Introduction

Perfluoroalkyl acids (PFAAs), also referred to as perfluorinated tensides (PFTs), are an industrially extensively used class of chemical compounds. Due to their hydro and oleophobic properties as well as their surface activity they find versatile application, for example in electroplating and semiconductor production, and as additives in medical and cosmetic products or aqueous film-forming foams (AFFFs) for firefighting (Hu et al. 2016).

They consist of a completely fluorinated carbon chain and a terminal functional group (e. g. carboxylate or sulfonate). Notably the high number of strong bonds between the carbon and fluorine atoms is responsible for the exceptional stability of PFAAs to chemical and thermal (Buck et al. 2011) as well as biological degradation processes, making them extremely persistent in the environment (Giesy and Kannan 2002; Liou et al. 2010; Scheringer et al. 2014; Ochoa-Herrera et al. 2016). Most widely produced representatives of this group of chemicals in the past have been perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) (USEPA 2016c).

PFAAs are released into the environment for several decades and since then globally detectable in diverse environmental compartments, for example soils, sediments, water bodies such as rivers and lakes, and even in seas and oceans (Senthilkumar et al. 2007; Ahrens et al. 2009b, 2009a; Naile et al. 2010; Wang et al. 2011; Llorca et al. 2012; Theobald et al. 2012; Pignotti et al. 2017; Joerss et al. 2019; Muir and Miaz 2021). Furthermore, these toxic and bio accumulating compounds are also provable worldwide in a variety of aquatic and terrestrial animal species, including humans (Giesy and Kannan 2002; Sinclair et al. 2006; Fromme et al. 2007; Zhao et al. 2012; Bossi et al. 2015; Riebe et al. 2016; Koch et al. 2019). They have various harmful effects on human health with correlations to kidney, bladder, testicular, and liver cancer (Alexander et al. 2003; Steenland and Woskie 2012; Barry et al. 2013; Vieira et al. 2013; Mastrantonio et al. 2018; Li et al. 2022; Goodrich et al. 2022; Messmer et al. 2022), thyroid disease (Melzer et al. 2010), female fecundity disorder (Fei et al. 2009; Di Nisio et al. 2020; Velez et al. 2015), decreased sperm quality (Vested et al. 2013; Louis et al. 2015; Sabovic et al. 2020), and a lower immune response after vaccination (Grandjean et al. 2012; Porter et al. 2022). Besides that, bioassays with mammalian species demonstrated their potential for an induction of pancreatic and also liver and testicular tumors in rats (Biegel et al. 2001), as well as an adverse immunomodulation in mice (DeWitt et al. 2008, 2012).

The proven high mobility of perfluorinated surfactants in the hydrosphere, especially that of short-chain PFAAs, lead to a prevalent contamination of drinking water supplies (Skutlarek et al. 2006; Thompson et al. 2011; Boiteux et al. 2012; Eschauzier et al. 2013b; Banzhaf et al. 2017; Kabore et al. 2018) which in view of an impending water scarcity pose a problem for the delivery of clean potable water. In addition to PFOA and PFOS, perfluorobutane sulfonic acid (PFBS) and perfluorohexane sulfonic acid (PFHxS) are among the most frequently detected PFAAs in drinking water of industrialized countries (Ateia et al. 2019; Boston et al. 2019) and it is well known that tap water is a not negligible source for the daily uptake of PFAAs through diet (Ericson et al. 2008; Hoffman et al. 2011; Noorlander et al. 2011; Eschauzier et al. 2013a; Zhang et al. 2019; Hu et al. 2019). Monitoring studies demonstrated that residents of areas with contaminated drinking water more likely show enhanced body burdens (Bruton and Blum 2017) indicated by increased serum or plasma levels (Hölzer et al. 2008; Brede et al. 2010; Pinney et al. 2014; Hurley et al. 2016; Gyllenhammar et al. 2019; Glynn et al. 2020).

In 2016, the US Environmental Protection Agency (USEPA) therefore first issued drinking water health advisories (HAs) of 70 ppt (70 ng L−1) for PFOS and PFOA in order to ensure a protection from health effects throughout a lifetime exposure (USEPA 2016b, a). Based on new studies, USEPA (2022a, b) recently published a preliminary lifetime HA for PFOS, which has been drastically reduced to 20 ppq, and for the first time a HA value of 2.0 ppb for the short-chain perfluorobutane sulfonic acid (PFBS) (USEPA 2022b, a). The European parliament and council (EUCO) passed a recasted directive on the quality of water intended for human consumption in 2020, that regulate the maximum drinking water level to in sum 100 ppt for 20 PFAAs of concern, including PFSA with a perfluoroalkyl moiety of four to 13 carbons (Directive 2020/2184, Annex III, Part B, 3) (EUCO 2022).

