Advances in cold atmospheric plasma for water purification: reactive species generation, diagnostics, and degradation of persistent organic contaminants

1 Introduction

Persistent organic contaminants (POCs), including pharmaceuticals, personal care products, pesticides, industrial chemicals, dyes, and per- and polyfluoroalkyl substances (PFAS), have emerged as a critical challenge for water security due to their widespread occurrence and persistence in aquatic environments. These chemically stable compounds frequently detected in surface water, groundwater, and even drinking water systems (Bayabil et al., 2022; Marotta and Paradisi, 2025). Continuously released into aquatic systems through industrial discharges, agricultural runoff, landfill leachate, and municipal wastewater effluents, POCs resist natural degradation, enabling their long-term persistence and accumulation in ecosystems and living organisms (Huaccallo-Aguilar et al., 2024; Kumar et al., 2026a; Li et al., 2026). Their persistence, toxicity, and bioaccumulation potential, raise increasing concerns for environmental sustainability and human health, driving the urgent need for innovative and effective treatment technologies capable of removing these challenging contaminants.

Conventional wastewater treatment plants (WWTPs), which mainly rely on biological processes such as activated sludge, are not capable of efficiently remove many of these contaminants (Arslan et al., 2017; Mukhopadhyay et al., 2022). Therefore, complementary advanced treatment strategies are required to overcome these limitations and achieve effective removal and degradation of recalcitrant contaminants (Kumar et al., 2026a; Tian et al., 2024; Massima Mouele et al., 2021). In Europe, the regulatory framework for water quality is evolving to address POCs. The recast Urban Wastewater Treatment Directive (UWWTD; Directive (EU) 2024/3019) introduces stricter requirements for large urban WWTPs, including quaternary treatment to achieve at least 80% removal of selected micropollutants by 2045, with earlier interim targets. In parallel, PFAS are receiving increasing regulatory attention through updated EU priority and watch lists and specific requirements under related legislation, including the Drinking Water Directive (DWD; Directive (EU) 2020/2184), which establishes monitoring requirements and limit values for PFAS in drinking water (Tasselli et al., 2025; Gautier, 2026).

Cold atmospheric plasma (CAP) has emerged as a versatile technology for environmental remediation, with applications spanning medicine, agriculture, and water treatment (Figure 1a) (Jablonowski and Von Woedtke, 2015; Laroussi, 2020; Kumar et al., 2021; Rao et al., 2021; Barjasteh et al., 2023; Karimian et al., 2025). In water treatment, CAP combines the characteristics of AOPs and ARPs, enabling the degradation and mineralization of a wide range of POCs (Figure 1b) (Kumar et al., 2023; Lamichhane et al., 2025; Mbanugo et al., 2025). CAP combines oxidative and reductive reaction pathways by generating a broad spectrum of ultrashort-lived, short-lived, and long-lived reactive species through CAP-liquid interactions (Figure 1c) (Foster, 2017; Kumar et al., 2021; 2025; Yang et al., 2023). These species include RONS, such as •OH, O•, O₃, H₂O₂, NO₂ ⁻, NO₃ ⁻, ONOOH, and e ^- _sol, among others (Gopalakrishnan et al., 2016; Liu et al., 2019; Kumar et al., 2022; Li et al., 2026). The combined action of these short- and long-lived species enables the transformation of diverse POCs, including pharmaceuticals, dyes, pesticides, industrial chemicals, and highly persistent PFAS (Kumar et al., 2021; 2025; 2026b; Stapelmann et al., 2024). However, complete mineralization, defluorination, and detoxification remain challenging and depend strongly on pollutant structure, reactor configuration, reactive-species generation, interfacial transport, and water matrix composition.

Recent advances in CAP-source engineering, discharge control, gas-liquid mass transfer enhancement, and reactor design have significantly improved the efficiency of CAP-based pollutant treatment. These developments have expanded the application of CAP to PFAS degradation, where CAP-generated reducing species, particularly e ^- _sol, promote cleavage of strong C–F bonds (Lewis et al., 2020; Mbanugo et al., 2025; Ganzallo et al., 2026; Kumar et al., 2026b). However, the mechanisms governing reactive-species generation, transport, and utilization at the CAP-liquid interface remain insufficiently understood, particularly under realistic water matrices. Recent advances in plasma diagnostics, multiphysics modeling, and machine-learning approaches have improved mechanistic understanding by correlating discharge characteristics, CAP-liquid interactions, reactive-species generation, and treatment performance. Building on these developments, this mini-review presents an integrated perspective by connecting discharge physics, CAP-liquid interfacial processes, reactive-species diagnostics, pollutant transformation mechanisms, reactor engineering, and computational optimization, providing insights for the rational design and scale-up of efficient CAP-based water purification technologies.

The literature discussed in this mini-review was selected through a structured survey of studies related to CAP-based water purification, CAP-liquid interactions, reactive-species generation, pollutant transformation mechanisms, diagnostics, and process optimization. Searches were conducted using Web of Science, Scopus, and Google Scholar with keywords including “cold atmospheric plasma”, “plasma water treatment”, “plasma-liquid interaction”, “plasma for reactive species generation”, “organic pollutant degradation”, “plasma diagnostics”, “plasma multiphysics modeling”, and “machine learning for plasma”. Peer-reviewed studies published mainly between 2015 and 2026 were considered, with earlier studies included to provide fundamental insights into CAP-liquid chemistry. Selected studies focused on reactive-species generation and characterization, spectroscopic diagnostics, pollutant transformation pathways, reactor design, energy performance, and computational approaches for process optimization.

