Forest litter was collected in a Norway spruce forest in Simlångsdalen, Sweden (N: 56°42′-3.17″; E: 13°6′57.51″). The litter layer was collected after removing the moss mat and avoiding the mineral layer of the soil. The collected litter was subsequently sieved through a 2 mm sieve to remove large roots, mineral particles and remaining mosses. 120 g of fresh litter was boiled for 1 h in 1 L of distilled water. The litter slurry was sieved through a cheesecloth and a sequence of filters with decreasing pore size (2.6, 1.6, 0.7, 0.45 and 0.22 µm). This yielded the DOM (denoted DOM_ini) that was used in subsequent experiments. The composition of a DOM sample extracted from the same soil following the exact same extraction protocol has been characterized in our previous work (Wang et al., 2017) and shown to contain a mixture of carbohydrates, aliphatics and aromatics as well as carbonyl/carboxyl-containing compounds. This overall composition was confirmed by ¹H nuclear magnetic resonance (¹H NMR, Bruker Avance II 400 MHz NMR spectrometer, Billerica, MA, United States) and Diffuse Reflectance Fourier Transform Infrared Spectroscopy (Vertex 80v, Bruker FTIR spectrometer) of DOM_ini (Supplementary Figures S1, S2). Total Organic Carbon (TOC) was determined using an organic C analyzer (TOC-V(CPH); Shimadzu, Kyoto, Japan) and a combustion catalytic oxidation method. Total Nitrogen (TN) was measured with the same apparatus equipped with a TNM-1 detector (Shimadzu, Kyoto, Japan). Samples were acidified to pH < 2 with 1 M HCl before TOC and TN measurements.

Because the DOM_ini was obtained from a forest soil, it contained various elements, in addition to the organic compounds. Most important for this study was the native content of iron and phosphate. Total Fe and P concentrations were analyzed in both fractions using Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) (Perkin Elmer Optima 8300, Waltham, MA, United States). The P and Fe concentrations are presented in Supplementary Table S1, re-calculated from the DOM stock solutions to the actual values in the different batch experiments. The iron concentrations were non-negligible and needed to be considered when the reductive dissolution effects of the DOM fractions were analyzed. The phosphate concentrations were low in comparison with the TOC concentrations in the experiments, corresponding to a surface coverage in the range of 0.008–0.012 μmol m⁻². As a result, only small competitive adsorption effects from phosphate were expected, and accordingly, very minor IR spectral contributions from phosphate bands to the carbohydrate region 900–1,200 cm⁻¹.

The ECM fungus P. involutus is commonly found in boreal and northern temperate forests with a soil pH of around 4 (Wallander and Söderström 1999; Hobara et al., 2016). For this reason, all batch and IR experiments were conducted at pH 4.0.

P. involutus (Batsch) Fr. (ATCC 200175) was grown for 9 days in the dark at 20°C on a monolayer of sterile 4 mm diameter glass beads and 10 ml of Fries medium (Supplementary Figure S3) (Fries, 1978). The medium used contained 3.74 mM NH₄Cl, 0.41 mM MgSO₄·7H₂O, 0.22 mM KH₂PO₄, 0.18 mM CaCl₂·2H₂O, 0.34 mM NaCl, 1.34 mM KCl, 0.24 mM H₃BO₃, 20 µM ZnSO₄·7H₂O, 5.01 µM CuSO₄·5H₂O, 50.29 µM MnSO₄·H₂O, 0.16 µM (NH₄)₆Mo₇O₂₄·7H₂O, 74 µM FeCl₃·6H₂O, 33.3 mM D-glucose, 55.51 µM myo-inositol, 0.3 µM thiamine·HCl, 0.1 µM biotin, 0.59 µM pyridoxine, 0.27 µM riboflavin, 0.82 µM nicotinamide, 0.73 µM p-aminobenzoic acid, and 0.46 µM Ca-pantothenate. The pH was corrected to 4.8. After 9 days, the Fries medium was removed and replaced with 10 ml Fries medium minus NH₄Cl to starve the cultures for N, which significantly stimulates the DOM-decomposition activity in P. involutus (Nicolas et al., 2018). After 24 h, the N-free medium was removed and replaced with a DOM_ini solution supplemented with 2.5 g L⁻¹ of glucose, and this DOM_ini-glucose solution was filter sterilized through a pore size of 0.22 µm prior to the incubation experiments. Glucose was added since previous experiments have shown that the supply of an easily available C-source is required by P. involutus to oxidize DOM (Rineau et al., 2013). Cultures were grown on this extract for another 7 days and culture media were subsequently sampled. After these 7 days, almost all glucose was consumed. The decomposed DOM, now also containing secondary metabolites secreted by P. involutus (Wang et al., 2017), was filtered through a 0.22 µm filter to remove any remaining hyphae and finally freeze-dried for further use. The decomposed DOM is further referred to as modified DOM (DOM_mod).

