Partitioning of arsenic between hydrothermal fluid and pyrite during experimental siderite replacement

Arsenian pyrite is the major host mineral of gold in Carlin type gold deposit (CTGD) and forms during the interaction of H2S bearing hydrothermal fluids with Fe-bearing carbonates (e.g., siderite, ankerite, Fe-rich dolomite) in the wall rock. Experimental replacement of siderite at hydrothermal conditions (i.e., 200 °C, Psat) was used to study the parameters that control the formation of CTGD pyrite (i.e., pH, time, aH2S) and the partitioning of As between the hydrothermal fluid and newly formed pyrite. Hydrothermal pyrite formed via replacement of siderite in experiments mainly occurs as two types: 1) porous, massive pyrite at alkaline pH conditions or 2) clusters of dense euhedral to anhedral pyrite grains at slightly acidic conditions (pH<6). Longer experimental run times (2weeks) lead to formation of type 2 pyrite at alkaline conditions and an increase in euhedral grain sizes in all conditions. Concentration of As in newly formed pyrite depends strongly on the As concentration of the hydrothermal fluid used in the experiment. Calculated partition coefficients vary between 300 and 1700 with lowest D values at high As concentrations of the fluid. In the studied system As−1 substitutes for S−2 as indicated by the correlation of high As with low S concentrations. Experimental findings agree well with previous studies in natural CTGD on As substitution and the partitioning of As between ore stage pyrite and hydrothermal fluids trapped as

Interaction of this H 2 S rich hydrothermal fluid with reactive iron that originates either from the wall rock (Cail and Cline, 2001;Hofstra et al., 1991;Stenger et al., 1998) or is introduced via an additional fluid (Kesler et al., 2003;Ye et al., 2002) leads to the formation of pyrite rich in Au and As. Reactive iron in the carbonaceous wall rocks is contained in Fe-oxides, Fe-hydroxides, clays and in particular Fe-rich carbonates i.e., ankerite, siderite, and Fe-rich dolomite (Hofstra et al., 1991;Stenger et al., 1998). Decarbonation, or, sulfidation of the wall rock (i.e., desulfidation of fluid) is considered to be one of the major ore forming processes in CTGD and functions via fluid mediated replacement. Although the processes governing CTGDs have been studied in detail in natural systems, experimental work on sulfidation reactions is very limited.
The ore-stage pyrite in CTGD has characteristically high As concentrations (up to 19 wt%), which control the occurrence of Au in the pyrite ore (Deditius et al., 2014;Fleet and Mumin, 1997;Reich et al., 2005). Arsenic in pyrite can occur in different oxidation states which correspond to different crystallographic sites in the lattice and different substitution mechanisms: As −1 substitutes for S under reducing conditions and is most commonly found in CTGDs (Fleet and Mumin, 1997;Reich et al., 2005;Simon et al., 1999a) whereas cationic As (As +2 , As +3 , As +5 ) substitutes for Fe under oxidizing conditions (Deditius et al., 2008;Qian et al., 2013). Incorporation of As into hydrothermal pyrite is generally described as a surface, and kinetically controlled, process (Deditius et al., 2008;Fleet and Mumin, 1997;Palenik et al., 2004). Although As in natural pyrite is intensively studied, and often used as a tracer for hydrothermal processes in natural rocks, studies investigating the partitioning of As between hydrothermal fluids and pyrite are rare (Deditius et al., 2014).
In this study, we investigated the interaction of siderite (FeCO 3 ) with a H 2 S-containing hydrothermal fluid at 200°C to mimic conditions of Carlin-type deposits showing wall rock sulfidation. On the one hand, the study investigated if the replacement of siderite is a critical process during formation of pyrite of Carlin-type deposits. On the other hand, the replacement of siderite was used to study the partitioning of As between the hydrothermal fluid and newly formed pyrite to better understand the interplay between As incorporation into pyrite and sulfidation of wall rock during formation of Carlin-type deposits.

