Is there a common alpha-efficiency in polymineral samples measured by various infrared stimulated luminescence protocols?

Measuring the infrared stimulated luminescence (IRSL) of K-feldspar or polymineral separates at elevated temperatures (>150°C) after initial IRSL stimulation at 50°C (IRSL₅₀) has gained considerable popularity for estimating the equivalent dose (D_e) in retrospective dosimetry (e.g., Thomsen et al., 2008; Thiel et al., 2011; Reimann and Tsukamoto, 2012). This technique – usually referred to as post-IR IRSL or pIRIR – offers access to feldspar-derived luminescence signals less affected by so-called anomalous fading (Wintle, 1973). The conceptual model explaining these lower fading rates builds upon an increased distance between trapping and recombination sites as stimulation temperature is raised, resulting in a decreased probability of tunneling recombination of opposite charge carriers (e.g., Jain and Ankjsrgaard, 2011; Buylaert et al., 2012). There is however a trade-off between low fading rates of pIRIR signals and signal bleachability, i.e., the pIRIR stimulation temperature is inversely related to the optical resetting rate of the respective signal, while part of the signal remains unaffected by sunlight exposure (Li and Li, 2011). This unbleachable residual has occasionally been determined in the laboratory by applying some sort of residual subtraction method in order to isolate the optically sensitive component from the bulk signal (e.g., Schatz et al., 2012; Li et al., 2013).

Commonly, either sand-sized (~90–250 μm) K-feldspar or silt-sized (~4–11 μm) polymineral grains are prepared for pIRIR measurements. While it is debatable whether the coarse grain size fraction should be HF-etched to remove the outer layer influenced by external α-radiation (e.g., Duller, 1992; Li and Li, 2011; Trauerstein et al., 2014; Kenzler et al., 2015), the α-dose rate has to be fully considered for fine grains (~4– 11 μm). To account for the different efficiency in producing luminescence of heavy particles such as α-particles and slightly ionising radiation such as β- and γ-radiation, the so-called α-efficiency must be incorporated when calculating the dose rate (Aitken, 1985a). Among the different systems presented for quantifying α-efficiency (see Aitken, 1985b, for a summary), the a-value system has most widely been used both for quartz and feldspar. This system is based on the findings of Aitken and Bowman (1975) indicating that the induced luminescence per unit of generated α-track length (in μm μm^–3 = μm^–2) is nearly independent from the α-particle’s energy. Hence, the luminescence recorded after generating a certain cumulative length of α-tracks is compared with the luminescence resulting from a known β-dose to calculate the a-value. In the case of quartz and mono-energetic 3.7 MeV α-particles (as delivered by many ²¹⁴Am sources), the a-value is by definition equal to the k-value, where the ratio of luminescence induced by an α-dose (in Gy) to that induced by a β-dose (in Gy) is calculated. For non-quartz materials, a correction factor r relates the material-specific a-value to the quartz a-value (Aitken, 1985b). This correction factor largely depends on the density of the used material. Since the density of feldspar, however, only deviates 2% from the density of quartz, we used uncorrected a-values in this study.

Compared to the vast body of literature published during the past years on the application of the pIRIR protocol to various sedimentary archives, only relatively few studies determined the a-value individually for the samples investigated. Many authors refer to Rees-Jones (1995) when using an a-value of 0.08 ± 0.02, although this value is based on three samples only and – more importantly – determined for the IRSL signal recorded using a multiple-aliquot protocol and under strongly varying preheat and measurement conditions than the ones used in the respective studies (e.g., Stevens et al., 2011; Thiel et al., 2011; Buylaert et al., 2012; Vasiliniuc et al., 2012; Schmidt et al., 2014). Only Biswas et al. (2013) reported measured pIRIR₂₉₀a-values ranging from 0.036 ± 0.003 to 0.055 ± 0.002 for volcanic ash samples, determined by administering a known α-dose to a bleached sample and recovering the equivalent β-dose using a single-aliquot regenerative dose protocol. These values are on average higher than the IRSL₅₀a-values of these samples by a factor of 1.28 ± 0.03 (Biswas et al., 2013). Successful dose recovery using either solely α-regeneration or solely β-regeneration always combined with a fixed β-test dose for normalisation led the authors to conclude that this approach of a-value determination is accurate. In a more comprehensive study, Kreutzer et al. (2014) determined pIRIR₂₂₅a-values for five loess samples (polymineral fine grains) from Saxony (Germany), revealing a systematic difference between IRSL₅₀ and pIRIR₂₂₅a-values. Furthermore, the mode of signal resetting (heating vs. bleaching) prior to α-irradiation appears to affect the size of the a-value (Kreutzer et al., 2014). Considering the results obtained in the referenced studies above, one might suspect that the pIRIR a-value of polymineral fine grains correlates positively with the pIRIR stimulation temperature. It therefore appears timely to further investigate the a-value of the pIRIR₂₉₀ signal for a range of different samples. Specifically, the present study aims at:

