This study applied magnetic nanoparticles (MNPs) to chemical mechanical polishing wastewater treatment using experimental design (Plackett–Burman methods) to select the key factors among pH, mixing, Polyaluminum chloride dosage, settling time, MNPs dosage, and temperature and using response surface methodology (RSM) to determine the optimal values of key factors. Research results showed that the key factors influencing processing performance were pH and rpm, and the optimal conditions were a pH of 4.9 and rpm of 68. The turbidity removal rate through RSM simulation was 90%; under this parameter, the actual turbidity removal rate in the experiment was 89%, which was extremely close to the simulation value; this value was also much higher than the nonoptimized removal rate of 61 ± 8%. Additionally, in the subsequent regeneration and reuse experiment involving mixing and ultrasound for desorption and regeneration, the number of recoveries were 4 and 5, respectively. The study showed that the average particle size of MNPs following ultrasonic vibration was reduced; the effect was optimal at 23 to 15 nm. Therefore, a removal rate of over 80% could be maintained for the fifth ultrasonic regeneration, and the energy of the mixing method may not have been sufficient, causing incomplete desorption and a turbidity removal rate of only 71%.

INTRODUCTION

In the semiconductor industry, the use of chemical mechanical polishing (CMP) technology generates large amounts of wastewater with nanoscale aerosols (particles ranging from 20 to 120 nm in size) (Roth et al. 2015). This wastewater has high turbidity, rendering treatment difficult. Most of the semiconductor industry uses traditional chemical coagulation to treat wastewater (Browne et al. 1999; Chou et al. 2009; Wan et al. 2011). The chemical coagulation generates large amounts of sludge, and because the change in turbidity in CMP wastewater is greater, the coagulant dosage is difficult to control (Chin et al. 2006; Kuan & Hu 2009; Wan et al. 2011), resulting in high processing costs.

Numerous scholars have used various methods to treat the wastewater resulting from CMP and backside grinding (BG) in an attempt to reduce processing costs. Heavy metal methods for treating wastewater have included activated carbon adsorption, electrocoagulation, membrane filtration, ion-exchange resin, ultrafiltration, and nanobubble flotation technology (Browne et al. 1999; Lai & Lin 2003, 2004; Huang et al. 2004; Den & Huang 2005; Chin et al. 2006; Tsai et al. 2007; Chou et al. 2009; 2010; Kuan & Hu 2009; Wang et al. 2009; Roth et al. 2015; Liu et al. 2016). These techniques have been either ineffective or costly (Lin & Yang 2004; Yang & Tsai 2006).

Because of their high adsorption ability, easy separation, low cost, high stability, and nontoxicity, magnetic nanoparticles (MNPs) have often been used in wastewater treatment (Li et al. 2006; Wan et al. 2011). For example, the use of MNP for the adsorption of BG wastewater treatment has produced effective results (Chin et al. 2006; Wan et al. 2011; Lai et al. 2013).

Experimental design is one of the most accurate methods for conducting experiments. It enables various interactions between influencing factors, obtaining optimized experimental designs such as response surface methodology (RSM), Taguchi design of experiments, and full or fractional factorial design (Bezerra et al. 2008). RSM has been statistically proven to be an optimal tool for statistical analysis (Micic et al. 2015). Numerous optimization designs for experiments have used the RSM method, because central composite design (CCD) has limited effective factors (Bezerra et al. 2008). RSM has been successfully used in MNPs and other areas of research (Ghazanfari et al. 2016).

This study applied MNP to CMP wastewater treatment using experimental design, specifically Plackett–Burman (PB) methods, to select the key factors among pH, mixing, polyaluminum chloride dosage, settling time, MNP dosage, and temperature and using RSM to determine the optimal conditions of key factors.

MATERIALS AND METHODS

Materials

This study used CMP wastewater from a semiconductor plant in Taiwan. The basic wastewater quality parameters are as shown in Table 1; the color of wastewater with pH values ranging between 8.4 and 8.9 was milky white. The turbidity of the wastewater ranged from 190 to 400 NTU. The main components were the anti-settling Si nanoparticles and other heavy metal ions. The average silica size was approximately 180 nm, and zeta potential ranged from −21 to −44 mV, relatively close to that of other studies (20 to 300 nm, −10 to −80 mV), as shown in Table 2. Inductively coupled plasma with atomic emission spectroscopy was used to measure the concentrations of the main components of CMP wastewater. The primary component of wastewater was Si, with a concentration of approximately 118 mg/L, followed by Ca, Al, and other trace metals.

