Radiocarbon dating of charcoal and bone collagen associated with early pottery at Yuchanyan Cave, Hunan Province, China

The bedrock of Yuchanyan Cave slopes steeply from the east, where it is about 2.0 m below datum, to the west, where it is 3.2 m below datum. The cave can be roughly subdivided into 3 main areas differentiated mainly by major rockfalls. The western area (mainly square T1) is composed of 2 major lithostratigraphic units: the uppermost intact unit is composed of approximately 80 cm of calcareous anthropogenic deposits resulting from numerous burning events. Specifically, they are stringers composed of white and light gray calcitic ash lenses that in some cases overlie discontinuous bands of red clay, which are approximately 1–3 cm thick by approximately 30–50 cm long. The many ashes and red bands are compact and massive, with millimeter-size aggregates of red clay (Figs. 2 and S1). Well-bedded lenses with varying white and red colored fine-grained sediments are separated by brown colored sediments. The major mineral components of these sediments are calcite, quartz, and clay. The central area (squares T3 and T4) contains brown colored sediments with fewer lenses. The sediments here are also composed mainly of calcite, quartz, and clay. The eastern part of the cave (square T5) contains massive brown sediments with almost no color differences, and stratification is not clearly visible. These sediments are also dominated by calcite, quartz, and clay. Micromorphological analyses of the sediments clearly show that the calcite is mainly composed of wood ash that has been weakly cemented. The ash is remarkably well-preserved, and in many samples, rectangular pseudomorphs of wood-derived calcium oxalate crystals can be observed. Furthermore, much of the red clay (Fig. S2) was purposefully brought into the cave, as there are no possible geological means for clay to accumulate as lenses within the cave. In fact, the massive lenses (e.g., the one shown in Fig. S1) are constructed surfaces and are virtually identical to similar features from the Paleoindian site of Dust Cave in Alabama (27). Infrared spectra of the red lenses show that some of them were exposed to temperatures between 400 and 500 °C based on the absence of absorption peaks around 3600 cm⁻¹. For kaolinite, one of the major clay components in these sediments, these peaks disappear when the clay is exposed to temperatures above 400 °C (28). Note, too, that the clay component extracted from white lenses also often showed these characteristics. Thus, the exposure to elevated temperatures was probably part of the normal use of fires and was not associated with the production of ceramics.

The distribution of bones was more or less uniform in all areas of the cave. In contrast, the charcoal was much less abundant in the eastern T5 square, especially in the deeper part of the section. Of the samples collected from throughout the cave, about 35% of the bones and about 45% of the charcoal were suitable for dating. For both bone and charcoal, the proportions of dateable samples in squares T4 and T5 were much less than for the squares in the eastern parts of the cave. The preservation conditions are clearly much better in the western part of the cave. The results of the prescreening procedure are presented in Table 2.

Seventy-five charcoal samples were selected, prescreened, and pretreated. After the pretreatment, 21 samples were found to contain clay based on their infrared spectra (strong absorptions at 1033 cm⁻¹ together with absorptions at 535 and 472 cm⁻¹). As clay is a potential carbon carrier and therefore a possible contaminant, these samples were excluded. Furthermore, an additional 8 samples dissolved completely during the procedure. The infrared spectra of the remaining samples showed only charcoal (peaks from 1718 to 1595 cm⁻¹) and thus could potentially be used for ¹⁴C analysis. Twenty of these samples that contained relatively large amounts of material were also analyzed by Raman spectroscopy (29). The average fluorescence intensity after the first and last HCl steps decreased in all samples except for 4 (YAS 237d, 540, 559, and T1E 6), indicating that most of the humic acid was removed during the acid-alkali-acid (AAA) treatment. These 4 samples were also rejected.

Sixty-seven bones were analyzed from the different areas in the cave. All were treated with 1N HCl, and an acid insoluble fraction was identified in 43 samples. This fraction was then isolated, and 25 samples were shown to produce a pure collagen infrared spectrum. The weight percentage of insoluble collagen ranged from 0.02% to 1.6%. The infrared splitting factor (IRSF) values of 4 samples were within 2.6–2.9, i.e., the IRSF values of fresh bones (30), while most of the samples had an IRSF value between 2.9 and 3.3. In some of the collagen spectra, the presence of humic acid was detected; therefore, after whole pretreatment, the collagen was again characterized by infrared spectroscopy before target preparation for accelerator mass spectrometry (AMS) dating (31).

A total of 27 samples were analyzed for their ¹⁴C contents. They were selected based on the quality of context and material preservation. Of these, 10 pretreated samples were separated into 2 parts and were prepared separately as duplicate analyses. Three samples (BA 95098, 95057a, 95057b) (12) were analyzed during the 1990s excavations when prescreening procedures were not used (Table 1). Table 3 lists the 40 radiocarbon dates according to excavation square, and within each square the samples are arranged according to increasing stratigraphic depth. The duplicate analyses are also listed. The uncalibrated and calibrated ages are shown. All of the radiocarbon dates were calibrated with OxCal 3.10 by Bronk-Ramsey (32, 33).

The reproducibility of the duplicate measurement analyses (Fig. 3) shows that the data distribution based on the analytical uncertainty follows a normal distribution. This result shows that there is no bias between the measurements, and that there is no consistent difference between charcoal and bone samples from the same depth or level.

