Combined high-precision Δ48 and Δ47 analysis of carbonates

High-precision analysis of the excess abundance (relative to the stochastic distribution) of mass 48 isotopologues in CO2 evolved from acid digestion of carbonates (Δ48) has not been possible until recently due to the relatively low natural abundance of O. Here we show that the 253 PlusTM gas source mass spectrometer equipped with Faraday cups and 10 Ω resistors can perform combined Δ47 and Δ48 analyses on carbonates with external reproducibilities (1SD) of 0.010 ‰ and 0.030 ‰, respectively. ~10 mg aliquots of five carbonate reference materials (ETH 1, ETH 2, ETH 3, ETH 4, and Carrara) are digested with phosphoric acid at 90 °C using a common acid bath. The evolved CO2 is purified using an automated gas preparation system (including cryotraps and a GC) and analyzed for its Δ47 and Δ48 compositions using the dual inlet system of a 253 PlusTM gas source mass spectrometer. Raw Δ47 and Δ48 values are finally normalized to the Carbon Dioxide Equilibrium Scale (CDES). In Δ47, CDES 90°C vs. Δ48, CDES 90°C space, calcite reference materials Carrara, ETH 3 and ETH 4 agree with the equilibrium curve for calcite after adding semi-empirically determined 90 °C acid fractionation factors of 0.196 ‰ (for Δ47) and 0.136 ‰ (for Δ48) to theoretical Δ63 and Δ64 data. Agreement between measured and theoretically expected Δ48, CDES 90°C highlights the accuracy of our high-precision clumped isotope analytical setup. Combined analysis of the abundances of mass 47 and mass 48 isotopologues in CO2 evolved from acid digestion of natural carbonates has excellent potential for the determination of accurate and highly precise paleotemperatures as well as for the identification of rate-limiting kinetic processes involved in biomineralization. A formation temperature of 15(± 2) °C is obtained on the 95 % confidence level for the Upper Cretaceous chalk


