NEW GEOCHEMICAL INSIGHTS INTO CENOZOIC SOURCE ROCKS IN AZERBAIJAN: IMPLICATIONS FOR PETROLEUM SYSTEMS IN THE SOUTH CASPIAN REGION

The Maikop Group and the Diatom Formation constitute the two main source rocks in the South Caspian Basin and onshore Azerbaijan where large‐scale oil production began more than 150 years ago. However, the stratigraphic distribution of the source rocks and the vertical variation of source‐rock parameters are still poorly understood. The aim of the present paper is therefore to investigate in high resolution the source‐rock distribution in the Perekishkyul and Islamdag outcrop sections, located 25 km NW of Baku, which provide nearly complete middle Eocene and lower Oligocene to upper Miocene successions. Bulk geochemical parameters of 376 samples together with maceral, biomarker and isotope data were analysed. In addition, new Re/Os data provide independent age dating for the base of the Upper Maikop Formation (30.0 ± 1.0 Ma) and the paper shale within the Diatom Formation (7.2 ± 2.6 Ma). The presence of steradienes in high concentrations demonstrates the thermal immaturity of the studied successions, limiting the application of some biomarker ratios.

(~25 m/Ma), which may have had a negative impact on organic matter preservation. Terrigenous organic matter occurs in variable but typically low amounts. If mature, the Maikop Group sediments at Islamdag could generate about 2.5 tHC/m².
The Diatom Formation includes a 60 m thick paper shale interval with high TOC contents (average 4.35 wt.%) of kerogen Type . The source potential is higher (~3 tHC/m²) than that of the Maikop Group. The organic matter is dominated by algal material including diatoms. High TOC/S ratios suggest deposition under reduced salinity conditions. Strictly anoxic conditions are indicated by the presence of biomarkers for archaea involved in methane cycling.
For oil-source correlations and a better understanding of the petroleum system, it will be necessary to distinguish oil generated by the Maikop Group from that generated by the Diatom Formation. This study shows that these oils can be distinguished based on the distribution of specific biomarkers e.g. C 30 steranes, C 25 highly branched isoprenoids (HBIs), and the C 25 isoprenoid pentamethylicosane (PMI).

INTRODUCTION
Onshore Azerbaijan is known for the presence of numerous oil and gas seeps. As self-ignition occurs frequently, Azerbaijan has become famous as the "land of eternal fire" and it has been suggested that the burning seeps formed the basis for the Zoroastrian religion (e.g. Selly and Sonnenberg, 2015). Oil seeps around Baku, the capital city, were described in the 13th century by Marco Polo; this area was the largest oil-producing region in the world during the second half of the 19th century and the early 20th century (e.g. Boote et al., 2018). Most oil production onshore Azerbaijan and in the South Caspian Sea comes from the Apsheron-Pribalkhan and the Lower Kura petroleum provinces (Boote et al., 2018;Fig. 1a). The main reservoirs are terminal fluvial fan sandstones of the Pliocene Productive Series (Hinds et al., 2004).
The Oligocene to middle Miocene Maikop Group and the upper Miocene Diatom Formation are generally considered to be the most important source rocks in this region, while older (e.g. Eocene) source units may also have contributed to the accumulation of hydrocarbons (e.g. Goodwin et al., 2020). In the offshore South Caspian Basin, these units are known to occur in the subsurface at depths of more than 10 km (Inan et al., 1997;Goodwin et al., 2020). However in onshore Azerbaijan, they are exposed at the surface in the Shamakhy-Gobustan area to the west of Baku (Fig. 1a) where outcropping source rocks have been the subject of detailed studies (TOC, Rock-Eval, organic geochemistry: Saint-Germès, 1998; Katz et al., 2000;Feyzullayev et al., 2001;Guliyev et al., 2001;Saint-Germès et al., 2002;Hudson et al., 2008Hudson et al., , 2016Johnson et al., 2010;Bechtel et al., 2013;2014;Alizadeh et al., 2017;Washburn et al., 2019). A synopsis of these studies suggests there is no consistent pattern in source rock data from the Maikop Group (Sachsenhofer et al., 2018a,b). For instance, it has been proposed that the highest quality source rock interval (~3 wt.% TOC; HI ~300 mgHC/gTOC) occurs in lower Oligocene units at Angeharan and Lahich (Bechtel et al., 2013(Bechtel et al., , 2014 Fig. 1a), whilst elevated TOC contents and HI values are also present in the upper Oligocene (TOC up to 4.7 wt.%; HI up to 350 mgHC/gTOC) and lower Miocene (TOC up to 5.8 wt.%; HI up to 400 mgHC/gTOC) at Siyaki and Khilmili, respectively (Saint-Germès 1998; see Fig. 1a for locations). Even higher TOC contents (max. 14.3 wt.%) and HI values (max. 800 mgHC/gTOC) were reported from a 14 m thick lower Miocene section at Gezdek (between Perekishkyul and Baku) by Katz et al. (2000).
Similarly, source rock parameters for the Diatom Formation vary greatly. TOC contents reported from outcrop samples are typically low (average 0.6 wt.%), but TOC contents up to 7.8 wt.% (average 1.0 wt.%) and HI values ranging from 107 to 807 mgHC/gTOC (average 308 mgHC/gTOC) occur in rock samples ejected from mud volcanoes (Isaksen et al., 2007), and in boreholes to the south of Baku (Alizadeh et al., 2017 and references therein). The low TOC contents in onshore samples were reproduced by Katz et al. (2000), Feyzullayev et al. (2001) and Johnson et al. (2010).
The lack of consistency indicates either major lateral and vertical variability in source-rock parameters, or issues surrounding the absolute chronology of the stratigraphy. Further problems are related to the general lack of integration of high-resolution (bio-)stratigraphy (e.g. Popov et al., 2008;Bati, 2015) and source rock studies. In this paper, to further the stratigraphic understanding of the Lower Kura Basin in Azerbaijan, we present representative source rock data from organic-rich intervals of Paleogene to late Miocene age from the Islamdag and Perekishkyul sections, located about 25 km NW of Baku (Fig. 1). The results offer new insights into the distribution of the main source rocks within the Eocene to Miocene sedimentary succession in eastern Azerbaijan and provide a basis for the identification of individual source rock intervals that may have generated hydrocarbons.

GEOLOGICAL SETTING
During Paleocene to Oligocene times, the territory of Azerbaijan evolved from being part of the northern Peri-Tethyan platform to being part of the Eastern Paratethys as a sea-level fall and intense tectonic activity separated the Paratethys from the World ocean (Popov et al., 2004). Restriction of marine gateways before and after this separation caused two major episodes of oxygen-depleted conditions and the deposition of hydrocarbon source rocks during middle Eocene (Kuma Formation) and Oligocene to early Miocene times (Maikop Group;Sachsenhofer et al., 2018b). Basin restriction reached a peak during early Solenovian (mid-Rupelian) time (early NP23; ~31 Ma) and triggered Paratethys-wide low salinity conditions (Solenovian Event; Fig. 2) (e.g. Zaporozhets and Akhmetiev, 2015). The Solenovian Event is often characterized by a carbonate-rich layer (the Polbian or Ostracoda Bed; Gavrilov et al., 2017;Sachsenhofer et al., 2017Sachsenhofer et al., , 2018a, but low salinity conditions continued during deposition of overlying carbonate-free sediments. The Tarkhanian flooding event, recently dated as 14.85 Ma in the western Caucasus (Palcu et al., 2019b), terminated the anoxic conditions in the Eastern Paratethys. Thereafter marine environments with varying salinity conditions continued during middle and late Miocene time. A major (Pontian) transgression has been dated as 6.1 Ma (late Messinian; van Baak et al., 2016;. In the South Caspian Basin, Miocene sediments are overlain by the fluvial lower Pliocene Productive Series (Hinds et al., 2004).

Cenozoic sediments in eastern Azerbaijan
The Cenozoic succession in eastern Azerbaijan is typically fine-grained and includes from base to top: the Paleocene Ilhidag and Sumgayit Formations, the

Sumgait Baku
Sites studied by Saint-Germès (1998), Bechtel et al. (2013Bechtel et al. ( , 2014 Neogene   Oligocene to middle Miocene sediments, attributed to the Maikop Group, are exposed along the flanks of Mount Islamdag (the "Islamdag section": Fig. 1b). The Maikop Group is dominated by laminated noncalcareous mudstones and claystones. These sediments have been studied previously using palynomorphs, calcareous phytoplankton and fish fossils (Popov et al., 2008) as well as rhenium-osmium (Re-Os) geochronology (Washburn et al., 2018;. Whereas calcareous (benthic) fossils are largely absent, fish remains are common. Conventionally the Maikop Group in eastern Azerbaijan is divided into the Lower and Upper Maikop Formations, with the Upper Formation further separated into three lithological units, labelled from base to top as members A to C (Weber, 1935) (Fig. 2). Member A is the thickest unit and is rich in secondary jarosite on weathered surfaces in outcrops giving it a whitish to yellow colour in the field. Samples from the uppermost part of member A on the eastern slope of Mount Islamdag yielded a Re-Os age of 17.2 ± 3.2 Ma (Washburn et al., 2018;. The age fits within the error range of the postulated early Miocene (Sakaraulian) age of the upper part of member A of the Maikop Group (Popov et al., 2008). Member B forms an important regional marker horizon and is composed of black, laminated non-calcareous mudstones with interbeds of platy siltstones (Popov et al., 2008). Member C is composed of laminated non-calcareous claystones to mudstones with jarosite weathering. Members B and C, characterized by impoverished endemic fish assemblages, may represent the Kozakhurian episode of salinity change. In terms of palaeogeography, the Maikop Group in eastern Azerbaijan represents deposition in one of the deepest parts of the Eastern Paratethys (Popov et al., 2008). Fish fossils indicate that oxygenated conditions extended down to a water depth of 600 m during early Miocene times (Akhmetiev et al., 2007).
The Chokrak-Spirialis Formation, weathering dark brown in outcrop, is composed of low-calcite claystones and calcareous claystones containing Limacina sp. planktonic microgastropods (Popov et al., 2008). The base of the interval with Limacina sp. is characterised by a distinct, light-grey marl bed rich in foraminifera (Weber, 1941). Dolostone interbeds up to 50 cm thick are commonly observed, particularly in the higher part of the unit. The average thickness of the formation in the study area is about 50 m. Based on the presence of carbonate components and the absence of jarosite in the Chokrak-Spirialis Formation, the boundary with the Maikop Group can readily be recognized in the field (Popov et al., 2008). The Chokrak-Spirialis Formation represents deposition during semi-marine conditions during Tarkhanian and early Chokrakian times (nannoplankton zones NN4-NN5; Akhmetiev et al., 2007).
Sediments of the Diatom Formation follow above those of the Chokrak-Spirialis Formation and contain multiple lithologically distinct units. The total thickness of the Diatom Formation varies greatly and can be in excess of 500 m. The basal part of the Diatom Formation is dated as Karaganian based on endemic fish fauna (e.g. Sardinella karaganica) implying a freshening phase compared to the underlying rocks (Popov et al., 2008). Higher in the Diatom Formation, which remains predominantly fine-grained, intervals with diatomites and organic-rich paper shale mudstones are found (Weber, 1941). Deposition of the Diatom Formation continued into late Miocene times (Johnson et al., 2010). This is further substantiated by the Diatom Formation being overlain by sediments dated as Pontian (late Messinian) elsewhere in the Gobustan region (van Baak et al., 2016;.

