Adsorption kinetics of 20 glucocorticoids at environmentally relevant concentrations in wastewater by powdered activated carbons and development of surrogate models

Abstract

Glucocorticoids (GCs) are widely used in the treatment of the coronavirus disease of 2019 (COVID-19), and the toxicity of GCs to aquatic organisms has aroused widespread concern. Powdered activated carbon (PAC) has proven effective in removing various trace organic pollutants. In this study, the adsorption behaviors of 20 typical GCs onto PACs were investigated at environmentally relevant concentrations (ng/L) in real wastewater, using four commercially available PACs (HDB, WPH, 20BF, PWA). The results showed that PAC adsorption was feasible for GC removal at ng/L concentrations. After adsorption for 60 min, the GC removal efficiencies obtained by HDB, WPH, 20BF, and PWA were 90–98 %, 89–97 %, 84–96 %, and 71–90 %, respectively. The adsorption processes of 20 GCs on PACs were well fitted by the pseudo-second-order kinetics model (with R² >0.98). Among the four PACs, HDB achieved the highest rates because of the electrostatic attraction between HDB (positively charged) and the complex of GCs and natural organic matter (GC-NOM, negatively charged). Among the 20 GCs, compounds with substitutions of halogen atoms or five-membered rings at C-17 achieved higher adsorption rates because of the enhanced formation of hydrogen bonds and a resulting increase in electron density. In addition, surrogate models with total fluorescence (TF) and ultraviolet absorbance at 254 nm (UV₂₅₄) were developed to monitor the attenuation trend of GCs during adsorption processes. Compared with the UV₂₅₄ model, the TF model showed better sensitivity to GC monitoring, which could greatly simplify the water quality monitoring process and facilitate online monitoring of GCs in water.

Graphical abstract

1. Introduction

Glucocorticoids (GCs), including endogenous GCs and synthetic GCs, are a class of hormones with anti-inflammatory and immunomodulatory effects involved in a series of physiological reactions in organisms, such as immunosuppression and the regulation of metabolism [1]. During the coronavirus disease outbreak of 2019 (COVID-19), GCs have been widely used in treating acute pneumonia in severe or critically ill patients, reducing patient mortality and the risk due to mechanical ventilation [2]. However, with the enormous consumption of these drugs, the GC concentrations in municipal wastewater and hospital wastewater increased considerably [3]. As the traditional activated sludge biodegradation processes used in wastewater treatment plants (WWTPs) are inefficient for GC removal [4], GC concentrations in WWTP effluent range from a few nanograms to hundreds of nanograms per liter (ng/L) [5], [6]. Due to WWTP effluent discharges, various exogenous GCs are ubiquitous in environmental waters worldwide, with concentrations from <0.01 to hundreds of ng/L [7]. GCs at μg/L concentrations could affect gene transcription of the hormone receptors and enzymes, interfering with gene expression of the hormone system and immune system, producing endocrine disrupting effects and other abnormal motor activities [8], [9]. Zhong et al. [10] showed that dexamethasone at 5–50 μg/L in water significantly reduced the transcriptional expression levels of some genes and caused liver tissue damage in mosquitofish. Hamilton et al. [11] reported that GCs at low concentrations (<6 ng/L) may pose potential risks to the cardiovascular and immune systems of vertebrates, including fish. As the potential adverse effects and threats of environmental trace GCs on aquatic organisms and human beings have increased, removing GCs in wastewater is critical to control GC pollution.

Biological processes cannot effectively remove GCs in water, and additional physical and chemical methods are necessary to improve the removal capacity. As a commonly applied process in water and wastewater treatments, the adsorption process has been reported to be effective in trace organic contaminant removal [12]. Compared with chemical treatment processes, adsorption processes do not form harmful transformation products or introduce secondary pollution [13], [14]. Common adsorbents in water treatment exist in powdered, granular and membrane forms [15]. Among these forms, powdered adsorbent materials are the most widely used because they have a high specific surface area, better adsorption performance and low material cost. Most powder adsorbents are carbonaceous [15], namely powder activated carbon (PAC), which shows excellent performances in removing micropollutants [16]. Egirani et al. [17] reported that PAC achieved higher pollutant removal rates and faster adsorption kinetics than GAC due to its smaller particle size. Woermann et al. [18] suggested that PACs adsorbed with micropollutants had negligible adverse effects on microbial communities or animals in sediments. High doses of PAC also had no adverse effects on settled water quality [19]. Therefore, PAC adsorption may benefit GC removal from water, favoring GC pollution control in water treatment plants. However, the adsorption behaviors of various GCs by different kinds of PACs remain unclear. It is crucial to understand the adsorption kinetics of GCs at environmentally relevant concentrations by various PACs to determine the optimal PAC type and adsorption time. The adsorption mechanisms of GCs by PACs are also need to be clarified to guide GC removal practices.

