Introduction

The ever-increasing production, consumption, and release into environment of pharmaceutical and personal care products (PPCPs) has become a global concern1,2. Carbamazepine (CBZ) is one of the four most widely prescribed pharmaceuticals for the treatment of epilepsy and psychosis3,4. The extensive use and long durability/low degradability of CBZ have resulted in its detection in sewage, surface water, groundwater, and drinking water4,5. CBZ cannot be treated adequately using conventional water treatments. Therefore, research efforts are underway to remove CBZ from the aqueous phase using advanced methods, such as filtration6,7,8, biological processes9, advanced oxidation methods5,8,9,10,11,12, coagulation/flocculation/sedimentation13,14, and adsorption3,14,15,16,17,18,19,20. Among the various methods being explored, adsorption is particularly attractive because it is simple in design, easy to perform, cost-effective, and free of byproducts15,16.

Since carbon-based materials (CBMs) have high specific surface areas and hydrophobic characteristics, they are being explored for use in various fields21,22,23. They have also been studied widely for use as adsorbents with high sorption capacities for organic compounds24,25,26,27,28. Zhu et al.29 reported that the octanol/water distribution coefficient with respect to dissociation at pH 7 is proportional to the adsorption performance of the porous adsorbent or CBM used. They suggested that hydrophobic or π–π interactions are the major mechanisms of PPCP adsorption. Thus, the hydrophobic interactions between the PPCP in question and the adsorbent used have a determining effect on the adsorption process. The pKa of an organic molecule determines the specific pH at which protonation or deprotonation occurs. Hence, PPCPs deprotonate and form negative ions at pH < pKa, which inhibits the hydrophobic interactions between the CBM used and the PPCP30,31. In addition, the ionic strength of the aqueous phase also affects the hydrophilic and hydrophobic interactions32. Therefore, the adsorption efficiency of CBZ, which is a representative persistent PPCP and does not readily undergo biological and physicochemical degradation33,34, can be controlled based on the hydrophobicity of the adsorbent used and the environmental conditions such as pH and ionic strength.

Santa Barbara Amorphous-15 (SBA) is a stable and mesoporous silica material. The effects of the length of the alkyl chain attached to quaternized SBA-15 (QSBA) on its hydrophobic and hydrophilic adsorption properties have been studied35,36. For example, QSBA with a long alkyl chain shows high adsorption for diclofenac (DCF) owing to both the hydrophobic interactions of the long alkyl chain and the electrostatic attraction of the N+ species36. For the nitrate ion, electrostatic adsorption on the N+ species of the quaternary ammonium occurs readily even in the presence of competing oxyanions, such as bicarbonate, phosphate, and sulfate ions. This is owing to the high nitrate selectivity of QSBA because of its long-alkyl-chain-based hydrophobicity from higher hydration energy of nitrate than one of the other oxyanions35. By controlling the hydrophobicity based on the length of the alkyl chain, the adsorption capacity and selectivity for the target contaminants can be improved, and the adsorption mechanism can be elucidated35,36. However, in previous studies, the adsorption characteristics have been investigated only with respect to PPCPs based on different functional groups37,38, and there has been no research on the adsorption characteristics of QSBA for PPCPs based on its hydrophobicity; the exception is DCF36. CBZ has a lower hydrophobicity than DCF but does not dissociate under general pH conditions because of its high pKa39,40,41,42. Therefore, its adsorption on QSBA is expected to be different from that of DCF in terms of the adsorption capacity, which would depend on the alkyl chain length. The results obtained for CBZ can be utilized to propose an appropriate alkyl chain length for QSBA in removing various PPCPs, considering their characteristics.

In this study, we examined the effects of the hydrophobicity of QSBA on CBZ adsorption by varying the alkyl chain length of QSBA as well as the aqueous conditions via batch experiments. The equilibrium adsorption capacity was evaluated by testing the effects of the initial CBZ concentration on its adsorption on QSBA with alkyl chains of different lengths (trimethyl, dimethyloctyl, and dimethyoctadecyl, which resulted in C1QSBA, C8QSBA, and C18QSBA, respectively). It was hypothesized that the changes in the hydrophobicity of QSBA would result in variations in the kinetics and equilibrium adsorption characteristics for hydrophobic CBZ. The pH and ionic strength of the test solution were also varied. This study provides additional insights into the molecular structure of QSBA for CBZ adsorption as well as the optimal characteristics of wastewater for the removal of CBZ.