Fig. 4 shows a plot of the calibrated ages obtained in each excavation square, and within each square the samples are arranged according to increasing depth. This shows that the upper part of each section contains sediments from around 14,600–13,800 cal BP. Older sediments were found close to the base of the sections in squares T1 D and E, as well as in squares T10–12. Most of these sediments are from around 18,000–16,400 cal BP. A major exception is a bone sample which was just above bedrock in T1 that gave an age of 21,000 cal BP. In square T9, near the western cave wall, the ages are similar and show no trend with depth.

In the field, samples from well-defined contexts (for example ash lenses) were collected with the associated sediments. All charcoal pieces were collected separately, placed in aluminum foil, and dried before closing. For the bones, preliminary tests were conducted on-site by dissolving a small bone fragment in 1N HCl and then determining if a light insoluble fraction was preserved. The light insoluble fraction indicates, but does not prove, that insoluble collagen is preserved. As only about half the bones did have an insoluble fraction, we collected many more samples for an extensive prescreening in the laboratory.

All of the samples that were selected in the field based on context and size for radiocarbon dating were subjected to further prescreening procedures to determine the state of preservation and their suitability for dating based on the quality parameters (defined in ref. 31).

Sixty-four bones from squares T1, T4, T5, T9, T11, T12, T14, and T15 were initially checked in the laboratory for mineral crystallinity based on their splitting factor (38). The splitting factors ranged from 2.6 to 3.0, which is close to the value of 2.7 ± 0.2 for modern bone (30). Only one sample had a splitting factor as high as 4. The HCl insoluble fraction was then quantitatively extracted and used to determine if any collagen was present based on infrared spectroscopy. The FTIR spectra of the 1N HCl insoluble fractions indicated that 23 samples showed good preservation of collagen as indicated by the 1645, 1545, and 1450 cm⁻¹ Amide I and II and proline peaks respectively. In some of the collagen spectra the presence of collagen and/or humic acid was detected. Therefore after the entire pretreatment procedure, the collagen was again characterized by infrared spectroscopy to ensure that it was pure (31) before target preparation for AMS dating.

Many charcoal samples were collected and prescreened in the laboratory before and after acid and alkali treatment using Raman microspectroscopy to assess humic acid contamination (29, 31) removal, infrared spectroscopy to assess clay contamination (31) and loss of weight. The latter proves to be a good indicator of charcoal preservation (24) and in practice determines the yield of clean charcoal and hence whether or not the sample can be dated. Only samples that were well-preserved and free of detectable contaminants were dated.

Sample pretreatment for bone and charcoal was performed at the Weizmann Institute according to the procedure presented in ref. 31. The cleaning procedure for the collagen samples chosen for dating was based on the AAA technique (39). The bone (2 to 4 g) was ground to powder and homogenized. Ten to 20 ml of 1N HCl were added and after 30 min the sample was centrifuged for 3 min at 3000 rpm. The supernatant was removed and the pellet was washed with distilled water (DW) to pH 7. The pellet was re-suspended in 7 ml of 0.1% NaOH for 15 min and centrifuged again for 7 min at 3000 rpm. The supernatant was removed and the pellet was washed with DW to pH 7. The atmospheric CO₂ adsorbed during the alkali treatment was removed by adding 7 ml of 1N HCl for 30 min and washing the pellet until the supernatant reached pH 3. A few milliliters of solution were left over the pellet.

Gelatinization was achieved by heating the pellet in acid solution pH 3 to 70 °C for 20 h (40). The solution was then filtered through a polyethylene filter (Eezi-filter^TM) and then by superfiltration (Vivaspin 20). The filtrate was lyophilized (Heto LyoLab 3000) to produce pure dry collagen (41). The quality of the collagen was checked again using infrared spectroscopy.

The cleaning procedure was based on the AAA procedure (39), except that after each step the pellets were dried at 60 °C, weighed, and a few milligrams were taken for infrared and Raman analyses. The alkaline step was repeated between 2 to 3 times depending on the solution color. In the last step after adding the 1N HCl, the solution was placed on a hot plate and heated slowly to 80 °C for an hour, centrifuged, and the pellet was washed with DW to pH 7 and dried at 60 °C.

Monitoring the removal of humic acids from the charcoal samples by Raman spectroscopy is based on the fact that humic acids tend to fluoresce strongly (42). Measurements were made using a Raman Imaging Microscope (Renishaw) through a 50× lens. The excitation at 632 nm was produced by a 25 mw He/Ne laser. Each homogenized sample was measured 10 times at different places, and the spectra were averaged. The spectral resolution was 4 cm⁻¹ and the range analyzed was1200–2000 cm⁻¹. For details of the method see ref. 29.

Bone and charcoal samples indicated the highest preservation and provided enough material for the AMS measurement.

The carbon content of the samples was analyzed by Elemental Analyzer-ELEMENTAR, vario EL. Samples were weighed according to their carbon contents and sealed with copper oxide and silver in quartz tubes under vacuum system. The combustion temperature was 850 °C. The CO₂ from the tubes was purified and transferred into the gas container separately. The reduction from CO₂ to graphite was performed with H₂/Fe in a new vacuum line. The new system has 2 graphitization lines, each line has 10 reactors. Fe catalyst is cleaned and activated under 450 °C with O₂ and H₂ separately before reduction (43). The graphite was formed at 540 °C. Magnesium perchlorate is used to trap water (44), and it was replaced for every new sample.

The AMS radiocarbon measurements were carried out on a NEC 1.5SDH-1 0.5MV Pelletron with 40-sample MC-SNICS ion source. The accuracy of this system is better than 0.4% and the machine background is lower than 0.03pMC.