2 Reactive species generation in CAP-liquid systems for POCs treatment

Recent advances in CAP reactive-species generation have been achieved through improvements in power supply technology, electrode design, reactor engineering, and gas-liquid flow management (Kumar et al., 2021; 2025; Mededovic Thagard, 2025; Ganzallo et al., 2026; Li et al., 2026). These developments enable better control of electron energy distribution and enhance plasma non-equilibrium, leading to more efficient electron-impact reactions at the gas-liquid interface. As a result, CAP systems can more selectively generate highly reactive short-lived species such as •OH, O•, •H, and e ^- _sol. However, the treatment efficiency depends not only on reactive-species generation but also on their transport across the CAP-liquid interface, lifetime in water, and interaction with specific pollutant structures.

CAP can be generated by various non-thermal discharge systems, including plasma jets, dielectric barrier discharges (DBD), corona discharges, gliding arcs, and radio-frequency plasmas, operated under different applied voltage waveforms such as DC, AC, radio frequency, microwave, or pulsed discharge condition (Kumar et al., 2021; Shahsavari and Zhang, 2023; Mbanugo et al., 2025; Papalexopoulou et al., 2026). The discharge configuration, applied power, and feed gas composition (air, Argon, O₂, N₂, or their mixtures) strongly govern plasma behavior through the reduced electric field (E/N: E: electric field strength and N: gas number density) and electron energy distribution function (EEDF), which determine electron-impact excitation, ionization, and dissociation reactions responsible for RONS formation (Lin and Keidar, 2021; Mladenović and Gocić, 2022).

In CAP-liquid treatment systems, CAP reactors can generally be classified into three major configurations: (i) plasma above the liquid surface, (ii) plasma confined within gas bubbles or at bubble-liquid interfaces, and (iii) direct plasma discharge generated inside the liquid phase (shown in Figure 1d). Representative reactor configurations and their applications for POCs treatment are summarized in Table 1. Among these approaches, gas-phase plasma above liquids and bubble-based plasma systems are the most widely investigated because they provide relatively stable operation, easier plasma ignition, and lower energy requirements compared with direct liquid discharges. The characteristics of the generated plasma, including electron energy, reactive-species density, and gas temperature, are strongly dependent on whether the discharge occurs in the gas phase or directly within the aqueous phase (Foster, 2017; Vanraes and Bogaerts, 2018; Kumar et al., 2021).

TABLE 1. Common CAP sources with various configurations and discharge types used for the treatment of POCs. CAP systems are generated using different power supplies and working gases.

2.1 CAP above liquid surface systems

In plasma above liquid systems, the discharge is sustained in the gas phase while the liquid acts as a secondary electrode or dielectric barrier. This configuration has been extensively studied because of its simple design, operational flexibility, and effective CAP-liquid interaction (Bruggeman et al., 2016; Kumar et al., 2021; Mededovic Thagard, 2025). Reactive species are generated in the gas phase and transferred across the gas-liquid interface, where short-lived radicals react near the interface, whereas long-lived species dissolve into the bulk liquid and drive secondary oxidation processes.

Common reactor designs include pin-to-liquid, DBD above water, wire-plate systems, and plasma jets impinging on liquid surfaces (shown in Table 1) (Foster, 2017; Kumar et al., 2021; 2026b; Mededovic Thagard, 2025). The efficiency of these systems is strongly controlled by interfacial area, mass-transfer resistance, and hydrodynamic conditions. Recent reactor developments, including falling films, thin liquid layers, sprays, aerosol systems, and continuous-flow configurations, have increased CAP-liquid contact area, reduced diffusion limitations, and improved reactive-species transfer into water (Ren et al., 2021; Eeso et al., 2023; Olson et al., 2025). These improvements are particularly important for surface-active contaminants such as long-chain PFAS, which preferentially accumulate at the air-water interface and can therefore interact more effectively with short-lived interfacial radicals (Kumar et al., 2026b).

2.2 Bubble-based CAP systems

Bubble plasma reactors represent an advanced approach for enhancing CAP-liquid contact by generating discharge inside gas bubbles or at the bubble-liquid interface (Kumar et al., 2021; Shahsavari and Zhang, 2023; Papalexopoulou et al., 2026). Owing to the strong dielectric contrast between the gas and liquid phases, the electric field becomes highly concentrated within the bubble, resulting in elevated local reduced electric fields (E/N), energetic electrons, and highly non-equilibrium plasma chemistry. Under these conditions, streamer-like or pulsed microdischarges form inside the bubbles, producing reactive radicals and excited species in close proximity to the gas-liquid interface, thereby enhancing their transfer and mixing into the surrounding liquid phase (Kumar et al., 2026b; Li et al., 2026).