Table 1

Water quality characteristics of CMP wastewater

Water quality characteristic Range 
pH 8.4–8.9 
Turbidity (NTU) 190–400 
zeta potential (mV) −21 to −44 
average particle sizes (nm) 180 
Suspended Solids, SS (mg/L) 180–230 
TDS (mg/L) 143–152 
COD (mg/L) 10–40 
Conductivity (μs/cm) 210–240 
Color Milky white 
Si(mg/L) 117.58 
Ca(mg/L) 3.68 
Mg(mg/L) 1.02 
Al(mg/L) 4.86 
Fe(mg/L) 0.81 
Zn(mg/L) 0.15 
Cu(mg/L) N.D 
Pb(mg/L) N.D 
Cr(mg/L) N.D 
Mn(mg/L) N.D 
Water quality characteristic Range 
pH 8.4–8.9 
Turbidity (NTU) 190–400 
zeta potential (mV) −21 to −44 
average particle sizes (nm) 180 
Suspended Solids, SS (mg/L) 180–230 
TDS (mg/L) 143–152 
COD (mg/L) 10–40 
Conductivity (μs/cm) 210–240 
Color Milky white 
Si(mg/L) 117.58 
Ca(mg/L) 3.68 
Mg(mg/L) 1.02 
Al(mg/L) 4.86 
Fe(mg/L) 0.81 
Zn(mg/L) 0.15 
Cu(mg/L) N.D 
Pb(mg/L) N.D 
Cr(mg/L) N.D 
Mn(mg/L) N.D 
Table 2

The particle sizes and zeta potential of CMP wastewater

Particle sizes (nm) Zeta potential (mV) Reference 
20–120 – Roth et al. (2015)  
55–300 (average 90) −10 to −65 Liu et al. (2016)  
148.6–166.7 −14.8 to −49.7 Wang et al. (2009)  
85–95 −28 to −35 Chou et al. (2009)  
20–80 – Browne et al. (1999)  
100 – Lai & Lin (2004)  
42 (Minimum particle size) −41.6 Kuan & Hu (2009)  
55–220 −45 to −55 Tsai et al. (2007
60 −47.6 Chin et al. (2006
90–130 −50 to −80 Den & Huang (2005
180 (average particle sizes) −21 to −44 This study 
Particle sizes (nm) Zeta potential (mV) Reference 
20–120 – Roth et al. (2015)  
55–300 (average 90) −10 to −65 Liu et al. (2016)  
148.6–166.7 −14.8 to −49.7 Wang et al. (2009)  
85–95 −28 to −35 Chou et al. (2009)  
20–80 – Browne et al. (1999)  
100 – Lai & Lin (2004)  
42 (Minimum particle size) −41.6 Kuan & Hu (2009)  
55–220 −45 to −55 Tsai et al. (2007
60 −47.6 Chin et al. (2006
90–130 −50 to −80 Den & Huang (2005
180 (average particle sizes) −21 to −44 This study 

Screening of key factors using the PB method

This study used the PB method because it could provide a large number of variables and effective identifying methods. In addition, it allowed the optimization of significant factors and disregarding of the interaction effects between variables. PB is a two-level (high-level, low-level) experimental design; it can rapidly screen key factors by using the least possible number of experiments (Deshmukh & Puranik 2010).

Performing RSM analysis

Based on the concept proposed by Box and Wilson, RSM analysis was used to optimize the two factors selected through the PB method, which was indicative of the interaction effects of the analysis variables (Jabeen et al. 2015). This method can reduce the experimental time, number of experiments, and experimental costs. More critically, it enabled an examination of the interaction effects of the factors. Therefore, it has been widely applied in various domains (Bezerra et al. 2008). RSM was applied in this batch of experiments; CCD tests were conducted on the lowest-combination experimental groups generated by the factors, and RSM was then used to predict the second-order regression (Kasiri et al. 2008; Zhang & Pan 2014).

