Demystifying mercury geochemistry in contaminated soil–groundwater systems with complementary mercury stable isotope, concentration, and speciation analyses

Interpretation of mercury (Hg) geochemistry in environmental systems remains a challenge. This is largely associated with the inability to identify specific Hg transformation processes and species using established analytical methods in Hg geochemistry (total Hg and Hg speciation). In this study, we demonstrate the improved Hg geochemical interpretation, particularly related to process tracing, that can be achieved when Hg stable isotope analyses are complemented by a suite of more established methods and applied to both solid- (soil) and liquid-phases (groundwater) across two Hg2+-chloride (HgCl2) contaminated sites with distinct geological and physicochemical properties. This novel approach allowed us to identify processes such as Hg2+ (i.e., HgCl2) sorption to the solid-phase, Hg2+ speciation changes associated with changes in groundwater level and redox conditions (particularly in the upper aquifer and capillary fringe), Hg2+ reduction to Hg0, and dark abiotic redox equilibration between Hg0 and Hg(ii). Hg stable isotope analyses play a critical role in our ability to distinguish, or trace, these in situ processes. While we caution against the non-critical use of Hg isotope data for source tracing in environmental systems, due to potentially variable source signatures and overprinting by transformation processes, our study demonstrates the benefits of combining multiple analytical approaches, including Hg isotope ratios as a process tracer, to obtain an improved picture of the enigmatic geochemical behavior and fate of Hg at contaminated legacy sites.

The site geology is made up of Palaeozoic granite underlying quaternary fluvial sediments that contain heterogeneously distributed organic materials derived from stratified peat lenses (Richard et al., 2016a;2016b). This has resulted in a complex two-layer (confined and unconfined aquifers) groundwater system with highly variable flow direction across the site (Richard et al., 2016a). Groundwater depths for the wells at site A are presented in Section S3. SCA1 and SCA2 were drilled in October 2018 and SCA3 was drilled in July 2019. Topsoils were collected in October 2018. All soil core samples were the composite of 5-7 subsamples taken across the depth range of each sample and homogenised by mixing and shaking in polypropylene containers or double Whirl-Pak® bags. Measured samples were taken from this homogenised solid material. Topsoils followed the same procedure except that the 5-7 subsamples were taken from a 1 m 2 area around each sampling site.

SITE B:
In the 1970s, the area of the former industrial facility was converted into a residential zone and a 50 cm layer of uncontaminated material was added to the surface of the soils in this area after discovery of the extent of contamination in the 1990s (Schöndorf et al., 1999;Brocza et al., 2019). Under this layer is a 1 -3 m thick layer of material artificially disturbed by industry and residential development made up of loess, loess/loam, and building rubble. A homogenous loess layer occurs until about 5 -6 m and represents the top of the natural soils/layered sediments, which contain very low organic material (0.5 -1%). Next is a layer of very low organic matter (<0.5 %) fluvial loose gravel deposits we term "Rhine sediments". The unconfined aquifer below this material is highly permeable weathered sand-gravel sediments (flow velocity of 3 -10 m d -1 ; again, low in organic matter: <0.5 %). Finally, the aquitard is encountered between ≈13 -16 m below the surface and is made up of low permeability weathered gravels (Schöndorf et al., 1999;Bollen et al., 2008). SCB1 and SCB2 were both drilled in December 2019, while SCB3 was drilled in May 2018. Soil core sampling at site B followed the methods outlined above for site A.

GROUNDWATER SAMPLING:
Groundwater was sampled five times at both sites: October 1 -2, November 21 -22, and December 11 -12, 2018, and April 1 -2 and July 29 -30, 2019; and additionally, in September 2015 and May 2018 at Site B. Site B wells were always sampled on the first date of each sampling and Site A wells on the second. Not all wells were sampled during each sampling campaign due to (i) well lids frozen in place, (ii) no groundwater -dry wells, and (iii) repeated measurements below detection limits.
A "pump-and-treat" groundwater remediation conducted by an independent consulting firm has been on-going at site B since July 2018. Water was removed from wells WB2 and WB8, treated with activated carbon and Hg chelating scrubbers, and the groundwater scrubbed of Hg was reintroduced into well WB26. This scrubbed solution was also sampled and always below detection limits of the instrumentation. Samples from WB2, WB8, and WB26 were collected directly from taps within the pump-and-treat facility. WB2 was constructed at the same location as SCB2 after removal of the soil core material.

