Modified Halophyte Biochar for Congo red Removal: Adsorption and Neural Prediction

Keywords

Congo red - Adsorption - Biochar - Suaeda monoica - Ferrite spinel

Significance

The present work delivers a sustainable wastewater treatment strategy using modified Suaeda monoica biochar. It encourages resource recovery by converting biomass waste into an inexpensive, environmentally friendly dye-removal adsorbent. Process optimization becomes predictive and effective by coupling with ANN modeling. This method provides a scalable, sustainable solution for real world water remediation, while reducing adverse effects on the environment and promoting natural resource reuse.

1

Introduction

Water pollution is a significant environmental issue that is faced worldwide.[1] The world is experiencing a clean water crisis due to increased industries, the citification of rural areas, and globalization.[2] On the other hand, as the world’s population increases rapidly, so do these necessities of clean water for human existence.[3] According to the latest United Nations prediction, the world population has reached 8.12 billion as of July 1, 2024, and will rise to 9 billion in 2037 and 10 billion in 2058.[4] Water scarcity affected 933 million urban people in 2016.[5] Around 1.693–2.373 billion people will live in water-scarce regions by 2050, which is a great number.[5] Water purification is essential for getting clean drinking water and for pharmaceutical, industrial, and medical, chemical purposes.[6] Wastewater contains a variety of wastes, including inorganic pollutants like heavy metals (Pb, Hg, Cd, Cu, etc.), nitrates, phosphates, and colloidal pollutant particles, as well as organic pollutants including dyes, pesticides, herbicides, human faces, oil, animal wastes, drugs, and phenols. Dye is one of the significant pollutants affecting human health, as it is extremely toxic and carcinogenic. About 10⁷ kilograms of dyes are utilized by textile industries with total synthetic dye manufacturing of 7 × 10¹⁰ worldwide annually.[7] Approximately 1.8 × 10⁸ kilograms of dye are released into wastewater annually due to textile dyeing.[8] Besides textiles, dyes are used in paper printing, printing, tanning, leather, and food-coloring industries. So these industries also expel a considerable amount of dye into water. This persistent throwing of dye wastewater into the environment without proper treatment has hazardous consequences.[9] The most immediate effect of dye wastewater is the disturbance in the aquatic ecosystems. It may kill aquatic plants, fishes, and other wildlife, reducing biodiversity. Chemicals and heavy metals in dyes can toxify the food chain causing severe contamination in potable water. Congo red (CR) (C₃₂H₂₂N₆Na₂O₆S₂) is a synthetic azo anionic dye and a pH indicator. It has a brownish-red appearance. It is widely used in the textile industry, paper industry, biological and medical applications, etc. CR dye is potentially considered to be carcinogenic under certain conditions, particularly due to its breakdown products. Its presence in wastewater can harm aquatic ecosystems.

Dyes are very stable and synthesized in a way to combat the degradation, making their repair a major challenge for scientists.[10] Furthermore, nowadays, most of the dyestuff industries make dyes with high levels of fastness, so that the stability and resilience of dyes improve. Dyes with high fastness have high resistance to fade or bleed their color under certain conditions like light, water, rubbing, washing, and perspiration. While the high color fastness of dyes is good for aesthetic and functional purposes, it makes it difficult to treat these dyes from wastewater. There are many methods for the treatment of dyes from wastewater like membrane filtration, biological and chemical oxidation, electrolysis, photocatalysis, adsorption, ion exchange, coagulation/flocculation, and catalysis.[11] [12] [13] [14] [15] [16] The adsorption technique has been proven superior to other techniques owing to its low cost, simplicity, flexible design, lack of harmful by-products, effectiveness even in the presence of toxic pollutants, ease of operation, and eco-friendly approach.[17] Other techniques have limitations such as pore occlusion and membrane clogging as in membrane filtration; production of intrinsic sludge and waste-management issues in coagulation/flocculation as in chemical methods; constraints like spatial demands, poor removal efficiency, and ineffectiveness in addressing stubborn dye components as in biological methods; plant’s pollutant tolerance capacity and the need for extensive land for establishing treatment facilities as in the phytoremediation technique of dye removal; use of costly organic solvents as in ion exchange methods; and high energy expenditures and the generation of secondary products as in the oxidation process of dye removal, etc.[18] [19] Adsorption exhibits significant efficacy in adsorbing dye molecules onto solid surfaces. Additionally, the diverse selection of adsorbents tailored for different water contaminants enhances its versatility. The significant advantage of adsorption is that the adsorbent can be recycled and further used. Simultaneously, through regeneration, we can minimize secondary pollution and also reuse it in alternative applications, which promotes cost-effectiveness and resource restoration.[20] Various adsorbent surfaces are modified by effective chemical and physical methods, which increase the affinity of adsorbate molecules toward adsorbents and increase the surface area of adsorbents. This in turn increases the adsorption capacity of adsorbents.[21] Therefore, adsorption is widely used for wastewater treatment.

Activated carbon is a commonly utilized adsorbent, but its effectiveness is often constrained by challenges such as high production costs, low selectivity, environmental concerns, and complications related to regeneration and reuse. Recently, activated carbon has been widely substituted by biochar due to cost-effectiveness, sustainability, low energy requirements, adequate adsorption for specific applications, versatility in usage, and reduced environmental footprint for adsorption purposes. One study showed that the q _max value for methylene blue dye removal was greater for biochar derived from rice straw than its activated carbon.[22] Biochar is a carbon-rich material produced by the pyrolysis of organic biomass (such as agricultural residues, forestry waste, or animal manure) under limited or no oxygen conditions.

In recent times, the incorporation of ferrite spinels onto biochar is an advanced technique that enhances the adsorption efficiency of biochar in wastewater treatment, i.e., wheat straw biochar–zirconium ferrite nanocomposite (BC–ZrFe₂O₅ NCs) showed higher adsorption capacity for tartrazine dye than the pristine wheat straw biochar.[23] [24] Ferrite spinels (MFe₂O₄, where M = divalent atoms like Cu, Zn, Ni, Co, Mg, Mn) have a face-centered cubic (FCC) structure. They crystallize in either normal or inverse forms, with M²⁺ ions in the tetrahedral sites and Fe³⁺ ions in the octahedral sites in the normal form, and vice versa in the inverse form. They have unique properties like high magnetism, chemical stability, surface active sites, high surface area, mechanical strength, electrical resistivity, and ease of modification. Therefore, they are extensively used in magnetic materials, electronics, catalytic reactions, energy storage, photocatalysis, and wastewater treatment. Recently, biochar–ferrite spinel composites have gained attention as a powerful adsorbent in wastewater treatment due to their excellent combination of desirable properties. Biochar assists in enhancing the durability and stability of spinels, while spinels increase the mechanical and chemical stability of biochar thereby resulting in the improvement of the adsorption performance of the adsorbent. Ferrite spinels enhance adsorption mechanisms through ion exchange, surface complexation, hydrogen bonding, and oxidation–reduction, owing to the presence of their metal ions and -OH groups. Their magnetic properties simplify the recovery process of used adsorbent after multiple cycles.

S. monoica is a succulent plant of the family Chenopodium growing wild in coastal areas, salt marshes, and saline desert regions of the Middle East and South Asia, Africa, the Arabian Peninsula, and Central Asia. It is widely distributed in the coastal areas of India. It is rich in salt. It is a shrubby or herbaceous plant, typically growing up to 1 m in height. Its leaves are fleshy and succulent and are adapted to conserve water. As a halophyte, S. monoica is adapted to high salinity conditions, making it capable of growing where most other plants cannot survive. Some parts of the plants are believed to have medicinal properties including anti-inflammatory and antibacterial. Many research studies have been done on its medicinal properties. Some studies have also been done on its role in soil stabilization and land reclamation, especially in saline areas. Seemingly, no studies have been done on the use of S. monoica in water remediation.

An artificial neural network (ANN) is a group of simple algorithms that are connected to one another and assess the data in response to external input. In a few ways, it is based on biological neural networks. It models complex nonlinear relationships between input and output variables. It assesses the performance using coefficient of determination (R ²) and mean squared error (MSE), where higher R ² and lower error values indicate better prediction accuracy.[25] In adsorption studies, ANNs are trained using experimental input–output data to model the relationship between process variables and adsorption capacity/removal percentage. By minimizing the difference between actual and predicted values, ANNs effectively forecast adsorption behavior without the need for explicit mechanistic equations.[26]

In the present work, we have synthesized biochar (BC300) from the S. monoica leaves powder (LP) followed by its pretreatment to yield base-treated biochar (BTBC). A ferrite composite of biochar (FCOB) was synthesized by incorporation of ferrite spinel NiCuZnFe₂O₄ into BTBC. FCOB was tested as an adsorbent for the removal of CR as a model anionic dye. ANN modeling was performed to analyze and predict the adsorption behavior. It is important to highlight that this is the first research study on the application of modified S. monoica biochar in toxic CR dye removal. So the synthesized adsorbent in the present work is novel and unique, representing a significant contribution to the field of adsorbent materials. The physicochemical characteristics of FCOB were investigated using powder X-ray diffraction (XRD), Fourier-Transform Infrared (FTIR) Spectroscopy, Scanning Electron Microscopy (SEM) with Energy Dispersive Spectroscopy (EDS), Atomic Force Microscopy (AFM), Zeta potential and particle size analysis by Dynamic Light Scattering (DLS), Thermogravimetry (TG) analysis, and the Brunauer–Emmett–Teller (BET) surface analyzer.