Central composite design

The RSM experimental design can be applied in pseudo-second surfaces, particularly in the optimization process (Ahmadi et al. 2006). All factors in this design could be sorted into the categories of the central point, factor point, and pivot point. The experimental results of the CCD were analyzed using Statistica software, and the analysis result was treated as the description of the change in response surface (Bezerra et al. 2008). The concept of the CCD is shown in Figure 1.
Figure 1

Central composite design.

Figure 1

Central composite design.

MNP regeneration and reuse

The most suitable factors and conditions obtained from this selection method were applied in the regeneration and reuse experiment. The regeneration and reuse conditions of this experiment were a residual turbidity of 27 NTU and removal rate of ≤80%; the pH value was also adjusted. Using the two desorption methods of ultrasonic vibration and rapid mixing to perform desorption on MNP and CMP wastewater could improve MNP regeneration and reuse times.

RESULTS AND DISCUSSION

Zeta potential and characterization of MNP

The zeta potential of silica particles was negative in semiconductor industry wastewater. The charge carried by the surface of MNP differed according to the pH conditions; this variable was related to the stability, adsorption ability, and colloidal dispersion of the particles. The effects of pH on the zeta potentials in CMP wastewater and MNP zeta potential are shown in Figure 2. The diagram demonstrates that in an acidic environment of pH = 3–6, CMP wastewater and MNP carried dissimilar potentials, generating attraction; greater adsorption or attraction indicated a significant coagulation effect. In an alkaline environment of pH = 7–11, CMP wastewater and MNP carried similar potentials, generating repulsion; this property could be applied in the desorption of MNP and silica particles as well as the regeneration and reuse of magnetic particles.
Figure 2

Zeta potential of MNP and CMP wastewater.

Figure 2

Zeta potential of MNP and CMP wastewater.

PAC and magnetic particles in the absorption of CMP wastewater coagulation experiments

The researchers further made tests by setting the optimum conditions at pH 6, the magnetic particle dose was 4.99 g/L, and the settling time was 30 minutes. The treatment was actualized in clear form as shown in Figures 3 and 4. On the other hand, a considerable degree of turbidity removal rate can be reached with the use of magnetic particles at 4.99 g/L with polyaluminium chloride (PAC) 0.01–0.04 g/L in 30 minutes after the residual turbidity of about 9 NTU (Figure 5). Therefore, PAC 0.04 g/L is the best PAC dose of 4.99 g/L for magnetic particles. If the PAC dose is increased again, the effect of flocculation will be deteriorated, presumably because the pellet is stabilized.
Figure 3

The effect of the different pH values on the magnetic particles.

Figure 3

The effect of the different pH values on the magnetic particles.

Figure 4

Diagram of flocculation of the different doses of magnetic particles at pH 6.

Figure 4

Diagram of flocculation of the different doses of magnetic particles at pH 6.

Figure 5

Effect of magnetic particles at 4.99 g/L with different PAC on flocculation.

Figure 5

Effect of magnetic particles at 4.99 g/L with different PAC on flocculation.

Experimental design

Using the PB method to select key factors

The most suitable conditions from the chemical coagulation experiment were selected and applied in the PB method. After setting the factor levels, seven were designed and key factors were screened; the factors were set as shown in Table 3. The baseline of the factors selected from the coagulation experiment were set as Level 0, and low (−1) and high (+1) factors were then screened. The factor settings were entered into the Statistica software to conduct simulation. The results showed that the most critical factors were pH and rpm (Table 4). Using the Plackett-Burman method in Table 4, the two factors: pH 0.007 and rotation speed .012 were selected for the central mixed design, and the number of experimental groups was designed. Further, if the p value is less than 0.05, the effect of this factor on the experiment is significant. Otherwise, if the p value is more than 0.05, the effect of this factor on the experiment is not significant.