S2. Quality Assurance and Quality Control (QA/QC) Section Total Hg Solid-Phase Analyses:
ERM-CC018 (contaminated sandy soil) was digested along with the solid-phase samples (n = 16) and the recovery was 92.9 ± 6.8 %. NIST-3133 was run throughout these analyses (n = 189), the average accuracy was 102 ± 4 % and RSD for individual session was between 0.8 -6.2 %. Detection and quantification limits were 0.03 ± 0.04 and 0.10 ± 0.13 µg L -1 , respectively. Three ultrapure water field blanks were prepared on-site (one in each of the first three sampling campaigns) and transported back to the lab for analysis with the rest of the samples. All were below detection limits and therefore this was not continued during the later sampling periods as distilled water field blanks were also analysed for the liquid-phase Hg speciation analyses. 1 % BrCl solution was also analysed throughout the analytical sessions and was generally below detection limits (n > 100). Low concentration samples used for sequential extraction were additionally analysed in triplicate on a DMA-80 atomic absorption spectrometer (AAS, Milestone Srl). For quality control ERM CC-141 (loam soil) and NIST-3133 were measured along with the samples with recoveries of 92 ± 10 % (n = 5) and 97 ±4 % (n = 18), respectively.

Total Hg Liquid-Phase Analyses:
Quality control of these analyses was adjudged by repeated analyses of a 0.5 µg L -1 Hg calibration standard run as sample throughout these analyses (n = 35), the average accuracy was 102 ± 3 % and RSD for individual sessions was between 1.2 -6.9 %.

Solid-Phase Hg Speciation Analyses -Pyrolytic Thermal Desorption (PTD):
Due to the need to generate sufficient signal peaks released during continuous temperature ramp of the sample, the detection limit of this method is ≈0.1 -0.5 mg kg -1 THg; samples below this THg concentration were not considered. Irrespective of the area of individual peaks we can quantify the cumulative integrated area of Hg 2+ peaks due to their distinct separation from Hg 0 peaks ( Figure 1). While there is overlap between Hg 0 and Hg 1+ species, calomel (Hg 2 Cl 2 ), the latter is rare and generally believed to occur only at trace levels (if present at all) as a metastable intermediate phase prone to disproportionation (see Section 3.2.3 for details) (Schuster, 1991;Morel et al., 1998;Hazen et al., 2012). Therefore, Hg 0 peaks, Hg 0 fraction (of THg), and Hg 0 concentrations (calculated from this fraction and THg concentrations) can be quantitatively or at least semi-quantitatively assessed.

Solid-Phase Hg Speciation Analyses -Sequential Extraction Procedures (SEP):
The F3 1M KOH extraction step from Bloom et al. (2003) was omitted in our analyses due to the low OM content of the soils and suspected low sorption to the OM as suggested by Bollen et al. (2008). Test were conducted using the Bloom et al. (2003) extraction procedures on eight selected samples. Figure S2.1 shows the F3 1M KOH (organo-chelated Hg species) was small and <6.5% of THg in all samples except SCB2 -180cm. SCB2 -180cm is a relatively low THg concentration (2.16 mg kg -1 ) with little surface Hg inputs except for some contamination associated with redistributed contaminated building materials. These data support the hypothesis that OM plays only a minor role in solid-phase Hg sorption at these sites.  Figure S2.1: SEP results based on Bloom et al. (2003) method that highlight the minor role OM plays in solidphase sorption at these sites. F3 (1M KOH) is typically associated with organo-chelated Hg species (Bloom et al., 2003); data from this fraction are labelled in the figure.
F2 extractant solution buffering by high pH solid-phase materials was also examined experimentally. The pH of the F2 extractant solution was 0.35 ± 0.03 (n = 4) and after addition of ≈1 g of sample from -2.75 m in SCB1 (pH = 11.1) and 20 h on an overhead shaker the pH rose only to 0.85 ± 0.03. This small change in pH is unlikely to have a significant effect on the species released in the F2 fraction. In contrast, previous extraction tests using the F2 reagent of the Bloom et al. (2003) method (0.01 M HCl / 0.1 M CH 3 COOH) on carbonate-rich soil samples from site B resulted in a pH increase from 2.5 (reagent) to up to 5.7 after overnight extraction (Brocza et al., 2019), which is likely to decrease the extraction efficiency of the F2 step. Thus, the chosen F2 extractant may reduce some of the reported SEP limitations associated with changes in physicochemical properties and mineralogy.
It should be noted though that the solid-to-liquid-phase ratio of the Brocza et al. (2019) method applied in this study was 1:25 to increase Hg concentrations for stable isotope analyses of the individual extracts. This is lower than the recommended ratio (1:100) to ensure complete Hg extraction (Bloom et al., 2003). This issue is of greater concern for the labile fractions, F1 and F2 (Bloom et al., 2003), and may contribute to the incomplete extraction of Hg exchangeable with water and generally low combined proportion of Hg extracted in these labile fractions (16 ± 22 % based on all SEP analysed samples).

