Introduction

Diclofenac, (2-[2′,6′-dichlorophenyl)amino] phenylacetic acid) (DCF) (Scheme 1), is widely used as a non-steroidal anti-inflammatory drug and pain-relieving analgesic in humans and domestic animals. Owing to the fact that DCF exhibits low biodegradability and is readily excreted with urine, it is frequently detected in sewage treatment plant effluents and fresh water resources with moderate levels of aquatic toxicity (Ziylan and Ince 2011). Hence, DCF has been classified as an “emerging water pollutant” due to its potential for long-term unpredictable health effects in humans. As such, a lot of research has been lately devoted to the degradation of the parent compound and its metabolites in water and advanced oxidation processes (AOPs) have been demonstrated to be viable solutions to the problem (Naddeo et al. 2009; Ravina et al. 2002; Ziylan et al. 2013). AOPs are based on the in situ production of hydroxyl radicals (•OH) and reactive oxygen species (ROS) in water upon irradiation of the sample by UV light, ultrasound, electromagnetic radiation, and/or the addition of ozone or a semiconductor. The advantages of •OH and ROS are that they are non-selective and highly reactive with a large range of organic compounds.

Scheme 1
scheme 1

Molecular structure of diclofenac (DCF)

Full size image

Reaction mechanisms for the decomposition of DCF by •OH-mediated oxidation have been proposed (Hartmann et al. 2008; Martinez et al. 2011; Perez-Estrada et al. 2005; Vogna et al. 2004; Ziylan et al. 2014), but need to be further explored. The results of previous studies show that the most common degradation intermediates are hydroxylated compounds and compounds that form upon cleavage of the C-N bond, such as dichloroaniline, dichlorophenol, and 2-indolinone. Subsequent oxidative ring cleavage was found to lead to carboxylic acid fragments via classic degradation pathways (Hartmann et al. 2008; Vogna et al. 2004). Further reaction of dichloroaniline with •OH was found to produce 4-amino-3,5-dichlorophenol, 2,6-dichloro-p-hydroquinone, and dichlorobenzene (Hartmann et al. 2008). In addition, a cyclic intermediate (1-(2,6-dichlorophenyl)-indolin-2-one) has been detected by some researchers at highly acidic conditions (Reddersen and Heberer 2003). The production of chloroanilines such as N-(2,6-dichlorophenyl-)2-aminotoluene has been reported as part of the reduction pathway during the degradation of chlorobenzene (Drijvers et al. 2000; Stavarache et al. 2002). Final products such as oxalic acid, ammonium, nitric acid, and hydrochloric acid were attributed to the mineralization process via complete cleavage of the NH bridge (Hofmann et al. 2007). More recently, it has been demonstrated that during •OH-mediated oxidation, the toxicity of a DCF sample follows an up-and-down pattern during the early stages of the reaction, but stabilizes to a non-toxic level at prolonged reaction times (Ziylan et al. 2014).

In recent years, photodegradation mechanisms, kinetics (An et al. 2015; Gao et al. 2014a; Gao et al. 2014b; Gao et al. 2016a; Gao et al. 2016b; Ren et al. 2013; Ren et al. 2012; Zeng et al. 2016), and toxicology assessments (Gao et al. 2015; Li et al. 2016) of several pollutants were explored by theoretical calculations. Sonochemical degradation of diclofenac has been investigated by several experimental methods; however, there is a little information on its degradation mechanisms (Hartmann et al. 2008; Sein et al. 2008; Vogna et al. 2004). Understanding the degradation pathway of diclofenac and the toxicological consequences of its degradation by-products is crucial for environmental studies.

The present work aims to extend and enrich our previous study (Ziylan et al. 2014) with a reaction mechanism proposed to the initial reactions of diclofenac with •OH to elucidate the mechanism of subsequent reactions yielding the most common degradation by-products reported in the literature for the sonochemical decay of the compound. The study also encompasses estimation of the acute toxicities of the intermediates to assess the relative risk of the parent compound and its oxidation by-products to the aquatic environment.

