Experimental

3.2.1 Samples and sample preparation

Two different standard materials were analyzed for their Δ47 signatures:

• NBS 19 (calcite, supplied by the IAEA, Vienna, Austria, and by the National Bureau of

Standards, Gaithersburg, MD, USA) is a coarse-grained white marble with δ¹⁸O = –2.20‰ (V-PDB) or +28.64‰ (V-SMOW) and δ¹³C = 1.95‰ (V-PDB). NBS

19 is used as an interlaboratory standard for clumped isotope analyses (Ghosh et al., 2006a; Schmid and Bernasconi, 2010; Dennis et al., 2011). For digestions at 25 °C, NBS 19 was crushed to a fine powder and homogenized using a mortar and pestle.

When not treated that way, the reaction with 104% H3PO4 was observed to be very slow and, sometimes incomplete, as indicated by δ¹³C values significantly lower than 1.95‰.

• Arctica islandica, a bivalve with an aragonitic shell, grew near Langanes, NE Iceland, in ca. 30 m water depth at an average temperature of 6.0 ± 0.5 °C. The specimen was collected alive in August 2006. Sample material was taken from the outer layer of the most recently formed part of the shell. For this purpose, the shell was cut into two parts along the direction of growth from the umbo to the ventral marging using a Buehler low-speed precision saw equipped with a 0.4 mm thick diamond-coated saw blade.

Subsequently, both shell slabs were cut perpendicularly to the growth lines. The youngest part of the valve was used for clumped isotope analyses. The periostracum and the inner shell layer were physically abraded before the two shell pieces were ground and homogenized in an oscillating disc mill. Some aliquots reacted at 25 °C were pretreated for 2 h with 1.5% H2O2 prior to the analysis to remove organic matter, as organic matter has been found to preclude accurate Δ47 analysis of some biogenic carbonates (Huntington et al., 2009).

3 Clumped isotope analysis of carbonates: comparison of two different acid digestion techniques

3.2.2 Acid digestion

The concentration of the phosphoric acid used for digestion reactions was 104%. This corresponds to a density of 1.91 g/cm³ (for the conversion of acid concentrations into acid densities we use the equation determined by Ghosh et al. (2005): y = 0.0114x + 0.723, where y is the acid density in g/cm³ and x is the acid concentration in %). The acid was prepared by slightly modifying the method of Coplen (1983): 99% H3PO4 (Merck KGaA, Darmstadt, Germany, ≥99%) was heated to 150 °C, and then P4O10 was slowly added to the stirred solution. After 10 h an acid aliquot was cooled to room temperature for about 1 h. The acid density was measured gravimetrically at 25 °C, and once a concentration of 104% was achieved, the acid was heated (at 150 °C) for at least three additional hours. Otherwise, either more P4O10 or distilled water was added until the concentration reached 104%. The acid was stored for at least two weeks to ensure that all the water had reacted with P4O10 to H3PO4

(Coplen et al., 1983). Before using the acid for digestion reactions, the density was controlled gravimetrically again.

Carbonate digestions at 25 °C were performed in McCrea-type reaction vessels (McCrea, 1950). 4 to 14 mg aliquots of carbonate were reacted with 7 ± 1 mL 104% H3PO4. The carbonate was placed into the side arm and the acid was filled in the main tube of the reaction vessel, before evacuating it for 2.5 to 3.5 h to reach a vacuum better than 10^–3 mbar. Before starting the reaction by tipping the acid over the carbonate, the vessels were placed in a common water bath at 25.0(±0.2) °C for 1 h to ensure thermal equilibration. During the reaction, the vessels were kept in the water bath at 25 °C. The reaction time was between 16 and 20 h.