Experimental methods
Hydrothermal experiments were performed using in house manufactured PTFE-beakers (~6 ml) that were closed by conical shaped lids and placed in steel autoclaves. Autoclaves were placed in muffle furnaces at 200°C for varying run durations of 22 to 450 h (Table 1). Weighed amounts of starting material (i.e., siderite or black shale containing siderite, 3-4 pyrite seeds, thioacetamide) were loaded together with~4 ml of experimental fluid into the pre-cleaned PTFE beakers. Thioacetamide (CH 3 CSNH 2 ) was used as the sulfidation agent as it breaks down upon heating to form H 2 S. Although most experiments were done with a H 2 S concentration of 0.05 m which is characteristic for Carlin type deposits (> 0.01; Simon et al., 1999b;Saunders et al., 2014;Hofstra and Cline, 2000), one experimental set (Sd2Py3-11) was done with H 2 S concentrations varying from 0.005-0.5 m to investigate the influence of H 2 S concentration on replacement. Hand-picked grains (n = 20-40, 125-250 μm, varying amounts) of natural siderite (Fe 0.6-0.9 (Mn,Mg) 0.1-0.4 CO 3 ) from two different localities (py1: Bad Schlema, Germany; py0: Schönbrunn, Germany) were used as starting material for each experiment. To promote epitaxial overgrowth, 3-4 grains of natural pyrite (< 0.1 mg, 125-250 μm) were added. Additionally to siderite experiments, two sets of experiments (Sd2Py12-14 and Sd2Py42-44) were conducted using mm-sized ships of natural black shales containing Fe-carbonate from the Selwyn Basin, Yukon, Canada (Magnall et al., 2016) to study siderite replacement in rocks that are similar to the host rocks of CTGD. Siderite replacement experiments were done at fluid buffered conditions and had high water-to-mineral ratios (by weight) of 200-2000 (Table 1). To control and fix the pH during experimental runs, most experiments were buffered in pH by using buffer solutions (NH 4 H 2 PO 4 / (NH 4 ) 2 HPO 4 ; CH 3 COOH/CH 3 COONa, H 3 BO 3 /NaOH) following Qian et al. (2010).
In partitioning experiments, As from an ICPMS standard solution (1000 ppm in 2% HNO 3; Sigma Aldrich) was added in concentrations ranging from 1 ppm to 100 ppm (see Table 1). Arsenic of the standard solution was introduced via dissolution of As 2 O 3 and, therefore, cationic As (i.e., As +3 ) was the dominant species. Additionally to As, we also added trace amounts of Au in a HCl based standard solution (0.05-10 ppm) to experimental runs. Partitioning of Au is not the focus of this paper and is, therefore, not discussed here. Here, all data for the As containing experiments are presented and experiments with the same As concentrations but varying Au concentration in the fluid can be seen as duplicates. Experimental run products (e.g., phase assemblage, pyrite appearance, progress of replacement) and partitioning of As are found to be independent of Au concentration in the experiment. In some experiments, beakers were flushed with Ar before closing to reduce fO 2 of the experiment which is induced by an airgap (Table 1). Oxygen fugacity of the experiments is controlled only by intrinsic phase stabilities and no phases were added to buffer fO 2 .
Experiments were ended by taking the steel autoclaves out of the furnace and letting them cool slowly in air for about 1 h before the beakers were removed from the autoclaves. Due to the relatively slow cooling it cannot be ruled out that additional nm sized mineral phases precipitated and altered the fluid composition after cooling the experiment. Immediately after ending the experiment, beakers were weighed to check for leakage, opened and pH was measured at 25°C. After separating fluid and solid run products, the fluids in the partition experiments were acidified using 50 μl concentrated suprapure HNO 3 and solids washed 3 times with deionized water and dried at 50°C.
In many run solutions, a yellowish gel separated after acidifying, likely due to the reaction of the pH buffer with HNO 3 . It was not possible to redissolve this gel which hindered the measurements of elements therein. Nevertheless, a clear filtrate without the gel component was measured for As, Au, Fe and Mn with ICP-MS using a ThermoFisher ELEMENT 2XR located at the GFZ Potsdam. Results can be found in the Supplementary table, but are not discussed further due to the problems encountered during acidification.

Electron microprobe analysis (EMPA)
Several grains of solid run products as well as 1-2 pyrite seeds were carefully picked and mounted in epoxy and analyzed for major and trace element composition. Major and trace elements (i.e., S (Kα), Fe (Kα), As (Lα)) were analyzed using a JEOL JXA-8230 superprobe located at the GFZ Potsdam. Analytical conditions for quantitative analyses were: 15 kV, 40 nA, spot size (1-5 μm) depending on zonation and porosity, and long measurement times of 100 s for As. Standardization was done using natural and synthetic minerals (Fe, S: natural pyrite; As: gallium arsenide). Detection limits for As in pyrite are around 70 ppm.

Mass spectroscopy (LA-ICP-MS)
Laser ablation ICP-MS was carried out using the Analyte Excite 193 nm ArF* excimer-based laser ablation (LA) system (Teledyne Photon Machines, Bozeman, MT, USA), coupled to the quadrupole-ICP-MS iCAP from Thermo Scientific. The LA-system is equipped with a HelEx II 2-volume ablation cell. Helium was used as a carrier gas for aerosol transport from the sample surface to the ICP and was mixed downstream with Ar as a make-up gas before entering the plasma. Operational parameters of the ICP-MS instrument and LA-unit were tuned for maximum sensitivity, low oxide formation based on the 232 Th 16 O/ 232 Th ratio and low laser-induced elemental fractionation based on the 232 U/ 232 Th ratio using NIST SRM 610. For analysis we measured the following isotopes: 24 Mg, 34 S, 55 Mn, 57 Fe, 75 As and 197 Au. We used 57 Fe as internal standard and the certified reference material MASS1 for calibration for all elements. Samples were ablated with spot sizes between 20 and 40 μm, for 30 s with a repetition rate of 10 Hz and an energy density of 2-3 J/cm 2 . The data was reduced using the commercial software Iolite (Paton et al., 2011) and the data reduction scheme X_trace_elemets_IS (Woodhead et al., 2007). Reproducibility of As concentrations is better than 7% based on multiple measurements of standard material (i.e., NIST610, MASS1).
The LA-ICP-MS spot size was usually bigger than individual grains of newly formed pyrite, and so the data represent an average of different grains and zones. Nevertheless, concentrations quantified by LA-ICP-MS are in agreement with concentrations measured by EMPA.

Microscopy and Raman
Additionally to compositional analysis of iron bearing phases, we used Raman spectroscopy (LABRAM HR Evolution, HORIBA located at the GFZ, Potsdam) on selected samples and petrographic microscopy to identify different types of newly formed pyrite. Obtained Raman spectra were compared to reference spectra of common sulfides provided by the RRUFF data base (see supplements for details). Spectra were taken using a 532 nm laser, 50× magnification and grating of 1800.

