Introduction

For the purpose of this review, reactive oxygen species (ROS) are defined as relatively short-lived molecules that contain oxygen atoms, with half-lives (t ½) in aquatic environments in the range of nanoseconds to hours (Bartosz 2006; Kearns 1971; Lu et al. 2006; Schmidt 2006; Zafiriou 1977; Zafiriou et al. 1984, 1990; Kieber et al. 2003; Waite et al. 1988). ROS are commonly found at picomolar to micromolar concentrations in environmental systems. This review focuses on the dominant ROS in surface waters and includes methods for the detection and quantification of singlet oxygen (1O2), superoxide (O ·−2 ) and its protonated form (hydroperoxyl radical; HOO·), hydrogen peroxide (H2O2), hydroxyl radical (HO·) and carbonate radical (CO ·−3 ).

In natural systems, 1O2, HOO·, HO·, H2O2, and CO ·−3 are capable of oxidizing a wide variety of molecules (including biomolecules) with relatively low selectivity and are involved in the attenuation of contaminants and the transformation of dissolved organic matter (DOM) in aquatic environments (Brezonik and Fulkerson-Brekken 1998; Canonica et al. 2005; Westerhoff et al. 1999, 2007). O ·−2 is more selective in its reactions with aqueous organic compounds than other ROS, but its reduction potential overlaps that of a range of biologically important metal ions (e.g., iron, copper, and manganese) that can themselves affect DOM oxidation (Goldstone and Voelker 2000). H2O2 is a thermodynamically powerful oxidant but its reaction rates with many compounds are typically slow compared to those of free radicals. Its conjugate base (HOO−) is also capable of acting as a reductant under some conditions, particularly for transition metal ions (Wood 1974; Koppenol and Butler 1985; Petlicki and van de Ven 1998).

ROS are usually generated by photolysis, electron transfer or energy transfer reactions (Bartosz 2006; Kearns 1971; Lu et al. 2006; Schmidt 2006; Zafiriou 1977; Zafiriou et al. 1984, 1990; Kieber et al. 2003). In the absence of other sinks, most free radical ROS undergo self-reaction (e.g., dimerization or disproportionation), while the electronically excited 1O2 rapidly decays through vibronic coupling with water (i.e., non-radiative decay). The steady-state concentration of 1O2 observed in natural waters is typically constrained by its interaction with water (resulting in a t ½ ~4 μs) with significant concentrations only observed in localized hydrophobic environments (Grandbois et al. 2008; Pogue et al. 2000). The bimolecular rate constants for the self-reactions of O ·−2 /HOO·, HO· and CO ·−3 are reasonably high (Czapski et al. 1994; Elliot et al. 1990; Zafiriou 1990; Scurlock and Ogilby 1996; Czapski and Dorfman 1964). However, in environmental systems, concentrations of these ROS are rarely high enough for self-reaction to be a significant sink for removal due to their large bimolecular rate constants with other sinks such as trace metal species and organic compounds. H2O2 does not react with itself but catalytically degrades through rapid reactions with trace metal ions and enzymes (Zepp et al. 1992; Rush and Bielski 1985; Duesterberg et al. 2005). Consequently, analyses of ROS have proven challenging because their lifetimes, with the exception of H2O2, are usually too short for ex-situ analysis.

Methods for ROS detection can be broadly classified as either direct or indirect. Due to the short lifetimes and typically low concentrations of ROS in aquatic systems, their direct observation is only possible on the sub-millisecond timescale, with the relatively stable H2O2 being an exception. Indirect methods involve the reaction of a particular ROS with a probe molecule to yield a more stable, long-lived analyte (Zafiriou et al. 1990). Such methods typically involve specific chemical derivatization (e.g., trapping a radical with a nitroxide or other spin trap) or are based on competitive kinetics. By virtue of introducing additional chemical reactions, all indirect techniques risk perturbing the observed system. In addition, both direct and indirect analyses suffer from the poor availability of standardized approaches to calibration (see Online Resource 1), particularly for use in the field.

Some important aspects to consider when choosing an ROS analysis method include: (1) the sensitivity of the method; (2) the selectivity and specificity of the method for the analyte of interest; and (3) the ability of the method to allow measurements with sufficiently fast time resolution. Specificity varies widely between methods, and should be carefully considered when choosing a method for ROS qualification and/or quantification. Additional analytical considerations are availability, robustness, portability (for field studies), the cost of the necessary instrumentation, and in some cases, the cost of the probe molecules. Largely due to these latter factors, much of the method development for aqueous ROS analysis has focused on ultraviolet (UV)/visible (Vis) light spectroscopic techniques and the use of relatively common and hence lower cost probe molecules. Spectroscopic detection strategies [including absorbance (UV/Vis), fluorescence (FL) and chemiluminescence (CL)] share a common approach with several other techniques for measuring rates of ROS formation and decay in laboratory experiments. These strategies are also compatible with methods such as steady-state kinetic analyses, stopped flow methods, time-resolved laser spectroscopy, flash photolysis and pulse radiolysis (Waite et al. 1988). However, applications of spectroscopic techniques for ROS analyses in natural waters are often limited by interference from DOM through background absorbance or FL, although the use of CL probes may circumvent these issues. All of these spectroscopic approaches can benefit, in certain circumstances, from the application of a preliminary “clean-up” technique such as the use of a concentrating resin, extraction, or chromatography [gas chromatography (GC) or high performance liquid chromatography (HPLC)] to remove interferences prior to analysis. Other analytical techniques for ROS detection, such as electron paramagnetic (spin) resonance (EPR), nuclear magnetic resonance (NMR), derivatization with attendant mass spectrometric (MS) analysis and liquid scintillation counting can also be quite useful but are less portable and often require considerable technical expertise to operate and can be expensive. For these reasons, when use of the aforementioned instrumentation is required, the observable species must be stable on timescales of days or more.

