Phosphate immobilization by mackinawite in freshwater sediments and the competitive role of organic carbon

Abstract

Iron(III) (oxyhydr)oxides are the major sink for PO₄³⁻ in freshwater sediments, helping to control the onset of eutrophic conditions when PO₄³⁻ is in excess; however, this sink is transient as ferrous iron and PO₄³⁻ are typically released when sediments become anoxic, triggering a self-sustaining reversible cycle as sediments become an internal source of phosphate. Ferrous iron can then react with dissolved sulfides to form iron sulfide minerals such as mackinawite (FeS), whose interactions with PO₄³⁻ remain largely unexplored. We investigate the role of FeS in the sequestration of PO₄³⁻ under anoxic conditions, with or without natural organic carbon (OC). We show that Fes has a high affinity for PO₄³⁻, with a maximum sorption capacity comparable to moderately crystalline FeO_x minerals. The presence of sorbed OC decreases the FeS affinity toward the PO₄³⁻ by ~ 75%, which is attributed to the lower surface area of FeS-OC compared to FeS, electrostatic repulsion, and the reduced accessibility to reactive Fe(II) sites. Moreover, we show that upon exposure of FeS to oxic conditions, PO₄³⁻ is efficiently recaptured by the FeS oxidation products (FeOx), hence highlighting a parallel self-sustaining reversible cycle resulting in the permanent sequestration of a small fraction of PO₄³⁻. Overall, these findings highlight FeS as a potentially important phosphate sink in anoxic freshwater sediments.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-52964-z.

Introduction

Phosphorus, commonly found in the form of phosphate (PO₄³⁻, used in this work to represent all forms of orthophosphate) and organic phosphorus, is a vital nutrient for all aquatic organisms. It is one of the key components of DNA and RNA and plays an essential role in energy transfer through adenosine triphosphate (ATP) molecules¹. However, when present in excessive amounts in aquatic systems, it can trigger harmful algal blooms and lead to poor water quality in nutrient-rich aquatic environments^2–4. In both freshwater and marine ecosystems, the distribution, transport and biological availability of PO₄³⁻ are heavily influenced by its interactions with iron^5–8. Iron is among the most prevalent metals found in soils and sediments, where it forms highly reactive minerals such as short-range-ordered and poorly crystalline Fe(III) (oxyhydr)oxides^9–11. These minerals exhibit a strong binding affinity for PO₄³⁻ in oxygen-rich surface sediments¹². Among the various FeOx minerals (in this work, FeO_x indiscriminately represents iron oxides, hydroxides and oxyhydroxides), ferrihydrite (Fe3⁺₁₀O₁₄(OH)₂·nH₂O, n < 1) and lepidocrocite (γ-FeOOH) have high affinities for PO₄³⁻ (approximately 2000 μmol g⁻¹ and 100–300 μmol g⁻¹, respectively^13–15), primarily due to their moderate to high surface areas (300–600 m2 g⁻¹ and 20–100 m2 g⁻¹, respectively¹⁶), which result from their poorly to moderately crystalline structures. In oxic soil and sediment environments, FeO_x minerals often form in the presence of PO₄³⁻, leading to the coprecipitation of PO₄³⁻ with FeO_x¹⁷. Unlike PO₄³⁻ adsorption onto preformed FeO_x, which does not alter its mineral structure^14,18, coprecipitation with PO₄³⁻ reduces the crystallinity of the resulting Fe(III) precipitate by interfering with crystal growth sites. In both cases, the interaction between FeO_x minerals and PO₄³⁻ leads to the transient immobilization of PO₄^3−19–21.

Although extensive research has been conducted on the role of FeO_x minerals in immobilizing PO₄³⁻ in oxic soils and sediments, little is known about the interactions between PO₄³⁻ and iron sulfide in freshwater sediments, particularly mackinawite (FeS), the most reactive iron mineral under anoxic conditions. FeS is the first iron sulfide phase that forms under the anoxic conditions owing to the rapid reaction between FeO_x minerals and sulfide ions produced by sulfate-reducing microorganisms^22–25. This mineral is a metastable iron monosulfide phase, often appearing as fine-grained, disordered particles (~ 5 nm in diameter as the primary precipitate) with a relatively high surface area (up to 350 m² g⁻¹)^23,26,27. The conversion of FeO_x to FeS under anoxic conditions involves a positive feedback loop, in which Fe2⁺ ions released during the reductive dissolution of FeO_x further promote FeO_x dissolution by electron transfer to the solid and, in turn, facilitate the subsequent formation of FeS². Given the high reactivity and surface area of mackinawite, a question remains as to whether iron sulfides can play a similar role as FeO_x minerals in binding or immobilizing PO₄³⁻ in sediments, potentially on much longer time scales. The majority of the research conducted so far, however, has focused on the indirect role of FeS in immobilizing PO₄³⁻, more specifically on its role as an iron source for the formation of vivianite (Fe₃(PO₄)₂·8H₂O)^28–30. Ma et al.³¹ for example demonstrated that nanoparticulate FeS can rapidly transform into vivianite under circumneutral conditions (pH 6–8) through FeS dissolution and PO₄³⁻-induced precipitation of ferrous iron. The rate of vivianite formation was governed by FeS dissolution kinetics, with PO₄³⁻ adsorption inhibiting FeS aging and thus favoring dissolution. These findings suggest that under low sulfide concentrations, FeS may act as an iron source for vivianite formation, indirectly promoting PO₄³⁻ preservation in anoxic sediments via redox-driven mineral transformations. Although not focused on FeS, experimental evidence by de Souza‐Barros et al.³² showed that pyrite (FeS₂) can adsorb PO₄³⁻ under acidic conditions, but this binding is reversible under higher pH or oxidizing environments. These findings suggest that PO₄³⁻ stabilization by pyrite in sediments is highly conditional and may be disrupted by changes in redox or pH conditions. In anoxic marine sediments, long-term PO₄³⁻ preservation is generally attributed to the formation of authigenic apatite minerals, which represent the dominant permanent sink for phosphorus and do not directly involve iron³³. Nevertheless, this pathway does not preclude additional or complementary mechanisms of PO₄³⁻ immobilization associated with reduced mineral phases, particularly during early diagenesis when redox-sensitive minerals actively regulate solute partitioning.

It was only recently that Ding et al.³⁴ demonstrated the importance of direct PO₄³⁻ interactions with FeS surfaces. They showed that even at a low concentration (0.3 mM), PO₄³⁻ strongly adsorb onto the surface of mackinawite nanoparticles, forming monodentate complexes and creating localized acidic microenvironments. This surface-bound PO₄³⁻ enhanced FeS-driven oxidation reactions, leading to the generation of hydroxyl radicals (•OH), which in turn promoted the degradation of organic compounds. Although explored in the context of contaminant breakdown, the findings reveal that FeS not only binds PO₄³⁻ but may also influence OC turnover through PO₄³⁻-mediated redox processes in sedimentary environments. Nevertheless, several important chemical and environmental questions regarding the FeS–PO₄³⁻ interactions in natural sediments remain to be addressed: (i) How does the affinity of FeS for binding PO₄³⁻ compare to that of FeO_x?, (ii) What is the impact of OC adsorbed on FeS surfaces on FeS–PO₄³⁻ interactions and immobilization?, and (iii) What is the effect of the FeS–PO₄³⁻ interactions on the release of PO₄³⁻ to the aqueous phase upon the onset of anoxic conditions in freshwater systems? Addressing these questions will help better understand the intertwined biogeochemical cycles of Fe, S, OC, and P in sediments, as well as the controls on PO₄³⁻-induced primary productivity in freshwater aqueous systems. We hypothesize that FeS plays a significant role in the cycle of PO₄³⁻ in freshwater sediments, and that extent of this influence should be assessed in detail to better constrain the cycle of this biologically essential element. To address the first question above, we prepared a series of FeS samples by reacting Fe2⁺ with HS⁻ in the absence and presence of varying concentrations of PO₄³⁻ with the aim of investigating the potential chemical and environmental implications of FeS–PO₄³⁻ interactions. Both coprecipitation and adsorption approaches were employed to simulate natural authigenic and detrital FeS formation pathways, respectively. We then carried out a similar PO₄³⁻ adsorption experiment in the presence of a constant concentration of OC to study the impact of FeS-sorbed OC on FeS–PO₄³⁻ interactions (question 2). Finally, we exposed FeS–PO₄³⁻ and FeS-OC-PO₄³⁻ particles to wet oxic conditions to determine the extent to which PO₄³⁻ is released to solution upon the oxidation of FeS (question 3). Crystallographic variations resulting from the incorporation of PO₄³⁻ into the FeS structure were monitored using powder X-ray diffraction (PXRD) and HRTEM-SAED analyses. The affinity and maximum theoretical capacity of FeS for PO₄³⁻ immobilization were determined through isotherms simulation.

