Adsorption is a surface phenomenon in which adsorptive (gas or liquid) molecules bind to a solid surface. However, in practice, adsorption is performed as an operation, either in batch or continuous mode, in a column packed with porous sorbents. Under such circumstances, mass transfer effects are inevitable. The three common steps includes: Film diffusion (external diffusion), Pore diffusion [intraparticle diffusion (IPD)] and Surface reaction), but, the seven classical steps involved are: 1) Diffusion of the reactants from the bulk phase (boundary layer) to the external surface of the catalyst pellet (film diffusion or interphase diffusion), 2) Diffusion of the reactant from the pore mouth through the catalyst pores to the immediate vicinity of the internal catalytic surface; the point where the chemical transformation occurs, (pore diffusion or intraparticle diffusion), 3) Adsorption of reactants on the inner catalytic surface, 4) Reaction at specific active sites on the catalyst surface, 5) desorption of the products from the inner surface, 6) Diffusion of the products from the interior of the pellet to the pore mouth at the external surface, and 7) Diffusion of the products from the external pellet surface to the bulk fluid (interphase diffusion), are generally used as the key for explanation (Figure 1) [37] . If external transport is greater than internal transport, rate is governed by particle diffusion. If external transport is less than internal transport, rate is governed by film diffusion and if external transport approximately equal to internal transport, the transport of adsorbate ions to the boundary may not be possible at a significant rate thus, formation of a liquid

film surrounding the adsorbent particles takes place through the proper concentration gradient [38] . In order to predict the actual slow step involved in the adsorption process, the kinetic data were further analyzed using the Boyd model equation [B_t = −0.4977 − ln(1 − F)].

Where, F Represents the fraction of solute adsorbed at any time, t(h), as calculated by, F = q_t/q_o. When B_t vs. t(h) plotted and if the linear lines pass through the origin the rate limiting step become particle diffusion if not it governed by external mass transport mechanism [39] [40] , reported the adsorption of 2, 4, 6-trichlorophenol on the prepared activated carbon from Boyd plot and mainly governed by particle diffusion.

During all the above mentioned processes, two main types of interaction, namely, physical and chemical, are possible between adsorbent and adsorbate [42] ; the former is known as physisorption, and the latter is chemisorption. Physisorption is a result of attractive forces between sorbent and adsorbate molecules, whereas chemisorption provides a stronger bond as it involves the transfer or sharing of electrons between adsorbent and adsorbate species. As a guideline, heat of adsorption with magnitude between 5 and 40 kJ/mol indicates physisorption as the dominant adsorption mode, while the other between 40 and 125 kJ/mol indicates chemisorption [43] .

There are six different types of adsorption isotherms as shown in Figure 2. Type one (I) adsorption isotherm is for very small pores or microporous adsorbents. In this case adsorption occurs by filling of the micropores. The adsorbate adsorption rate depends on the available micropore volume instead total interior surface area. Type two (II) and type four (IV) adsorption isotherms were detected for non-porous or macroporous adsorbents with unlimited monolayer-multilayer adsorption. First the adsorption volume quickly increases at low relative pressures due to contact of the adsorbate molecules with the higher energetic section followed by the interaction with less energetic section.

When the monolayer formation of the adsorbed molecules is complete, multilayer formation starts to take place corresponding to the “sharp knee” of the isotherms. As the relative pressure approaches unity a sudden rise shows the bulk condensation of adsorbate gas to liquid. Finally, type three (III) and type four (V) isotherms do not have the “sharp knee” shape indicating that a stronger adsorbate-adsorbate interactions than adsorbate-adsorbent interaction.

Adsorption equilibrium information is the most important piece of information needed for a proper understanding of an adsorption process. A proper understanding and interpretation of adsorption isotherms is critical for the overall improvement of adsorption mechanism pathways and effective design of adsorption system. But most of the studies apply only Langmuir and Freundlich

models and depending on coefficient of determination (R²) values the mechanism of adsorption judged by declaring the process as physical, chemical or both. Since most those models fits well with different experimental data, confirming with other models to know whether those models are perfect or not become the most important requirement. The selective mini review is presented in the Table 1.

