Fe₃O₄ Magnetic Biochar Derived from Pecan Nutshell for Arsenic Removal Performance Analysis Based on Fuzzy Decision Network ^†

Abstract

This study evaluates Fe₃O₄ magnetic biochar synthesized from pecan nutshells for arsenic removal. Surface modification with Fe₃O₄ significantly enhanced arsenic adsorption selectivity and efficiency compared to raw biomass (PM). Synthesis variables (precursor type, particle size, Fe/precursor ratio, N₂) and adsorption conditions (such as concentration, pH, agitation) were investigated. The modified biochar achieved >90% arsenic removal efficiency under various conditions, demonstrating the modification’s critical role. A fuzzy decision network was employed to analyze experimental results and identify optimal conditions for maximizing performance. This approach effectively leverages knowledge for scenario-specific optimization, offering a sustainable strategy for advanced water treatment materials.

1. Introduction

Groundwater arsenic (As) contamination poses severe public health and environmental risks. Adsorption, particularly using biochar derived from biomass waste like pecan nutshells, offers a promising remediation solution. Biochar’s porous structure, large surface area, and functional groups enable effective contaminant removal via mechanisms like ion exchange [1,2]. Its adsorption capacity is influenced by feedstock, production methods, and chemical modifications [3,4]. Magnetic biochar, incorporating magnetite (Fe₃O₄) nanoparticles, enhances adsorption efficiency and enables easy magnetic separation [4,5].

This study focuses on synthesizing magnetic biochar from pecan nutshells via co-precipitation, investigating variables (precursor type, particle size, Fe/precursor ratio, N₂ gas) to optimize As removal. Performance optimization in such complex systems requires advanced modeling. Fuzzy Decision Networks (FDNs), specifically the multi-input fuzzy rules emulated network (MiFREN), effectively handles uncertain, nonlinear relationships between input parameters (e.g., material properties, experimental conditions) and performance outcomes [6,7]. MiFREN integrates experimental data, constructs membership functions, and formulates a performance index (Q_r) to assess and optimize As removal efficiency.

The primary objective is to employ MiFREN-based analysis to enhance arsenic removal, leveraging magnetic biochar’s advantages (sustainable feedstock, efficient separation) while providing a robust framework for process optimization under uncertainty. This approach aims to advance cost-effective water treatment technologies for groundwater remediation.

2. Materials and Methods

2.1. Fe₃O₄-Pecan Biochar Synthesis via Co-Precipitation

Pecan nutshells were washed, dried (60 °C, 24 h), ground, and sieved (0.105–0.18 mm; 0.38–0.7 mm). Biochar was produced by pyrolyzing biomass under N₂ (500 °C, 1 h). Fe₃O₄ magnetic composites were synthesized via co-precipitation: FeCl₃·6H₂O (0.1 mol) and FeSO₄·7H₂O (0.05 mol) were dissolved in DI water (600 mL). NaOH was added dropwise (80 °C, 500 rpm) until pH 10. Biomass (PM) or biochar was added, and stirring continued (30 min, 80 °C, 500 rpm) with/without N₂. The nanocomposite was magnetically separated, washed to neutral pH, and dried (60 °C, 24 h).

Fe₃O₄ magnetic composites (FS1–FS10) were prepared via co-precipitation using pecan nutshell biomass (FS1–FS6, FS8–FS10) or biochar pyrolyzed at 500 °C (FS7) as precursors. Particle sizes varied: 0.105–0.18 mm (FS3, FS6), 0.18–0.38 mm (FS1, FS4, FS7–FS10), and 0.38–0.7 mm (FS2, FS5). Iron salts FeCl₃·6H₂O and FeSO₄·7H₂O were used for all samples except FS1 (FeCl₃ + FeSO₄(NH₄)₂SO₄·6H₂O). The Fe/precursor ratio was predominantly 1:1 (FS1–FS8), except FS9 (1:2) and FS10 (1:3). Nitrogen gas (N₂) was applied only during synthesis of FS4, FS5, and FS6; all other samples (FS1–FS3, FS7–FS10) were synthesized without N₂. This matrix systematically tested precursor type, particle size, Fe salt combination, Fe/precursor ratio, and N₂ atmosphere effects.

