Separation and Purification Technology 327 (2023) 124894

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Machine learning screening based strategy for the synthesis of a

molecularly imprinted ionic liquid polymer for specific adsorption of
perfluorooctanoic acid
Yiwei Zhang a, 1, Yingjie Luo a, 1, Jie Tong b, Xuesong Liu a, *, Yong Chen a, Tengfei Xu a, *
a
Key Laboratory of Advanced Drug Delivery Systems of Zhejiang Province, College of Pharmaceutical Sciences, Zhejiang University, Hangzhou 310058, Zhejiang, China
b
PET Center, Department of Radiology and Biomedical Imaging, Yale School of Medicine, New Haven, CT 06520, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Perfluorooctanoic acid (PFOA) is a persistent organic pollutant that poses a significant environmental threat due
Machine learning to its resistance to degradation in both aquatic and terrestrial environments. Ionic liquids (ILs) are a type of salt
IL that is liquid at or near room temperature and have been widely used in separation science, as well as showing
MIP
great potential in environmental applications. However, the large number and variety of ILs make them
Remove PFOA
Persistent organic pollutant
impractical to assess the suitability of all possible ILs for PFOA adsorption using purely experimental methods. To
address this issue, a high-throughput machine learning method was employed in this study to screen more than a
thousand ILs and identify the most appropriate molecular for designing IL based molecularly imprinted polymer
(MIP) with high affinity and selectivity of PFOA. The synthesized MIP, (Vim)C3(L-Pro)@MIP, has been proven to
effectively and specifically recognize and adsorb PFOA in water. It has a maximum adsorption capacity of
568.18 mg g− 1 and removal efficiency of over 99%, making it a highly efficient and selective option for PFOA
removal in environmental remediation applications. This study showcases the promising potential of integrating
high-throughput computer screening with experimental verification in the initial stages of MIP material design
and development. By doing so, it is possible to create MIP materials that can effectively remove various pol­
lutants, and potentially extend to other IL doped materials and applications.

1. Introduction [9–10], making it necessary to remove PFOA from contaminated water

sources. Conventional wastewater treatment methods, such as coagu­
Per- and polyfluoroalkyl substances (PFAS) are a kind of aliphatic lation, filtration, aeration, oxidation, and disinfection, have proven to be
organic fluorine compounds, in which hydrogen atoms are partially or ineffective for the removal of PFOA [11]. The traditional method of
completely replaced by fluorine [1]. The carbon–fluorine (C–F) bond is removing PFOA from water is to use granular activated carbon [12].
one of the strongest covalent bonds [2], which makes PFAS highly Advanced technologies such as bioremediation and advanced oxidation/
resistant to degradation in aquatic and terrestrial environments [3–4], reduction methods have also been developed, but they face challenges
making them persistent pollutants. Perfluorooctanoic acid (PFOA), is such as high costs and insufficient environmental protection [13–15]. As
commonly used in the manufacturing of industrial and household a result, there is a need for an environmentally friendly and cost-
products and is the most common PFAS found in the environment [5]. effective adsorbent to remove PFOA from water. Ionic liquids (ILs) are
PFOA has been designated as a Persistent Organic Pollutant (POP) due to organic salts with a low-melting point that consist of a bulky, asym­
its persistence in the environment [6]. PFOA has been detected in water metric cation and a weakly coordinated anion. They possess specific
sources contaminated by industrial facilities and sewage treatment physicochemical properties that make them suitable for various appli­
plants, as well as in finished drinking water [7–8]. cations, including sludge denitrification, control of irreversible fouling
Long-term exposure to low levels of PFOA in drinking water has been in reverse osmosis membrane processes, and passive sampling with ILs,
linked to toxic effects on various organs and systems in the human body among others [16–17].

* Corresponding authors.
E-mail addresses: [email protected] (X. Liu), [email protected] (T. Xu).
1
The authors contributed equally to this work.

https://doi.org/10.1016/j.seppur.2023.124894
Received 15 June 2023; Received in revised form 6 August 2023; Accepted 22 August 2023
Available online 26 August 2023
1383-5866/© 2023 Elsevier B.V. All rights reserved.
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

