Removal of Polycyclic Aromatic Hydrocarbons in a Heterogeneous Fenton Like Oxidation System Using Nanoscale Zero-Valent Iron as a Catalyst

Abstract

Oil and gas effluents contains highly toxic and harmful organic pollutants. Therefore, it is necessary to eliminate and/or reduced the concertation of organic pollutants to a technologically acceptable levels before their discharge into water streams. This study investigates the application of nanoscale zero-valent iron (nZVI), and hydrogen peroxide (H₂O₂) for removal of organic pollutants from real oily produced water. Batch studies were performed and effect of different operating parameters, including concentration of nZVI and H₂O₂, pH and reaction time were studied. Moreover, optimization of independent variables was performed using central composite design (CCD) in response surface methodology (RSM). The experimental set up provided maximum removal efficiencies of 89.5% and 75.3% for polycyclic aromatic hydrocarbons (PAHs) and chemical oxygen demand (COD), respectively. The optimum values of independent variables such as concentrations of nZVI, and H₂O₂, contact time and pH were obtained as 4.35 g/L, 1.60 g/L, 199.9 min and 2.9, respectively. Predicted PAHs and COD removal efficiencies at the optimum values of independent variables were found as 89.3% and 75.7%, respectively which are in line with the experimental values. The study indicates that application of heterogeneous Fenton like oxidation system using nZVI as a catalyst is an efficient treatment method for removal of organic pollutants from real produced water.

1. Introduction

Oil and gas industries are one of the important sectors producing hazardous byproducts in the form of oily produced water [1]. Produced water is usually present in deep underground rocks and is brought to the ground surface during oil and gas extraction. Composition of produced water depends on the geological conditions of the site, type of hydrocarbons involved in extraction processes, characteristics of hydrocarbons well and extraction methods [2]. Produced water typically contains copious hazardous organic, inorganic pollutants, oxides, metals and minerals [3]. Globally, approximately 250 million barrel/per day produced water is generated from oil extraction processes [1,4], whereas in Malaysia, its production is about 1800 thousand barrel/per day [5].

Produced water contains high concentration of hazardous polycyclic aromatic hydrocarbons (PAHs) which are classified as low molecular weight (LMW) and high molecular weight (HMW) compounds based on number of benzene rings. For instance, naphthalene and acenaphthylene are LMW PAHs with two or three benzene rings, where HMW PAHs include pyrene and chrysene consisting of four or more fused benzene rings [4]. Acute and chronic toxicity of produced water is mainly assigned to presence of these PAHs, phenols and higher chemical oxygen demand (COD) concentrations [6]. Various physio-chemical and biological methods, such as electrocoagulation, membrane-based technologies, electrodialysis, ion exchange, adsorption and activated sludge processes have been employed for the treatment of produced water [5,7]. Nevertheless, these conventional treatment methods have many drawbacks, such as low removal efficiencies, high maintenance cost, more energy consumption, membrane fouling, and production of toxic byproducts [5].

Advanced oxidation processes (AOPs) has gained considerable attention recently as an alternative to conventional chemical methods to remove the refractory organic pollutants from wastewater [8]. Several AOPs have been employed for treatment of petroleum industry effluents. Among AOPs, homogeneous Fenton oxidation is considered one of the best options for remediation of these pollutants [9], as it can degrade an extensive range of organic micropollutants by producing hydroxyl radicals (OH^•) and can transform non-biodegradable organic pollutants into biodegradable substances [10]. However, production of a high amount of iron, toxic sludge and usage of significant amount of reagents limit the use of homogeneous Fenton oxidation [11].

Nowadays, heterogeneous Fenton like oxidation systems are being investigated by utilizing iron supported catalysts such as magnetite (Fe₃O₄), zero-valent iron (ZVI), lepidocrocite (γ-FeOOH) and goethite (α-FeOOH). However, except nanomaterials such as nanoscale ZVI (nZVI), many catalysts require ultraviolet (UV) radiation, visible irradiation, or ultrasound as additional assistance for remediation of pollutants [12]. Whereas nZVI has shown potential to address the issues discussed above because of its novel properties, cost-effectiveness, abundance in nature, and environmentally friendly nature [13]. The reduction potential of nZVI is E = −0.44 V, which indicates it can effectively degrade wide range of refractory pollutants [14]. The nZVI rate of reaction is higher (up to 100 times) than other catalysts, such as conventional iron powders, γ-FeOOH and Fe₃O₄ [15,16]. Zhang et al. [17] examined application of nZVI particles (size 1–100 nm) for degradation of organic pollutants and showed excellent reactivity of nZVI because of its large specific surface area (33.5 m²/g). nZVI can be used as a catalyst in Fenton like oxidation system for the degradation of refractory organic pollutants. Heterogeneous Fenton like oxidation system using nZVI as a catalyst has been suggested for efficient degradation of organic contaminants in numerous studies [18]. In heterogeneous Fenton like oxidation system, nZVI generates H₂O₂ by losing two-electron during reaction with O₂, as shown in Equation (1). The OH^• radicals are also produced during the reaction of nZVI and H₂O₂, as presented in Equations (2) and (3). The OH^• degrades organic pollutants via radical combination, electron transfer, and hydrogen abstraction, as shown in Equations (8)–(10), respectively [19]. While already produced Fe³⁺ ions may be reduced reacting with H₂O₂ and regenerate Fe²⁺ ion, as shown in Equation (4). This mechanism known as Fenton like reaction, it allows regeneration of Fe²⁺ in an excellent cyclic process. Moreover, Equations (4)–(7) show the regeneration of Fe²⁺ from Fe³⁺ ion by consuming H₂O₂ [20,21].

