Sequencing Analysis and Radiocarbon Dating of Yarn Fragments from Six Paracas Mantles from Bundle WK12-382

Abstract

The Necrópolis de Wari Kayan, at the Paracas site in the coastal desert of south–central Peru, is a large archeologically excavated mortuary complex with fine textile preservation, dated approximately to 2000 BP. This study investigates loose yarns associated with textiles from Wari Kayan tomb 12 (bundle 382), collected by the late Dr. Anne Paul in 1985 at what is now the Museo Nacional de Arqueología Antropología e Historia del Perú (MNAAHP). Sequencing multiple state-of-the-art analyses, including direct analysis in real time mass spectrometry (DART-MS), high performance liquid chromatography (HPLC) with diode array detection, and accelerator mass spectrometry, on the same small sample, seeks to “squeeze out every drop” of information. Six mantles from the outer layer include different sets of color hues and values, representing either different time periods or different producer groups. Plasma oxidation at low temperature (<100 °C) prepared carbon dioxide for AMS radiocarbon analysis. Fibers remaining after oxidation were combusted for light-stable isotope analysis. The sequential analysis results in fiber and dye composition, radiocarbon age, and stable isotope fractionation values may suggest fiber origin, continuing and updating a project started over 40 years ago.

1. Introduction

The cemetery known as the Necrópolis de Wari Kayan, at the Paracas site in the coastal desert of south–central Peru, is the largest archeologically excavated mortuary complex of the pre-Hispanic Andes, and one of the best for preservation of organic materials, including textiles. Excavation of the tombs, dating from about 2300–1900 BP, was directed by Julio C. Tello in 1927–1928, and most of the associated materials are custodied by the National Museum of Anthropology, Archeology and History of Peru (MNAAHP). Eight percent of the mortuary bundles are large, conical, and multilayered, including between fifty and over two hundred artifacts arranged around a seated individual. Even the simpler bundles often include elaborately embroidered garments and complex headdress elements. The information available from other sites in the surrounding region indicates that both the persons and objects in this cemetery reflect a far-flung network of social relationships.

The Paracas site is located on the neck of a peninsula of the same name, on the desert Pacific coast of Peru. Its rich maritime environments are nourished by a combination of rocky headlands and sandy bays between the cold waters of the deep Nasca trench to the south and the warmer, shallow Bay of Paracas to the north. The dry Andean foothills to the east rise quickly to cool, green highland valleys and the high rolling plains and rocky peaks of the continental divide. The ‘catchment area’ of social communities who potentially contributed persons, textiles, and other artifacts found in tombs at the Paracas site includes parts of the Cañete, Chincha and Pisco watersheds to the north and the Ica and Nasca watersheds to the south, as well as regions further afield. The materials used to make those artifacts drew on diverse environments: the ocean shores and coastal wetlands, the rivers and forests of the coast, semi-arid and temperate Andean slopes, and high-altitude lakes and pastures. Metals and mineral ores could be sourced in deposits throughout the region, but in some cases traveled further [1,2,3,4,5]. Certain other resources traveled great distances, such as obsidian from Quispisisa and elsewhere [6,7], Spondylus shell from equatorial waters [8] and bird feathers from the Amazon tributaries [9].

At the Paracas site, communities associated with artifact styles defined as late Paracas, Topará and early Nasca met and interacted between about 350 BC and AD 50, in regional gatherings involving seasonal fishing encampments and the development of elaborate funerary practices. There is diverse evidence of social interaction, mutual influence and exchange [10,11], as well as interpersonal violence [12,13]. The tombs are renowned for their diverse, beautiful, and well-preserved textiles, as well as a wide range of other finery [14,15,16,17]. Garments, personal ornaments, tools, and weapons were placed in layers among thick wrapping cloths to form a conical bundle around each seated individual (Figure 1). At the time the bundle was placed in a tomb, other artifacts were grouped around it. Most of the Paracas tombs were excavated between 1925 and 1928, and the more elaborate funerary bundles were studied in the early 20th century by project director Julio C. Tello and his collaborators [14,18,19,20,21,22,23,24]. One elaborate bundle was more recently studied by a curatorial team at the Museo Nacional de Antropología, Arqueología e Historia [25].

Dwyer [26] and Paul [16,27] developed a style-based sequence of Paracas textiles and tombs, but its relationship to absolute chronology is not clear [28,29]. Anne Paul’s extensive photographic documentation of textiles from the Paracas tombs was part of the re-cataloging of the MNAAHP collection between 1978 and 1981. She went on to document tomb assemblages [16,28,30], analyze design principles in certain garment types [17,31,32], and study the social organization of mantle embroidery [33]. Paul’s interest in color led her to collaborate with chemists in dye analysis [34,35]. Subsequent research by Frame [36,37,38], Aponte Miranda [39], Thays [25], and Peters [10,11,40,41,42,43,44] has focused on analysis of archeological context, gender, garment forms, ritual practices, and changing sociopolitical identities and relationships. Collaborations with bioanthropologists [45,46,47,48,49,50,51,52] has been essential for better understanding the lives of artisans, fishermen, farmers, herders, warriors, and leaders buried at Paracas. Continuing archeological research on this period in the Paracas region [2,53,54,55] and the Chincha [56,57,58,59,60,61], Pisco [62,63,64,65], Ica [6,66,67,68,69,70,71,72], and Nasca [3,48,73,74,75,76,77,78,79] regions provides new data and analytical perspectives that inform this research.

There is an inherent strategic importance to directly dating textiles [80], especially multiple ones from a layered mortuary bundle such as those described herein. Directly dating the individual textiles provides for a more precise chronology of the creation and use of the Necropolis of Wari Kayan. First, radiocarbon dates based on the human remains show marked marine reservoir effect, so dates on textiles should more accurately document their creation, as this reservoir effect should not influence either the camelids or the plants from which they are made. Second, the creation of a textile may occur after the individual is deceased or may have occurred long before the object was placed in the tomb. Radiocarbon dates on textiles, then, will affect our modeling of ritual practices. Third, because textile fibers can be directly dated, such dates provide a firm cross-check on style analysis, allowing us to correct assumptions. Individual textile styles often contradict expectations of mortuary bundle chronology, with ‘late’ styles found near the body while ‘early’ styles have been placed on the outer layer. This strongly suggests that more dates, and more efficient dating performed in combination with other technical analyses, will have a great impact on understanding the directions of influence among textile producers and the processes of tomb assemblage creation.

Combustion is the standard method for converting organic material of any kind into CO₂, which is then usually converted into a graphite pellet for AMS radiocarbon dating. Though graphite has the advantage of providing a higher ion current, some laboratories utilize a gas ion source, allowing direct radiocarbon measurement of the CO₂ itself [81]. The combustion process completely consumes the sample: nothing remains after the standard acid–base–acid wet chemical pretreatment followed by heating the remaining material to 800 °C in a sealed ampoule with oxygen. The plasma-chemical oxidation process was originally intended to “extract” organic carbon from rock paintings that did not contain obvious organic material, allowing them to be directly dated for the first time [82]. In the intervening years, plasma-chemical oxidation (PCO) has been demonstrated as a minimally destructive approach to dating fragile and perishable materials like fibers and textiles [83,84] because it uses the chemical reaction of the oxygen plasma, consisting primarily of atomic oxygen, with the surface of the exposed organic material. Samples can thus be oxidized to produce CO₂ without consuming them in the process, though small samples may react rapidly depending on the plasma conditions, leaving too little for further analysis.

The most efficient approach to the analysis of the textiles is one that utilizes a single sample to obtain all of the desired information. As accelerator mass spectrometry requires only small amounts of carbon for dating, it has made the dating of single yarns not just possible, but routine. Dye analysis, too, can be completed on small samples without any sample preparation [85,86,87] or on low volumes of extracts from minute samples [88,89,90,91,92]. Combining sequential analysis and dating is not new. Smith et al. [93] combined HPLC-MS/MS with combustion and AMS sequentially to identify dyes and directly date yarns from a Nazca textile to determine its authenticity. Sequentially applying DART-MS and PCO, as reported by Armitage and Jakes [94], has both benefits and drawbacks: DART-MS is best for dye screening, rather than comprehensive characterization, while PCO can be performed so the yarn is not consumed in the process of making CO₂ for AMS dating.

This pilot study sequences analytical methods with PCO sample preparation and AMS radiocarbon dating on a single yarn fragment shed from a larger textile object, following the general scheme shown in Figure 2. The samples were originally collected decades ago with the intent of applying scientific studies to determine what could be learned about the textiles from Paracas funerary bundle contexts. Thus, these studies are one step in the culmination of many decades of work to characterize and date these objects.

2. Materials and Methods

2.1. WK12-382 Textiles

The materials in this study were collected from loose fragments and fibers closely associated with the funerary bundle textiles from Wari Kayan located at what is now the Museo Nacional de Arqueología, Antropología, e Historia del Perú (MNAAHP) in Lima, Peru, by the late art historian Dr. Anne Paul in 1985. The samples were to be used for an Earthwatch-funded grant project with one of us (KJ) to complete technical analysis of the fibers, textile structure, and dyes. Some preliminary studies were undertaken at that time [34,95,96], but the majority of the samples have remained unstudied until now. This pilot study focused on just six objects within a single funerary bundle, specifically the six outermost mantles from WK12-382, one of the largest bundles from the Necropolis of Wari Kayan.

Each textile layer and artifact was given an identifying number when the bundle was unwrapped in February 1929 by Peruvian archeologist J.C. Tello. Due to political pressure to send this large bundle to Spain, its number had been changed from 12 to 382 [97]. WK12-382 was described as a large conical bundle wrapped in a deteriorated thick woven matting, with the apex shaped to resemble a head. Each layer of textiles and objects was numbered roughly sequentially from the outermost to the inside. Under the matting were found two huarango (Prosopis Neltuma pallida) staffs, placed vertically (382-1 and -2), followed by a thick cotton cloth (382-3). Of the next seven layers in WK12-382, six were the large, embroidered mantles of this study; 382-8 consisted of a yellow feather fan.

Table 1. Materials from WK12-382 investigated in this pilot study.