Today, public water suppliers have to utilize sophisticated techniques for the removal of very low concentrated organic pollutants like PFAAs from contaminated drinking water resources, among them membrane separation methods, e.g. pervaporation, reverse osmosis or nano filtration, as well as anion-exchange, and adsorption, usually with activated carbon (AC) (Izák et al. 2006; Appleman et al. 2014; Crone et al. 2019; Belkouteb et al. 2020). Sorption based technologies are generally efficient and economic processes for the elimination of a multitude of hazardous organic compounds of lowest concentration from the water phase (Nassi et al. 2014). For the generation of a clean and safe drinking water supply, drinking water treatment plants thus widely apply water purification processes with integrated AC separation stages for the removal of organic contaminants, pharmaceuticals, flame retardants or PFAAs (Takagi et al. 2011; Delgado et al. 2012; Sim et al. 2021). However, long-chain PFAAs and notably those with shorter carbon chains (C4–C6) at concentration levels < 0.5 µg L−1 are only partially or not removed by AC, respectively (Eschauzier et al. 2012; Flores et al. 2013; Rahman et al. 2014; Boone et al. 2019).

This study is part of a project that aims on the development of an adsorptive drinking water treatment method for the elimination of very low concentrated PFAAs on the basis of reuseable fluorinated silica adsorbents. For this purpose, the putative high fluorophilicity of the newly synthesized adsorbents due to specific and potentially selective non-covalent interactions between PFAAs and fluorinated ligands on the functionalized silica surface (Marchetti et al. 2015) should be exploited. The main objectives beside the synthesis of suitable adsorbents were the subsequent investigation of their adsorption efficiency and behavior with regard to PFBS, PFHxS, and PFOS. We present data on the synthesis of a set of five silica gel-based adsorbents, functionalized with different polyfluorinated ligands and degrees of functionalization, as well as their subsequent characterization in terms of maximum removal efficiencies and equilibrium loads (isotherms) with respect to the individual PFSAs. Evaluated higher adsorption capacities at relevant drinking water concentrations compared to that of granular activated carbon are also discussed.

Materials and methods

Chemicals

A macroporous spherical silica (pore size 1000 Å, particle size distribution 200–500 µm, specific surface 25 m2 g−1) was applied as base material for the synthesis of functionalized adsorbents and obtained from SiliCycle (Québec, Canada). For silica functionalization, a polyfluorinated N-(alkoxysilylalkyl)alkanamide and four polyfluorinated alkylalkoxysilanes with different alkoxy moieties and chain lengths were purchased from abcr (Karlsruhe, Germany) and internally encoded as HSU54 and HSU55-HSU58, respectively. Acetonitrile (LiChrosolv, hypergrade), ethanol (LiChrosolv, gradient grade), methanol (Lichrosolv, hypergrade), n-octane (for synthesis), 2-propanol (p. A.), xylene (p. A.) and ammonium acetate (p. A.) have been procured from Merck Millipore (Darmstadt, Germany), as well as acetone (Rotipuran, p. A) from Carl Roth (Karlsruhe, Germany).

Certified methanolic standard stock solutions (c = 50 μg mL−1) of potassium perfluoro-1- butanesulfonate, sodium perfluoro-1-hexanesulfonate, perfluoro-n-nonanoic acid, sodium perfluoro-1-octanesulfonate) and isotopically labeled sodium perfluoro-1-[2,3,4-13C3] butanesulfonate, sodium perfluoro-1-[1,2,3-13C3] hexanesulfonate, perfluoro-n-[1,2,3,4,5-13C5] nonanoic acid, sodium perfluoro-1-[1,2,3,4-13C4] octanesulfonate for analytical purpose as well as analytical grade (> 98%) perfluorobutane-1-sulfonic acid, perfluorohexane-1-sulfonic acid (potassium salt), perfluorooctane-1-sulfonic acid (potassium salt), perfluorononanoic acid (as reference substance) for the preparation of acetonitrile working solutions were purchased from Wellington Laboratories (Guelph, Canada), and Sigma-Aldrich (Munich, Germany), respectively. All stock and working solutions were stored in amber glass vials at − 18 °C, and 5 °C, respectively. Ultrapure water (conductivity 0.055 µS cm−1), produced by a water purification system (AQUAlab, Höhr-Grenzhausen, Germany), was employed for the production of reagents and throughout the adsorption experiments. Pure oxygen (99.998%) and nitrogen (99.999%) for the operation of the thermobalance were obtained from Air Liquide (Düsseldorf, Germany).

