Abstract

In recent years, electrocoagulation has been extensively studied on the removal of refractory pollutants. However, the application of electrocoagulation in actual flocculation tank is limited because of its high energy consumption, especially under the condition of large electrode plate spacing. In this study, the computational fluid dynamics (CFD) software – ANSYS Fluent had been used to simulate the flow state of grid flocculation tank, for the purposes of optimizing the design parameters. The simulation results showed that vortex velocity gradient was stronger, the grid plate spacing was smaller when the velocity was 0.13 m s−1, perforation size was 25 × 25 mm, porosity was 31.25%. And the optimal grid plate spacing was 250 mm. Moreover, in order to prove the reasonableness of simulation results, the humic acid wastewater was treated by electrocoagulation process in the specific device which was built based on simulation results. The results showed that the optimal condition of orthogonal test were as follows: the initial pH was 8, the concentration of sodium chloride was 5 mmol L−1, the voltage was 15 V; and the power time was 60 min. This study greatly narrowed the grid plate spacing, optimized design parameters under the circumstances of strong turbulent intensity and provided a theoretical basis for the combination of electrocoagulation and hydraulic flocculation.

INTRODUCTION

Electrocoagulation (also called electrochemical) has been widely used for refractory pollutants treatment such as dyes and humic acid. Compared with conventional chemical coagulation, electrocoagulation has the following advantages: the flocculant produced by electrocoagulation has a stronger adsorption effect and higher activity; the sludge production is low and easy to disposal; the process of flocculation, air flotation, electrolytic oxidation and electrolytic reduction can be completed in a single device, which is simple and easy to operate and manage (Lai & Lin 2004; Essadki et al. 2008). However, the great energy consumption of electrocoagulation limits its practical application, especially in the case of large plate spacing. At present, the research of electrocoagulation is mainly in the stage of beaker test, in which the plate spacing is small and water purification capacity is limited.

Flocculation reaction equipment in conventional water treatment process is mainly divided into two categories: mechanical agitation and hydraulic agitation. Grid flocculation tank is one kind of efficient hydraulic mixing flocculation tank. Compared with other forms of flocculation tank, grid flocculation tank has the advantages of smaller head loss, shorter flocculation time and wider applicable conditions (Koh 1984; He et al. 2018). At present, the simulation researches of grid flocculation tank based on computational fluid dynamics (CFD) mainly focus on the structure optimization and parameters optimization, as well as evaluation indexes of flocculation effect (Klimpel & Hogg 1986; Wong et al. 1992; Bouyer et al. 2005). However, the researches are rarely involved in reducing the spacing of grid layers and combining it with electrocoagulation.

In this study, the flow field of grid flocculation tank had been simulated using the CFD software — ANSYS Fluent to optimize the design parameters and reduce the spacing of grid layers. Then the experimental device was designed according to the simulation results and the grid coagulation principle was combined with electrocoagulation technology to treat the humic acid wastewater. The optimum operation parameters of the device were determined through the dynamic test under the premise of low energy consumption and good effect, which could strengthen the complementary and synergistic effect of those two processes.

MATERIALS AND METHODS

Flocculation evaluation index

Vortex velocity gradient

A number of papers proposed (Hanhui et al. 2005; Bubakova & Pivokonsky 2012; Popovici 2017; Yang 2017) that velocity gradient (G) could not well explain some fluid mechanics phenomena. For example, time-average value, which could be regarded as isotropic, was used to describe the hydraulic characteristic of a certain area behind flocculation plate of the grid flocculation tank. But as a statistical parameter, the time-average velocity would tend to 0, which meant that G was 0, and this was obviously not consistent with the actual flow state of corresponding position.

However, vortex velocity gradient (G′) could better explain this phenomenon compared with velocity gradient(G). The velocity in turbulence could be divided into time-average velocity and pulsating velocity. Although the time-average velocity was 0 in the area behind flocculation plate, the floc particles still collided with each other under the action of the pulsating vortex velocity gradient. 
formula
(1)
where G0 is time-average velocity gradient(s−1), and is vortex velocity gradient (s−1).

It could be seen that the turbulence vortex velocity gradient was part of the velocity gradient, so the flocculation effect could be enhanced by increasing the vortex velocity gradient. At the same time, could be used as an index to evaluate the flocculation effect.

