Bioadsorption of Manganese with Modified Orange Peel in Aqueous Solution: Box–Behnken Design Optimization and Adsorption Isotherm

Abstract

Chemically demethoxylated and Ca-cross-linked orange-peel waste was engineered as a biosorbent for Mn(II) removal from water. A three-factor Box–Behnken design (biosorbent dose 3–10 g L⁻¹, initial Mn²⁺ 100–300 mg L⁻¹, contact time 3–8 h; pH 5.5 ± 0.1, 25 °C) required only 16 runs to locate the optimum (10 g L⁻¹, 100 mg L⁻¹, 8 h), at which the material removed 94.8% ± 0.3% manganese removal under the optimized conditions (10 g L⁻¹, 100 mg L⁻¹, 8 h, pH 5.5) of dissolved manganese and reached a Langmuir capacity of 29.7 mg g⁻¹. Equilibrium data fitted the Freundlich (R² = 0.968) and Sips (R² = 0.969) models best, indicating a heterogeneous surface, whereas kinetic screening confirmed equilibrium within 6 h. FTIR and SEM–EDX verified abundant surface –COO⁻/–OH groups and showed Mn deposits that partially replaced residual Ca, supporting an ion-exchange component in the uptake mechanism. A preliminary cost analysis (<USD 10 kg⁻¹) and > 90% regeneration efficiency over three cycles highlight the economic and environmental promise of this modified agro-waste for polishing Mn-laden effluents.

1. Introduction

Manganese (Mn) is an essential trace element involved in photosynthesis, antioxidant defense, and carbohydrate metabolism; nevertheless, excess Mn becomes toxic once environmental or physiological concentrations surpass threshold values. The World Health Organization currently sets a provisional health-based guideline of 0.08 mg L⁻¹ for total Mn in drinking water [1]. Anthropogenic releases from mining, metallurgical, and battery industries have produced Mn levels that exceed these limits and, under hypoxic or acidic conditions, the metal becomes more soluble and bio-available, thereby aggravating its adverse impacts on aquatic fauna and human health [2,3].

Manganese contamination adversely impacts surface water, groundwater, soils, and sediments. Leaching from rocks, soils, and industrial facilities elevates Mn concentrations, thereby degrading water quality and compromising the health of aquatic ecosystems [4]. Concentrations exceeding 0.4 mg L⁻¹ in groundwater—the World Health Organization (WHO) guideline—may induce oxidative stress, neurotoxicity, and additional adverse effects in humans [5]. Moreover, excessive manganese lowers dissolved oxygen levels in aquatic environments and impairs fish physiology, particularly by damaging the branchial epithelium [6]. Chronic exposure is associated with neurotoxic outcomes and cognitive deficits in exposed populations [7]. Manganese is an essential element whose normal body burden ranges from 12 to 20 mg, distributed across critical organs such as the brain, kidneys, pancreas, and liver, with notably high accumulation in the pituitary gland. Dysregulation of Mn homeostasis has been linked to neurodegenerative disorders, including Alzheimer’s disease and Huntington’s disease [8]. Elevated manganese levels also pose a significant risk to neuronal integrity in children [9]. Nevertheless, environmental legislation in several countries—Peru among them—remains insufficient with respect to Mn discharge limits in effluents [10], thereby exacerbating the problem.

Owing to the high redox potential of manganese, conventional water-treatment methods are frequently ineffective. Recent research has therefore concentrated on chemical oxidation pathways that employ strong oxidants under alkaline and catalytic conditions [11]. Although these strategies have advanced the field, they remain constrained by elevated operational costs and the risk of secondary contamination. Biological manganese removal, which exploits the natural metabolic pathways of microorganisms, offers advantages in terms of cost and environmental sustainability; however, it still incurs substantial expenses associated with the maintenance and engineering of bioreactors, while nonetheless reducing the demand for chemical reagents [12]. Multiple studies have reported manganese abatement through diverse treatment modalities, including adsorption onto modified potato peels [13], ozonation [14], pearl-shaped chitosan nanocomposites [15], pyrite–calcium sulfite systems [16], manganese-coated sand and manganese oxide [17], sodium alginate beads [18], biochar derived from apple pulp [19], beet pulp [20], quartz grains coated with hydroxyapatite [21], and natural zeolite [22]. Among the techniques investigated, physical adsorption has emerged as the most promising, owing to its capacity to remove contaminants from aqueous media, operational simplicity, cost-effectiveness, and the broad availability of natural and synthetic adsorbents [23].

Adsorption has emerged as a sustainable technique for the removal of heavy metals. The process employs modified residual biomass to capture metal contaminants from aqueous media [24]. In contrast to conventional treatments, adsorption relies on low-cost waste materials and is environmentally benign [25]. A range of agricultural by-products has been investigated as adsorbents, including rice husk [26], assorted fruit peels [27,28], [29], sugar-cane bagasse [30,31], corn stover [32,33], eggshells [34,35,36], and peanut shells [37,38], among others. Among these biomasses, orange peel—an abundant residue of the juice industry—is particularly attractive because it is rich in pectin, a polysaccharide bearing numerous functional groups (carboxyl and hydroxyl moieties) capable of chelating metal ions [39]. Previous studies have demonstrated that orange peel-based materials efficiently remove arsenic [40], chromium [41], lead [42], and other metals.