It has been shown that the main iron-reducing activity of DOM decomposed by P. involutus can be extracted by ethyl acetate (Shah et al., 2015). Herein, we adopted this extraction procedure to split the DOM samples into an ethyl acetate (DOM_ini-EtOAc or DOM_mod-EtOAc) and a water soluble (DOM_ini-H2O or DOM_mod-H2O) fraction. 60 mg of freeze-dried DOM_ini or DOM_mod collected from the P. involutus cultures were dissolved in 600 ml MilliQ water. Aliquots of 5 ml were acidified with 1 ml 1 M HCl. These mixtures were extracted with 5 ml ethyl acetate during 10 min of vigorous shaking, and phase separation was accomplished using a separatory funnel. This procedure was repeated three times. The resulting ethyl acetate and water fractions were pooled and freeze-dried. These freeze-dried materials were used in subsequent batch and IR experiments. TOC, TON, total Fe and P were determined, and the DOM_mod-EtOAc and DOM_mod-H2O fractions were also characterized by means of diffuse reflectance IR Spectroscopy and ¹H NMR as described in Section Collection of Forest Litter, and Preparation and Chemical Characteristics of Dissolved Organic Matter.

6-line ferrihydrite and goethite were synthesized and characterized according to Krumina et al. (2016). The obtained mineral particles were identified by X-ray diffraction (XRD) and transmission electron microscopy (TEM). The specific surface area (SSA) of the goethite was determined to 71.5 m² g⁻¹ using the BET N₂ adsorption method (Micromeritics Gemini VII 2390a, Norcross, GA, United States) (Brunauer et al., 1938). The SSA of the 6-line ferrihydrite was estimated to be 300 m² g⁻¹, according to a procedure previously described (Krumina et al., 2017). The concentrations of the stock suspensions of goethite and ferrihydrite were 12.8 and 2.3 g L⁻¹, respectively. These suspensions were purged with N₂ gas to remove carbonates and dissolved CO₂ and stored at 4°C. The ferrihydrite experiments were carried out with particles aged for 1–3 months. This material showed only an XRD pattern characteristic of 6-line ferrihydrite (Supplementary Figure S4). All experiments were performed in 0.1 M NaCl solutions, where pH was adjusted to the desired values with 40 mM NaOH or HCl solutions in 0.1 M NaCl.

Batch experiments were conducted at total organic carbon (TOC) concentrations of 4 μmol m⁻² (800 µM TOC), 8 μmol m⁻² (1,600 µM TOC) and 12 μmol m⁻² (2,400 µM TOC) of ferrihydrite and goethite surface area, respectively. All stock solutions and reactions were adjusted to and performed at pH 4.0 in 0.1 M NaCl medium, where pH was adjusted with 40 mM NaOH or HCl solutions in 0.1 M NaCl solution. Stock mineral suspensions were diluted to yield the same surface area per volume liquid of 200 m² L⁻¹ for all batch experiments. During the experiments, the reaction tubes (15 or 50 ml non-pyrogenic polypropylene, Sarstedt) were covered by aluminum foil to avoid any exposure to light and mixed using a Grant-bio PTR-35 360° vertical multi-functional rotator (Grant Instruments, Royston, United Kingdom). Batch experiments were performed in triplicates.

