The collected pine bark was first washed thoroughly with deionized (DI) water to remove surface debris, dried in an oven at 80 °C for 24 h, and then ground using a laboratory mill. The resulting powder was sieved to obtain a uniform particle size fraction (40 mesh). For pyrolysis, 50 g of the prepared feedstock was placed in a ceramic boat and loaded into a horizontal tube furnace (Model OTF-1200X; MTI Kejing Materials Technology Co., Ltd., Hefei, Anhui, China). The furnace was purged with high-purity nitrogen (N₂) gas at a flow rate of 200 mL min^-1 for 30 min to establish an anaerobic environment. Pyrolysis was then conducted by heating the feedstock to a final temperature of 600 °C at a ramp rate of 10 °C min^-1, and this temperature was held for 2 h. The 600 °C pyrolysis temperature was selected based on preliminary studies to optimize the balance between high specific surface area and the formation of a stable, graphitic carbon structure [17]. After cooling to room temperature under N₂ flow, the resulting biochar (designated BC) was collected, ground into a fine powder, and stored in a desiccator.

The magnetic composites were synthesized via an in-situ liquid-phase reduction method. In a typical synthesis, 2.0 g of the prepared BC powder was dispersed in a 300 mL solution of ethanol (99.7%; Aladdin Biochemical Technology Co., Ltd., Shanghai, China) and de-aerated deionized water (1:1 v/v) and sonicated for 30 min to ensure a uniform suspension. A specific mass of FeCl₃$\cdot$6H₂O (Aladdin Biochemical Technology Co., Ltd., Shanghai, China) was then dissolved into the suspension, which was stirred mechanically for 1 h under a continuous N₂ purge. This step facilitates the pre-loading of Fe(III) ions onto the biochar surface via ion-exchange and electrostatic interactions with surface functional groups. Following this, a 0.5 M NaBH₄ (Macklin Biochemical Co., Ltd., Shanghai, China) solution (prepared with a stoichiometric excess) was added dropwise into the suspension at a rate of 5 mL min^-1 using a peristaltic pump (Model BT100-2J; Longer Precision Pump Co., Ltd., Baoding, Hebei, China), which induced the rapid reduction of Fe(III) to Fe(0) nanoparticles [18]. The reaction is represented by Eqn. 1:

(1)$$4\mathrm{F}{\mathrm{e}}^{3+}+3\mathrm{B}{\mathrm{H}}_4^{-}+9{\mathrm{H}}_2\mathrm{O}\to 4\mathrm{F}{\mathrm{e}}^0\downarrow +3{\mathrm{H}}_2{\mathrm{BO}}_3^{-}+12{\mathrm{H}}^{+}+6{\mathrm{H}}_2$$

The appearance of a black precipitate and vigorous hydrogen bubbling indicated the formation of nZVI. The suspension was aged under stirring for an additional 1 h to ensure complete reaction. Finally, the black composite material was separated from the solution, washed repeatedly with de-aerated ethanol and water to remove residual ions, and dried in a vacuum oven at 60 °C for 12 h.

To effectively evaluate the synergistic effects within the composite, three control materials were prepared. (1) Pristine BC: This is the pyrolyzed pine bark as described in Section 2.2. (2) Bare nZVI: This material was synthesized using the exact liquid-phase reduction method described in Section 2.3, but without the addition of the biochar support. (3) Two different composite loading ratios were prepared and designated: MBC-10, synthesized using a 1:10 mass ratio of Fe (from the FeCl₃$\cdot$6H₂O precursor) to BC; and MBC-40, synthesized using a 4:10 mass ratio of Fe to BC. The MBC-40 composite, representing a high-load material, was the primary focus of subsequent performance and mechanistic studies unless otherwise noted.

