The iodine adsorption test is employed to determine the adsorption capacity of AC, and IN value is used to determine if an AC is microporous (0 - 2 nm). IN value is obtained as the quantity of Iodine (I₂) adsorbed per gram of AC on a milligram scale. The IN for all samples were obtained following ASTM D4907 − 94 method [10][20].

The MBN is a measure of the mesoporosity (2 - 50 nm) of an AC, and is obtained as the amount of dye adsorbed on 1 g of adsorbent [10]. The MBN test was performed by mixing 0.020 g of AC and 10 mL of 1000 mg∙g⁻¹ MB solution and stirring the solution for 24 h in order to achieve equilibrium state [32]. The concentration of MB was measured at 660 nm using a double beam UV-visible spectrophotometer (Thermo Scientific Evolution 300, USA).

The activated carbon yield was calculated as the dry weight of obtained AC to raw material according to Equation (1),

Yield ( % ) = m A C m 0 × 100 (1)

where m_AC and m₀ are dried mass of AC and the dried mass of raw material respectively.

The carbonization temperature of ACs production depends on the thermal behavior of precursor (ESS), therefore, ESS was subjected to thermogravimetric analysis (TGA), TGA profile is given in Figure 1. TGA suggests three stages of weight loss during pyrolysis. The first stage shows 9.98% weight loss in the temperature range of 27˚C - 160˚C, quite possibly due to removal of surface bound water and moisture. The second stage, characterized by a weight loss of 52.34%, from 160˚C to 400˚C, could be due to the decomposition of cellulose, hemicellulose

^bno detection; ^aby difference.

and few lignin [20][39]. In final stage of pyrolysis, 400˚C and above (gradually and continuously up to 900˚C), the weight loss (16.80%) corresponds to the decomposition of lignin and formation of carbon material. The residue was about 22.38%. Hence, the temperature range of 400˚C to 700˚C was considered as the carbonization temperature in this study.

The FTIR spectrum of ESS is presented in Figure 2. The broad peak appearing at 3389.6 cm⁻¹ is attributed to the O-H bond stretching vibration of water, cellulose, hemicellulose and lignin (main constituent of ESS) [18]. The two peaks at 2928.3 and 2862.7 cm⁻¹ are due to symmetric and asymmetric stretching vibration of methyl group, and the peak at 1737.6 cm⁻¹ indicates the stretching vibration of carbonyl group (C=O) of hemicellulose and lignin [18]. The bands found in the 1653.4 - 1515.6 cm⁻¹ range are responsible for N-H bond stretching. More vibrations were observed between 1459 and 1377.3 cm⁻¹ can be ascribed to aliphatic and aromatic C-H groups in the plane deformation vibration. The bands between of 1261.0 - 1041.7 cm⁻¹ could be assigned to the C-O stretching vibration of carboxylic acid and alcohols [18].

The results obtained from Iodine adsorption are given in Table 3. The IN values of ACP and ACZ are in range 342.27 - 479.49 mg∙g⁻¹ and 350.44 - 539.98 mg∙g⁻¹

respectively. Statistical analysis was carried out to determine the significant variables in IN values (Table 4). Two quadratic regression models were used to predict the optimal IN value of ACP (Y₁) and ACZ (Y₄) (Equations (9) and (12)) respectively).

Y 1 = 466.044 + 11.951 x 1 + 43.412 x 2 − 5.301 x 3 − 61.397 x 1 2 − 5.581 x 2 2     + 1.953 x 3 2 − 2.274 x 1 x 2 − 10.648 x 1 x 3 + 7.402 x 2 x 3 (10)

Y 4 = 502.395 + 44.342 x 1 + 29.882 x 2 + 9.678 x 3 − 62.992 x 1 2 − 1.575 x 2 2     + 13.016 x 3 2 − 4.481 x 1 x 2 − 20.184 x 1 x 3 − 10.62 x 2 x 3 (11)

