Mercury Spikes Indicate a Volcanic Trigger for the Late Ordovician Mass Extinction Event: An Example from a Deep Shelf of the Peri-Baltic Region

The Late Ordovician mass extinction (LOME) was the second largest Phanerozoic crisis, but its cause remains elusive. Several triggering mechanisms have been proposed over the years, including bioevolutionary events, oceanographic changes, and geotectonic processes. Here, we report the presence of Hg spikes in the Zbrza PIG-1 borehole from the Upper Ordovician deep shelf sections of the peri-Baltic region. A strong positive anomaly in the lower late Katian (Hg/TOC = 2537.3 ppb/wt%) was noted. No correlation between Hg and TOC (R2 = 0.07) was distinguished in the Hirnantian, although several positive anomalies were found. Because the Hg/Mo ratio showed trends very similar to those of Hg/TOC, it seems likely that TOC values reflect the redox conditions. In order to evaluate the role of anoxia in levels of Hg enrichment several redox indicators were measured. These showed that the elevated mercury values in the Hirnantian are not caused by anoxia/euxinia because euxinic biomarkers (maleimides and aryl isoprenoids) are present in very low abundance and pyrite framboids are absent. In total, positive Hg/TOC anomalies occur in the lower late Katian, at the Katian - Hirnantian boundary, and in the late Hirnantian. The lack of a strong Hg/TOC correlation, Ni enrichments, and the absence of ‘anoxic indicators’ (no biomarkers, no framboids, low Mo concentration) at these levels, supports the interpretation that Hg enrichment is due to enhanced environmental loading. We conclude that our Hg and Hg/TOC values were associated with volcanic pulses which triggered the massive environmental changes resulting in the Late Ordovician mass extinction.


Materials and Methods
Sixty-eight samples were collected from the Zbrza PIG-1 borehole (Fig. 1), representing different types of Upper Ordovician and lower Llandovery claystone, mudstone, and black shale. The investigated samples are characterised by a relatively low level of thermal maturity (ca 0.72% of the vitrinite equivalent reflectance measured on graptolites) 32 . Upper Ordovician strata of the Zbrza PIG-1 borehole consist of the Sandbian to Hirnantian mudrock facies that form up to 90 m of the succession (Fig. 2). This shows laminae of varying thickness and is intercalated with numerous K-bentonite beds 33 . The Sandbian to lower Katian dark grey claystones and shales of the Jeleniów Formation are dated by their graptolite fauna that includes Nemagraptus gracilis, Diplograptus multidens, Climacograptus wilsoni and Dicranograptus clingani. They are overlain by the upper Katian clayey and marly mudstones and claystones of the Wólka Formation, which is roughly correlated with the complanatus, complexus and pacificus graptolite biozones. The mostly grey and greenish mudrocks of the Wólka Formation are intercalated, in their upper part, with dark grey and black shales (Fig. 2). Hirnantian strata in the Zbrza PIG-1 consists of sandy mudstones, marls and shales of the Zalesie Formation dated by Mucronaspis trilobites and the Hirnantia brachiopod fauna. There is a conspicuous facies change at the Ordovician/Silurian boundary with the Hirnantian mudstones overlain by Rhuddanian black, siliceous shales and cherts of the Zbrza Member. This is up to 20 m thick and is a component of the Bardo Formation. Graptolites indicate the ascensus to cyphus biozones are present and the Aeronian triangulatus biozone has also been postulated (Smolarek et al. 12 and references cited there). www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ results were compared with analyses of international standard reference materials and several samples were analysed in duplicate. Results, precision and accuracy were better than ±0.05% (mostly ±0.01%) for the major elements and generally better than ±1 ppm for the trace elements.
Mercury determination. Two independent methods were used for mercury determination for 20 samples (in Table 1 marked by bold). Other Hg measurements were performed using inductively coupled plasma mass spectrometry. ICP-MS analysis was conducted commercially at AcmeLabs, Vancouver, Canada. The precision and accuracy of Hg determination were better than ±10 ppb.
For atomic absorption spectrometry (AAS), a pyrolyser-type Milestone DMA-80 Direct Mercury Analyzer was used, with a detection limit of 0.2 ppb. DMA analytical curves were prepared with dilution of a 1-mg L -1 standard solution (Merck Darmstadt, Germany). Duplicate measurements were taken of each sample; analyses were repeated when the coefficient of variability of the samples exceeded 5%. The instrument was calibrated prior to the measurement using certified reference material INCT-OBTL-5 (tobacco leaves) with a Hg content of 20.9 ppm. The measured error did not exceed 2% (for more details see Racki et al.) 34 . Analyses were performed at the Faculty of Earth Sciences, University of Silesia (Poland).
