Assessment of Hg Speciation Changes in the Sedimentary Rock Record From Thermal Desorption Characteristics

Sedimentary mercury (Hg) has become a widely used proxy for paleo‐volcanic activity. However, scavenging and drawdown of Hg by organic‐matter (OM) and sulfides are important non‐volcanic factors determining variability in such records. Most studies, therefore, normalize total Hg (HgT) to a Hg “host‐phase” proxy (e.g., HgT/TOC for OM, HgT/TS for sulfides), with the dominant host‐phase determined based on the strongest observed (linear) correlations. This approach suffers from various non‐linearities in Hg‐host‐phase behavior and does not account for succession‐level, let alone sample‐level, Hg speciation changes. Thermal desorption characteristics or “profiles” (TDPs) for many Hg species during pyrolysis analysis are well‐established with applications including distinguishing between OM‐bound Hg and different Hg sulfides and oxides in (sub‐)recent sediments. We explore the use of TDPs for geological sediment (rock) samples and illustrate the presence of multiple release phases (Hg species)—correlated to geochemical host‐phase—in (almost) all the 65 analyzed Tithonian (146–145 Ma) silt and mudrock samples. By quantifying the Hg in each release phase for every sample, we find TOC concentration may determine ∼60% of the variability in the first (lower temperature) Hg TDP release phase: a stark difference with the total Hg released from these samples, where ∼20% of variation is explained by TOC variability. TDPs provide insight on sample‐level Hg speciation and demonstrate that, while the common assumption of single‐phase Hg speciation in sedimentary rocks is problematic, differences in Hg speciation can be detected, quantified, and accounted for using commonly applied techniques—opening potential for routine assessment.


Introduction
Mercury is released into and cycled through the atmosphere-ocean system by various geological, anthropogenic and biological processes (Lyman et al., 2020).Geological release processes include natural heating of sedimentary rocks (coal/other hydrocarbon combustion), erosion of Hg-bearing rocks, and notably via volcanic and hydrothermal activity (Selin, 2009).Among these naturally occurring processes, volcanic Hg emissions are a major component with an estimated flux of 76-700 metric tons yr 1 (Bagnato et al., 2015;Geyman et al., 2023;Pyle & Mather, 2003).Most volcanic mercury emissions are in the form of gaseous elemental mercury, which has an atmospheric residence time of up to 2 years (Lyman et al., 2020), enabling hemispheric or even global distribution (Selin, 2009).Following deposition from the atmosphere (e.g., Jiskra et al., 2021), Hg is assumed to be scavenged and sequestered into sediments and held in an absorbed or chemically bound form by a host-phase.Together with the host-phase, typically assumed to be organic-matter (OM) and sulfur groups within OM (thiols) or sulfides, Hg is ultimately stored in the geological record (e.g., Amos et al., 2014;Fitzgerald et al., 2007;Ravichandran, 2004).Mercury is generally assumed to be primarily hosted by OM upon deposition in aquatic sediments (Wallace, 1982), and this is widely supported by Hg data from soft sediments.However, it is also wellestablished that, even when OM-associated, Hg can be hosted by other (OM-associated) phases (Pham et al., 2014) and that OM may not always be the dominant host in all environments.Furthermore, the association with diagenetic pyrites shows that Hg speciation can change after deposition and (burial) diagenesis (Shen et al., 2019;Zhao et al., 2022).
Mercury speciation determines how (im)mobile Hg might be during erosion (Manceau et al., 2018) or upon heating, such as during hydrocarbon formation, production and burning (Yudovich & Ketris, 2005), but also soils subjected to intense forest fires (Biswas et al., 2007), and sediments near intrusive and eruptive volcanic activity (Chen et al., 2022;Liu et al., 2022;Svensen et al., 2023;Zaputlyaeva et al., 2020).It is difficult to gauge the importance of geological Hg fluxes from any of these potential sources without knowledge of Hg speciation in (sedimentary) rocks.

