Critical Review of UV-Advanced Reduction Processes for the Treatment of Chemical Contaminants in Water

Abstract

UV-advanced reduction processes (UV-ARP) are an advanced water treatment technology characterized by the reductive transformation of chemical contaminants. Contaminant abatement in UV-ARP is most often accomplished through reaction with hydrated electrons (e_aq^–) produced from UV photolysis of chemical sensitizers (e.g., sulfite). In this Review, we evaluate the photochemical kinetics, substrate scope, and optimization of UV-ARP. We find that quantities typically reported in photochemical studies of natural and engineered systems are under-reported in the UV-ARP literature, especially the formation rates, scavenging capacities, and concentrations of key reactive species like e_aq. The absence of these quantities has made it difficult to fully evaluate the impact of operating conditions and the role of water matrix components on the efficiencies of UV-ARP. The UV-ARP substrate scope is weighted heavily toward contaminant classes that are resistant to degradation by advanced oxidation processes, like oxyanions and per- and polyfluoroalkyl substances. Some studies have sought to optimize the UV-ARP treatment of these contaminants; however, a thorough evaluation of the impact of water matrix components like dissolved organic matter on these optimization strategies is needed. Overall, the data compilation, analysis, and research recommendations provided in this Review will assist the UV-ARP research community in future efforts toward optimizing UV-ARP systems, modeling the e_aq^–-based chemical transformation kinetics, and developing new UV-ARP systems.

1. Introduction

Ultraviolet-based advanced reduction processes (UV-ARP) have received significant attention in recent years for the treatment of several classes of recalcitrant chemical contaminants in water.¹⁻⁷ Advanced reduction processes are based on production of highly reducing hydrated electrons (e_aq^–, Figure 1), which exhibit fast bimolecular reaction rate constants with inorganic and organic compounds.⁸ In UV-ARP, e_aq are produced by the illumination of e_aq^– sensitizers with UV lamps. Photoproduced e_aq can react with target contaminants (TC), leading to TC transformation. Given its high reactivity, e_aq^– is also scavenged by various chemical species naturally present in water.

Figure 1.

Summary of UV-advanced reduction processes (UV-ARP) generation of hydrated electrons (e_aq^–).

The need for UV-ARP is motivated by the fact that many pollutants commonly found in contaminated waters are not readily treated by UV-advanced oxidation processes (UV-AOP), one of the most widely used degradative technologies in advanced water treatment. The need for an alternative to in situ oxidation processes has been recognized for some time. A patent by Bolton and Cater in 1993 detailing the treatment of wastewater or groundwater containing chloroform as a model halogenated organic with UV/iodide, UV/thiosulfate, UV/sulfite, and UV/iodide/thiosulfate as the e_aq^– source was one of the first reports of this technology.⁹ Subsequent UV-ARP studies have focused on substrates that can undergo reduction by e_aq reaction such as oxyanions,^6,10−19 highly oxidized elements (e.g., Cr(VI)),^20,21 halogenated disinfection byproducts,^5,6,16 and per- and polyfluoroalkyl substances (PFAS).²²⁻³⁴ Several bench-scale studies²²⁻³¹ and a recent pilot-scale³² demonstration suggest that UV-ARP is one of the more promising technologies for degrading PFAS, which are resistant to direct photolysis and UV-AOP.^22,23,35

Despite the promise of UV-ARP, implementation of these technologies at the large scale has not been realized. Given the high level of research activity of UV-ARP for the e_aq^–-based transformation of contaminants over the past decade, there is a critical need to understand the limitation of these technologies from both chemical and engineering perspectives. While recent publications have reviewed various aspects of UV-ARP, like specific e_aq sensitizer systems (e.g., UV-sulfite^3,36,37) or applicability toward certain contaminant classes (e.g., PFAS^2,38), a comprehensive review of UV-ARP technologies is lacking. The objective of this Review is therefore to critically evaluate the factors influencing the effectiveness of UV-ARP for the degradation of contaminants in water. We summarize and review the UV-ARP literature for treatment of some important inorganic and organic contaminants, identify the knowledge gaps and potential limitations for these technologies, and suggest paths forward for research in this field. This is accomplished by presenting the photochemical kinetics of e_aq^– production and scavenging in UV-ARP, reviewing the sensitizers that have been employed to generate e_aq in UV-ARP, examining the classes of contaminants for which UV-ARP treatment has been studied, and highlighting novel UV-ARP technologies that seek to increase the rate of target contaminant degradation. The data compilation and perspectives provided in this review will assist the UV-ARP research community in future efforts to optimize UV-ARP systems, model the e_aq^–-based transformation kinetics, and develop novel UV-ARP systems.

2. Production and Scavenging of Hydrated Electrons in UV-ARP

In UV-ARP, target contaminants are degraded through direct photolysis or through reaction with photochemically generated e_aq^–9,39 (Figure 2). Considering the latter, indirect, photolysis pathway, the rate of target contaminant transformation in UV-ARP systems is a function of both the intrinsic reactivity of the contaminant with e_aq (its bimolecular reaction rate constant) and the concentration of e_aq^–. The bimolecular rate constant for a target contaminant with e_aq is a constant value under specific solution conditions (e.g., temperature, ionic strength). Thus, strategies to optimize UV-ARP must aim to increase the exposure of the contaminant to e_aq^–, by increasing either the irradiation time or the hydrated electron concentration, [e_aq]. Increasing [e_aq^–] can be accomplished through multiple means, such as using higher wattage lamps, increasing sensitizer concentrations, and minimizing the rate of interfering e_aq scavenging reactions. Understanding the factors that influence [e_aq^–] is therefore an important consideration in UV-ARP studies. The sections below address these factors by using established photochemical kinetic models to describe the rate of e_aq-based contaminant transformation, the chemistry of known e_aq^– sensitizers, and the scavenging kinetics of e_aq in water. We end this section by focusing on knowledge gaps in this area, primarily the lack of a standardized measurement protocol for quantifying [e_aq^–] in UV-ARP.

Figure 2.

Direct and indirect (i.e., sensitized) photochemical transformation of a target contaminant in UV-ARP. The sensitizer charge (n) increases by +1 upon photoionization.

2.1. Transformation of a Target Contaminant by Hydrated Electrons

The rate law for the loss of a target contaminant (TC) in UV-ARP, which can be due both to direct photolysis and e_aq^–- sensitized reactions, is given by

2.1

where k_d^′ is the pseudo-first-order rate constant of TC by direct photolysis (s^–1), k_TC,eaq^– is the TC-e_aq bimolecular reaction rate constant (M^–1 s^–1), [TC] is the concentration (M) of TC remaining, and [e_aq^–] is the concentration (M) of hydrated electron.⁴⁰Equation 2.1 can be simplified by assuming that [e_aq] remains at a steady-state throughout the treatment process, resulting in the expression

2.2

where k_TC^′ is a first-order rate constant (k_TC = k_TC,eaq^–[e_aq^–]_ss). Equations 2.1 and 2.2 forms the basis of all photochemical kinetic models for describing the time dependence of [TC]. Equation 2.1 can be further specified by including equations that describe the formation and scavenging of e_aq. For example, if [e_aq^–] cannot be approximated as steady-state, then the time-dependent [e_aq] can be described by

2.3

where R_f,t^e_aq– represents the formation rate of e_aq (M s^–1) and k_S,t^′ represents the total e_aq scavenging capacity, all as a function of time t.

In UV-ARP, e_aq^– is typically formed by illumination of a chemical sensitizer by UV lamps. Under monochromatic conditions (e.g., a low-pressure Hg lamp), eq 2.3 can be expanded to eq 2.4

2.4

where Φ_eaq^– is the e_aq^– quantum yield of the sensitizer (sens), I₀ is the photon irradiance (mmol photons cm^–2 s^–1), ε_sens is the molar absorption coefficient (M^–1 cm^–1) of the sensitizer, [sens]_t is the sensitizer’s concentration at time t (M), α_t is the absorption of the background water matrix (cm^–1) at time t, k_Si,eaq^– is the bimolecular reaction rate constant for transformation of scavenger S_i (M^–1 s^–1), [S_i]_t is the scavenger concentration (M) at time t, and l is the path length (cm).⁴¹ The units in eq 2.4 are consistent because the conversion from mmol to mol and cm³ to L cancel each other.⁴¹

Integrating eq 2.1 results in a complete expression (eq 2.5), which can be employed to model the change in [TC] as a function of time.

2.5

In eq 2.5, ∫₀^t[e_aq]_t dt is called the e_aq^– exposure and incorporates both R_f,t and k_S,t^′ for any given background water matrix conditions. While the above expressions are time-based, it is also possible to present target contaminant transformation kinetics using fluence-based rate constants, which has been discussed extensively by Bolton et al.⁴²

Direct photolysis plays a minor but non-negligible role in the transformation of some target contaminants. The contribution of direct photolysis to contaminant transformation in UV-ARP can be calculated in terms of the direct photolysis rate constant, k_d^′

2.6

where Φ_TC is the reaction quantum yield for direct photolysis of the TC, ε_TC is the molar absorption coefficient of the TC (M^–1 cm^–1), and other terms ([TC], I₀, l, and α_t) are as defined previously.⁴¹ A key point is that there must be overlap between the spectra of the light source and TC for direct photolysis to occur (Grotthus–Draper law⁴³). Contaminants containing chromophores that absorb in the UVC (e.g., nitrate, nitrite, and halogenated aromatic compounds) undergo direct photolysis.⁴⁴⁻⁴⁶ Given the typical minor role of direct photolysis in UV-ARP, the factors influencing this pathway are not discussed extensively in this Review. Section 3 will specify contaminant classes in which direct photolysis plays a role.

2.2. Kinetics of Hydrated Electron Scavenging Reactions

Background water matrix components that react with e_aq^– decrease its availability for reaction with the target contaminant (TC) according to the following competition

2.7

2.8

where S represents the sum of e_aq^– scavengers other than the TC. The expected impact of e_aq scavengers on target contaminant transformation can be estimated by determining the first-order scavenging capacity (k_S,t^′, s^–1), which is defined as the sum product of bimolecular e_aq scavenging rate constants (k_Si,eaq^–) and [S_i]_t (eq 2.9)

2.9

Bimolecular rate constants for e_aq^– reactions depend on temperature and aqueous ionic strength according to the Arrhenius (eq 2.10) and Brønsted–Bjerrum equations (eq 2.11), respectively.⁸

2.10

2.11

In these equations, k₂ represents the bimolecular rate constant at two different temperatures (T₁ or T₂, units of K) or ionic strengths (I = 0 or I, units of M), E_a is the activation energy (kJ/mol), R is the gas constant, and Z_AZ_B is the product of the charges of the reactants. Because e_aq^– bears a negative charge, its reactivity with anionic species will be enhanced at higher ionic strengths because the increased shielding of like-charged reactants decreases the distance over which Coulombic forces repel the ions. Many of the contaminant classes highlighted in this Review are anionic at pH ranges used for UV-ARP, so increasing ionic strength will increase the e_aq bimolecular rate constant.

Figure 3 shows the relative change in the e_aq^– bimolecular rate constant as a function of temperature (10–40 °C) and ionic strength (0.001–1 M), which covers the range of values that could be anticipated in waters treated by UV-ARP. An activation energy of 15 kJ/mol and Z_AZ_B of 1 were assumed for these calculations. The bimolecular rate constant is more sensitive to temperature than ionic strength, with much higher ionic strengths (hundreds of mol/L) needed to achieve the same impact on the rate constant as a 10–40 °C temperature increase. Treatment processes that concentrate contaminants (e.g., by filtration or ion exchange) prior to UV-ARP readily achieve high ionic strengths. Thus, bimolecular rate constants will need to be modified accordingly (eq 2.11) when applying quantitative kinetic models (e.g., eq 2.5) in these high ionic strength systems.

Figure 3.

Relative change in bimolecular rate constant (k₂) as a function of the relative change in A) temperature (T) or B) ionic strength (I). Note the difference in scale of the x axis. The temperature dependence calculation assumes an activation energy of 15 kJ/mol, which is typical for radical reactions. The ionic strength dependence calculation assumes that Z_AZ_B = 1 as this would cause an increase in k₂ with increasing ionic strength. The inset in (B) has the same axes labels as the outer figure.

2.2.1. Hydrated Electron Scavenging Reactions in Real-World Waters

In UV-ARP systems, there are many e_aq^– scavengers, all with varying concentrations and reactivities toward e_aq. Scavengers can arise from background water matrix components (H⁺, O₂, NO₃^–, NO₂^–, and DOM) and even the e_aq^– sensitizer itself (see section 2.3). A list of e_aq bimolecular rate constants used extensively in this Review is provided in Table 1. This list does not describe all e_aq^– reactions or those initiated by e_aq that could be occurring in UV-ARP. We refer readers to the Notre Dame Radiation Laboratory/National Institute of Standards and Technology solution kinetics database (https://kinetics.nist.gov/solution/) as a source of bimolecular rate constants for e_aq^– and other radical species.

Table 1. Bimolecular Rate Constants for Reaction of e_aq^– with Species Discussed in This Review^a.

^aValues obtained from the NDRL/NIST solution kinetics database (https://kinetics.nist.gov/solution/) unless otherwise noted. The reaction products are shown only if present in the NDRL/NIST database. See Table 4 for tabulated e_aq^– bimolecular rate constants for PFAS. The Supporting Information contains a larger list of e_aq bimolecular rate constants for aliphatic and aromatic compounds.

From, eq 2.4, we can see that [e_aq^–] is inversely proportional to k_S,t. In principle, it is possible to calculate k_S,t^′ a priori provided the concentrations and rate constants for all important e_aq scavengers are known. Figure 4 illustrates this fact by showing the scavenging capacity for selected water matrix components as a function of their concentration (eq 2.9) calculated using the values in Table 1. While bimolecular e_aq^– rate constants are widely available for important hydrated electron scavengers,⁸ conducting a full water chemistry analysis to determine scavengers' concentrations is not always practical. In addition, Rosenfeldt and Linden demonstrated that calculated ^•OH scavenging capacities in UV-AOP differed substantially from values measured with p-chlorobenzoic acid,⁴⁰ suggesting that the same may be true for UV-ARP.

Figure 4.

