CW-OSL, LM-OSL and TL Dating of Bricks from Karakorum, Mongolia: Insights from TL Spectra

The brick samples used in this study were collected at the archaeological site Karakorum (47°12′N/102°50′E; 1461–1462 m a.s.l.) located in Orkhon Valley, Mongolia. This site assumed major importance as the medieval capital of the Great Mongolian Empire and as one of the most important cities in the history of the Silk Road. Karakorum (Qara Qorum) founded in 1220 became a walled urban settlement under Chinggis Khaan's son Ögedei Khan in 1235. “The palace district” was first identified in 1948–1949 by a Russian-Mongolian archaeological team led by Sergei V. Kiselev (Kiselev et al., 1965). The Mongolian-German Karakorum Expedition (MDKE) conducted by the German Archaeological Institute (DAI) and the Mongolian Academy of Sciences (MAS) in the old Mongolian capital has radically revised the conventional image of Karakorum (Hüttel and Erdenebat, 2011). According to the inscriptions, the Wanan palace of Ögodei Khan and some temples built at the same time, among them a monumental Buddhist stupa temple, the “pavilion/temple of the origin of the Yüan”, are held as the oldest buildings of the city. They were built in 1235 at the same time as the city walls. While the palace was completed within a year in the early part of 1236, the great Buddhist stupa temple was only finalised in 1256. It is known that the first restoration of the temple took place in 1311, and the second one between 1342 and 1346 (Hüttel and Erdenebat, 2011).

About fourteen “Mantou” type kilns with round walled fire pits and fire canals were discovered, nine of which were almost entirely intact (Hüttel and Erdenebat, 2011); it is evident that that numerous architectural ceramics, including roof tiles, bricks and Buddhist figurines come from these factories. These bricks were initially classified based on their colour, which is indicative of the provenance, use of oxidising and reducing atmosphere and thermal history of the material. In most kilns, carbon monoxide serves as the primary reducing agent, but in some traditions, hydrogen (a potent reducing gas) may have been generated towards the end of the firing through the introduction of water to the firebox or the kiln-chamber (Rose Kerr, 2004). Red coloured bricks were produced in an oxidising atmosphere, while grey coloured bricks were produced using conventional “air-starved fuel” reduction or the generation of water gas. Firing temperatures in the “Mantou” type kilns ranged in the interval 800–1000°C (Rose Kerr, 2004).

The bricks from the Karakorum were previously subjected to CW-OSL measurements; quartz from grey coloured bricks, however, gave significant dose/age underestimations (Solongo et al., 2006a). The main reason for this was that the ratio of fast to medium components was too low and D_e's with the highest fast to medium ratio were considered; in addition, we performed post-IR-OSL (Banerjee et al., 2001), leading to better dose estimates. That led to the conclusion that only red coloured bricks were suitable for blue-stimulated CW-OSL measurements.

The environmental radiation dose rates were measured using high-resolution γ-spectrometry applied to the external layer of the sample to determine K, U and Th concentrations. In-situ water content measured shortly after sampling was taken into account. The contribution from cosmic radiation to the dose rate was calculated following Prescott and Hutton (1994), assuming an uncertainty of 5%. The radionuclide concentrations were converted to dose rate data using the conversion factors from Guerin et al. (2011). Table 1 summarises the sample description and dose rate data.

For luminescence procedures, the external layer of the ceramic fragments was removed; the obtained cores were carefully crushed and sieved. All crushed material was subjected to 15% H₂O₂, 30% H₂O₂ to remove organic material, to 10% HCl to dissolve carbonate minerals, prior to density separation using a heavy liquid and separated by 2.62 g cm⁻³ and 2.74 g cm⁻³ to obtain quartz fractions. For quartz, grains were etched with 40% hydrofluoric acid (HF) for 40 min. to remove the alpha irradiated outer rinds and HCl to eliminate the possible formation of fluorosilicates. For OSL and TL measurements 90–120 μm coarse grains were deposited onto disc using 1 mm and 2 mm spot, while for TL spectra measurements 8 mm aliquots were used.

