1 Introduction

Sulfonamide antibiotics (SAs), used for treatment and prevention of human and animal bacterial infections, are widely discharged into natural environments through domestic sewage, wastewater, and livestock manures without being metabolized (Hou et al. 2015; Li et al. 2018). The discharged antibiotics are long-standing and raise environmental risks such as resistant genes (Lye et al. 2019). Due to large discharge of SAs, sulfonamide-resistant genes have been identified as the most abundant resistant phenotypes in rivers of China (Dang et al. 2017). Several techniques have been applied to  the immobilization and removal of SAs, such as biodegradation (Chen and Xie 2018), electrochemistry (Deng et al. 2019), advanced oxidation (Yin et al. 2020) and adsorption (Mangla et al. 2022). Among them, adsorption is a promising method due to its high efficiency, simple operation, and low energy consumption. The partition and sorption of SAs on geo-sorbents determine the fate and mobility of antibiotics as well as their bioavailability (Gao and Pedersen 2005). Therefore, it is urgent to understand the sorption mechanism of SAs and further explore high-efficiency engineered materials to recover the SAs in water.

Biochar (BC), produced from pyrolysis of agricultural biomass, has been widely used as an excellent and low-cost adsorbent for environmental remediation (Zheng et al. 2013; Huang et al. 2021; Liu et al. 2023; Wang et al. 2022). BC can adsorb hydrophobic organic compounds due to its surface porosity, hydrophobicity and aromaticity (Jiang and Dai 2023). In contrast, the immobilization of ionizable organic compounds on BC is pH-dependent and efficient under environmentally relevant circumstances (Chen et al. 2019; Mangla et al. 2022). Nevertheless, the obtained BC is hard to recover and be applied in engineered systems. Recently, the modification of iron nanoparticles on carbonaceous materials has attracted much attention in view of practical application (Chen et al. 2023; Huang et al. 2022). Compared to original BC, the iron-BC composites not only improve the adsorption capacity of BC but also enhance the recovery efficiency of BC via magnetism (Son et al. 2018; Wan et al. 2020). Understanding the interaction mechanism between SAs and iron-BC composites is critical for exploring engineered materials to alleviate antibiotic pollution. Due to electron donor–acceptor interactions and template effects, the surface oxygenated groups and graphitized carbonaceous phase of BC change after the combination of iron nanoparticles (Yang et al. 2016; Zhu et al. 2017). Thus, the modification of iron-nanoparticles enhances the affinity of SAs towards magnetite-functionalized biochar (MBC) (Bai et al. 2021b). In addition, the  roles of pH-dependent sorption by actual soils vary among the molecular types of SAs (Kurwadkar et al. 2007). Chen et al. also found that the types of H-bond formation between different SAs and biochar vary with pKa, affected by the substituents of SAs (Chen et al. 2019). However, studies focusing on the sorption of ionizable organic compounds such as SAs on iron nanoparticle-modified BC are rather limited. Moreover, the sorption mechanisms of iron nanoparticle-modified BC are not understood considering the inherent properties of different SAs.

Difference in sorption of various SAs results from their varying molecular properties. Ahmed et al. studied the difference in sorption capacity of sulfamethazine (SMT) and sulfamethoxazole (SMX) on H3PO4 modified bamboo biochar (Ahmed et al. 2017b). The difference in sorption can be explained by the mechanisms of π–π interaction for positive species, and π–π electron donating-accepting interaction (EDA) and negative charge-assisted H-bond ((–)CAHB) depending on proton exchange with water molecules (Ahmed et al. 2017b). Sun et al. found that four parent SAs exhibited different sorption behaviors towards BC and their hydrophobic properties dominated the van der Waals forces responsible for the sorption capacity and rates (Sun et al. 2018). However, few studies focus on the structures of iron nanoparticle-modified BC dominating the interaction mechanism and difference in sorption behaviors of SAs with varying molecular properties. Such knowledge gap hinders the development of engineered carbonaceous materials and understanding of the fate of SAs during practical remediation application. Therefore, it is necessary to investigate the structure–activity relationship of sorption of different SAs on iron nanoparticle-modified BC.

