Screening techniques as a preliminary diagnostic tool for advanced oxidative processes on a laboratory scale

Abstract

This study introduces an innovative screening approach to evaluate advanced oxidation processes (AOPs) as a preliminary diagnostic tool for degrading emerging contaminants (EC). It includes the design, prototyping, and cost-benefit analysis of circular photochemical reactors with flat and spiral internal geometries. Three-dimensional (3D) printing was used for reactor prototyping, providing flexibility and economy, and this stage was assisted by the hydrodynamic analysis of the prototypes based on residence time distribution (RTD) and macromixing models. The research evaluates the degradation of a model contaminant of emerging concern, fluoxetine (FLX) hydrochloride, using the solar/persulfate (PS) process in two water matrices (i.e., ultrapure water and sewage treatment plant effluent) to optimize reactor performance. The study also proposes primary theoretical pathways for fluoxetine degradation involving hydroxyl and sulfate radicals, as well as predicting the toxicity of the parent compound and its primary metabolites using quantitative structure-activity relationship (QSAR) models. The spiral reactor exhibits improved hydrodynamic behavior, closely resembling continuous stirred and plug flow reactors in series. Despite a slightly lower specific degradation rate in real wastewater, the solar/PS process remains effective for both matrices. By-products generated via the sulfate radical pathway are expected to be less toxic than those formed by hydroxyl radicals (HO·) attack.

Graphical abstract

Keywords

emerging contaminants / advanced oxidation process / three-dimensional printed reactor / quantum chemical calculations / toxicity

1 Introduction

Advanced oxidation processes (AOPs), with their strong oxidative abilities, have demonstrated high efficiency in the decomposition of emerging contaminants (EC). They produce powerful oxidative radicals, especially hydroxyl radicals (HO·), known for their remarkable reactivity, offering technical, economic and environmental advantages. AOPs cover several categories, including chemical, photochemical, sonochemical, and electrochemical processes, offering a versatile arsenal for wastewater treatment and environmental remediation [1].

However, the application of AOPs is still largely focused on conducting experiments in conventional reactors, such as quartz vessels, beakers, or ductless reactors [2,3]. This approach imposes limitations related to reactor size, flow type, radiation field (in the case of photochemical reactors), and increased consumption of reagents [4,5]. In addition, another limitation regarding the application of AOPs is the possible generation of degradation products that are more toxic and less biodegradable than the original compounds [6]. Therefore, in conjunction with the selection of suitable designs and processes, the identification of transformation products (TP) is crucial to predicting the impact of degradation processes. In most cases, the literature is rather restricted in this regard, as experimental procedures aimed at revealing degradation mechanisms have proven to be analytically challenging, complex, expensive, and equipment-dependent.

The increasing use of AOPs for the removal of EC in aqueous media depends on increasing their effectiveness [1]. Therefore, process screening can serve as an attractive alternative to aid process selection, acting as an initial guiding analysis. This screening process can be grounded in an initial understanding of both reactor designs, along with the application of three-dimensional (3D) manufacturing techniques to create different reactor configurations, and process performance, facilitated by theoretical chemical methods to evaluate byproducts. These specific designs, with a smaller environmental footprint, allow for greater autonomy compared to studies in conventional batch systems (i.e., beakers or flasks) [7]. This eliminates the excessive use of expensive reagents, reduces the generation of chemical waste, and allows for more reliable results. At the same time, theoretical predictions can be employed to understand the reaction mechanisms, providing a detailed representation of the oxidation pathways, with outcomes that have been observed to be strikingly realistic [8]. AOPs primarily generate HO· and sulfate radical (SO₄·⁻), which are potent oxidants that can significantly reduce or mineralize ECs in wastewater. However, SO₄·^– has proven to be a particularly efficient oxidizing agent for EC mitigation. Compared to HO·, SO₄·⁻ offers several advantages: they require simpler raw materials, are less impacted by organic matter and water alkalinity, and exhibit a higher quantum yield and greater electrophilicity [9].

Nevertheless, the non-ideal nature of real reactors is a prominent feature of such equipment, thereby requiring a preliminary examination for practical application [10]. Few works in the literature seek to understand the relationship between the intrinsic flow characteristics of reactors (e.g., experimental and/or full-scale) and the oxidation process carried out in them [11]. Determining the residence time distribution (RTD) has been a widely used methodology for this purpose, as well as being a relatively simple and reliable experimental technique. Complemented by a strong grasp of design principles, the integration of theoretical approaches with toxicological predictions by means of quantitative structure-activity relationships (QSAR) may constitute an intriguing approach to initiate an understanding of EC-related environmental concerns.

In this context, for proposing an innovative screening approach to the study of AOPs applied to the degradation of EC in aqueous matrices, this study presents: (1) the design, prototyping, and cost analysis of circular photochemical reactors with two distinct internal geometries; (2) the hydrodynamic study of the prototyped reactors, encompassing an examination of RTD and macromixture models; (3) a comparison of the degradation of the model EC fluoxetine (FLX) in two different aqueous matrices by means of solar/persulfate (PS) processes; (4) the proposal of the primary photochemical degradation pathways of FLX in reactions with HO· and SO₄·⁻ by means of a theoretical approach; and (5) the evaluation of the toxicity of the parent compound and its primary metabolites using a QSAR model in combination with ecotoxicological risk assessment.