identifying a suitable method to accurately determine pIRIR a-values,

Many samples in our laboratories were recently measured with a distinct IRSL protocol (following the approach described in detail in Faust et al., 2015) which includes a high-temperature preheat and a 20 min pause prior to IRSL measurement, in order to minimise anomalous fading. In addition to pIRIR₂₉₀a-values, we therefore also report on the a-values determined with this procedure (termed IRSL_F henceforth) in order to assess the variability of the a-value as a function of the IRSL/pIRIR measurement protocol. Details on both protocols employed in this study are given below.

To evaluate both the variability of the pIRIR₂₉₀a-value between samples of different mineralogy and geographical origin and within a set of samples originating from the same outcrop, we studied a total of 47 polymineral fine grain samples from 10 different locations in Europe. Table 1 provides a summary of these samples.

Table 1

Summary of investigated samples. Sample codes BT… refer to the luminescence laboratory in Bayreuth, codes C-L… to the laboratory in Cologne; pIRIR₂₉₀ = post-IR IRSL protocol with 290°C stimulation temperature (following Thiel et al., 2011); IRSL_F = IRSL protocol according to Faust et al. (2015); MAAD = multiple-aliquot additive-dose protocol applied both to the pIRIR₂₉₀ and the IRSL_F emission (see main text for further details).

Prior to measurement, samples were dry or wet sieved to grain sizes <63 μm, soaked in 10% HCl and 10% H₂O₂ to dissolve carbonates and oxidise organic matter, respectively, treated with 0.01 N sodium oxalate to disperse aggregates (only Cologne laboratory) and subsequently settled in a water column for distinct periods to extract the target grain size range of ~4–11 μm (application of Stokes’ Law). All steps were carried out under subdued red light conditions (640 ± 20 nm). Readily prepared fine grains were pipetted onto aluminium or steel discs (~2 mg per aliquot in Bayreuth, ~0.9–1.0 mg in Cologne) for IRSL and pIRIR measurements. Experiments by Kreutzer et al. (2014; Table S3) have shown that pIRIR₂₂₅a-values resulting from both 1 mg and 2 mg of sample material per aliquot are identical within uncertainties.

All measurements were carried out on Risø TL/OSL DA15/DA20 readers, equipped with infrared (875 ± 80 nm) diodes for signal stimulation and an EMI 9235QB15 photomultiplier tube coupled with a Chroma D410/30x band pass filter (410 ± 15 nm) for signal discrimination and detection. Artificial β-irradiation was carried out with a built-in ⁹⁰Sr/⁹⁰Y β-source delivering dose rates to fine grains between 0.040 ± 0.002 and 0.133 ± 0.005 Gy s^–1. In Bayreuth, α-irradiation was conducted in vacuum (<10^–2 mbar) using either the built-in ²⁴¹Am source (0.144 ± 0.007 Gy s^–1) of one of the Risø readers or an external six-seater Littlemore ²⁴¹Am irradiation facility (type 721/B) with a dose rate of 0.021 ± 0.002 Gy s^–1. In Cologne, the built-in α-source of a Freiberg Instruments lexsyg research reader was used for that purpose (0.197 ± 0.004 Gy s^–1, <1 mbar). Information on the calibration of these α-sources is given in the appendix.

Acquired luminescence data were analysed using the program Analyst (v.4.31.9; Duller, 2015). IRSL and pIRIR₂₉₀ decay curves were measured for 300 s (Bayreuth) or 200 s (Cologne) in this study, and the initial 5 s were integrated to give the signal with which dose response curves were constructed. Instrumental background averaged from the last ~40 s of the decay curve was subtracted from the integrated signal.