Batch experiments were conducted to evaluate the arsenic immobilization performance of the materials under controlled aqueous conditions. Kinetic studies were performed by adding 20 mg of sorbent (e.g., MBC-40) to 50 mL of As(III) or As(V) solution (initial concentration C₀ = 10 mg/L) in 100 mL conical flasks. The flasks were sealed and agitated on a rotary shaker at 150 rpm and 298 K, with the solution pH maintained at 7.0 $\pm$ 0.1. Unless otherwise specified, arsenic test solutions were prepared in DI water without added dissolved organic matter (DOM)/natural organic matter (NOM), and no deliberate geochemical ageing or long-term passivation protocol (e.g., prolonged oxic/anoxic pre-exposure in groundwater-like media) was applied before sorption testing; therefore, the results primarily reflect short-term performance in simplified aqueous matrices rather than aged field conditions. Aliquots were withdrawn at specified time intervals (0, 5, 10, 30, 60, 120, 180, 240 min), filtered through 0.22 µm syringe filters, and the residual arsenic concentration (C_t) was determined. The sorption capacity at time t (q_t, mg g^-1) was calculated. The data were fitted to non-linear pseudo-first-order (PFO) and pseudo-second-order (PSO) kinetic models.

Adsorption isotherm experiments were conducted by adding 20 mg of each sorbent (BC, bare nZVI, MBC-10, MBC-40) to a series of 50 mL solutions with varying initial As(III) and As(V) concentrations (C₀ = 5, 10, 25, 50, 100, 150, 200 mg L^-1). The suspensions were agitated for 24 h at 298 K to ensure equilibrium was reached. The equilibrium sorption capacity (q_e, mg g^-1) was calculated, and the data were fitted to the non-linear Langmuir and the Freundlich isotherm models. Unless otherwise stated, all batch sorption experiments (kinetics, isotherms, pH, competing anions, and reusability tests) were conducted in triplicate (n = 3), and results are reported as mean $\pm$ standard deviation (SD). Where error bars are shown in figures, they represent $\pm$ 1 SD calculated from the independent replicate measurements.

The effect of initial solution pH on arsenic removal was investigated over a range of 2.0 to 12.0. The pH was adjusted using 0.1 M HCl or 0.1 M NaOH. The influence of common co-existing anions found in mining effluents (Cl^–, SO₄^2-, CO₃^2-, and HPO₄^2-) was tested by adding 100 mg L^-1 of their respective sodium salts to the 10 mg L^-1 As(V) solution. Reusability of the MBC-40 composite was assessed over five consecutive adsorption-desorption cycles. After each sorption cycle, the spent composite was separated using an external magnet, washed, and then desorbed by agitation in 50 mL of 0.1 M NaOH solution for 2 h. The regenerated material was then washed with DI water until neutral and reused for the next cycle.

To assess the composite’s practical applicability in a mining-relevant inorganic matrix, a SMW was prepared based on typical effluent compositions [19]. The SMW formulation represents a deliberately complex mixed-ion/multi-metal challenge (high sulfate and co-occurring Pb²⁺/Cd²⁺/Zn²⁺), but it does not include DOM/NOM and it does not simulate time-dependent geochemical ageing/passivation processes; these factors may influence long-term performance in natural systems and are noted as limitations and priorities for future field-oriented evaluation. The SMW contained a multi-pollutant mixture: 3.0 mg L^-1 As(V), 2.0 mg L^-1 As(III), 10.0 mg L^-1 Pb²⁺, 5.0 mg L^-1 Cd²⁺, 8.0 mg L^-1 Zn²⁺, and 1000 mg L^-1 SO₄^2-, at a pH of 5.5. The removal efficiency of MBC-40 (dose = 0.4 g L^-1) was tested in this complex matrix. Final concentrations of all contaminants were measured using an atomic fluorescence spectrometer (AFS) (Model AFS-9700, Beijing Haiguang Instrument Co., Ltd., Beijing, China) for arsenic and an Inductively Coupled Plasma-Optical Emission Spectrometer (ICP-OES) (Model Avio 200, PerkinElmer Inc., Waltham, MA, USA) for heavy metals.

The morphology of the BC and the synthesized composites was first investigated using scanning electron microscopy (SEM), as shown in Fig. 1. Fig. 1A reveals that the pristine BC, derived from pine bark, retained the natural, fibrous, and channel-like structure of its biomass precursor, exhibiting a porous and rough surface [20]. This inherent structure is beneficial for providing mass transfer pathways. In sharp contrast, Fig. 1B shows the morphology of bare nZVI particles synthesized without the biochar support. The particles exhibit severe, grape-like agglomeration, forming large micron-sized clusters. This aggregation, driven by strong magnetic interactions, drastically reduces the available reactive surface area and is a primary limitation of bare nZVI. Fig. 1C (MBC-10) shows that at a low loading ratio, some nanoparticles are visible decorating the biochar pores. Fig. 1D (MBC-40), however, demonstrates a dramatic change: the biochar surface is densely and uniformly coated with a layer of spherical nanoparticles [21], which still appear to maintain their discrete nature, confirming that the biochar scaffold effectively acts as a dispersant, preventing the severe agglomeration observed in Fig. 1B.