Scavenging capacities of (k_S,t=0^′) as a function of species concentration for some e_aq scavengers in UV-ARP. Bimolecular rate constants used for these calculations are shown in Table 1. A value of 1.2 × 10⁸ M_C^–1 s^–1 (∼10⁴ L mgC^–1 s^–1) was used for the DOM-e_aq^– reaction based on the DOM-^•OH rate constant for SRFA.⁵⁰ Filled circles represent the scavenging capacity of each species under the following conditions: [H⁺] = 10^–7 M, [O₂] = 250 × 10^–6 M (8 mg/L), [NO₃^–] = 1.61 × 10^–5 M (1 mg/L), [NO₂^–] = 2.17 × 10^–5 M (1 mg/L), [DOM] = 1.66 × 10^–4 molC L^–1 (2 mgC/L), and [HCO₃^–] = 10^–3 M (61 mg/L).

These scavenging capacities allow for comparison of the impact of different water matrix components on [e_aq^–]. For example, while the fastest bimolecular rate constant is for e_aq reaction with H⁺, the typical [H⁺] (pH > 9) is optimized to minimize this pathway, making the e_aq^– scavenging capacity of H⁺ a small contribution to k_S,t in most UV-ARP.⁴⁸ Dissolved oxygen (∼250 μM in aerated water) is a significant scavenger of e_aq^–. In UV-ARP treatment, removal of dissolved oxygen is accomplished through mechanical deaeration (sparging with N₂ or Ar) or reactive quenching by sulfite,^32,51 which also functions as an e_aq sensitizer. Figure 4 demonstrates that dissolved oxygen will exert a significant scavenging capacity (k_S,t=0^′ = 4.75 × 10⁶ s^–1 for [O₂] = 250 μM) if not removed prior to treatment, which is an important limitation of UV-ARP (see section 2.2.2). Nitrate is a potent e_aq scavenger, and the concentration of this species in natural waters (∼1 mg/L for 50% of natural waters in the United States⁵²) will exert a significant scavenging capacity (k_S,t=0^′ = 2.02 × 10⁵ s^–1), thus limiting the effectiveness of UV-ARP. This idea is illustrated by Figure 5 which shows that UV/sulfite treatment of perfluorooctanoic acid (PFOA) in a contaminated groundwater is limited until nearly all nitrate and nitrite (formed during nitrate photoreduction) have been degraded.

Figure 5.

Time courses of nitrate, nitrite, and perfluorooctanoic acid (PFOA) during UV/sulfite treatment of a contaminated groundwater. Line at ∼200 min illustrates the point at which e_aq^– scavenging by the water matrix becomes low enough for PFOA-e_aq reactions to occur. Experimental conditions: 18 W low-pressure Hg lamp, pH₀ = 9.5, 10 mM sulfite, 30 min of sparging with industrial grade N₂, ACE glass water-jacketed immersion well reactors. PFOA concentration was measured according to previously described literature methods.³²

Another probable e_aq^– scavenger is dissolved organic matter (DOM). Despite the ubiquity of DOM in waste streams for UV-ARP treatment, the reactivity between e_aq and DOM remains unexplored. DOM contains several oxygenated functional groups,⁵³ which, as individual chemical moieties, are known to have high reactivity with e_aq^–. These moieties include carbonyl-containing compounds (e.g., aldehydes, ketones, and quinones), alkoxyphenols, and amino acids.^8,54 The estimated value of 10⁴ L mgC^–1 s^–1 (∼10⁸ L molC^–1 s^–1) shown in Table 1 is based on the bimolecular rate constant for the reaction between DOM isolates and the hydroxyl radical reported by Westerhoff et al.⁵⁰

2.2.2. Hydrated Electron Scavenging by Dissolved Oxygen in Laboratory Studies

UV-ARP studies are often performed in buffered, laboratory grade water that has been sparged with an inert gas like nitrogen to remove dissolved O₂. However, prior research shows that a residual [O₂] of approximately 0.2 mg/L (∼6 μM) remains even after sparging with nitrogen (25 mL/s) for 1 h.⁵⁵ An [O₂] of 6 μM in deaerated solutions will still exhibit a high e_aq^– scavenging capacity (k_S,t=0 = 1.2 × 10⁵ s^–1), which will impact the initial [e_aq^–] and thereby the rate of target contaminant transformation. Therefore, variations in [O₂] in laboratory studies with different sparging methods may substantially impact the observed kinetics for target contaminant transformation, making it difficult to compare UV-ARP data for different sensitizer systems and experimental setups. In addition, the use of different e_aq sensitizers may impact the residual [O₂] since reducing species like sulfite can remove O₂.^32,51

The impact of residual [O₂] will also depend on the solution conditions and target contaminant. Residual [O₂] will have the largest impact on the degradation of contaminants that have an e_aq^– bimolecular rate constant comparable to O₂ and thus react on similar time scales. Examples include nitrate, nitrite, and most halogenated organic compounds. On the other hand, target contaminants with slower e_aq bimolecular rate constants will be less impacted because the residual [O₂] will be reduced by e_aq^– at early reaction times. For example, Bentel et al. observed similar degradation kinetics for PFOA in the UV/sulfite system regardless of whether the solution was initially sparged with nitrogen.³⁰ This concept is further illustrated in Figure 6, which shows the calculated concentration profiles for the transformation of several compounds in a typical UV-ARP. Fast-reacting compounds like oxygen, nitrate, and nitrite are completely removed within a few minutes, while perfluorooctanoic acid (PFOA) and perfluoroctanesulfonate (PFOS), two representative PFAS, are degraded at much longer time scales.

Figure 6.

Calculated concentration profiles of typical UV-ARP target contaminants based on eq 2.5, assuming [e_aq^–] = 4.76 × 10^–12 M, no direct photolysis for nitrate and nitrite, and bimolecular rate constants (M^–1 s^–1) of k_PFOA,eaq^– = 5.1 × 10⁷, k_PFOS,eaq^– = 7.3 × 10⁷, k_{NO3^–,eaq^–} = 9.7 × 10⁹ M^–1 s^–1, k_{NO2^–,eaq^–} = 3.5 × 10⁹ M^–1 s^–1, k_O2,eaq^– = 1.9 × 10¹⁰ M^–1 s^–1.

2.3. Chemical Sensitizers for Hydrated Electron Production

It is possible to generate e_aq^– from the photolysis of many organic and inorganic compounds if the appropriate illumination wavelengths are used.⁵⁶ However, only a subset of the chemicals capable of producing e_aq under UV irradiation have been explored in UV-ARP. Absorption of photons by sensitizer (sens) molecules will result in a portion of them ejecting an electron (eq 2.12).⁵⁶

2.12

The ejected e^– initially forms an ion contact pair with sensⁿ⁺¹, and only a fraction of e^– eventually escapes the contact pair to form hydrated electrons, e_aq^–. The overall efficiency of this process can be described by the quantum yield of e_aq formation (Φ_eaq^–), which represents the rate of e_aq^– formation (R_f, M s^–1) divided by the rate of light absorption by the sensitizer (photons L^–1 s^–1). Employing sensitizers with higher Φ_eaq^– therefore represents one possible approach to enhancing [e_aq^–] in UV-ARP. The UV photochemistry of many of these e_aq sensitizers was recently reviewed by Cui et al.²

Table 2 provides a summary of e_aq^– sensitizers that have been examined in the literature, their molar absorption coefficient (ϵ) at selected wavelengths, Φ_eaq^– values, and the method of Φ_eaq^– determination. There is some variation in ϵ and Φ_eaq^– and some sensitizers for which no data are available. Reported Φ_eaq^– values for sulfite vary by less than a factor of 2. The Φ_eaq^– value of 0.116 at 254 nm determined by steady-state photolysis⁵⁷ is in good agreement with the recent value of 0.108 at 248 nm measured by transient absorption spectroscopy.⁴⁸ Given the agreement between these two values, we expect that a Φ_eaq^– of ≈0.1 for sulfite is reliable. Uncertainty in Φ_eaq^– values impact the ability of kinetic models to accurately predict the change in [TC] with time according to eq 2.5.

Table 2. Summary of Hydrated Electron (e_aq^–) Sensitizers Used in UV-ARP, Their Quantum Yields (Φ_eaq^–), and the Method of Φ_eaq^– Determination if Available^{62,63,65−67}.

^aReferenced study refers to application of the sensitizer in UV-ARP systems.

^bϵ₂₅₄ estimated using Web Plot Digitizer (https://apps.automeris.io/wpd/).

^cMeasured for this Review.

^dValue is for indole (not indoleacetic acid) in ethanol.

^eValue is for indole (not indoleacetic acid) at 30 °C.

^fMeasurements are conducted in deaerated solutions.

To help address the uncertainty in ϵ₂₅₄ values, we have measured molar absorption coefficients for 3-indoleacetic acid (IAA), nitrilotriacetic acid (NTA), iodide, and sulfite (Figure 7), four of the e_aq^– sensitizers employed most frequently in UV-ARP studies. Our results agree well with most with literature data, where available. For sulfite, ϵ₂₅₄ values agree within ∼15%, and we suggest that the value measured by Tenorio et al. (17.8 M^–1cm^–1) is reliable.³⁴ Our measured ϵ₂₅₄ for iodide (172 M^–1cm^–1) agrees well with one literature report (162 M^–1cm^–1)⁶¹ but is substantially different from another (220 M^–1cm^–1).⁶⁰

Figure 7.

Absorption spectra for four commonly used hydrated electron sensitizers in UV-ARP studies, including 3-indoleacetic acid (IAA), nitrilotriacetic acid (NTA), iodide, and sulfite. Molar absorption coefficients at 254 nm (ϵ₂₅₄) are 17.6 M^–1cm^–1 (sulfite), 26.3 M^–1cm^–1 (NTA), 172 M^–1cm^–1 (iodide), and 2576 M^–1cm^–1 (IAA) and are listed in Table 2. Experimental conditions: All solutions besides iodide (potassium iodide) were prepared in pH 10 borate buffer (10 mM), whereas potassium iodide was prepared in unbuffered laboratory grade water (18.2 MΩ·cm). Sulfite solutions were prepared in N₂ sparged (30 min) water. Molar absorption coefficients are derived from a linear fit of absorbance to molar concentration with an absorbance (optical density) less than 2.0 measured with a Cary-100 Bio dual beam spectrophotometer in a 1 cm path length quartz cuvette.

2.3.1. Sulfite

Sulfite is one of the most studied e_aq^– sensitizers in UV-ARP systems. Sulfite absorbs in the UV region of the spectrum (ε₂₅₄ ≈ 18 M^–1cm^–1)⁵⁷ and its Φ_eaq^– has been measured in several previous studies^48,57,68

2.13

In the absence of oxygen, dithionate (S₂O₆^2–) and sulfate are the major end products of this reaction, while in oxygenated solutions only sulfate is observed.⁶⁸

e_aq^– generation in the UV/sulfite ARP is much less efficient at pH values near and below the pK_a of HSO₃^– (7.19) because photolysis of HSO₃^– has a low Φ_eaq^–.⁶⁹ In addition, HSO₃^– is a scavenger of e_aq.⁴⁸

2.14

Moreover, sulfite itself reacts with e_aq^–. Although the SO₃^2–-e_aq reaction is slow (the upper limit for the rate constant is 1.6 × 10⁶ M^–1 s^–1),⁸ the high sulfite concentrations typically employed in UV-ARP (ca. 10 mM) could mean that this reaction pathway is significant. For example, Figure 8 shows the calculated e_aq^– formation rate (R_f), scavenging capacity (k_S,t=0^′), and [e_aq] expected in an aqueous solution containing monochloroacetate (MCAA at a fixed concentration of 20 μM) and sulfite at various concentrations (between 0 and 20 mM). The e_aq^– scavenging capacity exerted by sulfite increases linearly with increasing sulfite concentration (eq 2.9). However, R_f increases asymptotically due to the (1–10^{–(ε_sens[sens]_t + α_t)l}) term in eq 2.4. Overall, this results in an asymptotic increase in [e_aq^–] with increasing sulfite concentration. Although there is not a consensus on an optimal sulfite concentration in UV-ARP, previous literature demonstrates diminishing increases in target contaminant transformation kinetics above ∼10 mM sulfite.^13,23,34 This is consistent with the calculations presented in Figure 8.

Figure 8.

Plot of calculated e_aq^– formation rate (R_f,t=0), scavenging capacity (k_S,t=0^′), and [e_aq]_t=0 () as a function of initial sulfite concentration. It is assumed that sulfite functions as the sole light absorber while both sulfite and another target compound (MCAA) act as e_aq^– scavengers. Conditions: I₀ = 1.13 × 10^–8 Es cm^–2 s^–1, Φ_eaq^– = 0.116 mol Es^–1, ε_sens = 18.14 M^–1 cm^–1, l = 2.85 cm, k_{sulfite,eaq^–} = 1.3 × 10⁶ M^–1 s^–1, k_MCAA,eaq^– = 1.0 × 10⁹ M^–1 s^–1, [MCAA]₀ = 20 μM, pH₀ = 9.5.

2.3.2. Iodide

Iodide photolysis produces e_aq^– with a Φ_eaq^– of ≈0.2 at 248 nm.⁵⁹ This is approximately a factor of 2 higher than the Φ_eaq^– for sulfite. Iodide also has a higher molar absorption coefficient at 254 nm than sulfite (Table 2). However, as will be discussed further below, several UV-ARP studies have demonstrated that target contaminant transformation rates are faster in the UV/sulfite system compared to the UV/iodide system. This indicates that the [e_aq] is higher in the UV/sulfite system, despite iodide’s higher Φ_eaq^– and ϵ₂₅₄ values. The lower [e_aq^–] in the UV/iodide system is likely due to e_aq scavenging reactions^70,71 of I-containing species, like iodine and triiodide, which are formed during photooxidation of iodide.

2.15

2.16

2.17

2.3.3. Nitrilotriacetic Acid

Nitrilotriacetic acid (NTA) produces e_aq^– at UV irradiation wavelengths, but with an unknown quantum yield. It has been proposed that the mechanism of e_aq formation from nitrilotriacetic acid photolysis does not involve photoionization of nitrilotriacetic acid (eq 2.1).⁵⁶ Instead, electronically excited nitrilotriacetic acid is thought to sensitize the photooxidation of water, concomitantly producing e_aq^–, ^•OH, and ^•H.^56,64

2.18

2.19

Photoproduced ^•OH can react with the C–H bonds in nitrilotriacetic acid through H atom abstraction (eq 2.20),⁶⁴

2.20

which reduces the ^•OH-e_aq^– recombination rate (k = 2.8 × 10¹⁰ M^–1 s^–1)⁷² thereby allowing more e_aq to react with target contaminants.