2

Results and Discussion

2.1

Characterizations of Samples

2.1.1

Scanning Electron Microscopy (SEM)

[Fig. 1a,b] represent SEM images of BTBC with higher and lower magnification, respectively. [Fig. 1c,d] represent SEM images of FCOB with higher and lower magnification, respectively. It is apparent from SEM images of BTBC that their surface was not flat and smooth, but it was filled with small cavities and rough.[27] FCOB has a smooth appearance. This was because of the incorporation of ferrite spinel particles into the small cavities of BTBC after its modification. The SEM image of FC-CR is shown in Fig. S1. From Fig. S1, it is evident that there is an agglomeration of some small spherical and irregularly shaped particles after CR adsorption onto FCOB. There is a strong deposition of large molecules of CR dye on FCOB through strong interaction, which may yield an agglomerated structure. Thus, before adsorption, FCOB had a smoother surface, and after adsorption, FC-CR has a rougher surface.[28]

Fig. 1 SEM images of BTBC (a, b) at higher (×5000), lower (×350) magnification, SEM images of FCOB (c, d) at higher (×1000), lower (×350) magnification (FEI Nova NanoSEM 450, 20 kV, det = LVD)

The elemental compositions of FCOB and FC-CR were confirmed by EDS analysis ([Table 1]) of FCOB and FC-CR. Their EDS spectra are given in Fig. S2. The presence of Fe, Ni, Cu, and Zn confirms the incorporation of ferrite spinel (NiCuZnFe₂O₄) in BTBC. FCOB shows other minerals like Na, Mg, Si, P, and Ca because halophyte S. monoica accumulates these minerals from the soil.[29] The concentrations of O, Ca, and Fe were decreased, and Cu was diminished from the surface of FCOB after adsorption, which is an indication of their contribution to the adsorption process of CR. The decrease in the percentage of Ca may be due to the cation exchange interaction between Ca ions and CR dye, which may also be the cause of the diminishment of Na, Mg, Al, and Si on FCOB after adsorption.[30] The greater C concentration of FC-CR was due to the binding of the CR dye organic framework on FCOB. The displacement of N-containing functional groups from FCOB by CR dye due to strong interaction led to the diminishment of N on the FC-CR surface.

Table 1 EDS elemental composition (%wt) before (FCOB) and after (FC-CR) CR adsorption

	Percentage (%)
Elements	Before adsorption (FCOB)	After adsorption (FC-CR)
C	59.50	76.9
N	10.96	–
O	23.98	21.8
Na	0.27	–
Mg	0.46	–
Al	0.06	–
Si	0.07	–
P	0.14	–
Ca	0.77	0.4
Fe	2.32	1.0
Ni	0.43	–
Cu	0.60	0.0
Zn	0.44	–

2.1.2

Thermal Stability Investigation-TGA

The TGA curves of BC300, BTBC, and FCOB are shown in [Fig. 2a]. As shown in the TGA curve of BC300, in the first stage of thermal decomposition, BC300 shows around 4.08% mass loss from 24 to 184 °C, attributed to the evaporation of water from it.[31] In the second stage, from 185 to 400 °C, mass loss of 4.10% was observed. It corresponds to the thermal degradation of hemicellulose and cellulose. In the third stage, from 401 to 691 °C, a mass loss of 29.21% was observed. This is attributed to lignin decomposition.[32] At higher temperatures, aromatic carbon in BC300 begins to oxidize releasing CO and CO₂, which leads to mass loss. Residual volatiles are also released at higher temperatures. Some of the undecomposed minerals, organic complexes (e.g., residual cellulose, hemicellulose, lignin, and bioactive compounds) also degrade at the last stage.

Fig. 2 (a) TGA curves of BC300, BTBC, and FCOB (SDT Q600, RT – 700 °C, 5 °C/min, N₂ flow 150 mL/min) and (b) FT-IR spectra plot of FCOB and FC-CR (Perkin Elmer IR spectrometer, KBr pellet, 400–4000 cm⁻¹, resolution 4 cm⁻¹, 32 scans).

As shown in the TGA curve of BTBC, in the first stage (from 27 to 150 °C), 13.38% mass loss is attributed to moisture loss, which is greater than that of BC300. This is because BTBC is an alkaline form of BC300. The alkaline treatment introduces hydrophilic functional groups (–OH, –C=O, –COOH, and –OC=O) onto BTBC, resulting in a higher number of such groups compared to BC300. Alkaline treatment may remove a portion of cellulose, hemicellulose, and lignin but also leave behind some intermediate organic compounds that are still volatile and prone to degradation at moderate temperatures from 151 to 400 °C causing mass loss of 13.38%. In the third stage, from 401 to 691 °C, a high mass loss of 66.43% occurs. This mass loss during this last stage is attributed to the degradation of lignin and residual organic complexes, alkali and other salts, and residual hydrophilic groups.

As shown in the TGA curve of FCOB, in the first stage, from 21 to 150 °C, 12.21% of mass loss occurs due to moisture degradation. Here, in the first stage, the mass loss is slightly lower than that of BTBC because the incorporation of ferrite particles in the BTBC may have blocked its pores inhibiting its ability to retain water through capillary forces. This results in a lower overall water-retention capacity. The second stage, which occurs between 151 and 400 °C, has a mass loss of around 14.68%. Degradation of residual organic intermediate compounds (hemicellulose, cellulose, bioactive compounds) may occur in this stage. Ferrite spinel (NiCuZnFe₂O₄) contributes minimally to the mass loss as they are thermally stable and do not decompose at these temperatures but may influence the degradation of organic compounds due to their catalytic activity. The third and last stage (from 401 to 691 °C) has a mass loss of around 47.24%. This stage is attributed to the decomposition of lignin, ferrite spinel, and residual components (like organic complexes, moisture, salts, and hydrophilic groups), which were undecomposed during lower temperature ranges. At these elevated temperatures, ferrite spinels also oxidize the carbon matrix of biochar into CO and CO₂, yielding mass loss.

2.1.3

Fourier Transform Infrared Spectroscopy (FT-IR)

[Fig. 2b] shows the FTIR spectra plot of FCOB and FC-CR. The FT-IR spectra of LP, BC300, and BTBC are shown in Fig. S3. The FT-IR spectra of BC300, BTBC, and FCOB closely resemble that of the LP, indicating the preservation of inherent properties of S. monoica leaves.

LP shows a broad peak at 3455 cm⁻¹ because of stretching vibrations of the OH group, which justifies the presence of lipids in the leaves. In LP, the peak at 2346 cm⁻¹ is due to atmospheric CO₂ interference.[33] While the same peak in BC300 may also be due to residual CO₂ from the pyrolysis process. The peak at 1641 cm⁻¹ is due to amides containing C=O bond, N-H bending, C=O stretching, or aromatic/conjugated nonaromatic C=C stretching vibration. The earlier studies show that this peak identifies the presence of proteins in the leaves. The peak at 1334 cm⁻¹ corresponds to SO₂ stretching, O–H bending (carboxylic acids, alcohols), or N=O stretching (nitro compounds), which identifies the presence of carbohydrates in the leaves. The peak at 1396 cm⁻¹ is due to COO⁻ symmetric stretching in carboxylate groups from organic acids or their salts, which indicates the presence of fatty acids or amino acids in leaves.[34] The peak at 1110 cm⁻¹ is attributed to C–N stretching (aliphatic amines), C–O stretching (ethers, alcohols), or O–H bending (carboxylic acids), and the peak at 780 cm⁻¹ corresponds to C–N stretching (amines), =C–H bending (benzene) or C–Br stretching vibration.[35] Earlier studies indicate that these two peaks are associated with cell wall components like cellulose, hemicellulose, pectin, lignin, and proteins. Only one peak at 1396 cm⁻¹ is absent in BC300 because COO⁻ groups are removed or transformed to CO₂ and CO gases during the pyrolysis process to form biochar. But all other peaks are similar in LP as well as BC300 because BC300 was prepared at a relatively low temperature of 300 °C, which might preserve most of the functional groups found in LP in it. Thus, both LP and BC300 exhibit similar characteristic peaks in the IR spectrum.

BTBC exhibits all the peaks similar to those of LP, except for the one at 1396 cm⁻¹, which is identical to that of BC300. BTBC lacks a peak at 1396 cm⁻¹ (COO⁻ stretching) because base treatment can also convert carboxyl groups to other forms, such as phenolic hydroxyls or other oxygen-containing groups.

In FCOB, a very broad peak at 3270 cm⁻¹ is due to OH stretching vibration. This peak has a frequency shift to a lower wavenumber compared to LP, BC300, and BTBC. Hydrogen bonding occurs between the hydroxyl group and the NiCuZnFe₂O₄. It causes a shift in the O–H stretching frequency toward lower wavenumbers (e.g., from 3455 to 3270 cm⁻¹). The interaction of hydroxyl groups with Fe/Ni/Cu/Zn ions in ferrite could also involve a change in the electron density around the O–H bond, leading to the downshift in the O–H stretching vibration. The peak at 1594 cm⁻¹ is due to the C=C stretching vibration (aromatic and nonaromatic)/ C=O bond of amide/C=O stretching or N–H bending vibration. This peak is shifted to a lower wavenumber in FCOB than LP, BC300, and BTBC (1641 cm⁻¹) because the functional groups on the biochar (such as C=O, N–H, and C=C) can coordinate or interact with the metal ions of Fe/Ni/Cu/Zn from the ferrite. These interactions often lead to a shift in the vibrational frequencies of the functional groups. Specifically, the C=O, N–H bending, and C=C stretch vibrations may undergo a redshift (lower wavenumber) due to changes in the electron density and bonding characteristics resulting from metal–ligand interactions. The frequency shift is due to biochar–ferrite spinel interaction. The interaction with the ferrite can also affect the C=C stretching vibration of aromatic and nonaromatic structures. If the ferrite composite induces changes in the conjugation or electronic structure of the aromatic system, this can lead to a shift in the stretching vibration of C=C bonds. These two peak shifts further confirm the successful incorporation of ferrite spinel onto BTBC. The peak at 1396 cm⁻¹ contributes to the symmetric stretching vibration of the COO⁻ group. The peak at 1110 cm⁻¹ is likely due to the C–O stretching/C–N stretching or O–H bending vibration. This peak may also reflect the interaction between the composite (NiCuZnFe₂O₄) and the biochar surface.

