Introduction
The widespread production and use of plastic products in modern society have led to their continuous release and eventual accumulation in the environment, causing detrimental human health effects. Over time, these plastics degrade into smaller particles through physical, chemical, and biological processes [1]. Plastic particles ranging in size from 1 μm to 5 mm are classified as microplastics (MPs) [2], while those between 1 nm and 1,000 nm are referred to as nanoplastics (NPs) [3]. These fragments can originate from either primary sources, such as micro- or nanoscale plastics intentionally manufactured for use in cosmetics or industrial materials [4], or from secondary sources of larger plastic items fragmented over time [5, 6].
MPs and NPs enter the aquatic environments via various pathways, including discharge from wastewater treatment plants and stormwater runoff, ultimately contaminating freshwater systems [7]. Commonly used commercial polymers, such as polystyrene (PS), polyethylene (PE), polyvinyl chloride (PVC), polypropylene (PP), and polyethylene terephthalate (PET), have been frequently detected in both untreated and treated waters [8]. The presence of these polymers is of increasing concern due to the potential health risks associated with accidental human ingestion and inhalation [9]. A global study detected an average of 325 particles per liter in 93% of bottled water samples, exposing the potential public health implications [10].
While MPs have been extensively studied in drinking water systems, NPs have received less attention, despite growing evidence suggesting they pose greater risks to both human health and ecosystems. NPs have a high surface-to-volume ratio, making them highly reactive and capable of adsorbing heavy metals and toxic organic pollutants [11, 12]. Furthermore, their nanoscale size facilitates their penetration into the biological systems and promotes bioaccumulation, provoking ecotoxicological problems [13, 14].
As with engineered nanoparticles, the environmental behavior of NPs in freshwater systems such as lakes and rivers can vary depending on water chemistry parameters, including pH, ionic strength, and the presence of naturally occurring inorganic colloids or natural organic matter (NOM). These factors influence the transport, fate, and surface characteristics of NPs, often leading to the formation of heteroaggregates [15, 16]. For instance, positively charged PS NPs have been shown to form stable heteroaggregates in the presence of Fe2O3 and NOM, with the stability of aggregates being highly dependent on the concentrations of these substances [17]. NOM enhances the stability of NPs in aquatic environments primarily through steric repulsion [18]. Another study has demonstrated that the aggregation of PET NPs is more pronounced in the presence of divalent cations (e.g., Ca2+ and Mg2+) compared to monovalent cations (e.g., Na+ and K+) [19].
As the amount of NPs released into water bodies continues to rise, it is essential to assess the effectiveness of existing treatment technologies to remove these particles. Commonly used techniques like sand filtration and the adsorption of activated carbon are widely known for their environmental compatibility in use and operational simplicity, which have been proven effective in removing various contaminants from water [20, 21]. Y. Zhang et al. reported that anthracite coal filtration achieved a high removal efficiency of up to 99% for particles larger than 100 nm, while conventional coagulation and sedimentation methods were much less effective [22]. Membrane-based filtration with pore sizes of 0.22 μm was shown to remove approximately 92% of the NPs [23]. Furthermore, Tong et al. demonstrated that the use of biochar modified with Fe3O4 increased the retention of plastic particles more filtered in porous media [24].
Activated carbon's strong adsorption capacity and vast surface area make it one of the most commonly used media in water treatment. It effectively removes a broad range of contaminants, including metal nanoparticles and hazardous organic compounds [25–27]. McGillicuddy et al. reported that milled activated carbon could capture 63.6% of silver nanoparticles, where its removal efficiency differed upon exposure time and initial adsorbate concentration [28]. Activated carbon has also been demonstrated to be an efficient adsorbent for removing engineered nanoparticles such as ZnO and CuO [29]. Studies have further reported that removal effectiveness is greatly impacted by the physicochemical properties of the type of nanoparticles and adsorbents, together with solution pH and ionic strength [30, 31].