Since iron nanoparticle-modified BC is structurally different from BC, it was hypothesized that the modification of iron nanoparticles can enhance the sorption of SAs and the specific sorption mechanism affecting the binding energy and affinity of SAs. In this study, magnetite-modified BC (MBC) was prepared using one-step method and the sorption of four common SAs including SMT, SMX, sulfamerazine (SMR), and sulfadiazine (SDZ) on MBC was investigated. The selected four SAs with different heterocyclic substituents are frequently detected in aquatic environments (Duan et al. 2022; Zeng et al. 2022). In addition, the difference in the substituents of selected SAs can be compared, making them suitable for exploring the varied sorption affinities and mechanisms. The aims of this study  were to: (1) demonstrate the relationship between regulated carbonaceous domains in MBC and sorption affinities of SAs using various characterization methods by comparing original BC  with magnetite (Fe3O4); (2) compare the sorption isotherms and kinetics of different SAs to identify the impact of their substituents on the adsorption efficiency; (3) explore the contributions of specific interactions via modeling pH-dependent curves of sorption; and (4) quantify the binding energies of the four SAs via density functional theory (DFT) calculation. The theoretical and experimental investigation of the four typical SAs can uncover how the coating minerals alter the sorption mechanisms of pyrogenic chars towards SAs. The information provided in this study would be beneficial for further developing efficient engineered materials. Using both experiments and calculations, the reasons for the superior sorption performance of MBC and mechanisms promoting the sorption of SAs can be understood.

2 Materials and methods

2.1 Property of SAs

Sulfamethazine (SMT), sulfamerazine (SMR), sulfadiazine (SDZ), and sulfamethoxazole (SMX) were purchased from Tokyo Chemical Industry (Shanghai) Development Co., Ltd. The chemical structure, molecular weight (MW), aqueous solubility (Sw), octanol–water partition coefficient (Kow) and dissociation constants (pKa) of these four SAs can be obtained from previous studies (Chen et al. 2019; Zhao et al. 2016) and are listed in Additional file 1: Table S1.

2.2 Preparation of BC samples

Rice husk (RH), recycled from rice field in Harbin (Heilongjiang Province, China), was used as the precursor of BC due to its potential environment risks without appropriate treatments, high pyrolysis yield and rich surface functional groups (Bai et al. 2020; Dai et al. 2023). Prior to being used, the RH sample was oven-dried overnight at 80 °C and then crushed by a high-speed rotary cutting mill. After sieving by a 0.096 mm sieve, 3 g of RH powder was added to FeCl3 solution containing 50 mL ethanol (99%). The appropriate impregnation mass ratio of Fe/RH was 0.5:1. The mixture slurry was stirred at 50 °C until it dried. The Fe-RH slurry residue was pyrolyzed at 700 °C with a heating rate of 15 °C min−1, then maintained for 90 min in a tube furnace under nitrogen atmosphere. The obtained composite was denoted as MBC and stored at 25 °C under dark condition.

2.3 Sorption and desorption experiments

Batch experiments were conducted at ambient temperature in glass vials with screw caps. Stock solutions of SAs were prepared by dissolving SAs in deionized (DI) water containing 0.01 mM NaCl and 200 mg L−1 NaN3 (bio-inhibitor). Experimental pH values were adjusted by adding 0.1 mM HNO3 or NaOH. For kinetic study, 0.5 g MBC was added into 300 mL diluted SA stock solutions with SA concentration at 0.1 mM. Samples were taken out at specific intervals during 500-h contact time. For isotherm study, the initial concentrations of SAs covered their 0.02 to 0.99-fold aqueous solubilities (Ci/Sw). The MBC dose was selected to ensure that the solute removal efficiency reached 20–80% (Chen et al. 2019). Accordingly, 2  mg L−1 MBC was used for the sorption of SMT, SMR, and SMX, while 1  mg L−1 MBC was used for sorption of SDZ. The sample from sorption solution (20 mL) was shaken at 160 rpm until equilibrium, and then filtered using 0.45 μm syringe filter. The concentration of SAs was determined by Waters e2695 Alliance HPLC equipped with a 2489 UV/Visible detector operating at a wavelength of 265 nm with a C18 column (250 × 4.6 mm, 5 μm) according to the method previously reported by Jiang et al. (Jiang et al. 2014). The mobile phase consisted of acetonitrile and purified water with 0.1% formic acid (40:60, v:v) at a flow rate of 1.0  mL min−1. Column temperature was 30 °C and injection volume was 10 μL. The retention time was 8.0 min.

