**3. Post-Production Modification of Biochar for Dye Removal**

Biomass breaks down thermally in an oxygen-limited environment to generate BCs composed of numerous refractory oxides based on the feedstocks such as CaCO3, Fe2O3, SiO2, Al2O3, and CaSO<sup>4</sup> [78]. Moreover, pristine BC surfaces are generally negatively charged and linked with oxygen-bearing functional groups, thus displaying specific adsorption towards cations (e.g., heavy metals ions) [79]. However, adsorption of anionic species (i.e., oxyanion, anionic dyes, and organics) of BC is inadequate [80]. This restricts BC applications in various fields, instigating scientists to introduce novel metal oxides to alter BC for targeted applications [81]. Table 1 enlists few literatures available onto the treatment of dye-contaminated wastewater using BC and its nanocomposites. Table 1 enlist some literature available on the BC utilization for dye removal from wastewater.


**Table 1.** Literature summary on modified biochar for dye removal.


**Table 1.** *Cont.*

Additionally, conventional pyrolysis generates BC with inefficient physicochemical properties such as surface area, surface-oxygenated groups, pore volume, and width. Different feedstock types have different BC production conditions and physicochemical properties [89]. However, the existing -COOH, -OH, and C=O functional groups are crucial in water treatment [90]. The hydrophilic or hydrophobic properties depend on the functional groups (nature and type) present on the BC surface. Various methods such as impregnation, steam activation, and chemical and heat treatments have been utilized to improve the abundant BC properties [91]. In the presence of other catalysts, BC acts as a photocatalyst due to the physicochemical properties (enhanced surface area, active site, charge separation, porous volume, functional groups, catalyzing ability, stability, and recoverability) [92]. However, photocatalyst-based BC production needs to be optimized [93]. For instance, the addition of ZnO to BC improves the nanocomposite adsorption capability and photocatalytic ability [94]. Figure 4 shows various modification and classification of engineered BC for the dye removal from wastewater. *Catalysts* **2022**, *12*, x FOR PEER REVIEW 14 of 29

**Figure 4.** Modification and classification of engineered biochar. acids and fatty acids increases the hydrophobicity of BC adsorbents [111,112]. **Figure 4.** Modification and classification of engineered biochar.

*3.1. Acid-Base Activation/Decoration* 

[99].

to be performed [102].

tion of methylene blue [108].

creased colloidal stability and mobility [98]. After BC oxidation, -OH and -COOH functional groups are enhanced to improve hydrophilicity. Acids such as HNO3/H2SO4, HNO3, HF/HNO3, KMnO4, H2O2, oxalic, and citric acids are used for chemical modification of BC

One of the approaches for the modification of BC is HCl treatment resulting in higher adsorption of concomitant organic compounds. This is because functional groups (carboxylic, phenolic, and carbonyl groups) are added to the BC surface after treatment, the ash content is reduced, and the adsorption sites (hydrophobic) are increased [100,101]. However, this method is not well established for BC modification and more studies are needed

The use of strong bases for BC modification increases functional groups, surface area, surface charge, and porosity [103]. However, the degree of improvement in BC properties varies depending on the base used. Frequently used bases include potassium oxide (KOH) and sodium hydroxide (NaOH) [104,105]. Few studies showed that the BC surface is significantly improved when using NaOH rather than KOH. For instance, Cazetta et al. [106] reported a better improvement in the surface area (49%) of coconut shell-derived BCs when using NaOH. Wakame-derived BC showed a surface area of 69.7 m2 g−1 and 1156 m2 g−1 for non-activated BC and KOH-activated BC, respectively [88]. Similarly, the BC surface area derived from rice husk improved from 132.9 m2 g−1 for non-activated BC to 1818 m2 g−1 for KOH-activated BC [107]. BC activation by KOH and KMnO4 results in large pore volume, pore channels, and percentage of aromatized structures, enhancing the adsorp-

In general, modification of BC via acid is carried out in two ways: activation and decoration. BC activation via chemical activating agent increases SSA and porous structure. In contrast, decoration enhances the surface activity of BC via zero-valent iron nanoparticles [109], Fe3O4, and FeOOH [110]. Studies have also shown that decoration using

Post-modification can be categorized into the following groups: (i) modifications to surge surface area and porosity; (ii) changes to decrease the positively charged BC surface; (iii) increase surface oxygen-containing functional groups; and (iv) magnetization of BC to enable recovery. Among all categories, any method can improve one or two BC properties simultaneously. For instance, the SSA and oxygen-containing functional groups can be increased by H2SO<sup>4</sup> treatment. In the literature, these modifications are classified based on BC physicochemical characteristics or the nature of the process (chemical, physical, or composite) [95,96].