Previous studies on the appropriateness of using single-aliquot regeneration (SAR) protocols for a-value assessment raised concerns on the validity of β-test dose correction of α-induced luminescence signals (Mauz et al., 2006). One way of bypassing this problem is to apply multiple-aliquot additive-dose (MAAD) protocols. In this study, natural aliquots were divided into four dose groups (four aliquots each) and given equally-spaced additive α-doses between 0 and 969 Gy. After measuring the pIRIR₂₉₀ and IRSL_F signals according to the protocol described in Table 2 (steps 2–4), the fitted dose response curve was extrapolated to the α-dose axis and the so-obtained α-dose related to the β-dose derived from ‘regular’ SAR D_e measurements to obtain the a-value. We adopted the equal pre-dose normalisation technique (Franklin and Hornyak, 1992) to reduce inter-aliquot scatter using solely α-regeneration doses and an α-test dose of 211 Gy. In the course of this normalisation method, all aliquots received the same cumulative dose and heat treatment prior to test dose irradiation to consider dose-dependent sensitivity changes. To account for the pIRIR₂₉₀ residual, the signal left after 24 h bleaching in a solar simulator (Osram Duluxstar 24 W) was subtracted (Fig. 1). In all cases, the laboratory bleaching procedure resulted in IRSL_F and pIRIR₂₉₀ signals close to instrumental background levels. For MAAD investigations, we chose two samples (BT1344, BT1345) with comparably small D_e values with the aim to enable a linear fit to the dose points. The MAAD dose response curve of sample BT1345 is shown in Fig. 1; resulting a-values are listed in Table 4.

Fig. 1

Table 2

Measurement protocols employed for a-value determination. Step 1 in the first SAR cycle is carried out using α-irradiation, while all subsequent irradiations refer to β-doses. The pIRIR₂₉₀ protocol follows Thiel et al. (2011), while the IRSL_F protocol was adopted from Faust et al. (2015). Stimulation times for IRSL₅₀ and pIRIR₂₉₀ signals were 300 s (Bayreuth) or 200 s (Cologne); an IRSL readout at 325°C for 600 s (Bayreuth) or 200 s (Cologne) intended to fully zero IRSL traps prior to the next regeneration cycle.

Table 3

Relative change of the L_α/T_β ratio after repeated L_α – T_β cycles (see main text for further information).

Table 4

Results of a-value determination. In contrast to Table 1, samples are grouped according to the measurement protocol used for a-value determination. n is the number of measured aliquots per sample. The a-value is derived as the arithmetic mean of individual aliquots of one sample; the averaged measurement uncertainty Δa is calculated using the formula Δa = [(Δa₁²+Δa₂²+…+Δa_n²)/n]^0.5 SD is the standard deviation. The low-temperature IRSL readout of sample C-L3789 was carried out both at 50°C and 80°C, as indicated in the first column. Systematic errors relating to α- and β-source calibration are not considered in this compilation.

Despite the fact that a-values derived from MAAD measurements may be regarded as reliable for the above mentioned reasons (no test dose correction necessary and normalisation using α-doses only), they are nevertheless tied to some impracticalities. First, comparatively large amounts of sample material are required (>12 aliquots, in addition to those needed for ‘regular’ D_e determination). Secondly, data scatter and extrapolation of the dose response curve can cause uncertainties of dose estimates usually larger than those in SAR measurements. The following experiments hence aim at testing whether SAR measurements are a practical alternative to the MAAD approach and which measurement procedures yield the most reliable a-values. The pIRIR₂₉₀ protocol following Thiel et al. (2011) and the IRSL_F protocol after Faust et al. (2015) on which the SAR measurements are based are outlined in Table 2.

Our experiments on the correction of sensitivity changes induced by an α-regeneration dose by means of a β-test dose indicate that for the majority of the investigated samples the SAR protocol (recovery of a known α-dose with β-SAR cycles) is suitable to determine reliable a-values. This procedure is less time- and material-consuming than MAAD protocols, for which sensitivity changes should not have any effect on the results. Comparative a-values derived from SAR and MAAD protocols for four samples (pIRIR₂₉₀ and IRSL_F for two samples each; Table 4) are not in agreement within 1σ uncertainties, but are consistent with the overall a-value distribution of the respective IRSL signal and laboratory (see Fig. 4; except for the MAAD pIRIR₂₉₀a-value of sample BT1344, which could be classified as a statistically significant outlier). Therefore, it can be concluded that SAR protocols are a valid procedure to determine IRSL a-values of polymineral silt-sized samples, at least for the majority of samples investigated in this study.