Choosing appropriate ILs for removing PFOA is a challenging task Machine learning is then used to optimize the selection of ILs based on
due to the random combination of cations and anions in ILs, resulting in their physicochemical properties. The selected ILs are then used to form
thousands of types of ILs. High-throughput screening using artificial MIPs through cross-linking polymerization, using PFOA as a template
intelligence has become a promising method to overcome this challenge. molecule. Finally, the experimental mechanism of action is explored to
Various methods have been used for screening, including quantitative understand the selective adsorption of PFOA by the MIPs. This involves
structure–property relationship (QSPR), UNIversal Functional Activity analyzing the hydrogen bonding interaction and template cavity syn­
Coefficent (UNIFAC), and hasten solubility parameter [16]. However, ergy between the MIP and PFOA, among other factors. The success of
these methods require experimental data to generate training sets, this strategy is confirmed by the experimental results, as the designed
which may not be feasible for collecting experimental data from a large ILs-based MIP, (Vim)C3(L-Pro)@MIP, can remove over 99% of PFOA in
number of ILs. In contrast, COnductor like Screening MOdel for Real aquatic environments.
Solvents (COSMO-RS) is a method that does not require experimental This study demonstrates the mechanism of MIP selective adsorption
data. It is calculated based on the principles of equilibrium thermody­ of PFOA from both theoretical calculation and experimental perspec­
namics and quantum chemistry and calculates the solubility of ILs using tives, which is based on hydrogen bonding interaction and template
weak intermolecular interactions such as static electricity, hydrogen cavity synergy. Therefore, this work highlights the strong potential of
bonding, and molecular hydrophobicity [17]. COSMO-RS has been used combining high-throughput computer screening with experimental
in practical research, such as predicting the surface charge density dis­ verification in the early stages of designing and developing MIP mate­
tribution of ILs and gas solubility in polymers [17–18]. Therefore, using rials for various pollutant removal applications, potentially expanding
COSMO-RS to screen ILs that are compatible with PFOA may be a time- to other IL-doped materials and applications.
saving and labor-saving method.
While ILs have good solubility in water, separating the IL-PFOA 2. Experiments
complex from water can be challenging. Solid adsorption is a feasible
alternative to liquid–liquid separation for industrial wastewater treat­ 2.1. Materials
ment. Molecularly imprinted polymers (MIPs) are often used as solid
adsorbents due to their selectivity [19]. MIPs are polymers formed by The chemicals used in the experiments included 1-bromopropane, 1-
imprinting based on the structure and geometric shape of the target ethylimidazole, L-proline, octanoic acid, N-[(Prop-2-enoylamino)
molecule, and they have been used effectively to selectively remove methyl]prop-2-enamide (MBA), acetonitrile, potassium persulfate
various pollutants in water, such as 2,4-dinitrophenol, chloramphenicol, (K2S2O8), sodium hydroxide (NaOH), hydrochloric acid (HCl), meth­
17 beta-estradiol, perfluorooctane sulfonate and norfloxacin [20–24]. anol, L-alanine, L-valine, L-leucine, D-glucose, and acetaminophen. All
Therefore, ILs can be designed as functional monomers of MIP solid of these chemicals were obtained from Aladdin Chemical Reagent Co.
adsorbents, with PFOA analogues selected as template molecules to Ltd (Shanghai, China). The 201 × 7 strong basic anion exchange resin
reduce pollution during the production process. The preparation of ad­ used in the experiments was purchased from Yuan Ye Biotechnology Co.
sorbents for selective adsorption of PFOA in water environments in­ Ltd (Shanghai, China). The Milli-Q water used in the experiments was
volves using IL obtained from high-throughput screening of COSMO-RS sourced from laboratory purification systems.
as functional monomers, PFOA structural analogue octanoic acid as
template molecules, and cross-linking polymerization to obtain MIP. As 2.2. COSMO-RS based PFOA solubility calculation
shown in Scheme 1, The process of preparing adsorbents for selective
adsorption of PFOA in water involves several steps, including database The solubility of IL to PFOA was calculated using the COSMO-RS
screening, machine learning, MIP formation, and experimental explo­ [25] module of the AMS software (https://www.scm.com), which uti­
ration of the mechanism of action. First, a database screening is per­ lizes the COSMO-RS default method and ADF Combi2005 parameters
formed to identify suitable ILs for use as functional monomers in MIPs. [26]. The process began with the geometric optimization of a molecule

Scheme 1. The process of preparing adsorbents for selective adsorption of perfluorooctanoic acid (PFOA) in water.

2
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

(either solvent or solute) through density functional theory (DFT), fol­ allow the interaction between the IL monomers and template molecule
lowed by the generation of an amsterdam density functional (ADF) to form a pre-polymerized solution. MBA (1.18 g, 11.5 mmol) was then
COSMO result file that contains the screening surface charge density on added, followed by K2SO4 (50 mg) as the initiator. The solution was
the molecule surface. The distribution of the screening charge density stirred under a nitrogen atmosphere for 10 min and heated to 60 ◦ C for
was then generated through quantum mechanical calculations, resulting 12 h to allow the reaction to occur. The resulting solid was ground,
in the σ-profile. By treating the solvent as a dielectric continuum, the screened, and washed with NaOH solution (0.1 mol/L) and water until
COSMO-RS model combined the σ-profile and thermodynamic proper­ no template could be detected in the washing solution. Finally, the
ties to predict the solubility of a molecule in the solvent [27]. The sol­ product was dried in a vacuum drying oven at 60 ◦ C to obtain the (Vim)
ubility of PFOA in IL was calculated through the method of “Solubility in C3(L-Pro)@MIP. As a control, (Vim)C3(L-Pro)@NIP were prepared using
Mixture”, utilizing the combined anion and cation result files. The the same procedures, except without the template molecule.
temperature was set to 298.15 K for all COSMO-RS calculations.
2.4. Characterization of the adsorbents
2.3. Synthesis of (Vim)C3(L-Pro)@MIP
The functional monomer IL and the prepared MIP and Non-
The synthesis route of the IL monomer used in the preparation of MIP molecularly imprinted polymer (NIP) were tested by FTIR and TGA,
is depicted in Fig. 1A. The process involved dissolving 1-Bromopropane and the surface structure of MIP and NIP was scanned by scanning
(24.60 g, 0.2 mol) and N-vinylimidazole (18.82 g, 0.21 mol) in aceto­ electron microscopy (SEM). Fourier transform–infrared (FT-IR) spectra
nitrile (100 mL) and stirring the mixture at 82 ◦ C for 24 h. The resulting were recorded by a transmission method on a Thermo Scientific NICO­
mixture was then distilled under reduced pressure to obtain the LET iS50FT-IR using KBr pellets. Thermogravimetric Analysis (TGA) is
yellowish liquid (Vim)C3Br. This mixture was further processed by dis­ used to characterize the decomposition and thermal stability of mate­
solving (Vim)C3Br (10 g, 0.046 mol) in water and passing the solution rials. In addition, the morphology of the material was observed by a
through a 201 × 7 strong basic anion exchange resin to obtain (Vim) Zeiss Sigma 300 SEM.
C3OH. L-proline (5.50 g, 0.046 mol) was added to the (Vim)C3OH so­
lution and stirred at 25 ◦ C for another 24 h. The mixture was distilled 2.5. Batch sorption experiments
under reduced pressure to obtain the wine-red liquid (Vim)C3(L-Pro).
MIP was synthesized and prepared according to the route shown in The batch adsorption experiments were conducted using a shaker at
Fig. 1B. First, a pore-forming agent solution was prepared by mixing a constant speed of 300 rpm. The objective was to study the effect of
(Vim)C3(L-Pro), water, and methanol were mixed in a 100 mL round- various parameters on the adsorption of PFOA, including contact time,
bottom flask. Octanoic acid was then added as the template molecule solid–liquid ratio, pH, and temperature. These parameters were opti­
to form a monomer-template mixture. (Vim)C3(L-Pro) (912 mg, 3.5 mized through a single factor experiment. The pH of the solution was
mmol) was dissolved in the water/methanol (1:5, v/v) solution, and adjusted using a sodium hydroxide/hydrochloric acid system. The
octanoic acid (64.89 mg, 0.45 mmol) was added. The mixture was adsorption kinetics experiment was performed by adding 60 mg of the
sonicated for 15 min and allowed to stand overnight at 25 ◦ C to fully MIP into 100 mL of PFOA solution, with an initial concentration of 300