After adsorption, the spent MBC was magnetically separated from the solution and dried at 40 °C overnight. Approximately 0.01 g of dried spent MBC was added to 20 mL deionized water in a glass vial. The vial was sealed and placed in a shaker to desorb the SAs on spent MBC for 24 h. Concentrations of SAs before adsorption, after adsorption and after desorption were measured. The desorption rate of SAs was determined through mass balance analysis.

2.4 Characterizations

Fourier transform infrared spectroscopy (FTIR; PerkinElmer SPECTRUM ONE, USA) was employed to obtain information on the surface functional groups of MBC before and after adsorption of SAs. Raman spectroscopy (HR800, France, JY) was used to analyze the molecular structure of MBC before and after adsorption of SAs. X-ray photoelectron spectroscopy (XPS, Thermo Scientific K-Alpha, USA) was used to determine the surface element composition and valence state of MBC before and after adsorption of SAs.

2.5 Sorption kinetic and isotherm models

Pseudo first-order kinetic model, pseudo second-order kinetic model, and two-compartment kinetic model were used for the experimental kinetic data. Langmuir model, Freundlich model, and Dubinin–Ashtakhov model were employed for the fitting of experimental isotherm data. In addition, the pH-dependent sorption of SAs on MBC was described using the two-compartment isotherm model as reported previously (Bai et al. 2021b). Details of the models are shown in Additional file 1: Texts S1 and S2.

2.6 Comparison of MBC, Fe3O4, demineralized MBC (DMBC) and BC

To explore the contributions of mineral and carbonaceous phases in MBC, the sorption performance of SAs on MBC, BC, demineralized MBC (DMBC), and Fe3O4 was compared. Fe3O4, purchased from Aladdin Co., Ltd, was the major mineral phase in MBC, as reported in our previous study (Bai et al. 2021b). DMBC was obtained by pickling with 1 M HCl (solid-to-liquid ratio: 1:50) several times. The acid-washed sample was then rinsed using DI water until the suspension reached neutral pH. The concentration of Fe3+ in HCl solution was analyzed by inductively coupled plasma-optical emission spectroscopy (ICP-OES, Optima 8300, Perkin-Elmer). When using Fe3O4 and DMBC as sorbents, their dosages can be calculated based on their mass ratios in MBC. This method can ensure the validity of the comparison among MBC, Fe3O4, and DMBC due to the exclusion of dosage effect. Details of Fe concentrations and dosages of different sorbents are shown in Additional file 1: Tables S2 and S3.

2.7 Computations

VASP (Kresse and Furthmuller 1996; Kresse and Furthmüller 1996) was employed for spin-polarized DFT calculations within the generalized gradient approximation (GGA) utilizing the Perdew-Burke-Ernzerhof (PBE) (Perdew et al. 1996) formulation. Projected augmented wave (PAW) potentials (Blöchl 1994; Kresse and Joubert 1999) were selected to describe the ionic cores and valence electrons were accounted for using a plane wave basis set with a kinetic energy cutoff of 520 eV. To handle partial occupancies of the Kohn–Sham orbitals, the Gaussian smearing method was applied with a width of 0.05 eV. The convergence criteria were set to 0.01 eV Å−1 and 10–6 eV for the residual force and energy during structure relaxation. Van der Waals interaction was considered by employing the DFT dispersion correction (DFT-D). The Brillouin zone was sampled by a 2 × 2 × 1 k-points grid for the structure optimizations. In all calculations, vdW interactions were accounted for at the D3 level (Grimme et al. 2010). The absorption energy (ΔEx) was calculated using the following formula:

$$\Delta {\text{E}}_{{\text{x}}} = {\text{E}}_{{{\text{surf}} - {\text{x}}}} {-}{\text{E}}_{{{\text{surf}}}} {-}{\text{E}}_{{\text{x}}}$$
(1)

where Esurf−x is the total energy of the surface covered with adsorbed molecule, Esurf is the total energy of the surface without adsorbed molecule, and Ex is the energy of free absorbed molecule.

3 Results and discussion

3.1 Sorption isotherms of  the four SAs on MBC

Sorption isotherms of the four SAs on MBC are shown in Fig. 1a. All the isotherm curves of SAs were nonlinear and can be well fitted by the Langmuir model (R2 > 0.988) (Additional file 1: Table S4 and Fig. S1). It was observed that the sorption isotherm curves of  all the four SAs were different and the sorption saturation capacity (Qm) of SAs followed the order of SMX > SDZ > SMR > SMT (Additional file 1: Table S4). According to the well fitted isotherms using DA model (R2 > 0.943) (Fig. 1b), the order of calculated Q0 was: SMX > SDZ > SMR > SMT, in agreement with that fitted by Langmuir model (Additional file 1: Table S4). It was also observed that the sorption affinities (Ed) of  the four SAs were different and followed the order of SMX ≈ SDZ > SMR > SMT (Additional file 1: Table S4), generally consistent with that of sorption capacity for different SAs. The results of sorption isotherm collectively demonstrated that the affinities of the four SAs, related to sorption energies (Yang and Xing 2010), determined their sorption capacities on MBC. Considering that SAs have identical sulfanilamide structure and different substituents, it can be concluded that the substituents of SAs exhibited significant impacts on the sorption capacity and affinity. In particular, the SAs substituted with five-membered ring (i.e., SMX) had higher adsorption capacity than those with six-membered ring (i.e., SDZ, SMR, and SMT). This result was consistent with the previous studies, which showed that SAs with five-membered ring substituents exhibited higher affinity towards BC (Ahmed et al. 2017b; Fan et al. 2021). One of the reasons for this phenomenon is that the SAs substituted with six-membered ring could not adsorb on the surface due to steric effect. In addition, for SAs with six-membered rings, more hydrophobic –CH3 substituent led to lower affinity and sorption capacity. Thus, it is reasonable to speculate that the difference in specific interactions instead of van der Waals forces (e.g., hydrophobic interaction) resulted in the different sorption affinities and capacities of SAs. This speculation can be verified by the observation that the affinities of SAs were negatively correlated to their Kow. It was also observed the desorption efficiencies of MBC for the four SAs ranged from 0.1% to 6.8% (Additional file 1: Fig. S2), which was similar with previous work (Febelyn et al. 2017). This result further indicated that some irreversible specific interactions dominated the binding between SAs and MBCs and the prepared materials did not pose a significant risk of secondary pollution.

Fig. 1
figure 1

a Sorption isotherms of four SAs on MBC and b isotherm data fitted using Dubinin-Ashtakhov model. c Comparison of adsorption content (Qe) for four SAs (0.1 mmol/L) on MBC, DMBC, Fe3O4 and BC