Replacement and texture formation
The sulfidation reaction i.e. replacement of Fe-rich carbonates by Fe-sulfides can be illustrated by 2 distinct reactions: dissolution of carbonate to release Fe into fluid or fluid interlayer under neutral to acidic pH conditions of experiments: (1) and the formation of sulfide to precipitate Fe and S into a mineral phase (i.e. pyrite) under reducing conditions (2) resulting in the overall redox reaction Siderite replacement by pyrite can be described as a coupled dissolution-reprecipitation reaction (Putnis, 2002) and is considered to be pseudomorphic at pH 25°C > 5 as the external dimensions of the carbonate grains are preserved. Depending on experimental conditions (i.e., run time, pH, H 2 S concentration, trace element concentrations) different newly formed Fe-bearing minerals (i.e., magnetite, pyrite, pyrrhotite, arsenopyrite) are observed (see Table 1) of which pyrite is the most abundant and most important phase. In general, two different pyrite types are formed; type 1) 'porous pyrite' appears massive under the transmitted light microscope but has a high density of nm sized pores under the electron microscope (Fig. 1a). Microprobe analyses give totals of 85-93% due to the highly porous nature of this material. This phase has a S/Fe ratio of~2 identical to pyrite. Raman spectra of the porous phase show characteristic peaks similar to pyrite that are shifted towards lower wavenumbers, which probably results from the nano/ microporous structure (see supplements Fig. S1). Type 2) 'dense pyrite' either appears as euhedral to sub-euhedral grains of a few μm size that tend to form clusters ( Fig. 1a) or appears as elongated anhedral clusters within pyrite of type 1. Porous and dense pyrite also forms an intermediate phase, producing a mesh of nm to μm sized pyrite grains with open pore space between individual crystallites.
Two experiments (Sd2Py 45, 55) formed pyrite that is rounded and best described as framboidal (Fig. 1c). This pyrite type is found mainly on the outside of pseudomorphically replaced grains.
Apart from pyrite that formed via replacement of siderite, pyrite formed as overgrowth on the pyrite seeds or as replacement rims of these seeds was rarely observed and occur only in experiments with lowest pH (Fig. 1b).

H 2 S concentration dependency
The influence of the H 2 S concentration on the replacement was studied at fluid-mineral ratios of~300. The H 2 S activity is the major control on the stability of different Fe-bearing phases in the system. Low starting H 2 S concentrations (0.005-0.01 m) lead to a preferred replacement of siderite by pyrrhotite and magnetite ( Fig. 2a). At H 2 S concentrations ranging from 0.01 to 0.05 m, a complex replacement texture involving pyrite, pyrrhotite and magnetite is formed. The former siderite grain has a mineralogical zonation inwards, with euhedral pyrite and pyrite clusters (Fig. 2b) on the rim, a zone of pyrrhotite needles, and then a zone of μm sized euhedral magnetite towards the core. In some experiments, siderite still exists as remnants in core regions. Euhedral magnetite is also found within existing siderite along the former cleavage planes system of the carbonate (Fig. 2a,b). With increasing H 2 S (> 0.05) pyrite becomes the dominant phase and magnetite is no longer formed. Pyrrhotite can still be present and is often found in the cores of replaced grains. In some experiments, pyrite and pyrrhotite are intergrown or euhedral pyrite forms as an overgrowth on pyrrhotite (Fig. 2c).

Run time and pH dependency
Time and pH dependency of siderite replacement and the formation of pyrite were studied by varying experimental run durations from 22 to 450 h and pH 25°C from~4.4 to~8.6. Conditions were always strongly fluid-dominated (f/m ratio~1600) and H 2 S concentration was kept constant at 0.05 m. The experimental system was always oversaturated in H 2 S with respect to pyrite formation.
At short runs and high pH 25°C (6.6-8.6) porous pyrite (type 1) dominates the run products, whereas in long runs and at alkaline conditions, euhedral pyrite (type 2) becomes more abundant and is also larger in size (Fig. 3). At low pH 25°C conditions of~4.4, only euhedral pyrite or clusters of euhedral pyrite exist and crystal size increases with duration (Fig. 3). Although all experimental conditions were chosen to stabilize only pyrite, small amounts of pyrrhotite (either intergrown with or as inclusions in euhedral pyrite) were observed independent of run duration.

Experiments using siderite-bearing black shale
The replacement behavior of siderite was not only studied in pure siderite-fluid systems but also in natural sedimentary rocks. We performed experiments using one or two, mm-sized chips of natural black shales containing siderite. These samples were chosen as an analogue of potential CTGD wall rocks that also contain reactive iron in the form of Fe-carbonates (i.e., ankerite and siderite). The grain size of siderite in this sediment is in the range of 50-200 μm and, therefore, similar to that used in siderite only experiments. Replacement textures from these experiments are very similar to those observed from experiments using pure siderite (Fig. 4). Siderite in black shale sediments is mainly replaced by porous pyrite. A small fraction of the porous pyrite recrystallizes into euhedral pyrite crystals. These observations are in agreement with observations from the siderite only experiments, where similar run durations and a relatively alkaline pH 25°C of 8-9 led to the same results. Although part of the experiments were done using a pH buffered solution, the quench pH 25°C was always higher than the desired pH 25°C (see Table 1). The rise in pH is probably due to the usage of a complex multiphase natural rock rather than a single phase leading to a more rock controlled system.