Globally, citrus processing generates a sizeable and under-utilized lignocellulosic stream. FAO’s most recent market analysis places world citrus output at 150 million t yr⁻¹, and juice manufacture discards 40–50% of that mass as peel and pulp [43]. Oranges alone contribute 76 million t yr⁻¹ to this figure [44]; hence, more than 30 million t yr⁻¹ of orange peel is potentially available for valorization. Peru is no exception: the country now produces 1.7 million t yr⁻¹ of citrus [45], about one-third of which are sweet oranges, so an estimated 0.22 million t yr⁻¹ of orange peel waste accrues locally after processing. Recent high-impact studies confirm that modified orange peel is far more than a low-value by-product. Acid/alkali-purified peel has achieved Cd(II) sorption capacities up to 31 mg g⁻¹ [46], while Ca²⁺-cross-linked peel removed Co(II) and Mn(II) with > 90% efficiency in batch tests [47]. Beyond classical biosorption, orange-peel biochars doped with Fe–Cu or NiO have been shown to catalytically degrade tetracycline and other emerging pollutants, evidencing dual adsorptive–redox functionality [48]. Comprehensive reviews published in 2024 underscore these advances and highlight orange waste as one of the most promising agro-industrial feedstocks for sustainable wastewater treatment technologies [49].

Recent literature underscores how the chemically tailored orange-peel adsorbent outperforms other Mn(II) biosorbents on a cost-to-capacity basis. Among recent lignocellulosic biosorbents for Mn(II), sugar-cane bagasse shows the lowest capacity (2.82 mg g⁻¹) [50] but is virtually cost-free (0.02 USD kg⁻¹) [51], whilst chemically activated rice husk achieves outstanding uptakes (170–270 mg g⁻¹) at roughly USD 0.5 kg⁻¹ [52], albeit with extra processing steps. NaOH-activated coconut-shell biochar delivers a moderate 10 mg g⁻¹ [53] and remains inexpensive thanks to abundant feedstock, and pyrolysed rice-straw biochar offers 26 mg g⁻¹ for less than USD 0.5 kg⁻¹ [54] without chemical activation. In this context, the chemically demethoxylated and Ca-cross-linked orange-peel biosorbent balances both metrics, providing a competitive capacity (29.7 mg g⁻¹) at a production cost below USD 10 kg⁻¹ [51]—well above the performance of unmodified residues and on par with low-temperature biochars, yet far cheaper than commercial activated carbons, underscoring its economic and technical advantage for Mn(II) removal.

Response surface methodology (RSM) coupled with the Box–Behnken design (BBD) expedites both process optimization and experimental planning, thereby enhancing efficiency while conserving resources [55]. The integration of BBD within an RSM framework substantially reduces the number of experimental runs required, leading to significant savings in materials, time, and funds [56]. In addition, the evaluation of non-linear interactions improves system representation accuracy, whereas statistical modelling ensures that conclusions drawn from experimental data remain reliable [57]. Contour plots and response surface graphs facilitate result interpretation, enabling researchers to elucidate the intricate interactions among the experimental factors under investigation [58]. Recent applications of response surface methodology (RSM) to metal biosorption continue to rely predominantly on central composite designs (CCD) or Taguchi orthogonal arrays, leaving Box–Behnken designs (BBD) comparatively under-exploited. For example, Adansonia digitata biomass optimized by a face-centered CCD achieved 91% Cu(II) removal but required 20 experimental runs and four axial points outside the practical pH domain [59]. Similarly, CCD matrices have been used to maximize Pb(II) or Cd(II) uptake onto walnut-shell charcoal and other agro-residues, yet still demanded ≥ 30 runs to fit the quadratic model [60]. Orthogonal Taguchi L-arrays provide a leaner alternative but are limited to main-effects screening and ignore interaction terms critical for sorption equilibria [61]. By contrast, BBD operates with three coded levels per factor, avoids extreme axial points that can degrade biomasses, and generates rotatable second-order models in just 2k(k − 1) + C₀ runs. Recent comparative studies confirm that BBD delivers similar predictive power while cutting experimental effort by 40% relative to CCD in Cu/Cd removal systems [62]. Building on these findings, the present work adopts BBD to optimize manganese biosorption with a chemically tailored orange-peel sorbent, thereby filling a methodological gap and providing a more resource-efficient framework for process scale-up.

Despite the extensive exploration of citrus wastes as sustainable adsorbents, previous studies have focused mainly on unmodified peels or on simple esterification/alkaline treatments, reporting Mn(II) capacities ≤ 15 mg g⁻¹ [63]. To our knowledge, no work has combined selective demethoxylation of pectic domains followed by Ca²⁺ induced reticulation to tailor carboxyl density and mechanical stability of orange peel for manganese removal; recent reviews of orange-derived sorbents confirm this research gap [64]. Here, we bridge that gap and, using response surface methodology based on a Box–Behnken design, demonstrate that the modified biosorbent attains a Langmuir capacity of 25.1 mg g⁻¹ with 66% higher than the best values reported for untreated orange peel and other agro-biopolymers [65]. Furthermore, the BBD required only 16 experimental runs, approximately 40% fewer than an equivalent CCD, while maintaining model rotatability and quadratic resolution [66]. These advances position the proposed adsorbent–design pairing as a novel and economically attractive option for polishing Mn-laden effluents.

The present study evaluates the bio-adsorptive capacity of orange peel for manganese removal from synthetic aqueous solutions under laboratory conditions. Specifically, the influence of operational parameters, biosorbent dosage, contaminant concentration, and contact time on Mn removal efficiency is assessed and optimized through a BBD-based experimental design.

2. Materials and Methods

2.1. Characterization, and Pre-Processing of Orange Peel

Approximately 6 kg of post-consumer orange peels were collected from juice stalls in the San Martín de Porres district, Lima, Peru. The peels were thoroughly rinsed with bidistilled water to remove residual pulp and surface contaminants, then dried in an electric convection oven model YR05259-2 (Kalstein,. Ltd., Paris, France) at 60 °C for 24 h. A 230 g aliquot of the dried material was submitted to the accredited laboratory (SATPERU, Ltd., Lima, Peru) for physicochemical characterization. Analyses followed standardized AOAC methods: total sugars (AOAC 968.28, 2023), ash content (AOAC 940.26, 2023), lipid fraction (AOAC 920.177, 2023), moisture (AOAC 920.151, 2023), crude protein (AOAC 920.152, 2023), and total solids (AOAC 920.151, 2023). The resulting compositional profile provides critical baseline information for subsequent activation processes and manganese adsorption experiments.

