Chemical Engineering Science 66 (2011) 982–997

Contents lists available at ScienceDirect

Chemical Engineering Science

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

A new experimental method for determining particle capture efficiency

in flotation
Z. Huang a,b,c,d,e, D. Legendre d,e, P. Guiraud a,b,c,
a
Université de Toulouse; INSA, UPS, INP; LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France
b
INRA, UMRA792 Ingénierie des Syste mes Biologiques et des Procédés, F-31400 Toulouse, France
c
CNRS, UMR5504, F-31400 Toulouse, France
d
Université de Toulouse; INPT, UPS; IMFT (Institut de Mécanique des Fluides de Toulouse) Allée Camille Soula, F-31400 Toulouse, France
e
CNRS, IMFT, F-31400 Toulouse, France

a r t i c l e i n f o a b s t r a c t

Article history: A single bubble experiment has been developed for the determination of the capture efficiency of particles
Received 6 August 2010 by bubbles in flotation under well-controlled hydrodynamics and physico-chemical conditions. In a glass
Received in revised form column, small single bubbles (db ¼0.22  1.16 mm) are produced in pure water and then rise at their
25 November 2010
terminal velocity through a suspension consisting of spherical glass particles ðdp ¼ 5256 mmÞ where
Accepted 8 December 2010
Available online 15 December 2010
bubble–particle capture takes place. The capture efficiency Ecapt is calculated as the ratio of the number of
particles captured by one bubble to the number of particles present in the volume swept out by this
Keywords: bubble. Images recorded at high optical magnification show that particles slip on the interface, then
Flotation adhere to air bubbles individually or as aggregates and cover the rear part of bubble surface. The bubble’s
Particle
effective density and interface contamination level are increased by captured particles. As a result,
Bubble
bubble’s rising velocity Ub is reduced along the experimental device. By establishing the relationship
Interface
Hydrodynamic between capture efficiency Ecapt, bubble rise velocity Ub and bubble clean angle yclean , a new approach to
Capture efficiency measure particle–bubble capture efficiency is proposed. This new experimental technique is applied to
provide a new set of data for capture efficiency in the case of bubbles with a clean interface. Ecapt is found to
grow as db decreases and dp increases, within the range between 0.02 and 0.20, which is in the order of
magnitude of experimental results of Ralston and Dukhin (1999) as well as of numerical results of Sarrot
et al. (2005). These data are favorably compared to numerical modeling of collision efficiency.
& 2010 Elsevier Ltd. All rights reserved.

1. Introduction The key of the flotation process is the bubble–particle capture,

which has been generally treated as a sequence of three subprocesses
Originated in the mineral industry, flotation as an efficient separa- (Jameson et al., 1977; Schulze, 1989; Ralston et al., 1999). The total
tion process is actually used in various industries such as petrochem- capture efficiency is usually seen as the product of these three
ical refining, pulp and paper manufacturing, food processing, water successive steps, Ecapt ¼ Ecoll  Eatt  Estab , where Ecoll is the collision
treatment, liquid metal purification, etc. In most of these situations, efficiency, Eatt is the attachment efficiency and Estab the stability
the separation process is realized by creating swarms of tiny air efficiency of the bubble–particle aggregate. The overall capture
bubbles mechanically in a mixer or a sparger, or electrolytically, in a efficiency can be determined via two different types of experiment:
liquid tank or column containing the particles to be removed. Bubbles direct methods and indirect methods.
preferentially attach to hydrophobized particles and transport them Direct observation concerns recording particle’s trajectory as it
to the froth layer at the liquid surface, where they are removed by passes a captive bubble (Schulze, 1977; Nguyen and Kmet, 1992;
skimming devices. There are two principal classifications of flotation Nguyen and Evans, 2004). The particle’s releasing point is gradually
depending on the bubble size. In microbubble flotation, the bubbles shifted to the axis of the axisymmetric hydrodynamic flow passing the
are typically 30270 mm in diameter. They are often produced by bubble until the settled particles begin to collide with the bubble and
depressurization of dissolved air liquid (DAF). In dispersed air to attach to the interface. The distance between the closest releasing
flotation, the bubbles are on the order of 1 mm in diameter. point and the axisymmetric axis is called the critical radius rc. Then the
capture efficiency can be calculated as the square of the radius ratio
Ecapt ¼(rc/rb)2. The direct measurement helps one to examine the
 Corresponding author at: Université de Toulouse, INSA, UPS, INP, LISBP, 135 whole capture process, for instance rebounding (Whelan and Brown,
Avenue de Rangueil, F-31077 Toulouse, France. 1955), sliding time (Schulze, 1977), or even liquid film rupture
E-mail address: [email protected] (P. Guiraud). (Nguyen and Evans, 2004). However, this method is limited to weak