Earlier reviews have focused on the detection of specific ROS in specific media (e.g., cellular, aqueous, or organic solvents) (Bartosz 2006; Lu et al. 2006; Zafiriou et al. 1990; Gomes et al. 2005). This review is focused more broadly on comprehensively listing published methods that in the authors’ considered opinion are relevant for qualifying and quantifying ROS in environmental settings, particularly fresh and marine waters, groundwaters, and atmospheric waters. However, ‘relevance’ is defined by the needs of the researcher, and it is the authors’ hope that this review will also be useful for scientists working in engineered aquatic systems and biological systems. Since the probes reviewed in the compilation are by their nature reactive, it is important to note that the methods listed in the following tables require control studies in purified water to evaluate the possibility of probe degradation or competing reactions that may confuse the ROS signal. The low aqueous solubility of some probes may dictate the conditions of their use or result in their inadvertent partitioning from the aqueous phase to organic microenvironments (e.g., micellar or intra-DOM) (Grandbois et al. 2008; Lissi et al. 1993; Latch and McNeill 2006). The authors have personal experience with many of the methods but not all compiled in the tables. Because of this, when personal experience was lacking, additional pertinent information inserted into the tables was kept as faithful to the original cited text as possible.

The review is sectioned by ROS, with a brief introduction highlighting the fundamental chemical properties of the particular ROS followed by a critical evaluation and tabulation of relevant methods for detection of that ROS. The table entry for each method is arranged to display the identity of the probe molecule (using the nomenclature from the method’s literature citation), observable (analytical signal associated with the technique; e.g., absorbance, FL or CL emission, etc.), limit of detection (LOD) (as reported or calculated based on the original citation), application notes, approximate number of literature citations for the method (as of 09/2011), reaction schemes, and references for the method.

An Excel spreadsheet has also been prepared summarizing the information in the tables, enabling all the ROS analytical methods to be selected in terms of specific criteria such as the type of ROS to be analyzed while listing the relevant methods for that ROS, for example, in order of increasing LOD. This spreadsheet can be downloaded from the Web site http://neon.otago.ac.nz/research/bmp/data/ros_database.xlsx.

Singlet oxygen

Ground state molecular oxygen exists as a triplet state with the lowest lying excited state of oxygen being a singlet state [O2(1Δg) or 1O2; Table 1] that lies 94 kJ/mol above the ground state (Table 1) (Khan et al. 1967; Wilkinson et al. 1995). The singlet state can be generated in solution by energy transfer from excited photosensitizers (S, e.g., humic substances or Rose Bengal; Eq. 1); or chemically, for example via the reaction between hypohalites and H2O2 (Eq. 2) (Zafiriou 1977; Khan and Kasha 1963; Schweitzer and Schmidt 2003; Schmidt 2006).

Table 1 Properties of singlet oxygen
Full size table
$$ {\text{S}} \mathop \to \limits^{h\upsilon } {\text{S}} ^{*} \mathop \to \limits^{{{\text{O}}_{2} }} {\text{S}} + {^{1}\text{O}}_{2} $$
(1)
$$ {\text{OCl}}^{ - } + {\text{H}}_{ 2} {\text{O}}_{ 2} \to {\text{Cl}}^{ - } + {\text{H}}_{2} {\text{O}} +{^{1}\text{O}}_{2} $$
(2)

In sunlit waters, singlet oxygen concentrations have been measured in the range of ~10−12 to 10−13 M (Table 1) (Zepp et al. 1977; Haag and Hoigne 1986; Larson and Marley 1999; Egorov et al. 1989; Merkel and Kearns 1971; Wolff et al. 1981; Wick et al. 2000; Shao et al. 1994). The lifetime of singlet oxygen in aqueous solution is constrained through quenching by water with its lifetime in pure water being ~4 μs (Faust 1999; Egorov et al. 1989; Merkel and Kearns 1971). In natural waters its lifetime may be shorter due to the presence of additional quenchers, such as DOM (Table 1) (Faust 1999).

Direct measurement of the concentration of 1O2 is possible through observation of its emission at 1,268 nm (Hessler et al. 1994; Nonell and Braslavsky 2000). However, the intensity of this emission is too weak to be useful at low 1O2 concentrations. Consequently, use of this technique has been restricted to transient luminescence studies initiated by laser irradiation. The desire to measure environmentally relevant concentrations necessitated the development of molecular probes that can trap 1O2 or be used in competition kinetics (Table 2).

Table 2 Methods for detecting singlet oxygen (1O2)
Full size table

A suitable probe compound for photochemically generated 1O2 has to meet several requirements that are well summarized by Nardello et al. (1997): “The trap must be highly reactive towards 1O2, specific, compatible with aqueous media and must not perturb the system under study. Moreover, it must be transparent in the spectral range of the incident light in order to avoid photosensitization by the trap itself.” However, while many of the currently available probes meet some of these requirements, they do not meet all. Typical probes include anthracene- and pyrene-based compounds which are poorly soluble in water and absorb strongly in the UV-A and UV-B ranges appropriate for aquatic photochemistry (Evans and Upton 1985; Botsivali and Evans 1979; Wasserman et al. 1972; Corey and Taylor 1964). Therefore, the suite of probes matching all of the above mentioned requirements specified by Nardello et al. (1997) is small and includes only furfuryl alcohol (furan-2-ylmethanol; FFA) and 1,3-cyclohexadiene-1,4-diethanoate. FFA has the benefit of being commercially available and has been one of the most widely used probe compounds for singlet oxygen in aquatic photochemistry (Braun et al. 1986; Haag and Hoigne 1986; Haag et al. 1984a, b).

Furan derivatives, such as FFA, react with 1O2 to yield the corresponding molozonide (Table 2), which is unstable in water and rapidly hydrolyzes to other products, including the corresponding dicarbonyl. Steady-state concentrations of 1O2 are determined by measuring the rate constant of the loss of the furan probe and dividing by the second-order rate constant (Haag et al. 1984a; Latch et al. 2003). The rate constant is ~108 M−1 s−1 and hence two orders of magnitude below the diffusion-controlled limit (1010 M−1 s−1). Combined with the picomolar and sub-picomolar concentrations of 1O2 observed in natural waters, this results in long measurement times (tens of minutes) being required to observe a readily detectable decrease (>5%) in the concentration of FFA (Braun et al. 1986; Haag et al. 1984a, b).

