The impact of contact time on the adsorption capacity of activated carbon (AC) for BPB removal was evaluated. In Figure 3, it clearly depicts that, the removal of BPB by AC increases as the contact time increases. The removal of BPB occurred in two main phases. The sorption of BPB is fast during the first 60 minutes, where about 85.2% of the initial concentration of BPB were removed; then slows down around 3 h, until the end of the run. The equilibrium is achieved almost in 3 h. At this time, the adsorption efficiency was 97.4%. A contact time of 3 h was thus established for the following runs. The fast adsorption could be due to the high availability of free active sites at the beginning of adsorption process, whereas after a specific period of time, actives sites will be gradually occupied which will make the process of adsorption slower [23] . Similar results have been highlighted in the literature on the removal of dye [14] [24] .

The effect of activated carbon dose on BPB (pH 7 and 10 mg・L⁻¹) removal was studied and the result obtained is depicted in Figure 4. It is observed that the adsorption efficiency of BPB by activated carbon increases with an increase in concentration of sorbent. The adsorption efficiency increases from 41% to 95% with increasing the adsorbent dose from 1 to 4 g・L⁻¹. This may be due to an increase in the availability of surface active centers. The same observation is also reported in literature [25] [26] . A contact time of 3 h and the dosage of activated

carbon of 4 g・L⁻¹, were thus considered for the following runs.

The influence of the initial aqueous solution pH value is an important parameter that affects the adsorption process [11] . The study of BPB adsorption on activated carbon coconut shell-based is carried out during 3 h, at different pH values (3 - 11) of 10 mg・L⁻¹ BPB solution. This pH range (Figure 5) allows to investigate the evolution of BPB in two chemical forms due to its pKa 8.24 [3] . At pH < pKa, BPB existed as a neutral molecule, while it mainly existed as a negative species when pH value was higher than pKa.

Figure 6 shows the initial pH effect of BPB adsorption by AC. The maximum percentage of BPB uptake (97.87%), occurs at pH 3. However, increasing pH to 11, the adsorption capacity decreases to 82.51%. As the pH of the system increases, the number of negatively charged binding sites would increase. Therefore, at low pH range, the surfaces of activated carbon might become positively charged due to the excess of H⁺ ion in solution. The electrostatic attraction between BPB molecule and the adsorbent surface increase the amount of BPB adsorbed. At the high pH range, the surface of activated carbon might become negatively charged due to excess of OH⁻ ions concentration, which will compete with BPB for the available positively charged sites on the adsorbent

surface, thus decreasing the removal efficiency of BPB. Figure 5 shows in addition that, at pH > pKa, BPB is in its deprotonated ionic (maximum wavelength is 297 nm) form due to the ionization of the hydroxyl group, increasing its hydropholicity. In these conditions, the adsorption process was governed by physical phenomenon. The high removal value (85.51%) was mainly due to the BPB ionization, that gave an ionic form with different wavelength (297 nm).

Some authors [27] [28] assumed that the presence of functional groups on the adsorbent surface influences the adsorption ability. The activated carbon efficiency could be depend on several mechanisms in the removal of the BPB. In this study, the adsoprtion process versus pH shows that ion-exchange, electrostatic interactions, capillarity and BPB ionization form are involved in the removal mechanism. Similar phenomena were observed in the work reported by Hawari et al. [29] .

The equilibrium adsorption isotherms are important in the development of adsorption process. Adsorption characteristics of an adsorbent are used to estimate the adsorbent/adsorbate interaction. In this study, four isotherm models were used to investigate and elucidate the adsorption behavior of BPB onto activated carbon. Data obtained from the experiments were fitted by Langmuir, Freundlich, Temkin and Dubinin-Radushkevich isotherm models.

1) Langmuir isotherm

The Langmuir isotherm model was chosen for the estimation of maximum adsorption capacity corresponding to complete monolayer coverage on the biomass surface [30] . The linearized form of the Langmuir isotherm is described by the following equation:

1 / Q e = 1 / Q m K L C R + 1 / Q m (3)

where, Q_e is the amount of BPB adsorbed per mass of adsorbent (mg・g⁻¹), C_r is the residual concentration of BPB, Q_m is the monolayer adsorption capacity (mg・g⁻¹) and K_L is the Langmuir constant. The isotherm parameters, Q_m K_L and the coefficient of correlation are presented in Table 4. The sorption capacity, Q_m, which is a measure of the maximum sorption capacity corresponding to the complete monolayer coverage showed that the activated carbon adsorption capacity for BPB was 7.519 mg・g⁻¹. The adsorption coefficient, K_L, related to the apparent energy of sorption is 0.093 L・mg⁻¹ and the correlation coefficient 0.979. This observation showed that the energy of adsorption is not very favorable to BPB, probably due to its weak interaction between BPB molecules and the activated carbon surface properties. The same capacity order was reported [28] for organic compounds sorption on coconut shell activated carbon. According to the Langmuir model this could suggest some homogeneity in the surface or pores of the activated carbon [31] .

