Separation and Purification Technology 237 (2020) 116319

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Analysis of the flocculation process of fine tailings particles in saltwater T

through a population balance model
⁎
Gonzalo R. Quezadaa, Jahir Ramosb, Ricardo I. Jeldresb, , Pedro Roblesc, Pedro G. Toledod
a
Water Research Center for Agriculture and Mining (CRHIAM), Concepción, Chile
b
Departamento de Ingeniería Química y Procesos de Minerales, Facultad de Ingeniería, Universidad de Antofagasta, PO Box 170, Antofagasta, Chile
c
Escuela de Ingeniería Química, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile
d
Department of Chemical Engineering and Laboratory of Surface Analysis (ASIF), Universidad de Concepción, PO Box 160-C, Correo 3, Concepción, Chile

A R T I C LE I N FO A B S T R A C T

Keywords: A population balance model is used to describe the flocculation of tailings particles in aqueous salt solutions. The
Clay tailings synthetic tailings, composed of quartz and kaolin particles, are flocculated in a jar at a constant shear rate where
Bridging flocculation in-situ FBRM determines the size of the aggregates. The model follows the dynamics of aggregation and breakage
Magnesium hydroxide processes and provides a good approximation to the temporal evolution of aggregate size. The fractal and
Population balance modelling
permeable nature of the aggregates are considered, while the depletion of the collision efficiency allows de-
Seawater
scribing the initial growth of aggregates and subsequent size reduction. The numerical solution requires five
parameters, which are obtained by minimizing the difference between experimental size data and model pre-
dictions. A specific aim is to study the effect of magnesium hydroxide that is formed at pH ca. 10, and its
interaction with flocculant, on the flocculation kinetics parameters. At pH ≥ 10 the aggregates grow less due to
the presence of the magnesium hydroxide gel that surrounds quartz, kaolin and flocculant. The fractal dimension
is quite stable at pH < 10 with a representative value of 2.7, typical of a clustered network, although in the
presence of magnesium at pH ≥ 10 the fractal dimension of the aggregates is only 2.2, typical of Gaussian
chains. Tailings aggregates in the presence of hydroxide are smaller and weakly three-dimensional and therefore
contribute little to the settling velocity. The aggregation and breakage parameters are largely constant for a
particulate system which composition remains unchanged over a pH range, and if the composition changes, for
example by precipitation of magnesium hydroxide, then the aggregation parameters are different but close to
constant.

1. Introduction destination in tailings dams. Experimental evidence shows that the

adsorption of the polymer on the mineral requires the mediation of
Low-grade ore deposits, shortage of water resources and seawater multivalent ions, otherwise, the adsorption is low [3,4]. This suggests
used directly or partially desalinated have created a challenging sce- that the adsorption mechanism is by means of salt bridges. There is
nario for the sustainability of the mining industry. A great challenge is plenty of experimental work on flocculation in saline water, although
the effective closing of the water circuit, ideally with a modest water the final impact of the electrolytes is not yet clear. Jeldres et al. [5] and
makeup, which involves optimizing the thickener stage for efficient later Quezada et al. [6] showed that an increase in salinity favors the
solid-liquid separation. Fine particles from old deposits that remain in adsorption of polyelectrolytes on the mineral surfaces, although at the
suspension require high molecular weight flocculants to collect them in same time the size of the flocculant is reduced by winding. These two
large flocs heavy enough to decant by gravity [1,2]. competitive mechanisms favor or deteriorate flocculation, producing
Commonly, copper concentration operations are carried out at al- different responses respect to the presence of electrolytes.
kaline pH (pH > 10.5), which implies that both mineral particles and In fresh water, it is common for the flotation and thickening op-
flocculant molecules develop negative surface charges. This determines erations to be carried out at pH 10–12. In seawater, at a strongly al-
that the mineral-flocculant affinity is low, although sufficient to pro- kaline pH, typical hydroxides are generated in the form of specific
duce porous aggregates of low density to easy the transport to the final complexes Ca(OH)+, Mg(OH)+ and corresponding precipitates Ca

⁎
Corresponding author.
E-mail addresses: gonzaloquezada@udec.cl (G.R. Quezada), ricardo.jeldres@uantof.cl (R.I. Jeldres), pedro.robles@pucv.cl (P. Robles),
petoledo@udec.cl (P.G. Toledo).

https://doi.org/10.1016/j.seppur.2019.116319
Received 5 August 2019; Received in revised form 4 November 2019; Accepted 15 November 2019
Available online 17 November 2019
1383-5866/ © 2019 Elsevier B.V. All rights reserved.
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