Synthesis of functionalized silica adsorbents

Initially, 10 g of silica per functionalization were dried for 24 h in 250 mL round-bottomed flasks and closed with a glass stopper after drying. Calculated amounts of each silane were then thoroughly dissolved in beakers in 60 mL organic solvent (xylene) and organic solvent mixtures, consisting of n-octane and xylene of different compositions, and subsequently added to a dried silica. For synthesis, the resulting reaction mixtures were attached to a water-cooled rotary evaporator that operated at normal pressure and a cooling water temperature of  T = 6 °C. Depending on the individual boiling temperatures of the silanes, the various reaction mixtures were heated to a maximum of T = 120 °C for 4 h under reflux and constant rotation (120 rpm) and the resulting reaction products are then purified sequentially with different solvents of increasing hydrophilicity in a five-stage wash cycle at 30 mL per stage, respectively. After transfer to a Buchner funnel and a final washing step with 10 mL acetone the drying of each product took place at T = 60 °C overnight in a drying cabinet. For sample archival, all synthesized adsorbents received an internal code based on the described ligand coding system (see section "Chemicals"), namely HSU00107954, HSU00107955, HSU00107956, HSU00107957, and HSU00107958, respectively. These are also used throughout the manuscript as the corresponding adsorbent designation. Beside the specific ligand the respective code additionally contains precise information on the type, manufacturer as well as particle, and pore size of the base material used.

Thermogravimetric analysis (TGA) of functionalized silica adsorbents

Prior to each thermogravimetric analysis on an STA 449 F3 thermobalance (Netzsch, Selb, Germany) in pure oxygen atmosphere, a second drying of the adsorbent at T = 60 °C under vacuum was performed for 24 h in order to evaporate any residual solvent. During TGA, aluminum oxide (Al2O3) pans, loaded with 100 mg of adsorbent were heated from room temperature to T = 800 °C at a rate of 5 °C min−1 and held at 800 °C for 1 h. In order to compensate the influences of possible water losses, e.g. humidity, on the results, each thermogravimetric measurement was corrected throughout the analysis of an adsorbent sample, that is, automatically adjusted with stored data from a previous analysis of the non-functionalized silica.

Individual analysis was executed in triplicate (n = 3) and subsequently evaluated with the Proteus 6.1 software tool (Netzsch, Selb, Germany). The resulting mass changes formed the basis for the calculation of surface coverages (µmol ligand molecules m−2), and functionalization degrees (FD, expressed as µmol ligand molecules per µmol−1 silanol (SiOH) groups × 100%) of the individual adsorbents.

Adsorption experiments

All adsorption experiments preceded a three-step purification procedure of the utilized 8 mL poly-propylene (PP) cartridges (Chromabond Flash DL, Macherey–Nagel, Germany) with equal volumes (4 mL) of acetone, ethanol, and ultrapure water, respectively. After drying, 100 mg of the macroporous adsorbent was weighed into a cleaned PP cartridge for each experiment and carefully conditioned with 2 × 1 mL of acetonitrile without the adsorbent running dry. Although the macroporosity of each adsorbent favors a rapid adjustment of the adsorption equilibrium, which is advantageous for further process development, its hence small surface area, i.e., low adsorption capacity, had to be compensated by the comparatively high adsorbent dosage. Subsequently, 4 mL ultrapure water was pipetted into the cartridge and spiked with a defined volume of a corresponding PFAA working solution in order to adjust a specific initial PFAA concentration level (20–1,000 ng L−1 for equilibrium loading experiments and 10,000 ng L−1 for adsorbent screening). Because of the reduction of the specified theoretical concentrations caused by self-adsorption of the PP cartridges, a corresponding number of cartridges without adsorbent were spiked in parallel for each PFAA concentration, allowing then the determination of the real initial concentrations (c0). Each concentration level was executed in triplicate (n = 3). After sample preparation, the cartridges were sealed with PP caps and mixed for 24 h at 45 rpm in an overhead mixer (Sunlab, Aschaffenburg, Germany).

Instrumental analysis of aqueous PFAA samples by LC–ESI–MS/MS

Analysis of PFBS, PFHxS, PFOS and PFNA in the aqueous samples was performed on a PFC-free 1290 Infinity II high performance liquid chromatography system (Agilent, Waldbronn, Germany) hyphenated to an Agilent 6495 triple quadrupole mass spectrometer. At first, sample aliquots of 100 µL were mixed in 2 mL PP vials with 250 µL PP micro inserts and PP caps (Agilent, Waldbronn, Germany) with 2 µL internal standard working solution (c13C-PFAA = 0.025 ng µL−1) on a vortex mixer (VWR International, Darmstadt, Germany).

Subsequently, sample volumes of 5 µL were injected into the chromatographic system and then analyzed on a Triart C18 ExRS HPLC column (3 µm, 8 nm, 50 × 3.0 mm) with guard column (10 × 3.0 mm) from YMC Europe (Dinslaken, Germany) at a constant flow rate of 0.5 mL min−1 and a temperature of T = 40 °C. Elution of the PFAAs occurred with an initial isocratic step at 10% mobile phase B (acetonitrile/1 mM ammonium acetate) for 0.5 min followed by a linear gradient from 10% B to 70% B in 3.5 min and an additional isocratic step at 70% B for another 4 min (mobile phase A: ultrapure water/1 mM ammonium acetate). According to Kaupmees and Rebane (2017) an Agilent PFC delay column (InfinityLab, 30 × 4.6 mm) was additionally installed inline between pump and auto sampler so that PFAA contaminants from the system eluted with a retardation of roughly 0.5 min in comparison to the corresponding analytes (Kaupmees and Rebane 2017).