Fe²⁺ concentrations were measured by means of the ferrozine assay (Goodell et al., 2006). 100 µl of 1% ferrozine (w/v) solution was added to each sample and mixed for 5 min. The measurements were performed at 562 nm using an Ultraspec 3000 UV/Visible spectrophotometer (Pharmacia Biotech, Uppsala, Sweden) with a limit of detection (LoD) of 2.0 µM). Blanks contained 100 µl 1% ferrozine solution in 0.1 M NaCl solution with a total volume of 1 ml Fe²⁺ concentrations were calculated from a standard curve covering a range of 0–100 µM Fe²⁺. In the reductive dissolution experiments, the total Fe concentrations measured by ICP-OES were close to the Fe²⁺ concentrations determined in the ferrozine assay (≤5%) and are therefore not reported here.

300 µM terephthalic acid (TPA) was used as a probe for measuring HO^• production. TPA is oxidized by HO^• to hydroxy terephtalic acid (hTPA) which can be readily detected by fluorescence spectroscopy (Jing and Chaplin, 2017). After sampling mineral suspensions, an excess of phosphate (3 μmol m⁻²) was added to the samples and mixed for 1 h in order to remove adsorbed hTPA from the iron oxide surfaces (Lyngsie et al., 2018). The fluorescence intensity was measured at an excitation wavelength of 315 nm and an emission wavelength of 425 nm with an LS50B luminescence spectrometer (Perkin Elmer, Waltham, MA, United States). The LoD of this analytical method was 0.5 nM hTPA. The hTPA concentrations in the batch samples were calculated from a linear standard curve of hTPA (0–500 nM) in 0.1 M NaCl. Two types of experiments were performed: 1) simultaneous addition of DOM, TPA and 100 µM (0.4 μmol m⁻²) H₂O₂ to the iron oxide suspensions and 2) pre-adsorption of DOM to the iron oxides for 24 h prior to addition of TPA and 100 µM H₂O₂. The H₂O₂ concentration was chosen to be substantially below the total number of surface sites on the iron oxides, and 100 μM H₂O₂ is only slightly above concentrations detected in natural waters e.g., in rainwater (Yu and Kuzyakov, 2021).

IR spectroscopic measurements were performed according to a method originally presented by Loring et al. (2009) and then slightly modified by Krumina et al. (2016). Iron oxide overlayers were prepared on an attenuated total reflectance (ATR) ZnSe crystal by evaporating 0.7 ml of goethite (2.0 g L⁻¹) or ferrihydrite (2.3 g L⁻¹) suspension. The ATR crystal, which forms the bottom of a titration vessel, was connected to an ATR accessory (FastIR, Harrick Scientific Products, Pleasantville, NY, United States) and placed inside an evacuated (1.42 mbar) Fourier transform infrared spectrometer (Vertex 80v, Bruker, Billerica, MA, United States) in a climate-controlled laboratory with a set temperature of 21°C. Either ferrihydrite or goethite suspensions in 0.1 M NaCl medium were added to the titration vessel and slowly stirred using a propeller stirrer. pH was adjusted to the desired value and kept constant with an automated and computer-controlled burette system (907 Titrando and Tiamo 2.4 software, Metrohm AG, Herisau, Switzerland) during all experiments. A background spectrum of 4096 co-added scans was collected of the overlayer and the iron oxide suspension. In accordance with the batch experiments, two different types of IR experiments were conducted, i.e., one where all reactants were added simultaneously to the iron oxide suspension, and one where DOM was pre-adsorbed for 24 h prior to addition of H₂O₂. During the first hour of either DOM or DOM + H₂O₂ addition, data were collected at a rate of 1 min per spectrum. After this initial period, data collection was increased to ∼7.5 min per spectrum. All IR spectra were recorded over the range 700–4,000 cm⁻¹ at a resolution of 4 cm⁻¹.

The IR spectral data sets were analyzed using the multivariate curve resolution-alternating least squares (MCR-ALS) formalism implemented in the MATLAB program PLS Toolbox v. 8.6 (Eigenvector Research Inc., Manson, WA, United States). Non-negative constraints were applied to both the concentrations and the spectra, and the spectral components were modeled with the same maximum intensity. A critical parameter in this analysis is the initial assumption of the number of components needed to explain the variation in the data sets, and this was assessed from singular-value decomposition (Tauler and De Juan, 2006).

