Introduction

Caffeine is a weak alkaloid of the methylxanthine family, 1,3,7-timetylxanthine. This substance is classified as a drug, is a nervous system stimulant, and causes transient changes in blood pressure. Caffeine is used as an adjuvant in many pharmaceutical combinations to increase its analgesic effects. It is present in beverages such as coffee, teas, chocolates, and soft drinks. The metabolism of this substance is rapid where only a small amount (1–10%) is excreted (Thorn et al. 2012; Portinho et al. 2017; Beltrame et al. 2018; Ptaszkowska-Koniarz et al. 2018; Yamamoto et al. 2018; González et al. 2019). Many pharmaceutical compounds, such as caffeine, are considered water contaminants and classified as emerging pollutants. This class of pollutants now attracted attention as an environmental problem due to its presence that had been recently detected in the environment. Several researchers consider caffeine as an indicator substance of human pollution due its resilience to conventional water and wastewater treatments. Therefore, this substance has often been found in surface water and groundwater (Álvarez-Torrellas et al. 2017; Portinho et al. 2017; Wang et al. 2017; Beltrame et al. 2018).

Many methods can be used to reduce caffeine concentration in water. Among these methods, adsorption stands out due to its simplicity of design and operation, minimal energy requirements, possibility of adsorbent regeneration, and no generation of dangerous by-products (Álvarez-Torrellas et al. 2017). The wide range of materials to be used as adsorbents is one of the great advantages of adsorption. This characteristic makes this technique still of great interest, especially in the search for materials with high efficiency of adsorption, low production costs, and high capacity of regeneration.

In order to conduct an efficient evaluation of DS removal and to maximize the adsorption yield, an optimization study should be developed (Wakkel et al. 2019). The response surface methodology (RSM) is a combination of statistical and mathematical techniques which allow investigating the effect of several independent variables. This approach enables to obtain empirical models and to design, improve, and optimize several types of processes. The use of RSM, a multivariate optimization technique, is convenient once it employs experimental data and permits to evaluate the interactive effect of variables on process performance. Thus, RSM overcomes situations requiring a large number of experiments minimizing additional chemicals, time, and expensive analysis, promoting a reduction of extra costs (Biswas et al. 2019; Şahan 2019). RSM advantages guaranteed its application in several works regarding the removal of contaminants from water to process parameter optimization, such as adsorption (Kaynar et al. 2018; Biswas et al. 2019; Deng and Chen 2019; Hasan and Setiabudi 2019; Kaur et al. 2019; Şahan 2019; Sharifpour et al. 2019; Wakkel et al. 2019), electrocoagulation (Barsç and Turkay 2016; Murdani et al. 2018; Rabahi et al. 2018; Deveci et al. 2019; Karamati-Niaragh et al. 2019), Fenton reaction (Saeed et al. 2015; Xie et al. 2016; Liu et al. 2018), electrochemical oxidation (Garg and Prasad 2015; Domínguez et al. 2016; Darvishmotevalli et al. 2019; Duarte et al. 2019), and photocatalytic degradation (Mirzaei et al. 2018; Galedari et al. 2019; Karimi et al. 2019).

Activated carbons are among the most used materials as adsorbent. Recently, several researches have been conducted to obtain new activated carbons from renewable sources, such as agroindustry by-products (Suzuki et al. 2007; Beltrame et al. 2018), papaya seeds (Weber et al. 2013), Syagrus oleracea endocarp (dos Santos et al. 2019a), Wodyetia bifurcata endocarp (dos Santos et al. 2019b), coconut shells (Chandana et al. 2019), rice husks (Lv et al. 2020), wood sawmills (Ramirez et al. 2020), and apple seed shells (Abatan et al. 2019). The palm (Elaeis guineensis), known as dendê, is a typical Brazilian palm tree that is the higher oil producer per unit of cultivated area among other oleaginous plants in Brazil, an average of 4–6 t of oil/ha-year (Queiroz et al. 2012) (De Azevedo et al. 2014). The palm endocarp is one of the largest by-products generated during palm oil production.

The main objective of this work was to evaluate the adsorption potential of the Elaeis guineensis activated carbon in the removal of the caffeine from water through batch adsorption studies. Response surface methodology (RSM) was used for the optimization of independent variables mass dosage, caffeine initial concentration, and pH to obtain the maximum caffeine removal. Kinetic, equilibrium, and thermodynamics studies were conducted to evaluate the adsorption mechanism.