Section descriptions
The Perekishkyul section The Perekishkyul section, or sections referred to as Perekishkyul (including other spellings), have been described in the literature numerous times (Khalilov, 1962;Saint Germès, 1988;Saint Germès et al., 2002;Hudson et al., 2008, Johnson et al., 2010Bati, 2015;Hudson et al., 2016). However, there is considerable uncertainty and/or inconsistency in the previous publications as regards the thickness of the section and the associated age model, and the precise location. In part, this is because exposures are common in the immediate vicinity of Perekishkyul village and range from Late Cretaceous to late Miocene in age. This makes the incorporation of these published datasets into the new results presented here highly ambiguous. Therefore, we prefer to limit the integration of data from previous papers when interpreting our work on the section.
The Perekishkyul section studied here (Figs 2, 3) comprises three main lithological intervals covering approximately 185 m true stratigraphic thickness. The base of the section is marked by a tight anticline with a prominent 1.2 m thick bed of bioclastic limestones. Bioclasts include planktic and benthic foraminifera, brachiopod spines, red algae and possible fragments of bryozoans and corals. The lowermost interval of relatively continuous exposure comprises 16 m of alternating beds of lighter and darker calcareous claystone/marls (Fig. 3b). Preliminary results from palynology and calcareous nannofossils suggests that these sediments were deposited during the Paleocene, although these data will be presented in more detail in a future paper.
The Paleocene section is overlain by a short interval (16-22 m) of highly deformed, dark grey calcareous mudstones, fractured marls/argillaceous limestones and red sheared mudstones, indicative of severe tectonic disturbance. The duration of missing time within this deformed interval is uncertain, but it likely accounts for a significant part of the latest Paleocene and early Eocene.
Overlying the deformed zone is a predominantly white-weathering interval attributed to the Eocene Koun Formation (22-81 m) which shows an alternation of two dominant claystone lithologies. The primary lithology is a recessive-weathering claystone which is typically non-calcareous. The second lithology comprises tabular and prominent claystone beds that retain some sedimentary structures such as ripples and lamination, as well as calcareous microfossils which are visible in hand specimen. A number of volcanic ash layers are also preserved within this part of the section. This interval contains dinoflagellates dated as Eocene (Bati, 2015). The age appears to be in line with preliminary age data obtained by 40 Ar/ 39 Ar dating of ash layers, which suggest a middle Eocene age for the lower part of the interval. These data will be published in full in a future paper. Here, we refer to this part of the section as "Middle Eocene I" (Fig. 3c). Further logging upwards of this part of the section was not possible due to a gap in exposure.
Above the gap in exposure (81-134.5 m), two intervals were logged that were taken to be representative of the stratigraphically highest unit at Perekishkyul that represents a major lithological change compared to the underlying section. The lower of the two studied intervals in this topmost unit is predominantly composed of dark greenish grey, noncalcareous claystones. The higher interval (Fig. 3d) contains additional white calcareous marl interbeds up to 30 cm thick. A volcanic ash age in this part of the section suggests a depositional age at the end of the middle Eocene. We refer to these two stratigraphically highest intervals as "Middle Eocene II". Of particular importance in both intervals are cm-thick, black organic-rich non-calcareous claystone beds that appear to be interbedded with the background sediments. Overall, these black claystone beds may contribute up to 5 % of the total sediment thickness (probably less than 2 m net thickness) in the section.
The thicknesses of gaps in exposure have been estimated using average bedding orientations and GPS locations. The presence of chevron folding in the highest stratigraphic unit provides evidence of significant deformation, which reinforces the need for caution concerning the assumption of stratigraphic continuity. The thickness of the non-exposed sediments shown in Fig. 2 is therefore subject to a degree of uncertainty. Although exposures continue for several hundred metres further along the outcrop, the presence of folding and regular faulting may prevent a continuous section upwards from being established.  salinities. This suggests correlation with the Rupelian Solenovian Event (Zaporozhets and Akhmetiev, 2015).
Within the Upper Maikop Formation, three lithologically distinct members can be identified (Figs 4c,5): the members A (16-324 m), B (324-347 m) and C (347-364 m) of Weber (1935). The Maikop Group is predominantly composed of dark grey, non-calcareous laminated mudstones with yellow sulphur-rich jarosite alteration on weathered surfaces. Sandstones are rare throughout and, when present, occur as isolated cmscale lenticular scour infills. The presence of diverse dinocysts with marine affinity in member A indicates increased salinities compared to the underlying interval (Popov et al., 2008). Member B comprises dark grey, non-calcareous laminated mudstones with a brown oxidised appearance. Due to the change in weathering colour, this unit stands out in the field as a distinct, dark-coloured interval. Batiacasphaera sp. and Botryococcus braunii occur abundantly in this unit, suggesting a return to lower salinities (Popov et al., 2008). This reduction in salinity in the uppermost Maikop Group would suggest a correlation with the early Miocene Kozakhurian regional stage of the Eastern Paratethys.   The base of the overlying Chokrak-Spirialis Formation (368 m: Figs 4c, 5) is marked by a sharp transition to a 50 cm thick grey foraminifera marl. This point represents the first appearance of calcareous rocks and fossils. By correlation with other sections elsewhere in the Eastern Paratethys (Palcu et al., 2019b), this level likely has an age of 14.85 Ma. The overlying 47.5 m of the Chokrak-Spirialis Formation is composed of weakly calcareous laminated mudstones containing fossils of planktonic microgastropods. Occasional dolostone intervals up to 30 cm thick occur in this part of the section.
Overlying this, extending upwards to the top of the section, is the Diatom Formation which has a total stratigraphic thickness of ~330 m. The base of the Diatom Formation is gradational and is assigned an age of ca. 13.8 Ma, via correlation to the Karaganian regional stage of the Eastern Paratethys (Popov et al., 2008). The Diatom Formation is predominantly composed of a variety of fine-grained sediments comprising, from the base up: a >200 m thick interval dominated by diatomites and diatomaceous rocks with common dolostone interbeds (up to 550 m in the outcrop section); a mildly calcareous mudstone interval (550-610 m); a black non-calcareous mudstone reminiscent of the Maikop Group facies (610-625 m); bituminous paper shale calcareous mudstones (625-677 m; Fig. 4a,b); and an uppermost interval of weakly calcareous mudstones (677-750 m). The top of the section is interpreted as older than 6.1 Ma based on the absence of the lithologically and biostratigraphically distinct sediments of the younger Pontian regional stage (van Baak et al., 2016). Mineralogy and element concentrations of the interval between 613 and 663 m have been studied by Abdullayev et al. (2021;their Pereküşkül-Diatom section). Based on questionable correlations, the authors assumed a Karaganian to Konkian age for the interval.

Samples
The present study focuses on samples from the Perekishkyul and Islamdag outcrop sections (Figs 3,4). The Perekishkyul section (N40.49806° E 49.5355°) is located three kilometres east of the Perekishkyul settlement on the right bank of the Sumgait River ( Fig.  1b) (e.g. Bati, 2015; his Kirmizitepe section) (Fig.  3). Sixty-seven samples were collected from ~200 m of stratigraphy encompassing the Paleocene-Eocene interval. The sampling interval in the studied parts of the section varied between 0.1 and 3.0 m.
At the Islamdag section, the Maikop Group as well as the overlying Chokrak-Spirialis and Diatom Formations were studied in six laterally-offset subsections located on the southwestern (N40.51628° E49.44321°), eastern (N40.51439° E49.47781°) and northeastern (N40.52461° E49.45959°) flanks of Mount Islamdag (cf. Popov et al., 2008). The subsections are labelled from base to top IDF, IDG, IDB, IDC, IDE and IDD (Figs 1b,2,4). Most adjacent sub-sections contain a significant overlap, but a gap between IDC and IDE is estimated to consist of roughly 22 m of stratigraphy based on GPS positions, bedding orientations and the general context in the field. In total, 300 samples were collected from ~750 m of section. The sampling interval varied between 0.3 and 6.0 m.
In order to minimize the effect of weathering, samples were collected from freshly exposed surfaces in hand-dug trenches up to 1 m deep. Nevertheless, the freshness of the sampled material depends on the initial exposure quality which varied significantly along the sections, and secondary effects on determined parameters cannot therefore be ruled out completely.