The essential factor K_L is used to evaluate the feasibility of the adsorption process as favorably and nature of the isotherm.

R L = 1 / ( 1 + K L C i ) (4)

where K_L is the Langmuir constant and is the initial concentration of BPB.

The R_L value for BPB adsorption on activated carbon is 0.64 (0 < R_L < 1), indicating that the adsorption process is favorable [32] .

2) Freundlich isotherm

The Freundlich model was chosen to estimate the adsorption intensity of the adsorbate on the adsorbent surface. The linearized form of the Freundlich isotherm is given by:

ln Q e = ln ln K F + ln C e × 1 / n (5)

where K_F is the Freundlich isotherm constant related to the adsorption capacity, whereas describes the adsorption intensity [33] . Table 4 shows that the adsorption data for activated carbon were found to fit with Freundlich model having a correlation coefficient of 0.973. The Freundlich adsorption equation assumes that a multi-layer adsorption could occur on the surface of the

adsorbent [26] . The value of n obtained (n > 1) indicates that BPB is adsorbed on the activated carbon. This observation seems to explain the co-adsorption phenomenon due to the attraction generated by the BPB molecule after recovery the activated carbon surface, creating a multi-layer. Djilani et al. [34] reported similar results, when they studied the removal of methyl orange and methylene blue using activated carbon prepared from apricot stones and commercial activated carbon.

3) Temkin isotherm

The Temkin isotherm [35] reflects the fact that the heat of adsorption of all the molecules of the cover layer decreases linearly with the recovery due to the decrease of adsorbent/adsorbate interactions. The linearized form of Temkin isotherm is expressed by:

Q e = R T ln K T × 1 / b T + R T ln C e × 1 / b T (6)

where, b_T is the Temkin constant related to the heat of adsorption (J mol⁻¹), K_T is the equilibrium binding constant (L g⁻¹), R the gas constant (8.314 J mol⁻¹・K⁻¹), T is the absolute temperature (K), Q_e is the amount of BPB adsorbed per mass of adsorbent (mg・g⁻¹) and C_e is the residual concentration of BPB at the equilibrium. The values of the parameters are given in Table 4. The correlation coefficient (R² = 0.909) indicates that Temkin’s model fits the BPB adsorption on the activated carbon. The constant of Temkin K_T (0.317 mg・g⁻¹) is less than unity, suggesting a low affinity between the adsorbate/adsorbent, and the positive value of b_T (24.49 kJ mol⁻¹) indicates that the heat of adsorption due to interactions with adsorbate decreases linearly with the recovery rate [35], and that adsorption reaction seems to be exothermic. Furthermore the surface of activated carbon can be considered homogeneous energy. This result is related to the Langmuir observation about the homogeneity of the activated carbon surface.

4) Dubinin-Raushkevich isotherm

Dubinin and Radushkevich (D-R) proposed another isotherm used for analysis [36] . D-R hypothesis involve heterogeneity of the activated carbon surface. This isotherm equation is applied to estimate the characteristics porosity and the apparent free energy adsorption [37] . The linear form of the D-R equation is represented below:

ln Q e = ln ln X m − K D − R ϵ 2 (7)

where, ϵ = R T ln ( 1 + 1 / C e ) , Q_e is the amount adsorbed per unit mass of adsorbent (mg・g⁻¹), X_m is the adsorption capacity (mg・g⁻¹), C_e is the equilibrium concentration of BPB in solution (mg・L⁻¹), R the gas constant (8.314 J mol⁻¹・K⁻¹), K_D−R is the constant related to the adsorption energy (mol²・kJ^-2) and T is the absolute temperature (K).

The plot of lnQ_e versus ϵ² gives a straight line. The values of X_m (0.309 mg・g⁻¹) and K_D−R (2.681 mol²・kJ^-2) are calculated from the intercept and slope (Table 4). The porosity factor (K_D−R) for activated carbon toward BPB was found to be more than unity, indicating that sorption of BPB on activated carbon may not be significant. That result confirms the weak adsorption capacity calculated by the Langmuir isotherm (Figure 7). From K_D−R value, the mean energy of adsorption E (kJ・mol⁻¹) is calculated using the following equation:

E = − 1 / 2 K D − R (8)

The adsorption mechanism estimated by the magnitude of E (0.43 kJ・mol⁻¹) is less than 16 kJ・mol⁻¹ for activated carbon, which indicates that adsorption seems to be given by particle diffusion.

Since there are no other published results on BPB removal by adsorption at our knowledge, the comparison with other adsorbents is done based on the adsorption capacity of activated prepared from coconut shell (Table 5).

The kinetics of the adsorption phenomenon are determined by the mass transfer

to the liquid-solid interface. Adsorption kinetics modelling provides best information about the reaction pathways. The results are fitted according to different kinetics models such as the pseudo-first order equation [40], pseudo-second order equation [41] and intra-particle diffusion [35] .