Ionization of the PFAAs took place in the jet stream electrospray ionization (ESI) source of the 6495 mass spectrometer in the negative ion mode (capillary voltage 2.4 kV, drying gas temperature T = 185 °C, drying gas flow rate 15 L min−1, nebulizer gas pressure p = 40 psi (2.8 bar), sheath gas temperature T = 350 °C, sheath gas flow rate 11 L min−1). Multiple Reaction Monitoring (MRM) parameters for precursor ion ([M-H]−) transmission and fragmentation were optimized automatically for each analyte with the optimization tools of the Agilent Mass Hunter software (Version 10.0.142). For PFAA quantification, MRM transitions for the native and corresponding isotopically labeled compounds were utilized as follows: 299 → 80 (PFBS), 399 → 80 (PFHxS), 463 → 419 (PFNA), 499 → 80 (PFOS), and 302 → 80 (13C3-PFBS), 402 → 80 (13C3-PFHxS), 468 → 372 (13C5- PFNA), 503 → 80 (13C4-PFOS), respectively. Individual limits of quantification (LOQ) of the investigated PFAAs (LOQPFBS 19 pmol L−1 [5.8 ng L−1], LOQPFHxS 20 pmol L−1 [8.1 ng L−1], LOQPFNA 14 pmol L−1 [6.3 ng L−1], LOQPFOS 14 pmol L−1 [6.9 ng L−1]) were determined with the calibration curve method according to DIN 32645 (DIN 2008).

Results and discussion

3.1 Surface coverage and functionalization degree of the silica adsorbents

First silica functionalizations with each ligand at a reaction temperature of T = 100 °C in pure xylene, except the xylene-insoluble ligand HSU54, led to synthesis products with maximum surface coverages (SC) between 0.06 (HSU00107957) and 0.80 (HSU00107956) µmol ligands m−2 adsorbent. Individual SC were calculated as the product of the respective normalized mass changes (µg m−2) from TGA measurements and the molar masses (µg µmol−1) of the corresponding split-off fluoroalkyl groups. Through variation of solvent composition from xylene to n-octane/xylene mixtures, the corresponding SC values could be significantly increased by 1.7 to 6 times, resulting in enhanced values of 0.37 to 1.36 µmol ligands m−2 adsorbent for HSU00107957, and HSU00107956, respectively. With regard to ligands HSU54 and HSU55, functionalization in a solvent mixture of n-octane/xylene (1:1, v/v) at modified reaction temperatures of T = 120 °C, and T = 83 °C, respectively proved to be the optimal synthesis conditions.

Additionally, the determined SC was converted into appropriate functionalization degrees (FD/%) on the basis of the Zhuravlev constant (2000), i.e., the maximum number of SiOH groups per unit surface area (α = 4.90 SiOH nm−2, and 8.14 μmol m−2, respectively). Assuming that a maximum of two leaving groups of a silane molecule (ligand) react with two SiOH groups on the silica surface, the maximum achievable surface coverage (SCmax) of the silica with ligand molecules is 4.07 µmol m−2 (namely 0.5α). Individual FDs were then calculated according to the following equation:

$${\text{FD}} = \frac{{{\text{SC}}}}{{ {\text{SC}}_{{{\text{max}}}} }} \times 100 = \frac{{2{\text{SC}}}}{\alpha } \times 100$$
(1)

Individual surface coverages, functionalization degrees as well as specific surfaces of the silica-based adsorbents are summarized in Table 1.

Table 1 Characteristic properties of the functionalized silica adsorbents
Full size table

The 2.5- to fourfold higher functionalization degrees obtained for the adsorbents HSU00107955 (36.6%) and HSU00107956 (33.5%) compared to HSU00107954 (14.6%), HSU00107957 (9.00%), and HSU00107958, respectively, were possibly due to a generally higher rate of hydrolysis of the applied alkoxysilanes (HSU55, HSU56) (Brochier Salon and Belgacem 2011) and steric hindrance.

Synthesis of fluorinated silica by silanization with fluorinated silanes has been carried out since the early 1980s. Berendsen et al. (1980) investigated the suitability of a (heptadeca-fluorodecyl)dimethylsilyl-bonded silica as material for the separation of fluorine-containing compounds in the liquid phase (Berendsen et al. 1980). In the following years, several applications employing fluorinated silica as stationary phases for the liquid chromatographic separation of proteins, peptides, biogenic amines as well as perfluoroalkyl compounds were described (Xindu and Carr 1983; Brittain et al. 2005; Hayama et al. 2012; Marchetti et al. 2012). Compared to the macroporous adsorbents investigated in this study, those fluorous reversed phase materials typically consist of smaller mesoporous particles with a particle size and pore size distribution of 5–60 µm and 60–300 Å, respectively (Marchetti et al. 2015). When applying such small particles in form of a packed separation column, working with aqueous media, especially at elevated flow rates, usually produces high back pressures that must be overcome with help of high pressure pumps. A flow experiment with an aqueous PFOS solution (feed volume: 0.05 L, feed concentration: 56 nmol L−1) and HSU00107956 as adsorbent in a miniaturized fixed-bed adsorber (2 g adsorbent, bed length 40 mm, bed diameter 12.4 mm) resulted in an only gravity-induced average flow rate of 0.35 L h−1, with a simultaneous 97% reduction of the PFOS concentration in the effluent to 1.7 nmol L−1. By virtue of the exemplary but promising flow properties of the investigated HSU00107956 such functionalized macroporous silica adsorbents could be useful materials for a PFAA elimination process of drinking water at necessarily high volume flows but manageable back pressures.