Table 3

Factors of the PB method and their level settings

Factors Levels −1 +1 
X1 pH 
X2 Rotation speed (rpm) 100 200 
X3 PAC(g/L) 0.02 0.04 0.06 
X4 Settling time (min) 20 30 40 
X5 MNP dosage (g/L) 3.74 4.99 6.23 
X6 Temperature (°C) 20 25 30 
X7 D1(nil factor) – – – 
Factors Levels −1 +1 
X1 pH 
X2 Rotation speed (rpm) 100 200 
X3 PAC(g/L) 0.02 0.04 0.06 
X4 Settling time (min) 20 30 40 
X5 MNP dosage (g/L) 3.74 4.99 6.23 
X6 Temperature (°C) 20 25 30 
X7 D1(nil factor) – – – 
Table 4

Simulation results of the PB method

Source Sum of squares dof Mean square F-value P-value 
pH 1,148.1 1,148.1 6,889.000 0.007 
Rotation speed 468.1 468.1 2,809.000 0.012 
PAC 52.9 52.9 317.400 0.035 
Settling time 59.2 59.2 355.267 0.033 
MNP dosage 8.1 8.1 48.600 0.090 
Temperature 96.0 96.0 576.000 0.026 
Error 0.1 0.1   
Overall sum of squares 2,042.8    
Source Sum of squares dof Mean square F-value P-value 
pH 1,148.1 1,148.1 6,889.000 0.007 
Rotation speed 468.1 468.1 2,809.000 0.012 
PAC 52.9 52.9 317.400 0.035 
Settling time 59.2 59.2 355.267 0.033 
MNP dosage 8.1 8.1 48.600 0.090 
Temperature 96.0 96.0 576.000 0.026 
Error 0.1 0.1   
Overall sum of squares 2,042.8    

Central composite design

Using optimal parameters, pH and rotation speed, obtained from the PB simulation results, this study designed a two-factor CCD involving three levels. The number of experimental groups is as shown in Table 5. Statistica software was then used to perform analysis of variance (ANOVA) and analysis of the regression coefficient, obtaining the equations for the response variable and operating factor. The optimal operating parameter was obtained through the generation of contour plots and 3D response surfaces, and the effect of each factor was subsequently determined.

Table 5

Number of CCD experimental groups

  pH Speed (rpm) Residual turbidity (NTU) Turbidity removal (%) 
Corner point 8 (1) 150 (1) 29.5 89.1 
8 (1) 50 (−1) 50.8 81.2 
4 (−1) 50 (−1) 30.1 88.9 
4 (−1) 150 (1) 20.5 92.4 
Central point 6 (0) 100 (0) 25.6 90.5 
6 (0) 100 (0) 25.5 90.6 
6 (0) 100 (0) 20.2 92.5 
6 (0) 100 (0) 27.9 89.7 
6 (0) 100 (0) 27.5 89.8 
6 (0) 100 (0) 28.4 89.5 
Pivot point 6 (0) 170 (1.4142) 4.15 98.5 
8.82 (1.4142) 100 (0) 54.6 80 
6 (0) 29 (−1.4142) 19.4 92.8 
3.17 (−1.4142) 100 (0) 23.5 91.3 
  pH Speed (rpm) Residual turbidity (NTU) Turbidity removal (%) 
Corner point 8 (1) 150 (1) 29.5 89.1 
8 (1) 50 (−1) 50.8 81.2 
4 (−1) 50 (−1) 30.1 88.9 
4 (−1) 150 (1) 20.5 92.4 
Central point 6 (0) 100 (0) 25.6 90.5 
6 (0) 100 (0) 25.5 90.6 
6 (0) 100 (0) 20.2 92.5 
6 (0) 100 (0) 27.9 89.7 
6 (0) 100 (0) 27.5 89.8 
6 (0) 100 (0) 28.4 89.5 
Pivot point 6 (0) 170 (1.4142) 4.15 98.5 
8.82 (1.4142) 100 (0) 54.6 80 
6 (0) 29 (−1.4142) 19.4 92.8 
3.17 (−1.4142) 100 (0) 23.5 91.3 

Selection of RSM most optimal parameter conditions

Using Statistica software for analysis, ANOVA was performed on residual turbidity, examining the lack of fit and p-value of each factor. The results, as shown in Table 6, showed the p-value of the lack of fit of turbidity removal to be ≥0.05, indicating turbidity removal passed the lack-of-fit test, and that its further discussion would be meaningful. The results also indicated that there was no human or mechanical error during the experiment. Next, the p-value of each factor was examined; those with p ≤ 0.05 were retained, indicating that the factor had a significant effect (for example, a pH of 0.007 and rpm of 0.012) By contrast, p ≥ 0.05 indicated was factor was insignificant and could be ignored. The results of the analysis of the regression coefficient are shown in Table 7; the regression equation is presented as Equation (1). 
formula
1
Note: pH(L) = pH, pH(Q) = pH2, RPM(L) = rpm, RPM(Q) = rpm2
Table 6