Due to the presence of GCs in water at μg/L or ng/L concentrations, the commonly applied analytical method using tandem mass spectrometry usually requires complex pretreatments which are time-consuming and laborious [20]. To rapidly predict GC attenuation by PAC adsorption, surrogate parameters (spectral and other organic parameters) should be explored to achieve immediate feedback on the adsorption process. Ultraviolet absorbance at 254 nm (UV₂₅₄) and total fluorescence (TF) have been widely used as surrogates to predict trace organic compound removal in water [21]. Aung et al. [22] reported the feasibility of ultraviolet absorbance at 254 nm (UV₂₅₄) and total fluorescence (TF) could be used as surrogates to predict the removal of per-/poly-fluoroalkyl substances (PFAS) by carbon adsorption. Sgroi et al. [23] also confirmed the correlations between UV₂₅₄ or TF breakthrough and adsorption removal efficiencies of emerging organic pollutants in various water qualities. Therefore, the spectral parameters (UV₂₅₄ and TF) may be used as surrogates to predict GC removal in water during PAC adsorption. As the unique structures and properties of GCs may influence the regression parameters of the predictive models [24], specific surrogate studies for determining the spectral parameters are necessary for predicting GC removal characteristics. Further theoretical understanding of the correlations between GC removal and these specific spectroscopic surrogates can provide references for the comprehensive application of spectral parameters.

Consequently, this study investigated the adsorption behaviors of 20 GCs on PACs at environmentally relevant concentrations (ng/L) in real wastewater. The adsorption mechanisms were preliminarily elucidated based on the obtained kinetic equations and structural characteristics of target GCs. Furthermore, the correlation between bulk organic parameters (TF and UV₂₅₄) and GC removal by PAC adsorption was demonstrated. The results support implications for an effective method to remove GCs in actual WWTPs and a perfect surrogate model system based on TF and UV₂₅₄ to monitor trace organic compound removal in real time.

2. Materials and methods

2.1. Materials

Analytical standards of 20 target GCs, i.e., corticosterone (CTC, ≥98.5 %), cortisone (COR, ≥98 %), hydrocortisone (HCT, ≥98 %), 6α-methylprednisolone (MPL, ≥98 %), prednisone (PNS, ≥98 %), fluorometholone (FML, ≥98 %), amcinonide (AMC, ≥98 %), budesonide (BUD, ≥99 %), fluocinonide (FLC, ≥98 %), fluocinolone acetonide (FCA, ≥97.5 %), flunisolide (FNS, ≥97 %), deflazacort (DFZ, ≥98 %), triamcinolone acetonide (TCA, ≥99 %), beclomethasone (BCM, ≥99 %), betamethasone (BET, ≥98 %), flumethasone (FMS, ≥98 %), dexamethasone (DEX, ≥98 %), clobetasol propionate (CBP, ≥98 %), clobetasone butyrate (CBB, ≥98 %), and fluticasone propionate (FTP, ≥98 %) were obtained as powders from Sigma-Aldrich (St. Louis, MO, USA). Their chemical properties are shown in Table 1 . The basic structure of GCs can be seen in Fig. S1. Aldosterone-d₇ (ALD-d₇, 95 %), betamethasone-d₅ (BET-d₅, 97 % D), budesonide-d₈ (BUD-d₈, >98 % D), cortisone-d₈ (COR-d₈, 97.9 % D), fluticasone propionate-d₅ (FTP-d₅, 98.1 % D), triamcinolone-¹³C₃ acetonide (TCA-¹³C₃, >98 % ¹³C) were purchased as powders from Toronto Research Chemicals Inc. (Ontario, Canada). Dexamethasone-d₄ (DEX-d₄, 96 % D), hydrocortisone-d₂ (HCT-d₂, 99 % D), 6αmethylprednisolone-d₂ (MPL-d₂, 95.3 % D), and prednisone-d₈ (PNS-d₈, 98.5 % D) were purchased as powders from C/D/N Isotopes Inc. (Pointe-Claire, Canada). Chromatographic grade acetonitrile (ACN), ethyl acetate (EA), hexane (Hex), and acetic acid (AA) were purchased from Fisher Scientific Co. (Fair Lawn, NJ, USA). Ultrapure water with a resistivity of 18.2 MΩ·cm was prepared using a Milli-Q system (Millipore, Billerica, MA). Stock solutions of each analyte and isotope surrogate standard were prepared at 1.0 mg/mL in methanol.

Table 1.

Chemical properties of the 20 target GCs.