Flow field distribution

According to Kolmogoroff's theory, the energy of the fluid was reflected by the vortex of the water and was proportional to the size of the vortex (Rosenfeld 1994). So the flocculation effect could be judged by the number and the size of micro-vortex in the flow field distribution.

Mathematic models

Continuity equation is the dominate equation for the transient flow of viscous incompressible fluid (Li et al. 2015; Robescu & Manea 2015). 
formula
(2)
where u, v are the components of the flow velocity on the x, y axis, respectively.
Nacier-Stokes equation: 
formula
(3)
where ρ is fluid density.

Boundary conditions

There are some boundary conditions available in the fluent, such as pressure-inlet, pressure-outlet, velocity-inlet, velocity-outlet and outflow (Li et al. 2001). In order to accurately simulate the flow field, appropriate boundary conditions were required to avoid incorrect flow field solution.

Entrance boundary conditions

Water was uniformly distributed at the entrance of the model, so velocity-inlet boundary was adopted as the entrance boundary conditions.

Export boundary conditions

The export boundary conditions used outflow boundaries. The boundary condition was applicable to the flow field where the velocity and pressure of outlet were unknown, and the flow at the outlet was required to approach the fully developed state (Golzarijalal et al. 2017).

Wall conditions

Non-slip boundary conditions were the default options for viscous flow in the fluent. The standard wall function method was adopted in the area near the wall.

Free surface condition

In this paper, the steel cover hypothesis was applied to describe the free surface as a smooth plane with a normal velocity of zero.

Simulation parameters

The design flow was 1,000 m3d−1. The density of water was 998.3 kg m−3. The viscosity coefficient of water was 1.005 × 10−3. The shaft flow velocity was 0.13 m s−1. The orifice flow velocity was 0.41 m s−1.

Experimental design

The channel of the grid flocculation tank is roughly the same, and each of them can be seen as a set of CSTR reactors (Bhattacharya et al. 2017). In order to verify the simulation result, a three-dimensional model fitted to the optimal simulation size was adopted. The length, width and height of this model was 300 mm, 300 mm and 1,000 mm, respectively. The experimental device is shown in Figure 1. The device, which effective volume was 45 L, was built by the material of plexiglass, and the circulation was realized by water pump during the entire process of the experiment. The flow was a continuous viscous incompressible fluid. The top of the model was the inlet boundary, while the bottom was the outlet boundary.

Figure 1

An illustration of experimental system.

Figure 1

An illustration of experimental system.

As shown in Figure 1, the aluminum grids with two layers, which were used as electrode at the same time. The space between the two grid plates was 250 mm (Figure 1) in the experimental device which was based on the simulation results. As shown in Figure 2, there were 45 (5 × 9) holes evenly distributed on the electrode plate, and the single size was 25 × 25 mm. The holes of upper and lower plates were staggered. Plate orifice ratio was 31.25%.

Figure 2

The illustration of grid plate.

Figure 2

The illustration of grid plate.

During the experiment, the power supply kept running all the time and the voltage remained constant. The experiment processes included two parts. First, the flocculant was mixed well with the raw water by keeping the submersible pump running for 5 min. Then, the submersible pump was powered off and the two peristaltic pumps were powered on at the same time running continuously with maximum flow until the end of the reaction. In the second stage, flocculation was carried out with electrochemical action simultaneously.

A certain amount of humic acid was accurately weighed, and then dissolved in the sodium hydroxide solution with a concentration of 0.1 mol L−1 for 24 h using a magnetic stirrer. The 0.45 um filter membrane was used to remove the insoluble impurities during the filtration. Then the filtrate was transferred to a volumetric flask as reserve liquid and sealed in a low temperature storage condition without lighting. The humic acid solution required by the experiment was prepared by diluting the stock solution to the corresponding concentration and adjusting its pH to a specific value.

UV254, which characterizes the humic acid content in the water, and A400, which characterizes the chroma of the water, were tested by an ultraviolet-visible spectrophotometer (DR6000, Hach, USA). The methods of measuring UV254 were as follows. First, cellulose acetate membranes were used to filter water samples. Then, ultrapure water was taken as blank sample and the absorbance value was determined by using quartz cuvette which optical path was 1 cm under the wavelength of 254 nm in the ultraviolet spectrophotometer. Finally, the filtered water samples were tested by repeating the same steps. The measuring method of A400 was similar to that of UV254 except that the working wavelength was 400 nm.

