Introduction

The end-Devonian Hangenberg Crisis (EDHC; 359.3 ± 0.1 Ma with a duration of 100–300 ka) is of the same scale as the so-called ‘Big Five’ first-order Phanerozoic mass extinctions, and severely impacted marine invertebrates, especially ammonoids, trilobites, conodonts, stromatoporoids, and ostracodes, and is associated with a major turnover from spore groups to seed plants on land1,2,3,4. Climate and depositional environment also underwent remarkable changes during the latest Devonian, including abrupt glacial-interglacial fluctuations3,5, widespread anoxia6,7,8,9, enhanced continental weathering10, widespread demise of carbonate factories5,8,11, and distinct perturbations in the global carbon cycle3,12,13. Enhanced terrestrial inputs due to land plant evolution could be a primary trigger for climate cooling and marine anoxia during the EDHC, through drawdown of atmospheric CO2 levels and elevated primary productivity stimulated by nutrient influx10. However, other studies have argued that enhanced volcanism played a significant role in driving climate and environment changes based on mercury (Hg) enrichments during the EDHC or in upper Devonian successions14,15,16,17,18. Therefore, the ultimate cause(s) of the EDHC remain controversial.

Marine anoxia has been commonly proposed as a potent killing mechanism for the EDHC3,6,7,9. Widespread marine anoxia associated with latest Devonian black shales has been documented by various organic and inorganic proxies8,9,19,20,21. However, it has to be noted that the Hangenberg black shales (HBS) occurred in the lower part of the Middle Siphonodella (S.) praesulcata conodont zone in Germany, whereas the EDHC lasted a much longer time spanning from the base of the Middle S. praesulcata zone to the base of the earliest Carboniferous S. sulcata zone3. A recent study in western Canada shows that the onset of anoxia and black shale deposition in western Laurentia happened much earlier (ca. 4 Ma) than other regions, and suggests the occurrence of multiple diachronous anoxic basins in the latest Devonian where biota was under stress22. Therefore, the explicit response of marine life to anoxia during the EDHC remains elusive and open for further evaluation.

Mercury (Hg) has been widely used as a proxy for large-scale volcanic events as they can result in widespread Hg enrichments in sediments23. However, other factors, such as soil erosion and sequestration of seawater Hg under anoxic/euxinic conditions, can also cause Hg anomalies in sedimentary rocks23,24,25,26. Thus, clarifying the sources of Hg in seawater and/or sediments is crucial to our understanding of the causal relationships between extinction events and coeval geological processes. Hg isotopes are a powerful tool to trace Hg sources and pathways in sediments as it exhibits both mass-dependent (MDF) and mass-independent fractionations (MIF)27. To date, Hg isotopes have been successfully applied to trace past-volcanic events25,28,29, enhanced soil erosion25,26, and development of photic-zone euxinia30,31,32,33,34 throughout the Earth history. Hg isotopes from an Upper Devonian and Lower Carboniferous drillcore section in Tennessee (USA) show multiple negative excursions in MIF (Δ199Hg) along with concurrent positive shifts in MDF (δ202Hg), providing evidence for photic-zone euxinia events in the late Devonian and early Carboniferous oceans31. However, the EDHC beds in the studied core are either missing or have not been identified by biostratigraphy31, making it difficult to link seawater redox conditions with the EDHC. Hence, detailed Hg isotopes across the EDHC are required to unravel if there is a volcanic signature across the EDHC and to investigate how variations in seawater redox conditions related to the EDHC.

Here, we selected two Devonian-Carboniferous boundary (DCB) sections in South China, namely the Longan section that lies on the Qianggui carbonate platform and the Duli section that is situated in the Youjiang basin (Fig. 1), and one section at Jura Creek in Western Canada that lies within an interpreted inner-ramp setting22. These varied depositional settings from different tectonic domains enable a globally spatial view of the EDHC. We used carbon isotopes to monitor carbon-cycle disturbances and for stratigraphic correlation, and Hg isotopes to determine the sources and depositional pathways of seawater Hg and its response to the marine anoxia. We also measured major and trace elements to provide insights into changes of compositions in the studied samples and applied Ce/Ce* as well as size distribution of pyrite framboids to trace the redox state of the water column in studied sites35,36,37,38, and used chemical index of alteration (CIA) data in Jura Creek to evaluate weathering intensity changes across the DCB39. This data set provides a window to observe the Hg cycle across the DCB and the causes and consequences of the EDHC.

Fig. 1: Latest Devonian paleogeographic reconstructions.
figure 1

A global paleogeographic reconstruction showing the location of South China and Western Canada (base map from https://deeptimemaps.com authorized by Colorado Plateau Geosystems Inc.); B Paleogeographic reconstruction of Western Laurentia showing location of the Jura Creek (JC) section61; C Paleogeographic map of South China craton showing localities of Longan (LA) and Duli (DL) sections12.