Compounds	Abbr	CAS	Structure	logKowc
Corticosterone	CTC	50-22-6		1.94
Cortisone	COR	53-06-5		1.47
Hydrocortisone (Cortisol)	HCT (CRL)	50-23-7		1.61
6α-Methylprednisolone	MPL	83-43-2		1.95b
Prednisone	PNS	53-03-2		1.46
Fluorometholone	FML	426-13-1		2.00
Amcinonide	AMC	51022-69-6		3.78a
Budesonide	BUD	51333-22-3		2.55b
Fluocinonide	FLC	356-12-7		3.19
Fluocinolone acetonide	FCA	67-73-2		2.48
Flunisolide	FNS	3385-03-3		2.51b
Deflazacort	DFZ	14484-47-0		1.97b
Triamcinolone acetonide	TCA	76-25-5		2.53
Beclomethasone	BCM	4419-39-0		2.19b
Betamethasone	BET	378-44-9		1.94
Flumethasone	FMS	2135-17-3		3.86
Dexamethasone	DEX	50-02-2		1.83
Clobetasol 17-propionate	CBP	25122-46-7		3.50
Clobetasone 17-butyrate	CBB	25122-57-0		3.76
Fluticasone propionate	FTP	80474-14-2		3.96b

^aCalculated based on EPI Suite from U.S. EPA (KowWIN v1.68).

^bCalculated based on XLOGP3.

^cHydrophobicity index is expressed as the logarithm of the partition coefficient in 1-octanol/water.

Four commonly used commercial PACs (HDB, 20BF, WPH, and PWA) were used in this study. Among the four PACs, HDB and 20BF were purchased from Cabot Corporation (Boston, MA, USA), while WPH and PWA were obtained from Calgon Carbon Corporation (Pittsburgh, PA, USA). The four PACs differ in the Brunauer-Emmett-Teller (BET) surface area, iodine number, and surface charge. The BET surface areas of HDB, 20BF, WPH, and PWA are 546, 864, 1027, and 1142 m²/g, respectively. The iodine number of HDB, 20BF, WPH, and PWA were 540, 800, 800, and 900 mg/g, respectively (manufacturer provided data). To understand the surface charge characteristics of PACs, their slurry pH values at which the net charge of the carbon surface is neutral were measured according to the methods described by Hou et al. [25]. The slurry pH values of HDB, 20BF, WPH, and PWA were 10.9, 8.5, 8.58, and 7.64, respectively, indicating that the carbon surfaces of the four PACs were all charged positively at pH 7.5 (the pH of the wastewater), while HDB was the most obviously positively charged one.

2.2. Sample collection and adsorption test

The secondary effluents (DOC = 6.5 ± 0.3 mg/L) were collected from the Agua Nueva Water Reclamation Facility (ANWRF) loaded in Arizona, USA, and filtered using 0.7-μm glass fiber filters (Whatman GF/F, GE Healthcare Bio-Sciences). The water samples were adjusted to pH 7.5 using a phosphorous buffer (10 mM), and 100 ng/L of each target GC was spiked. Consequently, the initial GC concentrations in tested wastewater samples ranged from 100 to 500 ng/L (Fig. S2). All samples were collected in 4 L amber glass bottles and stored at 4 °C in the dark.

The adsorption experiments were conducted in a jar test apparatus (PB-900 Programmable Jar Tester, Phipps & Bird, Fig. S3). The apparatus contains six glass vessels of 11.5 cm × 11.5 cm × 21 cm dimensions. 2 L of wastewater effluent with GCs was placed in a vessel and mixed at 300 rpm using a jar tester (Phipps & Bird PB900). Then, PAC at 50 mg/L was added into the vessel, and the adsorption process was started. Samples were collected at 0, 5, 10, 20, 30, 40, 50, 60, 90, 120, 150, 180, 240, 300, and 360 min and filtered using a 0.45-μm syringe filter (PVDF Millex-HV, Billerica, MA, USA). The filtered water samples were placed in amber glass scintillation vials and refrigerated at 4 °C until chemical analyses.

2.3. Analysis method of GCs

A 990 μL sample was spiked with 10 μL surrogate standard mixture (in 50:1 water/methanol solution) containing 20 μg/L of each isotopically labeled GCs (which made the final concentration of isotope surrogates as 200 ng/L) and filtered through 0.2-μm filters (Agilent Captiva PES filters; p/n 5190–5096). The samples were analyzed by an Agilent 6490 triple quadrupole mass spectrometer coupled to an Agilent Infinity 1290 ultra-high-performance liquid chromatography system (Agilent Technologies, Palo Alto, CA). Specific operating conditions of analysis refer to the published literature [20]. The instrument detection limits (IDLs) and method detection limits (MDLs) for the target GCs and alternative standards are listed in Table S1.

2.4. Spectral parameter analysis

UV absorbance and fluorescence spectra were analyzed using a Horiba Aqualog fluorometer (Horiba Scientific, Edison, NJ). UV₂₅₄ was measured by scanning from 200 to 800 nm wavelength at 4 nm increments. Fluorescence intensity was monitored using the excitation wavelength from 200 nm to 450 nm with an interval of 5 nm and emission wavelength from 220 nm to 580 nm with an interval of 1 nm. The deionized water was used to normalize the fluorescence intensity of samples. Internal filtering effects of fluorescence spectra were corrected by scanning UV–vis spectra [26]. The fluorescence excitation-emission matrix (F-EEMS) was generated, and TF was calculated by integrating the fluorescence spectra in the range of excitation and emission wavelengths.