Microbubble-based CAP systems provide enhanced reactive-species production because of their large interfacial area and improved gas-liquid mass transfer. These reactors promote the formation and delivery of oxidants such as •OH, O•, and H₂O₂ through intensified interfacial reactions (Shahsavari and Zhang, 2023; Panchal et al., 2025; Kumar et al., 2026b). The performance of bubble plasma systems is strongly influenced by operational and design parameters, including bubble size, gas composition, pulse width of the power supply, residence time, superficial gas velocity, electrode configuration, and hydrodynamic conditions (Alam et al., 2024; Panchal et al., 2025). Smaller bubbles generally provide higher surface-area-to-volume ratios and longer residence times, improving reactive-species transport and pollutant interactions. Therefore, bubble-based CAP systems show particular promise for treating dissolved organic contaminants, including short-chain PFAS, which are more difficult to remove using conventional interfacial approaches (Mbanugo et al., 2025; Mededovic Thagard, 2025; Kumar et al., 2026b).

2.3 Direct liquid-phase CAP discharge

In direct liquid-phase discharge systems, plasma is generated between submerged electrodes directly within water. Reactive species are primarily formed through electron-impact dissociation of water molecules, leading to the production of •OH, H•, and other short-lived intermediates. Although this configuration can achieve very high local radical densities, it requires significantly higher breakdown voltages and energy input compared to gas-phase discharges (approximately ∼1 MV/cm in liquids versus ∼30 kV/cm in the gas phase) (Foster, 2017). Consequently, direct liquid discharges remain less widely implemented for POCs treatment because of higher energy requirements and engineering challenges (Foster, 2017; Kumar et al., 2021).

Overall, the choice of CAP reactor configuration determines not only the generation of reactive species but also their transport, lifetime, spatial distribution, and accessibility to target pollutants. These factors collectively govern CAP-liquid interactions, reaction pathways, pollutant degradation efficiency, energy consumption, and overall treatment performance. Therefore, future reactor optimization should focus on maximizing the effective generation, transport, and utilization of reactive species at the CAP-liquid interface, while improving mass transfer and energy efficiency, rather than simply increasing plasma power. Such an approach will be essential for developing efficient, scalable, and economically viable CAP systems for practical water treatment applications.

3 Characterization and ultrafast spectroscopy of CAP generated reactive species

Reliable characterization of reactive species is essential for establishing quantitative relationships between plasma discharge characteristics, CAP-liquid interfacial transport, and pollutant degradation. Because CAP generates numerous oxidizing and reducing species with lifetimes spanning from nanoseconds to hours, no single analytical technique can simultaneously detect and quantify all relevant species. Consequently, complementary diagnostic methods are required to characterize excited gas-phase species, ground-state radicals, long-lived aqueous species, and ultrafast interfacial processes. Recent advances in diagnostic capabilities have provided deeper insights into reactive-species generation, transport, and transformation mechanisms. In the following sections (Table 2), the main analytical approaches are discussed according to the location and lifetime of the reactive species, including gas-phase species, aqueous-phase species, and CAP-liquid interfacial processes.

TABLE 2. Common diagnostic techniques for the detection and characterization of reactive species generated during CAP-liquid interactions and liquid treatment.

3.1 Measurement of excited and ground-state gas-phase species

Optical emission spectroscopy (OES) remains the most widely used diagnostic technique for atmospheric-pressure plasmas because it is rapid, non-intrusive, and enables real-time monitoring of plasma chemistry. OES detects light emitted by electronically excited species, thereby providing valuable information on discharge stability, excitation pathways, and spatial plasma structure. For example, OES spectra (Figure 2a) of argon CAP generated over a liquid surface revealed emissions from hydroxyl radicals (•OH, 309 nm), atomic hydrogen (H•, 656.2 nm), atomic oxygen (O•, 777.5 and 844.6 nm), and nitrogen species including the N₂ second positive system (SPS) and N₂ ⁺ first negative system (FNS), confirming active plasma chemistry (Kumar et al., 2022). However, OES emission intensity should not be interpreted as the absolute concentration of reactive radicals. Spatially resolved OES combined with intensified charge-coupled device (ICCD) imaging has enabled visualization of streamer propagation and reactive-species distributions near the CAP-liquid interface, providing insight into localized reaction zones and plasma transport processes (Figures 2a,b) (Kumar et al., 2023).

Many chemically important plasma species exist predominantly in their ground electronic states and therefore cannot be detected by conventional OES. Laser-induced fluorescence (LIF) and two-photon absorption laser-induced fluorescence (TALIF) provide highly selective and sensitive measurements of ground-state •OH, O•, and NO radicals with high spatial and temporal resolution (Yang et al., 2023). Recent studies have demonstrated micrometer-scale mapping of radical densities within CAP-liquid interfacial regions. Long-lived neutral species such as ozone, NOx, and stable molecular products are commonly quantified using optical absorption spectroscopy (OAS), Fourier-transform infrared spectroscopy (FTIR), and mass spectrometry. These techniques complement OES by directly measuring stable species concentrations or molecular compositions. OAS enables the qualitative and quantitative detection of reactive species through wavelength-specific absorption and is widely used for real-time monitoring of plasma-activated environments (Simoncelli et al., 2019; Sobota et al., 2025; Kumar et al., 2026b). FTIR spectroscopy and mass spectrometry, allow identification of long-lived species and plasma-generated ions (Chen and Liu, 2021; Ibba et al., 2024; Sobota et al., 2025).

