Wastewater produced from polymer flooding in oil production features high viscosity and chemical oxygen demand because of the residue of high-concentration polymer hydrolysed polyacrylamide (HPAM). In this study, steel slag, a waste from steel manufacturing, was studied as a low-cost adsorbent for HPAM in wastewater. Optimisation of HPAM adsorption by steel slag was performed with a central composite design under response surface methodology (RSM). Results showed that the maximum removal efficiency of 89.31% was obtained at an adsorbent dosage of 105.2 g/L, contact time of 95.4 min and pH of 5.6. These data were strongly correlated with the experimental values of the RSM model. Single and interactive effect analysis showed that HPAM removal efficiency increased with increasing adsorbent dosage and contact time. Efficiency increased when pH was increased from 2.6 to 5.6 and subsequently decreased from 5.6 to 9.3. It was observed that removal efficiency significantly increased (from 0% to 86.1%) at the initial stage (from 0 min to 60 min) and increased gradually after 60 min with an adsorbent dosage of 105.2 g/L, pH of 5.6. The adsorption kinetics was well correlated with the pseudo-second-order equation. Removal of HPAM from the studied water samples indicated that steel slag can be utilised for the pre-treatment of polymer-flooding wastewater.

INTRODUCTION

Given that many oilfields are in their middle or final stage of production, methods to enhance oil recovery (EOR) have been developed (Chen et al. 2015). As one of the EOR techniques, polymer flooding has been successfully applied in full-industrial scale around the world, particularly in China, because of its good viscous properties (Zhang et al. 2015). By adding polymers, especially polyacrylamide (PAM), into flooding water, water mobility decreases and oil recovery increases. After polymer flooding, hydrolysed polyacrylamide (HPAM) remains in the produced water, which is generated after separation from oil using three-phase separators (Chen et al. 2015). A part of the produced water is injected back underground for reuse, and the rest is discharged into the surrounding ecosystems. Considering that HPAM can remain in surface water and groundwater for a long time, it may endanger the health of local residents. In addition, the amount of HPAM residue in the fluid contributes to the total chemical oxygen demand of the produced water; therefore, achieving the discharge standard is more difficult than achieving it with that produced from water flooding (Bao et al. 2010; Zhu et al. 2016). Consequently, effective treatment of effluents has become urgent and important. Several traditional methods, such as flotation, gravitational settling and membrane separation, have proved to be ineffective in treating produced water from polymer flooding compared with treating water flooding (Wang et al. 2006; Gao et al. 2011). However, installing new equipment is expensive and inconvenient. With these considerations, the adsorption technique was introduced. Various adsorbents have been tested for PAM removal from aqueous solutions. These adsorbents include magnesium hydroxide particles, oxide minerals, gold and silver (Lee & Somasundaran 1989; And & Frank 2004; Li 2016). The application of these materials is usually limited because of their high cost and pre-treatment requirement. As a result, other adsorbents, such as inexpensive materials from industry waste with a simple pre-treatment step, are in demand.

Steel slag is a high-volume waste product or by-product discharged from steel manufacturing (Jian et al. 2015). Given that the production of steel has increased, the volume of slag produced has exceeded the demand for recycling. The production of steel slag in China in 2010 was estimated to be nearly 696 million tonnes (Zhang et al. 2011). A large amount of steel slag is merely disposed of in landfills or water bodies, so extra costs for disposal and environmental pollution problems have been increasing (Qiu et al. 2015). In recent years, several attempts have been made to use granulated steel slag as an adsorbent to replace high-cost materials, such as activated carbons. Steel slag has been utilised in wastewater treatment to remove different types of pollutants from water, such as phosphate, cadmium, copper, lead ions and ammonia–nitrogen (Liu et al. 2009; Jian et al. 2015; Qiu et al. 2015; Francisca & Glatstein 2016). According to an early laboratory testing conducted by Lu et al. (Lu et al. 2013), steel slag can be used to adsorb HPAM in a simulated water sample. However, no thorough studies have investigated parameter optimisation and the mechanism of the adsorption process. Considering the potential applicability of this technology in environmental industries, a systematic study of HPAM adsorption by steel slag needs to be performed.