As an example [45] (2012) a researcher applied Langmuir, Freundlich, Temkin and Dubinin-radushkevich isotherms studies of equilibrium sorption of Zn²⁺ onto phosphoric acid modified rice husk. Among those four adsorption isotherm models, the R² value of Langmuir isotherm model was the highest and the heat of sorption process was estimated from Temkin isotherm model to be 25.34 J/mol and the mean free energy was estimated from DR-K isotherm model to be 0.7 KJ/mol which vividly proved that the adsorption experiment followed a physical process. On this studies the well-fitting of Langmuir isotherm model were confirmed by Temkin and DRK. Different models and their uses as well as the linear and nonlinear sorption isotherms equations are presented in Table 2. During adsorption study to understand the machoism of adsorption it is advisable to apply at list adsorption isotherm models present in bold red in Table 2 and Table 3 for kinetics of adsorption. Common nomenclatures of equations are given at the end above the references.

Kinetics of adsorption using the statistical least squares method, several adsorption studies fit the PFO and PSO models. Therefore, without applying any work used to reduce/eliminate the diffusion-adsorption mechanism (transport influences; like intra particle diffusion), deciding the mechanism of adsorption only depending on adsorption-reaction models become wrong interpretation of mechanism of kinetics of adsorption. Generally suggested by earlier research group, [35] PFO and PSO is not a model of defined mechanism but, rather, they are a

highly flexible formula that combines many different models with different controlling mechanisms. The rate constants obtained was the result of complex interaction between different controlling mechanisms which are simply empirical

constants without physicochemical meaning and provide little insights into adsorption mechanisms and no meaningful mechanism can be confidently postulated from these models. In addition to the mechanism of reaction, it is necessary to calculate important sorption parameters like adsorption activation energy with the help of Arrhenius equation D s = D s o exp ( − E a / R T ) [61] . The linear and nonlinear formula of sorption kinetics and their use are indicated in Table 3.

The nonlinear and linear models were used to define the kinetics curves. Their validities can be determined using coefficient of determination (R²) and standard deviation (SD) Δq (%) [78] . In most of sorption studies the linearized form of PFO, PSO and other relevant models are determined based on the correlation of determination (R²) value, and the obtained model parameters are cross-checked with the experimental values. But, the linearized forms of the PFO and PSO model provide sometimes-widely different parameter values. Nonlinear regression is recommended to avoid this problem, and it has been shown to be better than linear regression. It gives more realistic values of q_e and k, and most of the times a R² value. But, Regardless of its popularity it does not penetrate deep into the community.

Most linear regression creates discrimination on PFO than PSO based on R² values decision. Some studies qualified to PSO by linear regression, surprisingly fits PFO by nonlinear regression [79] . In addition, PFO is prone to experimental error, whereas PSO is less sensitive to experimental error, but the nonlinear regression evaluates both PFO and PSO equally towards experimental error. Therefore comparison of the linear and nonlinear models in terms of fit quality and reasonability of the determined parameters before deciding on the best model becomes the most acceptable procedure to reduce statistical discrimination bias [80] .

Linear regression is a method used to model the association among a scalar dependent variable and one independent variable. The well-fitting line is the line that reduces the sum of the squared errors (SSE) of prediction. The strength of the linear association between two variables is quantified by the square of the correlation coefficient (R) which is known as the determination of the coefficient (R²). The higher R² value (lower values of error functions) indicates a stronger linear relationship.

Nonlinear regression can be a dominant substitute to linear regression because it suggests the most flexible curve fitting functionality. For nonlinear model, sum of square must be minimized by an iterative method. The nonlinear regression line is the line that minimizes the sum of squared deviations of prediction (also called the sum of squares error). The standard error of the estimate (S) is the square root of the average squared deviation. This parameter measures the accuracy of predictions. The smaller the standard error of the estimate indicates the more accurate prediction. In nonlinear regression, good of fitness method is used after error analysis.

Error functions (Table 4) are statistics that measure the error between the model parameters and experimental values. It was a standard method developed based on the least squares criterion. Model parameters are captured well by the slope and intercept, both of which are clearly defined as functions of the experimental data. The fitted parameters are set by regression to minimize the sum-of-squares errors between the predicted and experimental values.