Historically, the mechanism proposed for nitrilotriacetic acid is an alternative to the photooxidation mechanism previously described (eq 2.12) and may in fact be operative for other e_aq^– sensitizers. The main point is that both mechanisms (photoionization and NTA-sensitized H₂O ionization) yield e_aq.

2.3.4. Indole derivatives

It has been suggested that indoles produce e_aq^– through the photoionization mechanism described in eq 2.12 with Φ_eaq^– of ≈0.1 (Table 2). Given the high molar absorption coefficient of 3-indoleacetic acid at 254 nm (≈ 2600 M^–1 cm^–1) and Φ_eaq^– value comparable to that of sulfite, indoles should have high rates of e_aq formation in UV-ARP systems. However, previous studies of UV/indoleacetic acid treatment of PFOA show no difference in degradation rates between direct photolysis controls and in the presence of 1 mM 3-indoleacetic acid.²⁵ This observation could indicate a high rate of scavenging between e_aq^– and indole itself (k = 2.6 × 10⁸ M^–1 s^–1)⁴⁹ or indole photoproducts that limits the [e_aq] available for reaction with PFOA. As discussed later (see Section 3.2.3), indoleacetic acid does effectively defluorinate PFOA in a heterogeneous UV-ARP involving an organomontmorillonite clay surface.^25,26

2.3.5. Vacuum UV

An alternative to the use of e_aq^– sensitizers is vacuum UV photons (VUV, 100 nm < λ < 200 nm), which are capable of directly ionizing water molecules to form e_aq, H^•, and ^•OH.⁷³ For example, studies have shown that photolysis of aqueous PFAS with Hg lamps not manufactured to block VUV degrades these contaminants in the absence of any added e_aq^– sensitizer,^24,74 implying that photoionization of water by the small amount of VUV emitted from these lamps is occurring. Practically, VUV is challenging to implement, as VUV photons are absorbed within the first ∼5 mm of water.⁷³

2.3.6. Factors Influencing the Choice of Chemical Sensitizer

Several studies have attempted to optimize experimental conditions to maximize the rate of e_aq^– generation from these sensitizers. For example, Yu et al. demonstrated that reductive dechlorination of monochloroacetate in the UV/iodide system is enhanced by adding sulfite.⁶⁰ In this case, it is believed that sulfite reduces reactive iodine species (e.g., I₂, I₃^–) back to iodide, whereas in the absence of sulfite, these reactive iodine species significantly scavenge e_aq (eqs 2.15–2.17). In the Yu et al. study, monochloroacetate dechlorination was fastest in the UV/sulfite/iodide system, when compared to UV/iodide and UV/sulfite.⁶⁰ Similarly, Sun et al. observed faster degradation of perfluorooctanesulfonate by UV/nitrilotriacetic acid when compared to UV/sulfite.⁶⁴ In this case, nitrilotriacetic acid was hypothesized to concomitantly sensitize the formation of e_aq^– and scavenge photoproduced ^•OH (eq 2.18–2.20).

Another important aspect of selecting an e_aq^– sensitizer for UV-ARP is the sensitizer’s expected degradation products. For example, iodide photolysis generates iodine and triiodide,⁷⁵ which absorb light in the visible region of the spectrum. These iodine species would need to be removed (e.g., through reduction to iodide) after the UV-ARP. Photolysis of nitrilotriacetic acid produces a foul odor characteristic of amine compounds.³² Conversely, sulfite photooxidation produces inert sulfate. Although high concentrations of sulfate in drinking water should be avoided,⁷⁶ this anion is relatively benign compared to iodide and nitrilotriacetic acid photolysis products. Another benefit to using sulfite is that it can be regenerated following its oxidation to SO₃^•-. This is inferred from the fact that sulfite persists in UV-ARP systems remain high even after many hours of light exposure.³⁴ Conversely, when used alone, sensitizers like iodide and nitrilotriacetic acid are degraded quickly.⁶⁴

Collectively, sensitizer concentration, e_aq^– quantum yield, and molar absorption coefficient at the actinic wavelength employed all impact the rate of e_aq formation and are important for sensitizer selection in UV-ARP. Byproducts of e_aq^– sensitizer photolysis are also an important consideration. Taken as a whole, using sulfite as a sensitizer in UV-ARP appears to have the optimal benefits for treatment applications.

2.4. Knowledge Gaps

Recent UV-ARP studies have employed an array of e_aq^– sensitizers, varying intensity UV light sources,^77,78 and optimized solution conditions⁷⁹ in an attempt to enhance degradation rates of target contaminants. However, progress in UV-ARP research has been limited by the lack of a systematic framework for evaluating treatment performance between these individual studies and the effect of process variables such as the UV-ARP irradiation source, sensitizer identity and concentration, and water quality. Overall, there is a need for the development of quantitative tools to allow comparison between UV-ARP studies, such as methods for quantifying [e_aq] and modeling target contaminant degradation kinetics.

2.4.1. Probe Compounds for Studying UV-ARP

There is a need to apply e_aq^– probe compounds to quantify photochemical parameters related to UV-ARP systems. Although several e_aq probes have been used in the broader chemical literature (vide infra), and some have been applied in UV-ARP studies, they have not been used extensively to quantify e_aq^– formation rates (R_eaq^–), e_aq^– concentrations ([e_aq]), or e_aq^– scavenging capacities (k_S,t) in UV-ARP. This contrasts with research on UV-AOP, which has made extensive use of p-chlorobenzoic acid as an ^•OH probe,^40,80,81 allowing for a comparisons of treatment performance.⁸²⁻⁸⁵ When measured using a validated probe compound, R_f^e_aq–, [e_aq], and k_S,t^′ provide metrics for assessing the effects of different water matrices, solution conditions, and process variables on UV-ARP treatment performance.

e_aq^– probe compounds used in the literature include chloroacetate (monitoring the [parent] by ion chromatography or gas chromatography),⁵⁷ chloroethanol (monitoring the product [Cl^–] by ion chromatography or an ion selective electrode),⁸⁶ nitrous oxide (monitoring the [N₂] by gas chromatography), and 3-amino-2,2,5,5-tetramethyl-1-pyrrolidinyloxy (monitoring the derivatized reaction product).⁸⁷

Chloroethanol⁸⁶ and the nitroxide probe⁸⁷ were used for characterizing e_aq^– formation from chromophoric dissolved organic matter at UV–B and UV-A wavelengths. However, to be a successful e_aq probe in UV-ARP, the nitroxide would need to be stable under UV–C irradiation (e.g., at 254 nm, where the nitroxide is highly absorbing). Another issue with the nitroxide probe is that it requires millimolar concentrations of dimethyl sulfoxide, which absorbs at 254 nm. For these reasons it is unlikely that the nitroxide probe is suitable for quantifying [e_aq^–] in UV-ARP. Because chloroethanol relies on measurement of [Cl^–], which has a high background concentration in real-world waters, it is also unsuitable for quantifying [e_aq] in UV-ARP. However, chloroethanol could be employed to measure fundamental photochemical parameters like Φ_eaq^– in laboratory grade water containing low background [Cl^–].

A few studies have demonstrated success using chloroacetate as an e_aq^– probe.^57,60,88,89 Chloroacetate can be quantified via ion chromatography in aqueous aliquots withdrawn from UV-ARP experiments down to ∼2 μM.⁵⁷ Lower concentrations can be quantified via derivatization and gas chromatography.⁹⁰ The availability of these analytical methods has important implications for the chosen probe compound concentration. If ion chromatography is the only method available, this limits the initial probe concentration to the micromolar range, where it can still exert a significant scavenging capacity in the system.

When employing chloroacetate as an e_aq^– probe, it is also possible to quantify chloride as an e_aq reduction product, but this would again be impractical in natural waters with high background [Cl^–]. Alternatively, the use of fluoroacetate as an e_aq^– probe would enable detection of fluoride since the fluoride concentration in natural waters is typically orders of magnitude lower than that of chloride.⁵² Fluoroacetate is much less reactive with e_aq (k ≤ 1.2 × 10⁶ M^–1 s^–191) than chloroacetate (k = 1.0 × 10⁹ M^–1 s^–1). Thus, higher concentrations of fluoroacetate could be employed (aiding with detection limit issues) with minimal impacts to the natural e_aq^– scavenging capacity of the background water matrix. Use of fluoroacetate as an e_aq would require accurate measurement of its e_aq^– bimolecular rate constant (the current value is an upper limit) and the fluoride reaction yield (see Section 3.2.2).

While development of e_aq^– probe compounds is needed, several challenges unique to UV-ARP have made this difficult. First, many UV-ARP studies are conducted in buffered solutions of laboratory-grade water where there is minimal background scavenging of e_aq. This makes the system sensitive to potential e_aq^–-reactive impurities like trace levels of O₂ not removed during prior deaeration.⁹² Another potential challenge for a standard e_aq probe compound is the possibility that reactive species besides e_aq^– generated in UV-ARP systems could cause probe transformation. For example, SO₃^•– formed upon sulfite photooxidation has been reported to act as both an oxidant and reductant (see also section 3.3.2).¹³ If SO₃^•– were to reduce the employed probe compound like e_aq, then the probe would not be selective. One way to overcome this challenge is to verify the selectivity of the probe using an e_aq^– quencher. Compounds like nitrous oxide, nitrate, and nitrite have been widely employed for this purpose in prior studies.^8,57,60 For example, N₂O reacts quickly with e_aq (k = 9.1 × 10⁹ M^–1 s^–1) but not with SO₃^•– (k < 1 × 10⁶ M^–1 s^–1) (Table 1 and https://kinetics.nist.gov/solution/).

2.4.2. Computer Kinetic Models

An alternative approach for gaining a quantitative description of UV-ARP chemistry is by modeling the kinetics of the system through computer numerical integration, which has received considerable attention in UV-AOP systems.^93,94 The process involves collecting all the known reactions occurring in a system and their rate constants and inputting this information into a computer program. Free (e.g., Kinetcus⁹⁵) and commercial programs (e.g., FACSIMILE) are available for building and running these kinetic models. If all parameters are known, running such programs to produce time profiles of target contaminants, intermediates, and reactive radicals is straightforward. Alternatively, if there are some unknown parameters (e.g., bimolecular rate constants, quantum yields) the kinetic model can be fit to a given data set to estimate the unknown parameters.^10,96 Despite the abundance of bimolecular e_aq^– rate constants available,⁸ examples of such numerical computer models being applied in UV-ARP are limited.⁶⁰

One potential limitation of computer kinetics models is that they may not be readily transferable across different UV-ARP. They may also fail to adequately describe experimental data in complex reaction systems where it is simply not possible to know all the reactions.⁹⁷ For these reasons, computer kinetics models may be most beneficial for estimating unknown parameters for a specific UV-ARP and understanding the impact of design and process variables on treatment performance.

2.4.3. The Role of Dissolved Organic Matter

While some prior research has evaluated the impact of dissolved organic matter (DOM) on target contaminant transformation in UV-ARP, more work is needed to fully understand the role of DOM in e_aq^–-based treatment systems. While it is expected that DOM is an important e_aq scavenger, specific bimolecular rate constants are not yet established. Dissolved oxygen, nitrate, and nitrite, which all exhibit high reactivity with e_aq^–, will likely dominate the e_aq scavenging at low e_aq^– exposures (Figures 5 and 6). However, given the variety of structural moieties present, DOM may exert a significant e_aq scavenging capacity long into the UV-ARP process. In addition, the impact of e_aq^– exposure on DOM light absorption in UV-ARP systems needs to be further evaluated, although prior studies indicate that various reduction methods (borohydride,⁹⁸ electrochemical,⁹⁹ and UV/sulfite¹²) decreases DOM absorbance. Changes in DOM absorbance at the actininic wavelength(s) will impact the rate of light absorption by the sensitizer and therefore the rate of e_aq formation in the system.

3. UV-ARP Substrate Scope

Previous research has evaluated the suitability of UV-ARP for degradation of both inorganic and organic contaminants. Target contaminants that have been considered for UV-ARP treatment typically have a high oxidation state central atom (e.g., N in NO₃^–) or one or more carbon–halogen bonds as part of their structure. Indeed, previous computational studies showed that bimolecular e_aq^– reaction rate constants were inversely proportional to the energy of the lowest unoccupied molecular orbital (E_LUMO) for a large set (>100) of substituted aliphatic compounds.^100,101

As part of this Review, we have compiled literature data from UV-ARP treatment studies of important inorganic and organic contaminants. Table 3 presents a summary of these UV-ARP studies separated by compound class, which includes details like lamp power and type (low-, medium-, or high-pressure Hg), e_aq^– sensitizer, solution conditions, and degradation rates (final % degraded and reaction time). A searchable copy of Table 3 including additional details (e.g., average UV fluence rate and sparging conditions) is included in the Supporting Information. The sections below highlight key results from this analysis for oxyanions, halogenated organics, and PFAS.

