a high and stable beam intensity of 52 (Cr + ) (Figure 4, Table S3). The reaction rate did not clearly increase with increasing temperature but was coupled with the reaction time. Attempts to shorten the sample processing time at 120°C or higher did not yield the satisfactory effect of the eliminating organic interference. In addition, the degree of elimination of the organic matter with pure HClO 4 is limited without the cooperative effect of HNO 3 . The procedural blank from HClO 4 was negligible, as the HClO 4 was purified. As shown in Figure 5b, d 53/52 Cr NIST SRM 979 values at every point were obtained during repeated measurements of one sample on one loaded filament with repeated sample digestion. High-precision Cr isotope measurement results, with a typical high and stable beam of 52 (Cr + ) (> 4 9 10 -11 A), were achieved for every repeated geological sample after elimination of the organic component by HNO 3 and HClO 4 (Figure 5b). The long-term intermediate precision for both USGS BHVO-2 and JP-1 was less than 0.035‰ for measurements over two years ( Figure 5b). As suggested by Makishima et al. (2002), the addition of perchloric acid results in Cr loss by evaporation [Colour figure can be viewed at wileyonlinelibrary.com] 6 5 5 of Cr in the form of CrO 2 Cl 2 in the process of drying (Makishima et al. 2002, Li et al. 2017. The boiling point temperature of CrO 2 Cl 2 is 116.85°C (390 K). In this study, because of the low boiling temperature of concentrated HNO 3 (83°C), HClO 4 can evaporate the solution to dryness with HNO 3 at only 105°C to avoid the loss of Cr.
Geological reference materials such as BHVO-1, BHVO-2, and JP-1 all produced a stable ( Figure 5b) and high beam intensity of 52 Cr + (i.e., 4-12 V) after eliminating organic matter. One measurement result for BHVO-1 was slightly higher than the mean value of BHVO-1 in our laboratory, as the repeated measurements on the filament resulted in an excessive filament temperature (1380°C). The 56 Fe/ 52 Cr ion beam ratio of this BHVO-1 sample (2.74 9 10 -4 ) is two orders of magnitude higher than that of the pure Cr isotope reference material, and these data are consistent with mean value of d 53/52 Cr NIST SRM 979 of BHVO-1 in our laboratory within the given precision. Nevertheless, a small mass of Fe is ionised at this temperature, perhaps resulting of a shift of the d 53/52 Cr NIST SRM 979 of BHVO-1. This observation is consistent with our previous discussion. To achieve high-precision Cr isotope determination, a steady filament temperature should be maintained at approximately 1350°C during measurement.

Performance of the new measurement procedure
Our long-term instrumental stability was established by measuring the spiked NIST SRM 979 Cr reference material and NIST SRM 3112a Cr reference material over a period of sixteen months ( Figure 6). The d 53/52 Cr NIST SRM 979 values were determined by the double-spike technique and normalised by the daily mean of the d 53/52 Cr NIST SRM 979 value of the isotopic reference material NIST SRM 979 to eliminate the small offset from the instrument's mass bias and double-spike technique. The d 53/52 Cr NIST SRM 979 value of NIST SRM 979 was 0.000 ± 0.017‰ (2s, n = 163) and that of d 53/52 Cr NIST SRM 979 of NIST SRM 3112a was -0.069 ± 0.025‰ (2s, n = 169) relative to that of NIST SRM 979. Our NIST SRM 3112a d 53/52 Cr NIST SRM 979 value is in excellent agreement with previous studies (i.e., -0.067 ± 0.024‰, Schoenberg et al. 2008, -0.09 ± 0.04‰, Xia et al. 2017, -0.07 ± 0.04‰, Shen et al. 2018. To check the stability of the measurement results, we also analysed the mass-independent Cr isotope ratios (after internal normalisation using the 50 Cr/ 52 Cr ratio) of these two isotope reference materials without double-spike addition over a period of six months. The 53 Cr/ 52 Cr and 54 Cr/ 52 Cr ratios of NIST SRM 979 and NIST SRM 3112a were stable and uniform, with 53 Cr/ 52 Cr ratios of NIST SRM 979 and NIST SRM 3112a of 0.1134616 ± 0.00002 (2s, n = 22) and 0.1134617 ± 0.00002 (2s, n = 63) and 54 Cr/ 52 Cr ratios of NIST SRM 979 and NIST SRM 3112a of 0.028214 ± 0.00001 (2s, n = 22) and 0.028214 ± 0.00001 (2s, n = 63), respectively (Figure 7). d 53/52 Cr NIST SRM 979 of geological reference materials We report stable Cr isotope ratio of seven commercially available geological reference materials relative to NIST SRM 979 (Table 4 and Figure 8). All the d 53/52 Cr NIST SRM 979 of isotope reference materials are in good agreement with the published data and have the same or smaller measurement uncertainty (Table 4) Previous studies revealed that the Cr(II)/Cr(III) ratios of mantle melts can significantly vary depending on the oxygen fugacity and the composition of the melts. Divalent chromium is likely to be the dominant species in basaltic melts (Berry et al. 2006). Recently, studies have found that partial melting of the mantle may cause a small but detectable (~0.4‰ of d 53/52 Cr NIST SRM 979 ) stable Cr isotopic fractionation (Schoenberg et al. 2016, Xia et al. 2017. The oxygen fugacity variation may dominate the mass-dependent chromium stable isotope fractionation in high-temperature processes (Bonnand and Halliday 2018, Shen et al. 2018, Sossi et al. 2018. Thus, stable Cr isotopic fractionation may be observed between mantle and basalts, given the different Cr(II)/Cr(III) ratios of mantle melts and basaltic melts (Schoenberg et al. 2008). In this study, two basalt reference materials from the USGS, BHVO-1 (-0.120 ± 0.029‰, 2s) and BHVO-2 (-0.129 ± 0.032‰, 2s), gave slightly lower d 53/52 Cr NIST SRM 979 values than that of peridotite reference material JP-1 from JSG (-0.088 ± 0.034‰; 2s) with the mean results over a period of two years. The data obtained in this study for these basalt and peridotite reference materials are consistent with the d 53/52 Cr NIST SRM 979 values reported by previous studies within the precision quoted (Schoenberg et al. 2008, Li et al. 2016, Bonnand et al. 2016a, b, Wu et al. 2017, Xia et al. 2017, Zhu et al. 2018 (Figure 5b).