Sedimentary Hg Host-Phases and Use of Hg as a Paleo-Volcanic Proxy
In addition to environmentally relevant Hg sources, sedimentary Hg has become a widely applied proxy for largescale paleo-volcanic activity, particularly large igneous province (LIP) volcanism.To do so, most studies make the a priori assumption that, through time, variations in sedimentary Hg concentration are primarily controlled by the abundance of a Hg host-phase.The abundance of these host-phases is typically estimated from parameters such as total organic carbon (TOC, a proxy for OM), total sulfur (TS, often used as a proxy for sulfides but also includes organic-S and sulfates), or aluminum concentrations (a proxy for clay abundance).The host-phases for Hg are closely linked to Hg speciation but it is important to note that Hg speciation typically refers to the chemical species of Hg (e.g., methylated Hg, HgS, HgO) and their relative importance and that, for example, a single Hg "host-phase" such as OM or pyrite can be composed of several Hg-species.In order to seek out enhanced Hg loading, almost all studies to date have "normalized" total Hg (Hg T ) to a geochemical parameter that serves as a proxy for a particular host-phase.Identifying the host-phase has largely relied on linear correlation for multimillion-year successions and across severe environmental perturbations by assuming that the strongest correlation with another analyzed parameter (usually TOC or TS) reflects the dominant host-phase for all analyzed samples (e.g., Grasby et al., 2019;Percival et al., 2015;Zhao et al., 2022).
There is theoretical and empirical support for this approach.The abundance of particulate and dissolved OM is an important factor in controlling the Hg levels in seawater (Wallace, 1982) and during sequestration in sediments (Fitzgerald et al., 2007;Gehrke et al., 2009;Outridge et al., 2007).This relationship is due to the very high affinity of Hg 2+ for OM and organic sulfur functional groups (thiols) (see, e.g., review of Ravichandran (2004)), which commonly leads to strong correlations between TOC and Hg (Gehrke et al., 2009;Outridge et al., 2007).In these cases, the correlation between Hg and TOC supports OM as the host-phase of the Hg within the sediment (wherein the Hg is likely chemically bound in organic compounds and minerals).Provided redox and diagenetic conditions and the Hg and TOC flux are unchanged over the analyzed sequence (Frieling et al., 2023;Grasby et al., 2013;Shen et al., 2020), this should lead to a stable Hg T /TOC in soft and lithified sediments.Anomalies above a Hg T / TOC baseline have most commonly been attributed to an increased Hg flux, the strongest of which may then be attributed to elevated volcanic activity in large igneous provinces (LIPs) (Percival et al., 2015;Sanei et al., 2012).As the use of sedimentary Hg as a proxy for (LIP) volcanism has become more prevalent, several inherent complexities that may influence Hg T signals in the geological record have come to light.Many of these issues are centered around the Hg-host phase relationship.
Until now, studies have been mostly limited to statistical correlations to derive the (dominant) sedimentary hostphase of Hg.This is especially challenging as proposed Hg carriers (OM, sulfides, and clays) tend to be inherently coupled or co-varying in sedimentary records and at the same time are heterogeneous compounds.In reducing conditions, including anoxic and euxinic environments, Hg can be present as HgS or be hosted by sulfide minerals such as pyrite (adsorbed or as HgS) (Manceau et al., 2018;Pham et al., 2014), and for lithified samples, elemental mapping yielded evidence of pyrite-hosted Hg (Shen et al., 2020;Z. Wang et al., 2020).Lastly, in settings where other scavenging ligands are rare or difficult to analyze, other sediment constituents such as iron-oxides and clay fractions have been proposed to serve as Hg hosts (e.g., Percival et al., 2018).
Whilst normalizing to the best-correlating host-phase proxy parameter (TOC, TS, Al) is currently common practice, there are important shortcomings and challenges with this approach that affect the interpretation of the geological Hg T record.First, even a perfect linear correlation does not always signal that Hg loading was driven by host-phase abundance.Variable mixing of low-Hg T , low-TOC and high-Hg T , high-TOC phases would also impose such a correlation.End-member mixing effects could be imposed by, for example, carbonate or siliciclastic fractions and impact the geological record (Fendley et al., 2024;Frieling, Mather, et al., 2024).Second, the host-phase relation can be complicated by variable fluxes in Hg or TOC, S or clay, which could induce a system change from receptor (Hg burial limited by the availability of the scavenging/host-phase) to flux limitation (Hg burial limited by the availability of Hg) (e.g., Fendley et al., 2024;Frieling et al., 2023;Machado et al., 2016;Schuster et al., 2018).Such systematic changes can result in non-linearity of the Hg-host relation.Non-linearity typically expresses as strongly inflated Hg-host phase normalized values when host-phase abundance is low and suppressed normalized Hg when host-phases are abundantly present (Fendley et al., 2024;Frieling et al., 2023;Schuster et al., 2018).
In light of these complexities, perhaps the most problematic and fundamental assumption is that Hg is typically assumed to be bound within a single dominant host-phase throughout a succession and that this phase can be identified via (linear) correlation.As (sub-)recent sediments often show strong spatial and temporal heterogeneity in Hg speciation (e.g., Chakraborty et al., 2014;Gamboa Ruiz & Tomiyasu, 2015;Lim et al., 2020;Wilman et al., 2023) it is likely that the same applies to the geological record and that the importance of various Hg species and hence host-phases (subtly) changes through the analyzed successions.The likelihood of changing speciation increases when the analyzed sequences span substantial environmental perturbations as they are often accompanied by broad biogeochemical and lithological changes.Such perturbations are inherent to the vast majority of the records to which Hg T is applied as a paleo-volcanic proxy (e.g., Grasby et al., 2019).Single-variable normalization aids in removing part of the host-phase driven variability, but it inevitably also introduces false variability in other sections or for single samples.An extreme example of this would be Hg focusing in diagenetic pyrite beds (Gong et al., 2017;Shen et al., 2019).In this case, if one assumed that Hg was hosted by a single dominant (non-pyrite) phase throughout a succession and remained unaware of the focusing effect, then the natural conclusion would be that environmental Hg loading was variable (Gong et al., 2017).In another scenario, the dominant host-phase may switch from OM to pyrite between samples in a binary fashion within a section.If normalized exclusively to either host, it might look as if Hg levels varied, but in fact the conclusion that no change in environmental Hg loading occurred could be equally valid (Shen et al., 2019).
While linear single-variable host-phase normalizations for entire successions may be crude, they have provided insight into the most severe Hg cycle changes in deep time (Jones et al., 2019;Percival et al., 2017;Sanei et al., 2012).The complexities and pitfalls that surround host-phase identification based on correlation alone are however obvious and require close attention to increase our skill in identifying and quantifying more subtle changes in past Hg cycling.We therefore seek to expand knowledge on the Hg T host-phase relation by applying a methodology that is less dependent on succession-level characteristics and correlations with other data in an attempt to avoid some of the confounding factors that hamper proper Hg-host phase determination.Specifically, we aim to test whether constraints on Hg speciation can help inform whether host-phase correlations and normalizations are justified.

Hg Speciation Techniques
Existing techniques to directly determine Hg speciation can be broadly divided into categories that center on chemical (sequential) extraction, mineral identification, element mapping and thermal desorption.Sequential extraction techniques have been developed using solubilizing Hg from Hg-binding phases (Gleyzes et al., 2002;Issaro et al., 2009).The main strength of sequential extraction is that it can distinguish between many Hg species.However, the analyses are generally time-consuming, and can suffer from variable recovery due to volatility of Hg and solubility characteristics of some phases (Bloom et al., 2003;Issaro et al., 2009;Kim et al., 2003).
Studies have applied synchrotron or other mineral-identification techniques to Hg within rock samples, but these approaches are best suited to materials with >ppm-level Hg concentrations, typically encountered in ore minerals and very rare in sedimentary rocks (Kim et al., 2003).Energy dispersive spectroscopy (EDS) element mapping can be sensitive enough to detect concentrated Hg in samples with sub-ppm total Hg concentrations.Mapping is a powerful tool that may identify which (mineral) phases concentrate elements, and this has established, for example, that besides typical Hg ores, (diagenetic) pyrite in sediments may hold significant amounts of Hg (e.g., Shen et al., 2019;Y. Wang et al., 2023;Z. Wang et al., 2020).Translating the Hg from two-dimensional element mapping to relative proportions of Hg hosted by specific phases for an entire sample, however, remains a challenge.In addition, this approach is time-consuming, such that realistically it cannot be applied to the sometimes >100s of samples that are often analyzed for a single study.
To expand to the sample volume often required to make inferences of geological Hg cycling, a more promising category for rapid Hg speciation of rock samples is the use of thermal properties of Hg species to disentangle Hg speciation (e.g., Bombach et al., 1994).Many dedicated Hg analyzers are based on pyrolysis-based methods, whereby the sample material is rapidly heated to a temperature where most naturally occurring Hg species are desorbed (>700°C).Mercury is quantified through atomic absorption spectroscopy, with or without an intermediate step where Hg is amalgamated onto gold (gold trapping) (e.g., Enrico et al., 2020).Over the past decades, several studies have established and utilized thermal desorption profiles (TDPs) of Hg species, including, for example, elemental Hg, OM-bound Hg, HgS ((meta-)cinnabar), HgO (montroydite), HgSO 4 and HgSe (tiemannite) to determine their abundance in natural and polluted sediment samples (e.g., Bełdowska et al., 2018;Biester & Scholz, 1997;Bombach et al., 1994;Rumayor, Diaz-Somoano, et al., 2015;Rumayor, Lopez-Anton, et al., 2015;Rumayor et al., 2013;Saniewska & Bełdowska, 2017;Petranich et al., 2022; Table 1).With a focus on environmental Hg and the adverse ecological effects of certain Hg species, these studies have generally applied step-wise or gradual temperature ramps to monitor the release of Hg species with increasing temperature.A relatively shallow temperature ramp is potentially beneficial as it may limit concurrent Hg release from multiple species (Bełdowska et al., 2018;Reis et al., 2015).Although less precise in determining the exact Hg species compared to sequential extraction or element mapping, due to overlapping desorption temperatures of compounds, thermal desorption has proven potential to reliably detect and resolve changes in relative abundances of most common naturally occurring Hg species (Biester & Nehrke, 1997).Crucially, this technique requires limited or no (chemical) pre-treatment, analysis time is typically ∼ minutes per sample and the method is broadly and directly applicable as the majority of Hg data for geological proxy work are generated using pyrolysis-based methods.