Table 3. UV-ARP Organic and Inorganice Treatment Studies Summary^a.

ref	compound class	[compound]0 (μM)	heterogeneous or homogeneous system	UV lamp	UV sensitizer	water characteristics	parent compound degradation % (time)
(102)	halogenated organic	10 μM trichloroacetic acid	heterogeneous (TiO2)	LP-Hg 10 W	4.0 mM formate	ultrapure water (pH 6.5; 23 °C)	100% (150 min)
(45)	halogenated organic	10 μM 4-bromophenol	homogeneous	LP-Hg 10 W	1.0 mM sulfite	ultrapure water (pH 9.2; 20 °C)	100% (10 min)
(103)	halogenated organic	20 μM 1,2-dichlororoethane	homogeneous	LP-Hg 36 W	1.0 mM dithionite; 1.2 mM sulfite; 0.4 mM sulfide; 4.0 mM ferrous iron	ultrapure water (pH 11, 11, 9, 9; 25 °C)	100% (varies)
(104)	halogenated organic	10 μM decabromodiphenyl ethers	homogeneous	LP-Xe 50 W and LED	none	reaction performed in methanol	90% (60 min)
(105)	halogenated organic	10 μM MCAA	homogeneous	n.r.	1.0 mM sulfite; 1.0 g L–1 bismuth oxyiodide	ultrapure water (pH 7; 25 °C)	100% (20 min)
(57)	halogenated organic	50 μM MCAA	homogeneous	LP-Hg 10 W	1.0 mM sulfite	ultrapure water (pH 9.2; 25 °C)	100% (15 min)
(88)	halogenated organic	50 μM MCAA	homogeneous	LP-Hg 15 W	0.2 mM phenol	ultrapure water (pH 12; 25 °C)	100% (6 min)
(60)	halogenated organic	50 μM MCAA	homogeneous	LP-Hg 10 W	1.0 mM sulfite; 0.1 mM iodide	ultrapure water (pH 9.2; 25 °C)	100% (5 min)
(89)	halogenated organic	50 μM MCAA	homogeneous	LP-Hg 10 W	1.0 mM sulfite (UPW); 2.0 mM sulfite (river)	ultrapure water; Yangtze River, China (pH 9.2; 25 °C)	100% (15 min in ultrapure water; 75 min in river water)
(106)	halogenated organic	300 μM Diclofenac	homogeneous	LP-Hg (n.r. W)	8.0 mM sulfite	ultrapure water (pH 9.2; room temp)	98% (30 min)
(107)	halogenated organic	600 μM 2,4-dichlorophenol	homogeneous	LP-Hg (n.r. W)	8.0 mM sulfite	ultrapure water (pH 10; n.r °C)	99.45% (40 min)
(39)	halogenated organic	0.6 mg L–1 vinyl chloride	homogeneous	LP-Hg 36 W	20 mg L–1 sodium dithionite; 12.5 mg L–1 sodium sulfite; 8.4 mg L–1 sodium hydrosulfide; 49 mg L–1 ferrous sulfate	ultrapure water (pH 3,7,10; n.r. °C)	varies
(108)	halogenated organic	1 mg L–1 vinyl chloride	homogeneous	LP-Hg 36 W	0.16 mM sulfite	ultrapure water (pH 9; 34 °C)	100% (90 min)
(109)	halogenated organic	10 mg L–1 difluoroacetic acid	homogeneous	LP-Hg 8 W	1.0 mg L–1 iodide	ultrapure water (pH 9; 20 °C)	n.r.
(110)	halogenated organic	10 mg L–1 trichloroethylene	homogeneous	LP-Hg 36 W	10.0 mM sulfite	ultrapure water (pH 10; 20 °C)	100% (60 min)
(46)	halogenated organic	250 mg L–1 2, 4, 6-trichlorophenol	homogeneous	LP-Hg (n.r. W)	6.2 mM sulfite	ultrapure water (pH 7; 20 °C)	100% (80 min)
(111)	halogenated organic	30 mg L–1 Acid Yellow 17	homogeneous	LP-Hg 6 W	1.5 mM dithionite	ultrapure water (pH 9; n.r. °C)	90% (5 min)
(112)	halogenated organic	5 mg L–1 tetrachloromethane	homogeneous	LP-Hg 14 W	10.0 mM sulfite	ultrapure water (pH 9; n.r. °C)	100% (5 min)
(113)	halogenated organic	50 mg L–1 pentachlorophenol	homogeneous	LP-Hg 16 W	100 mg L–1 dithionite; 100 mg L–1 phenol	ultrapure water (pH 7; n.r. °C)	96% (180 min)
(114)	halogenated organic	2 μM iopamidol	homogeneous	LP-Hg 10 W	0.5 mM sulfite	ultrapure water (pH 7; 20 °C)	80% (30 min)
(115)	organic	21.16 μM carbamazepine	homogeneous	LP-Hg (n.r. W)	1.6 mM sulfite	ultrapure water (pH 8.4; room temp)	100% (10 min)
(116)	organic	30 μM dinitrotoluene	homogeneous	LP-Hg 10 W	0.05 mM toluene	ultrapure water (pH 7.5; 25 °C)	100% (20 min)
(20)	oxidized transition metal	5 mg L–1 Cr(VI)	homogeneous	MP-Hg 300 W	5.0 mM sulfite	ultrapure water (pH 5–10.5; 20 °C)	100% (20 min with pH 5)
(21)	oxidized nonmetal	10 mg L–1 Se(VI)	heterogeneous (Fe(III) coagulation)	MP-Hg 300 W	5.0 mM sulfite	ultrapure water (pH 9.2; 20 °C) (HA DOM spike)	99% (20 min without HA); 72% (20 min with HA)
(19)	oxyanion	100 μM bromate	heterogeneous (Fe(II)–Ti(IV) LDH Nanoparticles)	LP-Hg 15 W	0.5 g L–1 Fe(II)–Ti(IV)	tap water (pH 7; 25 °C)	90% (1 h)
(16)	oxyanion	4 μM bromate	homogeneous	LP-Hg 8 W; MP-Hg 150 W	varying sulfite	ultrapure water (pH varies; 26 °C)	100% (varies)
(14)	oxyanion	6.6 μM bromate	heterogeneous (superparamagnetic BiOCl)	LP-Hg 10 W	4.0 mM formate	ultrapure water (pH 7; 25 °C)	79% (15 min)
(117)	oxyanion	10 μM bromate	homogeneous	LP-Hg 10 W	1.0 mM sulfite/0.1 mM iodide	ultrapure water (pH 9.2; 20 °C)	100% (5 min)
(15)	oxyanion	39 μM bromate	homogeneous	LP-Hg 15 W	0.78 mM sulfite	ultrapure water (pH 9.8; n.r. °C)	86% (4 h)
(18)	oxyanion	100 μM bromate	homogeneous	LP-Hg 15 W; MP-Hg 100 W	2.0 mM sulfur(IV)	ultrapure water; tap water (pH 9; 25 °C)	82% (20 min in tap water); 99% (20 in ultrapure water)
(6)	oxyanion	2.75 μM bromate	homogeneous	LP-Hg 36 W	25.6 mg L–1 sulfite	ultrapure water (pH 7, n.r. °C)	100% (125 min in LP)
(118)	oxyanion	100 μM bromate	homogeneous	LP-Hg 10 W	1.0 mM sulfite	ultrapure water (pH 7; 23 °C) (HA DOM spike)	20% (40 min)
(119)	oxyanion	160 μM nitrate	homogeneous	LP-Hg 36 W	8.4 mM sulfite	ultrapure water (pH 9; 27 °C)	100% (20 min)
(113)	oxyanion	50 mg L–1 nitrate	homogeneous	LP-Hg 16 W	100 mg L–1 dithionite; 100 mg L–1 phenol	ultrapure water (pH 7; n.r. °C)	94% (180 min)
(11)	oxyanion	50 mg L–1 nitrate	homogeneous	LP-Hg (n.r. W) (254 and 185 nm)	none	ultrapure water (pH 6; 25 °C)	99% (80 min)
(120)	oxyanion	2 μM nitrate	homogeneous	LP-Hg 9 W	60 mM sulfite	ultrapure water (pH 9.2; 25 °C)	100% (90 min)
(12)	oxyanion	100 μM perchlorate	homogeneous	LP-Hg 36 W	11.0 mM sulfite	ultrapure water (pH 11; n.r. °C) (SRHA DOM spike)	10% (25 h, 50 mg/L SRHA) 80% (25 h, 0 mg/L SRHA)
(13)	oxyanion	100 μM perchlorate	homogeneous	LP-Hg 36 W	3.76 mM sulfite	ultrapure water (pH 11; 30 °C)	20% (10 h)
(64)	PFAS	10 μM PFOS	homogeneous	LP-Hg 14 W	2.0 mM nitrilotriacetic acid	ultrapure water (pH 10; 30 °C)	85.4% (10 h)
(121)	PFAS	20 μM PFOA	homogeneous	LP-Hg 15 W	0.6 mM iodide	ultrapure water (pH 9.2; 20, 25, 40 °C)	98% (6 h), 99% (6 h), 100% (6 h)
(28)	PFAS	24 μM PFOA	heterogeneous (self-assembled micelle with cationic surfactant)	HP-Hg 500 W	1.0 mM indoleacetic acid	ultrapure water (pH 4–8; 25 °C)	90% (2.5 h)
(22)	PFAS	25 μM PFOA	homogeneous	LP-Hg 15 W	0.3 mM iodide	ultrapure water (pH 9; room temp)	95% (6 h)
(122)	PFAS	25 μM PFOA	homogeneous	LP-Hg 15 W	0.25 mM iodide	ultrapure water (pH 5, 7, 10; n.r. °C)	9.1% (2 h), 74.6% (2 h), 97.9% (2 h)
(123)	PFAS	25 μM PFCA	homogeneous	LP-Hg 18 W	10.0 mM sulfite	ultrapure water (pH 12; 20 °C)	varies
(33)	PFAS	25 μM PFCAs, PFSAs, fluorotelomers	homogeneous	LP- Hg 18 W	10.0 mM sulfite	ultrapure water (pH 12; 20 °C)	varies (4 h)
(78)	PFAS	32 μM PFOS	homogeneous	HP-Hg 250 W	10.0 mM sulfite	ultrapure water (pH 9.2; 25 °C)	98% (0.5 h)
(124)	PFAS	37.2 μM PFOS	homogeneous	MP- Hg 500 W	none	ultrapure water (pH 7.4; 90 °C)	90% (11 days)
(77)	PFAS	38.7 μM PFOA	homogeneous	HP-Hg 250 W	10.0 mM sulfite	ultrapure water (pH 9.2; 25 °C)	100% (10 min)
(25)	PFAS	10 mg L–1 PFOA; 10 mg L–1 PFOS	heterogeneous (organo-montmorillonite)	LP-Hg 36 W	1.0 mM 3-indoleacetic acid	ultrapure water (pH 3–11; 25 °C)	100% (5 h)
(1)	PFAS	10 mg L–1 PFOS, PFOA, PFHS, PFBS, PFHA, PFBA	homogeneous	LP-Hg 8 W	10.0 mM iodide	ultrapure water (pH 6–8; n.r. °C)	5–30% (150 min)
(34)	PFAS	Varying AFFF (15 PFAS)	homogeneous	LP-Hg 18 W	10.0 mM sulfite	ultrapure water (pH 9.5; 20 °C)	varies
(23)	PFAS	PFOA	homogeneous	LP-Hg 10 W	10.0 mM sulfite	ultrapure water (pH 10.3; 25 °C)	100% (1 h)
(30)	PFAS	25 μM PFCAs	homogeneous	LP-Hg 18 W	10.0 mM sulfite	ultrapure water (pH 9.5; 20 °C)	100% (8–12 h)
(31)	PFAS	25 μM PFECAs	homogeneous	LP-Hg 18 W	10.0 mM sulfite	ultrapure water (pH 9.5; 20 °C)	varies
(125)	PFAS	30 μM PFOA	homogeneous	LP-Hg 14 W	0.3 mM iodide	ultrapure water (pH 10; 25 °C) (HA DOM spike)	67.5% (1.5 h)
(58)	PFAS	30 μM PFOS	homogeneous	LP-Hg 14 W	0.3 mM iodide	ultrapure water (pH 10; 25 °C) (HA DOM spike)	86% (1.5 h)
(24)	PFAS	37.2 μM PFOS	homogeneous	LP-Hg 10 W (254 and 185 nm)	20 mM sulfite	ultrapure water (pH 10; 25 °C) (HA DOM spike)	97.3% (4 h without HA); 85% (4 h with HA)
(27)	PFAS	24 μM PFOA	heterogeneous (self-assembled micelle with cationic surfactant)	LP-Hg 36 W	1.0 mM indole-acetic acid	ultrapure water (pH 6; 25 °C) (SRHA and SRFA DOM spike)	100% (6 h)
(29)	PFAS	24 μM PFOA	homogeneous	LP-Hg 36 W	1.0 mM indole-3-acetic acid; 1.0 mM indole-1-acetic acid; 1.0 mM indole-3-carboxylic acid; 1.0 mM indole; 1.0 mM skatole; 1.0 mM donaxine	ultrapure water (pH 6; 25 °C)	35.7% (8 h); 31.4% (8 h); 24.4% (8 h); 100% (8 h); 100% (8 h); 100% (8 h)
(26)	PFAS	24.1 μM PFOA	heterogeneous (organo-montmorillonite)	LP-Hg 36 W	1.0 mM 3-indole-acetic acid; 1.0 mM indole; 1.0 mM 3-indole-carboxylic acid; 1.0 mM 3-methylindole	ultrapure water (pH 6; 25 °C) (SRHA DOM spike)	100% (5 or 8 h with DOM spike; 3-indole-acetic acid)

^an.r. is used to indicate if information was not reported by authors. UV sensitizer specified as “none” refers to vacuum UV studies.

3.1. Oxyanions

UV-ARP technologies have been studied for the degradation of oxyanions such as NO₃^–, NO₂^–, ClO₄^–, and BrO₃^–,^6,10−19 which contain a highly oxidized central atom. The reaction mechanism involves e_aq^– addition to the central atom and loss of O^•- (the conjugate base of ^•OH). The conversion of bromate to bromite is shown in Figure 9 as an example.

Figure 9.

Reaction of e_aq^– with bromate (BrO₃^–) to form bromite (BrO₂^–) and O^•-. Although shown as a concerted process, a one-electron adduct (BrO₃^2–) may be formed prior to Br–O bond homolysis.

Many oxyanions have been shown to be degraded quickly by e_aq^– generated in UV-ARP (e.g., NO₃^–, NO₂^–, BrO₃^–),^16,18,120 while ClO₄^– reacts more slowly.^12,13,126 Conversion of oxyanions to benign end products requires further reduction of intermediate products.^119,127 For example, NO₃^– reacts with e_aq to produce NO₂^– which is reduced further through additional reactions with e_aq^– to eventually produce N₂.

3.1

Given sufficient e_aq^– exposure, it may be possible to quantitatively produce the free halides of halogen-containing oxyanions.^13,15,18 For example, bromite is completely degraded during UV/sulfite treatment and yields bromide as the only product. Conversely, even at higher [SO₃^2–], higher light intensity, and longer reaction times, ClO₄^– treatment by UV/sulfite treatment has not achieved 100% abatement. However, Vellanki et al. showed that perchlorate degradation during UV/sulfite treatment was significantly enhanced at pH 12 (∼75% abatement) relative to pH 9 (∼25% abatement⁷). In the case of ClO₄^–, Cl^– and ClO₃^– were the only transformation products observed. This suggests that the rate limiting step is either e_aq reduction of ClO₄^– or ClO₃^–, which is consistent with very slow e_aq^– bimolecular rate constants for these species (k < 10⁶ M^–1 s^–1) and the much faster values for ClO₂^– and ClO^– (>10⁹ M^–1 s^–1). Consequently, entries in Table 3 reveal huge differences in UV-ARP treatment effectiveness for BrO₃^– and ClO₄^–, consistent with the fact that ClO₄^– is frequently employed as an “inert” electrolyte.