Fig. 1. A: The synthesis route of (Vim)C3(L-Pro); B: The synthesis route of (Vim)C3(L-Pro)@MIP.

3
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

mg/L. The isothermal adsorption experiment was conducted at an initial 3. Results and discussion
PFOA concentration ranging from 100 to 1000 mg/L, and the adsorption
time was kept constant at 1 h, based on the results from the single factor 3.1. COSOMO-RS screening results
experiment.
The cation categories used in the COSMO database include imid­
azole, guanidine, choline, and pyridine, while the anion categories
2.6. Gauss calculation include halogen, carboxyl, sulfate, phosphate, cyanide, and metal ions,
all of which are the main categories for ILs. These 82 cations and 57
The molecular models are built with Gaussian view 6.0 [28] to assist anions obtained the affinity of their combined ILs for PFOA through
in the use of Gaussian 16 [29] software version. The frequency calcu­ COSMO-RS algorithm. The ranking results of affinity scores are shown in
lation and geometric optimization (OPT + freq) of IL and PFOA are Table S1. Among the top ten ILs with high affinity for PFOA, different
carried out at the 6–311 + G (d, p) level of theory by using DFT/B3LYP- combinations of cations and anions exhibited favorable affinity towards
D3 density functional method in water environment. The results are PFOA. Such as the scores for the ILs 1-Vinylimidazole@L(-)-Proline,
convergent without virtual frequency and energy minimum, and stable octylcholinium@nitrate, 1,3-dimethyl-imidazolium@lactate and tetra-
structure is obtained. The weak interactions calculations were corrected N-butylphosponium@nitrate (presented as cation@anion) is calculated
for basis set overlap errors using basis set superposition error (BSSE) as 0.90777977, 0.90776458, 0.90775831 and 0.90772887, respec­
correction. To visualize the interaction, the weak interactions electro­ tively. The anions and cations of this parent nucleus are prioritized. It is
static potential penetration diagram is built with VMD [30] visualization important to note that the selected cations and anions for subsequent
program and Multiwfn [31] software. adsorption of PFOA from water should be environmentally friendly and
possess potential for synthesizing MIP. Notably, the top-ranked cation in
the IL is 1-vinylimidazole and the anion is L-proline, both being rela­
2.7. Selective sorption experiments tively environmentally friendly materials that can reduce secondary
pollution during production and application. Furthermore, 1-vinylimi­
To verify the selectivity of (Vim)C3(L-Pro) @MIP to PFOA, the ab­ dazole contains an alkene side chain structure that is capable of poly­
sorption was studied in the presence of interferences. A mixture con­ merization. Taking into consideration these aspects, the IL (named as
taining 50 μM PFOA, L-alanine, L-valine, L-leucine, D-glucose and (Vim)C3(L-Pro)) was chosen as the subsequent research candidate.
paracetamol was prepared in 50 mL of water, and 0.01 g of (Vim)C3(L- It is speculated that (Vim)C3(L-Pro) should have good water solubi­
Pro)@MIP was added. The mixture was then shaken at 25 ◦ C for 1 h. To lity due to that there no inter hydrophobic site. Therefore, direct
further prove the significance of the cavity for selective adsorption, the application of (Vim)C3(L-Pro) for the adsorption of PFOA in water may
adsorption capacity of MIP and NIP was compared. 0.1 g of each ma­ present challenges in separating the resulting mixture from the aqueous
terial was added to 50 mL of PFOA aqueous solution of the same con­ medium. Covalent crosslinking of the IL into a MIP, which is typically
centration, and the mixture was shaken for 1 h at 25 ◦ C to compare the presented in solid form as a polymer, may address the challenge through
adsorption effect of (Vim)C3(L-Pro)@MIP and NIP. solid–liquid separation. To minimize pollution in the manufacture of
MIP, this study has innovatively employed octanoic acid, which shares
structural similarity with PFOA, as a template molecule. The resulting
2.8. Recycling experiment MIP, (Vim)C3(L-Pro)@MIP, was found to exhibit excellent adsorption
efficacy for PFOA upon preliminary application. Further exploration of
In the experiment, the adsorption of PFOA by (Vim)C3(L-Pro)@MIP the structure of (Vim)C3(L-Pro)@MIP and the adsorption mechanism of
was studied with an initial concentration of 200 μg mL− 1. 100 mg of PFOA in water is thus warranted.
(Vim)C3(L-Pro)@MIP was added to the solution and shaken at 300 rpm
for 1 h at 25℃ using a constant temperature shaker. After each round of 3.2. Characterization of (Vim)C3(L-Pro)@MIP
adsorption, the (Vim)C3(L-Pro)@MIP were cleaned with a 0.1 mol/L
NaOH/methanol solution by ultrasound, filtered, washed with deion­ Fig. 2 shows the characterization data of the final synthesized MIP
ized water three times, dried, and reused for the next round of adsorp­ material, including FT-IR and SEM images. The FT-IR spectrum of (Vim)
tion. The recovery experiment was repeated for five rounds. C3(L-Pro) can be found in Fig. S1. The structural similarities between the
(Vim)C3(L-Pro)@MIP (contains octanoic acid as the template molecule)
and the (Vim)C3(L-Pro)@NIP (devoid of any template molecule), were
2.9. PFOA detection investigated through analysis of their respective FT-IR spectra. Notably,
both spectra exhibited strong peaks corresponding to the stretching vi­
The adsorbed (Vim)C3(L-Pro)@MIP were removed by centrifugation brations of the C=O bond in ketones within the range of 1700 cm− 1 to
and the supernatant was filtered through 0.22 μm membrane. The pre­ 1600 cm− 1, as well as middle peaks in the range of 1300 cm− 1 to 1200
sent study employed the ThermoFisher HPLC-CAD technique to deter­ cm− 1 arising from the stretching vibrations of the C–N bond, suggesting
mine the concentration of PFOA, utilizing a Waters ZORBAX SB-C18 the successful crosslinking of the IL through the introduction of MBA
reverse-phase column (4.6 × 250 mm). The mobile phase contains 80% crosslinkers (Fig. 2A). Subsequent SEM imaging analyses revealed that
methanol and 20% water at flow rate of 0.8 mL min− 1, the column the (Vim)C3(L-Pro)@MIP possessed a more porous and fluffy surface
temperature was set as 35℃, and the injection volume was 0.3 μL. The relative to the (Vim)C3(L-Pro)@NIP, indicative of the influence of the
removal rate and adsorption quantity of PFOA were ascertained through template molecule, octanoic acid, on the morphology of the crosslinking
a comparative analysis of its concentration prior to and following the polymerization product (depicted in Fig. 2B), with corresponding higher
adsorption process. The adsorption capacity of the binding was deter­ specific surface area (Fig. 2B). Importantly, the removal of octanoic acid
mined using the equation: from (Vim)C3(L-Pro)@MIP was correlated with an increased exposure of
Q = (C0 − Ce ) V / m adsorption cavities, while thermal stability measurements indicated that
the (Vim)C3(L-Pro)@MIP containing the template molecule possessed
where Q (mg/g) represents the experimental adsorption quantity, C0 superior stability and showed successful removal of the template
(mg/L) is the initial analyte concentration, Ce (mg/L) is the equilibrium molecule from the crosslinked IL, as evidenced by the transition of the
analyte concentration, V (L) is the volume of the solution, and m (g) is decomposition temperature range from 350 ◦ C to 440 ◦ C before
the mass of MIP. desorption as compared with 250 ◦ C and 450 ◦ C after desorption