Full size image

To identify the impact of magnetite on the sorption of SAs, the sorption data for BC, Fe3O4, and DMBC were compared. Results showed that the adsorption capacities of the four SAs on Fe3O4 and BC were much lower than those on MBC (Fig. 1c). The Qm values of the four SAs on MBC were even 5 and 10 times higher than those on Fe3O4 and BC, respectively (Fig. 1c). The results demonstrated that the higher sorption of SAs on MBC was due to the synergetic effect of carbonaceous and mineral phases. However, the direct contribution of magnetite in MBC to the sorption of SAs was relatively limited. Aggregated magnetite nanoparticles in solution may exhibit low sorption ability (Fu et al. 2019; Karunanayake et al. 2019). Moreover, the Qm values of the four SAs on DMBC were  about 1.4-fold higher than those on MBC (Fig. 1c). It should be noted that DMBC excluded the direct contribution of magnetite, but retained the carbonaceous phases formed during co-pyrolysis with Fe3+. This result indicated that the synergistic effect, which was responsible for the enhanced sorption capacity of MBC, resulted from irreversible structural changes in carbonaceous phases. Such synergistic effect of MBC may result from the enhanced formation of oxygen-containing groups (e.g., –OH and –COOH) and graphitization of carbonaceous phases during co-pyrolysis with impregnated Fe3+ (Bai et al. 2021a, b; Zhu et al. 2017). These structural changes of carbonaceous phases in MBC can provide more active sites for the adsorption of SAs. The synergistic effect of minerals and carbonaceous phases can be further verified by comparing other common carbonaceous materials. As shown in Additional file 1: Table S5, MBC had higher adsorption capacity for SAs than other carbonaceous materials such as activated biochar, graphene oxide, and carbon nanotubes. This comparison also suggested that agricultural waste-derived MBC is a suitable material for the removal of SAs. It can be concluded that the direct interaction of SAs on the formed Fe3O4 minerals in MBC was limited and regulated carbonaceous matrix in MBC was mainly responsible for the enhanced sorption of SAs.

3.2 Sorption kinetics of four SAs on MBC

Sorption kinetics of the four SAs on MBC and BC are presented in Additional file 1: Fig. S3. Data of sorption kinetics can be well described using pseudo-second-order and two-compartment kinetic models (Additional file 1: Table S6 and Fig. S3). It was observed that the second-order sorption rates  of the four SAs on MBC were significantly higher than those on BC (Additional file 1: Table S6). This observation suggested that MBC was more promising than BC as remediation material when applied in engineered systems. Comparing the four SAs, the order of sorption rate on MBC was: SMX > SDZ > SMR > SMT, consistent with that of sorption capacity and affinity. This order was also negatively related to the Kow of SAs (Additional file 1: Table S1). It has been reported that the modification of magnetite on BC enhances the hydrophilic oxygen-containing groups and shields the hydrophobic organic components of original BC (Shi et al. 2023). Since the hydrophobic interaction could not explain the sorption of SAs, the specific interactions related to proton exchange with water molecules and the species of SAs may be more important. In that case, the interactions between SAs and dissociable structures of MBC may lead to faster sorption of MBC than BC.

According to the two-compartment kinetic model, the contribution of slow adsorption of SAs on BC was higher than that of fast adsorption (Additional file 1: Table S6). This was because the diffusion of SAs on the high-temperature BC was slow when the compact and orderly aromatic structures dominated the carbonaceous components (Chen et al. 2012). In contrast, the contribution of fast adsorption on MBC was several orders of magnitude higher than that of slow adsorption. The difference in sorption kinetic  properties of MBC compared with BC revealed important structural changes. When more oxygen-containing groups formed on MBC than BC, they can provide more sites for the formation of hydrogen bonds, rendering fast sorption (Bai et al. 2021b). In addition, compared to BC, the contribution of slow adsorption on MBC was higher. This phenomenon confirmed that the co-pyrolysis of magnetite promoted the condensation and graphitization of carbon structure in BC (Sun et al. 2013). Collectively, it was concluded that MBC exhibited faster sorption rates for SAs due to its higher portion of oxygen-containing groups as hydrophilic domain, while it possessed enhanced graphitization due to Fe-catalyzed effect.