Reactions at 90 °C were performed using an automated acid bath whose design is very similar to those used at Caltech and Johns Hopkins University (Passey et a, 2010). The acid bath was was placed in a copper block surrounded with a heating band to keep the acid at a constant temperature of 90.0(±0.1) °C. The carbonate samples were filled into small Ag capsules (IVA Analysentechnik e. K., Meerbusch, Germany, art. no. 184.9921.36). These were loaded into a Zero Blank Autosampler (Costech, Valencia,CA, USA). The acid bath and the autosampler were pumped by a turbomolecular pump backed by a Membrane pump (Pfeiffer, Aßlar, Germany). After a sample was dropped in the acid, the evolved CO2 was frozen in a U-trap cooled at liquid nitrogen temperature after passing another cold trap held at –80 °C to remove trace amounts of water. By monitoring the pressure of the evolved CO2 we observed that the digestion of the fine grained carbonate powder of A. islandica takes only ~10 min, while the larger crystals of NBS 19 reacted for ~20–30 min. After manometrically determining the yield,

3 Clumped isotope analysis of carbonates: comparison of two different acid digestion techniques

the extracted CO2 was frozen into a transportable glass finger equipped with high-vacuum valves (Louwers, Hapert, The Netherlands, art code 40.200.8).

3.2.3 CO2 cleaning procedure

The CO2 derived from phosphoric acid digestions at both 25 and 90 °C was cleaned following the procedure described in Ghosh et al. (2006a). Briefly, the transportable glass finger containing the sample CO2 was connected to a cryogenic vacuum extraction line, evacuated to

<10^–6 mbar using a turbomolecular pump supported by a membrane pump (Pfeiffer). The CO2

yield was determined with a capacity manometer, and then CO2 was passed twice over a trap held at –80 °C before being frozen back into the volume of the glass finger. Afterwards, the sample was passed through a gas chromatograph (GC) purification system to remove traces of hydrocarbons. The CO2 was entrained into a He carrier gas flow (18 mL/min; purity:

99.9999%) and purged through a 1.20 m x 2.15 mm i.d. stainless steel column packed with Porapak Q 80/120, kept at –20 °C. Before entering and after leaving the GC column the He-CO2 gas mixture passed additional water traps held at –77 °C. The CO2 effluent from the GC column was collected for a period of 30 min in a U-trap immersed in liquid nitrogen. After one sample had passed through the GC column the He flow was switched to a backflush mode, enabling the water traps and the column to be heated to 25 and 150 °C, respectively, for at least 15 min. In a final step, the He carrier gas was pumped away and the CO2 was cryogenically purified one final time using the vacuum extraction line described above. For isotopic analysis, the purified CO2 was transferred to a transportable glass finger that could be connected to the dual inlet system of the mass spectrometer.

3.2.4 Measurements

Isotopic analyses were performed at Goethe University (Frankfurt, Germany) on a MAT 253 gas source isotope ratio mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with dual inlet system and six Faraday cup collectors for masses 44 to 49 (resistors:

3 x 10⁸ Ω, 3 x 10¹⁰ Ω, 10¹¹ Ω for masses 44–46, respectively, and 10¹² Ω for masses 47–49).

The original stainless steel capillaries were replaced with 4 feet long electroformed nickel (EFNi) capillaries (VICI AG, Schenkon, Switzerland; 1/32" o.d., 0.005" i.d., art. no.

TEFNI.505, 122 cm x 0.127 mm ID; Passey et al., 2010).

Measurements were performed using the dual inlet system, after adjusting the sample and reference gas signals of mass 44 to (16000 ± 150) mV. Ten acquisitions consisting of ten cycles with an ion integration time of 20 s each, were used for all measurements,

3 Clumped isotope analysis of carbonates: comparison of two different acid digestion techniques

corresponding to a shot-noise limit of ~0.008‰ (Merritt and Hayes, 1994). Before each acquisition, peak centering, background determination and pressure adjustments to 16 V at mass 44 were carried out. The total analysis time of one sample was about 3 h. CO2 from Oztech (Safford, AZ, USA; δ¹⁸O = +25.01‰ vs V-SMOW; δ¹³C = –3.63‰ vs V-PDB) was used as the reference gas.