3.2 Measurement of aqueous reactive species

Direct measurement of aqueous reactive species remains considerably more challenging than gas-phase diagnostics because radicals undergo extremely rapid reactions, diffusion, recombination, and interconversion within the liquid phase. Consequently, most liquid-phase analyses rely on chemical probes, electron paramagnetic resonance (EPR) spin trapping, electrochemical methods, or spectrophotometric assays.

Hydroxyl radicals are commonly detected using terephthalic acid, coumarin, or DMPO spin trapping, whereas superoxide (O₂•^-), singlet oxygen (¹O₂), H•, and O• are measured using selective EPR probes. Solvated electrons are often quantified indirectly through chloride release from chloroacetic acid or related scavenger reactions. Long-lived oxidants including H₂O₂, O₃, NO₂ ⁻, and NO₃ ⁻ are typically determined using UV-visible colorimetric assays. However, UV-visible measurements require careful interpretation because absorption spectra of nitrate, nitrite, hydrogen peroxide, and dissolved ozone partially overlap, particularly in the ultraviolet region. Consequently, matrix composition, pH, dissolved organic matter, and coexisting oxidants may interfere with quantitative measurements unless appropriate calibration procedures or complementary analytical techniques are employed. Similarly, H₂O₂ should not be considered a direct marker of •OH concentration. Although H₂O₂ is frequently produced through •OH radical recombination, its final concentration also depends on secondary plasma chemistry, solution composition, scavenging reactions, decomposition kinetics, and reactor operating conditions. Therefore, H₂O₂ reflects the cumulative oxidative environment rather than instantaneous •OH production.

Recent studies have shown that optical absorption spectroscopy can monitor the accumulation of long-lived RONS, including H₂O₂, NO₂ ⁻, NO₃ ⁻, and dissolved O₃, in plasma-treated water (Figure 2c) (Kumar et al., 2025). However, the overlapping UV absorption spectra of H₂O₂, NO₂ ⁻, NO₃ ⁻, and dissolved O₃ can introduce significant analytical uncertainties. Therefore, quantitative measurements require careful calibration and, where appropriate, complementary analytical techniques to distinguish individual species. Oxidation-reduction potential (ORP) measurements further provide an integrated indicator of solution reactivity (Kumar et al., 2025). H₂O₂ is commonly quantified using titanium oxysulfate colorimetry or peroxide test strips, offering rapid semi-quantitative assessment, and is often used as an indirect marker of •OH generation (Kumar et al., 2022). EPR and fluorescence spectroscopy have also been widely used for peroxide detection (Myers et al., 2021; Tornin et al., 2021). Nitrite and nitrate are typically analyzed using the Griess assay, while dissolved ozone is measured via iodometry, indigo methods, or selective probes (Van Boxem et al., 2017; Cao et al., 2018; Kumar et al., 2025).

Direct measurement of ultrashort-lived radicals (in water) such as •OH remains difficult; therefore, fluorescence probes and EPR spin trapping remain the most widely adopted techniques. Fluorescent probes such as terephthalate (TPA), forming 2-hydroxyterephthalate (λ_{excitation/emission} = 310/425 nm), have also been applied (Tornin et al., 2021). Spin-trapping EPR techniques using DMPO and TEMP further enable detection of •OH and O• (Myers et al., 2021; Sobota et al., 2025). More recently, LIF has enabled direct measurement of •OH densities at CAP-liquid interfaces with substantially improved spatial and temporal resolution (Yang et al., 2024).

3.3 Ultrafast characterization of CAP-liquid interfacial species

Since many CAP-induced reactions occur on nanosecond-to-microsecond timescales at the CAP-liquid interface, ultrafast spectroscopic techniques have become essential for elucidating reaction mechanisms. Time-resolved infrared absorption spectroscopy (TIRAS), femtosecond transient absorption spectroscopy, ultrafast LIF, and femtosecond TALIF (Table 2) enable direct observation of transient intermediates such as e ^- _sol, O•, and •OH before secondary reactions dominate. These techniques have demonstrated that reactive-species generation is highly localized within several tens of micrometers of the CAP-liquid interface, highlighting the importance of interfacial transport for pollutant degradation and reactor optimization.

Among these transient species, e ^- _sol remain one of the least understood despite their central role in CAP-driven reductive degradation of persistent contaminants, particularly PFAS. Energetic gas-phase electrons can excite, dissociate, or ionize water molecules at the interface, whereas low-energy electrons may penetrate the liquid and become solvated through trapping within a hydrogen-bonded water cage (Foster, 2017). This picosecond-scale solvation process stabilizes the electron while suppressing avalanche formation in liquid water.

Due to their ultrashort lifetime and extreme reactivity, direct quantification of e ^- _sol remains challenging, and their presence is commonly inferred using electron scavengers (Cao et al., 2018; Sebastian et al., 2024; Kumar et al., 2026b). Recent advances include laser-based optical absorption spectroscopy based on the Beer-Lambert law to detect e ^- _sol at the CAP-liquid interface, taking advantage of their characteristic broad absorption band centered around ∼700 nm (Rumbach et al., 2015; Martin et al., 2021). In particular, Rumbach et al., 2015 reported one of the few direct measurements of e ^- _sol generated by atmospheric-pressure plasma in contact with water, using optical absorption spectroscopy to estimate both interfacial concentration and penetration depth via laser-based detection. Previous studies have demonstrated that NO₃ ⁻ can serve as an effective probe for e ^- _sol owing to its high electron-scavenging rate constant (9.7 × 10⁹ L mol^-1 s^-1). Its reduction to NO₂ ⁻ enables the indirect quantification of e ^- _sol production (Sebastian et al., 2024). More recently, femtosecond laser techniques have been explored for imaging transient species such as solvated oxygen atoms (O•), highlighting emerging ultrafast approaches for interfacial plasma chemistry (Myers et al., 2025).