In the literature, these error statistics were calculated in addition to the correlation of determination (R²) to confirm and support the model already discriminated by R². High R² values correspond to low error statistics in most cases. When R² values are too near to discriminate the PFO and PSO models, RMSE comes in convenient [81] . Roughly well-known error functions are listed in table. When its value is nearer to unity indicates a well fit.

The past study [78] compared the linear and nonlinear equations of pseudo-first order and pseudo-second order kinetic models on adsorption of Cd (II) onto chitosan and methyl orange onto bentonite. For validation standard deviation (SD) Dq (%) and the coefficient of determination R² values was used. The conclusion was that the nonlinear models are more appropriate than linear forms for the modeling of the kinetics of adsorption in liquid phase and it is noted that the determination of R² alone is inadequate to decide. Therefore, using the error functions like standard deviation Δq (%) in addition to R² value is supportive.

In the past substantial efforts such as: doping with impurities/dopant [90] , formation of nanocomposites [91] , surface modification, dye sensitization [92] , noble metal deposition [93] , non-noble metal deposition [94] , sensitizing with narrow bandgap absorption materials (Eg < 3.0 eV; like: CdS, ZnS, etc.), organic dyes, conducting polymers and inorganic metal complexes [95] and hydrogenation procedure (annealing under hydrogen atmosphere) are studied for allowing visible light absorption of high band gap metal oxides (such as: TiO₂, ZnO, …). The hetro-junction of those high band gap metal oxides with other low band gap materials get attention due to not only induces visible light sensitization for the composite but also for it enabling the oxidation and reduction reactions to arise at different components. But those methods are also not well-organized enough to induce enough visible light absorption. The combination of these hetro-junction leading to multiple alterations (such as: simultaneous doping-sensitization, doping-deposition, deposition-coupling, etc.) are also are become the alternative towards improvement of visible light absorption [96] . Also functionalizing of the metal oxide surface [using the bifunctional linking or with citric acid molecule] enables the connection of larger amount of low band gap semiconductor materials [8] .

In order to decrease the recombination of photogenerated electron-hole pairs (charge carrier separation enhancement), structural stability and for isolation of the redox sites at various modification methods (such as: integrating with equivalent bandgap materials, designing with exposed reactive facets, modification with carbon nanostructures and hierarchical morphologies coupling two semiconductors) have been through. Among those methods researchers choose coupling of two/more semiconductors metal oxides having related or different band gaps (TiO₂/ZnO, SnO₂/ZnO, SnO₂/ZnO/TiO₂ and Co₃O₄/ZnO) [6] [7] [14] [97] , due to continuous transfer of photo-generated electrons and holes from one semiconductor to another [98] [99] .

For the reason that, the nanocomposite improves light absorption effectiveness, prevention of photo corrosion, increases sun light absorption, new functionalities get up at the interfacial, change in surface acid-base behavior and better suppression of electron hole recombination [100] . Specially, the tricomposites are more energetic towards suppression of electron hole recombination, because, unidirectional charge transfer (interelectron transfer) within the stepwise electronic band structure between the catalyst [97] . On the other studies coupling of bare metal oxide with its doped counterparts (example: N-ZnO coupled ZnO) enhanced the charge carrier separation, free radical generation and visible light response [18] [19] . Explanation of different ways used to increase the visible light response, structural stability, altering the defect chemistry with improved charge carrier separation procedures are shown in Figure 3.

In addition to the electron-hole recombination and visible light response, deep awareness on factors affecting photo-degradation, as well as, adsorption effectiveness, such as: catalyst/dosage loading (which controls light scattering, screening effects, turbidity and agglomeration) [101] [102] , catalyst structure (surface area) [103] , solution pH(surface charge, positions of VB and CB) [104] , light intensity [105] , light wavelength (around 254 nm) [106] , temperature (20˚C - 80˚C) [107] , inorganic species (H₂O₂ [108] , S 2 O 8 2 − [33] , BrO 3 − [34] , SO 3 − [32] ) should also be taken in to consideration.