2H⁺ + Fe⁰ + O₂ → Fe²⁺ + H₂O₂

(1)

2H⁺ + H₂O₂ + Fe⁰ → Fe²⁺ + 2H₂O

(2)

H₂O₂ + Fe²⁺ → OH^• + OH⁻ + Fe³⁺

(3)

Fe³⁺ + H₂O₂ → Fe²⁺ + ^•O₂H + H⁺

(4)

Fe²⁺ + ^•OH → Fe³⁺ + OH⁻

(5)

Fe²⁺ + ^•O₂H → Fe³⁺ + OH⁻₂

(6)

Fe³⁺ + ^•O₂H → Fe²⁺ + O₂ + H⁺

(7)

OH^• + OH⁻ → H₂O₂

(8)

OH^• + Rⁿ → OH⁻+ Rⁿ⁻¹

(9)

OH^• + RH → H₂O + R^•

(10)

Response surface methodology (RSM) is a statistical and mathematical Design-Expert software which is commonly used for optimization purposes. It develops experimental design based on level of variables, which may help to minimize experimental work by providing optimal conditions [22]. In addition, it validates the interaction between independent and dependent factors [23]. Central composite design (CCD) is a standard, reliable, and accessible design in RSM. CCD improves statistical interpretations and gives fewer experimental runs with an overall experimental error [22].

Although heterogeneous Fenton like oxidation system using nZVI as a catalyst has been used for remediation of different pollutants in various wastewaters, such as groundwater [1], palm oil mill effluent [24], municipal wastewater [25], and metalworking fluids [26]. However, to date this has never been employed for PAHs removal from real oily produced water. Furthermore, no application of RSM has been observed which optimizes the combined PAHs and COD removal in produced water. Therefore, this work aims (i) to evaluate the potential of heterogeneous Fenton like oxidation system using nZVI as a catalyst for PAHs and COD removal from real produced water and (ii) to design experiments and evaluate optimal conditions of independent variables including, concentrations of nZVI and H₂O₂, pH and contact time using CCD and RSM.

2. Materials and Methods

2.1. Materials

Produced water was collected from oil and gas exploration site in South East Asia region and stored at 4 °C following standard protocol prescribed by American Public Health Association (APHA) [27]. The analytical grade chemicals H₂SO₄ (95–98%), NaOH (30% w/v), dichloromethane (DCM), acetonitrile and H₂O₂ (30%) were purchased from R&M Chemicals Malaysia and utilized as received. A syringe filter with 0.45 μm pore size (Whatman U.S.A, 25 mm dia) was used for filtration. Commercially available nZVI was used in this study without any surface modifications. The characterization of nZVI used was provided by manufacturer as more than 80% iron content, 50 nm particle size and 20–25 m²/g surface area. Many researchers have employed same nZVI (purchased from NANO IRON) for the treatment of copious types of wastewater and characterized using scanning electron microscope (SEM), X-ray diffraction (XRD) and transmission electron microscope (TEM). The results obtained were in line with the manufacturer’s reported values. For instance, Taha et al. [24] characterized commercially available nZVI (from NANO IRON, Židlochovice, Czech Republic) using BET, SEM, XRD, TEM analysis and reported that pore size, surface area and size of nZVI were 6.18 nm, 2.4 × 10⁻⁵ m³/kg, and 49 nm respectively. Further, they accredited the existence of elemental iron (Fe) via XRD analysis. Energy dispersive X-ray (EDX) connected with SEM showed that surface of nZVI was comprised of Fe, carbon, oxygen and silicon. TEM image showed the presence of shell layer and core of nZVI [24]. Almost similar results of nZVI (from NANO IRON, Židlochovice, Czech Republic) characterization were reported by several other researchers [28,29,30,31,32]. Based on comprehensive literature review of commercially available nZVI (from NANO IRON, Židlochovice, Czech Republic), it is expected that nZVI used in current study may have similar characteristics to nZVI used in previous studies illustrated above.

2.2. Heterogeneous Fenton Like Oxidation System

Produced water sample of 300 mL was taken in a reactor (1 L glass beaker). The reactor was covered with an aluminum sheet to protect the water sample from sunlight. pH of produced water was adjusted using an appropriate volume of 1 M H₂SO₄ and 1M NaOH solution. A known amount of nZVI and H₂O₂ were added to the water sample to initiate heterogeneous Fenton like oxidation reaction. A magnetic stirrer was used for homogeneous mixing. An increase in pH of reaction mixture was observed during the oxidation process due to reaction of nZVI with H₂O₂ as shown in Equation (11) [33].

