Quantifying Fenton reaction pathways driven by self-generated H2O2 on pyrite surfaces

Oxidation of pyrite (FeS2) plays a significant role in the redox cycling of iron and sulfur on Earth and is the primary cause of acid mine drainage (AMD). It has been established that this process involves multi-step electron-transfer reactions between surface defects and adsorbed O2 and H2O, releasing sulfoxy species (e.g., S2O32−, SO42−) and ferrous iron (Fe2+) to the solution and also producing intermediate by-products, such as hydrogen peroxide (H2O2) and other reactive oxygen species (ROS), however, our understanding of the kinetics of these transient species is still limited. We investigated the kinetics of H2O2 formation in aqueous suspensions of FeS2 microparticles by monitoring, in real time, the H2O2 and dissolved O2 concentration under oxic and anoxic conditions using amperometric microsensors. Additional spectroscopic and structural analyses were done to track the dependencies between the process of FeS2 dissolution and the degradation of H2O2 through the Fenton reaction. Based on our experimental results, we built a kinetic model which explains the observed trend of H2O2, showing that FeS2 dissolution can act as a natural Fenton reagent, influencing the oxidation of third-party species during the long term evolution of geochemical systems, even in oxygen-limited environments.

The rupture of S-Fe and S-S bonds over pyrite surface by mechanical fracture or during the dissolution process induces the formation of the non-stoichiometric defect sites (i.e., the dangling bonds) that trigger adsorption reactions [31][32][33] . The iron surface species formed cause a reduction in the band gap of the pyrite from 0.86 electron volts (eV) in the bulk to 0.55 eV on the surface 34 , catalyzing the dissociation of adsorbed oxygen (O 2 ) and water (H 2 O) molecules and leading to the formation of H 2 O 2 (equations 1-4) 19,25,26,35 :  2 3 Although this sequence of reactions describes the basic features of H 2 O 2 formation during pyrite dissolution (equations 1-4) and its further degradation to secondary ROS in solution (equations [5][6][7][8], various important aspects remain unclear, for instance: (1) can free radicals be formed by mechanism other than the photoactivation (i.e. by mechanical bond fracture or non-stoichiometry dissolution of pyrite)? (2) can the formation of H 2 O 2 occur in strictly anoxic conditions overcoming the energy required to split the water molecule and the further release of O 2 ? and (3) can the process of ROS generation by pyrite be sustained over long periods of time?.
Real-time measurements of this process are made difficult by ROS reactivity and the subsequent redox transformations of iron and sulfur species. Spectroscopic and fluorescence methods are commonly used to measure the concentration of H 2 O 2 ; however, these methods usually need the presence of dyes or chelating agents that are not well suited for the kinetic analysis of transient phases. In this study, we measured the real-time generation and decomposition of H 2 O 2 and dissolved O 2 induced by pyrite surfaces under different boundary conditions (i.e., dark/light, oxic/anoxic) to investigate the kinetic role of ROS during pyrite dissolution. In addition, we analyzed both the chemical evolution of dissolved species and surface pyrite oxidation with spectroscopy (UV-Vis, XPS), Cyclic Voltammetry (CV) and high-resolution transmission electron microscopy (HTREM) and performed specific experiments to evaluate the persistence of H 2 O 2 formation. Based on our results, we developed a kinetic model for the coupling between pyrite dissolution and H 2 O 2 generation/degradation through the Fenton reaction. When combined with the observed trends this model leads to the definition of constraints on the overall process of ROS oxidation mechanism induced by pyrite surfaces.

Results
H 2 O 2 and O 2 evolution: general pathway. Figure 1 shows the concentration of H 2 O 2 as a function of time in aqueous suspensions of pyrite starting at circumneutral pH, under oxic and anoxic conditions. H 2 O 2 concentration increased at the beginning of the experiment until a maximum value was reached, and decreased thereafter asymptotically towards a nearly stationary, residual value ([H 2 O 2 ]> 200 nM) still measurable at the end of the experiment (~22 h). This coupled generation-decay response was generally observed in every experiment, although there were variations in the particular shape of the curves and, some experiments showed characteristic shoulders or secondary maxima at intermediate stages of the process. The overall process followed a sigmoidal trend, suggesting a strong interaction between the H 2 O 2 generation and degradation rates characteristic of an autocatalytic process. H 2 O 2 formation and decomposition evolved more slowly under anoxic than under oxic conditions. We did not observe any correlation between pyrite loading and the H 2 O 2 yield between experimental runs under our experimental conditions (i.e., 0.5-0.3 g/L, unbuffered neutral pH) because of several factors (e.g., kinks, steps, lattice anisotropy) can determine the variability of the reactive surface between samples [36][37][38][39] .