Furthermore, physical (van der Waals forces, electrostatic forces, and hydrophobic interactions) and chemical (surface complexation, ion exchange, π-interactions, coprecipitation, partition, and pore filling) interactions occur during adsorption of organic and inorganic pollutants by BC [97].

#### *3.1. Acid-Base Activation/Decoration*

Acid treatment of BC results in enhanced surface area, better porosity, increased sorption capacity, created charged and hydrophilic surface functional groups, and increased colloidal stability and mobility [98]. After BC oxidation, -OH and -COOH functional groups are enhanced to improve hydrophilicity. Acids such as HNO3/H2SO4, HNO3, HF/HNO3, KMnO4, H2O2, oxalic, and citric acids are used for chemical modification of BC [99].

One of the approaches for the modification of BC is HCl treatment resulting in higher adsorption of concomitant organic compounds. This is because functional groups (carboxylic, phenolic, and carbonyl groups) are added to the BC surface after treatment, the ash content is reduced, and the adsorption sites (hydrophobic) are increased [100,101]. However, this method is not well established for BC modification and more studies are needed to be performed [102].

The use of strong bases for BC modification increases functional groups, surface area, surface charge, and porosity [103]. However, the degree of improvement in BC properties varies depending on the base used. Frequently used bases include potassium oxide (KOH) and sodium hydroxide (NaOH) [104,105]. Few studies showed that the BC surface is significantly improved when using NaOH rather than KOH. For instance, Cazetta et al. [106] reported a better improvement in the surface area (49%) of coconut shellderived BCs when using NaOH. Wakame-derived BC showed a surface area of 69.7 m<sup>2</sup> g −1 and 1156 m<sup>2</sup> g −1 for non-activated BC and KOH-activated BC, respectively [88]. Similarly, the BC surface area derived from rice husk improved from 132.9 m<sup>2</sup> g −1 for non-activated BC to 1818 m<sup>2</sup> g −1 for KOH-activated BC [107]. BC activation by KOH and KMnO<sup>4</sup> results in large pore volume, pore channels, and percentage of aromatized structures, enhancing the adsorption of methylene blue [108].

In general, modification of BC via acid is carried out in two ways: activation and decoration. BC activation via chemical activating agent increases SSA and porous structure. In contrast, decoration enhances the surface activity of BC via zero-valent iron nanoparticles [109], Fe3O4, and FeOOH [110]. Studies have also shown that decoration using acids and fatty acids increases the hydrophobicity of BC adsorbents [111,112].

#### *3.2. Persulfate Activation*

Lately, the activation of BC via persulfate (PS) has been extensively studied [113] owing to the enhanced contaminant removal efficacy from water by the combination of PS radicals with BC-SSA [114]. Studies have shown that the sulfate radical-based advanced oxidation processes (AOP) have higher redox potentials, extended half-life time, and extensive pH flexibility than traditional hydroxyl radical-based AOP. Liu et al. [115] studied food waste digestate-derived BC (FWDB) for peroxydisulfate (PDS) catalyst (radical-based oxidant) for the removal of azo dye (reactive brilliant red X-3B). They observed a 92.21% removal of X-3B within 30 min from solution at an initial content of 1 g/L. The reactive oxygen species SO<sup>4</sup> •−, •OH, O<sup>2</sup> •−, and <sup>1</sup>O<sup>2</sup> were found to be present in the FWDB/PDS system and the generation of reactive oxygen species is due to the presence of graphitized carbon, doped-N, oxygen-containing groups, and the defective sites on the FWDB surface. Onestep sol–gel pyrolysis was used to synthesize CuFe2O4@BC composite (CuFe2O4@BC) and assessed for the efficiency, stability, and mechanism of activating PS against malachite green degradation [116]. The malachite green removal efficiency in the CuFe2O4@BC/PS system was observed to be 98.9%, signifying that CuFe2O4@BC has better catalytic activity compared to other catalysts under the optimal conditions.