3.3 Sorption mechanisms of MBC

In order to determine the interactions between SAs and MBC, the sorbents before and after adsorption of four SAs were analyzed using FTIR, Raman spectra, and XPS. The FTIR results (Fig. 2a) showed that the -OH stretching vibration peak of MBC at 3411 cm−1 was significantly weakened and the peak of Fe–O at 562 cm−1 shifted to 550 cm−1. In addition, the peak of aromatic C=C/C=O vibrations around 1600 cm−1 shifted, while the peak of C=O vibration at 2361 cm−1 disappeared (Fig. 2a) (Xu et al. 2022). The results indicated that the oxygen-containing groups including hydroxyl and carboxyl groups present in MBC participated in the sorption of SAs. The surface –OH and –COOH groups of MBC could interact with different N-containing groups of selected SAs to form three types of hydrogen bonding (Bao et al. 2014). Such H-bonding may include the interactions with aromatic primary amine (HB1), secondary amine (HB2), and heterocycle-N on different substituents (HB3) in SAs. The interactions between SAs and benzene ring of MBC also formed π–π conjugation (Bao et al. 2014). The sorption induced by π–π interactions resulted from the electron donor–acceptor (EDA) interactions between oppositely polarized systems in the parallel-planar pattern (Teixidó et al. 2011; Xiao and Pignatello 2015). Thus, π–π EDA interactions can occur between electron-deficient and electron-rich domains in SAs and MBC in this study, including the associations of aniline group of SAs and aromatic carbon of MBC, the heterocycle of SAs and the aromatic carbon of MBC, as well as the N-heterocycle of SAs and the carbonyl group of MBC.

Fig. 2
figure 2

a FTIR spectra and b Raman spectra of MBC before and after sorption of four SAs

Full size image

Raman spectra showed two characteristic peaks at 1350 and 1600 cm−1, representing defective (D band) and graphitized (G band) phases in MBC (Fig. 2b). The D band is associated with the disordered hybridization of carbon atoms such as oxygen-containing functional groups, while the G band is related to ordered sp2 carbon hybridization (Ahmed et al. 2017a). The ratio of intensity of D to G band (i.e., ID/IG ratios) is typically used to describe the degree of defects and graphitization of carbonaceous materials. It was observed that compared to the original MBC, the ID/IG ratio for MBC with adsorbed SAs decreased considerably (Fig. 2b). This result indicates that the oxygen-containing groups in MBC interacted with SA molecules (Ahmed et al. 2017b). Moreover, the comparison of ID/IG of MBC among the four SAs can provide the insights into the difference in sorption behaviors to some extent. For example, the MBC after sorption of SMX exhibited higher ID/IG value than SMR and SMT. This observed order was consistent with the sorption affinity, i.e., the higher affinity of SAs along with the more disordered structures of MBC. The relationship of ID/IG vs. affinity may not directly reflect thermal property, but it implies that SAs with higher sorption energies on MBC formed larger defects of carbonaceous phases. The phenomenon also suggested that sorption of SAs on MBC did not depend on the hydrophobic interaction, but was due to the interactions with defects of MBC such as oxygen-containing groups.

XPS spectrum (Fig. 3 and Additional file 1: Fig. S4) was used to observe the difference in the binding patterns among   the four SAs. The C 1 s spectrum of MBC can be deconvoluted into three different peaks located at 284.8, 286.2, and 288.9 eV (Additional file 1: Fig. S4), corresponding to C–C/C=C, C–O, and O–C=O bonds (Guo et al. 2021). It was observed that the atomic percentage of C–C/C=C in MBC increased after sorption of SAs (Additional file 1: Table S7), indicating π–π EDA interactions between benzene/heterocyclic rings of SAs and graphite phases of MBC. Moreover, the atomic percentage of C-O in MBC decreased (Additional file 1: Table S8), consistent with FTIR results. The O 1 s spectrum was deconvoluted into three peaks located at 530.5, 532.2, and 533.7 eV (Fig. 3), which can be attributed to Fe–O/N–O, C–O, and O–C=O (Liu et al. 2018). After adsorption of SAs, the percentage of peak at 530.5 eV increased significantly, which mainly resulted from the sorbed SAs with O-N bonds. Additionally, the percentage of O–C=O in MBC markedly increased along with the decrease in percentage of C–O (Additional file 1: Table S8). The above results suggested that the sorption of SAs changed the inherent properties and distributions of oxygen-containing groups and sp2/sp3 hybridized structures of MBC. The order of structural change in MBC, i.e., SMX < SDZ < SMR < SMT, was in contrast to that of sorption affinity. This contrary tendency can be reasonable because the SAs with higher affinity have higher thermodynamic energy of specific interactions with individual group as active sites (discussed as below). Combined with the spectral results, it can be concluded that the dissociable oxygen-containing groups and π-delocalized systems of MBC contributed to the sorption of SAs and the sorption affinities of SAs were significantly related to the structural changes of MBC.