Full size image

Results

In the Longan section, values of δ13Ccarb are ~ 0 ‰ in the Lower S. praesulcate conodont zone, whereas they shift to negative values from the uppermost segment of the Lower S. praesulcate zone to the minimum of −1.63‰ during the Middle S. praesulcate zone, followed by a positive excursion to the maximum of 2.39‰ during the Upper S. praesulcate zone (Fig. 2). The δ13Ccarb values return to ~0‰ at the end of Upper S. praesulcate zone and show the second positive excursion during the Tournaisian (S. duplicate-S. sulcata zone; Fig. 2). Hg concentrations in this section vary from 1.3 ppb to 122.3 ppb with low values (less than 25 ppb) in Lower to Middle S. praesulcate conodont zones, but with elevated values at the end of Upper S. praesulcate zone, and higher values (up to 80.9 ppb) in the Tournaisian (Fig. 2). Both Δ199Hg and δ202Hg show large variations in Longan, with Δ199Hg ranging from −0.16 ± 0.06‰ to 0.49 ± 0.06‰ and δ202Hg ranging from −2.10 ± 0.07‰ to −0.19 ± 0.07‰ (Fig. 2). The background values of Δ199Hg in the Lower S. praesulcate zone are ~ 0.4 ‰ and show a shift from the uppermost Lower S. praesulcate zone to lower values (~0.2‰) at the base of the Middle S. praesulcate zone (Fig. 2). δ202Hg is more varied in the Lower S. praesulcate zone with values ranging from −2.10‰ to −0.42‰. During the Middle to Upper S. praesulcate zones, which correspond to the EDHC interval, Δ199Hg values show a continuous decline from ~ 0.20‰ to ~ −0.15‰ and δ202Hg values show a ~ 1.8‰ positive shift from ~ −2.0‰ to ~ −0.2‰ (Fig. 2). Values of Δ199Hg remain negative (~ −0.2‰ to 0‰) and values of δ202Hg retain high values (> −1‰) except for one value of −1.98 ± 0.07‰ in the Tournaisian (S. duplicata-S. sulcata; Fig. 2). The samples in the lower and middle part of the Lower S. praesulcate zone are largely devoid of pyrite framboids. However, framboid abundance increases and framboid size declines from the uppermost Lower S. praesulcate zone to the lower part of S. duplicata-S. sulcata zones (Fig. 2).

Fig. 2: Stratigraphic information and geochemical profiles in Longan from South China.
figure 2

Including lithology, conodont zones12, inorganic carbon isotope (δ13Ccarb), whole rock Hg concentration, carbonate-free Hg concentration (Hgcf), Δ199Hg, δ202Hg and pyrite framboids size (PFS) data. Se. Series, St. Stages, Fm. Formation, Lith. Lithology, CZ Conodont zones, L. S. praesulcate Lower Siphonodella praesulcate, M. S. p Middle Siphonodella praesulcate, U. S. p Upper Siphonodella praesulcate, EDHC End-Devonian Hangenberg Crisis, NCIE Negative carbon isotopic excursion, SD standard deviation.

Full size image

In the Duli section, values of δ13Ccarb are relatively constant at ~2‰ below Bed 7 (Lower S. praesulcate zone), but commence a negative shift at Bed 7 and reach the most negative value (−3.4‰) in Bed 9 (costatus-kockeli interregnum; Fig. 3). The values return to −2‰ to 0‰ in beds 11–15 of the Tournaisian (S. sulcata-S. crenulata zones; Fig. 3). Hg concentrations in this section vary from 40.7 ppb to 4330.3 ppb with relatively low values (mostly less than 200 ppb) in the limestones of both the Wuzhishan and Wangyou formations, whereas peaks (more than 2000  ppb) occur in calcareous mudstones and black shales at the top of the Wuzishan Formation (Fig. 3). Values of Δ199Hg vary from −0.19 ± 0.08‰ to 0.18 ± 0.08‰ through the Duli section, and show the onset of a negative shift at the bottom of Bed 8, corresponding to the onset of negative carbon isotopic excursion (NCIE; Fig. 3). The values shift to ~ −0.1‰ in calcareous mudstones and black shales at the top of the Wuzishan Formation and reach the most negative values in the earliest Tournaisian (Fig. 3). Values of δ202Hg display synchronous variations with those of Δ199Hg, with negative values (−1.42 ± 0.10‰ to −0.36 ± 0.10‰) in beds 1–7, and show the onset of a positive shift at the commencement of the NCIE (Fig. 3). They achieve positive values (up to 0.28 ± 0.10‰) in calcareous mudstones and black shales at the top of the Wuzishan Formation and show the most positive values (up to 0.52 ± 0.10‰) in the Wangyou Formation of the earliest Tournaisian (Fig. 3). The Ce/Ce* profile show a similar trend to that of δ202Hg, displaying elevated values from the upper part of Bed 7 to the peak values in the Tournaisian (Fig. 3).

Fig. 3: Stratigraphic information and geochemical profiles in Duli from South China.
figure 3

Including lithology, conodont zones16,62, inorganic carbon isotope (δ13Ccarb), whole rock Hg concentration, carbonate-free Hg (Hgcf) concentration, residual Hg (HgR), Δ199Hg, δ202Hg and Ce/Ce* values. CZ Conodont zones, ckI costatus-kockeli interregnum, k kockli, S. Siphonodella, EDHC End-Devonian Hangenberg Crisis, NCIE Negative carbon isotopic excursion, RRSE robust residual standard error, SD standard deviation.