Materials and methods

Preparation of FeS and FeS–PO₄³⁻

Pure FeS was prepared by continuously titrating 75 mL of 1 M HS⁻ (NaHS anhydrous; Sigma-Aldrich) with 75 mL of 0.4 M Fe2⁺ solution (FeCl₂·4H₂O; Supelco) both prepared using O₂-free Milli-Q water (Merck Millipore)³⁵. The resulting black suspension was continuously sparged with N₂ in an airtight amber bottle connected to a water trap under agitation for 4.5 h to prevent chemical and photochemical oxidation of the forming FeS. The pH of the suspension was adjusted to 7.5–8.0 using either 0.1 M NaOH or 0.1 M HCl, as needed. The resulting precipitate was then centrifuged at 35,890 rcf (15,000 rpm) for 10 min, followed by three rinses with O₂-free 0.7 M NaCl solution. Finally, the sample was freeze-dried for 48 h and kept under the vacuum for the subsequent experiments.

FeS–PO₄³⁻ samples were synthesized using two different sorption approaches to simulate the natural pathways of FeS–PO₄³⁻ formation. Co-precipitated samples—mimicking authigenically formed FeS–PO₄³⁻—were prepared by initially adding different concentrations of PO₄³⁻ (ranging from 2 × 10⁻5 to 1 × 10⁻¹ M, corresponding to a P/Fe molar ratio of 1 × 10⁻⁴ to 0.5, using Na₂HPO₄ salt; Sigma-Aldrich) added to the HS⁻ solution. For the adsorption approach—simulating detrital FeS–PO₄³⁻—PO₄³⁻ was added to pre-formed wet FeS within the same concentration range as the co-precipitated samples, 3 h after the start of FeS formation, and allowed to equilibrate for an additional 1.5 h. To investigate the effect of adsorbed OC on the immobilization of PO₄³⁻, the same aliquot of a phytoplankton-derived OC extract (Nannochloropsis; Reed Mariculture) was added to all samples 1.5 h after mixing the HS⁻ and Fe2⁺ solutions to achieve an OC/Fe molar ratio of 0.4, a ratio that theoretically corresponds to about half of the surface reactive sites of FeS particles occupied by sorbed OC, while the other half remains free to react with PO₄³⁻⁹. The resulting suspension was then agitated under N₂ purging for another 1.5 h before adding the PO₄³⁻ solution. The rest of the procedure followed that of the FeS–PO₄³⁻ experiment without OC addition. Finally, both FeS–PO₄³⁻ and FeS–OC–PO₄³⁻ samples underwent the same centrifugation/rinsing cycles and freeze-drying steps as described for the formation of pure FeS above. All the above-described syntheses were conducted in triplicate.

Preparation of natural organic carbon extract

As a source of natural marine OC, organic extracts were prepared from Nannochloropsis, a genus of microalgae commonly used as a phytoplankton model. The extraction involved multiple freeze–thaw cycles using liquid nitrogen, followed by centrifugation at 32,700 rcf to recover the natural OC supernatant. To remove residual particulates, the supernatant was vacuum-filtered using a glass fiber filter with a nominal pore size of 0.7 μm. The dissolved organic carbon (DOC) concentration of the extract was quantified using a TOC analyzer (model TOC-VCPH 5000A, Shimadzu Corporation)¹¹.

PO₄³⁻ and iron measurements

The PO₄³⁻ and total Fe (Fe_T) contents of the collected solids from the FeS–PO₄³⁻ and FeS–OC–PO₄³⁻ syntheses were measured using ICP-MS (Agilent 7900). To this end, approximately 100 mg of the resulting solids was first digested in 4 mL of concentrated HCl (Sigma-Aldrich), followed by dilution to 50 mL with ultrapure Milli-Q water. The digested samples (triplicates in each case) were then diluted tenfold to achieve a suitable concentration for P and Fe_T quantification by ICP-MS. A series of P and Fe standards, ranging from 1.6 × 10⁻⁷ to 3.2 × 10⁻⁴ M and 8.9 × 10⁻⁷ to 1.8 × 10⁻⁴ M, respectively, corresponding to 0.05 to 10 mg/L in both cases, were prepared prior to sample analysis to externally calibrate the instrument. The final phosphorus, expressed as PO₄³⁻ and iron concentrations, were calculated based on stoichiometric ratios and applied dilution factors.

Solid state organic carbon measurements

To quantify the OC content of the solids after the synthesis, samples were analyzed using an elemental analyzer (EuroVector EA3000). The instrument was calibrated using an in-house β-alanine standard and a certified sucrose standard (IAEA-C6), covering a mass range of 20 to 120 μg. To ensure accuracy and correct for potential instrumental drift in OC measurements, both standards were run between every set of six samples.

Oxidation incubations

Oxidation incubation experiments were conducted by weighing approximately 0.25 g of each synthesized sample into Oak Ridge Teflon tubes, followed by the addition of 20 mL of a 10 mM MOPS buffer to adjust the pH to ~ 7.5–8.0. These experiments were performed in triplicate. Prior to incubation, the samples were sparged with high-purity N₂ through the headspace for 2 h to establish equilibrium. The samples were then incubated at 4 °C, simulating the average bottom water temperature in large Canadian lakes in the winter, for 1 and 7 d, followed by centrifugation and rinsing with a 0.7 M NaCl solution. The combined supernatants were collected for PO₄³⁻ analysis by ICP–MS, while the remaining solids were freeze-dried and stored for subsequent analyses.

Citrate-bicarbonate-dithionite method

To quantitatively determine the FeO_x products formed upon the oxidation of FeS through incubation experiments, the citrate-bicarbonate-dithionite (CBD) method, modified by Lalonde et al.⁹, was employed. In this method, the reactive—poorly crystalline structures such as ferrihydrite and lepidocrocite— FeO_x fraction is reductively dissolved. Briefly, about 0.25 g of synthesized samples was transferred into centrifuge tubes (in triplicate). A 13 mL aliquot of a solution containing 0.11 M NaHCO₃ and 0.27 M citrate was added, and the suspension was preheated in a water bath at 80 °C. Subsequently, 0.25 g of sodium dithionite (Na₂S₂O₄) dissolved in 2 mL of the same NaHCO₃–citrate solution was introduced, and the mixture was agitated and kept at 80 °C for 15 min. Following centrifugation at 32,700 rcf for 10 min, the supernatant was collected, and the residual solids were rinsed three times with 15 mL of 0.7 M NaCl. The supernatants were then acidified and stored at 4 °C prior to analysis of total iron, [Fe]_t, and PO₄³⁻ concentrations by ICP–MS.