Liquid-Phase Hg Speciation Analyses:
Since three species/fractions, Hg 2+ A, Hg 2+ B, and Hg-part, are determined by difference and these calculated fractions include the uncertainty of multiple measurements, this can result in negative values for these fractions (Leopold et al., 2010;Richard et al., 2016b). Similar to Richard et al. (2016b), we allocate a criterion for valid samples as those >5 µg L -1 (quantification limits for THg by this method is ≈1 µg L -1 ) and containing positive sample fractions, or a single negative fraction that is within 5 % of the expected higher fraction (the negative fraction is assumed as 0 % in this instance).
Considering the high uncertainty of these speciation analyses, we deem them of a qualitative nature. Quality control for analysis by the Hg-254 Analyzer used a certified 115 µg L -1 Hg 2+ in 5 % HNO 3 standard solution (Sigma Aldrich) and the recovery was 97 ± 6 % (n = 44). The distilled water field blanks were also analysed for Hg species and the concentration for all four measured species was below the method detection limit for the Hg-254 Analyzer (detection limit: ≈0.3 µg L -1 ) in all 13 distilled water field blanks across the sampling campaign.

S3. Groundwater depths, THg and species concentrations, and other groundwater parameters from sampled wells
Liquid-phase pH, redox, conductivity, dissolved oxygen (DO) content, and temperature were measured on site with handheld probes during groundwater removal (Table S3.4 and Table S3.5). Groundwater depths were measured during each sampling using an electrical tape water level meter and are shown in Figure S3.1 and Figure S3.2.   Figure S4.2). Figure S3.3: Historical groundwater depth at well WB18 from site B (Schöndorf, 2020           Continued on next page.

S6. Solid-phase THg, pH, moisture content and Hg stable isotopes
Soil pH was measured with a calibrated glass pH electrode after 1 h equilibration of ≈5 -10 g of wet soil mixed and shaken in 50 mL of 0.01 M CaCl 2 solution (Section S6).
Note: In all tables, letters next to sample names denote replicates.

S7. Sequential extraction procedure (SEP) data & SEP isotope data
Tables S7.1 -S7.5 show wet weight (ww) Hg concentrations, fractions (of sum of extract Hg concentrations), and δ 202 Hg isotope signatures of individual SEP extracts. The ± values of the δ 202 Hg values are the 2 SD values from the MC-ICP-MS analyses. When discussing the recoveries on THg concentration of ΣSEP extracts against measured THg concentration of bulk sample we must consider these are different samples and there is considerable heterogeneity of Hg within the contaminated solid phase material (Schroeder and Munthe, 1998;Miller et al., 2013). Thus, we cannot assume differences in these THg concentrations are attributable solely to SEP artefacts. The δ 202 Hg value from the combined SEP extracts (δ 202 Hg ΣSEP extracts) was calculated from equation S7.1 below: Where [Hg] Fi is the Hg concentration of fraction i. The uncertainty term of δ 202 Hg ΣSEP extracts is fully propagated to include the 2 SD of the analyses, and the mean replicated variability (1 SD) from table S6.8. The uncertainty term for the "δ 202 Hg difference: ΣSEP extracts to measured bulk" is propagated to include the uncertainty of the δ 202 Hg ΣSEP extracts and the 2 SD uncertainty of the bulk measured δ 202 Hg values.    Figure S7.1: Relationship between SEP F3 fraction and PTD matrix Hg peak maximum release temperature for all solid phase soil core samples analysed by both SEP and PTD analyses. This shows a weak relationship between shifts in PTD matrix bound Hg peak towards the upper end of the range (≈190 -300 °C) and an increasing proportion of SEP F3 in the solid phase samples.

S8. Total carbon, organic carbon, and inorganic carbon analyses of solid phase materials
Solid-phase organic carbon (OC) content was determined by first removing the inorganic carbon fraction with concentrated HCl, "carbonate-bomb" method (Müller and Gastner, 1971), combustion of the dried digestate, and analysis by infra-red detection of CO 2 released using a DIMA 1000NT (Dimatec, Germany). The total carbon (TC) content was determined by combusting and analysing the untreated dried material.

S9. ICP-OES data from site B
Major metal cations were measured with inductively coupled plasma optical emission spectrometry (ICP-OES). Samples for ICP-OES were prepared by adding 0.5 g of sample to 50 mL vials capped loosely filled with 12 mL aqua regia. Samples were digested for 2 h at 85 °C on a heating plate. Aqua regia was then refilled to 12 mL and the samples further digested for 1 h. Then samples were filled to 50 mL with deionized water and filtered through 0.45 µm cellulose acetate filters and finally diluted with 2 % HNO 3 before analysis. Figure S9.1: ICP-OES data of other trace metals (zinc, lead, cadmium, and chromium) in SCB1. Graphs show evidence of trace metal enrichment at the bottom of the lower artificial filling layer (Artificial Filling 4), above the loess layer. This provides supportive evidence of "ponding" that likely occurs during rainfall or snowmelt events creating a temporary aquifer where soluble trace metals would accumulate by leaching through the overlying artificial filling layers.