Study Aims and Approach
It is evident that information on Hg speciation is required at the sample level to test whether and to what extent the assumptions on host-phase behavior and host-phase normalization that form the basis of Hg as a proxy for paleovolcanic activity are justified.We emphasize that it is our intention to develop a method to assess fundamental aspects of the Hg data that should be followed by-but not replace-further rigorous statistical testing for potentially anomalous Hg concentration values, appropriate host-phase normalized values and Hg mass accumulation rates to infer enhanced paleo-volcanic activity.
To this end, we test whether the detailed knowledge of the thermal behavior of Hg species can be widely employed to provide insight into Hg speciation in individual geological samples.First, we determine whether Hg thermal desorption from lithified sediment samples varies systematically, and how these variations might manifest.Subsequently, we examine whether the thermal desorption during total Hg measurement via the rapid pyrolysis-based methods that are commonly used to determine Hg T in geological samples (<3 min per sample), rather than the much longer (>>10 min) temperature ramps or thermo-scanning common in environmental sciences (e.g., Saniewska & Bełdowska, 2017), can be used as a screening tool or even a basic quantitative estimate of Hg speciation in geological samples.Establishing such a method could provide a feasible way to monitor Hg speciation (changes) for each analyzed sample with limited additional effort, at the same time as Hg T is obtained.
For the purpose of testing and establishing such a method and in order to understand the background variability before applying it to more complex time spans, we selected a set of sedimentary rock samples that are thought not to be associated with any known LIP activity.Thus, we analyzed multiple aliquots of three different standard materials to assess reproducibility, and a set of sedimentary rock samples from the Tithonian Kimmeridge Clay Formation (KCF) (Late Jurassic,.Astronomically paced environmental changes drive the main lithological variability in this sedimentary succession from the Swanworth Quarry borehole, UK (Huang et al., 2010;Weedon et al., 1999) and no large-scale volcanic events (e.g., Ernst et al., 2021) or local activity that may have resulted in variable environmental Hg loading is known for this period.

List of Hg-Species That May Occur in Geological Sediment Samples and Their Thermal Desorption Characteristics
Hg species

Materials
The natural geological samples analyzed for this study were obtained from a cored borehole at Swanworth Quarry (SQ-1) (50.604°N, 2.044°E).The SQ-1 bore hole, together with a second bore hole at the same location and one at Metherhills (50.613°N, 2.123°E), represents a near-complete record of the Kimmeridge Clay Formation (KCF) (Morgans-Bell et al., 2001) (Figure 1).Deposited in the Wessex Basin (present-day United Kingdom) during the Upper Jurassic Kimmeridgian and Tithonian stages (ca.153-145 Ma), it was inferred from various sediment proxy data (pyrite framboids, uranium enrichment and biomarkers signaling the presence of photosynthesizing green sulfur bacteria) that intervals of the KCF were deposited in anoxic-euxinic bottom waters (Raiswell et al., 2001;Tyson et al., 1979;Van Kaam-Peters et al., 1998).Besides strata deposited under anoxic conditions, the formation comprises a range of lithologies including mudstones, marls, and shales with lesser amounts of limestones, siltstones and dolostones.The cyclical deposition of these facies and the environmental changes they reflect are thought to be the product of orbital forcing (Huang et al., 2010;Weedon et al., 1999).The KCF represents a range of natural conditions deposited without significant local volcanic events or LIP emplacement, and we therefore assume changes in environmental Hg loading and speciation are the result of non-volcanic processes.
In the absence of strong perturbations to the Hg cycle, variations in Hg T and Hg speciation will primarily be controlled by host-phase availability, dilution by Hg-poor fractions (carbonates) and changes in redox conditions or diagenesis (e.g., Frieling et al., 2023;Outridge et al., 2007;Sanei et al., 2012).In this study, we focused on the 113-38 m composite depth (mcd) interval of the SQ-1 borehole (Figures 2 and 3; described by Gallois (2000)), which was deposited in the Tithonian (ca. 146-145 Ma;Hesselbo et al., 2020).The studied interval spans the entire fittoni and the upper parts of the rotunda ammonite zones and is marked by an overall regressive phase (Morgans-Bell et al., 2001).All samples used here were selected by previous studies at regular intervals and analyzed for TOC and total inorganic carbon (carbonate) (Jenkyns et al., 2001;Morgans-Bell et al., 2001).The studied succession is composed of interbedded siltstones and mudstones of variable thickness but also marked by organic-rich (bituminous) mudstone horizons at 71.45-72.25 mcd and 87.35 mcd and calcareous mudstones from 97.8 to 113 mcd.Based on the astronomical pacing of variations in the TOC concentrations and borehole measurements (Huang et al., 2010), our studied section spans ca.840 Kyr and starts ca.1.2 Myr younger than the material studied by Percival et al. (2015) from the same borehole (Figure 2).Percival et al. (2015) showed that Hg and TOC varied in parallel and are reasonably well correlated for the TOC-rich lower part of the succession (R 2 = 0.47), resulting in mostly steady, low Hg/TOC ratios (5-10 ppb/wt%), a value similar to that of subrecent, unpolluted marine organic-rich (≥1% TOC) sediments (e.g., Frieling et al., 2023;Leipe et al., 2013;Shen et al., 2020).

Methods-Total Hg Measurements
Total sedimentary Hg (Hg T ) was analyzed at the University of Oxford on 200-400 mg of finely powdered rock for 363 samples using a Lumex RA-915 Portable Mercury Analyzer paired with a PYRO-915+ pyrolysis unit.Sample material was placed in a furnace at ∼700°C, volatilizing Hg, which is then oxidized to Hg 0 in air and quantified continuously via spectral absorption, without any trapping or preconcentration.Duplicate sample analyses (n = 15) showed that typical analytical reproducibility was <10% and accuracy is ±6% based on longterm observations of a high-mercury (290 ± 9 ppb) soil standard (National Institute of Standards & Technology (NIST) 2587)(see, e.g., Frieling et al., 2023).Thermal desorption profiles were generated for a subset of 65 samples using a Lumex RA-915M also set-up with a PYRO-915+.The Hg yield from the soil standard (NIST 2587) was routinely checked with two other sediment NIST standards (#1944, New Jersey waterway sediment, 3,400 ± 500 ppb and #2782, Industrial sludge, 1,100 ± 190 ppb).Mercury is reported as sedimentary Hg concentrations (in ng g 1 or ppb) and normalized to weight per cent total organic carbon (Hg/TOC) (ppb/%).These NIST standard materials were also used alongside natural sediments to illustrate differential Hg desorption.