Computational methodology

Quantum chemical calculations

All computations were carried out with the Gaussian 09 program package (Frisch et al. 2009). The stationary points were located using the B3LYP/6-31+G(d) methodology (Becke 1993a; b; Domingo et al. 2007; Lee et al. 1988; Stephens et al. 1994) and characterized by frequency analysis, from which thermal corrections were obtained at 298 K. Local minima and first order saddle points were identified by the number of imaginary vibrational frequencies. Intrinsic reaction coordinate (IRC) (Gonzalez and Schlegel 1990) calculations were performed on the transition state structures to determine the connectivity of the transition state structure to the corresponding product and reactant. Conformational searches for the structures corresponding to reactants, transition states, and products were carried out with the same methodology. Energies were further refined with the MPWB1K/6-311+G(3df,2p) level of theory (Adamo and Barone 1998; Zhao and Truhlar 2004). The effect of the solvent environment was taken into account utilizing the integral equation formalism polarizable continuum model (IEFPCM) (Mennucci et al. 1997; Mennucci and Tomasi 1997), where water (ε = 78.36) was used as solvent. The stability of the radicals generated as a result of H abstraction and OH addition was calculated using Eq. (1) and Eq. (2) (An et al. 2014; Galano and Alvarez-Idaboy 2009) as:

$$ \begin{array}{c}\hfill \mathrm{H}\ \mathrm{abstraction}\hfill \\ {}\hfill \mathrm{DCF}+\bullet \mathrm{OH}\to {\mathrm{H}}_2\mathrm{O}+{R}_{\mathrm{abs}}\bullet \hfill \\ {}\hfill \Delta {G}_{\mathrm{rxn}}= G\left({\mathrm{H}}_2\mathrm{O}\right)+ G\left({R}_{\mathrm{abs}}\bullet \right)- G\left(\mathrm{DCF}\right)- G\left(\bullet \mathrm{OH}\right)\hfill \end{array} $$
(1)
$$ \begin{array}{c}\hfill \mathrm{OH}\ \mathrm{addition}\hfill \\ {}\hfill \mathrm{DCF}+\bullet \mathrm{OH}\to {R}_{\mathrm{add}}\bullet \hfill \\ {}\hfill \Delta {G}_{\mathrm{rxn}}= G\left({R}_{\mathrm{add}}\bullet \right)- G\left(\mathrm{DCF}\right)- G\left(\bullet \mathrm{OH}\right)\hfill \end{array} $$
(2)

The standard radical stabilization energies (RSE) of the generated radicals were also calculated using Eq. (3) (Coote et al. 2010; Degirmenci and Coote 2016), where ΔH is the enthalpy change of the following isodesmic reaction and R• is a carbon centered radical:

$$ \begin{array}{c}\kern1em \mathrm{R}\bullet +{\mathrm{CH}}_3{\textstyle \hbox{-}}\mathrm{H}\to \mathrm{R}{\textstyle \hbox{-}}\mathrm{H}+\bullet {\mathrm{CH}}_3\kern1em \\ {}\kern1em \Delta {H}_{298}^0={\Delta}_f{H}_{298}^0\left(\bullet {\mathrm{CH}}_3\right)+{\Delta}_f{H}_{298}^0\left(\mathrm{R}{\textstyle \hbox{-}}\mathrm{H}\right)-{\Delta}_f{H}_{298}^0\left({\mathrm{CH}}_3{\textstyle \hbox{-}}\mathrm{H}\right)-{\Delta}_f{H}_{298}^0\left(\mathrm{R}\bullet \right)\kern1em \end{array} $$
(3)

Note that when RSE is a positive number, R• is supposed to be more stable than •CH3 and if it is negative, it is less stable.

The Gibbs free energies of the fragments are utilized in order to find the most possible fragmentation sites on DCF/DCF−. The Gibbs free energies of the reactions were calculated by Eq. (4):

$$ \Delta {G}_{\mathrm{rxn}}= G\left(\mathrm{fragment}(1)\right)+ G\left(\mathrm{fragment}(2)\right)- G\left(\mathrm{DCF}\right) $$
(4)

The basis set superposition error (BSSE) approach according to the Boys counterpoise method (Boys and Bernardi 1970; Kobko and Dannenberg 2001) was also utilized to calculate the energies for transition state structures of the initial reactions of diclofenac with •OH. The BSSE did not alter the general reaction profiles; the barriers have only increased slightly (1–2 kcal/mol). The correction being within the error margin of DFT, the BSSE-corrected results have not been included in the overall discussion of the mechanisms (Table S1).

