*2.1. Non-Destructive Extraction Methodologies*

The vessels from both Mochlos and Tourloti were sampled according to ARCHEM extraction protocols [4–6], which were developed utilizing non-destructive precedents [10,11] with field extractions in mind and subsequently refined through years of continuous field-testing [13]. While they share the same fundamental approach, the protocol used for Tourloti [1], being more recent, reflects the latest methodology. This is the product of an additional decade and a half of experimental refinement, which has shown that attention to the preservation, history, and therein potential contamination of an object is essential to the acquisition of qualitatively and quantitatively significant samples. With proper consideration of these features (i.e., a pre-experimental "diagnosis"), sampling strategies can be adapted to improve the collection and contamination and degradation risks likewise factored into interpretive frameworks. Object biography—including excavation conditions in the field – thus was and remains one of the key considerations in our methodology.

The ARCHEM extraction protocols are of two types: "swishing" and "boiling", both of which are non-destructive processes typically utilizing heated absolute, analytical-grade ethanol. Given that some mineral tempers in clay (e.g., calcites) are lipophilic [14], preferentially adsorbing certain fatty organic residues, the introduction of heated solvents induces dissolution of organics from the ceramic matrix. Ethanol is the ideal ORA solvent for field sampling, as it is both the least toxic of all alcohols, mitigating the need for a fume hood, and readily procurable worldwide, even in remote locations. In the early years, ARCHEM employed a two-step extraction process utilizing non-polar and polar solvents in rapid succession (e.g., ethyl acetate and methanol) with the goal of minimizing compounds to separate in each chromatogram [4,10]. However, over a decade of research in the field has empirically demonstrated that the downsides of a two-solvent extraction process in the field (e.g., working with highly toxic solvents) typically outweigh any marginal benefits that the more unwieldy solvent systems might provide in certain limited circumstances with ancient organic residues [5,6]. Beyond its ready availability and low toxicity, ethanol's solvent properties obviate the need for a two-step extraction process [5], as its hydroxyl group attracts polar and ionic molecules, while its ethyl group attracts non-polar ones. These qualities facilitate the rapid and effective extraction of a broad range of organic residues while leaving the artifact undamaged, making it ideally suited to field extraction. On-site extraction not only allows the ORA specialist to adapt extraction methods

to the immediate archaeological environment and take regular soil samples as future controls, but also results in a comprehensive library of samples with detailed provenance. This is information that can offer uncommon insight into bigger issues such as the spatial distribution of a particular ancient organic commodity, or help to identify large scale contamination events across a site [13], which are almost always anthropogenically induced.

Boiling extractions are best suited for vessel fragments that have a clear chain of custody. This includes those that have a clean excavation and sampling history (e.g., handled with gloves, isolated in aluminum foil, unwashed) and subsequently have a low probability of contamination on any surface. The isolated fragments are placed in suitably sized Griffin beakers and are immersed in heated ethanol for at least several minutes. The type of pottery fragment makes a difference: For boiling extractions, vessel fragments are optimally chosen from the base, followed by the rim, especially for cookpots [15] or pouring vessels, and then the body, in that order. This is predicated on the assumption that the interiors of bases have the greatest exposure to organics, which pool at the bottom of a vessel or its lowest point at deposition, while certain diagnostic compounds float to the top during its history of use. Care must be taken to ensure that samples extracted from multiple fragments originate from a single object (i.e., that they all join). In cases of uncertainty, or where there are constraints upon resources (time or materials), which necessitate more limited sampling, then a single fragment chosen from the base at a location where it joins the body is usually preferable.

Swishing extractions are better suited for complete and near-complete vessels, or vessel fragments such as large bases, which, given their location on the vessel (as noted above), have typically had more prolonged exposure to organic contents, and therein greater opportunity for extant residues to permeate the matrix of the clay fabric. For the swishing of large vessel fragments, solvents are heated in a beaker as usual (to just below boiling), but they are then poured in small volumes onto the fragment and gently swished and/or stirred for approximately one minute on an interior surface, where the original organic contents would have come into prolonged contact for adsorption. For swishing of largely intact vessels, hot solvent is directly pipetted or poured into vessels, and is immediately swished and/or stirred around the object for up to a minute, and subsequently filtered directly into 20 mL master scintillation vials. Multiple swish extractions are typically executed.