A composite of stalk BC (SBC) and nanoscale zero-valent iron (nZVI) was biosynthesized to remove dye. The composite displayed a synergetic role in enhancing PS catalytic activity for dye removal [117]. SBC functional groups (-OH and -COOH) stimulated PS and nZVI improved catalytic activity compared to SBC. SBC enhances catalytic activity for PS and electron transfer efficiency by increasing nZVI dispersibility and protecting it from oxidation. Studies have shown that superoxide radicals (O<sup>2</sup> •−) and hydroxyl radicals ( •OH) play an important part in pollutant removal, but they are not primary radicals. The primary radical involved in contaminant removal was sulfate radicals (SO<sup>4</sup> •−). The SBCnZVI composite also displayed better reusability, storage stability, and various dye removal applications [117]. One-step pyrolysis of loofah sponge-based BC (LSB) was prepared at various pyrolytic temperatures. LSB was further used to activate PS to remove acid orange 7 (AO7) [118]. The LSB produced at 800 ◦C showed the best catalytic ability for AO7 degradation, i.e., 96% degradation within 30 min [118]. The enhanced adsorption capability of LSB is due to the high surface area, carbonyl groups, and graphitized structure. The AO7 degradation mechanism in the LSB-800/PS process is owing to free radical SO<sup>4</sup> •– and non-radical channels (electron transfer).

#### *3.3. Physical Activation*

In this process, different gases (air, steam, and CO2) are used to activate the BC surface [119]. In general, physical activation consists of two stages: carbonization (occurs at lower temperatures of 427–877 ◦C) and activation (occurs at higher temperatures of 627–927 ◦C) [120]. BC activation can be carried out via steam or gas purging after BC is carbonized. Although this process is simple and cost-effective, it is less efficient than chemical activation approaches. BC physicochemical properties can be enhanced via physical activation by eliminating incomplete combustion by-products and the carbon surface oxidation to enhance BC surface area [121,122].

Pristine BC is activated by steam with improved polarity, surface area, and pore volume [123]. After preliminary pyrolysis, BC was partially vaporized by steam for 0.5 and 3 h [124]. Steam activation improves BC growth and the reachability of internal pores [124]. During steam activation, pore development is primarily associated with water-gas shift reaction, reduction of carbon [125], and elimination of trapped components (aldehydes, ketones, and some acids from biomass) [126].

In the presence of heated steam, BC is activated and displays high adsorption ability owing to improved surface area, oxygen-containing functional groups, and micropore volume. This allows for efficient removal of pollutants from wastewater [127]. The porosity of BC can be improved by physical activation at <250 ◦C in the presence of air or >750 ◦C in the presence of CO2, steam, and blends. CO<sup>2</sup> and H2O become reactive at high temperatures, allowing chemical reaction with BC [128]. CO<sup>2</sup> treatment increases the pore volume and surface area of BC by 1.5-fold and 5.9-fold, respectively [129]. At low temperatures (≈275 ◦C), the activation process with air increased the sewage sludge-derived BC structure and texture [130]. Furthermore, Sewu et al. [131] observed an increase in BC SSA and porosity via steam activation and significantly improved the crystal violet adsorption capacity by 4.1-fold.

#### *3.4. Biochar-Based Composites*

BC-based composites allow for the combined advantages of BC and nanomaterials, resulting in composites with enhanced functional groups, pore size and volume, active surface sites, net surface charge, ease of recovery, and catalytic ability for degradation [132,133]. These composites are divided into three groups according to the synthesis process: (i) nanometal oxide/hydroxide-BC composites, (ii) magnetic BC composites [80], (iii) clay mineralbased BC [134].

