Geochronologic data provide essential information for understanding the rates of Earth's surface processes, environmental changes, and the evolution of life. Advances in dating techniques have fundamentally changed our capacity to piece together our evolutionary past over millions of years, with luminescence dating proving to be a powerful tool in this field, as it applies to various types of sediments and contexts. The technique determines an age estimate for when mineral grains were last exposed to daylight or heat. Luminescence dating methods rely on the properties of certain minerals to record the amount of radiation to which they have been exposed during burial and release energy when exposed to sunlight or high temperature (e.g. Aitken, 1985, 1998; Bateman, 2019). In the laboratory, the total amount of energy per unit mass stored in the mineral is measured (dose; with the unit Gy). The energy absorption rate per unit mass (dose rate; with the unit Gy a-1) is derived from knowledge of the natural radioactivity surrounding the sampled sediments. The quotient of these two values (dose/dose rate) gives the burial age.

Of the two minerals routinely used for luminescence dating of sediments, quartz and potassium (K)-rich feldspar, the latter allows for the routine dating of older deposits of up to ∼300 or ∼900 ka (considering a dose rate of 3 or 1 Gy ka-1, respectively), using infrared stimulated luminescence (IRSL; Hütt et al., 1988). The datable upper age limit is given by the IRSL signal saturation after exposure to radiation doses around 900 Gy (see the summary in Sect. 8.1 in Murari et al., 2021a). Over the past few decades, different methods have been proposed to extend this upper age limit with varying degrees of success. The infrared radiofluorescence (IR-RF) signal of K-feldspar is a promising candidate for such an extension. The RF signal arises from prompt radiative recombination of charge within crystalline materials during continuous exposure to ionizing radiation. The IR-RF emission at 880 nm (e.g. Kumar et al., 2018; Riedesel et al., 2021; Sontag-González et al., 2022) decreases in intensity with dose accumulation as the electron traps fill until saturation (Trautmann et al., 1999a). This saturation level constrains the time range over which IR-RF dating is applicable.

Murari et al. (2018) demonstrated that an accurate dose recovery of a known dose of 3600 Gy is possible (a dose recovery test is a laboratory performance check of the measurement protocol, and successful dose recovery is a prerequisite for any protocol). If we assume typical environmental dose rates of between 3 and 1 Gy ka-1, then IR-RF dating could produce age estimates ranging from 1.2 to 3.6 Ma, which is around 4 times greater than the upper dating limit of conventional luminescence dating methods. However, more recent studies (Murari et al., 2021b; Kreutzer et al., 2022) indicated a dose saturation at around 1500 Gy, reducing the previously predicted temporal limit of IR-RF dating. Hence, the uncertainty surrounding its upper age limit remains, and further studies on known-age samples are required to assess whether the sample/grain geochemistry influences the age limit. There is undoubtedly a gap in our current understanding of the IR-RF production processes in K-feldspar, and a revised conceptual model might be needed.

The vast majority of IR-RF studies have been performed on multi-grain aliquots, so the possible effects of variability in the IR-RF signal (e.g. differences in signal saturation or in proportions of RF emissions) from different grains has not received much attention in the literature so far, as detailed in Sect. 2. Here, we investigate the IR-RF signal of five samples from different locations at single-grain resolution and discuss the effect that the observed variability could have on multi-grain aliquots. To assess whether the grain geochemistry influences the IR-RF signal and potentially the age limit of IR-RF, we examined individual K-feldspar grains at the submicron resolution X-ray spectroscopy (SRX) beamline at the National Synchrotron Light Source II (NSLS-II) at Brookhaven National Laboratory. Measurements at such a high resolution may lead to a better understanding of the luminescence kinetics in feldspars. We report on the feasibility and practicality of using μ-X-ray fluorescence (μ-XRF) and μ-X-ray absorption near-edge fine structure (μ-XANES) techniques in investigating the IR-RF signal origin and kinetics in K-feldspar.

Identification of the defect type linked to the IR-RF signal and its concentration would enable us to better characterize the light emission (signal sensitivity) in different types of feldspar, while identification of the origin of RF emissions could help us to gain a better understanding about the apparent saturation or quenching of the IR-RF signal. μ-XRF and μ-XANES produce high-resolution maps of elements and their oxidation states and are well suited for the purposes of our study. μ-XRF elemental analyses are based on the characteristic fluorescence of atoms when stimulated with X-rays with a higher energy than their ionization energy. In the case of μ-XANES, initial measurements of standards are run by varying the incident beam energy to determine the specific energy equal to the absorption edge (binding energy of inner shell electrons) of the element or ion of interest. This is apparent by an abrupt rise in the resulting fluorescence, which is different between oxidation states, as they require different minimum stimulation energies before ionization and subsequent fluorescence. μ-XRF maps using the obtained absorption edge energies allow for maps of the different oxidation states of the same element.