The evolution of O 2 under oxic conditions was characterized by an asymptotic decrease followed by a slight increase and a steady stable period at the end of the experiment (Supplementary Fig. S1). The opposite trend was observed under anoxic conditions (i.e., an initial increase in O 2 followed by an asymptotic decrease). To clarify the role of O 2 over the formation of H 2 O 2 , we monitored simultaneously the evolution of O 2 and H 2 O 2 under oxic and anoxic conditions (Fig. 2). In oxic conditions (Fig. 2a), the amount of H 2 O 2 increased whereas O 2 was rapidly consumed at the beginning of the experiment. Under anoxic conditions O 2 and H 2 O 2 formed concomitantly (Fig. 2b).
During the experiments, pH became acidic under both oxic and anoxic conditions. Overall, pH dropped rapidly towards a nearly constant value approximately 2 to 3 pH units lower then at the initial pH value. The initial drop was more pronounced under anoxic than under oxic conditions (2 hours vs 10 hours, respectively), and in both cases the initial decline in pH values was accelerated with increasing pyrite loading ( Supplementary Fig. S2).

The Fenton reaction in anoxic conditions: spectroscopic experiments.
To determine if the Fenton reaction was actually the mechanism to degrade the H 2 O 2 forming under anoxic conditions, we monitored the production of Fe 3+ and OH · species in pyrite suspensions under anoxic conditions and in the dark (Fig. 3). First, we evaluated the intensity of the absorbance bands in the UV-Vis range of 260-700 nm to identify the formation of dissolved Fe 3+ -complexes that coexist in solution in the range 300-450 nm [40][41][42][43] (Fig. 3a). The first absorbance peak (λ max ~375 nm) evolved after three hours of reaction and was shifted to higher wavelengths (around 390-400 nm), as the reaction proceeded. Second, we monitored the formation of short-lived radicals, mainly OH · by measuring the decrease in the light absorption spectrum (λ max = 590 nm) of Crystal Violet used as a dye probe 44 (Fig. 3b). The absorption spectrum of CV showed a rapid decline indicating that OH · formation occurs concomitantly from the start-up of the experiment.
Tracking the reversibility of H 2 O 2 generation: cycling experiments. We performed these experiments to better understand the long-term evolution of H 2 O 2 during pyrite dissolution in an open system. In order to use the same physical pyrite particles and in the same geometry, pyrite microparticles were adhered over silicone strips to form a thin film which was placed on the internal wall of the reactor as in the spectroscopic experiments. We monitored the evolution of H 2 O 2 until the observable amount of H 2 O 2 attained a constant value or a concentration of zero ( Fig. 4). At that point, we replaced the solution inside the batch reactor with distilled H 2 O, and monitored the evolution of H 2 O 2 again (we repeated this procedure twice). After a full cycle of H 2 O 2 generation and decay, the formation of H 2 O 2 was found to resume after the addition of fresh distilled H 2 O, although the maximum H 2 O 2 yield was consistently lower than in the previous cycle (Fig. 4a). Assuming a zero-order kinetic model, the initial observed rate of the H 2 O 2 formation (k 0 obs ) showed a slight decrease between the first and the second cycle, which was more prevalent in the third cycle (Fig. 4b). This decrease could be attributed to the formation of oxide patches on pyrite surface that partially block some of the iron reactive centers.