Despite the promising research results on nanoparticle removal, the utilization of granular activated carbon (GAC) for NP adsorption has been relatively underexplored. Given the proven efficacy of activated carbon and growing concerns about NP contamination in potable water resources, this research focused on investigating the adsorption performance of palm-based GAC for removing amidine-functionalized PS NPs. Key operating parameters affecting adsorption capacity were determined, from which their primary adsorption mechanisms were elucidated. Subsequently, the NP removal efficiency in the hybrid coagulation-adsorption process was investigated by comparing their turbidity and total organic carbon (TOC) concentration. Additionally, nanoparticle tracking analysis (NTA) and Raman spectroscopy (RS) were also used to further verify the reduction of NPs. The results of this study provide valuable insight for water treatment facilities seeking to assess the suitability of their existing infrastructure for NP removal and to optimize treatment protocols accordingly.
Materials and Methods
Materials
Amidine-functionalized polystyrene (PS) latex nanospheres were purchased as a 4% w/v stock solution from Thermo Fisher Scientific (USA). According to the manufacturer`s specifications, the average particle diameter was approximately 93 nm, as determined by transmission electron microscopy (TEM). Ultrapure water was used to dilute the stock solution in order to prepare suspensions at 5, 10, 20, 30, and 40 mg/L concentrations. Before each experiment, the PS suspensions were sonicated for 15 min using a sonicator (POWER Sonic 610, Hwashin Tech, South Korea) to ensure even dispersion of the nanoparticles.
Granular activated carbon (GAC) and ferric chloride hexahydrate (FeCl3·6H2O) were obtained from Duksan Pure Chemicals Co., Ltd. (South Korea). GAC was thoroughly cleaned with ultrapure water to eliminate dust and impurities, and then dried in an oven at 60 ℃ for 12 h. A solution of ferric chloride was made by dissolving 5.0 g of FeCl3·6H2O in 1 L of ultrapure water, after which the pH was brought to 4.5 using 0.1 M hydrochloric acid (HCl, Biomedic, South Korea) and 0.1 M sodium hydroxide (NaOH) solution (Daejung Chemicals and Metals, South Korea) to prevent premature floc formation.
Characterization of PS NPs and GAC
The zeta potential and particle size distribution of PS NPs were measured using a zeta potential analyzer (ELS-Z, Otsuka Electronics Co., Ltd., Japan), operating in electrophoretic light scattering (ELS) and dynamic light scattering (DLS) modes, respectively. Zeta potential measurements were performed in phosphate buffer solution over a pH range of 3 to 11. To compare surface charge characteristics, the zeta potential of GAC was also measured under the same conditions.
The GAC's surface morphology, both before and after PS NP adsorption for 240 min, was analyzed using fieldemission scanning electron microscopy (FE-SEM, S-4300SE, Hitachi, Japan). Prior to imaging, GAC samples were oven-dried at 60 ℃ and sputter-coated with a thin layer of platinum. SEM observations were scanned at magnifications ranging from ×35 to ×40 k.
The distribution of pore sizes and specific surface area of GAC were determined by a high-performance gas adsorption analyzer (3Felx, Micrometrics, USA), for which nitrogen adsorption-desorption isotherms at -195.748 ℃ of liquid nitrogen temperature were used.
Batch adsorption tests
Tests of batch adsorption were carried out to look at the effects of GAC dosage (0.1, 0.2, 0.5, 1.0, and 2.0 g), PS NP concentration (5, 10, 20, 30, and 40 mg/L), and reaction time (10, 20, 30, 60, 120, 240, and 480 min) on the NP removal. In each test, 15 mL of PS NP suspension was prepared in a 100 mL glass beaker, and a pH adjustment of 7.0 was made using 0.1 M NaOH or 0.1 M HCl. After adding the designated amount of GAC, the mixture was stirred in an orbital shaker (SI-600R, Jeio Tech, Korea) at a constant temperature of 25 ℃ and a shaking speed of 100 RPM (revolutions per minute). Samples were withdrawn at predetermined time intervals and allowed to stand for 30 min to gravitationally separate the used adsorbent. The turbidity of supernatant was measured by a portable turbidity meter (2020e, Lamotte, USA) using the nephelometric method. Turbidity readings were converted to PS NP concentration using a pre-established calibration curve between them. Turbidity serves as a practical and indirect indicator of the residual PS NP concentration quantified in the supernatant [23, 26].
The adsorption capacity at time t (qt, mg/g) was measured using Eq. (1):
Where Ci and Ct are the initial and time-dependent PS NP concentrations (mg/L), respectively, V is the volume of the solution (L), and m is the mass of GAC used (g).