Here again, object biography makes a difference. For objects with a clear chain of custody, and for which there are no indications of contaminants, all swished extractions can be collected into a single master sample prior to instrumentation. However, as our earlier work has indicated, the longer an object has been unearthed, the more likely the first extraction—those taken from the exposed surface—will include organics suffering from some level of contamination and degradation. Therefore, for those objects whose condition, or biography, is less certain, the swishing process is repeated and each resulting solution sample isolated, in order to moderate the removal of potential contaminants in early stages and facilitate compound separation and identification in multiple chromatograms. This can be a successful method to filter out plasticizers such as phthalates, which tend to appear in the first extraction and diminish or disappear in subsequent ones, as the results from plastic-packaged vessels from other areas of Mochlos and other sites have shown [4].

As freshly excavated artifacts, the Mochlos vessels—which included both complete and fragmentary vessels—were considered at lower risk for environmental degradation and anthropogenic contamination. Four samples were taken from the Mochlos vat and two from the nearby undecorated amphora, for a total of six samples from Mochlos and three from Tourloti (Table 1). Being smaller and more fragmentary, sherds from the vat's rhyton-spout were boiled using the two extraction solvents in sequence, ethyl acetate first followed by ethanol, which produced two samples (Figure 3). The large pieces of the main body of the vat, being too large for beakers, were swished using the two extraction solvents in sequence, which produced two more samples (Figure 4). The undecorated amphora (P6313A) was fragmented, but as the full profile was preserved and the sherds from the base, body, and rim were clearly from the same vessel, these sherds were boiled together first in ethyl acetate followed by ethanol to produce two samples (Figure 5).


**Table 1.** Gas

Chromatography-Mass

Spectrometry

 data from Tourloti and Mochlos vessels. peaks in a chromatogram, ‡ Relative Abundance, or peak area as a percentage relative to the maximum peak.

6

**Figure 3.** Total Ion Chromatogram from: (**a**) ethyl acetate extraction of Mochlos vat spout; (**b**) ethanol extraction of Mochlos vat spout.

**(c)**

**Figure 6.** TIC from: (**a**) first extraction of MS4494; (**b**) second extraction of MS4494; **(c)** third extraction of MS4494.

The Tourloti jar, by contrast, was sampled over a century after its excavation. The top-flight archaeological museum in which it currently resides now boasts climate-controlled storage; however, the jar certainly experienced more environmentally variable conditions prior to our sampling. Although neither the museum records nor visual inspection showed any signs of cleaning or conservation, it was unclear whether any viable compounds could be extracted from a jar so long exposed after excavation. As an experimental precaution, additional measures were employed to account for the potential buildup of dust or other contaminants on the interior of the vessel, and to maximize the likelihood that residues trapped within the fabric could be chemically extracted: Three separate hot solvent washes (i.e., swishes) were performed in rapid succession despite using the same solvent (i.e., ethanol), resulting in three separate filtered solution samples (Figure 6).

#### *2.2. Instrumentation Methodology*

The low concentrations of organic compounds preserved in ancient samples demand analytical instruments of high sensitivity, which allow researchers to separate and characterize organic compounds, both qualitatively and quantitatively. Instruments of choice have been chromatographic, especially liquid (LC-MS) or gas chromatography (GC-MS) coupled with mass spectrometry [16–18], with the protocols for each refined for the best expression of fragile and low concentration compounds.