#### 3.4.1. Nano-Metal Oxide/Hydroxide-Biochar Composites

These composites have been synthesized via numerous chemical additives such as H3PO4, NaCl, K2CO3, ZnCl2, KOH, and FeCl3, and can be incorporated prior or later biomass carbonization [89]. According to Zhao et al. [135], the synthesis of nano-metal BC composites can be divided into the following categories: (i) impregnation, categorized by the ease and enormous capacity of the attained composites, (ii) chemical coprecipitation, categorized by inexpensiveness, high purity, homogeneous nanoparticles, but both process causes chemical contamination, (iii) direct pyrolysis, considered as easy process and biomass supplemented with desired heavy metals; however, controlling the nano-metal oxide/BC ratio in the composites is difficult, and (iv) ball milling, simple, cost-effective, no generation of chemical pollutants, and efficient decrease of metal oxide particle size. However, during particle preparation via ball milling, the particles are easily dispersed in water, resulting in the migration of contaminants out of the polluted place, causing a possible hazard to groundwater [135]. Figure 5 represents various pathways for synthesis of nano-metal oxide/hydroxide-BC composites. *Catalysts* **2022**, *12*, x FOR PEER REVIEW 17 of 29

**Figure 5.** Various pathways for synthesis of nano-metal oxide/hydroxide-biochar composites. Reprinted with permission from [32]. **Figure 5.** Various pathways for synthesis of nano-metal oxide/hydroxide-biochar composites. Reprinted with permission from [32].

Sludge-derived BC was modified via Fe/Mn by a heterogeneous activator to attain enhanced PS activation to degrade reactive blue 19 (RB19) [140]. Experiments based on scavenger quenching and electron paramagnetic resonance displayed that radical and non-radical mechanisms are involved in the degradation of RB19 by Fe/MnBC combined PS. Sulfate radicals (SO4•−), hydroxyl radicals (OH•), and singlet oxygen (1O2) were found to be involved in the process. It was also postulated that Fe(IV)/Mn(VII) (non-radical) was involved in the degradation process. These results provided a new understanding of the Preparation of BC sample via ball milling surges SSA, possibly by revealing pores [136] and enhances carboxyl, lactone, and hydroxyl groups (oxygen-containing functional groups) [137]. Zheng et al. [138] synthesized a dual-function MgO/BC nanocomposite by ball milling to remove both cationic and anionic contaminants and showed an increased (8.4-fold) methylene blue adsorption, probably due to enhanced surface area and pore volume. Synthesis of BC by co-pyrolysis from firwood biomass utilizing non-magnetic goethite mineral proved to be a green method in comparison to the conventionally used FeCl<sup>3</sup> [139].

mechanism of PS activated by metal-BC composite. In addition, fixed-bed reactor studies revealed that Fe/MnBC has significant PS activation capability to remove dyes. Further, the authors analyzed degradation process by central composite design-response surface

BC derived from pulp sludge was produced via ZnCl2 modification to effectively remove methylene blue (MB) from aqueous solutions [141]. The maximum (590.20 mg/g) adsorption of MB was observed within 24 h at pH 8 when the initial amount of Zn2PT350- 700 was 10 mg [141]. The primary mechanism involved in MB adsorption was electrostatic interaction between deprotonated functional groups and MB+, cation exchange, and π-

Substantial studies have been conducted to enhance BC adsorption capacity. However, the biggest limitation of BC is its removal and reuse after wastewater treatment owing to its small size and low density [59]. Various studies of magnetic BC for successful production have been reported [142], displaying impregnation-pyrolysis, chemical coprecipitation, solvothermal, and reductive co-precipitation processes. It was observed that the composites have a significantly efficient adsorption, easy removal and recycling after application of an external magnetic field [143]. However, the conventional magnetic medium loading process surges the sorbent cost [144]. The primary common processes of magnetic material synthesis include pyrolysis, co-precipitation, and calcination [145]. Fig-

electron interactions followed by physical adsorption.

ure 6 shows a brief schematic of different magnetic BC preparation.

3.4.2. Magnetized Biochar Composites

Sludge-derived BC was modified via Fe/Mn by a heterogeneous activator to attain enhanced PS activation to degrade reactive blue 19 (RB19) [140]. Experiments based on scavenger quenching and electron paramagnetic resonance displayed that radical and non-radical mechanisms are involved in the degradation of RB19 by Fe/MnBC combined PS. Sulfate radicals (SO<sup>4</sup> •−), hydroxyl radicals (OH• ), and singlet oxygen (1O2) were found to be involved in the process. It was also postulated that Fe(IV)/Mn(VII) (non-radical) was involved in the degradation process. These results provided a new understanding of the mechanism of PS activated by metal-BC composite. In addition, fixed-bed reactor studies revealed that Fe/MnBC has significant PS activation capability to remove dyes. Further, the authors analyzed degradation process by central composite design-response surface methodology (CCD-RSM) and ANN. Statistical analysis showed that the ANN model was better than the CCD-RSM model [140].