Methods
All samples were analyzed in duplicate for total sulphur (S), total carbon (TC) and total organic carbon (TOC; after acidification of samples) using an Eltra Helios C/S analyzer at Montanuniversitaet Leoben, Austria. The standard deviation of the C and S analyses were usually below 0.1 wt.%. This is in good agreement with the standard deviations of the standards used for calibration (e.g. S-standard: 2.27 % ± 0.06). The TC and TOC contents were used to calculate calcite equivalent percentages (Calc equi = [TC-TOC]*8.333).
A Rock-Eval 6 instrument was used to determine the amounts of free hydrocarbons (S1, mg hydrocarbons [HC]/g rock) and of hydrocarbons generated during thermal cracking (S2; mg HC/g rock). The hydrogen index (HI = 100*S2/TOC [mg HC/g TOC]) and the production index, PI = S1/(S1+S2), were calculated according to Espitalié et al. (1977). The temperature of maximum hydrocarbon generation (T max ) was recorded as a maturity parameter. The standard deviation of the Rock-Eval parameters S1 and S2 are below 0.07 and 0.5, respectively, which is in agreement with the results of the standards used (S1: 0.14 ± 0.07, S2: 12.43 ± 0.5). T max values show a standard deviation of 2°C at maximum.
In order to quantify the hydrocarbon potential of the potential source rock intervals, the amount of hydrocarbons which can be generated under a surface area of 1 m² was calculated using the Source Potential Index, SPI, which is defined as thickness*(S1 + S2)*bulk density/1000 (cf. Demaison and Huizinga, 1991).
Polished blocks were prepared for ten samples from the Perekishkyul section and 23 samples from the Islamdag section. Semi-quantitative maceral analysis used reflected white light and fluorescence light to distinguish aquatic macerals (e.g. alginite, liptodetrinite) from terrestrial macerals (e.g. vitrinite, inertinite). Vitrinite reflectance measurements were performed using an incident light Leitz microscope following established procedures (Taylor et al., 1998). In addition, pyrite contents and fossil remains are described. Some of the samples were also investigated using scanning electron microscopy (Tescan Clara field emission SEM; see Misch et al., 2018 for details). To support the occurrence of diatoms, biogenic silica contents were quantified following a dissolution technique described by Zolitschka (1998).
Based on bulk geochemical parameters, 30 organic -rich samples (six middle Eocene samples from the Perekishkyul section; eleven samples from the Maikop Group; one sample from the Chokrak-Spirialis Formation and 12 samples of the Diatom Formation from the Islamdag section) were selected for organic geochemical analyses. Representative aliquots of selected samples were extracted for ca. 1 h using dichloromethane in a Dionex ASE 200 accelerated solvent extractor at 75°C and 100 bar. After evaporation of the solvent to 0.5 ml total solution in a Zymark TurboVap 500 closed cell concentrator, asphaltenes were precipitated from a hexane:dichloromethane solution (80:1 according to volume) and separated by centrifugation. The hexane-soluble fractions were separated into NSO compounds and aromatic and saturated hydrocarbons using medium pressure liquid chromatography with a Köhnen-Willsch instrument (Radke et al., 1980). The hydrocarbon fractions were analysed by a gas chromatograph equipped with a 60 m DB-5MS fused silica capillary column (0.25 μm film thickness) coupled to a ThermoFisher ISQ mass spectrometer. The oven temperature was programmed from 40°C to 310°C at 4°C/min, followed by an isothermal period of 30 min. Helium was used as the carrier gas. The sample was injected splitless at an injector temperature of 275°C. The spectrometer was operated in the EI (electron ionization) mode over a scan range from mass-to-charge ratio (m/z) 50 to m/z 650 (0.7 s total scan time). Individual compounds were identified on the basis of retention time in the total ion current (TIC) chromatogram and by comparison of the mass spectra with published reference data. Identification of pentamethylicosane (PMI) is based on abundant mass fragments 239 and 267, and the presence of irregular PMI was indicated by m/z 239 > 253. The C 25 highly branched isoprenoid (HBI) alkane and thiophenes were identified by diagnostic fragments 238 and 265, respectively. Absolute concentrations of most compounds were calculated using peak areas in the TIC chromatograms in relation to those of internal standards (squalane and 1,1´-binaphthyl). Compounds present in insufficient intensities for peak integration in the TIC (e.g. steranes, sterenes and hopanoids) were quantified by integration of peak areas in appropriate mass chromatograms using response factors to correct for the intensities of the fragment ion used for quantification of the total ion abundances. The concentrations were normalized to TOC. The uncertainties in concentrations are in the range of 5 to 10% (relative error), the latter resulting from peaks of low intensities.
Stable carbon isotope measurements of specific compounds were performed on selected samples using a Trace GC instrument coupled to a ThermoFisher DELTA-V IR mass spectrometer via a GC isolink combustion interface. CO 2 was injected during each analysis as a monitoring gas. The GC column and temperature programme used were the same as above. The samples, saturated and aromatic hydrocarbon fractions, were placed into tinfoil boats for the bulk carbon isotope analyses and combusted in an oxygen atmosphere using an elemental analyser (Flash EA 1112) at 1020 °C. The evolving CO 2 was separated by column chromatography and analysed online using a DELTA-V IR-MS. The 13 C/ 12 C isotope ratios of CO 2 were compared with the monitoring gas. Stable isotope ratios are expressed relative to the Vienna Pee Dee Belemnite (V-PDB) standard in delta notation (δ 13 C = [(δ 13 C/δ 12 C) sample /(δ 13 C/δ 12 C) standard − 1]; Coplen, 2011).  Fig. 2). The large sampling interval was not considered to be a problem, as Washburn et al. (2019) observed relatively similar Re-Os values over several metres of stratigraphy. The selected samples were polished with silica carbide grit pads to minimize weathering effects and to remove any contamination by metal tools. Samples were then powdered in an alumina-ceramic shatterbox to yield 30-80 g of powdered sample. The Re-Os analysis was conducted at the Laboratory for Sulfide and Source Rock Geochemistry and Geochronology, and the Arthur Holmes Laboratory, Durham University. Each shale aliquot was dissolved in ~8 mL of 0.25g/g of CrO 3 in 4N H 2 SO 4 with a known amount of 185 Re+ 190 Os tracer (spike) solution at 220ºC for 48 hr. The Re-Os laboratory protocol follows that described in Selby and Creaser (2003) and Cumming et al. (2013). The Re and Os isotopic compositions were determined using negative thermal ionization mass spectrometry (NTIMS) using a Thermo Scientific Triton mass spectrometer (Völkening et al., 1991;Creaser et al., 1991). The running average of the isotopic compositions of in-house Re and Os standard solutions are 0.59863 ± 0.00203 (N = 495) for 185 Re/ 187 Re (ReStd 125 pg solution) and 0.16087 ± 0.0004 (N = 513) for 187 Os/ 188 Os (DROsS 50 pg solution). The total procedural blank for Re and Os are 0.10 ± 0.5 pg and 16.0 ± 1.0 pg, respectively, with a 187 Os/ 188 Os of 0.20 ± 0.05 (n =4). The Re-Os isotopic data including 2σ calculated uncertainties for 187 Re/ 188 Os and 187 Os/ 188 Os and the associated error correlation function (rho; Ludwig, 1980) were regressed using IIsochron program Isoplot (Ludwig, 2011) and IsoplotR (Vermeesch, 2018) using the 187 Re decay constant of 1.666e-11 ± 5.165-14a-1 (Smoliar et al., 1996).

Rhenium-osmium geochronology
The sample suite from the basal part (18.0-53.1 m) of member A (IDF sub-section) in the Islamdag section (1.2-2.9 wt.%TOC) has Re and Os abundances between 3.4 and 281.4 ppb and 44.7 and 383.7 ( 192 Os = 16.7-113.1) ppt, respectively (  Fig. 6a). The Model 3 regression and MSWD >1 indicate that the scatter about the best-fit line of the Re-Os data is beyond that described by analytical uncertainties, and is controlled by a normal distributed variation in the initial 187 Os/ 188 Os of 0.120 (2σ) (Ludwig, 2011) or a non-normal variation in the initial 187 Os/ 188 Os of between 0.040 and 0.081 (2σ) (IsoplotR; Vermeesch, 2018). The scatter about the best-fit line is mainly caused by three samples (IDF-11, IDF-12, IDF-18), which possess initial 187 Os/ 188 Os at 30 Ma of between 0.56 and 0.59 (Table 1) and exhibit between 5.2 and 9.3% deviation from the line of best-fit. Linear regression of the Re-Os data set without samples IDF-11, -12 and -18 exhibits significantly less scatter (MSWD = 69), yet yielded an identical Re-Os date (30.0 ± 1.0 Ma, 187 Os/ 188 Os = 0.70 ± 0.03) (Fig. 6a). The scatter about the best-fit line is a result of a normal distributed or non-normal variation in the 187 Os/ 188 Os of 0.061 (Isoplot, Ludwig, 2011) or 0.0195 to 0.0435 (IsoplotR, Vermeesch, 2018), respectively. The Re-Os date indicates a late Rupelian (Solenovian) depositional age for the base of the Islamdag section. The initial 187 Os/ 188 Os of 0.70 ± 0.03 is higher (contains more radiogenic 187 Os) than that of the contemporaneous global ocean (Peucker-Ehrenbrink and Ravizza, 2020).
The samples from the IDD sub-section of the paper shale of the Diatom Formation (633.4-646.5 m; 1.8-10.8 wt.%TOC) in the upper part of the Islamdag section possess Re and Os abundances of 6.4 to 23.6 ppb and 127.9 to 261.8 ( 192 (Fig. 6b). The scatter about the best-fit line is a result of a normal distributed or non-normal variation in the 187 Os/ 188 Os of 0.049 (Isoplot, Ludwig, 2011) or 0.014 to 0.029 (IsoplotR, Vermeesch, 2018), respectively. The scatter about the best-fit line is caused by three samples 14,21). Without these samples a Model 3 Re-Os date of 7.2 ± 2.6 Ma and an initial 187 Os/ 188 Os of 0.80 ± 0.02 (MSWD = 12) is obtained (Fig. 6b) TOC -total organic carbon; S -sulphur, S1 -amount of free hydrocarbons, S2 -amount of generated hydrocarbons, min -minimum, avg -average, max -maximum, Tmax -temperature of maximum hydrocarbon genera�on hydrogen index (HI) and TOC versus S 2 are shown in Fig. 7. Average, minimum and maximum values of bulk geochemical parameters for each stratigraphic unit are listed in Table 2.