4
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

Fig. 2. A: FT-IR spectra of (Vim)C3(L-Pro)@MIP before wash, (Vim)C3(L-Pro)@MIP after wash and (Vim)C3(L-Pro)@NIP; B: SEM of (Vim)C3(L-Pro)@MIP before
wash, (Vim)C3(L-Pro)@MIP after wash and (Vim)C3(L-Pro)@NIP.

(Fig. S2). and SEM characterization of the MIP. In this section, we conducted
single-factor experiments to optimize the adsorption conditions of MIP
3.3. Single factor investigation and explore its maximum adsorption potential. The main experimental
results about single-factor experiments are shown in Fig. 3. As depicted
The feasibility of PFOA adsorption by MIP was inferred from two in Fig. S3, (Vim)C3(L-Pro)@MIP achieved a PFOA adsorption capacity of
perspectives: functional IL monomer molecules and polymer spatial 100 mg g− 1 within 10 s, which is approximately one-third of the
structures, using high-throughput screening of IL functional monomers maximum adsorption capacity, and achieved complete equilibrium

Fig. 3. A: The impact of varying (Vim)C3(L-Pro)@MIP weight on the adsorption rate of PFOA at a constant concentration; B: The effect of constant (Vim)C3(L-Pro)
@MIP weight on the adsorption rate of PFOA at varying initial concentrations; C: The influence of different pH values on the adsorption rate; D: The effect of varying
temperature on the adsorption rate.