Object Number	Style	Description	Samples Collected by Dr. Paul in 1985
382-4	Linear	Mantle: light red plain weave, embroidered borders and transverse bands, med. green ground; interlocking feline motifs in 5 colors; yarn fringe	Brown–red ground cloth (woven fragments and loose strands, including some identified as warp and weft)
			Blue embroidery thread
			Border fringe and weft yarns in red, green, orange, gold, and black
382-5	Linear	Mantle: dark blue plain weave; embroidered borders and center panels, med. red ground; standing warrior motifs in 3 colors; yarn fringe	Black yarns (2 warp, 1 weft)
			Embroidery threads, loose, in black, reds (2)
			Red fringe
382-6	Linear	Mantle: ochre plain weave; embroidered borders and center panels, med. red ground; horizontal figures with serpentine appendages in 3+ colors; yarn fringe	Yellow–orange ground cloth (unidentified strands, plus some designated warp and weft)
			Red fringe
382-7	Block	Mantle: red/ochre plain weave; embroidered borders and center figures, dark purple ground; standing warrior motifs in 9 colors; yarn fringe	Red–orange yarns identified as warp and weft
			Fringe yarns in purple–brown, green, red, and yellow
382-9	Block	Mantle: dark blue plain weave, embroidered borders and center figures, med. green ground; horizontal figures with streaming hair in 10 colors; yarn fringe	Black yarns identified as warp and weft
			Green yarn from border background
			Embroidery yarns in orange, red, and brown
			Brown fringe yarns (were green)
			Blue yarn
382-10	Block	Mantle: plain weave bands in dark blue, yellow, green, red; embroidered borders, med. green ground; standing warrior motifs in 11 colors; tabs embroidered head motifs; yarn fringe	Green–brown disintegrating fibers from border ground cloth
			Loose threads in purple and green

describes the materials from those mantles that were studied herein. Figure 3 shows examples of two of the mantles from which shed yarn fragments were analyzed in this study; additional images are found in the Supplementary Materials.

The mantles fall into two distinct styles, with 4, 5, and 6 sharing the late Linear mode features defined by Rowe [98] and 7, 9, and 10 in the Block-Color style, which is assumed to be later than the “Linear” mode, even though it is found in this bundle underneath objects assigned to that mode. Many of the surrounding tombs have been studied [15,45], including bundles with earlier and later textile styles spanning from the late Paracas phase 10B to early Nasca phase 2A. Paul [16] characterized bundle 382 as contemporary with Nasca 1b. The imagery in each of these mantles recurs in a series of textiles documented in other Wari Kayan tombs. In addition, each of the mantles matches personal garments like tunics, wrapped skirts, or headcloths in this particular tomb, though the color schemes may differ somewhat. From these similarities arise a number of questions about correlations between garments in time and technology, as well as in imagery, which will be explored in future studies.

With those type of questions for the future, this pilot study sought to answer more specific scientific questions to provide some insight into the age and production processes of these six embroidered mantles. Dye analysis could suggest different producer communities, as different compositions may indicate differences in technology or access to dyestuffs. Radiocarbon analysis would determine if the mantles were approximately contemporary, or if different artifacts or their elements had different histories. Finally, analysis of the stable carbon and nitrogen isotopes in camelid fibers has been used by others to evaluate the environment (based on the dietary composition) in which the camelids used in the textile production were raised [99,100]. As a pilot study, measurement of the δ¹³C and δ¹⁵N values of the yarns can provide preliminary indications whether the fibers within one textile originated from the same environments, as well as whether the isotopic signatures vary among the six mantles.

2.2. DART-MS

A small (<1 mm) bundle of fibers was cut with a sterile scalpel blade from the end of the yarn fragment selected for sequential analysis and held with a pair of self-closing tweezers. The fiber bundle was then introduced into the space between the DART ionization source (IonSense, Saugus, MA, USA) and Orifice 1 of the AccuTOF mass spectrometer (JEOL USA, Peabody, MA, USA) with the He carrier gas at 400 °C. The voltages of Orifice 1 and Orifice 2 were set to ±30 V and ±5 V, respectively, with the polarity determined by the mode of operation. The AccuTOF peaks voltage was set to obtain the maximum intensity for the range of interest, 150–1000 Da. Spectra were collected in both positive and negative ion modes. Before each sample was introduced to the ion source, PEG-600 in methanol was run for calibration of the data. All data files were processed with TSS Pro 3.0 software (Shrader Analytical and Consulting Laboratories, Detroit, MI, USA) and then analyzed by Mass Mountaineer Software (various versions, RBC Software, provided by R.B. Cody).

2.3. HPLC-DAD

Following the DART-MS screening, the yarn fragments were extracted for analysis by high performance liquid chromatography with diode array detection (HPLC-DAD). Though not as selective or as sensitive as LC-MS/MS, this equipment was readily available and has been widely used in previous studies [101,102]. Three different extraction protocols were employed, two of which were derived from Smith et al. [93]. First, the yarns were treated with a 0.1 M oxalic acid–methanol solution (1:1 v/v) and heated at 70 °C for 30 min. This method preserved glycosidic linkages that may provide species-specific information about the dyes present; overall, this yielded a greater number of peaks but a lower overall signal in the HPLC-DAD. Yarns containing indigoids were extracted a second time using DMSO, with brief heating to 70 °C. It was noted that the indigoid extracts changed color if left for more than 24 h before HPLC analysis, leading to much lower signal and even conversion from blue (suggesting primarily indigotin) to purple (suggesting significant conversion to indirubin). The reason for this change is not well understood, but suggests keeping the time between DMSO extraction and analysis short.

Because the signal for the oxalic acid–methanol extracts was very low for the red dyes, another extraction protocol was applied to those samples using a solution of methanol and hydrochloric acid (30:1 v/v). A fresh fragment of yarn (no more than 2 mm in length) was sonicated for 30 min at room temperature in this solution. Though the presence of the strong acid likely cleaves the glycosidic linkages, the signal for the anthraquinone dyes was significantly higher with this method.

HPLC analyses were carried out on a Shimadzu LC-20AT pump paired with an SPD-M20A photodiode array detector. The flow rate was 1 mL/min and a 20-μL injection volume was used for both of the separation programs employed. The analytical column used was a Restek Ultra AQ C18 column from Phenomenex (25 cm × 4.6 mm and 5 μm particle size) with temperature control provided by a heat tape controlled with a variable AC transformer to maintain the column at 30 °C. For the oxalic acid–methanol and DMSO extracts, 0.1% (v/v) formic acid in deionized water (solvent A) and acetonitrile (ACN, HPLC grade) as solvent B served as mobile phases. The program for separation of these extracts was 90% B for 6 min, followed by a linear gradient to 10% B in 64 min, followed by 90% B in 0 min held for 10 min (total run time 80 min including re-equilibration). For the preparations in methanol/HCl (after Manhita et al. [103,104], the mobile phase consisted of acetonitrile (A) and 2.5% (v/v) aqueous acetonitrile with 0.5% (v/v) formic acid (B) in a gradient of 0–100% A from 0 to 10 min, 100% A from 10 to 15 min and a 5 min rinse of 100% solvent B between runs. UV-vis spectra were acquired in the range of 200–800 nm with a resolution of 4 nm. Data was processed using Shimadzu LabSolutions software.

The solvents used in the methods described here were acetonitrile (ACN) of HPLC grade (Sigma-Aldrich, St. Louis, MO, USA); deionized water (H₂O) from a Thermo Barnstead Smart2Pure water polishing system; methanol and dimethyl sulfoxide (DMSO), all of HPLC grade (VWR Chemicals, Radnor, PA, USA); formic acid (FA) of high purity grade (Amresco Inc., Solon, OH, USA); and hydrochloric acid (HCl) from laboratory stocks. Dye colorant reference compounds were obtained commercially from a variety of sources.

2.4. Plasma Oxidation and AMS Radiocarbon Dating

Yarns that were to be radiocarbon dated after dye extraction were pretreated by sonicating twice for 30 min each time in pH 8 phosphate buffer to remove any residual dye extraction solvents and possible humic acid contamination. Samples were then rinsed with deionized water to neutrality and placed in ceramic combustion boats that had been previously baked at 500 °C to remove all carbon. Samples in combustion boats were dried overnight at 100 °C, weighed, wrapped in clean aluminum foil and stored in a desiccator until oxidation could be performed.

For the plasma oxidation, each sample in its combustion boat was placed directly into the plasma chamber under a positive flow of ultrahigh purity (research grade, 99.999%) Ar gas. Once the chamber was sealed with a copper-gasketed metal flange, the Ar was pumped out and then refilled to a pressure of 200 mTorr for two 15 min argon plasmas at 30 W of 13.54 MHz radiofrequency (RF) power. The unreactive Ar plasma is used to remove surface adsorbed gases through physical interactions between the energetic Ar atoms in the plasma and any atmospheric gases adhering to the sample or chamber.

Following the completion of the Ar plasmas, the chamber was then evacuated to an ultimate pressure of ~10⁻⁷ Torr. Vacuum integrity checks (VICs) prior to plasma-chemical oxidation indicated that no significant leaks were present in the system. Oxygen gas of research grade (99.999+ %) was used to generate reactive plasmas at different RF wattages to oxidize the surface of the cotton (cellulose) or wool (keratin protein) yarn samples to yield carbon dioxide.

Generally, at least 20 µg of carbon as carbon dioxide is necessary to obtain a reliable radiocarbon date, though recent improvements in background correction and graphitization have reduced that amount to around 10 µg. Samples at or near that level are saved for AMS radiocarbon analysis. Carbon dioxide resulting from the plasma reaction was collected by cooling a glass finger on the plasma system manifold with liquid nitrogen. The glass tube was then flame sealed and sent to the Keck Center for Accelerator Mass Spectrometry at the University of California, Irvine for radiocarbon analysis. Samples were prepared and measured using standard AMS techniques [105,106].

Briefly, CO₂ sample tubes were cracked in a metal bellows on a vacuum line, cryogenically purified, reduced to graphite with hydrogen using an iron catalyst, and pressed into aluminum cathodes for AMS analysis. Sample isotope ratios were measured repeatedly over the course of a 24 h run and normalized to those of six aliquots of the OX1 radiocarbon standard (NIST SRM 4990B) and were corrected for isotopic fractionation using δ¹³C values measured online with the AMS spectrometer. Small aliquots of ¹⁴C-free coal were plasma oxidized and graphitized and measured along with the unknowns as process blanks, to evaluate background contributions from ¹⁴C-free carbon. Results are reported as conventional radiocarbon ages according to Stuiver and Polach [107] with quoted uncertainties that include contributions from the scatter in repeated runs plus the normalizing standards and background subtraction, as well as counting statistics.