Toxicity assessment

The acute toxicity of DCF and its oxidation by-products to green algae, daphnia, and fish were estimated in terms of EC50 (concentration of tested compound that inhibits the growth of 50% of green algae after 96 h of exposure) and LC50 (concentration of tested pollutant for 50% dead fish and daphnia after 96 and 48 h of exposures, respectively) using the Ecological Structure Activity Relationships (ECOSAR) software (ECOSAR 2014).

Results and discussion

Initial reactions of diclofenac with •OH

Diclofenac has two aromatic rings connected with an amine group. The previous experimental work of some of the authors show that during degradation, the solution pH decreases from 6.50 to 5.23 in 30 min and even lower as the degradation proceeds, indicating the formation of acidic intermediates (Ziylan et al. 2014). Because the medium is changing from a neutral to an acidic environment, the degradation of DCF is modeled in both its anionic and neutral forms (pKa of diclofenac is 4.1–4.5 (Scheytt et al. 2005)); their best conformers are depicted in Fig. 1. In both forms, there are strong intramolecular H-bonding interactions between the carboxylic oxygen and the amine hydrogen. As expected, these interactions are stronger in the anionic form (1.86 Å) than in the neutral form (2.08 Å) as a result of the excess negative charge on the oxygens in DCF−. Both ring systems in DCF (DCF−) are slightly twisted away from each other with a dihedral of −161.8° (164.3°) to avoid steric clashes, as seen in Fig. 1.

Fig. 1
figure 1

Optimized structures of diclofenac in neutral (DCF) and anionic forms (DCF −) (B3LYP/6-31+G(d) in water)

Full size image

AOPs produce highly reactive •OH, which are the main oxidants in the degradation processes (Glaze and Kang 1989; Glaze et al. 1987). In aqueous medium, •OH can either add to diclofenac or abstract a hydrogen from it to give either addition or abstraction intermediates, respectively. In Schemes 2 and 3, the reaction sites of DCF are labeled with numbers (1–7) and letters (a, b) in order to characterize their reactivities upon H abstraction and OH addition reactions. •OH can abstract a hydrogen from the aromatic (1–6) or benzylic positions (7), leading to the formation of a radical carbon center on the DCF molecule and the elimination of water (Scheme 2).

Scheme 2
scheme 2

H abstraction by •OH from DCF, generating radical intermediate (R abs)

Full size image
Scheme 3
scheme 3

Addition of •OH to DCF, generating radical intermediate (R add)

Full size image

Relative stabilities of the free radicals formed via H abstraction (R abs) during oxidation of diclofenac are summarized in Table 1. The most stable radical is the one corresponding to R abs7, which is almost 30 kcal/mol more stable than those on the benzene rings. H abstraction from the benzylic position is found to be more favorable—both thermodynamically and kinetically—when compared to H abstraction from the rings; this is expected based on the stabilization generated by conjugation of the benzylic radical in the former case. The preference of H abstraction from benzylic carbon over aromatic carbons is also reported by Sehested et al. (Sehested et al. 1975). Radicals on the ring are thermodynamically less favored than R abs7; however, their generation is still kinetically feasible in terms of their activation barriers (ΔG ‡ ~10 kcal/mol vs ~3 kcal/mol). Radicals on the second ring are more stable (RSE ~−8 kcal/mol) than those on the first ring (RSE ~−6 kcal/mol). When the two forms of diclofenac are compared, it is observed that DCF− is more prone to H abstraction than DCF with slightly lower activation barriers and higher exergonicities.

Table 1 Activation Gibbs free energies (ΔG ‡), activation enthalpies (ΔH ‡), reaction Gibbs free energies (ΔG rxn), reaction enthalpies (ΔH rxn) for the R abs intermediates calculated from Eq. (1), and RSE values calculated from Eq. (3) (MPWB1K/6-311+G(3df,2p)//B3LYP/6-31+G(d), kcal/mol). Values in parentheses have been calculated with B3LYP/6-31+G(d) in water
Full size table

In aqueous medium, •OH can readily add to aromatic molecules producing hydroxycyclohexadienyl-type radicals (Mohan and Mittal 1999). Hence, •OH could add to DCF forming R add intermediates as shown in Scheme 3. As opposed to H abstraction, there are two additional reactive sites (a and b) for hydroxyl radical addition.