2 Experimental

2.1 Design and prototyping

The photochemical reactors were conceptualized based on unit modular designs (Fig.1), elaborated in 2D CAD software (Autodesk, AutoCAD 2021) and edited in 3D Fusion 360 software (Autodesk, Fusion 360 2021), with circular geometry and transverse fluid inlet [12]. Then, the reactors were manufactured from acrylonitrile butadiene styrene, obtained by recycling waste from 3D printing and automotive parts (PrintGreen 3D), by deposition of molten filament in a heated chamber extruder (Creator Pro, Flashforge, Inc.). This extrusion was performed under a nozzle temperature of 220 °C, with a speed of 50 mm·s⁻¹ on a platform heated to 100 °C in layers of 120 µm. The geometry of each of the reactors was built as a single piece, which was processed in proprietary software (FlashPrint, Flashforge, Inc.) for geometry slicing and preparation of the printing instructions (G-code). A 90% infill (with hexagonal patterns) was used which, coupled with the use of five solid layers for the flow surface and five wall layers, ensures the tightness of the part under standard operating conditions. Each reactor is equipped with a circular borosilicate glass window for irradiation. The reactors are irradiated from the top using an HgI₂ lamp (Subsubsection 2.3.2). Reactor A (flat) contains a circular inner chamber with a diameter of 8 cm and a height of 5.0 mm, comprising an irradiated area of 38.5 cm² with a designed internal volume of 24.7 cm³. In contrast, the inner chamber of reactor B (spiral) is equipped with a spiral flow guide, restricting the flow to a single 1.2 m long channel with a rectangular cross-section (1.4 mm × 5.0 mm), resulting in an internal volume of 11.6 cm³ and an irradiated area of 22 cm².

Fig.1 Reactor modules with different flow geometries: (a) flat (volume = 25 mL), and (b) spiral (volume = 8 mL).

Full size|PPT slide

2.2 Hydrodynamic behavior

The RTD assays of the prototyped reactors were performed using the pulse tracer methodology. A concentrated methyl orange (Merck) solution of 8.51 × 10 ⁻⁴ mol·L⁻¹ was used as a colorimetric tracer. This solution was injected in pulse form directly into the reactor inlet with the aid of a syringe (volume = 100 µL) equipped with a needle. At the same time, the reactor was fed with water (transport fluid) by a peristaltic pump (Gilson, MINIPULS^® 3) at a temperature of approximately 25 °C. The reactor outlet was connected in-line with a transmission cell coupled to a UV-Vis spectrophotometer (Cary 50, Varian Co.), and the concentration profiles used for RTD calculations could be determined for two different flow rates (${\upsilon }_0=$ 5.0 mL·min⁻¹ and ${\upsilon }_0=$ 10 mL·min⁻¹). The methodology used in the calculations can be summarized as follows [10]. The space time is given by Eq. (1):

(1)$\tau =\frac{V}{{\upsilon }_0},$

where V (m³) is the reactor volume and υ₀ is the flow rate (m³·h⁻¹).

The average residence time t_m, in turn, is calculated as a function of the output curve E(t):

(2)$t_{\mathrm{m}}=\frac{{\int }_0^{\mathrm{\infty }}tE\left(t\right)\mathrm{d}t}{{\int }_0^{\mathrm{\infty }}E\left(t\right)\mathrm{d}t}={\int }_0^{\mathrm{\infty }}tE\left(t\right)\mathrm{d}t,$

where E(t) is the exit function, determined from the evolution of the tracer concentration (C(t)) over time (t), measured at the reactor outlet (Eq. (3)):

(3)$E\left(t\right)=\frac{C\left(t\right)}{{\int }_0^{\mathrm{\infty }}C\left(t\right)\mathrm{d}t}.$

From E(t), one can calculate the variance (σ²) and the distortion (s³) of the distribution, indicating the hydrodynamic quality of the flow:

(4)${\sigma }^2={\int }_0^{\mathrm{\infty }}(t-t_{\mathrm{m}}{)}^{2}E\left(t\right)\mathrm{d}t,$

(5)$s^3=\frac{1}{{\sigma }^{\frac{3}{2}}}{\int }_0^{\mathrm{\infty }}(t-t_{\mathrm{m}}{)}^{3}E\left(t\right)\mathrm{d}t.$

For a deeper understanding of the fluid behavior inside the reactors, the following macroscopic models, based on RTD data, were employed: (1) laminar-flow reactor (LFR) model, which has only one parameter (t_m) and assumes laminar flow in the reactors; (2) tanks-in-series (T-I-S) model, which divides the entire volume into a number (N) of stirred tanks of equal capacity with t_m; (3) continuous stirred-tank reactor and plug flow reactor (CSTR + PFR) in series, which describes the reactors as a CSTR coupled to a PFR in series, with t_m = t_{m CSTR} + t_{m PFR}, where t_m refers to the average residence time described by the CSTR (t_{m CSTR}) and the PFR (t_{m PFR}); (4) axial dispersion (AD) model, which measures material dispersion throughout the reactors using the Peclet number (Pe) and t_m.

2.3 Experimental investigation

2.3.1 Chemicals

FLX (C₁₇H₁₈F₃NO·HCl) was supplied by Campos compounding pharmacy in the form of the hydrochloride salt. Sodium PS (≥ 98%, Na₂S₂O₈) was purchased from Sigma-Aldrich. Acetonitrile (Sigma Aldrich) and trifluoroacetic acid (TFA, Sigma Aldrich) were used to prepare the mobile phases used in liquid chromatography. Methyl orange Merck was used as a tracer in the determination of the RTD. Pure water (18.2 MΩ·cm) was obtained from a Milli-Q^® Direct-Q system (Merck Millipore) and used throughout the experiments.

2.3.1.1 Real matrix

Experiments were also performed with a real matrix of the effluent from the sewage treatment plant with membrane bioreactor technology in the city of Campinas, Brazil and operated by SANASA (Water Supply and Sanitation Company). Samples were collected on January 17, 2023, and stored in amber glass bottles under refrigeration until analysis. The physicochemical parameters of this sample are presented in Table S1 (cf. Electronic Supplementary Material (ESM)) and were provided by SANASA.

2.3.2 Photodegradation experiments

First, 250 mL of a FLX solution ([FLX]₀ = 10.27 ± 0.26 mg·L⁻¹ = 29.76 ± 0.75 μmol·L⁻¹) was used in all experiments. FLX was selected as a model EC, considering the persistence of this drug [13] and its negative impacts on the ecosystems [14]; its initial concentration was based on values found in wastewater for antidepressants [15]. After the addition of PS ([PS]₀ = 0.07 mol·L⁻¹, according to the average concentration used in the literature [16]), the solution was pumped through the spiral reactor in a recirculated batch mode using a peristaltic pump (Gilson, MINIPULS^® 3) under optimal hydrodynamic conditions (flow rate (Q) = 5 mL·min⁻¹). The reactor was irradiated by an HgI₂ lamp (Master HPI-T Plus, Philips Co.), providing 30.66 mW·cm⁻² (290–800 nm). The temperature of the liquid was kept constant at about 25 °C. Samples (1.5 mL) were collected over selected times from the recirculation vessel (0, 5, 10, 15, 30, 45, and 60 min), filtered, mixed with methanol and analyzed by high-performance liquid chromatography (HPLC).