Many factors, such as dosage of the adsorbent, contact time and pH, can influence the efficiency of the adsorption process. Optimisation of these factors can significantly increase process efficiency. In most previous studies, optimisation was usually carried out by changing one independent variable whilst keeping the other factors fixed. This procedure is time and energy consuming. To overcome this problem, response surface methodology (RSM) based on central composite design (CCD) was employed in the current study to optimise the adsorption of HPAM in an aqueous solution onto steel slag in a batch experiment.

The main objective of this study is to investigate the removal of HPAM from an aqueous solution using steel slag. In the experiment, water-cooled slag was applied as an adsorbent in wastewater containing HPAM. The interactive effect of the process variables on removal efficiency was analysed with CCD–RSM. The maximum response value and its corresponding operation parameters were obtained from the mathematical model. In addition, kinetic study of HPAM onto steel slag under optimum conditions was performed.

MATERIALS AND METHODS

Materials

The solvent used in the experiment was deionised water. A HPAM solution of 300 ppm was prepared by dissolving 0.30 g of PAM in 1,000 mL of water. PAM with an average molecular weight of 12 million and hydrolysis degree of 27% was supplied by Huabei Oilfield Company in China. By using partially hydrolyzed polymer with higher molecular weight, water mobility decreases, sweep efficiency increases, thereby oil recovery increases (Maghzi et al. 2014; Zhou et al. 2015). The structural formula of PAM is shown in Figure 1. Steel slag was obtained from a steelmaking laboratory in the University of Science and Technology, Beijing. The chemical composition (expressed as oxides) of the slag used in this study is listed in Table 1. All of the other reagents were of analytical grade and used without further purification. The chemical compositions of steel slag were determined through X-ray fluorescence spectroscopy (72000S, ARL, Switzerland). The specific surface areas, calculated with the Brunauer–Emmett–Teller (BET) equation, were determined from N2 adsorption isotherms measured with a Beckman Coulter SA 3100 (USA). Scanning electron microscopy (SEM) was performed with an SU8010 microscope (Hitachi, Japan) operated at 3.0 kv.
Table 1

Physico-chemical characteristics of steel slag

General chemical composition (%)
 
Specific surface area (cm2/g) 
Fe2O3 FeO CaO MgO SiO2 MnO Al2O3 
19.54 22.15 37.72 4.03 13.37 1.95 1.24 750 
General chemical composition (%)
 
Specific surface area (cm2/g) 
Fe2O3 FeO CaO MgO SiO2 MnO Al2O3 
19.54 22.15 37.72 4.03 13.37 1.95 1.24 750 
Figure 1

Structural formula of PAM. R: hydrophobic side group, carbon number: 16–20

Figure 1

Structural formula of PAM. R: hydrophobic side group, carbon number: 16–20

The pHpzc (point of zero charge, PZC) is an important characteristic that determines the surface charges of a solid particle (Kosmulski 2002). As described in a previous study, 0.1 g of steel slag was added to 1,000 mL distilled water as a background electrolyte (Mercado-Borrayo et al. 2013). The PZC of the steel slag was evaluated by measuring the zeta potential at the pH range of 2–12. The pH of the solutions was adjusted with 0.01 M of HCl or NaOH. The parameter was calculated with a zeta potential analyser (Zetasizer Nano Z, Malvern Instrument Co., Ltd, UK). As shown in Figure 2, the pHpzc of steel slag is 6.4.
Figure 2

Zeta potential of the steel slag as a function of pH.

Figure 2

Zeta potential of the steel slag as a function of pH.

Experimental procedure

A series of experiments were conducted to study the HPAM adsorption process on steel slag. At a room temperature of 298 K, steel slag was added to a 200 mL conical flask with 100 mL of HPAM solution at a concentration of 300 mg/L. The flask contained steel slag with a particle size of 3–5 mm (the particle size range was optimised based on a previous study). The flasks were shaken at 250 rpm on a rotary shaker. Through a preliminary study, the data range for the operation parameters (e.g., adsorbent dosage, contact time and pH) was set. All experiments were conducted in triplicate, and the data were presented with the average value. Meanwhile, the solution pH was carefully adjusted with 0.01 M of NaOH or HCl solution when necessary.

Analytical methods

The HPAM concentration in the solution sample was determined with the starch–cadmium iodide method, which was mentioned in a previous study (Lu & Wu 2002). HPAM was measured at 585 nm with an ultraviolet spectrophotometer (UV2450, Shimadzu, Japan). The detection limit of the method is 0.14 mg/L, linear range is 0.2–40 mg/L (Wang et al. 2010). HPAM removal efficiency was calculated according to the differences in the concentration values, which were quantified before and after the adsorption process.