**most significant, *less significant.

By applying the statistical model ANOVA, (see Table 4), it is clearly seen that the experimental and predicted values are in good agreement, since the correlation coefficient of the model is close to unity for both ACs (R² = 0.93 and 0.91 for ACP and ACZ respectively). The significant terms are x₂ and x 1 2 for ACP, and x₁, x₂ and x 1 2 for ACZ, given that the p-value of these terms are less than 0.05. It is also noted that IN values are significantly affected by the concentration of activating agent (x₁) and carbonization temperature (x₂), while the effect of residence time (x₃) is insignificant. Figure 3 illustrates the response surface plots, obtained by plotting temperature and concentration against IN values, with the residence time kept constant at the center point (75 min). A consistent trend was observed in ACP and ACZ; the IN value increases as the temperature increases. This may be explained by considering that at higher temperatures, there is a greater chance of volatile matter to vaporize from the precursor, causing an increase in pore development and/or formation of new micropores (or active sites) [12][40], hence an increase in the IN value. Tounsadi et al. (2016) [10]reported similar results. Furthermore, an increase in concentration of activating agent leads to an increase up to a maximum of 6 M for ACP (Figure 3(a)) and 8 M for ACZ (Figure 3(b)) of the IN value. Any further increase in the concentration of activating agents, causes a decrease in the IN value. As the concentration of activating agent increase, the interaction between activating agents and raw material increases and generating porosity [31]. When the concentration of activating agent is too high, two possibilities occur: 1) new micropores do not form, while existing ones are continuously enlarged [31]and 2), the excess and/or unreacted activating agent blocks the pores [41]resulting to a decrease in microporosity, which is directly link to the Iodine absorption capacity. Lower concentration of activating agent resulted to lower IN values, probably due to incomplete reaction between activating agent and the raw material.

For the methylene blue adsorption, the values of MBN varied from 107.05 to 374.14 mg∙g⁻¹ for ACP and 149.88 to 374.25 mg∙g⁻¹ for ACZ (see Table 3). In both ACs, the MBN was significantly influenced by the concentration of activating agent (x₁) and activation temperature (x₂), see Table 4. For the prediction of MB adsorption capacity, quadratic regression model Equations ((11) and (12)) were used for ACP (Y₂) and ACZ (Y₅), respectively.

Y 2 = 265.797 + 82.82 x 1 + 26.932 x 2 − 5.419 x 3 − 59.555 x 1 2 + 24.366 x 2 2     + 59.133 x 3 2 + 7.83 x 1 x 2 − 12.806 x 1 x 3 − 12.387 x 2 x 3 (11)

Y 5 = 272.227 + 84.822 x 1 + 39.945 x 2 − 14.208 x 3 − 58.148 x 1 2 + 7.441 x 2 2     + 37.879 x 3 2 + 40.545 x 1 x 2 + 0.938 x 1 x 3 + 23.522 x 2 x 3 (12)

The correlation coefficients (R²) were found to be 0.94 and 0.93 respectively (see Table 4), which indicate good fitting of the described models. The significant terms of the selected models given in Table 4 are x₁, x₂, x 1 2 and x 3 2 (p-values are less than 0.05). Figure 4 depicts the response surface plots based on the interactions and effect of concentration (x₁) and temperature (x₂) on methylene blue adsorption value. The residence time was kept fixed at the center point for all plots. The value of MBN increases as a function of two significant variables (concentration and temperature) for both ACP and ACZ. The observed results are in agreement with that reported by Tan et al. [9], who indicated that the impregnation ratio and temperature have significant effect on the pore size and structure of AC from mango stem peel using KOH as activating agent. Similarly, Kouotou et al. [13]using H₃PO₄ and Sayğili and Güzel [18]using ZnCl₂, found well-developed mesopores of ACs with increasing value of studied factors (i.e. concentration and temperature). The increase in MBN value with the concentration and temperature, was also considered to be due to the depolymerization of cellulose and distillation of tar from the carbon at high temperatures

(above 500˚C), which generate more mesopores in the resultant ACs [40].

The resultant yield of ACs is a key response factor directed to production. A set of experiment were performed with variable experimental conditions, in order to determine the optimal conditions for obtaining maximum yield of product (i.e. ACP and ACZ). According to the results (Table 3), the maximum yield of ACP and ACZ was achieved at low concentration of activating agent with least carbonization time, (experiment 1), whereas, minimum yield was obtained at high temperature and longer residence time (experiment 12). The ACP yield was in the range 27.31% - 52.25% and similar to the ACZ yield (26.60% - 49.93%). The results of ANOVA for the response of ACP yield (Y₃) and ACZ yield (Y₆) (Equations (13) and (14)) and their correlation coefficients are given in Table 4.