The use of synchrotron μ-XRF allows us to improve the spatial resolution compared with a standard lab bench μ-XRF setup (e.g. Buylaert et al., 2018) by reducing the beam spot size from ∼25 to 1 µm or 0.5 µm. Both the grain geochemistry and crystallography should be investigated to characterize the defect type and its environment. In the present study, we focussed only on geochemistry, though our results should be complemented with crystallographic studies in future work.

The defect(s) responsible for the IR-RF emissions are still subject to debate. It has been suggested that IR-RF occurs as a result of the change in the oxidation state of the participating lead (Pb) defect via the following transition: Pb2+ → (Pb+)∗ → Pb+ (Nagli and Dyachenko, 1986; Erfurt, 2003). A similar transition has been suggested for amazonite (see Ostrooumov, 2016), but the direct connection between the Pb centre and IR-RF has not yet been evidenced. Other reactions involving higher oxidation states would also be possible but have not yet been observed or formally proposed. Additionally, the IR-RF signal is composed of at least two separate emissions. Previous publications placed the main IR emission at 1.43 eV (865 nm), based on Trautmann et al. (1999a, b) and Erfurt and Krbetschek (2003), but more recent work including corrections for the spectrometer efficiency places the IR emission closer to 880 nm (Kumar et al., 2018; Riedesel et al., 2021; Sontag-González et al., 2022). A second IR emission centred at 955 nm (1.30 eV) at lower intensity has also been identified (Kumar et al., 2018), which partly overlaps with the 880 nm peak.

The presence of iron (Fe) in feldspar is known to lead to red RF (e.g. Telfer and Walker, 1978; Brooks et al., 2002; Visocekas et al., 2014), with the maximum peak wavelength varying between 700 and 770 nm, depending on feldspar composition (Dütsch and Krbetschek, 1997; Krbetschek et al., 2002). Such observations are in line with the suggestion of more than one component in the red photoluminescence of K-feldspar (Prasad and Jain, 2018). Despite the occurrence of the red RF emission in Fe3+ state, its initial state remains a subject of many debates, with conflicting opinions suggesting either Fe2+ (Fe2+ + h+ → Fe3+; here “h+” stands for hole) or Fe4+(Fe4+ + e- → Fe3+; here “e-” stands for electron) (Kirsh and Townsend, 1988; Jain et al., 2015). Recently, Kumar et al. (2020) argued that the initial state must be Fe4+, based on their findings using cathodoluminescence microscopy. Spectral analyses showed that, with dose exposure, the red RF emission (∼710 nm emission in K-feldspar) increases, while the 880 nm emission decreases (Krbetschek et al., 2000; Erfurt and Krbetschek, 2003; Kumar et al., 2018; Frouin et al., 2019). The thermal stability of the ∼710 nm emission has been, however, questioned (e.g. Krbetschek et al., 2000). Such a reduced thermal stability might be an issue for IR-RF dating, as it has been suggested that the tail of the ∼710 nm emission overlaps with the 880 nm emission, thus potentially playing a role in the shape of the measured IR-RF. Such a contribution can be reduced to less than 5 % of the IR-RF signal, using a bandpass filter centred at 850 nm (full width at half maximum, FWHM, 40 nm), but can still affect the equivalent dose (De) value at doses near the signal saturation (see Sontag-González and Fuchs, 2022). De values are determined by sliding the IR-RF dose–response curve of grains containing the natural signal onto that obtained after a full bleach of the same aliquot. In summary, although previous studies have identified factors that may influence the IR-RF signal in several ways, e.g. whether the IR-RF signal originates from Pb and is affected by the presence of Fe2+ or Fe4+, a conclusive confirmation or comprehensive linkage between these factors is yet to be established.

A possible variability in the several RF emissions in individual grains has received scant attention so far. Trautmann et al. (2000) were the first to analyse the IR-RF signal of individual K-feldspar grains. Using spectral measurements of 21 to 42 grains from three samples, they observed up to four emissions (IR, red, yellow, and blue) with variable intensities (a fourth sample appears in their Fig. 3 but is not mentioned in the main text). An IR-RF dose–response curve was only reported for one grain, which had a similar shape, albeit a later onset of saturation, when compared to the response from the multi-grain aliquot of the same sample. More recently, Mittelstraß and Kreutzer (2021) analysed 60 grains from two samples, of which 55 % and 80 % emitted a detectable signal. In that study, between one and three grains per sample (∼9 % of signal-emitting grains for both samples) were rejected due to a bad match between the natural and regenerative curves, which might have been caused by equipment issues but also due to sensitivity changes (Varma et al., 2013). However, all grains that emitted a detectable signal displayed the expected decay shape for IR-RF (decreasing signal with increasing dose). Likewise, our own laboratory observations indicated that the signal varies in sensitivity across feldspar minerals and can be contaminated for various reasons, leading to spectral interference or quenching, ultimately influencing the saturation level and/or the shape of the IR-RF signal (Frouin et al., 2017, 2019; Kumar et al., 2020).