Experimental design

The statistical software Design Expert 8.0 (Stat-Ease Inc., Minneapolis, USA) was applied for the experiment design, analysis and optimisation. CCD was employed to investigate and optimise the experimental variables in the removal of polymer from aqueous solutions. Three independent factors, namely, the dosage of steel slag (X1), contact time (X2) and pH (X3), were studied at five levels with five repetitions at the central point using a circumscribed CCD.

The design matrix requires a 23 factorial design augmented by six axial points coded as ±α and three central points (all variables at zero level). The value of α was calculated with the equation: 
formula
1
where n is the number of variables in CCD. Therefore, α is equal to 23/4 = 1.682 according to Equation (1) (Chou et al. 2010). For each of the three variables studied, high (coded value: +1.681) and low (coded value: −1.681) set points were selected, as shown in Table 2. The range and levels of the three variables were measured with the equation: 
formula
2
where Xi stands for the coded value of the ith test variable, xi denotes the uncoded value of the ith test variable, xic is the uncoded value of the ith independent variable at the center point and Δxi is the step size.
Table 2

Central composite design

Variable name Range and coded factor level
 
−1.681 (low) −1 +1 +1.681 (high) 
Dosage, X1, (g/L) 84 90 100 110 116 
Contact time, X2, (min) 46.5 60 80 100 113.5 
pH, X3 2.6 9.3 
Variable name Range and coded factor level
 
−1.681 (low) −1 +1 +1.681 (high) 
Dosage, X1, (g/L) 84 90 100 110 116 
Contact time, X2, (min) 46.5 60 80 100 113.5 
pH, X3 2.6 9.3 
The experimental data were analysed using the response surface regression equation to fit the following quadratic model: 
formula
3
where Y is the response; β0 is a constant coefficient; ɛ is the residual value; and βi, βii and βij are the linear, quadratic and binary interaction coefficients, respectively. xi and xj are the coded independent variables. The subscripts i and j are the integer variables. The quality of the fit and the significance of the model were checked with the coefficient of determination (R2) and Fisher's F-test, which were performed by analysis of variance (ANOVA) with 95% confidence level.

Kinetic experiment

Adsorption kinetics experiments were performed under optimum conditions. At fixed time intervals, samples were obtained from the solution and analysed. The HPAM adsorption capacity, Qe (mg/g), was calculated with the following mass balance equation: 
formula
4
where C0 and Ce are the initial and equilibrium concentrations, respectively (mg/L), v is the volume of the HPAM solution (L) and m is the quality of steel slag (g).

To define the adsorption kinetics of HPAM on steel slag, kinetic data were analysed by using pseudo-first-order and pseudo-second-order models at optimum conditions. The linear form of these models are represented as:

Pseudo-first order kinetic model: 
formula
5
Pseudo-second order kinetic model: 
formula
6
where Qe and Qt (mg/g) denote the adsorption capacity at equilibrium and at time t, respectively. k1 (min−1) and k2 (g·mg−1·min−1) are pseudo-first and pseudo-second order rate constants, respectively.

RESULTS AND DISCUSSION

Model fitting and statistical analysis

The relationships between HPAM removal efficiency and the three independent variables (adsorbent dosage, contact time and pH) were studied. The experimental design shown in Table 3 also provides the removal efficiency for each experimental run. By applying multiple regression analysis on the experimental data, the obtained response value and the coded factors were related by the second-order polynomial equation as follows: the positive sign in front of the parameters indicates a synergistic effect, and the negative sign suggests an antagonistic effect. 
formula
7
The response surface model results from CCD are F-values and p-values along with the constant and coefficients shown in Table 4. ANOVA evaluations of this model imply that this model can describe the experiments. The parameters, such as p-value, F-value, R2 and adjusted R2, were determined to measure the effectiveness of the model (Bezerra et al. 2008). The F-value of 8.81 shown in Table 3 implies that the quadratic model is significant. The possibility that a ‘model F-value’ this large could occur because of noise is only 0.11%. ‘Prob > F’ less than 0.05 indicates that the model terms are significant. In this case, the model parameters (X1 and X2) and the interaction terms (X12, X22 and X32) are significant based on ANOVA.
Table 3

