Thirty-nine samples from the religious complex of Qâni’ which was excavated by the French Archaeological Mission, and two samples from a burnt-out warehouse in Sector VI, which was excavated by the Russian Archaeological Mission in Yemen, were investigated by screening geochemical techniques (Rock-Eval, dichloromethane extract, GC-MS of steranes and terpanes, carbon and deuterium isotopes of asphaltenes) completed by devoted GC-MS analyses in order to identify frankincense. Several types of frankincense (well preserved, thermally altered and carbonized), were identified. In addition sterane and terpane patterns reveal the occurrence of unexpected bitumen in potsherds. Bitumen samples were compared to some oil seeps and crude oils from Yemen but none of the potential sources showed geochemical characteristics matching the bitumen from Qâni’. The occurrence of 18a(H)-oleanane in an ozokerite suggests a possible origin from Khuzistan or Fars in Iran. A specific flowchart, giving access to the predominant triterpenic biomarkers, namely structures with oleanane, ursane and lupane skeletons allowed excluding Boswellia frereana as the frankincense present in the sample set. Various states of thermal degradation of frankincense were found among the frankincense samples of the religious complex through the occurrence of functionalized molecule which were formed by thermal processes.

Keywords: bitumen, frankincense, archaeological samples, crude oils, natural asphalts, gas chromatography-mass spectrometry, Rock-Eval, Yemen, religious complex

The religious complex of Qâni’ was excavated by the French Archaeological Mission in Jawf-Hadramawt in 1995 and 1996,¹ and a burnt-out warehouse in the Sector VI, was excavated by the Russian Archaeological Mission in Yemen between 1985 and 1994.² Thirty-nine samples were gathered during the various campaigns to be investigated by geochemical techniques in order to identify frankincense.¹⁻⁴ The archaeological site near the modern village of Bi’r’Ali on the coast of the Arabian Sea cost, 100 kms southwest of al-Mukalla, is located in a bay (Figure 1). The site was easily identified as ancient Qâni’ and its name appears in two inscriptions found there, CIH 621 and CIH 728. A Russian team excavated the site between 1972 and 1994. The French Archaeological Mission in Jawf-Hadramawt worked on the site for three seasons between 1995 and 1997 in collaboration with A. Sedov, head of the Russian team. An Italian Mission, directed by B. Davidde, ran two seasons of underwater excavations in the bay of Bi’r’Ali in 1996 and 1998.³ The environment of the site does not lend itself to agriculture and only barely to pastoral activities. The bedrock is basalt covered by sand. Subterranean sweet groundwater is extant and feeds wells that are still in use today. This was a royal foundation in a remote region and dependent on an organized supply system. In founding Qâni’, the intention of the rulers in Shabwa, capital of the Kingdom of Hadramawt, was to participate in the flourishing sea trade, as well as to create a maritime outlet for its own production of resins. Sailors believed, and this is borne out by maritime charts, that Bi’r’Ali was the bay with the best location along the whole Arabian coast between Aden in Yemen and Ra’s Al Hadd in Oman, and the black peak of Husn al-Ghurāb is an excellent navigational landmark.

Figure 1 The ancient city of Qâni’ (©Joël Suire and Hélène David/Mission archéologique Jawf-Hadramawt).

Since it was found in 1^st century BCE, Qâni’ maintained its position as the main harbour of South Arabia for six centuries. The site’s phase of expansion occurred between the very end of 1^st century AD-late 4^th or early 5^th century AD. Originally a royal warehouse,⁸ the site gradually evolved into a town with dwellings and warehouses spread along its coast bordering the well-protected bay. To the west of the residential zone, two or perhaps, three sanctuaries were located on a long, flat slab of lava. Qâni’ was at the heart of the aromatic resins trade. Resins were transported to its warehouses either by caravan routes from the interior of the Hadramawt, or by sea from foreign ports, including Sumhurân (Khor Rori), from where the Dhofar’s production was chipped.⁴ The samples presented here were found in situ in the archaeological layers of two excavated areas of the site.         

Sector VII, from where most samples originated, is an enclosed religious complex (40x42m, Figure 2).⁵ In its central area, the main sanctuary was reached by a straight staircase on a 7.2x7.3 m podium. To the north, an 11-metre-long long hypostyle hall was probably used for ritual meals and in the northwest corner of the enclosure, two rows of rooms were dedicated to the activities of the clergy (cooking equipment, common objects) which were later completed by a small cella which was likely to have sheltered a religious statue. Small stone platforms were built around the central sanctuary and the open space to the south contained approximately fifty hearths that were irregularly distributed in the sand. A warehouse (Sector VI, Figure 2),⁵ was razed by fire at the end of the 2^nd century or the beginning of the 3^rd century AD. While working there, the Russian Mission discovered a large quantity of frankincense that was present in coffins⁶ which were preserved by the fall of the roof. The frankincense is of exceptional quality, according to present-day Yemenites.

Figure 2 The sanctuary in sector VII.

