where R47 is the measured 47/44 ratio of the sample and R47* is the calculated stochastic abundance based on the bulk isotopic composition of the sample as measured by the δ13C, δ17O and δ18O values. Note that in low mass resolution IRMS instruments analytically the 47/44 ratio contains minor contributions from the species 18O12C17O and 17O13C17O. Conventionally the Δ value is multiplied by 1000 and expressed in per mille. Δ48 and Δ49 are defined by directly analogous relationships.
A contaminant species with a major ion peak at mass 47 need only be present at a relative partial pressure with respect to the analyte CO2 of just 4.4 × 10-10 to cause systematic offsets in the data of 0.01‰. These are at the measurement precision required for clumped isotope geothermometry and therefore significant.
The potential for contamination to bias results has been widely recognised and tests to evaluate the presence or otherwise of contaminant gases proposed {Eiler:2004ue, Affek:2006ws, Huntington:2009jw, Petersen:2016kt}. Whilst not as yet identified the most likely contaminants include halocarbons, hydrocarbons and sulfur compounds. Many of these have characteristic fragmentation patterns and mass spectra. For example chlorinated compounds, including CH3Cl and CH2Cl2 have strong peaks at 47 (12C35Cl), 48 (12C1H35Cl) and 49 (12C37Cl and 12C1H235Cl). Thus we may expect to see patterns of covariation amongst Δ47, Δ48 and Δ49 that are characteristic of the contamination. Moreover the relative abundance of CO2 isotopologues at masses 48 and 49 are just 4 × 10-6 and 4 × 10-8 respectively and thus should be very sensitive to the presence of contamination. For these reasons Eiler and Schauble (2004) recommended that the Δ48 and Δ49 values be routinely analysed and used as identifiers of potential contamination.
As a test for this procedure Eiler and Schauble (2004) added measured quantities of dichloromethane (CH2Cl2) to pure CO2 and monitored the variation of Δ47 and Δ48 as a function of the intensity of the mass 49 ion beam. Both are observed to vary as linear function of the 49 intensity.
Many published studies state that they monitor Δ48 and Δ49 values, or other parameters related to the mass 48 and 49 signals such as as…...as a test for contamination. In few, however, are the values reported. Inspection shows that where values are indicated, for example in data plots, measurement precision and accuracy is limited (e.g. references……). Undoubtedly this is related to the difficulty of making accurate and precise measurements of the higher mass isotopologues of CO2, particularly where the instruments are strongly non-linear as a result of the pressure base line (PBL) effect.
An example is the recent study by Petersen et al. (2016). This study was aimed at specifically evaluating aspects of the performance of a static porapak trap designed to eliminate contamination from CO2. Porapak in either packed GC columns, or used as a static trap in vacuo is used to remove contamination from samples for clumped isotope analysis (refs). Petersen et al. used the Δ48 value as represented by the difference in Δ48 between a sample and pure reference gas, or between an individual sample value and the mean value for samples run through the porapak trap at temperatures below -10°C. They argue that samples passed through their porapak trap had Δ48 values within acceptable limits whilst those that bypassed the trap were outside an unspecified cut-off value. Inspection of their Figure 2 shows that samples that were deemed clean had Δ48 values between -2‰ and +2‰ whilst those that were contaminated had values >2‰.
A second effect reported by Petersen et al. (2016) is due to fractionation of the CO2 isotopic composition as a result of adsorption onto the porapak resin and subsequent incomplete recovery. This effect becomes significant at trap temperatures less than -20°C. The fractionation pattern is systematic with an increasing depletion in δ45, δ46 and δ47 as temperature decreases. Maximum depletions of -0.2, -0.8 and -1.0‰ respectively were reported at a trap temperature of -40°C. The effect of fractionation on the final Δ47 value is a depletion of as much as -0.15‰. blah blah blah
Here we present new data obtained whilst evaluating and comparing the performance of different porapak traps for removing contamination. We were prompted to do this study after adapting our sample processing procedure to that used by Petersen et al. (2016) in order to reduce our sample processing time and increase sample throughput. In doing so we observed significant and systematic shifts in recorded Δ47, Δ48 and Δ49 values for standard samples that indicated contaminant breakthrough from the porapak trap. The pattern of co-variation between Δ47, Δ48 and Δ49 is consistent with chlorocarbon contaminant species in the ion source. Careful measurement of the mass difference between the mass 49 CO2 isotopologue and the contaminant species at 49 identifies the key species as CCl (present at masses 47 and 49) and CHCl (present at mass 48). In addition we have measured the effective adsorption coefficient of CO2 onto porapak at different trap temperatures and determined the minimum time for 100% recovery of the sample CO2 whilst maintaining efficient separation of CO2 and impurity species.