Despite these advances, quantitative characterization of ultrashort-lived species such as e ^- _sol, O•, and •OH within liquids remains highly challenging, particularly in bubble-assisted CAP systems where multiple plasma microenvironments and complex interfacial transport processes coexist. Consequently, the generation yields, transport pathways, and reaction kinetics of reactive species remain insufficiently understood. Continued development of highly sensitive, time-resolved diagnostic techniques will therefore be essential for establishing quantitative links between plasma physics, interfacial chemistry, and pollutant degradation mechanisms.

4 CAP-driven degradation of POCs

Recent advances in CAP have demonstrated its potential as a versatile, non-catalytic, and chemical-free technology for the treatment of POCs. Depending on the pollutant properties and reactor configuration, CAP generates a broad spectrum of oxidative and reductive reactive species that drive different degradation pathways and treatment outcomes. However, the disappearance of the parent compound does not necessarily indicate complete contaminant destruction. Parent-compound disappearance, transformation-product formation, defluorination, mineralization, and detoxification represent distinct treatment endpoints that provide different information about treatment performance and should not be considered equivalent. Therefore, a comprehensive evaluation of CAP requires complementary analytical techniques to assess not only contaminant removal but also transformation products, fluoride release (for PFAS), mineralization, and toxicity reduction.

4.1 Parent-compound disappearance and transformation pathways

Most CAP studies initially evaluate treatment performance by monitoring the disappearance of the parent compound using chromatographic techniques such as HPLC or LC-MS. Recent studies have demonstrated efficient removal of a wide range of POCs, including pharmaceuticals, dyes, pesticides, PFAS, and industrial chemicals under various CAP configurations (Kumar et al., 2021; Karimian et al., 2025; Lamichhane et al., 2025; Mededovic Thagard, 2025). Our recent studies demonstrated degradation of Perfluorooctane sulfonic acid (PFOS), Perfluorohexane sulfonic acid (PFHxS), diclofenac (DCF), Acid Blue 25 (AB25), and p-chlorobenzoic acid (pCBA) using different CAP reactor configuration (Figure 2). PFOS and PFHxS degradation was investigated with and without gas dispersion. Air gas dispersion enhanced hydrodynamic mixing and enriched PFOS and PFHxS at the CAP-liquid interface, promoting interfacial microdischarges and increasing the concentration of short-lived reactive species, which improved both oxidative and reductive degradation pathways. Higher degradation efficiencies were observed in the presence of gas dispersion (Figures 2d,e) (Kumar et al., 2026b). CAP treatment of AB25 resulted in significant decolorization, indicating cleavage of the chromophoric group (Figure 2f) (Kumar et al., 2022). Similarly, pCBA degradation provided indirect evidence for •OH formation during plasma treatment (Figure 2g) (Kumar et al., 2023). DCF degradation using a multi-needle plasma reactor with recirculation enhanced CAP-liquid interaction and reactive species generation, resulting in improved degradation efficiency (Figure 2h) (Kumar et al., 2023).

Although the disappearance of the parent compound is frequently used as an indicator of degradation efficiency, it does not necessarily indicate complete destruction of the contaminant. Instead, the parent molecule may be transformed into numerous intermediate products with different physicochemical properties, persistence, mobility, and toxicity.

Identification of transformation products is essential for understanding CAP reaction mechanisms. High-resolution mass spectrometry, particularly Orbitrap LC-MS, has become one of the most powerful analytical tools for elucidating oxidation and reduction pathways because it enables identification of intermediate compounds based on accurate mass measurements. For example, Orbitrap LC-MS analysis of CAP-treated DCF revealed successive hydroxylation, decarboxylation, dechlorination, ring-opening reactions, and formation of multiple oxygenated intermediates (Figure 2i) (Kumar et al., 2023). However, Orbitrap LC-MS identifies transformation products and reaction pathways rather than mineralization. Therefore, identification of degradation intermediates should not be interpreted as evidence of complete carbon mineralization. Furthermore, transformation products may remain persistent, mobile, or even exhibit greater toxicity than the parent compound. Consequently, CAP treatment should be evaluated not only by parent-compound removal but also by the persistence, mobility, and toxicity of transformation products whenever possible.

4.2 Defluorination and PFAS degradation

PFAS degradation requires separate evaluation because cleavage of the exceptionally strong C–F bond differs fundamentally from the oxidation of conventional organic contaminants. PFAS destruction should be assessed by combining transformation-product identification, stepwise chain shortening, fluoride release, total organofluorine removal, and mineralization. Direct evidence of PFAS destruction requires fluoride release and fluorine mass balance, while complete mineralization should be confirmed by carbon mass balance measurements.