Some preliminary observations of the samples, which were carried out by team members familiar with natural fragrances and the different varieties of Yemen’s aromatic mixtures, were sufficiently encouraging to justify undertaking chemical analyses. Incense burners excavated in South Arabia attested to this diversity: they were inscribed with the names of 13 different perfumes⁷ in South Arabian script. Names on Sabaean incense burners attested to the use of other products such as different varieties of woods, barks, roots, leaves and herbs. They may have been, as they still are, prepared and blended to be sold in souks as burhūr, or added to incense burners during fumigations. The results from resins found in Qanî’ will complete the previously published study that was carried out at Sharma on Middle Age samples that identified copal from Madagascar as the main resin excavated from the warehouse.¹⁵ Frankincense was not totally absent but its scarcity (confirmed in only two samples) raised the question of the real relationship of Sharma with the Hadrami hinterland. Some references to black amorphous residues suggest occurrences of burned materials, or mixtures of resin with bitumen, but the lack of analytical proof did not permit confirmation of this hypothesis. Geochemical analyses used for the identification of petroleum were applied as screening techniques to classify the samples and to characterize bitumen when recognized among the samples.¹⁶ Specific methods, corresponding to spectroscopic (FT-IR) and chromatographic tools (GC-MS), already applied for the characterization of frankincense, were used on a selection of samples presumed to be composed of this resin.¹⁷ In addition, mineralogical analyses were carried out to document the minerals used in mixtures with frankincense and bitumen. These mixtures are still prepared today to be used in fumigations.

Samples of so-called aromatic resins collected during excavations of Sector VII may be separated into four groups according to their geographic location (Table 1):

Group 1: in relation to sanctuary and altars

Around the large northern altar (18, 20, 21, 23, 24, 27, 31, 32)

Around the central podium (2, 3, 4, 6, 8, 9, 10, 11, 12, 13, 15)

Around the southern altars (19, 28, 36, 38)

Seven samples were collected around the main podium, four on stones supports, which were thought to be at the top of the staircase were they were burned. The path leading to the sanctuary was bordered by three stone altars on each side with another four altars located behind the sanctuary. Several samples were gathered around the eastern facilities and in the ballast from the altars, and these indicated offerings and fumigations.

Group 2: in the hypostyle hall (26, 29, 30).

This hall is interpreted as a refectory or a room used for ritual meals. Fish and mammal bones were found in the occupation level which partly validates this interpretation. The occurrence of a room for meals is attested elsewhere in Hadramawt in association with sanctuaries and were located within the cultural enclosure or integrated in a complex building at Raybûn.¹²

Group 3: in relation to the northwest corner of the building

Before the building (19)

Within the occupation levels (7, 14, 37)

The worship sets are currently enclosed by a wall and integrated buildings that were used by the clergy for their activities. The enclosure, with the two first western rooms, was built after the construction of the sanctuary and sample 19 corresponds to the utilization of the sanctuary before the building of the enclosure. The remnants of resin discovered in this area may be related to the cella, also housed in the building, where an incense burner was also preserved.

Group 4: in relation to hearths (17, 22, 33, 35? 39?)

Hearths, housed in soft sand, corresponded to ritual practices relating to the sacredness of the place but it was not possible to determine whether they were constructed during the main occupation of the sanctuary or afterwards, when it was partly abandoned. Flat shells (scallop or Pecten maximus) with ash and traces of combustion that may have been used to burn aromatic resins for offerings or during communal meals were found in these hearths.

Two samples from the burnt-out warehouse in Sector VII, excavated by the Russian Archaeological Mission (34, 40) should be added to this group. Macroscopic inspection of these show a diversified panel between white powder (14 and 41, Figure 3) which was thought to be resin mixed with mineral matter (flour of calcite, gypsum): yellow-to-orange organic materials which was undoubtedly resin (18, Figure 3): brown residue which was possibly burned resin (25, Figure 3) or something else (27, Figure 3) and pitch-dark carbonized matter (40, Figure 3). For comparison purposes, a sample of present-day frankincense (1) from the market at Sana’a was also incorporated into the sample set.

Figure 3 The different macroscopic aspects of samples analyzed.

No.14#1266: sample containing calcium sulphate. No.41 #1293: sample with calcite, gypsum and NaCl. No. 18 #1270: preserved frankincense. No 25 #1277: thermally altered frankincense. No 27 # 1279: bitumen inside a potsherd. No.40 #1292: “carbonized” organic matter.

Rock-eval pyrolysis

The Rock-Eval pyrolysis¹⁸ which is widely used by the petroleum industry to characterize source rocks, has been adapted as a screening technique for archaeological samples (pitch, bitumen, resins, tar, etc.).^19,20 A small amount of raw sample (100 mg) was heated under helium at 300°C for three minutes and then the temperature was increased up to 600°C at 25°C/min. The thermovaporization at approximately 300°C gave a first peak (S1) corresponding to free hydrocarbons and is expressed in mg HC/g of sample. Hydrocarbons and CO₂ formed by cracking between 300 and 600°C gave the second peak (S2) which is also expressed in mg HC/g sample, and a third peak (S3) refers to oxygen-bearing compounds transformed into CO₂ in mg CO₂/g sample. The temperature of the maximum of the S2 peak (Tmax) is expressed in °C. The organic carbon remaining after pyrolysis was obtained by oxidation under air (or oxygen) atmosphere at 600°C, and the resulting CO₂ peak (S4) is expressed in mg CO₂/g sample. The total organic carbon (TOC in weight %) is computed using S1, S2 and S4.

Reagents and solvents

All solvents and reagents were of analytical grade from Meck (Darmstadt, Germany). Crystalline reference samples of all triterpenic standards from frankincense were isolated and characterized in Laboratoire de Chimie Bioorganique.²¹ a- and b-amyrins and lupeol were purchased from Extrasynthese (Genay, France) and b-amyrenone and lupenone were obtained from b-amyrin and lupeol by oxidation with pyridinium chlorochromate (PCC) in dichloromethane. Derivatization reagents employed to treat samples were hexamethyldisilazane (HMDS) and chlorotrimethylsilane (TMSCI), and the solvents used were pyridine (Sigma-Aldrich, ACS reagent, >99.0%) and dichloromethane (Sigma-Aldrich, ACS reagent, >99.5%).