Materials and methods

Quaternized SBA-15

C1QSBA, C8QSBA, and C18QSBA were prepared using a previously reported method35. Briefly, C1QSBA was prepared according to the following procedure: 0.1 mol of trimethyl[3-(trimethoxysilyl)propyl]ammonium chloride (50% in methanol, Tokyo Chemical Industry, Tokyo, Japan) and 6 g of SBA (Sigma Aldrich, MO, USA) were stirred for 1 h in 100 mL of toluene (99.5%, Daejung, Siheung, Republic of Korea). The mixture was then refluxed with 1 mL of deionized (DI) water at 100 °C for 48 h. Next, the slurry was treated with 0.1 M NaCl, separated using a 0.45-μm polyvinylidene fluoride (PVDF) filter, and dried at 65 °C in a drying oven until use. C8QSBA and C18QSBA were also prepared using the same processes as that employed for C1QSBA; the difference was that dimethyloctyl[3-(trimethoxysilyl)propyl]ammonium chloride and dimethyloctadecyl[3-(trimethoxysilyl)propyl]ammonium chloride (42% in methanol, Sigma Aldrich, MO, USA), respectively, were used instead of trimethyl[3-(trimethoxysilyl)propyl]ammonium chloride. Dimethyloctyl [3-(trimethoxysilyl)propyl]ammonium chloride was synthesized by reacting 0.1 mol of (3-chloropropyl)trimethoxysilane (≥ 97%, Sigma Aldrich, MO, USA) and 0.1 mol of N,N-dimethyloctylamine (95%, Sigma Aldrich, MO, USA) at 85 °C for 48 h.

Test solution of CBZ

A 100 mg/L CBZ stock solution was prepared by dissolving 50 mg of CBZ (≥ 98%, Sigma Aldrich) in 10 mL of methanol and then diluting it to 500 mL with DI water. There was no shift in the λmax value (285 nm) of CBZ with the pH (Fig. S1). Therefore, the calibration curves of CBZ were obtained by measuring the absorbance of a serially diluted solution at 285 nm using an ultraviolet–visible spectrophotometer (Optizen POP, Mecasys, Korea) and 1-cm quart cells for each pH.

Batch experiments

Because CBZ has high hydrophobicity and low solubility in water, it is difficult to test a wide range of concentrations in aqueous solutions. Furthermore, the dissociation of molecules greatly influences their solubility. Therefore, in this study, we examined the solubility to determine the concentration range for dissociation based on Eq. (1)39:

$${S}_{H}={S}_{0}\left(1+\frac{{K}_{a}}{\left[{H}^{+}\right]}\right) $$
(1)

where SH is the solubility in water at a specific pH (mg/L), and S0 is the intrinsic solubility of an undissociated molecule. Because the pKa of CBZ is very high (13.9), CBZ remains intact with low solubility at pH < 13.9. Therefore, the batch experiments to confirm its adsorption were conducted at concentrations of 100 mg/L or less, as S0 is in the range of 112–236 mg/L.

Batch experiments were also performed for QSBA by varying the contact time, initial CBZ concentration, pH, and ionic strength. All these experiments were performed using 0.03 g of either C1QSBA, C8QSBA, or C18QSBA. A 30 mL of the CBZ solution was poured into a 50-mL conical tube and incubated at 150 rpm and 25 °C in a shaking incubator. For each batch condition, the experiments were conducted in duplicate. After the reaction, the QSBA and solution were separated using a 0.45-μm PVDF filter. The CBZ concentrations before and after the batch experiment were calculated using a calibration curve.

The reaction times, CBZ concentrations, pH, and ionic strengths are listed in Table S1, which also lists the reaction conditions. The reaction time was varied from 5 to 360 min at a fixed CBZ concentration of 40 mg/L. In the other experiments, the CBZ solution and QSBA were allowed to react for 24 h. The equilibrium adsorption capacity was measured for various initial concentrations of CBZ (2–100 mg/L). The effect of the pH was assessed by adjusting the initial pH of the CBZ solution (40 mg/L) to 2, 4, 6, 8, and 10 using 0.1 M HCl and 0.1 M NaOH. The effect of the ionic strength was evaluated by adding 0.1–100 mM NaCl to the CBZ solution (40 mg/L).