Crystallographic characterization

The crystallographic features of both pure FeS, FeS–PO₄³⁻ and FeS–OC–PO₄³⁻ species were examined using powder X-ray diffraction (PXRD). Diffraction patterns were recorded over a 2θ range of 10–80°, with a step size of 0.01° and an integration time of 16 s per step. Measurements were performed using Cu Kα radiation at 30 kV and 10 mA on a Bruker D2 Phaser diffractometer (Germany).

Crystallite size and surface area calculations

The average crystallite size of FeS phases was determined from PXRD peak broadening using the Scherrer equation (Eq. 1)²⁶. For each selected reflection, the full width at half maximum (FWHM) was obtained by identifying the local peak maximum within ± 0.7° of the expected 2θ position and interpolating the half-maximum crossing points.

1

where D is the crystallite size in nm, K is the shape factor (0.9), λ is the Cu Kα wavelength (1.5406 Å), β is the measured FWHM (radians), and θ is the Bragg angle. The specific surface area (SSA; m²/g) was then calculated from the crystallite size assuming spherical domains of density 4.2 g cm⁻3 (Eq. 2):

2

where ρ is the density and D the crystallite size obtained from Eq. (1).

Capacity and affinity of FeS for PO₄³⁻ immobilization

Adsorption isotherm models that best represent the experimental data offer insights into the affinity of the sorbent for the sorbate, the underlying sorption mechanism, and the nature of surface interactions. To evaluate model performance, experimental adsorption data were fitted to the Langmuir, Freundlich, and Redlich–Peterson isotherm models using Microsoft Excel. Nonlinear regression was applied for the Langmuir and Redlich–Peterson models, while the Freundlich model was linearized.

For the Langmuir and Redlich–Peterson models, predicted adsorption capacities were calculated using the respective nonlinear equations, and Excel’s Solver tool was employed to minimize the sum of squared errors (SSE) between observed and calculated values by adjusting the model parameters. Initial parameter estimates were manually entered, and constraints were applied where necessary to ensure convergence.

The Freundlich model was linearized by applying logarithmic transformation to both equilibrium concentration and adsorption capacity data. Linear regression was performed on the transformed data using Excel’s LINEST function to determine the slope and intercept, from which Freundlich constants were derived.

The Langmuir (Eq. 3) model assumes monolayer adsorption onto a surface with uniform binding sites, while the Freundlich (Eq. 4) model empirically describes adsorption on heterogeneous surfaces with a distribution of site energies and variable affinities, without assuming a finite adsorption capacity. The Redlich–Peterson (Eq. 5) isotherm incorporates features of both, effectively describing adsorption behavior across a broad concentration range by aligning with Langmuir behavior at low concentrations and with Freundlich characteristics at higher concentrations³⁶.

3

4

5

where q_e (μmol PO₄³⁻/g FeS) is the amount of sorbed PO₄³⁻ per gram of FeS, C_e (M) is the equilibrium concentration of PO₄³⁻, Q_max is the maximum sorption capacity of the adsorbate (μmol PO₄³⁻/g FeS), K_L (L/mol), K_F (L/g), and K_RP (L/g) are the Langmuir, Freundlich, and Redlich-Peterson constants, n is the adsorption intensity/surface heterogeneity constant, a and b are the Redlich–Peterson saturation-related constant and dimensionless exponent constant, respectively. The maximum buffer capacity of the adsorption for each material was calculated by multiplying the K_L and Q_max values obtained from Langmuir model. The free Gibbs energy (ΔG; kJ/mol) of the sorption processes was calculated through Eq. (5) as follows^13,37:

6

where R is the universal gas constant (− 8.314 J/(K mol)), T is the sorption temperature (298.15 K), and K⁰ is a dimensionless constant derived by multiplying K_L by pure water molarity (55.4 M). The dimensionless sorption equilibrium parameter called separation factor was calculated using the following Equation³⁷:

7

where K_L is the Langmuir’s constant (L/mol) and C₀ is the initial concertation of PO₄³⁻ (M). Based on the value of R_L, sorption behavior could be described as unfavorable (R_L > 1), linear (R_L = 1), favorable (0 < R_L < 1), and irreversible (R_L = 0).

Surface area measurement

In order to evaluate the impact of adsorbed OC on the preformed FeS particles on the adsorption behavior of PO₄³⁻, Brunauer–Emmett–Teller (BET) surface area analysis was conducted. Measurements were carried out using nitrogen gas adsorption at 77 K with a Micromeritics TriStar II Plus surface area and porosity analyzer to determine both the specific surface area and pore size distribution of the samples.

Morphological assessment

Transmission electron microscopy (TEM) was employed to investigate the morphological features of the synthesized FeS samples and to assess structural variations induced by the presence of OC and PO₄³⁻. Imaging was performed using a Talos™ L120C TEM (Thermo Fisher Scientific) operating at an accelerating voltage of 120 kV. For the analysis, a small aliquot of the FeS sample slurry was drop-cast onto a carbon-coated copper grid and allowed to dry under ambient conditions prior to imaging. This technique enabled direct visualization of particle shape, size, and aggregation state, providing insight into how OC and PO₄³⁻ incorporation may alter the morphology of FeS minerals.

Results and discussion

Crystallographic and microscopic characterization of the synthesized FeS samples

The formation of the synthesized phases and the potential structural differences between the pre- and post-incubated samples were investigated using PXRD (Fig. 1). The presence of low-intensity, broad diffraction peaks at 2θ ≈ 17°, 30°, 39°, and 50°, corresponding to the (001), (101), (110), and (200, 112) crystallographic planes, respectively, is characteristic of poorly crystalline mackinawite (FeS) across all samples³⁸ (Fig. 1; blue asterisks on diffractogram #1; JCPDS 15-0037). In the case of the FeS–PO₄³⁻-Ads. and FeS–OC–PO₄³⁻-Ads. samples, the diffractograms are nearly identical to that of pure FeS, indicating that the adsorption of OC and PO₄³⁻ does not significantly alter the bulk crystal structure of the pre-formed FeS particles at the macroscopic scale (Fig. 1; diffractograms #4 and 7). However, as discussed below, at the microscopic scale the presence of OC and PO₄³⁻, even when introduced after FeS formation, can alter the local crystallographic properties of FeS.

Fig. 1.

X-ray diffractograms of reference JCPDS cards (colored traces) and prepared samples (numbered traces) collected over the 2θ range of 10–70°. Blue: mackinawite; red: greigite; green: lepidocrocite; gold: vivianite; deep pink: pyrite; purple: sulfur. (1) FeS, pre-oxidation; (2) FeS after 1 d of oxidation; (3) FeS after 7 d of oxidation; (4) FeS–PO₄³⁻,–Ads., pre-oxidation; (5) FeS–PO₄³⁻–Ads. after 1 d of oxidation; (6) FeS–PO₄³⁻–Ads. after 7 d of oxidation; (7) FeS–OC–PO₄³⁻–Ads., pre-oxidation; (8) FeS–OC–PO₄³⁻–Ads. after 1 d of oxidation; (9) FeS–OC–PO₄³⁻–Ads. after 7 d of oxidation; (10) FeS–PO₄³⁻–Cop., pre-oxidation; (11) FeS–PO₄³⁻–Cop. after 1 d of oxidation; (12) FeS–PO₄³⁻–Cop. after 7 d of oxidation. Ads. and Cop. stand for adsorption and coprecipitation, respectively. The NaCl peaks arise from the presence of residual counterions trapped in the solid during its synthesis.