In FC-CR, the broad peak at 3455 cm⁻¹ is due to OH stretching vibration. The peak at 1641cm⁻¹ indicates the stretching vibration of aromatic ring C=C or nonaromatic C=C/N–H bending/C=O of amide/C=O stretching vibration. This peak is more prominent in FC-CR than in LP, BC300, BTBC, and FCOB. When CR dye adsorbs onto the biochar surface, this peak can become more prominent due to the interaction between the dye and the biochar, particularly at sites with aromatic or oxygenated functional groups on the biochar surface. The peak at 1396 cm⁻¹ in FC-CR is due to C–N stretching vibration or the symmetric bending vibration of C–H bonds in the aromatic structure of the CR dye, which confirms the interaction of CR dye with FCOB through the adsorption phenomenon. This peak may also be due to COO- stretching vibration due to carboxylic acid groups present on FCOB. The peak at 1110 cm⁻¹ is may be due to the sulfonic acid group (–SO₃⁻) of CR dye because, during adsorption, these groups may interact with the ferrite spinel or the biochar’s surface, leading to vibrations characteristic of sulfonic groups in this range (1000–1200 cm⁻¹).[36] This peak may also be due to the C–O stretching/C–N stretching or O-H bending vibration, which may also justify a successful adsorption of CR onto FCOB.

2.1.4

XRD

The crystalline substance exhibits strong, sharp, and well-defined peaks, while the amorphous substance exhibits broad hump-like peaks in the XRD spectrum.[37] The XRD spectrums of BC300, BTBC, and FCOB are shown in [Fig. 3]. In a manner, BC300 contains the crystalline inorganic phase-NaCl ([Fig. 3]). The S. monoica is a halophytic plant that stores high amounts of salt (NaCl) in its tissues. As the plant in the present study was collected from the saline coastal region of Kandla, NaCl was anticipated in BC300. The peaks at 27.36°, 31.67°, 45.51°, 56.54°, 66.23°, and 75.24° (2θ) correspond to crystal planes of NaCl (1 1 1), (2 0 0), (2 2 0), (2 2 2), (4 0 0), and (4 2 0), respectively.[38] NaCl has an FCC crystal structure with an Fm–3m space group. There is an absence of a broad amorphous carbon peak in BC300 because biomass S. monoica has high concentrations of salt (as mentioned above), which may mask the amorphous carbon halo. The base treatment of BC300 dissolves the inorganic crystalline minerals and these minerals are leached out, reducing or eliminating sharp crystalline peaks in the XRD spectrum (XRD spectrum of BTBC). Moreover, NaOH activates the carbon structure of BC300 by creating pores, which in turn increases disorder leading to an increase in amorphous carbon structure. There is a broad peak between 2θ  $\cong {15}^{°}\text{to}{30}^{°}$ in the BTBC XRD spectrum. This indicates the presence of an amorphous carbon structure with randomly distributed aromatic sheets.[39] The peak around $\cong {25}^{°}$ represents the (002) plane of graphitic carbon. This peak is due to the removal of impurities by NaOH, which sharpens the graphitic carbon peak. As seen from [Fig. 3], in FCOB, a broad peak is there between 2θ  $\cong {15}^{°}\text{to}{30}^{°},$which again indicates the presence of amorphous carbon. The XRD pattern of FCOB clearly shows the two most intense characteristic peaks of ferrite spinel NiCuZnFe₂O₄ approximately at 2θ ≈ 35.5° and 62.9°, corresponding to (3 1 1) and (4 4 0) crystal planes, respectively.[40] This further demonstrates that NiCuZnFe₂O₄ was successfully incorporated onto BTBC to yield FCOB. As FCOB has a mole ratio of BTBC: NiCuZnFe₂O₄ 2:1, the absence of additional distinctive peaks of NiCuZnFe₂O₄ in FCOB may be explained by its low concentration in comparison to BTBC. The crystalline size of the materials was calculated using Scherrer’s [Eq. (1)].

$$D=\frac{0.9\times \mathrm{\lambda }}{\text{β cos}\mathrm{\theta }}$$

Fig. 3 XRD plots of BC300, BTBC, and FCOB(PANalytical X’Pert Pro, Cu Kα λ = 0.154 nm, 2θ = 5–80°, 40 mA, 40 kV).

where D, λ, β, and θ represent crystalline size (nm), wavelength (Cu-Kα = 0.15406 nm), full-width at half maxima, and Bragg’s angle (2θ/2), respectively. The average crystalline sizes of BC300, BTBC, and FCOB were 17, 3, and 2 nm, respectively. The crystallographic information is given in Section S1 of the supplementary file.

2.1.5

BET Surface Area Analysis

A Brunauer–Emmett–Teller (BET) graph, also known as a BET plot, is used to measure the surface area of porous solid materials. The nitrogen adsorption–desorption isotherm and BJH adsorption cumulative pore volume curve of FCOB are given in [Fig. 4a,b], respectively. A hysteresis loop in adsorption–desorption isotherms refers to the phenomenon where the amount of gas adsorbed during adsorption differs from the amount desorbed at the same relative pressure (P/P ₀). This loop appears as a mismatch between the adsorption and desorption branches of the isotherm and is characteristic of certain types of porous materials. [Fig. 4a] reveals that FCOB has type IV(a) isotherm with hysteresis type H₃. In type IV(a) isotherm, capillary condensation is accompanied by hysteresis. This isotherm exhibits mesoporous solids in which capillary condensation takes place at higher pressures of adsorbate in addition to multilayer adsorption at lower pressures.[41] The surface area, pore size, and pore volume of FCOB were 44.64 m² g⁻¹ ± 0.2396 m²/g, 7.04 nm, and 0.078 cm³ g⁻¹, respectively. The pore size between 2 nm to 50 nm suggests the mesoporous nature of the FCOB adsorbent. FCOB has a 44.64 m² g⁻¹ surface area, which is in a moderate range. The BET surface area value was calculated following the recommended procedures and avoiding the common pitfalls outlined in the guidelines reported by Johannes W. M. Osterrieth et al. (2022), including appropriate selection of the relative pressure range, verification of the linearity of the BET plot, and adherence to the Rouquerol criteria.[42] The BET surface area was calculated from the linear region of the BET plot in the relative pressure range of 0.05–0.30, following the Rouquerol criteria. The measurement of BET surface area showed high precision, with a standard deviation of only 0.2396, confirming the accuracy and reliability of the obtained BET surface area values.

Fig. 4 (a) BET isotherm linear plot (b) BJH adsorption dP/dV pore volume a,b-Micromeritics 3Flex 3500, degassed at 400 °C, 5 h, 77.29 K), and (c) Zeta potential curve (HORIBA SZ-100, 0.01 wt% in distilled water, pH adjusted to 2–10, 25 °C).

2.1.6

Zeta(ζ) Potential and Particle Size Analysis

The Zeta potential of biochar is the measure of the electrostatic potential at the boundary between the charged particles of biochar and the surrounding bulk solution. It is the result of a build-up of charges near the biochar–bulk solution interface leading to the formation of an electric double layer. pH is an important factor that affects the zeta potential value. The pH at which zeta (ζ) potential value is zero is called as point zero charge (pHpzc)/isoelectronic point. The curve of the zeta (ζ) potential values of FCOB against pH is given in [Fig. 4c]. The curve depicts that FCOB has a negative surface charge in both acidic and alkaline conditions. FCOB showed the least negative value of zeta potential (−1.9 mV) at pH 2. Notably, the highest adsorption of the anionic dye CR also occurred at this pH (Section 2.2.1). It can be attributed to the reduced electrostatic repulsion between the anionic CR molecules and the negatively charged FCOB adsorbent surface due to increased H⁺ ion concentrations at pH 2.

2.1.7

Atomic Force Microscopy (AFM)

While SEM provides detailed information on the overall surface morphology, AFM was further employed to quantify surface roughness at the nanoscale. The surface roughness parameters were calculated using Gwyddion (64-bit) software. The surface roughness parameters like mean roughness (Sa), RMS roughness (Sq), and skew (Ssk) of LP, BC300, BTBC, FCOB, and FC-CR are given in Table S1. The roughness parameter values of FCOB are lower than those of BTBC. This is because of the fact that ferrite spinel particles get embedded into cavities of BTBC, making the structure of FCOB smoother. The SEM imaging (Section 2.1.1) also supports this observation, as BTBC exhibits a channelled morphology with embedded cavities, whereas FCOB is smoother. The AFM images (2D and 3D) of LP, BC300, BTBC, FCOB, and FC-CR are given in [Fig. 5a–e], respectively.