The reactivity of oxyanions with e_aq^– depends on several key factors. The affinity for oxyanion reduction by e_aq is based primarily on the availability of an electron orbital offered by the central atom, not the compound’s one-electron redox potential.¹²⁷ For example, NO₂^– reacts with e_aq^– more slowly (k = 3.5 × 10⁹ M^–1 s^–1)¹²⁸ than NO₃^– (k = 9.7 × 10⁹ M^–1 s^–1),⁸ which is opposite of what one would expect from the NO₂^–/NO redox couple¹²⁹ (E° = 0.46 V) vs (NO₃^–/NO₂^– of 0.10 V¹³⁰). The oxyanion’s molecular geometry and charge of the central atom can further impact their reduction by e_aq. For example, NO₃^– is more reactive with e_aq^– than ClO₄^– (k < 1.6 × 10⁶ M^–1 s^–1).¹²⁹ While the Cl atom in ClO₄^– is at a higher oxidation state (+7) than N in NO₃^– (+5), the slower reactivity of e_aq may be attributed to the planar structure of NO₃^–, facilitating e_aq^– interaction versus the inhibiting tetrahedral structure of ClO₄^–.¹²⁷

Additional radical species can also contribute to oxyanion degradation occurring in UV-ARP. For example, Vellanki and Batchelor proposed that SO₃^•- is the main reactive species responsible for the reduction of ClO₄^– in the UV/sulfite system.¹³ In this case, it was proposed that SO₃^•- abstracts an oxygen atom from ClO₄^– to form the sulfate radical and chlorate.¹³ Similarly, the carbon dioxide radical anion (CO₂^•-) was proposed to be the main reducing agent responsible for BrO₃^– and NO₃^– reduction in the UV/formate system.^10,14 Moussavi and Shekoohiyan discovered that hydrogen atoms (H^•) are responsible for a small percentage of NO₃^– reduction under VUV irradiation (185 nm radiation wavelength).¹¹ In addition, H^• are the main reactive species responsible for BrO₃^– degradation under acidic conditions^6,18 due to increased formation of H^• from the reaction of H⁺ and e_aq^– (Table 1).

3.1.1. Other Inorganic Species

UV-ARP have also been studied for the removal of oxidized metal ions such as chromium(VI) and nonmetals like selenium(IV).^20,21 The e_aq^– reduction lowers the oxidation state of these contaminants, which decreases their solubility and often induces precipitation. For example, UV/dithionite treatment completely removed selenite through this combined reduction/precipitation process after ∼120 min.²¹ In the case of chromium, UV/sulfite treatment between pH 5 and 10 effectively reduced Cr(VI) to Cr(III), followed by precipitation of Cr(III) salts.²⁰ While other approaches can also be used for reductive precipitation of Cr(VI) and Se(IV), reduction by e_aq occurs readily at neutral and basic pH, while other reductants are limited by the decreased reduction potential of Cr(VI) and Se(IV) with increasing pH.

3.2. Organic Compounds

In addition to inorganic contaminants, a number of studies have evaluated UV-ARP for the degradation of halogenated contaminants. Below, we discuss UV-ARP based treatment for halogenated organic compounds, including chlorinated and brominated contaminants. A separate section is devoted to PFAS given the currently high level of interest in e_aq^–-based treatment for these compounds. For readers interested in further understanding the reactivity of e_aq with organic compounds beyond halogenated contaminants, the book chapter by Anbar entitled “The Reactions of Hydrated Electrons with Organic Compounds” provides a thorough discussion of e_aq^– reactivity with various organic functional groups.¹³¹

3.2.1. Halogenated Organic Compounds

The reactivity of organic compounds with e_aq^– in terms of bimolecular rate constants has been studied for many decades.^8,131,132 Bimolecular rate constants for e_aq with organic compounds vary widely depending on compound class and the presence of halogen atoms (Figure 10). For example, aliphatic compounds bearing only N and O atoms, regardless of the functional group, have substantially lower reactivity than aliphatic compounds also bearing halogen atoms. Halogen atom substitution on carbonylic and aromatic compounds also generally increases their reactivity with e_aq^–. An exception is that, for nonaromatic compounds, substituting −H with −F tends to decrease the reactivity with e_aq (e.g., acetone vs fluoroacetone, Figure 10A). For aromatic compounds, substituting −H with −F slightly increases its reactivity with e_aq^– and this is further enhanced when multiple −F are substituted (e.g., benzene, fluorobenzene, 1,4-difluorobenzene, Figure 10B).

Figure 10.

Bimolecular rate constants for reaction of hydrated electron with nonaromatic (A) and aromatic (B) compounds. Rate constants are from Li et al.¹⁰⁰ and are tabulated by compound class in the Supporting Information. Compound classes are differentiated by the presence of O, N, and X = F, Cl, Br, I atoms. The number of compounds in each class is represented by n. Selected compounds are shown within each class to compare trends in reactivity.

3.2.1.1. Haloaliphatic Compounds

In a homologous series of halogenated aliphatic compounds, the reactivity with e_aq^– typically follows the trend F ≪ Cl < Br < I.¹³¹ For example, the relative reactivity of fluoroacetate, chloroacetate, bromoacetate, and iodoacetate is 0.002 < 1.0:6.2:12.0 (Figure 11).⁸ This trend demonstrates that the C–X bond is more reactive than the carboxyl group.⁹¹ Indeed, acetate’s reactivity with e_aq (k = 1.1 × 10⁶ M^–1 s^–1) is comparable to that of fluoroacetate.⁸

Figure 11.

Relative reactivity of monosubstituted haloacetate ions with hydrated electron in terms of their bimolecular reaction rate constant.

Reductive dehalogenation of haloaliphatic compounds is thought to occur via a two-step reaction, in which a reduced intermediate, RX^•-, is first formed which subsequently undergoes unimolecular dissociation:¹³¹

3.2

3.3

with the products of the second step being a carbon centered radical (R^•) and halide (X^–). The fate of the carbon centered radical depends on the water matrix composition. Since UV-ARP seek to maintain low [O₂], the formation of peroxyl radicals is not expected to be a significant sink of R^•. As an example, the following reaction pathway for UV/sulfite transformation of chloroacetate in laboratory grade water has been established (eqs 3.4–3.7⁵⁷)

3.4

3.5

3.6

3.7

where the percent yields in parentheses are based on the observed rates of product formation relative to the rate of chloroacetate degradation. Chloride is formed at near 100% yield. Several additional carbon-containing products were also observed with varying yields, including succinic acid (formed from coupling of two ^•CH₂COO^– radicals), sulfoacetic acid (radical coupling of SO₃^•- and ^•CH₂COO^–), and acetate (possibly due to reduction of ^•CH₂COO^– by SO₃^•-).

For haloaliphatic compounds, UV-ARP studies have largely focused on haloacetic acids and vinyl chloride.^{39,57,60,89,102,108} Direct photolysis is typically an unimportant pathway for haloaliphatic compounds. Liu et al. studied the dechlorination of trichloroacetic acid using UV/titanium dioxide with formate and suggested that both CO₂^•– and e_aq^– reactions were occurring.¹⁰² Other dechlorinated products such as di- and monochloroacetic acid are formed during the reaction between trichloroacetic acid and e_aq. Vinyl chloride can be degraded by UV-ARP through both direct photolysis (due to its unsaturated carbon–carbon bond) and reaction with radical species.^39,108 In the UV/sulfite ARP, Liu et al. found that the rate of vinyl chloride dechlorination was inversely proportional to solution pH.¹⁰⁸ At pH 9, UV/sulfite treatment resulted in ∼88% dechlorination. Since e_aq^– is the main product of sulfite photolysis at this pH (see Section 2.3), this result suggests that at pH 9 the major mechanism of vinyl chloride degradation is reductive dechlorination induced by e_aq. Conversely, at pH 7, only ∼18% of the total [Cl^–] possible was produced. Liu et al. proposed that the low dechlorination of vinyl chloride at pH 7 was due to the reaction of SO₃^•- or H^• (the products HSO₃^– photolysis) with the C=C bond to produce chloroethane (identified by GC-MS).¹⁰⁸

3.2.1.2. Haloaromatic Compounds

Haloaromatic compounds exhibit a similar trend in their reactivity with e_aq^– as haloaliphatic compounds, but the fluorinated analogues tend to be more reactive than their aliphatic counterparts. This is demonstrated in Figure 12, which shows the bimolecular reaction rate constants for the aryl halide series. Iodobenzene is approximately 3 orders of magnitude more reactive than the fluorobenzene (like haloacetic acids in Figure 11). However, the difference in reactivity between fluorobenzene and chlorobenzene (factor of ∼30) is less than for the haloacetate ions (factor of ∼1000).

Figure 12.

Relative reactivity of aryl halides with hydrated electron in terms of their bimolecular reaction rate constant.

Halogenated phenols such as 4-bromophenol, 2,4-dichlorophenol, 2,4,6-trichlorophenol, and pentafluorophenol can also be degraded by UV-ARP.^{45,46,104,107,133,134} Direct photolysis contributes to the transformation of these compounds in UV-ARP systems.^45,46 The e_aq^–-induced reaction pathways for halophenols have been studied in detail,^135,136 with halides being produced quantitatively upon reductive electron capture by the aromatic ring.¹³⁶ Phenoxyl radicals are believed to be the initial nonhalogenated products following halide loss.¹³⁵

From a treatment perspective, Xie et al. found that the UV/sulfite ARP completely debrominated 4-bromophenol within 10 min,⁴⁵ and quenching experiments with nitrate and nitrite suggested that this degradation was induced by the reaction with e_aq^–. Similarly, Yu et al. observed that 2,4-dichlorophenol could be degraded with the UV/sulfite ARP, demonstrating that the reaction with e_aq is the main reactive species responsible for its degradation.¹⁰⁷ For fluorinated phenols, Shoute et al. reported that the rate constant between e_aq^– and pentafluorophenol was fast (2.4 × 10¹⁰ M^–1 s^–1) and that a fluoride ion was eliminated to yield a tetrafluorophenoxyl radical.¹³⁴ This study only evaluated the bimolecular rate constant for e_aq reaction with pentafluorophenol and not its degradation or defluorination percentage under steady-state UV-ARP treatment.

3.2.2. Halide Yields

While bimolecular rate constants provide some important information regarding the e_aq^–-based transformation of contaminants in water, the halide ion yield (i.e., mol of halide formed per mol of e_aq-substrate reaction) is also an important factor, as this parameter defines the overall reaction efficiency. The rate of formation of the reduction product (e.g., halide) is given by¹³⁷

3.8

where RfX– is the rate of halide ion formation (M s–1), Rf is the rate of eaq– formation (M s–1), γeaq–,X– is the halide ion yield, and kTC,eaq– and kTC,eaq–q are the bimolecular rate constants (M–1 s–1) for transformation of eaq by the target contaminant and quenching of the target contaminant by eaq–, respectively. eaq bimolecular rate constants are typically measured by directly observing the decay kinetics of the eaq– transient in the near-infrared, which yields the quenching rate constant kTC,eaq–. For radical reactions, it is typically assumed that kTC,eaq–t is equivalent to kTC,eaq–.137

γ_{eaq^–,X^–} values are not readily available for most of the substrates (organic or inorganic) for which UV-ARP are considered a possible treatment technology. However, some halide ion yields are available from the radiation chemistry literature (Figure 13).¹³⁸ For benzyl halides, γ_{eaq^–,X^–} depends strongly on the halogen substituent and solution pH. At acidic pH (4.2), there is a clear trend of decreasing halide ion yield with increasing electronegativity of the halogen substituent. This trend largely vanishes at basic pH, with fluoro-, chloro-, and bromobenzene having near 100% halide ion yields.

Figure 13.

Halide ion yields for aryl halide reactions with hydrated electron.¹³⁸ Yields are calculated by G_X^–/G_eaq^– where G_X^– is μmol X^–/Gy radiation and G_eaq^– is the e_aq^– yield from water radiolysis (0.27 μmol/Gy⁸).

Halide ion yields may also depend on the location of the halogen with respect to other substituents on an aromatic ring (Figure 13). For example, the chloride yield from the reaction of e_aq^– with chloroaniline isomers decreases with an increasing number of sigma bonds between the amino and chloro substituents at circumneutral pH. However, at basic pH, γ_{eaq^–,X^–} are ∼100% for all isomers.

It is also possible to evaluate γ_{eaq^–,X^–} in steady-state photochemical experiments by measuring R_f^X– and R_f. For example, the yield of chloride from reaction with e_aq^– is ∼0.94 based on the rates of chloroacetate degradation (3.76 ± 0.07 μM/min) and chloride formation (3.52 ± 0.01 μM/min) in the UV/sulfite process (see eq 3.4).⁵⁷ The experimental conditions (110 μM chloroacetate, 1 mM sulfite, lab grade water) at which these rates were measured are such that most e_aq should be scavenged by chloroacetate (k_S,t=0^′ = 110 000 s^–1). Thus, for γ_{eaq^–,Cl^–} ≈ 1, the initial rate of chloroacetate degradation should be equal to R_f, as observed.

3.2.3. Per- and Polyfluorinated Alkyl Substances

Perfluoroalkyl substances are “aliphatic substances for which all of the H atoms attached to C atoms in the nonfluorinated substance from which they are notionally derived have been replaced by F atoms, except those H atoms whose substitution would modify the nature of any functional groups present”.¹³⁹ Polyfluoroalkyl substances are “aliphatic substances for which all H atoms attached to at least one (but not all) C atoms have been replaced by F atoms, in such a manner that they contain the perfluoroalkyl moiety C_nF_2n+1”.¹³⁹ These per- and polyfluoroalkyl substances (PFAS) are characterized by extremely strong C–F bonds, which, combined with their properties as surfactant molecules, makes them extremely difficult to remove from, or degrade in, water.¹⁴⁰

UV-ARP have increasingly been proposed as a promising technology for e_aq^–-based transformation of PFAS, although it is interesting to note that Anbar originally described fluorinated aliphatics as “unreactive” toward e_aq due to fluorine’s high electronegativity.¹³¹ Previous reports have demonstrated that it is possible to achieve near 100% reductive defluorination (production of nontoxic fluoride) of some PFAS through UV-ARP.^22,27,30,33 Although studies of PFOA and PFOS are more common, UV-ARP technology has now been evaluated for many PFAS subclasses, including short- and long-chain perfluoroalkanesulfonates (PFSA),^1,33 perfluoroalkylcarboxylic acids (PFCA),^30,31 perfluoroethercarboxylic acids (PFECA),³¹ and fluorotelomer sulfonates and carboxylic acids (Figure 14).³³ Additionally, Tenorio et al. investigated UV/sulfite treatment of aqueous film-foaming foam (AFFF), which contains thousands of PFAS, diluted in sodium carbonate buffer at a ratio of 1:60 000.³⁴ They observed that pseudo-first-order rate constants for degradation of specific PFAS in diluted AFFF were nearly identical to single-compound mixtures in the same sodium bicarbonate buffer system, suggesting that scavenging of e_aq^– by AFFF components did not significantly impact [e_aq].³⁴

Figure 14.