Due to difficulties in separating adsorbed DOM from minerals using conventional filtration and centrifugation techniques, we instead estimated roughly the surface saturation of the DOM from IR spectroscopy data. DOM fractions were added to the iron oxide suspension and adsorption was monitored as a function of IR intensity. When no, or a very slow change in IR absorbance was observed, another aliquot of the DOM fractions was added. This procedure was repeated until addition of DOM fractions resulted in no or very little change in IR absorbance, i.e., the final addition of a given DOM fraction increased the IR absorbance by < 5% within several hours of adsorption.

In agreement with results from our previous studies, N-starved P. involutus cultures consumed a substantial part of the total N present in DOM_ini (Supplementary Table S1) (Rineau et al., 2012; Wang et al., 2017). This should be compared to the much smaller decrease in P and Fe after incubation (Supplementary Table S1), which may be due to a limited fungal uptake and/or losses during extraction. The extraction recovered ca. 70% of the starting material (DOM_mod) and of this recovered material ca. 30% of the TOC was contained in DOM_mod-EtOAc and ca. 70% in DOM_mod-H2O. The combined IR and ¹H NMR analyses showed that aromatic compounds were extracted into the ethyl acetate phase while carbohydrates remained mainly in the water phase (Supplementary Figures S1, S2). Carboxylic acids were distributed over both solvents but seemed to be relatively more prevalent in the ethyl acetate phase as indicated by the IR data. In Wang et al. (2017) we showed that decomposition of DOM_ini by P. involutus caused a decrease in reduced sugars and an increase in aromatic and phenolic functional groups as well as an increase in functional groups containing oxidized carbon atoms, presumably carboxylic acids. Moreover, data from ¹³C-labelling experiments in Wang et al. (2017) showed that parts of these compounds was secreted by the fungus. Thus, both aromatics originally present in DOM_ini and secreted by the fungus were likely contained in DOM_mod-EtOAc whereas modified and unmodified carbohydrates were present mainly in DOM_mod-H2O.

Initial experiments showed that DOM_ini caused a small but detectable reductive dissolution of ferrihydrite and goethite (Supplementary Figure S5). In the first time points, though, the concentration of Fe²⁺ was undetectable or significantly lower than the total Fe contained in DOM_ini added. Thus, if Fe²⁺ was present in DOM_ini it was rapidly removed in the presence of both ferrihydrite and goethite, either via adsorption or re-oxidation (Krumina et al., 2017). The slow increase of Fe²⁺ that followed the addition of DOM_ini to the iron oxide suspensions was accordingly a result of reductive dissolution. The reduction of ferrihydrite was more extensive than that of goethite, which is consistent with the higher reduction potential of ferrihydrite (Gorski et al., 2016).

DOM_ini was extracted into DOM_ini-EtOAc and DOM_ini-H2O, and the reductive dissolution of ferrihydrite was compared with the effects caused by the corresponding fractions of DOM_mod. In the case of DOM_ini-EtOAc and DOM_ini-H2O the Fe²⁺ reducing activity was completely captured in the ethyl acetate fraction (Figure 1); note that the Fe²⁺ concentrations in the DOM_ini-H2O experiments were at or below the LoD of the ferrozine assay. The reductive capacity markedly increased in the DOM_mod-EtOAc and DOM_mod-H2O, and at the endpoint of the experiment the increase was ca. 2.6 times for both fractions as compared to the corresponding DOM_ini fractions. Although the extent of reduction promoted by DOM_mod-EtOAc was far greater than DOM_mod-H2O, the increase caused by the latter was significant (Figure 1). These results showed that DOM decomposition by the ECM fungus P. involutus generated a modified end-product that had a larger Fe-reducing capacity, and that this property was conserved in the lyophilized and re-dissolved culture filtrate.