ANOVA table for turbidity removal

Factors Sum of squares dof Mean square F-value P-value 
pH(L) 91.44 91.43 74.78 0.00034 
pH (Q) 70.45 70.45 57.62 0.00063 
rpm (L) 47.98 47.97 39.23 0.00152 
rpm (Q) 27.59 27.58 22.56 0.0051 
1 L by 2 L 4.84 4.84 3.96 0.103 
Lack of fit 19.70 6.56 5.37 0.051 
Pure error 6.11 1.22   
Factors Sum of squares dof Mean square F-value P-value 
pH(L) 91.44 91.43 74.78 0.00034 
pH (Q) 70.45 70.45 57.62 0.00063 
rpm (L) 47.98 47.97 39.23 0.00152 
rpm (Q) 27.59 27.58 22.56 0.0051 
1 L by 2 L 4.84 4.84 3.96 0.103 
Lack of fit 19.70 6.56 5.37 0.051 
Pure error 6.11 1.22   
Table 7

Analysis of regression coefficients for turbidity removal

Factors Regression coefficient Pure error P-value 
Interaction effects 75.6 4.06 18.6 0.000008 
pH(L) 7.6 1.24 6.1 0.00167 
pH(Q) −0.77 0.1 −7.6 0.00063 
rpm (L) −0.11 0.03 −3.2 0.0248 
rpm (Q) 0.0008 0.00016 4.8 0.0051 
R-sqr 0.89    
Adj R-sqr 0.84    
Factors Regression coefficient Pure error P-value 
Interaction effects 75.6 4.06 18.6 0.000008 
pH(L) 7.6 1.24 6.1 0.00167 
pH(Q) −0.77 0.1 −7.6 0.00063 
rpm (L) −0.11 0.03 −3.2 0.0248 
rpm (Q) 0.0008 0.00016 4.8 0.0051 
R-sqr 0.89    
Adj R-sqr 0.84    
From the 3D response surface analysis, the response surface diagram and contour map of turbidity removal could be obtained, as shown in Figures 6 and 7. The diagrams showed that out of various pH values and rotation speeds, the turbidity removal for MNP was the highest at a pH of 4.9 and rpm of 68; turbidity removal did not improve following the increase in pH and reduction in rotation speed. Turbidity removal is the amount of silica particles that MNP can adsorb from wastewater, and is therefore affected by pH value and rotation speed. According to the model, the optimal operating parameters were a pH of 4.9 and rpm of 68, and the estimated highest rate of turbidity removal was 90%.
Figure 6

Turbidity removal response surface of MNP under different pH values and speed coagulations.

Figure 6

Turbidity removal response surface of MNP under different pH values and speed coagulations.

Figure 7

Turbidity removal contour map of MNP under different pH values and speed coagulations.

Figure 7

Turbidity removal contour map of MNP under different pH values and speed coagulations.

In summary, pairing high rpm with low pH conditions (the zeta potentials of MNP and silica particles possessed opposite charges and attracted each other) increased the turbidity removal rate. By contrast, the turbidity removal rate was relatively lower under high pH conditions (the zeta potentials of MNP and silica particles were of similar charges and repelled each other). Verification of the experiment showed that the error between the predicted results of RSM and that of the experiment was approximately 1%; comparatively, the error in turbidity removal rate was much higher than that observed in nonoptimized experiments (Table 8). The optimal conditions were applied in subsequent regeneration and reuse experiments to examine the regeneration and reuse times.

Table 8

Simulated, actual, and nonoptimized turbidity removal

Factors Simulated Actual Pure error (%) Nonoptimized experiment 
pH 4.9 4.9 – 
Rotation speed (rpm) 68 68 – 100 100 100 
Turbidity removal (%) 90% 89% 1.2 64% 67% 52% 
Factors Simulated Actual Pure error (%) Nonoptimized experiment 
pH 4.9 4.9 – 
Rotation speed (rpm) 68 68 – 100 100 100 
Turbidity removal (%) 90% 89% 1.2 64% 67% 52% 

MNP regeneration and reuse experiment

The optimal conditions from the RSM simulation were applied in the MNP regeneration and reuse experiment. In the CMP wastewater adsorption experiment, the conditions were set as MNP at 4.99 g/L, an adjusted pH of 4.9, and mixing at 68 rpm. The optimal turbidity removal rate was 89%.