Material characterization was performed via X-ray Diffraction (XRD) and Scanning Electron Microscopy (SEM). A Philips PW3040 diffractometer (Almelo, the Netherlands) and a JEOL 7800F Prime field-emission SEM (Tokyo, Japan) were used for these analyses, respectively.

2.2. Adsorption Process for as Removal

Arsenic adsorption tests were performed under various conditions to evaluate the removal efficiency (% removal) and adsorption capacity (q_e, mg/g) as a function of arsenic initial concentration (1, 5, and 10 mg/L), pH (3, 4, and 5), adsorbent dose (1 and 2 g/L), and agitation speed (100 and 150 rpm) for FS1–FS10, including pecan nutshell (PM) as a control.

2.3. Analysis of Fuzzy Decision Network

The performance analysis employs a multi-input fuzzy rules emulated network (MiFREN). Inputs include normalized removal quality ($q_{eN}$) and fuzzified via membership functions (${\mu }_{L-Z}$). MiFREN integrates weight vectors ($w$) specific to experimental setups. The performance index ($Q_r$) is calculated as a weighted sum of basis functions (${\varphi }_j$), derived from membership function combinations, which capture nonlinear parameter interactions and model complex decisions. A computational framework using unity functions ensures generality in deriving ${\varphi }_j$. Validation compares $Q_r$ values across samples, revealing impacts of material properties, particle size, and preparation methods. This demonstrates MiFREN’s efficacy in modeling nonlinear relationships and optimizing arsenic removal.

3. Results and Discussion

3.1. Characterization

The XRD pattern (Figure 1) shows the diffractograms of synthesized composites (FS1–FS10), pecan nutshell biomass (PM), and a standard magnetite (Fe₃O₄) pattern. The characteristic peaks of cubic Fe₃O₄, corresponding to the (220), (311), (400), (422), (511), and (440) lattice planes, are evident in the composites, confirming the successful formation of Fe₃O₄ [8,9]. These composites often incorporate Fe₃O₄ (magnetite) nanoparticles, which can be formed during the pyrolysis process of iron-impregnated biomass [10]. Samples utilizing biochar as a precursor show sharper Fe₃O₄ peaks, especially FS7 compared to FS2 and FS4, indicating higher crystallinity or larger crystallite size. This enhances the results from the pyrolysis process during biochar production, which creates a stable carbonaceous framework that promotes Fe₃O₄ formation [11]. The properties of the resulting nanocomposites are influenced by the precursor choice (biochar or biomass), the iron salt-to-precursor ratio, and the type of iron salt. The effects of N₂ and particle size on Fe₃O₄ crystallinity are minimal.

SEM images (Figure 2) reveal the surface morphology of Fe₃O₄-biochar composites (FS4, FS7, and FS8). FS7 (Figure 2b) exhibits a highly textured surface with a porous biochar matrix and well-distributed nanostructured Fe₃O₄ particles [12]. This suggests a complex, hierarchical structure with a combination of larger and smaller particles. Spherical-like particles, likely Fe₃O₄ nanoparticles (17.59–22.72 nm), are observed embedded within the matrix, corroborating XRD analysis. FS4 (Figure 2a) and FS8 (Figure 2c) display aggregated Fe₃O₄ nanoparticles on a biomass matrix, differing from FS7’s more porous structure and homogeneous nanoparticle distribution. These morphological variations likely influence the adsorption capabilities of these composites.

3.2. Arsenic Removal by Fe₃O₄-Pecan Magnetic Biochar

Figure 3 illustrates the arsenic removal efficiency of various materials (FS1–FS10 and PM) under five experimental conditions (EXP I–V). Fe₃O₄-based composites (FS1–FS10) exhibited significantly higher arsenic removal (59.50–99.89%) than raw pecan nutshell biomass (PM, 3.05–28.4%). FS1, synthesized using FeSO₄(NH₄)₂SO₄·6H₂O, consistently demonstrated high removal rates, particularly at pH 3. Composites synthesized with FeCl₃·6H₂O + FeSO₄·7H₂O also showed excellent performance. Lower biomass ratios (e.g., FS1–FS7) generally resulted in better arsenic removal. Furthermore, Fe₃O₄ particles on biochar surfaces have been shown to enhance the adsorption of organic pollutants like methylene blue and improve thermal stability [10]. Arsenic removal efficiency was significantly influenced by pH, with optimal removal observed at lower pH (pH 3), likely due to enhanced surface protonation.