Distribution of arsenic
Partitioning of As between hydrothermal fluids and newly formed pyrite was studied by adding trace concentrations (1-100 ppm) of As to experimental fluids. Partitioning experiments using pure siderite were exclusively performed at high fluid-to-mineral ratios > 1300, slightly acidic conditions and H 2 S concentrations of 0.05 M (Table 1). Arsenic was measured in newly formed sulfides by EMPA and LA-ICP-MS. Each of the two methods has their strength and weaknesses with respect to the purpose of this study and gave complementary results. EMPA measurements facilitated a high spatial resolution (i.e., spotsize < 1-5 μm) and was used to characterize As concentrations of co-existing pyrite types (e.g., framboidal, porous, euhedral) as well as pyrrhotite at high concentration levels (> 70 ppm).
LA-ICP-MS was used to characterize larger areas (using a 20-40 μm spot size) of a representative replacement assemblage at concentration levels down to few ppm. The analyzed area of LA-ICP-MS measurements exceeds the size of individual crystals, hence As values are always an average of individual euhedral pyrite crystals or a mixture of euhedral, porous pyrite, and/or pyrrhotite. Special care was taken during LA-ICP-MS measurements to analyze areas with a high density of euhedral pyrite. Nevertheless, the large heterogeneity within the euhedral type of pyrite as observed by EMPA is able to explain also the variance in LA-ICP-MS measurements.
In general, averaged results for euhedral pyrite of each experiment quantified using both methods agree within individual standard deviation (Fig. 5). Furthermore, compositions of newly formed pyrite from duplicate experiments overlap well within their respective uncertainties. Only As concentrations of experiment Sd2Py35 quantified by LA-ICP-MS deviate from EMPA measurements and also from results of duplicate experiments. Pyrite grains of this particular experiment are smaller than usual and only three LA-ICP-MS spots could be set on the sample. Therefore, we consider the EMPA measurements of experiment Sd2Py35 more trustworthy and will use EMPA data rather than LA-ICP-MS in the discussion.

Arsenic in different sulfides and pyrite types
In individual samples containing more than one type of pyrite, porous pyrite, generally, has lower As values compared to euhedral pyrite (Fig. 6). This behavior can be explained by the non-equilibrium conditions of fast replacement reactions that form porous pyrite, in contrast to euhedral pyrite that grew more slowly under partially equilibrated conditions. Long run durations and low pH of the experiments lead to higher degree of recrystallization of the porous pyrite. Minor framboidal pyrite has a similar composition as euhedral pyrite in one experiment (Sd2Py55) but distinctively lower As values in the experiment (Sd2Py45). As framboidal pyrite in our experiments occurs only on the outside of former siderite grains and not as a direct result of replacement, framboidal pyrite may precipitate as a very late product during quenching of the experiment. In this case, it will form rather rapidly under disequilibrium conditions that might explain the inconsistent incorporation of trace elements. It may also form freely within the bulk fluid in the first place and is later attached to the replaced pyrite. However, framboidal pyrite is only a very minor pyrite type formed under experimental conditions. Pyrrhotite, occurring either as dense single crystals or as aggregated needles, has consistently lower As concentrations compared to coexisting pyrite (Fig. 6). This behavior is also observed in natural pyrrhotite coeval with pyrite (Thomas et al., 2011) and is the result of a lower solubility of As in the pyrrhotite structure compared to pyrite. The low As values of pyrrhotite implies that LA-ICP-MS measurements of areas with larger amounts of pyrrhotite will lead to an underestimation of As in the actual pyrite. Consequently, LA-ICP-MS concentrations of euhedral pyrite should only be seen as minimum values.

Arsenic partitioning between pyrite and hydrothermal fluid
The As concentration of newly formed pyrite depends strongly on the starting composition of the experimental fluid (Fig. 7). Lowest values of 250-1500 ppm As were measured in pyrite from experiments using solutions containing 1 ppm, whereas highest values of 7-8 wt% As were measured in pyrite from experiments using solutions containing 100 ppm As (Fig. 7). Measurements of different pyrite grains from individual experiments by EMPA show a large variability in As (Table 1).
Nernst partition coefficients (D values) were calculated using the following expression: with c As being the concentration by mass of As in pyrite and fluid, respectively. For the calculation of individual partition coefficients, we use the starting composition of experimental fluids and averaged As concentrations of euhedral pyrite measured with LA-ICPMS. Fluid-tomineral ratios of siderite replacement experiments studying As partitioning were in the range of 1300-2200 and, therefore, experiments were strongly fluid buffered. Hence, the fluid composition is stable over the whole duration of the experiment. Calculated partition coefficients of all experiments range between 286 and 1676, showing that As is highly compatible in hydrothermally precipitated pyrite. In general, highest D values of around 1000 are homogenous over fluid compositions ranging from 1 to 20 ppm and agree within individual errors (Fig. 8). The observed decrease in D values down to~300 at high As concentration 50-100 ppm As in fluid is caused by approaching the solubility limit of As in pyrite, which limits the As concentration to 7-8 wt% (see Section 4.2).