Spectral measurements were carried out in MPI Heidelberg using the TL/OSL-Spectrometer described by Rieser et al. (1999). Here a Czerny-Turner flat field spectrograph was equipped with a grating of 150 lines mm⁻¹ for the measurements, with the entrance slit of spectrograph being set to 0.5 mm. The detector used was a liquid nitrogen cooled CCD array (Princeton Instruments), which had a size 1100×330 pixel. In order to cut the intense red/IR thermal emission at higher temperatures, the heat-absorbing filters were used.

All OSL measurements were performed using an automated Risø TL/OSL-DA-15 reader, equipped with blue light-emitting diodes (470±30 nm, 50mW cm⁻²) for stimulating the quartz, a Thorn-EMI 9235 photomultiplier combined with three 2.5 mm U-340 Hoya filters (290–370 nm) for OSL signal detection. Laboratory irradiation was undertaken using a calibrated ⁹⁰Sr/⁹⁰Y beta source delivering (2.72±0.2 Gy/min) to coarse grain aliquot. LM-OSL measurements were carried out by continuously ramping the power from zero to maximum (90% of power), over a period of typically 3000 s. For TL measurements, the TL glow curves (180 – 460°C; 5°C/s) were recorded using “D410 transmission filter” (410±30 nm) for TL signal detection. The dose D_e was determined using SAR (Table 2) procedure (Murray and Wintle, 2003), which employed CW-OSL, LM-OSL and TL measurements.

For the measurement of TL emission spectra samples were irradiated artificially with 50 Gy using a 90Sr β-source and stored for four weeks at room temperature. Because luminescence of natural quartz is usually weak, some samples were irradiated with 100 Gy. The measurements were performed using a heating rate of 5°C/s, a spectral range of 300–1200 nm, and a temperature range of 20–500°C.

Fig. 1a, 1b show the representative TL contour plot obtained for the natural laboratory irradiated quartz samples HD3 (red brick) and HD2-57 (grey brick). Over the entire investigated temperature range, the spectrum of the detected glow peaks is dominated by a blue emission. Two TL peaks are clearly detected around 225°C and 350°C, respectively.

Fig. 1

Fig. 2a, 2b shows TL emission spectra plotted vs the wavelength. For quartz HD3, the emission band peak had maxima at ∼360, and 440 nm, in agreement with the wavelength peaks presented in summary by (Krbetscheck et al., 1997). For this sample, for the temperature range of 75 to 175°C, both the UV emission band peaking at ∼366 nm and the blue emission band peaking at 440–460 nm (∼2.8 eV) increased in intensity. The UV emission band is still observed in the range 200–275°C, but disappears at 300°C. In contrast, the blue emission at 470 nm (2.6–2.5 eV) emission increases for 200–275°C.

Fig. 2

In contrast, for quartz HD2-57 the UV emission band is missing. The emission had maxima at ∼478 nm and a low-intensity orange-red band at ∼600 nm. We observed an increase in the intensity of blue emission bands over a temperature range of 125 to 175°C, but this tends to decrease with the further rise in temperature from 225 to 350°C. The intense high-temperature peak at around ∼800 nm observed in both samples may be inherited from the thermal history.

The comparison of the TL emission results is a strong indication that annealing in the past has taken place at different temperatures. Similar treatments have been already described in the literature. Previously, (Bøtter-Jensen et al., 1995) showed that a high-temperature annealing prior irradiation could significantly increase the luminescence sensitivity, typically by three orders. Poolton et al. (2000) explained the OSL sensitivity increase and concluded that E′ centres are annealed out of the samples at temperatures in the range 500–700°C.

Radio-luminescence and thermoluminescence spectra were used by Schilles et al. (2001) as a means of investigating the luminescence sensitivity changes if quartz was heated beyond the first and second crystal phase transitions, in the temperature range 573 and 1050°C. The results lead them to conclude that luminescence is generally sensitised in quartz by removal of oxygen vacancy E’ centres, which act either as non-radiative recombination centres or as luminescence centres emitting in the deep UV-region. They concluded that after annealing material beyond the first phase transition, the UV (360 nm) emission became enhanced in their samples.