We have been unable to identify the precursor molecule or the source of the contamination. The fragmentation pattern is not dissimilar to that of dichloromethane and may indicate the presence of cleaning solvents. We have, however, not been able to rule out that a further potential source is the ortho-phosphoric acid and samples used in carbonate analyses. Irrespective of this, in order to adequately assess the level of contamination in samples it is necessary to make precise measurements of Δ48 and Δ49. Results for Δ48 and Δ49 are not of sufficient quality to rule out contamination as a significant feature of published data sets. The implications of these findings for sample preparation and the interpretation of existing studies are discussed.
## 2 Experimental##
### 2.1 Samples, reference gases and orthophosphoric acid
We used two each of natural inorganic and biogenic calcites in these tests:
*UEACMST* was prepared from a block of Carrara Marble by crushing, homogenising and then sieving to select the 125-180$\mu$m size fraction. It has been one of the SIL internal standards since 1989 and has $\delta^{13}$C = 1.95‰~VPDB~ and $\delta^{18}$O = -2.2‰~VPDB~. Its $\Delta_{47}$ value on the absolute reference frame, based on the long term average over the past two years, is 0.382 $\pm$ 0.012‰ (n=57).
*UEATHC* is a modern calcite that precipitated in a pipe supplying natural thermal waters (56$^\circ$C) to a spa in xxxx, Turkey. The calcite has a concentric precipitation pattern with layers picked out by varying degrees of discolouration due to the presence of oxidised iron.
*UEAOES* is a modern ostrich egg shell of unknown provenance.
*UEABEL* is a carbonate prepared from belemnites (*Belemnitella mucronata*) collected from the Campanian, Upper Cretaceous at Weyborne Hope, North Norfolk. Fragments were carefully selected avoiding any specimens that showed visible indication of alteration. These were cleaned in dilute HCl in an ultrasonic bath, rinsed in deionised water, dried and ground to a fine powder.
*BDHequilRT* and *BDHrefgas-C* are gases prepared by reaction of BDH marble chips with 85% orthophosphoric acid and equilibrated with the product water from the reaction at 20$^\circ$C for a period of approximately three months. For each gas approximately 5g of crushed BDH marble chips are reacted with 4 $\times$ 10^-2^ mols of orthophosphoric acid at 85% concentration. This ensures that the carbonate is in excess and at the end of the reaction the resulting solution is near neutral. The reaction is carried out on a shaking table in an evacuated 2L flask for a duration in excess of three months. The gas is then dried, initially by passing through a large volume cryotrap at -79$^\circ$C, then expanding into a 25L volume to reduce the CO~2~ pressure to <50hPa before passing through two further traps at -110$^\circ$C. The gas is then passed over a 22cm x 0.4cm i.d. porapak column at -2$^\circ$C. We test for purity of the gas by taking 2cm^3^ aliquots and passing these through the porapak trap a second time. These aliquots are compared with the parent gas to ensure that there is no further change in bulk or, reduction in clumped isotopic values, including $\Delta_{48}$ and $\Delta_{49}$ as a result of the
second stage cleaning of the aliquot.
We use BDHrefgas-C as the daily working reference gas (wrg) in the MIRA mass spectrometer and regularly measure single aliquots of BDHequilRT to check for drift in the wrg over time. During the duration of the experiments we detected no drift in the wrg isotopic composition or degree of clumping as represented by the $\Delta_{47}$, $\Delta_{48}$ and $\Delta_{49}$ values.
2.2 Processing of carbonate samples
Carbonate samples are reacted under static vacuum conditions in sealed reaction vessels. The initial experiments were carried out using 6-8mg of sample weighed into one leg of two-legged reaction vessels with 1mL of 102% orthophosphoric acid aliquoted into the other leg. The vessels are then evacuated at room temperature to a vacuum better than 7 × 10-6 hPa over a period of 4 hours, isolated, then placed in a water bath at 25°C and allowed to equilibrate to the reaction temperature for a period of 30 minutes. The acid is then tipped onto the sample and the reaction allowed to proceed for 14-20 hours.
Following a major upgrade to the mass spectrometer ion source with sensitivity gains of a factor of 3.5 in January 2016 the sample mass was reduced to 3-4mg. Otherwise all details of the reaction are the same.
Gas drying and purification is carried out on a manual preparation line, Figure 1. The line is maintained at a vacuum better than 3 × 10-7 hPa. For the experiments reported here we used three variants of a static porapak trap design.
1. Initial results (July-August 2015) were obtained using a static 210mm x 4mm i.d. U-shaped glass trap packed with porapak resin (details here…). The trap was wrapped in a copper jacket and cooled by evaporation of liquid nitrogen using a Brenninkmeijer type system {Brenninkmeijer:1982wq}. The nominal trap temperature was -20°C but control proved difficult and could vary between -20 to -30°C. 100% recovery of the CO2 could be achieved with transfer times of 80 minutes.