The stabilities of the R add intermediates (arenium radicals) as calculated by Eq. (2) are listed in Table 2 for DCF and DCF−. Unlike the H abstraction route, the radicals formed on the first ring are more exergonic than those on the second ring, but have slightly higher activation barriers. The addition of •OH to positions a and b is slightly less favorable than addition to the rings and the reaction is not likely since re-aromatization is not probable. However, those intermediates could lead to C-N bond cleavage and fragmentation (Gao et al. 2014a). Furthermore, similar to the H abstraction route, we found that OH addition to the anionic form of diclofenac is thermodynamically and kinetically more favorable than the neutral one. The complete reaction profiles for these two reactions are presented in Scheme 4.

Table 2 Activation Gibbs free energies (ΔG ‡), activation enthalpies (ΔH ‡), reaction Gibbs free energies (ΔG rxn), reaction enthalpies (ΔH rxn) for the R add intermediates calculated from Eq. (2), and RSE values calculated from Eq. (3) (MPWB1K/6-311+G(3df,2p)//B3LYP/6-31+G(d), kcal/mol). Values in parentheses have been calculated with B3LYP/6-31+G(d) in water
Full size table
Scheme 4
scheme 4

Activation Gibbs free energies (ΔG ‡) and reaction Gibbs free energies (ΔG rxn) of the reaction of •OH with DCF− to give H abstraction intermediates (R abs) on the left (red path) and OH addition intermediates (R add) on the right (orange path)

Full size image

The reactivity of DCF and DCF− show the same trends, nonetheless with lower energy barriers and higher exergonicities for DCF−. It is known that OH addition to benzene ring is kinetically more favored than H abstraction from the ring carbons (Song et al. 2009). Similarly, based on RSE results, one can see that OH addition radicals are more stable than •CH3; note also that the radicals generated by H abstraction on the ring carbons are less stable than the ones formed by OH addition. When the whole reaction path is examined (Scheme 4 for DCF− and Scheme S1 for DCF), the major H abstraction intermediate is found to be R abs7 and this may lead to the fragmentation of DCF/DCF−. R add2 and R add3 are the most easily formed radicals among others.

Fragmentation of DCF/DCF−

The fragmentation of DCF/DCF− is modeled in order to understand the most probable degradation paths and be able to predict the possible fates of the generated radical intermediates. All fragments are optimized and their reaction energies are calculated with Eq. (4) (Table 3). The fragmentation energies are displayed in an increasing order. Based on Hammond’s postulate, ring cleavages from either side of the nitrogen (entries 2 and 3) seem to be harder than decarboxylation (entry 1). The rupture of the C-Cl bond (entries 4 and 6) is almost equally difficult in both cases (Cl (1) and Cl (2)). On the other hand, the losses of OH (entry 7) in the neutral form and O in the anionic form are more endergonic than the others; decarboxylation is less endergonic for DCF− (50.6 kcal/mol) than for DCF (66.2 kcal/mol).

Table 3 Reaction Gibbs free energies (ΔG rxn) for the fragments of DCF/DCF− calculated with Eq. (4) (MPWB1K/6-311+G(3df,2p)//B3LYP/6-31+G(d), kcal/mol). Values in parentheses have been calculated with B3LYP/6-31+G(d) in water
Full size table

