Origin and function of bitumen coated torpedo jars from the Sasanian to early Islamic period fort of Fulayj in Oman

1. Introduction

A sample of 15 sherds of bitumen coated torpedo jars have been selected for analysis from the late Sasanian to early Islamic period fort of Fulayj in Oman. Torpedo jars form a distinctive class of handless transport container vessels with a narrow mouth and a pointed base coated with a waterproof lining. They are found very widely distributed across the Mesopotamian plain, southern Iran and along the Persian Gulf and western Indian Ocean littoral. Isolated examples are also attested from shipwreck assemblages in Southeast Asia. Inconviently, their chronology is broad spanning the period from at least the 3rd to 9th centuries CE with little obvious stylistic, morphological or technical change. Recent research on the composition of both the ceramic fabric and bitumen lining provides evidence for the production of this class within the general area of southern Iraq and/or southwest Iran [1,2,3,4]. The potential to further refine our knowledge of the production location(s) of this class remains open to investigation. The current contribution adds to this body of research by analysing samples from securely dated contexts from the fort of Fulayj in Oman.

2. Archaeological Context

Fulayj is the site of a small c. 30 × 30m, heavily defended, square fortification with projecting ‘U’ shaped corner and entrance flanking towers and a single entranceway facing to the east. It was built with a thick, neatly constructed stone base most likely supporting a much more substantial mudbrick superstructure. Radiocarbon dating indicates that the fort was built in the late pre-Islamic period sometime between the early 5th and mid-6th century CE [5]. The latest results of excavation and dating indicate that there was a brief horizon of related activity in the same location preceding the fort construction. The location of the site is c. 30km southwest of the major Indian Ocean emporium of Sohar, which may already have attained some prominence before the Islamic conquest. The dating of the fort, its clear architectural links with similar small Late Antique fortlets across the Middle East, and its expert construction all strongly suggest that it was built by an external military force, which we can associate historically with the projection of Sasanian imperial power and the settlement of a Persian population on the Batinah during the late Sasanian period [6]. It is not clear if this was a long-lived military occupation. Finds within the fort are generally scarce. Significantly, the building underwent certain modifications and continued to be occupied during the first century following the Islamic conquest [7]. At this time mudbrick rooms were inserted into one corner of the fort and the defensive aspect of the gateway was modified. It is possible that the site took on a less overtly military function. Torpedo jar samples derive from contexts spanning both the earlier and later occupation.

3. Materials and Methods

3.1. Samples

Fifteen samples of bitumen coating the interior face of potsherds were isolated by scraping the surface with a scalpel (Table 1). Four examples of the samples are presented in Figure 1. They illustrate different sitiations with variable crusts of bitumen. No bitumen occurs on the exterior face of the potsherds.

3. Analytical Procedures

3.2.1. Bitumen Analysis 

All archaeological and geological bituminous samples were subjected to the same analytical procedure at GeoMark Research Ltd. Detailed procedures have been described in previous papers [3,8].

3.2.2. Wine Detection 

The LC-MS/MS analysis follows a new analytical technique that targets ancient wine biomarkers developed at the UCLA Pasarow Mass Spectrometry Laboratory) (Elezi et al. in preparation), [9].

The archaeological samples were pulverized in a pestle and mortar, and 2 g of pottery powder was treated with 5 mL of Methanol/DI Water/Formic Acid mixture 50:50:0.1 (v/v/v) in a glass test tube. The samples were vortexed and centrifuged at 2000 xg for 15 min, and the supernatants were transferred into new tubes. The extraction procedure was repeated by adding 3 mL of the mixture, and the supernatants were pulled into the new tubes. 100 pmoles of Daidzin solution in Ethanol were added as an internal standard for potential quantification analysis, and then the samples were dried in a vacuum concentrator for 4 hours. The dried samples were resuspended in 100 μl of Methanol/Water/Formic Acid (95:5:01) (v/v/v), vortexed thoroughly, and centrifuged at 2000 xg for 5 min. The supernatant was transferred to HPLC polypropylene vials, and 25 μl was injected for analysis.

A targeted Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS) assay was developed on a Linear Ion Trap LTQ-XL (Thermo Scientific) mass spectrometer coupled to a Dionex Ultimate 300 HPLC system (Thermo Scientific) through a reversed-phase GL Science analytical column (Inert Sustain 2 μm, Phenyl 150 x 2.1 mm). The instrument was optimized in positive ion mode monitoring transitions for Malvidine 3-glucoside (Mv3g) 493-331 m/z, Vitisin A (VA) 561-399 m/z, Vitisin B (VB) 517-355 m/z, and Daidzin (DA) 417-255 m/z). The mobile phase consisted of A (99.9:0.1 v/v Water/Formic Acid) and B (99.9:0.1 v/v Acetonitrile/Formic Acid). The HPLC method utilized the linear gradient mixture of eluents A and B to elute the targeted compounds (min/%B: 0/5, 5/5, 12/40, 26/75, 28/5, 40/The m/z of a fragment ion from each compound was monitored at a specific LC retention time to ensure their specificity and accurate identification.