The removal efficiency (R, %) after a reaction time of 480 min was determined using Eq. (2):
Where Cf is the final PS NP concentration (mg/L) after 480 min.
To evaluate the behavior of adsorption of PS NPs, experimental data were fitted to kinetic and isotherm models. Curve fitting and correlation coefficient (R2) were obtained using Microsoft Excel. Results are reported as the mean ± standard deviation based on triplicate measurements to ensure statistical reliability.
Comparison of coagulation, adsorption, and combined processes
The removal efficiency of PS NPs was compared under three different treatment conditions: (A) coagulation, (B) adsorption, and (C) a combination of (A) and (B) processes. In the (A) process, 2 mL of 5.0 g/L ferric chloride solution was added to 50 mL of PS NP suspension (initial concentration = 20 mg/L). The suspension was rapidly stirred at 200 RPM for 2 min, followed by gently mixing at 80 RPM for 30 min to promote floc formation. Subsequently, it was then allowed to settle for 30 min, after which the supernatant was collected using a 5 mL pipette. The supernatant was kept at room temperature and measured immediately for turbidity analysis. For the adsorption test (B), 1 g of GAC was added to 50 mL of PS NP suspension (20 mg/L), which was then agitated for 240 min to ensure equilibrium of adsorption, after which it was settled down for 30 min. Although the batch adsorption test was conducted up to 480 min, more than 90% of adsorption was achieved within 240 min, and therefore, this duration was selected as a representative reaction time for subsequent experiments. The resulting supernatant was then collected using a 5 mL pipette. The combined test (C) was carried out sequentially. The coagulation has been done in the same manner as aforementioned. First, 2 mL of 5.0 g/L ferric chloride solution was mixed with 50 mL of PS NP suspension (for 2 min at 200 RPM fast mixing, at 80 RPM slow mixing for 30 min, followed by 30 min of settling). The supernatant was transferred to a separate vessel, which was then reacted with 1 g of GAC for 240 min. It was settled down for 30 min to separate the reactants and its supernatant was finally collected using a 5 mL pipette. The supernatant was kept at room temperature, and samples were analyzed to compare the removal efficiency of PS NP in terms of turbidity, total organic carbon (TOC), and nanoparticle concentration as described in the following.
Turbidity and TOC
The turbidity was measured immediately on the supernatant using a portable turbidity meter (2020e, Lamotte, USA). TOC was measured as non-purgeable organic carbon (NPOC) using a TOC-L analyzer (Shimadzu Corporation, Japan). Prior to TOC analysis, 0.45 µm syringe filters (Whatman, UK) were used to filter the samples to remove particulates and then acidified with 2 M HCl. The instrument was operated in NPOC mode, removing inorganic carbon and volatile organic carbon via high-purity synthetic air sparging (≥ 99.99%) with the sparge gas flow of 80 mL/min for 1 min 30 s. All measurements were carried out three times to ensure reliable accuracy and reproducibility.
Quantitative analysis of NPs: Raman spectroscopy and NTA
Quantitative analysis of nanoparticles on samples was performed using nanolens effect-enhanced Raman spectroscopy (RS) and nanoparticle tracking analysis (NTA). For RS, 50 µL of each sample was drop-cast onto a silicone (Si) substrate. To minimize the coffee ring effect and to ensure uniform particle distribution, the droplets were freeze-dried at -50 ℃ under a vacuum pressure of 0.003 mbar. Simultaneous quantitative and qualitative analyses were performed using an automated Raman measurement system (Nano Raman Analysis, Incheon, South Korea). Bright-field vision images were captured prior to spectral analysis to determine the particle sizes. Raman spectra were acquired using a 532-nm laser at a power of 4.2 mW, under an exposure time of 1,000 ms per particle and single acquisition mode. Along with this, they were quantified by NTA using a ZetaView instrument (Particle Matrix, Germany) equipped with a 488-nm laser. The target particle size range was set between 0 and 912.736 nm. All acquired data were processed to determine their size using Software ZetaView (version 8.05.16 SP3).