2.5. Stable Isotope Analysis

Samples were cleaned for stable isotope analysis by briefly vortexing in a 2:1 solution of chloroform–methanol to remove any residual oils or other contaminants resulting from handling. Samples were air dried, then manually cut into short lengths and weighed using a model WXTE balance (Mettler Toledo, Switzerland; 0.001 mg precision) into 5 mm by 8 mm tin capsules (EA Consumables; Marlton, NJ, USA) for analysis. Samples were analyzed for δ¹³C, δ¹⁵N, weight percent (wt %) C and wt % N using an Isotope Cube elemental analyzer (Elementar, Germany) coupled with a Vision isotope ratio mass spectrometer (Elementar, UK). Sample combustion was performed at 950 °C using a reactor packed with copper oxide and silver wool, with subsequent reduction (to reduce NO_x compounds) using a reactor maintained at 650 °C and packed with short copper wires. Raw data were corrected using in-house standards and applying a two-point normalization. The in-house standards included a protein standard (B2155; procured from EA Consumables: δ¹³C = −26.98 ± 0.13‰/δ¹⁵N = +5.83 ± 0.08‰) and ground salmon (δ¹³C = −19.51 ± 0.06‰/δ¹⁵N = +11.29 ± 0.06‰) which themselves were calibrated against certified international standards to enable reporting of results in standard δ notation referenced to Vienna Pee Dee Belemnite (VPDB) for δ¹³C (using standards NBS18, NBS19, and USGS24; [108]) and atmospheric nitrogen (AIR) for δ¹⁵N (using IAEA N1 and IAEA N2; [109]).

2.6. Colorimetric Analysis

A Nix Spectro 2 (2 mm) colorimeter from Nixsensor.com was used with the Nix Toolkit application for Android with default app settings on a single yarn from Mantle 9 to generate a single-scan reflectance spectrum and obtain a standardized color measurement, expressed in, among other readings, CIELAB space. There are no settings for the colorimeter.

3. Results

Not every sample described in Table 1 was subjected to the complete sequential analysis. Samples for sequential analysis were selected primarily on the basis of size, where only relatively large samples were chosen when possible. A second criterion was if the sample came from the ground cloth of the mantle, though this was not always possible. All of the samples were screened for dyes by DART-MS and then extracted as described in the Methods for HPLC-DAD. Both the extracts and the fibers that were not used for plasma-chemical oxidation and/or stable isotope analysis were retained for future studies and dating.

3.1. Dye Analysis by DART-MS and HPLC-DAD

The colors of the yarn samples are based on our own observations and those of Dr. Paul at the time of collection. This has led to some confusion in samples that could be described accurately as red, brown, orange, red–brown, etc., depending on the light and angle from which the color is observed. This situation is further complicated by changes in color that have occurred from when the mantles were first unwrapped from the bundles to when they were sampled to today. One example of this is the fringe from Mantle 9. In Tello’s archives [110] (pp. 464–468), the fringe is described as green (“franjo verde”), but today the fringe yarns collected by Dr. Paul appears brown. Because a color description can be subjective, a colorimetric measurement was taken on the fringe sample from Mantle 9 using a Nix Spectro 2 (2 mm) colorimeter from Nixsensor.com, resulting in a CIELAB value of 18.57, 1.51, 6.18, a medium brown color. In the future, this measurement will be taken for all yarns under study where possible as another nondestructive step in the sequential analysis.

As shown in Figure 2 and described in the Methods, a tiny portion of each color of yarn was first subjected to DART-MS for screening of the dyes present. Negative ion mode is generally more sensitive than positive, and the sensitivity can be further improved through the addition of a microliter or less of 88% formic acid to the sample after the initial exposure to the DART ion source. This is beneficial when indigo dyes are present, as they ionize readily and dominate the DART mass spectrum, while the secondary dye colorants can go undetected; adding the formic acid releases some of the dye from the fibers resulting in more signal for those compounds. Indigotin and indirubin are structural isomers of the molecular formula C₁₆H₁₀N₂O₂ and thus cannot be differentiated in the DART mass spectra, where compounds are identified only based on mass. Thus a peak at m/z 262.086 (the M⁻, likely formed directly through Penning ionization [111]) corresponds to the presence of either or both compounds. This is true of all structural isomers, such as alizarin and xanthopurpurin (both C₁₄H₈O₄), and the numerous flavonoid compounds that make up the colorants in yellow dyes, such as luteolin, kaempferol, and fisetin (all C₁₅H₁₀O₆), among many others. For pure compounds, it is possible to fragment the molecular ion and obtain a CID-like pattern; for complex mixtures; however, the result is a sea of ions that are difficult at best to interpret manually.

Table 2 shows the results from searching the DART spectra for dye compounds characteristic of those previously reported in Peruvian textiles [86,89,91,96]. The search list of compounds and other information is provided in the Supplementary Materials as Table S1. Unsurprisingly, the DART mass spectra of the red and orange yarns as well as many of the purple/black and brown ones showed ions consistent with the anthraquinones characteristic of dyes prepared from the roots of plants in the genus Galium, originally classified as Relbunium. Plant reds prepared from the roots of Galium spp. are well known from previous studies of Paracas textiles. Carminic acid, the diagnostic colorant for the presence of insect red dye prepared from cochineal insects (in the Americas, usually Dactylopius coccus), cannot be detected with DART-MS [86,112], though the addition of formic acid sometimes results in detectable signal for reference dyeing. No traces of cochineal dye were detected, consistent with previous studies of Paracas Necropolis textiles.

The presence of indigo is suggested by the m/z 262.086 peak, consistent with the M- ion of both indigoin and indirubin, as described above, was observed in the blue, black, purple and green yarns, usually as the base peak of the spectrum. This peak was also observed in some of the brown samples (unsurprising, given the color changes described above) as well as in some other colors where indigo was unlikely to have been used in the preparation of the dye, e.g., in red yarns like that from the fringe of Mantle 7. When the samples were collected in the 1980s, several yarns of different colors were often placed together into the same small zip-top storage bag, leading to significant cross-contamination with fiber fragments of different colors adhering to the yarns. This cross-contamination could also have come from the original textile, and inside the tombs among adjacent textiles. Because the indigoids ionize so readily in the DART ion source, even a tiny fiber of an indigo-dyed yarn adhering to the red sample would result in this ion being detected. However, it is also possible that some small amount of indigo is actually present, as has been observed by others [113] (pp. 345–346). Due to the chemical transformations inherent in the preparation of the indigo dye vat, it is generally thought that the original botanical source of the indigoids cannot be determined, though further studies with mass spectrometric methods may suggest the presence of diagnostic molecular species other than the colorants that might differentiate between the possible Peruvian sources that include Indigofera suffruticosa Mill. and Cybistax antisyphyllitica [91,113].

The sensitivity of DART-MS, particularly for indigoids, led us to examine the resulting spectra closely for any evidence of the brominated indigoids characteristic of shellfish purple. Though the signals are low, they do appear to be significant, with peaks apparent at the m/z values for the major isotopes for dibromoindigo (Figure 4). Minute traces of dibromoindigo were observed in the purple-brown fringe from Mantle 7, both the warp and weft of the black ground cloth of Mantle 9, and in the first, but not the second, of two consecutive spectra collected on the loose green yarn from Mantle 10. This latter result is not surprising, as indigo, which is precipitated onto the fibers but is not tightly bound like a mordanted dye, has been observed to vanish from blue fibers while in the DART ion source. In negative ion mode, indigoids tend to ionize directly via Penning ionization rather than through a proton-transfer reaction in the gas phase, resulting in the formation of an M⁻ ion. Though dibromoindigo has only tentatively been identified here in a few samples, this is a tantalizing result.

Identifying the sources of yellow dyes is much more complex than for either red or blue and their secondary colors [89]. There are many possible sources of yellow dyes, and many overlaps in composition, with luteolin being one of the most common flavonoids detected in yellow dyes prepared from plants. Further, yellow dyes, particularly those composed of flavonoids, decompose readily over time due to oxidation. Combining this with the fact described above that structural isomers will give the same molecular ion, DART-MS alone cannot differentiate the yellow dye colorants in the Paracas yarn fragments. Table 2 shows the most abundant ions for yellow dyes in pre-Columbian Andean textiles from previously cited studies, as well as some benzoic acid-based degradation products suggested by Zhang et al. [114] as being characteristic of flavonoids. Ions characteristic of these compounds (see Supplementary Table S1) were observed in nearly all of the yellow and related yarns, though they also appeared in some of the other seemingly unrelated colors. This may be indicative of a degraded flavonoid used in the production of those dyes, or they may arise from some other source. They could also be other unrelated compounds with the same molecular formula.

Based on the DART-MS screening, specific compounds were targeted for extraction with the alternative preparations (DMSO and HCl–methanol) for HPLC-DAD analysis. The spectra also suggested what compounds to specifically look for in the chromatograms, keeping in mind that the results from the yarns directly and those obtained on extracts will likely be different based on the solubility of the compounds in the extraction solutions as well as the differences in sensitivity between the two methods. One difference between DART-MS and the HPLC results would be due to the presence of glycosides in the extracts. Due to strong hydrogen bonding interactions, the hydroxy-rich glycosides are not readily desorbed with heat in the DART ion source, and thus are not usually detectable by this method.

The results from the HPLC-DAD analyses of the dye extracts are summarized in Table 3. Each chromatogram was examined at wavelengths specific to different classes and colors of dye compounds: 450 nm for reds, 350 nm for yellows, and 538/610 nm for the indigoids indigotin and indirubin, and 598 nm for their brominated derivatives that are characteristic of shellfish purple.

Table 2. DART-MS results for dye screening. Compounds are identified based on mass only as [M-H]⁻ ions (or M⁻ for indigotin/indirubin at m/z 262.08) in negative ion mode. Because many flavonoid yellows share the same molecular formula as structural isomers, most of the compounds are unidentified except by the m/z value listed in the table. Compounds/ions are listed in order of abundance. Flavonoid decomposition products include a number of compounds from Zhang et al. [114], identified by mass as [M-H]⁻ ions: HMBA, hydroxymethylbenzoic acid; HBA, hydroxybenzoic acid; DHBA, dihydroxybenzoic acid; DMBA, dimethoxybenzoic acid; DBI, dibromoindigo as M⁻.