3. Results and discussion

3.1. Removal of GCs by PACs

As shown in Fig. 1 , PAC adsorption was effective for GC removal. The adsorption removal efficiencies of all 20 GCs on the four PACs were all higher than 95 % at 360 min (Fig. 1a). With the increase in the adsorption time, the GC removal efficiencies increased for the four PACs (Fig. 1b–e). The fastest increase in GC removal efficiency occurred from 15 min to 60 min (Fig. 1b–e). Upon further increasing the adsorption time from 60 min to 120 min or 360 min, the increase in GC removal efficiency became small (Fig. 1b–e). Adsorption time is a primary operating parameter for the adsorption process. A longer adsorption time is not economical because of the increased cost of a large contactor. In an actual water treatment process, the adsorption time is usually set at 30–60 min [27]. In this study, after adsorption for 60 min, the GC removal efficiencies obtained by HDB, WPH, 20BF, and PWA were 90–98 %, 89–97 %, 84–96 %, and 71–90 %, respectively (Fig. 1b–e), indicating that PAC adsorption was a feasible method for the removal of GCs from wastewater.

Fig. 1.

Removal efficiencies of 20 target GCs by the four PACs at 360 min (a) and 15 min (f), and removal efficiencies of 20 target GCs at various times by 20BF (b), HDB (c), PWA (d), and WPH (e). The GCs on the circular axis were arranged clockwise in ascending order of their logK_ow values.

Different GC removal efficiencies were observed among the four types of PACs during adsorption (Fig. 1f). The GC removal by HDB, WPH, 20BF, and PWA at 15 min was 76–94 %, 68–89 %, 63–87 %, and 45–70 %, respectively (Fig. 1f), indicating that HDB was the most efficient for removing GCs, followed by WPH, 20BF, and PWA. It is worth highlighting that HDB had the smallest BET surface area and iodine number among the four PACs but achieved the highest GC removal at 15 min (Fig. 1f), suggesting that the specific surface area might not be the decisive factor for the PAC adsorption capacities of GCs. This unique adsorption phenomenon may be attributed to other strong interactions between HDB and GCs, which needs further exploration.

Hydrophobicity is also an important factor affecting the adsorption capacity. Previous studies reported that the adsorption capacity of activated carbon was positively correlated with hydrophobicity [14]. The removal rates of 20 target GCs by the four PACs are plotted in radar charts in Fig. 1, arranged clockwise in ascending order of the hydrophobicity (expressed as logK _ow) of GCs. No significant correlation was observed between logK _ow and GC removal (Fig. S4, R ² < 0.5), indicating that the hydrophobicity of GCs has little influence on PAC adsorption capacity. Similar results were observed by Wu et al. [28] who mentioned that the hydrophobic values of bisphenols were slightly correlated with their adsorption efficiencies. Liang et al. [29] also reported that hydrophobic interactions had almost no effect on the adsorption removal of sulfamethoxazole.

Considering the insignificant effects from the specific surface areas of PACs and hydrophobicity characteristics of GCs on the adsorption processes, the good performance of PACs on GC adsorption removal might be attributed to the formation of strong hydrogen bonds and electron donor-acceptor (EDA) interactions [30], [31]. The ketone group (-C=O) at C-3 and the ketone (-C=O) or hydroxyl group (-OH) at C-11 could provide electrons to the cyclopentanophenanthrene ring of GCs, thus increasing the electron density and making GCs an effective π-electron donor. Meanwhile, the ketone groups (-C=O) contained in PACs could function as π-electron acceptors. Thus, π-π EDA interactions could be formed between PACs and GCs, enhancing GC attachment to PACs. He et al. [32] reported that π-π EDA interactions were the adsorption mechanism for 17α-Ethinylestradiol (EE2) and bisphenol A (BPA) adsorbed on anthracite. Furthermore, strong hydrogen bonds might be formed between the oxygen-containing functional groups (-OH) on the PAC surface and the functional groups (-OH, -CH) on GCs, strengthening the adsorption of GCs to PACs. Tizaoui et al. [33] also suggested that the key adsorption interaction was hydrogen bonding between EDCs and polyamide 6 (PA6) particle sorbents. Many studies have reported that EDA interactions and hydrogen bonds are the key adsorption mechanisms between carbon materials rich in carboxyl and carbonyl groups and steroidal compounds [34], [35].