(OH)2, Mg(OH)2. It has been shown that these hydroxides are quite Table 1
harmful in the flotation of sulfur minerals, since they are deposited on Input parameters and conditions.
the mineral surfaces, making them hydrophilic and, therefore, sig- N 170 RPM
nificantly decreasing their flotation capacity [7,8]. In other contexts, it
has been shown that the incorporation of these salts at high pH has been Np 0.6
useful for the coagulation of organic matter, wastewater treatment, D 8.0 Cm
V 0.25 l
drinking water production, microalgae handling, etc. [9–12]. In the
ρs 2600 kg/m3
presence of polymeric flocculants, the behavior varies according to the
ρw 1000 kg/m3
system. Dupont et al. [13] studied the PAA/alumina/Ca2+ system and w 0.13
found that adsorption is favorable in the presence of calcium when the
pH increases. For their part, Peng and Di [14] studied the HPAM/
kaolinite/Al3+ system and Ca2+, finding the hydroxide complexes ad- into the flocculation of tailings particles that contribute to sustainable
sorbed both on the polymer and on the surface, thus decreasing the water management for a sustainable mining industry.
adsorption of the polymer on the surface of the kaolinite.
The phenomena involved during the particle aggregation process 2. Model
determine the efficiency of the operations of thickening and transport
to tailings dam. Therefore, flocculation studies have focused at the The population balance equation (PBE) used in this work is derived
microscopic level on the measurement of properties of aggregates, such from the work proposed by Spicer and Pratsinis [34] for particle
as size distribution. Several techniques can be used for this purpose grouping, which combines the discretization approach of Hounslow
[15], although the measurement with focused beam reflectance (FBRM) et al. [26] for aggregation and Kusters et al. [27] for breakage. In this
has important advantages. It can be applied to feed slurries without approach, the size distribution of the particles is discretized into a
solids dilution and also to in-line monitoring of aggregate size without number of size intervals, classes or bins, according to a geometric
sampling and off-line analysis [16]. At the same time, several re- progression of the volume distribution of the aggregates. Thus, the
searchers have described the kinetics of aggregate growth and frag- volume of the aggregates in bin i + 1doubles that of bin i , i.e.,
mentation [17–20]. These models make extensive use of the classic von Vi + 1 = 2Vi . The volume of aggregates in any bin i respect to the volume
Smoluchowski equation [21] for irreversible aggregation and various of aggregates in the first bin is Vi = 2i − 1V1. The PBE is:
heuristics additions with adjustable parameters for reconciling experi-
mental data with model. The modeling with these equations, called dNi
population balance modeling (PBM) has proved to be a great help to dt
i−2 i−1
understand the aggregation process and to quantify the effect of oper- 1
= ∑ 2 j−(i −1) Qi −1,j Ni −1 Nj + Qi − 1, j − 1 Ni − 1 Ni − 1 − ∑ 2 j−iQi,j Ni Nj −
ating conditions, such as shear rate, flocculant dose, concentration of j=1 2 j=1
solids, and pH [22]. PBM can be used to describe the temporal change max1 max 2
of a particle size distribution under diverse conditions. The PBM model ∑ Qi, j Ni Nj − Si Ni + ∑ Γi, j Sj Nj
have been used to describe a wide range of systems of industrial in- j=1 j=1 (1)
terest, for example, coagulation and flocculation [23], size reduction in
where Ni is the number concentration of aggregates in bin i , whereas N1
mills [24], and polymerization [25].
is the number concentration of primary particles in bin 1. Each term in
The PBM models are increasingly robust and predictive. To this end,
Eq. (1) represents different events involving aggregates of size i that
modifications have been introduced to the Smoluchowski equation to
occur at time t. The aggregation-breakage-restructuring time is t. The
represent more complex systems that include aggregation and dis-
first two terms on the right-hand side of Eq. (1) describe aggregate
aggregation, apart from numerical techniques that facilitate calcula-
formation of size i from collisions of smaller aggregates. The next two
tions. One of these improvements is the early work of Hounslow et al.
terms represent the loss of aggregates in the i-th bin that become larger
[26] together with the work of Kusters et al. [27]. The recognition of
aggregates. The fifth term accounts for the loss of aggregates in the i-th
the fractal nature of the structure of many aggregates has been key to
bin through fragmentation. The last term denotes the gain of aggregates
represent real systems [18]. In bridging flocculation of mineral tailings,
of size i by fragmentation of larger flocs. The superscript max1 and max2
the size of the aggregates is expected to increase to a maximum and
represent the maximum number of size intervals used to describe the
then decrease until reaching a stationary value [19,28,29]. Several
complete aggregate size spectrum, respectively, by aggregation and
authors have tried to describe this behavior by adding richer but not
fragmentation. Q , S and Γ are functions representing, respectively, ag-
fully developed physics that is therefore completed with parameters, for
gregation rate, fragmentation rate, and breakage distribution function
example, Selomulya et al. [30] incorporate a fractal dimension that
of particulate systems. The six functions in Eq. (1) have heuristic origin
changes during flocculation due to rupture and restructuring of ag-
and therefore the parameters involved must be determined by solving
gregates. Other approaches include permeability as in the work of
the equation against experimental data. The functions are described
Ahmad et al. [31]. Heath et al. [19] and recently Vajihinejad & Soares
below for permeable particles with fractal nature of which spherical
[32] have proposed a model for the efficiency of particle capture in the
and impermeable aggregates are particular cases.
formation of aggregates; a function that decreases with time and re-
presents the decrease in effectiveness of the flocculant over time due to
its irreversible adhesion to the particles, the effect is a decrease in the 2.1. Aggregation kernel
size of the aggregates [33].
In the present work, a population balance model is applied to the The aggregation kernel Q is the product of the collision frequency
flocculation of fine tailings particles with a high-molecular-weight (β ) and the collision (capture) efficiency (α ):
polymeric flocculant in aqueous solutions containing magnesium, in- Qi, j = βi, j αi, j (2)
cluding natural seawater, for a wide range of pH. The chosen flocculant
is commonly used in solid-liquid separation in the mineral processing
industry. The aim is to study the effect of magnesium hydroxide, which 2.2. Collision frequency β
is formed at a pH ca. 10, and the hydroxide-polymer interaction on the
kinetics of flocculation and the impact on the parameters of the model Is has been shown that fluid flow can penetrate through particle
and its predictive capability. The ultimate goal is to provide new insight aggregates [35]; this means that the actual collision frequency is

2
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

0.04 0.04
CaCl2 pH 7.4 MgCl2 pH 7.0
pH 8.5 pH 8.4
0.03 pH 9.7 0.03 pH 9.3
Volume fraction, V

Volume fraction, V
pH 10.6 pH 10.4
pH 11.5 pH 11.7

0.02 0.02

0.01 0.01

0.00 0.00
1 10 100 1 10 100

Chord length [μm], lc Chord length [μm], lc

(a) (b)
0.04

Seawater pH 7.4
pH 8.2
0.03 pH 9.2
Volume fraction, V

pH 10.2
pH 11
pH 12
0.02

0.01

0.00
1 10 100

Chord length [μm], lc

(c)
Fig. 1. Normalized initial volume distribution of particles for different pH. In 0.01 M CaCl2 (a), 0.05 M MgCl2 (b), and seawater (c).

considerably lower than that predicted from rectilinear flow models. To where C is a packing coefficient, it is assumed in this work that its value
add the effect of permeability of the aggregates, the so-called fluid is 1, as is common in the literature, and d 0 is the primary particle dia-
collection efficiency parameter η is included. η is the ratio between the meter. di and d 0 are related by the expression proposed by Mandelbrot
flow that passes through an aggregate and the undisturbed flow that [39]
approaches the same aggregate. Veerapaneni and Wiesner [36] pro-
1
posed Eq. (3) to calculate the collision frequency including permeability 2i − 1 ⎞df
and fractal dimension of the aggregates. di = d 0 ⎛ ⎜ ⎟

⎝ C ⎠ (7)
1
βi, j = ( ηi di + ηj dj )3G Collision frequency by gravity sedimentation is neglected because
6 (3)
densities are similar.
where di and dj are the diameter of the aggregates of size i and j re-
spectively, G is the shear rate and ηi is derived from the Brinkman’s
extension to Darcy’s law in function of a dimensionless permeability ξ : 2.3. Collision efficiency α
9(ξi − tanhξi )
ηi = There are several expressions for collision efficiency depending on
(38ξi + 2ξi 2 − 3tanhξi ) (4)
the type of aggregate and polymer used. Here, the depletion of ad-
where ξi = di /2 Ki and Ki the permeability, for the latter the expression sorbed high molecular weight polymer and its reorganization on the
from Li and Logan [37] is used, that is, surface of the particles is represented by an exponential decay between
two fitted parameters, as in the recent work of Vajihinejad & Soares
di2 ⎛ 3 8
Ki = ⎜3 + − 3 − 3 ⎞⎟ [32], that is,
72 ⎝ 1 − ϕi 1 − ϕi ⎠ (5)
α = (αmax − αmin ) e−kd t + αmin (8)
The porosity ϕ is related to the fractal dimension of the aggregates
by the expression of Vainshtein et al. [38], that is, where αmax is the maximum collision efficiency, αmin is the minimum
df − 3
collision efficiency at steady state conditions and kd is the collision
d efficiency decay constant.
ϕi = 1 − C ⎛ i ⎞ ⎜ ⎟

d
⎝ 0⎠ (6)

3
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

Aggregate diameter [μm],<dagg>

30 220

Particle diameter [μm], <d0>

Seawater
Seawater MgCl2 0.05 M
MgCl2 0.05 M 200
25 CaCl2 0.01 M
CaCl2 0.01 M
180
20
160
15
140

10 120
7 8 9 10 11 12 7 8 9 10 11 12
pH pH
(a) (b)

8 2.9
Settling velocity [m/h], Uh

7 2.8

Fractal dimension, df
6 2.7
5 2.6
4 2.5
3 Seawater 2.4
2 MgCl2 0.05 M 2.3 Seawater
MgCl2 0.05 M
1 CaCl2 0.01 M 2.2
CaCl2 0.01 M
0 2.1
7 8 9 10 11 12 7 8 9 10 11 12
pH pH
(c) (d)
Fig. 2. Experimental data needed to determine fractal dimension of tailings particles (a) mean primary particle diameter, (b) mean aggregate diameter, (c) settling
velocity, and (d) fractal dimension for a broad range of pH in several saline solutions. Mixing time for (b)–(d) is 30 s.

