Removal of sulphate from mine waters by electrocoagulation/rice straw activated carbon adsorption coupling in a batch system: optimization of process via response surface methodology

The removal of sulphate ions constitutes one of the main challenges in mining, metallurgical and other industries. This work evaluated sulphate removal from aqueous solutions by an electrocoagulation (EC)/raw straw activated carbon (RSAC) adsorption coupled process. The process parameters affecting sulphate removal ef ﬁ ciency were investigated: current density (0 – 100 mA/cm 2 ), RSAC dosage (0 – 0.8 g/L), initial pH (4 – 9) and reaction time (0 – 40 min). A central composite design coupled with response surface methodology (RSM) was used to construct a mathematic model of EC/RSAC process that considers three key variables, namely current density, RSAC dosage and reaction time. Under optimum conditions (current density of 75 mA/cm 2 , dosage of 0.46 g/L and reaction time of 19.2 min), the removal ef ﬁ ciency of sulphate reached 95.2%. The RSM predictive value was 94.08% with a small deviation (1.12%). Thus, the fundamental data and results can provide some useful information for further studies and applications of the EC/RSAC coupled system in sulphate-containing wastewater treatment.


INTRODUCTION
Mine water is a source of sulphate contamination of natural water and industrial wastewater (Najib et al. ). The presence of sulphate ) in mine water is associated with the oxidation of sulphides and elemental sulphur or chemical weathering of sulphur-containing minerals when the drainage is exposed to oxygen and water (Kefeni et al. ).
As a result, this drainage is often characterised by a low pH and high sulphate concentration (Papirio et al. ). This method has advantages over others for contaminated water remediation because it requires relatively simple equipment, is easy to automate, needs no additional reagents and has high efficiency with a short operation time (Kumar et al. ; Guvenc et al. ).

Materials
Mine water was collected from the storage dam of a mining area in Magu (located in Guizhou Province, southwest China, sampling date 5/12/16) and stored in a freezer at À20 C. The pH value was measured using an HQ11D pH meter (Hach Co., Ltd, Shenzhen, China). Total dissolved solids and electrical conductivity were monitored using a Eutech Instruments CON510 conductivity meter (Vensik Electronics Co., Ltd, Qidong, China) at the temperature of 25 ± 0.2 C. Cation analysis was performed using an inductively-coupled plasma-atomic emission spectrometry (ICP-AES, HK8100, Huakeyitong Co., Ltd, Beijing, China) and inductively-coupled plasma-mass spectrometry (ICP-MS, Agilent 7900, Beijing, China). The former was used for concentrations greater than 1 mg/L, and the latter was used for concentrations less than 1 mg/L. Anion analysis was performed using ion chromatography (IC, ICS1000, Dionex Co., Ltd, Sunnyvale, USA). All samples were filtered (Whatman No. 1) before the ionic analysis to avoid any interference by the deposit. The data are presented in Table S1 (available with the online version of this paper).
Rice straw was collected from the agricultural lands in Jingzhou City in the south of China. The agricultural wastes were minced in an electric mill and then washed three or five times with deionised distilled water. The RSAC was prepared using zinc chloride (purity !98%), as described by

Experimental setup
As shown in Figure 1, the EC/RSAC adsorption process was conducted in a 720 mL EC cell (12 × 10 × 6 cm) using an organic glass reactor. Aluminum (purity !99%) was used as sacrificial electrodes, and stainless steel was used for the cathode. The effective area of the each electrode plate was 20 cm 2 . Parallel anodes and cathodes were positioned vertically, and separated at a certain distance. The electrodes were connected to a digital direct current (DC) power supply (WKY-505, East Co., Ltd, Guangdong, China) with voltage and electrical current range of 0-50 V and 0-5 A, respectively. Before each run, the electrodes were rubbed with fine-grained emery paper, washed with 0.2 M HCl and then with distilled water. For every experimental run, 600 mL of mine waters were placed into the EC reactor.
The solutions were stirred constantly (200 rpm) using a magnetic stirrer. The current density, electrode gap, initial pH and reaction time were set to a desired value. The pH was adjusted by adding 0.1 M HCl or 0.1 M NaOH and measured using a HQ11D pH meter (Hach Co., Ltd, USA). All the runs were performed at room temperature (20 C). The sulfate removal efficiency is defined as follows: where C 0 and C t are sulfate concentrations (mg/L) at time 0 and time t (min), respectively.