2.2. Synthesis of the Bioadsorbent

Valorization of citrus-processing residues as cost-effective adsorbents aligns with current circular-economy engineering strategies. In this study, dried orange peels were coarse-milled and sieved to an average particle size of 850 µm according to ASTM particle-size standards. Sequential chemical activation was then performed: (i) de-methoxylation in 0.1 M NaOH at 25 °C for 2 h under continuous agitation to hydrolyze pectic methoxyl groups and increase the density of free carboxylate functionalities, and (ii) ionic crosslinking in 0.2 M CaCl₂ to promote calcium-mediated bridges between adjacent pectin chains and reinforce the polymeric network. After each treatment, the solid was vacuum-filtered and oven-dried. These combined modifications improved mechanical stability and produced a higher density of carboxyl and electrostatically active sites, enhancements that are expected to increase the bioadsorbent contaminant-uptake capacity.

All batch experiments were carried out at pH 5.5 ± 0.1, adjusted with 0.1 M HCl or 0.1 M NaOH and monitored continuously with a calibrated pH electrode. This pH was selected because speciation diagrams show that > 98% of dissolved Mn remains as Mn²⁺ below pH 6, whereas hydrolysis and precipitation of Mn(OH)₂ start at pH de 6.3, compromising mass-balance accuracy and obscuring biosorption mechanisms [67]. Conversely, at pH < 4.5 proton competition for –COO⁻ and –OH active sites markedly lowers uptake efficiency, as reported for citrus-based biosorbents by Schiewer [68]. Fixing the pH at 5.5 therefore maximizes the availability of Mn²⁺ while preserving both the integrity of the modified orange-peel matrix and the validity of Langmuir/Freundlich modelling.

2.3. Characterization of the Modified Shell

Several physicochemical and spectroscopic analyses of the bioadsorbent were performed: (a) FTIR-ATR infrared spectroscopy was performed to identify functional groups present in the shell (after crosslinking), following the ASTM [69] regarding general techniques for obtaining infrared spectra for qualitative analysis. A Perkin Elmer infrared spectrophotometer, equipped with Perkin Elmer Spectrum 10 software, was used, and spectra were recorded within a wavenumber range of 380 cm⁻¹ to 4000 cm⁻¹, allowing identification of the main functional groups present on the surface of the activated material. (b) Scanning electron microscopy (SEM) was performed to examine the surface morphology of the adsorbent. (c) Energy-dispersive X-ray spectroscopy (EDX), coupled with SEM, was performed to determine the elemental composition of the surface following Mn adsorption.

2.4. Box–Behnken Experimental Design

Response surface methodology (RSM) was applied along with a Box–Behnken design (BBD) to optimize manganese adsorption in aqueous solution. RSM allows evaluation of the optimization process, while BBD facilitates systematic variation of factors within the experimental design. The design, regression analysis, and response surface plots were performed using Design-Expert v13 [70]. In particular, BBD requires fewer experimental runs than central composite design (CCD), thus reducing reagent usage and lowering costs during optimization.

A Box–Behnken design (BBD) was selected instead of a central composite design (CCD) because it fulfils the quadratic modelling requirements of response surface methodology while being more economical and safer for biomass-based systems. For a three-factor study, BBD needs only N = 2k(k − 1)+C₀ = 16 runs, roughly 20–40% fewer experiments than the 20–32 runs typically demanded by a face-centred or rotatable CCD, yet it preserves model rotatability and orthogonality. Unlike CCD, BBD omits axial (±α) points that would expose the orange-peel matrix to extreme pH or metal-loading conditions likely to cause hydrolysis or pore collapse. Recent high-impact studies report that BBD delivered predictive accuracies comparable to or better than CCD while reducing laboratory effort in the optimization of Ni(II) adsorption onto corn-waste biochar [71] and multi-metal remediation with nanoscale zero-valent iron [72]. Consequently, BBD offers an optimal trade-off between statistical rigour and experimental economy for the present Mn(II) biosorption study.

For this study, three independent factors were defined: bioadsorbent concentration (Factor 1), contaminant concentration (Factor 2), and contact time (Factor 3), each set at three levels (low, medium, and high). This resulted in 16 experimental runs, as shown in the corresponding Table. RSM enabled analysis of manganese removal efficiency (% removal) and interaction among factors, aiming to determine optimal adsorption conditions. Based on this, Equation (1) presents the mathematical expression for the polynomial statistical model.

$$y={\beta }_0+\sum {\beta }_i{X}_i+\sum {\beta }_{ii}{X}_i^2+\sum \sum {\beta }_{ij}{X}_{ij}$$

(1)

where $X_i$ and $X_j$ represent individual adsorption variables, and Y is the simulated response (Mn removal). ${\beta }_i$, ${\beta }_{ii}$, and ${\beta }_{ij}$ denote regression constants for the intercept, linear effect, quadratic effect, and interaction effect, respectively.

Table 1. Experimental matrix.