Experimental design results for the removal efficiency

Run number Dosage (X1Contact time (X2pH (X3Removal efficiency (Y, %) 
−1 −1 40.5 
84.0 
1.681 83.8 
83.1 
83.5 
1.681 78.5 
−1 −1 −1 34.6 
1.681 83.8 
−1 −1 65.2 
10 84.4 
11 −1.681 42.7 
12 −1.681 72.7 
13 −1 67.2 
14 −1 51.2 
15 78.2 
16 −1 75.0 
17 84.2 
18 −1 −1 43.5 
19 83.8 
20 −1.681 67.0 
Run number Dosage (X1Contact time (X2pH (X3Removal efficiency (Y, %) 
−1 −1 40.5 
84.0 
1.681 83.8 
83.1 
83.5 
1.681 78.5 
−1 −1 −1 34.6 
1.681 83.8 
−1 −1 65.2 
10 84.4 
11 −1.681 42.7 
12 −1.681 72.7 
13 −1 67.2 
14 −1 51.2 
15 78.2 
16 −1 75.0 
17 84.2 
18 −1 −1 43.5 
19 83.8 
20 −1.681 67.0 
Table 4

ANOVA of the model equations

Source Sum of squares df Mean square F value p-value Prob > F 
Model 4,990.96 554.55 8.81 0.0011 
X1 345.13 345.13 5.49 0.0412 
X2 2,503.95 2,503.95 39.79 <0.0001 
X3 59.70 59.70 0.95 0.3530 
X1X2 0.18 0.18 0.0028 0.9584 
X1X3 1.12 1.12 0.018 0.8963 
X2X3 8.82 8.82 0.14 0.7159 
X12 484.06 484.06 7.69 0.0197 
X22 1,467.55 1,467.55 23.32 0.0007 
X32 472.32 472.32 7.51 0.0208 
Residual 629.22 10 62.92   
Lack of fit 622.35 124.47 90.61 <0.0001 
Pure error 6.87 1.37   
Cor total 5,620.18 19    
Source Sum of squares df Mean square F value p-value Prob > F 
Model 4,990.96 554.55 8.81 0.0011 
X1 345.13 345.13 5.49 0.0412 
X2 2,503.95 2,503.95 39.79 <0.0001 
X3 59.70 59.70 0.95 0.3530 
X1X2 0.18 0.18 0.0028 0.9584 
X1X3 1.12 1.12 0.018 0.8963 
X2X3 8.82 8.82 0.14 0.7159 
X12 484.06 484.06 7.69 0.0197 
X22 1,467.55 1,467.55 23.32 0.0007 
X32 472.32 472.32 7.51 0.0208 
Residual 629.22 10 62.92   
Lack of fit 622.35 124.47 90.61 <0.0001 
Pure error 6.87 1.37   
Cor total 5,620.18 19    

R2 = 0.9492, = 0.9075.

Figure 3(a) shows a plot revealing the normal probability of the Student's t-test residuals for HPAM removal efficiency. The plot is an important diagnostic tool to check model adequacy. Most of the obtained data points follow a straight line, suggesting that the data were normally distributed in the responses of the model. The coefficient of determination (R2) is defined as the ratio of the predicted variation. The high R2 value of 0.9492 showed a small difference between the predicted and actual values of the response. The value of the adjusted determination coefficient (Radj2 = 0.8975) in the regression model was also close to 1, indicating that the experimental values can be significantly predicted by the model. Moreover, the predicted versus actual value plot of the response surface model presented in Figure 3(b) revealed excellent agreement between the actual and predicted data obtained from the model. Therefore, the results indicate that the second-order polynomial model is statistically valid.
Figure 3

Normal probability plot of the internally studentised residuals (a) and predicted vs. actual values plot (b) in the response surface model.

Figure 3

Normal probability plot of the internally studentised residuals (a) and predicted vs. actual values plot (b) in the response surface model.

Response surface plots and interaction effect analysis

2D response surface contour plots were derived based on the mathematical regression model to obtain new insights into the interaction effect between the independent variables on the HPAM adsorption process.