Data analysis

The amount of CBZ adsorbed during the reaction time (qt) experiments was determined by fitting the data using various kinetic models (pseudo-first-order43, pseudo-second-order44, and Elovich45 models, Table S1), while the amount of CBZ adsorbed at equilibrium (qe) during the initial concentration experiments was fitted using the Freundlich46, Langmuir47, and Redlich–Peterson48 models (Table S2). The optimal parameters for each model for each QSBA were obtained via nonlinear regression using the solver function in Excel 2019 (Microsoft Corporation, WA, USA). The coefficient of determination (Eq. 2) and sum of the squared error (Eq. 3) were used as the error functions for model comparison.

$${R}^{2}= \frac{{\sum }_{i=1}^{m}{{\left({y}_{c}-\overline{{y }_{e}}\right)}_{i}}^{2}}{{\sum }_{i=1}^{m}{{\left({y}_{c}-\overline{{y }_{e}}\right)}_{i}}^{2}+{\sum }_{i=1}^{m}{{\left({y}_{c}-{y}_{e}\right)}_{i}}^{2}}$$
(2)
$$SSE = {\sum }_{i=1}^{n}{{\left({y}_{e}-{y}_{c}\right)}_{i}}^{2}$$
(3)

where R2, Coefficient of determination; SSE, Sum of the squared error; yc, Adsorption capacity calculated using the model; ye, Adsorption capacity measured experimentally; \(\overline{{y }_{e}}\), Average measured adsorption capacity.

Characteristics of quaternized SBA-15 and CBZ

Figure 1 shows a schematic of the procedure for preparing the QSBA samples and digital images of the samples. X-ray photoelectron spectroscopy (XPS), 13C solid-state nuclear magnetic resonance (NMR) spectroscopy, and Fourier-transform infrared (FT-IR) spectroscopy were performed on C1QSBA, C8QSBA, and C18QSBA in a previous study35. XPS confirmed that quaternary ammonium was well-crosslinked to the SBA surface (Fig. S2). The 13C NMR and FT-IR spectra also showed that the alkyl chains were well grafted, as intended (Figs. S3 and S4, respectively).

Figure 1
figure 1

Schematics of the procedures for preparing QSBA samples and their digital images: (a) C1QSBA, (b) C8QSBA, and (c) C18QSBA. Figure was modified from Kang and Kim35.

Full size image

Table 1 lists the major chemical characteristics of CBZ. The octanol–water partitioning coefficient (log Kow) value of 2.25 indicates that CBZ is hydrophobic. The high pKa (13.9) indicates that, in aqueous solutions, CBZ exists in an undissociated state at almost all pH.

Table 1 Characteristics of CBZ.
Full size table

Results and discussion

Adsorption kinetics

Fig. 2 shows the temporal trend for the adsorption of CBZ by QSBA. The adsorption process reached equilibrium within 120 min. Table 2 lists the parameters for the various kinetics models. The fitting quality as determined based on the coefficient of determination (R2) was the best in the case of the pseudo-first-order model (Fig. S5). The equilibrium qt (i.e., qe) values for C1QSBA, C8QSBA, and C18QSBA as calculated using the pseudo-first-order model were 0.618, 2.279, and 10.988 mg/g, respectively. Thus, the qt value varied with the alkyl chain length of QSBA. CBZ adsorption was enhanced by increasing the alkyl chain length of QSBA. In a previous study, C8QSBA showed an adsorption capacity as high as 593 mg/g for hydrophobic and dissociated DCF36. In this study, the hydrophobic adsorption of CBZ by C18QSBA was even greater, owing to the longer alkyl chain of the latter. Thus, it was confirmed that the adsorption of undissociated PPCP molecules can be significantly enhanced by using QSBA with a long alkyl chain.

Figure 2
figure 2

Effect of contact time on adsorption of CBZ by QSBA.

Full size image
Table 2 Kinetics model parameters for adsorption of CBZ by QSBA.
Full size table

The k1 value was calculated based on the fitted pseudo-first-order model to characterize the adsorption of CBZ by QSBA. The k1 value decreased with increasing alkyl chain length, with the values for C1QSBA, C8QSBA, and C18QSBA, being 3.623, 0.327, and 0.198 L/min, respectively. The trends for qe and k1 were opposite because an increase in the alkyl chain of QSBA meant more adsorption sites and thus more time required to reach equilibrium.