Results and Discussion
Surface characterization of PS NPs and GAC
The surface charge of PS NPs and GAC was characterized by measuring zeta potential across the pH range of 3 to 11 (Fig. 1(a)). PS NPs exhibited a highly positive zeta potential of +35.3 ± 0.2 mV at pH 3.0, indicating strong electrostatic repulsion among particles. At pH 7.0, it was slightly positively charged (+5.3 ± 0.2 mV), but beyond that, the charge shifted to negative, reaching -21.4 ± 0.6 mV at pH 9.0 and -28.7 ± 2.2 mV at pH 11.0. In contrast, GAC exhibited a weakly positive zeta potential of +0.8 ± 0.2 mV at pH 3.0, which became increasingly negative with increasing pH. The zeta potential at pH 7.0, 9.0, and 11.0 was observed to be -13.5 ± 3.5 mV, -18.2 ± 1.2 mV, and -28.7 ± 2.2 mV, respectively. This is explained by the deprotonation of functional groups on the GAC surface that contain oxygen, like carboxyl and hydroxyl groups [32]. The opposite surface charges of GAC and PS NPs at neutral pH suggest that electrostatic attraction may enhance adsorption. Therefore, all subsequent adsorption tests were conducted at pH 7.0.
The particle size distribution of PS NPs, measured by dynamic light scattering (DLS), showed a log-normal distribution within the range of 88.0 nm to 887.7 nm (Fig. 1(b)). The 10th, 50th, and 90th percentile diameters (D10, D50, and D90) were determined to be 147.4 nm, 293.1 nm, and 553.2 nm, respectively. The low polydispersity index (PDI) of 0.132 confirmed the stable and uniform nanoparticle dispersion in the suspension [33].
SEM images revealed that GAC had a rough and highly porous surface structure (Fig. 2(a, b)). The Brunauer-Emmett-Teller (BET) analysis revealed an average pore diameter of 31.01 Å and a high specific surface area of 882.82 m2/g, with pore sizes ranging approximately from 15 to 200 Å, indicating a well-developed mesoporous structure favorable for NP adsorption. After adsorption, SEM images (Fig. 2(c, d)) showed spherical PS NPs adhered to the external surface and inner pores of GAC, suggesting that adsorption took place in both surface interaction and pore-filling processes [34].
Batch adsorption studies
Effect of GAC dose, PS NP concentration, and reaction time
The impact of GAC dosage on PS NP removal effectiveness was examined for PS NP concentration of 20 mg/L upon different GAC additions. Different amounts of GAC (0.1, 0.5, 1, and 2 g) were added to 15 mL of PS NP suspension and shaken for 480 min. As the GAC dosage increased, turbidity in the supernatants decreased from an initial 72.7 NTU to 67.5, 55.6, 48.4, and 23.9 NTU, equivalent to residual PS NP concentrations of 15.9, 12.8, 11.0, and 4.7 mg/L, respectively. The removal efficiency increased proportionally with the GAC dosage, reaching a maximum of 72.7% under the test conditions for 2 g.
The adsorption capacity of GAC removing the PS NP upon reaction time is shown in Fig. 3. Adsorption was conducted at 1 g of GAC at pH 7.0, with varying PS NP concentrations (5, 10, 20, 30, and 40 mg/L) upon different reaction times of 10, 20, 30, 60, 120, 240, and 480 min. As shown in Fig. 3(a), the GAC adsorption capacity was increased rapidly within the first 120 min at all concentrations, thereafter gradually approached an equilibrium. The maximum adsorption capacity, 0.308 ± 0.187 mg/g, was observed at 40 mg/L (Table 1). Nonetheless, after 480 min, their removal efficiencies were 85.6 ± 35.8%, 62.6 ± 14.9%, 66.8 ± 26.2%, 52.5 ± 11.1%, and 34.0 ± 17.2% for correspondingly 5, 10, 20, 30, and 40 mg/L. Fig. 3 (b) shows that dosing 5 mg/L gave the highest, even though 40 mg/L dosing had the greatest adsorption capacity. The decrease in removal efficiency for higher concentrations may result from increased intra-particle interactions and exhausted available adsorption sites on the GAC. It simply implies that the removal efficiency of PS NPs onto GAC is highly dependent on both adsorbate concentration and reaction time [26].
Adsorption kinetics
To elucidate the adsorption mechanism of PS NPs onto GAC, three kinetic models were compared: the pseudo-first-order model, the pseudo-second-order model, and the intra-particle diffusion model [26, 35]. The corresponding kinetic equations are as follows:
Pseudo-first-order model:
Pseudo-second-order model:
Intra-particle diffusion model:
Here, the adsorption capacity time t is denoted by qt (mg/g), the equilibrium adsorption capacity by qe (mg/g), and the rate constants of the pseudo-first-order and pseudo-second-order models are denoted by k1 (min-1) and k2 (g/mg-min). kd (mg/g-min1/2) is the intra-particle diffusion rate constant, and C (mg/g) is a constant related to the boundary layer effect.