Further experiments were focused on the DOM_mod-EtOAc and DOM_mod-H2O fractions and the effects of TOC concentrations. In general, the extent of reductive dissolution of ferrihydrite and goethite was directly related to the TOC concentrations (Figure 2). The exception was the reaction between DOM_mod-EtOAc and goethite where the experiments performed at 8 and 12 μmol m⁻² yielded similar results. In agreement with the DOM_ini experiments, we observed a greater extent of dissolution for ferrihydrite than goethite, as suggested by the difference in reduction potentials of the minerals. However, comparison of the experimental endpoints showed that the relative effects of DOM_mod-EtOAc and DOM_mod-H2O with respect to ferrihydrite and goethite reduction were quite different. The reductive dissolution promoted by DOM_mod-EtOAc in the presence of ferrihydrite increased by a factor of ca. 7 compared to goethite while in the case of DOM_mod-H2O the increase was only a factor of ca. 1.5 (Figure 2). Hence, DOM_mod-EtOAc seemed to contain a relatively high concentration of compounds with a reduction potential sufficiently low to reduce ferrihydrite but not goethite to any greater extent. In contrast, DOM_mod-H2O contained a relatively low total concentration of reducing compounds but with a low reduction potential that allowed a more similar reduction of ferrihydrite and goethite. Thus, decomposition of DOM by P. Involutus and the concomitant secretion of extracellular metabolites seem to generate a modified DOM containing molecules with a range of reduction potentials. Previous experiments where P. Involutus was grown on either DOM or proteins in absence of iron oxides have also shown that this fungus produces an array of iron (III) reducing metabolites (Shah et al., 2015; Shah et al., 2021).

The extracts from the modified DOM unquestionably promoted significant reductive dissolution of ferrihydrite and goethite. However, compared to typical values of surface site concentrations on the minerals the extent of dissolution was small as can be seen from the Fe²⁺ concentrations recalculated to units of μmol m⁻² (Figures 1, 2). For instance, the concentration of singly coordinated surface groups (≡Fe-OH) on ferrihydrite has been estimated to ca 10 μmol m⁻² (Hiemstra and Van Riemsdijk, 2009), which should be compared to the ca. 0.17 μmol m⁻² Fe²⁺ dissolved from ferrihydrite by DOM_mod-EtOAc after 24 h (Figure 2A). Accordingly, the net-generation of Fe²⁺ in solution indicated that only small fractions of the ferrihydrite and goethite surfaces were affected by the reductive dissolution processes.

The Fe²⁺ concentration data in Figure 2 can be fitted with the equation: [ F e 2 + ] = A ( 1 − e − k f a s t t ) + B ( 1 − e − k s l o w t ) (2)

The rate constant describing the fast process, k _fast , was poorly constrained because of the limited time resolution at the early stages of our experiments whereas k _slow was better defined by the shape of the [Fe²⁺] increase after the initial 2–10 min of the reactions. As shown in Table 1, k _slow is practically independent of TOC concentration except for the DOM_mod-EtOAc-goethite system. In this case the rate constant increased with increasing TOC concentration, and as already mentioned the same system showed similar extent of reductive dissolution at 8 and 12 μmol m⁻² indicating surface saturation of the reductant (Figure 2C). Thus, the increase in rate constant may be related to the excess of reductant with respect to surfaces sites active in the dissolution process. In contrast, the almost constant k _slow obtained for the other systems indicated a high ratio of surface-active sites relative to reducing molecules.

In-situ IR spectra were collected during the simultaneous adsorption-reductive dissolution reactions between the DOM fractions and the iron oxides. Spectra representing the endpoints (i.e., reaction time 1,440 min) as well as experiments performed at a higher TOC concentration (60 μmol m⁻²) are shown in Figure 3. In accordance with the chemical characterization of the DOM fractions (Supplementary Figures S1, S2) and the results in Wang et al. (2017), IR spectra of adsorbed DOM_mod-EtOAc were characterized by features from carboxyl groups (bands (2) and (4), (Figure 3A) and aromatics (band (3), (Figure 3A). While those of DOM_mod-H2O displayed typical strong carbohydrate bands between ca. 900–1,200 cm⁻¹ and features indicating other oxidized functional groups such as carbonyls and carboxyls in the region between 1,400–1800 cm⁻¹ (Figure 3B). The IR spectral dependence on TOC concentration was different for ferrihydrite and goethite. In the case of ferrihydrite, an increase of the DOM_mod-EtOAc concentration resulted in a significant shift of the asymmetric carboxyl band (band (2)), marked changes in the carbonyl region around band (1) (Figure 3A (I) and (II)), and an increase in band (3) originating from aromatics (Figure 3A (I) and (II)). The goethite spectra, however, were practically unaffected by an increase of DOM_mod-EtOAc concentration, except for a stronger relative contribution from the aromatic band (band (3), Figure 3A (III) and (IV)). It is also notable that band 1) at 1,660 cm⁻¹ was not observed in the goethite spectra. The comparison of TOC concentration dependence was more difficult to make for DOM_mod-H2O due to a generally low signal-to-noise ratio with resulting baseline problems, in particular for the experiments with goethite performed at low TOC. Still, comparison of the intensities of bands (5) and (6), representing oxidized functional groups, to the carbohydrate signals around 1,030 cm⁻¹ suggested that the former bands displayed a relative increase for ferrihydrite but not for goethite (Figure 3B).