Among different metal oxide nanomaterial synthesis methods, the solution-based approach is the simplest, least energy consuming and easy to control the morphology, as well as, sizes of the nanomaterials by handling the experimental factors such as type of solvents with different dielectric constant, starting materials (precursor) concentration, reaction solution pH, mineral acids, etc. [21] [109] . The chemical approach normally involves reduction of metal ions into metal atoms in the existence of stabilizing agents, followed by the controlled aggregation of atoms. Those approaches used in the synthesis of modified metal oxides includes:

impregnation method [22] , hydrothermal [24] , sol-gel [110] , co-precipitation, flame spray pyrolysis, micro-emulsion, electrochemical deposition process, solvothermal, microwave, polyol, flux methods, electrospinning, solid state dispersion method etc. [111] [112] .

Among these approaches, the sol-gel procedure is the most attractive method towards low production cost, good, high reliability, repeatability, simplicity of the process, low process temperature, ease of control of physical characteristics & morphology of nanoparticles, good compositional homogeneity and optical properties [29] . In addition to advance the photoactivity, various methods such as: controlling adding rate of precursors (example: sodium hydroxide), adding of selective capping agents, have been used to adjust the chemical properties of the materials, elemental composition, shape, morphology, size and growth of the nanostructure materials [113] . As an example among different shapes of ZnO nanostructures, spherical nanostructures were reported to exhibit the highest degradation rate of organic pollutants ascribed to their large oxygen vacancies. Not controlling those different parameters/conditions the toxicity of the reagent added must be taken in to consideration. In this context, ongoing research has modified the existing sol-gel technique by using only water as solvent [114] . Due to the existence of M-O-M bonds in metal oxides nanoparticles, agglomeration occurs readily [115] . This agglomeration of nanomaterials to each other strongly limits the adsorption, as well as, photo catalysis of nanoparticle. Among different agglomeration inhibition methods such as, chemical treatment [116] coating [115] , grafting polymerization [117] , ligand exchange [115] and self-assembly [118] ) techniques, ligand exchange are the one which is nicely acknowledged.

Photo-catalysis defined as “catalysis driven acceleration of a light-induced reaction”, the detailed scheme of semiconductor photo-catalyst reaction was summarized in Figure 4(a). First before degradation starts the organic pollutants should diffuse and adsorbed on the external surface of catalyst/semiconductor. Therefore, selection/modification of the adsorbent for specific adsorbate become the key point, because, without adsorption no degradation takes palaces. After adsorption when ultraviolet/visible light energy (hυ) (equal or greater than the excitation energy/band gap) interacts with catalyst/semiconductor, as a result of excitation of electron (e⁻) from valance band (VB) to conduction band (CB) the production of e⁻/h⁺ pairs occurs. Migration of those e⁻/h⁺ pair on the surface of the catalyst and interacting with other species, like, water or oxygen redox reaction will takes palaces. On oxidation reaction H₂O ( OH − ) reacts with catalyst holes ( h VB + ) results in formation of hydroxyl radicals ( OH • ). And on reduction reaction the exited electrons reacts with oxygen from the air for the formation of superoxide radicals ( O 2 • − ). The superoxide radicals will further protonated to harvest hydro-peroxyl radical ( HOO • ) and subsequently hydrogen peroxide

(H₂O₂) and hydroxyl radicals ( OH • ). The resulting powerful oxidizing agents ( OH • ) leads to partial or complete mineralization (CO₂, H₂O and mineral acids) of organics pollutants adsorbed on the surface of the semiconductor [119] .

Depending on the CB and VB potentials and bandgap energies the electronic arrangement of the semiconductor heterojunctions are categorized into (a) straddling gap (b) staggered gap and (c) broke gap heterojunctions as shown in Figure 4(b). In Figure 4(b_i) (straddling gap type) the e⁻/h⁺ collected and recombine on shorter band gap semiconductor. Therefore, weakening of the photo catalysis efficiency [120] . In Figure 4(b_ii) (staggered band structure) e-moves to the low negative CB semiconductor and h⁺ transfer to less positive VB of large band gap semiconductor. Here the probability of e⁻/h⁺ separation becomes high then the photocatalysis improved [121] . In Figure 4(b_iii) (broken gap heterojunction) due to great variance in the positions of the VB and CB levels, provides higher motive energy to transfer the photo induced charge carriers. Thus, only the staggered gap most typical heterojunction system for photo catalysis purposes [122] and selecting well-matched energy band configuration, less lattice inconformity, and fitting conductivity to be combined with ZnO has a key for higher photocatalytic activity [31] . To accept e⁻ from ZnO using semiconductor with positive CB than ZnO becomes essential.