Introduction
Clumped isotope analysis of the sum of 13 C 18 O 16 O and 12 C 18 O 17 O isotopologues (Δ 47 ) in CO 2 evolved from phosphoric acid digestion of carbonates (Ghosh et al., 2006) has become a now increasingly utilized (paleo)thermometer (Eiler, 2011) to determine marine (e.g., Wierzbowski et al., 2018) and terrestrial (e.g., Methner et al., 2016) paleotemperatures, body temperatures of extinct vertebrates (e.g., Eagle et al., 2011), diagenetic (e.g., Huntington et al., 2011) and hydrothermal temperatures (e.g., Bristow et al., 2011) as well as closure temperatures of metamorphic rocks (e.g., Passey and Henkes, 2012). Since this type of thermometer is based on the internal fractionation of isotopes amongst carbonate isotopologues, it is independent on the isotopic composition of the water from which the carbonate-bearing phase precipitated. Nonetheless, discrepant Δ 47 -temperature calibrations have been obtained in different laboratories (Dennis and Schrag, 2010;Ghosh et al., 2006;Fernandez et al., 2017;Petersen et al., 2019). Accuracy of Δ 47 calibration data has been shown to be sensitive to the choice of parameters used for the correction of (1) contributions deriving from 17 O-bearing CO 2 (Daëron et al., 2016;Schauer et al., 2016), (2) pressure baseline effect (Bernasconi et al., 2013;Fiebig et al., 2016;He et al., 2012), (3) scale compression (Dennis et al., 2011;Huntington et al., 2009), and (4) acid fractionation (Petersen et al., 2019). Also, calibration regression lines can be biased if the number of samples and replicates per sample, as well as the investigated temperature range, is relatively low (Fernandez et al., 2017). The observed discrepancy in Δ 47 -T calibrations is slightly reduced if calibration data from eight experimental and six empirical studies are processed with the IUPAC 17 O parameters, recommended by Daëron et al. (2016), and with a unique temperature dependence of the acid fractionation factor (Petersen et al., 2019). However, even for these reprocessed calibrations the absolute spread in Δ 47 at a given temperature remains as large https://doi.org/10.1016/j.chemgeo.2019.05.019 Received 11 February 2019; Received in revised form 9 May 2019; Accepted 14 May 2019 as 0.08 ‰, with slopes ranging from 0.033 to 0.042 in Δ 47 vs. 1/T 2 space (Petersen et al., 2019). Most recent investigations suggest that much of the observed scatter could be due to kinetics occurring in the solution prior to carbonate precipitation and at the solution-carbonate interface Daëron et al., 2019). If relevant, kinetic departures from equilibrium would limit the accuracy of the clumped isotope Δ 47 paleothermometer.
The second most abundant carbonate isotopologue containing two heavy isotopes is 12 C 18 O 18 O 16 O. CO 2 derived from acid digestion of carbonate contains on average 4.1 ppm of 12 C 18 O 18 O (mass 48), which is an order of magnitude lower than its 13 C 18 O 16 O (mass 47) content (Ghosh et al., 2006). Another mass 48 isotopologue is 13 C 18 O 17 O that has an average natural abundance of 16.7 ppb only. The weighted sum of the excess abundances of mass 48 isotopologues (relative to their stochastic distributions) is reflected by the Δ 48 value: (1) where Ri and Ri* represent the measured and stochastic* i/44 isotopologue ratios. Like Δ 47 , the carbonate Δ 48 composition does not depend on the oxygen isotopic composition of the water from which the carbonate crystallized. The position of a given carbonate sample in Δ 47 vs. Δ 48 space is, therefore, determined by its formation temperature and, if kinetics prevailed during formation, by the extent of kinetic departure from equilibrium. Since kinetically induced departures from equilibrium follow for given rate-limiting exchange mechanisms characteristic trajectories in Δ 47 vs. Δ 48 space, it should be possible to determine accurate carbonate formation temperatures based on measured Δ 47 and Δ 48 compositions even if these were affected by kinetics (Guo and Zhou, 2019).
So far, the precision of Δ 48 measurements with common gas source mass spectrometers equipped with ≤10 12 Ω amplifiers has been inadequate to permit temperature estimates based on Δ 48 . Instead, laboratories that analyze Δ 48 use its magnitude for data quality assurance of Δ 47 data, i.e., as an indicator of potential isobaric interferences on m/ z 47 (Huntington et al., 2009). Here we demonstrate that high-precision Δ 47 and Δ 48 analyses of carbonates are possible using a common acid bath and the dual inlet of a Thermo Scientific™ 253 Plus™ gas source mass spectrometer. The external reproducibilities for these measurements are close to the predicted shot noise limits. Our results imply that internal isotopic equilibrium is closely attained in calcite reference materials ETH 1, ETH 2, ETH 3, ETH 4, and Carrara.

Samples
Calcite reference materials analyzed for this study were an in-house Carrara and the internationally distributed ETH 1, ETH 2, ETH 3, and ETH 4 standards (Bernasconi et al., 2018). ETH 1 (Carrara marble) and ETH 2 (synthetic high-purity calcite) were internally equilibrated at 600°C. ETH 3 represents Upper Cretaceous chalk from the Isle of Rügen, Germany, whereas ETH 4 is the synthetic high-purity calcite that was also used for the preparation of ETH 2. The formation temperatures of ETH 3 and ETH 4 calcites are unknown (Meckler et al., 2014;Bernasconi et al., 2018). Bernasconi et al. (2018) reported Δ 47 , CDES 25°C for ETH 1, ETH 2, ETH 3 and ETH 4 that were assigned using the dual inlet system for equilibrated gas measurements and a Thermo Scientific™ KIEL IV Carbonate Device for carbonate-derived CO 2 measurements.