The time evolution of the IR spectra during the coupled adsorption-reductive dissolution reactions was analyzed by means of the MCR-ALS formalism that resolves component IR spectra and the corresponding contribution profiles. The resolution of the different components relies on differences in their rate of spectral change as a function of time, i.e., if two species display very similar rate behavior they will not be resolved, and their spectral characteristics will be included in the same component. Two components were sufficient to explain the IR data sets collected during the reactions between DOM_mod-EtAOc and ferrihydrite or goethite yielding fits of 99.99% of the variance or better (Supplementary Figures S7, S8). The ferrihydrite spectra were deconvoluted into a dominant species containing the spectral signatures of both carboxylates and aromatics and a minor species displaying slower kinetics and characteristics of carboxylates together with the band at ca 1,660 cm⁻¹ identified in Figure 3, but no aromatics. The two goethite components were similar and both contained carboxylate and aromatic bands. The main difference between these spectra seemed to be caused by variations in the signal from adsorbed water contributing to the spectra around 1,620 cm⁻¹ (Supplementary Figure S7).

The IR spectra of DOM_mod-H2O on ferrihydrite and goethite were analyzed at TOC = 60 μmol m⁻² because of the poor signal-to-noise at lower TOC concentrations. On ferrihydrite MCR-ALS resolved two unique spectra, one representing carbohydrates and a second minor carboxylate species (Supplementary Figure S8). This second species also displayed a slow time evolution while the signal from the former species was saturated after ca. 400 min. The two goethite components were characterized by close to identical IR bands of carbohydrates and differed mainly in their baselines close to one of the main goethite bands at 890 cm⁻¹ (Supplementary Figure S8).

The integrated absorbance over the range 1,850–900 cm⁻¹ of the IR spectra provided an estimate of the total adsorption of the DOM fractions. Comparison between these integral values and the Fe²⁺ concentrations obtained from reductive dissolution revealed a correlation between total adsorption and dissolution (Figure 4). The correlation indicated that the rate of reductive dissolution was controlled by the total adsorption of DOM even though the organic molecules responsible for the actual reduction likely constituted only a minor part of each DOM fraction. Figure 4 also shows that adsorption of the DOM fractions seemed to be faster on goethite than ferrihydrite. Moreover, the goethite data indicated a steady state after ca. 1,000 min while the integrated absorbances for ferrihydrite displayed a continued slow increase.

The integrated IR absorbances were also used to analyze the adsorption capacity of ferrihydrite and goethite with respect to DOM_mod-EtOAc and DOM_mod-H2O (Supplementary Figure S9). Incremental additions of the DOM fractions showed that TOC surface saturation was approached around 50–60 μmol m⁻² for DOM_mod-EtOAc and the adsorption isotherms were similar for both minerals. The DOM_mod-H2O adsorption leveled off at much higher TOC concentrations of 150 μmol m⁻² and above, and the difference between the isotherms of ferrihydrite and goethite was quite large. The detailed interpretations of these isotherms are outside the scope of this work, but it is important to note that all dissolution experiments were performed at TOC concentration far below surface saturation.

The detection of hydroxyl radicals (HO^•) was accomplished with TPA, which has been shown to be a suitable probe molecule because it forms one product, is resistant to other oxidative processes that would yield false positives and is not easily reduced (Jing and Chaplin, 2017). All experiments were conducted at a fixed TPA concentration of 1.5 μmol m⁻² of ferrihydrite or goethite. A previous study has indicated surface saturation of TPA on ferrihydrite at pH 4.5 around 0.4–0.5 μmol m⁻² (Lyngsie et al., 2018), thus in our experiments TPA should be distributed over both the solid surfaces and the bulk water phase. In the presence of the DOM fraction, it is likely that competitive effects shift this distribution and drive more TPA into solution.

