The distributions of C₂₀ tricyclic diterpanes in oils from source rocks of varying age and depositional setting are shown in partial m/z 276→123 chromatograms in Figures 2–4. Traditional indicators of gymnosperm contributions—5β(H)-rimuane, 5β(H)-rosane and isopimarane—are only abundant in oils originating from coals and some of the deltaic and lacustrine shales (m/z 276→123 chromatograms in Figure 2). The relative abundance of individual diterpanes can vary significantly, depending upon the major families of contributing gymnosperms (Killops et al., 1995; Otto and Wilde, 2001; Diefendorf et al., 2019), as demonstrated by the dominance of either the tetracyclic 16β(H)-phyllocladane (βPhyl) or the tricyclic isopimarane (iPim) in the coal-sourced oils in Figure 2. The variability in abundance of gymnosperm markers in oils associated with deltaic and lacustrine sources in Figure 2 is most likely controlled by transport of plant debris to the depositional site and its subsequent degree of preservation.

Abietane (Ab) appears to be present at low levels in the coal-sourced oils and is seemingly the main component in the corresponding small peaks in other samples, based on multiple GC-MS-MS transitions for eight oils of varying source age and environment (7, 8, 17, 18, 20, 35, 44, and 51; Table 1), which were selected to test the accuracy of peak identifications based on m/z 276→123 responses. The presence of abundant 5β(H)-rimuane, 5β(H)-rosane and isopimarane in the pair of Late Cretaceous coal sourced oils (7 and 8) was confirmed. Under the conditions employed, the m/z 276→247 response factors relative to m/z 276→123 were ∼100% for 5β(H)-rimuane and 5β(H)-rosane, but ∼60% for isopimarane. These response factors indicate that small amounts of all three diterpanes are present in NSO-1 (oil 35, Figure 5). The oil from a Paleogene deltaic shale (17, Figure 5) contains a significant amount of 5β(H)-rosane, but only traces of 5β(H)-rimuane and isopimarane. It has a m/z 276→123 peak corresponding to iPim that is larger than expected from the m/z 276→247 response (Figure 5), suggesting a co-eluant is present. Small peaks at the relative retention times of the three diterpanes can be observed in the m/z 276→123 mass chromatograms of the Neoproterozoic-Cambrian (18 and 44) and Ordovician oils (20 and 51) in Figure 5, but there is no corresponding m/z 276→247 signal, suggesting that other compounds are responsible. Consequently, there is no evidence of the presence of 5β(H)-rimuane, 5β(H)-rosane or isopimarane among the study oils from sources older than Devonian.

A small signal at the retention time of 5β(H)-rosane in oil sample 2 (Figure 2), even if genuine, could derive from the Mid Jurassic marine shale source component rather than the Devonian lacustrine. Similarly small signals are recorded for Late Devonian oil 24 and Late Devonian-Early Carboniferous oil 26, both related to marine shales (Figure 3), and also for Late Carboniferous oil 13 originating from deltaic shale (Figure 2) may be attributable to 5β(H)-rosane, but on its own it cannot be considered conclusive, given the potential origin from rearrangement of pimaroids (Otto and Wilde, 2001). However, dominance of 5β(H)-rosane among gymnosperm-derived diterpanes in oils is not unknown, as evinced by some oils and associated coals within Paleocene-Eocene strata in the Assam Basin, E. India (Rudra et al., 2017). Whether it is characteristic of Devonian coal-forming flora remains to be established.

Oils from marine shales, carbonates and marls of all ages contain much lower abundances of a fairly uniform range of mostly unidentified tricyclic diterpanes in m/z 276→123 chromatograms (G–M, X and Y) than those associated with coals, with limited occurrences of gymnosperm indicators as minor components (Figures 3, 4). These oils mostly exhibit a dominance of tricyclics over tetracyclics; the latter can be below the detection threshold. The C₂₀ 13β(H),14α(H)-cheilanthane (20C) is often relatively abundant, particularly so in the carbonate/marl oils. This is not surprising, because cheilanthanes in general are known to be abundant in high salinity environments, such as hypersaline lacustrine and marine carbonate/evaporite settings (de Grande et al., 1993). There is potential for a contribution of 20C from bacterial isoagathene-containing diterpenoids, additional to that from the source(s) of the wider cheilanthane series. This may account for the variation in the relative abundance of 20C among the tricyclic diterpanes in oils from marine shales (Figure 3) as well as enrichment in oils from evaporitic sources (Figure 4).