Carbonate reaction and gas purification
The setup of the self-constructed automated carbonate acid digestion and gas purification system (Hofmann's Auto Line, HAL) used for sample preparation is displayed in Fig. 1. It consists of a Zero Blank Autosampler (Costech Analytical Technologies, USA), a common acid bath, several cryotraps and a gas chromatograph (GC; 2.0 m of 1/8″ stainless steel tube packed with Porapak™ Q 80-100 mesh, Merck, Germany) and two turbopumps (HiCube 80 Eco, Pfeiffer Vacuum, Germany) with cryogenic water traps upstream of the turbo pumps. One pumps the part of the extraction line in front of the GC, including the autosampler and the common acid bath, the other provides the high vacuum for the two cryotraps in front of and behind the GC. Per replicate,~10 mg of calcite is loaded into silver capsules (IVA Analysentechnik, Germany). These are then placed into the autosampler, located on top of the common acid bath. Once loading is completed, the autosampler and the common acid bath are turbo-pumped for at least 5 h. Slight rotation of the autosampler forces a sample-bearing silver capsule to drop down into the phosphoric acid (> 105 wt%). Carbonate samples are routinely reacted for 30 min, and the evolving CO 2 is continuously removed at −196°C in trap 2. Trap 1 is kept at −80°C to remove water. During the reaction, the pressure is continuously monitored at trap 2. After the reaction is complete, traps 2, 3 and 4 are set to −80°C, while trap 5 is cooled to −196°C. Helium (purity > 5.0; Al-phagaz™ He, Air Liquide, France, gas led through a Supelco® 27600-U, Merck, Germany, helium purifier) enters trap 2 at a flow rate of 15 ml/ min and purges the CO 2 through trap 3, the GC column (kept at −15°C ) and trap 4, before it is frozen out again in trap 5. After 45 min, the GC column is purged with He in the reverse flow, heated to 150°C and kept at this temperature until the next sample is being prepared. Helium is pumped away from trap 5. Afterwards, trap 5 is warmed up to −80°C and trap 6 cooled down to −196°C. The yield of CO 2 is determined in the volume of trap 5 and the CO 2 then frozen out in trap 6. Once freezing is complete, trap 6 is isolated from trap 5, warmed up to −80°C and the pure CO 2 introduced into the sample bellow of the mass spectrometer through expansion. HAL is controlled with the software LabVIEW (National Instruments, USA).

Equilibrated gases and gas purification
CO 2 , equilibrated at 1000°C for at least 2 h in a quartz break seal tube (GM Associates, USA), is taken off the muffle furnace and quenched to room temperature within 2 min. The break seal tube is mounted on a tube cracker that is connected to a high vacuum (water trap upstream the turbopump), manually operated gas extraction line. Within < 5 min after having left the muffle furnace, the break seal tube is cracked, and the CO 2 is passed twice over a water trap kept at −80°C . The CO 2 is finally quantitatively transferred into a quartz glass with a pneumatic diaphragm valve (Swagelok, USA) at its top ("autofinger"). The same high vacuum line is used to clean CO 2 equilibrated at 25°C and to transfer it to the autofinger. However, before CO 2 equilibrated at 25°C is allowed to enter the manually operated high-vacuum gas purification line, liquid water is stripped off on a separate, rotary vanepumped gas extraction line by passing the CO 2 five times over a cryogenic water trap kept at −80°C. This way we can avoid that any significant amounts of water enter the high-vacuum gas preparation line.
The autofinger is finally connected to HAL in front of trap 2 (Fig. 1). There, equilibrated gases follow the same preparation pathway as carbonate-derived CO 2 with the exception that they enter HAL isolated from the common acid bath. Equilibrated gases are prepared in amounts equivalent to the CO 2 derived from carbonates (between 80 and 100 μmol).