RESULTS AND DISCUSSION

Simulation of grid flocculation tank with different plate spacing

The grid plate spacing of 200 mm, 250 mm, 300 mm, 350 mm and 400 mm were simulated and analyzed, respectively. The images of vortex velocity gradient on Y = 0 plane are shown in Figure 3.

Figure 3

Vortex velocity gradient cloud image at different plate spacing.

Figure 3

Vortex velocity gradient cloud image at different plate spacing.

The various patches in Figure 3 represent different vortex velocity gradient intensities. The vortex velocity gradient can reflect the magnitude of the turbulence. The water flow becomes turbulent after passing through the grid. The more violent the turbulence, the higher the probability of flocculation collision, then the more compact flocculation generated (Young & Edwards 2004).

When the plate spacing was 300 mm, 350 mm and 400 mm, the effective flocculation section caused by the first-layer grid plate developed completely. The vortex velocity gradient was small and the flocculation effect was weak at the dark blue section, which was close to the lower plate. The flow distribution was equivalent to a single layer grid plate acted independently. When the plate spacing was 200 mm, the effective flocculation section was not fully developed, and intersected with the lower plate, which led to a lower energy utilization rate compared with the former ones. When the plate spacing was 250 mm, the effective flocculation section was partially extended to the lower plate, but the turbulence vortex velocity gradient, which value was 5.27 × 10−3s−1, was weak at the contact surface of the lower flocculation plate. It was considered that the effective flocculation zone was reasonable.

The large vortex became smaller and smaller due to the inertial effect of the water flow. The vortex continued to produce, develop and eventually decay and disappear under the effect of viscous force (Du et al. 2000; Moruzzi & de Oliveira 2013). The vector diagram of Y = 0 plane is shown in Figure 4 when plate spacing was 250 mm. The energy of the fluid was presented in the vortex and was proportional to the size of the vortex. It could be seen from the figure that there were plenty of micro-vortex behind each layer of plate, which played a vital role in enhancing the effect of flocculation.

Figure 4

Vector diagram when plate spacing was 250 mm.

Figure 4

Vector diagram when plate spacing was 250 mm.

In summary, when the plate spacing was 250, the vortex velocity gradient between two plates was larger, the number of micro vortices was more, and the effective flocculation section was more reasonable. Grid flocculation tank gave full play to flocculation and made full use of energy. The costs of infrastructure would reduce when the plate spacing decreased.

Simulation of grid flocculation tank with four-layer grid plates

In order to verify that there was no difference between the flow of four-layer grid plates and two-layer plates. A four-layer grid plates model was used to simulate the process while the plate spacing, number and size of grid size remained unchanged. The vortex velocity gradient cloud image and the velocity vector diagram are shown in Figures 5 and 6, respectively.

Figure 5

Vortex velocity gradient cloud image between four-layer grid plates.

Figure 5

Vortex velocity gradient cloud image between four-layer grid plates.

Figure 6

Vector diagram between four-layer grid plates.

Figure 6

Vector diagram between four-layer grid plates.

As can be seen from the Figure 6, the flow state between the four-layer plates model and the two-layer plates model were almost the same. The effective flocculation section accounted for almost the whole area between the two plates. The number of micro vortices was more, mainly in the post-plate area. The application of the multi-layer grid plates helped to improve space utilization and enhance the flocculation effect.

Orthogonal test

As a refractory organic matter, there were many factors including voltage, sodium chloride concentration, power time, and initial pH that affected the removal of humic acid through electrocoagulation process. The pH of natural water was generally between 6.5 and 7.8. In order to further study the effect of pH on the removal rate of humic acid, the pH range was chosen as 5–10. When the conductivity remained constant, the greater the voltage, the faster the electrolysis reaction rate. The hydroxyl radicals and Al3+ produced by electrodes increased with the prolonging of energization time, but the experimental energy consumption would increase accordingly. At the same time, the polarization and passivation phenomenon of the electrode plates was obvious (Sun et al. 2018). The experiment time was set to 30–75 min, and the voltage was 10–25 V.

The sodium chloride was added to the solution in order to improve its conductivity. The natural water conductivity was generally between 50–1,500 μs cm−1 (Hua et al. 2015), so the sodium chloride concentration was set to 0.5–10 mmol L−1. The conductivity of experimental water was set to 512–1,673 μs cm−1 and the poly-aluminum-chloride (PAC) dosage was 30 mg L−1. The value of each factor in the orthogonal test is shown in Table 1.