Transmission electron microscopy (TEM) analysis provided further insight into the nanoscale structure. Fig. 2A again confirms the severe aggregation of bare nZVI. Fig. 2B, a high-magnification image of MBC-40, clearly shows that the biochar matrix (the lighter, amorphous region) is decorated with discrete, near-spherical nZVI nanoparticles with an average diameter of 50–80 nm. This high dispersion is the key to maximizing reactive sites. Most critically, the High-Resolution TEM (HR-TEM) image in Fig. 2C reveals a distinct core-shell structure for a single nanoparticle on the MBC-40 composite. Analysis of the lattice fringes in Fig. 2D provides definitive identification: the core exhibits a lattice spacing of 0.202 nm, which corresponds perfectly to the (110) crystallographic plane of body-centered cubic $\alpha$-Fe(0). The outer shell, approximately 3–5 nm thick, shows a lattice spacing of 0.253 nm, matching the (311) plane of magnetite (Fe₃O₄). This core-shell structure (Fe(0)@Fe₃O₄) is the ideal morphology, combining the reductive power of the Fe(0) core with the sorptive and magnetic properties of the magnetite shell [22].

The intimate association and uniform distribution of the iron component were further confirmed by SEM–energy dispersive X-ray spectroscopy (EDX) elemental mapping (Fig. 3). The maps for Carbon (C), Oxygen (O), and Iron (Fe) on the MBC-40 composite show a near-perfect spatial overlap. The Fe signal is distributed evenly across the entire C-O matrix, with no evidence of large, isolated Fe-only regions. This confirms that the in-situ synthesis method was highly successful in anchoring the iron nanoparticles directly onto the biochar support.

The crystalline phases of the prepared materials were analyzed by X-ray diffraction (XRD) (Fig. 4A). The pristine BC pattern exhibits a broad diffraction feature centered at 2$\theta$ $\approx$ 26.5°, commonly assigned to the (002) reflection of turbostratic/poorly ordered aromatic carbon (graphitic-like stacking over short length scales) within an overall disordered biochar matrix. The pattern for bare nZVI shows a sharp, intense peak at 2$\theta$ = 44.7°, corresponding to the (110) plane of $\alpha$-Fe(0). The presence of additional peaks (e.g., at 35.4° for Fe₃O₄ (311)) confirms the partial surface oxidation of the nZVI, consistent with the TEM-observed core-shell structure [23]. The XRD patterns for the MBC-10 and MBC-40 composites compellingly display both the broad carbon peak at 26.5° and the sharp $\alpha$-Fe(0) peak at 44.7°. The intensity of the Fe(0) peak increases significantly from MBC-10 to MBC-40, correlating with the higher iron loading. This result provides unambiguous proof that crystalline zero-valent iron was successfully synthesized and immobilized onto the biochar scaffold.

Raman spectroscopy was employed to investigate the structural nature of the biochar support (Fig. 4B). The spectra for both pristine BC and the MBC-40 composite exhibit two prominent bands: the D-band at approximately 1350 cm^-1, which is associated with disordered carbon structures and defects, and the G-band at approximately 1590 cm^-1, which corresponds to the in-plane vibrations of sp²-hybridized graphitic carbon. The intensity ratio of these bands I(D)/I(G) is a common metric for the degree of graphitization [24]. The spectra show D (~1350 cm^-1) and G (~1590 cm^-1) bands typical of disordered sp² carbon. The I(D)/I(G) ratio (0.92) is consistent with a condensed aromatic carbon structure containing defects and finite crystallite sizes; however, I(D)/I(G) in biochar is not a definitive measure of ‘graphitization’ to true graphite and should be interpreted cautiously as reflecting disorder/ordering trends within amorphous/turbostratic carbon [25, 26]. This graphitic nature is hypothesized to be beneficial, potentially enabling the carbon matrix to act as an electron shuttle, which could facilitate the redox reactions between the Fe(0) core and adsorbed arsenic species [27].