Chromatographic analyses were carried out using a Varian Saturn 3900 gas chromatograph, with a Varian 1177 injector, coupled to a Varian 2100 T ion-trap mass spectrometer. Separation was performed on a capillary column Varian CP-Sil 5 CB Low Bleed/MS, 30 mx0.25 mm i.d.x0.25 µm. The carrier gas was helium at a constant flow of 1mL/ min^-1. The GC injector temperature was 250°C. The initial oven temperature was 50°C maintained for two mins during splitless injection. The injection volume was 1µL. A 8°C.min^-1 gradient was then applied to the oven until 250°C was attained, and another 3°C.min^-1 gradient was applied until 350°C was reached. No hold time was performed at the upper limit. Ionization was carried out with standard conditions at 70 eV. The mass spectrometer was scanned in the m/z 50-650 range, with a cycle time of one second.

Derivatization procedure

Five mg of the material was trimethylsilylated with a solution of pyridine (0.5 mL), HMDS (0.45 mL) and TMSCI (0.3 mL). This derivatization reaction occurred at room temperature (30 mins) before the solution was dried with a stream of nitrogen while heating (<40 °C). Thereafter, the residue was dissolved in 0.6 mL of dichloromethane, and after filtration on PTFE cartridge, it was directly injected into the gas chromatographic system.

FT-IR spectroscopy

All infrared spectra were collected at room temperature with a Nicolet (Madison, Wisconsin, USA) Avatar 360 ASP spectrometer equipped with a DTGS KBr detector and controlled by EZ OMNIC 60 software. The IR spectra were collected with 64 scans and 4 cm^-1 resolution. Samples were homogeneously crushed with anhydrous potassium bromide in a proportion of 1/20. The powder was then compressed under 10t/cm² to form translucent pellets.

Analysis of the extractable organic matter

The soluble portion of the sample was extracted with dichloromethane, and the emphasis was on the analysis of “saturated” hydrocarbon fractions. The dichloromethane extract was fractionated into three classes of compounds (“saturated hydrocarbons”, “aromatic hydrocarbons” and polar compounds) by IATROSCAN MK IV apparatus on Chromarods (5 µ silica SIII). The quantitative measurement of each fraction was carried out by a flame ionization detector to obtain a gross composition of the extract. After the precipitation of asphaltenes by n-hexane and their quantitative gravimetric measurement, an aliquot of deasphalted extract (100 mg) was fractionated by MPLC (medium pressure liquid chromatography) on a LOBAR MERCK column (300 mm x10 mm, Lichro-prep 30-60 µm), activated at 120°C under N₂ for 30 mins. The “saturated hydrocarbon” fraction was eluted with n-hexane. Recovery of the “aromatic hydrocarbon” fraction was achieved with n-hexane using a back-flush step on both columns (pre-column and analytical column LOBAR-MERCK). This step could have re-introduced some polar compounds (ketones, esters) into the so-called “aromatic hydrocarbon” fraction in the case of archaeological samples.¹⁹ The two “hydrocarbon” fractions in the asphalts were mainly C₁₅₊alkanes and C₁₅₊aromatics. Finally, the polar compounds were eluted with dichloromethane/methanol (85/15) in back-flush mode from the pre-column. Each recovered fraction was then weighed.

Gas chromatography-Mass spectrometry (GC-MS) analysis of “saturated hydrocarbon” was performed using a Hewlett-Packard 6890/5973 GC-MS instrument equipped with an on-column injector and operating at 70 eV with a mass range of m/z 50-550. A DB5 fused silica capillary column was used (J&W, 60 mx0.25 mm, film thickness 0.1 µm). The samples were injected at 75°C, and the oven was programmed at 2°C/min to 300°C, where the temperature was held for 66 mins. d¹³C were measured by an on-line process that couples a Carlo-Erba 2500 elemental analyzer with a Finnigan Delta Plus mass spectrometer via a Conflo II unit. The sample was transported by a helium flow (15 psi, 86 ml/min) and oxidized by pure oxygen (17.5 psi, 37 ml/min) at a temperature of 1020°C in a furnace containing tungsten oxide on alumina. An on-line magnesium perchlorate trap was used to remove the water formed by combustion. The generated CO₂ was introduced into the mass spectrometer. A pulse of reference CO₂ with known d¹³C was injected before the CO₂ samples peak. The NBS 127 standard was used to calibrate the instrument and control the whole methodology. Isotope values are provided with an uncertainty of ±0.2‰/VPDB for CO₂.

dD was measured by an off-line process. The sample was oxidized with copper oxide at 590°C for 15 hours. The water produced with copper oxide at 590°C for 15 hours. The water produced by the oxidation was subsequently reduced with zinc at 470°C for 30 mins and the resultant hydrogen introduced via a dual inlet into a Finnigan Delta Plus mass spectrometer. A known standard was used to calculate the D value of the sample. The NBS 22 standard was used to calibrate the instrument. Deuterium values are given with an uncertainty of ± 4‰/SMOW.

