Quartz luminescence dating of Anglian Stage (MIS 12) fluvial sediments: Comparison of SAR age estimates to the terrace chronology of the Middle Thames valley, UK

Quartz optically stimulated luminescence (OSL) is increasingly being used to constrain the depositional age of fluvial and glaciofluvial sequences over orbital (Milankovitch) timescales within the British Isles. Few of these previous studies have had any age control; however there is some evidence that OSL ages based on the single aliquot regenerative dose (SAR) protocol may be subject to systematic age underestimation as samples approach saturation. In this study, the age of 12 luminescence samples from a chronologically well-constrained site dating to 450 ka from the Thames terrace sequence, southern Britain, was measured using SAR in order to test the performance of the method close to its upper age limit. The characteristics, dose response and thermal stability of the OSL signal in these samples were assessed by investigating equivalent dose (De) as a function of stimulation time and component-resolved pulse annealing. Despite the fact that the samples are dominated by the quartz fast component, these results showed that both the medium and slow components have lower stabilities than the fast component, but with the unstable medium component most affecting the initial part of the OSL signal used in dating. Based on isolating the fast component either through curve fitting or eliminating the medium component using the early background subtraction method, OSL ages up to 450 Gy were found to compare well with the expected age of the site of 450 ± 23 ka. In contrast, a systematic age underestimation of 10% was manifested at lower doses when using the initial part of the OSL signal, contaminated by the medium component. These results suggest that the early background subtraction method should be used when dating in the non-linear part of the growth curve as it provides a better separation of thermally unstable signals and represents a more convenient approach than curve fitting in well-behaved samples.