Surface characterization (CV, XPS and HRTEM). Results of cyclic voltammetry under anoxic condi-
tions revealed the anodic peaks associated with H 2 O oxidation by one electron and by two electron transfer ( Supplementary Fig. S3). Interestingly, in the cathodic counterpart, the peak assigned to iron reduction is split into two peaks (0.1 V NHE and 0.2VNHE, respectively), indicating that a fraction of ≡ Fe 3+ is reduced in a nearly spontaneous manner. The analyzed XPS spectra of the (001) face of pyrite after aqueous reaction in the presence and absence of dissolved O 2 (g) resulted in significant differences, showing major surface oxidation under oxic conditions with the subsequent formation of patches ( Supplementary Fig. S4). In fact, XPS analysis of the (100) face of the pyrite after anoxic reaction only showed appreciable changes in the O1s orbital showing a shift to In order to facilitate the identification of ≡ S 2− and ≡ Fe 3+ dangling bonds, one sample was ion-sputtered, which promoted the breakage of the S-S dimers (as in the grinding procedure) ( Supplementary Fig. S6). The formation of iron oxidation patches was also observed with HRTEM. Figure 5 shows an image of the pyrite surface after 22 hours of reaction in a micromolar solution of H 2 O 2 in absence of O 2 (g). The presence of discrete oxidation patches was observed in the uppermost area of the micrograph (Fig. 5a). The FFT (Fast Fourier transform) proved that the lattice fringe spacing of low contrast clusters (~0.25 nm) were consistent with ferrihydrite nanocrystals (Fig. 5b), viewed down [001]. Additional HRTEM images ( Supplementary Fig. S7) also showed the interplanar spacing characteristic of two-line ferrihydrite 45,46 . The FFT of the crystalline part showed the interplanar spacing of pyrite but also of goethite, suggesting that ferrihydrite can be a precursor of goethite formation (Fig. 5c) 47 .   oxic-open conditions (pyrite load particle = 0.33 g/L, Δ pH 1 cycle = 6.8-3.8, Δ pH 2 cycle = 6.8-6.1, Δ pH 3 cycle = 6.8-6.7); (b) Initial observed rate of H 2 O 2 formation assuming a zero kinetic order rate. Where k 1 , k 2 and k 3 , represent the rate constants of each Fenton reaction step (Supplementary, Table S1). As shown in the equation, the H 2 O 2 formation was calculated assuming a first order dependence on the reactive surface and, both iron defect sites act simultaneously, but in anoxic conditions the H 2 O 2 produced by Fe 2+ sites is limited by the O 2 derived by the Fenton reaction. Minimization between experimental curves and the values calculated from equation 9 were made with a non-linear least squares approach using the Marquardt algorithm, with k oxic and k anoxic , the specific rate constants of H 2 O 2 surface generation, as adjustable parameters (Table 1). Figure 6 presents the fitted values corresponding to the evolution of H 2 O 2 in oxic and anoxic conditions together with the model derived results. Under oxic conditions (Fig. 6a), when pH was higher, the ferrous surface iron was oxidized by dissolved O 2 forming H 2 O 2 according the first order rate. Deviations of the linearity started to occur due to a change in the reaction stoichiometry by the simultaneous decrease in pH and increase in dissolved Fe 2+ and H 2 O 2 . From this point, the Fenton reaction became important and the oxidation rate of Fe 2+ by O 2 slowed down leading to a minor O 2 consumption. As a result, reactions forming ROS -catalyzed by Fe 2+ and Fe 3+ − became effective for H 2 O 2 degradation, which rapidly decreased, while iron species followed an opposite trend. Under anoxic conditions (Fig. 6b), the formation of H 2 O 2 proceeded more slowly. The decomposition of H 2 O 2 was also retarded because the concentration of dissolved [Fe 2+ ] supplied to solution by pyrite dissolution was lower. The Fenton reaction was initiated when pH values dropped below 4.5 and the ratio H 2 O 2 / Fe 2+ increased, catalyzing the H 2 O 2 decomposition similar to the oxic experiments. The analysis of ROS derived from the model in both oxic and anoxic conditions (Fig. 6c and d) shows that, the first reactive species formed was OH · acting as a chain initiator, forming additional free radicals. The majority of the OH · reacted with H 2 O 2 , generating HO 2 · , the conjugate acid of O 2 ·− , which is the limiting reagent to assist the redox cycling of Fe 3+ / Fe 2+ forming O 2 and helping to buffer the pH drop Contrary to the expectation, the maximum amount of H 2 O 2 measured was significantly different between experimental runs, and independent of particle loading. This was likely because the effect of increasing the reactive surface was twofold: (i) an increase in the formation of H 2 O 2 , and (ii) an increase in the release of dissolved Fe 2+ , accelerating the decomposition of H 2 O 2 by the Fenton reaction (Fig. 7). We plotted H 2 O 2 and Fe 2+ evolution  as predicted by the model for a set of different values of pyrite reactive surface area (in Fig. 7a). As a result of this coupling effect, higher values of reactive surface area tend to increase the amount of secondary ROS (i.e., OH · /O − , HO 2 · /O 2 ·− , O 2 ) rather than stabilize the presence of H 2 O 2 in solution (in Fig. 7b) 2,48 .