Figure 4(a) shows a contour plot drawn by maintaining pH (X3) at 6 (the centre level) whilst the two other factors (dosage of the steel slag, X1; contact time, X2) were varied in their range. At 45.6 min, removal efficiency increased by 16.43% (from 8.83% to 25.26%) as the dosage increased from 84 g/L to 116 g/L; at 113.5 min, removal efficiency increased by 17.68% (from 53.25% to 70.93%) when the dosage was increased from 84 g/L to 116 g/L. This trend can be explained by the fact that increased adsorbent dosage means a larger active surface area and more adsorption sites (Hadi et al. 2015).
Figure 4

Response surface plots for interactive effect among factors ((a)–(c): adsorbent dosage and contact time; dosage and pH; contact time and pH) on HPAM removal efficiency.

Figure 4

Response surface plots for interactive effect among factors ((a)–(c): adsorbent dosage and contact time; dosage and pH; contact time and pH) on HPAM removal efficiency.

Figure 4(b) describes the interaction between dosage and pH with a contact time of 80 min. Increasing the pH led to an increase in removal efficiency in the investigated dosage, and this effect was highly significant at a pH of 9.3. At a pH of 2.6, removal efficiency increased by 14.55% (from 41% to 55.55%) as the dosage increased from 84 g/L to 116 g/L; at a pH of 9.3, removal efficiency increased by 19.24% (from 45.51% to 64.75%) when the dosage increased from 84 g/L to 116 g/L. At the optimum dosage of 105 g/L, increasing the pH from 2.6 to 6.2 resulted in increased HPAM removal efficiency from 70.8% to 89.2%. However, with a further increase in pH up to 9.3, HPAM removal efficiency decreased (73.5%). This phenomenon can be explained by the fact that pH affected the surface charge of the steel slag. As demonstrated in the plot of pHpzc (Figure 2), positive surface charges associated with positive zeta potentials were determined at pH < 6.4. Under this condition, the adsorption of negatively charged HPAM on the slag particle became more favourable because of the interaction of the adsorbent surface and pollutant. By contrast, above the PZC, the surface of the steel slag attained a negative charge. Low adsorption was observed because of the enhanced electrostatic repulsion between HPAM and the surface of the adsorbent.

Figure 4(c) illustrates the individual/combined effects of contact time and pH when the dosage was at zero level (100 g/L). The plot shows that at 45.6 min, removal efficiency increased by 12.65% (from 11.3% to 23.95%) when the pH increased from 2.6 to 9.3; at 113.5 min, removal efficiency decreased by 5.36% (from 68.29% to 62.93%) when the pH ranged from 2.6 to 9.3. The maximum removal efficiency of 89.31% was obtained at a pH of 5.6 and contact time of 95.4 min. This result may be caused by the fact that vacant sites on the surface of the adsorbent were abundantly available at the beginning of contact time. With a further increase in time, the adsorption became slow because of the decrease in the number of active sites. A similar phenomenon was also observed in study of phenol and p-chlorophenol removal by magnetic recoverable carbon and Hg2+ removal by alizarin red-S-loaded amberlite IRA-400 resin (Naushad et al. 2016). Comparison of the two independent variables showed that contact time exerted a more significant influence on the response surface plot than pH. This observation is agreement with the results obtained in ANOVA (Table 4), which showed a larger F value of contact time, X2 (F value = 39.79), compared with pH, X3 (F value = 0.95).

Optimisation and validation experiment

The desirable point prediction function in the experimental design was applied to identify the optimum conditions to maximise HPAM removal efficiency within the studied experimental ranges. At the laboratory scale, the maximum removal efficiency of 89.31% was obtained at an adsorbent dosage of 105.2 g/L, contact time of 95.4 min and pH of 5.6. Future research on the optimisation of the method should be conducted on an amplification test that considers adsorption material consumption. To test the validity of the model, experiments were conducted under the optimum conditions and the five other predicted conditions. HPAM removal efficiency is described in Table 5. The experimental results are close to the predicted values but with a lower percentage of error (−0.12, −0.78, −1.04, 0.38, −1.11 and −0.70).

