Detoxi ﬁ cation of oil re ﬁ ning ef ﬂ uents by oxidation of naphthenic acids using TAML catalysts

alyzetheoxidationofNAsbyactivatinghydrogenperoxide — consideredenvironmentallyfriendlybecauseitreleases only water as by-product — under ultra-dilute conditions. We tested Fe-TAML/H 2 O 2 systems on (i) model NAs and (ii) a complex mixture of NAs in oil re ﬁ ning wastewater (RWW) obtained from a re ﬁ ning site in Colombia. Given the need for cost-effective solutions, this preliminary study explores sub-stoichiometric H 2 O 2 concentrations for NA mineralization in batch mode and, remarkably, delivers substantial removal of the starting NAs. Additionally, a 72-h semi-batch process in which Fe-TAML activators and hydrogen peroxide were added every 8 h achieved 90 – 95%removalwhenappliedtomodelNAs(50mgL − 1 )anda4-foldreductionintoxicitytowards Aliivibrio ﬁ scheri whenappliedtoRWW.ChemicalcharacterizationoftreatedRWWshowedthatFe-TAML/H 2 O 2 treatment(i)reduced theconcentrationofthehighlytoxicO 2 NAs,(ii)decreasedcyclizedconstituentsinthemixture,and(iii)preferentially degraded higher molecular weight species that are typically resistant to biodegradation. The experimental ﬁ ndings, together with the recent development of new TAML catalysts that are far more effective than the TAML catalysts deployed herein, constitute a foundation for cost-effective treatment of NA-contaminated wastewater.