2.3.3 Analytical methods

FLX concentrations were monitored by a HPLC system (Shimadzu, series 20) equipped with a diode array detector (SPD-M20A) using a C18 column (4.6 mm × 250 mm, 4 µm), 35% acetonitrile: 65% H₂O (0.1% TFA) as mobile phase, at a flow rate of 1.0 mL·min⁻¹. The injection volume was 50 µL, the column temperature was maintained at 40 °C and the wavelength was 230 nm. This analytical method resulted in a retention time of 11 min, limit of detection of 0.143 mg·L⁻¹ and limit of quantification of 0.435 mg·L⁻¹.

2.4 Theoretical approach

2.4.1 Quantum chemical calculations

Scan of calculations with larger basis sets and other DFT functionals were considered, with an analogous protocol applied previously in references [17,18]. These calculations were performed specifically for HO· and it was selected at the M06HF/6-31 + G(d) level for all quantum mechanical electronic structure calculations. Once the optimal method for HO· was determined, the same method was applied to the calculations for SO₄·⁻. This level of calculation has shown an excellent cost-benefit for the description of the mechanism and kinetics in oxidative degradation of organic compounds.

The characterization the majority of the reactive sites of FLX with HO· and SO₄·⁻ was evaluated by the Fukui function ($f$) [19,20] described by the equations $f^{+}=\sum _i|{c_i|}_{\mathrm{H}\mathrm{O}\mathrm{M}\mathrm{O}}^2$, $f^{-}=\sum _i|{c_i|}_{\mathrm{L}\mathrm{U}\mathrm{M}\mathrm{O}}^2$ (HOMO: highest occupied molecular orbital; LUMO: lowest unoccupied molecular, ci: frontier molecular orbital coefficients), and $f^0\approx (f^{+}+f^{-})$, demonstrates the main atoms susceptible to nucleophilic, electrophilic and radical attacks, respectively. The contributions of the atomic orbitals of the frontier molecular orbital are weighted by the $c_i$ coefficients of the HOMO and LUMO orbital. $f^0$ values were used to study HO· and SO₄·⁻ attacks to the antidepressant. The Fukui function was evaluated by the Multiwfn package program [21].

The electronic structure properties of the reactants, products, and transition states were calculated at the M06HF/6-31 + G(d) calculation level using the solvation model density (SMD) [22]. The SMD model has been widely used to simulate the aqueous environment to describe the mechanisms of antidepressant degradation. The stationary points were determined by analytic harmonic frequency calculations of the optimized structures. The absence or presence of one imaginary frequency characterizes the optimized structures as local minima or transition states, respectively. The zero-point vibrational energy contributions were considered in the calculation of the energy barrier. Quantum mechanical electronic structure calculations were performed using the Gaussian 16 package [23].

Furthermore, the contribution of the single-electron transfer (SET) mechanism [24,25] from HO· and SO₄·⁻ compounds were assessed by Marcus theory [26] through the calculation of the activation free-energy barrier expressed as:

(6)${\mathrm{\Delta }G}_{\mathrm{S}\mathrm{E}\mathrm{T}}^{\text{‡}}=\frac{{\left(\lambda +{\mathrm{\Delta }G}_{\mathrm{S}\mathrm{E}\mathrm{T}}\right)}^2}{4\lambda },$

where ∆G_SET is the free-energy of the reaction and λ is the nuclear reorganization energy.

2.4.2 Toxicity and risk assessment

Ecological Structure Activity Relationship (ECOSAR V2.0), a free downloadable software developed by the US Environmental Protection Agency, was used to carry out ecotoxicity predictions for green algae, daphnia and fish as target organisms. The program uses QSAR calculations on the basis of its own data set, mainly data from the literature; the toxicity estimates provided by ECOSAR have been validated by different studies [27,28]. In our study, the toxicity profile was evaluated using acute toxicity (lethal concentration of 50% (LC₅₀) or effective concentration of 50% (EC₅₀)) and chronic toxicity (ChV) values. Acute toxicity is defined as the chemical concentration (mg·L⁻¹) that causes 50% mortality in the fish population after 96 h of exposure and in daphnia after 48 h of exposure (LC₅₀). It also refers to the chemical concentration (mg·L⁻¹) that inhibits 50% of the growth of green algae after 96 h of exposure (EC₅₀). Subsequently, the potential hazards of FLX and its main metabolites for aquatic environments were assessed using the technical criteria from [29]. This approach is based on the risk quotient (RQ) method, which involves dividing the measured environmental concentration (MEC) of the contaminant by its predicted no-effect concentration (PNEC). Contaminants with RQ < 0.01 hardly represent a risk to the environment, while those with 0.01 < RQ < 0.1 represent low risk. Moderate risk is associated with 0.1 < RQ < 1, while RQ > 1 indicates high risk. The details in calculation methods used were based on the procedures described in Ref. [8].

3 Results and discussion

3.1 3D printing and reactor costs

The 3D printing results showed that greater complexity in the internal geometry of the reactor led to an increase in deviation from the design geometry, as shown in Tab.1. This deviation is expected, as the spiral reactor geometry requires the fabrication of channels with a very narrow width internal diameter (1.0 mm) and is more sensitive to printing inaccuracy limitations (0.12 mm thickness per layer) [30].

Tab.1 Deviations in the volumes of the designed reactors in relation to the real prototypes

Reactor	Designed volume/mL	Real volume/mL	Deviation
Spiral	11.6	8.0	−31.0%
Flat	24.7	25.0	1.10%

In this sense, a preliminary analysis was carried out to verify the impact of changing the internal geometry on the final cost of the reactor. Tab.2 presents the costs of each part of the reactors according to the main parameters for their calculation (i.e., printing time and material consumption), as well as additional parts (i.e., a borosilicate glass window). As the complexity of the geometry increases, the cost of the prototypes also increases. Nevertheless, the total cost values of the flat (US$10.72)andspiral(US$ 12.26) reactors are not significantly different.