Methods-Thermal Desorption
Desorption profiles of natural samples (n = 65), sediment replicates with variable weights (n = 15) and standard materials (n = 30) were obtained with a Lumex RA-915M paired with a PYRO-915+ pyrolysis unit at the University of Oxford.The 65 samples analyzed for thermal desorption profiles were selected to obtain a sample set proportionally representative of the range of Hg concentrations, lithologies and stratigraphy covered by the high-resolution data set.All Hg measurements were obtained in the highest-temperature mode (Mode 1) but without "boost," so that the temperature in the furnace remained stable throughout the analyses.The temperature of the pyrolysis chamber in this mode is specified as ∼700°C.Samples at room temperature (20°C) were introduced into the furnace and rapidly heated to 700°C, effectively imposing a short, steep, temperature ramp.Each sample typically released Hg during 1-3 min.In addition to replicates of standard materials, two randomly chosen samples from the lower KCF (previously analyzed by Percival et al. (2015)) for which sufficient material (several grams) was available were run in replicates (n = 5 and n = 10) to illustrate the effects of variable sample weights.To obtain consistent desorption profiles for all other sedimentary rock samples (with relatively low Hg concentrations), the weight of each sample was held constant at a weight sufficient to allow reproducibility of low-Hg samples (average 202 ± 5 mg).(Jenkyns et al., 2001;Morgans-Bell et al., 2001) as were sedimentary mercury data for the intervals below 200 mcd (Percival et al., 2015).Abbreviated ammonite zonations: Aulacostephanus euxodus = euxodus, A. autissiodorensis = autissiodor.,Pectinatites elegans = ele., P. scitulus = scitul.,P. wheatleyensis = wh., P. hudlestoni = hudlestoni, P. pectinatus = pectinatus, Pavlovia pallasioides = pallas.,Pavlovia rotunda = rotunda, Virgatopavlovia fittoni = fittoni, Progalbanites albani = albani.Abbreviated lithological units: Lower Hounstout Silt = LHS, Hounstout Clay = HC, Upper Hounstout Silt = UHS, Portland Sand = Portland.
The release profiles for each sample were obtained simultaneously with standard Hg T and the detector signal at 1 s increments was extracted post-run from the raw time-series output (Figure 5).As the timing and thereby the temperature ramp for each measurement needs to be identical to align the samples, we marked the "start" of a sample measurement immediately upon inserting the sample into the furnace (although this is marked at t = 10 s in terms of the sample run, see Figure 5).The raw sample data from the entire run were subsequently extracted using the same sample start and end point flags as recorded in the Lumex's "RAPID" software.The baseline intensity signal was calculated from two 10-s intervals before and after the sample was inserted into the furnace (Figure 5).The baseline is subsequently subtracted from the raw detector signal to obtain the desorption profile.
After this step, the area below the intensity curve can be integrated in a similar fashion as is performed via the "RAPID" software.The integrated area, which is converted to total analyzed Hg (in ng), from the offline processed data typically does not deviate more than a few tenths of % from the standard "RAPID"-estimates and can be converted to Hg release and concentration with a standard-calibrated detector sensitivity and the sample mass.The desorption profile data for each sample is isolated from the raw detector data and provided as separate comma-separated files along with example files and R-code to extract the profiles in Supporting Information S1.

Analytical Considerations and Limits
To examine the profiles, we calculated the signal intensity in Hg mass released per second for each gram of the (total) sample material (in ng g 1 s 1 ) (e.g., Figure 6).The absolute release intensities (in ng g 1 s 1 ) are suitable to assess at which stage most Hg is released from the sample material and to directly examine which fractions may be enriched within specific samples.Alternatively, an area-normalized signal (i.e., a fraction of total Hg release, Figures 7e and 7f) can facilitate direct comparison of profiles from samples with (strongly) varying concentrations, providing insight into changes in proportional importance of Hg release phases.
Analytical noise typically does not play an important role in the reproducibility of total Hg measurements; even at low Hg concentrations and sample weight the total Hg released from geological samples is usually well above the total manufacturer-specified detection limit (∼0.1 ng Hg).However, it is a more important consideration for desorption profiles as we used partial signals and corresponding lower overall signal and intensity for individual 1-s increments.The typical standard deviation (noise) on the instrument's detector signal was observed to be of the order of 0.8 units, which converts to ∼0.001 ng Hg s 1 .Even natural samples with low Hg concentrations (∼15 ppb) and relatively small sample weights (<100 mg) yield signal intensities at least an order of magnitude above this level (Figures 6d and 6e).
For lower concentrations and more complex release patterns in natural samples (Figure 6), we tested the impact of analyzed sample mass.We found that variable weights impact Hg release profiles and sample mass can appreciably influence the release fraction (Figure 6f).This is likely due to thermal inertia of the larger sample masses (>50 mg) combined with relatively higher abundance of high-temperature Hg phases compared to the standard materials.It is therefore recommended that sample weights are kept similar to compare samples, certainly when low-Hg samples (<10 ppb) are included in the analytical run as they often require larger sample masses (>100 mg) to reliably measure.Weight effects in standard materials were insignificant, because the sample weights are typically smaller (<50 mg) and Hg releases at low temperature resulting in generally narrower peaks (Figure 5).

