Methylmercury complexes: Selection of thermodynamic properties and application to the modelling of a column experiment

18 19 Complexation with methyl groups produces the most toxic form of mercury, especially 20 because of its capacity to bioconcentrate in living tissues. Understanding and integrating 21 methylation and demethylation processes is of the utmost interest in providing geochemical 22 models relevant for environmental assessment. In a first step, we investigated methylation at 23 equilibrium, by selecting the thermodynamic properties of different complexes that form in the 24 chemical system Hg-SO 3 -S-Cl-C-H 2 O. The selection included temperature dependencies of 25 the equilibrium constants when available. We also considered adsorption and desorption 26 reactions of both methylated and non-methylated mercury onto mineral surfaces. Then we 27 assessed the kinetics of methylation by comparing a dedicated column experiment with the 28 results of a geochemical model, including testing different methylation and demethylation 29 kinetic rate laws. The column system was a simple medium: silicic sand and iron hydroxides 30 spiked with a mercury nitrate solution. The modelling of methylmercury production with two 31 different rate laws from the literature is bracketing the experimental results. Dissolved 32 mercury, iron and sulfate concentrations were also correctly reproduced. The internal 33 evolution of the column was also correctly modeled, including the precipitation of 34 mackinawite (FeS) and the evolution of dissolved iron. The results validate the conceptual 35 model and underline the capacity of geochemical models to reproduce some processes 36 driven by bacterial activity.


Introduction
Mercury (Hg) is among the most toxic elements, and has many natural sources.Human activity, especially mining and the burning of coal, has increased the mobilization of mercury into the environment.For about 200 years, anthropogenic emissions have been greater than natural emissions (UNEP, 2013).Mercury occurs in various chemical forms.Most atmospheric Hg is gaseous elemental mercury (Hg 0 ).In surface water and soils it occurs as elemental mercury (droplets of liquid mercury) and as Hg(II) complexes (Kim et al., 2003).
Hg-containing minerals, such as cinnabar and metacinnabar (two polymorphs of HgS) and montroydite (HgO), can control its solubility (Kim et al., 2003).Much human exposure to mercury is through the consumption of fish and other marine foods, since mercury is mainly introduced into the food chain as methylmercury (MeHg).In soils, the presence of MeHg results from a balance between different competing processes (Skyllberg, 2012): methylation and demethylation (Cossa et al., 2014) reactions, formation of aqueous complexes and gaseous species and adsorption/desorption reactions onto inorganic and organic substrates.
The balance between those different mechanisms determines bioaccumulation and MeHg transportation.
Researchers have begun to use geochemical modelling to assess the fate of mercury in the environment.Bessinger et al. (2012) have proposed a comprehensive model to reproduce mercury and arsenic speciation in sediment caps and how it changes over time, including MeHg.Leterme et al. (2014) also developed geochemical modelling of a conceptual soil, to assess the relative proportion of mercury release in the atmosphere or transported through the soil column or trapped, either onto mineral surfaces or as minerals precipitated along the profile.Leterme et al. (2014) explained that in spite of being an important tool, geochemical modelling suffers from the lack of well characterized sites.Actually, the counterpart of being able to reproduce detailed reactive mechanisms is that those models may require an important input dataset, including site-specific parameters.Even if determined, the values may suffer from variability and heterogeneities.As an alternative, we propose building a model to reproduce the results of a less complex column experiment.The model itself would include all the complexity governing MeHg fate, including methylation/demethylation reactions, the formation of aqueous complexes and gaseous species and surface complexation reactions.Johannesson and Neumann (2013) conducted a comprehensive set of measurements in groundwater along a 13 km flowpath located within a confined aquifer in southeastern Texas, USA.Their biogeochemical model was able to highlight the main mechanisms (mineral dissolution and sorption onto oxide surfaces) responsible for the speciation of total Hg.
The aim of this study is to test geochemical modelling in a more controlled context like the column experiment performed by Hellal et al. (2015).The geochemical calculations on Hg fate can then be somehow constrained or compared with respect to those experimental results.In addition, taking advantage of the analyses performed by Hellal et al. (2015), the calculations are especially focused on methylmercury fate (methylation/demethylation and transportation), allowing testing of the methylation/demethylation rates available to date in the literature.Before comparing the calculations with the experimental results, we include a critical selection for MeHg complexes, consistent with the database built up by Leterme et al. (2014) for Hg-bearing species.The selection extended to surface complexation reactions and methylation/demethylation rates, following the review and case study proposed by Bessinger et al. (2012) and Cossa et al. (2014).