Introduction
Naphthenic acids (NAs, Fig. 1) are natural constituents of crude oil and bitumen, with the general formula C n H 2n−z O 2 , where n = number of C atoms and Z = hydrogen deficiency (Table S1). NAs are major corrosive agents of process equipment (Slavcheva et al., 1999) and toxic to aquatic organisms (Thomas et al., 2009;Jie et al., 2015;Scarlett et al., 2013;Scarlett et al., 2012), and often end up as toxic contaminants in wastewater from the oil sector, especially refining wastewater (RWW) and oil sands process-affected water (OSPW). Currently, most high profile, public environmental concerns of NA-contaminated wastewater relate to OSPW generated during the extraction of bitumen from the oil sands of northern Alberta, Canada. Under Canada's Fisheries act (Goverment of Canada, 1985), it is not allowed to discharge deleterious substances in water, hence a zero-discharge policy is in place for these effluents. Oil sands companies operate under a no-release strategy, resulting in on-site settling ponds of lake-sized proportions used to store OSPW for years, where NA concentrations range between 40 and 120 mg L −1 (Holowenko et al., 2001). Considering that NAs have an estimated half-life in nature of 12.8-13.6 years (Han et al., 2009), the need to accelerate their break-down in OSPW has become a priority for both the oil sands industry and Canada (Giesy et al., 2010;Wu et al., 2019).
In the case of RWW, NAs are associated with the refining of acidic crudes. After refining, NAs are released in wastewater and may reach the environment unless appropriate treatment techniques are applied. The removal of NAs from OSPW remains a priority but receives much less attention in the refining sector Pinzon-Espinosa and Kanda, 2020), making RWW an important source of NA environmental contamination (Wong et al., 1996). The composition and concentration of NAs in RWW depend on the type of feedstock, the nature of the extraction process, and the degradation profile over space and time (Toor, 2012). NAs have been reported between 2.8 and 11.6 mg L −1 in the effluent of RWW treatment (LC-MS, relative to ptoluene sulfonate) (Misiti et al., 2013), but can be present at much higher concentrations, as high as 135 mg L −1 in RWW resulting from the processing of highly acidic crudes (GC-MS, relative to 1chlorooctadecane) (Pinzon-Espinosa and Kanda, 2020).
Many methods have been investigated for the clean-up of NAcontaminated wastewater (Wu et al., 2019;Quinlan and Tam, 2015) but there has been limited success with applied technologies developed over the last decade. Biodegradation has proven to be the most costeffective technique so far but only some NAs are biodegradable under aerobic conditions (Clemente and Fedorak, 2005). Ozonation has proven to increase overall biodegradability (Vaiopoulou et al., 2015;Brown et al., 2013;Brown and Ulrich, 2015;B. Wang et al., 2013), and activated carbon-based adsorption has been shown to remove hydrophobic NAs (Zubot et al., 2012;Kannel and Gan, 2012). However, none of these treatments have advanced to scale to provide solutions for OSPW and RWW NA-contamination. There remains the urgent need for better processes that align with the ideal characteristics of a viable treatment technology, namely high environmental safety, low cost, low energy, mild conditions, and high efficiency. Catalysts can help providing such conditions. Iron-TAML activators ( Fig. 1) of H 2 O 2 have evolved through iterative design (Collins et al., 1998) as miniature (≤1% enzyme mass), mechanistically faithful replicas of peroxidase enzymes, but they greatly outperform the enzymes (Wong et al., 1996;Misiti et al., 2013;Collins, 2002). For the oxidation of NAs, this study utilizes Fe-TAML activators 1a and 1b (Fig. 1), respectively the best overall TAML catalyst for persistent pollutants and the prototype TAML. While 1b is ten times less reactive than 1a (Tang et al., 2016;Popescu et al., 2008), the fact that it is fluorine-free and relatively inexpensive are attractive features that lead us to also evaluate its properties in TAML/H 2 O 2 degradation of NAs.
Therefore, the purpose of this study has been to investigate the catalytic performance of 1/H 2 O 2 oxidation systems for the transformation of NAs in order to reduce the toxicity of NA-containing effluents. The study was conducted to (i) test 1/H 2 O 2 on model NAs (NA1 and NA2, Fig. 1) (Pinzon-Espinosa and Kanda, 2020); (ii) examine the impact of pH on the performance of 1/H 2 O 2 in NA-contaminated water in the range considered relevant for RWW and OSPW (7-9); (iii) test 1/H 2 O 2 on a RWW sample known to contain a complex mixture of NAs, and (iv) undertake toxicity testing of samples treated with 1/H 2 O 2 using the luminescent bacteria test (LBT) based on Aliivibrio fischeri to assess toxicity reduction. Bacteria-based tests are known to detect chemicals with several modes of actions, rendering a wide range of detection, and have low requirements of sample. A. fischeri comes in the form of a standardized lyophilized reagent providing a low coefficient of variation.

Chemicals and reagents
Model NAs (4-methyl-1-cyclohexanecarboxylic acid and 4-(4-tertbutylphenyl)butanoic acid; Fig. 1 27 Da] as orange crystalline Na + salts with one axial water ligand and one water of crystallization) were provided by Carnegie Mellon's Institute for Green Science. H 2 O 2 (30% w/w) was purchased from Sigma Aldrich and diluted to 0.3% using Milli-Q® water before use. The stock solution (30% w/w) was standardized every 48 h by redox titration (KMnO 4 , Sigma Aldrich). Na 2 C 2 O 4 for standardization of KMnO 4 was purchased from Fisher Chemical. Unbuffered Milli-Q® water was used as reaction medium. The pH of reaction mixtures was adjusted by addition of 0.01 M KOH or 0.01 M HCl. Catalase from bovine liver was obtained from Sigma Aldrich. The Internal standard (IS) solution (1-chlorooctadecane and o-terphenyl; 4000 μg mL −1 each) was obtained from Restek UK. Phenol and K 2 Cr 2 O 7 for toxicity testing were purchased from Sigma Aldrich. Stock solutions of 17α-ethinylestradiol (EE2) (10 mg L −1 ) and TAML catalysts (40 mg L −1 ) were prepared in water and stored at 4°C. Stock solutions of catalase (500 mg L −1 ) were prepared in water and stored at 4°C for a maximum of one week.
The experimental design entailed two controls (NA only, NA + H 2 O 2 ), two catalysts (1a and 1b), and two treatments (batch and semi-batch) for each model NA (NA1 and NA2) at each pH value (7, 8, and 9) (Fig. S1), for a total of 6 test sets and 18 reaction vessels per set. The semi-batch treatment corresponded to one aliquot of H 2 O 2 (0.3%, 134 μL) and one of 1/H 2 O 2 (4000 μg L −1 , 250 μL) every 8 h, for a total of 9 aliquots. This served as an optimization stage for the subsequent degradation of NAs in RWW samples. For quantification of NAs, aliquots of 500 μL were sampled every 8 h in 2-mL glass vials. An aliquot of catalase solution (500 mg L −1 , 50 μL) was transferred to each 2-mL glass vial and left for 5 min to quench the H 2 O 2 and stop the reaction. The pH was then lowered by adding HCl (1 M, 50 μL) to allow detection of the NAs acid derivatives via high performance liquid chromatography (HPLC) coupled to a diode array detector (DAD).