The emergence of four new diffraction peaks at 2θ ≈ 11°, 13°, 18°, and 20°, corresponding to the (110), (020), (200), and (–201) reflections, respectively, in the diffractogram of the FeS–PO₄³⁻–cop sample suggests the formation of vivianite³⁹ (Fe₃(PO₄)₂·8H₂O; JCPDS 75-1186; gold asterisks). Theoretically, the coprecipitation of vivianite with FeS may appear unusual, as the solubility of the latter is approximately one order of magnitude lower (~ 10^–9 M for FeS vs. ~ 10^–8 M for Fe₃(PO₄)₂·8H₂O)^22,24,40. Nevertheless, the solubility of the freshly formed FeS nanoparticles is much higher than initially thought (10⁻⁶ M vs. 10⁻⁹)^41,42. Accordingly, it is not surprising that the well-pronounced reflection peaks of Fe₃(PO₄)₂·8H₂O were observed, taking into account that the initial concentration of the PO₄³⁻ in this experiment was relatively high (3.75 × 10⁻² M, Table 1). This observation demonstrates that under sulfidic conditions with an excess of sulfide (Fe:S molar ratio of 1:1.5) and a high concentration of PO₄³⁻ (P:S molar ratio of 1:5), surface adsorption onto FeS and other minerals is probably not the sole mechanism for PO₄³⁻ capture in the sediments and vivianite formation also plays an important role in locking PO₄³⁻ in the sediments^28–30. Note that while the diffraction patterns observed here are clearly consistent with vivianite, peak broadening and overlap with FeS phases limit definitive phase identification. In future work, complementary spectroscopic confirmation approaches such as Mössbauer spectroscopy or synchrotron-based techniques should be used to confirm that the neoformed mineral is vivianite.

Table 1.

Partitioning of PO₄³⁻ between the solid and aqueous phases.

Sample	P/Fe (molar)	[PO43−]0 (M)	[PO43−]eq (M)	[PO43−]s (μmol g−1 FeS)
FeS–PO4-Ads	1 × 10−4	1.88 × 10−5	1.28 × 10−5	0.30 ± 0.01
	1 × 10–3	1.88 × 10−4	1.67 × 10−4	1.00 ± 0.01
	1 × 10−2	1.88 × 10−3	1.03 × 10−3	43.37 ± 0.10
	2 × 10−1	3.75 × 10−2	3.36 × 10−2	192.83 ± 0.28
	5 × 10−1	8.82 × 10−2	8.42 × 10−2	212.43 ± 0.18
FeS-OC-PO4-Ads	1 × 10−4	1.88 × 10−5	1.84 × 10−5	0.11 ± 0.01
	1 × 10−3	1.88 × 10−4	1.86 × 10−4	0.40 ± 0.02
	1 × 10−2	1.88 × 10−3	1.80 × 10−3	22.37 ± 0.07
	2 × 10−1	3.75 × 10−2	3.71 × 10−2	134.76 ± 0.10
	5 × 10−1	8.82 × 10−2	8.77 × 10−2	165.71 ± 0.31
FeS–PO4-Cop	1 × 10−4	2.0 × 10−5	1.10 × 10−5	0.32 ± 0.01
	1 × 10−3	2.0 × 10−4	1.28 × 10−4	2.64 ± 0.02
	1 × 10−2	2.0 × 10−3	6.96 × 10−4	46.47 ± 0.18
	2 × 10−1	3.75 × 10−2	3.04 × 10−2	238.39 ± 0.04
	5 × 10−1	9.30 × 10−2	8.45 × 10−2	283.19 ± 0.12

[PO₄³⁻]₀, [PO₄³⁻]_eq, and [PO₄³⁻]_s are the initial, equilibrium, and FeS-immobilized PO₄³⁻ concentrations, respectively. Ads. and Cop. stand for adsorption and coprecipitation.

High-angle annular dark-field (HAADF) imaging in STEM mode (HAADF-STEM) was used to provide Z-contrast images in which intensity scales with atomic number, enabling direct visualization of compositional heterogeneity and phase distribution at the nanoscale (Figures S1-S4). The HAADF image of pure FeS (Figure S1a) reveals extended, flake-like aggregates composed of thin, irregular domains with diffuse and relatively uniform Z-contrast, and no evidence of sharply defined crystalline facets or dense cores, consistent with an amorphous to nanocrystalline structure. Overlapped Fe and S elemental maps show nearly complete spatial co-localization across the entire aggregate, confirming intimate association of Fe and S at the nanoscale and ruling out the presence of Fe-rich or S-rich segregated domains, secondary Fe-bearing phases, or elemental sulfur accumulations. The individual Fe and S maps further demonstrate uniform elemental distributions that mirror the particle morphology observed in HAADF images. Together, these observations indicate that freshly precipitated FeS is structurally disordered and compositionally uniform, consistent with rapid, low-temperature precipitation under reducing conditions and in agreement with the HR-TEM and SAED results. HAADF–STEM imaging of FeS–PO₄³⁻–Ads. (Figure S2) shows aggregated, flake-like particles with diffuse and relatively uniform Z-contrast, closely resembling the morphology observed for pure FeS. The absence of sharply defined crystalline facets or dense internal domains indicates that PO₄³⁻ addition after FeS precipitation does not substantially alter the bulk structure or induce significant crystallization of the FeS particles. The combined Fe–S–P elemental map reveals strong spatial co-localization of Fe and S throughout the aggregate, confirming the preservation of the FeS framework following PO₄³⁻ addition. In contrast, phosphorus is present at much lower intensity and displays a sparse, discontinuous distribution. The individual Fe and S maps closely mirror the HAADF morphology and show uniform elemental distributions, whereas the P map exhibits scattered, isolated signals without evidence of concentrated P-rich domains or volumetric overlap with Fe. This spatial decoupling between P and the Fe–S matrix indicates that PO₄³⁻ is primarily surface-associated rather than structurally incorporated into the FeS lattice. The lack of P enrichment within particle interiors and the preservation of diffuse HAADF contrast together suggest that PO₄³⁻ adsorption occurs via surface complexation on pre-existing FeS particles, without promoting bulk crystallization or the formation of discrete Fe–P phases. These observations are consistent with the PXRD results, which show minimal changes in FeS crystallinity upon post-synthetic PO₄³⁻ adsorption.

HAADF–STEM imaging of the FeS–PO₄–Cop. sample (Figure S3) reveals compact, dense aggregates with markedly brighter and more heterogeneous Z-contrast compared to both pure FeS and FeS–PO₄–Ads. The presence of localized high-contrast domains suggests the formation of Fe-rich phases with increased mass–thickness and structural coherence, consistent with enhanced crystallinity induced during coprecipitation. Elemental mapping provides clear evidence for intimate association between Fe, S, and P at the nanoscale. The combined Fe–S–P map shows extensive spatial overlap of all three elements throughout the aggregate. Unlike the FeS–PO₄³⁻–Ads. sample—where P is sparse and discontinuous—the P map here exhibits strong, volumetric enrichment that closely follows the Fe and S distributions. This pervasive P–Fe–S co-localization indicates that PO₄³⁻ is not merely surface-bound but is structurally incorporated during particle formation. The individual Fe and S maps confirm preservation of the FeS framework, while the dense and continuous P signal may also suggest the presence of Fe–PO₄³⁻ domains intergrown with, or intimately associated with, FeS (as shown earlier in PXRD results). This spatial relationship is consistent with coprecipitation-controlled nucleation, where PO₄³⁻ participates directly in Fe(II) coordination during mineral growth. In combination with PXRD evidence for vivianite, the HAADF–STEM/EDS results support the formation of crystalline Fe(II)–PO₄³⁻ phases (vivianite-like domains) embedded within or attached to the FeS matrix.