Selection of thermodynamic properties for methylmercury aqueous complexes
Methylmercury is a strongly toxic complex that accumulates in the muscles and various living tissues of living organisms.After Thomassin and Touze (2003), the methylation of mercury is favored in anoxic environments by the presence of sulfate-reducing bacteria and of sulfur.
Generally speaking, the methylation reaction proceeds this way: Dimethyl products are found as both aqueous complexes (in basic solutions) and in gaseous form.Methylation of mercury is reversible.We selected MeHg species using thermodynamic data processed according to guidelines describes by Blanc et al. (2012) and consistent with the previous selection by Leterme et al. (2014).The thermodynamic parameters associated with complexation reactions have been collected and discussed from Erni (1977), Alderighi et al. (2003), Loux (2007) and Skyllberg (2012), for various chemical systems, at 25°C.Loux (2007) are especially important, resulting from an extrapolation to infinite dilution of large experiment datasets.Alderighi et al. (2003) have measured, by calorimetry, the heat exchanged during various complexation reactions involving methylmercury.From these measurements, we were able to calculate the entropy of complexes, which are reported in Table 1.Actually, they provided Sr for reactions involving CH3Hg + as primary specie.In order to use these measurements, we have considered the reaction: It was converted into an isocoulombic equilibrium using Ca++ and Ca(OH)+ species: The third law entropy was calculated considering the one term approximation method from Gu et al. (1994) and entropies from the Thermoddem database (Blanc et al., 2012).The result allows obtaining the third law entropies from Alderighi et al. (2003) measurements.The whole results are reported in Table 1.In addition to the aqueous complexes we have included a gas phase, Hg(CH3)2,g which represents an extreme stage of the methylation process (Thomassin and Touze, 2003).
The selection was integrated in the Thermoddem database (Blanc et al., 2012) in order to use it with geochemical codes GWB (Bethke, 2004) and PhreeqC (Parkhurst and Appelo, 1999).
A first test of the database consisted in calculating the production of methyl complexes which would arise from the speciation model detailed in Table 1.After Leterme and Jacques (2012), the amount of MeHg in waters in contact with soil systems is usually close to 2%.A speciation calculation is conducted, using PhreeqC-2 with the above database and considering a solution with [NaCl] = with 10 -3 M and [Hg] = 10 -9 M.This amount was reached applying two different corrections: -by modifying reaction (2) equilibrium constant from log10K = 3 to 2.5 when reactions from Table 1 a written with CH4,aq as primary specie -or by modifying CH4,aq = CH3 -+ H + equilibrium constant from log10K = -46 to -52.5 when reactions from Table 1 a written with CH3 -as primary specie.
Actually, we did not expect the relation between organic and inorganic carbon to be driven by thermodynamic equilibrium.The correction considered in the first test is still of small extent, especially because the value originally given by Erni (1977) was calculated and not measured and the uncertainty can be larger in that case.In addition, the fact that the results obtained for MeHg speciation may depend on the way the database is written is to be underline.For the rest of the calculations, we no longer consider the equilibrium between organic and inorganic reduced carbon.However it is important to note that when the database uses CH4,aq or CH3 -complexes as primary specie for the methylation reactions, this allows considering also the methylation for other metals like Sn or Pb (Stumm and Morgan, 1996), jointly with Hg methylation.The database in Table 1 was also tested using activity diagrams of which examples are displayed in Figure 1  The database set up previously is now tested by modeling the MeHg rate obtained experimentally by Hellal et al. (2015) in a companion.To our knowledge, such comparison is rather unique for Hg methylation, up to now.
Only a brief description is given here and the reader referred to Hellal et al. (2015) for additional details.The experiment design is reported in Figure 2. The lower half the column is filled with sterile sand and the upper half with a sterile mixture of sand and iron oxides, initially enriched with Hg(+2).The column is inoculated with a bacterial consortium and the inflowing solution is supplemented with magnesium sulfate and sodium lactate to enhance the growth and activity of sulfate-reducing bacteria (SRB).The water flow is ascendant.Five septa set regularly along the columns enable water sampling from the different layers of the column without perturbing water flow or in situ experimental conditions.After an abiotic rinsing period, the system is inoculated with a bacterial consortium, and physical, chemical and microbial parameters are monitored in time and space, up to 143 days.