Mantle	Color	Part of the Mantle	Anthraquinones	Yellows	Indigoids
4	Blue	Emb. thread	None detected	None detected	Indigotin/indirubin
	Red-brown	Ground cloth	Purpurin, xanthopurpurin, rubiadin, munjistin, lucidin	269, 283 (likely anthraquinones, not yellows), HMBA, HBA	None detected
	Black	Border fringe	Trace of xanthopurpurin	None detected	Indigotin/indirubin
	Green	Border fringe	None detected	285, 269, 299, (likely luteolin, apigenin, chrysoeriol)	Indigotin/indirubin
	Orange	Border fringe	Rubiadin, purpurin	299, HMBA, HBA	None detected
	Gold	Border fringe	None detected	HBA, DHBA, HMBA (with FA, 283, 271, 287, 269, likely apigenin ME, butein, okanin, and apigenin)	Indigotin/indirubin
	Red	Border fringe	Purpurin, xanthopurpurin, munjistin, lucidin	HBA	None detected
5	Purple-black	Emb. thread	None detected	HBA	Indigotin/indirubin
	Red	Emb. thread	Purpurin, munjistin	HBA, DHBA, HMBA, 299 (possibly pseudopurpurin?)	None detected
	Black	Mantle warp	Purpurin	299 (possibly pseudopurpurin?)	Indigotin/indirubin
	Red	Fringe	Purpurin, xanthopurpurin, munjistin	HBA, HMBA	None detected
6	Red	Fringe	Purpurin, rubiadin	HBA	None detected
	Yellow	Ground cloth	None detected	285, 269, 299 (luteolin, apigenin, chrysoeriol)	None detected
	Yellow-brown	Mantle warp	Traces of lucidin, rubiadin, purpurin	285, 269, 299, DHBA, DMBA, HBA (luteolin, apigenin, chrysoeriol)	None detected
	Brown	Mantle weft	Traces of lucidin, rubiadin, purpurin, munjistin	285, 269, 299, DHBA, DMBA, HBA (luteolin, apigenin, chrysoeriol)	Trace indigotin/indirubin
7	Red weft	Mantle weft	Pupurin, rubiadin	HBA	None detected
	Red warp	Mantle warp	Purpurin	None detected	None detected
	Red	Fringe	Traces of xanthopurpurin, purpurin, rubiadin	None detected	Trace indigotin/indirubin
	Green	Fringe	None detected	329 (rhamnazin?)	Indigotin/indirubin
	Purple-brown	Fringe	Traces of purpurin and xanthopurpurin, only with formic acid treatment	DHBA, HBA, 329 (rhamnazin?)	Indigotin/indirubin, DBI?
	Yellow	Fringe	Possible trace xanthopurpurin	DHBA, HBA, 285, 283 (luteolin, apigenin methyl ether/genkwanin)	None detected
9	Black	Mantle warp	None detected	HBA, DHBA, 299 only with FA (chrysoeriol plus other luteolin MEs)	Indigotin/indirubin, DBI?
	Black	Mantle weft	None detected	299, flavonoid degradation products (chrysoeriol plus other luteolin MEs)	Indigotin/indirubin, DBI?
	Green	Border background	None detected	299, DHBA, HBA (as above)	Indigotin/indirubin
	Brown	Emb. thread	Traces of munjistin and rubiadin	285, 299, 315, DHBA, HBA (luteolin, luteolin MEs, rhamnetin?)	None detected
	Red	Emb. thread	Purpurin, munjistin, rubiadin, pseudopurpurin, lucidin	None detected	None detected
	Brown fringe	Fringe	None detected	329, 299, 343, HBA, DHBA, HMBA (rhamnazin, chrysoeriol/luteolin MEs, trimethylquercetin?)	Indigotin/indirubin
	Blue/purple	Thread	None detected	None detected	Indigotin/indirubin
10	purple	Loose threads	None detected	299, HBA (chrysoeriol/luteolin MEs)	Indigotin/indirubin
	green	Loose threads	None detected	285, 269, 299, DHBA, HMBA, HBA (luteolin, apigenin, chrysoeriol)	Indigotin/indirubin, DBI?
	green-brown	Border groundcloth	None detected	285, DHBA, 269, 299, HBA, 329 (chrysoeriol/luteolin MEs, luteolin, apigenin, rhamnazin)	Indigotin/indirubin

Table 2. DART-MS results for dye screening. Compounds are identified based on mass only as [M-H]⁻ ions (or M⁻ for indigotin/indirubin at m/z 262.08) in negative ion mode. Because many flavonoid yellows share the same molecular formula as structural isomers, most of the compounds are unidentified except by the m/z value listed in the table. Compounds/ions are listed in order of abundance. Flavonoid decomposition products include a number of compounds from Zhang et al. [114], identified by mass as [M-H]⁻ ions: HMBA, hydroxymethylbenzoic acid; HBA, hydroxybenzoic acid; DHBA, dihydroxybenzoic acid; DMBA, dimethoxybenzoic acid; DBI, dibromoindigo as M⁻.

Mantle	Color	Part of the Mantle	Anthraquinones	Yellows	Indigoids
4	Blue	Emb. thread	None detected	None detected	Indigotin/indirubin
	Red-brown	Ground cloth	Purpurin, xanthopurpurin, rubiadin, munjistin, lucidin	269, 283 (likely anthraquinones, not yellows), HMBA, HBA	None detected
	Black	Border fringe	Trace of xanthopurpurin	None detected	Indigotin/indirubin
	Green	Border fringe	None detected	285, 269, 299, (likely luteolin, apigenin, chrysoeriol)	Indigotin/indirubin
	Orange	Border fringe	Rubiadin, purpurin	299, HMBA, HBA	None detected
	Gold	Border fringe	None detected	HBA, DHBA, HMBA (with FA, 283, 271, 287, 269, likely apigenin ME, butein, okanin, and apigenin)	Indigotin/indirubin
	Red	Border fringe	Purpurin, xanthopurpurin, munjistin, lucidin	HBA	None detected
5	Purple-black	Emb. thread	None detected	HBA	Indigotin/indirubin
	Red	Emb. thread	Purpurin, munjistin	HBA, DHBA, HMBA, 299 (possibly pseudopurpurin?)	None detected
	Black	Mantle warp	Purpurin	299 (possibly pseudopurpurin?)	Indigotin/indirubin
	Red	Fringe	Purpurin, xanthopurpurin, munjistin	HBA, HMBA	None detected
6	Red	Fringe	Purpurin, rubiadin	HBA	None detected
	Yellow	Ground cloth	None detected	285, 269, 299 (luteolin, apigenin, chrysoeriol)	None detected
	Yellow-brown	Mantle warp	Traces of lucidin, rubiadin, purpurin	285, 269, 299, DHBA, DMBA, HBA (luteolin, apigenin, chrysoeriol)	None detected
	Brown	Mantle weft	Traces of lucidin, rubiadin, purpurin, munjistin	285, 269, 299, DHBA, DMBA, HBA (luteolin, apigenin, chrysoeriol)	Trace indigotin/indirubin
7	Red weft	Mantle weft	Pupurin, rubiadin	HBA	None detected
	Red warp	Mantle warp	Purpurin	None detected	None detected
	Red	Fringe	Traces of xanthopurpurin, purpurin, rubiadin	None detected	Trace indigotin/indirubin
	Green	Fringe	None detected	329 (rhamnazin?)	Indigotin/indirubin
	Purple-brown	Fringe	Traces of purpurin and xanthopurpurin, only with formic acid treatment	DHBA, HBA, 329 (rhamnazin?)	Indigotin/indirubin, DBI?
	Yellow	Fringe	Possible trace xanthopurpurin	DHBA, HBA, 285, 283 (luteolin, apigenin methyl ether/genkwanin)	None detected
9	Black	Mantle warp	None detected	HBA, DHBA, 299 only with FA (chrysoeriol plus other luteolin MEs)	Indigotin/indirubin, DBI?
	Black	Mantle weft	None detected	299, flavonoid degradation products (chrysoeriol plus other luteolin MEs)	Indigotin/indirubin, DBI?
	Green	Border background	None detected	299, DHBA, HBA (as above)	Indigotin/indirubin
	Brown	Emb. thread	Traces of munjistin and rubiadin	285, 299, 315, DHBA, HBA (luteolin, luteolin MEs, rhamnetin?)	None detected
	Red	Emb. thread	Purpurin, munjistin, rubiadin, pseudopurpurin, lucidin	None detected	None detected
	Brown fringe	Fringe	None detected	329, 299, 343, HBA, DHBA, HMBA (rhamnazin, chrysoeriol/luteolin MEs, trimethylquercetin?)	Indigotin/indirubin
	Blue/purple	Thread	None detected	None detected	Indigotin/indirubin
10	purple	Loose threads	None detected	299, HBA (chrysoeriol/luteolin MEs)	Indigotin/indirubin
	green	Loose threads	None detected	285, 269, 299, DHBA, HMBA, HBA (luteolin, apigenin, chrysoeriol)	Indigotin/indirubin, DBI?
	green-brown	Border groundcloth	None detected	285, DHBA, 269, 299, HBA, 329 (chrysoeriol/luteolin MEs, luteolin, apigenin, rhamnazin)	Indigotin/indirubin

Table 3. HPLC-DAD results for the WK12-382 mantle dyes. Dye colorants were identified based on the UV-visible absorbance spectrum and retention time compared to standard compounds run separately. Chromatograms and UV-vis spectra are shown in the Supplementary Materials.