Fe-TAML/H 2 O 2 degradation of NAs in RWW
Degradation reactions were conducted in duplicate for 72 h in 500-mL conical flasks containing 200 mL of an RWW sample known to contain NAs (Pinzon-Espinosa and Kanda, 2020). Reactions were initiated by the addition of one aliquot of H 2 O 2 (3%, 270 μL) to samples containing TAML catalyst (100 μg L −1 : 1a = 186 nM, 1b = 206 nM). MilliQ® water, RWW alone, and an uncatalyzed reaction of RWW were used as controls. Reactions were conducted in semi-batch mode, corresponding to 9 added aliquots of H 2 O 2 (3%, 270 μL) and 1/H 2 O 2 (40 mg L −1 , 500 μL). Reactions were stopped by adding catalase (500 mg L −1 , 5 mL) and left for 15 min for quenching. An aliquot (5 mL) was taken from each flask for toxicity tests; the remaining sample underwent liquid extraction prior analysis, as described below.

Liquid-liquid extraction (LLE)
An aliquot of the IS solution (20 μL) was added to post-treatment RWW samples to a final concentration of 100 μg L −1 of each component, and the pH was then adjusted to 2 using 1 M HCl. Samples were transferred to separatory funnels and extracted (x3) with dichloromethane (20 mL); the resulting organic extracts were combined and reduced to incipient dryness (a small volume without fully drying them out, hence avoiding evaporative losses of the lower molecular weight compounds) in a TurboVap® LV workstation. Extracts were then split in half and re-dissolved in dichloromethane (100 μL) for analysis via gas chromatography coupled with mass spectrometry (GC-MS) for semiquantification of naphthenic acids and in methanol (100 μL) for liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) for identification of individual congeners within the NA mixture.

HPLC-DAD analysis
Depletion of model NAs in water was calculated by measuring the residual NAs remaining in the reaction medium after oxidation. Quantification was performed using HPLC-DAD on an Agilent 1260 HPLC instrument (Agilent 1260 pump, Agilent degasser, Agilent DAD, column oven, and auto-sampler) using a Hyperclone® C8 (150 × 2 mm, 5 μm, Phenomenex) column. Mobile phases consisted of HPLC-grade water (A) and acetonitrile (B), pumped at 0.7 mL min −1 . Gradient elution was 1% B 0-1 min, ramped to 100% B by 2 min and held until 3 min, returning to 1% B by 4 min and held for 2 min, for a total run of 6 min. Monitoring of NA1 and NA2 was carried out at 210 nm (RT: 3.95 min ± 0.01) and 254 nm (RT: 4.11 min ± 0.01), respectively. External calibration solutions were prepared in Milli-Q® water.