Full size image

In the Jura Creek section, Hg concentrations are low (~10 ppb) in the limestones of the Palliser Formation, however show peaks (up to 1605 ppb) at the bottom of the black shales. Moving up through the section, Hg concentrations progressively declined and have low values (less than 10 ppb) in calcareous mudstones, siltstones and sandstones (Fig. 4). Values of Δ199Hg and δ202Hg show synchronous changes through the Jura Creek section with a ~ 0.3‰ negative shift from 0.19 ± 0.06‰ in the Palliser Formation to −0.10 ± 0.06 ‰ at the base of the black shales of the Exshaw Formation in Δ199Hg and a ~ 1.35‰ positive shift from −1.18 ± 0.07‰ to 0.18 ± 0.07 ‰ in δ202Hg (Fig. 4). Upward through the section, Δ199Hg and δ202Hg return to less negative values (−0.05‰ to 0‰) and more negative values (−1‰ to −0.5‰) in the lower part of the Exshaw Formation, respectively. Hg isotope values are then more negative (as low as −0.12 ± 0.06‰) for Δ199Hg and less negative values (~−0.5‰) in δ202Hg near the DCB. In the Tournaisian siltstones and sandstones, values of Δ199Hg and δ202Hg shift to 0.01 ± 0.01‰ (n = 2) and −1.99 ± 0.08‰ (n = 2), respectively (Fig. 4). CIA varies from 56.5 to 76.7 through the Jura Creek section with relatively higher values across the DCB (Fig. 4).

Fig. 4: Stratigraphic information and geochemical profiles in Jura Creek from Canada.
figure 4

Including lithology, conodont zones22, δ13Corg22, whole rock Hg concentration, carbonate-free Hg concentration (Hgcf), residual Hg (HgR), Δ199Hg, δ202Hg and pyrite framboids data8. Se. Series, St. Stages, Fm. Formation, CZ Conodont zones, U. Unit, H. Height, Lith. Lithology, EDHC End-Devonian Hangenberg Crisis, RRSE robust residual standard error, PFS pyrite framboids size, SD standard deviation.

Full size image

Discussions

No evidence for overwhelming volcanism as a driver of the EDHC

Recent10,12,14 studies together with our work show that the EDHC is characterized by a negative excursion followed by a positive shift in δ13Ccarb in the Paleo-Tethys ocean (Figs. 2, 3 and S1). The negative excursion in δ13Ccarb observed at well-preserved stratigraphic successions in South China may be the result of enhanced continental weathering that stimulated organic matter respiration10. This process would contribute abundant 12C-enriched dissolved inorganic carbon to the oceans10. In addition, release of 12C-enriched gases from volcanic eruptions is thought to have also contributed to the negative shift in δ13Ccarb, given high Hg concentrations during the EDHC14. Below we assess the relative importance of the two processes on the EDHC from our and published Hg and weathering data.

Profiles of whole rock Hg concentration display inconsistent variations across the DCB. In Longan, the peaks occur at the end of the EDHC and in the early Tournaisian (the uppermost of Upper S. praesulcata zone and S. duplicata-S. sulcata zones), whereas those in Duli and Jura Creek, respectively, appear during the EDHC (costatus-kockeli interregnum and Upper S. praesulcata zone) and the basal Exshaw Formation that is at least 4 Ma earlier than the EDHC based on bentonite U-Pb ages22 (Figs. 2, 3 and 4). The EDHC in Jura Creek might be within the upper part of the Exshaw black shale, based on continuous black shale deposition (Fig. 4). The discordance in Hg enrichments in whole rock data likely are associated with lithological changes given that the shifts to higher Hg concentrations correspond to changes from carbonates to shales in both Duli and Jura Creek (Figs. 3 and 4). Analysis of Hg data on a carbonate-free basis is important because increasing CaCO3 concentration can dilute non-carbonate compounds (including Hg, total organic carbon (TOC), aluminum (Al), and total sulphur (TS))40. These increasing concentrations can force negative correlations of non-carbonate vs carbonate compounds as well as positive correlations among non-carbonate elements40. The clear negative correlations between CaO content and Hg concentration and positive correlations between Al content and Hg concentration in all studied sections indicate that increasing CaCO3 concentration played an important role in measured Hg concentrations (Figs. S2 and S3). Carbonate-free values (Hgcf, TOCcf, TScf and Alcf) are calculated based on measured CaO concentration (the computational formula based on previous study40 and outlined in Supplementary Note S1). Results show that there is a rise in Hgcf in the middle part of the EDHC in Longan (Fig. 2), however, the shale beds during the EDHC in Duli and the shale layers near the DCB in Jura Creek show insignificant enrichments when compared to background values of carbonate layers (Figs. 3 and 4). We also calculated residual (excess) Hg (HgR) in Duli and Jura Creek (follow previous method40 and outlined in Supplementary Note S2) based on a strong linear correlation between Hgcf and TOCcf in Duli and a strong linear correlation between Hgcf and TScf in Jura Creek (Fig. S4). Values of HgR across the DCB at both the sections are within ± 2 robust residual standard error (Figs. 3 and 4), which suggest that the Hg data is at “background” concentrations and not significantly elevated due to external processes such as volcanism40. The synchronous changes between Hgcf and TOCcf (or TScf) as well as the lack of obvious anomalies either in Hg/TOC or Hg/TS across the EDHC in Duli and Jura Creek (Figs. S5 and S6) further indicate Hg concentrations of both the sections were highly controlled by environmental changes rather than Hg loading. In Longan, Hgcf shows no correlations (R2 ≤ 0.03) with TOCcf, Alcf and TScf (Fig. S4), making it difficult to determine the main host-phase of Hg. However, the higher Hgcf values during the Upper S. praesulcata zone and S. duplicata-S. sulcata zones correspond to distinctly elevated values in TOCcf and Alcf concentrations (Fig. S7). Profile of Hg/TOC show insignificant enrichments during the EDHC (Fig. S7), though most of TOC concentrations in Longan are less than 0.2 % that would produce large uncertainty23. In addition, the onset of negative shift in δ13Ccarb and the EDHC clearly predated higher Hgcf values of the Longan section in South China (Fig. 2), which discounts volcanism as a major driver of the EDHC even if volcanic Hg loading could have contributed to the higher Hgcf values in Longan. An upper Devonian section in Illinois basin (USA) that is dominated by shales also lack Hg/TOC anomalies across the DCB (Fig. S8)31, and a well calibrated shale section (Sjol core) for the EDHC in Williston basin (USA) has elevated Hg concentrations but lacks obvious Hg/TOC anomalies (less than 30 ppb/% and similar to background values) during the EDHC (Fig. S8)9. Furthermore, terrestrial DCB sections in East Greenland also show no Hg enrichments across the extinction interval4. Therefore, it is unlikely that massive subaerial volcanism, including LIP activity, was present in the latest Devonian and triggered the EDHC, though local magmatic activity could have contributed to carbon isotopic fluctuations in Europe14. Our Hg isotopic data further indicate there was no overwhelming volcanic signature during the EDHC. Volcanic derived Hg has been shown to have original Δ199Hg values close to zero41,42,43. However, oxidized Hg (II) during photoreduction in cloud droplets can result in positive Δ199Hg values in the residue Hg (II) phase that is primarily received by seawater through wet or dry deposition27,44,45. Therefore, near-zero and positive Δ199Hg values with increased Hg input in sediments are considered to reflect signatures of direct deposition and atmospheric transport of enhanced volcanic Hg loading, respectively25,28,29,46,47,48. Values of Δ199Hg during the EDHC in Longan have a large range from 0.21‰ to −0.15‰ with a shift from positive to negative values through that time (Fig. 2). Those during the EDHC in Duli and around the DCB in Jura Creek are −0.08 ± 0.04 ‰ (sd, n = 6) and −0.10 ± 0.02 ‰ (sd, n = 2), respectively (Figs. 3 and 4). The negative values or shifts in Δ199Hg likely indicate that Hg cycle during the EDHC was dominated by environmental changes rather than recording a direct (or overwhelming) signature of volcanism.