Mass spectrometric measurements
Isotopic analyses are carried out on a Thermo Scientific™ 253 Plus™ (Thermo Fisher Scientific, Germany) gas source mass spectrometer using its dual inlet system. The 253 Plus™ at the Goethe University has Faraday cups for m/z 44, 45, 46, 47, 47.5, 48 and 49. Cups for m/z 44-46 have 3 × 10 8 Ω, 3 × 10 10 Ω, and 10 11 Ω resistors, respectively, while m/z 47-49 cups are amplified by 10 13 Ω resistors. The m/z 47.5 cup is used to continuously read in the pressure baseline next to m/z 47.
After the sample bellow has been loaded with CO 2 , an equivalent amount (80-100 μmol) of reference gas CO 2 (ISO-TOP, Air Liquide; δ 18 O VSMOW = 25.26 ‰, δ 13 C VPDB = −4.20 ‰) is introduced into the reference gas bellow. Reference and sample gas enter the ion source of the 253 Plus™ gas source mass spectrometer through stainless steel capillaries whose inner surface is coated with inert fused silica (Thermo Fisher Scientific; bre 00014684). Reference and sample gas intensity are automatically adjusted to an intensity of 16,000( ± 100) mV on m/ z 44. Sample and reference gas are measured in 10 acquisitions, consisting of 10 cycles each. Each cycle considers an integration time of 20 s both for the reference and the sample gas, yielding a total integration time of 2000 s per replicate. Idle time is set to 16 s.

Data processing
The common background correction with no gas in the source is not applied. The negative background that is read in on m/z 47.5 is directly added to the raw intensities obtained for m/z 47, m/z 48 and m/z 49 in Isodat using a scaling factor of -1. Acquisition files are then exported to Easotope (release 20190125) (John and Bowen, 2016) which performs the calculation of δ 13 C VPDB , δ 18 O VSMOW , δ 47 , δ 48 , Δ 47, raw and Δ 48, raw using the IUPAC parameters (Daëron et al., 2016).

Results
A total of 76 carbonate reference samples and 38 equilibrated gases were run within a period of 8 weeks in January and February 2019.   Table 1 Δ 47, CDES 90°C and Δ 48, CDES 90°C of carbonate reference materials analyzed in this study. All values are processed using the IUPAC parameters recommended by Daëron et al. (2016) and are reported in ‰-deviation relative to the stochastic distribution. 1 SD and 2 SE denote the 1 σ standard deviation and the 2 σ standard error, respectively. The shot noise limit represents the best analytical precision that can be obtained. It was calculated from Eq. (15) of Merrit and Hayes (1994), considering the total integration time represented by the total number of replicates (2000 s × number of replicates).