Sulfide stabilities during progressing siderite replacement
Experimental observations (i.e., different sulfide phases, sequence of replacement rims) can be explained by two different effects: 1) compositional evolution of the system under equilibrium conditions or 2) metastable formation of pyrrhotite under disequilibrium conditions. In the first case, the different zones (i.e., pyrite, pyrrhotite, magnetite) result directly from thermodynamic stabilities of Fe-bearing phases at low oxygen fugacity and a decreasing H 2 S fugacity (Fig. 9). Due to the relatively low fluid to mineral ratio, S will be consumed from the fluid (according to Eq. (3)) until finally the stability field of magnetite is reached (Fig. 9). As the coupled dissolution-reprecipitation reactions As LA-ICPMS (ppm) As EMPA (ppm)    Qian et al. (2013) and sorption experiments of Bostick and Fendorf (2003) as well as fluid inclusion and pyrite data for natural CTGD systems from China (Su et al., 2009;Su et al., 2008;Su et al., 2012) and Nevada (Large et al., 2016;Reich et al., 2005). penetrate from rim to core, early phases (i.e., pyrite) found in the rim regions form under high H 2 S fugacity, whereas phases in core areas (i.e., magnetite) represent low H 2 S conditions. Once the sulfides are formed under local equilibrium they are kinetically hindered to react further and will not re-equilibrate during experimental run times. Evolution of the H 2 S activity can be either 'global' for the whole fluid present in the beaker, or only local for the interface fluid present at the boundaries of replacing phases. Given the large voids between replacement phases, stability of new phases is likely the result of a compositional evolution of the whole fluid, rather than controlled by local equilibria or diffusion from the bulk to the interfacial fluid. The equilibrium formation and, in particular, the stability of pyrrhotite implies very low oxygen fugacity values (a minimum logfO 2 of −44) according to thermodynamic calculations (triple point at experimental conditions in Fig. 9). Thermodynamic stability fields of mineral phases were calculated as a function of fO 2 and aH 2 S using the PHREEQC software package (Parkhurst and Appelo, 2013) for experimental conditions and starting compositions (Fig. 9). Calculated logfO 2 for the reaction of siderite replacement is~−43 for all compositions used in our experiments and is at odds with the formation of Po in our samples (Fig. 9). The second explanation utilizes the observation that pyrite often forms from a precursor sulfide phase (Benning et al., 2000;Qian et al., 2010;Schoonen and Barnes, 1991a) via various pathways (i.e., polysulfide, Fe-loss, sulfidation). Metastable pyrrhotite was found to act as a precursor phase at high temperature (Qian et al., 2010) as it shows a high nucleation rate at a low oversaturation. In the case of siderite replacement, the localized release of Fe during dissolution of siderite (and additionally the subsequent precipitation of sulfides) will lead to conditions far from equilibrium, in which pyrrhotite might easily form as a metastable phase. With increasing reaction time, the system will equilibrate and pyrite will form as the stable sulfide phase. Nevertheless, transformation of the precursor phase pyrrhotite into pyrite is complex and remnants can form inclusions in newly growing pyrite as characterized by overgrowth and intergrowth textures (Fig. 2c).
As a result of the observations made from replacement experiments with varying H 2 S concentrations, we adjusted our experimental setup for further investigation towards higher fluid-mineral ratios to avoid a compositional evolution during the experiment (Fig. 9).

pH dependent replacement features
In general, formation of sulfides via replacement of siderite is a very fast process and complete replacement is achieved within a few hours at low pH. We interpret the larger crystals and the predominance of euhedral over porous pyrite as being the results of recrystallization and crystal growth that increases with time and decreases with pH (Fig. 3). Under alkaline conditions replacement is slower and remnants of siderite persist in the core region of the former grains. Nevertheless, replacement is effective also at higher pH and the vast majority of the carbonate is replaced within 48 h. Compared to the experimental replacement of magnetite or pyrrhotite done at similar conditions (i.e., pH, temperature, fluid-mineral ratio, grain size and fluid composition) (Qian et al., 2010;Qian et al., 2011), siderite replacement is at least one order of magnitude faster at all pH conditions. Furthermore, magnetite and pyrrhotite replacement rarely reached completion within the experimental time, whereas almost all of our experiments using siderite show a complete replacement.
The observed pH and time dependencies of replacement suggest either generally slower reaction rates at alkaline pH or a change in the rate limiting process as a function of pH. The rate of a coupled dissolution-reprecipitation reaction is controlled by the slowest rate of all processes involved (i.e., dissolution, transport, precipitation). At low pH 25°C we observe the formation of a blackish 'film' (likely composed of nm-sized Fe-sulfides) on the autoclave walls. The precipitation of this 'film' on autoclave walls is interpreted to represent a long-range solution transport of Fe, which is not instantaneously incorporated into replacement sulfides. Hence, it indicates a faster dissolution rate of Fecarbonate compared to the precipitation rate of pyrite (Qian et al., 2010). Contrastingly, no precipitation on the autoclave walls at pH 25°C > 5 indicate an immediate incorporation of dissolved Fe into newly formed pyrite. In this case, dissolution becomes the rate limiting step and replacement shows short length scale pseudomorphism (Xia et al., 2009). Unlike in the case of U scavenging during replacement of magnetite by sulfides (Li et al., 2015), we do not observe a dependency of replacement features (e.g., zonation, rates) on trace element concentrations.

