The Ngari plateau, covering western Tibet with an average elevation of over 4,500 m MSL (Ye et al., 2016), is reputed as “the top of the Roof of the World” (Tang et al., 2018). This region was once the heart of the ancient Zhang Zhung Culture till the seventh century and the Guge kingdom in the later historical period (10th–17th century) (Huo, 1997; Li, 2017). In the last two decades, a series of archaeological researches, including excavations and surveys, have been conducted in Ngari, mainly concentrated in the Zanda Basin, the cradle of the ancient cultures that originated in the Ngari area.

Qulong, a site from the Early Metal Age, is located around Qulong Village in Zanda County, Ngari Prefecture, Tibet. The site contains nine scattered loci (4,200–1,600 m MSL) on both banks of the mainstream of the upper Langqen Zagbo (Sutlej) River, encompassing an area over 100,000 m². A team of archaeologists from Shaanxi Provincial Institute of Cultural Heritage and Archaeology and Northwest University in China has carried out a long-term project at the Qulong Site beginning in 2018. Numerous archeological remains, including petroglyphs, cave dwellings, tombs, and stone house foundations, have been uncovered (The archaeological report of the Qulong site has not been published. The archaeological data from Qulong presented in this paper were provided by co-author Lin Xi, the team leader of the Qulong excavation for the 2019 season).

During the 2019 excavation season, 96 perforated shell beads were unearthed in three tombs at two loci of Qulong: Tomb 2019QSZM1 at Sazha Locus and Tombs 2019QSM1 and 2019QSM2 at Sailaqinbopu Locus. In these tombs, human skeletons were buried with numerous animal bones and shell objects. Pottery and artifacts made of stone, wood, bronze and iron were also recovered. Although no radiocarbon dates of the three tombs are available yet, they could be typologically dated to around 2,800–2,500 years ago. All these beads are made of the shell of scared chank, a large Indian Ocean conch Turbinella pyrum (Subba Rao, 2003) which has been intensively exploited as raw materials for ornaments production in South Asia since the Harappan period (e.g. Kenoyer, 1991). More than half of them are roughly disc-shaped, while around a quarter show profiles as flat collared beads (e.g., Figure 1C, QSZM1-26), followed by a few triangular, rounded square, and irregular shell pieces. The shell beads vary in size between 17 and 68 mm along the long axis. Most beads have a single central perforation, whereas others bear two or more, apparently used as pendants and/or buttons.

Although all the beads exhibit varying degrees of weathering, red substances remained on one or both sides of the beads. It is worth noting that about one-10th of the samples exhibit a relatively thick red layer, such as QSM1-11a, QSM1-13b, and QSM2-12 (see Figure 1C). At the same time, other shell artifacts only show red traces that could barely be detectable with the naked eye (e.g., Figure 1C, second row). As Figures 1A,B show, a red substance, unclearly paint coat or powder, was found covering the bodies of the tomb owners and the shell beads nearby. Thus, it could be reasonable that the red residue attached to the shell surface was the red substance intentionally splashed onto the deceased during the burial process, signifying an indigenous funeral practice. It may explain the situation of the shell beads with faint red traces (e.g., sample QSM1-17b, Figure 1C), but not all the objects. Consequently, if the composition of this thick layer differs from that of the weak residues that remained on other shell decorations, it is possible that these thick red layers were not left behind due to the burial practice. If the result is opposite, we cannot rule out the possibility that the thick layer was also formed during funeral rites. In addition, take QSM1-11a as an example. It shows different textures of the pigment layers on two sides: the inner surface (of the original conch) displays a thin layer of red paint that is homogenous and bright; the outside, on the contrary, is crusted by a much thicker and darker layer, although part of it has flaked away. The contrast in texture of the substances may suggest a discrepancy in their composition. If it is the case, the formation of the red layers on QSM1-11a could not all have resulted from the red pigment splashing behavior. In short, two questions arise: 1) Are the thick coatings on some of the shell beads comprised of the same material as the faint red pigment remnant on the remaining shell objects? 2) Are the compositions of the red layers on the inner and outer surfaces of sample QSM1-11a different?

To answer these questions, non-destructive methods of pXRF and Raman spectrum were used to analyze the chemical composition and phase composition of pigments on the inner and outer surfaces of eight shell beads. In addition, the powder scraped from the outer surface of sample QSM1-11a was detected by micro-invasive methods including XRD and FTIR.