Lithium isotope display a rapid negative shift immediately before the onset of the EDHC in South China10, suggesting a causal relationship between enhanced weathering and the EDHC. Our CIA values (the computational formula was outlined in Supplementary Note S3) of shales in Jura Creek also display elevated values approaching the DCB (Fig. 4), corresponding to the re-development of water column anoxia that is indicated by pyrite framboids data8. In spite of low resolution (only 5 data from Pseudopolygnathus granulosus to Siphonodella sandbergi zones), CIA values in Illinois basin (USA) show a minor increase across the DCB49 (Fig. S8). Hence, nutrient input along with enhanced weathering could have played a significant role in marine anoxia across the EDHC.

Photic zone euxinia across the EDHC

In South China, both the Longan and Duli sections begin to display negative shifts in Δ199Hg coupled with positive shifts in δ202Hg in the upper part of the Lower S. praesulcata zone, corresponding to the onset of the NCIE (Figs. 2 and 3). Hg–MIF (Δ199Hg) is mostly controlled by photochemical processes and is unlikely to be altered by post–depositional processes, whereas, the causes of Hg–MDF (δ202Hg) are more varied27. The lack of obvious correlation between δ202Hg with Mn/Sr and Mg/Ca at both the Longan and Duli sections (Fig. S9), indicates that the study samples are unlikely to have been altered by diagenetic transformation or dolomitization. The observations of a weak negative correlation (R2 = 0.11, P < 0.05, n = 34) in Longan and a strong correlation (R2 = 0.61, P < 0.0005, n = 31) in Duli between Δ199Hg and δ202Hg (Fig. S10) further suggest the Hg isotopic compositions represent primary signals and the changes in Δ199Hg and δ202Hg were mainly controlled by common mechanisms.

Modern observations show atmospheric Hg (0) has higher δ202Hg and lower Δ199Hg values than atmospheric Hg (II)50. Different fractions of uptake of atmospheric Hg (0) in vegetation and wet/dry deposition of atmospheric Hg (II) onto the Earth’s surface results in more negative δ202Hg and Δ199Hg values in terrestrial than in marine reservoirs50. Therefore, the observed negative shifts in Δ199Hg and concurrent positive shifts in δ202Hg cannot be explained only by increased terrestrial Hg input as that would contribute more negative δ202Hg values rather than significant positive shifts (more than 1‰) as observed in the study samples (Figs. 2 and 3). Instead, the changes in Hg isotopic composition likely reflect the development of photic zone euxinia (PZE) conditions because both photoreduction of Hg (II) complexed by reduced sulfur and enhanced sequestration and oxidation of atmospheric Hg (0) in a sulfide-rich photic zone would produce negative MIF and positive MDF in Hg (II)51,52,53. Such changes in Hg isotopic compositions (negative MIF and positive MDF) have been observed in deep-time oceans associated with PZE30,31,32,33,34.