5
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

within 1 h. The initial concentration of PFOA in each group is 1 mg ml− 1, 3.5. Adsorption kinetics of (Vim)C3(L-Pro)@MIP
the volume of PFOA solution in each group of Figures A and B is 20 mL
(Fig. 3A). In the group with a MIP of 150 mg, the removal rate of PFOA is Based on theoretical calculations, the hydrogen bond between the IL
greater than 99%, it can calculate that the mass ratio of PFOA to MIP is functional monomer and PFOA is the main driving force for the
approximately 2:15. To confirm this ratio, the next group (Fig. 3B) of adsorption of PFOA by (Vim)C3(L-Pro)@MIP. However, to enhance the
experiments set different initial concentrations of PFOA solutions, but credibility of this conclusion, it is necessary to include experimental data
fixed the weight of each group of MIP to 150 mg, and the volume of each on the adsorption process. The study of adsorption kinetics can deter­
group of solution was still 20 mL. The results showed that the group with mine which process adsorption is closer to by fitting the adsorption
a maximum mass ratio of 2:15 could achieve a removal rate of over 99%. process with the kinetic model. The pseudo-first order kinetic model
A lower pH showed a better adsorption effect, indicating that the assumes that the adsorption rate is controlled by the diffusion step,
adsorption driving force between (Vim)C3(L-Pro)@MIP and PFOA is while the pseudo-second order kinetic model assumes that the adsorp­
achieved through electrostatic interaction and hydrogen bonding tion rate is determined by the square of the unoccupied adsorption va­
(Fig. 3C), consistent with the principle displayed by the COSMO-RS cancies on the adsorbent surface. The adsorption process is controlled by
screening results. Moreover, the temperature of the aqueous solution the chemical adsorption mechanism, which involves electron sharing or
had no significant effect on the adsorption performance (Fig. 3D). transfer between the adsorbent and the adsorbed substance. The equa­
tion for the pseudo-first order kinetic model is expressed as follows:
3.4. Gauss calculation
ln(Qe − Qt ) = lnQe − k1 t

The screening mechanism of COSMO-RS and single factor experi­ Similarly, the equation for the pseudo-second order kinetic model is:
ments initially indicates that the adsorption driving force of the MIP on
t/Qt = 1/k2 Q2e + t/Qe
PFOA is attributed to the hydrogen bonding force. To further investigate
the intermolecular forces involved in the driving force, Gauss calcula­ Where Qe and Qt represent the adsorption amount of (Vim)C3(L-Pro)
tion was utilized to perform calculations on the weak intermolecular
@MIP at equilibrium and at a certain moment during the adsorption of
interactions between the (Vim)C3(L-Pro) functional monomers and PFOA, respectively, with units of mg/g. The constants k1 and k2
PFOA. The geometrical and electrostatic potential analysis of PFOA and
respectively represent the first-order and second-order kinetic constants,
(Vim)C3(L-Pro) revealed positively charged groups on the carboxyl with units of g mg− 1 min− 1.
hydrogen of PFOA and negatively charged groups on the carboxyl ox­ The experimental data on the adsorption kinetics of PFOA by (Vim)
ygen of L-proline (Fig. 4A and B), which could potentially function as C3(L-Pro)@MIP is presented in Fig. S4, which shows that the adsorption
donor and acceptor groups for hydrogen bonding. The results of the system reaches equilibrium within 30 min of contact between the
weak intermolecular interaction calculation between (Vim)C3(L-Pro) adsorbent and PFOA. The kinetics model was used to fit the experi­
and PFOA confirmed strong hydrogen bonding [32] with an angle of H- mental data, and the results are presented in Table 1. The results indi­
O–H at 173.83◦ , length of 1.62 Å, and binding energy of − 126.48 kcal cate that the pseudo-second order kinetic model fits better, as it has a
mol− 1 (Fig. 4C). Consequently, these Gaussian calculations further higher R [2] value (Fig. 5A and B). The maximum adsorption capacity of
corroborate that the driving force behind the MIP adsorption of PFOA is (Vim)C3(L-Pro)@MIP for PFOA was calculated to be 330.50 mg g− 1.
primarily due to hydrogen bonding between the IL functional monomers Therefore, the adsorption process is mainly driven by chemical in­
and PFOA. teractions between MIP and PFOA through electron transfer or sharing,
which is consistent with the principle of hydrogen bonding.

3.6. Adsorption isotherms of (Vim)C3(L-Pro)@MIP

Adsorption kinetics explains the chemical mechanism of adsorption,

whereas adsorption isotherms explain the physical aspects of the
adsorption process. The thermodynamic models used to analyze
adsorption isotherms include the Langmuir and Freundlich models. By
fitting experimental data to these models, it is possible to determine
which model is more suitable for the adsorption process. The Langmuir
model assumes a uniform single-layer adsorption process on a uniform
surface, while the Freundlich model assumes that the adsorption is
mainly heterogeneous multi-layer adsorption caused by non-covalent
bond interactions on the adsorbent surface. The equations for the
Langmuir and Freundlich models are as follows:
Ce /Qe = 1/Qmax Ce + 1/KF Qmax

lnQe = lnKL + 1/nlnCe

In the equation, Ce and Qe denote the PFOA concentration (mg/L)
and adsorption capacity (mg/g) at equilibrium, respectively. Qmax rep­
resents the saturation adsorption capacity (mg/g), while KL and KF are
the Langmuir and Freundlich equilibrium constants (L/mg and mg/g),
respectively. N is the Freundlich coefficient.
The experimental data was fitted to the Langmuir and Freundlich
Fig. 4. A: Optimal geometry and electrostatic potential diagram structure of models (Table 2), and the linear fitting is visualized in Fig. 5C and D. The
PFOA; B: Optimal geometry structure and electrostatic potential diagram of results indicate that the Langmuir model has a higher R2 fitting result,
(Vim)C3(L-Pro); C: Non-covalent binding structure and electrostatic potential indicating that the adsorption process is more in line with its assump­
penetration diagram between (Vim)C3(L-Pro) and PFOA. tions. It shows that the active site on the adsorbent surface are relatively

6
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

Table 1
Parameters in the pseudo-first-order and pseudo-second-order kinetic models.
Temperature (A) Pseudo-first order kinetic model (B) Pseudo-second order kinetic model
− 1 2
K1(min ) Qe R K2 Qe R2
1
(mg/g) (mg g− min− 1) (mg/g)

298 K 0.0851 152.63 0.8997 33765146.83 330.50 0.9983

308 K 0.0616 147.52 0.6808 44134551.5 326.00 0.9984
318 K 0.1093 100.68 0.8770 74577583.23 347.82 0.9995

Fig. 5. Linear fitting for the processing of (Vim)C3(L-Pro)@MIP adsorption of PFOA. A: Pseudo-first order kinetic model; B: Pseudo-second order kinetic model; C:
Langmuir model; D: Freundlich model.