2.4. Breakage kernel and fragment distribution function calculated from

1
The S term in Eq. (1) is the fragmentation kernel and the Γ term is ερ 2
G = ⎜⎛ sus ⎟⎞
the breakage distribution function. The term S is difficult to predict ⎝ μsus ⎠ (11)
since there is no theory available, common is to fit it to an experi-
mentally determined size distribution. Here, the power law function for where ε is the average energy dissipation rate
the aggregate mass proposed by Pandya and Spielman [40] is used, that
Np N 3D5
is, ε=
V (12)
Si = s1 G s2 di (9)
where Np is the impeller power number (0.6 in our case for a plane disk
where G is the spatially-averaged share rate, di is the aggregate dia- with gentle agitation [42]), N is the rotation speed, D and V are re-
meter, s1 and s2 are parameters dependent on specific experimental spectively the diameter of the impeller and the working volume of the
conditions, and s2 is dimensionless. Eq. (9) appears frequently in the vessel. The density of the suspension ρsus is calculated from
flocculation literature, in a recent paper Hellmuth et al. [41] used it in
−1
the analysis of shear-induced platelet aggregation and breakup in the w 1 − w⎞
ρsus = ⎜⎛ + ⎟
early stages of thrombosis. ⎝ ρs ρw ⎠ (13)
There are several ways to define the breakage distribution function
Γ . Here we use a binary breakage function that is easy to implement and where w is the mass solid fraction of the solution, ρs and ρw are re-
requires no additional parameters, that is, spectively the solid and water density. Finally, the viscosity of the so-
lution μsus is measured.
Vj
⎧ V forj = i + 1
Γi, j = i
⎨ 0forj ≠ i + 1 2.6. Fractal dimension
⎩ (10)

Heath et al. [43] related the hindered settling velocity to the mean
2.5. Shear rate aggregate size and the effect of aggregate porosity was taken into ac-
count by using fractal geometry. According to these authors the settling
The shear rate required by the aggregation and breakage kernels is velocity is described by

4
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

250 dimension of the aggregates considering its three-dimensionality,

Mean aggregate diameter [μm], d
CaCl2 0.01 M pH=7.4 which is not entirely possible through microphotographs. Experimental
8.5 evidence for concentrated slurries without aggressive mixing shows
200
9.7 that the fractal dimension of the aggregates in the flocculation process
10.6 of mineral particles is close to constant in the time of experimentation
150 11.5 [19,43].

100
2.7. Model solution

50 To solve the stiff ODEs in Eq. ation (1) we use ode15s from Matlab, a
solver based on numerical differentiation formulas. The numerical so-
lution of the PBE model requires adjustment of parameters, which are
0
0 50 100 150 200 250 obtained from minimizing the objective function (FO) in Eq. (16).
Time [s], t tf
FO (αmax , αmin, kd, s1, s2) = ∑ti (〈dagg, exp 〉 − 〈dagg, mod 〉)2
(16)
(a)
250 where the volume-weighted mean aggregate diameter is
MgCl2 0.05 M
Mean aggregate diameter [μm], d

pH=7.0 max
∑i = 1 Ni di4
8.4 〈dagg 〉 =
200 max
∑i = 1 Ni di3
9.1 (17)
10.1
150 10.5 and exp stands for experimental and mod for PBM determined particle
11.7 diameters. The summation runs over all the size classes. To solve Eq.
(16), we use the MATLAB function fminsearch, which uses the Nelder-
100 Mead direct search to find the minimum of an unconstrained multi-
variable function.
50 Two criteria validate the model fit and predictions, one is the
coefficient of determination (R2 ) that measure the closeness of the
model values to the experimental values, that is,
0
0 50 100 150 200 250 max
∑i = 1 (dagg, exp, i − dagg, mod, i )2
Time [s], t R2 = 1 − max
∑i = 1 (dagg, exp, i − 〈dagg, exp 〉)2 (18)
(b)
220 and the other is the goodness of fit
Mean aggregate diameter [μm], d

200 Seawater pH=7.4 〈dagg,exp 〉 − std

180 8.2 GoF (%) = 100
〈dagg,exp 〉 (19)
160 9.2
10.2
140 std stands for the standard error calculated from
11
120 12 1
1 tf 2
100 std = ⎜⎛ ∑ti (〈dagg, exp 〉 − 〈dagg, mod 〉)2⎞⎟
80 ⎝ n − f ⎠ (20)
60
where n is the number of data values and f is the number of parameters
40 to be fitted. A GoF of 90% or higher means the proposed model is able
20 to predict the flocculation kinetics.
0 Finally, conservation of the total volume of particles is verified after
0 50 100 150 200 250 every integration to ensure that the simulations are maintaining the
Time [s], t particle population.
(c)
Fig. 3. Experimental (symbols) vs. calculated (continuous lines) time evolution 3. Materials and methods
of the mean diameter of aggregates of tailings particles as a function of pH in (a)
0.01 M CaCl2 solution, (b) 0.05 M MgCl2 solution and (c) seawater. Constant 3.1. Materials
shear rate 200 s−1.
High purity kaolinite (Ward’s Science) with density of 2.6 g/cm3
〈dagg 〉 ⎞ df − 3 3 − df 4.65 was used. Quartz particles (Sigma-Aldrich) with shape close to sphe-
Uh =
g (ρs − ρw )
〈dagg 〉2 ⎜⎛
⎛ ⎛ 〈dagg 〉 ⎞ ⎞
rical, density of 2.6 g/cm3, and average diameter of 24 μm were used.
18μsus
⎟
⎜1 − ϕ ⎜ 〈d 〉 ⎟ ⎟
⎝ 〈d 0 〉 ⎠ ⎝ ⎝ 0 ⎠ ⎠ (14) The flocculant used, supplied by SNF Chile S.A, was an anionic poly-
acrylamide (SNF 704) with a high molecular weight of 18 MDa. The
where g is the gravity acceleration, 〈d 0 〉and 〈dagg 〉are mean diameters of
flocculant was prepared at 1 g/l (0.1 wt%) gently mixing for 24 h and
primary particles and aggregates respectively. ϕ is the volume fraction
used within two weeks to avoid aging effects. Once a day, a flocculant
of primary particles given by
aliquot of the polymer solution was diluted at 0.1 g/l (0.01 wt%) for use
wρsus in daily tests. CaCl2 and MgCl2 salts used were analytical grade (Merck).
ϕ=
ρs (15) Solutions were prepared in distilled water (specific conductivity < 2
µS/cm and pH 5–7 at 25 °C). Natural seawater from Antofagasta in
If the data is available, Eq. (14) can be used to determine the fractal Chile was used after filtering.

5
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

100 1.0

80 0.8

GoF [%]
60 0.6

R2
Seawater Seawater
40 0.4
MgCl2 0.05M MgCl2 0.05M
CaCl2 0.01M CaCl2 0.01M
20 0.2

0 0.0
6 7 8 9 10 11 12 13 6 7 8 9 10 11 12 13
pH pH
(a) (b)
Fig. 4. Goodness of fit (a) and coefficient of determination (b).

