LUMINESCENCE DATING OF AN ANCIENT WALLED SETTLEMENT IN ORKHON VALLEY , MONGOLIA

We investigated the potential of the optically stimulated luminescence (OSL) method to date young (<1000 years) samples collected in the Orkhon Valley Cultural Landscape, Mongolia. Quartz showed an infrared signal; therefore the post-IR OSL method was applied to small aliquots which are considered proxies for single grain measurements. Statistical analysis of the dose distribution produced CAM De of 5.14±0.10 Gy and over dispersion of 47.5%, and MAM De of 3.7±0.6 Gy. Since no partial bleaching was suspected, the analysis of signal composition was done and the fast quartz post-IR OSL lead to De of 4.9±0.2 Gy. Based on the quartz fast component and CAM De we propose the new chronology of ancient construction at 785±80 AD, rather than 906-1125 AD as suggested by archaeological evidence. However, the MAM age is in good agreement with independent age control for construction of the ramparts suggesting the date of reconstruction, collapse or reuse for the square walled enclosure MOR3 during 1090±80AD.


INTRODUCTION
DOI: https://doi.org/10.5564/pmas.v57i4.918 The luminescence technique [2] is based on the fact that many naturally occurring minerals, including quartz and feldspar are able to act as dosimeters for the amount of ionizing radiation they are exposed to.The optically stimulated luminescence (OSL) method dates the last sunlight exposure event for minerals.A basic assumption contained within the OSL approach is that the traps are completely empty of all trapped charge by sunlight during the event, which is being dated.This assumption appears to be the case with ceramics and bricks during the firing process; fired clay brick samples were successfully dated by the thermoluminescence (TL), optically stimulated luminescence (OSL) and post-infrared stimulated (pIRIR) techniques [21], [24], [20]; using different brick samples they were able to reconstruct the construction of the palatial complex, the palace and the city walls in Karakorum -the ancient capital of the Mongol Empire.However, for construction of walls and ramparts, sediment embedded within the wall may have been incorporated at various stages of the construction process, and from various sources, and may not necessarily have completely reset the OSL signal.In that case scatter in dose distributions of sedimentary samples is expected; the sources are incomplete or heterogeneous bleaching of grains [15]; [7]), post-depositional mixing of grains [18]; [13]; [23] and beta-dose heterogeneity in the natural burial environment [14]).Several statistical models [10] have been suggested, whose outcome is highly dependent on the uncertainties assigned to individual dose estimates.
Other sources for dispersion of the dose distribution include differences in the OSL signals response from different grains for identical treatments [4].Internal variability leads to the observation of grains with different luminescence properties, such as brightness, and the presence of quartz grains with different OSL components [3].For some quartz samples, isolation of the so-called 'fast component' of the OSL improves the accuracy of absorbed dose estimates [1].A quartz fast component [4] is the signal used for D e evaluation using the single aliquot regenerative-¬dose protocol by [25].The ratio of fast to medium component leads to the scatter in D e distributions [24,8].The effect of the medium component on the dose evaluation and the ratio of fast to medium was observed for heated bricks [24].
In this study, we apply OSL dating to the walled enclosure in Orkhon Valley.Small aliquots used as a proxy for single grain measurements were analyzed using two approaches.First, we apply statistical analysis to the post-IR OSL data.Then signal composition analysis is used to isolate the fast signal dominated dose estimates.The results will be compared and discussed in terms of choosing the appropriate model.