The red pigment of eight samples contains Fe, indicating that the coloration may be Fe₂O₃ (Table 1). To further confirm the composition of the pigment, samples QSM2-12 and QSM1-16a were analyzed by Raman spectroscopy, and the external surface of QSM1-11a was analyzed by XRD and FTIR. The Raman spectrum of QSM2-12 and QSM1-16a (Figure 2) shows bands at 222 cm⁻¹, 289 cm⁻¹, 406 cm⁻¹, and 606 cm⁻¹, which are entirely characteristic of hematite (Guglielmi et al., 2022). The infrared spectrum of QSM1-11a (Figure 3; Table 1) shows bands at 467 cm⁻¹ and 529 cm⁻¹ which may be assigned to hematite (Bikiaris, et al., 2000), while there are also obvious diffraction peaks of hematite in the XRD spectrum (Figure 4). Therefore, the chemical composition and phase analysis show that the red pigment used in the shell beads unearthed at the Qulong Site is hematite. In addition, XRD results show the presence of quartz, which may be derived from the soil on the surface of the sample.

Interestingly, Ca and P were detected in the outer surface pigments of samples QSM1-11a, QSM2-12, and QSM1-13b. The shell beads matrix is CaCO₃, which will have a certain impact on pXRF analysis. But theoretically, the influence of the matrix on the compositional analysis is relatively small, given the thick coating layer. Therefore, the possibility that a kind of substance rich in Ca and P, such as bone ash, was mixed into the pigment should be considered. The XRD result of samples QSM1-11a indicated that hydroxyapatite was indeed present in the red pigment. The FTIR spectra recorded on the same samples also show the peaks of hydroxyapatite. A strong peak at 1,039 cm⁻¹ is due to the asymmetric ν₃ stretching modes of (PO₄)^3-, the 562 and 598 cm⁻¹ peaks are due to the asymmetric bending ν₄ mode of (PO₄)^3- groups (Opris et al., 2022). Besides the phosphate groups peaks, we observed weaker peaks due to the carbonate groups as follows: the 875 cm⁻¹ peak is due to the symmetric ν₂ mode of (CO₃)^2- groups, and the 1,421 and 1,479 cm⁻¹ are due to the asymmetric ν₃ stretching modes of (CO₃)^2- groups (Opris et al., 2022). The results showed that the red pigment was indeed doped with bone powder. However, the inner surface of the sample QSM1-11a lacks P, suggesting that no bone meal was added during the production of the outer surface pigments.

To further understand the production process of the pigments, the SF value of the infrared spectrum were calculated to evaluate whether the bone meal was heated. The FTIR SF calculation for bioapatite is based on the degree of peak splitting in the PO₄ ³⁻ antisymmetric bending mode (Andrew and John, 1996), which is an important index when discussing bone crystallinity after long-term depositing or burning. First, a local baseline was drawn between 750 and 495 cm⁻¹. Then the heights of the υ₃PO₄ ^3- antisymmetric bending frequencies at approximately 565 cm⁻¹ (A) and 604 cm⁻¹ (B) above this local baseline were measured and divided by the height of the enclosed valley (C) (Weiner and Bar-Yosef, 1990; Todd and Mary, 2001). Finally, the SF value of the bone meal in the samples QSM1-11a was calculated to be 3.7 by the formula SF = (A+ B)/C. The SF index of modern bones is about 2.8–3.0, and that of archaeological samples is about 3.5–4.8 (Guo et al., 2017). SF values are relatively stable until 600°C, then strongly increase to reach values around 8–10 for samples heated from 700 to 900°C (Lebon et al., 2008). The SF of the bone meal in the QSM1-11a pigment is within the range of the SF value of the archaeological sample, indicating that it has not been heated at a high temperature above 600°C.

In general, the red pigments of the shell beads of the Qulong site are all hematite, and most of the samples are doped with bone powder that has not been heated at a high temperature (600°C). Unlike the outer surface, the pigment on the inner surface of sample QSM1-11a does not contain bone powder, which may indicate the different processes of pigment production.

Binding media is an important part of ancient painted cultural relics. In the process of making a color painting, mixing glue and mineral pigment can improve the force between pigment particles and strengthen the adhesion between the pigment and the painting matrix (Yang et al., 2011a). The study on the existence and types of binding media will help us to understand the production technology of the surface pigments of the shell beads at the Qulong site and the utilization of animal and plant resources by the ancients.