The specific surface area (SSA) and porous properties of the materials, crucial for adsorption processes, were determined by N₂ adsorption-desorption analysis (Table 1). The pristine BC exhibited a high SSA of 285.4 m² g^-1 and a total pore volume of 0.198 cm³ g^-1, confirming its potential as an excellent porous support. Conversely, the bare nZVI had a very low SSA of only 22.1 m² g^-1, a direct consequence of the severe agglomeration observed in SEM/TEM. Upon loading of nZVI, the SSA of the composites decreased to 210.3 m² g^-1 for MBC-10 and 180.9 m² g^-1 for MBC-40. This reduction is expected, as the dense iron nanoparticles (which have a low intrinsic surface area) occupy or block some of the micropores of the biochar [28]. However, it is crucial to understand that while the total surface area decreased, the reactive surface area was vastly increased. The biochar support ensures that the 180.9 m² g^-1 of the MBC-40 composite is populated with highly dispersed, accessible nZVI reactive sites, in contrast to the inaccessible, agglomerated 22.1 m² g^-1 of bare nZVI [29].

The surface charge characteristics of the MBC-40 composite were investigated via zeta potential analysis (Fig. 5). The curve shows that the point of zero charge (PZC), the pH at which the net surface charge is zero, was 6.8. This magnitude is consistent with reported pH-PZC values for iron-bearing carbonaceous composites, while also falling within the broad variability expected for Fe-modified biochars due to differences in biomass feedstock, thermal history, iron phase composition, and surface oxygen functionality. For example, a heat-treated biochar impregnated with nanoscale zero-valent iron (HBC/nZVI) exhibited a pH-PZC of 5.1, reflecting a more acidic surface-charge transition for that specific nZVI–biochar system (likely linked to its surface functional group distribution and iron speciation) [30]. In an iron oxide–hydrochar composite prepared from pomegranate peel waste, pH-PZC values of 5.3–5.6 were reported for iron oxide–carbon composites, again indicating that Fe–carbon materials can display moderately acidic PZC values depending on synthesis conditions and oxide/hydroxide surface chemistry [31]. Conversely, Fe-oxide-modified biochars can also present higher pH-PZC values; for rice husk biochar systems modified with Fe₃O₄ or FeS₂, pH-PZC values of 9.23 and 7.78, respectively, were reported, highlighting that mineral phase identity and post-processing can shift the surface-charge neutrality point to more alkaline values. Related magnetic biochars incorporating iron oxides likewise show pH-PZC spanning approximately 4.5–9.9, depending on the modification route and co-present metal oxides, and Fe₃O₄ itself is commonly reported to have a pH-PZC near ~7.0, supporting the reasonableness of the PZC $\approx$ 6.8 measured here for an Fe(0)@Fe₃O₄-bearing composite [32].

This value is critical for understanding the electrostatic interactions during sorption. At solution pH 6.8, the composite surface is net positive (e.g., protonation of Fe-OH groups to Fe-OH₂⁺). This positive charge will strongly favor the electrostatic attraction of anionic arsenic species, which are dominant in this pH range (e.g., H₂AsO₄^–, HAsO₄^2-). Conversely, at pH $>$6.8, the surface becomes net negative (e.g., deprotonation to Fe-O^–), which would lead to electrostatic repulsion of these same anions [33]. This suggests that electrostatic attraction is a key mechanism in acidic and neutral conditions, but other, non-electrostatic mechanisms (such as chemical complexation) must become dominant for any removal observed at higher pH [34].

Fourier transform infrared spectroscopy (FTIR) spectroscopy was used to identify the surface functional groups on the materials (Fig. 6). The pristine BC spectrum shows a broad peak at ~3430 cm^-1 (O-H stretching of hydroxyl groups), a small peak at ~2850 cm^-1 (C-H stretching), a prominent peak at ~1660 cm^-1 (C=C stretching in aromatic rings and C=O stretching of carboxyl/carbonyl groups), and a band around 1050 cm^-1 (C-O stretching). These groups are all characteristic of lignocellulosic biochar. In the spectrum for MBC-40, these biochar-related peaks are still present but are attenuated [35], suggesting they are partially shielded by the nZVI coating. More importantly, a new, strong and sharp absorption peak appears at ~580 cm^-1. This peak is the characteristic Fe-O stretching vibration associated with the magnetite (Fe₃O₄) lattice. Its presence is definitive confirmation of the successful formation of the iron oxide (magnetite) shell on the biochar support, corroborating the TEM and XRD data [36].