The presence of 1O2 can also be tested through the addition of quenchers that promote non-radiative decay of this ROS back to the ground state. The addition of these materials result in an effective reduction in the rate of an observed process (k q), to a degree that is predictable (and therefore testable) based on the known rates of reaction between the added quencher and 1O2. The addition of quenchers can potentially disrupt the nature of the system and their suitability must be based on the system’s needs. For example, the azide ion (N3 −) is both a 1O2 quencher and a microbial poison, and is thus not suitable for microbiological studies examining the role of 1O2. Examples of frequently cited quenchers are N3 −, I−, and diazabicyclooctane (DABCO; Table 2) (Rubio et al. 1992; Hasty et al. 1972; Ouannes and Wilson 1968; Saito et al. 1975; Zepp et al. 1977). As with the furan derivatives, the interpretation of the results of these experiments is complicated by the additional reactivity of the quencher with HO· (Motohashi and Saito 1993).

An assay for 1O2 that avoids interference from HO· is the use of D2O as the reaction solvent (Table 2). Singlet oxygen has a longer lifetime in D2O solutions than H2O as a result of the relatively poor vibronic coupling between 1O2 and D2O (k d = 1.8 × 104 s−1) relative to H2O (k h = 2.4 × 105 s−1) (Wilkinson et al. 1995). Singlet oxygen’s reduced decay rate in D2O and/or D2O–water mixtures results in higher steady-state 1O2 concentrations leading to higher rate constants for oxidation of singlet oxygen acceptors (Merkel and Kearns 1972a, b; Merkel et al. 1972; Zepp et al. 1977). The use of the kinetic isotope effect in this instance allows the qualitative determination of 1O2 (through comparison of experimentally measured rates in the presence and absence of D2O) and its quantitative determination through application of the steady-state approximation to the rate of loss of a second probe (e.g., FFA). It should be noted that if the behavior of 1O2 is being monitored in a lipid membrane or other micro-heterogeneous phase (e.g., within a DOM micro-phase), many of these analytical measurements are rendered ineffective, since the reaction in this case is insensitive to the composition of the solvent and to the presence of quenchers that are present in the aqueous phase (Latch and McNeill 2006).

Superoxide and hydroperoxyl radical

Superoxide (Tables 3, 4) is the one-electron reduced form of triplet O2 and the conjugate base of the hydroperoxyl radical (Eq. 3; Table 3) (Hoigne 1975; Adams and Willson 1969; Bielski 1978). Because of its relatively low pK a (4.69), the O ·−2 anion dominates over HOO· in the majority of aqueous environments. O ·−2 can be generated during the redox cycling of transition metals (Eq. 4) and the photodegradation of DOM (Eq. 5) (Richard and Canonica 2005).

Table 3 Properties of superoxide and hydroperoxyl radicals
Full size table
Table 4 Methods for detecting superoxide (O ·−2 )
Full size table
$$ {\text{HOO}\cdot} \rightleftharpoons {\text{O}}_{2}^{ \cdot - } + {\text{H}}^{ + } $$
(3)
$$ {\text{M}}^{\text{n + }} + {\text{O}}_{2} \to {\text{M}}^{{({\text{n + }}1) + }} + {\text{O}}_{2}^{ \cdot - } $$
(4)
$$ {\text{DOM}} + {\text{O}}_{2} \mathop \to \limits^{{{\text{h}}\upsilon }} {\text{DOM}}_{\text{oxid}}^{ + } + {\text{O}}_{2}^{\cdot - } $$
(5)

Superoxide concentrations have been reported over the range of 10−9 to 10−12 M in natural waters (Faust 1999; Fujiwara et al. 2006; Petasne and Zika 1987; Rose et al. 2008a, b; Voelker et al. 2000; Hansard et al. 2010; Heller and Croot 2010b; Shaked et al. 2010). The O ·−2 anion is relatively unreactive due to resonance stabilization, as reflected by the low rate constant for self-reaction of the anion (Eq. 6), but readily undergoes disproportionation through reaction with HOO· (Eq. 7). HOO· also reacts relatively rapidly with itself (Eq. 8) (Zafiriou 1977; Bielski 1978; Cooper and Zika 1983a).

$$ 2 {\text{HOO}\cdot} \to {\text{O}}_{2} + {\text{H}}_{2} {\text{O}}_{2} \quad k_{6} = 8. 3\times 10^{ 5} ({\text{M}}^{ - 1} \,{\text{s}}^{ - 1} ) $$
(6)
$$ {\text{HOO}\cdot} + {\text{O}}_{2}^{\cdot - } \to {\text{O}}_{2} + {\text{HOO}}^{ - } \mathop \to \limits^{{{\text{H}}^{ + } }} {\text{O}}_{2} + {\text{H}}_{2} {\text{O}}_{2} \quad k_{7} = 9. 7\times 10^{ 7} ({\text{M}}^{ - 1} \,{\text{s}}^{ - 1} ) $$
(7)
$$ 2 {\text{O}}_{2}^{\cdot -} \to {\text{O}}_{2} + {\text{O}}_{2}^{\cdot - } \mathop \to \limits^{{2{\text{H}}^{ + } }} {\text{O}}_{2} + {\text{H}}_{2} {\text{O}}_{2} \quad k_{8} \approx 10^{ 1} ({\text{M}}^{ - 1} \,{\text{s}}^{ - 1} ) $$
(8)

The apparent second-order rate constant for disproportionation in terms of \({\text{T}}_{\text{O}_{2}^{\cdot - }}\), k obs, is thus highly pH dependent (Zafiriou 1977; Cooper and Zika 1983b; Bielski 1978) (Eqs. 9, 10), with a maximum value of ~107 M−1 s−1 at pH 4.69 (the pK a of O ·−2 /HOO·) that decreases by an order of magnitude for each unit increase in pH where the pH > pK a (Bielski 1978).