0.012 0.012
Seawater Seawater
0.010 MgCl2 0.05 M 0.010 MgCl2 0.05 M
CaCl2 0.01 M CaCl2 0.01 M
0.008 0.008
αmax

αmin

0.006 0.006

0.004 0.004

0.002 0.002

0.000 0.000
7 8 9 10 11 12 7 8 9 10 11 12
pH pH

0.30
Seawater
0.25 MgCl2 0.05 M
CaCl2 0.01 M
0.20
kd [s-1]

0.15

0.10

0.05

0.00
7 8 9 10 11 12
pH
Fig. 5. Optimum aggregation parameters vs pH in different salt solutions.

3.2. Parameters and conditions performed by placing the stirrer 20 mm above the vessel bottom. A
volume of 232.2 ml of salt solution, either 0.01 M CaCl2, 0.05 M MgCl2
Table 1 summarizes parameters and conditions used in this study. or seawater (ionic strength 0.5 M), was added to the vessel, followed by
With these conditions and the viscosity of the different solutions, the addition of 27 g of a mixture of 92.3 wt% of quartz and 7.7 wt% of
5 × 10−3 kg/(m s), a shear rate of 200 s−1 was obtained for all the as- kaolinite as a dry powder. The total solids in the suspension reached
says. 10% by weight. The suspension was preconditioned by mixing at
450 rpm for 2 h. Then, the mixing rate was changed to 170 rpm and
kept constant, and 10.8 ml of diluted flocculant (equivalent to 20 g/
3.3. Flocculation assays
ton) were added. Temperature was ambient close to 25 °C. For in-situ
analysis of aggregate formation and size evolution, a Mettler Toledo
Flocculation experiments were conducted using a PTFE-coated
(Particle Track E25) focused beam reflectance measurement (FBRM)
30 mm turbine type stirrer, placed in axial position in a vessel of
instrument was used. The FBRM software provides chord length
100 mm in diameter and 1 l capacity. All the experiments were

6
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

0.012 0.012
CaCl2 0.01 M 7.4 MgCl2 0.05 M 7.0
0.010 8.5 0.010 8.4

Collision efficiency, α

Collision efficiency, α
9.7 9.1
0.008 10.6 0.008 10.1
11.5 10.5
0.006 0.006
11.7
0.004 0.004

0.002 0.002

0.000 0.000
0 50 100 150 200 250 0 50 100 150 200 250
Time [s], t Time [s], t
(a) (b)
0.012
Seawater 7.4
0.010 8.2
Collision efficiency, α

9.2
0.008 10.2
11.0
0.006
12.0
0.004

0.002

0.000
0 50 100 150 200 250
Time [s], t
(c)
Fig. 6. Time evolution of aggregation of tailings particles in 0.01 M CaCl2 (a), 0.05 M MgCl2 (b), and seawater (c) for a broad range of pH.

0.010 3.0 Fig. 7. Optimum breakage rate parameters s1 and s2

vs pH in different salt solutions.
Seawater 2.5
0.008 MgCl2 0.05 M
CaCl2 0.01 M 2.0
0.006
1.5
1

s2

0.004
1.0 Seawater
MgCl2 0.05 M
0.002 0.5 CaCl2 0.01 M

0.000 0.0
7 8 9 10 11 12 7 8 9 10 11 12
pH pH
(a) (b)

distributions in the form of histograms with user-defined intervals weighted chord lengths are comparable to conventional sizing over the
(channels). In this study, the data in a typical measurement were col- range from 50 to 400 µm. The FBRM probe was placed directly and
lected in the range of 1 µm to 1 mm for 2 s. Size distribution of primary vertically in the flocculation vessel, 10 mm above the stirrer and 20 mm
particles corresponded to the average of ten consecutive measurements. off-axis.
Size distribution of aggregates in the presence of flocculant was mea- Batch settling tests were conducted after 30 s of flocculant-suspen-
sured only one time. The square-weighted average sizes from FBRM are sion mixing in closed cylinders of 300 cm3 (35 mm internal diameter),
independent of suspension bulk flow velocity, laser focal position, and by gently pouring the slurry into the vessel once the mixing is over, and
solid fraction in the range 0.1–20% w/v. The latter was key to opt for then slowly inverting the cylinder two times by hand. The change of the
FBRM, instead of laser diffraction, given the high solid concentrations mud line, defined as the vertical position of the suspension-supernatant
in the pulps of interest. FBRM provides a distribution of particle chords, interface, was determined for one hour. Size of primary particles and
unlike a conventional distribution based on diameters, however Heath aggregates were also measured by FBRM.
et al. (2002) proved that the mean or mode averages of the square-

7
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

103 4. Results
Fragmentation rate [s ], Si
7.4 4.1. Initial particle size
102 8.5
-1

9.7 Solving the PBM equations for a given suspension requires the initial
101 number concentration of particles for each size class and per unit vo-
10.6 lume of suspension (N0, i ). Here we obtain such a distribution by a
100 11.5 transformation of the data from the FBRM from a volume distribution to
a number distribution according to equation (21),

10-1 v (d 0, i )
N0, i = ϕ
V0, i (21)
10-2
CaCl2 0.01 M where v (d 0, i ) is the experimental volume fraction of primary particles
10-3 with diameter d 0, i , obtained from Eq. (7), i is the size class. V0, i is the
volume of the primary particles following the geometric progression
0 5 10 15 20 25 30 Vi = 2i − 1V1. v (d 0, i ) is obtained by interpolating the data in Fig. 1 at the
Size interval, i required values of d 0, i . The initial particle diameter measured was 5 μ m,
the smallest primary particle from the volume distribution, the re-
(a) maining particle sizes were calculated from Eq. (7). Normalized volume
103 distributions are shown in Fig. 1 for a range of pH from ca. 7 to ca. 12 in
7.0 0.01 M CaCl2, 0.05 M MgCl2, and seawater.
Fragmentation rate [s ], Si

For all suspensions considered, at low pH ca. 10 the initial dis-

102 8.4
-1

tribution is skewed towards the smaller sizes, however at pH greater

9.1 than ca. 10 the distribution is skewed towards the larger sizes. The
101 10.1 change is slightly more gradual in CaCl2 than in MgCl2 (solution or
seawater) which is not strange considering that at pH 9.3 Mg(OH)2
10.5 precipitates (Kps = 1.8 × 10−11 at 25 °C) and thus at pH ≥ 9.3 the system
100 11.7 changes from tailings particles to tailings particles interspersed with a
gelatinous precipitate of magnesium hydroxide. In practice, massive
10-1 precipitation of the hydroxide occurs at pH 10.