Run	Factor 1	Factor 2	Factor 3	Results
	Bioadsorbent Concentration	Manganese Concentration	Contact Time	Mn Removal (%)
	(g L−1)	(mg L−1)	(h)
1	7.2	190	3	85.73
2	5.8	100	5.25	90.07
3	10	300	8	82.72
4	10	300	8	84.39
5	10	100	8	92.73
6	10	210	5.25	84.03
7	6.15	300	4.75	75.99
8	3	100	7	83.87
9	3	210	3	84.66
10	3	230	5.75	77.58
11	10	100	3	90.84
12	10	300	3	79.38
13	10	300	3	81.53
14	10	100	8	94.79
15	4.75	200	8	78.70
16	3	300	8	69.34

details the 16 experimental runs generated by the three-factor, three-level Box–Behnken design (BBD) used to optimize Mn(II) biosorption. Factor 1 (bioadsorbent concentration) was varied from the coded low level of 3 g L⁻¹ to the high level of 10 g L⁻¹, capturing the transition from dilute to particle-rich suspensions. Factor 2 (initial manganese concentration) spanned 100–300 mg L⁻¹, a range wide enough to probe both kinetic and capacity-limited regimes. Factor 3 (contact time) was adjusted between 3 h and 8 h to evaluate the interplay between mass-transfer and equilibrium constraints. The design includes four replicated center-point runs (10 g L⁻¹, 200 mg L⁻¹, 5.5 h) that permit pure-error estimation and rotatability assessment. Mn removal efficiencies varied from 69.34% (Run 16; low biosorbent/high metal, long contact) to 94.79% (Run 14; high biosorbent/low metal, intermediate contact), indicating that the chosen factor levels effectively bracketed both sub-optimal and near-optimal operating windows.

The summary statistics corroborate the balance and dispersion built into the BBD (

Table 2. Descriptive statistics for the Box–Behnken matrix.

Factor	n	Mean	Median	Standard Deviation	Minimum	Maximum
Bioadsorbent concentration (g L−1)	16	7.24	8.6	3.08	3	10
Manganese concentration (mg L−1)	16	208.75	210	85.63	100	300
Contact time (h)	16	5.69	5.5	2.18	3	8
Mn removal (%)	16	83.52	83.95	6.6	69.34	94.79

). Mean bioadsorbent loading (7.24 g L⁻¹) and mean manganese concentration (208.75 mg L⁻¹) lie close to their respective medians, reflecting the design’s symmetry around central levels. Standard deviations of 3.08 g L⁻¹ for bioadsorbent and 85.63 mg L⁻¹ for Mn represent roughly 43% of the variable ranges, confirming adequate exploration of factor space without excessive extrapolation. Contact time shows moderate spread (SD = 2.18 h), ensuring kinetic effects are discernible yet experimentally tractable. The response variable, Mn removal, exhibits a mean of 83.52% with a coefficient of variation of only 7.9%, indicating consistently high performance across runs and suggesting that the modified orange-peel matrix maintains robust sorption capacity throughout the tested domain.

2.5. Adsorption Isotherms

To generate equilibrium isotherms, 50 mL aliquots of Mn(II) solution with initial concentrations ranging from 50 to 500 mg L⁻¹ were contacted with the modified orange-peel biosorbent at a fixed dosage of 1 g L⁻¹. Suspensions were maintained at 25 ± 1 °C in an orbital shaker (150 rpm), and the pH was kept constant at 5.5 ± 0.1 using 0.1 M HCl or NaOH. After the predetermined contact time corresponding to equilibrium, samples were filtered through 0.45 µm membranes and analyzed for residual Mn(II) by ICP-OES.

Equilibrium data were interpreted with four adsorption isotherms. The Langmuir model (Equation (2)) assumes monolayer uptake on a homogeneous surface with finite, energetically equivalent sites; its parameters are the maximum sorption capacity q_max (mg g⁻¹) and the Langmuir affinity constant K_L (L mg⁻¹), which reflects binding energy. The Freundlich model (Equation (3)) describes multilayer sorption on heterogeneous surfaces and yields the Freundlich constant K_f and the heterogeneity factor n (dimensionless). To capture possible adsorbate–adsorbate interactions at moderate surface coverage, the data were also fitted to the Temkin equation (Equation (4)), where K_T (L g⁻¹) is the Temkin binding constant and B=RT/b relates to the mean heat of adsorption (b, J mol⁻¹). Finally, the dataset was examined with the Sips model (Equation (5)), a three-parameter equation that merges Langmuir behavior at low adsorbate concentrations with Freundlich heterogeneity at higher loadings; its parameters are the hybrid capacity q_m, the Sips affinity constant k_s (L mg⁻¹), and the heterogeneity exponent n (dimensionless), which approaches unity for a purely Langmuir surface. Non-linear least-squares regression was applied to Equations (2)–(5), and goodness-of-fit was assessed with adjusted R², root mean square error (RMSE), and reduced chi-square.

$$q_e={\displaystyle \frac{q_{max}{K}_L{C}_e}{1+K_L{C}_e}}$$

(2)

$$q_e=K_f{C}_e^{1/n}$$

(3)

$$q_e={B Ln (K}_T{C}_e), \mathrm{where} B={\displaystyle \frac{RT}{b}}$$

(4)

$$\ln ({\displaystyle \frac{q_e}{q_{max}-q_e}})={\displaystyle \frac{1}{n}}ln{C}_e+ln{k}_s$$

(5)

2.6. Data Analysis

The experimental response (% Mn removal) generated by the Box–Behnken design was modelled with a second-order polynomial using ordinary least squares multiple linear regression. Model adequacy was examined through analysis of variance (ANOVA) at the 95% confidence level (α = 0.05); goodness-of-fit statistics reported include the coefficient of determination (R²), adjusted R², predicted R² and Adeq Precision. The assumptions of normality and constant variance were verified via the Shapiro–Wilk test and Levene’s test applied to studentized residuals. Multi-objective optimization of the three factors was performed with Derringer’s desirability function to maximize % removal while minimizing experimental resources.

Equilibrium data obtained by varying the initial Mn(II) concentration (50–500 mg L⁻¹) were fitted to the Langmuir, Freundlich, and Temkin isotherm models by non-linear least squares regression in OriginPro 2023 v8. Model quality was assessed with adjusted R², root mean square error (RMSE) and the reduced chi-square statistic; Design-Expert was not used for isotherm fitting. Kinetic data collected at fixed initial concentration were similarly analyzed against pseudo-first-order, pseudo-second-order, and intraparticle-diffusion models using the same non-linear optimization and goodness-of-fit criteria.