Initial experiments performed in absence of DOM showed large differences in reactions between H₂O₂ and ferrihydrite or goethite (Figure 5). The reaction with ferrihydrite generated significant amounts of HO^• that increased almost linearly with time, while radicals generated in the goethite system were barely detectable (Figure 5, green lines).

Addition of H₂O₂ to the DOM-iron oxide suspensions effectively oxidized Fe²⁺, and the resulting Fe²⁺ concentration after addition of H₂O₂ was below the ferrozine assay LoD in all cases. Equilibration of DOM_mod-EtOAc with ferrihydrite or goethite 24 h prior to addition of H₂O₂ increased the generation of detectable hydroxide radicals (Figures 5A,B). Ferrihydrite was more efficient in generating HO^• compared to goethite, which is consistent with the more extensive reductive dissolution of ferrihydrite. Initially, the amounts of HO^• detected was correlated to the Fe²⁺ produced during pre-equilibration (Figure 5; Table 2), but in the case of ferrihydrite the HO^• concentrations converged almost to the same value after 1,440 min irrespective of TOC concentration showing that the total concentration of H₂O₂ was the limiting factor at extended reaction times. For goethite, however, the concentrations of HO^• depended on the TOC concentrations and the slow increase in HO^• after ca. 200 min was in agreement with the slow release of Fe²⁺ during extended reaction times between DOM_mod-EtOAc and goethite (cf. Figures 2, 5).

In line with the low reductive capacity of DOM_mod-H2O the amounts of HO^• generated from the pre-equilibrated suspensions were initially lower than for DOM_mod-EtOAc (Figure 5). For ferrihydrite, though, the HO^• concentrations approached those detected in the presence of DOM_mod-EtOAc, and showed similar increase with time as observed in the absence of DOM. In fact, simultaneous addition of DOM_mod-EtAOc and H₂O₂ produced identical results to the addition of H₂O₂ only, whereas pre-equilibration with DOM_mod-H2O resulted in lower linear rates that were independent of TOC concentrations. The lower rates in the presence of pre-adsorbed DOM are presumably due to surface competition between H₂O₂ and DOM. In the goethite system the slow increase of HO^• with time was dependent on TOC concentrations and again followed the trends of reductive dissolution (cf. Figures 2D, 5D).

The batch experiments above showed that HO^• was formed in ferrihydrite and goethite suspensions containing H₂O₂ in the absence and presence of DOM fractions, and that the extent and rates strongly depended on the properties of the DOM fractions and the iron oxides. In these experiments, the precise distribution of the TPA probe molecules were unknown but, as discussed above, these molecules were likely present both on the iron oxide surfaces and in bulk solution. Thus, from the TPA measurements we cannot ascertain if the radical oxidation took place at the surface or in solution. Further information was obtained from complementary in-situ IR spectroscopic measurements of the DOM fractions adsorbed to the iron oxides during the reactions with H₂O₂. In contrast to the TPA batch experiments, these measurements are sensitive to any changes of the adsorbed organic matter caused by the addition of H₂O₂. The IR spectra recorded of ferrihydrite and goethite surfaces pre-equilibrated with the DOM fractions for 24 h and subsequently reacted with H₂O₂ showed distinct trends for ferrihydrite but not for goethite (Figure 6). Overall, these results suggest that DOM fractions associated with ferrihydrite were more susceptible to radical oxidation than fractions associated with goethite. The DOM_mod-EtOAC-ferrihydrite system displayed the strongest response to H₂O₂ addition, and bands characterizing oxidized functional groups such as carboxyls and carbonyls increased in relative intensity while the intensities of aromatic/phenolic bands decreased (Figure 6A). The changes in the DOM_mod-H2O-ferrihydrite system also indicated a relative increase in oxidized carbon but in this case, it was accompanied with a decrease in carbohydrate bands (Figure 6C). The IR spectra of DOM_mod-EtOAc-goethite were practically unaffected by H₂O₂ addition. The low signal-to-noise of the DOM_mod-H2O-goethite spectra and the strong baseline variations close to a major goethite band at 890 cm⁻¹ prevented a closer analysis but there were no indications of systematic changes after H₂O₂ addition.