Similarities to natural Carlin type deposits
The main difference between the sediment experiments and pure siderite experiments is in the degree or progress of replacement. In the pure siderite system replacement was always complete at high H 2 S concentrations, in the natural rock samples siderite is either still present in cores of the reacted sediment chips or in the process of being replaced (Fig. 4b). Additionally, newly formed pyrite is mainly found in the outer part of rock chips (Fig. 4a). These findings suggest that overall replacement rates are slower when sediment is used in the experiments. In this case, the transport of S from the bulk fluid towards the reaction interface (i.e., siderite domain within the black shale) is likely the rate limiting step as S needs to be transported via diffusion in pores or along grain boundaries. Nevertheless, once replacement of the carbonate starts it creates porosity due to the lower molar volume of pyrite compared to siderite which then enhances the transport capabilities. This effect will also apply for formation of ore stage pyrite in CTGD, where sedimentary layers rich in Fe-carbonates will be preferred sites for replacement.
Experimentally derived replacement textures share similarities to pyrite textures of natural Carlin type deposits. Apart from overgrowth on preexisting pyrite, natural CTGDs exhibit 'spongy', 'massive-porous' and disseminated pyrite of hydrothermal origin forming the main stage ore (Kesler et al., 2003;Large et al., 2009;Simon et al., 1999b;Su et al., 2008). In general these textures are interpreted to be the result of replacement. As shown by our experiments, replacement of siderite in sedimentary host rocks can lead to highly porous pyrite but also to aggregations of μm-sized euhedral to anhedral pyrite due to recrystallization; these are very similar to common features of pyrite of CTGD (Saunders et al., 2014). Although some features of CTGD were reproduced experimentally, prominent overgrowths of arsenian pyrite were almost never observed, neither in pure siderite experiments where 1-3 grains of seed-pyrite were added, nor in black-shale experiments that contained preexisting pyrite. A μm sized overgrowth of irregularly shaped arsenian pyrite was only formed in low pH (~4.4) experiments (Fig. 1b). These are the same experiments that showed a precipitation of sulfides on the beaker walls (see Section 4.1.2) and were interpreted to represent conditions under which dissolution of siderite is faster than precipitation of pyrite. Consequently, slightly acidic conditions are needed to produce an overgrowth on existing seed pyrite. In all other experiments at higher pH values, the precipitation of pyrite is more effective in suppressing Fe mobility and transport and, therefore, suppressing overgrowth textures. In both experimental setups Fe-carbonate is the only source of reactive iron, whereas in natural CTDG systems, other phases or fluids contribute to the overall amount of reactive iron, which might then be used to produce overgrowths on existing pre-ore pyrite. Additionally, replacement of pre-existing pyrite by ore stage pyrite via coupled dissolution-reprecipitation can lead to textures that look similar to overgrowth textures (Fleet and Mumin, 1997).

Arsenic substitution and arsenic solubility
Micro analytical measurements show that As concentrations in newly formed euhedral pyrite strongly correlate with decreasing S concentrations (Fig. 10a). We, therefore, conclude that As mainly replaces S in the pyrite structure and occurs therefore as As −1 , which is also the common species in Carlin type pyrite (Fleet and Mumin, 1997;Reich et al., 2005;Simon et al., 1999a;Simon et al., 1999b). Although, the correlation is distinctive it deviates from the 1:1 line. Consequently, around 25% of the As follows a different substitution mechanism. Although, correlation of As with Fe is less clear (Fig. 10b), pyrite having highest As of 4 mol% has reduced Fe (32 mol%) compared to stoichiometric pyrite with 33.3 mol% Fe. Therefore, other more oxidized species like As +2 or As +3 must substitute for Fe. This observation agrees with findings in natural and experimental arsenian pyrite (Deditius et al., 2014;Deditius et al., 2008;Le Pape et al., 2017;Qian et al., 2013).
In comparison to other experimental studies, arsenian pyrite formed via replacement of siderite resembles results of arsenian pyrite formation via replacement of non-arsenian pyrite and pyrrhotite (Fleet and Mumin, 1997), rather than results of Qian et al. (2013), in which pyrite was formed via magnetite replacement. Maximum As concentration of our euhedral pyrite is similar to the experimental observations of Fleet and Mumin (1997) and does not exceed 7-8 wt%. This is in agreement with thermodynamic calculations that find a miscibility gap between pyrite and arsenopyrite in the system FeS 2 -FeAs 2 showing a solubility limit of As in pyrite at around 7 wt% at 500 K (Reich and Becker, 2006). End member arsenopyrite was rarely identified to coexist with arsenian pyrite in experiments with the highest As concentrations (50-100 ppm). Nevertheless, a few EMPA analyses show As values higher than 8 wt%; this potentially indicates an intergrowth of pyrite and arsenopyrite at the nanoscale (Reich and Becker, 2006;Simon et al., 1999aSimon et al., , 1999b. When As substitutes for Fe as As +2 its concentrations can be as high as 24 wt% due to a different substitution mechanism (Qian et al., 2013). In nature, the occurrence of different As species in hydrothermal pyrite can be attributed to a different oxygen fugacity of the individual system. Under oxidizing conditions, typical for high-sulfidation systems, cationic As (As +2,+3 ) is the dominant species in arsenian pyrite (Deditius et al., 2014;Deditius et al., 2008;Qian et al., 2013), whereas substitution of anionic As −1 is preferred in more reducing environments like that of Carlin type deposit. The similar behavior of As incorporation into pyrite during experimental siderite replacement and in pyrite from natural CTDGs confirm the general idea that ore stage arsenian pyrite of CTGSs is derived from replacement of Fe-carbonates present in the sedimentary wall rocks.