Analysis of the degradation by-products

Following the generation of radicals through H abstraction and OH addition, subsequent reactions are examined to understand the formation of the mostly encountered degradation by-products of DCF by sonolysis, namely P1-P9 (Scheme 5). The degradation mechanisms that follow are based on the outcomes of “Initial reactions of diclofenac with •OH”. Hydroxylated derivatives lead to the degradation by-products, P5 (Hartmann et al. 2008; Michael et al. 2014; Perez-Estrada et al. 2005), P6 (Hartmann et al. 2008; Vogna et al. 2004), P7, and P8 (Hartmann et al. 2008; Homlok et al. 2011; Michael et al. 2014; Perez-Estrada et al. 2005; Sein et al. 2008; Vogna et al. 2004). H abstraction intermediate R abs5 gives P4 (Encinas et al. 1998a; Encinas et al. 1998b; Lekkerkerker-Teunissen et al. 2012; Martinez et al. 2011; Poiger et al. 2001; Salgado et al. 2013). Decarboxylated fragment of DCF results in P3 (Homlok et al. 2011) by ring closure. Ring cleavage fragmentation produces P1 (Bae et al. 2013; Hartmann et al. 2008; Homlok et al. 2011; Perez-Estrada et al. 2005; Sarasidis et al. 2014; Sein et al. 2008; Vogna et al. 2004), P5, P6, and P9 (Hartmann et al. 2008; Vogna et al. 2004). The rearrangement gives cyclization product P2 (Bae et al. 2013; Hartmann et al. 2008). The mechanisms and the energetics of these products (P1-P9) are examined and displayed in Scheme 5. The detailed reaction mechanisms (Scheme S2-S10) and the reaction profiles (Fig. S1-S9) are reported in the supporting information.

Scheme 5
scheme 5

Schematic representation of the generation of degradation by-products by (I) rearrangement and (II) in the presence of •OH using DCF− (except for P2 where DCF has been utilized) (Gibbs free energies (kcal/mol) are calculated with MPWB1K/6-311+G(3df,2p)//B3LYP/6-31+G(d))

Full size image

Subsequent degradation mechanism of DCF involves •OH almost in every step because it is well known that high •OH concentrations are produced under extreme conditions in advanced oxidation processes (AOP) (An et al. 2010a; An et al. 2010b; Fang et al. 2013; Shu et al. 2013). Despite the high concentration of •OH, P2 and P3 are formed presumably by rearrangement of DCF before the attack of •OH (Scheme 5-I). P2 is the result of an intramolecular cyclization leading to water loss (Fig. S2, Scheme S3). Even though the activation barrier for P2 is very high (ΔG ‡ = 47.0 kcal/mol), its overall formation is exergonic (ΔG rxn = −12.7 kcal/mol). The mechanism suggested for the formation of P2 starts with cyclization and is followed by water elimination. A cyclic tetrahedral gemdiol intermediate would form by attack of nitrogen to the carbonyl carbon, and then the intermediate loses a water molecule aided by the solvent, yielding the amide P2.

Another degradation by-product that forms without •OH in its subsequent degradation mechanism is P3 (Scheme S4). The decarboxylation of DCF/DCF− is relatively facile as shown in “Fragmentation of DCF/DCF− ”. After decarboxylation, DCFdecarb is formed; the exocyclic methylene radical attacks the chlorinated carbon, ejecting a Cl• and yielding P3. The barrier for this reaction is easily overcome (ΔG ‡ = 16.7 kcal/mol) and the reaction is highly exergonic (ΔG rxn = −32.2 kcal/mol) due to the formation of the stable cyclic by-product, P3 (Fig. S3).

Besides P2 and P3, there is another cyclic degradation by-product P4, which is also encountered frequently in the degradation experiments of DCF. P4 is produced in two steps: in the first step, a radical center is generated by a H abstraction from position 3 (on ring II) and in the second step, R abs3 attacks the chlorinated carbon to form a five-membered cyclic product similar to the formation of P3 where a six-membered ring was formed (Scheme S5). The barrier to form P4 from R abs3 is 13.4 kcal/mol with very high exergonicity (ΔG rxn = −45.3 kcal/mol) (Fig. S4).