3.2. Adsorption kinetics

To further explore the adsorption behaviors of PACs for the removal of GCs from aqueous solutions, three kinetic models [36] (pseudo-first order (PFO), pseudo-second order (PSO), and intraparticle diffusion) were used to explain the experimental data (Table S2). The adsorption amounts of the 20 GCs on the four PACs were well fitted by the three kinetic models. The calculated values of the intraparticle diffusion model are listed in Table S3. For the 20 GCs, the intraparticle diffusion rate constant (k _id) values for the four PACs were in the same order: k _1d > k _2d > k _3d, and all values of C _i were not zero. This indicated that the GC adsorption process by PACs may include three phases: rapid external mass transfer, intraparticle diffusion, and adsorption equilibrium on the internal surface of PACs [37], [38]. The correlation coefficients (R ²) of the PFO and PSO models are illustrated in Table S4. Compared with the PFO model (R ² > 0.91), the PSO model achieved better fitting performances (R ² > 0.98) for the adsorption processes of GCs (Table S4), indicating that the adsorption processes for GCs on PACs were mainly chemisorption [39].

By fitting the adsorption capacities of 20 GCs on the four PACs over time by the PSO model, the adsorption equilibrium amounts (q_e) and pseudo-second-order adsorption rate constants (k₂) of the 20 GCs for the four PACs were obtained (Figs. 2 , 3 , and Table S5). Because of differences in the initial concentrations of the 20 GCs in real wastewater (ranging from 100 to 500 ng/L as shown in Fig. S2), the q_e values of the 20 GCs were different (Fig. 2 and Table S5). According to the differences in q_e values on the four PACs, the target GCs could be divided into two groups (PAC-type dependent group and PAC-type independent group) (Fig. 2). The PAC-type dependent group (q_e-relative standard deviation (q_e-RSD) of the four PACs >7) included CTC, DFZ, BCM, CBP, CBB, AMC, FTP, and FLC (Fig. 2 and Table S5). The PAC-type independent group (q_e-RSD of the four PACs <7) included PNS, COR, HCT, DEX, BET, MPL, FML, FCA, FNS, TCA, BUD, and FMS (Fig. 2 and Table S5). For the GCs in the PAC-type dependent group, the selection of an appropriate type of PAC had a significant influence on their adsorption equilibrium amounts, while for the GCs in the PAC-type independent group, the four types of PAC all had similar GC adsorption equilibrium quantities (Fig. 2 and Table S5). Comparing the chemical structures of GCs in the two groups, the main differences were that GCs in the PAC-type dependent group mostly have complex substituent groups at the C-17 containing halogen atoms (such as CBP, CBB, and FTP) or nitrogen atom (such as DFZ) or oxygen atoms with a closed five-membered ring (such as AMC) (Table 1).

Fig. 2.

Adsorption kinetic curves of the 20 target GCs fitted by pseudo-second-order model with q (ng/mg) representing the adsorbate amount in the solid phase.

Fig. 3.

The pseudo-second-order adsorption rate constants (k₂) of the 20 target GCs for HDB, WPH, 20BF, and PWA.

As shown in Fig. 3 and Table S5, the GCs in the PAC-type dependent group (such as CBB, FTP, AMC, CBP, BCM, DFZ, CTC, and FLC) also achieved higher adsorption rate constants than those in the PAC-type independent group. Halogen atoms could act as electron acceptors due to the uneven electron density distribution and react with hydroxyl and carbonyl groups to form halogen bonds [40], [41]. Strong halogen bonds might have formed between the halogen atoms of CBB, CBP, and FTP (Table 1) and the hydroxyl groups on PACs so that the adsorption rates of CBB, CBP, and FTP were significantly higher than those of other GCs (Fig. 3). On the other hand, the hydrogen bonds could be formed by the bonding of a hydrogen atom in the -OCH₃ group as the hydrogen bond donor center with the sulfur atom of the adjacent molecule and the oxygen atom of the ketone group (-C=O) [42]. Therefore, the halogen atoms of CBB, CBP, and FTP (Table 1) and the oxygen atoms on ketone groups (-C=O) of PACs may form hydrogen bonds (O-H… X) with the presence of hydrogen atoms in GC molecules or in an aqueous solution, thus accelerating the adsorption of CBB, CBP, and FTP on PACs. For AMC, the extra five-membered ring at C-16 and C-17 of the AMC molecule enhanced its electron density, thus promoting electron interactions between AMC and GCs. Therefore, the properties of the functional groups of GC chemical structures significantly influenced their adsorption behaviors.

HDB achieved the largest adsorption rate constants (k₂) of the PSO models for GCs, followed by WPH, 20BF, and PWA (Fig. 3). The k₂ values for the four PACs (Fig. 3) were in the same order as the GC removal at 15 min (Fig. 1f), confirming that HDB showed the best performance in adsorbing GCs. It is worth noting that the surface of HDB was positively charged since the slurry pH of HDB (10.9) was much larger than the solution pH (7.5), while the other three PACs were close to neutral. At the solution pH of 7.5, the ionization degrees of the 20 GCs were relatively low, and the existing forms of GCs were mainly neutral because of their high pK_a values (above 10) [43]. Therefore, no electrostatic attraction between HDB and GCs could be expected in their original forms. However, as real wastewater also containes natural organic matter (NOM), GCs may combine with NOM in wastewater and change their existing forms. Some studies have reported that the micropollutants in water may combine with the NOM through EDA interactions and form a pollutant-NOM complex [44], [45]. As NOM is always negatively charged in wastewater because of its acidic groups (carboxyl groups, and phenolic groups) [46], the GC-NOM complex may also be negatively charged and generate electrostatic attraction with the HDB surface. This could accelerate the binding of GCs to the active sites on HDB. Sun et al. [47] also confirmed that the surface electrification of carbon enhanced its ability to capture anions in aqueous solution, resulting in faster adsorption diffusion.