0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.12.006
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 983

liquid flow in order to avoid bubble oscillation. Moreover, because the and Kitchener (1977), Dai et al. (1998), Hewitt et al. (1994) and
bubble is held at bottom, the liquid motion near the bubble’s surface is used by Sarrot et al. (2007).
modified. We can also imagine that it is quite difficult to determine if
the particle trajectory remains in the symmetry plane of the bubble,
and, as a consequence, that rc is not so easily measured. 2. Experimental methods
Indirect methods allow us to determine the overall collection
efficiency of the whole capture process. By this way, the physico- 2.1. Materials
chemical and hydrodynamics parameters (such as hydrophobicity,
nature and size distribution of the particles, gas and liquid flow The water used for all the experiments is produced by an
rates, etc.) should be well defined and controlled, in order to Aquasource ion exchange device. Tap water is filtered for all
identify the respective influence of different parameters on the particles larger than 1 mm and then de-ionised through two stages
elimination of given solid particles. This is why single bubble ion exchange resin. The conductivity measured by a WTW LR235/
experiments are widely adopted (Reay and Ratcliff, 1973; Dai et al., 01 conductivity sensor ranges between 0.75 and 2:8 mS cm1 . The
1998; Hu et al., 2003; Sarrot et al., 2007). Air bubbles are injected surface tension is of 72.8 mN m  1 at 20 1C measured by a GBX 3S-
one by one into the flotation column and then they rise through a Bal285 instrument. Bubble coalescence time with the water–air
suspension at rest. So the local flow field around the bubbles is not interface is less than 1 s, that indicates no significant surfactant
disturbed and is closer to the real situation in flotation machines contamination of the de-ionised water (Dai et al., 1998; Nguyen
where bubbles rise freely in the liquid. Capture efficiency is defined et al., 2006).
as the ratio of the number of captured particles over the number of Solid particles are spherical glass micro-beads of density
particles initially situated in the volume that the bubble sweeps off. 2363 kg m  3 supplied by Marteau et Lemarie Company. The
By using strongly hydrophobic particles, the efficiency of attach- particle size ranges between 5 and 75 mm. After sieving, their size
ment approximately equals unity. Moreover, since the external distribution, determined by laser light scattering technique (Mal-
fluid agitation is absent, at high particle hydrophobicity and vern Mastersizer 2000) is shown in Table 1. The volume weighted
at rather high electrolyte concentration, where the hydrophobe mean diameter is d p ¼ 22 mm and 90% particles are smaller than
force is strongly attractive and the repulsive electrostatic forces are 31 mm. The advancing contact angle ya for water was measured by
largely reduced due to the double layer compression mechanism, the Washburn test. It is shown that without any cleaning surface
the particle–bubble aggregate stability is controlled only by ther- treatment, ya ¼ 583 (for the original glass beads). This angle is larger
modynamics; the stability efficiency reaches its maximum value than the value usually found for a clean glass surface. The same
for rather small particles (Anfruns and Kitchener, 1977; Hewitt et al., particles were washed twice by de-ionized water and later cleaned
1994; Dai et al., 1998; Ralston et al., 2002). in aqua regia. The measured contact angle decreased to 501 and 441,
However, among the few single rising bubble experiments, air respectively, confirming the presence of surface contamination
bubbles with an immobile surface have traditionally been considered, coming from the manufacture processing. However, low contact
because traces of surfactive substances (frothers or collectors) are angle, like ya ¼ 441, is not favorable for the particles bubble capture
always present and effectively immobilize the bubble surface. Dai (Ralston, 1992). Since all experiments were carried out in de-
et al. (1998) have used Milli-Q water (conductivity less than 0.5 mS ionized water to avoid the desorption of the contaminant from the
cm  1) to ensure that experimental water is free of impurities, in order particle’s surface to the water and to keep the contact angle
to compare the experimental results to their collision model for constant, we decided to work with particles cleaned up by de-
bubble with mobile surface. Nevertheless, in their experiments, air ionized water.
bubbles are produced directly in the particle suspension, so the Calibrated air bubbles are produced by glass capillary tubes
bubble may be contaminated during the bubbling stage. It easily leads with different inner size, the smallest one having an internal
that the rise velocity measured may not be that of a clean bubble. In diameter of approximately 30 mm measured by microscopy (LEICA
fact, the bubble–liquid interface contamination by surfactant agents DM1000). The injection air flow rate is fixed by an automatic
or by the captured particles themselves plays a major role in the syringe pump (Harvard PHD2000), so that the frequency of the
capture efficiency due to the interface immobilization that modifies bubble production can be controlled. Direct numerical simulation
the flow field around the bubble, from clean bubble to fully (DNS) for millimetric clean bubbles has shown that a distance
contaminated bubble (behaving as a solid particle). Research on separating two successive bubbles greater than 50 times the
bubbles with a mobile surface is driven by the observation that the bubble’s diameter is sufficient to avoid inter-bubble significant
surface contamination at a rising bubble surface is a dynamic process. hydrodynamic disturbances that may come from the wakes (Sarrot
The surface contaminants are swept to the rear surface of the bubble et al., 2007). The air flow rate is fixed to ensure this distance. As an
by adjacent water, and, as a result, the forward part of bubble surface example, the bubble production frequency is about 5 bubbles per
can remain mobile, while an immobile or stagnant cap is formed at minute, when db ¼1 mm.
the rear. In previous papers, Sarrot et al. (2005) and Legendre et al.
(2009) investigated the collision efficiency for such interface by using 2.2. Flotation device
direct numerical simulation (DNS) of the flow fields around the
bubbles. It is shown that the collision behaviors are given by the flux of The single bubble experiments were carried out in a glass
particles near the surface which is controlled by the tangential column type flotation cell with an internal diameter of 5 cm and a
velocity for mobile interface and by the velocity gradient for immobile
interface. Table 1
In this work, the whole particle–bubble interaction is investi- Glass beads size fractions.
gated in a single-bubble glass flotation cell by using visualization
Size range ðmmÞ Volume fraction (%)
with a high speed CMOS image sensor with efforts to measure
bubble size and velocity. The originality of the method presented 5–15 17.6
here is to use the rate of change in the bubble rise velocity along the 15–25 61.8
bubble rising trajectory through the particle suspension to deduce 25–40 19.8
40–56 0.8
the capture efficiency. Results are compared to that of classical 56–75 o0.1
counting method derived from the technique proposed by Anfruns
984 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