Effect of initial CBZ concentration

Figure 3 shows the effect of the initial concentration of CBZ on its adsorption by QSBA. The observed data were analyzed using various isotherm models such as the Freundlich, Langmuir, and Redlich–Peterson models (Fig. S6). Table 3 lists the parameters of the isotherm models. Similar to the trend seen in the kinetics, the QSBA samples with longer alkyl chains exhibited higher maximum adsorption capacities (Qm) in the case of the Langmuir model (3.14, 6.56, and 24.5 mg/g for C1QSBA, C8QSBA, and C18QSBA, respectively). As mentioned previously, the adsorption capacities of C8QSBA and C18QSBA were higher because of favorable hydrophobic interactions.

Figure 3
figure 3

Effect of initial concentration of CBZ on its adsorption by QSBA.

Full size image
Table 3 Equilibrium model parameters for CBZ adsorption by QSBA.
Full size table

Regardless of the length of the alkyl chain, the Redlich–Peterson model was the most suitable of the three isotherm models used, based on their SSE and R2 values (Table 3). All three models showed high R2 values (> 0.967), which were acceptable for model fitting. This was probably because the experiments were not performed using high CBZ concentrations. The Langmuir model assumes monolayer adsorption, while the Freundlich model assumes multilayer adsorption32. However, both the Freundlich and Langmuir models showed linear adsorption characteristics within a certain early concentration range. Detailed characterization of the adsorption models, such as considering single and multiple layers, is not possible owing to the limited solubility of CBZ. Thus, the standard Langmuir model is a suitable one because QSBA would have a larger surface available for adsorption compared with that of CBZ. The adsorption characteristics of a number of PPCPs with limited solubility have been described previously using the Langmuir model, including CBZ50,51,52, ibuprofen29, levofloxacin53, sulfamethoxazole54,55,56, tylosin55, and 17β-estradiol56.

Table S3 compares the adsorption capacities of the various adsorbents for CBZ. Most adsorbents are based on activated carbon, mesoporous silica, or metal–organic frameworks. The data in Table S3 suggest that the primary mechanism for the adsorption of CBZ is hydrophobic interactions16,17,18,19,20. Deng et al.19 and Jun et al.20 reported Qm values of 104.17 and 250.4 mg/g for CBZ using carbon-dot-modified magnetic carbon nanotubes and a metal–organic framework (Basolite A100), respectively. These values are much larger than that of C18QSBA (24.5 mg/g). However, no study has attempted to control the hydrophobicity of the adsorbent to confirm that hydrophobic interactions are indeed the adsorption mechanism responsible for the removal of CBZ. In this study, we show clearly that the removal of CBZ is improved owing to the higher hydrophobicity owing to the longer alkyl chains. We also analyzed the effects of the pH and ionic strength to confirm that the enhancement in the hydrophobic interactions is not affected by the various wastewater characteristics.

Effect of initial pH

In a previous study, the adsorption of DCF onto C8QSBA was reduced after an increase in the initial pH from 5 to 12. This suggests that DCF adsorption onto C8QSBA involves not only hydrophobic interactions but also an anion exchange with the N+ moiety of the quaternary ammonium group. Unlike CBZ, DCF dissociates into negatively charged molecules at pH > 4.15 (pKa = 4.15)35. Consequently, the pH controls both the hydrophobic and hydrophilic interactions and determines the adsorption efficiency of DCF by QSBA. Figure 4 shows the effect of the initial pH on CBZ removal by QSBA. Despite the variations in the initial pH, stable adsorption capacities were observed (0.41–0.92, 1.70–2.24, and 7.56–9.10 mg/g for C1QSBA, C8QSBA, and C18QSBA, respectively); the exception was when the pH was 2 and C18QSBA was used. Thus, the pH had a limited effect on the interactions between CBZ and QSBA. However, C18QSBA showed an improved adsorption capacity of 12.06 ± 0.07 mg/g at pH 2. The pH of the CBZ solution was adjusted using HCl and NaOH. Thus, it was expected that an extremely low pH would improve the hydrophobic interactions at high concentrations of H+ and Cl-. The H+ and Cl− concentrations at pH 2 were 10 mM. It is known that when ions and proteins are present in high concentrations, they compete to interact with the water molecules, and unreacted proteins are precipitated by the hydrophobic interactions57. Similarly, Bautista-Toledo et al.58 explained that the enhancement in the adsorption of sodium dodecylbenzenesulfonate (SDBS) on activated carbon with increasing ionic strength was owing to the decreased solvation of SDBS because of the high ionic strength, which increased the hydrophobic-interaction-based adsorption. Likewise, the high concentrations of H+ and Cl− may have enhanced the hydrophobic interactions between C18QSBA and CBZ in the present study. Similarly, the Na+ and OH− ions probably also aided the enhancement in the hydrophobic interactions when present in a high concentration at pH 10. However, the concentrations of Na+ and OH− were 0.1 mM at pH 10 and only 1/100th of those of H+ and Cl− at pH 2. The effect of the ion strength on the hydrophobic interactions between CBZ and QSBA was verified using NaCl, as described in the next section.