Screening of the silica adsorbents: PFAA removal efficiency

In order to prove the principle, the five newly functionalized adsorbents (HSU00107954-58) were first screened regarding their removal efficiencies at high, but realistic environmental PFAA concentrations as found in European surface and wastewaters (Skutlarek et al. 2006; Mazzoni et al. 2015). Removal efficiencies \(\left( {{\text{RE}} = \frac{{C_{0} - C_{{\text{e}}} }}{{C_{0} }} \times 100 } \right)\) for each individual PFAA were determined at initial concentration (c0) of 10 µg L−1 corresponding to molar concentrations of 20.0 to 33.4 nmol L−1. The results are demonstrated in Fig. 1.

Fig. 1
figure 1

Removal efficiencies (RE/%) of the functionalized silica adsorbents HSU00107954, HSU00107955, HSU00107956, HSU00107957, and HSU00107958 for perfluorobutane sulfonic acid (PFBS), perfluorohexane sulfonic acid (PFHxS), perfluorooctane sulfonic acid (PFOS), and perfluorononanoic acid (PFNA) after 24 h; initial PFAA concentrations (c0) 20.0–33.4 nmol L−1

Full size image

HSU00107957 reached REs for the investigated PFAA between 8.70 to 83.4% (PFBS 8.70%, PFHxS 9.80%, PFOS 40.8%, PFNA 83.4%) comparable to those achieved with HSU00107958 (PFBS 9.50%, PFHxS 16.2%, PFOS 38.5%, PFNA 78.9%). Attainable REs with HSU00107955 and HSU00107956 for PFBS (10.1 and 18.0%), PFHxS (22.2 and 15.3%), and PFNA (86.8 and 93.8%) exhibited the same magnitude as the corresponding RE values for HSU00107957 and HSU00107958 with exception of PFOS, which was eliminated about two times more effective by HSU00107955 and HSU00107956.

Although the carbon chains of the ligands bound to HSU00107957 and HSU00107958 are elongated by two to six fluorinated carbon moieties compared to those bound to HSU00107955 and HSU00107956, theoretically leading to stronger fluorine-fluorine (F-F) interactions with PFOS, the resulting twofold lower PFOS-REs of about 40% for HSU00107957 and HSU00107958 as opposed to approximately 80% for HSU00107955 and HSU00107956, respectively, were probably due to their three to four times lower degrees of functionalization (refer to Table 1). The associated higher hydrophilicity of the adsorbent surface was thus disadvantageous for the adsorption of the rather hydrophobic and nonpolar PFOS with a high log P value of 5.43 (Park et al. 2020). Octanol–water partition coefficient P is a measure for the hydrophobicity (log P > 1) and hydrophilicity (log P < 1) of a given molecular species and also reflects its polarity. In order to compare: PFNA with the same number of perfluorinated carbons (C8) as PFOS but a different functional group, showed a similar but significantly lower negative adsorption tendency from HSU00107956 to HSU00107958 because the higher hydrophobicity of PFNA (log P = 5.81) (Park et al. 2020) probably compensates the higher hydrophilicity of the lower-functionalized adsorbents HSU00107957 and HSU00107958. Overall, the determined REs for the more hydrophilic PFBS (log P = 2.63) (Park et al. 2020) and PFHxS (log P = 4.03) (Park et al. 2020) of the four adsorbents HSU00107955-58 were significantly lower (8.70–22.2%) and in fact almost independent of the adsorbent`s FD and the particular chain length of the immobilized ligand.

Of all adsorbents, HSU00107954 yielded the highest REs for the investigated PFSA, namely 46.3%, 79.1%, and 96.7% for PFBS, PFHxS, and PFOS, respectively. In comparison to HSU00107956, the RE for the examined long-chain PFAS could be improved by 10% for PFOS (86.5–96.7%) and more than 60% for PFHxS (15.3–79.4%) as well as for the short-chain PFBS remarkably by around 30% (18.0–46.9%). The enhanced adsorption performance of HSU00107954 possibly caused by the secondary carboxamide group (formally R1–(C=O)–NH–R2) of ligand HSU54. Because of the mesomeric stabilization of carboxamides, zwitterionic boundary structures are formed with positive charged nitrogen atoms probably capable of additional ionic interactions with the negatively charged PFSA molecules.