The band at 1,660 cm⁻¹ indicated H₂O₂-promoted oxidation of both DOM_mod-EtOAc and DOM_mod-H2O adsorbed on ferrihydrite. Notably, in the absence of H₂O₂ this band also increased during the adsorption of these DOM fractions on ferrihydrite but not on goethite (Figure 3; Supplementary Figures S7, S8). Using the intensity of the 1,660 cm⁻¹ band as a measure of oxidation and plotting this as a function of time showed that the oxidation rates of adsorbed DOM were linearly correlated to the rate of TPA oxidation by H₂O₂ in presence of ferrihydrite but in absence of DOM fractions (Figure 7) indicating the same mechanism of oxidation in both cases. This implies that the initial fast generation of HO^• that was detected with TPA was caused by the Fe²⁺ already released into solution during pre-equilibration between the DOM fractions and the iron oxides (Figure 3), and that the Fenton reaction occurring in bulk solution had no detectable effect on the adsorbed DOM.

To further explore the mechanisms of H₂O₂-mediated oxidation of ferrihydrite-associated DOM we conducted additional IR experiments where DOM_mod-EtOAc and H₂O₂ were added simultaneously and where DOM_mod-EtOAc was added 60 min prior to H₂O₂ i.e., H₂O₂ was added at a point when the dissolution rate of Fe²⁺ was high and the surface coverage of DOM was also considerable (cf. Figures 2A, 4A). In comparison with the DOM_mod-EtOAc experiment conducted in absence of H₂O₂ we observed a significant increase of the 1,660 cm⁻¹ band irrespective of the timing of H₂O₂ addition (Figure 8A). However, the extent and rate of the increase was strongly affected by the time and sequence that DOM_mod-EtOAc and H₂O₂ were added to the suspension (Figure 8B). The most pronounced increase in the 1,660 cm⁻¹ intensity was registered when H₂O₂ was added 1 h after DOM_mod-EtOAc, thus during conditions when reductive dissolution occurred in the presence of a high surface concentration of DOM_mod-EtOAc. Simultaneous addition of DOM_mod-EtOAc and H₂O₂ yielded an initial rapid increase that quickly leveled off whereas addition of H₂O₂ to a DOM_mod-EtOAc-ferrihydrite suspension that was pre-equilibrated for 24 h resulted in a slow, almost linear increase, as already described in the previous paragraph.

In addition to using a specific IR band intensity as indicator for radical oxidation, the complete IR spectral changes over the region 1,800–800 cm⁻¹ of DOM_mod-EtOAc adsorbed to ferrihydrite during reaction with H₂O₂ were analyzed using the MCR-ALS formalism. These analyses showed that spectral variations in the data sets were explained by two components (Figure 9). Experiments where H₂O₂ was added 1 or 24 h after pre-equilibration between DOM_mod-EtOAc and ferrihydrite were both characterized by one spectral component containing a strong band at 1,660 cm⁻¹ (Figures 9A,C, blue lines). Moreover, the same components increased in importance relative to the second MCR component, and the shape of these increases matched those describing the 1,660 cm⁻¹ intensity (cf. Figures 8B, 9B,D). Note that the MCR contributions in Figure 8 should not viewed as concentrations due to the constraints used in the modeling and the strong correlations between MCR spectral intensities and contribution profiles. Nevertheless, the ratios between the contributions highlight the strong increase of the component characterized by the 1,660 cm⁻¹ band relative to a component containing carboxylic and aromatic groups as previously discussed. The pronounced H₂O₂-mediated oxidation induced after 1 h pre-equilibration yielded an MCR spectrum that showed a very close resemblance to oxalate adsorbed to ferrihydrite under similar conditions (Figure 9C). Thus, extensive radical oxidation of the adsorbed DOM_mod-EtOAc generated substantial amounts of adsorbed oxalate, and the 1,660 cm⁻¹ band was a strong indicator of this process. At the same time, the difference between the MCR spectra of the oxidized component (Figures 9A,C, blue spectra) suggested that oxidation produced not only oxalate but likely a range of oxidized adsorbed DOM fragments.