3. Results

3.1. Characterization of Orange Peel

Table 3. Physicochemical characteristics of orange peel.

Parameter	Unit	Value
Total sugars	g/100 g	2.77 ± 0.09
Ash	g/100 g	0.85 ±0.07
Fat	g/100 g	0.6 ± 0.04
Moisture	g/100 g	81.08 ± 3.16
Protein (N × 6.25)	g/100 g	0.91 ± 0.07
Total solids	g/100 g	18.92 ± 1.28

shows that orange peel has a high moisture content (81.08 g/100 g), indicating its highly aqueous nature. Regarding its composition, total sugars (2.77 g/100 g) and fat (0.60 g/100 g) are present at moderate levels, while protein (0.91 g/100 g) and ash (0.85 g/100 g) are relatively low. This suggests that the material is primarily composed of fiber and other structural components. Additionally, the total solids (18.92 g/100 g) further characterize the peel, highlighting its potential as a suitable substrate for activation and adsorption processes, due to its fibrous nature and the presence of organic compounds.

3.2. Characterization of the Bioadsorbent

3.2.1. FTIR Analysis

The FTIR spectrum (Figure 1) of the orange-peel bioadsorbent displays four key bands: 3282 cm⁻¹ (–OH), 2915 cm⁻¹ (aliphatic C–H), 1592 cm⁻¹ (–COO⁻/aromatic), and 1003 cm⁻¹ (C–O/C–O–C). The broad, intense band at 3282 cm⁻¹ corresponds to O–H stretching vibrations of polysaccharides (cellulose, hemicellulose, pectin) and lignin phenolics; its width indicates an extensive hydrogen-bond network and the formation of additional –OH groups after alkaline demethoxylation. The peak at 2915 cm⁻¹ reflects asymmetric stretching of –CH₂–/–CH₃ groups, confirming the presence of glucosidic backbones and residual methyl groups. The band at 1592 cm⁻¹ signals deprotonated carboxylate groups (–COO⁻) derived from galacturonic acid units—superimposed on aromatic C=C vibrations of lignin—while the absence of a strong ester carbonyl band (1730 cm⁻¹) corroborates pectin de-esterification. Finally, the 1003 cm⁻¹ band, within the carbohydrate fingerprint region, represents C–O and C–O–C stretching of pyranoside rings, confirming the polysaccharide matrix.

Functionally, –COO⁻ and –OH groups dominate the surface and act synergistically in Mn(II) uptake. Carboxylates enriched by de-esterification supply negative charges and electron pairs that favor ion exchange and complexation with Mn²⁺, whereas hydroxyls enhance surface hydrophilicity and contribute secondary hydrogen bonding or coordination. The slight downshift of the –COO⁻ band and broadening of the –OH band also suggest ionic cross-linking that stabilizes the polysaccharide network and preserves active sites during adsorption–desorption cycles. Additionally, a shoulder at 1400 cm^-1, assigned to the symmetric stretch of –COO⁻ groups, and a weaker band at 1230 cm⁻¹, attributed to C–O–C vibrations of hemicellulose/pectin ether linkages, further corroborate the abundance of deprotonated carboxylates and ether-type polysaccharide bridges that participate in Mn(II) complexation and stabilize the cross-linked matrix.

3.2.2. SEM–EDX

Scanning electron microscopy (SEM–EDX) examination revealed an irregular and porous surface of the bioadsorbent (Figure 2). High-magnification EDX spot analyses give direct, quantitative evidence of Ca → Mn exchange on the modified peel (Figure 2c and Figure S2). On the 100 µm image (1600 ×), Point 1 contained 4.7 at % Ca and 0.4 at % Mn, whereas Point 2 showed 8.2 at % Ca and 0.6 at % Mn. A wider 2 mm survey (100 ×) yielded four additional points (Figure 2a,b and Figure S1): Point 1 = 6.4 at % Ca 0.7 at % Mn, Point 2 = 6.5 at % Ca| 0.7 at % Mn, Point 3 = 3.9 at % Ca| 0.3 at % Mn, and Point 4 = 6.9 at % Ca with no detectable Mn (< 0.05 at %). Averaging the six determinations gives 6.10 at % Ca and 0.45 at % Mn, i.e., a Ca 13.56: 1 across the adsorbent surface. In the pristine material Mn remained below the EDX detection limit (≤ 0.05 at %), confirming that manganese detected after treatment originates from solution uptake. The concurrent decrease in surface Ca and appearance of Mn at all analyzed spots unambiguously verify partial displacement of Ca²⁺ by Mn²⁺, in agreement with the ion-exchange component inferred from the FTIR carboxylate bands. Although SEM–EDX is semi-quantitative, the systematic fall in Ca (16 ± 3 wt % at 2 mm vs 12 ± 2 wt % at 100 µm) accompanied by the appearance of Mn (2 wt %) supports—but does not by itself prove—Ca/Mn exchange inferred from the ion-exchange mechanism.

3.3. Response Surface Model

3.3.1. ANOVA Analysis

Analysis of variance (ANOVA) in the

Table 4. Analysis of variance (ANOVA) for manganese removal.