$$ \frac{{{\text{dT}}_{{{\text{O}}_{2}^{\cdot - } }} }}{\text{dt}} = - k_{\text{obs}} ({\text{T}}_{{{\text{O}}_{2}^{ \cdot - } }} )^{2} $$
(9)
$$ k_{\text{obs}} = \left( {\frac{{[{\text{H}}^{ + } ]}}{{k_{8} + [{\text{H}}^{ + } ]}}} \right)^{2} k_{6} + \left( {\frac{{[{\text{H}}^{ + } ]}}{{k_{8} + [{\text{H}}^{ + } ]}}} \right)\left( {\frac{{{k}_{8} }}{{k_{8} + [{\text{H}}^{ + } ]}}} \right)k_{7} $$
(10)

At the typical pH and superoxide concentrations found in seawater (8.1, 10−9 to 10−12 M), the uncatalyzed second-order disproportionation rate predicts a decay half-life of hours to hundreds of days. Measured half-lives of superoxide in seawater are much faster, ranging from 10 to 300 s, presumably due to catalysis by enzymes or transition metal ions (Hansard et al. 2010; Heller and Croot 2010b; Rose et al. 2010; Shaked et al. 2010; Rusak et al. 2011; Saragosti et al. 2010). This accelerated reactivity of O ·−2 thus presents significant analytical challenges, because standards (Online Resource 1) and samples at natural pH are stable for only seconds to minutes.

Superoxide absorbs strongly in the 230–350 nm region of the UV/Vis spectrum and can be quantified directly at micromolar concentrations by measuring its absorbance (Tables 3, 4). However, such a method is of limited value for measuring naturally occurring concentrations of superoxide due to the strong absorbance exhibited by other components of natural waters in this wavelength range. Many of the superoxide decay rate measurements in pure water were measured by millisecond ultraviolet spectroscopy (Bielski 1978). Superoxide can also be determined spectrophotometrically by measuring the rate of loss of compounds such as ferricytochrome c (FC), nitrobluetetrazolium (NBT), and 4-chloro-7-nitrobenzo-2-oxa-1,3-diazole (NBD-Cl) with which it readily reacts (Heller and Croot 2010a). These techniques have largely been used for measurements of superoxide production rates in biological systems due to their limited sensitivity (LOD ~ 1 μM–0.1 μM) and their lack of specificity (Olojo et al. 2005).

Due to superoxide’s brief lifetime and low steady-state concentrations (<1 nM) in natural waters, it is typically measured using highly sensitive CL probe molecules. Successful probes for decay or steady-state measurements must react at rates of at least ten times greater than that of natural superoxide disproportionation. Luminol is the most widely used CL probe for natural water analysis. This reagent has been used for the analysis of iron (Rose and Waite 2001; Xiao et al. 2002), chromium (Xiao et al. 2000), hydrogen peroxide (Yuan and Shiller 1999), and superoxide (Fujiwara et al. 2006), among other species, where these analytes are the rate-limiting species in the oxidation of luminol by superoxide. Unfortunately, because so many species can promote the CL of luminol in the presence of dissolved oxygen and hydrogen peroxide, this reagent is problematic for the selective analysis of superoxide in complex matrices.

In contrast, the selectivity, CL intensity, and commercial availability of MCLA (2-methyl-6-(4-methoxyphenyl)-3,7-dihydroimidazo[l,2-a]pyrazin-3(7H)-one) and red-CLA([2-[4-[4-[3,7-dihydro-2-methyl-3-oxoimidazo[1,2-a]pyrazin-6-yl]phenoxy]butyramido]ethylamino]sulforhodamine101) make these probes particularly suitable for superoxide analysis in natural waters (Godrant et al. 2009; Hansard et al. 2010; Heller and Croot 2010b; Rose et al. 2008a, b; Shaked et al. 2010; Zheng et al. 2003). The entire class of CLA molecules react with superoxide through an expoxitane intermediate, with red-CLA and FCLA (3,7-dihydro-6-[4-[2-[N′-(5-fluoresceinyl)thioureido]ethoxy]phenyl]-2-methylimidazo[1,2-a]pyrazin-3-one) involved in a CL resonance energy transfer to shift the emission to longer wavelengths (Teranishi 2007). The CLA probes are also reactive with singlet oxygen (Suzuki et al. 1990), but selective analysis of superoxide is possible by first waiting ~100 μs for the singlet oxygen to decay. More significantly, the CLA reagents are not reactive to hydrogen peroxide, which is often present at concentrations in 100-fold excess to that of superoxide in natural waters. The selectivity of these probes is driven by their specific and relatively rapid second-order reaction rate with superoxide (~5 × 105 M−1 s−1). These rates of reaction are then 100 times that of the rate of disproportionation (Bielski 1978) and the rate of first order superoxide decay in natural samples (Hansard et al. 2010; Heller and Croot 2010b; Rose et al. 2008a, 2010; Shaked et al. 2010). The CLA reagents will react with superoxide and oxygen at pH and temperature dependent rates, thus requiring background CL measurements and standard additions of superoxide at ambient conditions with a constant oxygen concentration (creation of standards and correction for background CL described in Online Resource 1) (Godrant et al. 2009; Hansard et al. 2010). For superoxide flux measurements, a CLA probe is added several minutes before making chemiluminescence measurements to ensure complete reaction of the steady-state superoxide concentration in the samples. The CL signal is then assumed to arise solely from superoxide production in the sample with the chemiluminescence photon flux proportional to superoxide production rates (Godrant et al. 2009).

Of the analytical methods discussed above, MCLA and red-CLA have both been successfully used to measure steady-state concentrations and production/decay rates of superoxide in seawater. Detection limits of ~30 pM are reported for superoxide concentrations using MCLA (Hansard et al. 2010) while production/decay rates as low as ~1 pM/s can be determined using red-CLA (Godrant et al. 2009). Recently, there has been a substantial increase in the number of superoxide measurements being made in a wide range of oceanographic water types, most of which have utilized CL probes (Hansard et al. 2010; Heller and Croot 2010a, b; Rose et al. 2010; Shaked et al. 2010; Rusak SA 2011).