GC-MS analysis
Depletion of NAs in RWW was calculated by semi-quantifying NAs in LLE extracts after treatment with 1/H 2 O 2 . Analyzes were conducted using a Perkin Elmer Clarus® 500 instrument equipped with a DB-5 capillary column (30 m × 0.25 mm I.D.) coated with 0.25 μm film 5% phenyl polysilphenylene siloxane. Carrier gas was high-purity helium at 1.0 mL min −1 . The inlet was held at 250°C and injection volume was 1 μL. Column was held at 35°C for 4 min, ramped at 8°C/min to 310°C, and held for 10 min, for a total run time of 48 min. MS was operated in electron ionization mode at 70 eV (scan range 50 to 600 amu). Perfluorotributyl-amine (PFTBA) was used for calibration. Filament emission current was 0.06 pA.

LC-HRMS analysis
LLE extracts from post-treatment RWW samples were analyzed using HRMS on a Thermo Exactive® mass spectrometer using a Thermo Accela LC pump and a CTC autosampler. Separation was conducted using a Varian Pursuit XRs C18 (100 × 3.0 mm, 3 μm, 100 Å) column. Mobile phases consisted of 0.1% NH 4 OH in HPLC water (A) and 0.1% NH 4 OH in methanol (B), pumped at 600 μL min −1 . Gradient elution was 10% B from 0 min to 2 min, ramped to 70% B by 2.5 min and held until 6 min, ramped to 100% B by 6.5 min and held until 7.5 min, returning to 10% B by 9 min and held for 3 min, for a total run of 12 min. Detection was performed in negative ion mode (scan range 80-500 m/z) with the following settings for the heated-electrospray ionization (H-ESI) source: sheath gas flow rate 50 units; spray voltage A. Pinzón-Espinosa, T.J. Collins and R. Kanda Science of the Total Environment 784 (2021)

Toxicity evaluation
Toxicity of aqueous RWW after treatment with 1/H 2 O 2 was measured using a modified version of the LBT methodology described in BS EN ISO 11348-3:2008, adapting the procedure to 96-well plates (Supplementary information -Toxicity evaluation section). Samples were analyzed in duplicate using phenol (400 mg L −1 in saline solution) as reference substance (expected EC 50 = 13-26 mg L −1 ), saline solution (20% NaCl) as negative control, and Cr(VI) (105.8 mg L −1 of K 2 Cr 2 O 7 in saline solution) as positive control. A Promega GloMax™ luminometer was used for light readings, and incubation was performed in an A. Pinzón-Espinosa, T.J. Collins and R. Kanda Science of the Total Environment 784 (2021) 147148 Aqualytic thermostatic cabinet at 15°C ± 0.3. EC 50 values were determined using the linear regression approach.

Statistical analyzes
Two-way analysis of variance (ANOVA) and the Tukey's multiple comparison test were used to determine significant differences within and between treatments using Prism version 7.03 (Graphpad software, San Diego, CA). Significance level (α) was 0.05.
Results for activity checks of 1a and 1b aqueous solutions conducted before NA degradation are shown in Fig. S2.