Our pyrite framboid data in Longan and Ce/Ce* values in Duli further demonstrate that observed negative shifts in Δ199Hg coupled with positive shifts in δ202Hg were driven by water column redox changes. Pyrite framboid sizes have been widely utilized to diagnose redox condition of the water column, with a small size displaying a narrow distribution occurring in euxinic settings, whereas the sizes are more variable and generally larger in dysoxic environments35,36. Based on size distribution of framboids, criteria have been established to distinguish different redox states of the water column (Fig. S11)35,36. Ce anomaly (Ce/Ce*) of carbonate is another proxy for rebuilding redox changes of a water column during formation of the carbonate rocks37, with less negative and even positive Ce/Ce* values in the wake of increased marine anoxia (the computational formula and theory outlined in Supplementary Note S4). However, it should be noted that Ce/Ce* can only represent seawater information when Y/Ho are more than 36, as ratios of Y/Ho can be used to distinguish between seawater and non-seawater signatures38. Given that most values of Ce/Ce* in Longan are less than 36 (Supplementary Dataset 1), compared to most of those in Duli being more than 36 (with only four data within 30-36 (Fig. 2 and Supplementary Dataset 2), we propose that the variations in Ce/Ce* at Duli record the redox changes of the water column. Both the pyrite framboids in Longan and Ce/Ce* in Duli show anoxia gradually increased from the upper part of the Lower S. praesulcata to the S. sulcata-S. crenulata zones, as indicated by reduced sizes in framboids and elevated Ce/Ce* values (Figs. 2 and 3). The covariant relationships between Hg isotopes (Δ199Hg and δ202Hg) with mean framboid diameters as well as Ce/Ce* values, as shown by their correlations in scatter diagrams (Fig. 5), ultimately confirm the changes in Δ199Hg and δ202Hg were controlled by redox changes of seawater and thus reflected the PZE condition across the EDHC, further promoting Hg isotopes as a photic zone redox proxy.

Fig. 5: Scatter plots of Δ199Hg and δ202Hg with redox proxies in Longan and Duli.
figure 5

Values of Δ199Hg and δ202Hg show correlation with mean framboid diameter in Longan and Ce/Ce* values in Duli. SD standard deviation.

Full size image

The initial deposition of black shales and development of anoxic conditions in Jura Creek have been shown to start before or during the P. expansa conodont zone, ca. 4 Ma earlier than the EDHC22. The anoxia was suggested to be caused by a locally restricted basin driven by mountain building in the latest Devonian22, that persisted into the earliest Carboniferous although the bottom water was likely more oxic around the DCB than earlier as evidenced by trace metal proxies (e.g., Mo, U, and Cd)8,22. Our Hg isotopic data from Jura Creek also show correlation between Δ199Hg and δ202Hg (R2 = 0.18, p < 0.05, n = 15; Fig. S10), and a ~ 0.3‰ negative shift of Δ199Hg coupled with a ~ 1 ‰ positive shift of δ202Hg at the bottom of the Exshaw black shales (Fig. 4), corresponding to Mo concentration peaks22, supporting a shift to euxinic condition even in the photic zone. After a slight recovery (~0.1‰), Δ199Hg values show again a ~ 0.1‰ negative excursion coupled with a ~ 0.4‰ positive excursion in δ202Hg in the uppermost S. praesulcata zone (Fig. 4), suggesting the PZE had continued into the EDHC. This is supported by pyrite framboids data8 that show euxinic water column conditions during that time in Jura Creek (Fig. 4).

Apart from our Hg isotopic data, the presence of lipid biomarkers of green sulfur bacteria in the black shales from sections in central Poland and North America provides further evidence for photic-zone euxinia20,21. Spatiotemporal redox trends across the entire Williston Basin (USA) during the EDHC display not only expanded water column euxinia but also euxinic seawater into shallow environments9. Ratios of I/Ca for a carbonate platform in South China as well show expansion of the oxygen minimum zone into the ocean-surface layer at that time6. Although the DCB shale beds were extremely condensed and largely lack biostratigraphic constraint31, a drill core section in the Illinois basin (USA) shows negative shift in Δ199Hg coupled with positive shift in δ202Hg pre-the DCB (Fig. S8), suggesting a PZE condition during that time31. Therefore, the development of PZE was widespread and played a significant role in organism losses across the EDHC.

Implications for causal relationship between environmental change and the EDHC

The integration of our carbon and Hg isotopes with previously published latest Devonian-earliest Carboniferous weathering proxy (lithium isotope), palaeontologic data and paleotemperature records (Fig. 6), enable a holistic assessment of environmental conditions across the EDHC. Values of lithium isotope (δ7Licarb) as well as δ13Ccarb are relatively constant in lower-middle part of the Lower S. praesulcata zone (Fig. 6), however, both the isotopes show sudden negative shifts in the uppermost part of the Lower S. praesulcata zone (Fig. 6), ca. 90 ka before the EDHC10, indicative of enhanced silicate weathering and carbon perturbations immediately before the EDHC. Enhanced weathering is expected to increase nutrient inputs to the ocean that would contribute to oceanic eutrophication and anoxia54. Our Hg isotopic data support PZE conditions starting to spread at the onset of the NCIE and thus was likely a consequence of enhanced weathering (Fig. 6). A PZE event is deadly to organisms living in surface waters, such as stromatoporoid and rugose coral (Fig. 7), both of which show a decline in diversity before the EDHC (Fig. 6)55. There are small-scale but distinctive extinction episodes of conodonts and ammonoids in the immediate pre-crisis beds in Sardinia (Italy) and Rhenish Massif3, suggesting environment deterioration immediately preceding the EDHC.