As partition coefficients in nature and experiments
Our experimental partition coefficients largely agree with observations and calculated D values of other natural and experimental fluidpyrite systems that have similar conditions (i.e., T, major and trace element composition, pH, fluid to rock/mineral ratio, As speciation). In general, experimental and field studies reporting As trace element compositions of pyrite and coexisting/related fluid are rare and none of them report D values. Qian et al. (2013) published experimental data that can be used to calculate partition coefficients. As the authors did not publish fluid compositions but stated weights of all As containing solids they used, we estimated As concentration in their experimental runs from these data assuming reported complete dissolution of the As bearing phases. Concentration of As in their experimental fluids range from 24 to 220 ppm corresponding to maximum As of coexisting pyrite of 0.26 wt% up to 24 wt%, respectively. For the calculation of the D value we used reported maximum As concentrations. Arsenian pyrite in their experiments formed via a coupled dissolution-reprecipitation process of magnetite and highest As concentrations were always measured in the outermost rim, that was in contact with the hydrothermal fluids. Furthermore, this As-rich rim also shows features of recrystallization and was therefore in local equilibrium with the Asbearing fluid. Calculated D values range between 100 and 1600 for their experiments. Although their experimental conditions are different from ours, their calculated D values are very similar to our D values. Using data from the most similar experiment to ours (i.e., A11) from Qian et al. (2013) the calculated D value of 416 fits our results (Fig. 8). This is interesting as the substitution process of As into newly formed pyrite in the case of magnetite replacement (As for Fe) is different than in the case of siderite replacement (As for S).
In natural systems, partition coefficients can be inferred by using compositions of fluid inclusions that represent the hydrothermal fluid that coexisted with pyrite. In CTGD, the lack of large fluid inclusions, the high detection limits and low concentrations of trace elements in individual fluid inclusions means that fluid trace element data is rare and only a few studies exist that published As values for fluids and pyrite. In a series of papers Su et al. (2008Su et al. ( , 2009Su et al. ( , 2012 reported trace element data for gold-bearing arsenian pyrite as well as data for coexisting fluid inclusions from Shuiyindong and Yata CTG deposits (Guizhou, China). Ore stage pyrite contains between 3.4 and 14 wt% As that substitutes for S in the pyrite structure. Related ore stage fluid inclusions were analyzed by LA-ICP-MS and range from 80 to 250 ppm. Assuming that As in pyrite correlates positively with As in fluid, calculated partition coefficients are between 437 and 560. Similar values can be calculated for the Golden Quarry CTGD from Nevada using recently published fluid inclusion data by Large et al. (2016) combined with general data for pyrite from the Carlin trend deposits (Deditius et al., 2014;Arehart et al., 1993). Measured fluid inclusion data for As range from 24 to 540 ppm and span a slightly larger range than CTDGs from China. It should be noted, however that As concentrations in many of the analyzed fluid inclusions are below the detection limit and, therefore, true As concentrations of the hydrothermal fluid are likely lower than the averaged values. Concentration of As in coeval pyrite is stated as being~2-6 wt% (Large et al., 2016), which agrees with more precise data for ore stage pyrite (3.7-5.4 wt% As) from the same deposit (Arehart et al., 1993). In general, ore stage arsenian pyrite from CTGDs from Nevada has maximum As concentration of 19 wt%, that can decrease to~1 wt% in some deposits (Reich et al., 2005). Given these large variabilities in fluid and pyrite composition, D values for highest and lowest reported concentrations of ore stage arsenian pyrite are calculated to be in the range of 350 to 420. The inferred partition coefficients of our experimental study as well as of natural CTGD systems are in strong disagreement with the only other publication stating D values. Deditius et al. (2014) suggest very low D values for As in the order of 0.02 to 0.2 based on fluid inclusion and pyrite data from the Rosia-Poieni porphyry/epithermal Cu-Au deposit (Kouzmanov et al., 2010). Although element partitioning should be stable over a large range of conditions and composition, several differences between CTGD and epithermal deposits might lead to the discrepancy of 4 orders of magnitude in calculated D values: 1) the higher temperature of the epithermal systems (250-350°C) compared to CTGD systems (< 220°C) might lead to changes in the partitioning behavior as pyrite has retrograde solid solubility of As with temperature (Deditius et al., 2014). This behavior can be explained by kinetic factors which in turn might influence partitioning as they affect recrystallization of sulfides before reaching equilibrium. In particular, far from equilibrium conditions at low temperatures might lead to varying D values. Temperature dependent kinetics also influence fundamental processes involved during incorporation of As into pyrite like the rates of adsorption and growth. 2) An additional change in partition behavior might be caused by strongly differing pH values of epithermal (extremely acidic; Kouzmanov et al., 2010) and CTG (slightly acidic to neutral; Saunders et al., 2014) deposits. 3) Rosia-Poieni epithermal veins contain, apart from pyrite, large amounts of enargite (Cu 3 AsS 4 ) which acts as a potential sink for As and conflicts with equilibrium partitioning between fluid and pyrite. 4) A potential different speciation of As in the fluid can cause different partitioning. Qian et al. (2013) showed in their experiments that As dissolved from solids containing anionic As (i.e., loellingite, arsenopyrite) is incorporated much less into pyrite compared to As from sources containing cationic As (e.g., realgar, As 2 O 3 ). Although, this argument seems counterintuitive as conditions of epithermal deposits are more oxidizing and should favor cationic As +3,+5 , findings of Qian et al. (2013) are independent of oxygen fugacity and seem only to rely on the speciation of As in source phases. The chemical state of arsenic in enargite, which is the major As bearing phase in the epithermal sample, is very similar to arsenides and, therefore, similar to loellingite and arsenopyrite (Fantauzzi et al., 2006). Consequently, As dissolved from enargite is not incorporated into pyrite and calculated D values from the Rosia-Poieni epithermal system will be unusually low.