The Perekishkyul section
The average TOC and sulphur contents of the Paleocene interval are 0.90 and 0.25 wt.%, respectively. The sulphur content is low in the two lowermost samples and shows a general decrease upwards in the remaining part of the section. Apart from the basal layers, TOC/S ratios generally increase upward from 1.7 to 26.2 (Fig.  5). HI (26-58 mgHC/gTOC), T max (395-412°C; Table  2) and PI (<0.1) are low. Samples from the Middle Eocene I interval contain small amounts of organic matter (0.02-0.64

Abbreviations:
TOC -total organic carbon; S -sulphur; S1 -amount of free hydrocarbons; S2 -amount of generated hydrocarbons; min -minimum; avg -average; max -maximum; T max -temperature of maximum hydrocarbon generation.
wt.% TOC) and sulphur (0.08-0.67 wt.% S). A very low TOC content (<0.05 wt.%) occurs in samples from the upper 10 m of this section. TOC/S ratios show a general decrease upwards from 5.6 to 0.09. HI values are slightly higher than in the Paleocene section and typically range from 100 to 200 mgHC/ gTOC. Even higher values (up to 575 mgHC/gTOC) are found in samples with a TOC content below 0.2 wt.% and high PI values (>0.1). Hence, these data probably reflect contamination. T max values could be determined reliably only for samples with relatively high TOC contents, and are of the same order as those of samples from the Paleocene interval (394-426°C).
Whereas the TOC content is below 1.0 wt.% in most Middle Eocene II samples, significantly higher values (2.0-11.4 wt.%) are observed in thin black shales within the upper part of the exposed section. The sulphur content of organic-lean sediments ranges from 0.6 to 1.8 wt.%, whereas that in organic-rich sediments varies between 1.6 and 3.5 wt.%. TOC/S ratios increase with increasing TOC contents and reach a maximum of 3.5 in a sample with 11.3 wt.% TOC. HI values are below 200 mgHC/gTOC in samples with low or moderate TOC contents (<2.1 wt.%), but vary between 323 and 420 mgHC/gTOC in samples with high TOC contents (3.4-11.3 wt.%). The Chokrak-Spirialis Formation typically contains low amounts of organic matter (avg. TOC: 0.82%) and has a low HI value (avg. 58 mgHC/gTOC). However, the formation includes a number of thin (~30 cm), microgastropod-bearing laminated mudstones, two of which (samples IDC-48.2 and IDC-48.4) have been included in the present study. These mudstones have high TOC contents (2.37 and 2.46 wt.%) and have moderately high HI values (214 and 227 mgHC/ gTOC). The sulphur content varies widely from 0.10 to 2.95 wt.%, and maximum values occur in organic-rich samples. TOC/S ratios in these samples are below 1.0, but are significantly higher (up to 5.73) in low-TOC samples. The average TOC/S ratio and the average T max are 2.67 and 423°C, respectively. PI values range from 0.02 to 0.26 for the entire samples set and from 0.02 to 0.19 for samples with a TOC content exceeding 1 wt.%.
The paper shales are rich in organic matter (avg. TOC: 4.35 wt.%; max. 13.27 wt.%) with very high HI values (avg. 456 mgHC/gTOC; max. 770 mgHC/ gTOC). The very high TOC and moderately high sulphur contents (avg. 0.96 wt.%) result in high TOC/S ratios (avg. 5.14). Sediments above the paper shale contain low amounts of organic matter (avg. TOC: 0.64 wt.%) with low HI values (avg. 74 mgHC/gTOC). Both TOC and HI values show a general decreasing trend upwards in the section. Low sulphur contents (avg. 0.22 wt.%) result in moderately high TOC/S ratios (2.94). The average T max value of the Diatom Formation is 422°C. PI values strongly depend on organic matter richness, and range from 0.01 to 0.10 for samples with S1 peaks ≤0.02 mgHC/g rock.
The Paleocene samples contain large amounts of alginite (59-90 vol.%) dominated by lamalginite. Considerable amounts of telalginite (Fig. 8b) are found in the upper two samples. The percentage of vitrinite macerals increases upwards from 10 to 39 vol.%. Inertinite macerals occur in significant amounts in a single sample. The lowermost sample (OM-09-04) contains a considerable amount of pyrite (Fig. 8a).
The maceral composition of the samples from the Maikop Group is dominated by liptinite (Fig.  8f). Lamalginite is generally more abundant than telalginite, but telalginite dominates in some samples from the lower part of the Maikop succession. Inertinite occurs in small percentages (≤10 vol.%). Vitrinite is the main maceral group in samples IDB-09 (309.95 m) and IDB-16 (316.4 m; Fig. 8e). n.e. n.e.
Paper shales from the Diatom Formation contain large amounts of alginite (avg. 77 vol.%; Fig. 8l,o). The vitrinite percentage is on average 21 vol.%. Inertinite macerals are typically rare (max. 7 vol.%). With the exception of two samples from the base of the paper shale, lamalginite is more abundant than telalginite. Diatoms and fish remains are observed in many samples (Fig. 8k,m-p).
Vitrinite reflectance was determined for two samples from the Perekishkyul section (Middle Eocene II), five samples from the Maikop Group in the Islamdag section, and two samples from the Diatom Formation. All reflectance values are below 0.3 %R o .

Biogenic silica
Diatoms were observed in samples from throughout the section. Therefore, a technique proposed by Zolitschka (1988) was applied to determine biogenic silica contents. The determined concentrations are generally low, but concentrations up to 10 wt.% were detected in samples below and above member B of the Upper Maikop Formation and in some paper shales within the Diatom Formation (Fig. 8). However, in some samples with microscopically visible diatom remains, biogenic silica contents are low.

Molecular composition of hydrocarbons
Biomarker data have been determined for 30 samples with high organic matter (OM) contents (Table 3). Characteristic chromatograms are shown in Fig. 9, and stratigraphic plots of selected biomarker ratios in Fig. 10. The significance of the biomarker parameters determined is summarised in Table 4. Facies-and maturity-related parameters are presented separately for the most relevant organic-rich intervals.

Middle Eocene II (Perekishkyul)
Six samples with high TOC contents (2.0-9.0 wt.%) from the upper part of the Perekishkyul section (Middle Eocene II) were investigated. These samples yielded relatively small amounts of extract (average 18 mg/ gTOC). The n-alkane distributions of all of the samples show a predominance of long-chain n-alkanes (n-C 26-31 ) with an average value of 47 %, whereas mid-chain (23 %) and short-chain n-alkanes occur in significant lower amounts (12 %). CPI values range from 2.2 to 3.3.

Maikop Group (Islamdag)
One sample from the Lower Maikop Formation (IDF-07), together with nine samples from member A and two samples from member C of the Upper Maikop Formation in the Islamdag section were analysed. Sample IDC-48.2 from the Chokrak-Spirialis Formation is also included in this section because it is similar to the samples from member C. The extract yield increases upwards from 14 to 155 mg/gTOC within the Maikop Group. A strong positive correlation (R² = 0.72) with relative asphaltene percentages shows that high extract yields are mainly caused by high amounts of asphaltenes. A positive correlation also exists in the Maikop Group between extract yields and PI values (R²=0.54). A relatively high extract yield of 74 mg/gTOC characterizes the uppermost sample from member A (IDB-16; 316.35 m). The extract yield in the sample from the Chokrak-Spirialis Formation is 45 mg/gTOC. n-Alkane distributions are characterized by high percentages of mid-(avg. 41 wt.%) and long-chain n-alkanes (avg. 31 wt.%). Short-chain n-alkanes occur in minor amounts (avg. 13 wt.%). Long-chain n-alkanes dominate over mid-chain n-alkanes in the lowermost and uppermost sample. CPI values vary significantly between 1.39 and 5.39.
Koun F. 28%) generally occur in low amounts. The steroids/ hopanoids ratio is generally very high (max. 8.23), but lower in a few samples from near the top of the Maikop Group.

Diatom Formation (paper shale; Islamdag)
The Diatom Formation is represented by ten samples from the paper shale interval, as well as one underlying and one overlying sample. All samples yield considerable amounts of extract (avg. 47.5 mg/ gTOC). The maximum value (109.4 mg/gTOC) was observed in the low TOC sample (1.23 wt.%) above the paper shale (IDD-52). n-Alkane distributions are characterized by strongly varying percentages of short-(avg. 19 wt.%), mid-(avg. 34 wt.%) and long-chain n-alkanes (avg. 32 wt.%). CPI values are 1.92-5.18 and correlate well with the relative amount of long-chain n-alkanes (R² = 0.83).

Compound-specific C isotope ratios
Isotope data for n-alkanes, as well as for pristane [δ 13 C pristane ] and phytane [δ 13 C phytane ] were determined on the 30 biomarker samples ( Table 5). The δ 13 C values for pristane were obtained by integrating the overlapping peaks of n-C 17 , present in very low abundances, and pristane. δ 13 C-values of short-(n-C 19 ) and long-chain n-alkanes (n-C 31 ), as well as of pristane and phytane, are plotted versus depth in Fig. 10.
n-C 18 to n-C 24 alkanes derived from Middle Eocene II sediments in the Perekishkyul section are characterized by a decrease in δ 13 C values with increasing chain length from -29.0 to -31.5 ‰. No further decrease could be observed for longer-chained n-alkanes (Fig. 11a). n-Alkanes in samples from the Maikop Group and the Chokrak-Spirialis Formation in the Islamdag section are generally isotopically heavier (Fig. 11b-d), although three isotopically light samples (IDG-53; IDB-09; IDB-23) occur near the top of member A (Fig. 11c). The paper shales in the Diatom Formation show varying isotope patterns ( Fig. 11e-g) with strongly negative values in some samples from its upper part (Fig. 11g).
δ 13 C pristane and δ 13 C phytane values are similar (difference ≤1 ‰) in samples from the Perekishkyul section and several samples from the Islamdag section (Fig. 11h). This suggests a common precursor (i.e. chlorophyll). In contrast, phytane is significantly enriched in 12 C in the lowermost four Maikop Group samples from Islamdag and seven samples from the Diatom Formation, suggesting co-elution of phytane with isotopicallylight crocetane, an irregular C 20 -isoprenoid (Thiel et al., 1999).
δ 13 CHBI values in middle Eocene sediments range from -31.1 to -29.7 ‰ and are less negative in the Maikopian sediments (-29.6 to -28.5 ‰). The range of δ 13 CPMI values was also determined on some samples from the Diatom Formation, and yielded even more negative values (-34.2 to -36.5 ‰).