Rock-Eval screening analysis

Rock-Eval pyrolysis on whole samples provided access to total organic carbon (TOC in %/sample) and, therefore, evaluated the contribution of organic matter in the samples. Other pyrolytic parameters, deduced from the pyrograms that are shown for four samples in Figure 4, provided the Hydrogen Index (HI in mg HC/g TOC), the Oxygen Index (OI in mg CO₂/g TOC) and the Tmax (in °C) of the S2 peak, and gave some insight into the nature of the sample. The plot of the Hydrogen Index as a function of Total Organic Carbon (TOC in w. % /sample) in Figure 5 showed a wide variety of sample properties: TOC varies from 0.18% -73% when HI ranged from 39-746 mg HC/g TOC. These data suggested that the sample set contains some preserved raw resins (e.g. samples 15 and 22) but also burned or highly carbonized samples (e.g. samples 11 and 30) and even some residual ashes (sample 21, not shown). In addition, the samples may contain rich in organic matter as indicated by TOC higher than 50% but were also composed of organic matter diluted by mineral phases, which may even be dominant. The nature of this mineral phase was explored on some samples by using X-ray diffraction analysis and FTIR analysis. The plot of the Hydrogen Index as a function of Tmax and OI in figure 6a confirmed the diversity of the encountered situation and a tendency of increasing the Tmax when the HI was reduced. This property fits with the occurrence of resins that are both well preserved and burnt. The choice of a representative Tmax may be sometimes difficult to assess due to flat (sample 5, Figure 4) or bimodal curves (samples 5 and 17). However, subsequent molecular information clarified the raw parameters, enabling the two main groups of samples in the whole set to be distinguished: those that contained bitumen and those in which bitumen was lacking. Bitumen-bearing samples are characterized by lower Oxygen Index and higher Tmax (Figure 6a & Figure 6b). A Tmax between 420°C and 432°C is currently seen in archaeological samples.^20,21 Sample 13, which is bitumen-bearing, displayed a typical bitumen value at 434°C (Figure 4). The detailed molecular analysis shown below provides some insight into observed differences among bituminous samples by indicating that Sample 17 is pure paraffinic bitumen called ozokerite. This specific character explains the bimodal pyrogram with the main peak related to paraffin contribution at 392 °C and the shoulder at 438 °C to polymeric macromolecules (Table 2).

Figure 4 Examples of pyrograms obtained by the Rock-Eval device.

Figure 5 Plot of HI (mg HC/g TOC) vs. TOC (%/sample).

Figure 6 Rock-Eval data: (A) plot of HI (mg HC/g TOC) vs. Tmax (°C). (B) plot of HI (mg HC/g TOC) vs. OI (mg CO₂/g TOC).

Analysis of the mineral phase

Investigation of the mineral content of the samples was carried out by X-ray diffraction, Fourier transform infrared spectroscopy (FT-IR) and also chemical means on both extracted residues after dichloromethane treatment and whole sample. The results obtained are compiled in Table 3. The selected samples covered a wide range from organic-rich powder (e.g. Samples 6 and 7) to those that were mineral dominant ones (e.g. Samples 28 and 41). The three samples (13, 27 and 29) that were identified as bituminous-bearing samples, were preferentially examined to be incorporated into a wider research program dealing with archaeological bitumen in the Near East and the Gulf.²³ Insoluble organic residues were mainly composed of mineral matter (I.O.C. = Insoluble Organic carbon between 0.3% and 3.5 %) with calcite and clays as dominant constituents. Quartz was generally low, except in Sample 34, where it reached 28% with 25% NaCl! Occurrence of aragonite brought by carbonated bioclasts and of NaCl related to sea proximity, favoured a local contribution of bioclastic sands⁷ However the identification of gypsum (and even anhydrite in one sample) in both extracted and whole sample suggested another source that has yet to be located. Gypsum, found as a major constituent in the white powder of Sample 14 and as crystals with NaCl in Sample 41, was possibly added to carbonated mixtures available locally or from another source located elsewhere.

Four archaeological samples (#1258, #1263, #1264 and #1283) contained mineral fraction characterized by FT-IR. The interpretation of the obtained spectra showed that these contain inorganic sulphate and/or carbonate groups (Figure 7). Indeed, the infrared absorption spectra of inorganic carbonate and sulfate compounds were very distinctive. Normal inorganic carbonate linkages presented strong characteristic absorptions that could be observed around 1450-1410 cm^-1, 880-860 cm^-1 and 750-700 cm^-1. The last two lower frequency bands were very sharp. The last was not systematically observed but it is often useful to identify minerals such as calcite and dolomite. In calcium carbonate, these absorption bands occurred at 1430 cm^-1 and 876 cm^-1. For inorganic sulfates, there were two characteristic bands, the first one, which was the strongest and was relatively broad, was shown in the range 1130-1080 cm^-1 while the second one at around 680-610 cm^-1, was weaker and very narrow. In calcium sulfate the very strong band was at 1140cm^-1.²⁴ We can conclude that samples #1263 and #1264 contain inorganic carbonate compounds corresponding to calcium carbonate and materials #1258, #1263 and #1283 show the presence of calcium sulphate. In these last samples, the simultaneous occurrence of triterpenoids, fatty substances and mineral matter could be a probable result of an anthropic mixture corresponding to a real formulation.

Extractable organic matter

Extractable organic matter (EOM) was isolated from a selection of diversified samples including yellow-to-orange resin-like powder and brown-to-dark brown mixtures, which were free of any support or stuck on potsherds. The quantified data are listed in Table 4. The GC-MS analysis of their respective alkane fractions has enabled us to distinguish samples corresponding to pure, well-preserved frankincense, burned frankincense and bitumen (Figure 8). The discovery of bitumen, not expected when the study was undertaken, entailed a geochemical analysis of the oil seeps in Yemen which outcrop at Balhaf (Wadi Harubya, Figure 9) in the vicinity of Qâni’, from two salt mines at Shabwa (Shabwa 1 and 2, Figure 9),²⁵ and from a tar sand in the neighborhood of the Mintaq 1 well (Figure 10). A plot of the gross composition of extracts in a ternary diagram (% saturates + aromatics vs. % resins, vs. % asphaltenes, Figure 11) shows four differentiated groups corresponding to preserved frankincense, burned frankincense, archaeological bitumen and finally, oil seeps. As usual, oil seeps appear as the most enriched in hydrocarbons, namely “saturates” + “aromatics”, whereas the bitumen of Qâni’ falls within the current area of archaeological oxidized bitumen with an abundant amount of asphaltenes, except for the sample #1269, which clearly belongs to the oil seep family. Its high organic carbon (TOC=72.8 w. %/sample) combined with its exceptional richness in high molecular weight n-alkanes and its distribution peaking at n-C₃₇ refine the diagnostic and indicate that the sample may be classified among geological paraffinic residues called ozokerites.^26,27