height of 150 cm, as shown in Fig. 1. The column is divided into the particles in suspension. At the top of the column, a clean zone
three parts by two glass three-way taps specially constructed. The prevents particles carried by the wake of the rising bubble from
bottom of the column is a 10 cm high bubbling zone where the reaching the water surface, where the particles captured by the
injection capillary is installed. This section is always filled with bubble are released. A special device is used to collect the particles. It
clean water, so that bubbles are produced in clean water, instead of consists of an inclined cylindrical jar whose bottom is replaced by a
being directly created in the particles suspension. By this choice, conical funnel (see Fig. 2 in detail). When bubbles reach at top of the
we intend to avoid the capture of a number of particles during the conical funnel, solid particles carried by the bubbles are released at
bubbling process without knowing the amount. This is one of the the free surface. At the bottom of this device, pure water is injected
important differences compared to the apparatus developed by continuously at a low flow rate to prevent the particles from settling
Hewitt et al. (1995), already adopted in the experiment device and bring them to the outside zone of this recipient, which is
developed by Sarrot et al. (2005). The 10 cm height ensures that connected directly to a particle counter.
bubbles have enough time to achieve their terminal rise velocity
before entering into the suspension. Note that a db ¼ 1 mm air
bubble needs approximately 75 ms to go through 2 cm height of 2.3. Experimental procedure
water to reach 99% of its terminal velocity. The second part of the
device corresponds to the capture zone, where the bubbles encounter Before each experiment, the whole column is totally cleaned
with de-ionized water until the particle concentration is less than
18 part ml  1. As shown in Fig. 3, dilute (0.03 wt%) particle in water
suspension is prepared in a separated glass mixing tank, mean-
while the whole volume of the flotation cell is filled with de-ionized
water. Single air bubbles are generated at the bubbling zone with
the three-way taps in open position. Bubbling flow rate is set to the
operation condition and a series of images of the bubbles is
recorded in order to measure bubble diameter db and rise velocity
Ub. Next, the two three-ways taps are closed. Bubbles continue to be
released, but they are stored temporarily in a small cavity at bottom
of the three-ways tap instead of entering into the column. The
particle suspension is introduced by a vane pump (Johnson Pump
F3B-19). This suspension (105–5  105 part ml  1) is recirculated
between the capture zone and the preparation tank until a constant
concentration in particle number is measured by the particle
counter PCX-Model (MET ONE). Once the concentration of particle
suspension becomes stable, recirculation pump is stopped and the
three-ways taps are re-opened. Calibrated bubbles rise through
the capture zone and then pass the clean zone, finally arrive at the
water surface and release the captured particles. Water containing
the captured particles in the recipient is sent directly to another
particle counter (ChemTrac) and examinated.
A high speed CMOS camera (Photron APX, 2  105 im s  1,
1024  1024 pixels2) with a Nikon 105 mm lens was used to record
bubbles in the bubbling zone so as to obtain the size and the
terminal rise velocity. Another camera (PCO 1200 HS, 1357 im s  1,
1280  1024 pixels2) with a Nikon 55 mm lens was installed to
record the bubble rise velocity all along its passing through the
suspension. Pixel resolution was calibrated by imaging a ruler
located in the plane of the rising bubbles. The horizontal Dx and
vertical Dy lengths of the bubble are measured on the image and
bubble’s equivalent diameter dexp b is taken as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dexp
3
b
¼ ðDxÞ2 Dy: ð1Þ

particles released
at the water surface

to particle counter

low flowrate
water injection

Fig. 1. Experimental glass device. Fig. 2. Particle collecting system.

 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 985

Particle counter
Particle collecting system MET ONE

Collection section Pure water injection

(Clean water)
Three−Way Tap

Suspension
Capture zone Container
(Suspension)
video camera
PCO1200HS Particle
Counter Pump
ChemTrac
Three−Way Tap

Bubbling zone LightSource

(Clean water)
video camera Syringe pump
Photron APX Air Bubble Injection

Fig. 3. Experimental device for the measure of collection efficiency.

Measurement error on Dx and Dy is 2 pixels. A bubble of 1 mm in

diameter measuring 55 pixels, so the accuracy of the diameter
measurement is about 0.036 mm or 3.6%.
Bubble’s rise velocity Uexp
b is obtained from measuring the
distance Dh between the centers of gravity of bubble on successive
images. Noting f the frequency of image capture:

Ubexp ¼ Dh=Dt ¼ Dh  f : ð2Þ

Inaccuracy in the measurement of Uexp

comes from the mea-
b
surement of Dh. Measurement error for Dh is about 2 pixels. So for a
clean bubble of 1 mm in diameter, with a rise velocity of 300 mm s  1
and image capture frequency f¼100 image s  1, the measurement
error of Uexp is about 3.6 mm s  1 or 1.2%. Fig. 4. Bubble capture system using nylon lines.
b