P1 was found experimentally to be formed at early stages of the reaction; furthermore, it is the precursor of the subsequent oxidative products, e.g., P5 and P6 (Ziylan et al. 2014). The C-N bond cleavage in DCF is a very common fragmentation site (Hartmann et al. 2008; Perez-Estrada et al. 2005; Vogna et al. 2004). P1 is the first by-product that forms upon cleavage of the bond between C of ring II and N (Scheme S2). This cleavage is initiated with •OH addition to position a (formation of R adda with 9.3 kcal/mol) and disruption of the aromaticity of ring II (22.6 kcal/mol) (Fig. S1). After formation of R adda, the C-N bond is broken with a proton transfer from OH to NH giving two products (ΔG rxn = −36.9). This step requires 22.6 kcal/mol energy for the anionic form. The same transition state cannot be located for neutral form of diclofenac. The final products are P1 and a radical. These first cleavage products of P1 (Fig. S1) give other degradation by-products (P5 and P6) by further reacting with •OH. •OH addition to P1 produces an arenium radical on the ring and this radical captures another •OH resulting in a diol with broken aromaticity in the ring (Scheme S6). In order to rearomatize, a water molecule is lost and hydroxylated aniline is generated, which is found experimentally as one of the degradation by-products of diclofenac, P5 with very high exergonicity (−105.9 kcal/mol relative to P1, Fig. S5). Another attack of an •OH to P5 results in substitution of OH with NH2 to give P6. •OH attacks to the carbon where NH2 is attached and forms easily (4.5 kcal/mol) an arenium radical on the ring (Scheme S7); the latter rearomatizes as the NH2 radical leaves (11.8 kcal/mol). P6 is found to be 6.7 kcal/mol more stable than P5 (Fig. S6).

P7 and P8 are considered as the main degradation by-products of DCF/DCF−. •OH adds to the ring (ΔG ‡ = 7.2 kcal/mol for P7 and ΔG ‡ = 5.7 kcal/mol for P8) forming an arenium radical which is followed by a barrierless addition of •OH (Scheme S8 and S9). The latter intermediate is expected to lose a water and yield the final product in an exergonic path (ΔG rxn = −108.4 kcal/mol for P7 and ΔG rxn = −108.5 kcal/mol for P8) (Fig. S7 and S8). Another •OH addition to position b in P8 results in the C-N bond breakage and formation of P9 (ΔG rxn = −33.6 kcal/mol). The mechanism is similar to the formation of P1 but with a lower activation barrier (ΔG ‡ = 6.8 kcal/mol) (Scheme S10, Fig. S9).

Comparison of the suggested reaction mechanisms with experimental sonolytic findings

As reported by Hartmann et al., the attack of •OH immediately leads to the formation of the hydroxylated DCF derivatives P7 and P8 (compounds 5 and 6 in Hartmann et al. (2008)). In the proposed pathway, •OH addition is followed by H2O elimination and the whole process is highly exergonic (Fig. S7 and S8). We also propose the formation of P9 (compound 20 in Hartmann et al. (2008)) by further hydroxylation of P8 followed by C-N bond cleavage.

Dichloroaniline (P1) is the main by-product observed within 30 min of sonication by Hartmann et al. (2008). The attack of hydroxyl radical leads to the formation of the hydroxylated diclofenac derivative which by attack of another •OH and water elimination leads to the formation of an intermediate (R addb). The latter leads to P1 by C-N bond cleavage. P1 is also the major by-product characterized by Ziylan et al. (compound D1) after 90 min (Ziylan et al. 2014). Note that the Gibbs free energy of activation of P1 is as low as 22.6 kcal/mol justifying its detection as the major product in both studies (Fig. S1). Hartmann et al. have claimed that the small degradation rate for P1 at low pH values (<4) is in good agreement with the sonolysis of aniline which is stable under acidic conditions (Jiang et al. 2002). The mechanism suggested in this study corroborates this finding since the formation of P1 is exothermic by 36.9 kcal/mol (Fig. S1).

The reaction of hydroxyl radical with 2,4-dichloroaniline (P1) can easily yield P5 in which the NH2 group can be further replaced by OH to yield P6. Both reactions occur easily as P1 is already formed (compounds 16 and 17 in Hartmann et al. (2008)). Ziylan et al. have identified the same species (D8, D9) after 90 min of irradiation (Ziylan et al. 2014).

The cyclic intermediate 1-(2,6-dichlorophenyl)-indolin-2-one (P2) was previously observed during sample preparation of diclofenac under acidic conditions (Reddersen and Heberer 2003). P2 was classified as an immediate by-product in the sonolysis of Hartmann et al. (compound 2) (Hartmann et al. 2008) We propose the ring closure to take place with the interference of an explicit water molecule. The activation Gibbs free energy is higher than the ones required in the other pathways (47.0 kcal/mol). However, this barrier could have been overcome under reaction conditions. The formation of the other cyclic by-products (P3 and P4) is less demanding. P3 which occurs by decarboxylation followed by cyclization has also been detected by Ziylan et al. (compound D4) (Ziylan et al. 2014).