PFAS subclasses referred to frequently in this review, including their acronym and a representative structure.

In one of the earliest UV-ARP PFAS studies, Park et al. investigated the reductive defluorination of perfluorocarboxylic acids and perfluorosulfonates in the UV/iodide system.¹ They found that the headgroup (e.g., −SO₃^– vs COO^–) and −(CF₂)_n– chain length influenced the reductive defluorination kinetics. Specifically, the degradation rate of perfluorosulfonates increased with increasing chain length whereas perfluorocarboxylic acid chain length had little impact on the observed kinetics. Interestingly, Park et al. observed that UV/iodide degradation of PFOS occurred faster than that of PFOA, but several later studies employing UV/sulfite demonstrate an opposite trend in reactivity.^30,34 This result from Park et al. is probably the result their PFAS/KI solutions being at different initial pHs, where the initial pH of PFOS solution was ∼7 while for PFOA it was ∼4.5. Therefore, quenching of e_aq^– by H⁺ in PFOA solution (pH 4.5, see Table 1) is a likely explanation of its slower transformation than PFOS. Conversely, UV/sulfite studies buffer solution at pH > 9 to maintain SO₃^2– as the major S(IV) species,^23,34 which also minimizes e_aq quenching by H⁺.

3.2.3.1. PFAS Transformation Mechanisms

The mechanistic understanding of PFAS reductive defluorination has improved in the past few years. In their investigations of UV/sulfite treatment of PFAS sub classes,^30,31,33 Bentel et al. proposed that reductive defluorination can proceed by one of two main mechanisms, as summarized in Figure 15.

Figure 15.

Proposed reaction pathways for the transformation of perfluorocarboxylic acids (PFCA), perfluorosulfonates (PFSA), and fluorotelomercarboxylic acids or fluorotelomer sulfonates (FTCA + FTSA) by hydrated electron (solid, black lines) and hydroxyl radical reactions (dashed, gray lines). Based on work of Bentel et al.^2,30,79 and Liu, Bentel et al.³³

The first pathway is H/F atom exchange, in which a −CF₂– group is transformed to a −CH₂– group. This reaction takes two e_aq^– and two H⁺ per H/F exchange. For perfluorocarboxylic acids (like PFOA), H/F exchange is thought to occur preferentially at the carbon atom adjacent to the carboxylate headgroup. For perfluorosulfonates (like PFOS), the H/F exchange is thought to occur toward the middle of the chain. This H/F exchange mechanism is supported by the formation of polyfluorinated intermediates identified through nontarget analysis during UV/sulfite treatment of PFAS.³⁰ Computational chemistry results are also supportive of this mechanism. For example, based on spin density isosurfaces, Van Hoomissen and Vyas showed that an excess electron added to PFOA is localized around the carboxylate headgroup whereas for PFOS^•2– the extra electron is distributed throughout the fluoroalkyl chain in a fashion similar to perfluorooctane.¹⁴¹ The location of e_aq reaction during H/F exchange also relates to the positive correlation between chain length and perfluorosulfonate reactivity with e_aq^– reported in several studies.^1,30,34 For example, aqueous vertical attachment energies correlated positively to the chain length of perfluorosulfonates but not perfluorocarboxylic acids.¹⁴¹ Additionally, Bentel et al. demonstrated that perfluorosulfonate C–F bonds with the lowest bond dissociation energy tended to be toward the middle of the perfluoro chain whereas the lowest energy C–F bonds in perfluorocarboxylic acids tended to be adjacent to the carboxylate group.³⁰

The product of H/F exchange is a polyfluorinated substance and UV/sulfite studies show that polyfluorinated substances like fluorotelomercarbxylic acids (FTCA) are less reactive toward e_aq^–.^30,34 Thus, reaction via H/F exchange limits the complete defluorination of PFAS. To address this problem, Liu, Bentel et al. recently demonstrated that greater reductive defluorination of PFAS could be achieved by performing both reductive and oxidative treatments in sequence.³³ These reaction pathways are highlighted in the context of UV-ARP treatment in Figure 15 with additional details relevant to the study of Liu, Bentel et al.³³ shown in Figure 16. In brief, polyfluorinated compounds formed during pH 12 UV/sulfite treatment of perfluorocarboxylic acids or perfluoroalkanesulfonates (H/F exchange) could be further defluorinated by subsequent ^•OH oxidation via high temperature NaOH + K₂S₂O₈. When employing fluorotelomer sulfonates and carboxylates as the starting contaminant, a sequence of oxidation–reduction–oxidation was able to achieve near 100% defluorination for most chain lengths (Figure 16).

Figure 16.

Reaction scheme for the oxidation–reduction–oxidation treatment scheme developed by Liu, Bentel et al. for the near complete defluorination of fluorotelomer carboxylates and sulfonates.³³

The second reaction pathway is decarboxylation-hydroxylation-elimination-hydrolysis (DHEH) and is thought to occur only for perfluorocarboxylic acids. The DHEH mechanism is proposed to shorten the perfluoroalkyl chain by one −CF₂– group in each step. The decarboxylation leads to a perfluorinated alcohol, which is susceptible to HF elimination. The DHEH mechanism results in chain shortening and so-called “deeper” levels of defluorination than H/F exchange.³⁰ Note that perfluorosulfonates do not undergo DHEH directly but could do so after desulfonation, which has been proposed to explain the formation of PFOA and other perfluorocarboxylic acids during UV/sulfite treatment of PFOS.³⁰ The mechanism is supported in part by the significantly faster kinetics and greater defluorination percentage of perfluorocarboxylic acid abatement during UV/sulfite treatment at pH 12 relative to pH 9.5.¹²³ Perfluoroether carboxylic acids (e.g., GenX) were proposed to undergo a similar reduction pathway, with cleavage of the C–O ether bond helping to enhance formation of the unstable perfluoroalcohol.³¹

Collectively, the reductive defluorination mechanisms discussed above suggest that the abatement of parent PFAS and ultimate fluoride production is determined by more than just the availability of e_aq^–. The reactivity of PFAS transformation products with e_aq appears to be an important controlling factor. Thus, biasing solution conditions to favor the DHEH reaction pathway (e.g., high pH) could enhance reductive defluorination.

3.2.3.2. Optimizing PFAS Treatment by UV-ARP

Multiple studies have investigated the conditions needed to optimize PFAS treatment by increasing its e_aq^– exposure.^{1,2,22−25,58,64,121,122,124,125} For example, UV irradiation of sulfite, iodide, nitrilotriacetic acid, and several indoleacetic acid derivatives have all been shown to degrade PFOA or PFOS through reductive defluorination by e_aq, all with varying kinetic rates.^22,23,25,64 However, no single study has compared all these e_aq^– sensitizers under equivalent photochemical conditions. Sun et al. demonstrated that the UV/nitrilotriacetic ARP led to faster defluorination of PFOS than UV/sulfite at an equivalent concentration (2 mM), but it is important to note that nitrilotriacetic acid has a higher molar absorption coefficient (ε₂₅₄ = 26.3 M^–1 cm^–1 at pH 10⁶⁴) at the actinic wavelength (254 nm) than sulfite (ε₂₅₄ ≈ 18 M^–1 cm^–157) (see Table 2). In addition, Yu et al. showed that combining sulfite and iodide into one system yielded faster rates of reductive dechlorination of a monochloroacetate system⁶⁰ (this UV/sulfite/iodide system has not yet been applied to PFAS degradation). The UV/indoleacetic acid system is ineffective at degrading PFOA in bulk, aqueous solution (see further below).²⁵

Additional treatment conditions studied for PFOA and PFOS degradation includes the initial solution pH, temperature, and the presence of e_aq^– scavenging species like O₂, NO₃^–, NO₂^–, and DOM.^{1,2,22−24,58,121,122,124,125} Higher temperatures resulted in faster degradation kinetics for PFOA and PFOS, with apparent activation energies of ∼50 and 25 kJ/mol, respectively, in the UV/iodide system.^121,124 Qu et al. demonstrated the major scavenging impact of O₂ on PFOA degradation; when continuously sparging solutions with O₂, the percent defluorination decreased from ∼98% (under N₂ saturation) to less than 5% (O₂-saturated solution).²² However, Bentel et al. showed that initial sparging of reaction solutions with N₂ made no difference in the removal kinetics of PFOA,³⁰ which could be explained by O₂ removal through reaction with sulfite, either through a thermal⁵¹ or photochemical process⁶⁸ at early UV-ARP treatment times (see also Figure 6).

Gu et al. found that PFOS degradation in the UV/sulfite ARP was significantly inhibited by humic acid (≥2.64 mgC/L) and NO₃^– (≥0.5 mM NO₃^–), whereas HCO₃^– had minimal impact at a concentration of 1.0 mM.²⁴ The decreased reactivity of PFOS in the UV/sulfite system in the presence of humic acid (a surrogate for DOM) could be due to the following three mechanisms: (1) shielding of the e_aq^– sensitizer from UV light by DOM chromophores, (2) DOM molecules acting as a scavenger of e_aq, or (3) through the production of e_aq^–-quenching reactive oxygen species (e.g., ^•OH, ¹O₂, or H₂O₂) by DOM photolysis.¹⁴² Interestingly, in the UV/iodide system,⁵⁸ Aldrich Humic Acid enhanced the degradation and defluorination rate of PFOS when present at ≤10 mg/L, but it had an inhibitory effect at 30 mg/L. Sun et al. provided several hypotheses for this enhancement,⁵⁸ including the possibility that phenolic electron donors reduce photochemically formed reactive iodine species (I₂, I₃^–) back to I^–, thereby decreasing the e_aq scavenging by these species (eq 2.15–2.17). Although DOM also generates e_aq^–, this pathway is likely an insignificant source of e_aq in UV-ARP due to its low quantum yield (Φ_eaq^– ∼ 10^–5).¹⁴³ Further research is needed to identify the role of DOM in UV-ARP treatment of PFAS, especially in terms of e_aq^– scavenging and potential DOM-PFAS interactions.^58,144

One of the more significant factors influencing PFAS degradation efficiencies is solution pH. Multiple studies have reported a clear trend of increasing degradation rates and improved defluorination percentages with increasing pH for perfluorocarboxylic acids and perfluorosulfonates,^2,122 with parent contaminant abatement defluorination percentages being especially rapid at pH 12.^33,79 The [H⁺] influences multiple aspects of UV-ARP systems, including e_aq^– production from the sensitizer (see section 2.3), e_aq scavenging (see Table 1 and eq 2.14), and secondary chemistry following the initial PFAS-e_aq^– reduction event. Scavenging of e_aq by H⁺ does not contribute significantly to at pH 7 or higher (Figure 4). At circumneutral pH, HSO₃^– quenching reaction may be the main e_aq^– scavenger in UV/sulfite ARP,⁴⁸ where scavenging by H⁺ should be minimal. In addition, HSO₃^– has a low quantum yield of e_aq generation, so e_aq^– formation rates in the UV-sulfite system will be reduced with decreasing pH. Finally, it is interesting to observe the pH dependence of γ_{eaq^–,X^–} values for halobenzenes and chloro-substituted anilines (Figure 12), in light of the discussion above on the effect of pH on the DHEH mechanism for reductive PFAS defluorination.

3.2.3.3. Nonhomogeneous UV-ARP

So far, we have discussed UV-ARP studies in which e_aq^– and PFAS are homogeneously distributed throughout aqueous solution. An alternative approach that has received attention recently is to enhance PFAS transformation by creating a microheterogeneous system.²⁵⁻²⁸ In principle, these microheterogeneous reactors enhance the rate of reductive defluorination relative to the homogeneous system by bringing the e_aq and PFAS in closer proximity, thereby increasing the frequency of e_aq^–-PFAS reactions (Figure 17).

Figure 17.

Illustration of the microheterogeneous reductive defluorination of perfluorooctanoic acid (PFOA) via a cationic micelle with a negatively charged hydrated electron sensitizer.

Reported UV-ARP microheterogeneous technologies include UV/organomontmorillonite clay and UV/micelle systems, both of which use indoleacetic acid as the e_aq^– sensitizer.^25,26,28,29 The UV/organomontmorillonite clay system works by first adsorbing indoleacetic acid and PFAS on the cationic clay surface.^25,26 Tian et al. demonstrated that the UV/organomontmorillonite clay system allows indoleacetic acid to generate e_aq and radical cations within the clay layer.²⁵ This system increased the reductive defluorination of PFOA from approximately 20% (PFOA and indoleacetic acid only, no different from direct photolysis control) to 90% (PFOA, indoleacetic acid, and organomontmorillonite clay) in 10 h. The authors suggested that the clay system further stabilized the radical cations produced from 3-indoleacetic acid and thus minimized radical cations recombination with e_aq^–.²⁵ In the UV/micelle system, a self-assembled micelle was created by a cationic surfactant (e.g., cetyltrimethylammonium bromide), indoleacetic acid (the e_aq sensitizer), and PFOA.²⁸ The self-assembled micelle increased the reductive defluorination of PFOA from approximately 10% (PFOA and indoleacetic acid only) to 90% (PFOA, indoleacetic acid, and cationic surfactant) in 2.5 h. Both microheterogeneous technologies demonstrate enhanced reductive defluorination kinetics when compared to the homogeneous UV/sensitizer system. Both technologies also reduce the e_aq^– scavenging impacts of H⁺, allowing treatment over a wider range of pH conditions (pH 4 to 8). Chen et al. found that the UV/micelle system could still be effective for PFOA transformation in the presence of high ionic strength and low concentrations of fulvic acid (1 mg/L).²⁷

A related microheterogeneous approach is the combination of electrochemical and UV methods for ARP degradation of PFAS. Recent work by Rao et al. demonstrated that PFAS defluorination by e_aq^– is enhanced by PFAS sorption to carbon nanotube electrodes at low applied potentials (−0.58 V vs Ag/AgCl).⁷⁴ It was hypothesized that this reaction occurs via a two-electron process in which the first electron (derived from the cathode) reduces the C–F bond strength, thereby allowing the second electron (from VUV photolysis of water) to more effectively defluorinate the PFAS than UV treatment alone. In this combined electrochemical/photochemical ARP, PFAS defluorination exhibited a chain-length dependence similar to UV-ARP, with degradation rates of perfluorocarboxylic acids being relatively independent of chain length compared to perfluorosulfonates.^1,30,34

While these microheterogeneous reactors appear promising for enhanced PFAS treatment, additional research is needed to investigate the application of these systems in actual surface-and groundwater matrices, especially given the impact of water matrix components (e.g., nitrate, DOM) that could also interact with micelles, minerals, and carbon nanotube electrode materials. It is challenging to compare the treatment effectiveness between microheterogeneous systems as experimental conditions vary significantly. Future research should explore utilizing a probe compound to measure key parameters such as R_f^e_aq–, [e_aq], and k_S,t^′ to compare between systems.