Performance of Fe-TAML catalysts for the oxidation of model NAs
Due to the particularly high volumes of NA-contaminated wastewater in the oil sector, extreme efficiency is essential to favor low costs and the eventual practicality of any potential real-world process. Thus, this study became all about discovering if effective degradation of NAs could be achieved with relative concentrations of NAs, catalyst, and hydrogen peroxide that would have commercial viability. Fig. 2 shows the performance of 1a/H 2 O 2 and 1b/H 2 O 2 in the catalytic degradation of NA1 and NA2. Because of the critical importance of the relative amounts of NAs, peroxide, and TAML catalysts, their relative quantities for both batch and semi-batch processes are presented in Table 1 and the resulting molar ratios in Table 2. The peroxide mineralization requirement ([H 2 O 2 ]/[NA]) for NA1 = 21 and for NA2 = 36, which were exceeded in semi-batch reactions, as observed in Table 2. Oxidation reactions conducted under batch conditions (1 at 100 μg L −1 , 1a = 186 nM, 1b = 206 nM; H 2 O 2 at 40 mg L −1 , 1176 μM) proved to be insufficient to fully degrade model NAs, removing <45% (Table S2). However, it is noteworthy that these reactions were conducted under sub-stoichiometric conditions (Table 2) and, yet TAML catalysts delivered a substantial removal under such demanding circumstances. Further research would be required to understand the mechanistic principles behind these results.
As has been found for metaldehyde, particularly persistent compounds, such as NAs, can be degraded by repeated treatments with TAML/oxidant (Tang et al., 2016). Analysis of semi-batch oxidation reactions revealed that lower final concentrations of model NAs were consistently achieved under 1a/H 2 O 2 vs 1b/H 2 O 2 at all pH values (Fig. 2), which was anticipated due to the increased oxidative aggression provided by the electron-withdrawing capacity of the NO 2 group (Popescu et al., 2008). Moreover, superiority of 1a over 1b was statistically significant at all pH values for both NAs (p = 0.9997-<0.0001) (Popescu et al., 2008). As shown in Fig. 2, semi-batch reactions with 1a achieved degradation rates of approximately 95% for NA1 and 90% for NA2 within 72 h, demonstrating the suitability of Fe-TAML/H 2 O 2 systems to decompose model NAs under laboratory conditions. The corresponding values with 1b are 75.6% at pH 7, 68.3% at pH 8, and 70.7% at pH 9 for NA1, and 48.4% at pH 7, 49.0% at pH 8, and 49.1% at pH 9 for NA2.
3.3. Effect of initial pH on the performance of 1 in catalyzing the oxidative degradation of model NAs by H 2 O 2 The catalytic reactivity of TAML activators is highly pH dependent with the highest rates having been found near pH 9 for 1a and near pH 10 for 1b for multiple substrates (Collins et al., 2014;Su et al., 2018;Ellis et al., 2010). Previous studies with TAML catalysts have been conducted using buffered solutions as a matrix, but we conducted the oxidation reactions in unbuffered water to approximate real-world water treatment. Degradation took place in milliQ® water with initial pH values of 7, 8, or 9; these pH values were selected based on the native pH of the RWW samples (~7) and the reported pH values for OSPWtypically between 8 and 9 (Toor, 2012)-because these effluents are important sources of NAs into the environment.
Previous studies using ozone to oxidize NAs have also reported higher efficiency at basic pH (Perez-Estrada et al., 2011;Afzal et al., 2015), which has been linked to the decomposition of ozone at high pH values and the resulting formation of hydroxyl radicals (•OH) (Perez-Estrada et al., 2011;Glaze et al., 1987). The non-selectivity of •OH plays a key role in the removal of NAs by hydrogen abstraction (Glaze, 1987;Meshref et al., 2017). However, •OH seem not to be significantly involved in Fe-TAML oxidation reactions, and the pH-dependent efficiency is related to the acidity of the axial water ligand which, when deprotonated, promotes the activation of H 2 O 2 (Ryabov and Collins, 2009;Ghosh et al., 2003) to form the Fe(V)(oxo)-TAML as the most aggressive reactive intermediate (Ryabov and Collins, 2009;de Oliveira et al., 2007).

By-product formation from degradation of NA2
HPLC-DAD analysis of the reaction medium after 1a/H 2 O 2 degradation showed the formation of a by-product with increasing peak area over time (Fig. S4), which was not present in untreated samples. LC-HRMS results confirmed the presence of an additional compound that Table 1 Relative amounts of NAs, H 2 O 2 , and TAML catalysts during batch and semi-batch reactions.

Reagent
Reaction process co-eluted with NA2 in an unresolved peak (Fig. 3). The peak was composed of 3 unresolved peaks corresponding to NA2 in two isomeric forms (