Materials and methods

Materials

The adsorbent used in the present study was a commercial activated carbon obtained from Elaeis guineensis endocarp (Pelegrini Carbon). First of all, it was triturated and sieved in order to obtain particles with a medium diameter of 337.5 nm. A stock solution of caffeine (1000 mg/L) was prepared by dissolving the analytical standard in water, which was used to prepare all the work solutions by appropriated dilution. Caffeine quantification was performed using a spectrophotometer Shimadzu UV mini-1240, with absorbance measurements at 273 nm. Calibration curve was plotted with concentrations ranging from 0.5 to 6.0 mg/L and used to determine the adsorbate concentration after all adsorption assays.

Adsorbent characterization

For the determination of the adsorbent point of zero charge (pHpzc), 0.02 g of the activated carbon was added to Erlenmeyer flasks with 20 mL of NaCl 0.1 mol/L solution. The pH values were adjusted to 1.0 up to 12.0 using HCl or NaOH solutions (1.0 mol/L). The samples were stirred at 140 rpm (25 °C) for 24 h. After that, the mixtures were filtered and the final pH values were measured (Regalbuto 2006). Fourier transform infrared spectroscopy (FT-IR) was performed using a spectrophotometer Shimadzu/IRPrestige-21 through the KBr method. Spectra were obtained in the range of 4000 to 400 cm−1 with transmittance of 50 scans. Thermogravimetric analysis (TGA) was performed using the term scale model Shimadzu DTG-60H, in which 7 mg of the adsorbent was heated until 900 °C at the rate of 10 °C/min in an inert atmosphere (nitrogen gas) with a flow rate of 50 mL/min. The surface morphology of the adsorbent material was analyzed by the SEM Shimadzu SSX-550 model. N2 adsorption/desorption analysis was performed in a micrometrics equipment (ASAP 2020) at − 196 °C (77 K), treating the sample previously by degassing for 12 h, under vacuum (2 μm of mercury) at 350 °C, in order to remove any species on its surface. A surface external area was determined by the BET method, and the volume of pores and the distribution of their size were specified by the BJH method.

Kinetic studies

Kinetic studies were performed using a Dubnoff (SPLabor/SP-158/22/A) bath with orbital agitation. A total of 0.1 g of the adsorbent was added to Erlenmeyer flasks with 25 mL of adsorbate solution (20 mg/L) and stirred at 135 rpm (30 °C). At the end of the adsorption process, the samples were centrifuged (Solab/SL-700) at 2000 rpm for 5 min and caffeine final concentration was measured. Samples were collected at 5, 10, 15, 30, 60, 120, 180, 240, and 300 min for the construction of the kinetic curve.

In order to evaluate the adsorption capacity of the adsorbents, the adsorbed amount (qt) in milligrams per gram and the caffeine removal (R) in percentage were calculated using Eqs. 1 and 2, respectively.

$$ {q}_{\mathrm{t}}=\frac{\left({C}_0-{C}_{\mathrm{e}}\right)}{m}\times V $$
(1)
$$ R=\frac{C_0-{C}_{\mathrm{e}}}{C_0}\times 100 $$
(2)

where C0 and Ce are the initial and concentration values (mg/L), respectively, m is the mass (g) of the adsorbent, and V the volume (L) of the adsorbate solution.

The experimental data were adjusted with pseudo-first-order (Eq. 3) and pseudo-second-order (Eq. 4) models (Lagergren 1898; Ho and McKay 1999).

$$ {q}_{\mathrm{t}}={q}_{\mathrm{e}}\left(1-{\exp}^{-{k}_1\mathrm{t}}\right) $$
(3)
$$ {q}_{\mathrm{t}}=\frac{k_2t{q_{\mathrm{e}}}^2}{\left(1+{k}_2t{q}_{\mathrm{e}}\right)} $$
(4)

where k1 and k2 are the first- and second-order adsorption kinetics (min−1 and g mg−1 h−1), respectively, and qt and qe are the adsorbed adsorbent (mg g−1) in equilibrium time, respectively.