The porous structure of biochar in FS7 likely enhances adsorption by facilitating greater interaction with arsenic. SEM analysis revealed that the lower porosity of the biomass matrix in FS4 and FS8 may hinder adsorption performance. FS7, with its high surface area, better dispersion of Fe₃O₄ particles, and advantageous biochar precursor, demonstrated the highest arsenic adsorption capacity among all materials.

3.3. MiFREN Performance Analysis

MiFREN utilizes the normalized removal quality (q_eN) as its key input for assessing adsorption performance. The network employs two node functions. The second function incorporates a defined weight vector w = [0.5, 0.6, 0.7, 0.85, 1]^T, assigned to each experimental setup, allowing the model to differentiate arsenic removal efficiency across conditions. Membership functions μ_L−Z are specifically designed to characterize the fuzzy logic relationships within MiFREN. These functions, graphically presented in Figure 4, define the input–output mapping crucial for performance assessment using the fuzzy decision network.

Thereafter, the quality index $Q_r$ is determined as

$$Q_r={\displaystyle {\sum }_{j=1}^{18}}{\beta }_j{\varphi }_j$$

(1)

where ${\beta }_j$ denotes as the weight parameters of MiFREN and ${\varphi }_j$ is the basis function of MiFREN at the $j^{th}$ rule such that ${\varphi }_j={\mu }_L\left(f_1\right){\mu }_S\left(w{f}_2\right)$. Without loss of generality, in this work, both functions $f_1\left(q_{eN}\right)$ and $f_2\left(q_{eN}\right)$ are selected as unity functions. Therefore, the basis function can be simplified as

$${\varphi }_j={\mu }_L^j\left(q_{eN}\right){\mu }_S^j\left(w{q}_{eN}\right)$$

(2)

$${\varphi }_j={\mu }_L^j\left(q_{eN}\right){\mu }_S^j\left(w{q}_{eN}\right)$$

(3)

Figure 5 demonstrates the nonlinear relationship between $Q_r$ and $q_{eN}$, highlighting MiFREN’s strength in capturing complex interactions governing arsenic removal efficiency. This nonlinearity reveals intricate dependencies between material properties, particle size, and preparation methods that simple linear models miss. MiFREN effectively models these interactions using adaptive membership functions and rule-based inference, providing a more accurate representation of performance variations under different conditions. Compared to conventional regression, MiFREN offers superior flexibility in handling uncertainties and dynamic parameter changes, making it highly suitable for optimizing remediation processes. The analysis identifies critical thresholds in $Q_r$ where specific parameter adjustments significantly impact efficiency, offering valuable insights for fine-tuning adsorption conditions. Furthermore, MiFREN’s ability to model these complex dependencies suggests strong potential for real-time monitoring and adaptive control in water treatment applications, enhancing its utility for performance optimization.

Figure 6 presents the performance index $Q_r$, showing FS7 achieved the highest value, followed by FS2 and FS5. FS7’s superior arsenic removal is attributed to its thermally treated biochar base, enhancing surface area and porosity. FS2’s high performance stems from its specific physical structure and chemical composition. Conversely, FS10 exhibited significantly lower $Q_r$, indicating poor efficiency. These results confirm that material properties (e.g., surface area), synthesis conditions (Fe/precursor ratio, particle size, N₂ incorporation), and preparation methods critically influence active site formation and adsorption performance. MiFREN effectively models these relationships, demonstrating utility for process optimization.

.

4. Conclusions

Fe₃O₄ magnetic biochar composites, synthesized via co-precipitation from agricultural residues, demonstrate effective arsenic removal. Key performance factors are Fe₃O₄ loading, precursor type (biochar > biomass), particle size (smaller preferred for biomass), N₂ presence during synthesis, and optimal Fe/precursor ratio (1:1). These composites are recoverable and eco-friendly. Performance analysis using the MiFREN fuzzy decision network effectively evaluated efficiency under varying conditions. By modeling normalized removal quality (q_eN) with tailored weights and membership functions, MiFREN captured nonlinear interactions and generated the performance index Q_r. Results confirmed FS7 as the top performer, followed by FS2/FS5, while FS10 showed limited effectiveness, validating MiFREN’s sensitivity to material properties and synthesis conditions.