UV254 removal rate calculation formula: 
formula
(4)
where A0 is solution absorbance before treatment, A1 is solution absorbance after treatment.
Energy consumption formula: 
formula
(5)
where U is the voltage (V), I is the current (A), t is the time(s) and V is the solution volume (m3) (Garcia-Garcia et al. 2014).

It can be seen from Figure 7 that the removal rate of UV254 and A400 was positively correlated, and the difference was small, which indicated that the main chroma material of the experimental raw water was humic acid. Among the experiments, the removal rate of UV254 and A400 of No. 2, No. 3, No. 6 and No. 10, No. 11, No. 15, and No. 16 experiment was above 82%. The sodium chloride concentration of No. 2, No. 3, and No. 16 experiment was 10 mmol L−1. The conductivity was about 1,600 μs cm−1, which was higher than the average value of natural water body, so its representative was poor. At the same time, the excess dose might lead to the increase of subsequent desalination units, thereby increasing the cost of water purification. The electric time of No. 6 and No. 15 experiment was 75 min, which meant a longer flocculation channel was required, and thus would inevitably bring the high cost of infrastructure.

Figure 7

Result of orthogonal test.

Figure 7

Result of orthogonal test.

From the energy consumption formula, we could see that there was no difference between the two removal rates when the energy consumption value of No. 10 and No. 11 experiment was 0.166 kWh/m3 and 0.372 kWh/m3, respectively, so the result of No. 10 experiment was optimal. The experimental condition was as follows: the initial pH was 8, the concentration of sodium chloride was 5 mmol L−1, the voltage was 15 V, and the power time was 60 min.

The range R was used to measure the importance of the factors. The larger the R, the greater the change in the experimental result. The results of orthogonal test are shown in Table 2. The range R1, R2 represented the removal rate of UV254 and energy consumption, respectively.

Table 1

Factor and level of orthogonal test

Factors Levels 
Initial pH 
Sodium chloride concentration/(mmol L−10.5 10 
Voltage/V 10 15 20 25 
Power time/min 30 45 60 75 
Factors Levels 
Initial pH 
Sodium chloride concentration/(mmol L−10.5 10 
Voltage/V 10 15 20 25 
Power time/min 30 45 60 75 
Table 2

Result of orthogonal test

Test number Initial pH NaCl concentration/mmol L−1 Voltage/V Electrolysis time/min UV254 removal rate/% A400 removal rate/% Current/A Energy consumption/kWh m−3 
20 45 76.83 70.80 0.35 0.125 
10 10 45 82.50 82.35 0.42 0.075 
10 20 60 85.42 86.33 0.96 0.457 
10 60 50.51 50.74 0.136 0.032 
10 75 75.19 71.53 0.286 0.085 
0.5 20 75 85.96 87.79 0.31 0.185 
0.5 10 30 49.08 48.06 0.146 0.017 
20 30 59.49 57.25 0.577 0.137 
0.5 25 60 75.43 67.41 0.466 0.277 
10 15 60 84.68 85.82 0.465 0.166 
11 25 45 84.07 84.14 0.834 0.372 
12 0.5 15 45 27.65 22.76 0.24 0.064 
13 10 15 30 56.04 48.18 0.759 0.136 
14 25 30 46.92 38.85 0.473 0.141 
15 15 75 84.32 87.03 0.299 0.133 
16 10 25 75 88.89 89.55 1.229 0.914 
K1 70.87 59.53 64.32 52.88     
K2 75.72 64.65 63.17 67.76     
K3 75.02 75.86 76.93 74.01     
K4 56.64 78.21 73.83 83.59     
Maximum 75.72 78.21 76.93 83.59     
Minimum 56.64 59.53 63.17 52.88     
R1 19.09 18.68 13.75 30.70     
K1 0.16 0.14 0.05 0.11     
K2 0.25 0.11 0.12 0.16     
K3 0.14 0.19 0.23 0.23     
K4 0.29 0.40 0.43 0.33     
Maximum 0.29 0.40 0.43 0.33     
Minimum 0.14 0.11 0.05 0.11     
R2 0.15 0.29 0.37 0.22     
Test number Initial pH NaCl concentration/mmol L−1 Voltage/V Electrolysis time/min UV254 removal rate/% A400 removal rate/% Current/A Energy consumption/kWh m−3 
20 45 76.83 70.80 0.35 0.125 
10 10 45 82.50 82.35 0.42 0.075 
10 20 60 85.42 86.33 0.96 0.457 
10 60 50.51 50.74 0.136 0.032 
10 75 75.19 71.53 0.286 0.085 
0.5 20 75 85.96 87.79 0.31 0.185 
0.5 10 30 49.08 48.06 0.146 0.017 
20 30 59.49 57.25 0.577 0.137 
0.5 25 60 75.43 67.41 0.466 0.277 
10 15 60 84.68 85.82 0.465 0.166 
11 25 45 84.07 84.14 0.834 0.372 
12 0.5 15 45 27.65 22.76 0.24 0.064 
13 10 15 30 56.04 48.18 0.759 0.136 
14 25 30 46.92 38.85 0.473 0.141 
15 15 75 84.32 87.03 0.299 0.133 
16 10 25 75 88.89 89.55 1.229 0.914 
K1 70.87 59.53 64.32 52.88     
K2 75.72 64.65 63.17 67.76     
K3 75.02 75.86 76.93 74.01     
K4 56.64 78.21 73.83 83.59     
Maximum 75.72 78.21 76.93 83.59     
Minimum 56.64 59.53 63.17 52.88     
R1 19.09 18.68 13.75 30.70     
K1 0.16 0.14 0.05 0.11     
K2 0.25 0.11 0.12 0.16     
K3 0.14 0.19 0.23 0.23     
K4 0.29 0.40 0.43 0.33     
Maximum 0.29 0.40 0.43 0.33     
Minimum 0.14 0.11 0.05 0.11     
R2 0.15 0.29 0.37 0.22     