RSM methodology

The effects of selected independent process variables were evaluated by the response surface methodology (RSM). Single and synergetic effects of 3 variables, i.e., X1 adsorbent dosage (g), X2 caffeine initial concentration (mg/L), and X3 pH, were evaluated at 2 levels with the experimental response Y caffeine removal (Eq. 2). The total number of experiments of the 23 full experimental design was given as the sum of the 2k + n0 (2k, factorial runs; k, the number of independent process variables; and n0, the center runs). Then, it was conducted 8 experiments + 8 duplicates + 3 central point runs, consisting of 19 experiments. The experimental levels of independent process variables are presented in Table 1. Equation 5 was used to predict the optimum condition of DS removal related to the interaction between dependent and independent variables. Besides, the analysis of variance (ANOVA) was used to validate the adequacy of model.

$$ Y={b}_0+{\sum}_{i=1}^n{b}_i{x}_i+{\sum}_{i=1}^n{b}_{ii}{x}_i^2+{\sum}_{i=1}^{n-1}{\sum}_{j=2}^n{b}_{ii}{x}_i{x}_j+\upepsilon $$
(5)
Table 1 Experimental levels of independent process variables
Full size table

where Y is response; b0 represents the intercept; bij, bii, and bi are coefficients; n, number of variables; xi and xj, independent variables; and ε, the error (Kaynar et al. 2018).

Equilibrium studies

Equilibrium studies were performed using the contact time obtained in the kinetic studies, at 30, 40, 50, and 60 °C and DS concentrations of 50, 100, 200, 500, 750, and 1000 mg L−1. The experimental data obtained were adjusted through the nonlinear regression using Langmuir (Eq. 6) (Langmuir 1918), Freundlich (Eq. 7) (Freundlich and Freundlich 1906, Redlich-Peterson (Eq. 8) (Redlich and Peterson 1959), and Sips (Eq. 9) (Sips 1948) models.

$$ {q}_{\mathrm{e}}=\frac{\mathrm{ce}Q{k}_{\mathrm{L}}}{1+\left(\mathrm{ce}{k}_{\mathrm{L}}\right)} $$
(6)
$$ {q}_{\mathrm{e}}={k}_{\mathrm{F}}{\mathrm{ce}}^{\frac{1}{n}} $$
(7)
$$ {q}_{\mathrm{e}}=\frac{\mathrm{ce}{k}_{\mathrm{rp}}}{\left(1+{a}_{\mathrm{rp}}{\mathrm{ce}}^{b_{\mathrm{rp}}}\right)} $$
(8)
$$ {q}_{\mathrm{e}}=\frac{q_{\mathrm{S}}{\left({k}_{\mathrm{S}}{C}_{\mathrm{e}}\right)}^{m_{\mathrm{S}}}}{1+{\left({k}_{\mathrm{S}}{C}_{\mathrm{e}}\right)}^{m_{\mathrm{S}}}} $$
(9)

where Q is the maximum adsorption capacity (mg g−1); kL is the Langmuir constant (L/mg); kF is the Freundlich constant (mg g−1)(mg L−1)−1/n; 1/n is the heterogeneity factor; krp (L mg−1), arp (L mg−1), and β are Redlich-Peterson constants; qS is the maximum adsorption capacity from the Sips model (mg g−1); KS is the Sips constant (L mg−1); and mS is the exponent of the Sips model.

In order to determine the accuracy of the models, the experimental data were evaluated using the correlation coefficient (R2) and the relative mean error (ARE), presented in Eqs. 10 and 11, respectively (Piccin Jr et al. 2017).

$$ {R}^2=1-\frac{\sum_{i=1}^n{\left({y}_{\mathrm{i},\exp }-{y}_{\mathrm{i},\operatorname{mod}}\right)}^2}{\sum_{i=1}^n{\left({y}_{\mathrm{i},\exp }-\overline{y_{\mathrm{i},\operatorname{mod}}}\right)}^2} $$
(10)
$$ \mathrm{ARE}=\frac{100}{n}{\sum}_{i=1}^n\left|\frac{y_{\mathrm{i},\exp }-{y}_{\mathrm{i},\operatorname{mod}}}{y_{\mathrm{i},\exp }}\right| $$
(11)

where yexp is the value obtained experimentally, ymod is the value predicted by the model, np is the number of parameters of the model, and n is the number of experimental points.

Adsorption assays using real matrixes

Tests were performed with real water samples, preparing solutions of 20 mg L−1 from tap, ultrapure, and mineral water. The parameters used were based on the previous study, 0.20 g of adsorbent, 25 mL of solution volume at 20 mg L−1 of caffeine initial concentration, 30 °C, pH 2, and 4 h. Samplings were performed in 15, 30, 45, 60, 120, 180, and 240 min, in duplicate. Solution concentration was determined by spectrophotometer UV-Vis (Shimadzu/UV-1800).