Current evidence suggests that e ^- _sol are the primary reductive species responsible for initiating PFAS degradation through one-electron transfer reactions that promote C–S and subsequently C–F bond cleavage (Kumar et al., 2026b). For PFOS, degradation is proposed to begin with desulfonation of the sulfonate head group, producing perfluoroalkyl radicals that are subsequently converted into perfluorocarboxylic acids (e.g., PFOA). These intermediates then undergo sequential decarboxylation and stepwise shortening of the perfluorinated carbon chain, accompanied by progressive fluoride release. However, these pathways are inferred primarily from transformation-product analysis and scavenger experiments rather than direct observation of individual elementary reactions. Hydrogen atoms may also contribute to reductive reactions, whereas short-lived oxidative species, including •OH and O•, are generally considered to play a greater role in the oxidation of intermediate products than in direct C–F bond cleavage (Mbanugo et al., 2025; Ganzallo et al., 2026). Likewise, CAP-generated UV photons, strong interfacial electric fields, energetic ions, and localized thermal microzones may facilitate initial activation of PFAS molecules, particularly at the CAP-liquid interface where long-chain PFAS preferentially accumulate because of their surface-active nature (Mededovic Thagard, 2025; Kumar et al., 2026b). Their individual contributions, however, remain difficult to quantify due to the simultaneous generation of multiple reactive species and coupled CAP-liquid processes.

Strategies such as gas bubbling, plasma generated within microbubbles, direct plasma exposure above the liquid surface, and surfactant-assisted interfacial enrichment have been shown to improve PFAS degradation by increasing CAP-liquid interfacial area and reactive-species transport (Aranzales et al., 2025; Mededovic Thagard, 2025; Kumar et al., 2026b). Bubble-based plasma systems may be particularly advantageous for short-chain PFAS because enhanced mass transfer improves contact between dissolved contaminants and plasma-generated reactive species. Nevertheless, the relative importance of oxidative and reductive pathways for long- and short-chain PFAS remains an active area of research, requiring advanced time-resolved diagnostics and comprehensive fluorine mass balance analyses to establish definitive reaction mechanisms.

4.3 Mineralization versus pollutant removal

Complete mineralization represents a substantially higher level of treatment than parent-compound disappearance, as it implies the transformation of organic compounds into inorganic end products. For carbon-containing pollutants, mineralization is generally evaluated through the conversion of organic carbon into CO₂, while heteroatoms such as fluorine, chlorine, nitrogen, or sulfur require additional monitoring through the release of inorganic species (e.g., F⁻, Cl⁻, NO₃ ⁻, SO₄ ^2-) and corresponding elemental mass balances. Therefore, demonstrating true contaminant destruction requires complementary analytical measurements including TOC, DOC, COD, inorganic carbon, fluoride or chloride release, and overall carbon/fluorine mass balances.

This distinction is particularly critical for CAP-based treatment, where highly reactive plasma-generated species can rapidly transform parent compounds while generating a wide range of intermediate products with different persistence, toxicity and biodegradability. Several studies have reported high pollutant removal efficiencies while observing limited mineralization, indicating accumulation of transformation products (Kumar et al., 2021; He et al., 2022; Mbanugo et al., 2025). Consequently, future CAP investigations should distinguish clearly between degradation efficiency and mineralization efficiency and report appropriate carbon and halogen mass balances to verify complete contaminant destruction.

4.4 Detoxification and treatment effectiveness

Ultimately, the goal of water treatment is not only pollutant removal but also the reduction of environmental and human health risks. Detoxification therefore represents the highest treatment endpoint, as transformation products may differ from the parent compounds in toxicity, persistence, and mobility. Several studies have shown that CAP treatment reduces the overall toxicity of various POCs, with transformation-product mixtures generally being less toxic than the parent compounds (Kovačič et al., 2023; Ni et al., 2024). However, toxicity reduction should be confirmed using appropriate bioassays and ecotoxicity tests rather than inferred solely from parent-compound removal. Therefore, future CAP studies should combine chemical analyses with toxicity assessment to comprehensively evaluate treatment performance.

4.5 Energy consumption

From a process engineering perspective, CAP performance is commonly evaluated using the electrical energy per order (EEO), defined as the energy (kWh) required to reduce a contaminant concentration by one order of magnitude in 1 m³ of water. Reported EEO values vary depending on pollutant type and reactor configuration, with typical values for selected POCs ranging from approximately <1 to several tens of kWh/m³ per order (Vanraes et al., 2017; Miklos et al., 2018; Papalexopoulou et al., 2024). Compared with conventional advanced oxidation processes such as ozonation, which can exhibit selectivity toward specific compounds, CAP offers broader reactivity due to the simultaneous generation of multiple reactive species. Miklos et al., 2018 classified AOPs into three energy categories: <1 kW h/m³ per order (e.g., ozonation and UV-based AOPs), 1–100 kW h/m³ per order (including plasma, photo-Fenton, and electrochemical oxidation), and >100 kW h/m³ per order (e.g., photocatalysis and ultrasound). Under optimized conditions, several CAP studies have also reported EEO values at or below 1 kW h/m³ per order, demonstrating its potential for energy-efficient treatment of selected pollutants (Chen et al., 2023; Meropoulis and Aggelopoulos, 2024). However, direct comparison among treatment technologies remains difficult because studies employ different contaminants, water matrices, operating conditions, and performance endpoints. In addition, auxiliary energy demands, including gas production, pumping, recirculation, cooling, and off-gas treatment, are often excluded from reported EEO values.