Table 2
The parameters in the fitted Freundlich and Langmuir models.
Temperature R2 A) Langmuir isotherm parameters B) Freundlich isotherm parameters

Qmax KL R2 1/n KF
(mg g− 1) (mg g− 1) (L mg− 1)1/n (L mg− 1)

298 K 0.9305 568.18 3777.051241 0.6579 7.8259 203.12

308 K 0.9240 571.43 3456.70904 0.6310 7.9321 201.14
408 K 0.9375 559.66 3577.479623 0.6532 8.0128 201.72

uniform, and the adsorption of PFOA from water follows the single-layer 3.7. Adsorption selectivity
adsorption mode. The maximum adsorption capacity at room tempera­
ture was calculated to be 568.18 mg g− 1, and compared with the Previous theoretical calculations and experimental results suggest
experimental value of 542.60 mg g− 1, it further confirms that the that hydrogen bonding is the driving force for (Vim)C3(L-Pro)@MIP
adsorption process conforms to Langmuir model. To compare the adsorption of PFOA, it is well-known that hydrogen bonds can exist as
adsorption performance of (Vim)C3(L-Pro)@MIP with other materials, long as there are suitable donors and receptors, which limits their
we have compiled the adsorption capacity of materials that adsorb PFOA contribution to the selective adsorption of PFOA by adsorbents. Here, we
reported in other literature (Table S2). The (Vim)C3(L-Pro)@MIP investigate the adsorption selectivity of the synthesized (Vim)C3(L-Pro)
exhibited better absorption capacity for PFOA than the materials pre­ @MIP. Interfering molecules with common hydrogen bonding donor
pared in previous reports, such as the powdered activated carbon can groups, such as carboxyl and hydroxyl groups, but different geometric
reach a maximum adsorption capacity of 277.4 mg g− 1, while the FCX4- configurations from PFOA, were selected for co-adsorption. The struc­
P presented a maximum adsorption capacity of 188.7 mg g− 1 [33,34]. tures of these interfering molecules are shown in Fig. 6A. The amino acid
The results demonstrate that (Vim)C3(L-Pro)@MIP has a higher molecules selected share a carboxyl structure and have relatively small
adsorption capacity, indicating its superior performance as an adsorbent side chain hindrance, which enables them to potentially enter the cavity.
for PFOA. Glucose was also chosen because of its multiple hydrogen bond donor
hydroxyl groups and soft long chain structure, which is similar to PFOA.
Aspirin was selected due to the rigidity of the benzene ring in its

7
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

Fig. 6. A: Comparison of adsorption efficiency between (Vim)C3(L-Pro)@MIP and NIP. B: The structure of the interfering substances. C: Adsorption selectivity of
(Vim)C3(L-Pro)@MIP for molecules with different structures. D: Regeneration of the (Vim)C3(L-Pro)@MIP.

structure, which prevents it from entering the pores through the soft side cycles of adsorption and desorption, demonstrating the excellent
chain structure like PFOA. These five molecules were added to water regeneration performance of (Vim)C3(L-Pro)@MIP, which is advanta­
together with PFOA as interference factors, kept at a concentration of 50 geous for its practical production and application.
µM for all the molecules, and 0.01 g (Vim)C3(L-Pro)@MIP was subse­
quent added (according to the single factor experiment, the added 4. Conclusion
adsorbent can exactly adsorb PFOA). The results in Fig. 6B indicate that
the adsorption efficiency of (Vim)C3(L-Pro)@MIP for PFOA is almost In our proposed design of a PFOA adsorbent, we first utilize machine
100% (with no residual in the solution), while (Vim)C3(L-Pro)@MIP has learning to screen for ILs with relatively high affinity for PFOA. Next, we
almost no adsorption for other interfering substances, indicating that use crosslinking agents to preserve the location of the affinity sites in the
MIP has extremely good adsorption selectivity for PFOA. functional monomers. Finally, the cavity formed in the MIP should have
Furthermore, we conducted an additional set of experiments to a matching spatial structure to PFOA. Using a high-throughput com­
investigate the importance of (Vim)C3(L-Pro)@MIP cavities in adsorb­ puter screening method, we identified a highly compatible IL with PFOA
ing PFOA. Under identical conditions, the efficiency of (Vim)C3(L-Pro) as a functional monomer, and subsequently used octanoic acid as a
@MIP and (Vim)C3(L-Pro)@NIP (without inside cavities) in adsorbing template molecule to crosslink and polymerize the IL into MIP. The
PFOA was separately tested. As depicted in Fig. 6C, the adsorption ef­ synthesized MIP efficiently and selectively removes PFOA from water
ficiency of (Vim)C3(L-Pro)@MIP is about two fold higher than that of due to its absorption capacity and specific cavity. The mechanism of
(Vim)C3(L-Pro)@NIP. (Vim)C3(L-Pro)@NIP can only rely on the (Vim)C3(L-Pro)@MIP adsorption of PFOA was explored using theoret­
hydrogen bonding force between the IL functional monomer and PFOA ical calculations and experimental methods, revealing a non-covalent
to generate adsorption, resulting in a significant reduction in the multilayer adsorption process through hydrogen bonds and material
adsorption efficiency. In summary, the results demonstrate that (Vim) voids. This approach not only developed a PFOA-specific absorbent
C3(L-Pro)@MIP can selectively adsorb PFOA in water. This also confirms material in a more efficient and less labor-intensive way, but can also
that selecting octanoic acid as a template molecule is reasonable and potentially be applied to other IL materials and environmental pollut­
feasible, as it provides suitable cavity for PFOA. Furthermore, the pre­ ants, even for other classes of chemicals.
sent results preliminarily suggest that the existence of (Vim)C3(L-Pro)
@MIP molecular cavity and hydrogen bonds, contribute to its Declaration of Competing Interest
selectivity.
The authors declare that they have no known competing financial
3.8. Regeneration of the (Vim)C3(L-Pro)@MIP adsorbents interests or personal relationships that could have appeared to influence
the work reported in this paper.
The desorption performance and stability of materials are critical
factors that determine their regenerative performance and practical Data availability
applications. The desorption process following (Vim)C3(L-Pro)MIP
adsorption of PFOA is straightforward, and the material can be regen­ Data will be made available on request.
erated by washing it with readily available solvents and reagents such as
methanol, water, and sodium hydroxide. As illustrated in Fig. 6D, the
efficiency of removing PFOA remains largely unchanged over the five