10-2 MgCl2 0.05 M 4.2. Fractal dimension

10-3
To determine the fractal dimension, we simply follow the metho-
0 5 10 15 20 25 30 dology of Heath et al. [43], that is, minimizing the relative difference
Size interval, i between the experimental hindered settling velocity and the predicted
hindered settling velocity from Eq. (14). Data required includes hin-
(b) dered settling velocity, mean primary particle size, and mean aggregate
size. Fig. 2 summarizes the experimental results of the various settling
103 tests. In CaCl2 brines, the mean primary particle size increases slightly
7.4
Fragmentation rate [s ], Si

from 15 to 20 µm. At low pH, kaolinite agglomerates acquire card-

102 8.2 house structure but do not interact significantly with quartz, however at
-1

high pH the kaolinite disperses and both its surface charge and that of
9.2 quartz are neutralized by the Ca ions present giving rise to strong at-
101 10.2 tractions that lead to larger agglomerates. In MgCl2 brines and in sea-
11.0 water at pH ≥ 10 the aggregates grow less due to the presence of the
100 12.0
magnesium hydroxide gel that surrounds the quartz and kaolinite
particles [44]. The data in Fig. 2b also show a significant drop in mean
aggregate diameters and in settling velocity above pH 10 for seawater
10-1 and MgCl2 brines. This can be accounted by the presence of Mg(OH)2
that hinders aggregation. The resulting fractal dimension for MgCl2
10-2 brine and seawater is quite stable at pH < 10 with a representative
Seawater value of 2.7, typical of a clustered network, and also very stable at pH ≥
10-3 10 with a representative value of 2.2, typical of Gaussian chains. The
fractal dimension value of 2.7 in CaCl2 brine for the entire pH range is
0 5 10 15 20 25 30 adequate as an average value between 2.8 at the lower pH and 2.6 at
Size interval, i the higher pH. From these results, it is inferred that the appearance of a
new solid phase, magnesium hydroxide, at high pH involves new pat-
(c) terns for the growth of flocs, which lead to structures that are smaller
and weakly three-dimensional and therefore contribute much less to the
Fig. 8. Breakage rate vs aggregate diameter. In 0.01 M CaCl2 (a), 0.05 M MgCl2 settling velocity. It can be anticipated that the permeability of these
(b), and seawater (c).
aggregates will be greater than the truly three-dimensional aggregates
in the absence of Mg(OH)2.

8
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

103 103
102 7.4 102 7.0
101 101 8.4

Permeability [cm ], K

Permeability [cm ], K
8.5
100 9.7 100 9.1

2
10-1 10-1 10.1
10-2 10.6 10-2
11.5 10.5
10-3 10-3 11.7
10-4 10-4
10-5 10-5
10-6 10-6
10-7 10-7 MgCl2 0.05 M
10-8 CaCl2 0.01 M 10-8
10-9 10-9
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Size interval, i Size interval, i
(a) b)
103
102 7.4
101 8.2
Permeability [cm ], K

100 9.2
2

10-1 10.2
10-2 11.0
10-3 12.0
10-4
10-5
10-6
10-7
10-8 Seawater
10-9
0 5 10 15 20 25 30
Size interval, i
(c)
Fig. 9. Permeability as a function of aggregate size class in 0.01 M CaCl2 (a), 0.05 M MgCl2 (b), and seawater (c) for a broad range of pH.

4.3. Flocculation kinetics examination shows some order in the chaos. For example, in CaCl2
brine, except for the result at the lower pH of 7.4, all three parameters
Fig. 3 shows the effect of different brines and seawater, and broad are insensitive to the change in pH, they are clearly constant. We be-
pH range, on the flocculation kinetics of tailings particles with a high lieve that the problem at pH 7.4 is the quality of the fit of the mean size
molecular weight polyacrylamide at ambient temperature. Fig. 3 shows of aggregates in Fig. 3, which fails to correctly represent the experi-
experimental FBRM data versus results from our modeling. The ag- mental data at the highest size frequency due to the dispersion of the
gregate size first increases, reaches a maximum and then decreases to a data. The same occurs in MgCl2 brine and in seawater at pH < 10, the
stationary size. At short times the flocculant chains reaches the particle parameters are not sensitive to pH except for one or two points that
surfaces and dominates the aggregation process over the fragmentation could be better fitted with better data. In seawater, except for the result
of the aggregates. At longer times the aggregation-fragmentation at pH of 9.2, the parameters follow similar trends as in MgCl2 brine
equilibrium is attained and the particle aggregates reached a stable size; below and above pH 10. The problem at pH 9.2 is the model that over
the stationary aggregate size is considerably smaller than the maximum predicts the highest size frequency. Thus, the parameters of the ag-
size, this is typical of aggregation mediated by mechanisms of polymer gregation kernel for the system of tailings particles can be adequately
bridging. In brine of MgCl2 and in seawater at pH ≥ 10, the aggregates represented by αmax 0.0055, αmin 0.0025, and kd 0.025. In MgCl2 brine
develop smaller sizes because the precipitated magnesium hydroxide and in seawater at pH ≥ 10, magnesium hydroxide precipitates and the
hinders the action of the flocculant. Fig. 3 shows that the model cor- nature of the particulate system changes to tailings particles inter-
rectly describes the experimental data. The quality of the fits up to pH spersed with a hydroxide gel. Under this condition imposed by the
11.5 in every salt solution is better than 95% (GoF) and 0.95 (R2).(See hydroxide gel, the collision efficiency is expected to decrease, because
Fig. 4) the collision frequency is also expected to decrease. Fig. 5 shows that
the three parameters of the aggregation kernel change drastically, the
maximum and minimum collision efficiency acquire very low values,
4.4. Collision efficiency
practically zero, and the collision efficiency decay constant grows sig-
nificantly, indicating that the collision efficiency is practically zero. A
Of great interest is the effect of different brines and pH on the op-
set of aggregation parameters that can represent the system of tailings
timized parameters of the model. Fig. 5 shows the impact on the
particles and magnesium hydroxide is αmax 0.001, αmin 0 , and kd 0.2.
parameters of the aggregation kernel, that is, maximum and minimum
These results are particularly interesting because they suggest that the
collision efficiency (αmax , αmin ) and collision efficiency decay constant
aggregation parameters are largely constant for a particulate system
(kd ). At first glance the behavior of each of the parameters regarding the
that does not change its nature under diverse operating conditions, and
pH for each salt solution seems very erratic, however a close

9
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

Fig. 10. Time evolution of permeability of aggregates of tailings particles in 0.01 M CaCl2 (a), 0.05 M MgCl2 (b), and seawater (c) for a broad range of pH.