The GC-MS analysis of alkanes of the four samples from Qâni’, which were expected to be bitumen-bearing, confirms the occurrence of bitumen that exhibit markedly different origins. The #1269 ozokerite is separate from other Qâni’ samples with abundant diasteranes (cf. m/z 217, Figure 12) and the occurrence of 18α(H)-oleanane (cf. m/z 191, Figure 12). Other archaeological samples show various degrees of biodegradation in steranes and terpanes (Figure 12 & Figure 13) and changes in diagnostic ratios (Table 5) such as Ts/Tm, Gammacerane/C30abhopane, C29αααR/C29αααS, etc. suggesting different potential sources which are researched using data from oil seeps that were analyzed and by referring to data on crude oil from various fields (Figure 9). Selective biodegradation of the C29αααR, C28αααR and C27αααR steranes, which are the steranes with a biological configuration (#1281, Figure 13) were noticed among the recorded properties of biomarkers. This feature, which was also observed in other archaeological bitumen has been reproduced under laboratory conditions using Nocardia and Arthrobacter genera.²⁸ A plot of several molecular ratios currently used for correlation purposes, i.e. 27Sdia/29αααR vs. Ts/Tm (Figure 14a), GA/31αβHR vs. Ts/Tm (Figure 14b), 18α(H)-oleanane vs. Ts/Tm (Figure 14c), and a ternary diagram with ββ-steranes (Figure 14d) did not show a match with either oil from oil fields or oil seeps especially Balhaf, which was, according to its location, thought to be the likely source. Diasteranes were lacking in three samples (Figure 15a) from Qâni’, but were abundant in all crude oil. The ozokerite (#1269) possesses special properties: abundant diasteranes, higher Ts/Tm (Figure 14a & Figure 14c) and the occurrence of 18α(H)-oleanane (Figure 14c). This last characteristic suggests that #1269 may have originated from Khuzistan in Iran.²³

In summary, archaeological bitumen from Qâni’ did not obviously match the analyzed oil seeps of Yemen and may have been imported from Khuzistan and Fars in Iran as suggested by the occurrence of 18a(H)-oleanane. Such a trade link has been highlighted at the ruins of Al-Baleed near Salalah on the southern coast of Oman, near Yemen, through the analysis of caulking residues on some wooden planks, which date from the 11^th-12^th century, that were used in the walls of the Husn fortress. These planks are re-used materials that came from a sewn boat likely to have been Iranian in origin as shown by the abundance of 18a(H)-oleanane in the bitumen that was analyzed.

The plot of isotopic data of asphaltenes, dD vs. d¹³C in Figure 15, allowed three groups of samples to be distinguished:

1. Frankincense (#964, #1271, #963, #1286, #962) was, as expected, depleted in deuterium. Within this group the detection of the effect of alteration (thermal?) on the isotope values may be possible. Thermal degradation entails a shift of both dD and d¹³C. Thermally altered samples (#963, #962, #1286 and #1271) are enriched in deuterium and ¹³C, as usually reported
2. Oil seeps (Balhaf and Minqah), the deuterium values of which approach those of archaeological bitumen and may have been affected by oxidation
3. Archaeological bitumen that tend, as seen elsewhere (As-Sabiyah, Failaka, Umm an-Namel, Akkaz)²⁰ to be enriched in deuterium due to severe oxidation.

The characterization of frankincense by GC-MS analysis of selected samples

Archaeological frankincense has only been identified in various samples originating from sites located in Egypt, Yemen, France and Belgium.^{29−33,30-32} This study offers a novel opportunity to add chemical information on preserved archaeological frankincense as well as remains from incense burners that were affected by thermal treatments of various intensities. Thirteen archaeological samples were selected for the research and identification of frankincense. They were analyzed by GC-MS (Gas Chromatography-Mass Spectrometry). Their chromatograms showed the presence of predominant triterpenic biomarkers with oleanane, ursane and lupane skeletons. The occurrence of boswellic and lupeolic acids and their O-acetyl derivatives indicate that these samples contain a triterpenic resin belonging to the Burseraceae family named olibanum (Boswellia spp.) or frankincense.³⁴⁻⁴² Previous research demonstrated that it is possible to chemically distinguish Boswellia frereana, an endemic species from North Somalia, from other botanical species of Boswellia.⁴³ Indeed, B. frereana has a dominant triterpenoid named 3-epi-lupeol whereas the other Boswellia species also contain this component but in lower concentration and in association with a larger triterpenic population. Moreover, the usual biomarkers (a- and b-boswellic and lupeolic acids and their O-acetyled derivatives) are not detected in B. frereana (Figure 16). Since 3-epi-lupeol was not the predominant component in the analyzed samples (Figure 17), B. frereana can be excluded as the source of the frankincense present in the sample set.