2H₂O + Fe⁰ → Fe²⁺ + H₂ + 2OH⁻

(11)

After a certain contact time, reaction was stopped by increasing pH of reaction mixture up to pH 10.

2.3. Produced Water Characterization

Total suspended solids (TSS) and five days biochemical oxygen demand (BOD₅) was measured following the standard methods (APHA, 1992) [33]. Total organic carbon (TOC) was determined using TOC analyzer (TOC-VCSH, Shimadzu, Japan). COD analysis was performed using HACH method following U.S. Environmental Protection Agency (USEPA) method [1,34]. The presence of ferrous ions (Fe²⁺) in aqueous solution during COD analysis may interfere the COD value [35]. Therefore Fe²⁺ ions were eliminated in the form of insoluble ferrous hydroxide (Fe(OH)₂) by increasing pH of the aqueous solution up to pH 10. Treated produced water was centrifuged (Heraeus, MEGAFUGE 16-Centrifuge, Hanau, Germany) at 5000 rpm for 15 min to separate the insoluble Fe(OH)₂. Finally, the supernatant was taken for COD and PAHs analysis. The COD and PAHs were calculated using Equations (12) and (13), respectively [36,37]. All experiments were performed in triplicate and average values were reported.

$$\mathrm{COD} (\%)=\frac{CO{D}_i-CO{D}_f}{CO{D}_i}\times 100$$

(12)

$$\mathrm{\Sigma }\mathrm{PAHs} (\%)=\frac{\Sigma PAH{s}_i-\Sigma PAH{s}_f}{\Sigma PAH{s}_i}\times 100$$

(13)

where in Equations (12) and (13) the subscripts “i” and “f” stand for initial and final concentration of COD and PAHs which were taken before and after treatment of produced water. Produced water samples for gas chromatography–mass spectrometry (GC-MS) analysis were prepared using liquid-liquid extraction (LLE) following USEPA method (3510C LLE) with few modifications. Briefly, 250 mL of produced water was taken in 500 mL separating funnel and 75 mL of DCM was added followed by vigorous shaking of about 5 min. The reaction mixture was kept at room temperature until a separate layer of aqueous and organics was formed [27]. Bottom organic phase was collected into a flask. The extraction was performed thrice using same aqueous layer and top supernatant was discarded. After mixing the three extracts into same flask, a calculated amount of anhydrous sodium sulfate was added into conical flask to absorb the water and then sample was filtered. The organic phase was concentrated up to 0.5 mL by rotary evaporator and transferred into 1.5 mL glass GC-MS vials and topped up using acetonitrile.

Prepared samples were analyzed via GC-MS (Agilent, model G7035A, combination of 7820A GC system with 5977E MSD, Santa Clara, CA, USA) with 30 m Elite 5MS column (inner diameter and film thickness were 0.25 mm and 0.25 μm, respectively) with helium as a carrier gas. The temperature of column was enhanced from 60 °C to 175 °C with 6 °C/min followed by the temperature of 240 °C with 3 °C/min and then finally held for 7 min at 300 °C. Temperature of injector and transfer line was 280 °C and 300 °C, in turn [1]. Identification was accomplished in selected ion monitoring (SIM) mode along with molecular masses of individual PAHs. A 16 PAHs mix standard solution with concentration of 2000 mg/L (Cat. No. 47930-U, Sigma Aldrich, Malaysia) was used for quantification of PAHs. The mixed standard solution was used by taking five-point calibration with a range from 10–300 μg/L. GC-MS chemstation software was utilized for assembling and treating the chromatographic data.

2.4. Statistical Modelling and Design of Experiments Using Response Surface Methodology (RSM)

Statistical modelling and optimization of system was conducted by RSM and CCD based on Design-Expert software (Version 11, Stat-Ease, Minneapolis, MN, USA). RSM evaluated optimal values of independent parameters such as concentrations of nZVI (A), and H₂O₂ (B), contact time (C) and pH (D). PAHs and COD removal was referred as responses. Performance of modelling technique was assessed by studying PAHs and COD removal efficiencies. For more straightforward statistical design, temperature and stirring speed were kept constant at 25–29 °C (room temperature) and 200 rpm, respectively to prevent from influence of other factors. All independent variables were changed into dimensionless codes, such as X₁ (nZVI concentration), X₂ (H₂O₂ concentration), X₃ (contact time), and X₄ (pH) for comparison of variables with different units. It helps to reduce the error in statistical analysis (Equation (14)) [38].

$$\mathrm{XI}=\frac{X_i-X_o}{\mathsf{\Delta }\mathrm{X}}\mathrm{i} =1, 2, 3, \dots ,\mathrm{k}$$

(14)

In Equation (14), XI is the variable’s coded values. While subscript “i” and “o” stand for ίth value of independent parameter, and X_i value at the center point, respectively. Whereas, ΔΧ is representing step change value. The ranges of variables like −1, 0 and +1 were designated to minimum, center and maximum level of variables according to the CCD principle, respectively. Ranges of all variables were finalized via preliminary experiments based on literature review. The CCD model levels (−1, 0 and +1) and all independent variables used in the current study are shown in

Table 1. Actual and coded values for experimental design for the Fenton process.