BC derived from pulp sludge was produced via ZnCl<sup>2</sup> modification to effectively remove methylene blue (MB) from aqueous solutions [141]. The maximum (590.20 mg/g) adsorption of MB was observed within 24 h at pH 8 when the initial amount of Zn2PT350-700 was 10 mg [141]. The primary mechanism involved in MB adsorption was electrostatic interaction between deprotonated functional groups and MB<sup>+</sup> , cation exchange, and π-electron interactions followed by physical adsorption.

#### 3.4.2. Magnetized Biochar Composites

Substantial studies have been conducted to enhance BC adsorption capacity. However, the biggest limitation of BC is its removal and reuse after wastewater treatment owing to its small size and low density [59]. Various studies of magnetic BC for successful production have been reported [142], displaying impregnation-pyrolysis, chemical co-precipitation, solvothermal, and reductive co-precipitation processes. It was observed that the composites have a significantly efficient adsorption, easy removal and recycling after application of an external magnetic field [143]. However, the conventional magnetic medium loading process surges the sorbent cost [144]. The primary common processes of magnetic material synthesis include pyrolysis, co-precipitation, and calcination [145]. Figure 6 shows a brief schematic of different magnetic BC preparation.

Further, traditional heating in electric furnaces [146] and microwave heating in modified furnaces have also been studied. Among them, the co-precipitation method is easy because mixing and heating processes are simple. It emphasizes the transition metal ion molar ratios used in mixing to improve magnetic BC characteristics concerning porosity and magnetism [145]. The co-precipitation process is used for BC surface coating with various iron oxides such as Fe3O4, c-Fe2O3, and CoFe2O<sup>4</sup> particles to impart magnetic characteristics to the iron oxide active sites for the pollutant removal [147,148]. It was observed that the high decomposition ability of additives such as AlCl3, FeCl2, and MgCl<sup>2</sup> displays enhanced porosity, high catalytic and sorption activity, and creation of positively charged adsorption sites [89,149].

Fe-tanned collagen fibers were pyrolyzed to synthesize magnetic BC catalyst (Fe@BC) utilized as a persulfate activator to remove a refractory dye (methylene blue) [150]. Results have shown that within 20 min, methylene blue was completely degraded at a constant rate of 0.2246 min−<sup>1</sup> , much higher than when using pure BC (0.0497 min−<sup>1</sup> ). The enhanced catalytic activity and exceptional recycling ability of the Fe@BC/PS system can be attributed to the homogeneously dispersed Fe ions, enhanced functional groups (oxygen containing), and deformed carbon matrix. SO<sup>4</sup> •− was found to be the primary free radical for the degradation of methylene blue. Table 2 enlists literature on surface-modified BC and its nanocomposites for dye removal from wastewater.

adsorption sites [89,149].

Further, traditional heating in electric furnaces [146] and microwave heating in modified furnaces have also been studied. Among them, the co-precipitation method is easy because mixing and heating processes are simple. It emphasizes the transition metal ion molar ratios used in mixing to improve magnetic BC characteristics concerning porosity and magnetism [145]. The co-precipitation process is used for BC surface coating with various iron oxides such as Fe3O4, c-Fe2O3, and CoFe2O4 particles to impart magnetic characteristics to the iron oxide active sites for the pollutant removal [147,148]. It was observed that the high decomposition ability of additives such as AlCl3, FeCl2, and MgCl2 displays enhanced porosity, high catalytic and sorption activity, and creation of positively charged

**Figure 6.** Brief schematic of different magnetic biochar preparation. Reprinted with permission from [32]. **Figure 6.** Brief schematic of different magnetic biochar preparation. Reprinted with permission from [32].