Fig. 5 2D and 3D images of (a) LP, (b) BC300, (c) BTBC, (d) FCOB, and (e) FC-CR (Bruker Dimension Icon, tapping mode, scan area 10 × 10 μm).

2.2

Effect of Different Parameters on the Adsorption Performance of FCOB

2.2.1

Effect of pH

pH is a key parameter in the adsorption process as it changes the structures of dye (CR) and also elevates the surface charge density of the adsorbent material (FCOB). As shown in [Fig. 6a], the adsorption of CR (C ₀ = 50 mg L⁻¹; m = 0.02 g; V = 0.04 L; T = 27 °C; t = 360 min) onto FCOB increased with increase in pH from 2 to 10. The highest dye removal was observed at pH 2, which was ≅ 80.5%. From pH 4 to 8, there is a slight decrease in the R% of CR dye. However, pH 10 shows the lowest removal of CR dye (≅50.20%).

Fig. 6 (a) Effect of pH [m = 0.02 g, t = 360 min, T = 27 °C, V = 0.04 L, C _o = 50 mg L⁻¹], (b) FCOB adsorbent dose [pH = 2, t = 360 min, T = 27 °C, V = 0.04 L, C _o = 50 mg L⁻¹], (c) CR dye concentration [pH = 2, m = 0.02 g, t = 360 min, T = 27 °C, V = 0.04 L], and (d) effect of equilibrium time at temperature 85 °C [pH = 2, m = 0.02 g, T = 85 °C, V = 0.04 L, C _o = 50 mg L⁻¹].

As shown in [Fig. 4c], FCOB has negative zeta potential values in both acidic and alkaline conditions, which may reduce the adsorption of anionic CR dye, but this is not the case here. At pH 2, FCOB has the lowest negative value of zeta potential, which is −1.90 mV. Therefore, there is a greater possibility of adsorption of anionic CR dye at pH 2 rather than higher pH values, because at higher pH, zeta potentials of FCOB have higher negative values leading to higher repulsion between anionic dye and highly negatively charged FCOB surface. Moreover, there is a high number of H⁺ ions at pH 2, which ultimately reduce the repulsive force between the negatively charged FCOB and anionic CR dye. This increases the adsorption of CR onto FCOB through electrostatic interaction.[43]

Another reason for comparatively higher adsorption at lower pH is hydrogen bonding between the protonated functional groups of FCOB and the hydrogen bond acceptor functional groups present in CR dye.

R% of CR decreases with increases in pH. As seen in [Fig. 4c], from pH 7 to 14, there is a decreasing trend in the negative values of the zeta potential of FCOB. So as seen in lower pH conditions, R% should increase with an increase in pH according to zeta pt values but this is not the case here. The increase in the concentration of OH⁻ ions in alkaline conditions overcompensates the decrease in the negative values of zeta potentials in basic pHs.

2.2.2

Effect of Adsorbent (FCOB) Dosage

As seen in [Fig. 6b], FCOB dose has great effect on CR dye adsorption (pH = 2; V = 0.04 L; C _o = 50 mg L⁻¹; T = 27 °C; t = 360 min). The R% effectively increases with an increase in FCOB dose, while adsorption capacity decreases with an increase in FCOB dose. With an increase in FCOB dose, there are more available active adsorption sites and hence more surface area for adsorption of CR dye. Hence, from 20 mg to 35 mg, R% increases from ≅80.5% to ≅94%. Hence, there is no accumulation of FCOB particles with an increase in its dose, which may increase the available active surface area for CR dye adsorption.[44] On the other hand, with an increase in FCOB dose, the same concentration of CR dye is dispersed over a greater mass or area of FCOB, which decreases the number of CR dye molecules adsorbed on the unit mass of FCOB. Hence, adsorption capacity decreases (from 84.52 to 30.53 mg/g) with an increase in FCOB dose from 20 to 70 mg.

At the 50 and 70 mg FCOB dose, R% is 100%, which is the highest dye removal % any adsorbent can have. This means that after the 50 mg FCOB dose, increasing the FCOB dose does not affect the R% of CR dye because of the saturation point. 20 mg of FCOB dose was further optimized.

2.2.3

Effect of Adsorbate (CR) Concentration

As seen from [Fig. 6c], CR concentration has a significant effect on its removal by FCOB (pH = 2; V = 0.04 L; m = 0.02 g; T = 27 °C; t = 360 min). Varying the CR concentrations from 52.48 to 600 mg/L, there is a substantial decrease in R% from 80.20% to 18.30%. This is due to the fact that with an increase in CR concentration, there is a saturation of active adsorption sites on FCOB making it difficult for all CR dye molecules to get adsorbed. Moreover, at higher CR concentrations, there is rivalry between CR molecules for available active sites of FCOB, which makes some CR molecules unadsorbed.

Adsorption capacity (q _e) substantially increases (from 84 to 226 mg/g) with increasing CR concentration (50–600 mg/L). High CR dye concentration provides a higher concentration gradient or driving force between bulk solution and FCOB surface, which ultimately increases q _e.[45] Moreover, at high concentrations of CR, the available active adsorption sites are more efficiently used, which enhances the amount of CR adsorbed per unit weight of FCOB.

As seen from the graph in [Fig. 6c], FCOB shows good R% and q_e at 52.48 mg/L CR concentration. For 600 mg/L, q _e is high but R% is very low and for other CR concentrations, q _e is comparatively higher than 52.48 mg/L, but R% is very low compared to that of 52.48 mg/L CR concentration. As a result, 52.48 mg/L CR concentration was optimized for further adsorption experiments.

2.2.4

Effect of Time and Temperature

[Fig. 6d] depicts the effect of contact time between FCOB and CR dye at temperatures 85 °C having conditions (pH = 2; V = 0.04 L; m = 0.02 g).

The adsorption experiments were also performed at 27 and 50 °C. The detailed information and plots are given in Section S2 and Fig. S4 of the supplementary file, respectively. The equilibrium time for 85 °C was 320 min ([Fig. 6d]). After 320 min, R% and q values remained constant, which were 99.75% and 104.70 mg/g, respectively.

The equilibrium time for CR adsorption decreases with an increase in temperature because the kinetic energy of CR dye molecules increases at high temperatures leading to an increase in turnover number between the CR dye and the FCOB adsorbent active sites, which ultimately slows down the equilibrium time of adsorption. Moreover, we can see that the R% and q_e significantly increased with an increase in temperature from 27 to 85 °C. This indicates the endothermic nature of the adsorption of FCOB on CR.[46]

2.2.5

Leaching Behavior

The UV–visible spectra of leachates collected at various pH values after 320 min (equilibrium time), 700 min and 5 days of contact are shown in Fig. S16. At equilibrium time, no detectable peaks were observed in the pH range of 2–8, confirming the stability of the FCOB composite and the absence of leaching. However, slight leaching was detected only under highly basic conditions (pH 10–13). After 700 min, minor leaching was observed at highly acidic conditions (pH = 2). After prolonged contact (5 days), slight leaching was observed even in the pH 3–8 range. The leaching under extreme pH is likely due to partial degradation of the biochar matrix and/or detachment of NiCuZnFe₂O₄ nanoparticles from the composite surface. Overall, the FCOB composite demonstrates excellent stability for repeated adsorption–desorption cycles within the practical pH range.

2.2.6

ANN Training

Regression plots are valuable tools for assessing the performance of ANN regression modeling. These plots compare the predicted output values generated by the ANN with the actual (target) values from the dataset. A high close R-value (close to 1) indicates that ANN has effectively learned the mapping between inputs and outputs, and the regression plot shows most data points clustered near the diagonal line. For both models (ANN-LM, ANN-BR), MSEs versus the epochs number plot (performance plot) is shown in Figs. S12 and S14, which signifies that method performance did not change notably after 44 and 255 epochs for LM and BR algorithms, respectively. A reduction in the MSE for the training set indicates effective learning from the training data. Optimal performance is achieved at the point where the MSE value is minimized. The regression plots of ANN-LM and ANN-BR models are shown in Figs. S11 and S13. The LM algorithm is faster than the BR algorithm as the LM model requires less epochs to achieve convergence than the BR model. The respective MSE and R-squared values are given in supplementary file (Table S5). The ANN was successfully trained on the experimental dataset to precisely predict the adsorption capacity (q) and estimate the removal efficiency (%) for CR uptake using FCOB. Both ANN-LM and ANN-BR models demonstrate the capability to process new inputs under diverse conditions, including variations in pH, temperature, time (t), FCOB mass, and CR concentration. The comparison plots for actual and ANN predicted data at these conditions are given in Fig. S8. The ANN model developed in Simulink, along with its training dataset, is available on GitHub and has been appropriately referenced.[47]

2.3

Adsorption Isotherm Models and Thermodynamics Parameters

The interaction between adsorbates and adsorbents is described using adsorption isotherm models, which also aid in understanding the adsorption process, surface characteristics, and capacity. In the present work, four adsorption isotherm models were examined: (1) Langmuir isotherm, (2) Freundlich isotherm, (3) Temkin isotherm, and (4) Redlich Peterson isotherm.[48] The specific equations for each model are given in [Eqs. (2)–(5)], respectively.

$$\frac{C_{\mathrm{e}}}{q_{\mathrm{e}}}=\frac{1}{q_{\mathrm{m}}}{C}_{\mathrm{e}}+\frac{1}{K_L{q}_{\mathrm{m}}}\dots .\text{Langmuir linear form}$$

$$log{q}_{\mathrm{e}}=log{K}_{\mathrm{F}}+\frac{1}{n_{\mathrm{F}}}log{C}_{\mathrm{e}}\dots .\text{Freundlich linear form}$$

$$q_{\mathrm{e}}=B_{1}ln{A}_{\mathrm{T}}+B_{1}ln{C}_{\mathrm{e}}\dots .\text{Temkin linear form}$$

$$ln\left(\frac{C_{\mathrm{e}}}{q_{\mathrm{e}}}\right)=Bln{C}_{\mathrm{e}}-lnA\dots .\text{Redlich Peterson linear form}$$

where, q _m = Maximum adsorption capacity (mg/g)

q_t = Adsorption capacity at time t (mg/g)

C _e = Equilibirum concentration of adsorbate (dye) solution (mg/L)

K _L = $\text{Langmuir constant}$ (L/mg)

K _F = $\text{Freundlich constant}$(L^1/n mg^1−1/n g⁻¹)

n _F = Represents the heterogeneity of the adsorption surface and the adsorption intensity. It explains the degree to which the adsorption process is favorable. It is dimensionless.