Thermogravimetric analysis (TGA) was conducted in an air atmosphere to determine the thermal stability and quantify the actual iron loading in the composites (Fig. 7). Pristine BC shows a major weight loss of over 60% between 200–500 °C, corresponding to the combustion of the carbon matrix. In contrast, the nZVI-loaded samples show significantly enhanced thermal stability and a much larger residual mass. At 800 °C, after all carbon has been burned off and all iron has been oxidized to its most stable form (Fe₂O₃), the final residual mass was 3.1% for BC (ash content), 14.5% for MBC-10, and 41.2% for MBC-40. By back-calculating from the mass of Fe₂O₃, these residuals confirm that the actual iron loadings in the composites are highly consistent with the theoretical 10% and 40% (by Fe) targets of the synthesis [37].

The magnetic properties of the materials, essential for the “magnetically assisted separation” component of this study, were measured by vibrating sample magnetometry (VSM) (Fig. 8). As expected, pristine BC exhibits no significant magnetic response. MBC-40, however, displays a strong “S”-shaped magnetic hysteresis loop, which is characteristic of superparamagnetic or soft ferromagnetic materials [38]. The saturation magnetization (M_s) was determined to be 32.5 emu g^-1 (Table 1). This value is substantially higher than the ~16 emu g^-1 commonly referenced as a practical threshold for rapid magnetic solid–liquid separation and sorbent recovery from aqueous suspensions.

The practical implication of this strong, superparamagnetic behavior is demonstrated visually in Fig. 9. A suspension of MBC-40 in water (Fig. 9A) is turbid and black. Upon the application of an external neodymium magnet to the vial wall, the black MBC-40 particles are rapidly pulled from the solution [39]. Within 30 s (Fig. 9B), the supernatant is rendered completely clear and transparent, with all the composite material aggregated at the magnet’s location. This demonstration validates the potential for easy, rapid, and non-energetic recovery of the sorbent from treated aqueous matrices, which is a practical advantage over conventional solid–liquid separation steps such as filtration or centrifugation.

The synergistic effect of the composite structure was evaluated by comparing the arsenic removal performance of the four prepared materials under identical conditions (Fig. 10). For As(V) removal, the pristine BC showed negligible efficacy, removing only 12.2%. Bare nZVI, despite its high intrinsic reactivity, performed moderately, removing 45.7%; its effectiveness was clearly limited by the agglomeration and passivation observed in the characterization. The composites, however, showed dramatically improved performance. MBC-10 removed 62.5%, while the high-load MBC-40 composite achieved an outstanding 98.5% removal of As(V). The trend of MBC-40 $>$ MBC-10 $>$ Bare nZVI $>$ BC provides unequivocal evidence of synergy. Both the biochar support (to disperse nZVI) and a high nZVI loading (to provide sufficient reactive sites) are essential for optimal performance [40].

The solution pH is a master variable in arsenic chemistry and sorption processes. Fig. 11 shows the effect of initial pH (2.0–12.0) on the removal of As(V) and As(III) by MBC-40. The removal of As(V) was highly efficient ($>$95%) in the acidic-to-neutral range (pH 3.0–7.0). However, as the pH increased into alkaline conditions (pH $>$8.0), the removal efficiency decreased sharply. This behavior correlates perfectly with the zeta potential analysis. In the pH 3.0–7.0 range, the solution pH is below the PZC of 6.8, meaning the composite surface is positively charged and strongly attracts the anionic As(V) species (H₂AsO₄^–, HAsO₄^2-). At pH $>$8.0, the surface becomes strongly negative, leading to electrostatic repulsion. In stark contrast, the removal of As(III) (which exists as a neutral species, H₃AsO₃, across a broad pH range) was much less pH-dependent, remaining high ($>$80%) across the entire pH 3.0–10.0 range. This differential behaviour indicates that electrostatic attraction is an important contributor for anionic As(V) species when pH PZC, whereas the comparatively weak pH sensitivity for neutral As(III) is consistent with the involvement of non-electrostatic interactions (e.g., specific binding/ligand-exchange type interactions) on iron (hydr)oxide surfaces commonly associated with nZVI core–shell particles. We note that confirming inner-sphere complexation and distinguishing among potential surface precipitates versus adsorbed complexes typically requires spectroscopic/speciation tools that were not applied here; accordingly, the present interpretation is framed as plausible rather than conclusive [41].