Fig. 6: Generalized chemostratigraphic trends of the latest Devonian and the relationship to biotic change.
figure 6

Integrated excursions of inorganic carbon isotope (δ13Ccarb; Longan and Duli (this study); Longmenshan10), lithium isotope10, Hg isotopes (Δ199Hg and δ202Hg), genera number of stromatoporoid and rugose coral55, and δ18Oapatite56 across the end-Devonian Hangenberg crisis. ckI costatus-kockeli interregnum, k kockli, SD standard deviation.

Full size image
Fig. 7
figure 7

Scenario of environmental and biological changes along with Hg cycle across the end-Devonian Hangenberg Crisis (ca. 359.3 Ma).

Full size image

A continuous and more rapid weathering was present during the Middle S. praesulcata zone (or costatus-kockeli interregnum)10, corresponding to the shifts to lower values in δ13Ccarb (Fig. 6), likely indicating enhanced 12C-enriched dissolved inorganic carbon inputs along with increased weathering. Meanwhile, efficient CO2 drawdown due to strong weathering of rocks would regulate the climate into a cooling stage as shown by δ18Oapatite values from conodonts (Fig. 6)56. Oberved negative shift in δ13Ccarb during the Middle S. praesulcata zone supports the hypothesis that the peak or onset of Hangenberg positive carbon isotopic excursion postdated the HBS because the HPICE records burial of abundant organic matter in the deep ocean rather than in epeiric seas57,58. Following the decline in δ13Ccarb values, positive shifts of δ13Ccarb have been recorded within the Upper S. praesulcata zone, consistent with cessation of increased weathering (though the weathering was still stronger than pre-the EDHC) (Fig. 6), supporting the inference of large-scale sequestration of organic carbon within the Upper S. praesulcata zone12. Our Hg isotopic data in South China show progressively negative shifts of Δ199Hg coupled with positive shifts of δ202Hg from the uppermost of the Lower S. praesulcata zone to the Tournaisian (Fig. 6), indicative of an enhanced PZE process through the EDHC, which is further confirmed by data from elevated Ce/Ce* values and pyrite framboids during that time (Figs. 2 and 3).

In these circumstances, the PZE was likely driven by nutrient influx and elevated primary productivity, and played an important role in the collapse of marine invertebrates during the EDHC (Fig. 7), as shown by the final demise of stromatoporoids and a large decline of biodiversity in rugose corals (Fig. 6). The development of PZE condition near the DCB triggered by increased weathering input is also indicated by the coincidence between nutrient input and Hg isotopic shifts in Illinois basin (USA)31 and by Our CIA and Hg isotope data in Jura Creek (Fig. 4). The ultimate cause of enhanced silicate weathering across the EDHC is still unknown. The absence of a global fingerprint for volcanic Hg and any record of a large igneous province during that time suggests the driving force is not linked to large-scale volcanism. In contrast, the tectonically driven basin restriction model22 and/ or the arguably rapid spread of seed plants59,60, likely were a more possible trigger for the increase in weathering intensity. Emerging studies disagree with the idea that plant expansion caused anoxia, and even argue that instead it led to oxidation of the oceans60. In any case, our Hg isotopic data from successions show a global PZE condition across the EDHC that could be an important killing mechanism for the DCB mass extinction.

Methods

Studied sections

During the Late Devonian to Early Carboniferous, paleogeographic reconstructions indicate the South China craton occupied a paleo-latitude of 0–15° S and the northwest coast of Laurentia occupied a near equatorial location of 10–15°N (Fig. 1). At that time, the southern (paleo-eastern) margin of South China developed a complex facies architecture spanning from siliciclastic platform, to carbonate platform to deep water basin (Fig. 1), whereas in western Canada, a shallow epicontinental sea covered the broad cratonic platform61.

The Longan section (23°10′34.8′′N, 107°27′48.0′′E) is located in Dujie Village of Long’an County in Guangxi Province, of which the lithologic, carbon isotopic and bio-stratigraphy has been well studied11. The DCB succession in this section contains the Yonghsien Formation that consists mainly of massive bioclastic grainstone, and the overlying Longan Formation that is composed of thin-medium bioclastic packstone and packstone (Fig. 2). The DCB defined by the first appearance of the conodont S. sulcata, is located ~6 m above the base of the Longan Formation12. Based on fossil assemblage and carbon isotopic stratigraphy, the base of the Upper S. praesulcata and Middle S. praesulcata zones are placed at ∼2 m and ∼3.3 m below the base of the Longan Formation, respectively7,12. The Longan section was situated on a platform separated from the Yangtze paleo-landmass by the Youjiang Basin.