General Observations on Natural Samples
Mercury concentrations are low (7-15.5 ppb) for the section studied here, and always well below the shale average (∼63 ppb) of Grasby et al. (2019), although it should be noted that we focused on a period without known LIP volcanism, unlike most of the works included in their compilation.Total organic carbon values are typically between 0.6% and 1.2% for the same samples, resulting in an average Hg T /TOC of 7.9-16.6ppb/%.The Hg T concentration of the section analyzed for this study is lower than that of the interval analyzed by Percival et al. (2015), as is the TOC (Table 1), likely due to better ventilation in the gradually shallowing environment (Morgans-Bell et al., 2001).In contrast to the lower Hg T and TOC, the Hg/TOC is higher (7.9-16.6 compared to 4.0-10.7 ppb/%) (Figure 2d).In general, a higher Hg/host-phase at lower host-phase abundance and vice versa (Figure 4c) are indicative of non-linearity in the Hg-host-phase relation, which has been observed previously (Fendley et al., 2024;Frieling et al., 2023;Grasby et al., 2019).Besides analytical limitations on the host-phase measurements (Grasby et al., 2019), these non-linearities can stem from oxidation effects on host-phase abundance on the low-end and evasion of Hg or dilution by the host-phase at the high-end (Fendley et al., 2024;Frieling et al., 2023).Regardless, the Hg T -TOC correlation for the entire succession (combined data from Percival et al. (2015) and our study, n = 604, Table 1) is strong (R 2 ∼ 0.66) and thus, overall, TOC variability appears most closely aligned with Hg T variability.Morgans-Bell et al. ( 2001) estimated inorganic carbon (TIC) content from the difference between total carbon (untreated material) and TOC (acid-treated, decarbonated) and assuming that all inorganic carbon resides in CaCO 3 , multiplied TIC by 8.333 to obtain carbonate content.Carbonate shows a weak but significant negative correlation with both TOC and Hg, which likely indicates a degree of carbonate dilution of these phases (Fendley et al., 2024).Carbonates in the KCF comprise a wide range of both primary and secondary phases (calcite, dolomite, siderite; Morgans-Bell et al., 2001;Scotchman, 1991) that are likely to have a diluting effect on non-carbonate compounds.We therefore also show the carbonate-free fractions (Figures 2b,2c,3b,3c,and 4b,4d) but note that this does not affect Hg/TOC as both are generally assumed to be exclusively hosted in the non-carbonate fraction.
When the data from the succession are examined in subsections and specific lithologies, we find that the Hg T -TOC correlation is much weaker for the interval analyzed for this study (R 2 = 0.22, n = 363) than for the entire KCF succession.Although the correlation for a small subset of our samples characterized as organic-rich (bituminous) mudstones is somewhat stronger (R 2 = 0.42, n = 10), the same cannot be said for siltstone (R 2 = 0.13) and carbonate-rich mudstone lithologies (R 2 = 0.04) (Table 2).In the following paragraphs we explore whether thermal desorption profiles can shed light as to why the Hg-TOC appears to be much weaker in the here studied interval.

Insights From Thermal Desorption Profiles
Thermal desorption profiles (TDPs) were analyzed for a representative subset of 65 samples from the upper interval (113-30 m) (Figures 7-11).We employ TDPs to test whether the poorer correlation in Hg-TOC in the  (d and e) Two low-Hg samples (16.6 ± 0.2 ppb, n = 5 and 21.2 ± 0.6 ppb, n = 10) from the Kimmeridge clay formation.For (a-e), mercury release is shown as ng per gram of sample material per second (ng g 1 s 1 ).(f) Mercury is released as a fraction of total Hg before 40 s in the analysis timeline (dashed vertical line in panels (a)-(e)) versus sample mass in mg.Kimmeridge sediment 2 shows significantly reduced release with increasing mass (significance of the linear correlation p ∼ 0.002).Profiles in panels (a)-(e) are given different colors to illustrate that they represent individual samples.Kimmeridge sediments 1 and 2 (Swanworth Quarry 1-94-22 (222.36 mcd) and 92-18, (216.68 mcd)) (panels d-f) were randomly chosen lower Kimmeridge clay formation samples below the stratigraphic interval focused on here (Percival et al., 2015).studied interval is due to the occurrence of multiple Hg-species.First, the thermal desorption profiles are separated by lithology and visually examined (Figure 7).The profiles show a degree of systematic behavior between the various lithologies, strongly suggesting multiple phases of Hg release in most samples, although the organic-rich mudstones are typically dominated by a single phase.Inspection of individual profiles in the mudand siltstones and carbonate mudstones indicated that many have two separate maxima, that is, multiple (double) peaks.The first peak, with a maximum at around 30 s after start of the run, often dominates in amplitude whereas the second peak is more protracted with maximum release around ca. 45 s.When TDPs from different lithologies are compared quantitatively, the organic-carbon rich mudstones have the most straightforward release profile, as most Hg is released in a single dominant peak that centers around ∼30 s and >90% of the total Hg is released before 50 s.For the carbonate-mudstones and mudstone lithologies late-stage Hg release (>50 s) is proportionally much more important (average ∼30%) and the siltstones are in between (∼25% release after 50 s).There is however also significant variability within these lithological groups, which indicates that the lithology is not the only factor in shaping the release profiles.There is however significant variability within these lithological groups, which indicates that the release profiles cannot be easily predicted from lithological changes and raises Figure 7. Thermal desorption profiles for natural samples and standard aliquots analyzed in this study.The sample is inserted into the furnace at t = 10 s (t 0t 10 is used for the baseline) and the area below the curves is integrated to calculate the total and phase-specific Hg concentration (see also Figure 5).questions as to what (syn-or post-depositional) processes control Hg speciation in these samples.Future work specifically designed to detect how specific environmental conditions and resulting lithologies relate to preserved Hg TDPs is required to resolve this potentially crucial unknown.

Evidence for Hg Species, Host-Phases and Relation to Thermal Desorption Peaks
The evidence for multiple release peaks during heating that signals the presence of multiple Hg species (Figure 8) prompts us to consider the host-phase relation for the individual peaks.We therefore examine the release intensity per gram of sample material for specific intervals of the TDPs.In Figure 8, two approaches are shown.First, the cumulative release up to a certain point in the TDP is compared to the total TOC measured in each sample, which shows a distinct peak in correlation-strength (r ∼ +0.75, R 2 ∼ 0.6) at 39 s.After this point, the correlation strength with TOC decreases.Second, we examine Hg release over 5, 10, and 20 s intervals shifting the window by 1-s increments over the duration of the sample run and correlate it with TOC.This confirms that the interval from 20 to 40 s has the strongest positive correlation to TOC (r > +0.75, R 2 > 0.6), as also indicated by the cumulative release up to 39 s.When focused into shorter intervals, we find that the maximum strength of the correlation is further increased to r values around +0.85 (R 2 > 0.7) in the intervals 28-38 s and 30-35 s, suggesting >70% of variance in Hg released in this TDP peak can be explained by TOC variability.This remarkable difference with Hg T /TOC (r ∼ +0.44, R 2 ∼ 0.18, n = 65, Table 3) appears to be due to a secondary peak in Hg release beyond 40 s into the sample analyses (Figure 7).For Hg released >50 s, the correlation for Hg-TOC is weakly (r ∼ 0.3, R 2 ∼ 0.1) but significantly (p < 0.05) negative (Figure 8b).
Although a correlation with TOC may not directly signal that the first Hg peak in the TDPs relates to OM-bound Hg in the TDPs, multiple lines of evidence suggest it is indeed most likely OM or an OM-associated host-phase being desorbed at this stage.First, all analyzed samples contain this phase, and as Hg is presumed (dominantly) OM-bound upon deposition we would thus expect such a phase to be omnipresent unless all OM is broken down (oxidized) after deposition, which is not the case here as TOC rarely drops below 0.5%.Second, for the organicrich (bituminous) mudstone, where the TOC is high, Hg release is focused as a single peak and the Hg T -TOC relation is comparatively strong (R 2 = 0.42).This suggests more closely coupled Hg T -TOC behavior as was found for the more organic-rich KCF interval analyzed by Percival et al. (2015).Finally, previous Hg speciation and thermal desorption studies that have shown that OM-bound Hg in natural sediments is usually the phase from which Hg desorbs at the lowest temperature (Biester & Scholz, 1997;Rumayor, Diaz-Somoano, et al., 2015;Saniewska & Bełdowska, 2017).Other Hg-species (e.g., elemental Hg, Hg-chlorides and Hg-bromides) with lower desorption temperatures are considered more labile and their (abundant) occurrence appears limited to ore materials or polluted sediments (Table 1) (Rumayor et al., 2017).The KCF in places has abundant pyrite (Matthews et al., 2004;Morgans-Bell et al., 2001) and is notable for sulfurization of OM (van Kaam-Peters et al., 1998); when considering OM-bound Hg release, we therefore also consider that a portion of this Hg may in fact be associated with organic sulfur or present as inclusions in (framboidal) pyrite or metacinnabar (black HgS).These Hg species can (appear to) be coupled and correlated with TOC in sediments as they are either directly associated with OM and OM-breakdown processes (Berner, 1982(Berner, , 1984;;Raven et al., 2023;Werne et al., 2003) or preferentially form in the presence of OM (such as HgS, e.g., Pham et al., 2014).This connection implies that the controlling factor for the sedimentary abundance of these phases during deposition might have been organic-matter flux, so that even if other chemical species (e.g., sulfides) now bind Hg, the abundance of these phases may still link to OM-drawdown.We hereafter refer to the first release phase as organic-matter and organic-matter associated sulfur Hg (Hg OMS ) and the secondary release as "other" Hg species (Hg OTH ) (Figure 9).