HAADF–STEM imaging of the FeS–OC–PO₄³⁻–Ads. sample (Figure S4) shows aggregated, flake-like particles with diffuse and relatively uniform Z-contrast, indicating that sequential adsorption of OC followed by PO₄³⁻ does not lead to the formation of dense crystalline Fe–P domains or extensive bulk recrystallization of FeS. Nevertheless, consistent with the HR-TEM and SAED observations (see below), the presence of OC, even when introduced after FeS formation, appears to enhance the local structural ordering of FeS relative to both pure FeS and FeS–PO₄³⁻–Ads. samples. This OC-induced increase in apparent crystallinity may reflect kinetic attenuation of FeS surface restructuring through Fe(II)–OC complexation, which can stabilize surface motifs, reduce defect density, and facilitate rearrangement toward more ordered Fe–S domains. Such surface-mediated ordering provides a plausible explanation for the enhanced crystallinity observed in the OC-bearing system without invoking the formation of new crystalline Fe–P phases.

Elemental mapping reveals strong spatial co-localization of Fe and S throughout the aggregate, again confirming preservation of the FeS framework. Importantly, the phosphorus signal broadly overlaps the Fe–S footprint and is more homogeneously distributed than in the FeS–PO₄–Ads. sample, where PO₄³⁻ appeared sparse and discontinuous. This improved PO₄³⁻ dispersion could speculatively reflect the presence of P-bearing functional moieties (56.58 μmol PO₄³⁻.g⁻¹ OM) within the phytoplankton-derived organic carbon itself (e.g., organically bound PO₄³⁻ groups), which can act as intrinsic P reservoirs and promote nanoscale P retention within the FeS–OC assemblage. As a result, PO₄³⁻ in FeS–OC–PO₄³⁻–Ads. appears spatially associated with the composite FeS–OC assemblage through a combination of weak FeS surface interactions and association with the organic matrix, rather than through strong surface complexation or the formation of discrete crystalline Fe–PO₄³⁻ phases. This nanoscale co-distribution likely reflects physical proximity and OC-mediated dispersion—potentially involving P-bearing organic moieties—rather than enhanced binding affinity to FeS. Accordingly, the apparent incorporation-like P distribution observed by HAADF–EDS does not imply stronger FeS–PO₄³⁻ interactions but instead is consistent with reduced thermodynamic affinity (next section) and the absence of vivianite or other Fe–P crystalline phases in PXRD.

FeS sorbtion efficiency towards PO₄³⁻?

The amount of sorbed PO₄³⁻, hereafter referring to both adsorbed and coprecipitated species, was quantified by analyzing the solutions collected after the digestion of the synthesized samples (Table 1). The amount of non-sorbed PO₄³⁻ ([PO₄³⁻]ₐ_q × volume) remaining in solution following the sorption experiment was determined by mass-balance calculations as the difference between the initial amount of PO₄³⁻ in solution ([PO₄³⁻]₀ × volume) and the amount of PO₄³⁻ associated with the solid phase ([PO₄³⁻]ₛ × mass) after the adsorption and coprecipitation experiments. Moreover, the PO₄³⁻ content of phytoplankton-derived OC was measured by ICP-MS, and the amount of PO₄³⁻ associated with FeS–OC was corrected to account for internal PO₄³⁻ loading of the organic matter (OM; 56.58 μmol PO₄³⁻ g⁻¹ OM).

As shown in Table 1, in all cases an increase in the [PO₄³⁻]₀ leads to a corresponding increase in the sorbed PO₄³⁻ concentration ([PO₄³⁻]s). To quantify the sorption capacity and affinity of the synthesized FeS samples under different experimental conditions, the adsorption data were fitted using Langmuir, Freundlich, and Redlich–Peterson isotherm models (Fig. 2), and the corresponding regression parameters calculated (Table 2). Notably, for all systems investigated, the Langmuir and Redlich–Peterson models (Eq. 1 and 2, respectively) provided nearly identical fits to the experimental data, whereas the Freundlich (Eq. 3) model failed to adequately describe the observed sorption behavior (χ² values in Table 2). The Freundlich model shows relatively high R2 values (0.85–0.94), however R2 alone is not sufficient for assessing nonlinear isotherm fits because it reflects variance but it is insensitive to the magnitude and distribution of residuals. By contrast, χ2 directly evaluates the deviation between observed and predicted values across the full dataset and is more sensitive to systematic deviation from good fit. The higher χ2 values for the Freundlich model therefore indicate poorer agreement with the data, supporting the dominance of Langmuir-and and Redlich-Peterson-type behavior.

Fig. 2.

Experimental (black symbols) and fitted adsorption isotherms using the Langmuir (red solid line), Freundlich (blue dashed line), and Redlich–Peterson (RP; green dash–dot line) models for (a) FeS–PO₄³⁻–Ads., (b) FeS–OC–PO₄³⁻–Ads., (c) FeS–PO₄³⁻–Cop., and (d) combined experimental isotherms for all three systems. Ads. and Cop. stand for adsorption and coprecipitation, respectively.

Table 2.

Isotherm parameters for PO₄³⁻ sorption to FeS and FeS–OC structures.

	FeS–PO43−-Ads	FeS-OC-PO43−-Ads	FeS–PO43−-Cop
Langmuir fitting
KL (L mol−1)	219.00	63.77	214.69
Qmax (μmol g−1)	221.79	194.09	288.53
MBC (L kg−1)	48.57	11.86	61.94
ΔG (kJ mol−1)	− 23.15	− 20.04	− 23.11
R2adj	1.00	1.00	0.65
χ2	6.33	1.91	6.60
Freundlich fitting
Kf (L mol−1)	2592.66	2529.74	3324.48
1/n	0.80	0.92	0.77
R2adj	0.85	0.90	0.94
χ2	219.45	87.66	363.97
Redlich–Peterson
KRP (L mol−1)	54,698.2	14351.52	85004.43
a	224.50	64.36	219.63
b	0.96	0.94	0.89
R2adj	1.00	1.00	1.00
χ2	6.63 × 10−8	1.97	6.08

K_L, K_F, and K_RP are the Langmuir, Freundlich, and Redlich-Peterson constants, respectively. Q_max is the maximum sorption capacity, MBC is the maximum buffer capacity, ΔG is the Gibbs free energy of sorption, 1/n is the heterogeneity (adsorption strength) constant. The parameters a, b, and c correspond to the Redlich–Peterson constants. R²_Adj. and χ² denote the goodness-of-fit parameters. Ads. and Cop. stand for adsorption and coprecipitation, respectively.

The similarity between the Langmuir and Redlich–Peterson fits arises from their shared ability to describe adsorption processes that approach finite saturation at high solute concentrations^43,44. The Langmuir model assumes monolayer adsorption onto a surface with a finite number of energetically equivalent sorption sites and no lateral interactions between adsorbed species⁴⁵. This framework is particularly appropriate for PO₄³⁻ binding to FeS, where sorption is dominated by specific surface complexation at discrete Fe(II) sites, ultimately leading to site saturation as surface coordination environments become fully occupied³⁴.