Hydrogen peroxide

Hydrogen peroxide (Table 5) is a weak acid (pK a 11.62) that is a ubiquitous component of surface and atmospheric waters (Eqs. 11–13). It is an important component of natural waters due to its impact on redox chemistry and biological processes, its role as an indicator of photochemical oxidation of DOM, as a photic zone tracer in the ocean, and its potential utility for in situ degradation of pollutants. Its presence in natural waters typically arises from the disproportionation of superoxide and the hydroperoxyl radical (von Sonntag and Schuchmann 1991). H2O2 production often occurs in association with the photoexcitation of DOM or the thermal oxidation of reduced transition metal ions (Eqs. 12–13), along with production from biological sources (Petasne and Zika 1987; Cooper et al. 1987; Thompson and Zafiriou 1983; Cooper and Zika 1983a).

Table 5 Properties of hydrogen peroxide
Full size table
$$ {\text{H}}_{2} {\text{O}}_{2} \rightleftharpoons {\text{HO}}_{2}^{ - } + {\text{H}}^{ + } $$
(11)
$$ {\text{M}}^{\text{n + }} + {\text{O}}_{2}^{\cdot - } \to {\text{M}}^{{({\text{n + 1}}) + }} + {\text{H}}_{2} {\text{O}}_{2} $$
(12)
$$ {\text{DOM}} \mathop \to \limits^{{{\text{h}}\upsilon }} {\text{DOM}}^{*} $$
(13)

The lifetime of H2O2 in the environment is dependent not only on pH, but also on the presence of transition metal ions, biological enzymatic decay, and some organic species that can catalyze its decomposition (Eq. 5) (Petasne and Zika 1997; Moffett and Zafiriou 1990, 1993; Moffett and Zika 1987). In ocean waters, its lifetime is on the order of days but in coastal waters it is much shorter (Hakkinen et al. 2004; Shaked et al. 2010). Hydrogen peroxide can be detected directly using spectrophotometric techniques, although its molar absorption coefficient (189 M−1 cm−1; Table 5) is low, thus limiting this analytical method to relatively pure solutions where the H2O2 concentration is high or the analytical pathlengths are long (e.g., atmospheric measurements) (Hochanadel 1952; Morgan et al. 1988).

Hydrogen peroxide has been measured at concentrations between 10−7 and 10−11 M (Table 5) in natural surface waters (Peake and Mosley 2004; Moore et al. 1993; Szymczak and Waite 1988, 1991; Cooper and Zika 1983a) and is normally in the micromolar range in atmospheric water (Kok et al. 1978; Zika et al. 1982). Concentrations in surface waters are generally highest in the photic zone and diminish toward the detection limit of most methods in the dark waters below the mixed surface layer, reflecting its dominant formation process by photochemical reactions.

Of the more than 30 analytical methods listed in Table 6, only 10 have a LOD that is useful for the H2O2 concentrations found in surface waters (LOD ≤ 50 nM). Of these, six employ a peroxidase enzyme to achieve the specificity required for H2O2. Because H2O2 is ubiquitous in water due to equilibrium with gas-phase H2O2 in the atmosphere (even in laboratory water), most methods require addition of catalase (an enzyme that decomposes H2O2 into water and O2) to the sample to eliminate H2O2 for analytical blanks prior to analysis.

Table 6 Methods for detecting hydrogen peroxide (H2O2)
Full size table

Use of horseradish peroxidase (HRP) to provide specificity for the peroxide functional group can result in measurement of not only H2O2, but also other peroxy species such as peroxyacetic acid, methyl hydroperoxide, hydroxymethylperoxide, ethylhydroperoxide, and several propylperoxides formed through HRP-catlayzed reactions. These have all been shown to activate HRP and allow subsequent reaction with electron donors used to either develop (e.g., p-hydroxyphenylacetic acid or POHPAA) or diminish (e.g., scopoletin) FL. Using a post column chromatographic method based on the POHPAA technique, Miller et al. (2005) showed that any interference from organic peroxides is likely to be insignificant in the open ocean (Miller et al. 2005; Lee 1995). However, users of any peroxide method employing HRP should be aware of the potential contribution of organic peroxides in coastal and fresh waters.

While the 2-electron oxidation of HRP provides specificity for the peroxide functionality, it subsequently requires an electron donor to return HRP to its ground state. This second set of redox reactions is much less specific. This property of the enzyme is exploited in HRP-based methods, whereby the oxidized HRP subsequently reacts with a probe molecule to yield a product that is easily quantified, typically using spectroscopic methods such as absorbance or FL. This allows flexibility in terms of choosing the most suitable substrate for detection of H2O2 under particular measurement conditions. Phenolic compounds have fast reaction rates with the activated enzyme and all three FL methods discussed here take advantage of this fact. However, Miller and Kester (1988) have demonstrated that DOM in natural waters can also act as electron donors, likely via phenolic moieties which may need to be considered when using HRP-based methods under some conditions (Miller and Kester 1988).

While much research has been done with absorbance methods, the most commonly used and highly cited methods for determination of H2O2 in natural waters involve the HRP-catalyzed oxidation of probe compounds to yield products that either exhibit FL (e.g., p-hydroxyphenylacetic acid) or whose FL is diminished (e.g., scopoletin) after oxidation (Table 6). These fluorometric methods make use of readily available fluorophores, do not require specialized equipment other than a reliable fluorometer, and generally afford greater specificity, sensitivity and lower limits of detection compared to absorbance-based methods.