4. Results

4.1. Gross Composition

Gross composition data are listed in Table 2 Extractable Organic matter in dichloromethane (EO % by weight/sample), which represents the amount of bitumen, ranges between 9.8 and 51% by weight with an average value of 37.9%. All samples are rather rich in bitumen.

A plot of % sat (saturated hydrocarbons) vs. % aro (aromatic hydrocarbosn) vs. % polars (= resins + asphaltenes) in Figure 2, shows that most samples are extremely rich in polars. One sample (No. 3462) is slightly different for enriched in hydrocarbons (9.3 %).

A plot of % sat + aro vs. % resins (NSO’s) vs. % asphaltenes shows a diversified situation with samples extremely rich in asphaltenes (more than 80%) and some others (No. 3462, 3606 and 3451) in which resins are higher, between 18.6 and 37.1 %. The compositions of archaeological samples characterized by an enrichment in asphaltenes has been reported in many publications [10,11,12,13,14,15,16,17,18,19,20,21,22,23,24].

4.2. Isotope Data

Carbon and hydrogen isotopes are presented in Table 3 The plot of 13Csat vs. 13Caro and 13CNSO vs. 13Casp in Figure 3, clearly shows that there is at least two main groups of bitumen, group one with 13Casp between -27.1 and -26.8 ‰/VPDB and group two with 13Casp between -28.3 and -27.2 ‰/VPDB. Comparison of these values with data acquired on oil seeps from Iran (Figure 4) confirms the occurrence of two main group of samples originating from different geographical locations. Group one corresponds to samples originating mainly from Khuzestan and Bushehr, group two to samples from Kermanshah, Ilam, and Lorestan.

The plot of Dasp vsDNSO in Figure 5a does not show any relationship between D values and their group, defined by their carbon isotope ratios. This feature is not surprising for these parameters are not source dependant but reflects oxidation intensities among bitumens [25]. Comparison of Fulayj data (Figure 5a) with those of 15 oil seeps from Iran (Figure 5b) shows that the Fulayj asphaltenes are generally more oxidized than those of oil seeps. Surprisingly and not explained, resins (NSO) of Fulayj are comparable to those of oil seeps of Iran.

4.3. Steranes and Terpanes

Steranes and terpanes were used as biomarkers to characterize the source of bitumen. Mass fragmentograms of terpanes (m/z 191) and steranes (m/z 217) of two archaeological samples (No. 3463 and No. 3460) are reproduced in Figure Terpanes of the No. 3463 sample show a complete pattern of the hopane family (C28 H to C35 H) with a high Tm/Ts ratio, a rather high gammacerane, almost no tricycloplyprenanes. Their steranes is dominated by regular steranes which are slightly biodegraded. Terpanes from No. 3460 show also the complete hopane family with occurrence of tricyclopolyprenanes (28/3 and 29/3), a low gammacerane and the occurrence of 18(H)-oleanane which reflects the contribution of the Pabdeh Tertiary formation in the source rocks which have produced this type of bitumen. One should notice that the occurrence of 18(H)-oleanane is matching with samples having 13C of asphaltenes between -26 and -27 ‰/VPDB (Figure 6 and Figure 7).

The plot of some characteristic ratios namely Ts/Tm vs.13C of asphaltenes (Figure 7), Ts/Tm vs. Diasteranes/Regular steranes and Ts/Tm vs. Gammacerane/C31 RHopane confirms the occurrence of at least two bitumen families. The first one (No. 3458, 3459, 3460, 3461, 3464, 3602 and 3606) corresponds to bitumen containing 18(H)-oleanane, high diasteranes/regular steranes, high Ts/Tm ratio, low gammacerane/C31 RHopane and occurrence of tricyclic terpanes. This family was generated from a source rock (likely the Eocene-Oligocene Pabdeh formation) in which the organic matter has a vegetal contribution, and is deposited in a marine environment under anaerobic conditions [26]. The second family (No. 3462, 3463, 3465, 3603, 3604, 3605, 3607, 3608) has a low Ts/Tm ratio, mainly regular steranes , a high gammacerane/C31 Rhopane and originates likely from the Albian-Aptian Khazdumi formation deposited in a very anoxic hypersaline marine environment [27,28].

Figure 9. Plot of 18α(H)-oleanane / C30αβHopane vs. δ¹³Casp (‰ / VPDB) in oil seeps from different provinces of Iran.

Figure 9. Plot of 18α(H)-oleanane / C30αβHopane vs. δ¹³Casp (‰ / VPDB) in oil seeps from different provinces of Iran.