Results and discussions

Adsorbent characterization

The pHpzc is obtained when the final pH is independent of initial pH (buffer effect) or when final pH is equal to initial pH. The pHpzc indicates the pH at which the adsorbent has a net zero surface charge. The adsorbent has a positive charge when the solution pH is lower than the pHpzc. On the other hand, when the solution pH is higher than the pHpzc, the adsorbent is negatively charged. Figure 1 depicts the result obtained for the activated carbon pHpzc determination, in which the value obtained was around 6.3 (Kong et al. 2013).

Fig. 1
figure 1

Determination of pHpzc

Full size image

The FT-IR spectrum, presented in Fig. 2, was obtained in the range of 4,000 to 400 cm−1. The band between 3600 and 3200 cm−1, with peak around 3461 cm−1, is characteristic of group stretch vibrations -OH that may be related to the presence of hydroxyl and the water chemisorbed on carbon surface (Álvarez et al. 2015). The bands evidenced in 2956, 2924, 2850, 1425, and 460 cm−1 can be attributed to the presence of aliphatic groups, such as alkanes and alkenes, corresponding to C-H bonds (Sotelo et al. 2012; Álvarez et al. 2015). The band indicated in the region between 1650 and 1558 cm−1 is assigned to links C=C and C=O, present in carboxyl, carbonyl, and aromatic carbon radicals. Bands in 1425 and between 1118 and 1233 cm−1 are characteristics of phenolic and lactam groups (Fonts et al. 2009; Royer et al. 2009; Sotelo et al. 2012; Foletto et al. 2013; Larous and Meniai 2016).

Fig. 2
figure 2

Palm endocarp activated carbon FT-IR spectrum

Full size image

SEM images at different magnifications are shown in Fig. 3. The activated carbon is a fine granular material confirming the diameters around 337.5 nm obtained by sieving. The material presented an irregular and heterogenic surface with the presence of pores, swellings, and canals. These characteristics are propitious for adsorption since the interaction between liquid and solid may occur in the internal and external surfaces (Georgin et al. 2019).

Fig. 3
figure 3

SEM images of the palm endocarp activated carbon surface

Full size image

N2 adsorption/desorption isotherm presented in Fig. 4 a may be classified as type IV according to IUPAC (Thommes et al. 2015). This type of isotherm is characteristic of mesoporous materials with an evident hysteresis, due to increased pressure that causes an increase in the volume of N2 adsorbed (Beltrame et al. 2018). In addition, the desorption process presents a very marked and open hysteresis, indicating the occurrence of a sudden nitrogen desorption or its enclosure in the material pores, and a gas condensation may occur, something common in mesoporous materials (Zhang et al. 2015). According to the distribution of mesoporous sizes in Fig. 4 b, varying from 21 to 40 Å (2.50–4.0 nm), with an average diameter of 30.51 Å. Macroporous sizes appear in the range of 90 to 130 Å (9.0–13.0 nm) and 220 to 320 Å (22.0–32.0 nm). In addition, the material presented a specific area of 407.66 m2 g−1 and total pore volume of 0.169 cm3 g−1. Ferreira et al. (2015) obtained different values, 672 m2 g−1 and 0.369 cm3 g−1, for dendê mesocarp activated carbon, which considering the variability of climate, soil, harvest, and other agricultural characteristics, it is totally expected.

Fig. 4.
figure 4

a N2 adsorption/desorption isotherm. b Pore diameter distribution

Full size image

RSM analysis

The final experimental design matrix for the three independent variables with response is presented in Table 2. The experimental mathematical model in terms of coded variables and with more significant coefficients is presented by Eq. 12.

Table 2 Experimental design matrix
Full size table
$$ Y=\underset{\left(\pm \mathrm{0,570}\right)}{\mathrm{67,748}}+\underset{\left(\pm \mathrm{0,669}\right)}{\mathrm{9,149}}{X}_1-\underset{\left(\pm \mathrm{0,669}\right)}{\mathrm{16,258}}{X}_2-\underset{\left(\pm \mathrm{0,669}\right)}{\mathrm{4,330}}{X}_3-\underset{\left(\pm \mathrm{0,669}\right)}{\mathrm{2,909}}{X}_1{X}_2 $$
(12)

Analysis of variance (ANOVA) of mathematical/statistical model is shown in Table 3. The model was statistically significant since R2 was high and values of lack of fit and pure error were low indicating the model can predict successfully the experimental data. F test showed calculated F (96.33) was higher than the standard F (3.03); besides, the ration between calculated F and standard F was higher than 1.0 (Fcalculated/Fstandard = 31.79) confirming the model validation.