In many ways, current interest in the role of ROS in marine chemistry was inspired by early studies that used the HRP-catalyzed oxidation of scopoletin to analyze H2O2 in seawater (Perschke and Boda 1961; Zika et al. 1985a, b). While no longer the most commonly used method for the quantitative determination of peroxide, it is the seminal method from which many current methods evolved and so a presentation of some methodological detail is appropriate for any review. Specifically, when HRP, phenol, scopoletin, and H2O2 are together in a sample, the activated HRP enzyme catalyzes the production of a phenolic radical that then oxidizes scopoletin to a non-fluorescent product. This results in a stoichiometric decrease in scopoletin FL proportional to the concentration of H2O2 in the sample. Some studies have omitted phenol and still observed a decrease in scopoletin FL in the presence of HRP and H2O2. For example, Holm et al. (1987) substituted NaN3 for phenol to act as a more effective bactericide and H2O2 was still effectively measured using changes in scopoletin FL. It was noted, however, that the stoichiometry of the reaction varied without phenol. This is consistent with the greatly enhanced ability of naturally occurring phenolic compounds to compete as electron donors for activated HRP. In the absence of the fast-reacting phenol, a variable and wide variety of oxidized compounds having different reaction rates with scopoletin would generate variability in the observed decrease in FL. Additional details for the scopoletin method include the use of narrow-band optical filters and an excitation shutter (Donahue 1998), placing the sample in a dark fluorometer cell compartment for several minutes before measuring FL to minimize erroneous readings due to excitation from ambient light, storing reacted samples in the dark to reduce photobleaching, and carefully controlling pH. In fact, scopoletin FL is highly pH dependent and a buffer (usually phosphate buffer at pH 7) is required to ensure a meaningful and consistent relationship between the decrease in scopoletin FL emission intensity and the concentration of H2O2.

Hydrogen peroxide detection through the HRP-catalyzed dimerization of POHPAA to create a FL compound has been used extensively in rain, oceanic, and fresh water studies. In this method, two POHPAA radicals created via electron exchange with activated HRP dimerize to form a product that is FL at high pH using excitation and emission wavelengths of 313 and 400 nm, respectively (Miller and Kester 1988). The background FL of DOM in natural waters is variable and often pronounced at these wavelengths, thus requiring careful measurement of the natural FL in the sample prior to adding analytical reagents. This blank subtraction allows peroxide analysis in solutions with varying levels of naturally occurring fluorophores. As for all HRP-based methods, in samples with varying DOM content, standard additions are required to account for changes in stoichiometry due to competition between POHPAA and DOM for H2O2.

CL methods, while historically somewhat less useful in natural waters due to interferences from other ROS, have an inherent capacity for great sensitivity and have been used successfully in natural waters for measurement of H2O2. Both the luminol (Rose and Waite 2001) and acridinium ester methods (Cooper et al. 2000), when paired with portable and stable CL systems, have been demonstrated to be robust methods for analysis of H2O2 in oceanographic systems (Miller et al. 2005).

Luminol CL has been used most often for analysis of trace metals in natural water samples. However, in the presence of a metal catalyst, luminol (5-amino-2,3-dihydro-1,4-phthalazinedione) is oxidized, yielding the luminol radical, and subsequently a diazaquinone species. The diazaquinone reacts stochiometrically with H2O2, forming an intermediate that decays spontaneously to 3-aminophthalate while emitting luminescence that can be detected at ~425 nm (Rose and Waite 2001). In solutions with an excess of metal catalyst and carefully controlled pH, H2O2 becomes the limiting reagent in the luminescence reaction, and the photon signal can be quantitatively related to the concentration of H2O2 (Yuan and Shiller 1999). Luminol-based determinations of H2O2 are generally performed using flow injection analysis, and offer subnanomolar detection limits in natural waters. Using a cobalt catalyst, Yuan and Shiller (1999) achieved a LOD of <1 nM, with a precision of 17 pM. The method is linear up to 300 nM H2O2; above 300 nM, the slope of the calibration curve decreases. As an excess of Co(II) must be added to the sample prior to analysis, the method does not exhibit interference from most other oxidants present in natural waters. Iron species, however, do compete with H2O2 in a secondary oxidation reaction that leads to a positive interference (Yuan and Shiller 1999; Rose and Waite 2001).

Similar to the use of luminol, acridinium ester reacts stochiometrically with H2O2 to form an intermediate structure that upon addition of base, forms a second intermediate which rapidly decays and emits luminescence that can be detected at ~470 nm (Cooper et al. 2000). Acridinium ester method advantages are that it minimizes interference from colored and FL organic compounds and metals naturally present in natural waters and it does not require the addition of metal complexes or other catalysts for the method to work. Miller et al. (2005) demonstrated that the use of the acridinium ester CL method and the POHPAA FL method in the open ocean gives indistinguishable and accurate results.

Hydroxyl radical

The hydroxyl radical (Table 7) is a non-selective oxidant that can be generated from a variety of sources, including photolysis of nitrate and nitrite ions (Eq. 14), transition metal complexes, and DOM, as well as Fenton-type reactions involving peroxides, hypohalites and several transition metal ions (Eq. 15), on photo-excited transition metal oxide surfaces and as a result of the aqueous decomposition of ozone (Zafiriou 1977; Zafiriou and True 1979a, b; Alfassi 1999; Weissler 1953; Makino et al. 1983; Dixon and Norman 1963; Zepp et al. 1992).

Table 7 Properties of hydroxyl radical
Full size table
$$ {\text{NO}}_{3}^{ - } + {\text{h}}\upsilon \mathop \to \limits^{{{\text{H}}_{ 2} {\text{O}}}} {\text{NO}}_{2} + {\text{OH}}^{ - } + {\text{HO}\cdot} $$
(14)
$$ {\text{M}}^{\text{n + }} + {\text{H}}_{2} {\text{O}}_{2} \to {\text{M}}^{{({\text{n}} + 1) + }} + {\text{OH}}^{ - } + {\text{HO}\cdot} $$
(15)

HO· reacts at near-diffusion-controlled rates with many substrates, resulting in low steady-state concentrations of HO· in sunlit natural waters (10−15 to 10−18 M) (Zepp et al. 1987; Brezonik and Fulkerson-Brekken 1998; Haag and Hoigne 1985). The corresponding low concentrations and brief lifetimes (~μs) for HO· pose significant challenges in quantifying this ROS. While HO· absorbs light in the UV region (Table 7), direct observation is not typically possible because of its limited lifetime and the presence of other chromophores absorbing in a similar wavelength region. Therefore, HO· is quantified either through the loss of a reagent or accumulation of a product. The key challenge is obtaining a compound that will react selectively and unambiguously with HO·, which either does not unduly influence the other aspects of the chemistry of the system under study, or alters the system in a predictable and well-defined manner.