Martini et al. (2009) reported that samples annealed at around 500°C showed dramatic enhancement of the 3.4 eV (360 nm) band followed by a decrease. This is consistent with our observation for quartz HD3 which exhibits similar TL spectra suggesting that quartz experienced in the past the annealing around the first transition phase. Deconvolution into Gaussian components of the thermoluminescence emission spectra of HD3 indicates the presence of 363nm band.

Further, Martini et al. (2012) showed that “hydrogen-swept” quartz has no UV emission. Hydrogen is important due to the supposed participation in various generation processes of defects, both intrinsic and extrinsic, and to the high mobility. Hashimoto et al. (2001) reported that the atomic hydrogen derived from the radiolysis of OH-related impurities could operate as a “killer” of radiation-induced Al-centers.

The study of the relationship between the luminescence emissions and specific defects in quartz is beyond the scope of this study. However, it can be inferred that the technological parameters of manufacturing grey coloured bricks, including high-temperature annealing and cooling using hydrogen reduction, could have affected luminescence properties such as the decrease in the UV emission band.

In the following sections, we addressed the question of whether the TL spectra affect luminescence properties and the dose assessment using different SAR measurements.

In OSL dating protocols, OSL signal is usually obtained under during optical stimulation at steady stimulation power, resulting in a continuous wave OSL (CW-OSL) signal. It is well-known that the decay of signal during the OSL measurement of quartz does not form the simple exponential, it is composed of components generally referred to as fast, medium and slow, each of which has different optical and thermal luminescence properties, indicating the existence of three optically active traps according to Bailey et al. (1997).

Alternatively, the OSL signal measured using a linearly increasing stimulation power to obtain the linearly modulated OSL (LM-OSL, (Bulur, 1996)) signal tends to separate the individual OSL components involved in the OSL production, and it was initially used as an analytical tool (Bulur et al., 2000, 2002) for discriminating the different OSL components (Jain et al., 2003), (Singarayer and Bailey, 2003).

Representative blue-stimulated CW-OSL decay curves of heated quartz are shown in Fig. 3. Note that HD3 was stimulated for 20 s after preheating at 220°C, HD2-57 for 100 s after preheating at 260°C. The test dose CW-OSL signal was fitted to a sum of first-order exponential components, and the fitting procedure revealed that the OSL signal of HD3 comprised the dominant fast component. The contribution of fast, medium and slow components are shown in the inset. In contrast, HD2-57 does not have a fast OSL component, but medium and slow components.

Fig. 3

A simple mathematical transformation to convert the CW-OSL decay curves to the LM-OSL curves was performed according to Bulur (2000), and pseudo-LM-OSL curves constructed are shown in Fig. 3b, 3d for quartz HD3 and HD2-57, correspondingly. The inset shows the relative contribution of fast, medium and slow components derived from the pseudo-LM-OSL.

The fitting of multi-exponential functions to the CW-OSL decay curves from HD3 revealed the presence of fast, medium and slow OSL components correspondingly. The corresponding photoionisation cross-sections values were 2.47 ± 0.33 × 10⁻¹⁷ cm² (fast), 5.05 ± 1.44 × 10⁻¹⁸ cm² (medium), which are in agreement with the values of 2.32 ± 0.16 × 10⁻¹⁷ cm² and 5.59 ± 0.44 × 10⁻¹⁸ cm² (Jain et al., 2003).

The fitting procedure revealed that the LM-OSL signal of HD3 (Fig. 3e) is dominated by fast, and LM-OSL of K02/02 is not dominated by fast OSL component, but by the medium and slower components, these results are consistent with the CW-OSL results. The corresponding photoionisation cross sections for fast, medium and slow components were 2.15×10⁻¹⁷, 3.14×10⁻¹⁸, 4.64×10⁻¹⁹ (cm²), respectively, giving rise to b values of 1.59, 0.52 and 0.04 (s⁻¹), which is in agreement with (Choi et al., 2006).