Thermal maturity and its influence on biomarker ratios
Many geochemical parameters used for the reconstruction of depositional environment and the assessment of source rock potential are influenced by the thermal maturity of the rocks. Hence the       Table 5. Compound specific carbon isotope (δ 13 C, ‰ V-PDB) of Eocene, Oligocene and Miocene sediments from the Perekishkyul and Islamdag sections. thermal overprint is discussed first, and depositional environment and hydrocarbon potential are then considered in the following sections.
The low vitrinite reflectance values (<0.3 %R r ) and the low average T max values obtained for samples from all of the studied sedimentary units (405-424°C; Table  2) are evidence of very low maturity. The low maturity is also supported by other geochemical parameters, such as high CPI values, high concentrations of sterenes and steradienes, the absence of ααα steranes with an S configuration, and the absence of αββ steranes. Isomerisation of C 31 hopanes is low in samples from the Perekishkyul section and in the Maikop Group samples from the Islamdag section, but significantly higher in samples from the Diatom Formation. This upward increase in isomerisation ratios is a facies effect rather than a maturity effect. Similarly, higher isomerisation ratios are observed in Jurassic rocks containing kerogen Type I than in rocks with kerogen Type II (Neumeister et al., 2015).
Classical maturity parameters such as vitrinite reflectance, T max values and hopane isomerisation ratios do not show an increase with the age (and the burial depth) of the sediments. However, the fact that monoaromatic steroids occur (in very low concentrations) exclusively in Eocene sediments and in the Maikop Group, and that steradiene concentrations are higher than sterane concentrations only in the Diatom Formation, may reflect a subtle downwardincreasing thermal overprint.The very low maturity may influence facies-dependent molecular proxies. For example, the incomplete reduction of steradienes and sterenes may be responsible for the lack of diasteranes and may also influence the relative sterane percentages. This is evidenced by differences in the relative abundances of C 27 , C 28 and C 29 homologues of steradienes and steranes, respectively, in samples from the Diatom Formation (Fig. 9). Moreover, the observed low concentrations of DBT and the resulting very low DBT/Phen ratios (Table 3) may be due to the very low maturity rather than the limited availability of free H 2 S (e.g. Hughes et al., 1995). This interpretation is supported by the high concentration of other sulphur-containing compounds (e.g. isoprenoid thiophenes in the Diatom Formation; HBI thiophenes in the Eocene and Maikop Group samples). The sample overlying the paper shale (IDD-52) is the only sample with a high DBT/Phen ratio (0.73). Moreover, Pr/Ph ratios may be influenced by the low maturity of the samples as phytane is released preferentially during early diagenesis (Volkman and Maxwell, 1986;Peters et al., 2005). Pr/Ph ratios in the marginally mature Maikop Group at Lahich range from 1 to 3 (Bechtel et al., 2014), whereas ratios in immature Maikop Group sediments at Angeheran were constantly below 1 (Bechtel et al., 2013).

Depositional environment of organic-rich sediments Middle Eocene II (Perekishkyul section)
Organic-rich layers with TOC contents between 2 and 12 wt.% are intercalated with low-TOC shaly sediments in the Perekishkyul section. These alternations were studied in detail between 171.5 m and 178.5 m (Figs 5, 10), but individual, unstudied, organic-rich layers were also observed between 135 m and 142 m, suggesting that they cover a much greater stratigraphic interval than that currently sampled. Thus, it is difficult to estimate the cumulative thickness of the organic -rich layers, but they probably do not exceed 2 m at the studied location.
Maceral analysis shows that the organic matter is dominated by aquatic biomass but with a significant input of land plant material. In addition to the abundant vitrinite and inertinite macerals, the presence of the terrigenous organic matter fraction is indicated by high percentages of long-chain n-alkanes with a significant odd-even predominance. Land plantderived biomarkers (diterpenoids) occur in small amounts with a slight upward-increasing trend. The dominance of diterpenoids suggests that vegetation was dominated by gymnosperms (e.g. Bechtel et al., 2008). This is in agreement with the observation that gymnosperm pollen are notably more abundant than angiosperm pollen (Bati, 2015).
The aquatic biomass is represented by high percentages of lam-and telalginite, indicating contributions from both algal/bacterial mat material and planktonic algae. Very high C 27 sterane percentages also indicate a predominance of marine algal material. Although the percentage of biogenic silica determined with the technique proposed by Zolitschka (1998) was low (Fig. 8), significant concentrations of isotopically light (-30.5 ‰) HBIs indicate the high productivity of marine diatoms. This is further evidenced by elevated relative abundances of 24-methylcholesta-5,22-diene and 24-methylcholesta-5,24(28)-diene, derived from sterols found in high concentrations in the extracts of diatoms (Rampen et al., 2010). The presence of the C 30 sterane 24-n-propyl-5α-cholestane (20R) indicates a marine depositional environment, as 24-n-propylcholestanes have as yet only been identified in marine Chrysophyte algae (Moldowan et al., 1990).
δ 13 C values of Pr and Ph (Fig. 11h) indicate their similar source (i.e. chlorophyll), and Pr/Ph ratios may therefore be used as a redox parameter (Didyk et al., 1978). Even considering maturity effects (see above), the very low Pr/Ph ratios obtained are consistent with anoxic conditions. The observed depth trend suggests that the most oxygen-deficient conditions occurred during deposition of the middle part of the succession. The presence of PMIs, most likely including minor amounts of irregular PMI characteristic of methanogenic and methanotrophic archaea (e.g. Schouten et al., 1997;Greenwood and Summons, 2003), also supports strictly anoxic conditions. The presence of isoprenoid and HBI thiophenes, although in generally low concentrations (0.83 and 0.31 µg/gTOC), confirms the availability of free H 2 S during early diagenesis.
HI values of organic-rich sediments vary significantly (Table 2). However, S2 and TOC values correlate well (R² = 0.99; Fig. 7). The positive intercept and the slope of the regression line show significant retention of S2 hydrocarbons on the shale matrix (mineral matrix effect; Espitalié et al., 1984), and that all organic-rich samples contain a similar kerogen Type II with a true HI (sensu Langford and Blanc-Valleron, 1990) of 412 mgHC/gTOC. The homogeneity of the organic matter is also reflected by the similar isotope patterns of the n-alkanes (Fig. 11). A strong mineral matrix effect was also indicated by analyses of kerogen concentrates from interlayered low-TOC shaly rocks. Whereas whole rock samples had low HI values (avg. 59 mgHC/gTOC; Table 2), isolated kerogens yield HI values similar to those of organic-rich sediments (>400 mgHC/gTOC; Saint-Germès et al., 2002). This shows that despite the widely differing TOC contents, the type of organic matter is similar in organic-rich and organic-poor sediments. Hence, layers with high TOC contents may represent the background sediment, whereas low TOC contents may result from dilution by clastic material.
Maikop Group (Islamdag Section) The exposed part of the Maikop Group in the Islamdag section is 364 m thick. Significant progress has been achieved in age-dating this unit, and the new Re/Os age from the base of member A (30.0 ± 1.0 Ma) shows that sediment deposition in the Islamdag section began during the late Solenovian (Fig. 2). The nominal Re/Os age determined near the top of member A is 17.2 ± 3.2 Ma (Washburn et al., 2019). The Tarkhanian flooding event, which resulted in the cessation of deposition of the Maikop Group, was recently dated as 14.85 Ma (Palcu et al., 2019b). Using these age constraints, remarkably low sedimentation rates (~25 m/Ma) can be determined for the Upper Maikop Formation. The low sedimentation rates agree with the palaeogeographic setting of the study area in the deepest part of the Eastern Paratethys (Popov et al., 2004;. The Maikop Group in the Islamdag section begins with a 16 m thick interval rich in Botryococcus braunii and Batiacasphaera sp. indicating deposition in a basin with reduced salinity. This interval represents low salinity conditions of the Paratethys-wide Solenovian Event (Zaporozhets and Akhmetiev, 2015). Therefore, this interval is attributed to the Lower Maikop Formation (sensu Weber, 1935). The early Solenovian Event itself (early NP23), often characterized by carbonate-rich sediments (Polbian or Ostracoda Bed; Gavrilov et al., 2017;Sachsenhofer et al., 2017Sachsenhofer et al., , 2018a, is not exposed but may occur several tens of metres below the base of the section. The exposed sediments have varying and partly high TOC contents, but very low HI values (≤90 mgHC/gTOC; Table  2); in the case of sample IDF-07, this is despite the high alginite content (82 vol.%). This suggests poor conditions for organic matter preservation during the low salinity event, which differs from equivalent sediments at the type locality of the Maikop Group (e.g. Lower Morozkina Balka Formation; Sachsenhofer et al., 2017).
Following Weber (1935), the overlying Upper Maikop Formation is divided from base to top into members A (308 m thick), B (23 m) and C (17 m). Members A (avg. 1.82 wt.%) and C (avg. 1.71 wt.%) contain significant amounts of organic matter. Even higher TOC contents (~3-6 wt.%) were reported by Baldermann et al. (2020). Despite moderately high TOC contents, HI values are typically low (avg. 121, and 114 mgHC/gTOC), but vary significantly (28-491 mgHC/gTOC). The lack of a good correlation between S2 and TOC values (Fig. 7) indicates that the varying HI values are due to different organic matter types rather than due to a mineral matrix effect (e.g. Langford and Blanc-Valleron, 1990) or due to a constant proportion of inert organic matter (Dahl et al., 2004). The generally low HI values are probably related to the negative effect of low sedimentation rates on organic matter preservation (e.g. Stein, 1990). The presence of abundant framboidal pyrite in the sediments and high contents of elemental sulphur in the Maikop extracts indicate that bacteria of the sulphur cycle were important for OM decomposition.
Member A represents the long time interval between the late Solenovian and the Sakaraulian (approximately 13 Ma). The member displays cyclic variations in TOC contents, but the HI shows a uniform upward-decreasing trend between 18 m and 300 m from values of ~200 to ~50 mgHC/gTOC. In addition, slightly elevated TOC contents and HI values are also observed in the uppermost part of the member (303-325 m).
The initial 187 Os/ 188 Os ratios of high TOC samples with (moderately) high HI values near the base of member A (0.70 ± 0.03) and near its top (0.80 ± 0.14 and 0.56 ± 0.13, Washburn et al., 2019) are either higher or lower than that of the contemporaneous global ocean (0.61 ± 0.05 and ~0.72; Peucker-Ehrenbrink and Ravizza, 2020). This indicates that there was limited connectivity between the Eastern Paratethys and the global ocean during deposition of these sediments, and suggests that basin restriction resulted in (slightly) higher organic matter preservation as reflected by the higher HI values. The reduced salinity of the surface waters during deposition of the uppermost part of member A, suggested by the presence of Batiacasphaera sp. (Popov et al., 2008), may be another consequence of basin isolation. Salinity stratification is indicated by the presence of a fullymarine fish fauna, which occurred to depths of at least 300 to 400 m (Popov et al., 2008).
TOC/S ratios are generally very low in member A, which indicates anoxic conditions (e.g. Berner, 1984). Slightly elevated TOC/S ratios between 195 and 300 m are probably due to weathering of subsection IDG (Fig. 4d), and should not therefore be used for environmental interpretations. Weathering of the sulphur-rich sediments gives the rocks a yellowish colour in outcrop due to jarosite formation.
The black-coloured member B, 22 m thick, contains surprisingly low amounts of TOC (avg. 0.69 wt.%) and sulphur (avg. 0.28 wt.%). Whereas TOC contents do not show a depth trend, HI values decrease gradually upwards from 60 to 20 mgHC/gTOC suggesting that conditions for OM preservation were poor and deteriorated through time. This interpretation is supported by the observation of strongly corroded palynomorphs in member B (Popov et al., 2008).
With the exception of the uppermost samples, TOC/S ratios are high. This reflects low salinity conditions, at least in the surface waters, as suggested by Popov et al. (2008) based on the impoverished palynoflora dominated by Batiacasphaera sp. Low TOC contents and low HI values show that increasing basin isolation during the middle/late Kozakhurian did not result in improved OM preservation. It should be noted that, based on its black colour, member B has formerly been described as organic-rich (Weber, 1935;Popov et al., 2008). However, the new data show that the dark surface colour is due to the absence of jarosite resulting from pyrite weathering, a consequence of the very low sulphur content. Member C is generally poor in organic matter, but includes a few intervals with TOC contents of 2.1 -5.7 wt.%TOC.
Samples from Maikop sediments with high TOC contents (1.64-5.72 wt.%) and strongly varying HI values (42-491 mgHC/gTOC) were used to determine the OM type based on maceral analysis and biomarkers. High vitrinite percentages show that there was a significant input of terrigenous organic matter into the basin. This agrees with the observation of abundant wood fragments in the Islamdag section reported by Popov et al. (2008 cum lit.) as well as of silicified wood in member A. Despite this, concentrations of land plant biomarkers (mainly gymnosperm-derived diterpenoids) are low. Moreover, although angiosperm wood has been detected besides the more common gymnosperm wood (Popov et al., 2008), angiospermderived triterpenoids could not be identified.
Hop-17(21)-ene is present in all samples. Based on its slightly 13 C-depleted isotope composition (Table 5), an origin from heterotrophic bacteria is suggested. High steroid/hopanoid ratios suggest that bioproductivity in the photic zone outcompeted bacterial activity.
The n-alkane distribution patterns of most samples are dominated by long-chain n-alkanes with a marked odd-over-even predominance (e.g. IDF-07; Fig. 9), indicating the contribution of plant waxes (Eglinton and Hamilton, 1967). However, in three samples (IDB-09, IDB-23, IDG-53) near member B, different n-alkane distributions were obtained. In these samples, short-chain n-alkanes dominate (Fig. 9), indicating high contributions of alkanes derived from algae and/ or micro-organisms (Cranwell, 1977). Furthermore, the n-alkanes are depleted in 13 C compared to the rest of the samples from the Maikop Group (Table 5, Fig.  11c). The data may indicate algal or bacterial blooms in the photic zone of the water column. Short-chain n-alkanes have been found to predominate in cultures of cyanobacteria adapted to salinity (NaCl) stress (Bhadauriya et al., 2008). Significantly lower steroid/ hopanoid ratios suggest accelerated bacterial activity during OM deposition in these samples.
High concentrations of HBIs, especially at 106.90 m (IDF-39) and 188.30 m (IDF-67), indicate high productivity of marine diatoms. HBIs (and n-alkanes) in the Maikop Group from the Islamdag section are less depleted in 13 C than those in the older sediments in the Perekishkyul section. The presence of diatoms and other siliceous organisms is also indicated by the high biogenic silica contents in samples from the uppermost member A and from member C. HBIs are known exclusively from marine diatoms (Table 4). Hence, the lack of HBIs in these sediments may reflect the low salinity. Unfortunately, the MTTC ratio, a biomarker-based salinity proxy (e.g. Sinninghe Damsté et al., 1987), could not be determined. The presence of the 24-norcholestane has been used to trace the contribution of dinoflagellates or diatoms (Holba et al., 1998;Rampen et al., 2007).
The difference in isotope ratios of pristane and phytane is remarkably high (up to 3.5 ‰) in samples Fig. 11h). This suggests that the phytane peak is overlain by another isotopically-light compound, probably crocetane. Hence, the Pr/Ph ratios determined for these samples (0.11-0.18) should not be used for the assessment of redox conditions. However, the presence of HBI and isoprenoid thiophenes as well as (minor amounts of) DBT suggest highly oxygendeficient conditions for most studied samples. Minor amounts of crocetane (co-eluting with phytane in the chromatograms) and irregular PMI, as evidenced by their low δ 13 C values, suggest the presence of methanotrophic archaea beside sulphate-reducing bacteria in the anoxic part of the water column (Elvert et al., 2000).