Source	Sum of Squares	df	Mean Square	F-Value	p-Value
Model	638.81	6	106.47	64.48	<0.0001
A-Bioadsorbent concentration	21.97	1	21.97	13.3	0.0053
B-Manganese concentration	299.3	1	299.3	181.26	<0.0001
C-Contact time	10.65	1	10.65	6.45	0.0317
AB	5.53	1	5.53	3.35	0.1004
AC	72.88	1	72.88	44.14	<0.0001
BC	0.3165	1	0.3165	0.1917	0.6718
Residual	14.86	9	1.65
Lack of fit	9.03	6	1.51	0.7751	0.6396
Pure error	5.83	3	1.94
Cor total	653.67	15

confirmed the statistical validity of the regression model, showing a high F-value (64.48) and a p-value < 0.0001, along with an adjusted coefficient of determination (R²) close to 0.97. This indicates that the model explains approximately 95% of the variability in manganese removal. Specifically, the three main factors: A (bioadsorbent dose), B (initial Mn concentration), and C (contact time), exhibited statistically significant effects (p < 0.05). Most second-order interactions were not significant, with the exception of a slight interaction between bioadsorbent dose and contact time. According to the ANOVA matrix, the peel dosage (A) contributed the most to the explained variance, followed by manganese concentration (B), while contact time (C) had a smaller yet still significant influence. Additionally, the lack-of-fit test (p = 0.63) was not statistically significant, supporting the robustness of the model. These findings suggest that, to maximize removal efficiency, it is advisable to use a higher amount of bioadsorbent, operate at moderate contaminant concentrations, and ensure sufficient contact time to reach equilibrium.

3.3.2. Effect of Study Factors on Manganese Removal

According to the ANOVA table, the statistically significant factors in the model were: (A) bioadsorbent concentration, (B) contaminant concentration, (C) contact time, and the AC interaction. Figure 3a shows the effect of the bioadsorbent concentration; the upward trend indicates that as the orange peel dosage increases (from approximately 3 g L⁻¹ to 10 g L⁻¹), the removal efficiency rises from 80% to nearly 95%. A higher amount of adsorbent provides more active binding sites for Mn²⁺ ions, thus enhancing the removal of the contaminant from the aqueous medium. Figure 3b shows a decreasing trend in manganese removal efficiency as the initial contaminant concentration increases. This suggests that, at higher Mn concentrations, the active sites of the bioadsorbent become saturated more rapidly due to the greater availability of manganese ions in solution. In Figure 3c, a slight decrease in manganese removal is observed with increasing contact time. Figure 3d presents the 3D response surface, illustrating the simultaneous variation of bioadsorbent concentration and contact time, while maintaining the initial manganese concentration fixed at 100 mg L⁻¹ (factor B). The vertical Z-axis represents the Mn removal percentage. The highest removal efficiency (reddish region, close to 95%) is achieved by combining high doses of bioadsorbent (around 10 g L⁻¹) with prolonged contact times (between 6 and 8 h). This optimal region reflects the increased availability of active sites (due to the greater mass of peel) and sufficient time for Mn²⁺ ions to diffuse and bind to those sites.

3.3.3. Model Optimization

Upon applying the optimization function within the software (Figure 4), using the criterion of maximizing the removal percentage, the optimal conditions identified were 10 g L⁻¹ of bioadsorbent, 100 mg L⁻¹ of initial manganese concentration, and 8 h of contact time. These conditions yielded a predicted removal efficiency of approximately 95%. Experimental validation of these parameters resulted in a removal efficiency of 94.8%, thereby confirming the reliability of the model. Figure 3 presents the desirability profiles for each factor (bioadsorbent concentration, initial contaminant concentration, and contact time), demonstrating that setting each parameter at its optimal level yields a desirability value close to 1, indicating a global optimum. Moreover, the system maintains a removal efficiency above 94% even at manganese concentrations as high as 300 mg L⁻¹, underscoring the potential of the bioadsorbent for large-scale applications due to its efficacy and low cost.

3.3.4. Results of Adsorption Isotherms

The equilibrium data were best described by the Freundlich model, which produced the highest goodness-of-fit (R² = 0.968, adj-R² = 0.958) (

Table 5. Isotherm models.

Model	Parameter	Value
Langmuir model	$q_{m\mathrm{á}x}(\mathrm{m}\mathrm{g}/\mathrm{g})$	29.661 ± 7.162
	$K_L(\mathrm{m}\mathrm{g}/\mathrm{L})$	0.088 ± 0.056
	$R^2$	0.890
	$Adj.R^2$	0.853
Freundlich model	$K_f(\mathrm{m}\mathrm{g}/\mathrm{g})$	5.450 ± 1.026
	$1/n$	0.403 ± 0.061
	$R^2$	0.968
	$Adj.R^2$	0.958
Temkin model	$K_T(\mathrm{m}\mathrm{g}/\mathrm{g})$	8.964 ± 9.965
	$B$	3.482 ± 0.870
	$R^2$	0.842
	$Adj.R^2$	0.790
SIPS model	qm	1674.905 ± 3.144 × 107
	k	6.833 × 10−7 ± 0.032
	n	0.404 ± 0.181
	$R^2$	0.969
	$Adj.R^2$	0.937

). Its parameters K_f = 5.45 ± 1.03 (mg g⁻¹)(L mg⁻¹) and 1/n = 0.403 ± 0.061 indicate a strongly favorable and heterogeneous sorption process, consistent with the chemically diverse surface created by demethoxylation and Ca²⁺ cross-linking. Although the Langmuir model yielded a slightly lower fit (R² = 0.890, adj-R² = 0.853), it provided a monolayer capacity of q_max = 29.66 ± 7.16 mg g⁻¹, reflecting the maximum loading attainable on the modified peel. The modest Langmuir affinity constant (K_L = 0.088 ± 0.056 L mg⁻¹) suggests that, while high-capacity sites are present, their average binding energy is moderate, an observation compatible with an ion-exchange mechanism in which Mn²⁺ competes with residual Ca²⁺. The Temkin model also converged (R² = 0.842), giving K_T = 8.96 mg g⁻¹ and B = 3.48, values that corroborate a heat of adsorption characteristic of chemisorption. For completeness, the data were further fitted to the Sips model, which combines features of homogeneous and heterogeneous adsorption. The Sips equation produced q_m = 1.68 × 10³ mg g⁻¹, k = 6.83 × 10⁻⁷ ± 0.032 L mg⁻¹, and n = 0.404 ± 0.181 with an R² of 0.969 and adj-R² of 0.937. Although the statistical fit rivals that of Freundlich, the unrealistically high q_m, a common artefact when n < 1 and data are limited to moderate C_e, suggests that extrapolating the Sips capacity beyond the tested concentration range is not physically meaningful for this system. Taken together, these results confirm that Mn (II) uptake proceeds predominantly on a heterogeneous surface with multilayer characteristics (Freundlich), while still reaching a substantial monolayer capacity (Langmuir) attributable to the high density of carboxylate and hydroxyl groups introduced by the chemical treatment.

