Application of U/Th and 40Ar/39Ar Dating to Orgnac 3, a Late Acheulean and Early Middle Palaeolithic Site in Ardèche, France

Refined radio-isotopic dating techniques have been applied to Orgnac 3, a Late Acheulean and Early Middle Palaeolithic site in France. Evidence of Levallois core technology appeared in level 4b in the middle of the sequence, became predominant in the upper horizons, and was best represented in uppermost level 1, making the site one of the oldest examples of Levallois technology. In our dating study, fourteen speleothem samples from levels 7, 6 and 5b, were U/Th-dated. Four pure calcite samples from the speleothem PL1 (levels 5b, 6) yield ages between 265 ± 4 (PL1-3) and 312 ± 15 (PL1-6) thousand years ago (ka). Three samples from the top of a second stalagmite, PL2, yield dates ranging from 288 ± 10 ka (PL2-1) to 298 ± 17 ka (PL2-3). Three samples from the base of PL2 (level 7) yield much younger U/Th dates between 267 and 283 ka. These dates show that the speleothems PL1 and PL2 are contemporaneous and formed during marine isotope stage (MIS) 9 and MIS 8. Volcanic minerals in level 2, the upper sequence, were dated by the 40Ar/39Ar method, giving a weighted mean of 302.9 ± 2.5 ka (2σ) and an inverse isochron age of 302.9 ± 5.9 ka (2σ). Both 40Ar/39Ar dating of volcanic sanidines and U/Th dating of relatively pure and dense cave calcites are known to be well established. The first parallel application of the two geochronometers to Orgnac 3 yields generally consistent results, which point to the reliability of the two methods. The difference between their age results is discussed.


Introduction
The Orgnac 3 site is located at a place called Mattecarlinque, at an altitude of 320 m, on the southwest fringe of an Urgonian karstic plateau (lower Cretaceous), in southern Ardèche, central France [1][2][3][4] (Figure 1). The site was initially a cave with human settlement, later changed into a rock shelter, and finally became an open-air site [5] (Figure 1). The depositional sequence is 11m thick. The lower archaeological levels (8 to 4a) were deposited in a cave context while the upper levels 2-1 were accumulated in an open-air environment. Seven hominin teeth, in levels 6, 5b and 5a, assigned to Homo heidelbergensis [6], about 50,000 stone artefacts and abundant mammal fossils have been discovered [1]. Bone assemblages indicate the predominance of carnivores in lower levels (8 and 7), cervids in levels 6-5a, bovids in levels 4b-3 and equids in upper levels 2 and 1. According to biostratigraphical correlation, the lower levels (8 to 3) are attributed to the Middle Pleistocene (MIS 9) and the upper levels 2 and 1 to the late Middle Pleistocene (MIS 8). Levallois debitage, marking the beginning of the Middle Palaeolithic, appears in the middle strata and becomes predominant at the top of the sequence, producing changes in tool kits, raw material procurement and subsistence strategies [1,5]. A reliable chronology for this site is thus particularly important for understanding human cultural evolution and the onset of Neandertal culture. The aim of this study is to refine the age intervals using high-precision U/Th dating on intercalated speleothems and the 40 Ar/ 39 Ar method on well-preserved volcanic minerals in the upper strata. Note that both of these methods are considered as reliable for establishing a temporal frame for human evolution.
The lowermost unit I includes five levels (Ia -Ie) composed of bedded-sandy-clay with angular gravels [7,9]. This unit, containing mainly carnivore and reindeer remains (archaeological levels 8, 7), including small sized Canis lupus, Crocuta crocuta spelaea, Ursus thibetanus, Vulpes vulpes, Panthera (Leo) spelaea, Ursus deningeri, Ursus arctos, appears to have been deposited under a generally cold climate [1,4] (Figure 2). Unit II, divided into three archaeological levels (6, 5b and 5a) with a preponderance of Cervus elaphus, Dama clactoniana, Capreolus sussenbornensis and Sus scrofa fossils, is composed of silty deposits with eroded gravels, large fallen blocks and speleothem formations [7], corresponding to a humid and temperate climate (MIS9, [1]) (Figure 2). Further up, unit III is composed of three sedimentary levels (IIIa to IIIc) of clayey sand with angular gravels and blocks, with abundant Bovidae fossils, corresponding to a cool and humid climate. Three archaeological horizons (4b, 4a, 3) can be identified (Figure 2). The uppermost unit IV, including two archaeological levels (1 and 2), is composed of clayey deposits with some gravels [7]. This unit marks the last human occupation of the site, and contains predominantly Equus steinheimensis remains, corresponding to a cooler climate and an open landscape [1] (Figure 2).
Evidence of the emergence of Middle Palaeolithic-type behavior can be observed throughout the depositional sequence with the development of long and complex flaking reduction sequences. In the lower levels (7 to 5a), debitage is mainly represented by centripetal cores. In the middle levels (4b and 4a), the first evidence of Levallois cores can be observed. In the top levels (3 to 1), Levallois cores on flakes are dominant. Two groups of levels may be distinguished by observing the flake-tool kit; levels 8-3 (with a broader diversity of flake-tools) and levels 2-1 (containing a majority of scrapers with thinner retouch). The shaping reduction sequences are limited throughout the whole sequence (bifaces and pebble tools). In levels 2 and 1, the frequency of bifaces is very low (less than 1%), and these are mainly bifacial tools with few