Mantle	Color	Part of the Mantle	Compounds Identified in	Compounds Identified in	Compounds Identified in
			MeOH-OxA	DMSO	HCl-MeOH
4	Blue	Emb. thread	350 nm: caffeoylquinic acid, gallic acid, indirubin	Indigotin, indirubin	Not run
	Brown	Warp	Xanthopurpurin, purpurin	Not run	Not run
	Red-brown	Ground cloth	Xanthopurpurin, purpurin	No compounds observed	Purpurin
	Black	Border fringe	350 nm: caffeoylquinic acids (6), luteolin, luteolin methyl ether, luteolin-like compounds (methyl ethers), indirubin	Indigotin, indirubin, traces of yellow luteolin-like compounds (methyl ethers?)	350 nm: chrysoeriol, luteolin-like compounds (MEs) indirubin
	Green	Border fringe	Indirubin	Indigotin, indirubin	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, genkwanin (apigenin methyl ether), luteolin methyl ether, other flavonoid glycosides and methyl ethers, cafeoylquinic acid, indirubin
	Orange	Border fringe	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, genkwanin (apigenin methyl ether); 450 nm: traces of xanthopurpurin and rubiadin.	Not run	450 nm: Purpurin, rubiadin, xanthopurpurin, indigotin (contamination?); 350 nm: luteolin, xanthopurpurin
	Gold	Border fringe	350 nm: rutin?, lutelin 7-O-glycoside, luteolin; 500 nm: indirubin, traces of indigo	Not run	350 nm: luteolin
	Red	Border fringe	Xanthopurpurin, purpurin	No compounds detected	Xanthopurpurin, purpurin
5	Purple-black	Emb. thread	Indirubin, indigotin	Indirubin, indigotin; 350 nm: luteolin, chrysoeriol, indirubin	350 nm: Indigotin, indirubin
	Red	Emb. thread	Xanthopurpurin, purpurin	Not run	Purpurin
	Black (33)	Mantle warp	Indirubin	Indigotin (tr), indirubin	Indigotin, indirubin, luteolin
	Red	Fringe	No compounds detected	Not run	Purpurin
6	Red	Fringe	No compounds detected	Not run	purpurin
	Yellow	Ground cloth	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, luteolin glycosides, luteolin-like compounds (methyl ethers)	Not run	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, luteolin glycosides (2), luteolin-like compounds (methyl ethers)
	Yellow-brown	Mantle warp	350 nm: luteolin 7-O-glucoside, apigenin glycoside, chrysoeriol glycosides?, luteolin, apigenin, chrysoeriol, rhamnetin; 450 nm: Xanthopurpurin	Not run	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycosides (2), luteolin glycosides (3)
	Brown	Mantle weft	350 nm: luteolin 7-O-glucoside, apigenin glycoside, chrysoeriol glycosides (2)?, luteolin glycoside, luteolin, apigenin, chrysoeriol	350 nm: luteolin 7-O-glucoside, apigenin glycoside, luteolin glycoside, luteolin, apigenin, chrysoeriol	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, ellagic acid, luteolin glycosides (2), luteolin-like
7	Red weft	Mantle weft (39)	Purpurin	Not run	Purpurin, xanthopurpurin, possible luteolin glucoside
	Red warp	Mantle warp (41)	No compounds identified.	Not run	Not run, too little material
	Red	Fringe (43)	Xanthopupurin	Not run	350 nm: Purpurin, xanthopurpurin, ellagic acid, possible luteolin; 450 nm: purpurin, xanthopurpurin, other unidentified anthraquinones
	Green	Fringe	350 nm: quinic acid-like, caffeic acid, rutin, caffeoylquinic acids, quercetin glycoside?, chrysoeriol-like (5), isorhamnetin, indirubin	Indigotin; 350 nm: luteolin?, unknown luteolin-like compounds, indigotin, indirubin	350 nm: luteolin, chrysoeriol, indigotin, indirubin
	Purple	Fringe	350 nm: indirubin; 450 nm: Indirubin, purpurin	350 nm: Indirubin, indigotin, luteolin, luteolin 7-O-glucoside; 598 nm: dibromoindigotin.	350 nm: luteolin-7-O-glucoside, purpurin, indirubin; 450 nm: purpurin, indirubin
	yellow	Fringe	350 nm: Quinic acid-like compound, chrysoeriol glycosides? (2), Luteolin-7-O-glucoside, luteolin	Not run	Luteolin-7-O-glucoside, luteolin, apigenin, chrysoeriol
9	Black	Mantle warp	Indirubin	538/606 nm: indigotin, indirubin, pseudoindirubin; 598 nm: monobromoindigo, dibromoindigo	350 nm: luteolin, indigotin, indirubin, caffeoylquinic acid
	Black	Mantle weft	Indirubin, cafeoylquinic acid-like compound	538/606 nm: indigotin, indirubin, pseudoindirubin; 598 nm: dibromoindigo	350 nm: indigotin, indirubin, no significant yellows
	Green	Border background	350 nm: dihydrobenzoic acid, caffeic acid, cafeoylquinic acid-like compounds, luteolin, flavonoid methyl ethers	indigotin, indirubin	350 nm: caffeic acid derivatives, luteolin, luteolin methyl ether, indigotin, indirubin, many luteolin-like compounds (probably methyl ethers, check MS); 450 nm: indigotin, indirubin
	Brown	Emb. thread	350 nm: luteolin, luteolin 4-methyl ether, other favonoid methyl ethers	indigotin, indirubin	350 nm: quinic acid derivatives, luteolin and other flavonoid methyl ethers; 606 nm: trace indigotin; 538 nm: trace indirubin
	Red	Emb. thread	xanthopurpurin, purpurin	Not run	purpurin, xanthopurpurin, pseudopurpurin?
	Brown fringe	Fringe	350 nm: cafeoylquinic acids (5), luteolin, luteolin-like, rutin-like, flavonoid methyl ethers	indigotin, indirubin, luteolin methyl ethers	350 nm: luteolin, luteolin 4-methyl ether, luteolin methyl ethers, genkwanin, indigotin
	Blue/purple	Thread	350 nm: quinic acid-like compounds (3), rutin-like and luteolin-like late-eluting compounds (methyl ethers), indirubin; 450 nm: pseudoindirubins, indirubin	indigotin, indirubin	Not run
10	Purple	Loose threads	Indirubin, ellagic acid @ 350 nm	350 nm: genkwanin, indirubin, caffeic acid-like, and quinic acid-like compounds; 538 nm: indirubin; 606 nm: trace indigotin; 598 nm: dibromoindigo	450 nm: purpurin, indigotin, indirubin
	Green	Loose threads	350 nm: luteolin 7-O-glucoside, luteolin, indirubin	350 nm: Indigotin, indirubin, luteolin 7-O-glucoside; 598 nm: dibromoindigo	350 nm: luteolin 7-O-glucoside; 450 nm: indigotin, indirubin
	Green-brown	Border ground cloth	No dyes detected	350 nm: Indigotin, indirubin, luteolin 7-O-glucoside	350 nm: quinic acid-like compounds; 538 nm: indirubin; 606 nm: indigotin

Table 3. HPLC-DAD results for the WK12-382 mantle dyes. Dye colorants were identified based on the UV-visible absorbance spectrum and retention time compared to standard compounds run separately. Chromatograms and UV-vis spectra are shown in the Supplementary Materials.

Mantle	Color	Part of the Mantle	Compounds Identified in	Compounds Identified in	Compounds Identified in
			MeOH-OxA	DMSO	HCl-MeOH
4	Blue	Emb. thread	350 nm: caffeoylquinic acid, gallic acid, indirubin	Indigotin, indirubin	Not run
	Brown	Warp	Xanthopurpurin, purpurin	Not run	Not run
	Red-brown	Ground cloth	Xanthopurpurin, purpurin	No compounds observed	Purpurin
	Black	Border fringe	350 nm: caffeoylquinic acids (6), luteolin, luteolin methyl ether, luteolin-like compounds (methyl ethers), indirubin	Indigotin, indirubin, traces of yellow luteolin-like compounds (methyl ethers?)	350 nm: chrysoeriol, luteolin-like compounds (MEs) indirubin
	Green	Border fringe	Indirubin	Indigotin, indirubin	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, genkwanin (apigenin methyl ether), luteolin methyl ether, other flavonoid glycosides and methyl ethers, cafeoylquinic acid, indirubin
	Orange	Border fringe	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, genkwanin (apigenin methyl ether); 450 nm: traces of xanthopurpurin and rubiadin.	Not run	450 nm: Purpurin, rubiadin, xanthopurpurin, indigotin (contamination?); 350 nm: luteolin, xanthopurpurin
	Gold	Border fringe	350 nm: rutin?, lutelin 7-O-glycoside, luteolin; 500 nm: indirubin, traces of indigo	Not run	350 nm: luteolin
	Red	Border fringe	Xanthopurpurin, purpurin	No compounds detected	Xanthopurpurin, purpurin
5	Purple-black	Emb. thread	Indirubin, indigotin	Indirubin, indigotin; 350 nm: luteolin, chrysoeriol, indirubin	350 nm: Indigotin, indirubin
	Red	Emb. thread	Xanthopurpurin, purpurin	Not run	Purpurin
	Black (33)	Mantle warp	Indirubin	Indigotin (tr), indirubin	Indigotin, indirubin, luteolin
	Red	Fringe	No compounds detected	Not run	Purpurin
6	Red	Fringe	No compounds detected	Not run	purpurin
	Yellow	Ground cloth	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, luteolin glycosides, luteolin-like compounds (methyl ethers)	Not run	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, luteolin glycosides (2), luteolin-like compounds (methyl ethers)
	Yellow-brown	Mantle warp	350 nm: luteolin 7-O-glucoside, apigenin glycoside, chrysoeriol glycosides?, luteolin, apigenin, chrysoeriol, rhamnetin; 450 nm: Xanthopurpurin	Not run	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycosides (2), luteolin glycosides (3)
	Brown	Mantle weft	350 nm: luteolin 7-O-glucoside, apigenin glycoside, chrysoeriol glycosides (2)?, luteolin glycoside, luteolin, apigenin, chrysoeriol	350 nm: luteolin 7-O-glucoside, apigenin glycoside, luteolin glycoside, luteolin, apigenin, chrysoeriol	350 nm: Luteolin 7-O-glucoside, luteolin, apigenin, chrysoeriol, apigenin glycoside, ellagic acid, luteolin glycosides (2), luteolin-like
7	Red weft	Mantle weft (39)	Purpurin	Not run	Purpurin, xanthopurpurin, possible luteolin glucoside
	Red warp	Mantle warp (41)	No compounds identified.	Not run	Not run, too little material
	Red	Fringe (43)	Xanthopupurin	Not run	350 nm: Purpurin, xanthopurpurin, ellagic acid, possible luteolin; 450 nm: purpurin, xanthopurpurin, other unidentified anthraquinones
	Green	Fringe	350 nm: quinic acid-like, caffeic acid, rutin, caffeoylquinic acids, quercetin glycoside?, chrysoeriol-like (5), isorhamnetin, indirubin	Indigotin; 350 nm: luteolin?, unknown luteolin-like compounds, indigotin, indirubin	350 nm: luteolin, chrysoeriol, indigotin, indirubin
	Purple	Fringe	350 nm: indirubin; 450 nm: Indirubin, purpurin	350 nm: Indirubin, indigotin, luteolin, luteolin 7-O-glucoside; 598 nm: dibromoindigotin.	350 nm: luteolin-7-O-glucoside, purpurin, indirubin; 450 nm: purpurin, indirubin
	yellow	Fringe	350 nm: Quinic acid-like compound, chrysoeriol glycosides? (2), Luteolin-7-O-glucoside, luteolin	Not run	Luteolin-7-O-glucoside, luteolin, apigenin, chrysoeriol
9	Black	Mantle warp	Indirubin	538/606 nm: indigotin, indirubin, pseudoindirubin; 598 nm: monobromoindigo, dibromoindigo	350 nm: luteolin, indigotin, indirubin, caffeoylquinic acid
	Black	Mantle weft	Indirubin, cafeoylquinic acid-like compound	538/606 nm: indigotin, indirubin, pseudoindirubin; 598 nm: dibromoindigo	350 nm: indigotin, indirubin, no significant yellows
	Green	Border background	350 nm: dihydrobenzoic acid, caffeic acid, cafeoylquinic acid-like compounds, luteolin, flavonoid methyl ethers	indigotin, indirubin	350 nm: caffeic acid derivatives, luteolin, luteolin methyl ether, indigotin, indirubin, many luteolin-like compounds (probably methyl ethers, check MS); 450 nm: indigotin, indirubin
	Brown	Emb. thread	350 nm: luteolin, luteolin 4-methyl ether, other favonoid methyl ethers	indigotin, indirubin	350 nm: quinic acid derivatives, luteolin and other flavonoid methyl ethers; 606 nm: trace indigotin; 538 nm: trace indirubin
	Red	Emb. thread	xanthopurpurin, purpurin	Not run	purpurin, xanthopurpurin, pseudopurpurin?
	Brown fringe	Fringe	350 nm: cafeoylquinic acids (5), luteolin, luteolin-like, rutin-like, flavonoid methyl ethers	indigotin, indirubin, luteolin methyl ethers	350 nm: luteolin, luteolin 4-methyl ether, luteolin methyl ethers, genkwanin, indigotin
	Blue/purple	Thread	350 nm: quinic acid-like compounds (3), rutin-like and luteolin-like late-eluting compounds (methyl ethers), indirubin; 450 nm: pseudoindirubins, indirubin	indigotin, indirubin	Not run
10	Purple	Loose threads	Indirubin, ellagic acid @ 350 nm	350 nm: genkwanin, indirubin, caffeic acid-like, and quinic acid-like compounds; 538 nm: indirubin; 606 nm: trace indigotin; 598 nm: dibromoindigo	450 nm: purpurin, indigotin, indirubin
	Green	Loose threads	350 nm: luteolin 7-O-glucoside, luteolin, indirubin	350 nm: Indigotin, indirubin, luteolin 7-O-glucoside; 598 nm: dibromoindigo	350 nm: luteolin 7-O-glucoside; 450 nm: indigotin, indirubin
	Green-brown	Border ground cloth	No dyes detected	350 nm: Indigotin, indirubin, luteolin 7-O-glucoside	350 nm: quinic acid-like compounds; 538 nm: indirubin; 606 nm: indigotin