The Duli section (25°23′ 50.1′′ N, 107°20′ 05.5′′ E) is well-exposed along the X921 road, near the Duli village of Nandan County, Guangxi Province. The lithology and biostratigraphy have been described in detail56. At this locality, the DCB succession is divided into the Wuzhishan Formation and the overlying Wangyou Formation. The Wuzhishan Formation is mainly composed of massive micrite, with rare calcareous mudstone and a ~ 15 cm-thick black shale at the top separating the overlying Wangyou nodular and banded limestone (Fig. 3). Conodont biostratigraphy show the Wuzhishan Formation in this section contains abundant pelagic conodonts of the Lower S. praesulcata Zone, whereas the Wangyou Formation spans from the Upper S. praesulcata Zone to the Lower S. Crenulate Zone with the DCB placed at ~20 cm above the top of the black shale based on the first occurrence of typical S. sulcata62. The fine grained, condensed limestones observed in this section indicate deposition in quiet water below storm wave base, in a basinal setting62.

The Jura Creek section (51°05′26″N, 115°09′29″W) is well-exposed along a tributary to the Bow River, western Alberta, Canada. The DCB succession in this stratigraphic section is composed of the Palliser Formation carbonates and the overlying Exshaw Formation that mainly consists in ascending order of black shales, calcareous mudstones, and siltstones and sandstones (Fig. 4). The conodont fauna indicates the uppermost Palliser Formation is within the Palmatolepis (P.) trachytera-lower P. expansa zone, and the deposition of black shales ranges from lower P. expanse zone to S. praesulcata zone (Fig. 4)22,51,63. The DCB was placed at the top of black shales, corresponding to the base of the S. sulcata zone63 (Fig. 4).

A total of 267 samples (178 for Longan, 49 for Duli and 40 for Jura Creek) were analyzed geochemically. Samples were collected from outcrop with weathered surfaces first removed. In the laboratory any remaining weathered surfaces were removed and then samples were powdered to ~200 mesh in a freeze grinding apparatus for later geochemical analysis. All analytical data of Longan, Duli and Jura Creek sections are reported in Supplementary Datasets S1, S2 and S3, respectively.

Major and trace element analyses

Whole rock major element abundances of the Longan and Duli sections were determined by wave length dispersive X–ray fluorescence analysis of pressed powder pellets using a Panalytical Zetium XRF spectrometer at the State Key Laboratory of Geological Processes and Mineral Resources (SKLGPMR), China University of Geosciences (CUG), Wuhan. Before analysis, a muffle furnace was used to determine the loss on ignition. Experimental error was less than 5%, as monitored by measuring national reference sample GBW07132. Trace elements of the Longan and Duli sections were analyzed at SKLGPMR via ICP–MS (Agilent 7500a). The dissolution procedure follows previous method64. The sample powder was weighed into Teflon bombs followed by adding 1 mL HNO3 and HF. Then the sealed bombs were heated at 190 °C in oven for 48 h. The solution was steamed on a 115 °C electric heating plate, followed by adding 1 mL HNO3 and evaporating to the second round of dryness. The resultant salt was re-dissolved by adding ∼3 mL of 30% HNO3 and resealed and heated in the bomb at 190 °C for 12–24 h. The final solution was diluted to 100 g with mixture of 2% HNO3 for ICP–MS analysis. Blanks and national standards AGV-2, BHVO-2, BCR-2, RGM-2 were selected for calibrating elemental concentrations with an analytical uncertainty of better than 2% for all elements of interest. Whole rock elemental data of the Jura Creek section were measured on powdered material digested in a 2:2:1:1 acid solution of H2O-HF-HClO4-HNO3 and analyzed using a PerkinElmer mass spectrometer with ± 2% analytical error22.

Inorganic carbon isotopes analysis

Inorganic carbon isotopes of the Longan and Duli sections were analyzed at SKLGPMR using a MAT 253plus mass spectrometer. In preparation for carbonate carbon isotope (δ13Ccarb) analysis, 80 to 120 μg of powder were placed in a 10 mL Na-glass vial, sealed with a butyl rubber septum, and reacted with 100% phosphoric acid at 72 °C after flushing with helium. The Evolved CO2 gas was analyzed for δ13C. Isotope values are reported as per mile variation (‰) relative to Vienna Pee Dee belemnite standard. The analytical precision was better than 0.04‰ for δ13C based on duplicate analyses of the national reference standard GBW-04416 (δ13C = +1.61 ‰), TB-2 (δ13C = −6.06 ‰).

Total organic carbon (TOC) and total sulphur (TS) analysis

TOC contents of the Longan and Duli sections were analyzed by an Elementar micro cube analyzer at the SKLGPMR in CUG. Before analysis, sample powder was digested in 2 N HCl at 50 °C for 12 h to dissolve any carbonate minerals. After multiple centrifugal and lyophilization steps, the residue was prepared for later TOC measurement. Data quality was monitored via multiple analyses of standard samples DP–1 (65.44%) and GSS-8 (1.92%), yielding an analytical precision of ± 5 % of reported values. Values of TOC at the Jura Creek section have been published21. Total sulfur (TS) contents of the Longan and Duli sections were measured on a CS-902T carbon–sulfur analyzer at the State Key Laboratory of Geomicrobiology and Environmental Changes in CUG. Analytical error was within 5% based on replicate analyses of Alpha Resources standard SJSM (TS = 1.02%).

