Using BCN nanostructure as anode electrode for photoelectrocatalytic degradation of organics: a statistical approach

There is too much water inside the oil and gas reservoirs which could appear during the extraction and production operations. This wastewater includes injection, condensed and formation water, and a few treatment chemicals which is called produced water (Xu et al. 2016). Produced water contains organic compounds, dissolved, and suspended solids (Guerra & Drewes 2008); hence, it is very hazardous to the environment and its treatment has become a challenge when brought to the surface. The largest volume of wastewater that deals with oil, gas, and petroleum industry can be produced water (Fan et al. 2018). About 250 million barrels of produced water was reported to be produced each day in the world in 2015 and its equivalent production of oil was 80 million barrels, which showed that its volume was three times more than the oil volume. For all kinds of produced water, different items such as organics and inorganics, high total dissolved solids (TDS) and consequently high electrical conductivity, and some fatty acids and fouling oil are the common concerns independently of the location (Xu et al. 2016).

Advanced types of reactors with an immobilized photocatalyst have been recently studied for photoelectrocatalysis. Some researchers showed that these kinds of reactors could degrade different groups of harmful components efficiently (Quan et al. 2005; Suhadolnik et al. 2016).

Different parameters as independent variables can have effects on pollutants removal from wastewaters as dependent variables (Malvestiti et al. 2019). The optimization of the dependent variable is the goal of many studies so does ours. For optimization in conventional methods, all of the independent variables should be kept constant except the one that can vary only one at a time (Hassani et al. 2015; Badli et al. 2017; Baig et al. 2020). So many experiments are needed while using these methods which take more time and cost more (Sharif et al. 2016). Experimental design for the modeling and optimization of produced water treatments is an important procedure to promote pollutants removal efficiency while decreasing the number of experiments considerably (Chen et al. 2019; Sheikholeslami et al. 2019; Shahriari & Hosseini 2020).

In this study, the modeling and optimization of organic pollutants removal processes in produced water treatment were described. Photoelectrocatalysis was used as the treatment process for pollutants degradation, and boron carbon nitride (BCN) nanomaterials were applied as photocatalyst because it has a low bandgap energy (0.5 eV) with respect to current semiconductors (3.02 eV for TiO₂) (Tu et al. 2017; Ebadi et al. 2021). By using a 12 run Plackett–Burman design as the first step, 2 parameters were screened with 7 main parameters. Following the first step, the steepest ascent method was used for shifting to an optimum region, and finally, a central composite design (CCD) was planned to maximize the response described as chemical oxygen demand (COD) removal efficiency.

For photocatalyst (BCN) synthesis, Poly(ethylene glycol)-block-poly(propylene glycol)-block-poly(ethylene glycol) (P123), methanol (CH₃OH > 99.8%), melamine powder, and boric acid as reagent were purchased from Sigma Aldrich. Sulfuric acid and nitric acid (for functionalization of photocatalyst), hydrochloride acid (purity = fuming 37%), sodium hydroxide (for adjusting pH of the wastewater), and acetic acid (for cleaning copper coils before immobilizing of photocatalyst on them), were all Merck index. Copper wire with 250 μm diameter was used as anode and cathode electrodes. The wastewater was sampled from Iranian central oilfields.

Using a MIRA3 TESCAN electronic microscope, electron micrographs were scanned. The analysis of COD was done in a HACH reactor (USA) using K₂Cr₂O₇ as oxidant and a DR/200 portable data logging spectrophotometer (HACH, Germany) according to the standard methods for the examination of water and wastewater (Song et al. 2016). To determine the structure of the crystal phase of BCN, X-ray diffraction (XRD) was measured using an X'Pert Pro MPD Diffractometer (Panalytical, Netherlands) with radiation of Cu target (K∝1, λ = 1.54 nm) at 25 °C and 30% RH.

BCN nanosheets and nanorods were prepared by the chemical vapor deposition (CVD) reactor as follows: P123 (3 g) was added into a stirred solution of demineralized (DM) water and methanol (1:2 volume ratio). When this polymer became completely dissolved, boric acid (3.71 g) and melamine (60.5 g) were added to the solution gently. Approximately 2 h later, a white slurry with high viscosity was gradually formed. During the stirring for 24 h, the temperature of the mixture should be maintained at 50 °C. Finally, obtained white wet cake was transported to the CVD reactor. The temperature of the reactor was set to reach 800 °C with the ramp of 5 °C/min and then was maintained at 800 °C for 3 h. During the reaction, nitrogen gas purging was used to prevent oxidation reactions at high temperatures. Then the reaction product was cooled overnight by continuing nitrogen gas purging to obtain BCN nanostructures.