Discussion
Aqueous suspensions of pyrite form H 2 O 2 , in the presence and absence of dissolved O 2 , following a generation-decay trend (Fig. 1)   have suggested that in the presence of dissolved O 2 , pyrite slurries form H 2 O 2 by an electron transfer between the ferrous iron defect sites (Fe 2+ − S) and the adsorbed O 2 through a Habber-Weiss reaction, involving the O 2 ·− radical formation (equations 1 and 2). This hypothesis is consistent with the observation that there is an inverse relationship between H 2 O 2 and O 2 at the beginning of the oxic experiment (Fig. 2a). The mechanism of H 2 O 2 formation in anoxic conditions remains more controversial. Some studies have suggested that in absence of O 2 the formation of H 2 O 2 is driven by the oxidation of adsorbed H 2 O catalyzed by the pyrite surface 4,19,24,26 , whereas other studies have considered that this reaction is unlikely due to energetic considerations 15,25 . A possible reconciliation comes from considering the presence of ≡ Fe 3+ dangling bonds generated from the cleavage of S-S bonds. Briefly, the rupture of S-S bonds generate S − species which are highly instable and, to compensate its charge disequilibrium, donate one electron to the nearest iron atom by the autoredox reaction 26,31,32,[49][50][51][52][53][54] : These dangling bonds could decrease the energy requirements for the chemisorption of H 2 O molecules, some authors even talk about "ferryl" iron dangling bonds ≡ Fe 4+ 55 , leading to H 2 O splitting into H + and OH · , with the subsequent formation of H 2 O 2 , as described by 19 equations 3 and 4. Our experiments showed that H 2 O 2 was generated in the absence of both dissolved O 2 and light (Fig. 1b). Besides, the estimated ratio of S/Fe (< 2) in unreacted pyrite particles (Supplementary Fig. S8) indicated the presence of S vacancies.
Experiments with pyrite slurries under both oxic and anoxic conditions showed a sudden drop of pH during the first hours ( Supplementary Fig. S2). This increase of H + concentration together with the progressive accumulation of H 2 O 2 and Fe 2+ in solution is expected to trigger the Fenton reaction. UV-Vis spectroscopy confirmed the presence of Fe 3+ -complexes during anoxic pyrite dissolution (Fig. 3a) whose maximum wavelengths were compatible with the following reaction 41 : The detection of both Fe 3+ -complexes and OH · in solution (Fig. 3) indicated that the Fenton reaction occurred even in the absence of dissolved O 2 , supporting the idea that the suite of Fenton reactions conditioned the decay period of H 2 O 2 curves. Additionally, the formation of OH · concomitantly with the formation of H 2 O 2 from the start-up of the experiment under anoxic conditions (Fig. 3b) together with the splitting of the cathodic peak associated with the nearly spontaneous iron reduction ( Supplementary Fig. S3), suggest that ≡ Fe 3+ dangling bonds, actually catalyze the H 2 O oxidation by one single electron, forming OH · as described by equation 3. In principle, the overall process of aqueous pyrite oxidation -under oxic conditions-involves only O 2 (g) consumption, according to: However, considering a free radical mechanism, a simultaneous uptake and release of dissolved O 2 it to be expected, because it can be both a product and a reactant. In oxic experiments, O 2 concentration rapidly dropped and then reached a more constant value (Fig. 2a). Since the amount of Fe 2+ in solution is presumably low at these early stages, the consumption of O 2 during this period could be attributed to Fe 2+ oxidation at surface sites to form the superoxide anion (O 2 ·− ) and to the subsequent production of H 2 O 2 (equations 1 and 2). An interesting result was the formation of O 2 (aq) as a by-product in anoxic experiments ( Fig. 2b and Supplementary Fig. S1b). Although O 2 is not a direct product in the equation 3, since H 2 O 2 is formed, the formation of O 2 can occur via several pathways such as the Fenton-like reaction, the "catalase-like reaction" 25 or via non-radical disproportionation of H 2 O 2.