The correlation coefficient between ICP-MS and AAS mercury data was very good (R 2 = 0.83). In this study we present Hg data obtained by the ICP-MS method.
Analysis of pyrite framboid diameter. Forty-four rock samples from the Zbrza PIG 1 borehole ( Fig. 1) were selected for framboidal pyrite diameter assay. Small chips of all samples were polished, and examined for pyrite framboids in an uncoated state using back-scattered electron (BSE) imaging with the aid of Philips XL30 Environmental Scanning Electron Microscope (ESEM) at the Faculty of Earth Sciences (Sosnowiec, Poland). The ESEM internal measuring device was used for measurement of the framboid diameters (in µm). At least 100 framboids were measured in each sample when possible. The results were presented in the traditional form of box and whisker plots 35 . organic matter extraction and separation. An extraction of crushed (to ca 100 mesh) samples was made in an accelerated Dionex ASE 350 solvent extractor with a mixture of dichloromethane (DCM)/methanol (5:1 v:v). The column chromatography was used for the extracts separation into aliphatic, aromatic, and polar fractions 36 . Firstly, silica gel was activated at 120 °C for 24 h, then cooled and poured into Pasteur pipettes. The fraction collection was made by three eluents, namely, n-pentane for the aliphatic fraction, n-pentane and DCM (7:3) for the aromatic fraction, and DCM and methanol (1:1) for the polar fraction.
Maleimide separation and derivatisation. Separation of the polar fraction obtained via column chromatography was made on activated silica gel in Pasteur pipettes. A fraction eluent a DCM/acetone mixture (8:2) was used. To obtain maleimide tertiary-butyl-dimethylsilyl derivatives, derivatisation of maleimide fraction was performed with MTBSTFA (N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide). Samples were derivatised with MTBSTFA dissolved in super-dehydrated DCM, heated at 50 °C for 1 h, and run directly after derivatisation (see also Grice et al.) 37 .

Results
Mercury concentrations were obtained for selected samples of the entire Zbrza PIG-1 profile (Figs 1 and 2; Table 1). The Sandbian claystones were characterised by low to elevated values of Hg ranging from 40 to 190 ppb, with a mean value of 105 ppb. The Hg concentrations in the Katian ranged from 30 to a very high value of 460 ppb (the mean value was 187 ppb). In the Hirnantian mudstones, mercury concentrations were generally low (approximately 10 to 90 ppb) with a few spikes of ~100 to 280 ppb (Fig. 2, Table 1). The youngest part of the profile recorded high values of mercury ranging from 160 to the highest value of the entire profile, 530 ppb (Table 1).
Because Hg concentration and organic matter in sediments are closely correlated, the presentation of Hg results as a ratio of mercury to total organic carbon (Hg/TOC, ppb/wt%) is typically used. The Zbrza PIG-1 section is characterised by a low degree of correlation between Hg and TOC in the Sandbian (R 2 = 0.2), with no positive Hg/TOC anomalies (Fig. 2, Table 1). No correlation exists between Hg and TOC in the Katian (R 2 = 0.01), but strong positive anomalies were noted in the lower late Katian (Hg/TOC = 2537.3 ppb/wt%) and at the Katian-Hirnantian boundary (Hg/TOC = 681.3 ppb/wt%). No such correlation (R 2 = 0.07) was distinguished in the Hirnantian, apart from several positive anomalies (Figs 2 and 3). The Silurian (Rhuddanian) shales and cherts showed a moderate correlation (R 2 = 0.56) between Hg and TOC, with no positive anomalies (Figs 2 and 3). The Ni/TOC (ppm/wt%) ratio shows trends similar to those noted for Hg/TOC (Fig. 2, Table 1).
Additionally, in order to exclude an increase in Hg connected with anoxic conditions supporting TOC and Hg elevation, we calculated the ratio of Hg to Mo (Table 1, Figs 2 and 3). This ratio shows trends very similar to those of Hg/TOC values, with the same correlations and positive anomalies ( Table 1, Figs 2 and 3).
In order to further check the possibility that Hg enrichment was caused by anoxic conditions, we assessed additional redox indicators. Thus, pyrite framboid diameters were measured for selected samples of the entire Zbrza PIG-1 profile (Fig. 2). Three intervals were found to display a large number of tiny (<5 μm) framboids, (in the Sandbian, the upper late Katian, and the Rhuddanian (Fig. 2). In contrast, no framboids were found in two intervals: the upper Sandbian to the lower late Katian and the entire Hirnantian (Fig. 2). Biomarkers of anoxic/ euxinic conditions were also assessed for the Hirnantian and the O/S boundary intervals (Fig. 3, Table 1). The upper late Katian and the Rhuddanian, both characterised by elevated TOC values, contain both maleimides and aryl isoprenoids (Fig. 3, Table 1). The Hirnantian interval shows elevated mercury values, but euxinic biomarkers are absent or present in very low level of abundance (Fig. 3, Table 1).