For immobilizing of BCN nanostructures on the copper electrode, first, they should be functionalized to be able to dissolve in water. For functionalizing of these nanostructures, a solution of nitric acid and sulfuric acid (200 mL, 1:3 volume ratio) was evaporated in a balloon and nanostructures got in touch with the vapor on a sintered disc above the balloon for 24 h. Then nanostructures (0.14 g) were dissolved in DM water (50 mL) and stirred for 40 min in an ultrasonic cleaning bath (Clifton) to be dissolved thoroughly. Previously cleaned coil-type copper electrode by acetic acid was immersed in this suspension for 1 h. Finally, the coil was kept inside the oven at 120 °C for 4 h to fabricate anode electrode. Figures 1 and 2 show the XRD and scanning electron microscopy (SEM) of BCN, respectively.

The wastewater used in this study was a real case from an Iranian gas field which was distilled first in an evaporation process before using for photoelectrocatalysis. Some characteristics of the wastewater are shown in Table 1.

The experiments were carried out in an experimental setup. Figure 3 shows the schematic diagram of the experimental setup for COD removal from a gas field produced water by photoelectrocatalysis. The most important section of this setup is the microreactor (3) that includes three coil-type copper electrodes which were placed on a 5 mm outer diameter glass rod inside a UV-transparent 6 mm-inner diameter Plexiglass tube. Two electrodes (12 mm length) were used as cathodes on the sides of the glass rod and one another (64 mm length) which was coated by BCN semiconductor nanostructures was used as an anode on the middle of the glass rod. All of these three electrodes were separated from each other by two 2 mm-length plastic pieces. Wastewater was charged in a 5 mL syringe with a specific diameter and was injected into the microreactor by a syringe pump (SP1000HOM Fnm, Iran) (1). Microreactor was illuminated by two 8 W UV lamps (4 W–365 nm and 4 W–254 nm, VILBER LOURMAT VL-4.LC) (2). An MCH-310D switching DC power supply (China) was used for creating voltage differences between the electrodes (5). The purified wastewater was collected in 10 mL vial bottles (4).

For each process, different kinds of parameters can have effects on its results. Studying the effect of all parameters separately, which is called one factor at a time, is time-consuming and would be more expensive (Fan et al. 2018). However, experimental design can be applied for the identification of significant factors (Kıranşan et al. 2015). The Plackett–Burman design is a statistical method that is applied for screening a large number of factors to specify the most effective parameters.

Optimum and valid results with a minimum effort, time, and resources are the primary objectives of applying the experimental design in the analytical process (Kıranşan et al. 2015). In an experimental design, response surface methodology (RSM) is a mathematical approach that fits the experimental data to a full-quadratic model, statistically (Nair et al. 2014; Myers et al. 2016).

The XRD pattern of photocatalyst in this study (Figure 1) shows that there are nine main reflections in this chart. Reflections that are observed in 2θ = 26.5, 43.5, and 50.5 belong to BCN (Tu et al. 2017). Three other peaks that occurred in 2θ = 18.5, 30.5, and 35.5 are related to hexagonal boron nitride (h-BN), and other peaks that are seen in 2θ = 16.5, 57.5, and 63 show that there is some boron and graphite oxide in the sample (He et al. 2013).

The SEM of the nanomaterial shows that most of the particles are in nanosheet and nanorod forms. Nanosheets and nanorods were tied together and made high surfaces, so high energy was adsorbed from UV light by the photocatalyst. Therefore, good results of photoelectrocatalysis efficiency were reached by this combination (Zhou et al. 2000; Tu et al. 2017).

In this study, the experimental output (⁠⁠) was COD removal efficiency (%), and parameters for studying their effects on y were initial COD (mg/L), electrical conductivity (μS/cm), applied cell voltage between anode and cathode (V), illuminating UV lamp wavelength (nm), H₂O₂ concentration (mM), hydraulic residence time (min), and initial pH of wastewater (Vinodgopal et al. 1996; Cornish et al. 2000; An et al. 2002; Mao et al. 2019). Determining low and high values of each parameter depends on the knowledge about the process. These boundary levels were selected to study if the parameters were effective on y or not. On the other hand, after the Plackett–Burman design, the data should be taken to the optimum region, so boundary levels in this step should be selected around the center of overall boundary levels.