Previous chronological studies
The first dating of Orgnac 3 was carried out in 1985 [10]. Four speleothem samples from archaeological levels 7 and 6, and between levels 6 and 5b were dated with the alpha spectrometric U/Th method (Figure 3). Based on the results obtained, the author proposed that the mean age of four age results 339 −42 +76 ka, should be taken as the best age estimate for the speleothem formations. One of the four calcite samples was also analyzed by the electron spin resonance (ESR) method, yielding an age of 309 ± 34 ka [11]. At about the same time, Debard and Pastre [8] described and analyzed fallout volcanic ashes in the upper archaeological level 2, which is composed of lightly brown silty sand [7]. The volcanic ashes there are yellowish inclusions several tens of centimeters in diameter (an example of such an inclusion is given in Figure  4A, marked as ORG-C1). The authors [8] extracted wellpreserved angular green pyroxenes, which are characteristic of one of the last eruptions of the Puy de Sancy volcano (Mont-Dore, Massif Central, France) [12]. They proposed an age of about 300 ka for the upper level 2. With the fission track (FT) technique, Khatib [7] analyzed 22 zircons from volcanic ashes from the same level, obtaining an age of 298 ± 55 ka ( Figure  3). Masaoudi [13] presented the results of U/Th and ESR dating of bones and teeth and ESR dating of calcite and quartz samples from different levels (Figure 3). The measured dates are overdispersed, and do not conform well to the stratigraphy. Recently 40 ,Ar/ 39 Ar dating was carried out on 16 sets of sanidine grains [9]. Four of them yield ages too old to be acceptable because of contamination by inherited K feldspar grains, while the remaining 12 ages are between 276 and 326 ka with a weighted mean of 308.2 ± 6.8 ka. Roger et al. [14] considered that the tephra layers at the Praclaux and du Bouchet maars (French Massif Central) and at Orgnac 3 all came from the eruption of the Sancy volcano centre. Based on the weighted mean of 40 Ar/ 39 Ar dates on a series of sanidines from the Praclaux and Bouchet lakes, they assigned an age of 275 ± 5 ka to the Sancy eruption. Recently, Nomade et al. [12] recalculated the age according to ACR-2 at 1.193 Ma and obtained an age of 279 ± 5 ka. Roger et al. [14] also determined a step-heating plateau age of 300 ± 2 ka from sanidine grain populations (Figure 3). For their experiments, the neutron fluence (J) was monitored with a biotite Bern B4B with an age of 17.25 Ma and a sanidine Draz with an age of 24.99 Ma [14], recently recalibrated to 25.42 Ma [15].