3.3. Prediction of GC attenuation using UV₂₅₄ and TF

The adsorption amount data of UV₂₅₄ and TF were also fitted with the PSO model (Fig. 4 ). The adsorption trends of UV₂₅₄ and TF by the four PACs were similar to those of GCs (Figs. 2 and 4). HDB achieved the highest adsorption rate for TF values, followed by WPH, 20BF, and PWA (Table S6), which was the same order as the GC removal efficiency. Furthermore, compared with UV₂₅₄, TF showed better goodness of fit with high R ² values (TF: R ² = 0.964–0.975, UV₂₅₄: R ² = 0.951–0.967, Table S6), indicating that TF was more suitable as a surrogate for GCs during adsorption processes than UV₂₅₄.

Fig. 4.

Solid-phase amounts (q) of TF (A) and UV₂₅₄ (B) during adsorption by 20BF, HDB, PWA, and WPH.

Theoretical equations of the correlation between GC concentrations and bulk organic parameters (TF and UV₂₅₄) were developed. Considering the influence of NOM in real wastewater, Eq. (1) could be used to depict the sorption of GC_i by activated carbons in wastewater [48].

$$\frac{d{q}_i}{\mathit{dt}}=k_{a,i}^{\ast }{q}_e{c}_i\left(1-\theta \right)-k_{d,i}{q}_i$$ (1)

where k _{a, i} ^∗ (L · ng i ⁻¹ s ⁻¹) and k _{d, i}(s ⁻¹) are the adsorption and desorption rate constants of GC_i, respectively. In this study, k _{a, i} ^∗ is expressed by the second-order kinetic constant (k₂) of GC_i, c _i(ng/L) is the initial concentration of GC_i, and θ is the occupied fraction of active sites (0 ≤ θ ≤ 1). Since no desorption was observed in this study and the PAC dose used was enough for GC adsorption at environmentally relevant concentrations (100–500 ng/L), Eq. (1) can be simplified to Eq. (2) by ignoring the tiny desorption and adsorption sites occupied at dt:

$$\frac{d{q}_i}{\mathit{dt}}=k_{a,i}^{\ast }{q}_e{c}_i$$ (2)

By replacing the adsorption rate constants of GC_i (k _{a, i} ^∗ (L · ng i ⁻¹ s ⁻¹)) with the adsorption rate constants of bulk organic parameters (TF and UV₂₅₄) (k _{a, s} ^∗ (L · ng i ⁻¹ s ⁻¹)), Eq. (3) was obtained, using TF and UV₂₅₄ as surrogates for GCs:

$$\frac{d{q}_s}{\mathit{dt}}=k_{a,s}^{\ast }{q}_e{c}_s$$ (3)

where c _s(ng/L) is the concentration of bulk organic parameters (TF and UV₂₅₄).

By combining Eqs. (2), (3), Eq. (4) is obtained, which reflects the relationship between the bulk organic parameters (TF and UV₂₅₄) and GCs:

$$\mathit{ln}\left(\frac{c_i}{c_{i,0}}\right)\begin{array}{l}=\frac{k_{a,i}^{\ast }{q}_{e,i}}{k_{a,s}^{\ast }{q}_{e,s}}\mathit{ln}\left(\frac{c_s}{c_{s,0}}\right)\\ =\mathit{\alpha ln}\left(\frac{c_s}{c_{s,0}}\right)\end{array}$$ (4)

where c _{s, 0}(ng/L) and c _{i, 0}(ng/L) are the initial concentration of the bulk organic parameters (TF or UV₂₅₄) and GC_i, respectively, and α is k _{a, i} ^∗ q _{e, i}/k _{a, s} ^∗ q _{e, s}.

As shown in Eq. (4), the relationship between the bulk organic parameter (TF or UV₂₅₄) removal ($\frac{c_s}{c_{s,0}}$) and GC removal ($\frac{c_i}{c_{i,0}}$) mainly depend on the relative adsorption rate and adsorption equilibrium concentration of GC to surrogate (α), which is a constant at specific adsorption conditions.