The rate of arsenic uptake by MBC-40 was investigated (Fig. 12). The removal process was extremely rapid for both arsenic species. More than 80% of the total arsenic uptake occurred within the first 30 minutes, with the reaction reaching equilibrium by approximately 120 minutes. This rapid kinetics is highly desirable for practical water treatment applications, as it implies shorter contact times and smaller reactor volumes [42]. The data were fitted to PFO and PSO kinetic models (Fig. 13), with the calculated parameters summarized in Table 2.

As shown in Table 2, the PSO model provided a far superior fit to the experimental data for both As(V) and As(III). The correlation coefficients (R²) for the PSO model were $>$0.998, significantly higher than those for the PFO model. Furthermore, the calculated equilibrium capacity (q_e,calc) from the PSO model was in excellent agreement with the experimentally determined q_e. The PSO model provided a superior empirical description of the uptake profiles for both As(V) and As(III), as reflected by the high R² values and the agreement between q_e,calc and q_e,exp. We clarify, however, that pseudo-order kinetic fits (including PSO) are widely used as macroscopic models and do not, by themselves, uniquely diagnose chemisorption or a specific rate-limiting step; diffusion or mixed-control processes can yield comparable fits depending on the system and fitting practices. Therefore, the kinetic modelling in this work is used to compare uptake behaviour across conditions/materials rather than to assert a definitive mechanistic pathway. Mechanistic hypotheses (electrostatic contributions, possible redox involvement in nZVI systems, and specific interaction with iron (hydr)oxide surfaces) are discussed below as plausible interpretations that would require dedicated speciation/spectroscopic verification [43].

Adsorption isotherms (Fig. 14) were studied to determine the maximum immobilization capacity of the materials. The equilibrium data were fitted to the Langmuir and Freundlich models, and the resulting parameters are presented in Table 3. The Langmuir model assumes monolayer adsorption onto a homogeneous surface, while the Freundlich model describes multilayer adsorption on a heterogeneous surface.

For the MBC-40 composite, the Langmuir model provided a superior fit for both As(V) and As(III), with R² values $>$0.995, suggesting that the arsenic immobilization primarily occurs as a monolayer on a finite number of homogeneous reactive sites on the composite surface. The calculated maximum adsorption capacity (Q_max) from the Langmuir model for MBC-40 was 65.4 mg g^-1 for As(V) and 82.1 mg g^-1 for As(III). The high capacity for As(III) is particularly noteworthy, as this more toxic species is often more difficult to remove with conventional sorbents. These capacities are quantitatively synergistic, representing an 8-fold and 2.7-fold increase in As(V) capacity compared to pristine BC (8.2 mg g^-1) and bare nZVI (24.5 mg g^-1), respectively.

Real-world mining effluents are not clean systems; they contain a complex mixture of co-existing ions. Fig. 15 illustrates the impact of common anions on As(V) removal by MBC-40. The presence of high concentrations (100 mg/L) of Cl^– and SO₄^2- had a negligible impact, with removal efficiency remaining $>$95%. Carbonate (CO₃^2-) showed a slight inhibiting effect, reducing removal to 88.2%, likely by competing for surface sites or altering surface charge. Phosphate (HPO₄^3-), however, caused a significant reduction in As(V) removal, with efficiency dropping to 61.5%. This is expected and unavoidable; phosphate and arsenate are chemical analogs with similar geometries and binding affinities [44]. They compete directly for the same inner-sphere complexation sites on the iron oxide surface. This finding is critical for field applications, as high phosphate levels would necessitate a higher sorbent dose.