Sampling site
The Orkhon Valley Cultural Landscape includes numerous archaeological remains including Karakorum, the 13 th century capital of Mongolia and Karabalgasun, the 8 th -9 th centuries Uighur capital.In the course of the field work of the German-Mongolian archaeological team in 2009 and 2010, the key region between the supposed spring palace of Ögödei Khan and the urban area of Karabalgasun and its archaeological structures were documented via aerial photograph analysis and ground survey [6].Altogether 310 archaeological sites were documented in the middle and upper Orkhon Valley; these sites, the majority of which was unknown to archaeologists prior to the project, encompass more than one thousand individual features such as burials, settlements, find scatters, walled enclosure, petroglyphs etc.All periods from the Middle Paleolithic through to Post Medieval and modern times are represented [6].One of this sites is a square walled enclosure (E102°35'33'', N47°28', figure 1) located to the north-west of Karabalgasun presented walled ramparts that were constructed from rammed earth; an ancient method of wall construction using local materials well represented in the Orkhon Valley.For construction changing layers of sand, clay and gravel were moistened and then pressed into a wooden framework.The rampart itself was constructed in horizontal layers (Fig. 2); construction progressing until one layer is complete and compaction of the next layer begins.The test trench showed that the borders between the bands are very clear-cut which indicated that the rampart was erected from layers of rammed earth.The event to be dated is the preparation of the ramparts.Sample MOR3 was provided with independent age control suggesting the construction during 906-1125 AD based on archaeological evidence.Clay sediment samples were collected from this ancient settlement in Orkhon Valley for the luminescence dating.

Sample preparation
Coarse grains (>100µm) were treated with 10% HCl to dissolve carbonates, then dispersed using sodium oxalate and 10% H 2 O 2 to remove organic material, with a heavy liquid (2.58 g cm -3 , 2.62 g cm -3 and 2.70g cm -3 ) to obtain quartz fractions.Quartz is finally etched with 40% HF for 40 min to remove the alpha irradiated outer rinds of the quartz grains, treated with HCl to remove insoluble fluorides and re-sieved.Coarse grains (212µm) were chosen for single grain measurements.Coarse grains (150µm) 2mm aliquots were prepared for OSL and post-IR OSL luminescence measurements.

Instrumentation
Luminescence measurements were made on an automated Risø TL/OSL-DA-20 reader equipped with 90 Sr/ 90 Y beta source; with blue diodes (~470 nm) for quartz stimulation.Quartz OSL and TL were detected through a 7.5mm Hoya U-340 filter (290-370 nm) and D e values were determined using a SAR protocol and a double-SAR protocol.
Single grain measurements were stimulated with a green laser Nd:YVO 4 diode-pumped laser (532 nm) delivering a power density of ~50 W cm -2 .

OSL measurement
The single-aliquot regenerative-dose (SAR) procedure [25] was used for equivalent dose evaluation.Additionally, a double SAR procedure [5] employed IR stimulation prior to blue stimulation.Fitting of the OSL decay curves was executed using Origin 8.6.

Dose rate measurement
The samples for dose rate determination were dried at 50°C until air dry and finally homogenized.157 g sample material was filled into a Marinelli beaker and stored for four weeks to allow equilibrium reestablishment of 226 Ra and its daughter nuclides; the samples were analyzed using high-resolution gamma spectrometry.The external and internal dose rates were calculated using the conversion factors [12] and the cosmic dose rates were calculated following Prescott and Hutton [16].In situ water content, measured shortly after sampling, was used to calculate water attenuation factors.The beta-ray attenuation factors were calculated for the grain size of 150 µm.

Dose rate results
The nuclide activities of 238 U, 232 Th, 40 K were converted to dose rates using the conversion factors given in [12] and shown in Table 1.

D e results
The methodology is far more advanced and better understood for quartz in comparison to feldspar and tends to give reliable estimates, however several studies dealing with OSL quartz dating reported problems with low luminescence efficiency observed most commonly in young samples [17].To obtain an accurate D e we investigate the samples using two approaches: statistical analysis and signal composition analysis.

D e using OSL on quartz
Preliminary green laser stimulated single grain OSL measurements have shown that grains are dim giving around 100 counts; only 4 out of 100 grains gave doses from 1.48±0.33Gy to 5.15±1.05Gy.Therefore, small aliquots A small aliquot is an average of up to 100 grains; were used as proxy for single grains measurements.
Blue stimulated OSL curves are shown in Fig. 3. To examine whether the obtained OSL signal originates from more than one electron trap [25], the OSL decay curves were fitted with a sum of exponential components using the equation [22]: where n i -concentration of electrons at the i-th trap; p i -detrapping probability of electrons from the i-th trap; c -constant.The detrapping probability is the product of the photoionization cross-section and the stimulation intensity.
The fitting results indicate that the fast OSL component and the medium OSL dominate for the initial time period and the slow component is contributing to the overall signal.Furthermore, the quartz sample with a dominant medium component showed an infrared contaminated signal.The IRSL and blue stimulated post-IR OSL signals are very similar in intensity.To further investigate the behavior of quartz, the corresponding TL glow curves where the D e were already evaluated were recorded using small aliquots.Fig. 3b shows the TL curves obtained when heating to 450°C, immediately after beta irradiation (4 Gy) and following 10 s preheat at 70°C.The TL glow curves show that there are two dominant TL peaks, one at 110°C and an additional intermediate peak around 200°C was obtained for sample MOR3.It can be speculated that there is a correlation between the presence of an infrared signal and the presence of a significant medium component.