Figure 3.
Stratigraphic distribution of trace metals and biomarkers for pre-LOME and LOME intervals (see Table 1; maleimide ratio -Me, n-Pr/Me, Et).
www.nature.com/scientificreports www.nature.com/scientificreports/ Discussion Mercury vs total organic carbon. Phanerozoic mass extinctions have been linked to the emplacement of LIPs [38][39][40] and the associated release of SO 2 and CO 2 6,41,42 . Because mercury is the major pre-anthropogenic product of volcanism in Earth's surface environments 43 , its accumulation in sediments is commonly used as a proxy for ancient volcanic events 18,22,25 . Changes in the content, early diagenetic degradation, or evaluated environmental loading of organic matter may influence both Hg and Hg/TOC enrichments 30 . In the case of the Zbrza section, in both the Sandbian and Rhuddanian stages, Hg was correlated with TOC content (R 2 = 0.2 and 0.56, respectively), which does not indicate environmental Hg loading, especially in the case of Rhuddanian organic-rich shales. The opposite was true in intervals connected with LOME. No correlation was found between Hg and TOC content for either the late Katian (R 2 = 0.01) or the Hirnantian (R 2 = 0.07) (Fig. 2), suggesting that this pattern of Hg concentration was not a result of changing TOC content 30 . Moreover Jones et al. 30 emphasised that the Hg/TOC ratio may be diagenetically inflated by loss of TOC during burial. In the Zbrza section all samples exhibit a rather high TOC value (specifically, ≥0.2 wt%), with the exception of sample ZB 28.45 (at the Katian-Hirnantian boundary), which has a low TOC content (0.1 wt%). However, due to the generally low level of thermal maturity of the Zbrza samples 33 , the loss of carbon during diagenesis is unlikely (Fig. 2, Table 1).
Moreover, we used Ni/TOC (ppm/wt%) to support the use of mercury as a volcanic indicator. Recently, Rampino et al. 44 suggested a nickel anomaly as a link between the Siberian Traps eruptions and the latest Permian mass extinctions. For the Zbrza section, the Ni/TOC ratio corresponded to Hg/TOC data, with anomalies in the late Katian (Ni/TOC~967 ppm/wt%; Ni~175 ppm, while the background is approximately 65 ppm), on the Katian-Hirnantian boundary (Ni/TOC~404 ppm/wt%), and in the late Hirnantian immediately below the Ordovician-Silurian boundary (Figs 2 and 3; Table 1).

Mercury's independence of redox-sensitive indicators.
The strongest evidence for the occurrence of euxinic (persistently sulfidic lower water column) conditions in the photic zone of the water column is the identification of biomarkers from green sulphur bacteria (GSB) 45,46 . The most powerful and frequently used indicators of euxinia are isorenieratane and its diagenetic degradation products [47][48][49][50][51] and Me,i-Bu maleimides, which are degradation products of bacteriochlorophyll pigments related to autotrophic sulphur bacteria (Chlorobiaceae) [52][53][54] . Framboidal pyrite populations provide another redox-sensitive indicator 35 . Based on Wignall & Newton 35 , we interpreted euxinic conditions for the samples with high number of very low-diameter pyrites (mean 3-5 μm; narrow size range), anoxic conditions for the samples with high number of low-diameter pyrites (mean 4-6 μm; some larger framboids), samples where framboids were moderately common with broad range of sizes were interpreted as dysoxic conditions and oxic conditions for the samples lacking framboids. It is noteworthy that intervals with strong positive Hg/TOC anomalies (late Katian, Katian-Hirnantian boundary, late Hirnantian) are characterised by the absence of 'anoxic indicators' such as biomarkers (aryl isoprenoids, maleimides) or framboidal pyrite (Figs 2 and 3; Table 1). Thus, these Hg spikes appear to be unconnected with organic matter accumulation in anoxic/ euxinic conditions. Moreover, the Zbrza section displays the high Hg/Mo ratio. Mo, as a redox-sensitive element, is associated with pyrite 55 and increases in euxinic conditions. The three highest Hg/Mo spikes occur in the lower late Katian, at the Katian-Hirnantian boundary, and in the late Hirnantian, showing Hg's independence of Mo and other redox-sensitive elements. Gong et al. 31 in their study of Ordovician-Silurian strata from South China, presented Hg and Mo concentrations. As was true of the Zbrza section, their Hg/Mo spikes were related to Hg/TOC anomalies, which they interpreted as Hg flux to the ocean basin.