To investigate these issues, first, we recorded IR-RF curves from individual grains in our luminescence dating laboratory at the Research Laboratory for Archaeology and the History of Art (RLAHA) at the University of Oxford (UK). Then, during our beam time (96 h), and as a proof of concept, we optimized the measurement conditions and obtained compositional maps of the individual K-feldspar grains. We paid particular attention to K, Ca, Fe, and Pb. After analysing the μ-XRF maps, μ-XANES measurements were done at selective spots where Fe occurred in greater concentrations. Note that the atomic number of sodium (Na; another end member of the feldspar ternary system) is too low to be measured at the current SRX beamline.

A total of five samples were selected to represent a diversity of (i) geological context, (ii) geochemistry, (iii) shape of the IR-RF signal, and (iv) age. Sample Gi326 from a Triassic sandstone from Bayreuth, Germany, is composed of 89 % of K-feldspar (Sontag-González and Fuchs, 2022) and has previously been used as a reference sample in a laboratory comparison of IR-RF dating (Murari et al., 2021b). X7343 was collected from a Pliocene sediment at the Nyayanga site in the Homa peninsula, Kenya (Plummer et al., 2023). X7363 was taken from the Gele Tuff in the Turkana Basin, Kenya, dated by Ar / Ar at 1.315±0.002 Ma (Phillips et al., 2023). Previous compositional analyses of Gele Tuff pumice feldspars (crushed clasts without density separation) indicate that they are mostly composed of anorthoclase, with smaller proportions of sanidine and plagioclase; K, Na, and Ca contents ranged ∼1 wt %–6 wt %, ∼5 wt %–6 wt %, and ∼0 wt %–3 wt %, respectively, without appreciable differences between the grains' cores and rims (Phillips et al., 2023). Relatively high Ba contents of up to 0.8 wt % were also reported in that study, with a positive correlation between Ba and Na contents. X7368 is a sediment sample collected above the Silbo Tuff (0.751±0.022 Ma; McDougall and Brown, 2006) and below the Kale Tuff (younger than the Silbo Tuff but not directly dated) in the Turkana Basin, Kenya. Sample H22550 is a coastal marine sample from Sula, Russia, dated by quartz single-aliquot-regenerative optically stimulated luminescence (OSL) at 103±8 ka (Murray et al., 2007) and was used in the past as a reference sample to test the accuracy of IR-RF dating (Buylaert et al., 2012). All samples were prepared following conventional treatments (e.g. Preusser et al., 2008), including wet-sieving to isolate the desired grain size fraction, chemical treatment with HCl at 10 % to remove carbonates and H2O2 at 30 % for a few hours to a few days to remove organic matter, and density separation at 2.58 g cm-3, using a heavy liquid solution to enrich K-feldspar grains. Sample H22550 was then etched with diluted HF (10 %, 40 min). All grains were exposed under a solar simulator (Dr Hönle AG, model UVACUBE 400) for a few days to reset their signal.

Unmeasured grains of sample X7343 were placed on a stub mount on a piece of carbon tape and then imaged with a scanning electron microscope (SEM) equipped for energy-dispersive X-ray spectroscopy (EDS) at Archéosciences Bordeaux (FR) (JEOL JSM-6460LV; detector from Oxford Instruments X-Max (51-XMX0002); software from Oxford Instruments, INCA version 4.11). The SEM was operated at 20 kV and 55 µA beam current. Sample X7343 is referred to as BDX22338 in the Archéosciences Bordeaux system.

IR-RF measurements were recorded with a lexsyg research luminescence reader fitted with an annular 90Sr / 90Y beta source (Richter et al., 2013), using a bandpass filter centred at 850 nm (FWHM 40 nm) mounted in front of a Hamamatsu H7421-50 photo-multiplier tube. Measurements were performed at 70 °C, following Frouin et al. (2017). Multi-grains and single grains were measured on stainless steel cups. High-resolution compositional analysis of the grains was undertaken at the SRX beamline at NSLS-II (Chen-Wiegart et al., 2016). After IR-RF measurements, the grains were removed from the stainless steel cups and fixed on a polymer microscope slide (UVT acrylic; Agar Scientific) with a small piece of carbon tape to avoid misplacement during measurement (Supplement Fig. S1). μ-XRF maps were obtained by scanning across pre-selected regions on the grains with low topographic changes (90×90 µm maps, with a step size of 0.67 µm and an integration time of 0.1 s). The incident X-ray beam was focussed by a pair of Kirkpatrick–Baez mirrors. An incident beam energy of 13.5 keV was used for the μ-XRF measurements. The excited elements' characteristic fluorescence was obtained from the sum of the four elements of a silicon drift detector. All μ-XRF measurements were normalized to the corresponding incident X-ray flux (I0) (Fig. S2).