The initial COD of wastewater was 750 mg/L (this is the wastewater that was obtained from the distillation of the main wastewater) and it could be decreased by dilution of the wastewater by DM water. It reached 562.5, 375, and 187.5 mg/L by adding the following amounts of DM water: 25, 50, and 75% of wastewater initial volume, respectively. Finally, 375 and 562.5 mg/L were selected as low and high values of COD. The initial electrical conductivity of wastewater was 1,100 μS/cm and increased to 1,800, 2,500, and 3,300 μS/cm by adding 350, 700, and 1,050 mg/L of NaCl to the wastewater. An average interval should be selected, so 1,800 and 2,500 μS/cm are used as boundary levels. The DC power supply used in this study could create 0–30 V potential differences between electrodes continuously, so 10 and 20 V were used as low and high levels, respectively. Because of the specification of the UV lamp, just two wavelengths (254 and 365 nm) were selected as boundary values of this parameter. According to recent studies, the optimum concentration of H₂O₂ in photoelectrocatalysis is 4–10 mM (An et al. 2002; Mao et al. 2019). Therefore, it was obvious that 6–8 mM was the best choice of average interval for H₂O₂ concentration. Photoelectrocatalysis is a fast process for the degradation of organic pollutants (Wei & Wan 1991), so the overall time interval was considered to be 5–20 min and between these, 10 and 15 min were considered as boundary levels for the Plackett–Burman design. For safety issues, minimum and maximum values of pH were considered to be 3 and 11, respectively. In this region, the pH values of 5–9 was selected as the central interval for the experimental design.

A 12-experimental PB design was used here, so there were 4 dummy factors labeled as A, C, F, and I which were included randomly in Table 3. The last column in Table 3 represents experimental responses at specified levels of parameters placed in their rows.

By using Equations (1) and (2), the analysis of variance (ANOVA) for the data in Table 3 is summarized in Table 4.

Each of the sum of squares (SS) in Table 4 has just one degree of freedom, so their mean square values (i.e., variances) are the same as the SS ones. The SS values for the dummy factors A, C, F, and I are similarly found to be 6.46, 75.18, 12.88, and 54.04, respectively. The mean SS for these estimates of the random measurement errors is thus 37.14: this has four degrees of freedom, as there are four dummy variables. Each of the individual factors can now be compared with this estimated random error using a one-tailed F-test at the p = 0.1 significance level. So for factor initial COD, the value of F is 1.4/37.14 = 0.04. The critical value of F_1,4 at p = 0.1 is 4.54, so it can be concluded that the effect of changing the level of the initial COD factor is not significant (Myers et al. 2016). The same approach shows that the initial pH factor has a significant effect and the residence time factor seems to have a little significant effect. Table 4 shows that the effects of residence time and pH are positive and negative, respectively, which are in agreement with the last research studies (Andand et al. 2018; Mansouri et al. 2018; Xu et al. 2018).

Table 5 shows the different steps of the steepest ascent.

Considering that pH could not be decreased to less than 3, doing the experiment correspondent to origin + 3Δ was impossible. The last row of Table 5 belongs to a pH value of 3 which was obtained by interpolation between origin + 2Δ and origin + 3Δ. It is obvious from Table 5 that y is continuously increased with increasing residence time and decreasing pH. Finally, when residence time is 15.88 min and pH is 3.00, y starts to decrease so a step before that (residence time: 15.00 min and pH: 4.04) is the optimum region and a CCD should be designed around this point for the response optimization.

The predicted responses by the regressive model (Equation (11)) are shown in the last column of Table 6 that indicates good adaptability with experimental y.

ANOVA for this design (CCD) is summarized in Table 7.

Analytical optimization by RSM predicts optimum y. Some independent variables, which resulted in maximum y (89.76%), were estimated by RSM. These variables were applied to fix the optimum conditions for maximum y. The greatest amount of y was predicted at a residence time of 15.86 min and an initial pH value of 3.33. An experiment was done in these conditions and resulted in 89.00% of COD removal efficiency which was in good agreement with the predicted value.

Contour and surface plots of COD removal efficiency versus residence time and pH are shown in Figures 4 and 5, respectively.

In this study, the design and manufacture of a microreactor with BCN as photocatalyst was explained thoroughly. This paper is aimed to use experimental design for the optimization of organics removal from a gas field produced water by photoelectrocatalysis. A 12 run Plackett–Burman experimental design was used for determining the most influential parameters by screening 7 different factors and results, which indicated that the following two parameters had significant effects on this process: residence time and initial pH. A linear predictive model was derived from the experimental results of Plackett–Burman design which its predicted values were in good agreement with experimental results. As COD removal efficiency was the target parameter for reviewing the influence of the photoelectrocatalysis process on organic pollutants removal, so the steepest ascent was used to maximize this parameter. The steepest ascent was used as a technique for taking selected parameters to the optimum region.

A CCD with 11 runs was applied to optimize the significant parameters obtained by the Plackett–Burman design. A polynomial regression model was obtained from CCD experimental responses which included individual parameters and their interactions. This model predicted values that were also in good agreement with experimental responses. The predicted results showed that the maximum COD removal of 88.79% could be achieved with a residence time of 15.68 min and a pH value of 3.54. The predicted values were in agreement with experimental values with a coefficient of determination (R²) of 0.90. The model was validated by subsequent experimentations at the optimized conditions.

We thank Mr Akbar Irandokht for his assistance in preparing this article. The authors are also grateful for the support provided by the Research Institute of Petroleum Industry (RIPI) in Iran.

All relevant data are included in the paper or its Supplementary Information.