The kinetic parameters were determined by adjusting experimental results to the respective models. The pseudo-first-order model was fitted using the linear plot of log(qe−qt) versus t (Fig. 4(a)), while the pseudo-second-order model was simulated with a plot of t/qt versus t (Fig. 4(b)). The intra-particle diffusion model parameters kd and C were obtained from the slope and intercept of the plot of qt versus t1/2 (Fig. 4(c)).
Among the three different kinetic models evaluated, the experimental data were best fitted by the pseudo-second-order kinetic model, as evidenced by high correlation coefficients (R2) ranging from 0.991 to 0.999 across all PS NP concentrations (Table 1). In contrast, the pseudo-first-order kinetic model yielded lower R2 values (0.874-0.982). These results suggest that chemisorption could be the main rate-controlling mechanism, involving electron sharing or exchange between the GAC surface and the PS NPs [36, 37]. Several minor interactions also contribute to the overall adsorption process. Specifically, under the experimental pH conditions, electrostatic attraction facilitated the rapid approach of PS NPs toward the GAC surface due to their opposite surface charges (Fig. 1). This was further stabilized by π-π interactions between the aromatic benzene rings of PS NPs and the GAC surface [37], hydrophobic forces, and physical trapping within the micropores [38].
The adsorption process typically comprises three sequential mass transfer steps [39]. The first action is to move the adsorbate molecules from the bulk solution to the adsorbent's outer surface (external or film diffusion). The second step entails diffusion into the internal pore structure (intra-particle diffusion). Finally, adsorption occurs on both the internal and external surfaces. To assess whether intraparticle diffusion was the sole rate-limiting step, the Weber and Morris model was applied [40]. As shown in Fig. 4(c), the plot of qt versus t1/2 exhibits two distinct linear phases, indicating that intraparticle diffusion is not the exclusive rate-controlling step. The initial steeper slope corresponds to the external mass transfer stage, while the second linear region is associated with the subsequent intra-particle diffusion into the GAC pores [41]. The corresponding intraparticle diffusion rate constants kd1 and kd2, derived from the slopes of the respective regions, are summarized in Table 1. In all cases, the kd1 exceeded kd2, indicating that faster initial external diffusion occurred, followed by slower diffusion into the internal pores. Consequently, these findings suggested that the initial adsorption stage might be governed by surface interactions, while intra-particle diffusion becomes the dominant rate-limiting step in the later stages of PS NP adsorption onto GAC.
Adsorption isotherms
The Langmuir and Freundlich isotherm models were simulated for the adsorption equilibrium of PS NPs on GAC. These models are expressed by the following Eqs. (6) and (7).
Langmuir isotherm:
Freundlich isotherm:
Where RL is the dimensionless separation factor (0 < RL < 1 indicating favorable adsorption; RL = 0: irreversible; RL = 1: linear; RL > 1: unfavorable), Ce is the equilibrium concentration of PS NPs (mg/L), kL is the Langmuir constant associated with adsorption energy (L/mg), and qe and qm are the equilibrium and maximum monolayer adsorption capacities (mg/g). [42]. kF is the Freundlich adsorption capacity constant ((mg/g) (L/mg)¹/ⁿ), and the adsorption intensity parameter is denoted by n.
Monolayer adsorption on a uniform surface with a finite number of identical binding sites is assumed by the Langmuir isotherm model, while heterogeneous surface adsorption using multilayer is represented by the Freundlich isotherm model. As shown in Fig. 5, both models described the experimental data fairly well. However, in contrast to the Freundlich model (R2 = 0.927), the Langmuir model displayed a better correlation coefficient (R2 = 0.985), suggesting that monolayer adsorption was dominant under the given experimental conditions. The Langmuir model yielded a maximum adsorption capacity (qm) of 0.370 mg/g and a constant (kL) of 0.198 L/mg. The calculated separation factor (RL) was all less than 1, which was decreased with increasing PS NP concentration, suggesting that adsorption became increasingly escalated and even less reversible at higher concentrations. For the Freundlich model, the constants were determined to be kF = 0.078 (mg/g)(L/mg)1/n and n = 2.151, which also indicates favorable adsorption behavior possessed (since 1 < n < 10) [43, 44]. Overall, the Langmuir isotherm better describes the equilibrium adsorption of PS NPs onto GAC, supporting the conclusion that the process could be primarily governed by monolayer adsorption on a relatively homogenous surface.