XPS and HRTEM results reported herein showed the formation of Fe 3+ − O patches over pyrite surfaces. The implication of these Fe 3+ − patches during aqueous pyrite oxidation is not clear. Some studies argued that the electron cycling of Fe 2+ and Fe 3+ between unoxidized and oxidized areas favors the electron transfer from the surface of the pyrite to molecular O 2 , increasing the oxidation rate of pyrite 56 . However, this mechanism is based on atmospheric oxidation of pyrite and does not explain the formation of H 2 O 2 . In contrast, other studies have suggested that the formation of these Fe 3+ − patches interrupts the redox cycling of Fe 2+ /Fe 3+ , thereby inhibiting the formation of H 2 O 2 and decreasing the rate of pyrite dissolution 25 . Our cycling experiments showed that most of the defect sites of the pyrite microparticles remained active at the end of the experiment, even when H 2 O 2 was no longer observable in solution (Fig. 4). Moreover, it is expected that oxidized patches will desorb at low pH during the pyrite dissolution process. HRTEM images showed an appreciable increment of the pyrite alteration layer ( Supplementary Fig. S9), suggesting that these Fe 3+ − patches failed to completely block the surface renewal during pyrite dissolution.
Based on our experimental results, we propose a model that explains the generation-decay trend of H 2 O 2 in terms of a kinetic competition between (1) the formation of H 2 O 2 by the self-oxidation of iron-sulfur cluster defect sites; and, (2) the degradation of H 2 O 2 by the Fenton reaction triggered by pyrite dissolution. Accordingly, the evolution of H 2 O 2 in solution can be summarized as follows: The model allowed us to estimate the rate constants of H 2 O 2 formation under both oxic and anoxic conditions (Table 1). Peak production of H 2 O 2 was shifted towards longer times in anoxic conditions ( Supplementary Fig. S10), pointing to slower oxidation kinetics of pyrite in the absence of O 2 , a result also supported by our XPS analysis ( Supplementary Figs S4 and 5) and by the sulfate and iron released by pyrite dissolution ( Supplementary Fig. S11). When compared with experimental data, the model reproduces qualitatively well the observed trends for pH, O 2 , Fe total and SO 4 2− , as these parameters are calculated using the overall equations of pyrite dissolution 17,57,58 (Fig. 6a-b). An additional feature of the model is that provides a way to reconcile the classical dissolution approach with the free radical assumption, opening a new pathway to analyze the flux of secondary ROS resulting from the degradation of H 2 O 2 through the Fenton reaction (Fig. 6c-d). OH · radicals were the first species produced, rapidly decaying to HO 2 · /O 2 ·− . These ROS species counterbalances the decrease in pH and promoted the so-called "Fenton like" reaction, which resulted in the formation of O 2 (g). Fe 2+ regenerated through this sequence, and also through the dissolution of pyrite, makes the Fenton reaction more efficient 59 . Therefore, the disappearance of H 2 O 2 in solution was likely due to a fast transformation into ROS, catalyzed by dissolved Fe 2+ , rather than by the cessation of the generation mechanism itself. This result is consistent with the continuous production of H 2 O 2 and OH ·23,26,60 during pyrite dissolution. We hypothesize that as a result of the progressive acidification of the solution together with iron recycling by the Fenton reaction, the abiotic dissolution of pyrite microparticles can be considered as a natural and auto-catalytic Fenton reagent, useful to understanding long-term oxidation processes even in oxygen limited environments.

Methods
Sample preparation and characterization. Natural pyrite samples (Logroño, Spain) were ground to obtain particles with average diameter of 1.4 μ m, Laser Diffraction Particle Size Analyzer (LS13320) and BET (Brunauer -Emmett -Teller) surface area of 1.46 m 2 /g (multi-point N 2 adsorption). Prior to use, freshly ground pyrite microparticles were washed by sonication in ethanol (96%) and hydrochloric acid (HCl 0.25 M), to remove organics and oxide surface coatings. Samples were then rinsed with deoxygenated deionized water (MilliQ) and dried in a low vacuum chamber purged with nitrogen (N 2 ) until used. Minor and trace elements in the acid washed samples were evaluated by scanning electron microscopy (SEM) using X-ray mapping (XRM) (Supplementary Fig. S8). X-ray diffraction (XRD) was used to assess the presence of secondary phases and to characterize the degree of structural disorder in the samples using the Rietveld method ( Supplementary Fig. S12).
Batch kinetic experiments. The kinetics of H 2 O 2 formation and degradation on pyrite slurries was investigated in batch reactors utilizing amperometric sensors for H 2 O 2 (ISO-HPO-100, World Precision Instruments, Inc.) and dissolved O 2 (Unisense DK), and a glass electrode for pH determination (Vernier FPH-BTA). Batch reactors were designed to fit with microsensors, spectroscopic probes, ports for pyrite inlet under N 2 atmosphere, and valves. The valves allowed fluid circulation and solution sampling in a closed system configuration.
