Chemical separation and MC-ICPMS analysis of U, Th, Pa and Ra isotope ratios of carbonates†
Diagenetic alteration of fossil reef corals may have severe effects on the reliability of 230Th/U-ages. The widely applied criteria introduced to test for the reliability of coral 230Th/U-ages are often not sufficient to identify all altered ages. The combination of 230Th/U- with 231Pa/U-dating has been suggested as a promising method to identify altered ages. Furthermore, 226Ra/230Th ratios can provide information about the diagenetic history during the last 10 000 years. Here we present a method to chemically separate U, Th, Pa and Ra from the same aliquot of a carbonate sample. The isotope ratios of our mixed U–Th–Pa–Ra spike are calibrated using a secular equilibrium material treated in the same way as the samples during chemical separation and mass spectrometric analysis. This approach does not only circumvent corrections for the radioactive decay of the short-lived 233Pa spike, but also enables us to estimate the reproducibility of the spike calibration. The relative standard deviation (RSD) of the spike ratios is 0.27% for 229Th/236U, 4.2% for 228Ra/236U and between 0.6 and 4.0% for 233Pa/236U. The RSDs of the final ratios are 1.2% for 231Pa/235U and 3.4% for 226Ra/230Th. We assess the individual sources of uncertainty (e.g., mass spectrometric corrections, counting statistics, and spike calibration) in detail and quantify their contribution to the total uncertainties of the final isotope ratios. Most corrections contribute only moderately to the final uncertainties. However, in the case of a large abundance of natural isotopes in the spike, this correction can have a large influence on both the ratios and their uncertainties. Another parameter affecting the final uncertainties and the reproducibility of the method is the sample size and, thus, the concentration of the natural isotopes. This discussion provides useful guidelines for future applications and can be adjusted to the individual requirements of a specific user.