The binders in ancient painted cultural relics include proteins, lipids, polysaccharides, resins, etc. For QSM1-11a, the strong absorption bands in the 1,628–1,636 cm⁻¹ were attributed to amide I (C=O stretching), a characteristic of proteinaceous binders (Pellegrini et al., 2016). Additionally, the asymmetrical N-H stretching vibration occurring near 3,300 cm⁻¹, the methyl (CH₃) and methylene (CH₂) groups produce asymmetric stretching vibrations respectively at 2957 and 2925 cm⁻¹ and small symmetric stretching bands at 2,854 cm⁻¹ (CH₂) (Invernizzi et al., 2018). Common ancient proteinaceous binders include animal glue (bone glue, skin glue, fish bladder glue), eggs (whole egg, egg yolk, egg white), and milk. Among them, the whole egg, egg yolk, skin glue, and milk are rich in lipids, and there are C=O stretching bands near 1745 cm⁻¹ of their infrared spectra (Ma et al., 2020). There is no band near 1745 cm⁻¹ in the infrared spectrum of sample QSM1-11a, which may indicate that the proteinaceous binders were not from the whole egg, egg yolk, skin glue, milk, and other lipids rich materials, but the degradation of lipids during the burial process increases the uncertainty of the deduction. Previous studies showed that during the evolution of organisms from the sea to the land, some amino acids in collagen, such as serine (Ser), were gradually converted into hydroxyproline (Hyp) to adapt to the new higher temperature living environment (Song et al., 2008). Therefore, the content of hydroxyproline (Hyp) in the collagen of terrestrial mammals is significantly higher than that of fish, while the content of serine (SER) is lower than that of fish. It is shown that there is an absorption peak generated by the stretching vibration of the C-O on the serine side chain in the 1,076 cm⁻¹ of fish bladder gum, which is not significant in the infrared spectrum of terrestrial mammalian gum (Yang et al., 2011b). There is no peak around 1,076 cm⁻¹ in the infrared spectrum of sample QSM1-11a, which may indicate that the proteinaceous binders are not fish bladder glue, but the effect of protein degradation cannot be excluded.

In addition, due to the presence of bone powder that has not been heated at a high temperature, it is still necessary to further explore whether the protein detected on the surface of the shell beads was attributed to bone powder or intentionally added glue. The amide-to-phosphate ratio (amide I/PO₄) has been used to evaluate the preservation of collagen in bone (Trueman et al., 2004; Fredericks et al., 2012; Kontopoulos et al., 2020). LeBon et al. (2016) studied 42 bone samples from the Pleistocene to modern times. The results showed that the amide I/PO₄ ratio had a good correlation with the nitrogen content, and thus established the relationship between the amide I/PO₄ ratio and N wt% and Collagen wt%: N w t % = 20.6 a m i d e I / P O 4 + 0.31 (1) C o l l a g e n   w t % = 113.13   a m i d e I / P O 4 + 1.69 (2)

According to the guidance of Bouchard et al. (2019), the ratio of the area under the Amide I peak to the area under the phosphate ν₃ peak was calculated, and the amide I/PO₄ value of QSM1-11a was about 0.23. Thus, the nitrogen content is about 5.05% and the collagen content is about 27.71wt%. Depending on the species and type of bone, fresh bone contains 20–35% organic matrix, of which ∼90% is collagen, giving bone a %N of ∼3.5–4.5% (Brock et al., 2012; Currey 2012; Richter et al., 2022). Kontopoulos et al. (2020) tested 266 human and animal bone samples from 10,000 BC to 1850 AD, and found that the amide I/PO₄ values were lower than 0.2 and the collagen content was lower than 25%. Meanwhile, the mean value of amide I/PO₄ value in 42 bones was measured by LeBon et al. (2016) was 0.08, the highest value was 0.18, the mean value of N content was 2.29%, and the highest value was 4.25%. In the case of the degradation of ancient proteins and the reduction of the thermal stability of collagen by hematite (Pellegrini et al., 2016), proteinaceous binders should be added to the surface of shell beads at the Qulong Site, resulting in such high nitrogen content and collagen content, even higher than that of fresh bones.

Therefore, FTIR results showed that the outer pigment of sample QSM1-11a was doped with bone powder and then added with proteinaceous binders. There are no characteristic peaks of the whole egg, egg yolk, skin glue, milk, and other lipids-rich binders and fish bladder glue in the spectrum. However, the degradation limits the exploration of the specific ancient binder.