The assessment whether or not there is a common a-value for polymineral samples measured with several infrared stimulated luminescence protocols is not straightforward. While a-values appear to be consistent within one laboratory environment, there are discrepancies when comparing a-values that were obtained using different measurement setups. Although comparative measurements on the same samples in the two laboratories showed differences in a-values, these were not significant (but n = 5 only). However, considering the entire data set it appears that the a-values measured in this study are less dependent on sample origin and lithology, but rather on the measurement equipment and laboratory routines. Since slight variations in experimental conditions may affect the characteristics (anomalous fading, residual signals) of (post-)IRSL signals, measured a-values are probably only valid for this respective experimental setup (type of sample carriers, luminescence reader etc.), and D_e and a-value measurements on a sample should be carried out with the same setup. Specifically, possible technical reasons for the trend of difference in a-values between laboratories could be the calibration of the α- and β-sources, thermal lag or the temperature-calibration of the heating elements in the readers, which seem to be offset, as indicated by the comparative measurements described in the last section. Similar effects were observed in a previous interlaboratory comparison study of the 110°C thermoluminescence peak of quartz (Schmidt et al., 2018). The findings from the pIRIR₂₉₀D_e comparison of the same samples within two laboratory environments (Fig. 7) suggest that the α-source calibration is not the sole cause responsible for the discrepancy in a-values between laboratories. For samples with IRSL₅₀ signals just high enough to permit determination of rough a-value estimates, there is the tendency of higher average IRSL₅₀a-values in Cologne (0.086 ± 0.012, n= 15) as compared to Bayreuth (0.074 ± 0.009, n = 10). Since temperature deviations at 50°C stimulation temperature should be negligible, it appears that systematic shifts in temperature cannot explain the observed a-value differences alone. Rather, an interaction of the influencing factors described above seems to cause the observed variation in a-values. The aliquot size (~1 mg vs. ~2 mg per aliquot) and the mode of signal resetting, however, do not appear to influence the a-value significantly. It is noteworthy that the pIRIR₂₉₀a-values derived after annealing (Cologne) and after hot bleach (Bayreuth) match well with the pIRIR₂₂₅a-values of Kreutzer et al. (2014) obtained with comparable resetting mechanisms. Although the comparison of the two resetting mechanisms in our study did not reveal any significant differences in the resulting a-values, a consistent pattern with higher a-values after annealing emerges. For future research, it might be advisable to test this with more samples and interlaboratory comparisons.

The slight differences in the pIRIR₂₉₀ measurement protocol (cf. Table 2) are unlikely to be the source of a-value variations between the two laboratories, because β-regeneration doses and hence signal intensities were usually low. Therefore, pIRIR₂₉₀ signals always reached instrumental background levels at the end of a regenerative measurement and signal carry-over into the test dose cycle as observed by Colarossi et al. (2018) appears improbable.

Contrasting the a-values obtained from different profiles revealed no provenance-related difference for the samples measured with the IRSL_F protocol. However, these samples were all taken on Fuerteventura in similar geomorphological settings and show comparable geochemical composition. Contrarily, we detected some provenance-related differences between the samples measured with the pIRIR₂₉₀ protocol. It is noteworthy though that these are mainly associated with the only lacustrine archive (Vrsac, Serbia) tested. This might suggest different a-values for different types of lithology or sedimentary archives; a statement, however, which needs further investigation. Furthermore, it is possible that slight provenance-related differences in a-values are masked by the variation induced through the differences in experimental setup. Therefore, even if there were systematic trends in a-values among samples with different lithology, they would not be discernible given the measurement uncertainties. This is supported by the fact that within one laboratory, the a-values for most of the samples are statistically the same.

The study of Kreutzer et al. (2014) showed for five loess samples from Saxony that the a-value appears to rise with increasing sample temperature during measurement (by 0.023 ± 0.012 for IRSL₅₀ and pIRIR₂₂₅, respectively, on polymineral fine grains). This trend could not be confirmed with the current data set where the average a-values obtained with the IRSL_F and pIRIR₂₉₀ protocols are statistically indistinguishable. Considering a-values employed in previous applications of the pIRIR₂₉₀ protocol (e.g., 0.08 ± 0.02 in Thiel et al., 2011; or 0.07 ± 0.02 in Preusser et al., 2014) and the dataset compiled here, it is conceivable that continuing use of literature values could lead to slight, but systematic age overestimation.

Taken previous and the current data together, the a-value of IRSL signals from polymineral loess samples appears to range between 0.08 and 0.11, and whenever an individual assessment of a-values is not feasible (what is, however, highly encouraged), an increased uncertainty level of 30% could account for the variation in a-values caused by differences in measurement equipment and laboratory routines. Nevertheless, to avoid systematic errors in age determination, a-values should be measured as accurately as possible for each set of samples and specific measurement setup.

The analyses conducted in the frame of a-value measurements of 47 polymineral silt-samples with two different IRSL measurement protocols (pIRIR₂₉₀ and IRSL_F) lead to the following conclusions:

–

The SAR protocol appears to be appropriate for determining IRSL a-values.