The production of ferric iron (Fe 3+ ) and hydroxyl radical (OH · ) species by pyrite slurries in absence of dissolved O 2 were measured using UV-vis spectroscopy by monitoring absorption bands at specific wavelengths. To prevent data masking due to particle absorption, the pyrite microparticles were deposited onto silicone strips as a thin film adhered to the inner reactor walls. Spectroscopic data were collected with a fiber optic UV-Vis spectrometer (Black-comet, Stellarnet or USB4000, Ocean Optics) and acquired with the SpectraWiz ® or loggerpro3 codes. In addition, total dissolved iron and sulfate released during pyrite dissolution were measured at different time intervals using inductively coupled plasma mass spectrometry (ICP-MS) and ion chromatography (IC).
All experiments were made under continuous magnetic stirring (500 rpm), at room temperature (T° ∼ 22 °C) and in the dark, unless other conditions are specified. Due to the autocatalytic nature of Fenton chain reactions, kinetic experiments were carried out in unbuffered distilled water. Further details of the batch reactors, electrochemical sensors, experimental procedures and test analysis are given as supplementary material ( Supplementary Figs S13-18). Surface analysis. Cyclic Voltammetry (CV) was employed as an additional way to assess the sequence of redox reactions involving free radicals during H 2 O adsorption at pyrite interface. CV was performed using Pt/ Pyrite-Np's/ Nafion©/ electrodes in N 2 purged solutions vs Ag/Cl 3 M KCl. A detailed description of the experimental set-up is given as Supplementary Material. X-ray photoelectron spectroscopy (XPS) was used to analyze the surface oxidation states of (001) faces of pyrite (single-crystal, ~1 cm 2 × 2 mm) after aqueous reaction in oxic and anoxic conditions. Platelets parallel to (001) faces were cut (1 cm 2 × 2 mm) and cleaned following the same procedure described above and allowed to react with water in oxic and anoxic conditions. Samples were dried and stored in N 2 -filled tubes until introduction into the XPS vacuum chamber. The oxidation states were analyzed at three different stages of the dissolution process: (t 1 ) unreacted sample; the pyrite crystal was acid-washed to generate oxide-free surfaces by removing the normal contaminants, carbon, nitrogen, and oxygen (C, N, O) due to atmospheric oxidation; (t 2 ): sample after 22 hours immersed in oxic water. (t 3 ): sample after 22 hours immersed in anoxic water. Prior to this stepped analysis, the XPS spectra of a pyrite surface was analyzed after argon ion (Ar + ) sputtering to verify the formation of Fe 3+ and S 2− surface species as occurs during mechanical fracture by preferential sulfur removal. XPS Spectra were collected from a take-off angle of 90° relative to the sample surface in a Thermo Scientific K-Alpha ESCA analyzer using monochromatic Al Kα (1486.6 eV) radiation and pass energies of 100 eV and 20 eV for survey spectra and narrow region spectra, respectively. Spectra were aligned by setting the C1s peak to a binding energy of 285 eV 61 . Deconvolution and fitting of experimental data were done Scientific RepoRts | 7:43703 | DOI: 10.1038/srep43703 with the XPSpeak4.2 software (http://www.phy.cuhk.edu.hk/~surface/XPSPEAK/). The Shirley method was used for background subtraction and the binding energies of the species identified were assigned using values taken from literature (Supplementary Tables S2 and 3).
High resolution Electron Transmission Microscopy (HTREM) analysis were performed to identify nano-domains of secondary oxidation products at the pyrite interface. HTREM images were acquired on a JEOL JEM-3011 microscope with accelerating voltage of 200 kV using a Gatan Ultrascan 1000 CCD camera and Digital Microgrograph software. Data processing was performed with the GADDS and image-J codes. Pyrite lamellas were prepared using a focused ion beam (FIB) with a high resolution JEOL JSM-6700 f.
Kinetic model. The model was run using the computer code Copasi 4.8 (COmplex PAthway SImulator) 62 . We assume that the experimental trend of H 2 O 2 , in the presence and absence of O 2 (g), is shaped by three main processes: (i) the rate of H 2 O 2 generation by the iron defect sites on the surface of pyrite particles; (ii) the production rates of Fe 2+ and SO 4 2− by pyrite dissolution; and, (iii) the kinetics of H 2 O 2 degradation by the Fenton reaction. Fits to the experimental curves that describe the evolution of H 2 O 2 were performed by using the Marquardt algorithm employing as adjustable parameters the rate constants of the H 2 O 2 formation. A detailed explanation of the modeling set-up is included in the supplementary material.