The study involved two types of desorption experiments (mixing and ultrasonic-assisted). The pH was set at 7 and 9 under a mixing rate of 200 rpm (25 W) for 3 min, and ultrasound (235 W) for 3 min; desorption and reuse were repeated in the experiment. When reuse turbidity removal was ≤80%, MNP was not regenerated and reused.

To summarize the two desorption methods (Table 9 and Figure 8), at pH 7, only one reuse cycle was observed in the rapid mixing and ultrasonic vibration methods. The number of reuses of pH 7 and pH 9 differed because of their zeta potentials. At pH 7, the differences between the zeta potentials of MNP and silica particles were smaller, and thus the mutual repulsion for desorption was also lower. This resulted in relatively incomplete desorption, which was far less effective than the interactions observed in an environment of pH 9; thus, the pH value for desorption should be high to achieve an optimal result. In an alkaline environment, the charges carried by MNP and silica particles were mutually repulsive, meaning that adjusting the pH to 9 would substantially increase the regeneration and reuse times. Thus, pH is one of the key factors in desorption. Following the increase in regeneration and reuse times, the number of silica particles that could be adsorbed by MNP was gradually reduced because adsorption ability became saturated, and therefore the target removal rate of 80% could not be reached. When the pH was 9, the desorption was incomplete because of insufficient energy for mixing and number of times of reuse was only 4; under ultrasonic vibration desorption, the number of reuse cycle reached 5, and a turbidity removal rate of 81% was still achieved compared to the 70% turbidity removal rate of rapid mixing. The average particle size of MNP declined after ultrasonic vibration and the surface area increased, so the reusable adsorption capacity was superior to that of rapid mixing desorption. In contrast, rapid mixing did not change the particle size of MNP; the particle size distribution is shown in Figures 9 and 10.
Table 9

Comparison of MNP regeneration and reuse times

  Mixing (25 W) Ultrasound (235 W) 
pH 7 
pH 9 
  Mixing (25 W) Ultrasound (235 W) 
pH 7 
pH 9 
Figure 8

Using mixing 200 rpm (25 W) and ultrasound (235 W) reuse cycles.

Figure 8

Using mixing 200 rpm (25 W) and ultrasound (235 W) reuse cycles.

Figure 9

MNP(Fe3O4) before the use of ultrasound regeneration distribution (with an average particle size of 23 nm).

Figure 9

MNP(Fe3O4) before the use of ultrasound regeneration distribution (with an average particle size of 23 nm).

Figure 10

MNP(Fe3O4) after using ultrasound regeneration particle size distribution (with an average particle size of 15 nm).

Figure 10

MNP(Fe3O4) after using ultrasound regeneration particle size distribution (with an average particle size of 15 nm).

CONCLUSIONS

In the present study, the MNP were first successfully used in CMP wastewater treatment, and afterwards usefully regenerated between rapid mixing and ultrasound for CMP wastewater. Therefore, some conclusions are presented as follows.

  • 1.

    Through the PB experimental design, pH and mixing were selected as the two key factors of silica particles adsorption by MNP. RSM was then used to obtain the optimal conditions (pH 4.9 and 68 rpm), under which a simulated turbidity removal rate of 90% could be reached. Compared with an actual turbidity removal rate of 89% under these conditions; thus, the results conformed to the simulation value. The actual removal rate was significantly higher than that of a non-optimized condition (approximately 61 ± 8%).

  • 2.

    MNP was applied in the CMP wastewater adsorption experiment. pH and mixing were the two key factors in the experiment. Under acidic conditions, the zeta potentials of MNP and silica particles differed, causing mutual attraction; if appropriate mixing was conducted, the two could be evenly mixed, resulting in enhanced adsorption. By contrast, under neutral or alkaline conditions, the zeta potentials of MNP and silica particles were the same and the adsorption effect was poor.

  • 3.

    The regeneration experiment showed that pH was a key factor. When the pH value was high, because the zeta potentials of MNP and silica particles were the same, mutual repulsion existed and the desorption effect was superior; desorption was more difficult when the environment was neutral or acidic.