3.2.5 Data processing

We report Δ47 values on the absolute scale of Dennis et al. (2011). In comparison to the data correction procedure described by Huntington et al. (2009), absolute scaling of raw Δ47 values has the advantage that the processed data becomes comparable between labs as the Δ47

composition of the CO2 reference gas (which may vary between labs) is considered. Briefly, raw data is corrected in two steps. (1) To correct for non-linearities of the mass spectrometer, CO2 gases of different bulk isotopic compositions were measured. Prior to isotopic analysis, these were heated in quartz break-seal tubes to 1000 °C for more than 2 h to reach the characteristic distribution at this temperature. After quenching to room temperature, the gases were purified cryogenically and by GC like carbonate samples. (2) Δ47 values corrected for non-linearity were then converted to the absolute scale using the empirical transfer function (ETF). The ETF is determined by plotting the intercepts of linearity lines (derived from measurements of CO2 gases of distinct bulk isotopic compositions equilibrated to at least two different temperatures) against the corresponding theoretically expected values (Wang et al., 2004). We have determined our ETF using the intercepts of CO2 gases heated at 1000 °C and equilibrated with water at 25 °C. CO2 and H2O were enclosed in glass tubes, which were placed in a waterbath for at least three days at 25.0 ± 0.2 °C. After quenching with liquid nitrogen, the glass was held in an ethanol/dry-ice slush at –80 °C to release CO2, while H2O remained frozen. Separation between water and CO2 was improved by passing the gas at least five times over a trap held at –80 °C using the vacuum extraction line. Afterwards, water-equilibrated gases were cleaned and measured like sample gases (Dennis et al., 2011).

Carbonates were reacted at 25 and 90 °C between April 2011 and January 2012. Instrument non-linearity was continually monitored by measuring heated gases daily and comparing the slopes of heated gas regression lines determined from blocks of nine consecutive heated gas analyses. As long as slopes of these single blocks were identical within the standard errors, all the analyses were used for the determination of the heated gas slope. If the slopes of the different blocks of analyses were changing significantly, new correction parameters were determined for each day using the running average of the heated gas line of nine consecutive

3 Clumped isotope analysis of carbonates: comparison of two different acid digestion techniques

analyses. [For example, for linearity correction of the NBS 19 sample analyzed on 04 November 2011 the slope of the heated gas line (m = 0.0250) was derived using heated gas analyses carried out between 24 October 2011 and 16 November 2011 (Supplementariy Table S3.1, see Supporting Information)]. The δ⁴⁷ and Δ47 values of heated gases were, however, highly correlated, with R²-values always better than 0.99. The slopes of the heated gas lines which were used to correct carbonate data are listed in Table 3.1 along with raw and corrected carbonate Δ47 data.

For the determination of the ETFs we considered the intercepts of linearity lines of heated gases analyzed over a period of two months. The intercepts for April/May and June/July 2011 are statistically indistinguishable (Supplementary Table S3.1, see Supporting Information).

After source cleaning a significant change in the heated gas intercept between September/October 2011 and November/December 2011 was observed, while no change occurred between November/December 2011 and January/February 2012 (Supplementary Table S3.1, see Supporting Information). Gases equilibrated at 25°C were measured frequently in November/December 2010 and, after source cleaning in September 2011. In addition, some gases were analyzed temporarily (February, March, April, July and November 2011, Supplementary Table S3.1, see Supporting Information) to test the consistency of the Δ47

composition of the reference gas. The intercept of the line determined by 25 °C water-equilibrated gases was constant at –0.03‰ between November 2010 and July 2011,

while it was –0.06‰ after source cleaning. We therefore applied three ETFs to correct our

carbonate data: y = 1.1094x + 0.9585 between November 2010 and July 2011, y = 1.0959x + 0.9910 for September 2011 and y = 1.1521x + 0.9943 from November 2011 to

January 2012 (with x: intercepts of the equilibration and heated gas lines and y: theoretical equilibrium value for Δ47 at the corresponding equilibration temperature).

For the normalization of Δ47 values to acid digestions at 25 °C a difference in acid fractionation factors of +0.081‰ (Passey et al., 2010) was applied to data obtained from carbonates reacted at 90 °C, as used by Dennis et al. (2011). It should be noted that the difference of 0.081‰ was determined on the internal ‘Caltech scale’. If this value is projected in the absolute reference frame using the slope of the secondary reference frame transfer function provided by Dennis et al. (2011), a value of 0.084‰ is obtained instead. Considering the precision of Δ47 analysis as represented by our shot noise limit, this value is indistinguishable from the +0.081‰ reported by Passey et al. (2010).

3 Clumped isotope analysis of carbonates: comparison of two different acid digestion techniques