Implication of high partitioning for pyrite systems
Although the preferential incorporation of As into pyrite has been suggested for different types of ore deposits, this study quantifies partition coefficients for sulfidation systems at moderate temperatures like CTGD, sediment hosted massive sulfide and epithermal deposits. The derived D values can not only be used to estimate the composition of ore fluids from the As concentration of pyrite, but can help to explain the large variation of As in pyrite observed in these deposits. In general, μm scale variations of As (and other trace elements) in pyrite producing a growth zonation are explained by changes in fluid composition due to various processes like mixing, boiling or exsolution (Deditius et al., 2009;Feng et al., 2017;Peterson and Mavrogenes, 2014;Tardani et al., 2017;Yan et al., 2018). Taking the high D values into account, parts of the zonation showing decreasing As concentrations can be attributed to compositional evolution of the fluid caused by the effective incorporation of As into the pyrite overgrowth. Repeated replenishment with an enriched fluid will lead to oscillating zonations which are frequently observed in pyrite. On the scale of individual CTGDs, the overall variation of As found in different types of ore stage pyrite (i.e., porous, overgrowth, disseminated) ranging from~1 up to 19 wt% might also be caused by the evolution of As in the hydrothermal fluid as a result of equilibrium fractionation of a single fluid event.

Control of incorporation of As into hydrothermal pyrite
Partitioning experiments were designed to mimic conditions and compositions of natural CTGDs and the calculated partition coefficients are purely empirical. They allow only limited insights into the exact processes that controlled the partitioning of As into pyrite during growth. In a simplistic way, one can think of two individual factors and related processes that govern the incorporation at low to hydrothermal temperatures: 1) the solid solubility of As in pyrite (maximum around 7-8 wt% at conditions of interest) which is controlled by the crystallographic structure; 2) surface controlled sorption of As onto the growing pyrite. The first factor is a function of intrinsic variables and ultimately controlled by Gibbs free energy minimization of the solid solution and is, therefore, the major control for partitioning at high temperatures (e.g., melt systems, sub solidus diffusional exchange) as described for example by the lattice strain model (Blundy and Wood, 1994). In low temperature, fluid mediated environments the adsorption capability onto mineral surfaces becomes more important for partitioning as it controls the availability of the substituent during crystal growth. Given that the intrinsic factors controlled by the crystal lattice of pyrite allow for perfect solid solution (< 7 wt% As), sorption of As onto the reactive pyrite surface controls partitioning in our experiments as well as in natural systems. Often the formation of arsenian pyrite (and coeval Au enrichment) in low T settings is referred to as a surface and kinetically controlled process (Deditius et al., 2014;Fleet and Mumin, 1997;Palenik et al., 2004;Qian et al., 2013) that enables the reduction of the dissolved species As 3+ (either as As 3+ (OH) 3 or as thiocomplexes in presence of H 2 S and low pH (Keller et al., 2014)) to As −1 in pyrite. Sorption experiments at ambient temperatures show a high adsorption capacity of As onto sulfide surfaces that can be described by common sorption isotherms (i.e., Freundlich and/or Langmuir isotherms). Isotherms, in general, relate the amount of an adsorbed element onto a mineral to concentration of sorbent in fluids (Bostick and Fendorf, 2003;Farquhar et al., 2002;Han et al., 2013;Wolthers et al., 2005) and simple Nernst partition coefficients can be seen in this context to be one form of a linear isotherm expression. For As, the observed maximum adsorption on pyrite at ambient temperatures is around 2310-270 μmol/g relating to a surface coverage of 5.6 μmol/m 2 (Bostick and Fendorf, 2003;Han et al., 2013) which correspond to a concentration of~2 wt% As in pyrite assuming that all adsorbed As is incorporated during growth. Interestingly, the empirically gained adsorption maxima and the thermodynamically calculated solubility limit (~4 wt% As at ambient conditions (Reich and Becker, 2006)) are very similar and differ only by a factor of 2. The shape and maxima of the isotherms, and therefore the relationship of As adsorption to As concentration in fluid, strongly depends on pH but also on H 2 S and likely temperature (Bostick and Fendorf, 2003). In the absence of high temperature experiments, sorption behavior of As at hydrothermal conditions is largely unknown. Although, it is therefore impossible to extract exact partitioning values from low temperature sorption experiments, the linear correlation in As adsorption observed by Bostick and Fendorf (2003) for H 2 S-rich systems can be used to calculate D values of around 250 which agrees with D values of our experiments.
The similarities between adsorption onto, and partitioning into, pyrite indicate a potential coupling of the two processes working during trace element distribution in hydrothermal systems like Carlin. Consequently, the high variance in composition of pyrite from these systems is a direct result from changes in the local environment (compositionally and kinetically) that in turn changes the adsorption behavior onto a growing surface.

Conclusion
Fluid mediated replacement of siderite by pyrite is a fast and effective process to transform Fe-bearing carbonates into sulfides that works via a coupled dissolution-reprecipitation reaction. New pyrite forms either as cluster of dense euhedral grains under acidic pH conditions or as a porous mass under alkaline pH that recrystallizes with time to form euhedral to anhedral pyrite grains or clusters.
Experimental partitioning of As between hydrothermal fluid and newly formed euhedral pyrite have partition coefficients in the range of 286 to 1676 that are similar to D values calculated from literature data for natural and experimental pyrite systems. Arsenic in pyrite formed via siderite replacement is mainly substituting for S and, therefore, bound as anionic As −1 in the pyrite structure.
Findings of this experimental study are similar to observations made in natural Carlin-type gold deposits and confirm the general idea that arsenian pyrite of these deposits forms via replacement of Fe-carbonates like siderite. The newly constrained partition coefficients for As are able to explain the large heterogeneity in CTGD pyrite compositions which is caused by a compositional evolution of the fluid due to the strong fractionation of As into the forming pyrite and, therefore, depletion in the fluid.