Other Mercury Release Phases
The later peaks represent Hg species with higher desorption temperature.This higher-temperature release may relate to a range of chemical species, of which sulfides (cinnabar, HgS) are typically considered as important hostphases for geological samples.For example, the various studies of reference compounds (Table 1) show that red HgS (cinnabar) has higher desorption temperatures and broader peaks than OM-bound Hg, which may contribute to the secondary peak in our samples.However, some caution is warranted as the potential diversity in sedimentary Hg-S species (Table 1) poses a major challenge; Hg may be associated with sulfur via an OMassociated sulfur fraction (see also Section 3.4) but also with various sulfides and sulfates.This implies that it is not straightforward to examine the Hg-S relation in low-Hg sediments in detail with any of the currently available statistical or analytical (Section 1.2) techniques.Besides sulfur-minerals, it is also conceivable that the Hg is present as, for example, Hg-oxide (HgO; montroydite), Hg-selenide (HgSe; tiemannite), or Hg-sulfate (HgSO 4 ) Figure 8. Mercury desorption intensity per gram of sample material (in ng g 1 s 1 ) and correlation to the total organic carbon concentration across all 65 samples.(a) Position of the optimum integration windows (as determined via correlations shown in panel b) relative to desorption profiles of the Kimmeridge clay formation samples.(b) Mercury release for rolling windows shifted at 1-s increments (i.e., 1-6 s, then 2-7 s) was integrated and then correlated to TOC.Windows cover intervals of 5, 10, and 20 s.The cumulative integration grows by 1 s from t 1 -t 2 to t 1 -t 3 , t 1 -t 4 etc., in which the final step t 1 -t end is equivalent to standard Hg T /TOC.Note the significant negative correlation coefficients for >t 60 identified in the 5, 10, and 20 s intervals.
Optimal integration for the 20, 10, and 5 s windows are found at t 20 -t 40 , t 28 -t 38 and t 30 -t 35 , defined as the position of maximum r-values.For the cumulative integration, the optimum window is defined as the interval up to and including the highest r-value (t 1 -t 39 ).

Figure 9.
Example of Hg release phases within a sample divided based on the optimum correlation to TOC (cumulative, Figure 8).Dashed line signifies the cut-off between organic-matter and OM-associated sulfur bound Hg (Hg OMS ) and other Hg release phases (Hg OTH ) (here 39 s) (see also Section 3.4).Total Hg (Hg T ) is equal to the sum of Hg OMS and Hg OTH .
although reports of their occurrence in unpolluted sediments are limited and some of these Hg-species are generally considered rare in nature (e.g., HgO; Biester et al., 1999) or unstable as they are easily dissolved or decompose in water (e.g., HgSO 4 ; Haynes, 2014).Lastly, the late-stage Hg release may represent a Hg fraction locked in more refractory mineral grains, which may include pyrite or other metal sulfides (e.g., PbS; galena, ZnS; sphalerite, HgS; cinnabar) but also silicates or (absorption into) diagenetic carbonates (cf.Biester et al., 1999).At the concentrations (<10 ppb Hg) analyzed in this study, resolving and quantifying which phases are present besides Hg OMS will be limited by our ability to directly measure these phases or locate the (rare) minerals that contain them via element mapping.To summarize, while the TDPs have limitations in identifying host-phases beyond OMS (if rapidly heated), we can be reasonably confident that Hg OMS can be separated from other more temperature-resistant Hg-species.