The Redlich–Peterson model, which is a three-parameter hybrid between the Langmuir and Freundlich models, retains the capacity to account for slight surface heterogeneity while still predicting saturation behavior at high concentrations⁴⁶. The Redlich–Peterson exponent approaches unity, effectively reducing the model to Langmuir-type behavior⁴³. This convergence indicates that PO₄³⁻ sorption onto FeS is largely governed by a finite number of high-affinity sites and that deviations from ideal surface homogeneity are minor. Consequently, both models capture the same underlying adsorption mechanism and therefore yield nearly superimposed fits. In contrast, the Freundlich model assumes adsorption on a highly heterogeneous surface with a continuous distribution of site energies and, critically, does not predict a maximum sorption capacity⁴³. Instead, it implies that sorption increases indefinitely with increasing solute concentration. This assumption is inconsistent with the experimental observations, which clearly show a tendency toward saturation at elevated PO₄³⁻ concentrations. The poor performance of the Freundlich model therefore reflects its inability to represent the site-limited nature of PO₄³⁻ uptake by FeS. In systems such as FeS, where sorption is controlled by specific, stoichiometrically constrained Fe–O–P surface complexes³⁴ and possibly surface precipitation at higher loadings, the absence of an explicit saturation term leads to systematic deviations between Freundlich predictions and the measured data. The mismatch between the empirical, non-saturating nature of the Freundlich model and the mechanistically constrained, site-specific behavior of phosphate sorption on FeS explains both its poorer statistical performance (e.g., elevated χ2) and its lack of physical interpretability in this system.

According to the Langmuir model fitting results, the calculated Q_max values for the three different species are in the following order: FeS–PO₄³⁻-Cop. (288.53 μmol PO₄³⁻/g FeS) > FeS–PO₄³⁻-Ads. (221.79 μmol PO₄³⁻/g FeS) > FeS-OC-PO₄³⁻-Ads. (194.09 μmol PO₄³⁻/g FeS). Although these values are much smaller than for ferrihydrite¹⁴ (2000 μmol PO₄³⁻/g Fh), they are still comparable to those of more crystalline FeO_x minerals such as lepidocrocite^13,15 (100–300 μmol PO₄³⁻/g Lp). This comparison highlights that FeS can act as a quantitatively important PO₄³⁻ sink, particularly in anoxic and sulfidic environments where Fe(III) oxides are thermodynamically unstable⁴⁷. The ability of FeS to immobilize PO₄³⁻ at levels comparable to crystalline Fe(III) oxides underscores its previously underappreciated role in regulating phosphorus availability. Moreover, because FeS is abundant and persistent in reduced sediments⁴⁸, its cumulative contribution to PO₄³⁻ retention may be substantial, even if its per-mass sorption capacity is lower than that of amorphous FeO_x minerals. Therefore, this observation suggests that FeS represents an important sink for PO₄³⁻ under reducing conditions, with direct implications for phosphorus burial, nutrient cycling, and redox-controlled phosphorus release in natural sediments.

Comparing the calculated Langmuir affinity constants (K_L; Table 2) shows that the presence of OC markedly decreases the PO₄³⁻ sorption affinity of FeS. Consistent with this trend, the Gibbs free energy (ΔG) values for FeS–OC–PO₄³⁻–Ads. are less negative than those obtained for FeS–PO₄³⁻–Ads. and FeS–PO₄³⁻–Cop., indicating that PO₄³⁻ binding is thermodynamically less favorable in the presence of OC. Together, these results demonstrate that the interaction between FeS–OC and PO₄³⁻ is weaker than that observed for bare FeS.

This interpretation is further supported by the maximum buffer capacity (MBC) values (Table 2), which quantify the ability of the solid phase to regulate aqueous PO₄³⁻ concentrations by integrating both sorption affinity and sorption capacity⁴⁹. Among the three systems investigated, FeS–OC–PO₄³⁻–Ads. exhibits the lowest MBC, indicating a diminished capacity to buffer changes in porewater PO₄³⁻ concentrations and highlighting the attenuating effect of OC on PO₄³⁻ retention by FeS. In addition, across the entire concentration range, the dimensionless separation factor (R_L) is consistently higher for FeS–OC–PO₄³⁻–Ads. compared to the other samples (Fig. 3a). Because R_L values approaching unity indicate less favorable adsorption, these results further corroborate the reduced affinity of PO₄³⁻ for OC-coated FeS surfaces. A plausible mechanistic explanation for the inhibitory effect of OC on PO₄³⁻ sorption involves electrostatic interactions and surface charge modification. Bare FeS has a potential of zero charge (PZC) of ~ 7.5; thus, under the experimental pH conditions (≈8), its surface is expected to be slightly negatively charged but still relatively close to electrostatic neutrality, allowing access of PO₄³⁻ to reactive Fe(II) surface sites. In contrast, organic carbon (OC), particularly of biological origin, contains abundant acidic functional groups such as carboxylic moieties (pKₐ ≈ 4–5) and amino acid residues with average isoelectric points (pI ≈ 6). At pH ≈8, both carboxyl groups and amino acid functionalities are predominantly deprotonated, imparting a net negative charge to the OC matrix. When OC is associated with FeS, it effectively shifts the surface away from near-neutral conditions toward a more negatively charged interface. This OC-induced surface charge modification enhances electrostatic repulsion between the FeS–OC composite and PO₄³⁻, which also exists as a negatively charged species under these conditions. As a result, PO₄³⁻ access to reactive Fe(II) sites is hindered, and sorption domains become partially blocked or energetically unfavorable. In addition to electrostatic effects, the physical coverage of FeS surfaces by OC further reduces the availability of active sites. Collectively, these mechanisms lead to a decrease in both PO₄³⁻ sorption affinity and maximum buffering capacity in the FeS–OC system⁵⁰.

Fig. 3.

(A) R_L plots for FeS–PO₄³⁻-Ads. (black square-line), FeS-OC-PO₄³⁻-Ads. (red circle-line), and FeS–PO₄³⁻-Cop. (blue triangle-line) over the [PO₄³⁻] range from 0.00 to 0.10 M. (B) Post-oxidation [PO₄³⁻] and [Fe]_t concentration after 1 and 7 days (d) of incubation. Ads. and Cop. stand for adsorption and coprecipitation, respectively.

The value of 1/n obtained from the Freundlich model reflects the adsorption strength, where smaller values indicate more favorable adsorption⁵¹. As 1/n approaches zero, adsorption becomes increasingly favorable and suggests a higher degree of surface heterogeneity⁵². Accordingly, the 1/n values of the different species in this study varied in the following order: FeS–PO₄-³⁻Cop. (0.77) < FeS–PO₄³⁻-Ads. (0.80) < FeS-OC-PO₄³⁻ (0.92). These results suggest that bare FeS has a more heterogenous surface compared to FeS-OC complex which, in turn, results in a more effective interaction between the adsorbent materials and PO₄³⁻. To support this interpretation, the surface areas of the FeS and FeS–OC adsorbents were quantified using both BET and PXRD-based approaches. BET measurements yielded surface areas of 1.35 and 0.75 m2 g⁻¹ for FeS and FeS–OC, respectively, indicating that the FeS–OC sample exhibits an approximately 45% lower surface area than bare FeS. This decrease is consistent with the lower adsorption observed for FeS–OC toward PO₄³⁻. In parallel, the SSAs were also estimated from PXRD data by calculating the SSA associated with individual mackinawite crystallographic reflections ((001), (101), (110), and (200/112)) using Bragg- and Scherrer-based equations ⁵³ (Eqs. 1 and 2). This approach yielded average SSAs of 327.40 and 263.36 m2 g⁻¹ for FeS and FeS–OC, respectively, again indicating a lower surface area for the OC-associated FeS and reinforcing its reduced adsorption affinity for PO₄³⁻ relative to bare FeS. Notably, the PXRD-derived SSA values are approximately two orders of magnitude higher than those obtained from BET analysis. This discrepancy arises from the fundamentally different physical states and surface accessibility probed by the two techniques. The BET measurements reflect the surface area of freeze-dried and aggregated solids, in which extensive platelet stacking and flocculation during drying effectively seal off a large fraction of reactive faces from gas adsorption. In contrast, the adsorption and coprecipitation experiments conducted in this study were performed using wet FeS suspensions, where a substantially larger proportion of particle surfaces remains exposed to the aqueous phase. Under these conditions, the XRD-derived SSA—representative of the nanometer-scale crystallite dimensions of freshly precipitated, poorly ordered mackinawite—provides a more realistic estimate of the reactive surface accessible to dissolved OC and PO₄³⁻. Similar discrepancies between BET- and XRD-derived surface areas have been widely reported²⁶, with BET measurements systematically underestimating accessible surface area due to drying-induced aggregation, while PXRD-based approaches capture the intrinsic nanoscale dimensions of disordered FeS by reflecting the external surface area of coherent crystalline domains. Moreover, the sulfur-rich synthesis conditions employed here (0.4 M FeCl₂ and 1.0 M NaHS) likely promoted rapid nucleation followed by aggregation and stacking of platy nanocrystals, further enhancing flocculation and reducing gas-accessible surface area in dried samples. Taken together, these observations indicate that while BET-derived surface areas provide useful information on the aggregated state of dried FeS, the PXRD-derived SSA is a more appropriate descriptor of surface reactivity under the aqueous conditions relevant to the present adsorption experiments since it is more representative of the effective reactive surface area of wet FeS suspensions, as found in natural systems.