For a sorbent to be economically viable, it must be reusable. The MBC-40 composite was subjected to five adsorption-desorption cycles, with regeneration achieved using a 0.1 M NaOH wash (which desorbs arsenic) and recovery via an external magnet. As shown in Fig. 16, the composite demonstrated good reusability. After five cycles, the removal efficiency for As(V) remained at 81.3%, a moderate decrease from the 98.5% in the first cycle. This slight decrease in performance is likely due to two factors: (1) incomplete regeneration, as some arsenic may be irreversibly bound in inner-sphere complexes, and (2) the progressive consumption and loss of the Fe(0) reductive capacity, which is not restored by the NaOH wash. Nonetheless, an 81.3% efficiency after five uses is highly promising.

The final and most critical test was the performance of MBC-40 in the complex, multi-pollutant SMW, designed to mimic a real-world challenge. The results, summarized in Table 4, are outstanding. Despite the extremely high background concentration of sulfate (1000 mg L^-1) and the presence of competing heavy metals, the composite demonstrated excellent efficacy. It removed 97.0% of the total initial arsenic (5.0 mg L^-1), reducing the concentration to 0.15 mg L^-1. It was equally effective against both As(V) (97.3% removal) and As(III) (96.5% removal). Furthermore, the composite exhibited excellent co-removal capabilities, removing $>$99.9% of the toxic Pb²⁺ and 95.6% of the Cd²⁺. This confirms that the MBC-40 composite is not just an arsenic sorbent but a broad-spectrum remediation agent, capable of effectively treating the complex, multi-pollutant effluents typical of mining-impacted sites [45].

To contextualize these outcomes, we emphasize that several macroscopic adsorption behaviors observed here (e.g., pseudo-second-order fitting, Langmuir-type isotherms, phosphate competition, and magnet-assisted recovery) have been reported previously for iron–carbon adsorbents and magnetic composites. For example, Singh and Mohan reported that adsorption behavior consistent with pseudo-second-order kinetics and Langmuir/Freundlich modeling, identified phosphate as a strong competitor, and noted that the sorbent could be recovered by a magnet [46]. The present study therefore contributes primarily at the applied systems level: (i) Fe-loading-controlled synthesis (MBC-10 vs MBC-40) is linked to quantified magnetization and arsenic uptake, (ii) the Fe(0)@Fe₃O₄ core–shell architecture is verified directly by HR-TEM lattice spacing assignments, and (iii) performance is validated under mining-relevant mixed contamination (5.0 mg L^-1 total As with 1000 mg L^-1 sulfate and co-occurring Pb²⁺/Cd²⁺/Zn²⁺), demonstrating robust arsenic removal alongside strong co-removal of Pb²⁺ and Cd²⁺.

While the MBC-40 composite demonstrates high efficacy for arsenic immobilization in the tested environments, several limitations of the current study should be noted:

Mechanistic uncertainty: The proposed immobilization mechanisms are interpreted as plausible based on macroscopic adsorption data rather than direct speciation measurements. This study did not utilize advanced spectroscopic tools such as X-ray absorption spectroscopy (XAS) or XPS to definitively distinguish between surface adsorption, inner-sphere complexation, and the formation of specific surface precipitates.

Aqueous matrix simplification: Experimental evaluations, including the SMW tests, were conducted in the absence of DOM or NOM. Because DOM can compete for reactive sites or alter arsenic mobility, these findings primarily reflect material performance in simplified inorganic matrices rather than complex natural or field conditions.

Geochemical ageing and passivation: The study focused on short-term performance and did not account for time-dependent geochemical ageing or the long-term passivation of the nZVI core under oxic or anoxic environmental conditions. Assessing the long-term stability and reductive lifespan of the Fe(0)@Fe₃O₄ architecture remains a priority for future field-oriented research.

Kinetic and structural interpretations: The application of PSO kinetic models serves as a macroscopic comparison of uptake behavior across different conditions but does not uniquely diagnose a specific rate-limiting step or provide definitive proof of chemisorption. Additionally, the I(D)/I(G) ratio obtained via Raman spectroscopy is a measure of relative disorder and should not be interpreted as a definitive quantification of true graphitization within the biochar matrix.

Loss of reductive capacity: Reusability experiments showed a moderate decline in arsenic removal efficiency after five cycles (from 98.5% to 81.3%), which is likely attributed to the progressive consumption of the Fe(0) core’s reductive capacity and the irreversible binding of arsenic to certain reactive sites.