Table 3 ANOVA
Full size table

The response surface plots are presented in Fig. 5. Figure 5 a shows the effect of caffeine initial concentration and mass dosage on the percentage of removal. Removal increases with the decrease of concentration and increase of pH. In Fig. 5 b is presented the influence of pH and mass of adsorbent on the removal of caffeine. Removal increases with the augment of mass dosage; however, it is not observed a significant influence of pH. A similar behavior is observed in Fig. 5 c where the caffeine initial concentration had a more significant effect on the caffeine removal than pH. The augmentation of adsorbate concentration is proportional to the increase in the number of molecules in the medium competing for the available active sites on the adsorbent surface. When the adsorbent surface is saturated by adsorbates, the caffeine molecules remain in solution. The increase in dosage provides an increase in the removal. However, this direct relation is intimately linked with caffeine concentration and pH, since for low amounts of adsorbate, it was obtained high values of removal (> 74.55%). pH was the parameter with the lowest influence; the highest values of removal were obtained in acid medium. When the solution pH is higher than the pHpzc the adsorbent is negatively charged. In this condition, the caffeine is attracted by the adsorbent surface since they have opposite charges (Couto et al. 2015).

Fig. 5
figure 5

Response surface plots

Full size image

Kinetic study

Kinetic studies are fundamental to better understand the adsorption mechanisms involved as well as to evaluate the efficiency of the separation process. Assays performed from 5 to 300 min of contact between adsorbent and adsorbate (20 mg/L) showed that the equilibrium of caffeine adsorption onto palm endocarp activated carbon was reached after 5 h. Figure 6 shows the kinetic curve and the adjustment of the experimental data with the pseudo-first-order and pseudo-second-order models. The statistic parameters are presented in Table 4. According to the results obtained, the experimental data fitted better with the pseudo-secondo-order model, due to the higher determination coefficient (R2 = 0.96) and lower error (ARE = 8.77), when compared with the pseudo-first-order parameters (R2 = 0.91 and ARE = 11.87). It suggests that chemisorption is the dominant adsorption mechanisms involved, justifying the slow kinetics observed.

Fig. 6
figure 6

Kinetic curve obtained for caffeine adsorption onto palm endocarp activated carbon and adjustments to pseudo-first-order and pseudo-second-order models

Full size image
Table 4 Kinetic parameters obtained for the adjustment of the experimental data with pseudo-first-order and pseudo-second-order models
Full size table

Equilibrium studies

Isotherm curves were performed at 30, 40, 50, and 60 °C in order to evaluate the type of interaction between the adsorbent and the adsorbate. The experimental data, as well as the adjustments for Langmuir, Freundlich, Redlich-Peterson, and Sips models, are depicted in Fig. 7. The parameters obtained are shown in Table 5. From the results achieved, the temperature increase favored the adsorption, indicating an exothermic process for caffeine concentrations up to 500 mg/L, reaching the adsorptive capacity of 13.17 mg/g. However, for all isotherms, adsorption decreased when the initial adsorbate concentration was 750 mg/L. This result can be explained by saturation of adsorbent which, when the equilibrium was reached at 500 mg L−1, almost all active sites are occupied, reducing the removal of the dispersed molecules, decreasing adsorption efficiency.

Fig. 7
figure 7

Isotherms obtained for caffeine adsorption onto palm endocarp activated carbon at 30 °C (a), 40 °C (b), 50 °C (c), and 60 °C (d) and adjustments to Langmuir, Freundlich, Redlich-Peterson, and Sips models

Full size image
Table 5 Equilibrium parameters obtained for the adjustment of the experimental data with Langmuir, Freundlich, Redlich-Peterson, and Sips models
Full size table

Among all models, Redlich-Peterson and Sips were the ones who presented the best fit to the experimental data, based on the highest R2 values and the lowest ARE. Once both isotherms are hybrids of Langmuir and Freundlich’s models, they can overcome some limitations of the two parameter models. Therefore, in general, experimental data fitted better to the Sips model, presenting higher coefficient of determination (R2) and lower average relative error (ARE), predicting a maximum adsorption capacity of 13.5 mg g−1, closed to the experimental value. The isotherm profile indicates they can be classified as L-2 type, according Giles et al. (1960). It is evident a marked initial rise and a concavity in relation to the x-axis at low equilibrium concentration. This result is characteristic of systems in which the adsorbate has a strong affinity with adsorbent, and there is no significant competition of the solvent for active sites, allowing the formation of a monolayer on the surface (Gil et al. 2018). Besides, this type of isotherm indicates that more solute loading can be carried into the solid, as long as it is at lower concentrations (Couto et al. 2015).