B ₁ = Heat of adsorption constant (J/mol). Moreover, B ₁=$\frac{\mathit{RT}}{b_T}$…where, b _T = Temkin constant, which is related to the heat of adsorption (J/mol), R = Universal gas constant = 8.31 (J mol⁻1 K⁻¹), T = Absolute temperature (K)

A _T = Temkin equilibrium binding constant (L g⁻¹)

B (β) = indicates whether the adsorption process is closer to Langmuir or Freundlich behavior. It is dimensionless.

A = It is a constant related to the adsorption capacity and energy of adsorption, obtained from the intercept. It has unit of $\frac{\left(C_{\mathrm{e}}^B\right)}{q_{\mathrm{e}}}$. Its unit depends on the value of B.

These parameters of the four adsorption isotherms were calculated from the slope and intercept values of C _e/q _e vs. C _e (Langmuir isotherm plot, [Fig. 7a]), log q _e vs. log C _e (Freundlich isotherm plot, [Fig. 7b]), q _e vs ln C _e (Temkin isotherm plot [Fig. 7c]), and lnC _e/q _e vs. lnC _e (Redlich Peterson isotherm, plot-[Fig. 7d]). The calculated parameters for the three models are given in [Tables 2]. R ² value (0.99) is the highest for the Redlich Peterson model than the Langmuir, Freundlich, and Temkin models, which explains that the adsorption of CR on FCOB can be best described by the Redlich Peterson model. The value of B for the Redlich Peterson model was 0.7468, indicating that the adsorption mechanism is a mix that combines features of both Langmuir and Freundlich models. This suggests a heterogeneous adsorption surface with partial monolayer coverage. The maximum adsorption capacity (q _max) was 239.80 mg/g, which was calculated using the Langmuir model. A considerably high q _max value indicates that FCOB has a high number of adsorption sites available to bind CR dye molecules and FCOB is highly efficient in adsorbing target CR dye molecules. R _L factor (Eqs. (6)) known as the separation factor is a parameter that is derived from the Langmuir model. It helps in determining if the adsorption process is favorable or not. It is a dimensionless quantity. The favorable, unfavorable, linear, and irreversible nature of the adsorption process is identified by 0 < R _L <1, R _L > 1, R _L = 1, and R _L = 0 values. The R _L values for each CR dye concentration were between 0.07 and 0.5 (0 < R _L < 1) indicating favorable adsorption of CR on FCOB.

$$R_{\mathrm{L}}=\frac{1}{1+K_{\mathrm{L}}{C}_{\mathrm{o}}}----\text{where}{C}_{\mathrm{o}}=\text{Initial concentration of}\mathrm{dye}\left(\mathrm{mg}/\mathrm{L}\right)$$

Fig. 7 (a) Langmuir, (b) Freundlich, (c) Temkin, (d) Redlich Peterson, and (e) thermodynamic parameter plots.

Table 2 Adsorption isotherm parameters for adsorption of CR on FCOB and ANN predicted data

Model	Parameter	Experimental	LM	BR
Langmuir	q (mg/g)	239.80 ± 0.0003	239.80 ± 0.0003	236.40 ± 0.0003
	K L (L/mg)	0.021 ± 0.095	0.02 ± 0.095	0.02 ± 0.088
	R 2	0.97	0.97	0.97
Freundlich	K F (mg. g−1 (L mg−1)1/n )	44.77 ± 0.052	43.21 ± 0.052	42.16 ± 0.046
	1/n	0.25 ± 0.025	0.25 ± 0.025	0.26 ± 0.022
	R 2	0.96	0.97	0.96
Temkin	B T (J/mol)	36.80 ± 6.05	36.76 ± 6.04	37.28 ± 6.12
	A T (L g−1)	0.63 ± 28.93	0.64 ± 28.88	0.57 ± 29.34
Redlich Peterson	R 2	0.89	0.89	0.90
	B	0.74 ± 0.028	0.74 ± 0.024	0.73 ± 0.028
	A	44.56 ± 0.13	44.56 ± 0.11	41.14 ± 0.13
	R 2	0.99	0.99	0.99

The value of n_F for CR adsorption was 3.88, which indicates favorable CR adsorption as n _F = 1, n _F > 1, and n _F < 1 corresponding to linear, favorable, and unfavorable adsorption. The higher n _F value (3.88) aligns with the high q _max value (239.80 mg/g). The high values of both these parameters suggest that adsorbent FCOB has great affinity toward anionic CR dye leading to higher removal percentage and adsorption capacity. The R ² (0.89) value is the lowest for the Temkin model, suggesting that CR adsorption by FCOB cannot be well explained by this model. The comparison plots between actual and ANN predicted data for Langmuir, Freundlich, and Redlich Peterson models are given in Fig. S9.

Thermodynamic parameters like free energy change (ΔG°), enthalpy change(ΔH°), and entropy change (ΔS°) were studied to determine feasibility-spontaneity, nature, and randomness of adsorption reaction. ΔG° values for each temperature were derived using the following [Eq. (7)], while ΔH° and ΔS° were determined using the slope and intercept of plot of ln K _e vs. 1/T ([Fig. 7d]), respectively, as in [Eq. (8)]. The thermodynamic parameter plot is shown in [Fig. 7d].

$$\mathrm{\Delta }{G}^{°}=\mathrm{\Delta }{H}^{°}-T\mathrm{\Delta }{S}^{°}$$

$$ln{K}_{\mathrm{e}}=\frac{\mathrm{\Delta }{S}^{°}}{R}-\frac{\mathrm{\Delta }{H}^{°}}{\mathit{RT}}\text{where},K_{\mathrm{e}}=\frac{q_{\mathrm{e}}}{C_{\mathrm{e}}}$$

ΔH° for CR adsorption was positive with a value of 71.02 ± 1.41 kJ/mol indicating the endothermic nature of adsorption. This is also consistent with the increasing R% with increasing temperature. ΔS° was also positive with a value of 0.256 ± 4.34 kJ mol⁻¹ K⁻¹. This indicates increased randomness during CR adsorption on FCOB. The values of ΔG° for 27, 50, and 85 °C were −5.77, −11.66, and −20.62 kJ/mol, respectively. The negative value of ΔG° signifies the spontaneity of CR adsorption on FCOB. This means that adsorption occurs on its own (naturally) without the need for outside energy.

2.4

Adsorption Kinetics

Adsorption kinetics models explain the rate of adsorbate release or retention from an aqueous solution to a solid phase boundary. In the present work, four adsorption kinetics models were examined: (1) pseudo-first-order (PFO), (2) pseudo-second-order (PSO), (3) intra-particle diffusion (IPD), and (4) Elovich model. The equations for each model are given in the supplementary file Section S3.

The kinetic parameters were calculated by the slopes and intercepts of the plots t/q_t vs t (PSO), log (q _e − q_t ) vs. t (PFO), q_t vs. t ^0.5 (IPD), and q_t vs. lnt (EM). The CR adsorption kinetics result at temperature 85 °C with ANN data is summarized in Table S2. The PSO model had a higher R ² value (R ² = 0.99) than that of the PFO model (R ² = 0.82), and the calculated q _e,cal value using the PSO model was 111.11 mg g⁻¹, closer to the experimental q _e = 104.70 mg g⁻¹ value, proving the better fit for the PSO model for adsorption of CR on FCOB rather than the PFO model. From the obtained data, it was concluded that the CR adsorption on FCOB was chemisorption and the rate of adsorption was proportional to the square of the difference between q _e and q_t .[49] The IPD model does not exhibit a single linear line but three discrete linear lines, which indicate that multiple processes controlled CR adsorption on FCOB, each governed by different mechanisms. Moreover, the line did not pass through the origin indicating that intraparticle diffusion was not the sole rate-controlling step. The first initial linear line was observed in the first 50 min of contact between CR and FCOB. Around 64% of CR was adsorbed in this initial stage. In this stage, CR molecules diffuse from the bulk solution to an external surface of the FCOB adsorbent. The second linear line was observed between 51 min and 265 min of the adsorption experiment, and %CR adsorbed in this stage was around 23%, which is less than that of the first stage (K _p1 > K _p2). In the second stage, the rate is controlled by intraparticle diffusion. The third or last stage corresponds to equilibrium and it is between the contact time of 266–500 min. Around 13% of the CR was adsorbed in this stage. The correlation coefficient for CR adsorption on FCOB (R ² = 0.93) was less for the Elovich model than the PSO model indicating that the Elovich model was not the best fit for explaining the CR adsorptionon on FCOB. The plots for PFO, PSO, IPD, and Elovich models at 85 °C temperature are shown in Fig. S5. Their comparison plots with ANN predicted data are given in Fig. S10. All model plots for 27 and 50 °C are given in Figs. S6 and S7, respectively.