The extracted Mochlos ORA samples, evaporated to dryness, were held in climate-controlled storage until export permissions from the Hellenic Ministry of Culture were granted, after which they were sent at first opportunity to the University of Pennsylvania Department of Chemistry in spring of 2005 [4]. Once safely in Philadelphia, preparations were made to inject the samples into the department's most advanced GC-MS housed in its Mass Spectrometry Facility. These preparations were made at the Roy and Diana Vagelos Laboratories at the invitation of G. Palladino, executive director, and G. A. Molander, the Hirschmann-Makineni Professor of Chemistry. The Mochlos ORA samples were processed and reintroduced into solution using ethyl acetate. Before injection into the Mass Spec Facility GC-MS, general conditions were determined using a HP 5890A Gas Chromatograph interfaced with a HP 3395 Integrator, housed in the Vagelos research laboratories of the Molander Group. This protocol was then further optimized in collaboration with the Molander Group over the course of several weeks using the actual GC-MS at the Mass Spec Facility, an Agilent Technologies 6890N Network 6C System used in series with an Agilent Technologies 5973N Network Mass Selective. Samples were automatically injected with an Agilent Technologies 7683 Series Injector with 10 μL syringe, which allowed automated overnight runs and the possibility to analyze hundreds of samples in a compressed time frame. Before the injection of each 2 μL sample, the computer automatically cleaned the syringe with 10 μL of methylene chloride twice and also primed it twice with the ORA sample pre-injection. Post-injection, the syringe was cleaned four times using 10 μL of methylene chloride. The GC-MS had a capillary column (30.0 m × 250 mm × 0.15 mm) set to constant flow mode and front inlet. The carrier gas flow (helium) was set to 0.5 mL/min with an average velocity of 41 cm/s. The EPC Split-Splitless Inlet was ultimately set to splitless as experimentation demonstrated that the small quantities of ancient compounds present in the circumstances at hand were better expressed in such fashion. The GC oven temperature had a set point of 110 ◦C held for two minutes, and increased by 6 ◦C/minute until it hit 250 ◦C, and then held for 20 min, with a total run time of 45.33 min. Solvent blanks were utilized before and after the set to verify that no contaminants existed from previous runs or that components were lost in the column.

The Mass Spec Facility GC-MS was interfaced with a HP Vectra VL420 DT computer (P4 2.20 GHz, 256 mb RAM) running Microsoft Windows 2000 Professional SP 2 (5.00.2195) and Agilent Technologies MSD Enhanced Chemstation (G1701DA version D.00.00.38). The compounds were initially identified using the National Institute of Standards and Technology Mass Spectral Database (NIST 02) with its Mass Spectral Search Program (1.7a) before being manually checked for final peak assignation. In order to help ensure the survival of the molecular ions in sample sizes that were anticipated to be relatively minute, the mass spectrometer was eventually set up for "soft" chemical ionization (CI), instead of

the harsher electron ionization (EI) more prevalent in labs. This approach made it more laborious to interpret mass spectra since most reference spectra and the computer library searches are EI-based, but the results proved this approach worthwhile in the context of ancient organic residue analysis.

The Tourloti ORA solution samples were transported back to the Brandeis University Department of Chemistry immediately after extraction in August 2015 and processed for instrumentation the following day [1]. The samples were concentrated to solid by rotary evaporator and then redissolved in uninhibited tetrahydrofuran (THF) to produce ~300 μL GC-MS analytes, which supplied 4 μL for auto injection into an Agilent 7890A GC with a HP-5MS column and a 5975C VL MSD Triple Axis Detector. Before the injection of each sample, the computer automatically cleaned the syringe with 10 μL of THF twice, and also primed it twice with the ORA sample pre-injection. Post-injection, the syringe was cleaned two times using 10 μL of THF. The pulsed split injector and interface were both set to 250 ◦C. The initial oven temperature was set to 100 ◦C and held for two minutes before reaching 250 ◦C at a rate of 10 ◦C/minute, at which time it was held for an additional 11 min, giving the total program time of 28 min/sample. As at Penn Chemistry, solvent blanks were utilized before and after the set to verify that no contaminants existed from previous runs or that components were lost in the column. Within the first sample (ARCHEM 4426), any anticipated dust and dirt were removed by hot filtration during the extraction process. The compounds were initially identified using the NIST Mass Spectral Database (NIST 11) with its Mass Spectral Search Program (2.0 g) before being manually checked for final peak assignation.