Experimental design and model development
The statistical software Design Expert 8.0 (Stat-Ease Inc., Minneapolis, USA) was applied for the experimental design, analysis and optimization. Central composite design (CCD) was employed to investigate and optimize the experimental variables in the sulfate removal from mine waters. Three independent factors, the current density (X 1 ), dosage of RSAC (X 2 ) and reaction time (X 3 ), were studied at five levels with five repetitions at the central point using circumscribed CCD.
CCD consists of 2 3 factorial designs augmented by six axial points coded as ±α and three central points (all variables at zero level). The value of α was calculated as follows: where n is the number of variables in CCD. Therefore, α is equal to 2 3/4 ¼ 1.682 according to Equation (7). Each parameter was coded at three levels, -1.681 (minimum), 0 (central), and þ1.681 (maximum), which covered the entire study range. The range and levels of the three variables were measured with the equation: where X i stands for the dimensionless value of the ith test variable, x i denotes the real value of the ith test variable, x ic is the real value of the ith independent variable at the center point and Δx i is the step size. Table 1 shows the CCD matrix.
The experimental data were analysed using the response surface regression equation to fit the following quadratic model: where Y is the response; β 0 is a constant coefficient; ε is the error; and β i , β ii and β ij are interaction coefficients of linear, quadratic and second-order terms, respectively. After fitting the model, the generated data were used for 3D response surface optimization.

RESULTS AND DISCUSSION
Characterization of RSAC Figure S1 (available with the online version of this paper) shows the FTIR spectrum of raw rice straw and activated carbon prepared from rice straw. The broad adsorption band at around 3,419.12 cm À1 can be attributed to the stretching of H-bonded hydroxyl groups. The adsorption peak at 2,926.20 cm À1 is assigned to C-H stretching vibration. The peak at 1,632.86 cm À1 is due to the bending vibration mode of the adsorbed water molecules. In contrast to raw rice straw, a sharp adsorption peak of the stretching vibration of C-N bond is observed at 605.57 cm À1 in RSAC (Farahmand et al. ). Representative SEM images of the raw and activated rice straw are shown in Figure S2 (available online). The surface morphologies of the activated material are clearly irregular and porous, which could favour a high uptake of sulphate.

Effect of current density
The current density is a crucial parameter that affects not   Therefore, the optimum dosage was 0.4 g/L and was used for the rest of the single factor experiment.

Effect of initial pH
The effect of pH on the adsorption process is significant because pH affects the solubility of metal hydroxide and surface characteristic of the adsorbent material (Sundaramurthy et al. ; Zhu et al. a). Batch experiments were conducted for the EC/RSAC coupled adsorption process at pH 4, 5, 6.5, 7, 8 and 9. The results presented in Figure 4 show that the maximum percentage sulphate removal (90.5%) was at neutral conditions, that is, pH 7. The sulphate removal efficiency was also low either at low or high pH. This is because at low pH (<4), the primary form of aluminum's occurrence is Al 3þ , at high pH (>8), the primary form of aluminium hydroxide is Al(OH) 4 -. These products are not useful for adsorption.
( Figure S3, available online, shows that the Al(III) species distribution could be estimating under a wide variety of pH values using MINTEQ 3.0.) A decline in sulphate removal efficiency (from 90.5 to 89.0%) was observed with an increase in pH (from 7 to 9). The pH can also affect the surface charge of RSAC. The suppressed removal efficiency is partly due to the electrostatic repulsion between the negatively charged surface of RSAC (pH > pH pzc ) and sulphate ions. According to  the preceding analysis, the optimum pH for EC/RSAC coupled adsorption process was 7.  Table S2 (available online). The relationship between sulfate removal efficiency and the three independent variables (i.e. current density, dosage and reaction time) was fitted to the second-order polynomial equation as given below:
The suitability of the selected model for providing adequate estimations of real systems was also confirmed by normal probability plots of the studentised residuals and the predicted versus actual value plots, as shown in Figure S4(a) and S4(b), respectively (available online). It can be deduced from the plots that the data were evenly distributed. The model was tested using the determination coefficient (R 2 ).
An R 2 value close to 1 indicates that the model is strong and gives good predictions of sulphate removal efficiency.
The determination coefficient (R 2 ¼ 0.9253) showed that only 7.47% of response variability was not explained by the model. In addition, the value of the adjusted determination coefficient (R adj 2 ¼ 0.8836) was extremely high, thereby showing a high significance of the model. Moreover, the value of the predicted R 2 was high, thereby supporting the significance of the model. Thus, the predicted R 2 of 0.9075 for the model was in reasonable agreement with the adjusted R 2 of 0.8836. These results illustrate that the data prediction capability of the response surface model was satisfactory.

Interactive effects analysis and optimization
Contour plots of the mathematical regression model were described using Design Expert software to study the  The interaction between dosage and reaction time at the current density of 70 mA/cm 2 is illustrated in Figure 5(b).

Economic and technical analysis
The calculated cost of current, voltage and reaction time was considered in the assessment of the described EC/RSAC coupled system. In this study, current density and reaction time at the optimum operating conditions were 75 mA/cm 2 and 19.2 min, respectively. For this current density, current