To verify the theoretical relationship (Eq. (4)) between GCs and the bulk organic parameters (TF or UV₂₅₄), an analysis of the experimental data was conducted by exploring the data correlation between GC removal efficiencies (ln (C/C₀)) and bulk organic parameter (TF or UV₂₅₄) removal efficiencies (ln (TF/TF₀) or ln (UV₂₅₄/UV_254,0)). As shown in Fig. 5 , ln (C/C₀) showed an obvious linear relationship with ln (TF/TF₀) or ln (UV₂₅₄/UV_254,0) for PNS, BCM, and FMS independent of PAC type, indicating that the theoretical equation obtained in this study (Eq. (4)) was practicable. The natural log-log plots showing the relationship between GC removal and TF removal or UV₂₅₄ removal for the other GCs are shown in Figs. S5 and S6. By fitting the log-log curves (Figs. 5, S5, and S6), the linear correlations are showed in Table 2 . It was found that the linear intercepts were all above zero, which indicated that part of the NOM breakthrough was faster than the GCs, so UV₂₅₄ reduction and TF removal had occurred without the breakthrough of GCs [22]. Hence, to improve the predictive model, the Eq. (4) was empirically corrected as follows:

$$\mathit{ln}\left(\frac{c_i}{c_{i,0}}\right)=\mathit{\alpha ln}\left(\frac{c_s}{c_{s,0}}\right)+\beta$$ (5)

Fig. 5.

Natural log-log plots of the removal of PNS, BCM, FMS versus TF removal and UV₂₅₄ removal during adsorption by 20BF, HDB, PWA, and WPH.

Table 2.

The parameters of the prediction formula with TF and UV₂₅₄ for the 20 target GCs.

Compound	Surrogate	20BF			HDB			PWA			WPH
		α	β	R2	α	β	R2	α	β	R2	α	β	R2
AMC	TF	5.441	0.809	0.954	5.147	0.371	0.973	4.945	0.826	0.921	4.348	0.613	0.961
	UV254	10.233	1.167	0.877	8.478	0.374	0.970	10.404	1.098	0.881	7.627	0.765	0.949
BCM	TF	4.810	0.529	0.970	4.474	0.498	0.951	4.514	0.703	0.934	3.922	0.306	0.979
	UV254	9.134	0.881	0.910	7.361	0.498	0.946	9.498	0.951	0.893	6.896	0.451	0.971
BET	TF	4.871	1.024	0.905	4.758	0.862	0.896	4.590	0.944	0.885	4.504	0.985	0.923
	UV254	9.062	1.303	0.814	7.830	0.863	0.891	9.613	1.182	0.839	7.927	1.156	0.918
BUD	TF	4.506	1.017	0.870	4.202	0.573	0.919	4.062	0.844	0.876	4.122	0.953	0.902
	UV254	6.996	0.872	0.785	7.447	0.731	0.878	5.089	0.322	0.875	5.476	0.495	0.893
CBB	TF	4.228	0.268	0.976	4.598	0.354	0.938	3.265	0.354	0.952	3.580	0.383	0.967
	UV254	8.813	0.807	0.895	7.647	0.390	0.953	6.898	0.542	0.918	6.308	0.522	0.964
CBP	TF	5.135	0.955	0.764	5.336	0.572	0.968	4.599	0.789	0.926	4.138	0.573	0.963
	UV254	9.662	1.103	0.879	8.325	0.366	0.972	9.663	1.037	0.884	7.269	0.724	0.954
COR	TF	4.808	1.070	0.889	4.613	0.841	0.886	4.699	0.955	0.881	4.439	0.957	0.927
	UV254	8.918	1.334	0.795	7.593	0.841	0.882	9.848	1.201	0.836	7.806	1.121	0.920
CTC	TF	5.436	0.750	0.945	4.742	0.337	0.949	5.337	0.811	0.923	4.337	0.325	0.916
	UV254	10.243	1.115	0.872	7.834	0.352	0.952	11.124	1.070	0.866	7.697	0.522	0.926
DEX	TF	4.951	1.123	0.885	4.833	0.951	0.878	4.486	0.966	0.866	4.589	1.180	0.891
	UV254	9.196	1.400	0.794	7.960	0.954	0.875	9.392	1.197	0.820	8.087	1.359	0.888
DFZ	TF	4.034	0.456	0.942	3.638	0.470	0.923	3.644	0.593	0.911	3.067	0.375	0.911
	UV254	7.355	0.663	0.866	5.972	0.468	0.920	7.681	0.798	0.874	5.393	0.485	0.877
FCA	TF	4.356	0.993	0.886	4.476	0.789	0.902	3.867	0.751	0.876	3.997	0.950	0.896
	UV254	8.090	1.236	0.794	7.362	0.788	0.897	8.123	0.959	0.835	7.044	1.105	0.893
FLC	TF	5.337	1.111	0.914	5.502	0.910	0.921	4.601	0.879	0.890	4.445	0.942	0.915
	UV254	9.935	1.419	0.823	9.045	0.906	0.919	9.663	1.126	0.848	7.822	1.111	0.909
FML	TF	4.544	0.892	0.906	4.617	0.863	0.885	4.287	0.908	0.866	4.334	0.951	0.915
	UV254	8.449	1.150	0.814	7.602	0.865	0.881	8.994	1.135	0.823	7.608	1.105	0.905
FMS	TF	4.834	1.062	0.894	4.781	0.900	0.898	4.436	0.953	0.873	4.472	1.027	0.913
	UV254	8.992	1.338	0.804	7.870	0.901	0.893	9.275	1.178	0.825	7.881	1.202	0.910
FNS	TF	3.919	0.901	0.864	4.116	0.828	0.867	3.445	0.698	0.865	3.730	0.939	0.878
	UV254	7.267	1.115	0.772	6.775	0.829	0.863	7.223	0.882	0.824	6.592	1.094	0.880
FTP	TF	5.313	0.611	0.963	5.087	0.152	0.977	4.296	0.552	0.956	3.484	0.063	0.976
	UV254	10.143	1.023	0.912	8.368	0.150	0.971	9.041	0.788	0.914	6.108	0.183	0.962
HCT	TF	4.703	1.093	0.880	4.544	0.844	0.881	4.403	0.900	0.881	4.436	1.075	0.898
	UV254	8.735	1.356	0.789	7.478	0.845	0.876	9.240	1.134	0.838	7.833	1.256	0.898
MPL	TF	4.654	1.038	0.886	4.570	0.859	0.889	4.234	0.903	0.874	4.694	1.192	0.853
	UV254	8.658	1.304	0.798	7.525	0.861	0.886	8.860	1.120	0.826	8.338	1.409	0.864
PNS	TF	4.714	1.050	0.890	4.436	0.864	0.873	4.591	0.896	0.887	4.419	0.964	0.916
	UV254	8.747	1.310	0.797	7.302	0.864	0.869	9.614	1.133	0.840	7.789	1.137	0.914
TCA	TF	4.240	0.916	0.888	4.378	0.826	0.882	3.820	0.793	0.869	4.309	1.148	0.855
	UV254	7.884	1.157	0.798	7.199	0.824	0.877	8.003	0.992	0.824	7.631	1.335	0.860