  • 4.

    Regarding recycle times and effect, ultrasonic vibration was the superior desorption method for MNP regeneration and reuse because ultrasonic vibration could reduce the particle size of MNP (23 nm to 15 nm), and its adsorption capability was also superior. Recycle times of the MNP by ultrasonic vibration was 5 close to that of the rapid mixing method (4 times).

ACKNOWLEDGEMENTS

Financial support for this study was graciously provided by the National Yunlin University of Science and Technology, Taiwan.

REFERENCES

REFERENCES
Ahmadi
M.
Vahabzadeh
F.
Bonakdarpour
B.
Mehranian
M.
2006
Empirical modeling of olive oil mill wastewater treatment using loofa-immobilized Phanerochaete chrysosporium
.
Process Biochemistry
41
(
5
),
1148
1154
.
doi: https://doi.org/10.1016/j.procbio.2005.12.012
.
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
.
doi: https://doi.org/10.1016/j.talanta.2008.05.019
.
Browne
S.
Krygier
V.
O'Sullivan
J.
Sandstrom
E. L.
1999
Treating wastewater from CMP using ultrafiltration
.
Micro
17
,
5
14
.
Chin
C. J. M.
Chen
P. W.
Wang
L. J.
2006
Removal of nanoparticles from CMP wastewater by magnetic seeding aggregation
.
Chemosphere
63
(
10
),
1809
1813
.
doi: https://doi.org/10.1016/j.chemosphere.2005.09.035
.
Chou
W. L.
Wang
C. T.
Chang
S. Y.
2009
Study of COD and turbidity removal from real oxide-CMP wastewater by iron electrocoagulation and the evaluation of specific energy consumption
.
Journal of Hazardous Materials
168
(
2–3
),
1200
1207
.
doi: https://doi.org/10.1016/j.jhazmat.2009.02.163
.
Chou
W. L.
Wang
C. T.
Chang
W. C.
Chang
S. Y.
2010
Adsorption treatment of oxide chemical mechanical polishing wastewater from a semiconductor manufacturing plant by electrocoagulation
.
Journal of Hazardous Materials
180
(
1–3
),
217
224
.
doi: https://doi.org/10.1016/j.jhazmat.2010.04.017
.
Den
W.
Huang
C.
2005
Electrocoagulation for removal of silica nano-particles from chemical–mechanical-planarization wastewater
.
Colloids and Surfaces A: Physicochemical and Engineering Aspects
254
(
1–3
),
81
89
.
doi: https://doi.org/10.1016/j.colsurfa.2004.11.026
.
Deshmukh
D. V.
Puranik
P. R.
2010
Application of Plackett–Burman design to evaluate media components affecting antibacterial activity of alkaliphilic cyanobacteria isolated from Lonar Lake
.
Turkish Journal of Biochemistry
35
,
114
120
.
Ghazanfari
M. R.
Kashefi
M.
Jaafari
M. R.
2016
Optimizing and modeling of effective parameters on the structural and magnetic properties of fe3o4 nanoparticles synthesized by coprecipitation technique using response surface methodology
.
Journal of Magnetism and Magnetic Materials
409
,
134
142
.
doi: https://doi.org/10.1016/j.jmmm.2016.02.094
.
Huang
C.
Jiang
W.
Chen
C.
2004
Nano-silica removal from IC wastewater by pre-coagulation and microfiltration
.
Water Science and Technology
50
,
133
138
.
Jabeen
H.
Iqbal
S.
Anwar
S.
Parales
R. E.
2015
Optimization of profenofos degradation by a novel bacterial consortium PBAC using response surface methodology
.
International Biodeterioration & Biodegradation
100
,
89
97
.
doi: https://doi.org/10.1016/j.ibiod.2015.02.022
.
Kasiri
M. B.
Aleboyeh
H.
Aleboyeh
A.
2008
Modeling and optimization of heterogeneous photo-Fenton process with response surface methodology and artificial neural networks
.
Environmental Science & Technology
42
(
21
),
7970
7975
.
doi: https://doi.org/10.1021/es801372q
.
Kuan
W. H.
Hu
C. Y.
2009