2.4. Observation device

Some experiments have been performed for the purpose of

observing the interaction between an air bubble and micro-glass
beads. The experimental apparatus was initially designed to study
the mass transfer from a single bubble to the surrounding liquid
(Dani, 2007). In a rectangular Plexiglass column (10  20  60 cm3),
a special capture system constituted of four 80 mm diameter thin
Nylon lines is installed (Fig. 4). These lines (3) cross each other to
form a small diamond grid (4), where the bubble is blocked. The
screws on the left side (1) are used to stretch the threads, while the
screws on the right side (2) are used to adjust the size of the grid for
various bubble sizes (1–4 mm). Small air bubbles were released by
a needle just under the blocking device as shown in Fig. 5. A uniform
flow field was produced by falling down the pure water in stock
through a funnel. The water flow generated by gravity can provide
an average velocity up to 10 cm s  1. Above this value, the bubble
become unstable. Particle Image Velocimetry (PIV) measurements
have shown that the presence of the blocking system slightly
decrease the water velocity near the crossing points of the threads Fig. 5. Bubble blocked by the crossing nylon lines ðdb  3 mmÞ. Captured particles
(less than 15%). can be observed at the bottom.
Compared to the bubble fixing system used by Schulze (1977)
and later by Nguyen and Kmet (1992), in which bubble was held at system leaves free the bottom of the bubble where the layer of
the bottom, the capture grid blocked only bubble’s rising motion captured particles can be observed. The particles suspension was
but do not affect bubble’s interface mobility. Furthermore, this injected at the top of the apparatus far away from the captured
986 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

100
Microscope

Glass lamellar 200

Air bubble

Water 300

y [pixel]
Attached solid particles
400
Fig. 6. Visualization system at the top of the flotation cell.

bubble. The injection flow rate was very low in order to avoid the
disturbance of the local flow field or the bubble oscillation. The 500
whole particle–bubble capture process was recorded at high
optical magnification by using a 24X zoom coupled to a Nikon
105 mm lens and to the CMOS video camera PCO 1200HS. The
1
highest resolution of digital image was 5 mm pixel . The recorded 600
s
images were analyzed by image processing using Matlab in order
to obtain particle trajectory. Additional observations were per-
formed in the single bubble flotation cell. Instead of the four lines
bubble fixing system, a simple glass slice is placed at the upper 700
clean zone to block the bubble particles aggregates before visua-
lization with the help of a microscope (see Fig. 6).
600 650 700 750 800 850
x [pixel]
3. Preliminary studies Fig. 7. Trajectories of particles moving at Us ¼ 1.0 mm s  1 past a captive bubble
db ¼ 3.0 mm.
In this section, we present some preliminary investigations
concerning the bubble–particle interaction. Firstly, the formation
few particles (Fig. 8c); in some cases, a second layer of particles may
of the bubble–particle aggregates is discussed. Secondly, a simple
attach to the first one (Fig. 8d). But it is impossible to know if these
model of interface covering is proposed, which will be used in the
clusters agglomerate in the liquid before the attachment or at the
third part to link the variation of bubble’s rise velocity to the
bubble interface, which means captured by other individual particles
number of captured particles. In order to facilitate the visualization,
already present on the interface. However, it is known that particles
these preliminary experiments have been carried out in a rectan-
are subjected to competitive heterocoagulation with bubble and
gular glass column similar to the flotation device with a height of
agglomeration with other particles (Vinke et al., 1991). If particle-to-
the capture zone being about 300 mm.
particle cohesion dominates, large clusters will be formed and they
are unable to adhere to bubbles in water, while if particle-to-bubble
3.1. Formation of particle–bubble aggregates adhesion dominates, bubble’s surface can be totally covered by the
particles.
Several observations of the bubble–particle collision have been The layer of particles on one of these particle–bubble aggregate
performed. The size of captive bubble was in the range of 1–3 mm has been observed using a microscope (see Fig. 6). It concerns a
and the liquid velocity Ul varies between 0 and 12.8 mm s  1. The bubble with a monolayer of glass beads. The microscopic exam-
trajectories of particles moving around a captive bubble are illustrated ination shows in detail that the bottom of the air bubble is covered
in Fig. 7. by a monolayer of particles (see Fig. 9). Highly packed arrangement
Depending on their initial position, some of the particles of the attached particles at the bubble surface is observed. However
succeed in colliding with the bubble. After collisions, particles slip there is still a small empty space between the particles. The ratio of
on the interface and then attach to the bubble once the liquid film the surface occupied by particles to the total surface area (surface
was drained out. Adhered particles slide along the interface under fraction f) depends on the particle’s size distribution and its
s
the effect of the surrounding liquid flow, but also simply under the packing arrangement. The image processing by using Matlab
effect of their weight when there is no liquid flow. At the bottom of shows that the ratio of the surface occupied by the particles to the
the bubble, particles keep moving on the surface and attempt to total surface area f can reach to 80% (see Fig. 10). So the interface
rearrange their position and by this way minimize the fully covered may be considered as immobile within this area and a no-slip
surface area. When the bubble is rising, we can easily imagine that condition is offered to the liquid flow on this part of the bubble–
the sliding and the accumulation of particles to form a ‘‘particle- liquid interface.
cap’’ are more efficient due to the liquid velocity at the bubble
interface.
Different patterns of particles–bubble aggregates have been 3.2. Modeling of bubble surface coverage
observed, as shown in Fig. 8: the solid particles may adhere to a
bubble individually or form a monolayer (Fig. 8a and b, respectively); To describe the bubble’s surface coverage layer, Vinke et al.
some particles’ clusters attach to the bubble surface through one or (1991) introduced the concept of bubble coverage defined as the
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 987

Fig. 8. Images of air bubbles covered with micro-glass beads (db ¼ 1 mm, dp ¼ 556 mm). (a) adhesion of a single particle to a gas bubble, (b) monolayer adhesion, (c) adhesion
of a cluster of particles (d) multilayer adhesion.