Combined coagulation-adsorption for PS NP removal
Coagulation dose optimization
Before combining the coagulation-adsorption process for PS NP removal, a series of vessel tests using varying coagulant concentrations (1.0-9.0 g/L) were conducted to determine the optimal dosage of coagulant. The pH of each solution was adjusted to 4.5 with 0.1 M NaOH or HCl. For each test, 2 mL of the coagulant was added to 50 mL of PS suspension (20 mg/L), followed by rapid mixing (200-300 RPM, 2-3 min), slow mixing (80 RPM, 30 min), and settling for 30 min.
At low concentrations (1.0-2.0 g/L), the turbidity of supernatant decreased from an initial 55.5 NTU to 48.7 and 40.2 NTU, respectively, with weak or nearly invisible floc formation. At 3.0 g/L, small, fragile flocs began to appear. The best performance was achieved at 5.0 g/L, with a significant reduction in turbidity to 35.5 NTU and the formation of dense, stable flocs, indicating effective destabilization and flocculation of PS NPs. When the coagulant dosage was further increased to 7.0-9.0 g/L, the turbidity of the supernatant increased again to 48.4-55.5 NTU, and floc formation diminished. This was attributed to charge inversion and particle-to-particle restabilization under overdosage conditions [45, 46]. Therefore, 5.0 g/L of coagulant (corresponding to a final concentration of 200 mg/L in reaction solution) was identified as the optimal concentration that provided a balance between turbidity reduction and stable floc development. The relatively high dose is consistent with the strong surface charge and colloidal stability of ~90 nm PS NPs, which require a larger amount of Fe3+ hydrolyzed species for charge neutralization and aggregation [45]. Furthermore, the small-volume batch system (50 mL) using concentrated coagulant stocks ensured reproducible dose dosing and measurable analytical precision. Previous studies have reported that the optimal Fe-based doses for PS NPs are 5-20 mg/L under favorable conditions [45]. Thus, further fine-tuning of the coagulant dose in the lower mg/L range is recommended for future work.
Reduction of turbidity and TOC
The removal efficiency of PS NPs was observed using coagulation, adsorption, and their combined process, with the PS NP concentration of 20 mg/L. In the coagulation-sole process, turbidity was decreased from 49.1 ± 1.5 NTU to 45.8 ± 0.4 NTU, resulting in a removal efficiency of 6.6 ± 2.3% (Fig. 6). This lower limited performance suggests that coagulation alone was not sufficient for removing nano-sized PS particles. This poor performance could be attributed to several factors, including the tiny particle size of PS NPs not being agreeably coagulated, high colloidal stability, incomplete charge neutralization, and limited interparticle bridging. Although coagulation alone is inefficient, it can still function as an effective pretreatment step by destabilizing the colloidal suspension. The addition of a coagulant reduces zeta potentials of PS NPs, thereby mitigating the electrostatic repulsion between PS NPs and the GAC surface, which subsequently enhances the attachment probability of the particles onto the adsorbent [47]. In parallel, the adsorption-only process using a palm-derived GAC achieved a significantly higher removal efficiency of 66.7 ± 4.4% (Fig. 6). When coagulation and adsorption were combined, the removal efficiency further increased to 90.4 ± 3.3%, with the final turbidity reduced to 4.7 ± 1.4 NTU (Fig. 6). This significant improvement demonstrates the synergistic effect between the two distinct mechanisms. The initial coagulation promotes the formation of micro-flocs and destabilized aggregates, which not only accelerates mass transfer toward the GAC surface but also prevents severe pore clogging by macro-sized particles. By retaining the availability of GAC micropores for subsequent deeper diffusion, the combined process maximizes the utilization of the adsorbent`s capacity [48]. It strongly supports the use of a combined coagulation-adsorption approach as a promising and effective strategy for removing PS NPs from aqueous environments.