2.5

Plausible Mechanisms for the Adsorption of CR by FCOB

The efficient adsorption of CR (q _max = 239.80 mg/g) by FCOB results from the integrated physicochemical properties of BTBC and NiCuZnFe₂O₄ ferrite. From zeta potential measurements, it is evident that FCOB has a negative charge in both acidic and alkaline conditions. Moreover, FCOB showed the highest CR dye removal in highly acidic condition (pH = 2) because CR is an anionic dye and there is an electrostatic attraction between the –SO₃⁻ group of CR dye and the protonated biochar surface at pH = 2.[50] Similar results have been reported by Nitish Semwal et al. The surface hydroxyl groups present on NiCuZnFe₂O₄, which were formed during the coprecipitation synthesis method from aqueous solutions, become positively charged, attracting the anionic CR dye through electrostatic attraction. OH groups present on the FCOB surface form hydrogen bonds with nitrogen atoms (from protonated –NH₂ and –N=N– groups) and the -SO₃⁻ group of CR. The aromatic moieties of biochar have π-electrons that interact with the aromatic ring of CR dye through π–π stacking interactions assisting in adsorption. The FCOB is rich in carboxyl (–COOH), hydroxyl (–OH), phenolic (–OH) groups, metal salts, and metal ions due to the presence of S. monoica biochar and NiCuZnFe₂O₄ ferrite spinel. These functional groups and metal salts/ions of FCOB can coordinate with the N=N and –SO₃⁻ group of CR increasing the adsorption efficiency of it. FCOB surface may also exchange ions (e.g., H⁺, Na⁺, K⁺) with anion (–SO₃⁻) of CR in acidic conditions where the functional groups of FCOB are protonated. This also facilitates CR adsorption on FCOB. As observed from BET analysis, the FCOB surface is mesoporous having a pore size of 7.0429 nm. So CR dye molecules may diffuse into these mesopores leading to physical adsorption. The adsorption mechanism diagram is shown in [Fig. 8].

Fig. 8 Adsorption mechanism of CR by FCOB.

2.6

Recycling Studies of FCOB

FCOB was regenerated and reused five times to analyze its stability and recyclability following its initial use. Firstly, it was regenerated by stirring with 40 mL 0.1 NaOH for 30 min. CR dye was removed immediately after stirring it with the NaOH. Rapid removal of CR from the FCOB surface was attributed to the fact that it showed the least CR adsorption ($\cong$50%) in highly alkaline pH ([Fig. 6a]). The deprotonation of the FCOB surface in high OH⁻ concentration yielded more negative charge on it, which reduced the electrostatic attraction between FCOB and CR dye; hence, the dye desorbed rapidly from FCOB. Thus, the surface charge of the adsorbent and pH of the highest dye removal influence the choice of solvent used for the desorption of dye from the adsorbent surface.

As shown in the [Fig. 9], the R% of CR reduced drastically from 99.7% to 77% in the first recycling process. Progressively, the R% decreased in the second cycle to 71.6%, the third cycle to 68.5%, the fourth cycle to 67.1%, and the fifth cycle to 64.9%. This was due to a strong attractive force between the CR dye and FCOB surface that some of the CR dye molecules were firmly attached to active sorption sites of FCOB even after regenerating FCOB with an efficient desorbing solvent. Desorption efficiency (%) decreases after multiple cycles because the FCOB adsorbent’s active sites become less available due to incomplete removal of the CR dye or gradual structural degradation. Additionally, some CR dye molecules may irreversibly bind to the FCOB adsorbent, reducing its regeneration efficiency over time. The FCOB showed $\cong 65\%$ removal after five regeneration–recycling studies making it a fairly good adsorbent material considering the fact that a readily accessible and cheaper desorbing solvent was used for regeneration of FCOB. The adsorption efficiency (%), desorption efficiency (%), and q _e were calculated from [Eqs. (11), (9), and (12)].

$$\text{Desorption efficiency}=\frac{q_{\mathrm{d}}}{q_{\mathrm{a}}}\times 100$$

Fig. 9 Adsorption performance of FCOB after recycling (adsorption: 20 mg adsorbent, 50 ppm CR, 40 mL, pH 2, 85 °C, 320 min, 250 rpm; desorption: 40 mL 0.1 N NaOH, pH 13, 27 °C, 30 min, 250 rpm).

where, q _d = amount of adsorbate desorbed (mg/g)

q _a = amount of adsorbate initially adsorbed during the adsorption cycle (mg/g).

2.7

Comparison with Other Biochar Adsorbents

Many researchers have investigated the removal of CR by biochar adsorbents. The adsorbent prepared by modifying the orange peel biochar with hexadecyl trimethyl ammonium bromide (CTAB) had a high BET surface area of 618.44 m²/g and a high Langmuir adsorption capacity of 609.8 mg g⁻¹ for CR dye as depicted in [Table 3]. The mesoporous nano-zerovalent manganese (nZVMn) and Phoenix dactylifera leaves biochar (PBC) composite had a very low BET surface area of 57.74 cm²/g. Despite having a low BET surface area, it removed CR efficiently with a q _max of 117.647 mg g⁻¹ ([Table 3]). Thus, the advantages and limitations of some of the biochar adsorbents compared to FCOB are reported in [Table 3].

Table 3 Comparison of various biochar adsorbents utilized in the literature for the removal of CR

Biomass	Adsorbent	q max (mg/g)	Reference
S. monoica leaves	FCOB	239.80	Present work
Orange peel	NOBC	609.80	[43]
Phoenix dactylifera leaves	nZVMn/PBC	117.64	[51]
Corncobs agriculture waste	AMBC	89.30	[52]
Green pea peels	ZnO/GPBC	114.94	[53]
Pine needles	PEI-BC	294.11	[54]
Breadfruit leaf	Biochar made from breadfruit leaves	17.81	[55]
Musa acuminata stem	Biochar prepared from Musa acuminata stem	135.15	[56]
Rice husk	Biochar derived from rice husk	42.91	[57]
Green pea husk	CGPH	2.69	[58]
Sugarcane bagasse	Fe3O4/SBB	74.07	[59]

3

Conclusions

The present work proposes an efficient, eco-friendly, cost-effective biochar adsorbent that offers enhanced toxic CR dye removal. The novel FCOB adsorbent prepared via pyrolysis and coprecipitation method had a negative charge. The examination of structural and morphological properties confirmed the mesoporous structure of FCOB. The excellent fit to the Redlich–Peterson model with a value of B = 0.7468 indicates that the adsorption follows a mixed mechanism involving both Langmuir-type monolayer coverage and Freundlich-type heterogeneous surface interactions. The isotherm studies revealed that FCOB has a high affinity toward CR dye (q _m = 239.80 mg g⁻¹; equilibrium time = 320 min). The adsorption thermodynamics study reveals the endothermic nature of adsorption. PSO kinetics was best fitted for the adsorption of CR on the FCOB surface. The electrostatic attraction is the key factor of high CR adsorption as R% is highest at the pH when the surface is least negatively charged (pH = 2). FTIR studies also reveal the stronger ionic interaction between FCOB and anionic CR dye. Thus, CR adsorption is highly dependent on the surface charge of the FCOB surface. Both BR and LM algorithms accurately predicted removal efficiency and adsorption capacity with high R and low MSE value. The FCOB retained $\cong 65\%$ removal after five regeneration recycle studies making it a profitable adsorbent. FCOB was synthesized by incorporating ferrite spinel into S. monoica biochar making it sustainable compared to other synthetic adsorbents. Thus, FCOB can be utilized in dye industries and other organic sectors for the adsorption and separation of anionic molecules.

4

Experimental Section

4.1

Materials

The information about chemicals used in the experiment is given in supplementary file (Section S3). The halophyte S. monoica (Genus: Suaeda, Species: Monoica, and Family: Amaranthaceae) was collected from Nh141, Kandla Port Road, Gandhidham, Gujarat 370210, India (23.026038°N, 70.192009°E). This saline halophyte is widely located in the intertidal zones of Kandla. Images of biomass from the collection site, dried plant, and LP are shown in [Fig. 10a–c].

Fig. 10 Images of (a) biomass from collection site, (b) dried plant, and (c) leaves powder.