Factors	Independent Variables	Coded Values			Star Point α = 1
		−1	0	1	−α	+α
A	nZVI conc. (g/L)	1.0	3.50	6.0	1.0	6.0
B	Hydrogen peroxide (g/L)	0.4	1.625	2.85	0.4	2.85
C	Contact time (min)	30	115	200	30	200
D	pH	1.0	3.0	5.0	1.0	5.0

.

According to the principle of CCD, design consists of fractional factorial point (2^k), axial point (2k) and center point (1). K represents total number of independent variables (4) defined earlier. Therefore, 30 experimental runs were conducted with fractional factorial point 16 (2⁴), axial point 8 (2 × 4) and central points (1). In addition, five replications of center point were performed for the final estimation of error. The values of experimental design are presented in

Table 2. Experimental runs for PAHs and chemical oxygen demand (COD) removal.

Experiment No.	Independent Variables
	nZVI Conc. (g/L)	H2O2 Conc. (g/L)	Contact Time (min)	pH
1	6.0	2.85	200	1.0
2	6.0	0.40	30	1.0
3	3.5	1.62	115	3.0
4	3.5	2.85	115	3.0
5	3.5	1.62	115	3.0
6	3.5	1.62	115	3.0
7	1.0	2.85	200	1.0
8	3.5	1.62	115	3.0
9	3.5	0.40	115	3.0
10	3.5	1.62	200	3.0
11	1.0	0.40	200	1.0
12	6.0	1.62	115	3.0
13	1.0	2.85	30	1.0
14	1.0	2.85	30	5.0
15	3.5	1.62	115	1.0
16	6.0	2.85	30	5.0
17	3.5	1.62	30	3.0
18	1.0	0.40	30	5.0
19	6.0	0.40	30	5.0
20	3.5	1.62	115	3.0
21	3.5	1.62	115	5.0
22	6.0	2.85	200	5.0
23	1.0	0.40	200	5.0
24	6.0	2.85	30	1.0
25	1.0	1.62	115	3.0
26	3.5	1.62	115	3.0
27	1.0	2.85	200	5.0
28	6.0	0.40	200	5.0
29	1.0	0.40	30	1.0
30	6.0	0.40	200	1.0

. The behavior of experimental design was analyzed using CCD regression model which is represented by Equation (15) [39,40].

Y = b₀ + b₁A + b₂B + b₃C+ b₄D + b₁₂AB + b₁₃AC + b₁₄AD + b₂₃BC + b₂₄BD + b₃₄CD+ b₁₁A² +b₂₂B² + b₃₃C²+ b₄₄D²

(15)

where Y is a predicted dependent variable and b₀, b₁, b₂, b₃, b₄, b₁₂, b₁₃, b₁₄, b₂₃, b₂₄, b₃₄, b₂₂, b₃₃ and b₄₄ are regression coefficients. A, B, C and D are independent variables that delineate linear effects, whereas, AB, AC, AD, BC, BD, and CD, are independent variables which describe interactive impact, while A², B², C² and D² show quadratic effect.

The quadratic regression Equation (15) can be used as a function of independent variables to predict the responses (PAHs and COD removal) and also can elucidate the interaction between independent parameters. The analysis of variance (ANOVA) interprets complicated relationship between independent and dependent factors of complete data. The model’s significance is usually assessed by ANOVA results. Similarly, R-squared values of both models are also used for the confirmation of model significance [41]. In addition, sensitivity of four independent variables for PAHs and COD removal efficiencies are demonstrated by three dimensional (3D) plots.

3. Results and Discussion

3.1. Produced Water Characteristics

Produced water contains large amount of organic and inorganic compounds resulting in higher concentrations of COD, TSS, turbidity and TOC. Therefore, characterization of water is important for designing an efficient treatment system.

Table 3. Characteristics of produced water.

Characteristics	Concentration (mg/L)
BOD	50.0
COD	2213.0
TSS	139
Turbidity	85 NTU
NH3-N	28.0
TOC	759.4

depicts the characteristics of produced water used in this study.

The COD and TOC concentrations of 2213 mg/L and 759.4 mg/L, respectively, indicate presence of higher number of organic compounds in produced water [42]. Figure 1 portrays characterization of 15 different PAHs compounds present in real produced water. Naphthalene was presented at highest concentration of 201.46 μg/L and benzo (g, h, i) perylene was in lowest concentration of 23.15 μg/L as compared to other PAHs. Our results are in line with earlier studies where various researchers have reported PAHs and COD of produced water in the range of 124–1000 μg/L and 1220–2600 mg/L, respectively [43,44,45,46].