if its nature changes then the aggregation parameters will be different between the empirical parameter s1, the breakage frequency coefficient,
but constant. Fig. 6 supports these deductions, aggregation predictions and the shear rate is also linear. The parameters of Pandya & Spielman
with optimal parameters reveal that it remains in a narrow range for all [40] are different from ours, but Hellmuth et al. [41] used s2 = 2 to
saline solutions in the absence of magnesium hydroxide and that it is demonstrate that the aggregation efficiency can be modeled as a shear-
significantly reduced in the presence of magnesium hydroxide, even in independent constant when breakage is considered. The closeness of the
seawater It collapses to zero. In Fig. 6, the highest values for the col- value of s2 used by Hellmuth et al. [41] with the value in Fig. 7b could
lision efficiency are not always for the lowest pH, except for CaCl2; we be fortuitous; what matters, however, is that the parameters are con-
argue that it has to do with the fitting of flocculation kinetics to widely stant for the same system, which makes it possible to think of a PBM
dispersed data and not to a physicochemical effect. At any given pH, the that is much less intense in variables and easier to solve.
aggregation decreases slowly over time until reaching a stationary A more detailed analysis of s2 for 0.05 M MgCl2 and seawater shows
value, except in the presence of magnesium hydroxide in which the a change from ca. 2.5 to ca. 1.5 at pH between 9 and 11 which is the pH
constant value is reached rapidly. These results obtained by invoking range at which Mg(OH)2 precipitates and the system changes to tailings
Eq. (8) make sense due to flocculant depletion and its apparently irre- particles interspersed with a gelatinous precipitate of magnesium hy-
versible adhesion to the particles. It is interesting to note that the de- droxide. Thus, the value of s2 2.5 is sufficient to represent the breakage
crease in aggregation in MgCl2 solutions is gradual as the pH increases rate of tailings particles at pH < 10 while s2 1.5 is sufficient to re-
from 10, and very sharp in seawater. The effect of magnesium hydro- present the breakage rate of particles of tailings with magnesium hy-
xide in both brines is not the same, its impact on seawater seems en- droxide at pH ≥ 10.
hanced by the cooperative effect of other ions that compose it. The breakage rate in terms of aggregate size is shown in Fig. 8 for
particles of tailings in 0.01 M CaCl2 solution, 0.05 M MgCl2 solution and
seawater. Clearly the relation is linear. In 0.01 M CaCl2 at pH < 12
4.5. Breakage rate there are no solid species other than tailings particles, so at these
conditions the breakage rate grows with the size of the aggregates in-
The results obtained with the PBM on the flocculation process of dependent of pH between 7.4 and 11.5. In 0.05 M MgCl2 and in sea-
tailings particles show an interesting result (Fig. 7); the parameters s1, water the relation remains linear for each pH. The amplification factor
particularly, and s2 each maintain practically constant values suggesting is very much the same for all pH in MgCl2 solution and in seawater, and
that the breakage rate, for a specified shear rate, depends linearly on the exponent s2 is also the same for the two solutions up to a pH
the size of the aggregates, at least in the range of variables studied here. slightly < 10, at which s2 decays significantly. Thus, the value of the
This dependence was first found by Pandya and Spielman [40] for ag- breakage rate in MgCl2 solution decreases significantly only at pH 11.7,
gregates of kaolin-hydrous ferric oxide in agitated batch systems for a at pH > 10 and < 11.7 the presence of Mg(OH)2 does not significantly
specified shear rate. These authors also found that the relationship

10
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

103 103

Collision frequency [cm /s], βii

Collision frequency [cm /s], βii

102 7.4 102 7.0
101 8.5 101 8.4

3
100 9.7 100 9.1
10-1 10.6 10-1 10.1
10-2 11.5 10-2 10.5
10-3 10-3 11.7
10-4 10-4
10-5 10-5
10-6 10-6
CaCl2 0.01 M MgCl2 0.05 M
10-7 10-7
10-8 10-8
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Size interval, i Size interval, i
(a) (b)
103
Collision frequency [cm /s], βii

102 7.4
101 8.2
3

100 9.2
10-1 10.2
10-2 11.0
10-3 12.0
10-4
10-5
10-6
10-7 Seawater
10-8
0 5 10 15 20 25 30
Size interval, i
(c)
Fig. 11. Collision frequency as a function of particle size in 0.01 M CaCl2 (a), 0.05 M MgCl2 (b), and seawater (c) for a broad range of pH.

250 aggregate of tailings particles formed in the various salt solutions at

Mean aggregate diameter [μm], d

Seawater different pH in terms of their sizes. Results are shown in Fig. 9. Once the
model prediction pH 7.8 fractal dimension of the aggregates is defined, the calculation is simple
200 model prediction pH 8.4
model prediction pH 8.7
and direct, however the results reveal non-trivial characteristics of the
experimental data at pH 7.8 flow through the aggregates depending on the pH. For all the solutions
150 experimental data at pH 8.4
and pH considered, the largest aggregates are several orders of mag-
experimental data at pH 8.7
nitude more permeable, the singular thing is that at pH ≥ 10 in the
100 presence of Mg ions, in MgCl2 brine and in seawater, the permeability
of the larger agglomerates is even higher, for example, for aggregates of
size class 30 the permeability acquires values three orders of magnitude
50
higher. In contrast, in 0.01 M CaCl2 solution and in the absence of Mg
ions the permeability of the larger agglomerates shows a lower relative
0 increase with pH, for agglomerates of the size class 30 this increase is
0 50 100 150 200 250 approximately only an order of magnitude.
Time [s], t These results are a consequence of the truly tree-dimensional clus-
tered network of tailings particles that forms at pH < 10, which is
Fig. 12. Experimental (symbols) vs. calculated (continuous lines) time evolu-
highly permeable especially if the size of the aggregates is large, but not
tion of the mean diameter of aggregates of tailings particles as a function of pH
as much as the weakly three-dimensional structure that it is formed in
in seawater. Constant shear rate 200 s−1.
MgCl2 solutions and in seawater at pH ≥ 10 which is ultra-permeable.
Next, the temporal evolution of the permeability of the tailings
affect the breakage rate. In contrast, in seawater the breakage rate aggregates is determined by replacing the diameter of aggregates of
drops sharply as soon as pH 10 is reached. These results reinforce the given size class with the measured mean diameter of aggregates as a
idea that the impact of Mg(OH)2 generated in seawater is improved in function of pH in brines and seawater. Fig. 10 shows the results,
the presence of the other ions in seawater to distinctively decrease both naturally the permeability follows the evolution of the diameter of the
aggregation efficiency and breakage rate. aggregates, starts at zero, reaches a maximum in a short time, and then
decreases until reaching a stable value. Again, it is verified that the
4.6. Permeability aggregates of tailings particles interspersed with a gelatinous pre-
cipitate of magnesium hydroxide are more permeable than the
Eqs. (5) and (6) are used to estimate the permeability of the

11
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

aggregates of tailings particles and, broadly speaking, that the perme- calculations, revealed a novel result, the aggregation and breakage
ability is the highest when the aggregate size is the highest (see Fig. 8). parameters are to a large extent constant for a particulate system which
composition remains unchanged over a range of pH, and if the com-
4.7. Collision frequency position changes, for example by precipitation of magnesium hydro-
xide, then the aggregation parameters are different but close to con-
Fig. 11 shows the frequency of collision between particles of the stant. The predictive capacity of the model can be used as a support in
same size in different salt solutions for wide ranges of pH. The most practical decisions on the best operating conditions in the flocculation
frequent particle size according to Fig. 3 is ca. 200 µm in the absence of of tailings and water clarification.
Mg(OH)2 and ca. 140 µm in the presence of Mg(OH)2, which corre-
spond to size intervals between 10 and 15 in Fig. 11. In general, the Author contributions
frequency increases with the size of the aggregates, but it is clear that
for particles in size intervals ≤ 15 it maintains small values (< 10-3 The manuscript was written through the contributions of all au-
cm3/s). The high collision frequencies observed for large particles in the thors. JR, PR and RIJ designed and performed the experiments, GRQ
presence of Mg(OH)2 (pH > 10 in Fig. 11b and c) are not of great performed the calculations, RIJ, GRQ and PGT analyzed the results and
interest because these large particles are almost inexistent. At any given wrote the manuscript.
pH (pH ≥ 10) in the absence of magnesium the size of the aggregates,
or size interval, is much larger than the size that the aggregates reach in Declaration of Competing Interest
the presence of magnesium, see Fig. 2. Therefore, a correct reading of
Fig. 11 shows that for the same pH (≥10), the collision frequency be- The authors declare that they have no known competing financial
tween particles of the same size without magnesium is higher than in interests or personal relationships that could have appeared to influ-
the presence of magnesium. This result directly impacts the aggregation ence the work reported in this paper.
efficiency that decreases in the presence of Mg(OH)2 as anticipated in
section 4.4. Acknowledgements