The EPR technique is often used to measure the formation of stable paramagnetic aminoxyl radicals (spin adducts) by reaction of HO· with diamagnetic nitrone or nitroso (spin trap) compounds. The use of these diamagnetic compounds can also be beneficial in that they may allow for simultaneous quantification of several ROS through formation of different spin adducts, each with a characteristic EPR spectrum. However, the stability of the spin adduct in a given system needs to be carefully evaluated and controls established to confirm that radical trapping is the source of aminoxyl formation (Finkelstein et al. 1980c). For example, the adduct formed from the reaction of HO· with dimethylpyrrolidine oxide (DMPO, Table 8) can be readily transformed to a diamagnetic species by both Fe(III) and superoxide (Mizuta et al. 1997; Samuni et al. 1988); also, the superoxide DMPO-adduct decays to form either the HO· adduct or HO· itself (Finkelstein et al. 1979, 1982). Similarly, under acidic conditions the alternative spin trap (E)-2-methyl-N-((1-oxidopyridin-4-yl)methylene)propan-2-amine oxide (4-POBN, Table 8) hydrolyzes to the hydroxylamine, which can be readily oxidized to form the HO· adduct (Brezonik and Fulkerson-Brekken 1998; Janzen et al. 1978). pH can also significantly impact the stability of the hydroxyl spin adducts, with dramatic increases in stability observed for phenyl/pyridlyl-butylnitrone hydroxyl adducts when the pH is reduced from 8 to 6 (Janzen et al. 1992b). Spin trapping-type compounds have also been employed using alternative methods of quantitation to EPR. For example, 19F-NMR has been employed with a fluorinated-DMPO derivative, albeit with similar drawbacks to those already described for DMPO (Khramtsov et al. 2001). Although employed extensively in the biomedical literature, spin trapping with EPR detection appears to have had minimal use in natural systems. Of course, apart from these chemical limitations of the method, one of the biggest draw backs is the cost of EPR instrumentation, which is much greater than typical absorbance or fluorescence spectrometers.

Table 8 Methods for detecting hydroxyl radical (HO·)
Full size table

An alternative to direct spin trapping of HO· is the inclusion of an additional reagent to convert HO· to a carbon-centered radical (such as the use of DMSO to form methyl radicals), which is then trapped and quantified. 3-aminomethyl-2,2,5,5,-tetramethyl-1-pyrrolidinyloxy (3-AMP, Table 8) and 3-amino-2,2,5,5,-tetramethyl-1-pyrrolidinyloxy (3-AP, Table 8) trap methyl radicals formed from the reaction of DMSO and HO·. The resultant complex can then be derivatized with fluorescamine and fluorometrically quantified after HPLC separation (with detection limits on the order of 10 nM) (Alaghmand and Blough 2007; Li et al. 1997a, 1999; Vaughan and Blough 1998). It is also possible to use a pre-fluorescamine derivatized aminoxyl probe in a similar fashion (Pou et al. 1993). This method can be employed under both oxic and anoxic conditions and has also been used to examine photochemical processes with minimal background production of the HO· derived product (Vaughan and Blough 1998).

The ability of HO· to undergo H-abstraction reactions has been employed for detection, typically using aliphatic alcohols/acids or halogenated alkanes as probe compounds (see Blough and Zepp 1995 for discusison of further probe compounds). Typical probe compounds employed in early studies involved quantitatively monitoring the loss of 1-chlorobutane (Haag and Hoigne 1985) or the reaction of 2-propanol with HO· yielding acetone, which can be quantified using HPLC after derivatization with 2,4-dinitrophenylhdrazine (Warneck and Wurzinger 1988).

Aromatic hydroxylation is another technique that is often employed for HO· quantification, particularly in natural aqueous environments. A wide variety of compounds have been reported for use in this type of HO· assay, including terephthalate, benzoic acid, p-chlorobenzoic acid, benzene and phthalhydrazide (Table 8) (Backa et al. 1997; Haag and Yao 1993; Miller et al. 2011; Page et al. 2010; Qian et al. 2001; Saran and Summer 1999; Vione et al. 2006; Zhou and Mopper 1990). These methods utilize the ability of HO· to add to an aromatic ring to initially form a hydroxycyclohexadienyl radical which can be further oxidized to a phenolic moiety by a range of oxidants. Dissolved oxygen is a suitable oxidant for this process, presumably proceeding via the mechanism suggested by Dorfman et al. (1962) involving HO2·/O ·−2 elimination to yield the diamagnetic hydroxylated species (Dorfman et al. 1962). The presence of other oxidants can alter the product distribution, and, as such, care must be taken to ensure consistent conditions for such a procedure to be analytically useful. In the hydroxylation of terephthalate, the yield of 2-hydroxyterephthalate increases by fivefold when IrCl6 2− is employed as the oxidant instead of the more abundant O2 (Fang et al. 1996). When benzoic acid is used as the probe compound, the more benign Fe(III)-EDTA (Ethylenediaminetetraacetic acid) oxidant is at least an order of magnitude less efficient than O2, and should not interfere with this procedure under oxic conditions (Maskos et al. 1990). Other routes to achieve aromatic hydroxylation are also possible, e.g., the hydroxylation of benzene to phenol by cytochrome P450 is a key process in the metabolism of benzene (Medinsky et al. 1995) with such processes representing potential interferences to aromatic hydroxylation probes. As O2 is necessary to oxidize the intermediate hydroxycyclohexadienyl radical, such probe systems are not suitable for investigating anoxic systems, where the absence of O2 would result in altered product distributions.