In order to link the number of particles at the surface to the angle

yclean , we consider a spherical gas bubble with diameter db ¼2rb,
partially covered by identical particles with diameter dp that form a
monolayer at the bubble surface. The surface of the immobile area
can be approximately calculated as below:

Sim ¼ 2prb2 ð1 þ cosyclean Þ: ð3Þ

Bubble’s surface contamination level is an important parameter

for the bubble’s motion. On the one hand, adhering particles
increase bubble’s apparent density, in other words, bubble’s buoy-
ancy. On the other hand, the immobilization of a part of the
interface may appreciably increase the drag of the bubble. From the
hydrodynamics point of view, at y ¼ yclean , the condition of interface
of the bubble changes from the zero tangential stress t  n ¼ 0 to the
no-slip condition u ¼0. For creeping flow (Reb ¼0), Sadhal and
Johnson (1983) have proposed an expression for the drag coeffi-
cient for a spherical bubble with partially contaminated surface
Cd ðyclean Þ, in which Cd ðyclean Þ is related to the drag coefficients for
fully contaminated bubbles (Cim m
d ) and to that for clean bubbles (Cd )
Fig. 9. Particles layer at bubble surface. 
via a normalized drag coefficient Cd ðyclean Þ:

Cd ðyclean ÞCdm
fraction of bubble surface covered with adhering particles to model Cd ðyclean Þ ¼ , ð4Þ
Cdim Cdm
the particle to bubble adhesion, when a monolayer of particles
covers the rear of a bubble. This bubble coverage is very similar to where Cm im 
d ¼ 16/Reb, Cd ¼24/Reb and Cd ðyclean Þ is written as follows:
the concept of interface contamination by reference to the surface  
covered by surfactants or other liquid contaminants, where the 1 1
Cd ðyclean Þ ¼ 2ðpyclean Þ þ sinyclean þ sin2yclean  sin3yclean :
bubble surface is divided into two parts (see Fig. 11): at the front 2p 3
part, bubble’s surface is clean and can move with the liquid, so this ð5Þ
area is called mobile surface noted as Sm; meanwhile at the rear
More recently, DNS studies have shown that Eq. (4) can correctly
part, bubble’s surface is contaminated and behaves as a ‘‘stagnant
evaluate the drag coefficient for the case of bubble’s Reynolds
cap’’. This surface is called immobile and noted Sim (Sadhal and
number up to Reb ¼ 300 (Cuenot et al., 1997; Dani et al., 2006), by
Johnson, 1983). In this model, bubble surface contamination is
using the correlations of Mei et al. (1994) and of Schiller and
characterised via the angle yclean , measured from the front stagnant
Nauman (1935) to calculate, respectively, Cm im
d and Cd :
point to the boundary of the contaminated zone. yclean varies from
01 to 1801. yclean ¼ 03 means that the bubble is totally covered by " #
16 Reb
particles, while yclean ¼ 1803 means that the whole bubble surface Cdm ¼ 1þ , ð6Þ
remains free of particles.
Reb 8 þ 0:5ðReb þ 3:315Re0:5
b Þ
988 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

Fig. 10. Particles distribution at bubble surface with a high compacity. (a) raw image and (b) treated image.

300

290

280

Sm 270

clean 260
Ubexp [mm.s−1]

250
cap
240

230
S im
220

210

200
0 50 100 150 200 250 300
h [mm]

Fig. 12. Evolution of bubble’s rise velocity Uexp

b in the flotation cell (db ¼0.97 mm).
&: bubble’s rise velocity in pure water, 3: bubble’s rise velocity in particle
suspension.

3.3. Bubble rise velocity variation during capture process

During each experiment, bubble trajectories in the particle

capture section of the column have been registered by the video
camera. It is possible to extract the evolution of the bubble rise
velocity Ub in the flotation column. Surface coverage angle
Fig. 11. Schematic view of particles adhering to a gas bubble. (a) schema of stagnant ycap ¼ pyclean is deduced by the number of captured particles
cap model and (b) distribution of particles on the rear part of the bubble. determined by the particle counter. The velocity variation of a
bubble of diameter db ¼0.97 mm is shown in Fig. 12. The rise
velocity in pure water is plotted in the same figure for comparison.
24 We can firstly notice that the rise velocity in the de-ionized pure
Cdim ¼ ð1 þ 0:15Re0:687
b Þ: ð7Þ
Reb water remains almost constant along the height of capture zone. A
slight decrease of Ub has been observed, since the water contains
By applying this model, we discuss in the next paragraph certainly little impurities despite all the cleaning techniques that
about how this surface covering affects the bubble rise velocity have been used. On the contrary, a significant reduction of the
and can be used to obtain quantitative information about the cap- bubble rise velocity Ub has been recorded after the bubble has
ture efficiency. passed through the suspension. Table 2 reports the bubble’s rise
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 989