#### **Table 2.** *Cont.*

## 3.4.3. Modification via Clay Mineral

Pre-mixing of the feedstock prior to pyrolysis or post-mixing of the generated BC and clay minerals produces a clay-BC composite [95]. Arif et al. [134] studied the clay mineral-BC synthesis process and characteristics including pollutant interactions with the composite. This study displayed that clay-mineral amendment increases BC micropore area (>200%), whereas slight growth was detected in the mesopore area [163]. BC stability can also be enhanced by clay organo-mineral layers and avoid BC decomposition when applied to soil [164].

The interaction of the BC-clay/mineral composites with the pollutants involves various interactions such as ion exchange, precipitation, complexation, electrostatic interactions, hydrogen bonding, partitioning, electron–donor–acceptor interactions, hydrophobic interactions, and pore-filling [165]. Layered porous clay-BC (C-BC) was produced via layered bentonite clay intercalating properties comprising high cation exchange capacity with BC supporting redox-sensitive zero-valent iron (nZVI) nano-trident particles [166]. The synthesized C-BC-nZVI displayed a heterogeneous pore distribution on the C-BC-nZVI surface with dispersed un-oxidized nZVI particles (20–30 nm). Methylene blue was selectively adsorbed (52.1 mg/g) by nano-trident in the dye mixture. Further, pH and dissolved organic matter showed no influence on methylene blue sorption [166].

## **4. Application of Machine Learning and Artificial Neural Networks into Biochar-Facilitated Wastewater Remediation**

The process initiates from BC production. Various techniques are then carried out to ensure the continued use of BC for wastewater treatment until it is landfilled or recycled. Precisely, the produced BCs are characterized as SSA, elemental composition and pH, used as input and aided in implementation [36]. Machine learning (ML) algorithms such as support vector regression, ANN, or random forests (RF) are applied at this stage. The data is processed via algorithm processes and the variables are transformed into files

using weights [167,168]. The decision criterion is made from the index to choose BC for wastewater treatment. After BC selection, it is passed on for wastewater treatment, where it is otherwise activated or functionalized by chemical or physical approaches. BC is recharacterized after activation/functionalization, and ML algorithms reprocess the data to determine whether the functionalized BC fulfils the standards required for wastewater treatment [36,169]. Moosavi et al. [170] compared the model performance of activated carbon derived from agricultural waste to remove dyes using gradient boosting, decision tree, and RF models. Model accuracy can be evaluated as correlation coefficient (*R*), mean squared error (*MSE*), and root mean squared error (*RMSE*) according to the following equations:

$$R = \sqrt{1 - \frac{\sum\_{l=1}^{N} (\mathcal{Y}\_l - y\_l)^2}{\sum\_{l=1}^{N} (\mathcal{Y}\_l - \overline{y})^2}},\tag{1}$$

$$MSE = \frac{1}{N} \sum\_{l=1}^{N} (\check{y}\_l - \overline{y})^2 \,\text{.}\tag{2}$$

$$RMSE = \sqrt{MSE}.\tag{3}$$

The linear dependences amid any two variables or each attribute and the target variable are calculated by the following equation.

$$r\_{xy} = \frac{\sum\_{i=1}^{n} (\mathbf{x}\_i - \overline{\mathbf{x}}) \ \sum\_{i=1}^{n} (y\_i - \overline{y})}{\sqrt{\sum\_{i=1}^{n} (\mathbf{x}\_i - \overline{\mathbf{x}})^2} \sqrt{\sum\_{i=1}^{n} (y\_i - \overline{y})^2}} \tag{4}$$

where *x* or *y* is the mean of the factors *x* or *y*, respectively, and the data for each variable are normalized to a range of 0 and 1 using the equation below.

$$y = \frac{(\mathbf{x}\_i - \mathbf{x}\_{\min})}{(\mathbf{x}\_{\max} - \mathbf{x}\_{\min})} \text{ \tag{5}$$

where *y* is the normalized value of the initial *x<sup>i</sup>* , and *xmax* and *xmin* are the maximum and minimum values of *x<sup>i</sup>* , respectively.