3.2. Preliminary Batch Experiments

The preliminary experiments were conducted to optimize the operating parameters, such as concentrations of nZVI and H₂O₂, pH and contact time for COD removal. The nZVI was tested in range of 1.0 to 6.0 g/L while keeping pH and contact time constant at 3.0 and 120 min, respectively. Initially, 32.4% of COD was reduced at 1.0 g/L of nZVI. As the concentration of nZVI increased from 1.0 g/L to 5.0 g/L, removal efficiency of COD was also increased. The highest COD reduction of 53.5% was obtained at 5.0 g/L of nZVI. Furthermore, at nZVI concentration higher than 5.0 g/L, a decrease in COD removal was observed and at 6.0 g/L of nZVI, COD removal efficiency reduced to 45.7%. The cause for this phenomenon may be linked to rise in the concentration of Fe²⁺ iron associated with nZVI concentration increment. Therefore, the increase in the concentration of Fe²⁺ led to enhance scavenging of OH^• which led to the production of Fe³⁺ and OH⁻. The Fe³⁺ iron and OH⁻ passivated the nZVI surface [47]. As a consequence, the Fe²⁺ iron substituted the refractory organic pollutants present in aqueous solution as the primary sink for OH^•, leading to a significant reduction in degradation rate of organic pollutants at higher concentration [48]. Accordingly, 1.0–6.0 g/L of nZVI was chosen for further analysis of PAHs and COD removal using CCD. Even though COD started to decline after 5.0 g/L of nZVI concentration, 1.0 g/L and 6.0 g/L were selected as a maximum and minimum values for RSM, respectively, to design comprehensive experimental run.

The effect of H₂O₂ was studied by changing the concentration in the range of 0.40 to 3.0 g/L while keeping pH, contact time and concentration of nZVI fixed at 3.0, 120 min and 5.0 g/L, respectively. At 0.5 g/L of H₂O₂, 45% of COD removal was obtained and it increased up to 71% at 2.0 g/L of H₂O₂. Further increases in H₂O₂ concentration resulted in decreases in COD removal efficiencies. After investigating the impact of H₂O₂ values on COD removal via screening test. It was noticed that a slight increase beyond 2.0 g/L of H₂O₂ concentration decreased COD removal efficiencies. Thus, for comprehensive experimental design 0.40 g/L, and 2.85 g/L H₂O₂ concentration was selected for RSM.

The effect of pH on COD removal was studied using pH values from pH 1.0 to pH 5.0 as shown in Figure 2c, while keeping the concentration of nZVI and H₂O₂ fixed at 5.0 g/L and 2.0 g/L, respectively. The maximum COD removal of 71.2% was attained at pH 3.0. Further increase in pH resulted in reduction of COD removal efficiencies. Hence, pH 1.0 and 5.0 were chosen as a minimum and maximum values for Design Expert (CCD/RSM) to study pH influence on organic pollutants removal.

The effect of contact time on COD removal was investigated from 30 to 200 min, while keeping all other operating parameters constant as shown in Figure 2d. Initially after 30 min of reaction time, 50% of COD removal was achieved. Contact time showed significant impact on organic oxidation and maximum COD removal of 70.5% was obtained at contact time of 200 min. No significant increase in COD removal was observed by increasing the contact time higher than 200 min. Thus, time ranges of 30 to 200 min were used for optimization purpose using Design-Expert software (Stat-Ease, Minneapolis, MN, USA). In all screening tests, the COD removal efficiency was considered as response factor (output) which evaluated the organic pollutant’s degradation potential present in produced water. Based on the findings of organic pollutants removal, further tests were designed via RSM for the removal of PAHs and COD from produced water.

3.3. Batch Experiments Based on RSM Experimental Design

By performing batch experiments, the ranges of independent variables such as A, B, C and D were chosen for RSM optimization as shown in Table 1. These ranges of independent variables were utilized in CCD/RSM to design the experimental run for optimization work. CCD produced rotatable 30 experimental runs having different values of independent variables as shown in Table 2. All 30 sets of experiments proposed by CCD were performed in lab to study the removal of PAHs and COD for each set. The responses were predicted and the interaction between independent variables was also investigated using second order quadratic model as shown in Equations (16) and (17).

COD removal efficiency (%) = 73.08 + 11.49A + 2.08B + 1.21C − 1.81D + 8.91AB − 1.28AC − 0.7812AD − 2.66BC − 1.41BD + 1.28 CD − 12.40 A² − 6.10B² − 0.4500C² − 18.60D²

(16)

PAHs removal efficiency (%) = 85 + 7.53A − 3.79B + 3.34C − 3.93D + 5.78AB − 1.15AC − 0.100AD − 1.61BC − 3.99BD + 2.61CD − 11.78A² − 7.68B² + 0.17C² − 14.98D²

(17)

The experimental and predicted removal of PAHs and COD varied for each set of tests depending on different values of independent parameters. The maximum removal of PAHs (89.5%) and COD (76.3%) was observed in 10th experimental run, as depicted in Figure 3. The predicted and experimental values of all responses were in good agreement.

3.4. Analysis of Variance and Fit Summary

The analysis of variance (ANOVA) is a tool which presents descriptive statistics and statistical tests. ANOVA interprets complicated relationship between independent and dependent factors of complete data. The results are assessed using different statistical analysis and tests, such as alpha (0.05) at 5%, probability (p-value) at 95% confidence level and Fisher’ test (F-test). Typically for higher F-values (greater than 4.0) and lower p-values (less than 0.05), ANOVA table identifies the parameter’s significance in the model. In this study the results obtained from ANOVA indicated that all four independent factors and their interactions were substantial and significantly influenced the removal of PAHs and COD. The higher F-values of PAHs (16.87) and COD (37.86) models showed that both models were significant. Likewise, p-values of both models were less than 0.05, which confirmed significance of both models. Thus, these values are acceptable and all independent parameters have a significant impact on models.