4.2

Synthesis of BC300

The collected biomass was thoroughly washed with distilled water thrice to eliminate all the dust particles and impurities. It was then dried naturally in sunlight for seven days to remove the moisture. Sun drying was preferred over oven drying because it preserves plant tissues’ cellular structure and bioactive compounds with its gentler drying process. In contrast, oven drying may cause the deformation of plant tissues with its rapid drying process, which could impact the research studies on it. The leaves, twigs, stems, and roots were manually separated from the dried plant and cut into small pieces. The small pieces were pulverized into fine powder by a mortar pestle. Afterward, the raw powders of leaves, twigs, stems, and roots were stored in an airtight container. The twigs, stems, and roots were stored for future research. In the present work, LP was further utilized to synthesize biochar (BC300). About 18 g of LP was pyrolyzed at 300 °C for 2 h at a heating rate of 2.50 °C/min without oxygen in a muffle furnace to obtain BC300. The yield of BC300 was 11.90 g. Here the yield of BC was less than the reactant biomass due to loss of moisture, volatile light gases, and breaking of bioactive compounds like protein, resins, tannins, cardiac glycosides, terpenoids, flavonoids, phenols, acidic compounds, and glycosides present in succulent leaves.[60] BC300 was stored in an oven overnight to remove moisture. The BC300 yield can be calculated from [Eq. (10)].[61]

$$\mathrm{BC}300\text{yield}\left(\%\right)=\frac{\text{Mass of}\mathrm{BC}300\left(\mathrm{g}\right)}{\text{Mass of}\mathrm{LP}\left(\mathrm{g}\right)}\times 100$$ 4.3

Synthesis of BTBC

About 11.90 g BC300 was immersed in 1.19 L of 0.10 N NaOH (pH = 13) under continuous mechanical stirring for 4 h as a pretreatment of BC300. Then, the immersion was filtered and washed with distilled water to remove impurities. It was followed by drying the precipitates in the oven at 65 °C overnight. The resulting BTBC was stored in an airtight container. The alkaline pretreatment of BC300 was performed to improve porosity and unclog the partially blocked pores of BC300, which increased the total surface area of biochar and hence the adsorption capacity.[62]

4.4

Synthesis of FCOB

FCOB was prepared by the same method as reported in our previous work.[63] A total of 4 ferrite composites of biochar (FCOB) having different ratios of BTBC: NiCuZnFe₂O₄ were prepared and tested for CR dye adsorption (See supplementary file, Section S6). The best-performing FCOB (F3) (Table S4) had a 2:1 ratio of BTBC: NiCuZnFe₂O₄, so it was selected for bulk synthesis of FCOB. Briefly, 50 mL of distilled water was added to 9 g BTBC, and the mixture was irradiated with an ultrasonic system for 20 min. Totally, 0.03, 0.005, 0.005, and 0.005 moles of Fe(NO₃)₃·9H₂O, Ni(NO₃)₂·6H₂O, Cu(NO₃)₂·3H₂O, and Zn(NO₃)₂·6H₂O metal salts were respectively mixed with 10 mL of distilled water and stirred for half an hour to make a homogenous mixture. Further coprecipitation method was used to make a composite of NiCuZnFe₂O₄ – ferrite spinel and BTBC.[64] The BTBC solution was added to the homogenous mixture of metal salts. NaOH was added dropwise into the mixture until its pH reached 11. Then, the mixture was stirred for 2 h with heating at 60 °C. Afterward, the mixture was filtered and washed with distilled water. The black-colored precipitates were dried in an oven overnight at 60 °C. Then it was calcined at 300 °C for 2 h and dried in an oven. The schematic diagram for the synthesis of FCOB from LP is shown in [Fig. 11].

Fig. 11 Schematic diagram for synthesis of FCOB from LP.

4.5

Characterizations

All the characterizations of LP, BC300, BTBC, and FCOB were done before adsorption, and characterizations of FC-CR were done after adsorption. The information about the characterizations performed is given in the supplementary file (SectionS5).

4.6

Adsorption and Regeneration Experiments of FCOB

Throughout all the adsorption and regeneration experiments, the concentration of adsorbate CR dye was measured using UV–vis spectroscopy (TCC-240A UV–Vis spectrophotometer, by Shimadzu) in the λ range of 200–800 nm. The absorbance of the CR dye solution was measured at λ _max = 497.5nm.

The aqueous solution of CR with 600 mg L⁻¹ was prepared and used as a stock solution throughout the experiment. It was prepared accurately by weighing 0.600 g of CR dye and dissolving it in a small volume of distilled water using a beaker. Once it was fully dissolved, the solution was transferred to a 1 L volumetric flask and diluted to the mark with distilled water and mixed thoroughly to ensure a homogeneous solution. It was stored in a properly labeled container for further use.

A 100 mL conical flask was used for each parameter of the adsorption experiment. The effect of different pH (2, 4, 6, 8, 10) on adsorption was studied at fixed parameters like 20 mg FCOB dose, 50 mg L⁻¹ CR dye concentration, 450 min contact time, and 27 °C temperature. All the necessary pH changes were done using 0.10 N HCl and 0.10 N NaOH and pH was measured using a pH meter (EI Deluxe). The effect of varying amounts of FCOB (20, 30, 35, 50, 70 mg) was also studied at fixed parameters like 2 pH, 50 mg L⁻¹ CR dye concentration, 450 min contact time, and 27 °C temperature. The effect of different CR dye concentrations (52.48, 80, 184, 517, 600 mg L⁻¹) was also studied at fixed parameters of 2 pH, 20 mg FCOB dose, 450 min contact time and 27 °C temperature. The effects of temperature were studied at fixed parameters like 2 pH, 20 mg FCOB dose, and 50 mg L⁻¹ CR dye concentration. The temperature effect study was performed at three different temperatures of 27, 50, and 85 °C for the contact time of 450 min. FCOB was separated from CR dye using an electrical centrifuge machine. The stirring speed of magnetic stirrer was kept constant at 250 rpm throughout all the adsorption experiments.

The initial concentration of the CR dye (C _o, mg L⁻¹) and the final concentration of the CR dye (C _e, mg L⁻¹) were used to calculate the removal percentage (R%) of CR as shown in [Eq. (11)]:

$$R\%=\frac{\left(C_{\mathrm{o}}-C_{\mathrm{e}}\right)}{C_{\mathrm{o}}}\times 100\%$$

where C _o and C _e are in mg L⁻¹ unit.

C _o, C _e, V (volume of CR dye, L), and m (FCOB dose, g) were used in the calculation of adsorption capacity at equilibrium (q _e, mg g⁻¹) as shown in [Eq. (12)]:

$$q_{\mathrm{e}}=\frac{\left(C_{\mathrm{o}}-C_{\mathrm{e}}\right)\times V}{m}$$

Regeneration of the adsorbent is essential because of its sustainability, resource conservation, cost-effectiveness, and environmental control. For regeneration studies, 0.1 N NaOH was used as the desorbing agent. In each cycle, 40 mL of NaOH solution was added to 20 mg of dye-loaded adsorbent (FC-CR). Thus, the desorption was carried out at a pH of approximately 13, as 1 N NaOH was used. The mixture was stirred on a magnetic stirrer (250 rpm) at a fixed temperature of 27 °C for 30 min. After desorption, the adsorbent was filtered, washed thoroughly with distilled water, dried, and reused for the adsorption of CR dye under the same optimized conditions (pH = 2, FCOB adsorbent dose = 20 mg, CR dye concentration = 50 ppm, 85 °C temperature, contact time = 320 min, stirring speed = 250 rpm). Further, CR dye was desorbed from FC-CR (ferrite composite of biochar after CR adsorption) using the same desorption method and reused for CR adsorption. After each adsorption and desorption cycle, FCOB was washed with distilled water. Thus, FCOB adsorbent was regenerated and reused for five cycles.

4.7

Adsorbent (FCOB) Leaching

To assess the potential leaching of particles or matrix components from the FCOB adsorbent composite during the adsorption–desorption process, UV–vis spectroscopy was employed. Solutions with a composite-to-water mass ratio of 0.00075:1 (0.75 mg mL⁻⁻¹) were prepared across a pH range of 2–12, adjusted using 0.1 M HCl or 0.1 M NaOH. The mixtures were stirred for a predetermined period, followed by filtration. UV–vis measurements were conducted with baseline corrections applied to eliminate any interference from peaks originating from the pH-adjusting agents.

4.8

Artificial Neural Networks (ANNs)

ANNs are a machine learning technique that mimic the structure of neurons in the human brain. They consist of input layers, hidden layers, and output layers. By adjusting connection weights during training, an ANN model is created that identifies patterns and trends in the input data.[65] With proper training, ANNs can efficiently capture complex multivariate functions without extensive computations. This capability allows them to quickly interpret intricate, nonlinear relationships within data. The hidden layers are responsible for processing the input data and using it to make predictions. The number of hidden layers varies depending on the complexity of the model.

In this study, the ANN architecture includes five input variables (FCOB mass, CR dye concentration, temperature, time, and pH), 6 hidden layers for predicting the R% output, and adsorption capacity (q). There are two output layers for the adsorption capacity (q) and R%. The relationship between the experimental dataset and the ANN is crucial; the dataset forms the foundation that allows the ANN to learn complex correlations between the inputs and outputs. During training, the ANN understands these correlations, enabling it to predict outcomes for new input values. This results in faster predictions for specific experimental conditions.

The experimental dataset used in this study consists of 44 data points, which are divided as follows: 15% for validation, 70% for training, and 15% for testing. This division ensures that the model learns effectively, provides accurate predictions, and improves its generalization ability. The ANN architecture is represented in Fig. S15.