Conventional modelling and optimization methods, i.e., RSM and statistical analysis, need a varied range of experiments, which are expensive and time consuming. Additionally, they do not precisely elucidate the association amid different factors affecting removal efficacy. ML is more robust and effective compared to statistical models when modelling complex data with possible nonlinearities or partial data [171]. ML offers stage for mapping associations amid input and output parameters during pollutant remediation. It is grounded on computer algorithms that are automatically developed and adjusted to specified conditions, with the ability to forecast the properties of concern via the knowledge and interpretation carried out by machines. It is used to study complex associations amid adsorbent properties (e.g., surface area, total pore volume, average pore diameter, and elemental composition), adsorption conditions, and adsorption performance without empirical assumptions [172]. Moreover, the foremost goal of system modelling and simulation is to obtain data on the progress of the system without conducting experiments. Thus, it is essential to develop a multidimensional nature model that can adapt to each process. Regarding the problem of overlooking variable relationships, ML models have demonstrated beneficial tools for method design, regression models, and response optimization [173].

Non-linear associations between independent and dependent variables can be created by ANN models grounded on a set of test results. Recently, the use of ANN models for modelling dye adsorption has become very common. Based on the ANN model type, dye adsorption can be categorized into the following groups: (i) multilayer feedforward neural networks, (ii) adaptive neuro-fuzzy inference systems, (iii) support vector regression, and (iv) hybrid models. Various ANN models were used to assess dye removal [174]. A three-layer feedforward backpropagation network (FFBPN) was used to study the removal of acid orange from an aqueous medium by powdered activated carbon [175]. In an input

layer with three neurons (consisting of initial pH, dye concentration, and contact time) and an output layer with one neuron (dye concentration after time 't'), they established FFBPN. As a result of evaluating the model's performance using the mean relative error (MRE), it was found to be 5.81%. This MRE value signifies that the ANN model had good analytical performance and can be used in place of kinetic studies.

#### **5. Conclusions and Future Perspectives**

BC has been considered an efficient adsorbent to treat wastewater owing to the presence of abundant functional groups and large surface area. Few studies assessed expenses during the production of BC-based adsorbents. A cost-effective technique and the reuse of multiple cycles are possible. However, additional investigation of numerous outlooks is obligatory to confirm BC efficiency and inexpensiveness, mainly for pilot-scale implementation in subsequent fields.


Modified clay/BC composites displayed significant benefits compared to modified clays alone, owing to their cost-effective, high adsorption ability, and anionic dye removal effect. However, few studies have been carried out on a pilot scale. The advancement of magnetic BC research with organic pollutant removal is a vital and new research area, and more research in this field should be conducted in the future. However, generation of noxious components during the synthesis of magnetic BC should be careful.

BC delivers a robust interaction between the BC surface and various pollutants existing in water and soil. However, as BC loaded with pollutants may pose serious hazard associated with disposal, a new challenge for researchers concerning the prospects for the transformed environmental movement of these pollutants is to change the context of pollutant hazard valuation in the milieu. Implementation of a sustainable management strategy for contaminant loaded-BC is life threatening. This approach should follow the concept of a circular economy and should allow for an eco-friendly and cost-effective recycling of these BCs in the forthcoming adsorption cycles. In addition, during laboratory scale studies, knowledge of the mechanisms of chemical and physical interaction of BC with pollutants is required. Promising technologies always have a secondary hazard, and the utilization of BC loaded with pollutant for bio-oil production in the future could be a fascinating area that must be explored.

ANN and ML predictive models reduce workload, budget, space requirements, and pollutant remediation time. For instance, ML can be used to develop models to improve pyrolysis process and forecast the yield of BC depending on raw material characteristics and process conditions, which can help develop a sustainable wastewater treatment system. Further, the commencement of more compacted reactors capable of accommodating BC-supported catalysts are required for implementation. Future research should be devoted to the application of BC-based catalysts in large-scale reactors, which can be implemented at the industrial level. It is urgent to further explore the potential of BC in real challenging systems.

**Author Contributions:** L.G.: writing—original draft preparation; A.K.: writing—review and editing; S.R.K.: writing—review and editing; B.-S.K.: writing—review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by National Research Foundation of Korea (NRF-2019R1I1A-3A02058523).

**Data Availability Statement:** The study did not report any data.

**Conflicts of Interest:** There are no financial conflict of interest to disclose.

## **References**