The results showed that the sequences of influential factors which affected the humic acid removal efficiency were as follows: electrolysis time > initial pH > NaCl concentration > voltage. As for energy consumption: voltage > NaCl concentration > electrolysis time > initial pH. Time and the voltage played a considerable role in the removal of humic acid and energy consumption, respectively.

Comparison experiment of grid flocculation electrochemical technology and coagulation

In order to discuss the relationship between flocculation and removal efficacy of organic matter, the comparison experiment of grid flocculation-electrochemical technology and coagulation on humic acid removal was conducted. The experimental parameters were as follows. The initial pH was 8, the concentration of sodium chloride was 5 mmol/L, the voltage was 15 V in the grid flocculation-electrochemical experiment, the reaction time was 60 min and the PAC dosage was 30 mg/L.

According to Figure 8, the removal rate of humic acid was higher in the grid flocculation-electrochemical experiment. The removal rate of the two experiments was not so obvious within the first 20 min. After that, there was an absolutely clear difference. The analyses of the results above were as follows. The removal mechanisms of humic acid by grid flocculation-electrochemical technology contained three parts. First, PAC performed the function of charge neutralization and adsorption bridging. Second, the Al3+ produced by anodic dissolution was hydrolyzed and polymerized to form polyhydroxyl complexes and metal hydroxides, thus exerting the effect of electric flocculation. Third, the H2 released from the cathode and the O2 released from the anode could form tiny bubbles to remove the floc. But humic acid was removed only by charge neutralization and adsorption bridging in the coagulation experiment. Therefore, the effect of humic acid removal was better when electrified. As the reaction was prolonged in the grid flocculation-electrochemical experiment, anodic would form corrosion pits on the surface because of dissolution. When the aluminum plate was used to a certain extent, the dissolution rate of aluminum plate would be affected, as well as the formation rate of bubbles, thus weakening the electro flocculation and electrical flotation.

Figure 8

Result of comparison test.

Figure 8

Result of comparison test.

CONCLUSIONS

This paper used the fluent to simulate the grid flocculation tank shafts, then treated the acid wastewater by flocculation-electrochemical technique and obtained the following conclusions:

  1. When the velocity was 0.13 m s−1, perforation size was 25 × 25 mm, porosity was 31.25% and the grid plate spacing was 250 mm, the vortex velocity gradient between the plates and the number of micro vortices were larger, and the effective flocculation section was reasonable. There was almost no difference in flow state between the four-layer plates model and the two-layer plates model. The application of the multi-layer grid plate helped to improve space utilization and enhance the flocculation effect.