Fig. 3
figure 3

XPS spectrum of O 1 s and its deconvoluted peaks of MBC before sorption a and after sorption of b SMX, c SDZ, d SMR and e SMT 

Full size image

3.4 pH-dependent sorption of SAs

Sorption of SAs on MBC under different pH conditions is shown in Fig. 4. The pH-dependent distribution coefficient (Kd) can be used to explore the possible interactions between ionized organic compounds and carbonaceous materials. It was observed that the Kd of the four SAs changed significantly as the pH increased from 1 to 12, which involved all pKa values of SAs. The two-compartment isotherm model was employed to fit the pH-dependent curves of the four SAs, but it failed to describe the data. Only considering the change in species of SAs (i.e., cationic, neutral, and anionic species) at different pH, the change in Kd just resulted from the hydrophobic effect of SAs (Bai et al. 2021b; Chen et al. 2019). As described in the previous studies (Bai et al. 2021b), the additional term ΔKd was added to the model and the data can be well fitted (Fig. 4). Such description of model considered the dissociation of carbonaceous materials. The well-fitted data indicated that the H-bonding of dissociable groups between SAs and MBC contributed greatly to the sorption. As observed in Fig. 4, the modeled data exhibited two peaks of Kd at different pH. First peak of Kd was at pH 1.94 for SDZ, 1.67 for SMR, 2.17 for SMT, and 1.95 for SMX. The pH values at  peaks of Kd approached the corresponding pKa1 values of SAs. Similarly, the pH values at second peak of Kd for SAs ranged from pH 5 to 7, approaching the corresponding pKa2 values. The ionized organic compounds can interact with BC by stronger forces than neural species of organic compound, which provided additional stabilization on the BC surface (Teixidó et al. 2011). According to the previous studies (Chen et al. 2019; Gilli et al. 2009), the dissociable functional groups of MBC (e.g., hydroxyl/ carboxyl groups) and SAs (e.g., amino and sulfamido groups) can be protonated and the hydration effect is outcompeted when the pH approaches pKa-2 and pKa + 2. In such case, H-bonding or π-bonding assisted H-bonds can form between SAs and MBC via proton exchange with water. From the results in Fig. 4, when pH approached pKa1/pKa2, proton exchange occurred between protonated and deprotonated forms, resulting in hydrogen bonding interaction. Considering the obvious π-delocalized systems of MBC and observed π–π EDA interaction, it was speculated that π-bonding assisted H-bonds dominated the sorption of SAs.

Fig. 4
figure 4

Impact of equilibrium pH on distribution coefficient (Kd) of a SMX, b SDZ, c SMR and d SMT. The gray dashed line represents the fitted data using only speciation model and the black dashed line represents the fitted data using two-compartment model

Full size image

Based on the above analysis, H-bonding between SAs and MBC dominated the sorption, and the peaks of Kd can be used to compare the sorption strength (Chen et al. 2019). As seen from Fig. 4, the Kd at peak 1 was significantly greater than that at peak 2 for all the four SAs. This result indicated that the amino group of SAs was more favorable to form H-bonding with MBC than sulfonamido group. It should also be noted that the Kd values at peaks, especially at peak 2, were different among the four SAs due to their inherent differences in dissociation properties. It was observed that the order of Kd at peak 2 was: SMX > SDZ > SMR > SMT. The difference in strength of H-bonding with MBC was related to the pKa. When pKa of SAs was lower, the protonation between SAs and MBC was closer and H-bonding was more favorable to form. In addition, the order of strength of H-bonding was consistent with the order of sorption capacities and affinities in Fig. 1. This observation reveals that the strength of H-bonding formed between SAs and MBC determined the sorption affinities related to changes in surface thermodynamic energy.