The TOC analysis further confirmed the turbidity reduction trend, as shown in Fig. 6. The coagulation-sole process showed a low TOC removal efficiency of 12.8 ± 9.8%, suggesting a limited removal capacity for dissolved organic matter [49]. In contrast, adsorption using GAC significantly reduced TOC, achieving a removal efficiency of 81.2 ± 4.6%. The combination of them showed the highest performance with a TOC removal efficiency of 92.0 ± 3.5%, emphasizing the enhanced effectiveness resulting from the integration of the two treatment systems.
Quantitative analysis of NPs
The NTA technique uses a laser to illuminate freely diffusing particles and track their Brownian motion, allowing for the number of particles counted and size determined [50]. However, as a purely quantitative method, NTA does not provide information about the chemical species, making it difficult to confirm the identity of detected unknown particles. To overcome this limitation, RS was employed in parallel. RS is a non-destructive technique that can identify individual particles based on their unique molecular fingerprint, even in aqueous media. However, conventional RS methods still struggle to keep up with this hurdle to detect nanoscale particles due to the inherently lower scattering intensity of the cross-section of nanoparticles. Overcoming this barrier, we have developed an advanced Raman analysis incorporating the nanolens effect, which significantly improved the detection limit beyond conventional RS (typically ~1 µm). This approach enabled both qualitative and quantitative analysis of NPs in complex matrices to be simultaneously conducted for plastic recycling wastewater and bottled water [51].
Fig. 7 shows the particle concentrations (per mL) determined by NTA and RS. NTA measured the total particle concentration regardless of composition, such as PS, chemical flocs, etc., while RS selectively quantifies PS particles. NTA results showed that both the adsorption alone and the combined processes removed more than 90% of the total particles. A similar trend was observed in the RS data as of NTA. The removal efficiency for coagulation, adsorption, and the combined process in NTA was 30.0%, 98.1%, and 99.4%, equivalent to -131.7%, 87.6%, and 100.0% for RS, respectively. The negative removal efficiency observed for coagulation in RS reflects an apparent increase in detectable PS particles compared to the untreated sample. This artifact arises from limitations in the image-based analysis using optical microscopy and OpenCV, which cannot detect particles smaller than 250 nm. Many PS NPs (approximately 90 nm) in the untreated sample were initially undetectable. However, aggregation promoted particle size increase, thereby enhancing RS detectability.
Limitations and future perspectives
In this study, the adsorption mechanisms of PS NPs were elucidated, and the removal efficiency of a coagulation-adsorption hybrid process was investigated. Despite being conducted under the controlled laboratory conditions using a fixed coagulant dosage, neutral pH, and ultrapure water, the results demonstrate the potential of integrating coagulation with adsorption for the effective removal of PS NPs. However, several limitations constrain the broader applicability. Future studies investigating the effects of types of coagulants and their variable dosages, a wider pH range, and more complex water chemistries are needed to optimize and validate this approach for practical applications. In particular, performance evaluations in real-world water matrices containing NOM, multivalent ions, and competing pollutants will be critical to assessing its robustness and scalability for drinking water treatment applications.
Conclusions
This study demonstrated that a hybrid treatment approach combining coagulation and GAC adsorption is highly effective in removing PS NPs from water. SEM visually confirmed successful adsorption, revealing that PS NPs were embedded on and within the GAC surface. According to kinetic modeling, the adsorption process followed a pseudo-second-order model, indicating that the predominant mechanism is chemisorption. An intraparticle diffusion model further supported a multistep mass transfer, wherein PS NPs rapidly transport to the GAC surface and then diffuse into the internal pore structure as the surface sites become saturated. Equilibrium data were best described by the Langmuir isotherm model, indicating monolayer adsorption onto a homogenous surface. Meanwhile, the hybrid coagulation-adsorption process achieved a removal efficiency of up to 99.4% for PS NPs, significantly superior to coagulation alone (30%), as verified by NTA analysis. Reductions in turbidity, TOC, and particle counts further corroborated the effectiveness of the combined treatment. It supports the hybrid approach as a promising and scalable strategy for NP removal in water treatment systems. Future research should focus on optimizing the coagulant type and dosage, evaluating system performance under various water quality conditions to further enhance removal efficiency.
PDF Links
PubReader
XML Download
Full text via DOI
Download Citation 
Print 