HPLC-DAD provided additional clarification to the DART mass spectra for confirming the compounds detected by DART-MS. The yarn samples described as red and red-brown from all of the mantles showed the presence of purpurin, which is consistent with the presence of Relbunium or Galium, as has been the case in nearly all previous studies of Paracas textiles. The predominant colorant was in all cases purpurin, with xanthopurpurin the second most abundant anthraquinone detected, though the extraction procedure affected which colorants were detected. Alizarin was not observed by HPLC, suggesting that the ion at m/z 239.05 in the negative ion DART mass spectra are most likely due to the xanthopurpurin, though alizarin may still be present below the lower limit of detection for the diode array detector. Though the oxalic acid–methanol extraction was chosen to preserve the glycosides, the number and amount of aglycones extracted were so low that no anthraquinone glycosides were detected. With only two compounds reliably detected by the HPLC methods, it was not possible to suggest a specific plant source for the red dyes other than Galium/Relbunium. There were no noted differences between the red dye colorants in the cotton ground cloths, the wool embroidery threads or even in the mixed wool and cotton fibers from Mantle 7. If found, such differences might have suggested that the mantle fabric and decoration occurred in different locations or were prepared by different communities of practice.

Though only two yarn fragments, from Mantles 4 and 9, were classified as “blue”, indigotin and indirubin were identified in those samples as well as in the purple and black yarns from the other mantles. Because traces of the brominated indigoids were detected in the DART mass spectra for some of the black yarns, the chromatograms at 598 nm were also examined for signals at the appropriate retention times for these compounds. Only the warp from Mantle 9 showed both mono- and dibromoindigotin, while the weft from the same mantle showed only a trace of the dibrominated species. Dibromoindigotin was further detected in the purple–brown fringe from Mantle 7 and the green threads from Mantle 10, as seen in the DART-MS, but also in the purple threads from Mantle 10. Chromatograms and UV-visible spectra for these brominated species are provided in the Supplementary Materials.

In addition to indigoids, compounds characteristic of tannin dyes (caffeic, caffeoylquinic, feruoylquinic, gallic, and ellagic acids) were also detected in some of the blue/purple/black yarns (see Table 3). Other compounds characteristic of yellow dyes and with UV-vis spectra resembling that of luteolin were also observed in these colors; these will be discussed in detail with the rest of the yellow dyes. Purpurin and other Relbunium/Galium anthraquinones were present in some of the yarns that appeared purple as well, indicating a purple likely prepared from overdyeing.

Yellow dyes are the most complex, and thus are the most difficult to identify due to the vast number of possible plant sources, the overlaps in composition between many of those plants, and how the dye colorants prepared from a single plant source can vary due to location, harvesting time, preparation of the dye, degradation, etc. Boucherie has described in detail the composition of the myriad sources of yellow dyes available to the ancient dyers of southern Peru [113]. In the outer mantles of WK12-382, only two different yellow compositions are apparent from the HPLC results, which for simplicity are referred to here as Yellow 1 and Yellow 2.

Yellow 1 contains luteolin, apigenin, and chrysoeriol in ratios consistent with the composition of Salix humboldtiana Willd. (Figure 5a) described by Boucherie et al. [102]. This Yellow 1 composition was observed in the orange and gold yarns from Mantle 4, both the warp and weft of the ground cloth yarns from Mantle 6, as well as in Mantle 7 (the yellow yarn only) and perhaps in the Mantle 10 border ground cloth, though the yarn was quite degraded. Yellow 1 was present alone or with Relbunium red to yield orange, but was not unequivically identified in the presence of indigo. Though these results would suggest that Yellow 1 is derived from Salix humboldtiana Willd., there were differences that need to be considered in the identification of the components of this dye. The large peak identified by Boucherie et al. as luteolin in the reference dye prepared from Salix humboldtiana Willd. (Figure 5a) appears instead to be the 7-O-glucoside of luteolin, with a retention time that matched that of a standard of that compound. Luteolin aglycone elutes later than its glycosides, as the polar sugar(s) cause(s) the glycosides to interact less with the nonpolar C18 stationary phase. This was confirmed by spiking the extract from one of the Mantle 6 samples with luteolin standard (Figure 5b); the later-eluting peak increased in area, whereas no change (except that due to dilution) was observed in the earlier-eluting peak. The luteolin 7-O-glucoside did not hydrolyze in the oxalic acid–methanol extractant, nor did it completely hydrolyze in the HCl–methanol extraction solution, as shown in the ESI spectra of solutions of the glucoside standard (Figure 5c). This is consistent with the results of Sabatini et al. [91], where HPLC-DAD and HPLC-ESI-QToF-MS was used in the study of Paracas dyes. One of the major disadvantages of HPLC-DAD is that the chromophores are characteristic of the flavone backbone, meaning that the glycosides and aglycones of luteolin, for example, both have effectively the same UV-visible spectrum, leading to the difficulty in identifying the specific compounds present in yellow dyes. How this impacts the interpretation of the analytical results will be addressed further in the Section 4.

Yellow 2 is distinctly different from Yellow 1 in that little or none of the luteolin glucoside is present in the chromatogram. This dye was observed in all cases alongside indigo to yield green, black, and what today appears brown in Mantle 4, Mantle 9 and the green yarn from Mantle 7. The purple embroidery yarn and blue/black ground cloth from Mantle 5 showed some evidence of flavonoids, but not in sufficient quantity to suggest which of the two yellows might be present. Yellow 2 is further marked by the presence of a number of tannin-like compounds such as caffeoylquinic acid derivatives, along with gallic and ellagic acids in some cases, as well as a series of strong peaks at retention times significantly longer than the flavonoid aglycones that also showed UV-visible spectra characteristic of luteolin, chrysoeriol and apigenin, in particular. These compounds appear to be methyl ethers of these flavonoids based on their later retention times relative to the aglycones.

Boucherie [113], working with indigenous people in Peru to identify and select native dye plants from the Andes, prepared reference samples from several of the more than 70 different species of Baccharis found in Peru. Of these, she described in detail the composition of five species: Baccharis salicifolia (R. & P.) Pers. (which has been reclassified from B. lanceolata (L.) Kunth); Baccharis genistelloïdes (Lam.) Pers.; Baccharis odorata Kunth.; Baccharis tricuneata (L.f.) Pers.; and Baccharis latifolia (R. & P.) Pers. All of these Baccharis dyes also had similar flavonoids, including luteolin, quercetin, apigenin, chrysoeriol, kaempferol, rhamnetin, and genkwanin. Another study of reference dyes prepared from Peruvian plant dyes in the 1970s by Kay Antunez de Mayolo for Max Saltzman [92]) included a comprehensive analysis of Baccharis floribunda. This dye was characterized by the presence of a number of caffeoylquinic acids (and feruoylquinic acids identified with tandem mass spectrometry), flavonoids including luteolin and quercetin as well as their glycosides and methyl ethers.

Although our laboratory did not have access to more than one reference dyeing of this genus (in this case, only the B. floribunda both with and without alum mordant), botanical material was provided by Kay Antúnez de Mayolo from B. genistelloides (Lam.) Pers. and B. lanceolata Kunth (now reclassified as B. salicifolia) that were collected in 1975 during her work with Max Saltzman [115]. Antunez de Mayolo refers to the dye from B. lanceolata as chilca, while B. genistelloides was identified as espadilla or kuchu kuchu; these are generally consistent with the Quechua designations for dyes prepared from these species by Boucherie [115] (pp. 591–592). Antúnez de Mayolo collected two samples of B. genistelloides, one from Huancapi in the Department of Ayacucho, and another from the Q’ero village of Chuwa Chuwa. Using a few leaf fragments from these botanical specimens, small (~10 mL) dye baths were prepared, with the addition of a few milligrams of sodium carbonate, by simmering the material in water for about an hour. Dyeing samples were prepared from alpaca wool yarn both with and without alum mordant from these three botanical specimens. Portions of these reference dyes were extracted using the oxalic acid:methanol process described in the Methods and separated under the same conditions by HPLC-DAD. Figure 6 shows the result obtained on the brown fringe from Mantle 9 compared to the reference dyes also prepared on alum mordanted wools with the B. genistelloides plant material from Q’ero and B. floribunda. The chromatogram for the dye prepared from B. genistelloides seems consistent with the results obtained by Wouters and Rosario-Chirinos, with the “luteolin-like” compounds eluting later than luteolin being what are likely to be the corresponding methyl ethers.

It appears that our Yellow 1 designation aligns with the LUTE group 1 of Wouters and Rosario-Chirinos [101] and Salix humboldtiana Willd., also designated LUTE-1 by Boucherie [102,113]. Yellow 2 is generally consistent with the LUTX and Baccharis genistelloides or a mixture of Baccharis and some other yellow. Other sources of flavonoids, and tannins also abound in the south-central Andes, such as the aforementioned Neltuma pallida (huarango) [116]. The difficulties associated with identifying yellow dyes will be addressed further in the Discussion, as stated above.