#### **3. Results**

Table 1 presents the comparative results and quantities of the diagnostic compounds recovered from each extraction of the Mochlos and Tourloti vessels. Note that palmitic and stearic acid, although usually represented in all samples, have been omitted from the list of compounds, as they are so ubiquitous in organic samples as to be non-diagnostic. In the Mochlos samples, oleic acid, cinnamic acid, linalool, manoyl oxide, docosane, and camphor were identified after initial peak assignation using the NIST Mass Spectral Database, NIST 02 (Figures 3–5). Each of these compounds can be connected to a high degree of certainty with known botanical sources local to Crete and the eastern Mediterranean. In the Tourloti jar, azelaic acid, oleic acid, linalool, and manoyl oxide were tentatively identified in the first extraction (ARCHEM 4426, Figure 6a) after initial peak assignation using NIST 11 with the understanding that they could be byproducts or remains of modern contamination until confirmed by subsequent samples. The second extraction (ARCHEM 4427, Figure 6b) clearly produced azelaic acid, cinnamic acid, camphor, and linalool from the ceramic matrix. In the third extraction (ARCHEM 4428, Figure 6c), azelaic acid, linalool, and manoyl oxide were present again confirming their presence as components of the original ancient residue. Mass spectrum fragments for oleanolic acid also clearly came out at 20.07 min from this third extraction. These identifications were confirmed by running chemical reference samples for each individual constituent compound, and also through comparison with the chromatographic signatures from ethnobotanical samples extracted from *Artemisia*, *Cistus*, and *Tilia* sourced from Crete, and from ethnographic samples extracted from modern 18th century pithoi housed in the Museum of Cretan Ethnology Research Centre, which oral history and visual inspection documented had been used to store olive oil. More data can be seen in the Supplementary Materials.

The slate of ingredients suggests that both the Mochlos and Tourloti jars held complex perfumed oils. The combination of botanicals employed suggests a surprisingly early understanding of perfume design through the creation of balanced tripartite blend, as each jar contained constituents that appear to have functioned in the perfumes as lower ("base"), middle ("heart"), and top ("head") notes, respectively. This fragrance framework of base, heart, and head notes is a governing principle in the craftsmanship of perfumes throughout history [19] and its importance was apparently recognized millennia ago.

### *3.1. Oil from Olea Europaea (Olive)*

Azelaic Acid (C9H16O4, 188 MW, 11.05 min at Brandeis, 12.90 min at Penn) Oleic Acid (C18H34O2, 282 MW, 16.70 min at Brandeis, 24.50 min at Penn)

Oleic acid is a monounsaturated fatty acid (18:1) that is a pale to brownish-yellow liquid at room temperature. It is the most abundant unsaturated fatty acid found in nature, but occurs in high concentrations (55–85%) only in olive oil from *Olea europaea*. Oleic acid is rarely found in animals, so it serves as a good biomarker for olive when detected in sufficient quantity in the proper context. Azelaic acid in sufficient quantity can also be a secondary biomarker for olive oil, in certain archaeological contexts. After a decade of ORA research, the project has verified that legacy objects suspected of containing olive oil in antiquity often contain large quantities of azelaic acid, a saturated dicarboxylic acid, in lieu of oleic acid. This is because oleic acid degrades to azelaic acid through oxidation scission processes that can occur once artifacts are exposed to the open environment [20–22] and discussion below. *O. europaea* is abundant on the island of Crete and was cultivated by 3000 B.C., or Early Minoan I [5]. In addition to being a widely available product of the Mediterranean region, textual sources from the Bronze Age and later classical periods demonstrate that olive oil was frequently used as a liquid base for perfumes [23], likely preferred in antiquity for its ability to be stored at ambient temperatures for prolonged periods without evaporating or spoiling.