4.8. Prediction capability The authors thank financial support from Centro CRHIAM Project
Conicyt/Fondap/15130015. RIJ also thanks CONICYT Fondecyt
To test the predictive capacity of PBM with the parameters de- 11171036 for financial support. Pedro Robles thanks the Pontificia
termined in the preceding sections, new experiments were carried out Universidad Católica de Valparaíso for the support provided.
to determine the kinetics of flocculation of tailings particles in seawater
at pH 7.8, 8.4 and 8.7. Pulps with the same composition and percentage References
of solids were considered as in the trials in section 4.3. The size of
primary particles is again 5 µm. Also, the same flocculant and the same [1] R. de Kretser, P.J. Scales, D.V. Boger, Improving clay-based tailings disposal: case
operating conditions are used, that is, temperature, flocculant dose, and study on coal tailings, AIChE J. 43 (1997) 1894–1903, https://doi.org/10.1002/aic.
690430724.
agitation time and intensity. Fig. 12 shows the experimental and [2] A.T. Owen, P.D. Fawell, J.D. Swift, D.M. Labbett, F.A. Benn, J.B. Farrow, Using
modeling results. The parameters for PBM are interpolated from Fig. 2d turbulent pipe flow to study the factors affecting polymer-bridging flocculation of
for the fractal dimension, Fig. 5 for the aggregation kernel and Fig. 7 for mineral systems, Int. J. Miner. Process. 87 (2008) 90–99, https://doi.org/10.1016/
j.minpro.2008.02.004.
the breakage rate for each pH. The adjustments are generally good with [3] L. Burlamacchi, M.F. Ottaviani, E.M. Ceresa, M. Visca, Stability of colloidal TiO2 in
GoF of 87.36 for pH 7.8, 93.38 for pH 8.4, and 93.01 for pH 8.7, which the presence of polyelectrolytes and divalent metal ions, Colloids Surf. 7 (1983)
demonstrates the predictive capacity of PBM. From the modeling point 165–182, https://doi.org/10.1016/0166-6622(83)80045-6.
[4] L. Nabzar, E. Pefferkorn, R. Varoqui, Polyacrylamide-sodium kaolinite interactions:
of view there are no important differences in the kinetics at one or the flocculation behavior of polymer clay suspensions, J. Colloid Interface Sci. 102
other pH, but the truth is that there are also no important differences in (1984) 380–388, https://doi.org/10.1016/0021-9797(84)90240-6.
the experimental kinetics at the different pH. That is why we also adjust [5] R.I. Jeldres, E.C. Piceros, W.H. Leiva, P.G. Toledo, N. Herrera, Viscoelasticity and
yielding properties of flocculated kaolinite sediments in saline water, Colloids
the experimental kinetics with the average parameters defined in sec-
Surfaces A Physicochem. Eng. Asp. 529 (2017) 1009–1015, https://doi.org/10.
tions 4.2, 4.4 and 4.5 for pH < 10 (without Mg(OH)2), that is, df 2.7 1016/j.colsurfa.2017.07.006.
for the fractal dimension, αmax 0.0055, αmin 0.0025, and kd 0.025 for the [6] G.R. Quezada, R.I. Jeldres, P.D. Fawell, P.G. Toledo, Use of molecular dynamics to
aggregation kernel, and s1 0.0007 and s2 2.5 for the breakage rate. The study the conformation of an anionic polyelectrolyte in saline medium and its ad-
sorption on a quartz surface, Miner. Eng. 129 (2018) 102–105, https://doi.org/10.
results are confused with the adjustments in Fig. 12 and are therefore 1016/j.mineng.2018.09.025.
not included. [7] T. Hirajima, G.P.W. Suyantara, O. Ichikawa, A.M. Elmahdy, H. Miki, K. Sasaki,
Effect of Mg2+ and Ca2+ as divalent seawater cations on the floatability of mo-
lybdenite and chalcopyrite, Miner. Eng. 96–97 (2016) 83–93, https://doi.org/10.
5. Conclusions 1016/j.mineng.2016.06.023.
[8] R.I. Jeldres, M.P. Arancibia-Bravo, A. Reyes, C.E. Aguirre, L. Cortes, L.A. Cisternas,
The flocculation of fine particles of tailings with a high-molecular- The impact of seawater with calcium and magnesium removal for the flotation of
copper-molybdenum sulphide ores, Miner. Eng. 109 (2017) 10–13, https://doi.org/
weight anionic polymer flocculant in aqueous salt solutions, including 10.1016/j.mineng.2017.02.003.
natural seawater, revealed that in the presence of ionic magnesium at [9] J. Leentvaar, M. Rebhun, Effect of magnesium and calcium precipitation on coa-
pH ≥ 10 the aggregates grow less due to the formation of magnesium gulation-flocculation with lime, Water Res. 16 (1982) 655–662, https://doi.org/10.
1016/0043-1354(82)90087-2.
hydroxide gel that surrounds quartz, kaolinite and flocculant. The [10] M. Megersa, W. Gach, A. Beyene, A. Ambelu, L. Triest, Effect of salt solutions on
fractal dimension of the tailings aggregates was ca. 2.7 at pH < 10, coagulation performance of Moringa stenopetala and Maerua subcordata for turbid
typical of a clustered network, although in the presence of magnesium water treatment, Sep. Purif. Technol. 221 (2019) 319–324, https://doi.org/10.
1016/j.seppur.2019.04.013.
at pH ≥ 10 the fractal dimension of the aggregates was only 2.2, typical
[11] L. Semerjian, G.M. Ayoub, High-pH–magnesium coagulation–flocculation in was-
of Gaussian chains. Smaller and weakly three-dimensional aggregates in tewater treatment, Adv. Environ. Res. 7 (2003) 389–403, https://doi.org/10.1016/
the presence of magnesium hydroxide contributed very little to the S1093-0191(02)00009-6.
settling velocity. The PBM model used followed the dynamics of the [12] D. Vandamme, I. Foubert, I. Fraeye, B. Meesschaert, K. Muylaert, Flocculation of
Chlorella vulgaris induced by high pH: Role of magnesium and calcium and prac-
aggregation and breakage processes underlying particle flocculation tical implications, Bioresour. Technol. 105 (2012) 114–119, https://doi.org/10.
thus providing a good approximation to the temporal evolution of ag- 1016/j.biortech.2011.11.105.
gregate size. The parameters of the model, which were obtained by [13] L. Dupont, A. Foissy, R. Mercier, B. Mottet, Effect of calcium ions on the adsorption
of polyacrylic aid onto alumina, J. Colloid Interface Sci. 161 (1993) 455–464,
minimizing the difference between experimental size data and model