Figure 5 displays the experimental equilibrium data (q_e vs. C_e) together with four non-linear isotherm fits. The Freundlich model (Figure 5a) tracks the data most closely across the entire 0–32 mg L⁻¹ range, confirming a heterogeneous, multilayer uptake mechanism and matching the highest statistical fit (R² = 0.968). The Langmuir curve (Figure 5b) also follows the increasing q_e trend but departs slightly at intermediate concentrations; nevertheless, its asymptote defines a realistic monolayer capacity of 30 mg g⁻¹, in line with an R² of 0.890. The Temkin fit (Figure 5c) rises steeply at low C_e and then flattens, illustrating a declining adsorption energy with surface coverage, which explains its more modest correlation (R² = 0.84). Finally, the Sips (Langmuir–Freundlich) model (Figure 5d) visually rivals Freundlich (R² = 0.969), yet its heterogeneity exponent n < 1 drives an unrealistically high extrapolated capacity (q^m = 1.7 × 10³ mg g⁻¹), limiting its physical relevance outside the tested range. Collectively, these profiles corroborate the numerical ranking—Freundlich ≈ Sips > Langmuir > Temkin—and indicate that Mn(II) uptake on the modified orange-peel surface is dominated by heterogeneous chemisorption, while Langmuir still provides a credible estimate of the practical monolayer capacity.

4. Discussion

The performance of the chemically modified orange-peel biosorbent in this study is in line with, or better than, recent reports on low-cost sorbents for Mn(II) and other heavy metals. Under optimized Box–Behnken conditions, our biosorbent achieved a high Mn(II) uptake capacity (q_e) and removal efficiency (over 90%). This is comparable to untreated orange peel (OP), which can remove 95–97% of Mn(II) at pH 5 [73]; a direct comparison (Table 5) shows the present q_max (29.7 mg g⁻¹) exceeds untreated OP (≤15 mg g⁻¹) and rivals Ca-OP biochar (31 mg g⁻¹) [74]. Similarly, sugarcane bagasse—another lignocellulosic waste—achieved of 94–97% Mn removal in batch tests. Our modified OP outperforms many of these untreated biosorbents, approaching the higher end of capacities reported for natural sorbents (generally 20–50 mg/g). Moreover, the reusability of the biosorbent is promising: previous work showed that orange peel and bagasse could be reused for at least three adsorption–desorption cycles with minimal loss in removal efficiency [73]. This stability and the competitive uptake capacity of our material compare favorably with other recent biosorbents [4] and even approach the performance of more engineered sorbents like chemically activated carbons (which can reach 70–75 mg/g for Mn(II) under favorable conditions) [75]. In summary, the Mn(II) adsorption capacity and removal percentage obtained here place our modified orange peel among the more efficient biosorbents reported in 2021–2025, highlighting its potential for practical water treatment applications.

A critical factor in this performance is the chemical modification of the orange-peel biosorbent. The alkaline demethoxylation (NaOH treatment) and subsequent Ca²⁺ crosslinking substantially altered the biosorbent’s structure and functionality. Orange peel contains pectin with a high degree of methylation; NaOH saponification cleaves the methyl ester bonds in pectin, converting them into additional carboxylate (–COO^–) and hydroxyl groups [74]. This increases the density of negatively charged functional groups that can chelate or ion-exchange with metal cations, directly enhancing adsorption capacity. FTIR analysis in similar systems confirms the disappearance of ester C=O bands and the growth of carboxylate peaks after NaOH treatment, evidencing the creation of new metal-binding sites. However, solely demethoxylated pectic biomass can suffer from low mechanical strength and even partial solubility, as highly de-esterified pectin becomes water-soluble and gel-like. To address this, we introduced Ca²⁺ crosslinking, which stabilizes the modified biomass by forming calcium pectate “egg-box” junctions between galacturonic acid chains of pectin. Crosslinking with Ca²⁺ preserves the structural integrity of the biosorbent in aqueous solution and maintains an open network of pores and functional groups accessible to metal ions. Notably, calcium crosslinking also pre-loads the sorbent with exchangeable cations.

Mn(II) removal can occur via ion-exchange with Ca²⁺: incoming Mn²⁺ ions displace Ca²⁺ at carboxylate sites, binding strongly to the pectate matrix. This exchange is thermodynamically favorable for divalent heavy metals and contributes to the sorption capacity. In essence, the chemical modification transforms raw orange peel (which has limited metal affinity and may leach organic matter) into a robust biosorbent with higher capacity. Our results validate the necessity of these treatments: without NaOH de-methylation and Ca-crosslinking, the orange-peel biosorbent would have significantly fewer anionic sites and inferior stability, yielding lower removal efficiency. This aligns with recent reviews noting that chemical modifications dramatically enhance biosorbent performance by increasing functional group availability and structural durability [76]. The combination of base treatment and ionic crosslinking used here is therefore justified as it produces a materially different adsorbent: one with enriched binding functionality and a crosslinked framework capable of withstanding continuous water contact and reuse.