Model development
A specific reactive transport model was developed to reproduce the experimental conditions.
Methylation is described in the model following a suite of chemical reactions described in Figure 3, where MeHg formation arises from a rather complex process including different steps: -a first step corresponds to the reduction of Fe(+3) from the dissolution of ferrihydrite.
It is reproduced using the model developed by Poulton (2003) -meanwhile, sulfate undergoes a reduction induced by the activity and growth of SRB bacteria.The reduction is modeled using a first order rate law which rate constant is extracted from Bharati and Kumar (2012) experiment (Table 2) -both Fe(+2) and S(-2) are consumed to precipitate mackinawite (FeS,cr) -cobaltihexamine (CoB-CH3) complexes with mackinawite surfaces, the methyl group CH3 -is released and combines in solution with dissolved mercury Hg ++ to form eventually the methylmercury complex CH3Hg + .Two rate laws were tested, a first order proposed by Bessinger et al. (2012) and a second law based on the work by Heyes et al. (2006) (Table 2) and which rate corresponds to a balance between both methylation and demethylation rates.
The possible formation of complexes at the ferrihydrite surface is implemented, using capacity of ferrihydrite, using the Dzombak and Morel (1990) model and database.Numerical modelling uses PhreeqC-2 code (Parkhurst and Appelo, 1999) and the Thermoddem database (Blanc et al., 2012).The conceptual model is based on 1D-cartesian geometry, as displayed in Figure 4.A tracer test has been successfully simulated to verify the hydrodynamic set of parameters used.

Surface complexation
Ferrihydrite surface

Results and discussion
The results were first verified at the column breakthrough point (Figure 5) and along the column (Figure 6) for the longest reaction time (143 days).The reduction of sulfates was correctly reproduced, globally.In that regard, Bharati and Kumar (2012) experiment was selected because they were performing sulfate reduction by using lactate as substrate and carbon source, as in our case.However, sulfate reduction appears somewhat underestimated by the model, especially concerning sulfide production.The redox conditions in the column (and departure to equilibrium conditions) could be questioned in that regard.
Dissolved Fe(+2) at the outlet was quite correctly predicted and calculations corresponding to the concentrations measured in the different septa even matched the abrupt increase observed in the ferrihydrite-loaded part of the column.Such increase happens during a period when calcite precipitates.It also corresponds to the precipitation of Mackinawite, observed in the experiments by Hellal et al. (2015), using Raman spectroscopy.From the reaction displayed in Table 2, mackinawite precipitation happens to consumes protons which increases pH and can lead induce the precipitation of carbonates.
For dissolved mercury (Figure 5), the total concentrations were correctly predicted.
Methylmercury concentrations analyzed were somewhat bracketed by the calculation performed with the model proposed by Bessinger et al. (2012) and Heyes et al. (2006) (Model 1 and Model 2 in Figure 5, and  from the thermodynamic dataset.In addition, it enhances possible connections between inorganic and organic dissolved carbon forms, from the point of view of methylation processes.Such relations do probably not involve equilibrium reaction.There is a strong need for additional experiment data to test such hypotheses, including measurements that does not seems to be related, at first sight, like dissolved CH4 concentrations, for instance.

Figure 1 -
Figure 1 -Stability relation between MeHg aqueous complexes at 25°C: A -for the carbonate sub-system; B -depending on sulfide and chloride activity; C -depending on sulfide activity and pH experiment(Hellal et al., 2015)

Figure 5 -
Figure 5 -Concentration in dissolved elements analysed at the outlet of the column

Table 1 -
Selection for thermodynamic properties of MeHg-bearing species

Table 2 -
Main parameters for reactive transport modeling