Furthermore, coefficient of determination (R²) for PAHs and COD were 0.94 and 0.97, respectively which displayed reliability in PAHs and COD removal estimation. The higher values of R² exhibited a good relationship between predicted and observed values [49]. Adequate precision (AP) for PAHs and COD removal efficiency were found 15.07% and 17.7%, respectively which showed adequate signal. The predicted and experimental values of responses models correlated with each other and indicated a significant agreement as shown in Figure 4 and Figure 5, where a linear behavior has been demonstrated by plots which suggest that prediction of the models was accurate.

3.5. 3-Dimensional Surface Plots for PAHs and COD Removal Efficiency

3D plots carrying different peak values and curves represent the impact of independent variables and their interaction on PAHs and COD removal. In 3D plots, two independent variables were systematically varied while keeping other independent variables fixed. The most promising results of PAHs and COD removal efficiencies were chosen as shown in Figure 6 and Figure 7, respectively.

The impact of pH on PAHs and COD removal efficiencies is shown in Figure 6a–c and Figure 7a–c, respectively. In Figure 6 and Figure 7 it can be seen that by increasing the pH value up to a certain value has a positive influence on PAHs and COD removal efficiencies. While varying the pH value above optimum value have negative impact on removal efficiencies. In these Figures, nZVI concentration, H₂O₂ concentration, and contact time were kept fixed at 3.0 g/L, 1.38 g/L and 120 min, in turn. Initially at pH of 1.0, the PAHs removal efficiencies were 73.0%, 74.2% and 74.5%, whereas, COD removal efficiencies were 54.2%, 55.7% and 56.5%. An increment in organic removal was found as the value of pH was increased from 1.0 to 3.0. Further, at pH 2.0 the PAHs removal efficiencies surged up to 78.5%, 79.4% and 81.2% (Figure 6). Similarly, COD removal efficiencies followed the same trend for pH 2.0 and increased up to 58.3%, 59.7% and 61.7% (Figure 7). Likewise, at pH 3.0, PAHs removal efficiencies increased up to 83.4%, 85.8%, and 85.2% (Figure 6). Similarly COD removal increased up to 70.3%, 72.0%, and 73.2% (Figure 7). The removal efficiencies trend started to decline as the pH value was increased beyond pH 3.0. pH of the aqueous solution is a very important parameter that controls the production of Fe²⁺ and OH^• during the reaction. In both Figure 6a–c and Figure 7a–c pH below 3.0 decreases the PAHs and COD degradation due to the generation of complex Fe species and oxonium ions (H₃O⁺). The pH values above pH 3.0 may inhibit the hydrogen peroxide decomposition due to the absence of H⁺ ions and therefore decrease the formation of OH^• radical, which may decrease the removal efficiencies of PAHs and COD above pH 3.0.

Besides, at higher pH value of aqueous solution, Fe³⁺ converts into Fe(OH)₃ in the form of precipitate, which hinders the generation of Fe²⁺ [50]. While at pH 3.0, nZVI/H₂O₂ generates a higher amount of Fe²⁺ and OH^• ions, which are the main ions for the degradation of organic pollutants in produced water. Shafieiyoum et al. [33] reported that the increase in pH above 3.0 decreases OH^• oxidation potential. Similar results have been reported by other researchers showing pH 3.0 as optimal value for heterogeneous Fenton like oxidation system using nZVI as catalyst for maximum removal of organic pollutants [51]. Yaqub et al. [52] reported that highest PAHs degradation was attained at pH 3.0 in produced water. Viardi et al. [53] achieved maximum COD removal efficiency from tannery wastewater at pH 3.0. Similarly, Yu et al. [50] reported pH 3.0 as the optimum value for maximum removal of organic pollutants. In another stidy, Kallel et al. [54] found that acidic medium (between pH 2.0–4.0) is favorable for the production of Fe²⁺ and OH^• which are primary ions for elimination of carbon-based pollutants from wastewater. The effect of H₂O₂ concentrations on PAHs and COD removal has been shown at pH 4.0 (Figure 6b and Figure 7b), and 166 min of contact time (Figure 6e and Figure 7e), respectively. Initially, at 0.40 g/L concentration of H₂O₂, the PAHs reduction efficiencies were 78.6% and 84.0%, while COD removals were about 60.0% and 67.1%. In addition, 3D plots showed that increase in the concentration of H₂O₂ from 0.40 g/L to 1.87 g/L improved PAHs and COD removal efficiencies. It was observed that the value of PAHs removal efficiencies surged up to 80.3% and 85.1% at 1.38 g/L dosage of H₂O₂ (Figure 6). Similarly, COD removal efficiencies were improved from 60.0% to 70.0% (Figure 7). Moreover, the increment in PAHs and COD removal efficiencies were reached up to their maximum values 87.0% and 75.3%, respectively, at 1.87 g/L concentration of H₂O₂ (Figure 6 and Figure 7). Further increases (beyond 1.87 g/L) in H₂O₂ concentration decreased the removal efficiencies of PAHs and COD. The influence of H₂O₂ concentration can be illustrated by reaction of H₂O₂ and Fe²⁺ ions. As concentration of H₂O₂ increased, it generated a larger amount of OH^• by reacting with Fe²⁺ increasing removal efficiency of organic pollutants [35]. The effect of H₂O₂ concentration on PAHs and COD removal depends on the dosage of nZVI also. An excessive amount of H₂O₂ in the solution can cause ^•OH scavenging effect and produces hydroperoxy radical (^•HO₂) which have less oxidation potential than OH^• as given in Equation (18) [43]. Few studies have reported approximately 1.87 g/L of H₂O₂ concentration as optimum value for maximum removal of organic pollutants from different types of wastewater which is in line with our findings [20,35]. Taha et al. [24] reported similar results, and they obtained a maximum 75.0% removal of organic pollutants from palm oil mill wastewater at 1.84 g/L of H₂O₂ concentration. Bogacki et al. [20] achieved highest organic pollutant removal (76.0%) from wastewater at 1.90 g/L concentration of H₂O₂.