Investigation of the equilibrium loadings: Adsorption isotherms

Subsequently, further investigations of the equilibrium loadings of three of the five adsorbents (HSU00107954, HSU00107955, and HSU00107956) with the researched PFSAs were conducted in a concentration range from 0.04 (PFOS) to 3.34 (PFBS) nmol L−1, corresponding to concentrations of 20–1000 ng L−1. The applied concentration range was chosen taking into account the recommended drinking water standard of the German Federal Environment Agency (Umweltbundesamt) for PFHxS (0.25 nmol L−1) as well as PFOS (0.20 nmol L−1) (Umweltbundesamt 2017) and representative literature levels for PFSAs in ground and drinking water. Boone et al. (2019) recently studied 25 drinking water treatment plants across the United States and reported PFBS, PFHxS, and PFOS concentrations in treated drinking water up to 0.04, 0.05, and 0.07 nmol L−1, respectively (Boone et al. 2019).

Experimental data were then fitted to the linear forms of the Langmuir model, the Scatchard linearization (Eq. 2), and the Freundlich model (Eq. 3):

$$\frac{{q_{{\text{e}}} }}{{C_{{\text{e}}} }} = - K_{{\text{L}}} q_{{\text{e}}} + K_{{\text{L}}} q_{{\text{m}}}$$
(2)
$$\log q_{{\text{e}}} = n\log C_{{\text{e}}} + \log K_{{\text{F}}}$$
(3)

These models describe the adsorption of solved compounds (adsorbates) to solid surfaces (adsorbents), where qe is defined as the equilibrium loading of an adsorbent with an adsorbate and qm (in Eq. 2) as the maximum monolayer adsorption capacity of an adsorbent. ce represents the adsorbate concentration in solution at the equilibrium and KL as well as KF represents the Langmuir, and Freundlich coefficients, respectively. n (in Eq. 3) denotes the dimensionless Freundlich exponent, that specifies the extent of surface heterogeneity of an adsorbent (Tran et al. 2017). A linear correlation (coefficient of determination r2 > 0.7) indicates a good fit between the experimental results and the corresponding model. Within the examined low concentration range including the elevated screening level, all experimental derived PFSA isotherms for the three investigated adsorbents strictly followed the linearized Freundlich model (r2 > 0.95), depicted in Figs. 2, 3 and 4.

Fig. 2
figure 2

Linearized Freundlich isotherms for PFSA adsorption onto HSU00107954 (filled square perfluorooctane sulfonic acid [PFOS, r2 = 0.95], filled triangle perfluorohexane sulfonic acid [PFHxS, r2 = 0.98], filled circle perfluorobutane sulfonic acid [PFBS, r2 = 0.99]; initial concentrations (c0) 0.04–33.4 nmol L−1). Resulting PFOS and PFHxS equilibrium concentrations (ce) for c0,PFOS < 0.40 nmol L−1 and c0,PFHxS = 0.05 nmol L−1 were below the corresponding limits of quantification (LOQ) and therefore excluded from the linear regressions

Full size image
Fig. 3
figure 3

Linearized Freundlich isotherms for PFSA adsorption onto HSU00107955 (filled square perfluorooctane sulfonic acid [PFOS, r2 = 0.98], filled triangle perfluorohexane sulfonic acid [PFHxS, r2 = 0.97], filled circle perfluorobutane sulfonic acid [PFBS, r2 = 0.96]; initial concentrations (c0) 0.04–33.4 nmol L−1)

Full size image
Fig. 4
figure 4

Linearized Freundlich isotherms for PFSA adsorption onto HSU00107956 (filled square perfluorooctane sulfonic acid [PFOS, r2 = 0.99], filled triangle perfluorohexane sulfonic acid [PFHxS, r2 = 0.96], filled circle perfluorobutane sulfonic acid [PFBS, r2 = 0.98]; initial concentrations (c0) 0.04–33.4 nmol L−1). Resulting PFOS equilibrium concentrations (ce) for c0,PFOS < 0.20 nmol L−1 were below the corresponding limit of quantification (LOQ) and therefore excluded in the linear regression

Full size image

Deduced KF values from the ordinate intercepts (= log KF, see Figs. 2, 3 and 4) and the n values from the slopes of the calculated linear equations, respectively, varied between KF = 0.33–52.4 × 10–3 nmol(1−n) Ln m−2 and n = 0.82–1.08, where n = 1 signifies adsorption linearity (Tran et al. 2017). Nonlinearity (n < 1) may results from adsorption site heterogeneities or adsorbate–adsorbate interactions (Yu et al. 2009) such as molecular aggregations, that here are rather unlikely due to the low PFSA concentrations used (Cheng et al. 2009). In terms of the surface-bound ligands, the demonstrated virtual adsorption linearities (n > 0.8) of the PFSAs possibly indicate an approximate surface homogeneity of the three functionalized adsorbents analyzed. According to the type of PFSA, a detailed summary of the calculated KF and n values for each silica adsorbent is provided in Table 2.