Table 5

Verification test of the fitting model

Test number Operation parameters
 
Response value (Y)
 
X1 X2 X3 Observed value (%) Predicted value (%) % Error 
0.82 0.14 0.21 85.71 86.38 −0.78 
0.13 0.82 0.15 88.24 89.17 −1.04 
0.45 0.49 −0.41 88.2 87.87 0.38 
−0.10 0.13 0.09 84.52 85.47 −1.11 
0.97 0.07 0.26 84.12 84.71 −0.70 
Optimum 0.52 0.77 −0.18 89.2 89.31 −0.12 
Test number Operation parameters
 
Response value (Y)
 
X1 X2 X3 Observed value (%) Predicted value (%) % Error 
0.82 0.14 0.21 85.71 86.38 −0.78 
0.13 0.82 0.15 88.24 89.17 −1.04 
0.45 0.49 −0.41 88.2 87.87 0.38 
−0.10 0.13 0.09 84.52 85.47 −1.11 
0.97 0.07 0.26 84.12 84.71 −0.70 
Optimum 0.52 0.77 −0.18 89.2 89.31 −0.12 

Adsorption kinetics

Figure 5 shows the kinetic experiment of HPAM adsorption on steel slag with an adsorbent dosage of 105.2 g/L, and pH of 5.6. The removal efficiency of HPAM significantly increased (from 0% to 86.1%) with increasing contact time from 0 min to 60 min. HPAM removal efficiency was slightly enhanced after 60 min. After 100 min, the removal efficiency declined (from 86.2% to 85.7%) due to the limitation of the adsorption sites on the surface of the steel slag. The pseudo-first-order and pseudo-second-order models were employed to simulate the adsorption kinetics of HPAM on steel slag. The fitted parameters of the two models are summarised in Table 6. In terms of the correlation coefficients (R2), the pseudo-second-order model (R2 = 0.999) describes the adsorption kinetics better than the pseudo-first-order model (R2 = 0.949). Moreover, the calculated values of Qe,cal2 (0.31 mg/g) related to the pseudo-second-order model agree very well with the experimental data (Qe,exp = 0.27 mg/g). The fitted kinetic model suggests that the adsorption of HPAM may be due to the electrostatic attraction between the charged surface of the steel slag and HPAM. This finding was supported by the zeta analyses mentioned above. A similar phenomenon was also observed previously in the case of adsorption of crystal violet dye onto a surfactant-modified magnetic nanoadsorbent (Muthukumaran & Thirumarimurugan 2016).
Table 6

Pseudo-first-order and pseudo-second-order adsorption model parameters for HPAM adsorption by steel slag (adsorption condition: initial HPAM concentration of 300 mg/L, adsorbent dosage of 105.2 g/L, pH of 5.6)

Pseudo-first-order
 
Pseudo-second-order
 
Experimental value
 
Qe,cal1 (mg/g) k1 (min−1R2 Qe,cal2 (mg/g) k2 (g(mg min)−1R2 Qe,exp (mg/g) 
0.56 0.0252 0.949 0.31 0.2189 0.999 0.27 
Pseudo-first-order
 
Pseudo-second-order
 
Experimental value
 
Qe,cal1 (mg/g) k1 (min−1R2 Qe,cal2 (mg/g) k2 (g(mg min)−1R2 Qe,exp (mg/g) 
0.56 0.0252 0.949 0.31 0.2189 0.999 0.27 
Figure 5

Kinetic study on the HPAM removal efficiency with the adsorbent dosage of 105.2 g/L, pH of 5.6, initial HPAM concentration of 300 mg/L.

Figure 5

Kinetic study on the HPAM removal efficiency with the adsorbent dosage of 105.2 g/L, pH of 5.6, initial HPAM concentration of 300 mg/L.

SEM analysis

The morphology of steel slag before and after HPAM adsorption under optimum conditions was observed through SEM. Figure 6(a) and 6(b) show a significant difference between the surfaces of the two samples. The adsorption effect of steel slag was proven by the micrograph. In plot (a), the steel slag surface is rough and contains many irregular microspores. Figure 6(b) shows that the surface of the steel slag is smooth. The solid particle is compact and dense with low porosity because of the coverage of HPAM after adsorption.
Figure 6

SEM of the steel slag (before (a) and after (b) adsorption of HPAM).

Figure 6

SEM of the steel slag (before (a) and after (b) adsorption of HPAM).

CONCLUSIONS

Optimisation of HPAM adsorption on steel slag was performed to maximise HPAM removal efficiency. A response surface method based on a three-variable, five-level central composite experiment design was successfully employed in this optimisation study. The effects of the independent variables, e.g., adsorbent dosage, contact time and initial pH, on HPAM removal efficiency were considered. Regression analysis revealed the good fit of the experimental data to a second-order polynomial model with a coefficient of determination (R2) value of 0.9492 and an F value of 8.81. The maximum removal efficiency of 89.31% was obtained under the following optimal parameters: adsorbent dosage of 105.2 g/L, contact time of 95.4 min and pH of 5.6. The ANOVA results showed that contact time (X2) possessed a low p-value (<0.0001) and a high F value (39.79), which implies that it was the most significant factor affecting HPAM removal efficiency. The adsorption kinetics were fast and followed a pseudo-second-order kinetics model. The results also revealed that steel slag has a promising application in the pre-treatment of polymer-flooding produced water. However, further studies are required to explore the behaviour of the function group on the steel slag surface involved in the adsorption process.