Tab.2 Reactor costs as a function of prototyping efficiency parameters and additional parts

Reactor	Printing time/h	Material consumption/g	Printing costa)/US$	Glass cost/US$	Total cost/US$
Spiral	17.72	113.86	3.26	9.00	12.26
Flat	7.11	64.53	1.72	9.00	10.72

a) The printing cost was calculated as a linear combination of the costs of material and electricity, considering the average commercial tariff in August 2022 in Sao Paulo, Brazil.

3.2 Hydrodynamic evaluation

Regarding the results of RTD, the parameters of each system were obtained to characterize the distributions in the different reactor modules and examined conditions, as shown in Tab.3. A clear deviation from ideality was observed in the flat reactor when compared to the spiral reactor. This is evidenced by the fact that the t_m values were smaller than the τ in both studied conditions studied (Tab.3). Furthermore, the values of σ² and distribution distortion in relation to the mean (s³) were more significant in the flat reactor (Tab.3). These results can be attributed to the greater distance of the flat reactor from an ideal reactor since this design generates an internal geometry characterized by the strong presence of preferential paths or short circuits (Fig. S1, cf. ESM). Therefore, due to these characteristics, the flat reactor would not be a suitable choice for a photochemical setup in practical applications, since this particular flow behavior can promote combinations of chemical species present in distinct stages of the degradation process, generating by-products with the same or greater potential for toxicity compared to target compounds [1].

Tab.3 Calculated RTD parameters for the photochemical reactors

Reactor	Flow/(mL·min−1)	tm/min	σ2/min2	s3/min3	τ/min	(tm – τ)/min
Spiral	10	1.26	0.79	1.77	0.8	0.46
	5	2.64	3.10	5.71	1.6	1.04
Flat	10	1.95	2.03	2.42	2.5	−0.55
	5	3.39	3.12	17.1	5.0	–1.61

As a result, the spiral module (Fig.1(b)) exhibited more promising results in terms of flow pattern and photochemical application and, consequently, was selected for photodegradation experiments. The optimal flow rate was determined based on an analysis of the RTD curves (Fig. S1, cf. ESM), taking into account the distribution with the lowest number of subcurves that most closely approximated a single peak (Fig. S1(b), cf. ESM). It was concluded that at this flow rate (υ₀ = 5 mL·min^–1), the smallest amount of backmixing would occur in the reaction volume [10].

Fig.2 allows observing that the combination of CSTR and PFR best fits the experimental RTD data for both reactor modules, exhibiting the smallest deviations. In the case of the spiral reactor, this combination is attributed to an inlet that induces minimal mixing (CSTR), followed by a plug flow influenced by spiral geometry (PFR) (Fig. S1(b), cf. ESM). In contrast, the preferential path formed in the flat reactor causes the fluid to rapidly exit the reactor in a piston-like behavior (PFR), with the remaining fluid undergoing mixing (CSTR). Comparable results were obtained for a larger reactor with inlet and outlet positioned tangentially to the reactor body [12].

Fig.2 Comparison of macroscopic strategies to describe the flow in the circular photochemical reactors by predicted residence time (spiral (υ0 = 10 mL·min−1, blue columns); spiral (υ0 = 5 mL·min–1, orange columns); flat (υ0 = 10 mL·min−1, gray columns); flat (υ0 = 5 mL·min−1, yellow columns); Exp: experimental; LFR: laminar-flow reactor; T-I-S: tanks-in-series; CSTR+PFR: continuous stirred-tank reactor and plug flow reactor; AD: axial dispersion model).

Full size|PPT slide

3.3 Effect of the aqueous matrix on the degradation of FLX by the solar/PS process

Compared to Ultraviolet (UV) light and electron beams, sunlight is considered a cost-effective and environmentally-friendly activation alternative, as well as being efficient in the degradation of organic contaminants [31]. However, the application of these methods is restricted by several consequential issues, including the possible generation of hazardous disinfection by-products [32] and the unstable nature of the oxidants (i.e., H₂O₂). A promising way to apply sunlight-driven AOPs is the use of PS as a source of SO₄·⁻ [33]. PS has several advantages, such as the stability of its salts (favoring storage and transport); less possibility of generating more toxic by-products; higher redox potential (E⁰ = 2.5–3.1 V vs. standard hydrogen electrode); and higher molar absorption coefficient compared to H₂O₂ in the UVA region (i.e., ${\epsilon }_{{\mathrm{H}}_2{\mathrm{O}}_2}$ = 0.01 L·mol^–1·cm^–1 (360 nm) [34]; ε_PS = 0.35 L·mol^–1·cm^–1 (300–400 nm) with photolysis quantum yield ϕ_PS = 0.34 mol·E^–1 [35−37]). In addition, SO₄·⁻ is more selective, has a lower scavenging effect, and is suitable for various pH values [36]. Although PS can be activated by different methods, including transition metals, ultrasound, heating, photoirradiation, and alkaline activation [36], photoactivation is more economical and does not require additional chemicals. This activation can result in the formation of SO₄·⁻ and HO·, as described by the equations below [36]:

(7)${\mathrm{S}}_2{\mathrm{O}}_8^{2-}+hv\to 2{\mathrm{S}\mathrm{O}}_4{\cdot }^{-}$

(8)${\mathrm{S}\mathrm{O}}_4{\cdot }^{-}+\mathrm{O}{\mathrm{H}}^{-}\to \mathrm{H}\mathrm{O}\cdot +\mathrm{S}{\mathrm{O}}_4^{2-}$

(9)${\mathrm{S}\mathrm{O}}_4{\cdot }^{-}+{\mathrm{H}}_2\mathrm{O}\to \mathrm{H}\mathrm{O}\cdot +\mathrm{S}{\mathrm{O}}_4^{2-}+{\mathrm{H}}^{+}$