8
Y. Zhang et al. Separation and Purification Technology 327 (2023) 124894

Acknowledgements [15] M. Boleij, T. Seviour, L.L. Wong, M.C.M. van Loosdrecht, Y. Lin, Solubilization and
characterization of extracellular proteins from anammox granular sludge, Water
Res. 164 (2019), https://doi.org/10.1016/j.watres.2019.114952.
The authors acknowledge the financial support provided by the [16] D.J.C. Constable, C. Jimenez-Gonzalez, R.K. Henderson, Perspective on solvent use
National Natural Science Foundation of China (Grant No.:82204584), in the pharmaceutical industry, Org. Process Res. Dev. 11 (2007) 133–137, https://
Zhejiang Province Human Resources and Social Security Department doi.org/10.1021/op060170h.
[17] W. Tu, et al., An ionic fragments contribution-COSMO method to predict the
(Grant No.ZJ2022074) and Chinese Pharmacopoeia Commission (Grant surface charge density profiles of ionic liquids, J. Mol. Liq. 282 (2019) 292–302,
No.: 2022Y12). https://doi.org/10.1016/j.molliq.2019.03.004.
[18] R. Zhu, Z. Lei, COSMO-based models for predicting the gas solubility in polymers,
Green Energy Environ. 6 (2021) 311–313, https://doi.org/10.1016/j.
Appendix A. Supplementary data gee.2021.03.009.
[19] M. Arabi, et al., Molecular imprinting: green perspectives and strategies, Adv.
Supplementary data to this article can be found online at https://doi. Mater. 33 (2021), https://doi.org/10.1002/adma.202100543.
[20] Z.M. Idris, et al., Amino-functionalised silica-grafted molecularly imprinted
org/10.1016/j.seppur.2023.124894. polymers for chloramphenicol adsorption, J. Environ. Chem. Eng. 8 (2020),
https://doi.org/10.1016/j.jece.2020.103981.
References [21] T. Jing, et al., Highly effective removal of 2,4-dinitrophenolic from surface water
and wastewater samples using hydrophilic molecularly imprinted polymers,
Environ. Sci. Pollut. Res. 21 (2014) 1153–1162, https://doi.org/10.1007/s11356-
[1] R.C. Buck, et al., Perfluoroalkyl and polyfluoroalkyl substances in the environment:
013-2007-0.
terminology, classification, and origins, Integr. Environ. Assess. Manag. 7 (2011)
[22] M. Le Noir, A.-S. Lepeuple, B. Guieysse, B. Mattiasson, Selective removal of 17
513–541, https://doi.org/10.1002/ieam.258.
beta-estradiol at trace concentration using a molecularly imprinted polymer, Water
[2] R. Natarajan, R. Azerad, B. Badet, E. Copin, Microbial cleavage of C–F bond,
Res. 41 (2007) 2825–2831, https://doi.org/10.1016/j.watres.2007.03.023.
J. Fluor. Chem. 126 (2005) 425–436, https://doi.org/10.1016/j.
[23] Q. Yu, S. Deng, G. Yu, Selective removal of perfluorooctane sulfonate from aqueous
jfluchem.2004.12.001.
solution using chitosan-based molecularly imprinted polymer adsorbents, Water
[3] O. Quinones, S.A. Snyder, Occurrence of perfluoroalkyl carboxylates and sulfonates
Res. 42 (2008) 3089–3097, https://doi.org/10.1016/j.watres.2008.02.024.
in drinking water utilities and related waters from the United States, Environ. Sci.
[24] Y. Hu, et al., Surface-imprinted polymer microsphere based on self-stabilized
Tech. 43 (2009) 9089–9095, https://doi.org/10.1021/es9024707.
precipitation polymerization for selective removal of norfloxacin, Appl. Surf. Sci.
[4] M. Park, et al., Adsorption of perfluoroalkyl substances (PFAS) in groundwater by
574 (2022), https://doi.org/10.1016/j.apsusc.2021.151706.
granular activated carbons: Roles of hydrophobicity of PFAS and carbon
[25] C, p. c., t, z., v, l. e. & n, l. j., An implementation of the conductor-like screening
characteristics, Water Res. 170 (2020), https://doi.org/10.1016/j.
model of solvation within the Amsterdam density functional package—part II.
watres.2019.115364.
COSMO for real solvents 1, Can. J. Chem. 87 (2009) 369–408.
[5] K. Steenland, T. Fletcher, D.A. Savitz, Epidemiologic evidence on the health effects
[26] R, x., i, s. s. & i, b. r., An improvement to COSMO-SAC for predicting
of perfluorooctanoic acid (PFOA), Environ. Health Perspect. 118 (2010)
thermodynamic properties, Ind. Eng. Chem. Res. 53 (2014) 8265–8278.
1100–1108, https://doi.org/10.1289/ehp.0901827.