ACKNOWLEDGEMENTS

This work is supported in part by grants from the International Joint Key Project of the Chinese Ministry of Science and Technology (2010DFB23160) and the National Natural Science Foundation of China (41273131, 41273092 and 41430106), Public Welfare Project of Chinese Ministry of Environmental Protection (201509049).

REFERENCES

REFERENCES
Bezerra
M. A.
Santelli
R. E.
Oliveira
E. P.
Villar
L. S.
Escaleira
L. A.
2008
Response surface methodology (RSM) as a tool for optimization in analytical chemistry
.
Talanta
76
(
5
),
965
977
.
Chen
H. X.
Tang
H. M.
Gong
X. P.
Wang
J. J.
Liu
Y. G.
Duan
M.
Zhao
F.
2015
Effect of partially hydrolyzed polyacrylamide on emulsification stability of wastewater produced from polymer flooding
.
J. Pet. Sci. Eng.
133
,
431
439
.
Francisca
F. M.
Glatstein
D. A.
2016
Influence of pH on cadmium, copper, and lead removal from wastewater by steel slag
.
Desal. Water Treat.
57
(
45
),
21610
21618
.
Hadi
P.
To
M. H.
Hui
C. W.
Lin
C. S. K.
McKay
G.
2015
Aqueous mercury adsorption by activated carbons
.
Water Res.
73
,
37
55
.
Jian
Y.
Liang
W.
Li
W.
Li
F.
Zou
Y.
Wang
H.
2015
Phosphate removal from domestic wastewater using thermally modified steel slag
.
J. Environ. Sci.
31
(
5
),
81
88
.
Kosmulski
M.
2002
The pH-dependent surface charging and the points of zero charge
.
J. Colloid Interf. Sci.
253
(
1
),
77
87
.
Lee
L. T.
Somasundaran
P.
1989
Adsorption of polyacrylamide on oxide minerals
.
Langmuir
5
(
3
),
854
860
.
Liu
S. Y.
Jin
G.
Yang
Y. J.
Yang
Y. C.
Ye
Z. X.
2009
Adsorption intrinsic kinetics and isotherms of lead ions on steel slag
.
J. Hazard. Mater.
173
(
1–3
),
558
562
.
Lu
L.
Lu
X.
Zhao
D.
Liu
Q.
2013
Advanced treatment of polymer-flooding wastewater using steel slag
.
Chin. J. Environ. Eng.
7
(
11
),
4355
4360
(in Chinese)
.
Mercado-Borrayo
B. M.
Schouwenaars
R.
González-Chávez
J. L.
Ramírez-Zamora
R. M.
2013
Multi-analytical assessment of iron and steel slag characteristics to estimate the removal of metalloids from contaminated water
.
J. Environ. Sci. Heal. A.
48
(
8
),
887
895
.
Qiu
R.
Cheng
F.
Wang
X.
Li
J.
Gao
R.
2015
Adsorption kinetics and isotherms of ammonia-nitrogen on steel slag
.
Desal. Water Treat.
55
(
1
),
142
150
.
Wang
B.
Chen
Y.
Liu
S.
Wu
H.
Song
H.
2006
Photocatalytical visbreaking of wastewater produced from polymer flooding in oilfields
.
Colloids Surf. A.
287
(
1–3
),
170
174
.
Wang
W.
Zhang
X.
Yang
Z.
Wang
X.
Ding
Z.
Li
Y.
2010
Determination of polyacrylamide applied with Starch-Cadmium Iodide method for low oil polluted water
.
Water Purif. Technol.
29
(
5
),
33
35
(in Chinese)
.
Zhou
Y.
Li
W.
Wan
W.
Zhang
R.
Lin
Y.
2015
W/Mo co-doped BiVO4 for photocatalytic treatment of polymer-containing wastewater in oilfield
.
Superlattices Microstructures
82
,
67
74
.