Background correction
Fig. 2 displays the correlation between δ 47 and Δ 47, raw for CO 2 equilibrated at 1000°C and 25°C, respectively, as well as for ETH 1 and ETH 2. The slopes of the corresponding regression lines are almost zero, demonstrating that the m/z 47.5 cup accurately reads in the negative background below m/z 47. However, since there is a slight residual slope for the heated gases that is distinguishable from zero, an additional slope correction of Δ 47, raw data needs to be applied (Huntington et al., 2009). To determine this slope, 1000°C and 25°C gas data are merged, adding the average difference between the two corresponding regression lines to the heated gas Δ 47, raw values. This way, a residual slope of 0.00012 ( ± 0.00008) is obtained that exactly matches the residual slope displayed by ETH 1 and ETH 2 raw data (Fig. 2) where m 47 is the residual slope, in our case 0.00012. Correlation plots of δ 48 vs. Δ 48, raw , however, yield significant steeper residual negative slopes (Fig. 3). In addition, the slopes vary slightly between −0.01435( ± 0.00036) for 25°C gases, −0.01474( ± 0.00030) for ETH 1 and ETH 2 and − 0.01560( ± 0.00048) for heated gases. These residual slopes result from an overcorrection of the negative background occurring below m/ z 48 when using the intensity monitored by the m/z 47.5 cup. The merged data set of equilibrated gases is characterized by a slope of −0.01479( ± 0.00027), which is indistinguishable from the slope displayed by ETH 1 and ETH 2 data. As was done for slope correction of Δ 47, raw data, we, therefore, use the slope obtained from the ETH 1 and ETH 2 data (Fig. 3) to correct all sample Δ 48, raw data for the overcorrection induced by the m/z 47.5 intensity:  Dennis et al. (2011), by plotting Δ 47, sc and Δ 48, sc values of equilibrated gases against their corresponding theoretical values. Δ 47, sc and Δ 48, sc values of equilibrated gases are reflected by the intercepts of equilibrated gas data regression lines in Δ 47, raw vs. δ 47 space (Fig. 2) and Δ 48, raw vs. δ 48 space (Fig. 3), respectively. Theoretical Δ 48, 1000°C and Δ 48, 25°C values are not directly available but can be computed from the data provided by Wang et al. (2004) After processing the data of Wang et al. (2004) (their Table 4(I)) through Eq. (4), a best fit polynomial regression yields: According to Eq. (5), the theoretical Δ 48 should be 0.345 ‰ and 0.000 ‰ for CO 2 at temperatures of 25°C and 1000°C, respectively. At these temperatures, the corresponding theoretical Δ 47 values are 0.9252 ‰ and 0.0266 ‰, respectively (Dennis et al., 2011).
In Δ 47, raw vs. δ 47 space, the difference in intercepts between 1000°C (−0.8527( ± 0.0020)) ‰ and 25°C CO 2 (0.0406( ± 0.0021)) ‰ matches the theoretical difference of 0.90 ‰ (Fig. 2). This observation strongly implies that the 253 Plus™ installed at the joint Goethe University -Senckenberg BIK-F stable isotope facility is devoid of any significant scale compression for Δ 47 , contrary to the MAT 253™ that was run in the same laboratory using electropolished nickel capillaries Fig. 2. Plot of Δ 47, raw vs. δ 47 for CO 2 equilibrated at 1000°C and 25°C, respectively, as well as for carbonate reference materials ETH 1 and ETH 2. Δ 47, raw and δ 47 values are relative to the working gas composition. Slopes (m) and intercepts (b) of the corresponding regression lines, as well as their errors, are provided for further information. Fig. 3. Plot of Δ 48, raw vs. δ 48 for CO 2 equilibrated at 1000°C and 25°C, respectively, as well as for carbonate reference materials ETH 1 and ETH 2. Δ 48, raw and δ 48 values are relative to the working gas composition. Slopes (m) and intercepts (b) of the corresponding linear regression lines, as well as their errors, are provided for further information. . For the determination of Δ 47 , CDES 90°C the following empirical transfer function is obtained: In Δ 48, raw vs. δ 48 space, intercepts of −0.3348( ± 0.0076) ‰ and 0.0082( ± 0.0134) ‰ are obtained for CO 2 equilibrated at 1000°C and 25°C, respectively (Fig. 3). Again, the difference between the two intercepts is in perfect agreement with the theoretically expected value of 0.345‰. The hypothesis that the 253 Plus™ installed in our laboratory is not affected by any significant scale compression is, therefore, confirmed by Δ 48 data of equilibrated gases. It follows: Δ 47 , CDES 90°C and Δ 48 , CDES 90°C computed for carbonate reference materials ETH 1, ETH 2, ETH 3, ETH 4 and Carrara are listed in Table 1. For these, it becomes obvious that Δ 48 , CDES 90°C increases with Δ 47 , CDES 90°C . Such behavior is expected if both compositions are largely controlled by temperature. External reproducibilities (expressed as 1 σ standard deviation, 1 SD) range from ± 0.008 ‰ to ± 0.011 ‰ for Δ 47 , CDES 90°C and from ± 0.028 to ± 0.036 ‰ for Δ 48 , CDES 90°C (Table 1). These values are only slightly higher than the predicted shot noise limits of 0.008 ‰ and 0.027 ‰, respectively, which represent the best attainable precision for the chosen analytical conditions of 2000 s integration time per replicate (Merrit and Hayes, 1994). Compared to the MAT 253™ (1SD of 11-16 ppm, see Fiebig et al. (2016)), the external reproducibility for dual inlet-based Δ 47 measurements with the highohmic 253 Plus™ is improved by a factor of 1.5.