These findings demonstrate that in natural anoxic and sulfidic freshwater sediments, where FeS commonly coexists with organic-rich matrices, FeS can act as an effective sink for PO₄³⁻, thereby limiting porewater PO₄³⁻ mobility and suppressing diffusive fluxes back to the overlying water column. The presence of OC coatings moderately attenuates this PO₄³⁻ retention capacity, yet does not eliminate the ability of FeS to immobilize PO₄³⁻ under reducing conditions. As a result, FeS remains a quantitatively important buffer of porewater PO₄³⁻, even in OC-rich sediments, although OC-mediated surface passivation may partially modulate its buffering efficiency. This interplay highlights FeS as a stabilizing control on sedimentary phosphorus cycling, while emphasizing that organic matter influences the magnitude, rather than the existence, of this retention pathway.

The release and recapture of FeS-sorbed PO₄³⁻ under oxic conditions

To simulate transient oxygenation of anoxic sediments and assess its potential effects on FeS and the fate of immobilized PO₄³⁻, a series of oxic incubation experiments were conducted on the three types of synthesized samples ( FeS–PO₄³⁻-Ads., FeS–PO₄³⁻-Cop. and FeS–OC–PO₄³⁻–Ads.). In all cases, following 1 and 7 day exposure to oxic conditions, the [Fe]ₜ and [PO₄³⁻] released to the aqueous phase were undetectable, showing that there was no significant mobilization of PO₄³⁻ from the solids. This result suggests that the FeS oxidation products efficiently scavenged the PO₄³⁻ in-situ before it was released to the water column. We used PXRD to characterize the FeS oxidation products. As shown in Fig. 1, even after one day of oxidation, FeS (trace 2 in Fig. 1) exhibits new diffraction peaks centered at 2θ ≈ 15°, 25.5°, and 30°. The slight shifts in peak positions are attributed to the concurrent formation of multiple oxidation products during FeS oxidation, resulting in overlapping and partially superimposed reflections. These peaks can be assigned to a mixture of greigite, with reflections corresponding to the (111), (220), and (311) planes³⁸ (red asterisks in Fig. 1; JCPDS 16-0713), and lepidocrocite, with reflections corresponding to the (020) and (110/103) planes¹¹ (green asterisks in Fig. 1; JCPDS 44-1415). After 7 days of oxidation, these reflections become more pronounced and increase in intensity, indicating progressive growth and/or enhanced crystallinity of the oxidation products over time. In contrast to bare FeS, oxidation of the FeS–OC–PO₄³⁻–Ads., FeS–PO₄³⁻–Ads., and FeS–PO₄³⁻–Cop. systems appear to be noticeably retarded, with only minor changes observed following 7 days of incubation. The diffraction peaks associated with greigite and lepidocrocite remain substantially less pronounced than those observed for bare FeS, indicating suppressed formation and/or growth of these oxidation products. Instead, in the FeS–OC–PO₄³⁻–Ads. and FeS–PO₄³⁻–Ads. samples, new reflections attributable to vivianite begin to emerge, while in the FeS–PO₄³⁻–Cop. system these vivianite peaks become markedly more intense (gold asterisks in Fig. 1). This trend suggests that, in the presence of sorbed PO₄³⁻, iron is preferentially incorporated into iron–PO₄³⁻ phases, such as vivianite, rather than following oxidation pathways leading to the formation of greigite and lepidocrocite. FeS would then not only act as a Fe(II) reservoir for the kinetically slow formation of vivianite^28–30, but also for PO₄³⁻, suggesting a role as a PO₄³⁻ shuttle for FeS, analogous to the role of reactive Fe hydroxides as an OC shuttle during its transfer from oxic to anoxic sediment layers, where it is more efficiently preserved.

To quantify the amount of PO₄³⁻ associated with the oxidation products and the extent of Fe oxidation, we applied the citrate-bicarbonate-dithionite (CBD) extraction method to the residual solids pre- and post-oxidation. This method selectively dissolves the reactive Fe(III)-bearing phases (FeO_x) and associated compounds while remaining largely inert toward Fe(II)-bearing minerals such as FeS, greigite or vivianite, which in our case allows measuring the extent of FeS oxidation and the partitioning between PO₄³⁻ associated to FeS particles and PO₄³⁻ associated with the newformed FeO_x particles. The supernatants were collected and analyzed for [Fe]_t and [PO₄³⁻] (Fig. 3b). Irrespective of oxidation time, the [Fe]ₜ follows the order: FeS–OC–PO₄³⁻–Ads. < FeS–PO₄³⁻–Cop. < FeS–PO₄³⁻–Ads. A one-way (single-factor) analysis of variance (ANOVA) confirmed that these differences are statistically significant (p < 0.05), reflecting the influence of system type as the sole independent variable (Table S2-S5).

This trend indicates a markedly lower extent of FeS oxidation in the FeS–OC–PO₄³⁻–Ads. system compared to the other treatments. Such behavior can be attributed to the combined effects of OC shielding and reduced surface accessibility. Specifically, surface complexation between Fe(II) ions and oxygen-containing functional groups within the OC matrix (e.g., carboxylate, phenolic, and hydroxyl groups) likely promotes the formation of inner-sphere Fe(II)–OC complexes. These interactions sterically and electronically shield reactive Fe(II) sites, limiting their accessibility to dissolved oxygen and thereby suppressing oxidation kinetics over short timescales⁵⁴. This effect is further reinforced by the lower specific surface area of the FeS–OC material relative to bare FeS. In the case of FeS–PO₄³⁻–Cop., the lower total iron concentration ([Fe]ₜ) compared to FeS–PO₄³⁻–Ads. can be attributed to vivianite formation, which increases the overall crystallinity of the system and stabilizes Fe(II), thereby enhancing its resistance to oxidation. However, the trend in released PO₄³⁻ differs from that of [Fe]ₜ, following the order: FeS–OC–PO₄³⁻–Ads. < FeS–PO₄³⁻–Ads. < FeS–PO₄³⁻–Cop. The relatively low PO₄³⁻ concentration in the FeS–OC–PO₄³⁻–Ads. system is consistent with the previously discussed effects of organic carbon, including surface site shielding and reduced FeS reactivity, which limit both PO₄³⁻ loading and its subsequent release. In contrast, although FeS–PO₄³⁻–Cop. exhibits a lower extent of FeS oxidation, it shows the highest PO₄³⁻ release. This behavior likely reflects the intrinsically higher phosphate content immobilized in this system, as well as the formation of metastable Fe–P phases during coprecipitation. These phases are more susceptible to dissolution under CBD extraction, thereby contributing to the enhanced PO₄³⁻ release despite comparatively lower Fe mobilization. To support this hypothesis, HR-TEM and SAED were employed to evaluate the structural organization and crystallinity of the synthesized FeS samples under different chemical conditions (Fig. 4). The combined analyses reveal a progressive increase in structural order from pure FeS to FeS–OC and, finally, to FeS–PO₄–Cop., highlighting the distinct roles of organic carbon and PO₄³⁻ in governing FeS crystallization.