2. Following publication of the Petersen et al. (2016) study we changed the porapak trap for a 70mm x 4mm i.d. straight glass trap inserted into an Al block and cooled by peltier heat pumps. The first experiments were carried out at a temperature of -7°C (limited by the air cooling of the peltier devices) and later at -20°C (after modifying the trap design to include water cooling). 100% recovery was achieved with a 30 minutes transfer time.
3. A final trap design consisted of two 120mm x 4mm i.d. glass traps arranged in series and inserted into a single Al block. The trap is cooled by peltier heat pumps to -20°C with temperature control better than ±0.1°C. The transfer time through this trap is 100 minutes.
Between samples the traps are baked at 120°C for 30 minutes with a longer overnight bake at the same temperature. The vacuum whilst baking recovers rapidly to better than 6 × 10-7 hPa.
The steps in purification are: (i) distill the sample CO2 into trap 1 with liquid N2 for 5 minutes, then remove non-condensable gases; (ii) sublimate the CO2 from trap 1 at -110°C through trap 2 (-110°C) whilst cryo-distilling with liquid nitrogen into the cold finger and Barocell volume for a period of 5 minutes; (iii) pump away non-condensable gases, isolate the gas in the Barocell volume, warm to room temperature and measure the sample pressure; (iii) cry-distill the CO2 via the porapak trap into the gas tube for transfer to the mass spectrometer. The duration of this final cryo-distillation varies depending on the length of the trap (see above).
2.3 Mass spectrometry
All samples were measured on the UEA MIRA dual-inlet isotope ratio mass spectrometer. This instrument is an in-house design with a 120° extended geometry, Nier type ion source and 6 faraday collectors with a working mass resolution of 250. Signal gains of 108, 1010 and 1012 are used for masses 44, 45 and 46, and 47-49 respectively. The instrument is linear with respect to Δ47, Δ48 and Δ49 values.
Measurements are made with accurately matched reference and sample gas pressures and volumes thus ensuring identical sample depletion rates through each acquisition cycle. The measurements are made at an initial major beam intensity of 7.5 × 10-8A and consist of 4 acquisitions, each of 20 sample-reference pairs. Signal integration for each sample and reference cycle is 20s with a dead-time of 10s following switching of the change-over valve. Between each acquisition the sample and reference beams are balanced to the initial signal strength of 7.5 × 10-8A. Total depletions through a complete acquisition cycle are on the order of 10%. Internal precisions are close to the shot noise limit for Δ47 and Δ48 at ± 0.008‰ and ± 0.025‰ respectively. For Δ49 precisions are typically ± 4-5‰.
Δ values are calculated as:
where R13, R17, R18 are the measured 13C/12C, 17O/16O, 18O/16O ratios. R47 is the measured 47CO2/44CO2 ratio assuming a stochastic distribution of isotopes in the mass spectrometer working reference gas. The data processing algorithms are implemented in a Mathematica™ program (see supplementary information).
We report the raw Δ47, Δ48 and Δ49 values measured on the local reference frame with respect to the mass spectrometer working reference gas _BDHrefgas-C_. In April 2016, part way through this study, we started to use a new working reference gas, _BDHrefgas-D_. This gas differs in its Δ47 value from reference gas C by -0.041 ± 0.012‰. To account for this step change in the working reference gas composition all measurements made with reference gas _BDHrefgas-D_ have been corrected using the equation:
‰
Direct comparison of the data reported on the local scale for Δ47, Δ48 and Δ49 is facilItated by the fact that both the clumped values of the working reference gases and the mass spectrometer scale compression (= 1.291) remained stable throughout the 12 month duration of this study.
3 Results
3.1 CO2 adsorption onto porapak resin
The results evaluating the adsorption and desorption of CO~2~ onto porapak resin are presented in Figure 2(a) and (b). Approximately 2 cm^3^ of CO~2~ was expanded through to the 24cm long porapak trap (trap 3 described above) and the pressure monitored as a function of the trap temperature. In the absence of porapak the initial CO~2~ pressure of 29.6 hPa is estimated from the known volumes of sections of the sample preparation line used in the experiment. At room temperature, 20$^\circ$C, the observed pressure is 27.6 hPa, Figure 2(a). We infer that approximately 7% of the CO~2~ is adsorbed onto the porapak. Cooling the trap temperature in increments of between 5 and 10$^\circ$C to close to -30$^\circ$C results in a rapid fall in pressure that is commensurate in rate with the fall in trap temperature. The new temperature and pressure stabilises within approximately one minute. Reversing the temperature cycle and warming the trap in increments shows adsorption to be completely reversible. At -20$^\circ$C the pressure of CO~2~ in equilibrium with gas adsorbed on the porapak is 17.5 hPa, corresponding to adsorption of 41% of the total amount of gas.