By comparing the α and β values (Table 2) obtained by fitting the log-log curves (Figs. 5, S5, and S6), we found that the α and β values of the TF models were all smaller than those of the UV₂₅₄ models, indicating that the TF model was more sensitive to GC concentration changes than the UV₂₅₄ model [49], [50]. According to the obtained parameters of the prediction formula (Table 2) and the removal efficiencies of the bulk organic parameters (TF or UV₂₅₄) during adsorption by the four PACs, the predicted removal efficiencies of the 20 GCs at various contact times by the four PACs were calculated (Fig. 6 ). As shown in Fig. 6, the calculated values of GC removal by both TF models and UV₂₅₄ models (Table 2) were in good agreement with the measured values (TF: R ² = 0.980–0.999, UV₂₅₄: R ² = 0.971–0.999), indicating that the TF and UV₂₅₄ models (Table 2) could accurately predict the GC removal characteristics during PAC adsorption processes.

Fig. 6.

Prediction of GC removal using the developed TF-involved and UV₂₅₄-involved surrogate models during adsorption by 20BF, HDB, PWA, and WPH.

Therefore, TF or UV₂₅₄ could be used as a surrogate index to predict the concentration change of GCs during PAC adsorption. The developed models (Table 2) can quickly monitor and qualitatively analyze the real-time GC attenuation in the PAC adsorption process. The surrogate models (Table 2) can be used to build a practical monitoring system of TF and UV₂₅₄ for predicting GC removal.

4. Conclusions

In this paper, the adsorption capacities of four commonly used commercial PACs on 20 GCs in real secondary effluent of WWTPs were investigated. The results showed that PAC adsorption was feasible for GC removal at environmentally relevant concentrations (ng/L). After adsorption for 60 min, the GC removal efficiencies obtained by HDB, WPH, 20BF, and PWA were 90–98 %, 89–97 %, 84–96 %, and 71–90 %, respectively. The adsorption processes of the 20 GCs on the four PACs were well fitted by the pseudo-second order model with R ² values >0.98.

The good performance of PACs on GC adsorption removal can be attributed to the formation of hydrogen bonds and electron donor-acceptor interactions between GCs and PACs, which could be further enhanced by the substitution of halogen atoms or five-membered rings at C-17 and the resulting increase in electron density. The positively charged surface of HDB enhanced its adsorption potential by generating electrostatic attraction with the GC-NOM complex.

The correlation between the breakthrough of surrogate parameters (TF and UV₂₅₄) and GC removal has been demonstrated (TF: R ² = 0.964–0.975, UV₂₅₄: R ² = 0.951–0.967). The surrogate models developed in this manuscript provide accurate prediction of GC attenuation during PAC adsorption processes (TF: R ² = 0.980–0.999, UV₂₅₄: R ² = 0.971–0.999). This work provides a possible method for real-time online monitoring of GC attenuation, and reference information for the development of a relatively complete and accurate online monitoring system based on surrogate parameters (TF and UV₂₅₄).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.