Peaks H, J, M and 20C are usually the most abundant in Figures 3, 4, suggesting that the diterpanes either derive from the same widespread group of organisms, among which bacteria are obvious candidates, or represent the most stable products of diagenetic alteration of what may be a wide range of diterpene precursors from an equally wide range of organisms. Mass spectra of the previously unidentified tricyclics and possible structural inferences are discussed in Section 3.2.3. Among tricyclics known to synthesized by bacteria are cassane, cleisthantane and noscomane (Figure 1B), but neither their mass spectra nor their GC relative retention times appear to have been reported.

Peak H approximately corresponds to the relative retention time of pimarane (Pim), as reported by Zinniker (2005) but its mass spectrum (Figure 6) lacks the characteristic m/z 247 fragment ion. Minimal m/z 276→247 responses indicate that pimarane is no more than a trace component in any of the eight samples subjected to detailed GC-MS-MS analysis (Figure 7A), including the coal-associated oils. So, despite the ubiquity of pimaroid-type skeletons among known microbial diterpenes, pimarane seems not to be a diagenetic product in most sedimentary environments. Zinniker (2005) observed only very low relative abundance of the compound from the m/z 276→247 transition for a Beaufort Sea, coal-sourced oil, but higher relative abundance in the Merced Formation and oil from the San Joaquin Valley (North California). Another rare report of relatively abundant pimarane (again from the m/z 276→247 transition) is in a Late Permian Mongolian coal (Peters et al., 2005). Pimarane gives the lowest response factor for the m/z 276→247 transition of 5β(H)-rimuane, pimarane, 5β(H)-rosane and isopimarane (Zinniker, 2005), but m/z 123 responses from traditional selected ion recording GC-MS analysis are fairly similar for these four diterpanes. Small peaks in the relative retention window of H/pimarane can be observed in such m/z 123 mass chromatograms of coaly samples in the literature, often slightly broader than expected for a single compound, as observed in this study. An example is provided by an Early Eocene, NE Indian coal (Figure 9 of Chattopadhyay and Dutta, 2014). This may be caused by closely eluting peaks H* and H we observe in oils from coaly sources (Section 3.2.3).

Among possible diagenetic transformations of diterpane precursors which may explain the virtual absence of pimarane is acid-catalyzed rearrangement (Hall and Oehlschlager, 1972; Zinniker, 2005). Cyclization of pimarene via carbocation rearrangement to yield tetracyclics such as beyerene is a known biosynthetic pathway (e.g., scheme 10 of Peters, 2010), which might have an equivalent among potential diagenetic modifications.

A factor possibly impacting the range and abundance of C₂₀ tricyclic diterpanes observed in oils and rock extracts is competition between aromatization and reduction of precursors. Anoxia tends to favour reduction to diterpanes, whereas oxidizing conditions generally leads to aromatization via oxidation and decarboxylation/dehydration (Diefendorf et al., 2014 and references therein). Determining the extent to which varying degrees of aromatization may affect diterpane distributions in oils is problematic, because specificity tends to be lost with methyl group elimination during aromatization. Solvent extract studies of the Zeit Formation (Eocene, Saxony) have provided evidence for the preferential diagenetic formation of diterpanes in conifer fossils, whereas aromatic species dominate the host sediment (Otto and Simoneit, 2001). Our results are consistent with this model in that diterpanes are most abundant where coaly sources, rich in preserved gymnosperm macerals, are involved.

There is no clear evidence that alteration of triterpenoids is a source of the ubiquitous unidentified tricyclics. The similar distributions in the relatively immature Latrobe tasmanite analysed by Greenwood and George (1999) to those in the diverse oils in our study suggests the absence of significant rupture of ring systems of alkyl side-chains during catagenesis. Although hopanoid origins would satisfy the requirement for widespread occurrence of precursors, these compounds preferentially undergo cleavage of the C ring at the 8(14) C-C bond, and ultimately tending to yield the dicyclic members of the drimane family (Schmitter et al., 1982; Alexander et al., 1984; Wang et al., 1990). Any A,B,C-ring fragment that might be formed would likely share the basic isoagathane/cheilanthane structure.