Fig.3 illustrates the photochemical degradation FLX as a result of the activation of PS under simulated solar radiation in different aqueous matrices. Adsorption control tests were carried out on the material from which the reactors were manufactured (acrylonitrile butadiene styrene) and it was found that this material did not significantly adsorb the target contaminant. The findings show a low rate of FLX photolysis in 60 min of exposure (FLX degradation rate constant in pure water: k_{obs-pure water} = 4.1 × 10⁻³ min^–1; FLX degradation in pure water: 17.7%) in pure water, attributable to the low direct photolysis quantum yield, as previously discussed in Ref. [13]. Conversely, irradiation in a real wastewater matrix resulted in an increase in the rate of photolysis and FLX degradation (FLX degradation rate constant in real wastewater: k_{obs-real water} = 8.2 × 10⁻³ min^–1; FLX degradation in real wastewater: 32.4%). This suggests the occurrence of additional indirect photochemical processes due to species present in the real aqueous matrix (i.e., nitrates/nitrites) (Table S1, cf. ESM), which can generate radical species (i.e., HO·) and facilitate the degradation process [8], as also observed in Ref. [13]. Control experiments performed in the dark and with PS in pure water only exhibited low rates of FLX degradation (k_{obs-pure water} = 5.5 × 10⁻³ min^–1; 23.3%), a pattern comparable to that observed during photolysis in pure water. In contrast, the control test conducted with PS in real wastewater exhibited a slight improvement in degradation efficiency (k_{obs-real water} = 6.0 × 10⁻³ min^–1; 28.6%), which could be attributed to the slight increase in pH from pure water (pH = ~ 6.2) to real wastewater (pH = ~ 6.8), allowing for slight PS activation [36].

Fig.3 FLX degradation in pure and real wastewater by the solar/PS process ([FLX pure water]0 = 10.36 ± 0.29 mg·L–1; [FLX real water]0 = 10.18 ± 0.09 mg·L–1; natural pH: pure (~6.2) and real (~6.8) water matrices).

Full size|PPT slide

As shown in Fig.3, the degradation of FLX by the solar/PS process was slightly slower in the real wastewater matrix compared to pure water (k_{obs-pure water} = 2.83 × 10⁻² min^–1; FLX degradation in pure water (60 min) = 76.8%; k_{obs-real water} = 2.43 × 10⁻² min^–1; FLX degradation in real wastewater (60 min) = 69.9%). This is attributed to the presence of organic and inorganic components in real wastewater (Table S1, cf. ESM), which can act as scavengers of photogenerated sulfate radicals, leading to a reduction in the extent of degradation reactions [13]. Nevertheless, differences in k_obs values were relatively small and can therefore be considered practically negligible. Similar findings were reported for the degradation of the antidepressant sertraline when comparing the effectiveness of the UV/PS process under UVC radiation in a real wastewater system [16]. Comparable results were also observed in a study involving the antibiotic tinidazole [38] whose degradation increased with the addition of K₂S₂O₈ under sunlight compared to the direct photolytic process.

3.4 Theoretical prediction of the intermediates generated from FLX oxidation by HO· and SO₄·⁻

Fukui index values for reactions involving the two main radicals generated in the solar/PS process (SO₄·⁻ and HO·) are shown in Fig. S2 (cf. ESM), along with the chemical structure of FLX (Fig. S3, cf. ESM). In this study, the thermodynamic and kinetic properties for initiated SO₄·⁻ and HO· radicals reaction are investigated to find the effective degradation pathways at the molecular level. The calculations were performed for the addition and H-abstraction reactions highlighted in blue in Fig. S3 (cf. ESM), as the atoms involved have the highest Fukui indices, suggesting that they are the most reactive. As depicted in Fig.4, HO· attack on C₁₀ (TP₁) is expected to be the most kinetically favored reaction mechanism, leading to the substitution of the trifluoromethyl group. Although the trifluoromethyl substituent of an aromatic compound is a persistent site in the molecule [39], the electron density resulting from the hyperconjugation effect of the CF₃ group at C₁₀ makes the HO· attack kinetically favorable [40]. Furthermore, the addition of the HO· radical to carbon atoms is more kinetically favorable than reactions involving the abstraction of hydrogen atoms.

Fig.4 Proposed initiated mechanisms for hydrogen abstraction and addition reactions between FLX and HO· and SO4·− in an implicit aqueous medium (The reaction energy barriers (E0) and free energies (ΔG), in kcal·mol–1, were calculated at the M06HF/6-31 + G(d) level; RAF: radical adduct formation; HT: hydrogen transfer).

Full size|PPT slide

Fig.4 shows the initiated degradation pathways where the SO₄·⁻ addition and H-abstraction reactions are most likely to occur. As the reaction with HO·, the attack of SO₄·⁻ on C₁₀ (TP_1A) is also the most kinetically favored reaction mechanism, presenting the lowest energy barrier. In turn, addition at C₁₂ leads to a thermodynamically more favorable product and the second lowest energy barrier. The apparent negative energy barrier is observed by SO₄·⁻ attack which is clearly connected with the formation of van der Waals complexes (Fig. S4, cf. ESM) promoted by charge alignment [41,42]. Recent works on oxidation-based systems explained this free energy profile in the degradation of micropollutants, which it is a fingerprint of anti-Arrhenius kinetics [8,43]. Based on these findings, the addition of SO₄·⁻ can be inferred to be a more favorable reaction compared to H-abstraction in terms of both kinetics and thermodynamics [44]. In this work, the addition of SO₄·⁻ to C₁₂ (TP_2A) was identified as the main pathway of FLX degradation. Furthermore, we found that the SO₄·⁻ reaction is more likely to occur than the HO· attack, due to the low energy barrier.

