Trace element analysis

Cleaned and dissolved planktonic G. ruber were analysed for a suite of trace element to calcite ratios at three laboratories. This resulted from splitting samples into three batches. These include trace element-only samples, boron-isotope samples for scoping out the potential of the core using a conventional method, and boron isotope samples using a new boron separation method. To gain a goodδ¹¹B-based seawater pH reconstruction, trace elements must be measured on direct splits of the boron samples. We therefore combined the trace element data from the trace-element only samples with trace element data fromδ¹¹B analysis.

Trace element-only samples (n=190) were analysed on a Thermo Scientific ELE-MENT XR mass spectrometer at the Cardiff University CELTIC laboratories, while trace element data from δ¹¹B analyses were collected at the STAiG laboratories in St. Andrews, and the Foster Lab laboratories in Southampton. In St. Andrews, trace element samples (n=198) were analysed on an Agilent Triple Quadrupole mass spectrometer. In Southampton samples (n=36) were measured on a Thermo Scientific Element XR mass spectrometer. In all three laboratories, samples were spun in a centrifuge prior to analysis to reduce risk of undissolved particle contamination.

Then an aliquot of 10µL in Cardiff, 3µL in St. Andrews, and 20µL in Southampton (Henehan et al., 2015), was diluted and pre-run for calcium (Ca) concentrations. In Cardiff, matrix matched standards were created for each sample below a concentra-tion of 2mM calcite. Samples with a concentraconcentra-tion >2mM calcite were diluted to set standard concentrations at 2mM, 2.5mM, 3mM, 3.5mM, and 4mM. In St. Andrews,

and Southampton (Henehan et al., 2015), samples were matched to a 0.5mM or 1mM Ca standard, depending on boron concentrations of the sample. The following ratios were collected at all three institutions: Mg25/Ca43, B11/Ca43, Sr88/Ca43, Cd111/Ca43, U238/Ca43, Al27/Ca43, Mn55/Ca43, Li7/Ca43, Nd146/Ca43, and Ba138/Ca43 (Figure 5.7).

Figure 5.7 Benthicδ¹⁸O with a complete set of U1476 trace element/Ca ratios across the last 1.6Ma measured for this thesis. Samples were analysed in three locations including Cardiff University (filled circles with dark blue outline), University of St Andrews (squares with black outline), and National Oceanographic Centre Southampton (triangles with pink outline).

Long-term reproducibility from in-house standards for Cardiff, St. Andrews, and Southampton are listed in Table 5.1 at the end of this subchapter. For comparison, average relative standard deviations (RSD) from the sample batches are listed in Table 5.2. Long-term reproducibility for the Foster Lab in Southampton is also referenced in Henehan et al. (2015).

Raw-data quality was tested by calculating the signal/noise ratio for B11/Ca43 as

5.3 Sample preparation and analyses 69

B11_{Sam ple}[cps]

B11previousblank[cps] (5.1)

and the sample-standard matrix match in [%] as

Ca43_{Sam ple}[cps] − Ca43previousst andard[cps]

Ca43previousst andard[cps] ∗ 100 (5.2)

where [cps] stands for counts per second.

I chose these quality checks, because boron concentrations can be very small in samples and provide a good measure of signal/noise ratio when compared to blank measurements. Additionally, it is important that sample and bracketing standard are matrix matched well, since poor matrix matching can lead to matrix effects on the mass spectrometer and may influence the final measurement of concentration or isotopic signature if the plasma conditions and long-term machine drift are somewhat unstable. Samples which had a signal/noise ratio for B11 <5, or a matrix match >30%

were scrutinised and considered for exclusion (Figure 5.8). Data points were only excluded in circumstances where all element ratios were well out of the naturally occurring range of that ratio.

Some samples experienced difficulties during the analysis (see Figure 5.9). These included samples that were (i) run with a dirty blank pot, (ii) left uncapped for 24 hours due to plasma shutdown, (iii) showed signs of dilution problems or high contamination tracer concentrations, or (iv) were re-picked and re-run after sample loss during preparation. Dirty blank pot samples (i) were blank corrected using an average of the clean blanks also used in the run as a better representation of blank.

During one run, the plasma shutdown due to a pressure drop in argon gas. These samples (ii) were re-run in their respective beakers 24 hours later using the blank and standard pots from the original run. As such, contamination accumulation over the 24 hours was also representative in the blanks and standards and was corrected

Figure 5.8 sigal/noise ratio for B11, and standard-sample matrix offset for the three sample batches run in Cardiff (orange circles), St. Andrews(blue squares), and Southampton (green triangles). The red line, and grey bars indicate boundary levels, respectively. Samples that fall below the red line, or within the grey bars were subject to quality scrutiny.

for through the bracketing method. Samples with dilution problems (iii) can be problematic, because of differences in the standard and sample calcium concentration.

This can lead to errors in the sample measurement, as not all matrix effects are accounted for by the standard bracketing. If matrix effects are small during certain runs, moderate sample-standard differences may be acceptable. Therefore, I did not correct or exclude any wrongly diluted samples. Samples with high contamination tracers (Al/Ca, Fe/Ca, Mn/Ca) were only excluded if a strong correlation existed between the contamination tracer and the climate variable. This was never the case.

Some samples experienced high rates of carbonate loss during the preparation steps.

These samples were re-picked and re-run. No further issues occurred to the re-picked samples(iv). Again, data points were only excluded if trace element ratios plotted well outside the naturally occurring range of that ratio. In the final record, one Mg/Ca data point was excluded due to plotting four times higher than the natural range, and one B/Ca data point was not analysed by the mass spectrometer.

5.3 Sample preparation and analyses 71

Figure5.9Plotofsampleshavingexperiencedmeasurementdifficulties.Tohighlighttherespectivesamplesintheirtemporalspace,theywereplotted representativelyasvaluesofδ18 Obenthic(a).17largeorangecirclesshowsamplessubjecttodirtyblankpots.14largegreycirclesshowsamplessubject tomassspectrometershutdownandthereforelongexposuretolaboratoryair.2largeyellowcirclesrepresentsampleswithconcentrationmismatch betweensamplesandstandards.1largelightbluecirclerepresentsasamplesubjectedtohighcontamination.1largelightgreencircleshowstheexcluded datapoint.3largedarkbluecirclesrepresentdatapointsthatwerere-run.1largebrowncircleshowsadatapointwhereB/Cawasnotmeasured.For comparison,smallgreycirclesshowalldatapointsmeasuredfortraceelementsinthisthesis,andsmallbluecirclesshowallsamplesmeasuredforstable isotopes.Thisalsorepresentsthesamplingdensityofthecore.InfluenceofsamplemeasurementdifficultiesonMg/CaandB/Caarevisiblein(b)and(c).