H₂O₂ + OH^• → ^•HO₂ + H₂O

(18)

The effect of nZVI dosage on PAHs and COD degradation was noticeable as shown in Figure 6a,f and Figure 7a,f, respectively. When nZVI dosage was increased from 1.0–6.0 g/L at pH 2.0 and 90 min of contact time, PAHs and COD removal varied accordingly. Figure 6a and Figure 7a depicts that at 1.0 g/L of nZVI dosage PAHs and COD removal were 65.6% and 45.3%, respectively. The PAHs and COD removal efficiencies surged by increasing the concentration of nZVI from 1.0 g/L until 4.0 g/L. At 4.0 g/L concentration of nZVI, the removal efficiencies of PAHs and COD were 84.0% and 71.0%, respectively (Figure 6a and Figure 7a). However, beyond 4.0 g/L of nZVI, PAHs and COD removal started to decrease. At 6.0 g/L of nZVI, removal of PAHs and COD were 79.0% and 68.3%, respectively (Figure 6a and Figure 7a). In Figure 6f and Figure 7f, similar trends were followed by nZVI concentration for PAHs and COD removal efficiencies, respectively. At 1.0 g/L dosage of nZVI, the PAHs and COD removal efficiencies were at their minimum level, 60.1% and 45.5% respectively. The reduction of PAHs and COD increased as the dosage of nZVI was increased from 1.0 to 4.0 g/L. At 4.0 g/L of nZVI concentration the PAHs and COD elimination reached up to 70.3% and 55.8%, respectively. In addition, the removal efficiencies of PAHs and COD declined as the concentration of nZVI was increased beyond 4.0 g/L. At initial concentration of nZVI (1.0 g/L), less amount of Fe²⁺ was produced but as the nZVI concentration increased, the amount of Fe²⁺ also increased, more cites were available, which favored the higher generation of OH^• [26]. This phenomenon enhanced removal of organic contaminants. Nevertheless, increase in nZVI concentration beyond optimum value releases more Fe²⁺ ions in aqueous solution that may cause scavenging of OH^• by formation of Fe₂O₃ and OH⁻ and these oxides passivate on the surface of nZVI as shown in Equation (19) [55]. The Fe²⁺ ions may replace the pollutants present in aqueous solution and react with OH^• instead of organic compounds, which causes reduction of organic pollutants degradation at higher concentrations of nZVI.

Fe²⁺ + OH^• → Fe³⁺ + OH⁻

(19)

Several researchers have reported optimum value of nZVI concentration for maximum removal of organic pollutants is equal or close to 4.0 g/L [35]. Bogacki et al. [20] obtained maximum organic pollutants removal from automotive wastewater at 4.0 g/L of nZVI concentration. Taha et al. [45], reported 3.91 g/L of nZVI concentration as optimum value for maximum removal of organic pollutants, and they achieved highest removal (75%) of organic pollutants. The best PAHs and COD reduction efficiencies were obtained in the range of 60.0–89.5%, and 40.0–76.3%, respectively, which is based on significant interaction between four input factors. In this study, higher removal of PAHs and COD was obtained using heterogeneous Fenton like oxidation system as compared to previously reported studies [18,35,54,56,57]. Some studies used synthetic wastewater and have reported higher removal of organics than the current work. However, in synthetic wastewater just a certain amount of targeted organic pollutants are added, and reagents do attack only on targeted organic contaminants rather than on other unwanted compounds [58,59], as shown in

Table 4. Comparison of previously conducted heterogeneous Fenton like oxidation studies with the current study.