3.3. Knowledge Gaps

Although UV-ARP have been studied for the e_aq^–-based transformation of organic and inorganic compounds has been studied for over a decade, several knowledge gaps are still apparent.

3.3.1. Degradation Byproducts

A question that all degradative advanced water treatment technologies need to address is “are the target contaminants completely mineralized and, if not, are the end products more toxic, less toxic, or just as toxic as initial substrate?” UV-ARP processes also need to address this question. Given the focus on halogen-containing contaminants, the goal of UV-ARP is not complete mineralization but to reduce all halogens to a −1 oxidation state. For inorganic oxyanions, the reaction pathways are well established, with each e_aq^– reaction leading to successive reduction of the central atom; for example, bromate can be reduced all the way to bromide given sufficient e_aq exposure.¹¹⁷ More work is needed to establish the operating conditions required to achieve the e_aq^– exposure necessary for complete transformation of the less-reactive ClO₄^– to Cl^– ion.

Haloaliphatic compounds react by reductive electron capture according to eqs 3.2 and 3.3, the products of which are X^– and a carbon-centered radical (R^•). Halogen atoms will be the main site of e_aq^– reactivity until all C–X bonds are reduced, with the fate of the carbon centered radical being highly dependent on the water matrix. The reaction pathways for haloaromatic compounds are also established, with halides being produced upon reductive electron capture by the aromatic ring.¹³⁶ The secondary chemistry of the aromatic is dependent on additional (nonhalogen substituents). For example, e_aq reduction of chlorophenol is believed to produce phenoxy radicals.¹³⁵ Nonhalogenated organic products (e.g., phenol) tend to be less reactive with e_aq^– and thus more persistent following dehalogenation of the initial contaminant. For halogenated aromatic and aliphatic compounds, a remaining knowledge gap is the ultimate fate of nonhalogenated end products.

As highlighted herein, recent reports have demonstrated that a variety of PFAS structures (see Figure 14) can be degraded with near 100% defluorination by sequential UV-ARP and oxidation reactions. Continued research efforts are needed to evaluate how these proposed treatment schemes are impacted by water matrix components such as dissolved organic matter.

3.3.2. Role of Additional Reactive Species in UV-ARP Systems

Prior literature indicates that reactive species other than e_aq^– may play an important role in contaminant degradation in UV-ARP.^{22,108,145−149} For example, SO₃^•- was hypothesized to be responsible for the transformation of vinyl chloride at acidic pH, resulting in a lower percentage dechlorination than obtained using e_aq.¹⁰⁸ This observation is consistent with the view of Neta and Huie¹⁵⁰ who suggested that SO₃^•- may act as a mild oxidant (eq 3.9)

3.9

or strong reductant (eq 3.10)

3.10

where E is the reduction potential vs NHE. Equation 3.10 demonstrates that SO₃^•- has a potential comparable to that of e_aq^– (−2.9 V). Based on computational chemistry estimates of one-electron reduction potentials of perfluoropentanoic acid and perfluorobutanesulfonate made by Van Hoomissen and Vyas (E < −2.0 V),¹⁴¹ reductive defluorination of PFAS by SO₃^•- should be thermodynamically feasible. Future research should evaluate this possibility.

The superoxide radical anion (O₂^•-) is another reactive species of interest in e_aq^–-based UV-ARP¹⁴⁵ since this species is formed by reaction of O₂ with e_aq. There is disagreement in the literature about the role of O₂^•- in the degradation of perfluorocarboxylic acids (like PFOA^146−148), with some reports suggesting that O₂^•- can displace fluorine in C–F bonds yielding F^–. Yet other studies demonstrate that O₂^•- is completely unreactive with PFOA.¹⁴⁹ The latter conclusion is consistent with the low oxidation potential of O₂^•- (E = 0.16 V) and low nucleophilicity of O₂^•- in water.¹⁵¹ The UV-ARP studies employing e_aq^– surveyed in this Review indicate that O₂^•- is not involved in the reductive defluorination of PFAS. If it were, then significant PFAS abatement should be expected in UV-ARP systems containing dissolved oxygen. However, this is not the case; for example, Qu et al. observed minimal changes in PFOA concentration when continuously sparging O₂ (which should favor O₂^•- formation) in the UV/iodide ARP over a 10 h irradiation period.²²

3.3.3. Availability of Hydrated Electron Bimolecular Rate Constants

Bimolecular rate constants for e_aq^– and many of the organic and inorganic contaminants listed in Table 3 are widely available. There are at least two circumstances in which this is not the case. First, some e_aq reactions with contaminants (e.g., perchlorate), background scavengers (e.g., bicarbonate), or e_aq^– sensitizers (e.g., sulfite) have reported rate constant values that can be considered only as upper limits. This situation arises when the species of interest is “unreactive” toward e_aq, which makes the transient absorption monitoring of e_aq^– more subject to errors from e_aq reactive impurities present in either the solvent used (i.e., pure water) or the added chemical solutes.⁸ The importance of these slow-reacting species in the total scavenging capacity will increase with increasing e_aq^– exposure as more reactive e_aq scavengers are abated. Thus, the lack of solid e_aq^– bimolecular rate constants for these species limits estimation of the scavenging capacity under these conditions.

Second, although many studies have demonstrated reductive defluorination of diverse PFAS structures by UV-ARP, absolute bimolecular rate constants for e_aq^–-PFAS reactions are limited. For the few values that have been reported, there is substantial variability between different studies, even when using the same experimental methods (Table 4). For example, Huang et al. using laser flash photolysis and potassium ferricyanide (K₄Fe(CN)₆) as an e_aq sensitizer reported bimolecular rate constants for e_aq^– and PFCA of ∼10^6–7 M^–1 s^–1 (1.7 × 10⁷ M^–1 s^–1 for PFOA).¹⁵² A recent report by Maza et al. employed the same e_aq sensitizer and detection method but measured much faster rate constants, not only for perfluorocarboxylic acids (6.4 × 10⁸ M^–1 s^–1 for PFOA) but also for perfluorosulfonates (2.4 × 10⁹ M^–1 s^–1 for PFOS).¹³² A pulse-radiolysis study measured PFOA-e_aq^– and PFOS-e_aq bimolecular rate constants of 5.1 × 10⁷ and 7.3 × 10⁷ M^–1 s^–1,¹⁵³ respectively. The pulse radiolysis values are faster than the value measured by Huang et al.¹⁵² but still an order of magnitude slower than measured by Maza et al.¹³²

Table 4. Summary of Literature Hydrated Electron (e_aq^–) Bimolecular Rate Constants for Per- And Polyfluoroalkyl Substances (PFAS).

compd	acronym (chain length)	biomolecular rate constant, k (M-1s-1)	method	ref
trifluoroacetic acid	TFA (2C)	1.9 × 106	LFPa,b	(152)
		5.0 × 108	LFPa	(132)
perfluoropropanoic acid	PFPrA (3C)	5.8 × 108	LFPa	(132)
perfluorobutanoic acid	PFBA (4C)	7.1 × 106	LFPa,b	(152)
		5.4 × 108	LFPa	(132)
perfluoropentanoic acid	PFPA (5C)	5.2 × 108	LFPa	(132)
perfluorohexanoic acid	PFHxA (6C)	5.4 × 108	LFPa	(132)
perfluoroheptanoic acid	PFHpA (7C)	7.5 × 108	LFPa	(132)
perfluoroocatanoic acid	PFOA (8C)	1.7 × 107	LFPa	(152)
		5.1 × 107	PRc	(153)
		7.1 × 108	LFPa	(132)
perfluorononanoic acid	PFNA (9C)	6.4 × 108	LFPa	(132)
perfluorobutanesulfonate	PFBS (4C)	1.8 × 109	LFPa	(132)
perfluorohexanesulfonate	PFHxS (6C)	1.7 × 109	LFPa	(132)
perfluorooctanesulfonate	PFOS (8C)	2.4 × 109	LFPa	(132)
		7.3 × 107	PRc,d	(153)

^aLaser flash photolysis, K₄Fe(CN)₆ as e_aq^– sensitizer, transient absorption monitoring of e_aq.

^bCorrected to zero ionic strength.

^cPulse-radiolysis generation of e_aq^–, transient absorption monitoring of e_aq.

^dMeasured at concentration greater than critical micellular concentration.

It is difficult to understand the wide variability in these reported values between different studies using similar experimental approaches. This was noted by Maza et al. in their recent study,¹³² who commented that the reason for the discrepancy was not immediately apparent and probably not due to differences in ionic strength, which tend to produce only small differences in e_aq^– bimolecular rate constants within the ranges studied (see also Figure 3). Maza et al. did point out that Huang et al. did not report their range of [PFOA] used in their study, which could have facilitated a closer comparison. It is interesting to note, however, that the e_aq bimolecular rate constant measured for trifluoroacetic acid by Huang et al. (1.9 × 10⁶ M^–1 s^–1) agrees with the upper limit reported previously by Anbar and Hart (<1.4 × 10⁶ M^–1 s^–191), whereas Maza et al.’s value is 2 orders of magnitude faster (5.0 × 10⁸ M^–1 s^–1). One potential reason for this scatter could be the level of residual [O₂] in these solutions; as noted previously, some dissolved oxygen remains no matter what deaeration techniques are used. This [O₂] might not have been consistent between studies. Based on the fast rate constant of O₂ reduction, even micromolar levels will dominate the PFAS-e_aq^– chemistry. Alternatively, the PFAS purity might be different, with even trace impurities impacting the PFAS reduction chemistry. However, the disparate values for trifluoroacetic acid suggest that the variability in e_aq bimolecular rate constants for longer chain PFAS are not caused by the presence of branched isomers, which tend to be more reactive with e_aq^–.³⁴

Practically, if values from the recent report by Maza et al.¹³² are correct and e_aq^– rate constants are on the order of 10⁸–10⁹ M^–1 s^–1, then it is difficult to understand the slow degradation kinetics of PFAS in UV-ARP systems. These data suggest that there is a need to resolve the uncertainty in absolute e_aq bimolecular rate constants for PFAS.

For PFAS, recommended approaches to measuring e_aq^– bimolecular rate constants, such as laser flash photolysis or pulse radiolysis, are only possible for those compounds that have high-purity sources at high concentrations. This represents a small number of the thousands of relevant PFAS structures that may be present in contaminated waters. Another consideration is the isomeric form of the compound of interest. Branched PFAS isomers are known to be much more reactive with e_aq,³⁴ and varying percentages of branched isomers have been quantified in commercially available PFAS reagents.¹⁵⁴ If branched and linear isomers are present, then measured e_aq^– bimolecular rate constants will represent an average of individual values weighted by the fractional composition of each isomer.

One path forward apart from experimental methods is the use of computational chemistry tools for the prediction of e_aq^– bimolecular rate constants. This could include the use of quantitative structure–activity relationships (QSARs) to relate known e_aq rate constants to molecular descriptors calculated by quantum chemical techniques. Li et al. recently employed this approach and found that the energy of the lowest unoccupied molecular orbital (E_LUMO) was a good predictor of k_TC,eaq^–.^100,101 Similarly, for hydroxyl radical reactions, density functional theory (DFT)-based methods have been used by Minakata et al. to calculate the aqueous-phase free energies of activation,¹⁵⁵ which in turn were correlated to measured bimolecular reaction rate constants through linear free energy relationships. This latter approach has been utilized extensively for UV-AOP but has not yet been applied to UV-ARP, highlighting this possible area for future research.

3.3.4. Fluoride Yields during Reductive Defluorination of PFAS

The efficiency of defluorination of PFAS by e_aq^– is limited by three main factors: (1) the amount of e_aq generated for the reaction, (2) the fraction of e_aq^– reacting with PFAS relative to water matrix species, and (3) the yield of fluoride per e_aq-PFAS reaction event. This latter probability can be described as a fluoride ion yield. Despite the importance of these yields, no data have been reported for e_aq^–-PFAS reactions. Acquiring these data are important, because the fluoride ion yield describes the intrinsic limits of e_aq-based PFAS transformation efficiency. It is important to note that the commonly reported percent defluorination (number of F^– produced per total F in parent PFAS) is different from the fluoride yield as it does not account for the reaction efficiency. This is illustrated by the equations below for the two terms. For % defluorination (deF%)

3.11

where [F^–] is the molar fluoride concentration at the end of the UV-ARP, [PFAS]₀ is the initial molar concentration of the parent PFAS, and N_C–F is the number of C–F bonds in the parent compound. When using steady-state gamma radiolysis, the fluoride yield is calculated as

3.12

where G_F^– and G_eaq^– are the radiolysis yields of fluoride (unknown) and e_aq^– (0.27 μmol Gy^–18), respectively. The assumption in eq 3.12 is that all e_aq formed in the system are scavenged by the PFAS. This typically requires high concentrations of reagents (millimolar) so that all radiolytically formed e_aq^– are scavenged by the added substrate, in which case it is important that the reagents are free as possible from e_aq-reactive impurities. To isolate e_aq^– reactions, solutions can be spiked with an alcohol such as tert-butanol (to scavenge ^•OH), sparged with N₂ (to limit the e_aq-O₂ reaction), and aliquots withdrawn at varying time intervals (irradiation doses) to measure the concentration of target contaminant and/or transformation product(s).