Chemical evidences for the optimal coagulant dosage and pH adjustment of silica removal from chemical mechanical polishing (CMP) wastewater
.
Colloids and Surfaces A: Physicochemical and Engineering Aspects
342
(
1–3
),
1
7
.
doi: https://doi.org/10.1016/j.colsurfa.2009.03.019
.
Lai
C. L.
Lin
S. H.
2003
Electrocoagulation of chemical mechanical polishing (CMP) wastewater from semiconductor fabrication
.
Chemical Engineering Journal
95
(
1–3
),
205
211
.
doi: https://doi.org/10.1016/j.seppur.2013.02.020
.
Lai
C. L.
Lin
S. H.
2004
Treatment of chemical mechanical polishing wastewater by electrocoagulation: system performances and sludge settling characteristics
.
Chemosphere
54
(
3
),
235
242
.
doi: https://doi.org/10.1016/j.chemosphere.2003.08.014
.
Lai
B. H.
Chang
C. H.
Yeh
C. C.
Chen
D. H.
2013
Direct binding of Concanvalin A onto iron oxide nanoparticles for fast magnetic selective separation of lactoferrin
.
Separation and Purification Technology
108
,
83
88
.
doi: https://doi.org/10.1016/j.seppur.2013.02.020
.
Li
L.
Fan
M.
Brown
R. C.
Van Leeuwen
J.
Wang
J.
Wang
W.
Song
Y.
Zhang
P.
2006
Synthesis, properties, and environmental applications of 87 nanoscale iron-based materials: a review
.
Critical Reviews in Environmental Science and Technology
36
(
5
),
405
431
.
doi: https://doi.org/10.1080/10643380600620387
.
Lin
S. H.
Yang
C. R.
2004
Chemical and physical treatments of chemical mechanical polishing wastewater from semiconductor fabrication
.
Journal of Hazardous Materials
108
,
103
109
.
doi: https://doi.org/10.1016/j.jhazmat.2004.01.014
.
Liu
Y. H.
Lin
C. Y.
Huang
J. H.
Yen
S. C.
2016
Particle removal performance and its kinetic behavior during oxide-CMP wastewater treatment by electrocoagulation
.
Journal of the Taiwan Institute of Chemical Engineers
60
,
520
524
.
doi: https://doi.org/10.1016/j.jtice.2015.11.023
.
Micic
R. D.
Tomić
M. D.
Kiss
F. E.
Nikolić-Djorić
E. B.
Simikić
M. Ð.
2015
Optimization of hydrolysis in subcritical water as a pretreatment step for biodiesel production by esterification in supercritical methanol
.
The Journal of Supercritical Fluids
103
,
90
100
.
doi: https://doi.org/10.1016/j.supflu.2015.04.026
.
Roth
G. A.
Neu-Baker
N. M.
Brenner
S. A.
2015
SEM analysis of particle size during conventional treatment of CMP process wastewater
.
Science of The Total Environment
508
,
1
6
.
doi: https://doi.org/10.1016/j.scitotenv.2014.11.075
.
Tsai
J. C.
Kumar
M.
Chen
S. Y.
Lin
J. G.
2007
Nano-bubble flotation technology with coagulation process for the cost-effective treatment of chemical mechanical polishing wastewater
.
Separation and Purification Technology
58
(
1
),
61
67
.
doi: https://doi.org/10.1016/j.seppur.2007.07.022
.
Wan
T. J.
Shen
S. M.
Siao
S. H.
Huang
C. F.
Cheng
C. Y.
2011
Using magnetic seeds to improve the aggregation and precipitation of nanoparticles from backside grinding wastewater
.
Water Research
45
(
19
),
6301
6307
.
doi: https://doi.org/10.1016/j.watres.2011.08.067
.
Wang
C. T.
Chou
W. L.
Chen
L. S.
Chang
S. Y.
2009
Silica particles settling characteristics and removal performances of oxide chemical mechanical polishing wastewater treated by electrocoagulation technology
.
Journal of Hazardous Materials
161
(
1
),
344
350
.
doi: https://doi.org/10.1016/j.jhazmat.2008.03.099
.
Yang
G. C.
Tsai
C. M.
2006
Performance evaluation of a simultaneous electrocoagulation and electrofiltration module for the treatment of Cu-CMP and oxide-CMP wastewaters
.
Journal of Membrane Science
286
,
36
44
.
doi: https://doi.org/10.1016/j.memsci.2006.09.007
.