Table 2 Freundlich coefficients (KF, nmol(1−n) m−2 Ln) and exponents (n, dimensionless) derived from experimental data and literature as well as corresponding coefficients of determination (r2) of the linear regressions
Full size table

Exemplary Scatchard plots (qe ce−1 vs. qe) of the Langmuir isotherms for the adsorption of PFHxS, and PFBS onto HSU00107954 (Fig. 5), and HSU00107955 (Fig. 6), respectively, clarify the poor agreement between the experimental data and the Langmuir model for all adsorbents and becomes evident due to the calculated low r2 values of 0.28, and 0.35, respectively. However, it must be taken into account that the correlation (r) between qe and qe ce−1 is often underestimated in the Scatchard linearization, i.e., Eq. 2 may simulate a poor fit to data even though they actually correspond to the Langmuir model (Tran et al. 2017).

Fig. 5
figure 5

Linearized Langmuir isotherm (Scatchard plot) for the adsorption of perfluorohexane sulfonic acid [PFHxS, r2 = 0.28] onto HSU00107954; initial concentrations (c0) 0.12–25.1 nmol L−1

Full size image
Fig. 6
figure 6

Linearized Langmuir isotherm (Scatchard plot) for the adsorption of perfluorobutane sulfonic acid [PFBS, r2 = 0.35] onto HSU00107955; initial concentrations (c0) 0.07–33.4 nmol L−1

Full size image

Overall, the calculated Freundlich coefficients (KF) as a measure of the adsorption capacities (Tran et al. 2017; Yu et al. 2009) of HSU00107954, HSU00107955, and HSU00107956 for PFOS were higher than for PFHxS and PFBS (refer to Table 2). These results were in agreement with the results of the screening experiments (refer to Fig. 1), confirming that in general the fluorinated silicas adsorb PFOS roughly two to eight times more efficiently from the aqueous samples (RE = 76.0–96.7%) as PFHxS (RE = 15.3–79.4%) and PFBS (RE = 10.1–46.9%). However, comparing the KF values of the functionalized adsorbents with high FDs (see Table 1) for each individual PFSA, notably adsorbent HSU00107956 (FD = 33.5%) exhibited a 20 times higher capacity for PFOS (KF = 7.23 × 10–3 nmol(1−n) Ln m−2) than for PFHxS (KF = 0.38 × 10–3 nmol(1−n) Ln m−2) and PFBS (KF = 0.33 × 10–3 nmol(1−n) Ln m−2) while the adsorption capacities of HSU00107955 (FD = 36.6%) were 5 and 7 times higher for PFHxS (KF = 0.53 × 10–3 nmol(1−n) Ln m−2) and PFBS (KF = 0.33 × 10–3 nmol(1−n) Ln m−2), respectively, compared to that of PFOS (KF = 2.49 × 10–3 nmol(1−n) Ln m−2). Regarding PFOS, the distinct KF difference between HSU00107956 and HSU00107955 (~ 200%; refer to Table 2) at approximately same FDs most likely reflected a strongly increased fluorous affinity of HSU00107956 for long-chain PFSA, possibly due to the extended fluorinated carbon chain of its immobilized ligand HSU56 (C8) in contrast to that of HSU00107955 (ligand HSU55).

Although HSU00107954 showed a twofold lower FD (14.6%) to that of HSU00107956 (33.5%), here the additional ionic interaction of the amide group with the anionic PFSA (see section "Screening of the silica adsorbents: PFAA removal efficiency") provoked five to 20-fold higher KF values in particular for the short-chain PFBS and PFHxS, respectively (refer to Table 2). This can be explained by a usually higher adsorption enthalpy of ionic bonds compared to the enthalpy of instantaneous dipole–dipole bonds that are described as the cause of the specific fluorine interaction (Xu and Oleschuk 2014). HSU00107954 demonstrated the best adsorption performance over all PFSAs. Its high removal efficiencies (REs) reached for PFSA during the screening experiments (46.9–96.7%) were reproducible also in the investigated low nano to picomolar concentration range, illustrated in Fig. 7. Missing REs for PFOS and PFHxS at the lowest initial concentrations in Fig. 7 are due to the fact that the corresponding equilibrium concentrations (ce) laid below their LODs and therefore were not considered in the calculation of the REs. Nevertheless, defining of the respective LODs for PFHxS (0.02 nmol L−1) and PFOS (0.014 nmol L−1) as the maximum achievable equilibrium concentrations (ce = LOD) and using them for RE calculations, results in values of the same order of magnitude, namely 60% for PFHxS and between 65 and 93% for PFOS. Maximum standard deviations of the triple determined REs (n = 3) at the highest initial concentrations studied (2.00–3.34 nmol L−1) were 10% and 14% for PFBS and PFHxS, respectively and for PFOS only 1% over the entire depicted concentration range.