Chokrak-Spirialis Formation
Deposition of the Maikop Group sediments ended with the Tarkhanian flooding event, which caused oxygenation of the Eastern Paratethys (Palcu et al., 2019b). The occurrence of the deep-water fish Vinciquerria merklini in the lower part of the Chokrak-Spirialis Formation suggests oxic deepmarine conditions which allowed fish populations to survive down to the depth of 700 to 1000 m (Popov et al., 2008). Consequently, TOC contents in the Chokrak-Spirialis Formation are generally low. The first occurrence of (calcareous) Limacina sp. (= Spirialis sp.) fossils several metres above the base of the Chokrak-Spirialis Formation reflects restricted connection with the open ocean and variable salinities (Popov et al., 2008).
Although the Chokrak-Spirialis Formation in general has TOC contents below 1.5 wt.%, a few samples contain more than 2 wt.%. Sample IDC-48.2 (TOC: 2.37 wt.%; HI: 214 mgHC/gTOC) from the upper part of the formation has been analyzed in detail. This sample contains the highest amount of vitrinite (64 vol.%) within the studied sample set and a very high proportion of long-chain n-alkanes (57 %) with a pronounced odd-even predominance (CPI: 5.10). This demonstrates a high proportion of terrigenous organic matter influx, and low amounts of HBI alkanes reflect the (minor) diatom biomass. The very low Pr/Ph ratio (0.18), the presence of PMI and isoprenoid thiophenes, although in low amounts, together with a TOC/S ratio (0.91) which is significantly lower than the average for the formation (2.18), indicate short-term events with strongly oxygen-depleted bottom waters.