Figure 4
figure 4

Effect of pH on CBZ adsorption by QSBA.

Full size image

Effect of ionic strength

Figure 5 shows the effect of the ionic strength on the adsorption of CBZ by QSBA. The adsorption capacity increased with increasing ionic strength. The amounts of CBZ adsorbed were 0.70 ± 0.09, 2.25 ± 0.14, and 9.27 ± 0.42 mg/g for C1QSBA, C8QSBA, and C18QSBA when 0.1 mM NaCl was used. In contrast, the adsorption capacities increased to 0.96 ± 0.33, 2.77 ± 0.29, and 14.94 ± 0.17 mg/g for C1QSBA, C8QSBA, and C18QSBA, respectively, in the case of 100 mM NaCl. These values are 1.38, 1.23, and 1.61 times higher, respectively, than the corresponding ones for 0.1 mM NaCl. The high salt concentration enhanced the hydrophobic interactions and shielded the electrostatic attraction between CBZ and the N+ moiety of QSBA32,59. Interestingly, only in the case of C18QSBA did the adsorption capacity exhibit high sensitivity to the ionic strength, with C18QSBA showing an enhanced adsorption capacity of 10.60 ± 0.92 mg/g when 10 mM NaCl was used. The ionic concentrations in this case were similar to those at pH 2. Thus, it was confirmed that the hydrophobic interactions are predominantly controlled by the ionic strength and not the pH. The reason for the highest sensitivity of C18QSBA was not clear. However, we believe that its longer alkyl chain exhibits stronger hydrophobic interactions in saline conditions. This means that the alkyl chain length can be increased to improve the efficiency of CBZ removal from wastewater samples with high ionic strength. The critical ionic strength of 100 mM NaCl (≒ 12–20 mS/cm60,61) is higher than that of surface water (0.9 mS/cm), sanitary sewage (1.5–3.0 mS/cm), and treated water (0.22–0.37 mS/cm) but similar to that of industrial wastewater (35–70 mS/cm) and sea water (30 mS/cm)62,63,64. This characteristic should allow for a high CBZ adsorption capacity in strongly ionic wastewaters. However, additional investigations are required to confirm the applicability of QSBA for use with wastewaters with high ionic strength.

Figure 5
figure 5

Effect of ionic strength on CBZ adsorption by QSBA.

Full size image

Conclusions

Batch experiments were performed to investigate the primary mechanism of CBZ adsorption on QSBA with different alkyl chain lengths. The efficiency of CBZ absorption by QSBA increased with increases in the reaction time and alkyl chain length of QSBA. Based on the Langmuir isotherm model, the maximum sorption capacities for C1QSBA, C8QSBA, and C18QSBA were determined to be 3.14, 6.56, and 24.5 mg/g, respectively. Regarding the effect of the initial pH, the adsorption capacity was mostly stable within the pH range of 4–10; the exception was a pH of 2. Because the pka value of CBZ is 13.9, undissociated CBZ does not interact with the N+ ions of QSBA. Consequently, it was assumed that hydrophobic interactions would be dominant in the pH range investigated in this study. The increase in CBZ adsorption at higher ionic strengths (similar to those of actual wastewater) is also attributable to the hydrophobic interactions between the alkyl chains of QSBA and CBZ. Thus, by controlling the alkyl chain length of QSBA, we were able to elucidate the CBZ adsorption mechanism in detail. Moreover, by comparing the adsorption characteristics of various PPCPs by adjusting the alkyl chain length of QSBA, it should not only be possible to determine the adsorption mechanism but also use the appropriate adsorbent for PPCPs based on their characteristics such as their hydrophobicity.