velocity at the entry Uexp exp

b (h¼ 0) and at the exit Ub (h¼300) of the The bubble’s apparent density can be calculated as
capture zone filled with the particle suspension. Knowing the 2 !3 3
bubble’s diameter db, the corresponding Reynolds number and drag
X5 dip
rb ¼ rb0 þ 4 i5
rp np : ð11Þ
coefficient Cd may be calculated by i¼1
db
!
4 db g r This increase of bubble’s apparent density breaks the balance between
Cd ¼ 1 b : ð8Þ the bubble’s buoyancy force and the drag force. Consequently,
3 ðU exp Þ2 rf
b bubble’s rise velocity is reduced to achieve a new equilibrium. This
new velocity can be calculated by the following relationship:
By using the correlations (6) and (7), one obtains the normalized drag " !#1=2
coefficient and the corresponding surface contamination degrees 4 db g rf rb
yclean through Eqs. (4) and (5). In the case of slightly deformed bubble, Ubcal ¼ : ð12Þ
3 Cd rf
Eq. (6) should be replaced by Moore (1965) correlations:
Note that the bubble’s relaxation time t  Oð0:01Þ (see Table 3)
!
48 HðwÞ remains small in comparison with the characteristic times of the
Cdm ¼ GðwÞ 1þ ð9Þ capture ts  0:150 s (Nguyen and Evans, 2004), so that a quasi static
Reb Re0:5
b
equilibrium between buoyancy and drag can be assumed.
In Table 3, the mass increase mp has been calculated as well as
with GðwÞ and HðwÞ are functions of the deformation w given in Moore
the corresponding bubble’s final rise velocity Ucal b . The bubble’s
(1965).
surface contamination level is supposed to be the same as its initial
The calculations performed for three different sizes of air
value yclean ðh ¼ 0Þ that is used to calculate the drag coefficient. The
bubbles are also reported in Table 2. Although these bubbles have
calculated terminal velocity at the exit of the capture zone Ucal b is
been slightly contaminated at the entry of the capture zone,
largely greater than the experimental one Uexp b (h¼300). In fact, it
bubbles lost 15–25% of their initial rise velocity after having passed
can be noted that Ucal b remains close to that at the entrance of the
through the particle suspension. The measurement error (less than
capture zone Uexp b (h¼0). So the additional mass due to the captured
3%) is not significant comparing to the velocity variation. This
particles cannot by itself explain the total decrease of the bubble
bubble’s rise velocity decrease can be explained by the two reasons
velocity. However, bubble’s effective density can reach more than
for which the relative contribution are evaluated in the following.
350 times its initial density for a bubble of diameter db ¼1 mm
Additional mass: Firstly, particles adhering to the rear of the
totally covered by solid particles of diameter dp ¼ 25 mm and
bubbles have a certain mass. Compared with the bubble’s own
density rp ¼ 2360 kg m3 . So the increase in bubble density may
mass, this additional mass mp may be important depending on the
not be neglected in the evaluation of Ub and has to be accounted for.
number nip of captured particles with a diameter dip:
Interface covering: As discussed above, the captured particles
5   slide along the interface and accumulate at the rear of the bubble
X 1
mp ¼ nip rp pðdip Þ3 ð10Þ due to the interface mobility. Following the bubble surface covering
i¼1
6 presented in Section 3.2, the interface, as a result, can be supposed
to be divided into a mobile part and an immobile part. This change
with i representing different particle size ranges (5–15, 15–25, of surface condition results in an augmentation of drag force. Since
25–40, 40–56, 56275 mm). the bubble’s relaxation time t being very short (Table 3), by
Thanks to the particle collecting system and to the particle laser supposing equilibrium between buoyancy and drag forces sub-
counter, the number nip of particles captured by one bubble and jected to the air bubble, it is possible to deduce the drag coefficients
released at surface have been recorded for each experiment, Cd at each instant by measuring the bubble size db and its rise
helping us to calculate the increase of bubble’s density. The total velocity Uexp
b (Eq. (8)). The corresponding contamination level yclean
volume of the hetero-aggregate was considered to be identical to can be then deduced from Eqs. (4) and (5) when the bubble exits the
the initial bubble volume, because the volume of particles attached capture zone (h¼300). Meanwhile, the angle yclean can also be
to the bubble remains small in comparison with the bubble volume. evaluated by another way. Assuming that particles adhere to a
bubble as a monolayer with a high packing arrangement as shown
Table 2 in Fig. 9, the surface fraction can be evaluated to fs  0:8 as
Variation of bubble’s rise velocity Uexp
b in the capture zone. discussed previously in section 3.1. Knowing the number of
particles nip captured by each bubble given by the particle counter,
No db (mm) Uexp (mm s  1) Reb (–) Cd (–) yclean (1)
b
the surface covered by particles is then
" #
Entry h ¼0 mm
im
X5 nip pðdip Þ2
1 1.157 0.05 2907 5 361 7 27 0.179 7 0.011 145 7 3 S ¼ : ð13Þ
2 0.977 0.04 2707 4 307 721 0.174 7 0.009 154 7 4 i¼1
4f
3 0.597 0.03 1037 2 66 75 0.725 7 0.037 128 7 4
For a bubble of diameter db, the clean angle ycal can then be
Exit h ¼300 mm
estimated as
1 1.157 0.05 248 75 309 724 0.244 7 0.014 135 7 3
2 0.977 0.04 223 74 254 7 20 0.256 7 0.015 139 7 3
 im m 
ðS S Þ
3 0.597 0.03 78 7 2 50 75 1.265 70.108 93 710 ycal ¼ arccos im m : ð14Þ
ðS þS Þ

Table 3
Correction of bubble’s rise velocity at the exit (h ¼ 300 mm) by accounting bubble’s effective density.