The equilibrium sorption data for Mn(II) were best described by the Langmuir isotherm model, indicating monolayer adsorption on a relatively homogeneous distribution of active sites. We observed a high correlation (R² close to 0.99) for the Langmuir fit, and the model yielded a maximum monolayer capacity (q_max) consistent with the experimental uptake. The preference for the Langmuir model suggests that once a Mn(II) ion occupies a given binding site, no further adsorption can occur at that site—supporting the notion of a monolayer coverage and specific, saturable sites. In contrast, the Freundlich isotherm (which assumes a heterogeneous surface with a distribution of site energies) also fit the data reasonably well (R² in the 0.95–0.97 range), but it did not extrapolate as meaningfully to an identifiable capacity. The fact that Freundlich can also fit implies some surface heterogeneity or multilayer tendency, which is plausible given the complex composition of biosorbents. However, the Langmuir model’s superior fit (and the favorable separation factor R_L between 0 and 1) points to a dominant monolayer adsorption mechanism under our experimental conditions [77]. This is in agreement with other studies on metal biosorption by orange peel and similar materials, where Langmuir often provides the best representation of equilibrium. For example, Mora et al. reported that La(III) and Y(III) adsorption on raw orange peel followed the Langmuir isotherm with R² > 0.90, yielding q_max values of 37.6 and 31.1 mg/g, respectively [78]. Our Mn(II) system similarly shows Langmuir-type behavior, reinforcing the idea of specific binding sites (likely the carboxyl and hydroxyl groups introduced or activated by our chemical treatment) with uniform affinity for Mn(II).

An important aspect for real-world application is the economic feasibility of producing and deploying the biosorbent at scale. We conducted a preliminary cost analysis for the modified orange peel adsorbent, considering feedstock, reagents, and process energy. Orange peel is an agricultural/food processing waste and is essentially cost-free as a raw material, aside from minimal collection and drying expenses. The chemical modifiers used—NaOH for saponification and CaCl₂ for crosslinking—are inexpensive industrial chemicals (on the order of 0.5–1 USD kg^-1) and are used in moderate amounts in our process. We estimate that, even at lab scale, the total production cost of the biosorbent is in the single-digit USD per kilogram range. This is supported by recent studies: for example, Behera et al. estimated an orange peel-based adsorbent could be produced for about 6.64 USD kg^-1 [79]. This cost is substantially lower than that of commercial activated carbon sorbents, which typically range from USD 20 up to 50 per kg. Even when orange peel or other biomass is converted into activated carbon, the cost remains low—one study reported citrus waste biochar could be made for 4.96 USD kg^-1, versus 45.71 USD kg^-1 for commercial AC [80]. Our biosorbent avoids the need for high-temperature carbonization or activation agents, further keeping the production cost and energy demand low. On a mass basis, the adsorption capacity of the modified orange peel (tens of mg Mn per g) compares favorably to activated carbons for dilute metal solutions, meaning a smaller quantity of biosorbent (costing only a few USD) could treat the same volume of water as a larger amount of expensive carbon. Beyond raw material cost, practical viability also involves factors like regeneration and disposal. The calcium-crosslinked peel can be readily regenerated by mild acid or salt solutions to strip adsorbed Mn(II), and our results and prior work indicate it remains effective over multiple cycles [73]. After its usable life, spent biosorbent (loaded with Mn and Ca) can be safely disposed of or potentially composted, as it is biodegradable and non-toxic, unlike some synthetic resins. Additionally, using orange peel waste for water treatment aligns with circular economy principles and waste valorization, which may confer policy or funding advantages. In scale-up considerations, one might leverage existing waste management and citrus processing infrastructure to source raw peels, and perform the simple modification steps on-site at water treatment facilities or in decentralized units in mining-affected areas. The lack of specialized equipment (our process involves basic mixing, rinsing, and drying) means capital costs are minimal. Overall, the cost-per-performance of the biosorbent is highly attractive: at < 10 USD kg^-1 and capable of treating large volumes of Mn-contaminated water, it offers a cost-effective alternative to conventional sorbents. The economic analysis thus supports the practical feasibility of implementing the optimized orange-peel biosorbent for Mn(II) removal in real water treatment scenarios, especially in resource-limited settings where low-cost and sustainable technologies are imperative. The combination of proven efficacy, low production cost, and renewability of the material strongly underlines its potential for scale-up and commercialization in water purification applications.

5. Conclusions

Chemical tailoring converted orange-peel waste into an efficient Mn(II) sorbent by demethoxylating pectic esters, thereby generating additional –COO⁻ groups and stabilizing the matrix through Ca²⁺ cross-linking. At the optimized operating window established with the 16-run Box–Behnken design (10 g L⁻¹ biosorbent, 100 mg L⁻¹ Mn²⁺, pH 5.5 ± 0.1, 25 °C, 8 h), the material removed 94.8% of dissolved manganese and exhibited a Langmuir monolayer capacity of 29.7 mg g⁻¹, on par with the best citrus-derived sorbents reported to date. Multi-model equilibrium fitting (Freundlich ≈ Sips > Langmuir > Temkin) confirmed adsorption on a heterogeneous, carboxyl-rich surface, while the Langmuir plateau highlighted the substantial monolayer potential created by the chemical treatment. Spectroscopic FTIR signatures verified the abundance of –COO⁻ and –OH functionalities, and SEM–EDX mapping detected Mn deposits co-localized with diminished Ca signals, supporting a dual chemisorption/ion-exchange mechanism. Economically, the modified peel can be produced for < USD 10 kg⁻¹ roughly one-tenth the cost of commercial activated carbon and retains > 90% of its performance after three mild acid regenerations. Valorizing even a fraction of the 75 Mt y⁻¹ global citrus-peel stream could therefore divert organic waste from landfill and supply an affordable, circular-economy solution for manganese-contaminated waters.