  2. The effect of grid flocculation-electrochemical technique on the treatment of humic acid raw water was studied through the orthogonal experiment. The optimal conditions were as follows: the initial pH was 8, the concentration of sodium chloride was 5 mmol L−1, the voltage was 15 V, and the power time was 60 min. The results showed that the sequences of influential factors which affected the humic acid removal efficiency were as follows: electrolysis time > initial pH > NaCl concentration > voltage. As for energy consumption: voltage > NaCl concentration > electrolysis time > initial pH.

  3. The results of numerical simulation showed that grid flocculation tank with smaller plate spacing had a reasonable flow field and flocculation effect. The experimental study showed that low energy consumption and good treatment effect could be achieved simultaneously using grid flocculation tank combined with electrocoagulation to treat humic acid wastewater.

ACKNOWLEDGEMENTS

Financial support by the 111 Project (No. B13041) and the Fundamental Research Funds for the Central Universities of China (No. 106112015CDJXY210002) is acknowledged.

REFERENCES

REFERENCES
Bhattacharya
A.
,
Malik
A.
&
Malik
H. K.
2017
A mathematical model to describe the fungal assisted algal flocculation process
.
Bioresource Technology
244
(
1
),
975
981
.
Bouyer
D.
,
Coufort
C.
,
Line
A.
&
DO-Quang
Z.
2005
Experimental analysis of floc size distribution in a 1-L jar under different hydrodynamics and physiochemical conditions
.
Journal of Colloid & Interface Science
292
,
413
428
.
Garcia-Garcia
P.
,
Arroyo-Lopez
F.
&
Rodriguez-Gomez
F.
2014
Partial purification of iron solutions from ripe table olive processing using ozone and electro-coagulation
.
Separation and Purification Technology
133
,
227
235
.
Hanhui
Z.
,
Xiaoqi
Z.
&
Xuehui
Z.
2005
Study on mixed shear stage in physical model of coagulant
.
Environmental Science & Technology
S1
,
4
6
.
He
W.
,
Xue
L.
,
Gorczyca
B.
,
Nan
J.
&
Shi
Z.
2018
Experimental and CFD studies of floc growth dependence on baffle width in square stirred-tankreactors for flocculation
.
Separation and Purification Technology
190
,
228
242
.
Hua
L.-C.
,
Huang
C.
,
Su
Y.-C.
,
Nguyen
T.-N.-P.
&
Chen
P.-C.
2015
Effects of electro-coagulation on fouling mitigation and sludge characteristics in a coagulation-assisted membrane bioreactor
.
Journal of Membrane Science
495
,
29
36
.
Klimpel
R. C.
&
Hogg
R.
1986
Effects of flocculation conditions on agglomerate structure
.
Journal of Colloid & Interface Science
113
(
1
),
121
131
.
Li
Y.
,
Liu
Z.
&
An
Y.
2001
Introduction to computational fluid mechanics – FLUENT
.
Chinese Journal of Hydrodynamics
16
(
2
),
254
258
.
Li
Y.
,
Gu
A.
&
Li
J.
2015
Existence and continuity of bi-spatial random attractors and application to stochastic semilinear Laplacian equations
.
Journal of Differential Equations
258
(
2
),
504
534
.
Moruzzi
R. B.
&
de Oliveira
S. C.
2013
Mathematical modeling and analysis of the flocculation process in chambers in series
.
Bioprocess and Biosystems Engineering
36
(
3
),
357
363
.
Robescu
L. D.
&
Manea
E. E.
2015
Using CFD techniques in teaching rectangular settling tank hydrodynamics
. In:
Conference Proceedings Balkan Region Conference on Engineering Education & MSE
,
1 (4)
.
Rosenfeld
Y.
1994
Density-functional of molecular fluids-free-energy model for the inhomogeneous hard-body fluid
.
Physical Review E
50
(
5
),
R3318
R3321
.
Wong
K.
,
Lixon
P.
,
Lafuma
F.
,
Lindner
P.
,
Charriol
O. A.
&
Cabane
B.
1992
Intermediate structures in equilibrium flocculation
.
Journal of Colloid & Interface Science
153
(
1
),
55
72
.
Yang
X.
2017
The lateral variation of P-n velocity gradient under Eurasia
.
Journal of Geophysical Research-Solid Earth
122
(
5
),
3677
3688
.
Young
J. C.
&
Edwards
F. G.
2003
Factors affecting ballasted flocculation reactions
.
Water Environment Research
75
(
3
),
263
272
.