There are three steps in an adsorption process [42] . The first one is the external mass transfer of the adsorbate from the bulk solution to adsordent. The second one is the internal diffusion of the adsorbate to the sorption sites and the last step is the sorption itself. The pseudo-first order and the pseudo-second order are based on the fact that the sorption is the rate limiting step in the adsorption process.

1) Pseudo-first order model

The adsorption of pseudo-first order equation is established by Lagergren in 1898 [40] and the corresponding equation is given as below:

ln ln ( Q e − Q ) = ln Q e − k 1 t (9)

where k₁ (min⁻¹) is the rate constante of the adsorption, Q and Q_e are the amount adsorbed at any time and at equilibrium respectively. The plot of ln(Q_e−Q) versus t give a straight line with correlation coefficient of 0.988, indicating the applicability of this equation. The values of k_ad and Q_e are calculated from the slope and intercept. The calculated value Q_theo (0.31 mg・g⁻¹) found from pseudo-first order equation (Table 6) is lesser than the experimental Q_exp (0.39 mg・g⁻¹).

2) Pseudo-second order model

The experimental data are tested for pseudo-second order, according to Ho and McKay’s [41] equation:

t / Q = 1 / K 2 Q e 2 + t / Q e (10)

where k₂ is the rate constant of sorption (g mg⁻¹・min⁻¹); Q_e is the amount of adsorbate at equilibrium (mg・g⁻¹); Q is the amount of adsorbate on the surface of the adsorbent at any time (mg・g⁻¹). The plot of t/Q versus t gives a straight line with correlation coefficient 0.995. The calculated adsorption capacity from the pseudo-second order equation (Table 6) is more than the experimental Q_exp (0.39 mg・g⁻¹).

3) Normalized standard deviation equation (ΔQ(%))

According to the correlation coefficient, the two models are close to the unity which accurately shows that the adsorption of BPB on this activated carbon follows pseudo second order. A normalized standard deviation, ΔQ(%) is

calculated to compare the efficiency of adsorption isotherms.

Δ Q ( % ) = 100 × ( ∑ ​ [ ( Q t exp − Q t c a l ) / Q t exp ] 2 ) 1 / 2 / ( n − 1 ) (11)

where, Q t exp is the experimental amount adsorbed at different time t, Q t c a l is the calculated amount adsorbed at different times and n is the number of observations. The ΔQ(%) value of pseudo second order (ΔQ(%) = 10.2) is smaller than that of the pseudo first order (ΔQ(%) = 20.4), confirming that the adsorption of BPB on activated carbon follows the pseudo second order equation (Figure 8) which is in good agreement with experimental Q_e for the activated carbon (Table 6).

4) Intra-particle diffusion model

The intra-particle diffusion model (Figure 9) was used to verify the effect of mass transfer resistance on the binding of BPB to activated carbon. This model was described by [35] :

q t = K p t 1 / 2 + C (12)

where, q_t is the amount of BPB adsorbed at time t, C is the resistance to the mass transfer in the film and K_p is the intra-particle diffusion rate constant (mg g⁻¹・min^1/2).

The value of (K_p), C and the correlation coefficient are given in Table 7. The correlation coefficient (R² = 0.943) indicates that the intra-particle transport is not the only rate of BPB adsorption onto activated carbon. This could be due to the adsorption of the solvent on the activated carbon area. Which would prevent the molecules of BPB to penetrate into the pores of activated carbon.

The effect of temperature on the adsorption of BPB by AC is shown in Figure 10(a). The amount of BPB adsorbed decreased slightly with increasing temperature. In order to understand more deeply the adsorption process of BPB on activated carbon from coconut carbon, thermodynamic parameters such as changes in standard enthalpy (ΔH˚), stardard entropy (ΔS˚) and standard free energy (ΔG˚), were evaluated.

Thermodynamic adsorption parameters are determined from the following relationships:

Δ G ˚ = − R T ln K L (13)

Δ G ˚ = Δ H ˚ − T Δ S ˚ (14)

where, K_L is Langmuir constant determined from the Langmuir isotherm.

The results are summarized in Table 8. The values of ΔH˚ and ΔS˚ can be determined from the intercept and the slope values, respectively, of the straight line of the plot of ΔG˚ against T (K) (Figure 10(b)). The positive values of ΔG˚ indicate that the adsorption of BPB on the activated carbon is not spontaneous. This result is in accordance with the intra-particle diffusion observation. The endothermic nature of the process was verified by the positive value of ΔH˚ (0.0002 kJ・mol⁻¹). The positive value of ΔS˚ (0.019 kJ mol⁻¹・K⁻¹) explains that the degree of randomness at the solid-liquid interface increased during BPB adsorption onto the activated carbon. Similar behavior was also reported for adsorption of Reactive Blue-19 on coconut shell based activated carbon [28] .