Fig. 4.

HR-TEM images of (A) pure FeS., (B) FeS-OC-PO₄³⁻-Ads., and (C) FeS–PO₄³⁻-Cop. SAED patterns of (D) FeS–PO₄³⁻-Ads., (E) FeS-OC-PO₄³⁻-Ads., and (F) FeS–PO₄³⁻-Cop. Ads. and Cop. stand for adsorption and coprecipitation, respectively.

Freshly precipitated FeS (Fig. 4a and d) exhibits a predominantly amorphous to poorly ordered structure. HR-TEM images show diffuse, poorly defined aggregates lacking continuous lattice fringes over extended domains. Instead, contrast variations are smooth and irregular, indicating only short-range order²⁶. Any lattice fringes that can be discerned are localized, discontinuous, and weak, suggesting the presence of small nanocrystalline or proto-crystalline mackinawite domains embedded within an amorphous matrix. Consistent with these observations, the corresponding SAED pattern is dominated by a broad diffuse halo rather than discrete diffraction rings or spots, confirming the absence of long-range crystallographic order. The lack of sharp diffraction features indicates extremely small coherent scattering domains, in agreement with the pronounced peak broadening observed in PXRD, and reflects rapid, low-temperature precipitation under reducing conditions.

In contrast, the FeS–OC sample (Fig. 4b and e) displays a markedly higher degree of structural ordering. HR-TEM images reveal well-defined, parallel lattice fringes extending over several nanometers, indicative of nanocrystalline FeS domains. Although these fringes are locally aligned, their orientations vary across the sample, suggesting the presence of multiple nanocrystallites with preferred local ordering rather than a single crystalline phase. The particles also exhibit elongated, plate-like or ribbon-like morphologies, consistent with reduced nucleation rate, directional growth or oriented attachment processes promoted by OC. Correspondingly, the SAED pattern shows distinct concentric diffraction rings, confirming enhanced crystallinity relative to pure FeS. While the rings remain broadened, reflecting nanocrystalline dimensions, their clear definition indicates a transition from amorphous FeS toward a more ordered, yet still polycrystalline, structure. Although more work is needed to confirm this interpretation, these observations suggest that adsorbed OC acts as a structure-modifying and stabilizing agent³⁵ by binding to Fe(II) surface sites, slowing rapid precipitation, and facilitating partial reorganization of Fe–S motifs into ordered nanocrystalline domains.

The highest degree of crystallinity is observed in the FeS–PO₄³⁻–Cop. sample (Fig. 4c and f). HR-TEM images show the most pronounced and continuous lattice fringes among all three systems, with fringes extending over larger domains and exhibiting well-defined angular relationships, likely introduced through vivianite formation. The particles display more faceted and anisotropic morphologies, indicative of crystal growth under stronger structural control. Consistent with these observations, the SAED pattern exhibits sharp, well-resolved diffraction rings with higher intensity and contrast than those observed for FeS–OC. The narrower ring widths indicate larger coherent scattering domains and improved long-range order, characteristic of a polycrystalline but well-ordered structure. These features reflect the strong crystallinity-enhancing effect of coprecipitated PO₄³⁻, likely arising from its incorporation during FeS nucleation, a reduction in nucleation rate, and stabilization of Fe(II) coordination environments.

Geochemical implications

This study shows that FeS can play a much more important role in sedimentary phosphorus cycling than is commonly assumed. Our results demonstrate that FeS is capable of immobilizing PO₄³⁻ through both surface adsorption and coprecipitation pathways, with sorption capacities comparable to those of more crystalline Fe(III) (oxyhydr)oxides such as lepidocrocite. In strongly reducing environments such as below the oxic/anoxic transition in sediments or in late summer when stratified bottom waters become anoxic, where reactive FeO_x phases are no longer stable, FeS therefore represents a viable and potentially widespread sink for PO₄³⁻, hence limiting its remobilization and potential re-diffusion to the water column (Fig. 5).

Fig. 5.

Conceptual model of iron–sulfur–phosphorus cycling across the oxic–anoxic sediment interface, highlighting the role of FeS in PO₄³⁻ immobilization. In oxic sediments, Fe²⁺ is oxidized to Fe(III) (oxyhydr)oxides (FeOₓ) that adsorb PO₄³⁻ and OC. Upon burial, reductive dissolution (IRB) releases Fe²⁺ and PO₄³⁻, while sulfate reduction (SRB) produces S²⁻, promoting FeS formation. PO₄³⁻ is then immobilized via adsorption or coprecipitation (e.g., vivianite), and FeS may further transform to FeS₂. OC influences these processes by modifying FeS reactivity and PO₄³⁻ binding.

Under conditions of elevated PO₄³⁻ availability, FeS not only acts as a reactive Fe(II) source for vivianite^28–31, but it could also constitute a PO₄³⁻ shuttle that sustains the kinetically slow formation of this mineral, providing an efficient mechanism for longer-term phosphorus stabilization in anoxic sediments.

At the same time, our findings clearly show that PO₄³⁻ retention by FeS is not an intrinsic or fixed property but is strongly controlled by the surrounding chemical environment. OC moderately reduces the affinity and buffering capacity of FeS for PO₄³⁻, primarily by decreasing accessible surface area and limiting PO₄³⁻ access to reactive Fe(II) sites. In organic-rich freshwater sediments, this attenuation may lead to transiently elevated porewater PO₄³⁻ concentrations under reducing conditions. Importantly, however, these results also indicate that internal phosphorus recycling in eutrophic systems is not necessarily self-sustaining during redox oscillations. When external PO₄³⁻ inputs are reduced, FeS-mediated retention, together with Fe redox cycling and secondary mineral formation, can progressively re-sequester phosphorus within sediments, facilitating system recovery over time. Thus, FeS acts not only as a regulator of short-term PO₄³⁻ mobility but also as a stabilizing control that limits runaway internal phosphorus loading in anoxic freshwater systems.

Finally, the oxidation experiment suggests that short-term oxygen exposure does not necessarily result in a net release of PO₄³⁻ to the water column. Instead, PO₄³⁻ can be efficiently retained in situ through rapid scavenging by newly formed iron oxidation products and/or preferential partitioning into iron–PO₄³⁻ phases such as vivianite. Taken together, these results highlight that the role of FeS in regulating phosphorus mobility depends on the interplay between redox conditions, organic matter associations, and iron mineral transformations. Rather than acting as a simple sink or source, FeS emerges as a dynamic component of sedimentary phosphorus cycling, with important implications for nutrient retention and long-term phosphorus burial in both freshwater and marine environments. Our results highlight the need for a reassessment of the cycle of phosphorus in freshwater environments, particularly in organic-rich sediments undergoing regular redox oscillations, representing a high proportion of lakes and other freshwater bodies located in forested areas.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

M.E.: Original ideal, experimental design, analyses, writing of the first draft. Y.G.: Supervision, experimental design, editing, funding.

Data availability

All the data is available in the paper and Supplementary section.

Declarations

Competing interests

The authors declare no competing interests.