Comparison of Fulayj data to those of oil seeps and solid bitumen analyzed in the different provinces of Iran shows that the first group matches data from natural asphalts from Karmanshah and Lorestan-Illam provinces whereas the second group is not matching the present day acquired data on oil seeps from Lorestan-Illam-Khuzestan-Busher-Fars and Hormozgan provinces. Utilization of data on archaeological sites as proxis in Figure 10, reveals that data from Fulayj are matching data from Susa in Khuzestan and Chogah Ahowan in Ilam.

Plot of data on steranes in Figure 11 discriminates again the two groups of samples. Samples with 18(H)-oleanane are more biodegraded than samples without 18(H)-oleanane as indicated by the reduced C27 steranes . Samples without 18(H)-oleanane have less C28 steranes as is often the case in samples with high gammacerane which display a V-pattern (C28<C27 and C29). Interesting, samples with 18(H)-oleanane shows the selective removal of the C29R sterane, a characteristic biodegradation effects seen elsewhere in archaeological samples: at Kuriki Höyük in Turkey, Qala’t al Bahrain and Saar in Bahrain, Hummal in Syria, Anuradhapura in Sri Lanka [1,29,30,31]. This selective biodegradation of the biological configuration of steranes has been observed in crude oils at depth [32] and has been reproduced in 15 days under laboratory conditions using grams-positive strains belonging to Nocardia and Arthrobacter genera [33,34].

4.5. Identification of Wine

In an ongoing research program conducted by Gazmend Elezi at the UCLA Pasarow Mass Spectrometry Laboratory, a targeted LC-MS/MS assay was developed to detect wine residues [9]. The method targeted wine biomarkers such as Malvidine 3-glucoside (Mv3g), Vitisin A (VA), and Vitisin B (VB) monitoring transitions from precursors (Mv3g 493 m/z, VA 561 m/z, VB 517 m/z) to product ions (OE, 331 m/z, VA 339 m/z VB 355 m/z) (Figure 12). These molecules are characteristics of red wine [35,36].

Positive answers were recorded for 5 samples: 3 (Nos 3459, 3460 and 3464) with bitumen containing 18(H)-oleanane and 2 (Nos 3453 and 3465) with bitumen without 18(H)-oleanane (Table 4, Figure 13). Sample No. 3459 comes from within a room securely dated to the 7th century AD, seemingly directly following the conversion to Islam.

Figure 14. LC-MS/MS chromatogram of the archaeological sample 3464 depicting the signals of the monitored precursor-product ion transitions for the targeted compounds and the internal standard.

Figure 14. LC-MS/MS chromatogram of the archaeological sample 3464 depicting the signals of the monitored precursor-product ion transitions for the targeted compounds and the internal standard.

5. Conclusions

Fifteen samples of bitumen coated torpedo jars have revealed two main compositional groups originating in southwest Iran, one from the Karmanshah- Lorestan provinces and the other from the Khuzestan-Illam provinces. Five of these potsherds were identified as wine containers on the basis of specific biomarkers associated with red wine. The interpretation of this relatively small sample is complicated. Both bitumen composition groups are associated with the detected traces of wine. This suggests a consistent function regardless of the location of production. Of course, an important caveat here is that the origin of the bitumen and location of pottery production may not necessarily have been the same. Chronologically, the samples are associated with different parts of the occupation sequence (Table 1), the dating of which is constrained by radiocarbon dating evidence. Not all of this is fully published, but the outline can be summarised. One of the sherds that provided a positive identification for the presence of wine comes from Phase 2a (No. 3460). This is a thin but particularly rich occupation horizon that predates the construction of the fort indicating some kind of earlier activity. The absolute dating evidence is broad but delimited within the Sasanian period from the c. mid-3rd to mid-6th century CE. It would be surprising if the dating is in fact very much earlier than the construction of the fort. A second fragment (No. 3465) belongs to Phase 2c, which comprises the ephemeral occupation directly connected with the use of the fort sometime between the early 5th to mid-6th century CE. A third sample (No. 3459) belongs to the re-occupation of the fort and the modification of its interior with the insertion of a domestic mudbrick buiding in Phase 3, which can be dated between the late 6th to late 7th century certainly covering the first decades of the Islamic transition. The remaining samples are associated with the gradual collapse of the fortification and we assume that these fragments are residual from any part of the earlier occupation. In terms of differing sources of bitumen composition, there does not appear to be any strong chronological correlation. One should be cautious in drawing any firm conclusions from this limited dataset. What the results do seem to point to is both a consitency in the pattern of usage of this vessel class through time, and to the enduring acquisition of bitumen from multiple sources. This might plausibly be connected with alternative manufacturing centers and the distribution of different regional varieties of wine as outlined in the contemporary historigraphy. These features require further systematic investigation at scale.