To help facilitate the attribution of biological sources to previously unidentified diterpanes, full mass range GC-MS analysis of oil sample 19 (Table 1) was performed; this oil having one of the highest relative abundances of these compounds among the study samples. It should lack any contribution from gymnosperm-derived diterpanes, because it is believed to originate from an Ediacaran anoxic marine shale, based on biomarker characteristics such as: C₂₉/(C₂₇–C₂₉) steranes 0.90; Pr/Ph 0.99; 28,30-dinorhopane/17α-hopane 0.91; DBT/P 0.09; 24-iso-propyl/24-n-propyl cholestanes 0.96; A-ring methylated steranes dominated by 3-methyl isomers; and lack of detectable dinosteroids (aliphatic and triaromatic). A comparison of the total ion current (TIC) and m/z 276 and 123 chromatograms from this analysis with the corresponding GC-MS-MS m/z 276→123 chromatogram is shown in Figure 8. Mass spectra for compounds G–N are shown in Figure 6. These compounds exhibit effectively identical relative retention times and mass spectra to the C₂₀ tricyclic diterpanes reported in Latrobe tasmanite by Greenwood and George (1999), whose labelling system is used here to aid comparison. Peak N (as shown in Figure 8) corresponds to the C₂₀ 13β(H),14α(H)-cheilanthane (20C) and J is the previously proposed 8β-methyl-13α-ethylpodocarpane (8M13EPod; Wang and Simoneit, 1995). In oils from younger sources, J can contain a pair of co-eluting compounds, the resulting composite mass spectrum seems to explain why it was not considered to represent 8M13EPod in the tasmanite analyzed by Greenwood and George (1999). The other compounds remain unidentified, although suggestions of possible structural elements to account for major fragment ions have been made by Greenwood and George (1999).

None of the mass spectra in Figure 6 exhibit fragment ions corresponding to the facile loss of ethyl (m/z 247) or iso-propyl (m/z 233) groups from the C ring, which characterize the pimarane and abietane structural types, respectively (Zinniker, 2005). Molecular ions at m/z 276 and loss of a single methyl group (m/z 261) are observed in all of the mass spectra, often accompanied by abundant fragment ions at m/z 123, 163 and 191. Such a pattern could suggest an A ring structure identical to that of pimarane/isopimarane (Zinniker, 2005), with a pair of methyl groups at C-4 and one at C-10 (Figure 1A). Among possible candidates are cassane, cleisthantane, noscomane and staminane (Figure 1B), although their mass spectra are unknown. The A-ring fragment from such structures is represented by m/z 123, whereas m/z 191 can originate from an A-B fragment, as in cheilanthanes (Heissler et al., 1984), or from a B-C fragment (involving neutral loss of the A ring from the molecular ion), as in the 8β-methyl-13α-n-alkylpodocarpanes (Wang and Simoneit, 1995). In the latter series, further loss of the C-13 alkyl group from the C ring of the m/z 191 fragment, as shown in Figure 1C, results in the sequence of major ions at m/z (135+n14), where n = 0, 1, 2, 3 …, confirming that the alkyl group on the C-ring forms part of this fragment (e.g., in the C₂₀ member, n = 4, resulting in a m/z 191 fragment ion). This process has been suggested to account for the m/z 163 ion in the mass spectra of pimarane/isopimarane (Noble, 1986; Zinniker, 2005). A likely contributing factor to the significant m/z 247 response in pimarane/isopimarane is the stability of the tertiary ion (involving the methyl group at C-13) formed upon ethyl loss.

Interpretation of diterpane mass spectra is complicated by the profound influence of the position of substituents and stereochemistry, as demonstrated by a base peak at m/z 123 in 8β,13α-dimethylpodocarpane but at m/z 191 in 16-nor-isoagathane (Aquino Neto et al., 1983), these being C₁₉ compounds differing only in the position and stereochemistry of a C-ring methyl group. Stereochemical differences alone can result in variations in major fragment ion responses in cheilanthane isomers, with the m/z 191/123 ratio being significantly greater in the 13β-methyl than the 13α isomers (Aquino Neto, 1986). Adjacent 8β-methyl and 14β-alkyl groups on the C-ring yields a m/z 191 base peak, regardless of whether or not there is a 13α-methyl group (e.g., isoagathane, N, Figure 6), whereas the presence of both 8β-methyl and 13α-alkyl groups, without an alkyl group at C-14, results in a m/z 123 base peak (e.g., 8β-methyl-13α-ethylpodocarpane, J, Figure 6). Because of such complexities, our suggestions of structures compatible with the mass spectra in Figure 6 are limited.