No db (mm) Uexp
b (mm s  1) mp (mg) rapp
b
(kg m  3) Reb (–) Cd (–) yclean (1) Ucal
b (mm s
1
) t (s)

1 1.15 248 1.49  10  2 19.5 328 0.209 145 284 714 0.041
2 0.97 223 7.40  10  3 15.2 277 0.211 154 266 712 0.032
3 0.59 78 1.15  10  2 107.8 60 0.773 128 947 4 0.013
990 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

Table 4
Correction of bubble’s rise velocity at the exit (h ¼ 300 mm) by accounting for bubble’s surface coverage effect on drag force.

No db (mm) Uexp
b (mm s  1) Sim (mm2) Sm (mm2) yclean (1) Reb (–) Cd (–) Ucal
b (mm s
1
)

1 1.15 248 0.553 3.601 137 319 0.229 256 7 10

2 0.97 223 0.318 2.680 142 263 0.238 231 7 8
3 0.59 78 0.453 0.637 100 52 1.167 81 73

Table 5
Correction of bubble’s rise velocity at the exit (h ¼ 300 mm) by accounting the effects of bubble surface coverage on the bubble apparent density and on the drag force.

No db (mm) Uexp
b (mm s  1) yclean (1) rb (kg m  3) Reb (–) Cd (–) Ucal
b (mm s
1
) DUb =Ub (%)

1 1.15 248 137 19.5 306 0.187 2527 12 1.61

2 0.97 223 142 15.2 258 0.107 2277 8 2.07
3 0.59 78 100 107.8 48 0.597 757 3 3.97

Table 6 4. Results and discussions

Comparison between different particle packing levels f at the bubble surface for the
calculation of the bubble velocity at the exit of the capture zone.
A direct relationship between the number of captured particles,
db f ¼ 0:7 f ¼ 0:8 f ¼ 0:9 the bubble size and its rise velocity is established firstly in this
(mm) section and this new method is then applied in the experimental
yclean Ubcal yclean Ubcal yclean Ubcal device (Fig. 1) to determine the capture efficiencies by measuring
(1) (mms  1) (1) (mms  1) (1) (mms  1) the variation of bubble rise velocity. The comparisons with various
theoretical predictions and with experimental results of the
1 134 241 137 252 140 262
2 139 219 142 223 144 234 literature are discussed at last.
3 93 72 100 78 105 78

4.1. Experimental conditions

In Table 7, the experimental conditions are reported. The rise

In Table 4, we have recalculated Ucal b , the bubble’s rising velocity at velocity of air bubbles has been measured in pure water at the top of
the exit of the capture zone, by using ycal obtained from surface the bubbling zone to ensure that bubble has achieved its terminal
coverage (Eqs. (13) and (14)), while the bubble’s density is kept velocity. The variation of the water density and its dynamic visco-
constant rb ¼ 1:18 kg m3 . Compared to the experimental value, sity due to the temperature fluctuation (between 221 and 281) are
Ucal
b is now closer to but still greater than Ub
exp
at the exit. accounted for. Note that, the difference in temperature may cause an
These calculations show that both the bubble’s effective density obvious modification on the rise velocity. For example, Ub increases,
and its surface coverage due to the particles capture are important respectively, about 15% and 12% for an air bubble db ¼0.6 and 1.0 mm
to influence the bubble’s motion. In Table 5, these two parameters in pure water, as T increases from 201 to 301. The terminal rise velocity
are both taken into consideration for the evaluation of the bubble’s Ub is plotted vs. the bubble’s diameter db in Fig. 13. These experi-
final rise velocity. mental results are also compared to those in the literature (Schiller
The values of bubble’s final rise velocity Ucalb reported in Table 5 and Nauman, 1935; Moore, 1965; Mei et al., 1994; Duineveld, 1995)
can then be compared to Uexp b (h¼300). For all the three sizes of for the average temperature T¼251. Firstly, we can note that Ub is
bubbles tested, a rather good agreement is achieved with a slight closed to that of a clean bubble and much far away from that of a solid
difference between the measured value Uexp b and the calculation Ub
cal
sphere, so the bubbles produced in our experiment can be considered
(less than 4%). In addition to all the hypothesis that are necessary to with a clean or quasi-clean interface. Secondly, Ub is observed to begin
perform this estimation, a discrepancy may also comes from the to deviate from a spherical bubble to an ellipsoidal bubble at
agglomeration of particles. If the clusters formed at the rear of the db ¼0.8 mm and the difference increases with the bubble size. This
bubble surface is attached through only one or few particles is confirmed by measuring the bubble’s deformation w defined as the
(Fig. 8c) or if an aggregate of particles adheres through successive ratio between the longer and smaller axes (Dx and Dy, respectively).
layers (Fig. 8d), effects of the surface covering could be significantly As it is shown in Fig. 14, even for small bubble (db o 0:6 mm), its shape
reduced. The determinated bubble’s rise velocity is also sensitive to is not perfectly spherical. Our measurement are found to be in good
the packing level f. As it is shown in Table 6, when f varies from agreement with the solution for clean bubble (Moore, 1965). This
low packing level 0.7 to high packing level 0.9, the covering surface small deformation reduces the bubble’s terminal velocity.
decreases and bubble rise velocity increases consequently. How- Bubble’s Reynolds number varies from Reb ¼8 (db ¼0.22 mm)
ever, this increase may achieve 5% between two levels, for instance, to Reb ¼456 (db ¼1.16 mm). It is noted that for db  0:6 mm, the
when db ¼1.15 mm. bubble deformation becomes observable. The dimensionless settling
This preliminary study suggests that there is a direct relation- velocity us ¼Up/Ub remains very low O(10  3), except for the small
ship between the number of captured particles, the bubble size and bubbles db r0:25 mm with us ¼0.01. The numerical studies of
its rise velocity during the capture process. Following this observa- Heindel and Bloom (1999) show that gravitational effect is no longer
tion, a new experimental method is then developed to determine negligible only for us Z 0:01. We will then suppose in the following
the capture efficiency by the measurement of bubble size and that settling effect on capture efficiency can be neglected. The particles
velocity variation along the capture zone. Stokes number Stp, defined as the particle’s relaxation time tp to the
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 991