To sum up, the lack of strong Hg/TOC correlation, in combination with high levels of TOC concentration, the presence of Ni enrichments, and the absence of 'indicators of anoxia/euxinia' (no specific biomarkers, no framboids, very low Mo concentration), supports the interpretation that the positive Hg/TOC anomalies in the lower late Katian, on the Katian-Hirnantian boundary, and in the late Hirnantian (Figs 2 and 3) record enhanced environmental Hg loading.
pre-LoMe and LoMe volcanic event. The Katian and Hirnantian strata in the Zbrza section display both Hg and Hg/TOC anomalies with a rather high concentration of TOC (≥0.2 wt%) at the same stratigraphic interval (Table 1, Fig. 2). For the lower late Katian, Hg reached approximately 3 times the background level (Hg/ TOC ~2500 ppb/wt%) and can be roughly correlated with the mid-Boda change from cool to warm climate conditions 4 . The spike detected precisely on the Katian-Hirnantian boundary shows high values of Hg/TOC (~681 ppb/wt%), as well as a positive anomaly of Ni/TOC (~404 ppm/wt%). The late-Hirnantian anomaly is approximately double the Hg background level (Hg/TOC~470 ppb/wt%) and appears to be coeval with the termination of the Hirnantian glaciation (Table 1, Figs 2 and 4). Similar positive Hg/TOC anomalies were also reported in Wangjiawan (South China) and the Monitor Range (West Laurentia) by Jones et al. 30 , who distinguished the ornatus (=complexus) anomaly in the late Katian, coincident roughly with the lower late Katian anomaly from Zbrza, but characterised by a lower concentration of Hg and moderate TOC content (0.10-0.19%). The Katian-Hirnantian anomaly from the Zbrza section can be compared to the first pulse of the pacificus anomaly 30 with extreme Hg enrichment (500 times background levels). Jones et al. 30 postulate that Hg enrichments are products of enhanced environmental loading driven by LIP emplacement. Thus, the Hg enrichment in the Katian geochemical record (the ornatus anomaly) is interpreted as a volcanic event, that triggered severe cooling 30,56,57 . It has been suggested that the upper pacificus anomaly is connected with a volcanic eruption which triggered an albedo catastrophe and the rapid expansion of ice sheets 30 . The presence of numerous K-bentonites in the Upper Ordovician of the Zbrza section suggests that the Hg spikes may be related to increased delivery of volcanic ash. Trela et al. 32 postulate that pyroclastic material was transported to the HCM by westerlies from the Avalonian volcanoes. Alternatively, Shen et al. 58 claimed that expanded eukaryotic algal production contributed to a higher level of export efficiency during the late Katian, resulting in increased organic carbon burial and drawing down of CO 2 which ultimately triggered the Hirnantian glaciation. Enhanced algal production may have been connected www.nature.com/scientificreports www.nature.com/scientificreports/ with the delivery of nutrients to the ocean from volcanic sources. In the Zbrza section, the late-Hirnantian Hg anomaly was probably connected with the second phase of LOME, which can be linked with the global environmental effects of LIP-associated CO 2 (marine transgression, warming, anoxia) 4,42 . Jones et al. 30 argued that the late-Hirnantian warming was caused by CO 2 release during the later phase of LIP emplacement and reduction of SO 2 emissions 59 as aerosol albedo forcing waned. www.nature.com/scientificreports www.nature.com/scientificreports/

Conclusions
We have identified three mercury spikes from the Katian and Hirnantian mudstone facies in a deep shelf area of the peri-Baltic area (Holy Cross Mountains, Poland) which we link to volcanic pulses that triggered massive climate changes during the Late Ordovician mass extinction. A comparison of Hg concentration with biomarker, pyrite framboid, and molybdenum data showed that the intervals of organic matter enrichment are associated with anoxia, but are not related to the late-Katian and late-Hirnantian Hg anomalies. As in the case of other mass extinctions, we recorded Hg/TOC anomalies immediately below the interval of the second pulse of the Ordovician extinction. Two volcanic pulses are directly related to warm periods during the Ordovician and Silurian, while the third, precisely on the Katian-Hirnantian boundary, corresponds to Late Ordovician cooling (glaciation).
Although no late-Katian-Hirnantian LIPs have been identified to date, we conclude that our Hg and Hg/ TOC anomalies in the late Katian and Hirnantian were associated with volcanic pulses which triggered major environmental changes and, consequently, the Late Ordovician mass extinction. Since there is no direct geological or geochronological evidence for late-Katian-Hirnantian LIPs, more studies are required. Identification of a Late Ordovician LIP would indicate that the LOME was due to a volcanism-driven climate changes mechanism.