It is possible that the compounds responsible for peaks I and M are isomers of J. Peak K may be structurally related to M, although the dominance of the m/z 123 fragment might be explained by its formation from both A and C rings, such as appears possible in noscomane (Figure 1B). Peak L has major fragment ions at m/z 149 (base peak), 137 and 177 (Figure 6), which might result from a similar structure to I and M, but in which a methyl group from the B-C rings is relocated to the A ring (Figure 1C). The m/z 123 fragment from L is of relatively low abundance compared to the other major diterpanes, so the compound is quantitatively of more significance than suggested by its signal in the m/z 276→123 transition (Figure 8). Peak G may be affected by co-elution. Its mass spectrum is dominated by fragment ions at m/z 137 and 191 (Figure 6), which might be explained by a structure similar to L. Although it has a major m/z 191 fragment ion, it elutes too early to represent one of the four potential isomers of the C₂₀ cheilanthane (Chicarelli et al., 1988), as previously noted by Greenwood and George (1999).

Two very closely eluting C₂₀ tricyclanes are commonly present in peak H, resulting in some broadening of m/z 123 SIR and 276→123 MS-MS responses. The slightly earlier eluting of the pair of compounds (H*, Figure 7A), characterized by an intense m/z 276→163 signal, is more prominent in the Late Cretaceous, coal-sourced oils (7 and 8), but a definitive mass spectrum could not be obtained because of its low abundance and additional co-elution problem with a C₁₉ sterane. The later-eluting compound (H of Greenwood and George, 1999) dominates in the Neoproterozoic-Cambrian oils (18 and 44). The other four oils are consistent with intermediate mixtures (Figure 7B). Consequently, care is required when attributing an identity to the peak in m/z 123 or m/z 276→123 chromatograms. From the data provided by extended GC-MS-MS analysis, the major peaks M and 20C appear to result from single compounds, without any obvious co-elutants.

Peak J elutes on the leading edge of 5α(H),14β(H)-androstane (14βAnd; Bender et al., 2015), when the latter is present, as in the Neoproterozoic-Cambrian samples. In such samples it corresponds to 8β-methyl-13α-ethylpodocarpane, based on its reported relative retention time and mass spectrum (Figures 6, 8A, respectively; Wang and Simoneit, 1995). However, in the coal-sourced oils a different, unidentified compound is present, in which m/z 163 dominates the transitions monitored and the m/z 123 and 261 responses are depressed, as in the mass spectrum reported for J by Greenwood and George (1999). Reliable mass spectra could not be obtained for the early eluting peaks X and Y.

The C₂₀ tetracyclic diterpanes present a simpler distribution pattern than the tricyclics. In oils sourced by coals and some from Tertiary deltaic shales (16 and 17) in which gymnosperm contributions are inferred to be significant, tetracyclic diterpanes are at least as abundant as tricyclics and often dominated by 16β(H)-phyllocladane (βPhyl) in Figure 2. The second eluting of the pair of (16R)-14,15-cyclo-8,14-seco-atisane isomers (csAti2) identified by Zinniker (2005) appears to be present, at very low relative abundance, in the oils from coals and Eocene deltaic shale 16, based on relative retention time. Presence of the earlier eluting seco-atisane isomer (csAti1) could not be confirmed in any sample because of its reported co-elution with atisane isomer Ati1 (Zinniker, 2005). Sample 1, from an Eocene hypersaline lacustrine shale, appears to contain abundant 16β(H)-kaurane, rather than the thermodynamically more stable 16α(H) isomer, which is consistent with its early oil-window maturity [e.g., for 5α,14α,17α-24-ethylcholestane, 20S/(20S+20R) = 32%].

The relatively enriched tetracyclanes in general, together with the presence of gymnosperm-related 16β-phyllocladane, in NSO-1 oil (sample 35, Figure 3), may reflect a minor contribution from Mid Jurassic coaly sources, which is more prominent in the Sigyn Field oil from the Norwegian North Sea (sample 14, Figure 2). NSO-1 represents Oseberg B-18 production oil, from Mid Jurassic (Brent) reservoir at ∼3 km depth. Within the study set, the presence of C₂₀ tetracyclic diterpanes in samples predating the appearance of gymnosperms is limited to traces of Kau-Ati-Bey combination in the Ordovician marine shale oil (sample 20, Figure 3). Whether the diterpanes in this oil are augmented by those from a younger source rock must be considered. Contributions to marine shale, carbonate and marl source rocks in general from kaurene precursors to microbial gibberellins is a possibility, if they are sufficiently abundant. Spores and pollens may represent another source of gibberellin precursors.