The μ-XANES maps cover 60×60 µm in steps of 0.5 µm, thus creating a grid with 120×120 data points (i.e. 14 400 μ-XRF spectra). To obtain maps of Fe states in our samples, we varied the incident beam energy according to the absorption edge values obtained from the μ-XANES spectral measurements of Fe standards (Fe foil, pyrite, and hematite). The μ-XANES maps were measured three times to obtain μ-XRF emission spectra restricting the Fe species to either (i) the total Fe (at 7275 keV), (ii) the sum of Fe3+ and Fe2+ (at 7134 keV), and (iii) only from Fe2+ (at 7122 keV). The difference between the intensity levels of the latter two measurements can qualitatively give the intensity levels of Fe3+; i.e. IFe2+&Fe3+ – IFe2+  = IFe3+, where I (arbitrary unit) refers to intensity and, thus, the μ-XANES map of Fe3+. We also attempted to record Pb states; however, the Pb standard available at the SRX beamline was fully oxidized, which hindered establishing the correct beam energy for mapping. Therefore, no Pb oxidation state maps were possible. μ-XRF and μ-XANES data were analysed using the open-source software PyXRF v1.0.23 (Li et al., 2017) and Athena v0.9.26 (Ravel and Newville, 2005), respectively. Maps and plots were created using R (R Core Team, 2022).

We demonstrated that individual K-feldspar grains of the same five samples display different IR-RF behaviour, illustrated by different signal decay shapes (i.e. increasing or decreasing with dose and different saturation levels). These behaviours are cumulative (see Fig. 1), and therefore, the IR-RF signal of a multi-grain aliquot can lead to inaccurate equivalent doses. Despite the use of chemical preparations to remove contaminants, manually picking individual grains was necessary to isolate K-feldspar grains, which is unrealistic for routine dating applications in a low-light laboratory. A more realistic way to remove such contamination is by selecting K-feldspar grain populations by isolating the emission signal of individual grains with an imaging system. Here, we wanted to gain a further understanding of the production and origin of the emission signal, which ultimately will help us design a more appropriate imaging system for IR-RF dating. For sample X7343, we show through SEM-EDS analyses that the different emissions can be linked to different grain mineralogy. Since K-feldspar grains are known to be heterogeneous on a subgrain level, we propose synchrotron-based X-ray spectroscopy to characterize the grains on a submicron scale and investigate the origin of the IR-RF and other linked emissions. Information on the oxidation states of, for example, Fe possibly allows for the characterization of the reactions behind the electronic changes leading to radiofluorescence.

In the preliminary work presented here, we successfully applied μ-XRF and μ-XANES at the SRX beamline (NSLS-II) to obtain element and oxidation state maps of regions of interest within individual K-feldspar coarse grains previously used for IR-RF measurements. We were able to correlate the desired IR-RF signal shape (category no. 1) with compositions of high proportions of K, Pb, and Ba and low proportions of Fe. High proportions of Fe in the μ-XRF spectra were found in grains of category nos. 2 and 3, but the possible role of Fe as a contaminant remains unclear. During our next beam time, we will polish the grains down to a uniform surface prior to μ-XRF and μ-XANES measurements to limit surface effects. Such a set-up will also allow us to directly test the hypothesis that the contaminating IR-RF signal is coming from an element present at the surface of the grain (e.g. iron coating possibly due to weathering) but not within the grain.

The relation between the chemical composition, crystal structure, and the shape of the IR-RF signal in individual K-feldspar grains is still poorly understood and efforts should be made to identify and quantify, at high resolution, the element responsible for producing the IR-RF signal with the highest dynamic range (i.e. saturation at high dose). Our future work will include implementing a second detector to simultaneously measure μ-XRF/μ-XANES and the IR-RF signal induced by the X-rays. Though not widely used, X-rays are a suitable alternative to radioactive sources for luminescence dosimetry including RF. The dual detection will allow us to isolate emissions from different mineral inclusions and directly correlate them to the elemental composition, thereby assessing the extent of overlap of the desired IR-RF emission centred at 880 nm and contaminating ones such as the possible unstable red emission associated with Fe3+.