Degradation of NAs in RWW using Fe-TAML/H 2 O 2
Reactions with model NAs showed the superiority of 1a over 1b and the higher catalytic performance of 1a at pH 9 for degrading NA2.
Hence, these conditions were tested in an environmental mixture of NAs using an RWW sample collected at a refining site in Colombia. Reactions were conducted at pH 9 for 72 h in a semi-batch process with final concentrations of 900 μg L −1 (1674 nM) for 1a and 360 mg L −1 (10,584 μM) for H 2 O 2 . To determine technical performance, four endpoints were monitored: (i) toxicity as measured by the LBT, (ii) concentration of NAs as determined using total ion chromatograms (TICs) from GC-MS, (iii) relative abundance of classic and oxy-NAs identified using LC-HRMS and, (iv) NA profiles based on n values and Z families obtained from HRMS data.
Toxicity results confirmed that no toxicity was exerted by the reaction matrix after quenching with catalase. Results also showed that treatment with 1a/H 2 O 2 reduced the toxicity of RWW by 4-fold after only 72 h (Fig. 5). Ozone alone has been reported to reduce toxicity by 3.3-fold, but this required the continuous supply of ozone throughout the reaction (60 min; constant ozone concentration of 25-35 mg L −1  in the gas-phase) (Vaiopoulou et al., 2015) implying high operational costs in any scaled-up process. Semi-quantification of NAs was based on the peak area of the unresolved hump assigned to NAs in GC-MS TICs. The single point external standard method was followed using 1-chlorooctadecane (RSD = 18.3%, spiking concentration 100 μg L −1 ) as IS for calculation purposes, with the formula below:

Concentration of NAs ¼ Peak area of NAs Peak area of IS Â Concentration of IS
NAs were estimated to be present in test samples in the concentrations shown in Table 3. These results indicate that the total NA content did not decrease after treatment. However, the reason behind the variability of the untreated and uncatalyzed samples is unclear. These concentration values, however, must be interpreted with caution because the detector does not respond identically to 1-chlorooctadecane and NAs, so an accurate quantitation would require a multiple point standard method using known amounts of the NAs present in the UCM. However, LC-HRMS profiles (Figs. S5 and S6) did show that the semibatch reaction with 1a/H 2 O 2 changed the distribution and abundance of NA species in RWW, explaining the transformation in toxicological properties regardless of the total NA content.
LC-HRMS analysis revealed that the relative abundance of O 2 NAs decreased after treatment with 1a/H2O2, with significant differences with respect to untreated and uncatalyzed RWW (α = 0.05; p = <0.0001), whereas relative abundance of O3 and O4 NAs increased with statistical significance (α = 0.05; p <0.0001) (Fig. S5). This finding seems to corroborate the result obtained with NA2 (Fig. 4), suggesting that semi-batch reactions with 1a (900 μg L −1 , final concentration) and H 2 O 2 (360 mg L −1 , final concentration) induce the oxidation of O 2 species into oxy-species, which is crucial for the treatment of RWW and OSPW because O 2 NAs have been regarded as the most toxic (Morandi et al., 2015;Yue et al., 2015). Considering that the most probable mode of action for acute toxicity of NAs is narcosis (Frank et al., 2009), which depends on hydrophobicity and size of a molecule, the reduction in toxicity towards A. fischeri might be linked to the reduction of hydrophobicity of alkyl groups due to the oxidation of O 2 into O 3 species by 1a/H 2 O 2 .
The NA profiles for classic (O 2 ) and oxy-NAs (O 3 , O 4 , and O 5 ) (Fig. S6) revealed that samples undergoing uncatalyzed treatment or no treatment at all showed the same distribution of NA congeners. Treatment with 1a/H 2 O 2 changed the overall profiles of oxy-NAs, with an evident increase in the intensity of O 4 NAs with C11-13 and Z = −2, and of Z = −2 and −4 in O 5 NAs. Statistical analysis (Fig. 6A) revealed significant differences in Z = 0, −2, −4, and −8 for O 3 , and Z = 0, −2, −4, −6, and −8 for O 4 in treated samples with respect to controls. Overall, relative abundance of species with high Z values declined and those with low Z values increased, possibly due to ring opening (Vaiopoulou et al., 2015;Afzal et al., 2015;Meshref et al., 2017;Scott et al., 2008). This suggests that 1a/H 2 O 2 under the conditions described herein decreased the degree of cyclization of the NA mixture, which follows the trend found with oxidation using electro-oxidation (Abdalrhman et al., 2019) and ozonation (Perez-Estrada et al., 2011). This has important practical implications because cyclization remains a major factor contributing to persistence, with faster biodegradation rates in NA mixtures with lower degrees of cyclization (Han et al., 2008).
As for the carbon content, we used n = 15 as margin to classify NAs into low and high molecular weight (MW) (Meshref et al., 2017;Sohrabi et al., 2013) and determine whether treatment with 1a/H 2 O 2 had any effect on high MW species, which are reported to be less biodegradable. As set out in Fig. 6B, both uncatalyzed and catalyzed treatments reduced the relative intensity of high MW species within O 2 NAs from 11% to 4%, which can be linked to oxidation by hydroxyl radicals from H 2 O 2 . In contrast, low MW NAs in O 3 and O 4 species increased with respect to controls only after treatment with 1a/H 2 O 2 , reaching 90% and 43% respectively. In alignment with other oxidation processes (Perez-Estrada et al., 2011;Afzal et al., 2015;Meshref et al., 2017;Abdalrhman et al., 2019), 1a/H 2 O 2 showed preferential oxidation towards high MW NAs, which can be attributed to the increasing number of oxidation sites and H atoms vulnerable to abstraction by hydroxyl radicals (Perez-Estrada et al., 2011  Evidence indicates that a combination of chemical oxidation and biological degradation could tackle the complex nature of NAs mixtures and their structure-specific resistance to treatment (Zhang et al., 2019;Martin et al., 2010), but the cost of upscaling such technologies remains the major fallout of this approach. Therefore, catalysts can play a key role in increasing oxidation efficiency, reducing the requirements of reagents and resulting costs. Fe-TAML/H 2 O 2 systems are showing a great potential to do so.
Ozone has been reported to be an effective advanced oxidation technology to oxidize model NAs and NAs from OSPW, increasing their biodegradability and reducing toxicity towards A. fischeri (Vaiopoulou et al., 2015;Afzal et al., 2015;Hwang et al., 2013;Gamal El-Din et al., 2011). For instance, complete detoxification of OSPW (based on the Microtox® assay) has been achieved after only 50 min of ozonation . However, when used alone, ozone is frequently applied in semi-batch mode with the gas running continuously (Vaiopoulou et al., 2015;Al-jibouri et al., 2015), thereby providing excess ozone for the duration of the reaction . The high cost of ozonation (due to the high consumption of energy and ozone) for the large quantities of NA-contaminated wastewater typically produced in the petroleum sector has been acknowledged by previous studies, questioning the economic feasibility of ozone-based technologies for removal of NAs in realistic scenarios Gamal El-Din et al., 2011). In fact, effective concentrations of ozone reported in previous research can go up to 150 mg L −1 (Afzal et al., 2015;Gamal El-Din et al., 2011;N. Wang et al., 2013). Dosing has been referred to as intensive at concentrations equal or above 80 mg O 3 /L, mild at 30-50 mg O 3 L −1 (Xue et al., 2016), and light at approximately 25 mg L −1 (Vaiopoulou et al., 2015), but even the latter is cost-prohibitive taking into account the large volumes of RWW or OSPW produced daily (Diya'uddeen et al., 2011). For instance, it has been estimated that for municipal wastewater implementing ozonation at 7.7 mg O 3 /L with a retention time of 25 min (10 kWh/kg of produced ozone per hour) would increase treatment cost by 0.18-0.22 €/m 3 in the Netherlands and from 0.10-0.18 €/m 3 in Germany (Mulder et al., 2015). It is noteworthy that these costs do not account for recalcitrant contaminants, such as NAs, which might increase operational costs even more. Consequently, ozone alone is hardly applicable for the treatment of RWW or OSPW. Instead, it is now seen as a potential pre-treatment for biological degradation (Xue et al., 2016;Oller et al., 2011) because it targets NAs with greater molecular weight (higher possibility for hydrogen abstraction) and number of rings (more tertiary carbons, which have higher reactivity), which tend to be more resistant to biodegradation (Perez-Estrada et al., 2011;Afzal et al., 2015). Likewise, adsorption has been proposed as a pre-treatment step of ozonation to reduce the levels of applied and utilized ozone (Gamal El-Din et al., 2011).
Our results show the effectiveness of a semi-batch reaction (72 h total reaction time, room temperature (~15°C)) with 1a (final concentration 900 μg L −1 ) and H 2 O 2 (final concentration 360 μg L −1 ) to detoxify RWW by a factor of 4, thereby comparing favorably with a similar approach undertaken previously (Vaiopoulou et al., 2015), where ozone at 25-35 mg O 3 /L resulted in a 3.3-fold reduction in toxicity. Further reduction in toxicity towards A. fischeri was achieved in the same study when ozone was used in combination with biodegradation, as this approach targets a wide range of NAs, and therefore further work is required to determine if the same benefit can be obtained by using Fe-TAML/H 2 O 2 and biodegradation combined. New generations of TAML activators have been developed recently whose catalytic performance is currently under testing for NA degradation and detoxification of NA-contaminated wastewater, aiming at lower concentrations of TAML catalysts and lower requirements of H 2 O 2 .