Fig. 7
figure 7

Removal efficiency (RE/%) of HSU00107954 as a function of different initial concentrations (c0) of perfluorobutane sulfonic acid [PFBS, 0.07–3.34 nmol L−1], perfluorohexane sulfonic acid [PFHxS, 0.05–2.51 nmol L−1], and perfluorooctane sulfonic acid [PFOS, 0.04–2.00 nmol L−1]

Full size image

Resulting Freundlich coefficients of the three examined adsorbents were finally matched to modeled KF values for GAC derived by Hansen et al. (2010) from experimental data of corresponding PFSAs (Hansen et al. 2010). In that study, natural PFAA contaminated well water at concentration levels very similar to our study was treated with an activated material originated from anthracite coal (average particle size 226 µm, specific surface 1200 m2 g−1). We converted their published KF values (released in ng(1−n) Ln g−1) for the individual PFSA to amounts of substance (nmol) and normalized to the GAC surface area (m2) in order to obtain values comparable to ours in nmol(1−n) Ln m−2. In sum, individual determined Freundlich coefficients for the adsorption of the investigated PFSA onto HSU00107954, HSU00107955, and HSU00107956 showed much higher values as corresponding KF values for GAC adsorption of PFHxS, and PFOS in the Hansen study (Hansen et al. 2010) and published (Crone et al. 2019) and also converted values for PFBS adsorption in deionized water (refer to Table 2). In particular, the KF values for the adsorption of PFBS (1.60 × 10–3 nmol(1−n) Ln m−2), PFHxS (7.43 × 10–3 nmol(1−n) Ln m−2), and PFOS (52.4 × 10–3 nmol(1−n) Ln m−2) onto HSU00107954 were remarkably 50–60 times higher to those for GAC adsorption (0.03 (Crone et al. 2019), 0.13 (Hansen et al. 2010), and 0.90 (Hansen et al. 2010) × 10–3 nmol(1−n) Ln m−2, respectively) and also three to 10 times higher in case of HSU00107955 and HSU00107956. With respect to the individual PFSA investigated, our findings confirm significantly increased adsorption capacities of the three newly developed functionalized silica adsorbents compared to granular activated carbon (GAC).

The results prove that the newly developed adsorbent HSU0010954 is able to reduce nanomolar as well as picomolar PFSA concentrations below actual drinking water regulation values. For PFBS at an initial concentration range of 0.07–3.34 nmol L−1 (corresponding to 21–1,000 ng L−1) and REs of around 50%, equilibrium concentrations of 10–500 ng L−1 were obtained and are therefore significantly below the USEPA regulation value for PFBS of 6.7 nmol L−1 (corresponding to 2,000 ng L−1) (USEPA 2022a). For PFHxS at an initial concentration range of 0.05–2.51 nmol L−1 (corresponding to 20–1,000 ng L−1) and REs of at least 80%, equilibrium concentrations of 4–200 ng L−1 were detected and therefore predominantly below the regulation value of the German Federal Environment Agency for PFHxS of 0.25 nmol L−1 (corresponding to 100 ng L−1) (Umweltbundesamt 2017).

However, our experiments were carried out in ultrapure water with single PFSA, while Hansen et al. utilized natural well water for their investigations that was natively contaminated with a mixture of PFAA. In more complex samples/matrices the contained compounds compete for adsorption sites, leading to a reduced adsorption of individual substances.

Conclusion

Highly efficient adsorbents for drinking water purification help to achieve the worldwide recommended drinking water limits for perfluorosulfonic as well as carboxylic acids (PFSAs, PFCAs). The herein presented functionalized macroporous silica adsorbents (HSU00107954, HSU00107955, HSU00107956) demonstrate up to 60 times higher PFSA adsorption capacities compared to granular activated carbon (GAC), especially at sub-nanomolar concentrations that are common for raw and drinking water.

Furthermore, our study demonstrates the high efficiency of the novel fluorinated macroporous silica adsorbent (HSU00107954) especially for the removal of short-chain perfluorobutane sulfonic acid (PFBS) as well as the long-chain PFSAs perfluorohexane sulfonic acid (PFHxS), and perfluorooctane sulfonic acid (PFOS) from aqueous solutions at concentration levels also relevant for drinking water. In batch experiments, HSU00107954 was able to reduce picomolar PFSA concentrations by 56.4%, 88.2%, and 97.4% for PFBS, PFHxS, and PFOS, respectively. Determined adsorption capacities of HSU00107954 for each PFSA in the investigated very low concentration range were significantly higher compared to corresponding literature values for granular activated carbon (GAC). Due to the similar particle size distribution and macroporosity to that of GAC, which is commonly employed for drinking water treatment, HSU00107954 performed as an effective adsorbent material for PFSA removal in laboratory experiments and thus could be highly efficient in flow-through systems of drinking water treatment plants.

Nevertheless, extensive further research of adsorption kinetics and dynamics of the new adsorbents are necessary for the development and design of a sufficient adsorbent-based water treatment process under real conditions, e.g., investigations in suitable fixed-bed adsorber arrangements using technically conventional flow rates (from L to m3 h−1) as well as natural water matrices like groundwater.