The HPLC-DAD results for Yellows 1 and 2 are useful for interpreting the DART-MS spectra, suggesting what compound or compounds correlate with each of the m/z values observed as reported in Table 2. For Yellow 1, the comparison is fairly straightforward, though the glycosides present in the chromatogram (Figure 7) were not detected in the DART spectrum, which here shows only peaks at m/z values for the aglycones luteolin, apigenin, and chrysoeriol at the m/z values of the characteristic [M-H] anions in negative ion mode. Yellow 2 is more complex due to the many possible isomeric methyl ethers observed in the chromatograms, as shown in Figure 6 top. A peak in the DART mass spectrum at an m/z of 299.06 could correspond to a number of different isomers, including chrysoeriol, which is a methoxyapigenin itself, as well as rhamnocitrin or kaempferide, both methyl ethers of kaempferol. Though there may be co-elution of some isomeric species in the HPLC, there will more likely be several peaks in the chromatogram that correspond to a single m/z in the DART mass spectrum.

One significant discrepancy between the HPLC and the DART-MS was found for the gold embroidery yarn in Mantle 4. Though the HPLC results were generally consistent with that of Yellow 1, significant, albeit low, signal was observed in the negative ion DART at m/z values of 271.07 and 287.06 which could correspond to butein and okanin. These compounds are chalcones, previously identified in a reference dye prepared from Bidens andicola as well as in a Paracas yarn also described as gold [89]. The best approach to evaluate if either or both dyes are present in this sample would be to combine HPLC and tandem mass spectrometry for separation and fragmentation of each of the colorant molecules, providing a comprehensive analysis of the dye. Indeed, this approach is widely utilized in other laboratories, and should be available at EMU in the future.

Until then, DART-MS may have some advantages over HPLC for the classification of dyes rather than their comprehensive identification. Using the information from this pilot study a multivariate approach to use DART spectra collected under the same conditions may be developed to determine if a new yarn sample falls within a group defined by the HPLC-DAD and DART-MS results as well as those from previous studies using modern methodologies. This will save time while still permitting comparisons between possibly related objects and dyes.

3.2. Plasma Oxidation and AMS Radiocarbon Dating

Table 4. Plasma treatments and yields for the yarn samples from Mantles 4-10.

Mantle	Sample	Oxygen Plasma Conditions	Mass, µg C	EMU Sample ID	Sample Consumed?
4	Red ground cloth fragment	50 W, 7 min	30	3P187	Yes
5	Red embroidery yarn	40 W, 15 min	10	3P193	No
6	Yellow-orange warp	40 W, 20 min	67	3P188	Yes
7	Red mantle weft	40 W, 15 min	8	3P189	Yes
7b	Green emb. thread	50 W, 18 min	160	3P207	No
9	Brown fringe	50 W, 15 min	46	3P191	Mostly
10g	Green-brown border ground cloth	60 W, 20 min	21	3P195	Mostly
10b	Green-brown border ground cloth	30 W, 12 min	10	3P186	Mostly

shows the plasma treatments and yields for the WK12-382 yarn samples. All of the samples used for dating were first extracted for dye analysis by HPLC. The amount of carbon produced from the oxygen plasma ranged from a high of 160 µg of carbon as CO₂ to a low of just 8 µg. The average amount of carbon yielded was 44 µg of carbon as CO₂, what is normally considered a large enough amount for a reliable AMS radiocarbon date. Whether or not the sample was consumed was in part related the applied RF power and the time of exposure to the reactive plasma. Figure 8 shows the before and after plasma images for three of the samples, showing the varying effects of the plasma process. The samples used in this pilot study were at the limit of what should be possible to date. Clearly this demonstrates that the time and RF power must be optimized further to preserve the samples for additional sequential studies. Though this would require using more sample initially, it appears that the benefit of completing all of the analyses on the same material outweighs the use of a larger yarn fragment. Considering that most dating studies of textiles utilize much larger samples (see [93]), this may not be a significant problem moving forward, assuming sufficient sample is available initially. This is particularly important when considering that the combustion method consumes the entire sample of yarn already. The potential for PCO as one step in a sequential analysis and dating process is clear, though much additional work is needed to optimize all of the parameters to preserve enough sample for further studies.

The CO₂ from the plasma process was graphitized and underwent radiocarbon analysis at the Keck Center for Accelerator Mass Spectrometry as described in Section 2.4, the results of which are shown in

Table 5. Radiocarbon dating results for Mantles 4–10 from WK12-382.

UCIAMS#	EMU Sample ID	Mantle	Mass, µg C	Radiocarbon Years Before Present, Uncalibrated	Calibrated Age Range
297733	3P187	4	30	2085 ± 45	343 BCE–21 CE
297739	3P193	5	10	2090 ± 140	416 BCE–247 CE
297734	3P188	6	67	2070 ± 25	167 BCE–7 CE
297735	3P189	7	8	4420 ± 300	Not calibrated
303849	3P207	7b	160	2140 ± 15	344-61 BCE
297737	3P191	9	46	2055 ± 35	168 BCE–55 CE
297732	3P186	10b	10	2130 ± 140	515 BCE–219 CE
297741	3P195	10g	21	2030 ± 70	343 BCE–201 CE

. Samples with low carbon yields from the plasma-chemical oxidation, from Mantles 5, 7, and 10, resulted in radiocarbon ages with large measurement uncertainties. Further, the first sample from the weft of Mantle 7, was both extremely small at just 8 micrograms and much older than all the others. Another yarn fragment was prepared (8b) and dated, resulting in a much larger carbon yield and a date consistent with the others. The suspicious date appears to be the result of plasticizer contamination from the plastic bags in which the samples have been stored for more than 40 years having a marked influence on such a small sample.

The measured ages, expressed in radiocarbon years before present, were calibrated using OxCal v.4.4.4 revision 5 online [117] with the most recent Southern Hemisphere radiocarbon calibration curve (SHCal20). Replacing the date presumed to be contaminated with the new sample from Mantle 7, the rest of the ages from the mantles were combined in OxCal to yield a date range of 173-53 BCE at a 95.4% confidence level. Recently published radiocarbon ages for the Paracas Necropolis using Bayesian modeling [118] has the highest probability for this cemetery and artifact complex spanning a range of 360-190 BCE/120-300 CE. Indeed, this range exactly corresponds to the highest probability for Necropolis at Paracas and Ica [119] (Figure 3). The measured ages, but especially the combined age for the six mantles, are all consistent with this range. Figure 9 shows the plotted calibrated ranges for the reliable dates obtained for the yarns from the six mantles.

3.3. Light Stable Isotope Results

Due to the poor preservation of the samples in the PCO process through this pilot study, additional fragments of yarn were selected for measurement of the light stable isotopes. This further permitted study of both wool and cotton yarns (and in the case of Mantle 7, yarn made from blended animal and plant fibers) from different parts of the mantles. Supplementary Table S2. shows the results of the stable isotope analyses. Some of the samples were large enough to be measured in multiple aliquots on the elemental analyzer.

The results from the stable isotope analysis (SIA) are preliminary at this stage, as there is not sufficient carbon and nitrogen isotope data available for the south coast of Peru to permit the types of comparisons used on the north coast to establish or at least suggest an origin for the textile yarns. Figure 10 shows the results for the wool mantle yarns for which stable isotope results were obtained. The raw data is provided in Supplemental Table S2. When compared to modern camelid wools from Samaca (corrected by 1.5‰ for the Suess effect) and surface finds of llama pelts from disturbed archeological contexts in the same region, the three yarns from Mantle 4 (ground cloth, embroidery thread and border fringe) seem to be generally consistent with wool from the lower Ica valley. The others are significantly more enriched in 13C, as well as in 15N. This could be related to the use of maize as animal fodder (affecting the carbon isotopes) or the use of guano fertilizers (affecting nitrogen), both of which have been identified in pre-Hispanic Andean archeology [121,122,123]. However if the results are compared to those obtained on Virú textiles from approximately the same time period (also Early Intermediate Period, [100]), but from the North Coast, the spread of isotopic values for the six mantles is generally consistent, with only some of the samples from Mantle 9 (including the brown fringe and both warp and weft from the dark blue-black ground cloth) and the single wool yarn from the red fringe of Mantle 6 differing significantly.

Only three cotton yarns were available for SIA, all from the ground cloths of Mantles 4, 5, and 6. Comparing the measured δ¹³C values for the mantle yarns with modern cotton from Samaca (corrected for the Suess effect) as shown in Figure 11 indicates that only Mantle 5 seems more depleted in ¹³C relative to the other yarns, As with the wool yarns from Mantle 4, the carbon isotope ratio was generally in the range of the corrected modern cotton. This is somewhat different from what was observed for the wool yarns in Mantle 6, which were much more positive than their modern comparative materials.

4. Discussion

Table 6. Summary of dyes identified in the six mantles from WK12-382.

Object	Style	Source of Yarns	Dyes Identified
382-4	Linear	Brown-red ground cloth	Relbunium
		Blue embroidery thread	Indigo + tannin?
		Border fringe and weft yarns in red, green, orange, gold, and black	Red: Relbunium; Green and black: indigo + yellow 2; Orange: Relbunium + yellow 1; Gold: Yellow 1
382-5	Linear	Black yarns from ground cloth	Indigo + yellow trace?
		Black and red embroidery threads	Black: Indigo + yellow 1; Red: Relbunium
		Red fringe	Relbunium
382-6	Linear	Yellow-orange ground cloth	Yellow 1
		Red fringe	Relbunium
382-7	Block	Red-orange yarns from ground cloth	Relbunium
		Fringe yarns in purple-brown, green, red, and yellow	Red: Relbunium; Green: indigo + yellow 2; Purple: indigo + relbunium + possible yellow 1? + shellfish purple; Yellow: Yellow 1
382-9	Block	Black yarns from ground cloth	Indigo + possible tannin or yellow 2 + shellfish purple
		Green yarn from border background	Indigo + yellow 2
		Embroidery yarns in orange/red and brown	Red: Relbunium; Brown: Yellow 2 + indigo
		Brown fringe yarns	Yellow 2 + indigo
		Blue loose yarn	Indigo + yellow 2
382-10	Block	Green-brown disintegrating fibers from border ground cloth	Indigo + possible yellow?
		Loose threads in purple and green	Purple: Indigo + tannin? + Relbunium; Green: Indigo + yellow 2; both contain shellfish purple