[27] M. Lotfi et al., in: 4th International Conference on Process Engineering and
[6] C. Chen, Y. Fang, X. Cui, D. Zhou, Effects of trace PFOA on microbial community
Advanced Materials (ICPEAM), 2016, pp. 459–466.
and metabolisms: microbial selectivity, regulations and risks, Water Res. 226
[28] R.K. Ksg. Dennington, A. Todd, John M. Millam, Semichem Inc., Shawnee Mission,
(2022), https://doi.org/10.1016/j.watres.2022.119273.
version 06, 2016.
[7] L. Liu, et al., Removal of perfluorooctanoic acid (PFOA) and perfluorooctane
[29] M.J. Frisch, G.W. T., H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, G.
sulfonate (PFOS) from water by carbonaceous nanomaterials: a review, Crit. Rev.
Scalmani, V. Barone, G.A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A.V.
Environ. Sci. Technol. 50 (2020) 2379–2414, https://doi.org/10.1080/
Marenich, J. Bloino, B.G. Janesko, R. Gomperts, B. Mennucci, H.P. Hratchian, J.V.
10643389.2019.1700751.
Ortiz, A.F. Izmaylov, J.L. Sonnenberg, F Ding Williams, F. Lipparini, F. Egidi, J.
[8] L.J.L. Espartero, et al., Health-related toxicity of emerging per- and polyfluoroalkyl
Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V.G. Zakrzewski, J. Gao,
substances: comparison to legacy PFOS and PFOA, Environ. Res. 212 (2022),
N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J.
https://doi.org/10.1016/j.envres.2022.113431.
Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K.
[9] V. Barry, A. Winquist, K. Steenland, Perfluorooctanoic acid (PFOA) exposures and
Throssell, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro, M.J. Bearpark, J.J. Heyd,
incident cancers among adults living near a chemical plant, Environ. Health
E.N. Brothers, K.N. Kudin, V.N. Staroverov, T.A. Keith, R. Kobayashi, J. Normand,
Perspect. 121 (2013) 1313–1318, https://doi.org/10.1289/ehp.1306615.
K. Raghavachari, A.P. Rendell, J.C. Burant, S.S. Iyengar, J. Tomasi, M. Cossi, J.M.
[10] R.A. Dickman, D.S. Aga, A review of recent studies on toxicity, sequestration, and
Millam, M. Klene, C. Adamo, R. Cammi, J.W. Ochterski, R.L. Martin, K. Morokuma,
degradation of per- and polyfluoroalkyl substances (PFAS), J. Hazard. Mater. 436
O. Farkas, J.B. Foresman, D.J. Fox, G16_C01. (ed Inc. Gaussian), Wallingford, CT,
(2022), https://doi.org/10.1016/j.jhazmat.2022.129120.
USA, 2016.
[11] S.C.E. Leung, et al., Emerging technologies for PFOS/PFOA degradation and
[30] W. Humphrey, A. Dalke, K.V.M.D. Schulten, Visual molecular dynamics, J. Mol.
removal: a review, Sci. Total Environ. 827 (2022), https://doi.org/10.1016/j.
Graph. 14 (1996) 33–38.
scitotenv.2022.153669.
[31] T. Lu, F. Chen, Multiwfn: a multifunctional wavefunction analyzer, J. Comp. Chem
[12] J.D. McNamara, R. Franco, R. Mimna, L. Zappa, Comparison of activated carbons
33 (2012) 580–592.
for removal of perfluorinated compounds from drinking water, J. Am. Water Works
[32] T. Steiner, The hydrogen bond in the solid state, Angew. Chem.-Int. Ed. 41 (2002)
Assn. 110 (2018) E2–E14, https://doi.org/10.5942/jawwa.2018.110.0003.
48–76, https://doi.org/10.1002/1521-3773(20020104)41:1<48::Aid-anie48>3.0.
[13] S. Khulu, et al., Development and application of a membrane assisted solvent
Co;2-u.
extraction-molecularly imprinted polymer based passive sampler for monitoring of
[33] D. Shetty, et al., Rapid and efficient removal of perfluorooctanoic acid from water
selected pharmaceuticals in surface water, Water Res. 225 (2022) 119145, https://
with fluorine-rich calixarene-based porous polymers, ACS Appl. Mater. Interfaces
doi.org/10.1016/j.watres.2022.119145.
12 (2020) 43160–43166, https://doi.org/10.1021/acsami.0c13400.
[14] S.-J. Choi, et al., Ionic fluid as a novel cleaning agent for the control of irreversible
[34] Q. Yu, R. Zhang, S. Deng, J. Huang, G. Yu, Sorption of perfluorooctane sulfonate
fouling in reverse osmosis membrane processes, Water Res. 224 (2022), https://
and perfluorooctanoate on activated carbons and resin: kinetic and isotherm study,
doi.org/10.1016/j.watres.2022.119063.
Water Res. 43 (2009) 1150–1158, https://doi.org/10.1016/j.watres.2008.12.001.