D e using post-IR OSL on quartz
For MOR3, 24 single aliquots were measured using post-IR OSL which is a technique that stimulates the same aliquot with infra-red (IR) light followed by blue light stimulation [5].

Effect of the precision criteria
The single-aliquot D e distributions are shown using Abanico plots in Figure 4.The sample showed a broad dose distribution: the D e varies from ~1 Gy to ~16 Gy.Grain-tograin variations are assumed to contribute to the D e scatter.Overdispersion (OD) in a D e distribution is a quantitative measure that refers to the relative standard deviation of the distribution of true single grain D e values from a central D e value, after having allowed for estimation of the statistical error [11].The choice of model, e.g.central age (CAM), minimum age (MAM) or finite mixture model (FMM) [11], is based on the OD value.This choice assumes that the precision assigned to each D e measurement is correctly calculated.Many studies have reported up to 20% overdispersion among D e estimates for single aliquots that have been well bleached [13] [9].
Individual aliquots with precision on D e <30% were analyzed using statistical tools [11], CAM D e is 5.14±0.54Gy and overdispersion of 47.5%.MAM D e was 3.7±0.6Gy .. The effect of the acceptance criteria on dose estimate was analyzed for the precision on D e <20% and <10%, respectively.With the increase on the precision the number of aliquots decreased from n=23 to n=11 (<10%).These precise dose estimates gave CAM D e of 4.21±0.22Gy and over dispersion of 17.7%.The CAM and MAM D e values, the precision on D e , the number of aliquots, over dispersion values are shown in Figure 4 and summarized in Table 3.

Effects of signal composition on D e
The post-IR OSL signals are shown in Fig. 5.We observed that the luminescence efficiency varied between aliquots; and the aliquots showing high luminescence efficiency showed a contribution from other components than the fast component.The corresponding growth curves were fitted using

DISCUSSION AND CONCLUSION
For the site MOR-3 single grain analysis failed due to dim signals, it was not possible to carry out reliable single grain measurements for the young sample under study.The smallaliquots comprise of 100 grains from which up to 3-4 grains might contribute to the dose estimate.Furthermore, the small-aliquots showed an infrared signal relating either to the presence of feldspar (REF) or to the medium OSL component.TL peak around 200°C which is indicative for feldspar was apparent for grains showing a dominant medium component.
D e distributions of post-IRSL OSL measurements on quartz were highly overdispersed: the CAM D e is 5.14±0.10Gy with OD of 47.5%.To improve the accuracy of the D e estimation, the acceptance criteria based on the precision on D e were applied and the most precise 11 aliquots yield the CAM D e of 4.21±0.22Gy and OD of 13.3%.Because of this apparent discrepancy further statistical analysis were made.

Figure 2 .
Figure 2. Test trench showing the sampling site

Figure 3 .
Figure 3. a) OSL signal showing different components; b) TL peaks at 110°C and 200°C; the latter is an indication of presence of feldspar

Figure 4 .
Figure 4. Dose distributions displayed using the Abanico plots for the precision on D e <30, <20, <10 % Aliquots with dominant fast component and dominant medium component yield D e of 4.9±0.2Gy and 7.39±0.87Gy, respectively.In concordance with previous speculations, medium OSL component gave higher doses than the fast OSL component.Fitting results on 23 single aliquots of sample MOR3 revealed that only 4 aliquots comprised the dominant fast OSL component.

Table 2 .
Fitting parameters obtained for two different OSL signals

Table 3 .
Summary of dose, dose rate and age results for samples.The selected De and ages are indicated bold