Implications for Hg Normalization
The desorption profiles and optimal correlation window for the Hg release in the individual phases in our samples revealed that in approximately half of the samples (35 out of 65), this first release phase (Hg OMS ) is not dominant (Hg OMS is <50% of Hg T ) (Figure 11c).Hg OTH is roughly equally important as OM-associated Hg at a successionlevel but strongly varies in proportion from sample to sample (Figures 11c and 11d).Moreover, Hg OMS and Hg OTH show a similar correlation with Hg T and the proportion of Hg OMS and Hg OTH in samples does not correlate with Hg T (Figure S1 in Supporting Information S1).These data thus show clear conflict with the common assumption that sedimentary Hg is hosted by a single dominant phase.Though the first phase correlates well with TOC, normalization of Hg T to TOC therefore cannot be justified for these samples.
When following a traditional approach to remove enhanced Hg sequestration by OM, the strong positive correlation found for the entire succession between Hg T and TOC (R 2 = 0.66) might support the normalization of Hg T by TOC.Although much weaker, a correlation remains for the Tithonian subsection analyzed here (R 2 = 0.2), which might be used to support a similar normalization approach.However, the TDP analyses show that the OMassociated Hg cannot be demonstrated to dominate in many samples, and that the fractional dominance of Hg OMS is not confined to specific sequential sample sets or portions of the succession.Thus, the simple Hg T /TOC approach would introduce variability through the section and inflate Hg T /TOC for all samples but to a variable extent.As multiple release phases are present in most samples, the same would likely apply to any other Hg T hostphase normalization.As all samples where Hg OTH dominates Hg T are at relatively low TOC concentrations, proportional differences between Hg OMS and Hg OTH in samples here may also impact assessment of linearity of Hg T -TOC.In this case, the secondary phase appears to anticorrelate with TOC (Figure 8b), which further complicates the interpretation of a TOC-normalized Hg T signal.Aside from the TDP characteristics and other techniques that resolve Hg speciation, there is no obvious way to tell the dominant Hg species or host-phase at a sample-level; the samples where Hg OTH dominates are within the normal TOC range (0.5%-1.5%) for this part of the succession and appear randomly distributed throughout (Figures 10 and 11).These samples also would not be discarded because of low host-phase abundance or analytical limitations and neither do their Hg T concentrations differ from samples for which Hg OMS dominates Hg T .The lithology of the samples may have some impact as the frequency of samples where Hg OMS is not dominant is highest in the carbonate mudstones.Nevertheless, the degree of dominance of Hg OMS impacts Hg T / TOC highlighting the advantage of separating the Hg OMS and Hg OTH phases.A greater proportion of Hg OTH correlates with higher Hg T /TOC.When only the first phase is normalized to the TOC (Hg OMS /TOC) values of all samples approach the Hg T /TOC values of the more organic-rich mudstones and the organic-rich lower parts of the KCF (Figure 10d).
If OM burial was the primary driver of Hg sequestration and abundance in our succession, the variability in TDPs for the more organic-lean samples may reflect Hg-speciation changes during syn-or post-depositional processes.
Mercury speciation changes in sediments may not be uncommon, as for example, Hg enrichment in diagenetic pyrite has been documented in the geological record (Shen et al., 2019;Z. Wang et al., 2020).Here, the occurrence and frequency of multiple phases in carbonate mudstone lithologies is noteworthy.The processes driving carbonate dissolution and (authigenic) precipitation in the KCF and in general are closely linked to OM breakdown (Irwin, 1980;Irwin et al., 1977;Scotchman, 1991), which generates dissolved inorganic carbon that can be reprecipitated with free metal ions (of Ca, Mg, Mn, Fe).If Hg is released during OM breakdown, which may have substantially and variably affected parts of the KCF (e.g., Kodrans-Nsiah et al., 2009), Hg may relocate into a secondary phase.Any diagenetic effect that affects either the host-phase or Hg warrants careful consideration as the potential for relocation implies Hg may (appear to) have been concentrated in certain stratigraphic levels.Such effects may be of a similar magnitude as is associated with some perturbations proposed to be volcanic in origin (Frieling et al., 2023;Mercone et al., 1999;Shen et al., 2019).As mentioned previously (Section 3.5), it is also conceivable that Hg is "locked" in (non-sulfide) mineral grains, which may be facilitated by, for example, sediment incorporation into authigenic carbonates and/or absorption into these phases (cf.Biester et al., 1999).We therefore emphasize that, while we demonstrate that the Hg OMS phase is of variable importance in these sedimentary rocks at the time of our analyses, Hg drawdown may have been facilitated by OM-fluxes at the time of deposition (146-145 Myr ago).

Mercury Speciation and Enhanced Hg Loading
Without the use of TDPs or other sample-level Hg speciation, we follow previous studies (Fendley et al., 2024;Frieling, Mather, et al., 2024;Percival et al., 2021) in their interpretation and recommendation that anomalous Hg loading should still be evidenced by at least two of the following: (a) exceptional Hg concentrations, (b) exceptional normalized or excess Hg, when appropriately corrected for host-phase effects and, ideally, (c) exceptional Hg mass accumulation rates.
The simple Hg speciation approach used here is principally a test to what extent Hg T is impacted by multiple Hg species or host-phases and whether single-phase normalization of Hg is justified.The key strength of the approach is in pinpointing which samples may be false positives or false negatives due to Hg normalization to an unrelated host-phase.Such constraints are particularly relevant in the normalized Hg T data that are ubiquitously used in efforts to trace LIP-volcanic activity in ancient rocks that encompass large-scale environmental changes.This ranges from extreme end-members such as enhanced Hg release from specific fractions in discrete stratigraphic levels (e.g., diagenetic pyrite) (Gong et al., 2017;Shen et al., 2019), which may have focused Hg, to subtler variations between samples such as those observed in our Tithonian sample set.
The Hg TDPs measured on the KCF material show a substantial degree of variability in Hg speciation that is not simple to predict based on available parameters (lithology, carbonate, TOC, from Gallois (2000), Jenkyns et al. (2001), andMorgans-Bell et al. (2001)).Such complexity in Hg speciation is not unexpected as data from (sub-)recent sediments show that Hg speciation can strongly vary through time and space.Moreover, our data indicate that some Hg speciation changes may be of complex diagenetic (post-depositional) or other (syndepositional) environmental origin, as evidenced by the documented speciation behavior in the Kimmeridge clay formation that also contains abundant diagenetic products (carbonates, sulfides).Overall, our results underline the value of acquiring Hg speciation data as a starting point for further study of the origin of the Hg T and Hg speciation signal in the sediments.
Sedimentary Hg speciation impacts our ability to infer paleovolcanic activity as the use of Hg as a proxy for LIP activity strongly depends on detection of anomalous Hg loading through means of host-phase normalization, determination of statistical outliers, or (multivariate) correspondence analyses (Fendley et al., 2024;Kovács et al., 2024;Sanei et al., 2012).The key assumptions (linear, predictable, single-host-phase Hg) that (implicitly) underlie all these methods have so far lacked independent validation in the form of sample-level constraints on Hg speciation.Qualitative and quantitative information on Hg speciation, such as through TDPs, provides a means to better guide the use of these techniques and to test whether the key assumptions to use them are justified.As general examples, we would recommend cautious interpretation of samples with both strongly elevated (normalized) Hg and different Hg speciation compared to surrounding samples, as these may have a greater likelihood of being "false positives."On the other hand, any elevated Hg loading that coincides with a peak in the host-phase proxy (TOC, TS or clays) would be easily missed by aggressive (linear) normalization (heightened chance of false negatives).If such samples are not characterized by a change in Hg speciation, those Hg peaks may require further investigation as to why Hg has accumulated.

Method Applicability and Adaptability
The approach used for this study can be applied directly with Lumex Instruments RA-915M (used here) and the RA-915Lab, with or without auto-sampler, as it is built around the data generated and output from these instruments' software.The main advantage of continuous-flow and continuous Hg recording is that the signal does not require further processing or (pre-)concentration that may impact recovery when analyzing thermal desorption (see also other studies using a Lumex RA-915, e.g., Rumayor, Lopez-Anton, et al., 2015).However, for pyrolysis based direct mercury analyzers with trapping (gold amalgamation), a similar result may be obtained through (stepped) temperature ramping of single samples or analyzing multiple aliquots of the same sample at different temperatures to obtain a thermal release-sequence (cf.Bełdowska et al., 2018;Rumayor, Diaz-Somoano, et al., 2015;Rumayor, Lopez-Anton, et al., 2015;Saniewska & Bełdowska, 2017).If the details of Hg speciation in geological samples become critical to constrain, more laborious and specialized but potentially also more accurate Hg speciation methods such as element mapping, mineral identification or sequential extraction of Hg phases (Section 1.2) can be used alongside the aforementioned pyrolysis-based methods to resolve the exact Hg containing phases (Biester & Nehrke, 1997;Biester et al., 1999;Bombach et al., 1994).