For the degradation of FLX, besides radical adduct formation (RAF) and hydrogen transfer (HT) pathways, the SET mechanism also plays a role due to the high reactivity potential of SO₄·⁻ and HO·. These reactions result in the formation of the FLX radical cation and sulfate and hydroxyl anions. The reaction barrier for the SET-SO₄·⁻ step, represented by ΔG^‡ and calculated using Marcus theory, is determined to be 1.43 kJ·mol ⁻¹, which is significantly lower compared to HO· under the same conditions (4.71 kJ·mol⁻¹). Additionally, the process via SO₄·⁻ is spontaneous, with a ΔG of –143.48 kJ·mol⁻¹ and an exothermic nature (ΔH of –143.01 kJ·mol⁻¹). In contrast, HO· has low spontaneity and an endothermic character (ΔG of –0.05 kJ·mol⁻¹ and ΔH of 0.70 kJ·mol⁻¹, respectively). The results for the SET mechanism are favorable, and SO₄·⁻ remains more favorable due to its higher reactivity. This is because SO₄·⁻ radical has a higher capacity to accept electrons, facilitating the degradation of FLX. Furthermore, the presence of the CF₃ group in FLX increases the molecule’s tendency to undergo electron transfer, which further enhances the effectiveness of the SET mechanism with SO₄·⁻ compared to HO· [25].

3.5 Toxicity and risk analysis

After studying the main degradation pathways, we estimated the toxicities of FLX and its main TP formed by HO· (TP₁) and SO₄·⁻ (TP_2A) attacks, in relation to fish, daphnia, and green algae (Tab.4). FLX has been shown to be toxic to plants, resulting in a significant decrease in root growth, daily growth rate, and asexual reproduction, even at environmental concentrations [45]. However, a complete lifecycle analysis of fathead minnows revealed virtually no effect [46]. Nevertheless, the potential for tissue absorption in fish exposed to this antidepressant is evident, as the highest values of the bioconcentration factor in three different fish organs were found in the group exposed to the lowest FLX concentration (0.1 μg·L⁻¹) [47]. This supports the predicted log K_{OW FLX} = 4.64 value from ECOSAR, which indicates potential bioaccumulation (log K_OW > 4.0) [48].

Tab.4 Toxicity and risk assessment of FLX and its main stable metabolites generated via HO· (TP1) and SO4·− (TP2A) radical attacks

Parameter/organisms	Compounds
	FLX/(mg·L–1)	TP1/(mg·L–1)	TP2A/(mg·L–1)
LC50 (fish 96 h)	1.08	1.39	656
LC50 (daphnia 48 h)	0.175	0.393	66.4
EC50 (green algae 96 h)	0.079	0.349	75.7
ChV (fish, chronic)	0.025	0.021	62.8
ChV (daphnia, chronic)	0.019	0.072	4.65
ChV (green algae, chronic)	0.033	0.072	22.3
RQ	1.27	0.37	0.002
Log KOW	4.64	3.21	0.59

FLX degradation via the TP₁ channel produces a metabolite that may potentially be more toxic, particularly in terms of acute toxicity to fish and algae (Tab.4). In this sense, the predicted value of log K_{OW TP1} = 3.21 suggests a potential for bioconcentration, although not high enough for bioaccumulation (log K_OW > 4.0). Conversely, degradation via the TP_2A channel results in the formation of a metabolite with low toxicity, both acute and chronic. This result is supported by the much lower value of log K_OW = 0.59, indicating a low potential for bioconcentration and bioaccumulation.

Tab.4 also presents the calculated RQ for FLX, TP₁, and TP_2A. MEC data collection was performed using a non-site-specific approach, utilizing the highest concentration of FLX in surface water reported in the literature [49] and a 1.3-fold ratio for metabolites concentrations [50]. PNEC values for EC₅₀ (green algae 96 h), EC₅₀ (green algae 96 h), and LC₅₀ (daphnia 48 h) were used for FLX and its metabolites (TP₁ and TP_2A), respectively, with assessment factor of 1000 in both cases. The results are consistent with toxicity predictions, revealing a high risk for FLX (RQ > 1), moderate risk for TP₁ (0.1 < RQ < 1), and no risk (RQ < 0.01) for TP_2A.

4 Conclusions

The use of 3D printing prototyping for laboratory devices allowed flexibility and cost-effectiveness in the production of alternative designs for a photochemical reactor with circular window illumination, especially by using a recycled polymer material in its manufacture. The RTD approach proved to be an efficient methodology for elucidating the fluid flow pattern in the reactors, which was better understood using the macromixture models, allowing reactor characterization based on the combination of ideal CSTR + PFR reactors. The spiral reactor was found to produce superior results for photochemical applications; therefore it was chosen for the photodegradation experiments with the model contaminant FLX, whose degradation by direct photolysis was more effective in a real water matrix (sewage treatment plant effluent). The addition of PS to the real water matrix combined with solar irradiation proved to be an effective combination for FLX degradation, with almost 70% removal in 1 h of treatment and a pseudo-first-order specific degradation rate of 2.43 × 10 ⁻² min⁻¹. This performance is only slightly lower compared to that of pure water, due to the scavenging of photogenerated sulfate radicals by organic and inorganic species present in real water. Nonetheless, this does not prevent the process from being applied to real matrices, as the degradation values obtained for the different matrices were not significantly discrepant. The predicted toxicity of the by-products generated with different radical attack pathways indicated that SO₄·⁻ has the potential to produce by-products with a lower ecotoxicological impact. In summary, this study presents a promising alternative for a screening investigation into the degradation of EC using solar radiation and offers flexibility in the configuration of the operating system, since the reactors can be arranged in modules and built with recycled material.

References

[1] Oppenländer T . Photochemical purification of water and air: Advanced oxidation processes (AOPs): principles, reaction mechanisms, reactor concepts. Chemistry International-Newsmagazine for IUPAC, 2003, 25(4): 31

[2] Wols B A , Harmsen D J H , Wanders-Dijk J , Beerendonk E F , Hofman-Caris C H M . Degradation of pharmaceuticals in UV (LP)/H₂O₂ reactors simulated by means of kinetic modeling and computational fluid dynamics (CFD). Water Research, 2015, 75: 11–24

[3] Psaltou S , Kaprara E , Tsaragklis A , Mitrakas M , Zouboulis A . Investigation of the removal of several micropollutants presenting different ozone reactivities from natural potable water matrix by the application of ozonation with the use of SiO₂ and Al₂O₃ as catalysts. Separations, 2022, 9(7): 173

[4] Sundar K P , Kanmani S . Progression of photocatalytic reactors and it’s comparison: a review. Chemical Engineering Research & Design, 2020, 154: 135–150