Chromatographic profiles of triterpenic structures differed in the various samples: the distinction mainly concerned the most polar compounds eluting between 45 and 50 min. The sample location and history of Qâni’ suggest that resinous materials may have been exposed to various degrees of heating resulting in thermal degradation such as combustion and burning processes that may have affected their original composition. Polar compounds were degraded into non-polar molecules via thermal alteration and oxidation. The altered samples may be distinguished according to their degree of degradation corresponding to the extent of the effects of combustion. The specific chemical markers of frankincense are a-(18) and b-boswellic (19) and lupeolic (20) acids and their corresponding O-acetyl-derivatives (21-23). Molecular structures are given in Appendix 1. These molecules can undergo chemical transformation and form degradation products, i.e. 24-noroleanane-3,12-diene (3), 24-norursa-3,12-diene (5) and 24-norlupa-3,20(29)-diene (6). They were identified by their mass spectra and by referring to specialized literature (Table 6, Appendix 1).^43,44 Other degradation compounds were also characterized: these are not functionalized and have one or two double bonds. They were identified with the NIST library of mass spectra as olean-11-12(18)-diene (1), ursa-11,13(18)-diene (2), olean-12-ene (4) and urs-12-ene (7). These molecules were generated by dehydrogenation of triterpenic precursors. Aromatic compounds can also be formed by further dehydration.⁴⁵

No	Compound		tR (min)		Skeleton type	R1		R2	Double bond(s)
1	oleana-11,13(18)-diene		37.72		O	H		CH3	C11-C12 and C13-C18
2	ursa-11,13(18)-diene		37.92		U	H		CH3	C11-C12 and C13-C18
3	24-noroleana-3,12-diene		38.12		O	H		-	C3-C4 and C12-C13
4	olean-12-ene		38.31		O	H		CH3	C12-C13
5	24-norursa-3,12-diene		39.12		U	H		-	C3-C4
6	24-norlupan-3,20(29)-diene		39.32		L	H		-	C3-C4 and C20-C29
7	urs-12-ene		39.57		U	H		CH3	C12-C13
8	3-epi-β-amyrine		41.49		O	a -OH, b -H		CH3	C12-C13
9	3-epi-α-amyrine		41.94		U	a -OH, b-H		CH3	C12-C13
10	3-epi-lupeol		42.07		L	a -OTMS, b-H		CH3	C20-C29
11	ursa-9(11),12-dien-3-one		42.91		U	O		CH3	C9-C11 and C12-C13
12	β-amyrone		43.54		O	O		CH3	C12-C13
13	β-amyrin		44.06		O	a -H, b-OH		CH3	C12-C13
14	α-amyrone		44.4		U	O		CH3	C12-C13
15	α-amyrin		44.61		U	a -H, b-OH		CH3	C12-C13
16	lupenone		44.66		L	O		CH3	C20-C29
17	lupeol		44.95		L	a -H, b-OH		CH3	C20-C29
18	α-boswellic acid		45.15		O	a -OH, b-H		CO2H	C12-C13
19	β-boswellic acid		45.63		U	a -OH, b-H		CO2H	C12-C13
20	lupeolic acid		45.75		L	a -OH, b-H		CO2H	C20-C29
21	O-acetyl α-boswellic acid		48.55		O	a-OAc, b-H		CO2H	C12-C13
22	O-acetyl β-boswellic acid		49.17		U	a-OAc, b-H		CO2H	C12-C13
23	O-acetyl lupeolic acid		49.28		L	a-OAc, b-H		CO2H	C20-C29

Table 6 Triterpenoic compounds in archeological samples

By referring to the chemical composition of terpenic derivatives (Table 7), it may be possible to establish a classification of archaeological samples as a function of their degree of alteration. Samples #1270, #1258, #1259 and #1283 did not show any degradation compounds (1-7) and contained native biomarkers of olibanum (18-23), so could be considered as well-preserved frankincense. However, samples #962, #963, #1263, #1277 and #1292, were devoid of the O-acetylated frankincense biomarkers (21-23) but their degradation derivatives (1-7) were ascribed as burned olibanum samples. In addition, a third class was defined with resins that contained a- (18), b-boswellic (19) and lupeolic (20) acids as well as O-acetylated derivatives (21-23) and their altered corresponding compounds (1-7). This last class (#964, #1264, #1270, #1286, #1287) is interpreted as a partly degraded resin due to a mild thermal degradation and/or ageing processes.

Apart from frankincense, fatty acids -myristic, palmitic and stearic acids- were identified in a number of other archaeological samples, i.e. #1258, #1263, #1264 and #1283 however their presence was not significant due to low concentrations. Their origin is questionable as they may originate from either from plants or animals that were incorporated with the minerals added to the frankincense. As discussed above these samples contained calcium carbonate and sulphate, and were anthropic mixtures that explain the occurrence of fatty acids as possible contaminants.

Analysis of three of the bitumen samples (#1265, #1279 and #1281) was undertaken using the frankincense flowchart in order to check if they were mixed with other ingredients, especially frankincense. The resulting chromatograms of samples #1265 and #1279 showed the occurrence of molecules with an hopane skeleton between 35-42 min in m/z 191 (Figure 18) but no other diagnostic terpenes. The bitumen samples that were analyzed were not therefore mixtures including frankincense but were mainly archaeological bitumen as found elsewhere.²³ Sample #1281 is slightly different with the occurrence of the same hopane derivatives as well as triterpene close to epi-lanosterol (t_R = 42.99 min), a-amyrin (15, t_R = 44.86 min) and lup-12-en-3-ol (t_R = 45.32 min), characterized with m/z = 189 and 203 and which may have originated from the introduction of vegetal matter into the bitumen (Figure 18). The presence of α-amyrin indicated the occurrence of a triterpenic resin belonging to the Burseraceae family.