The upper Devonian Munindalen cannel coals from Svalbard are rich in spores, primarily from lycopsids, but have been found to contain no detectable Kau-Ati-Bey, unlike the neighbouring Pyramiden Carboniferous humic coals (Blumenberg et al., 2018). In contrast, the cuticular liptobiolite Devonian coals from China, primarily derived from lycopsids, do contain Kau-Ati-Bey, and what appear to be 17-nor and possibly dinor homologues (Song et al., 1997; Song et al., 2022), but the nor-phyllocladanes suggested to be present by Sheng et al. (1992) appear to be misidentified, so the earliest sedimentary occurrence of phyllocladane remains at Late Mississippian (Versteegh and Riboulleau, 2010). The biological origin of the Devonian diterpanes is unclear; it may be lycopsid cutinite, or the result of microbial activity or even other plant types, such as the progymnosperms (Rothwell, 1996; Retallack, 2021). Devonian oils in our data set exhibiting minor Kau-Ati-Bey are from carbonate (sample 45, Figure 4), marine shale (sample 24, Figure 3) and mixed lacustrine and marine shale (sample 2, Figure 2). It is possible that the diterpanes in the last sample originate from the Mid Jurassic marine shale contribution to the Beatrice Field oil, rather than the co-sourcing Devonian lacustrine unit, as noted for a possible trace of 5β(H)-rosane in that oil (Section 3.2.2).

In oils not dominated by gymnosperm contributions, tetracyclics are mostly minor components or undetectable, although their m/z 274→123 responses can sometimes rival those of equally sparse tricyclic m/z 276→123 signals, such as for beyerane in the Jurassic marine shale sourced samples 29 and 35 in Figure 3. Small amounts of what appear to be phyllocladane isomers are present in the Triassic marine carbonate (46), otherwise Mesozoic carbonates and marls contain only traces of beyerane (Bey), atisanes (Ati1 and Ati2) and a previously unidentified compound, Z (Figure 4), which is discussed below. Oils from older carbonate and marl sources are devoid of these tetracyclics, except for the Mid Devonian sample (45) from the Western Canada Basin (Figure 4). The oils derived from marine shales exhibit broadly similar distributions to those from carbonates and marls (Figure 3).

Relative responses for the m/z 274 to 123, 245 and 259 transitions suggest that beyerane, where detected in five of the eight oils subjected to extended GC-MS-MS analysis (7, 8, 17, 20, and 35), does not suffer from significant co-elution problems as far as those responses are concerned (Figure 9). However, a closely eluting, unidentified compound affects the m/z 274→189 response. Similar evaluation of 16α(H)-kaurane (αKau) and the pair of atisane isomers (Ati1 and Ati2) was inconclusive because of the potential co-elution of αKau with Ati1 and of Ati2 with the earlier eluting of a pair of (16R)-14,15-cyclo-8,14-seco-atisane isomers (csAti1; Zinniker, 2005). However, there is no obvious evidence for significant contributions from compounds with distinctly different mass spectra.

Clay-catalyzed inter-conversion of kaurene, beyerene and atiserene has been observed in the laboratory and suggested to account for sedimentary distributions of βKau, Ati2 and Bey (Zinniker, 2005). The common occurrence of these three tetracyclics in the study samples could be argued to provide circumstantial evidence of such diagenetic control, although it may well be superimposed on varying contributions of individual tetracyclics related to differences in microbial communities.

Compound Z is a C₂₀ tetracyclane with an intense m/z 203 fragment ion, and it shares the mass spectrometric characteristics of 5α(H),14β(H)-androstane (Figure 10), so we suggest it is a homologue of that compound. The additional methyl group is likely to be found on the A-ring, because if it were on the D-ring the response of m/z 203 would be reduced and m/z 217 enhanced. This proposition is based on the observation of Tökés and Djerassi (1969) that 90% of the dominant m/z 203 signal originates from loss of the A-ring, because the absence of a side-chain at C-17 (which is present in regular steranes) confers greater stability on the D-ring, so the m/z 217 fragment ion is only minor. No higher homologues could be conclusively identified. The A-ring methyl group is suggested to be most likely at C-4, given that Z was detected in post-Paleozoic oils only, the methyl steranes of which are generally dominated by 4-methyl species.