Diatom Formation (Paper Shale)
The lithology of the Diatom Formation justifies its subdivision into a lower interval (429-629 m), the middle paper shale interval (590-629 m), and an upper interval (629-750 m). In the lower and upper intervals of the Diatom Formation, TOC contents are very low indicating depositional environments which were not suitable for organic matter accumulation. This also applies for the basal layers, which contain an endemic fish fauna (e.g. Sardinella karaganica, Mugil karaganicus) indicating deposition in the Karaganian basin with reduced salinity (Popov et al., 2008;Palcu et al., 2019b). The slightly increased TOC/S ratios (Fig. 5) in the basal part of the Diatom Formation may result from this freshening event.
Detailed investigations were focused on the paper shale interval and its lower and upper transition zones. Although, the Re/Os age of the paper shale has significant uncertainty (7.2 ± 2.6 Ma; Fig. 6B), it nominally indicates deposition in the late Tortonian/ early Messinian; but including the uncertainty, the age range is mid-Tortonian to early Zanclean. However, the absence of Pontian sediments exclude an age younger than 6.1 Ma (van Baak et al., 2016). Irrespective of the precise depositional age, the initial 187 Os/ 188 Os (0.80 ± 0.02) is slightly lower (less radiogenic) than that observed in the late Miocene global 187 Os/ 188 Os oceans (~0.85; Peucker-Ehrenbrink and Ravizza, 2020), indicating restriction of the Kura Basin during deposition.
Reflected light microscopy shows that organic matter is dominated by alginite macerals (62-90 vol.%). Lamalginite (including liptodetrinite) is typically more common than telalginite, which dominates over lamalginite only in the lowermost two samples. This suggests that the organic matter is mainly composed of algal or bacterial mats. The relative amount of short-chain n-alkanes is negatively correlated with δ 13 C values of n-C 19 (R² = 0.54) suggesting that shortchain n-alkanes derived from algal organic matter are isotopically light.
Diatoms are observed frequently in reflected light and SEM photomicrographs (Fig. 8). Considering the abundance of the diatom remains, the concentration of biogenic silica is low (Fig. 8). HBI compounds, characteristic of some (usually marine) diatom assemblages, were not detected (Fig. 10). However, high abundances of C 28 steradienes which are structurally similar to 24-methylcholesta-5,24(28)dien-3β-ol and 24-methylcholesta-5,22-dien-3β-ol (brassicasterol) indicate the contribution of diatoms to bioproductivity. The lack of C 25 HBI alkane suggests differences in diatom assemblages compared to the organic matter in the Upper Maikop Formation.
Apart from aquatic liptinite macerals, detrital vitrinite macerals also occur in significant amounts (6.6 -35.0 vol.%). The relative amounts of long-chain n-alkanes, which correlate well with CPI values (R² = 0.83 or even 0.90, if sample IDD-52 above the paper shale is neglected) are geochemical indicators for the terrigenous fraction. Thus, it is not surprising that vitrinite percentages correlate positively (R² = 0.50) with both the amount of long-chain n-alkanes and with the CPI. Concentrations of gymnosperm-derived diterpenoids are low. Triterpenoids are not measurable, except arborene which is of controversial origin (e.g. IDD-10); arborane-type triterpenoids may be derived from angiosperms, grasses or bacteria (Hauke et al., 1992). The 13 C-enriched isotopic composition (-24.9‰; Table 5) of the C 30 arborene measured in sample IDD-10 does not support a microbial origin.
PMI and crocetanes occur in enhanced amounts, suggesting the importance of archaea involved in methane cycling (e.g. anoxic methane oxidizers). The diminishing content of 24-n-propylcholestane, and lower abundances of framboidal pyrite and elemental sulphur in the extracts provide evidence for decreased activity of sulphur cycle bacteria and probably also brackish-water conditions due to freshwater inflow. Reduced salinity, probably related to the limited connectivity between the basin and the global ocean, is also suggested by high TOC/S ratios (avg. 5.24; Table 2). In contrast to sediments of the Maikop Group, C 30 4-methylsteranes are present in the rock extracts and are most probably derived from dinoflagellates (Robinson et al., 1984).
The dominance of hydrogen-rich algal material is also reflected by the slope of the regression line in the cross-plot of S2 versus TOC (Fig. 7), which shows that the reactive organic matter (sensu Dahl et al., 2004) in all paper shale samples has an average HI of 764 mgH/ gTOC (kerogen Type I/II). The positive intercept of the regression line shows that the inert organic matter fraction is about 1.1 wt.%. Hence, varying HI values reflect different relative contributions of algal and inert organic matter.
Strictly anoxic conditions during deposition of the paper shale interval are proven by the preservation of sediment lamination, the excellent preservation of organic matter, and biomarker proxies (e.g. PMI, crocetane). Hence, oxic conditions, as postulated recently by Abdullayev et al. (2021) based on redoxsensitive elements, can be excluded.

Petroleum Potential
In this section, the petroleum potential of the sediments studied in the Perekishkyul and Islamdag sections is discussed using cross-plots of the generation potential (Rock-Eval S1+S2) versus TOC content (Fig. 12). The potential of stratigraphic units which are not exposed in the study area will be evaluated in a following section below.
Paleocene sediments in the Perekishkyul section contain moderately high TOC contents (avg. 0.90 wt.%) with very low HI values. Although this unit contains surprisingly high percentages of alginite, it does not hold any commercial source-rock potential. The same applies for the Middle Eocene I sediments which are exposed in the same section (see Fig. 12).
Within the upper part of the Perekishkyul section (Middle Eocene II), very good to excellent petroleum potential is limited to cm-thick, organic-rich intervals that occur within dm-thick, low-TOC successions. The determined high (true) HI of 412 mgHC/gTOC (Fig. 7) shows that the sediments are oil prone. As mentioned above, the cumulative thickness of the organic-rich layers is difficult to determine but probably does not exceed 2 m. Assuming a net thickness of 2 m and an average density of 2.3 t/m³ for the largely carbonatefree sediments results in an SPI value of 0.11 t/m². This shows that the hydrocarbon potential of the middle Eocene in the study area is very low.  Based on the TOC contents, a good to very good petroleum potential can be attributed to members A and C in the Islamdag section, whereas member B has only fair potential (Fig. 12). However, as the generation potential (S1+S2) is typically low (Table  2), the Maikopian sediments have only fair to good petroleum potential. This discrepancy between high TOC contents and low generation potential is reflected by low HI values indicating the prevalence of Type II-III kerogen. Hence, the Maikopian sediments in the Islamdag section are oil-to gas-prone. The SPI (Demaison and Huizinga, 1991) has been calculated for members A and C, while member B is not considered a potential source rock of significance. Assuming an average density of 2.3 t/m³, the SPI is determined as 1.98 and 0.12 tons of hydrocarbons/m² for members A and C, respectively. Thus, the upper Solenovian to lower Miocene part of the Maikop Group may generate a total of 2.10 tHC/m². The hydrocarbon potential of the Pshekhian and lower Solenovian part of the Maikop Group remains unknown. However, data from the Siyaki (Pshekhian) and Shikhzairli (lower Solenovian) sections located to the west of Mount Islamdag (Saint-Germès, 1998) suggest that the Pshekhian and lower Solenovian could probably contribute an additional 0.28 and 0.05 tHC/m² to the entire Maikop Group. Thus, the Maikop Group in the Islamdag area may generate about 2.5 tHC/m². This total is of the same order as that suggested by Sachsenhofer et al. (2018b) for the Maikop Group in Azerbaijan.
The Chokrak-Spirialis Formation deposited after the Tarkhanian flooding event contains layers with high TOC contents, and a few samples can be classified as good source rocks (Fig. 12). However, the average generation potential is low (0.82 mg HC/g rock) and the formation is not considered a potential source rock.
In the Diatom Formation, whereas the intervals below and above the paper shales cannot be considered as potential source rocks, the paper shale interval itself (590-629 m) comprises highly oil-prone source rocks with good to excellent petroleum potential (Fig. 12). The SPI of the paper shale interval has been calculated for rock densities of 2.0 t/m³ (2.89 tHC/m²) and 2.3 t/m³ (3.33 tHC/m²). The values obtained show that independent of the applied rock density, and despite the significantly lower thickness of the paper shale, its hydrocarbon potential is higher than that of the Maikop Group. Considering the differences in generation potential, the expulsion efficiency of the paper shale is also probably much higher than that of the Maikop Group (cf. Mackenzie and Quigley, 1988).

Distinguishing the oils generated by the Maikop Group and the Diatom Formation
It is widely accepted that the Maikop Group and the Diatom Formation are the main source rocks in the South Caspian Basin (e.g. Wavrek et al., 1998;Smith-Rouch, 2006;Goodwin et al., 2020), although it is not clear whether the Diatom Formation has reached oil window maturity (Inan et al., 1997). Therefore, it is useful to check if isotope and/or biomarker data allow a distinction to be made between the oils generated by these stratigraphic units.
A systematic decrease of δ 13 C values of the saturated and aromatic hydrocarbon fractions with stratigraphic age has been reported by Abrams and Narimov (1997). Whereas extracts from Lower Oligocene source rocks are isotopically light (δ 13 C sat : ~-28‰; δ 13 C aro : ~-28‰), extracts from the Diatom Formation are relatively enriched in 13 C (δ 13 C sat : ~-23‰; δ 13 C aro : ~-24‰). Using compound specific isotope data of n-alkanes, a similar trend was observed for the lower and upper Oligocene sediments at Angeheran (for location see Fig. 1; Bechtel et al., 2013).
Lowermost Oligocene sediments are not exposed in the studied sections. Hence, the complete stratigraphic succession of the Maikop Group could not be investigated. However, results from the part of the Maikop Group exposed in the Islamdag section (upper Solenovian to lower Miocene; Fig. 11b-d) show more differentiated results with strongly varying compound specific isotope data, especially in the lower Miocene (Kozakhurian) sediments (Fig. 11c). Moreover, no systematic difference in the δ 13 C isotopic composition of n-alkanes exists between the investigated upper Oligocene to lower/middle Miocene samples from the Maikop Group and the Diatom Formation ( Fig.  11e-f). Hence, it might be possible to distinguish middle Eocene oils from upper Solenovian to Miocene oils, but based on isotope data alone it is impossible to distinguish oils derived from the Upper Maikop Formation and the Diatom Formation.
In contrast to the carbon isotope data, there are differences in the presence of specific biomarkers between the samples of the Maikop Group and the Diatom Formation paper shales. The investigated samples from the Diatom Formation contain C 30 4-methylsteranes in elevated contents and no C 30 steranes, whereas the rest of the sample set is characterized by the presence of 24-n-propylsterane with lower amounts of methylsteranes. The generated oils should therefore differ in the presence of methylsteranes (paper shales) versus the presence of additional C 30 steranes (Maikop Group, Middle Eocene II). Furthermore, the C 25 HBI compounds (alkane and thiophenes) were exclusively found in the extracts from samples of the Maikop Group (and the Middle Eocene II), and early-generated oils from the Maikop Group should therefore include the C 25 HBI alkane. The oils derived from the Diatom Formation may be also distinguished by the higher relative contributions of the C 25 isoprenoid (PMI).