Y 3 = 43.363 − 0.529 x 1 − 5.893 x 2 − 3.119 x 3 + 1.78 x 1 2 − 2.348 x 2 2     − 2.27 x 3 2 + 2.625 x 1 x 2 + 3.048 x 1 x 3 − 1.805 x 2 x 3 (13)

Y 6 = 40.707 − 1.211 x 1 − 6.084 x 2 − 2.62 x 3 + 2.26 x 1 2 − 1.555 x 2 2     − 1.527 x 3 2 + 2.907 x 1 x 2 + 1.47 x 1 x 3 − 2.155 x 2 x 3 (14)

All the terms, x₂, x₃, x₁x₂ and x₁x₃ and x₁, x₃, x₁x₂ for ACP and ACZ respectively (see Table 4) are significant. Figure 5 illustrates the response surface graph of carbonization temperature (x₂) and residential time (x₃) against the obtained yield. The concentration was maintained at the center point (6 M). It is observed from the graph that an increase in activation temperature leads to a decrease in the obtainable yield, whereas, residence time does not significantly affect the response. This decrease in yield may be attributed to the fact that an increase in activation temperature increases the quantity of carbon burn-off and the continuous liberation of the volatile matter which is reflected as a loss of adsorbent

material. The maximum yield was achieved when the temperature and residence time were maintained at their lowest value and the concentration was kept constant at the central point. The obtained results are in accordance with those from the literature [9][20][23]and it can therefore be concluded that, the activation temperature has a greater impact compared to the residence time on the AC yield.

The optimization process and method validation are quite essential in setting up the optimum conditions for maximum AC yield, high iodine and methylene blue adsorption capacity from the precursor. However, it is difficult to optimize these three responses under the same conditions because the zone of interest of the factors is different. The desirability function was then applied using Minitab 16 software in order to consolidate the three factors by considering the same weight for all the factors on the three responses [13][29]. The results of this optimization are presented in Table 5. The experimental responses are close to the predicted value, there was no significant difference between predicted and experimental data. A residence time of 30 min led to maximum yield irrespective of the activating agent. The optimum conditions for ACP are H₃PO₄ concentration of 7 M at a carbonization temperature of 633˚C, and ZnCl₂ concentration of 10 M and carbonization temperature of 698˚C for ACZ. These results show that, the AC obtained through H₃PO₄ activation require less concentration at comparative low temperature compared to that obtained from ZnCl₂.

The surface area of ACs was determinate through multipoint BET method by keeping the relative pressure (P/P₀) in the range of 0.02 to 0.3. The pore size parameters were calculated using Dubinin-Astakhov method. The results are summarized in Table 6. The linear plot of the multipoint BET (Figure 6) gave a surface area (S_BET) of 1058.91 m²∙g⁻¹ and 1008.99 m²∙g⁻¹ for ACP and ACZ respectively. The total pore volume was found to be 0.663 and 0.594 cm³.g⁻¹, and the average pore size of ACP and ACZ were obtained as 1.84 and 2.02 nm respectively. The average pore size indicates that the prepared ACs are microporous according to IUPAC classification [14]. Hence, both ACs show similar characteristic even though prepared at different carbonization temperature and concentration of activating agent.

^acalculated by difference; ^bno detection; ^cpH of point of zero charge.

The surface morphology of the precursor (Figure 7(a)), ACP (Figure 7(b)) and ACZ (Figure 7(c)) was assessed by AFM. The images show the presence of well-developed pores and cavities on the surface of ACP and ACZ, and no or undeveloped pores on the precursor’s surface. Therefore, surface modification and pore development occurs during activation and carbonization process [42].

The optimized ACs were subjected to Boehm’s titration and FTIR spectroscopy in order to determine the surface functional groups. The results from Boehm’s titration show that ACP and ACZ contain higher number of acidic groups (Table 6). The basicity character in ACP is comparatively weaker (0.43 mmol.g⁻¹) than that for ACZ (0.93 mmol.g⁻¹). The pH of point of zero charge, pH_PZC was found to be 4.00 and 7.02 for ACP and ACZ respectively.