Δ 47, CDES 25°C values of ETH carbonate reference materials
Addition of a 25-90°C acid fractionation factor of 0.088 ‰ (Petersen et al., 2019) to the Δ 47 , CDES 90°C values reported in Table 1 yields Δ 47 , CDES 25°C values ( ± 1 SD) of 0.302( ± 0.009) ‰ for ETH 1, 0.303( ± 0.011) ‰ for ETH 2, 0.707( ± 0.009) ‰ for ETH 3, and 0.545( ± 0.009) ‰ for ETH 4. These values are significantly different from those reported in Bernasconi et al. (2018), even if 0.004 ‰ are added to the latter data to account for the unique temperature dependence of the acid fractionation factor (note that Bernasconi et al. (2018) reacted at 70°C and that a 25-70°C acid fractionation of 0.062 ‰ was applied to their data, which slightly deviates from the 0.066 ‰ reported in Petersen et al., 2019). We speculate that the observed differences may arise from the circumstance that in the analytical setup used by Meckler et al. (2014) and Bernasconi et al. (2018) CO 2 gas derived from the ETH standards did not enter the ion source of the gas source mass spectrometer through the same stainless steel capillary as the equilibrated gases. If small amounts of water were adsorbed at the inner surface of the sample gas capillary, partial re-equilibration of CO 2 at ambient temperature may happen while the CO 2 passes through the capillary, finally introducing a compression of the Δ 47 -scale. Partial reequilibration may attain a steady state as long as the water content and the gas flux through the capillary are constant. Under such conditions, sample-derived CO 2 can still be accurately projected onto the CDES if the equilibrated gases enter the ion source through the same capillary as the sample gases. However, the same correction procedure may deliver inaccurate results if two different capillaries with distinct water contents and fluxes are used, as already stated by Bernasconi et al. (2018 Should equilibrium really have been attained in ETH 3, ETH 4 and Carrara, it would have been established at temperatures of 15( ± 2)°C, 86( ± 4)°C, and 210( ± 7)°C, respectively, as reflected from measured Δ 47 , CDES 90°C values (Fig. 4). However, considering the error in Δ 48 , CDES 90°C , we alternatively cannot rule out that the true Δ 47 , CDES 90°C and Δ 48 , CDES 90°C compositions of Carrara and ETH 4 plot slightly below or above the equilibrium line, respectively (Fig. 4). Generally, departures from equilibrium would be introduced by rate-limiting kinetic effects. It has recently been shown that the most prominent rate-limiting processes involved in carbonate precipitation, CO 2 absorption and CO 2 degassing, follow specific trajectories in Δ 47 , CDES 90°C vs. Δ 48 , CDES 90°C Fig. 4. Plot of Δ 48 , CDES 90°C vs. Δ 47 , CDES 90°C for ETH 3, ETH 4 and Carrara. Within errors (2 SE, 95 % confidence level), all data points agree with the theoretical equilibrium curve after adding the experimentally determined phosphoric acid fractionation factors of 0.136 ‰ and 0.196 ‰, respectively, to the theoretical Δ 64 and Δ 63 values for calcite (Hill et al., 2014).
space (Guo and Zhou, 2019). Hence, even if kinetics were of any importance during carbonate mineralization, it should be possible to determine accurate formation temperatures, along with the rate-limiting step governing isotopic exchange, solely based on the position of measured data in Δ 47 , CDES 90°C vs. Δ 48 , CDES 90°C space (Guo and Zhou, 2019).

Conclusions
We describe an analytical setup that allows accurate and precise determination of Δ 48 along with Δ 47 in CO 2 evolved from phosphoric acid digestion of carbonates. External reproducibilities for Δ 48 and Δ 47 analyses occur close to the corresponding shot-noise limits. There is no indication that the chosen setup introduces any significant artificial biases in the measured abundances of mass 47 and mass 48 CO 2 isotopologues.