Implications for source rock distribution in Azerbaijan and the Eastern Paratethys
The Perekishkyul and Islamdag sections provide information on the stratigraphic distribution of source rocks in Paleocene and middle Eocene intervals and in a largely continuous lower Oligocene to upper Miocene succession. However, there are gaps in sedimentation between the Paleocene and middle Eocene in the Perekishkyul section, as well as between the middle Eocene and the upper Solenovian in the Islamdag section (Fig. 2). The possibility of the presence of organic-rich horizons within these identified stratigraphic gaps is discussed in this section, together with the lateral distribution of organic-rich intervals with the upper Middle Eocene II and the paper shale interval in the Diatom Formation.
With the exception of thin layers in the upper part of the middle Eocene (Middle Eocene II), Paleocene and Eocene sediments in the Perekishkyul section are poor source rocks. However, three prominent black shale horizons may be present in the Paleocene to Eocene succession that appear to be missing in the studied sections. These include sediments deposited during the Paleocene/Eocene thermal maximum (PETM) and the Early Eocene Climatic Optimum (EECO), as well as the middle Eocene Kuma Formation (see Fig. 2 for stratigraphic positions).
PETM sediments deposited in the northeastern Peri-Tethys are often only a few decimetres thick, but can be traced from Crimea to Central Asia (Gavrilov et al., 2003;2018). In Crimea and the northwestern Caucasus foreland, these sediments contain low amounts of organic matter (Gavrilov et al., 2018). Elsewhere, including in the Rioni Basin in Georgia and in the central and northeastern Caucasus foreland, they are rich in organic matter and TOC contents may exceed 10 wt.% (Gavrilov et al., 2003;. HI values may reach 500 mgHC/gTOC near the base of the sapropelic interval but are more typically low (Gavrilov et al., 2009;Shcherbinina et al., 2016). Similarly, sapropelic intervals related to the Early Eocene Climatic Optimum (EECO), 7 to 30 cm thick, occur in the marly lower Eocene (Ypresian) succession in the northern Caucasus (e.g. Shcherbinina et al., 2020). Although it is likely that both horizons are present near the eastern end of the Greater Caucasus, to the authors' knowledge neither PETM nor EECO sediments have been described in Azerbaijan.
The middle Eocene Kuma Formation comprises organic-rich marly rocks which were deposited in a dysoxic to anoxic basin extending from Crimea to the Aral Sea (Beniamovski et al., 2003). It has been suggested that volcanic material contributed to the increased organic matter content (Vincent and Kaye, 2017). The succession, several tens of metres thick, is a prolific hydrocarbon source rock in Georgia and southern Russia (Sachsenhofer et al., 2018b;2021 this issue;Oblasov et al., 2020). Time equivalents of this formation have likewise not yet been described in Azerbaijan. "Oil shale layers" within the middle part of the Koun Formation mentioned in previous publications (e.g. Abbasov, 2005) may correlate with the Kuma Formation. However, the available data are not adequate to judge either the age of the sediments or the amount and type of organic matter. The Eocene part of the Perekishkyul section is generally poor in organic matter, but includes cm-thick layers with high TOC contents. As new radiometric age data suggest a middle Eocene age, these layers may represent equivalents of the Kuma Formation. In contrast to the Kuma Formation, the organic-rich layers in the Perekishkyul section are thin and interbedded with largely carbonatefree claystones, whereas marly interbeds are restricted to the upper part of the succession. Similar sediments, about 14 m thick, have been described from outcrops 5 km south of the Sumgait railway station (Saint-Germès, 1998; for location see Fig. 1). This suggests that the interval containing thin organic-rich layers within the Middle Eocene II unit is laterally persistent. The net thickness of the organic-rich layers is difficult to determine but is probably below 2 m, suggesting that the source rock potential is minor (~0.1 tHC/m²). However, there may be an increase in the thickness of these layers towards the South Caspian Basin.
Another major stratigraphic gap exists between the top of the Middle Eocene II (~38 Ma) and the base of the Islamdag section (~30 Ma; Fig. 2). The lowermost carbonate-free sediments in the Islamdag section were deposited following the low-salinity "Solenovian Event" (e.g. Popov et al., 2004). Elsewhere in the Eastern Paratethys, the Solenovian Event is represented by a carbonate-rich layer (e.g. the Ostracoda (= Polbian) Bed) overlain by shaly sediments that have the highest hydrocarbon potential of the entire Maikop Group (Sachsenhofer et al., 2017;2018a,b). Until now, the Ostracoda Bed has not been described in the Lower Kura Basin of Azerbaijan, but a short section (~7 m) with low carbonate sediments attributed to the Solenovian Event based on palynological data has been described from an outcrop 2 km south of Shikhzairli (Akhmetiev et al., 2007; for location see Fig. 1). TOC contents (0.33-0.73 wt.%) and HI values (61-182 mgHC/gTOC) of these samples are low (Saint-Germès, 1998).
Source rock parameters of lower Oligocene sediments were also investigated near Siyaki, Angeheran and Lahich. However, none of these sections provided evidence for the presence of the Solenovian Event (Saint-Germès, 1998;Bechtel et al., 2013Bechtel et al., , 2014. Thus, the source rock potential of (Pshekhian and) lower Solenovian sediments in the Lower Kura Basin of Azerbaijan remains unclear.
Major parts of the lower Oligocene (upper Solenovian) to middle Miocene (Kozakhurian) part of the Maikop Group have fair to good petroleum potential. It is not disputed that these sediments are of wide lateral extent. Moreover, higher sedimentation rates in the South Caspian Basin to the east may have resulted in improved organic matter preservation and the increased generation potential of the Maikop Group.
The thickness of the Maikop Group in the Eastern Paratethys varies significantly. For example, its thickness increases from about 600 m in the Sea of Azov to more than 4 km in the Taman peninsula (Popov et al., 2019). Similarly, the thickness of the Maikop Group increases dramatically from onshore areas to the South Caspian Sea (Green et al., 2009). However, thickness variations are also observed within the Lower Kura Basin. Lower Oligocene sediments, about 700 m thick, occur near Lahich ( Fig. 1a; Bechtel et al., 2014). In contrast, the thickness of the lower Oligocene succession near Angeheran is only some 175 m. The lower Oligocene succession at these sites was deposited in a strongly oxygen-depleted environment (Bechtel et al., 2013(Bechtel et al., , 2014. Consequently, the source rock quality is relatively good (~3 wt.% TOC; HI ~300 mgHC/gTOC). In contrast, upper Oligocene rocks at Angeheran, more than 250 m thick, are poor source rocks (Bechtel et al., 2013). This may indicate that the source rock quality of upper Oligocene (and lower Miocene) units increases eastwards.
The exposed part of the Diatom Formation in the Islamdag section is 320 m thick. Most of the sediments have low TOC contents (429-629 m; 688-750 m), but a 59 m thick interval of paper shales with very high TOC contents and HI values occurs about 250 m above the base of the formation. Both Khersonian and Maeotian ages for these black shales are possible. The great variability of source rock properties of samples of the Diatom Formation has been noted previously (Feyzullayev et al., 2001;Alizadeh et al., 2017). However, based on the characteristics of rock samples ejected from mud volcanoes in the Shamakhy-Gobustan area and of cores from boreholes located south of Baku (Baku archipelago), Feyzullayev et al. (2001) assumed that the petroleum potential of the Diatom Formation increases with depth and to the south (see also Alizadeh et al., 2017). Hence, marked variations in petroleum potential were interpreted in terms of lateral facies variability. The presence of a single organic-rich black shale horizon within organicpoor sediments of the Diatom Formation, which is evident in the Islamdag section, has seemingly so far been overlooked. Middle Miocene "oil shales" with Konkian to Maeotian ages were mentioned by Abbasov (2015). Furthermore, it is possible that the 14 m thick succession at Gezdek with TOC contents exceeding 10 wt.% and HI values up to 800 mgHC/gTOC, described as consisting of lower Miocene Maikopian sediments (Katz et al., 2000), belongs to the Diatom Formation. For an improved understanding of the petroleum system, it will be important to understand the lateral distribution of these intervals.
No new source rock data are available for finegrained intervals within the lower Pliocene Productive Series. However, Alizadeh et al. (2017) showed that TOC contents (average 0.47 wt.%; max. 2.71 wt.%) and HI values (average 147 mg HC/g rock; max. 334 mg HC/g rock) are typically low.

CONCLUSIONS
This study of Paleocene to Miocene sediments exposed in the Perekishkyul and Islamdag sections in the Gobustan area west of Baku provides new information on the distribution and quality of source rocks in eastern Azerbaijan and the South Caspian Basin. In addition, it helps us to understand changes in the regional palaeodepositional environment during Oligocene and Miocene times. The very low thermal maturity of the samples analysed may influence some biomarker ratios (e.g. the Pr/Ph and diasterane/sterane ratios) which can be used as environmental proxies. The most important results of the study are as follows: The upper part of the middle Eocene succession (Middle Eocene II) includes an approximately 40 m thick interval composed of alternations of decimetrethick organic-poor layers and cm-thick organic-rich layers. The latter contain very large amounts of kerogen Type II with abundant marine organisms. Similar alternations have been described from several outcrops and are probably of significant lateral extent. Because of the low net thickness of the organic-rich layers, the middle Eocene cannot be considered as a significant source rock in onshore eastern Azerbaijan.
The exposed part of the Maikop Group in the Islamdag section is 364 m thick. A new Re/Os age from the base of the Group in this section (30.0 ± 1.0 Ma) shows that the exposed sediments were deposited during early Oligocene (late Solenovian) to middle Miocene times, and that sedimentation rates were low (~25 m/Ma). Low sedimentation rates related to the deep-marine setting promoted organic matter degradation. The lowermost sediments represent the uppermost layers of the Solenovian Event. TOC contents are often high, but HI values are typically low (kerogen Type II/III). The Maikop Group (including the Pshekhian and lower Solenovian units) could generate about 2.5 tHC/m². Apart from the dominant aquatic micro-organisms (including diatoms, methanotrophic archaea and sulphate-reducing bacteria), the organic matter includes varying amounts of terrigenous land plant material.
The Diatom Formation includes a paper shale interval, about 60 m thick, with very high TOC contents (max. 13.3 wt.%) and HI values (max. 770 mgHC/ gTOC; kerogen Type II-I). This interval may generate more hydrocarbons than the Maikop Group (~3 t/m²). The organic material is dominated by algae (including diatoms) deposited in a basin with reduced salinity. Strictly anoxic conditions are indicated by the presence of biomarkers for archaea involved in methane cycling. Dinoflagellates also contributed to the biomass.
Carbon isotope data allow a distinction to be made between oil generated from middle Eocene source rocks, and that from upper Solenovian to upper Miocene source rocks. Oil generated from the Maikop Group and the paper shales within the Diatom Formation can be distinguished based on the presence of specific biomarkers. For example, oils generated from the Maikop Group should include C 30 steranes and C 25 HBI compounds (alkane and thiophene) that are absent in oils generated by the paper shales. The oils derived from Diatom Formation may be also distinguished by the higher relative contributions of the C 25 isoprenoid (PMI) and of C 30 methylsteranes.