[5] Ramos B , Carneiro J G M , Nagamati L I , Teixeira A C S C . Development of intensified flat-plate packed-bed solar reactors for heterogeneous photocatalysis. Environmental Science and Pollution Research International, 2021, 28(19): 24023–24033

[6] Sharma A , Ahmad J , Flora S J S . Application of advanced oxidation processes and toxicity assessment of transformation products. Environmental Research, 2018, 167: 223–233

[7] Kowalska E , Rau S . Photoreactors for wastewater treatment: a review. Recent Patents on Engineering, 2010, 4(3): 242–266

[8] Souza L P , Sanches-Neto F O , Junior G M Y , Ramos B , Lastre-Acosta A M , Carvalho-Silva V H , Teixeira A C S C . Photochemical environmental persistence of venlafaxine in an urban water reservoir: a combined experimental and computational investigation. Process Safety and Environmental Protection, 2022, 166: 478–490

[9] Ye T , Wei Z , Spinney R , Tang C J , Luo S , Xiao R , Dionysiou D D . Chemical structure-based predictive model for the oxidation of trace organic contaminants by sulfate radical. Water Research, 2017, 116: 106–115

[10] FoglerH S. Elements of Chemical Reaction Engineering. 5th ed. New Jersey: Prentice Hall, 2016

[11] Peres J C G , Tambani P C , Teixeira A C S C , Guardani R , Vianna A S Jr . Study of an annular photoreactor with tangential inlet and outlet. II. The UV/H₂O₂ reactive flow. Chemical Engineering & Technology, 2019, 42(2): 316–326

[12] Peres J C G , de Silvio U , Teixeira A C S C , Guardani R , Vianna A S . Study of an annular photoreactor with tangential inlet and outlet: I. Fluid dynamics. Chemical Engineering & Technology, 2015, 38(2): 311–318

[13] Souza L P , Carneiro J G M , Lastre-Acosta A M , Ramos B , Teixeira A C S C . Environmental persistence of the antidepressant fluoxetine and its pharmaceutical alternative: kinetics of oxidation and mathematical simulations. Water (Basel), 2022, 14(21): 3536

[14] Singh A , Saidulu D , Gupta A K , Kubsad V . Occurrence and fate of antidepressants in the aquatic environment: insights into toxicological effects on the aquatic life, analytical methods, and removal techniques. Journal of Environmental Chemical Engineering, 2022, 10(6): 109012

[15] Lester Y , Mamane H , Zucker I , Avisar D . Treating wastewater from a pharmaceutical formulation facility by biological process and ozone. Water Research, 2013, 47(13): 4349–4356

[16] Drossou C , Petrakis Y , Tyrovola K , Xekoukoulotakis N P . Photochemical degradation of the antidepressant sertraline in aqueous solutions by UVC, UVC/H₂O₂, and UVC/S₂O₈^2–. Water Research, 2022, 217: 118442

[17] Sanches-Neto F O , Coutinho N D , Palazzetti F , Carvalho-Silva V H . Temperature dependence of rate constants for the H(D) + CH₄ reaction in gas and aqueous phase: deformed transition-state theory study including quantum tunneling and diffusion effects. Structural Chemistry, 2020, 31(2): 609–617

[18] Sanches-Neto F , Coutinho N , Aquilanti V , Silva W , Carvalho-Silva V . Mechanism and kinetics of the degradation of nitazoxanide and hydroxychloroquine drugs by hydroxyl radicals: theoretical approach to ecotoxicity. Journal of the Brazilian Chemical Society, 2023, 34(8): 1119–1129

[19] López P , Méndez F . Fukui function as a descriptor of the imidazolium protonated cation resonance hybrid structure. Organic Letters, 2004, 6(11): 1781–1783

[20] MelinJAyersP WOrtizJ V. Removing electrons can increase the electron density: a computational study of negative Fukui functions. The Journal of Physical Chemistry A, 2007, A 111: 10017–10019

[21] Lu T , Chen F . Multiwfn: a multifunctional wavefunction analyzer. Journal of Computational Chemistry, 2012, 33(5): 580–592

[22] Marenich A V , Cramer C J , Truhlar D G . Universal solvation model based on solute electron density and on a continuum model of the solvent defined by the bulk dielectric constant and atomic surface tensions. Journal of Physical Chemistry B, 2009, 113(18): 6378–6396

[23] FrischM JTrucksG WSchlegelH BScuseriaG ERobbM ACheesemanJ RScalmaniGBaroneVPeterssonG ANakatsujiH, . Gaussian 16, Revision C.01. Gaussian, Inc., Wallingford CT, 2016

[24] Iuga C , Campero A , Vivier-Bunge A . Antioxidant vs prooxidant action of phenothiazine in a biological environment in the presence of hydroxyl and hydroperoxyl radicals: a quantum chemistry study. RSC Advances, 2015, 5(19): 14678–14689

[25] Zhang H , Chen L , Du P , Li F , Liu W . Unraveling different reaction characteristics of alkoxy radicals in a Co(II)-activated peracetic acid system based on dynamic analysis of electron distribution. Environmental Science & Technology, 2024, 8(7): 3506–3519

[26] Marcus R A . Electron transfer reactions in chemistry. Theory and experiment. Reviews of Modern Physics, 1993, 65(3): 599–610

[27] Reuschenbach P , Silvani M , Dammann M , Warnecke D , Knacker T . ECOSAR model performance with a large test set of industrial chemicals. Chemosphere, 2008, 71(10): 1986–1995

[28] Zhou L , Fan D , Yin W , Gu W , Wang Z , Liu J , Xu Y , Shi L , Liu M , Ji G . Comparison of seven in silico tools for evaluating of daphnia and fish acute toxicity: case study on Chinese Priority Controlled Chemicals and new chemicals. BMC Bioinformatics, 2021, 22(1): 151

[29] Leeuwen K. Technical Guidance Document on Risk Assessment, 2003

[30] Msallem B , Sharma N , Cao S , Halbeisen F S , Zeilhofer H F , Thieringer F M . Evaluation of the dimensional accuracy of 3D-printed anatomical mandibular models using FFF, SLA, SLS, MJ, and BJ printing technology. Journal of Clinical Medicine, 2020, 9(3): 817