U/Th dating
Isotopic measurements of the fourteen speleothem samples are presented in Table 1 Figure 5). Note that PL1-2 and PL1-2a are taken from exactly the same position. But compared with PL1-2a (with a [ 230 Th/ 232 Th] activity ratio of 17.6), the more contaminated PL1-2 (10.9) gives a significantly younger age result (218 ka << 275 ka, Table 1). For this, we tend to consider that the samples with a low [ 230 Th/ 232 Th] activity ratio may have undergone metamorphism leading to underestimated age results. Therefore these three (ICP-MS) U/Th ages were excluded (Table 1, Figure 5). The precision of ICP-MS isotopic ratio measurements is much better in comparison with the previous alpha measurements, as shown in Figure 6. For the base of the upper part of the first speleothem PL1, U/Th dates range from 265 ± 4 ka (PL1-3) and 295 ± 8 ka (PL1-4a) (levels 5b-6)   biased by post-depositional diagenesis and is thus excluded from this study. U/Th ages indicate that both speleothems from levels 7-6-5b range from 265 ka (marine isotope stage 8, MIS 8) to 312 ka (MIS 9).

Ar/ 39 Ar dating
Fifty-seven sanidine grain populations were analyzed with the total-fusion method (Figure 7). In order to obtain precise data, 50-150 sanidine grains were analyzed for each measurement 40 Ar/ 39 Ar ages with counting errors at 2σ are shown in Tables 2, 3, 4. Ten samples with determined ages older than 550 ka are most likely contaminated by inherited K feldspar grains. These samples, representing 17.5% of the results, were not taken into account (Tables 2, 3, 4). The abnormally old age results may be explained by the presence of old minerals, such as plagioclases or sanidines, extracted from the base rock during the Sancy volcano eruption.
Note that if 10 out of 57 samples are heavily contaminated, unless only one "inherited K feldspar" grain is enough to cause an abnormally old result, we should consider the possibility that the rest may be more or less contaminated.
The measurement of 36 Ar in 40 Ar/ 39 Ar dating allows for the plotting of the inverse isochron graph. The results are a series of data points ranging from pure atmospheric argon to pure radiogenic argon. A regression line through these data points forms an inverse isochron and the point at which the isochron intercepts with the x-axis yields the 39 Ar/ 40 Ar* of the samples and therefore the age. In Figure 9, a regression line was plotted using 36 Ar/ 40 Ar and 39 Ar/ 40 Ar ratios (n=43/57), excluding 14 samples. Ten contaminated samples, K346-20 and K346-11 with high 39 Ar/ 36 Ar ratios, K346-21 with a negative 36 Ar/ 40 Ar ratio and K346-51 were eliminated in order to attain the atmospheric ratio (298.56; [25]) (Figure 9). In these conditions, the atmospheric ratio obtained ( 40 Ar/ 36 Ar) 0 is 299.0 ± 41.8 (2σ) and the intercept inverse isochron age is 302.9 ± 5.9 ka (2σ) (MSWD=2.57). This is in agreement with the weighted mean of 302.9 ± 2.9 ka (2σ) (Figure 8).

U/Th and 40 Ar/ 39 Ar dating comparison
The 40 Ar/ 39 Ar weighted mean age of 302.9 ± 2.9 ka and the inverse isochron age of 302.9 ± 5.9 ka are older than the upper limit of the measured U/Th age interval 265 to 312 ka ( Figure  10). The 40 Ar/ 39 Ar apparent ages determined in the higher level 2, should be younger than the U/Th apparent ages obtained in the median levels. This small difference in age (about 30 ka, which is not a major difference but is nonetheless statistically significant), could be explained by a slight contamination of sanidine grains or by a minor excess of 40 Ar. Another hypothesis which may explain the age difference between the two methods may be that the volcanic minerals were transported to the site tens of thousand years after the Sancy eruption. In order to improve our 40 Ar/ 39 Ar dating results, we have tentatively used the step-heating method to highlight a possible 40 Ar excess. In spite of the presence of excess 40 Ar incorporated in minerals during crystallization [26], Renne et al. [27] have demonstrated in other studies with this method that a sanidine sample less than 2000 years old can be dated with 5% precision. However, for Orgnac 3, this step-heating method, using a VG 3600 mass spectrometer with a Daly detector, required a significant amount of material (up to 1,000 grains, 200 μm) which considerably increased the proportion of inherited K feldspar grains and thus increased the probability of an unreliable apparent age.
In conclusion, taking into account our total fusion multigrain analyses, U/Th dates and errors, it seems reasonable to conclude that the Orgnac infilling is contemporaneous with MIS 9 and 8 ( Figure 11). The U-series date of 265 ka may mark a minimum age for the level 5b, when the Levallois flaking technique began to appear at the site (Figures 2, 5 and 10). This date is concordant with the biostratigraphical pattern which attributes levels 2 and 1 to MIS8 [1].
Thus, according to our U/Th ages, preliminary 40 Ar/ 39 Ar and the comparison between the two dating methods, Orgnac 3 is one of the oldest sites with the systematic use of Levallois knapping. As evidenced in [5,1], this temporal framework indicates the emergence of new technological behavior in southern France and Europe during MIS 8. Standardized core technology such as Levallois knapping can be observed in a few well-dated European sites close to the limit between MIS 9 and MIS 8, such as La Micoque (L2/3) (France), Gran Dolina (TD11/10), Bolomor (Spain) and la Baume Bonne (France) attributed to MIS 8 [1].