3.5 Calculation data of H-bonding

It was identified that H-bonding formed between the dissociated groups and determined the sorption of SAs. According to the structural changes from FTIR and XPS spectra, –OH and –COOH groups of MBC were mainly involved in the H-bonding formation. Thus, to quantify the contributions of H-bonding to sorption, the thermodynamic energy of hydrogen bonds with various functional groups (–OH, –COOH, Fe3O4) was calculated using DFT (Fig. 5 and Additional file 1: Table S7). The optimal sorption configurations of SAs with MBC modeled structures are shown in Fig. 5a. Due to the difference in the hydrophobic property, the calculated sorption energy was normalized using Kow to exclude the influence of hydrophobic effect. Normalized sorption energy can be used to qualitatively compare the H-bonding strengths among the four SAs or various functional groups. As seen from the results in Fig. 5b, normalized total H-bonding energies differed significantly among the four SAs and the order of normalized energy was: SMX (2.61 eV) > SDZ (1.67 eV) > SMR (0.63 eV) > SMT (0.14 eV). This order of normalized energy was consistent with that of sorption affinity and capacity. The calculation results further indicated that the difference in H-bonding energy resulted in the different sorption affinities and capacities of SAs on the MBC. In addition, it was observed that the normalized H-bonding energies of –OH and –COOH accounted for more than 50% of the total energy of  the four SAs. This observation was consistent with the above findings that the direct contribution of iron nanoparticles in MBC was minor compared to the oxygen-containing groups. Based on the calculation results, it can be concluded that the H-bonding energies between SAs and MBC determined the sorption affinities and capacities and the oxygen-containing groups of MBCs played the main role of regulating the H-bonding interactions with SAs.

Fig. 5
figure 5

a Optimal adsorption configuration, b total normalized sorption energies of four SAs, and c normalized sorption energies between different functional groups of MBC and four SAs. Blue balls in a represent nitrogen (N) atoms, yellow balls stand for sulfur (S) atoms, red for oxygen (O) atoms, gray for carbon (C) atoms, white for hydrogen (H) atoms, and orange for iron (Fe) atoms

Full size image

4 Conclusion

This study found that functionalized MBC enhanced the adsorption of SAs compared to BC. The synergistic effect during co-pyrolysis with Fe3+ improved the sorption performance of MBC towards SAs compared to original BC. Regulated carbonaceous matrix in MBC rather than formed Fe3O4 mineral in MBC was mainly responsible for the enhanced sorption of SAs. The selected four SAs (i.e., SMT, SMR, SMX, SDZ), with varied hydrophobicity (Kow) and substituents, exhibited different sorption capacities and affinities towards MBC. The sorption isotherm followed Langmuir and DA models, while the order of sorption capacities and affinities of the four SAs followed: SMX > SDZ > SMR > SMT. The SAs substituted with five-membered ring possessed higher adsorption capacity than those with six-membered ring. For SAs with six-membered rings, more hydrophobic –CH3 substituents led to lower affinity and sorption capacity. In addition, the sorption rate of SAs was negatively correlated to their Kow. Spectral methods of FTIR, Raman spectra, and XPS showed that the dissociable oxygen-containing groups and π-delocalized systems of MBC contributed to the sorption of SAs and the sorption affinities of SAs were significantly related to the structural changes of MBC. Combined with the results of modeling pH-dependent sorption, it was found that the H-bonding or π-bond assisted H-bonding determined the sorption affinities and capacities of SAs. In particular, the SAs with lower pKa were thermodynamically favorable to form H-bonding with MBC via proton exchange with water molecules. Quantum calculation results indicated that the H-bonding energies between SAs and MBC determined the sorption affinities and capacities, and the oxygen-containing groups of MBCs played an important role in regulating the H-bonding interactions with SAs.