Pyrite framboid morphology analysis

We studied pyrite framboids from 31 sedimentary rock samples at the Longan section to reconstruct oceanic redox variation history through the EDHC. A polished surface (~1 × 1 cm) of each sample was examined for pyrite number and sizes using a scanning electron microscope (SEM; Hitachi SU8000; its morphology under SEM please see Fig. S12) at the State Key Laboratory of Geomicrobiology and Environmental Changes in CUG. A minimum number of 100 pyrite framboid sizes was measured for each sample if possible.

Hg concentration analysis

Hg concentration analyses at Longan and Duli were conducted by a Direct Mercury Analyzer (DMA80) at the SKLGPMR in CUG. About 100 mg sample powder for limestone and ~50 mg for shales was weighed for analyses in this study. Data reliability was assessed through measurement of a standard coal sample (502–685; 40 ppb) and a repeat after 10 unknowns, yielding an analytical precision of ± 5% of reported Hg concentrations. Hg concentrations at Jura Creek were measured by a LECO AMA254 mercury analyzer with an analytical uncertainty of better than 10%.

Hg isotopes analysis

Hg isotopic experiments of the samples in Longan and Jura Creek were conducted by a Nu–Plasma multi–collector ICP–MS at the Institute of Geochemistry, Chinese Academy of Sciences, Guiyang, China, whereas those in Duli were analyzed using a Nu–Plasma multi–collector ICP–MS at the School of Earth System Science, Tianjin University, following the previous methods31,65. The pyrolysis method was used to extract Hg from samples before Hg isotopic measurement. Sample powder was heated at 950 °C for ∼3 h in a thermal decomposition furnace with oxygen injected for blowing the decomposed gaseous Hg0 into an absorption bottle, in which 5 mL of 40% reverse aqua regia (HCl: HNO3 = 1:3, v–v) was loaded in order to oxidize Hg (0) to Hg (II) and preserve Hg (II) with Hg recoveries of 90–110%. Concentrations between the diluted samples and bracketing standards (NIST SRM 3133) were matched within 10% that was monitored by MC–ICP–MS using 202Hg signals. NIST SRM 997 Tl was the internal standard used for simultaneous instrumental mass bias correction of Hg. The signal for 202Hg was ~1.2 V for 1 ng/mL digested solutions. Mass–dependent fractionations (MDF) of Hg isotopes are reported in δ202Hg notation referenced to the NIST–3133 Hg standard66:

$${{{{\rm{\delta }}}}}^{200}{{{\rm{Hg}}}}({{\textperthousand }})=\left[\left({\scriptstyle{202}\atop} \!{{{\rm{Hg}}}}/{\scriptstyle{198}\atop} \!{{{{\rm{Hg}}}}}_{{{{\rm{sample}}}}}\right)/\left({\scriptstyle{202}\atop} \!{{{\rm{Hg}}}}/{\scriptstyle{198}\atop} \!{{{{\rm{Hg}}}}}_{{{{\rm{standard}}}}}\right)-1\right]\times 1000$$
(1)

mass–independent fractionations (MIF) of Hg isotopes are expressed in Δ notation (ΔxxxHg), which describes the difference between the measured δxxxHg and the theoretically predicted δxxxHg value, using the following equations:

$${{{\Delta }}}^{199}{{{\rm{Hg}}}}\approx {{{{\rm{\delta }}}}}^{199}{{{\rm{Hg}}}}-({{{{\rm{\delta }}}}}^{202}{{{\rm{Hg}}}}\times 0.2520)$$
(2)
$${{{\Delta }}}^{200}{{{\rm{Hg}}}}\approx {{{{\rm{\delta }}}}}^{200}{{{\rm{Hg}}}}-({{{{\rm{\delta }}}}}^{202}{{{\rm{Hg}}}}\times 0.5024)$$
(3)
$${{{\Delta }}}^{201}{{{\rm{Hg}}}}\approx {{{{\rm{\delta }}}}}^{201}{{{\rm{Hg}}}}-({{{{\rm{\delta }}}}}^{202}{{{\rm{Hg}}}}\times 0.7520)$$
(4)

To ensure data quality, standard reference SRM GSS–5 prepared during the pyrolysis dissolution produced results (n = 2) of δ202Hg = −1.33 ± 0.04‰ (2 sd); Δ199Hg = −0.29 ± 0.03‰ (2 sd); Δ200Hg = 0.01 ± 0.02‰ (2 sd); Δ201Hg = −0.31 ± 0.02‰ (2 sd) that are comparable with previous results65. Replicate analysis of the NIST 3177 Hg intra lab isotope reference standard produced results (n = 16) of δ202Hg = −0.50 ± 0.07 ‰ (2 sd); Δ199Hg = −0.01 ± 0.06 ‰ (2 sd); Δ200Hg = 0.00 ± 0.05‰ (2 sd); Δ201Hg = −0.03 ± 0.04‰ (2 sd) that agree well with previous results67,68,69. The averages for all analyses of NIST 8610 were (n = 11): δ202Hg = − 0.55 ± 0.08‰ (2 sd); Δ199Hg = −0.02 ± 0.05‰ (2 sd); Δ200Hg = 0.01 ± 0.03‰ (2 sd); Δ201Hg = −0.04 ± 0.04 ‰ (2 sd), consistent with published values31.