Conclusions
Application of the sedimentary Hg proxy for volcanic activity is often complicated by the poorly constrained influence of the host-phase on Hg concentration and Hg speciation.Identification of host-phase abundance effects (normalization) is commonly performed by choosing the strongest positive linear relation to common host-phase proxy data (e.g., TOC and TS, the proxies for organic-matter and sulfides).Critically, this approach often relies solely on (linear) correlations to infer which is the most likely host-phase and does not account for any complexity in Hg speciation within or between samples.
Here, we explore the use of thermal desorption characteristics for geological sediment (rock) samples.Thermal desorption profiles (TDPs) for many Hg species are well-established and a proven tool to distinguish between OM-bound Hg and different Hg sulfides as well as Hg-oxides and Hg-selenide (e.g., Biester & Nehrke, 1997;Bombach et al., 1994;Rumayor et al., 2013;Saniewska & Bełdowska, 2017).Using only the rapid (<3 min) desorption that is obtained as standard for each sample analyzed in continuous-flow direct Hg analyzers, we illustrate the presence of multiple release-phases (Hg species) in (almost) all of the 65 analyzed sedimentary rock samples of Tithonian (late Jurassic) age .When examining subsections of the long-term data set, we find that the overall correlation between Hg T and TOC is much weaker in the here studied interval compared to the more organic-rich shales below.
Invariably, the sediment TDPs that were analyzed show multiple phases of Hg release and these phases occur in different proportions in different samples.We quantify the Hg release in each phase and show that the first lowtemperature release phase is likely Hg associated with organic-matter and organic-matter associated sulfides (Hg OMS ), whereas a secondary higher temperature Hg phase that anti-correlates with TOC is of a more enigmatic origin (Hg OTH ).Separating Hg OMS and Hg OTH confirms that the weak Hg T (R 2 ∼ 0.2) correlation to TOC corresponds to the variable abundance of Hg OTH (Figure 10b).Specifically, the greater proportion of Hg OTH in carbonate-mudstones may suggest a link to the diagenetic effects of Hg and Hg speciation and could suggest that the Hg OTH currently hosted in the secondary phases may have been deposited with an OM fraction that has since been oxidized.Such diagenetic effects, that were also observed to focus Hg in pyrite can produce significant stratigraphic enrichment of Hg and warrant caution when interpreting the (normalized) Hg record.
Most importantly, the approach used here demonstrates that it is feasible to test for multiple Hg release phases (Hg species) and provides a fast, data-rich, quantitative perspective on Hg behavior in individual sediment samples.Our method allows offline quantification of the individual peaks, which means they can be used flexibly, for example, to test whether Hg release phases may or may not align with certain host-phases.As this approach resolves changes in Hg speciation on a sample level, the "host-phase" identification and quantification does not overly depend on the existence of positive correlation for entire successions and avoids the assumption that all Hg is always hosted in a single phase.Although crude in terms of Hg species identification, the approach can be applied with pyrolysis-based continuous-flow Hg analyzers to any sedimentary Hg record virtually without additional effort and can be extended to pyrolysis-based analyses with instruments that apply gold-trapping prior to quantification.

Figure 1 .
Figure 1.Paleogeographic map (Kimmeridgian) and core location.The location of the Swanworth Quarry and Metherhills boreholes in the Wessex Basin is illustrated by the red star (adapted from Williams et al. (2001)).

Figure 3 .
Figure 3. Stratigraphic overview for the Kimmeridge clay formation cores focused on the 30-115 mcd interval used in this study.(a) Carbonate percentage data.(b) Total organic carbon percentage, and TOC on a carbonate-free basis.(c) Mercury concentration, and on a carbonate-free basis.(d) Mercury normalized to TOC.Note that the x-axis scales of panels (a)-(c) are adjusted relative to those in Figure2to illustrate variability.All carbonate and total organic carbon percentages were obtained from previous work(Jenkyns et al., 2001;Morgans-Bell et al., 2001).Abbreviated ammonite zonations: Pavlovia rotunda = rotunda, Virgatopavlovia fittoni = fittoni, Progalbanites albani = albani.

Figure 4 .
Figure 4. Mercury versus total organic carbon concentration for the combined Kimmeridge clay formation data set.(a) x-y plot of Hg T and TOC, (b) as (a) but on a carbonate-free basis (CFB), (c and d) as (a and b) but zoomed in on the Tithonian samples.Dark colored symbols in each panel denote the Tithonian subset analyzed for this study.

Figure 5 .
Figure 5. Schematic overview of simultaneous acquisition of Hg T and thermal desorption profiles for each sample.Profiles from each of the illustrative desorption analyses are given different colors to indicate that they represent individual samples.
Figure7.Thermal desorption profiles for natural samples and standard aliquots analyzed in this study.The sample is inserted into the furnace at t = 10 s (t 0t 10 is used for the baseline) and the area below the curves is integrated to calculate the total and phase-specific Hg concentration (see also Figure5).(a) Release profiles for each of the carbonate mudstone samples.(b) As panel (a) for mudstones.(c) As panel (a) for siltstones.(d) As panel (a) for organic-rich (bituminous) mudstones.Profiles in panels (a-d) are given different colors to illustrate that they represent individual samples.(e) Average release intensity for lithologies shown in (a)-(d), normalized to total Hg release for each sample.(f) As panel (e) but for the NIST reference sediment replicate analyses shown in Figure 6.Note that the narrower peak-width for standard materials compared to all natural samples relates to smaller weights used for high-Hg standard materials.

Figure 10 .
Figure10.Desorption characteristics for the Tithonian samples.(a) All samples analyzed for this study (gray) and the representative subset analyzed for thermal desorption profiles (black) with weak overall correlation.(b) Cumulative mercury release before 39 s (Hg OMS , top, black) and after 39 s (Hg OTH , bottom, purple, note inverted y-axis scale) for individual samples.(c) Frequency histogram of Hg T normalized to TOC for all samples analyzed in this study compared to the organic-rich lower Kimmeridge clay formation samples(Percival et al., 2015).(d) Hg OMS /TOC and Hg T /TOC of the TDP-informed samples.

Table 2
Mercury (in ppb) and Total Organic Carbon (in Weight Percentage) Data and Correlation for the Entire Succession and by Lithology FRIELING ET AL.

Table 3
Thermal DesorptionProfile-Informed Mercury Release, Fraction of Hg (in ppb) in Total Organic Carbon (in Weight Percentage)