12
G.R. Quezada, et al. Separation and Purification Technology 237 (2020) 116319

https://doi.org/10.1006/jcis.1993.1489. [29] M.G. Rasteiro, F.A.P. Garcia, P. Ferreira, A. Blanco, C. Negro, E. Antunes, The use of
[14] F.F. Peng, P. Di, Effect of multivalent salts-calcium and aluminum on the floccu- LDS as a tool to evaluate flocculation mechanisms, Chem. Eng. Process. Process
lation of kaolin suspension with anionic polyacrylamide, J. Colloid Interface Sci. Intensif. 47 (2008) 1323–1332, https://doi.org/10.1016/j.cep.2007.04.009.
164 (1994) 229–237, https://doi.org/10.1006/jcis.1994.1161. [30] C. Selomulya, G. Bushell, R. Amal, T.D. Waite, Understanding the role of re-
[15] L. Liang, Y. Peng, J. Tan, G. Xie, A review of the modern characterization techni- structuring in flocculation: the application of a population balance model, Chem.
ques for flocs in mineral processing, Miner. Eng. 84 (2015) 130–144, https://doi. Eng. Sci. 58 (2003) 327–338, https://doi.org/10.1016/S0009-2509(02)00523-7.
org/10.1016/j.mineng.2015.10.011. [31] A.L. Ahmad, M.F. Chong, S. Bhatia, Population Balance Model (PBM) for floccu-
[16] C. Biggs, Activated sludge flocculation: on-line determination of floc size and the lation process: simulation and experimental studies of palm oil mill effluent (POME)
effect of shear, Water Res. 34 (2000) 2542–2550, https://doi.org/10.1016/S0043- pretreatment, Chem. Eng. J. 140 (2008) 86–100, https://doi.org/10.1016/j.cej.
1354(99)00431-5. 2007.09.014.
[17] Y. Sang, A. Lin, X. Liu, Population balance modeling of cationic polyacrylamide [32] V. Vajihinejad, J.B.P. Soares, Monitoring polymer flocculation in oil sands tailings:
(CPAM) induced flocculation process for lignin recovery from the pre-hydrolysis a population balance model approach, Chem. Eng. J. 346 (2018) 447–457, https://
liquor of kraft pulping process, Sep. Purif. Technol. 221 (2019) 152–158, https:// doi.org/10.1016/j.cej.2018.04.039.
doi.org/10.1016/j.seppur.2019.03.090. [33] B.J. Lee, E. Toorman, F.J. Molz, J. Wang, A two-class population balance equation
[18] D.N. Thomas, S.J. Judd, N. Fawcett, Flocculation modelling: a review, Water Res. yielding bimodal flocculation of marine or estuarine sediments, Water Res. 45
33 (1999) 1579–1592, https://doi.org/10.1016/S0043-1354(98)00392-3. (2011) 2131–2145, https://doi.org/10.1016/j.watres.2010.12.028.
[19] A.R. Heath, P.A. Bahri, P.D. Fawell, J.B. Farrow, Polymer flocculation of calcite: [34] P.T. Spicer, S.E. Pratsinis, Coagulation and fragmentation: universal steady-state
Population balance model, AIChE J. 52 (2006) 1641–1653, https://doi.org/10. particle-size distribution, AIChE J. 42 (1996) 1612–1620, https://doi.org/10.1002/
1002/aic.10749. aic.690420612.
[20] R.I. Jeldres, F. Concha, P.G. Toledo, Population balance modelling of particle [35] A. Thill, S. Moustier, J. Aziz, M.R. Wiesner, J.Y. Bottero, Flocs restructuring during
flocculation with attention to aggregate restructuring and permeability, Adv. aggregation: experimental evidence and numerical simulation, J. Colloid Interface
Colloid Interface Sci. 224 (2015) 62–71, https://doi.org/10.1016/j.cis.2015.07. Sci. 243 (2001) 171–182, https://doi.org/10.1006/jcis.2001.7801.
009. [36] S. Veerapaneni, M.R. Wiesner, Hydrodynamics of fractal aggregates with radially
[21] M. v. Smoluchowski, Versuch einer mathematischen Theorie der varying permeability, J. Colloid Interface Sci. 177 (1996) 45–57, https://doi.org/
Koagulationskinetik kolloider Lösungen, Zeitschrift Für Phys. Chemie. 92U (1918). 10.1006/jcis.1996.0005.
doi: 10.1515/zpch-1918-9209. [37] X.-Y. Li, B.E. Logan, Permeability of fractal aggregates, Water Res. 35 (2001)
[22] R.I. Jeldres, P.D. Fawell, B.J. Florio, Population balance modelling to describe the 3373–3380, https://doi.org/10.1016/S0043-1354(01)00061-6.
particle aggregation process: a review, Powder Technol. 326 (2018) 190–207, [38] P. Vainshtein, M. Shapiro, C. Gutfinger, Mobility of permeable aggregates: effects of
https://doi.org/10.1016/j.powtec.2017.12.033. shape and porosity, J. Aerosol Sci. 35 (2004) 383–404, https://doi.org/10.1016/j.
[23] V. Runkana, P. Somasundaran, P.C. Kapur, A population balance model for floc- jaerosci.2003.09.004.
culation of colloidal suspensions by polymer bridging, Chem. Eng. Sci. 61 (2006) [39] B.B. Mandelbrot, Self-affine fractals and fractal dimension, Phys. Scr. 32 (1985)
182–191, https://doi.org/10.1016/j.ces.2005.01.046. 257–260, https://doi.org/10.1088/0031-8949/32/4/001.
[24] A. Datta, R.K. Rajamani, A direct approach of modeling batch grinding in ball mills [40] J.D. Pandya, L.A. Spielman, Floc breakage in agitated suspensions: Effect of agita-
using population balance principles and impact energy distribution, Int. J. Miner. tion rate, Chem. Eng. Sci. 38 (1983) 1983–1992, https://doi.org/10.1016/0009-
Process. 64 (2002) 181–200, https://doi.org/10.1016/S0301-7516(01)00044-8. 2509(83)80102-X.
[25] C. Kiparissides, A. Alexopoulos, A. Roussos, G. Dompazis, C. Kotoulas, Population [41] R. Hellmuth, M.S. Bruzzi, N.J. Quinlan, Analysis of shear-induced platelet ag-
balance modeling of particulate polymerization processes, Ind. Eng. Chem. Res. 43 gregation and breakup, Ann. Biomed. Eng. 44 (2016) 914–928, https://doi.org/10.
(2004) 7290–7302, https://doi.org/10.1021/ie049901x. 1007/s10439-015-1409-1.
[26] M.J. Hounslow, R.L. Ryall, V.R. Marshall, A discretized population balance for [42] C. Pretorius, E. Wicklein, T. Rauch-Williams, R. Samstag, C. Sigmon, How oversized
nucleation, growth, and aggregation, AIChE J. 34 (1988) 1821–1832, https://doi. mixers became an industry standard, Proc. Water Environ. Fed. 11 (2015)
org/10.1002/aic.690341108. 4379–4411.
[27] K.A. Kusters, S.E. Pratsinis, S.G. Thoma, D.M. Smith, Ultrasonic fragmentation of [43] A.R. Heath, P.A. Bahri, P.D. Fawell, J.B. Farrow, Polymer flocculation of calcite:
agglomerate powders, Chem. Eng. Sci. 48 (1993) 4119–4127, https://doi.org/10. relating the aggregate size to the settling rate, AIChE J. 52 (2006) 1987–1994,
1016/0009-2509(93)80258-R. https://doi.org/10.1002/aic.10789.
[28] T. Li, Z. Zhu, D. Wang, C. Yao, H. Tang, The strength and fractal dimension char- [44] S. El-Manharawy, A. Hafez, Study of seawater alkalization as a promising RO pre-
acteristics of alum–kaolin flocs, Int. J. Miner. Process. 82 (2007) 23–29, https://doi. treatment method, Desalination 153 (2003) 109–120, https://doi.org/10.1016/
org/10.1016/j.minpro.2006.09.012. S0011-9164(02)01110-4.

13