Apart from radiolysis methods, it may be possible to estimate γ_{eaq^–,F^–} from UV/sulfite PFAS transformation data. Using this approach, γ_{eaq^–,F^–} can be calculated from the ratio of the e_aq^– formation rate (R_f,t=0) and rate of fluoride ion formation (R_f,t=0^F–), the latter of which can be obtained from the commonly reported deF%). We used the PFOA deF% data from Bentel et al.³⁰ to calculate R_f,t=0 and then calculated R_f,t=0^e_aq– from Figure 8 (similar reactors, sulfite concentrations, and lamp wattages were used) to estimate γ_{eaq^–,F^–} for PFOA (eqs 3.13–3.15):

3.13

3.14

3.15

The low fluoride ion yield calculated here for PFOA is based on several assumptions (e.g., all e_aq^– are scavenged by 25 μM PFOA) and future research measuring fluoride yields from PFAS of diverse structure is needed. Further, given the pH-dependence of halide ion yields for halobenzenes and haloanilines (Figure 13), it is possible that a similar trend of increasing yield with increasing pH could be occurring for PFAS defluorination. Acquiring these yields for different PFAS would enhance our understanding of the reductive defluorination chemistry.

4. Synthesis of Existing Information to Optimize UV-ARP Treatment

Development of an optimized UV-ARP system for e_aq^–-based contaminant transformation is needed to advance these technologies. The above sections have reviewed literature on e_aq generation (in terms of the specific sensitizer) and e_aq^– scavenging in UV-ARP systems. The following optimized experimental conditions were common to all substrates reviewed.

• Sensitizer selection: Sulfite is the most used sensitizer and appears to have the optimal benefits for UV-ARP treatment. A concentration of ∼10–15 mM likely provides the maximum level of e_aq^– generation

• pH conditions: Maintaining a pH above 7 is highly desired to mitigate the impact of scavenging by H⁺ while maintaining at least 2 pH units above 7 when using sulfite as a sensitizer is essential to eliminate the scavenging impact of HSO₃^–.⁴⁸ UV/sulfite treatment at pH 11–12 seems especially rapid for PFAS and perchlorate, and it would expedite treatment if pH adjustment is feasible.¹²³

• Minimizing scavengers: For substrates with a fast-biomolecular rate constant (≥10⁹ M^–1 s^–1), it is important to remove dissolved oxygen and minimize other scavengers with fast biomolecular rate constants (e.g., NO₃^–, NO₂^–). Additional research is necessitated to optimize UV-ARP treatment for real-world waters with DOM and to evaluate the impacts of changing water background matrix conditions on substrate degradation.

Recent PFAS treatment literature has focused on the electrical energy per order (E_EO) as a metric for comparing degradative technologies.^32,156E_EO is defined as the energy needed to reduce the concentration of contaminant by 1 order of magnitude in 1 m³ of water.¹⁵⁷ Comparing E_EO values between different UV-ARP systems in this way reveals the differences in efficiency of these processes. For example, although the microheterogeneous UV/indoleacetic acid/micelle system reported recently was described as “efficient,” a very high-powered lamp (500 W) was used to treat only 10 mL of solution,²⁸ resulting in a high E_EO (Table 5). A recent study by Liu et al. combined nanofiltration (NF) with UV-sulfite treatment at the pilot scale and found an E_EO of 3.7 kWh m^–3 for PFOA,³² which is approximately 4-fold lower than the E_EO for PFOA in a bench-scale photochemical system (15.8 kWh m^–3) reported by Bentel et al.^30,31 The much lower value reported by Liu et al. results from the fact that they considered the total volume of water treated in the NF-UV/sulfite process, which was approximately an order of magnitude greater than the volume treated by UV-sulfite. E_EO values for PFOA treatment via UV-ARP in Table 5 tend to be lower than most other degradative technologies.¹⁵⁶ UV-ARP at pH 12.0 is comparable in efficiency to electrochemical oxidation; for example, electrochemical oxidation with a Ti₄O₇ electrochemical membrane yielded E_EO values of 5.1 and 6.7 kWh m^–3 for PFOA and PFOS, respectively.¹⁵⁶

Table 5. Electrical Energy per Order (E_EO) for the Degradation of Perfluorooctanoic Acid (PFOA) for Different UV-ARP.

technology	EEO (kWh/m3)a	ref
UV/sulfite, pH 12.0, 18 W, 600 mL, kobs = 4.4 h–1 (bench)	15.8	(79)
UV/sulfite, pH 11.2, 1320 W, 568 L, kobs = 1.5 h–1 (pilot)	3.7	(32)
UV/indoleacetic acid/micelle, 500 W lamp, 10 mL treated, kobs = 1.7 h–1 (bench)	67 600	(28)

^aE_EO calculated as E_EO = 2.303/k_obs × power/volume, where k_obs (h^–1) is the first-order rate constant, power is the lamp power in kW, and volume is the treated volume in m³.

Minimizing the background scavenging of e_aq^– by water matrix components will enhance [e_aq] and thereby increase the rate of target contaminant degradation. Such minimization may only be important when the treatment time scale is short (less than a few hours). For example, e_aq^–-based transformation of PFAS occurs over a period of >12 h during which fast-reacting e_aq scavengers (e.g., O₂, NO₃^–) will be abated quickly through their reaction with e_aq^–. Additional background scavengers with slow bimolecular rate constants, such as HCO₃^– and CO₃^2–, may become more impactful for PFAS treatment in groundwater systems with high alkalinity, as these compounds become the main scavengers once the fast-reacting e_aq scavengers are removed. Liu et al. recently employed nanofiltration (NF) to remove PFAS from contaminated groundwater while concomitantly producing a concentrated waste stream that was treated by UV-sulfite.³² The advantage of NF compared to other membranes like reverse osmosis is that NF does not substantially concentrate ions like nitrate and nitrite.¹⁵⁸ Thus, concentrating PFAS in a small-volume waste stream while at the same time not concentrating e_aq^– scavengers would effectively treat more mass per unit time than if the UV-ARP were applied to the contaminated groundwater directly.

Another way to optimize the rate of target contaminant degradation in UV-ARP is to increase the rate of e_aq^– formation. This can be done by selecting sensitizers with a higher e_aq quantum yield or by increasing the fraction of incoming photons absorbed by the sensitizer. Most e_aq^– sensitizers absorb in the UV region of the spectrum, where background absorbance by DOM is significant. However, an alternative approach developed by Goez and co-workers is to utilize visible-absorbing e_aq sensitizers.¹⁵⁹ These visible-absorbing e_aq^– generation systems utilize low-cost green light emitting diodes, a ruthenium catalyst ([Ru(bpy)₃²⁺]), and ascorbate (Asc^2–) as the electron donor. Following absorption, [Ru(bpy)₃²⁺] rapidly forms a triplet metal-to-ligand charge transfer state that oxidizes Asc^2–. The one-electron reduction product of [Ru(bpy)₃²⁺] is encapsulated in a cationic micelle, which facilitates lifetimes long enough to undergo another green photon absorption, thereby producing e_aq and reforming ground state [Ru(bpy)₃²⁺]. This photocatalyst system demonstrates one possible route toward a more energy-efficient generation of e_aq^– by using a green light emitting diode; however, it is not practical for water treatment at this stage given the need for a transition-metal-containing catalyst and self-assembled cationic micelle. Future research efforts to generate visible-absorbing e_aq sensitizers could improve the energy efficiency of UV-ARP.

5. Outlook and Recommendations for Future Research

UV-ARP technologies offer a unique advantage to treating recalcitrant chemical contaminants in water that are resistant to UV-AOP or other oxidative processes. Numerous UV-ARP studies have demonstrated effective degradation of oxyanions, high valence transition metals, disinfection byproducts, and PFAS. Yet, the utilization of these technologies in real-world waters is limited. Based on this review of the UV-ARP research, we suggest the following research needs and opportunities to advance UV-ARP research and clarify the role of this technology in water treatment. Note that these research needs are discussed in depth in sections 2.4 and 3.3.

• Create standard methods for measuring e_aq^-in UV-ARP. Future research should focus on developing a probe compound capable of quantifying photochemical parameters important in UV-ARP, such as [e_aq^–], background e_aq scavenging capacity, and rates of e_aq^– formation. The deployment of a probe compound will provide researchers a way to compare treatment effectiveness between studies and optimize treatment in real-world waters. Measuring the [e_aq] will also facilitate modeling target contaminant transformation rates in UV-ARP.

• Understand the role of DOM. The impact of DOM on the e_aq^– scavenging capacity and light absorbing capacity in UV-ARP systems needs further investigation. Future research should aim to determine e_aq biomolecular rate constants for DOM and the impact of e_aq^– exposure on DOM’s absorbance. Addressing these research gaps will help to evaluate the impact of DOM on UV-ARP treatment of real-world waters where DOM is an important e_aq scavenger and light absorber.

• Clarify the role of additional reactive species in UV-ARP. Studies have suggested that SO₃^•- contributes to target contaminant transformation in the UV/sulfite system, for example, perchlorate and nitrate. Future research should explore whether reactions with SO₃^•- contribute to the reductive defluorination of PFAS. This reaction is thermodynamically favorable based on the reduction potentials of SO₃^•- (experimental)¹⁵⁰ and perfluorobutanesulfonate (computational).¹⁴¹ Direct, time-resolved kinetic measurements of bimolecular rate constants between PFAS and SO₃^•- could help answer this question. A computational assessment of the reactivity between PFAS and SO₃^•- may also provide insight.

• Resolve the uncertainty in e_aq^-bimolecular rate constants for PFAS. There is an abundance of bimolecular rate constants for e_aq^– and various organic and inorganic contaminants. PFAS are an exception to this, however, due the large number of PFAS and the difficulty of obtaining high quantities of pure compounds needed for e_aq kinetic experiments. In addition, available PFAS-e_aq^– rate constants in the literature are highly variable (k = 10⁷–10⁹ M^–1 s^–1 for PFOA; see Table 4), and thus, the intrinsic reactivity of PFAS with e_aq is still a major knowledge gap. Future work should investigate this discrepancy, potentially by incorporating the reaction yield (eq 3.12), as has been done previously for halogenated aromatic compounds (e.g., Figures 11 and 12). We suggest that direct, time-resolved methods such as laser flash photolysis and pulse radiolysis be employed for measuring e_aq^– bimolecular rate constants given the complex background photochemistry occurring with e_aq sensitizers. In addition, computational chemistry approaches should be explored to discern structural trends in biomolecular rate constants with the more than thousands of relevant PFAS structures that may be present in contaminated waters.

• Determine PFAS fluoride yields. Quantifying the fluoride ion yield for e_aq^–-PFAS reactions is important, as it describes the intrinsic limits of e_aq-PFAS transformation efficiency. Equation 3.12 demonstrates how these values can be obtained from steady-state gamma radiolysis experiments. While steady-state gamma radiolysis is our recommended method for measuring these values, other experimental approaches may be possible. For example, the preliminary calculations presented above using UV/sulfite data suggest that the fluoride ion yield for PFOA is ≈5%. Future work to verify this value and expand fluoride yield measurements to additional PFAS structures is needed.

We would like to conclude by giving some final thoughts on the question of whether there is there any practical role for UV-ARP in water treatment. This question is important considering the attention given to UV-ARP for treatment of PFAS. While e_aq^–-based processes appear to be one of the more promising methods for PFAS transformation, it is difficult to see how UV-ARP could be implemented at the same scale as UV-AOP. The main barriers include the large UV doses required, the undesirable byproducts from e_aq sensitizers in the treated water, and the rapid scavenging of e_aq^– by O₂. The use of UV-ARP for contaminant treatment appears to be most promising for groundwater applications with low background e_aq scavengers (e.g., little to no dissolved oxygen or nitrate present). Alternatively, the treatment of small volumes of concentrated contaminant waste streams, like from NF, RO filtration or ion exchange, could justify the large energy requirements of UV-ARP. Additional applications could include modular pump and treat systems for contaminated sites or temporary cleanup from chemical spills. Exploring these UV-ARP applications requires continued bench level experiments and more emphasis on pilot studies. Future research should seek to determine where UV-ARP and other e_aq^–-based treatment schemes fit into the management of chemical contaminants in water, most notably PFAS.

Glossary

Abbreviations and Symbols Used

α

(absorption coefficient of background water matrix, cm^–1)

CO₂^•–

(carbon dioxide radical anion)

e_aq^–

(hydrated electron)

ε_sens

(molar absorption coefficient, M^–1cm^–1)

E_EO

(electrical energy per order, kWh m^–3)

k_d^′

(first-order direct photolysis rate constant, s^–1)

k_S,t^′

(e_aq^– scavenging capacity, s^–1)

kTC,eaq–q

(bimolecular rate constant for e_aq^– quenching by the target contaminant, M^–1 s^–1)

kTC,eaq–t

(bimolecular rate constant for target contaminant transformation, M^–1 s^–1)

k_TC,eaq^–

(bimolecular target contaminant-hydrated electron rate constant, M^–1 s^–1)

I₀

(photon irradiance, mmol photons cm^–2s^–1)

IAA

(3-indoleacetic acid)

^•H

(hydrogen atom)

l

(path length, cm)

LED

(light emitting diode)

LP-M

(low-pressure mercury lamp)

MP-M

(medium-pressure mercury lamp)

MCAA

(monochloroacetic acid)

NTA

(nitrilotriacetic acid)

^•OH

(hydroxyl radical)

Φ_eaq^–

(quantum yield of e_aq^– formation)

PFAS

(per- and polyfluoroalkyl substances)

PFCA

(perfluorocarboxylic acid)

PFECA

(perfluoroethercarboxylic acid)

PFSA

(perfluorosulfonate)

R_f^e_aq–

(hydrated electron formation rate, M s^–1)

R_f^X–

(halide formation rate, M s^–1)

sens

(sensitizer)

SO₃^•–

(sulfite radical anion)

S

(scavenger)

TC

(target contaminant)

UV

(ultraviolet)

UV-ARP

(UV-Advanced Reduction Process)

VUV

(vacuum ultraviolet)

W

(watt)

γ_{eaq^–,X^–}

(halide yield)

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsenvironau.1c00042.

• Excel sheets containing bimolecular rate constants for hydrated electron reaction with organic compounds and a summary of UV-ARP treatment literature (XLSX)

The authors declare no competing financial interest.