2,6-dichloro-4-hydroxy aniline (P5) has also been reported as the by-product in the transformation pathway of DCF degradation by photocatalysis under UV-A and simulated solar irradiation and sonophotocatalyis (Ziylan et al. 2014).

In light of the experimental progression of the pH mentioned earlier (decrease from 6.50 to 5.23 in 30 min), in the late stages of the reaction when neutral DCF is the prevalent form, it is likely that energy demanding processes, e.g., fragmentation and cyclization rates, will increase. At the beginning of the AOP when pH is close to neutral and DCF− is predominant, H abstraction and OH addition are more likely to occur, leading to subsequent degradation reactions yielding by-products P1 and P4–P9.

Toxicity

The acute and chronic toxicity of diclofenac and its by-products to aquatic organisms was predicted using the ECOSAR software at three trophic levels (green algae, daphnia, and fish). The octanol–water partitioning coefficient (log K ow) was predicted as 4.02 for diclofenac, and the value is in good agreement with the experimental finding, 4.51 (Avdeef et al. 1998). As such, DCF is potentially toxic to aquatic organisms based on the literature classifying chemicals as toxic with log K ow < 6.0 (Konemann 1981; Mayer and Reichenberg 2006). The toxicity classification based on ranges of LC50 and EC50 (Table S2) was used to rationalize the acute toxicity of DCF and its by-products (Table 4). The estimated value of EC50 (41.41) is slightly higher than that found experimentally (33.26) (Ziylan et al. 2014). LC50 of diclofenac for daphnia and fish was found as 25.75 and 37.66 mg/L, respectively. The finding is in agreement with the European Union criteria (Table S2), which classify diclofenac as a harmful compound for the aquatic environment at all three trophic levels.

Table 4 Toxicity values of DCF/DCF− and its degradation by-products P1-P9 to aquatic organisms (mg/L) using ECOSAR
Full size table

The aquatic toxicities upon degradation of diclofenac were tracked for three trophic levels (Table 4). As •OH-mediated degradation proceeds, toxicities fluctuate between harmful and toxic levels. Aquatic toxicities seem to be dependent on the presence of chlorine: P9, the only by-product that is classified as not harmful for all three organisms, is free of chlorine. P4, P7, and P8, which possess an acidic group, are also classified as not harmful for green algae. The toxicities of P1, P2, P3, P5, and P6 are higher than DCF/DCF− in regard to all three organisms. However, P5 is less toxic to fish (LC50 = 53.61 mg/L) and P6 is less toxic to daphnid (LC50 = 42.53 mg/L). Overall, the assessment of EC50 and LC50 shows that DCF and most of its oxidation by-products are harmful for aquatic organisms.

Conclusions

The study has highlighted a computational means of proposing the •OH-mediated reaction mechanism for the degradation of the emerging water pollutant, diclofenac (DCF/DCF−) in water. It was found that the reactions by OH addition and H abstraction occurred almost readily; the methylene hydrogens were more readily abstractable than the others, and the outcome is most likely the decarboxylation of the ring. The radical stabilization energies of the arenium radicals are higher than those of the radicals formed by H abstraction, indicating their longer lifetimes. The by-products (P1-P9) identified experimentally for the sonochemical degradation of DCF/DCF− were confirmed in this computational work. As reported previously, the intermediates, P5, P6, and P9, formed at a later stage from their precursors, P1 and P8. P1, P7, and P8 are the most readily formed by-products (Ziylan et al. 2014).

ECOSAR toxicity analysis revealed that DCF is toxic, while some of the oxidation by-products (P4, P7, P8, and P9) are less toxic, such that elimination of the compound by AOPs is a safe method as it provides less harmful products than the parent compound. This study is the first to provide a detailed mechanistic account of the degradation of diclofenac in aqueous medium by •OH-mediated reactions, complementing our previous work (Ziylan et al. 2014).