Conclusions
In this study, we provided an insight into the efficacy of Fe-TAML catalysts (1a and 1b)/H 2 O 2 systems to oxidize model NAs and detoxify  Fig. 6. Changes in the NA mixture in RWW after treatment with 1a/H 2 O 2 . A. Heat map showing statistically significant changes in relative abundance of Z families for O 2 to O 5 NAs in treated RWW with respect to controls; B. variations in the carbon number of NAs in untreated, uncatalyzed, and catalyzed RWW using C15 as margin to separate NA congeners into low and high MW species.
A. Pinzón-Espinosa, T.J. Collins and R. Kanda Science of the Total Environment 784 (2021) 147148 RWW containing NAs. It was demonstrated that only very low concentrations of 1a (900 μg L −1 ) were required to oxidize aqueous solutions of NAs at high concentration (50 mg L −1 ), achieving up to 95% degradation of model NAs within 72 h in semi-batch mode. An oxidation by-product revealed hydroxylation of the parent NA by means of LC-HRMS. Catalyst 1a proved to be superior to 1b, which is explained by the increased oxidative aggression provided by the electronwithdrawing capacity of the NO 2 group (Popescu et al., 2008;Collins et al., 2014) in its ligand structure. Treatment with 1a/H 2 O 2 of RWW reduced abundance of O 2 NAs and increased that of oxy-species, reduced degree of cyclization of the NA mixture, and showed preferential oxidation towards high MW NAs. This is comparable with ozonation, which is one of the most effective oxidation technologies for the removal of NAs from OSPW (Quinlan and Tam, 2015;Vaiopoulou et al., 2015;Afzal et al., 2015;Meshref et al., 2017). However, the concentrations of 1a required for treatment are significantly lower than the required ozone dose to achieve the same results, which is highly relevant for up-scaling purposes. Further work is required to evaluate additional Fe-TAML/H 2 O 2 systems in combination with biodegradation to achieve lower residual NAs concentrations in wastewater and lower toxicity. Future studies also need to expand the range of toxicity bioassays to assess the environmental quality of the resulting effluent, and expand on the analysis of by-products, thus providing a more comprehensive study of the fate of NAs after advanced oxidation via Fe-TAML/H 2 O 2 . Further research is underway to evaluate new generations of Fe-TAML activators aiming at reducing H 2 O 2 concentration and reaction times. Overall, our findings suggest that Fe-TAML/H 2 O 2 systems are good candidates for reducing the toxicity of effluents containing these pollutants, moving a step forward in water sustainability of the petrochemical sector.

Funding sources
This research was supported by the PhD Scholarship program from COLCIENCIAS (National Department of Science, Technology, and Innovation of Colombia), Grant 568/2012. COLCIENCIAS had no involvement in the design or development of the study, nor in the writing of this manuscript.