In order to compare the effectiveness of activated carbon obtained from Elaeis guineensis endocarp in the removal of caffeine from aqueous solution, in Table 6 is presented a comparison between the maximum adsorption capacities (qm) of different adsorbents. Activated carbon obtained from Elaeis guineensis endocarp seems to be a promising material for caffeine removal since it presents qm values similar with other activated carbons as well as much more complex materials such nanotubes and composites.

Table 6 Comparison between the activated carbon obtained from Elaeis guineensis endocarp and other materials for caffeine removal
Full size table

Thermodynamic studies

Thermodynamics studies are important to understand the mechanism, spontaneity, and nature of adsorption and the adsorbent surface characteristics. Considering that the Sips model presented the best fit to the experimental data, Ks, the equilibrium thermodynamic constant, was used to calculate ΔG0 for all temperatures studied using Eq. 13. Then, by the plot of ln (Ks) versus (1/T), van’t Hoff plot, the values of ΔH0 and ΔS0 were found. Table 7 summarizes the thermodynamic parameters (ΔG0, ΔH0, and ΔS0) obtained.

Table 7 Thermodynamic parameters obtained for caffeine adsorption onto palm endocarp activated carbon
Full size table
$$ {\Delta G}^0=- RTln{K}_{\mathrm{s}} $$
(13)

where R is the universal gas constant (8.314 J mol−1 K−1), T is the temperature (K), and Ks is the thermodynamic equilibrium constant.

From the data obtained, the negative values of ΔG0 indicate that adsorption was spontaneous at all temperatures studied. In addition, more negative values are obtained with higher temperatures indicating adsorption is favored by temperature. By analyzing the ΔH0, the negative value indicates the process was exothermic; this behavior is related to the desolvation of water molecules from the solid surface for adsorption of caffeine molecules. The enthalpy value between 0 and − 40 kJ mol−1 presupposes a physisorption phenomenon. According to the positive S0, possibly some structural changes or readjustments in the adsorbate-adsorbent complex occurred demonstrating that the interaction is entropy-controlled (Silva et al. 2017; Lütke et al. 2019; Meili et al. 2019; dos Santos Lins et al. 2019). Similar results were obtained by Beltrame et al. (2018) and Anastopoulos and Pashalidis (2019), using activated carbon fibers from pineapple leaves and oxidized activated carbon derived from Luffa cylindrica for caffeine removal, respectively.

Adsorption in real water matrixes

Figure 8 shows the results of caffeine concentration behavior versus time according to the aqueous matrix. A higher percentage (98.54%) of caffeine removal was obtained using tap water, with lower and similar percentages for ultrapure water (91.64%) and mineral water (91.30%). In the tap water, several dissolved anions, such as sulfate and chloride, from water treatment process, may have interacted, electrostatically, with caffeine molecules, leaving it with an opposite load in relation with carbon surface, favoring the adsorption. This characteristic may be proven by the conductivity of water matrixes (Wu et al. 2020). Although mineral water presented higher conductivity than ultrapure water, dissolved anions presented milder action. Satisfactory results indicate that palm endocarp activated carbon has a high potential to treat the water contaminated with caffeine.

Fig. 8
figure 8

Adsorption in real matrixes

Full size image

Conclusions

The present contribution demonstrated the feasibility of using the palm endocarp activated carbon for caffeine removal from water. Assays performed through a complete experimental design 23 showed that the most significant tested parameter was adsorbate concentration, followed by mass of adsorbent and solution pH. The best response was achieved using 0.20 g of adsorbent, initial caffeine concentration of 20 mg/L, and pH 2, reaching a total removal of 95.8%. From the kinetic studies, equilibrium was attained after 5 h and the experimental data presented the best fit to the pseudo-second-order model, suggesting chemisorption as the predominant adsorption mechanism. The isotherms studies provided best fit to the Redlich-Peterson and Sips (adsorption capacity of 13.5 mg/g) models. Finally, thermodynamics calculations indicated an exothermic and spontaneous adsorption mechanism with structural modifications in the adsorbate-adsorbent interface. The positive results obtained regarding adsorption in different water matrixes demonstrated the ability to be used as an adsorbent agent in complex conditions. The adsorbent features and efficient removal results suggest a suitable material for caffeine withdrawn from water.