The results of the present study lead to the following observations

Firstly, the identification of bitumen in several samples is remarkable: a likely hypothesis is that it was used as a fuel and, when mixed with aromatic resin, facilitated fumigations related to religious practices. However, no such mixture was identified within the temple area. In addition, it is possible that the bitumen was used as a binder in elaborate mixtures to create diverse texture and density. Several samples were found to be mineral matter mixed with frankincense which was always the major constituent. It is important to underline that all samples that were analyzed were frankincense (Boswellia), never myrrh and it is significant that 35 samples of fumigation and offering residue found in religious complex were basically from the same resin and used in relation to a god or a religious practice. Thus, the use of some aromatics was likely to have been precisely defined. Fatty acids of either of animal or vegetal origin were identified in some mixtures indicating that anthropic mixtures comparable to those called now known as bukhur, associated to resins with other aromatic ingredients (flowers, wood, oils, etc.) were in use. Fumigations in a religious context could have been made with these preparations and not exclusively with pure resins.

Secondly, the analysis of bitumen samples shows that they did not originate from a local source (Balhaf or elsewhere) but probably from the Khuzistan area in Iran. The exchanges between Qâni’ and the Arabo-Persian Gulf are attested (mostly through the spread of pottery containers) but it is of particular interest to confirm the importance of a resource that was available locally. Bitumen was certainly available in the port of Qâni’ and reused from the waterproofing of boats. A dedicated trade of bitumen is attested by jars from south Mesopotamia which showed a thick deposit of Iranian bitumen when analyzed: these were excavated from coastal sites in eastern Arabia, including Qâni’.⁴⁶

The geochemical study of thirty-one samples from Sector VII of Qâni’, which were excavated by the French Archaeological Mission and from the Sector VI by the Russian mission, has led to the identification of frankincense and four bitumen samples. The frankincense was fully preserved in the burnt-out warehouse excavated by Sedov and in the religious complex of Sector VII, and was partly altered by thermal treatments associated with incense burners. No other type of resins, such as copal from Africa as identified at Sharma, was found at Qâni’. Some samples were also analyzed for their mineral content and NaCl, gypsum and calcium carbonate were identified. In addition to frankincense, bitumen samples were found in potsherds in the religious complex. The use of the bitumen was not understood but may have been a substance for caulking ceramics or lighting fires in incense burners. The geological origin of the bitumen samples was not confirmed, despite a comparison with Yemeni crude oil and oil seep. More precisely, the bitumen samples did not match the Balaf oil seeps, which are not far from Qâni’. The occurrence of 18a(H)-oleanane in an ozokerite and in some other bitumens suggests an import from Iran, as was shown to be the case in Salalah with another set of samples.

The authors are indebted to the authorities of Yemen and especially to the General Organisation of Archaeology, Museum and Manuscripts who authorized the export of samples for analysis. John Zumberge who allowed us to use used the data on crude oil from Yemen, and Thomas Van de Velde who drew the maps. Thanks are also due to Jean-Bernard Berrut, Beatrice Ruiz, Marie-Hélène Doumergue and Daniel Dessort who carried out the analytical work leading to the characterization of bitumen. We are deeply indebted to Michael H. Engel for his comments and suggestions which were of significant value in the preparation of the manuscript. Helen Kirkbride and Tom Vosmer are highly acknowledged for their serious review of the manuscript and for their numerous suggestions and improvements to the text.

We are alsoss indebted to Michel Mouton who provided all the archaeological details related to the studied area and who collected the samples in the sanctuary of Qâni’.

Authors declare that there is no conflict of interest.