summarizes the dye compositions obtained from the DART-MS and HPLC-DAD analysis of the samples from the six mantles studied. Dye practices employing various species of Indigofera to produce blue and related hues of green, purple, and black and the use of Galium species roots to obtain hues of red are well established for Paracas as described in the Results. The plant-derived indigoids are in some cases complemented by the presence of dibromoindigotin (and in one case, by monobromoindigotin); these brominated species are considered markers for the use of shellfish purple. Dyeing yarns or textiles purple can be accomplished in multiple ways, either through overdyeing indigo blue on top of red or perhaps by using a mordant like a copper salt to shift the color to reflect at longer wavelengths. Purple derived from mollusks, well known from the ancient Near East in the Old World, is also produced from a small number of shellfish species found on the Pacific Coast of South America. Recent work by Cárcamo-Vega et al. [124] have tentatively identified this dye in textiles from the coastal region of Chile from the Late Intermediate Period (1100–1450 CE). A previous study of a wide range of Andean textiles [89] found no evidence for the use of shellfish purples in the form of the brominated indigoids, though that may be a reflection of the small samples used in that study. This is consistent with previous analyses by Wallert and Boytner [125] of textiles from the same southern coastal region of Peru as the Paracas materials, but from later time periods and cultures, as did Michel et al. [126]. Using solution spectrophotometry, Saltzman [35] identified shellfish purple in a South Coast textile, though the sample was described only as “purple painted on white cotton–Ocucaje, Peru ca. 100 BC.” King [127] describes four Paracas tradition textiles from Ocucaje painted with a hand and dot design in a reddish purple on white or natural cotton. Saltzman et al. [128] tested this paint with their spectrophotometric method, concluding it is indeed shellfish purple. This reddish-purple “paint” was prepared from the local Muricidae mollusk, Concholepus concholepus, widely exploited for food in the past, based on the frequent appearance of their shells throughout the region. Though shellfish purple was used as a paint, it does not appear to have been reported as a dye during that time period. Boucherie [115] found no evidence of dibromoindigo in the samples studied from Wari Kayan (including several from bundle 451).

Further, our results appear to be consistent with those of Wouters and Rosario-Chirinos and Boucherie [101,102] that two different yellows are in use during the Paracas Necropolis period, both containing luteolin and its various derivatives. The LUTX composition is consistent with our Yellow 2, which also is only found in conjunction with indigoids, as has also been reported by those authors previously. It should be noted that the composition of the dye in the extractant solution is not static, as described already for the case of indigoids extracted into DMSO. Though the oxalic acid:methanol and HCl–methanol extractants do not significantly hydrolyze glycosides in the short term, if the samples are left to sit in those solutions, they can undergo significant changes. Here there were two yellow dyes observed in the six Paracas mantles, which appear to correspond roughly to those prepared from Salix humboldtiana Willd. and Baccharis genistelloides (Lam.) Pers. The Salix dye is marked by the presence of both luteolin and luteolin 7-O-glucoside (as determined by retention time of a standard), whereas the Baccharis dye lacks the glucoside and is instead marked by the flavonoid methyl ethers at longer retention times. Figure 12 shows the chromatograms obtained from a fresh extract of the yellow dye from the ground cloth of Mantle 6 in the 30:1 methanol–HCl extractant (12a) along with a sample of the same composition that was left on the lab bench for an extended time (12b). The glucoside that suggested the dye in this sample was our Yellow 1 hydrolyzed over time to the aglycone, which was then methylated due to the presence of both methanol and acid to yield the ethers characteristic of Yellow 2. This complicating factor can be avoided by preparing samples immediately prior to HPLC analysis, or perhaps by storing the extracts at low temperatures. It is, however, something to keep in mind when interpreting the results from the HPLC-DAD chromatograms. This observation also suggests that the chemical environment of the indigo vat (acidic during fermentation and basic for dyeing) may influence the observed composition of the yellow dye as well. Perhaps the yellow dye in any given green or brown yarn could undergo acid or base hydrolysis as described above when placed into the indigo vat. This would be an issue if the yarns were dyed initially with yellow and then overdyed with indigo. Such possibilities make the already difficult task of identifying yellow dyes yet more complex.

The results from this pilot study illustrate that it is possible—though not always ideal—to use a single small sample for multiple sequential analyses and dating. Further, the difficulties in specifically identifying the source of the dyes cannot be overstated. Other aspects of the production of the textiles besides dye composition, age, and environment (related to diet in the case of camelids) have not been addressed in this pilot sequential study. This includes the use of mordants, which can affect the composition of the dye and influence what is extracted. Though XRF [91] and SEM-EDS [113,129] have been used to identify metals present in Andean textiles with the assumption that these metals were used as mordants, a major complicating factor is contamination from handling and the burial environment. Aluminum is a light element to which XRF is not especially sensitive; both Al and Fe are widely used mordants that are also significantly present in soils. Just because a metal is detected in a particular yarn fragment does not necessarily mean that it was used as a mordant in the preparation of the dye. XRF line and area scans have been valuable in identifying pigments in painted artworks [130]; applying this approach to entire textile fragments to correlate color and metal composition has great potential for distinguishing between deliberate use of mordants compared with inadvertent soil contamination. Unfortunately this is not generally applicable to small samples like the yarns in this study, though Sabatini et al.’s use of XRF to yarn fragments from Paracas textiles did suggest possible mordanting metals.

The chemistry of the dyes is complex enough, even ignoring the influence of mordants, dyebath pH, or dyeing practice as outlined above. Even in a case like that of Nathalie Boucherie’s remarkable and detailed dissertation work, where all due care was taken to select dye plants with the aid of indigenous knowledge and the preparation of the dye references followed locally informed practices, there are too many variables to account for in comparing ancient materials to ones prepared today. The variations in composition observed in the Baccharis dyes are consistent with Boucherie’s observations ([115], p. 601, translated from French):

“Consequently, several factors can modify the dye composition of a botanical species: the age of the plant, its growth phase, the natural environment in which it grows (soil type, altitude), the season (climate), its condition (infestation [as with fungi in the case of Baccharis]) and preparation, the dyeing process, etc. It is therefore difficult to distinguish in a single analysis all the dye molecules that a single species can produce. If one or more of these criteria is/are not identical to those that prevailed for the [ancient] dyers, it is possible that the chromatographic results obtained on our dye samples do not correspond, or only imperfectly, to the results from the archeological fabrics.”

Biological systems—in this case, plants used in the production of dyes—are complex and affected by myriad variables. Adding in the actions of humans and the passage of 20 centuries, attempting to identify to the species what specific plant was used in the past may be an impossible goal. However, as with the mordant situation described above, the overall patterns of usage may provide insight into the dye practices and human agency. An example in this study is that the shellfish-derived purple brominated indigoids were only detected in Mantles 7, 9, and 10, suggesting that perhaps this dye recipe is unique to the Block Color style. This could also be a reflection of the colors Dr. Paul collected, as black, purple and green are not well represented in the samples from the Linear Style mantles. Replication studies and sharing of reference collections to be characterized by any and all analytical methodologies should be encouraged and maybe even facilitated by organizations like Dyes in History and Archaeology through workshops or repositories of materials. After all, it is an archeological question to be answered through the use of analytical chemistry, not a chemical answer.

The incorporation of radiocarbon dating into this sequential analysis has provided a more nuanced understanding of the relationships between sample size, precision of the dating method, and its impact on other compositional analysis goals. Because good dates were produced with smaller samples than has been possible with other methods, these results can be used to develop guidelines for future yarn sampling for sequential analysis. Because the dates obtained here can be compared with other recent AMS dating of Paracas Necropolis contexts and a Bayesian analysis of a broader comparative sample [118], our results both support and are supported by that prior analysis. As a result, this method can be employed with confidence to obtain dates for other textiles whose contexts and styles have not been well defined in time.

The preliminary results of the light stable isotope analysis are encouraging, but highlight the importance of obtaining and correcting other samples from animals and plants living in documented locations of the south-central Peruvian Andes. Ideally, the sourcing of both cotton and camelid hair used to create textiles placed in the Paracas Necropolis tombs can be determined. This sourcing would have been based on ongoing relationships between herding and agricultural communities and the textile producers, as well as the relationship between those producers and the placement of these garments, often newly made, in a particular tomb. Much work remains to be performed to delineate, on the one hand, the relationship between isotope signatures and the coastal and highland landscapes and, on the other hand, the patterns of consistency and variation among textiles in the same or different tombs.

The results of this preliminary analysis provide interesting patterns to be tested in further analysis. The outermost mantle, Specimen 4, provides highly consistent results among yarns differing in color and raw material. The similarity among both the cotton and camelid hair isotopic ratios and the corrected modern and archeological samples obtained in the lower Ica valley suggests that the component materials in this large mantle were all sourced in this region or a very similar landscape, such as the lower branches of the Nasca watershed immediately to the south. The samples from mantles 5, 6, 9 and 10 provide a modest range of variation that highlights the consistency of Mantle 4. In contrast, mantle 7 provides camelid hair yarns whose isotope ratios contrast as a group with the other textiles and also are more internally diverse, suggesting that the animals whose hair was spun and dyed for this mantle had lived in several different environments, and at higher altitudes, although the mixing of cellulose and animal fibers in some of these samples complicates the picture as well. Current work with indigenous llama and alpaca herders to sample hair from animals living in the Andean highlands to the east will better inform these comparisons. The analysis of yarns from garments that create a matching set with several of these mantles will also provide a broader comparative sample.

5. Conclusions

In summary, the pilot study was successful in that it provided the desired information about the six Paracas mantles and that the amount of sample needed to measure all the parameters was delineated. In the future, this work will expand to use the existing data and look for correlations between similar textiles within and among the textiles from secure archeological contexts like those from Wari Kayan. This information may then aid in the contextualization of the many fragmentary Andean textiles for which no context remains that reside in museum collections around the world.

The sequential analysis has already provided data pertinent to our interest in assessing consistent and variable textile production practices among objects placed in the Wari Kayan tombs at the Paracas site. AMS dating confirms the general time frame for the Necrópolis of Wari Kayan, and particularly a tomb contemporary with Nasca 1 styles, classified as Early Intermediate Period phase 1. No significant difference in radiocarbon age was detected among the textiles sampled. However, the fibers that compose these textiles have been sourced in different places of the Pacific watersheds and Andean highlands. At least one mantle indicates that both cotton cultivation and camelid herding occurred in regions of the south coast similar to the lower Ica valley, while another suggests camelid herding in diverse highland regions. This work has provided more detailed documentation of the chemical signatures associated with dye practices similar to those documented by other researchers for Paracas Necropolis and early Nasca, as well as note the potential effects of degradation processes and analytical techniques. The presence of at least two different sources of yellow dyes and their tendency to correlate with differences in design principles and embroidery procedures suggest production processes that integrate the spinning and dyeing of fine yarns with their subsequent use to create elaborate embroidered imagery. The layering of these mantles in a single funerary bundle indicates the contemporary contributions of different textile production communities in a sequence of funerary rituals carried out to honor the man at its core.