With aromatic hydroxylation probes, quantification can be undertaken by monitoring either the loss of the probe compound or the formation of one of the hydroxylated products, with the latter typically considered to be more selective for HO· than other strong oxidants and to offer sensitivity advantages. Since the hydroxylated product of analytical interest is only a fraction of the total amount of probe that has reacted with HO·, the yield of the hydroxylated product must be known in order to determine the quantity of HO· trapped. For benzoic acid, three hydroxybenzoic acid isomers are formed (Zhou and Mopper 1990). The proportion of ortho-, meta-, and para-substituted hydroxybenzoic acids has been found to be 36, 34, and 30%, respectively, under marine conditions (Zhou and Mopper 1990). When decarboxylation and ring fission products are accounted for, the fraction of HO· that reacts with benzoic acid to yield the para-isomer is determined to be 17%. Likewise, for phthalhydrazide, 20% of the reacted HO· yields the desired 5-hydroxy product (Miller et al. 2011) and 35% of the reacted terephthalate yields hydroxyterephthalate (Page et al. 2010). The higher yield from terephthalate is due to the symmetry of the probe molecule which yields only one possible hydroxy-isomer, thereby offering clear advantages in using this method.

For photochemical studies, it is essential to establish the stability of both the probe compound and the quantified product species with respect to direct photolysis by the light source. For the benzoic acid method, o-hydroxybenzoic acid has been shown to be stable with respect to direct photolysis using irradiation with λ > 313 nm and pH < 12 (Yang et al. 2004). The base (but not acid) form of m-hydroxybenzoic acid, however, is known to degrade in the presence of solar irradiation (Anastasio and McGregor 2001), whereas the para isomer is reportedly photo-stable (Zhou and Mopper 1990). With terephthalate, the probe compound itself is stable to solar irradiation, however, the hydroxylated product is directly photolyzed by 365 nm light with quantum yield (Φ) of (6.3 ± 0.1) × 10−3 (Page et al. 2010). In contrast, phthalhydrazide as a probe compound is directly photolyzed to the 5-hydroxy analyte, which is itself then seemingly stable to solar radiation (Miller et al. 2011). Although aromatic hydroxylation assays are advantageous with regards to sensitivity and selectivity, it is clear that caution must be used when applied to photochemical systems.

Further complications can arise when other oxidants are present in the system that are able to degrade the probe compound rendering it unavailable for reaction with any HO· that may be formed. For example, in studies of ozonation chemistry, precautions need to be taken to ensure the probe is unreactive towards ozone. In such cases, 4-chlorobenzoic acid has been found to be a suitable probe that is readily degraded by HO· yet is relatively unreactive towards O3. As such, HO· can be determined in the presence of O3 by monitoring the loss of 4-chlorobenzoic acid using HPLC (Haag and Yao 1993; Jans and Hoigné 1998).

There are significant challenges to determining HO· with all the methodologies described to date. In general, these all have some flaws or drawbacks that must be carefully considered and controlled as much as possible in any given system. Regardless of these potential problems, rigorous application of an appropriate method or indeed, potentially, several different methods (Table 8), enables quantitative insights to be made in the analysis of many systems. A discussion on the generation of standards to quantitatively determine HO· can be found in Online Resource 1.

Carbonate radical

The carbonate radical (Table 9) is produced in natural aqueous systems primarily through the oxidation of carbonate or bicarbonate ions by a one-electron oxidant such as HO· (Eq. 16) or the photoreactions of metal–carbonato complexes.

Table 9 Properties of carbonate radical
Full size table
$$ {\text{HO}\cdot}+{\text{{CO}}_{3}}^{2 - } \to {\text{OH}}^{ - } + {\text{{CO}}_{3}}^{\cdot - } $$
(16)

This radical has not been as widely measured in the aqueous environment as other transient oxidants. This is unfortunate, as its reduction potential overlaps that of I− and Br− and subsequent reactions of CO ·−3 could be a major source of reactive halogens in seawater. Its concentration in sunlit waters has been estimated at 10−13 to 10−15 M (Table 9) (Czapski et al. 1999; Faust 1999; Huang and Mabury 2000; Sulzberger et al. 1997; Larson and Zepp 1988). CO ·−3 reacts relatively slowly with itself, likely due to Coulombic repulsion (Behar et al. 1970a). It is much more selective than HO· with respect to reactions with organic species. For example, it reacts rapidly with phenols, anilines, and some amino acids (Busset et al. 2007; Chen and Hoffman 1973; Chen et al. 1975; Elango et al. 1985; Larson and Zepp 1988; Mak et al. 2007; Moore et al. 1977), but relatively slowly with saturated alkanes, aromatic hydrocarbons, etc. It primarily oxidizes organics through electron transfer as opposed to addition or atom transfer, and its role in environmental systems is underestimated because it has been so rarely measured. However, its impact may be greater than previously estimated, and it is interesting that the functional groups that CO ·−3 radicals are kinetically apt to react with are also those depleted most rapidly in the terrestrial DOM signature during mixing of fresh and marine waters.

CO ·−3 is strongly absorbing in the visible region (λ = 600 nm) and this is the primary method used for its detection in pump and probe experiments of its kinetics and reactivity (Weeks and Rabani 1966; Zuo et al. 1999). Its lifetime (~8 ms) is typically too brief to enable direct measurement in natural waters (Table 9) (Bonini et al. 1999; Canonica et al. 2005) and instead indirect quantitation methods are employed for this ROS. The presence of this radical is usually inferred through observation of the effects of the carbonate/bicarbonate ion on the oxidation of organic compounds by more oxidizing radicals such as HO· (Glaze et al. 1995; Glaze and Kang 1989). When the rate of oxidation of an HO· probe is reduced as a result of addition of carbonate ionic species, it is assumed that HO· has been scavenged by the carbonate anion resulting in the generation of CO ·−3 (Table 10). Although it is possible to trap this radical using nitrones with subsequent EPR detection (Table 10), the resulting adducts are subject to hydrolysis and so are difficult to characterize or quantify directly (Villamena et al. 2006, 2007; Wolcott et al. 1994; Yoon et al. 2002).

Table 10 Methods for detecting carbonate radical (CO ·−3 )
Full size table