1. Mouton M, Sanlaville P, Suire J. Comptes Rendus des séances de l’Académie des Inscriptions et Belles-Lettres. 2006 ;150(2):777−808.
2. Mouton M, Sanlaville P, Suire J. Arabian Archaeology and Epigraphy 2008;19(2):197−208.
3. Sedov AV. Arabian Archaeology and Epigraphy. 1992;3(2):110−137.
4. Sedov AV.Tradition and Archaeology. In: Ray HP, Salles JF, editors. Early Maritime Contacts in the Indian Ocean, Techno-Archaeological Perspectives of Seafaring in the Indian Ocean, 4th cent. B.C -15th cent. AD. New Delhi: Manohar; 1994. p.11−35.
5. Salles JF, Sedov AV. Qāni. Le port antique du Hadramawt entre la Méditerranée, l’Afrique et l’Inde. Fouilles russes 1972,1985−1989,1991,1993−1994. Indicopleustoi, Archaeologies of the Indian Ocean. Brepol: Turnhout; 2010.
6. Mouton M, Sanlaville P, Suire J. Comptes Rendus des séances de l’Académie des Inscriptions et Belles-Lettres. 2006;150(2):777−808.
7. Mouton M, Sanlaville P, Suire J. Arabian Archaeology and Epigraphy. 2008;19(2):197−208.
8. Davidde B, Petriaggi R, Williams DF. New archaeological investigations at Abu Dhabi Airport, United Arab Emirates. Proceedings of the Seminar for Arabian Studies. 2004;34:85−100.
9. Avanzini A. Journal of Oman Studies 2002;12:17−24;
10. Avanzini A. The Indian Ocean in the Ancient Period. Definite places, translocal exchange. In: EH Seland, editor. London: BAR International Series. Archaeopress, 2007. p. 23−31.
11. Inizan ML, Mouton M. Mission archéologique française dans le jawf-hadramaout. Chroniques Yéménites. 1997;4:27−28.
12. Sedov AV. Temples of Ancient Hadramawt. Arabia Antica. Pisa: Edizioni Plus, 2005. 3 p.
13. Sedov AV. New archaeological and epigraphical material from Qana (South Arabia). Arabian Archaeology and Epigraphy. 1992;3(2):110−137.
14. Audouin R, Robin C. Saba. 1994;1: 25−30,32−35.
15. Regert M, Devièse T, Le Hô AS, et al. Reconstructing ancient yemeni commercial routes during the middle ages using structural characterization of terpenoid resins. Archaeometry. 2008;50(4):668−695.
16. Connan J, Nissenbaum A. The organic geochemistry of the Hasbeya asphalt (Lebanon): comparison with asphalts from the Dead Sea area and Iraq. Organic Geochemistry. 2004;35(6):775−789.
17. Mathe C, Connan J, Archier P. Analysis of frankincense in archaeological samples by gas chromatography-mass spectrometry. Annali di Chimica. 2007;97(7):433−445.
18. Bordenave ML, Espitalié J. Applied Petroleum Geochemistry. Editions Technip, chapter II.2, Paris: 1993. p. 237−261.
19. Connan J. Géologie de la Préhistoire : Méthodes, Techniques Applications. In: JC Miskovsky, editor. Association pour l’étude de l’environnement géologique de la préhistoire. Paris: Maison de la Géologie, 77 rue Claude Bernard; 2002. p. 1045−1066.
20. Connan J, Carter JR. A geochemical study of bituminous mixtures from Failaka and Umm an‐Namel (Kuwait), from the Early Dilmun to the Early Islamic period. Arabian Archaeology Epigraphy. 2007;18:139−181.
21. Colioli GC, Mathe P, Archier C. Et al. A lupane triterpene from frankincense (Boswellia sp., Burseraceae). Phytochemistry. 2003;62(4):537−541.
22. Connan J. Akh Puratim 1. In: Margueron JC, Rouault O, Lombard P, editors. Série «Akh Purattim-Les Rives De L’euphrate». Mémoires d’archéologie et d’histoire régionales interdisciplinaires; 2007. p. 165−206.
23. Connan J. Le bitume dans l’antiquité. Actes Sud: Editions Errance; 2012. 272 p.
24. Bellamy LJ. The infra-red spectra of complex molecules. London: Fourth edition, Methuen edition; 1966. p. 344−345.
25. JF. Breton. Syria. 1991;68(14):59−75.
26. Marschner RF, Winters JC.Advances in Chem Series. 151. 1976;166−178.
27. Nissenbaum A, Aizenshtat Z. Geochemical studies on ozokerite from the Dead Sea area. Chemical Geology. 1975;16(2):121−127.
28. Chosson P, Lanau C, Connan J. Dessort, Biodegradation of refractory hydrocarbon biomarkers from petroleum under laboratory conditions. Nature. 1991;351:640−642..
29. Mathe C, Culioli G, Archier P, et al. Characterization of archaeological frankincense by gas chromatography-mass spectrometry. J Chromatogr A. 2004;1023(2):277−285.
30. Van Bergen PF, Peakman TM, Leigh-Firbank EC, et al. Archaeological frankincense. Nature. 1997;390:667−668.
31. Hamm S, Bleton J, Connan J, et al. A chemical investigation by headspace SPME and GC–MS of volatile and semi-volatile terpenes in various olibanum samples. Phytochemistry. 2005;66(12):1499−1514.
32. Charlier P, Poupon J, Jeannel GF, et al. The embalmed heart of Richard the Lionheart (1199 A.D.): a biological and anthropological analysis. Sci Rep. 2013;3:1296.
33. Baeten J, Deforce K, Challe S, et al. Holy smoke in medieval funerary rites: chemical fingerprints of frankincense in southern Belgian incense burners. PLoS ONE. 2014;9(11):e113142.
34. Fattorusso E, Santacroce C, Xaasan CB. 4(23)-Dihydroroburic Acid from the resin (incense) of Boswellia carterii. Phytochemistry. 1983;22(12):2868−2869.
35. Fattorusso E, Santacroce C, Xaasan CB. Dammarane triterpenes from the resin of Boswellia freerana. Phytochemistry. 1985;24(5):1035−1036.
36. Pardhy RS, Bhattacharyya SC. Tetracyclic Triterpene Acids from Resin of Boswellia. Indian J Chem. 1978;16:174−175.
37. Mahajan B, Taneja SC, Sethi VK, et al. Two triterpenoids from Boswellia serrata gum resin. Phytochemistry. 1995;39(2):453−455.
38. Snatzke G, Vértesy L. Über die neutralen Sesqui- und Triterpene des Weihrauchs. Monatsh Chem. 1967;98:121−132.
39. Winterstein A, Stein G. Untersuchungen in der Saponinreihe. X. Mitteilung, Zur Kenntnis der Mono-oxy-triterpensäuren. Z Physiol Chem. 1932;208:9−25.
40. Savoir R, Tursch B, Huneck S. Triterpènes XI(*): Présence d'acide 11‐ceto‐β‐boswellique dans l'encens. Bull Soc Chim Belges. 1967;76:368−370.
41. Corsano S, Iavarone C. Gazz Chim Ital. 1967;94:328−339.
42. Proietti G, Strapaghetti G, Corsano S. Triterpenes of Boswellia frereana. Planta Med. 1981;41(4):417−418.
43. Mathe C, Culioli G, Archier P. High-Performance Liquid Chromatographic Analysis of Triterpenoids in Commercial Frankincense. Chromatographia. 2004;60:(9/10):493−499.
44. Van Bergen PF, Peakman TM, Leigh-Firbank EC, et al. Chemical evidence for archaeological frankincense: Boswellic acids and their derivatives in solvent soluble and insoluble fractions of resin-like materials. Tetrahedron Lett. 1997;38(48):8409−8412.
45. Oros DR, Simoneit BRT. Identification and emission factors of molecular tracers in organic aerosols from biomass burning Part 2. Deciduous tres. Applied Geochemistry. 2001;16:1545−1565.
46. Jasim S, Yousif E. Dibba: an ancient port on the Gulf of Oman in the early Roman era. Arabian archaeology and Epigraphy. 2014;25(1):50−79.