The FTIR spectra of the ACs are noticeably different from that of the precursor

(see Figure 2). After activation and carbonization, there is the disappearance of some peaks in the FTIR spectra of the ACs and considerable decrease in the intensity of some peaks due to decomposition of cellulose, hemicellulose and lignin during the carbonization process [43]. However, there is an outstanding similarity in the spectra of both ACs. The band at 3421.6 cm⁻¹ is attributed to the O-H bond stretching vibration of absorbed water molecules [2]. The weak bands at 2923.6 and 2853.4 cm⁻¹ indicate the C-H stretching vibration of methyl and methylene group(s) [31][43], while the two bands appearing at 1743.9 and 1701.0 cm⁻¹ are attributed to the carbonyl group (C=O) of esters and carboxylic acid respectively [12]. The presence of C=C of aromatic ring stretching vibration is recognized by the band observed at 1541.4 cm⁻¹. The bands between 1400 and 1070.6 cm⁻¹ signifies the presence of primary, secondary and ternary alcohol, phenol, ether and esters showing C-O stretching and O-H bond deformation vibration [31]. The peak at 673.6 cm⁻¹ is indicates the out-of-plane O-H vibration [15][44].

The results of the elemental analysis are provided in Table 6. The carbon content of ACP and ACZ are significantly higher compared to that of the precursor (47.02%). This could be due to an increase in the degree of aromaticity [15]. A decrease in the quantity of non-carbon species such as hydrogen, nitrogen and oxygen was observed, probably due to the release of moisture and volatile substances during carbonization process [17]. The difference in elemental analysis between ACP and ACZ could be explained by the relative low concentration of activating reagent and carbonization temperature involve in the preparation of ACP (Table 5).

An initial nitrate ions concentration of 50 mg∙L⁻¹ was used to study the effect of pH on removal of nitrate ions in the range of 3 - 11. The maximum quantity of nitrate ion removed occurred at pH 3, and this amount decreases with increase in pH (Figure 8). The surface of ACs at low pH (pH ˂ pH_PZC) is predominantly positively charged as a result of the protonation of surface functional group(s), hence, more adsorption of nitrate ions by electrostatic attraction [42]. An increase in pH causes reduction of number of positive sites on the surface of the adsorbents, resulting to a reduction of adsorption capacity of adsorbent toward nitrate ions.

The experimental data obtained was analyzed using Freundlich, Langmuir, D-K and Tempkin non-linear isotherm models (see Figure 9). The best fit of the experimental data was analyzed based on R², RMSE and χ². Hence, high R² value, low RMSE and χ² value established the best fitting adsorption isotherm [6].

According to the presented result in Table 7, some discrepancies were observed in the Langmuir isotherm adsorption; lowest R² values, high RMSE and χ² values, unreasonably large value of Q_m and insignificant K_L, rendering this model unsuitable to explain the adsorption phenomenon on ACP and ACZ. The Freundlich and D-K isotherms both fit the experimental data. In addition, the n values obtained from the Freundlich adsorption isotherm model were 0.687 and 0.730 for ACP and ACZ, respectively. These values indicate that physisorption is dominant the process for both adsorbents [6]. The energy associated with the adsorption process can be calculated from the D-K model using the equation below

E ( kJ ⋅ mol − 1 ) = 1 2 k a d

From the model, if the magnitude of the energy (E) lies in the range of 8 - 16 kJ∙mol⁻¹, the sorption process is said to take place via ion exchange, whereas, if E < 8 kJ∙mol⁻¹, the sorption process is said to be controlled by physical adsorption. The values of E were found to be 0.021 and 0.027 kJ∙mol⁻¹ for ACP and ACZ, respectively, implying physical adsorption is dominant [2].

The adsorption process was analyzed using three kinetic models, pseudo-first-order, pseudo-second-order and intraparticle diffusion kinetic model.

From Figure 10 the amount of nitrate ions adsorption on ACP and ACZ increases rapidly in first 10 minutes. This fast kinetics of adsorption may be due to the availability of large number of active sites on the adsorbent. Hence, equilibrium is rapidly achieved for both ACP and ACZ after 20 min, as compare to chitosan hydrogel requiring 10 h [7], and activated carbon magnetic nanoparticles 1 h [5].

The pseudo-first-order kinetic model has a high R² value and low RMSE and χ² values among all models as shown in Table 8. This implies that the pseudo-first-order kinetic model best describes the adsorption of the nitrate ions on ACP and ACZ. In addition, there was no significant variation between the experimental and calculated values of the quantity adsorbed (q_e) of the pseudo-first-order kinetic model for both adsorbents. A similar observation was reported by Afkhami et al. [45]for adsorption of nitrate ions on acid modified carbon cloth. The low R² value and high values of RMSE and χ² obtained from the intraparticle diffusion kinetic model reveals that the pore diffusion is not the

rate-controlling step. In addition, the value of C_i ≠ 0, suggest the adsorption is a complex process and involve more than one diffusive resistance process [5].