For PFAS treatment in particular, recent studies indicate that CAP is among the most energy-efficient destructive technologies currently under development, especially when compared with high-temperature incineration and several conventional AOPs (Mededovic Thagard, 2025; Kumar et al., 2026b). Nevertheless, comprehensive techno-economic analyses and life-cycle assessments remain limited and are required to accurately assess the feasibility of large-scale CAP implementation.

5 Advances in multiphysics modeling and machine learning for CAP-based systems

Computational modelling plays a crucial role in understanding CAP processes and predicting the generation of reactive species during CAP-liquid interactions. By overcoming key experimental limitations, it provides valuable insights into species dynamics and underlying mechanisms that are difficult to probe or quantify experimentally (Liu et al., 2021; Yang et al., 2023; Karimian et al., 2025). Numerical simulations are therefore widely combined with experimental studies to investigate plasma chemistry, reactive-species transport, and pollutant degradation mechanisms. Among the available modeling platforms, COMSOL Multiphysics is widely used for CAP simulations because it can couple plasma discharge physics with fluid dynamics and mass transport (Momtazzadeh et al., 2025). This allows simulation of gas flow, electric field, electron density, plasma potential, and the formation and transport of reactive species. Its practical value lies in its ability to describe complex, coupled phenomena at the CAP-liquid interface and to support validation of experimental observations. Using such models, researchers can estimate key plasma parameters, including electron density, temperature distribution, and concentrations of short- and long-lived reactive species (Vanraes et al., 2018a; Heirman et al., 2019; Yang et al., 2023). Several modeling approaches have been applied depending on the required level of complexity (0D, 1D, 2D, and 3D) (Heirman et al., 2019; Gangawane et al., 2023).

For example, non-dimensional kinetic models are often used to analyze detailed reaction pathways and reactive-species formation, especially near the CAP-liquid interface, while multidimensional fluid models can describe spatial distributions of electric fields, charged species, and gas flow. In DBD systems, numerical simulations have also been used to examine species generation and transport under different operating conditions (Arserim et al., 2021). In many CAP studies, COMSOL is coupled with BOLSIG+, a Boltzmann equation solver that calculates the EEDF, electron transport coefficients, and electron-impact reaction rate coefficients from cross-sectional data, often obtained from databases such as LXCat (Vanraes et al., 2018a; Semenov and Weltmann, 2020). These data are then incorporated into plasma fluid models to improve the description of non-equilibrium discharge chemistry. Such combined COMSOL and BOLSIG + approaches provide a more reliable framework for predicting RONS generation and are therefore valuable for optimizing CAP reactors for water treatment. However, the reliability of CAP simulations strongly depends on model assumptions and boundary conditions. In CAP-liquid systems, maintaining uncertainty within an acceptable range requires careful selection of interfacial reaction schemes, transport parameters, and Henry’s law constants, since these factors strongly affect species transfer across the gas-liquid interface (Liu et al., 2021).

Recent advances have highlighted machine learning (ML) as a powerful tool for both mechanistic understanding and optimization of CAP-based water treatment systems (Fang et al., 2024; Gao et al., 2024; Shen et al., 2025). CAP-liquid interactions involve highly complex reaction networks, making conventional numerical simulations computationally expensive and often limited in resolving coupled physical and chemical processes. In this context, ML has emerged as an effective approach to model CAP behavior and extract insights into the underlying physicochemical mechanisms governing CAP-liquid interactions, including the prediction of RONS generation under varying operating conditions. In parallel, ML is widely used for process optimization, where data-driven models are employed to identify optimal operating parameters for enhanced treatment performance. Several recent studies demonstrate this dual capability. For instance, Shen et al., 2025 applied ML techniques to optimize the plasma degradation of the antibiotic norfloxacin while minimizing energy consumption. Similarly, Gao et al., 2024 utilized ML to optimize microbubble-enhanced plasma activation processes, and Xue et al., 2024 developed predictive models for antibiotic removal in plasma reactors.

Further advances include the work of Wang et al., 2025b, who proposed a ML-driven reverse-design framework for precise control of ROS generation in plasma reactors enhanced with Fe₃O₄ particle electrodes, contributing to more sustainable wastewater treatment technologies. Fang et al., 2024 employed ML to quantify the relative contributions of reactive species in a multistage oxygen-enhanced plasma reactor, significantly improving the removal efficiency of organic dyes. These studies collectively highlight the growing role of ML in accelerating reactor optimization and enhancing reactive species utilization in CAP-based systems. Beyond process optimization, ML has also been widely applied in plasma diagnostics. It has been used to analyze diverse data sources, including OES, current and voltage signals, LIF, mass spectrometry data, and visual imaging. In this regard, He et al., 2024 provided a comprehensive overview of ML integration with CAP, emphasizing its transformative impact on modeling, diagnostics, and applications. Additionally, Srikar et al., 2024 employed advanced ML techniques combined with OES and collisional-radiative (CR) models to enable non-invasive, real-time prediction of electron temperature in non-thermal argon plasma jets. Similarly, ML-based approaches have been developed for predicting optical emission spectra in scalable non-thermal atmospheric-pressure argon plasma multi-jet systems, allowing detailed analysis of operating conditions and their influence on reactive species generation (Srikar et al., 2025). Overall, the integration of ML with CAP systems represents a promising pathway toward intelligent plasma control, real-time diagnostics, and efficient wastewater treatment.