The heterojunction composed of an n-type semiconductor as a donor and a p-type semiconductor as an acceptor is so-called p-n heterojunction. As shown in (Figure 4(c_i)) the fermi level (FL) is positioned near the VB in p-type semiconductors place holes constitute major carriers. Moreover in the n-type, the FL is closer to CB and the concentration of electrons is higher than holes as carriers [123] . Hence, once they come into contact with each other, the FL of p-type shifts to higher and n-type to lower energies until attainment equilibrium level. Consequently, a depletion layer is formed at the interfacial distinct that produces an electric field [124] . The resulted electric field from the induction of negative and positive charges on the p- and n-sides will diffuse the photogenerated electrons and holes out of the depletion layer [125] .

Similarly, n-n heterojunctions (Figure 4(c_ii)) has been introduced as another slant engaged visible-light-driven (VLD) nice enactment photocatalysts. Like that of p-n type n-n system the fermi levels of both semiconductors should be equilibrium to make a steady and electrostatically balanced system. Subsequently, a depletion layer is created which helps for exploiting visible light and inhibiting the recombination of the photoinduced e⁻/h⁺ pairs [126] [127] .

Apart from the binary nanocomposites, recently combination of large band gap materials with various semiconductors to develop ternary or multi-component heterostructures resulted in great improvement in photocatalytic performance of photocatalysts under visible-light/solar irradiation. As indicated the studied ZnO based ternary composites are grouped into four classes, based on: 1) band gap engineering, 2) plasmonic, 3) p-n-n and n-n-n heterojunctions, and 4) magnetic properties.

According to band gap engineering the difference between the energy (CB) of the joint semiconductors forces a rapid transfer of the photoexcited electrons from one semiconductor to other, thus accelerating the separation of e⁻/h⁺ pairs and enhancing the photocatalytic efficacy. The enhancement photocatalytic efficacy was confirmed by Chen et al. [12] on ZnO/CdS/Ag₂S and Li et al. [13] on Ag₃PO₄/ZnFe₂O₄/ZnO ternary nanocomposites. The schematic representations for separation of the charge carriers in ZnO/CdS/Ag₂S and Ag₃PO₄/ZnFe₂O₄/ZnO nanocomposites are shown in (Figure 4(d)).

The synergistic effects of the internal electric fields formed in the n-n-n or p-n-n heterojunctions between constituents of the ternary nanocomposites are more efficient than the binary n-n or p-n heterojunctions for substantially enhanced separation of the photogenerated e⁻/h⁺ pairs and further prolonging lifetime of the charge carriers. Thus, the enhanced photocatalytic performances for the photocatalysts with n-n-n or p-n-n heterojunctions are much more than those of the photocatalysts with n-n or p-n heterojunctions [31] .

Now a days since recovering of the engaged photocatalysts from the solution is practically reduce cost and avoids the secondary contamination, especially on large-scale applications, magnetically separable photocatalysts are promising materials. In sight of this, a number of researchers have been investigated Fe₃O₄, owing to its outstanding magnetic properties, being environmentally kind and not expensive. Some of ternary composites investigated based on ZnO and Fe₃O₄ includes: ZnO/Ag₂O/Fe₃O₄ [128] , ZnO/Ag₃VO₄/Fe₃O₄ [129] , Fe₃O₄@AgBr/ZnO [130] , Fe₃O₄/ZnO/CoWO₄ [131] , Fe₃O₄@ZnO/AgCl [132] , ZnO/AgI/Fe₃O₄ [133] , and Fe₃O₄/ZnO/NiWO₄ [134] .