4.9

Statistical Analysis

All batch adsorption experiments were performed in triplicate, and the results are presented as mean values ± standard deviation (SD) in [Fig. 6] with error bars. Statistical analysis, including calculation of error bars and regression coefficients, was carried out using OriginPro (Version 8.5.) and Microsoft Excel (Version 2507). The isotherm and kinetic model parameters were fitted based on the linear regression method, and the goodness-of-fit was assessed using R ² _adj (adjusted determination coefficient) values. The formulas for R ² _adj and SD are given in the [Eqs. (13) and (14)], respectively. A small value of SD and a value of R ² _adj close to unity indicate a reliable, good curve fit by a model.[66]

$$R_{\mathrm{adj}}^2=\left\{1-\left(1-R^2\right)\right\}\left\{\frac{n-1}{n-P-1}\right\}$$

$$\mathrm{SD}=\sqrt{\left\{\frac{1}{n-p}\right\}\sum _i^n{\left(q_{i,exp}-q_{i,\text{model}}\right)}^2}$$

where, q _i,exp = An individual, i, data obtained from the batch experiment,

q _i,model = The estimation of the corresponding q _i,exp generated by each model,

n = Number of data points, and

p = Number of parameters in the model.

• References

• 1 Srivatsav P, Bhargav BS, Shanmugasundaram V, Arun J, Gopinath KP, Bhatnagar A. Water 2020; 12 (12) 3561
• 2 Goswami L, Kushwaha A, Kafle SR, Kim B-S. Catalysts 2022; 12 (08) 817
• 3 Sirach R, Dave PN. J Hazard Mater Adv 2023; 10: 100300
• 4 World population 2024 – StatisticsTimes.com https://statisticstimes.com/demographics/world-population.php (accessed June 25, 2024)
• 5 Zhong R, Chen A, Zhao D. et al. J Clean Prod 2023; 385: 135740
• 6 Water purification|Description, Processes, & Importance | Britannica https://www.britannica.com/topic/water-purification (accessed June 25, 2024)
• 7 Al-Tohamy R, Ali SS, Li F. et al. Ecotoxicol Environ Saf 2022; 231: 113160
• 8 Chequer FMD, de Oliveira GAR, Ferraz ERA. et al. Eco-Friendly Textile Dyeing and Finishing. IntechOpen; 2013
• 9 Teo SH, Ng CH, Islam A. et al. J Clean Prod 2022; 332: 130039
• 10 Sallam SA, El-Subruiti GM, Eltaweil AS. Catal Lett 2018; 148 (12) 3701-3714
• 11 Gholami M, Nasseri S, Alizadehfard M-R, Mesdaghinia A. Water Qual Res J 2003; 38 (02) 379-391
• 12 Türgay O, Ersöz G, Atalay S, Forss J, Welander U. Sep Purif Technol 2011; 79 (01) 26-33
• 13 Anantha MS, Olivera S, Hu C. et al. Environ Technol Innov 2020; 17: 100612
• 14 Joseph J, Radhakrishnan RC, Johnson JK, Joy SP, Thomas J. Mater Chem Phys 2020; 242: 122488
• 15 Gadekar MR, Ahammed MM. Desalin Water Treat 2016; 57 (55) 26392-26400
• 16 Arciniega Cano O, Rodríguez González CA, Hernández Paz JF. et al. Catal Today 2017; 282: 168-173
• 17 Banu HAT, Karthikeyan P, Vigneshwaran S, Meenakshi S. Int J Biol Macromol 2020; 154: 188-197
• 18 Kathing C, Saini G. Recent Prog Mater 2022; 4 (04) 1-15
• 19 Patel F, Patel P, Patel N, Patel S. Color Removal from Dye Wastewater: A Review. 2014
• 20 Baskar AV, Bolan N, Hoang SA. et al. Sci Total Environ 2022; 822: 153555
• 21 Girish CR. Int J Eng Technol 2018; 7: 330-334
• 22 Sutar S, Jadhav J. Bioresour Technol Rep 2024; 25: 101726
• 23 Environ Pollut 2023; 323: 121318
• 24 Perveen S, Nadeem R, Nosheen F, Asjad MI, Awrejcewicz J, Anwar T. Nanomaterials 2022; 12 (16) 2828
• 25 Zhang J, Fu K, Zhong S, Luo J. Environ Sci Technol 2025; 59 (07) 3603-3612
• 26 Rushdi IW, Hardian R, Rusidi RS. et al. Chem Eng J 2025; 510: 161595
• 27 Lafi R, Montasser I, Hafiane A. Adsorpt Sci Technol. 2018
• 28 Wang J, Tan Y, Yang H, Zhan L, Sun G, Luo L. Sci Rep 2023; 13 (01) 21174
• 29 Ibraheem F, Al-Zahrani A, Mosa A. Plants 2022; 11 (04) 537
• 30 Zein R, Satrio Purnomo J, Ramadhani P, Safni, Alif MF, Putri CN. Arab J Chem 2023; 16 (02) 104480
• 31 Tomczyk A, Sokołowska Z, Boguta P. Rev Environ Sci Biotechnol 2020; 19 (01) 191-215
• 32 Panizio R, Castro C, Pacheco N. et al. Heliyon 2024; 10 (18) e37882
• 33 Norooz Oliaee J, Dehghany M, McKellar ARW, Moazzen-Ahmadi N. J Chem Phys 2011; 135 (04) 044315
• 34 Chermahini ME, Ghiaci M, Chermahini AN, Shirvani M. Heliyon 2024; 10: e38780
• 35 Li X, Wang J, Li S, Li Z, Zheng Z, Zhang Y. Pharmaceutics 2019; 11 (09) 469
• 36 Asker FW, Mahamad ZZ, Eliwei AG, Nief OA. Int J Appl Chem 2017; 13 (02) 169-177
• 37 Ngernyen Y, Petsri D, Sribanthao K, Kongpennit K. RSC Adv 2023; 13 (21) 14712-14728
• 38 Rabiei M, Palevicius A, Dashti A. et al. Materials 2020; 13 (19) 4380
• 39 Hassaan MA, Yılmaz M, Helal M, El-Nemr MA, Ragab S, El Nemr A. Sci Rep 2023; 13: 12724
• 40 Sadhana K, Praveena K, Bharadwaj S, Murthy SR. J Alloys Compd 2009; 472 (01/02) 484-488
• 41 Schneider P. Appl Catal A Gen 1995; 129 (02) 157-165
• 42 Osterrieth JWM, Rampersad J, Madden D. et al. Adv Mater 2022; 34 (27) 2201502
• 43 Hua Z, Pan Y, Hong Q. RSC Adv 2023; 13 (18) 12502-12508
• 44 Rápó E, Tonk S. Molecules 2021; 26 (17) 5419
• 45 Idan IJ, Abdullah LC, Choong TS, Jamil SNABM. Adsorpt Sci Technol 2018; 36 (01/02) 694-712
• 46 Sumanjit, Rani S, Mahajan RK. Arab J Chem 2016; 9: S1464-S1477
• 47 dishamehta2307/ANN-training-dataset. https://github.com/dishamehta2307/ANN-training-dataset (accessed August 11, 2025)
• 48 Wang J, Guo X. Chemosphere 2020; 258: 127279
• 49 Emara AM, Elsharma EM, Abdelmonem IM. J Radioanal Nucl Chem 2025; 334 (01) 227-237
• 50 Semwal N, Mahar D, Chatti M, Dandapat A, Chandra Arya M. Heliyon 2023; 9 (11) e22027
• 51 Iqbal J, Shah NS, Sayed M. et al. J Hazard Mater 2021; 403: 123854
• 52 Faheem, Du J, Bao J, Hassan MA, Irshad S, Talib MA. Arab J Sci Eng 2019; 44 (12) 10127-10139
• 53 Rubangakene NO, Elwardany A, Fujii M, Sekiguchi H, Elkady M, Shokry H. Chem Eng Res Des 2023; 189: 636-651
• 54 Pandey D, Daverey A, Dutta K, Arunachalam K. Environ Monit Assess 2022; 194 (12) 880
• 55 Laxmi Deepak Bhatlu M, Athira PS, Jayan N, Barik D, Dennison MS. Adsorpt Sci Technol 2023; 2023: e7369027
• 56 Jadhav SK, Thorat SR. Biosci Biotechnol Res Asia 2022; 19 (01) 141-151
• 57 Wijaya A, Yuliasari N. Indones J Mater Res 2023; 1 (01) 1-7
• 58 Elsherif KM, Alkherraz AM, Edwards H, Abdulsalam Mutawia BY. Environ Health Eng Manag 2024; 11 (03) 273-284
• 59 Dharmendra G, Sahoo JK, Hota A, Sahoo SK. ECS Trans 2022; 107 (01) 5127
• 60 Lakshmanan G, Rajeshkannan C, Kavitha A, Mekala B, Kamaladevi N. J Pharmacogn Phytochem 2013; 2 (03) 49-152
• 61 Elhenawy Y, Fouad K, Bassyouni M, Al-Qabandi OA, Majozi T. Energy Convers Manag: X 2024; 22: 100583
• 62 Ben Ali M, Bakhtaoui Y, Flayou M. et al. E3S Web Conf 2025; 601: 00087
• 63 Mehta D, Dave PN, Kumar VV. Toxic Crystal Violet Dye Removal by Novel, Eco-Friendly Seablite Biochar–Ferrite Composite: Adsorption Isotherm, Kinetics, and Artificial Neural Network. RSC Adv 2025; 15 (40) 33189-33208
• 64 Fito J, Abewaa M, Nkambule T. Appl Water Sci 2023; 13 (03) 78
• 65 Ghaedi AM, Vafaei A. Adv Colloid Interf Sci 2017; 245: 20-39
• 66 Coconut husk-raw clay-Fe composite: preparation, characteristics and mechanisms of Congo red adsorption | Scientific Reports. https://www.nature.com/articles/s41598-022-18763-y (accessed August 6, 2025)