6 Discussion and future outlook

CAP has emerged as a promising technology for the treatment of POCs. CAP-liquid interactions generate a wide range of oxidative and reductive reactive species capable of degrading complex contaminants. The efficiency of CAP strongly depends on reactor configuration, plasma operating conditions, feed gas composition, and CAP-liquid interaction, all of which govern reactive-species generation, transport, and utilization (Liu et al., 2021; Gao et al., 2024; Kumar et al., 2025). Recent advances in plasma diagnostics, multiphysics modeling, and ML have substantially improved the mechanistic understanding and optimization of CAP systems.

Despite these advances, one of the major challenges remains the limited understanding of the ultrashort-lived reactive species that govern CAP-driven chemistry. Although significant progress has been achieved using optical diagnostics and chemical probes, direct characterization of species such e ^- _sol, O•, and •OH at the CAP-liquid interface remains scarce because of their extremely short lifetimes and localized formation. This challenge becomes even greater in gas-dispersion CAP systems, where CAP generated inside bubbles creates complex interfacial microenvironments. Future advances in ultrafast spectroscopy, laser-based diagnostics, and high-resolution imaging, combined with multiphysics simulations and ML-assisted analysis, are expected to provide deeper insight into reactive-species generation and transport.

PFAS degradation remains one of the most important emerging applications of CAP. Although long-chain PFAS can be effectively degraded through interfacial plasma processes, short-chain PFAS remain more difficult to treat because of their higher aqueous solubility and lower interfacial enrichment. Recent reactor developments, including thin-film and falling-film reactors, gas bubbling, microbubble generation, flotation, and surfactant-assisted interfacial enrichment, have improved degradation and defluorination efficiencies, but further optimization is required to achieve energy-efficient treatment under realistic conditions.

Future research should focus on validating CAP under continuous-flow operation and in complex real-water matrices while integrating transformation-product analysis, mineralization, toxicity assessment, and comprehensive mass balances into performance evaluation. In parallel, hybrid treatment systems combining CAP with hydrodynamic cavitation, photocatalysis, electrochemical oxidation, flotation, membrane separation, and other advanced oxidation processes offer promising opportunities to enhance degradation efficiency, improve energy efficiency, and broaden the range of treatable contaminants. The development of portable, low-energy CAP systems could further facilitate deployment in industrial wastewater treatment, municipal water and wastewater treatment plants, and remote or resource-constrained areas. Finally, techno-economic analysis and life-cycle assessment will be essential to evaluate the feasibility, sustainability, and scalability of CAP technologies for large-scale water treatment applications.

7 Concluding remarks

CAP has emerged as a promising advanced water treatment technology because it simultaneously generates oxidative and reductive reactive species under ambient conditions, enabling the degradation of a broad range of POCs. Current evidence indicates that plasma-above-liquid reactors are particularly effective for treating surface-active contaminants such as long-chain PFAS through enhanced interfacial reactions, whereas bubble-plasma and microbubble reactors show greater potential for dissolved pollutants and short-chain PFAS by increasing the gas-liquid interfacial area and improving reactive-species transport into the bulk liquid. To overcome the challenges associated with short-chain PFAS degradation, recent studies have combined gas dispersion with flotation or surfactant-assisted interfacial enrichment to concentrate PFAS at the CAP-liquid interface, thereby improving degradation and defluorination efficiencies. Overall, advances in reactor design and CAP-liquid interaction have significantly expanded the applicability of CAP for the treatment of diverse classes of POCs.

The review highlights that reliable identification of the dominant degradation pathways requires complementary diagnostic techniques rather than a single analytical method. While OES is valuable for monitoring excited gas-phase species and plasma stability, techniques such as LIF/TALIF, EPR, chemical probes, ultrafast spectroscopy, and mass spectrometry are required to characterize ground-state radicals, e ^- _sol, transformation products, and CAP-liquid interfacial processes. Despite recent progress, direct measurement of ultrashort-lived reactive species, particularly e ^- _sol, O•, and •OH at the CAP-liquid interface, remains very limited. This challenge is even greater in gas-dispersion CAP systems, where plasma generated inside bubbles creates complex interfacial microenvironments that are highly relevant to PFAS degradation and defluorination.

This review also emphasizes the need for standardized evaluation of CAP performance, including electrical energy consumption, transformation-product identification, mineralization (TOC, DOC, COD, or inorganic carbon), halide or fluoride release, elemental mass balances, toxicity assessment, and validation under continuous-flow operation using real water matrices. Although CAP has demonstrated considerable potential for the treatment of pharmaceuticals, dyes, pesticides, and PFAS, further advances in energy efficiency, comprehensive by-product characterization, advanced diagnostics, real-water validation, and techno-economic and life-cycle assessments are required before large-scale implementation can be realized.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The author(s) declare that generative AI was used primarily for grammatical editing of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Pradeep Lamichhane, Imperial College London, United Kingdom

Reviewed by: Sohail Mumtaz, Gachon University, Republic of Korea