Conclusion
For the first time, U/Th and 40 Ar/ 39 Ar dating methods have been applied together with greater precision than in previous studies for dating a Middle Pleistocene site. The 40 Ar/ 39 Ar dating gives a weighted mean age of 302.9 ± 2.9 ka (2σ) for upper level 2 of the Orgnac infilling while the U/Th method yields an age range of 265-312 ka for middle levels 7-6-5b. The age results from the two dating methods are generally consistent, which underlines their reliability. On the other hand, the difference between them is statistically significant taking into account the stratigraphical location of the samples. There are two possible explanations for an older 40 Ar/ 39 Ar age of 302.9 ± 2.9 ka (2σ). The first is that the volcanic minerals were transported to the site tens of thousand years after the Sancy eruption. The other possibility is that the analyzed sanidine grain populations have been systematically contaminated by inherited K feldspar grains. To check for the second
The selected bulk subsamples were physically cleaned with ultrasonic methods [30]. U/Th chemistry was conducted in a class-10,000 metal-free clean room with class-100 benches at the High-precision Mass Spectrometry and Environment Change Laboratory (HISPEC), Department of Geosciences, National Taiwan University [30,31]. U-Th isotopic compositions and concentrations were determined on a sector-field inductively coupled plasma mass spectrometer (SF-ICP-MS), Thermo Fisher ELEMENT II [17] or a multi-collector ICP-MS (MC-ICPMS), Thermo Fisher NEPTUNE, with a dry introduction system, Cetac ARIDUS [16]. Uncertainties in all ICP-MS U/Th isotopic data were calculated at 2σ level and include corrections for procedure blanks, multiplier dark noise, abundance sensitivity, mass discrimination, and the occurrence of isotopes of interest in spike solution. Age was off-line calculated [17] with decay constants of 9.1705 × 10 -6 yr -1 for 230 Th and 2.8221 × 10 -6 yr -1 for 234 U [19], and 1.55125 × 10 -10 yr -1 for 238 U [20].

Ar/ 39 Ar dating
Volcanic sediment samples were collected from level 2, about 1m below the top of the depositional sequence ( Figure  4). As sanidine is a proven chronometer [32], the largest possible and well-preserved sanidine grains (200-300 µm) were extracted using standard heavy liquid methods and then hand picked under a binocular microscope (Figure 7). The obtained sanidine grains are angular and quite well preserved. Their chemical composition was estimated using scanning electron microscopy (Figure 7) with Energy Dispersive X-ray     Table 4).   7). Gas was extracted with an infrared continuous laser and purified in stainless and glass extraction line using two Al-Zr getters and a N 2 cold trap. System blanks were run for every two or three analyzed samples. The mass spectrometer is a VG 3600 with a Daly detector. Mass discrimination was monitored by regularly analyzing one air pipette volume. The ultimate accuracy of the 40 Ar/ 39 Ar method depends on welldated homogeneous standards [15,23]. J values were calculated using an age of 1.194 Ma [22] for ACS and the total decay constant of [33].       [28] after [29] with modifications. The shaded vertical envelope (± 2σ) shows the occurrence of the Orgnac infilling after U/Th and 40 Ar/ 39 Ar dating, close to the transition between MIS 8 and MIS 9.