[31] Ferro G , Polo-López M I , Martínez-Piernas A B , Fernández-Ibáñez P , Agüera A , Rizzo L . Cross-contamination of residual emerging contaminants and antibiotic resistant bacteria in lettuce crops and soil irrigated with wastewater treated by sunlight/H₂O₂. Environmental Science & Technology, 2015, 49(18): 11096–11104

[32] Sun P , Lee W N , Zhang R , Huang C H . Degradation of DEET and caffeine under UV/chlorine and simulated sunlight/chlorine conditions. Environmental Science & Technology, 2016, 50(24): 13265–13273

[33] Tan C , Jian X , Wu H , Sheng T , Sun K , Gao H . Kinetics degradation of phenacetin by solar activated persulfate system. Separation and Purification Technology, 2021, 256: 117851

[34] Cataldo F . Hydrogen peroxide photolysis with different UV light sources including a new UV-LED light source. New Frontiers in Chemistry, 2014, 23: 99–110

[35] Herrmann H . On the photolysis of simple anions and neutral molecules as sources of O^–/OH, SO_x^– and Cl in aqueous solution. Physical Chemistry Chemical Physics, 2007, 9(30): 3935–3964

[36] Wang B , Wang Y . A comprehensive review on persulfate activation treatment of wastewater. Science of the Total Environment, 2022, 831: 154906

[37] Nunes R F , Teixeira A C S C . An overview on surfactants as pollutants of concern: occurrence, impacts and persulfate-based remediation technologies. Chemosphere, 2022, 300: 134507

[38] Velo-Gala I , Pirán-Montaño J A , Rivera-Utrilla J , Sánchez-Polo M , Mota A J . Advanced oxidation processes based on the use of UVC and simulated solar radiation to remove the antibiotic tinidazole from water. Chemical Engineering Journal, 2017, 323: 605–617

[39] Yale H L . The trifluoromethyl group in medical chemistry. Journal of Medicinal and Pharmaceutical Chemistry, 1959, 1(2): 121–133

[40] Szabó L , Mile V , Kiss D J , Kovács K , Földes T , Németh T , Tóth T , Homlok R , Balogh G T , Takács E , Wojnárovits L . Applicability evaluation of advanced processes for elimination of neurophysiological activity of antidepressant fluoxetine. Chemosphere, 2018, 193: 489–497

[41] Coutinho N D , Silva V H C , de Oliveira H C B , Camargo A J , Mundim K C , Aquilanti V . Stereodynamical origin of anti-Arrhenius kinetics: negative activation energy and roaming for a four-atom reaction. Journal of Physical Chemistry Letters, 2015, 6(9): 1553–1558

[42] Coutinho N D , Sanches-Neto F O , Carvalho-Silva V H , de Oliveira H C B , Ribeiro L A , Aquilanti V . Kinetics of the OH + HCl→H₂O + Cl reaction: rate determining roles of stereodynamics and roaming and of quantum tunneling. Journal of Computational Chemistry, 2018, 39(30): 2508–2516

[43] Ramos B , Vaz W F , Diniz L F , Sanches-Neto F O , Ribeiro J C O , Carvalho-Silva V H , Teixeira A C S C , Ribeiro C , Napolitano H B , Carvalho P S Jr . Kinetics, mechanism, and tautomerism in ametryn acid hydrolysis: from molecular structure to environmental impacts. Chemosphere, 2023, 324: 138278

[44] Yang J , Lv G , Wang Z , Sun X , Gao J . Mechanisms, kinetics and eco-toxicity assessment of singlet oxygen, sulfate and hydroxyl radicals-initiated degradation of fenpiclonil in aquatic environments. Journal of Hazardous Materials, 2021, 409: 124505

[45] Amy-Sagers C , Reinhardt K , Larson D M . Ecotoxicological assessments show sucralose and fluoxetine affect the aquatic plant, Lemna minor. Aquatic Toxicology, 2017, 185: 76–85

[46] Parrott J L , Metcalfe C D . Assessing the effects of environmentally relevant concentrations of antidepressant mixtures to fathead minnows exposed over a full life cycle. Science of the Total Environment, 2019, 648: 1227–1236

[47] Pan C , Yang M , Xu H , Xu B , Jiang L , Wu M . Tissue bioconcentration and effects of fluoxetine in zebrafish (Danio rerio) and red crucian cap (Carassius auratus) after short-term and long-term exposure. Chemosphere, 2018, 205: 8–14

[48] Arnot J A , Gobas F A P C . A generic QSAR for assessing the bioaccumulation potential of organic chemicals in aquatic food webs. QSAR & Combinatorial Science, 2003, 22(3): 337–345

[49] Castillo-Zacarías C , Barocio M E , Hidalgo-Vázquez E , Sosa-Hernández J E , Parra-Arroyo L , López-Pacheco I Y , Barceló D , Iqbal H N M , Parra-Saldívar R . Antidepressant drugs as emerging contaminants: occurrence in urban and non-urban waters and analytical methods for their detection. Science of the Total Environment, 2021, 757: 143722

[50] Figuière R , Waara S , Ahrens L , Golovko O . Risk-based screening for prioritisation of organic micropollutants in Swedish freshwater. Journal of Hazardous Materials, 2022, 429: 128302

Acknowledgements

This study was financed in part by the Coordination for the Improvement of Higher Education Personnel (CAPES)—Finance Code 001. The authors thank the supports of the São Paulo Research Foundation (FAPESP) (Grant Nos. 2018/21271-6, 2019/24158-9 and 2022/13583-3), the Goiás Research Foundation (FAPEG) (Grant No. GSP2019011000037) and the National Council for Scientific and Technological Development—Brazil (CNPq) (Grant Nos. 311230/2020-2 and 309154/2023-5).

Electronic Supplementary Material

Supplementary material is available in the online version of this article at https://doi.org/10.1007/s11705-024-2517-y and is accessible for authorized users.

Competing interests

The authors declare that they have no competing interests.

RIGHTS & PERMISSIONS

2024 Higher Education Press