Method	Type of Effluents	Pollutants	Removal (%)	References
nZVI/H2O2	Synthetic wastewater	Polyvinyl alcohol	94.0	[55]
nZVI/H2O2	Synthetic wastewater	pentachlorophenol	57.0	[56]
nZVI/H2O2	Tannery wastewater	COD	75.5	[57]
nZVI/H2O2	Synthetic wastewater	Benzotriazole and COD	73.4 and 40.0	[58]
nZVI/H2O2	Synthetic wastewater	Amoxicillin and COD	86.5 and 71.2	[11]
nZVI/H2O2	Contaminated soil	PAHs	80.0	[52]
nZVI/H2O2	Tannery wastewater	TOC, COD and phenol	70.0, 73.0 and 88.0	[59]
nZVI/H2O2	Landfill Leachate	COD	74.0	[60]
nZVI/H2O2/microwave irradiation	Landfill Leachate	Organic compounds	58.0	[61]
nZVI/H2O2/biodegradation	Synthetic wastewater	COD and naphthalene	55.2 and 99.0	[47]
nZVI/H2O2	Produced water	COD and PAHs	76.3 and 89.5	This Study

.

3.6. Optimization and Validation of Experimental Study

Optimal values for every individual experiment (response) is considered very hard to obtain. Therefore, numerical optimization is the best option to get optimal values. Numerical optimization is RSM based technique which is usually applied after ANOVA model validation to find the optimal conditions of all independent variables for desired removal percentage of responses. It also predicts removal percentage of responses based on proposed optimal values of independent variables which need to be validated by conducting an additional experiment [59]. The principle of this optimization is based on desirability function and comprehensive explanation of desirability function has been explained by X.Y. Shi et al. [62]. For this purpose “numerical optimization” based on quadratic models (Equations (16) and (17)) and the parameters in their critical ranges were practiced. For numerical optimization of operating conditions for the current study, “in range” options were selected for independent variables (nZVI concentration, H₂O₂ concentration, contact time and pH) while for PAHs and COD removal “maximize” options were chosen. The PAHs and COD removal efficiencies predicted by numerical optimization were 89.38% and 75.74%, respectively. Besides, these removal efficiencies were predicted at the optimum values of all four-independent variables such as nZVI (4.35 g/L), H₂O₂ (1.60 g/L), contact time (199.9 min) and pH (2.94).

For the validation of optimal conditions and the entire model an additional experiment was performed using the proposed optimal values of independent parameters. The responses obtained from experiment were same as the models predicted with a slight difference. The removal efficiencies of PAHs and COD gained from the experiment at predicted optimal conditions were 87.4% and 73.5%, respectively. There was a minor difference between predicted and experimental results at the optimal conditions which verified the model validity. Besides, all HMW PAHs were removed entirely, while it partially removed LMW PAHs such as naphthalene, acenaphthylene, acenaphthene, and fluorene. The concentration of naphthalene, acenaphthylene, acenaphthene, and fluorene after the treatment of produced water at optimal conditions were 64.13 μg/L, 30 μg/L, 42.23 μg/L, and 23.21 μg/L, respectively, as shown in Figure 8.

This shows that these LMW PAHs were more stable and resistant against the heterogeneous Fenton like oxidation system than the HMW PAHs. The small difference between predicted and experimental values indicated that CCD design is a useful tool to obtain the best optimal conditions for maximum PAHs and COD removal from produced water.

In summary, a heterogeneous Fenton like oxidation system using nZVI as a catalyst was utilized for the removal of PAHs and COD from produced water. Design-Export software was practiced to investigate the optimal conditions of four independent variables with significant p-values of PAHs and COD models (<0.05). In the validation experiment, there was a minor difference between predicted and experimental values, which indicated that the model was significant. The maximum removal of PAHs and COD at optimal conditions was 87.4% and 73.5%, respectively. Interestingly, real produced water was utilized and all experiments were performed at lab scale in current study. Based on the excellent results of heterogeneous Fenton like oxidation system obtained for the treatment of PW, it is expected that this system can efficiently treat PW in a bench scale setup. Therefore, in future, this system can be extended in a bench scale reactor built according to full scale plant configuration for the treatment of real produced water. Furthermore, the obtained optimized operating conditions from bench scale reactor could be employed for full-scale plant.

4. Conclusions

This study investigated heterogeneous Fenton like system using nZVI as a catalyst for removal of organic and inorganic carbon pollutant from real produced water. Independent variables were studied via preliminary experiments and then used in RSM optimization for the removal of PAHs and COD. RSM based CCD was used to design the experimental sets. CCD developed total 30 experimental runs with different values of four independent variables. Experimental conditions were optimized using numerical optimization in CCD/RSM to achieve highest removal of PAHs and COD. The optimal values were 4.35 g/L of nZVI and 1.60 g/L of H₂O₂ at reaction time of 199.90 min and 2.94 pH. The maximum predicted values of PAHs and COD removal at optimal conditions were 89.3% and 75.7%, respectively. While experimental values of PAHs and COD removal were 87.4% and 73.5%, respectively. The small error between predicted and experimental responses confirms validation of the designed model. All HMW PAHs were removed completely, whereas, LMW PAHs such as naphthalene, acenaphthylene, acenaphthene, and fluorene were partially removed. Based on obtained results, heterogeneous Fenton like oxidation system using nZVI as a catalyst is presented as promising method for the treatment of produced water, which can be extended to the removal of other micropollutants such as Endocrine-disrupting chemicals (EDCs), and pharmaceuticals.