Table 7
Experimental conditions.

db (mm) Ub (mm s  1) w (–) rf (kg m  3) mf (Pa s) Reb (–) Stp (–) us (–) Cd (–) ymax
clean (1)
C (p ml  1)

1 0.227 0.03 30.8 70.3 1.00 996.31 0.000833 8 0.018 0.010 2.991 156 7 5 25840
2 0.257 0.03 34.3 7 0.7 1.00 997.38 0.000911 9 0.016 0.009 2.825 161 7 9 21905
3 0.397 0.04 66.1 7 0.7 1.00 996.91 0.000873 30 0.021 0.005 1.170 144 7 4 31250
4 0.537 0.04 95.6 7 0.9 1.02 996.95 0.000891 58 0.021 0.003 0.751 137 7 6 46414
5 0.577 0.03 129.5 7 1.7 1.04 996.86 0.000871 84 0.028 0.002 0.447 168 7 5 25505
6 0.597 0.03 128.2 7 1.5 1.04 996.86 0.000871 87 0.026 0.002 0.472 150 75 31023
7 0.687 0.03 138.0 7 1.4 1.03 997.13 0.000891 105 0.024 0.002 0.464 139 7 5 26695
8 0.807 0.03 209.0 72.4 1.04 996.59 0.000852 200 0.033 0.001 0.228 158 7 8 32858
9 0.827 0.03 214.5 7 1.8 1.05 996.86 0.000871 195 0.031 0.001 0.244 155 7 4 31024
10 0.977 0.05 270.0 72.7 1.10 996.95 0.000891 299 0.033 0.001 0.174 154 7 4 46420
11 0.987 0.04 233.5 7 3.7 1.07 997.38 0.000911 251 0.028 0.001 0.234 144 7 4 46049
12 1.047 0.04 303.8 7 1.9 1.16 996.31 0.000833 378 0.037 0.001 0.147 160 74 15981
13 1.077 0.04 295.9 7 4.2 1.17 997.38 0.000911 345 0.032 0.001 0.160 165 7 3 36069
14 1.087 0.05 303.4 7 5.4 1.19 997.13 0.000891 368 0.033 0.001 0.154 163 7 2 32855
15 1.17 70.05 342.7 7 4.9 1.22 996.86 0.000871 456 0.035 0.001 0.130 161 7 2 14965

0.4 1.5

0.35 1.4

0.3 1.3

0.25 1.2
Ub [m.s−1]

χ = Δx/Δy

0.2 1.1

0.15 1

0.1 0.9

0.05 0.8

0
0.7
2 4 6 8 10 12 14
2 4 6 8 10 12 14
db [m] x 10−4
db [m] x 10−4
Fig. 13. Terminal velocity of an air bubble in pure water at T¼251. 3: Measurements
Uexp
b (h ¼0), —: spherical bubble (Mei et al., 1994), –––: ellipsoidal bubble (Moore, Fig. 14. Theoretical and experimental determinations of w as a function of the
1965), : ellipsoidal bubble Duineveld (1995),   : solid particle (Schiller and equivalent bubble radius at 251. 3: Measurements at h ¼ 0, : Moore (1965).
Nauman, 1935).
with the same assumptions that have been used in the previous
section:
characteristic time of the disturbed flow tb ¼ 2rb =ub , varies from 0.016
to 0.037. Note that for a value of Stp greater than 0.1, particle’s inertia  Spherical bubble of diameter db;
has a positive effect on the bubble–particle collision that increases Ecoll,  Spherical particles with identical size dp and same physical and
while for 0:01 oSt p o 0:10, particles inertia induces a negative effect physico-chemical properties;
and the collision efficiency is slightly reduced (Nguyen, 1998; Ralston  Homogeneous particle concentration C in the suspension;
et al., 2002). The particle concentration is between 14 965 and  Monolayer adhesion of particles on the bubble surface with a
46 420 p/ml and corresponds to 0.005–0.015% in weight or 0.002– packing level f ¼ 0:8.
0.006% in volume. Using the Einstein’s relation mðCÞ=m0 ¼ 1 þ 2:5C, the
effect of the particles on the liquid viscosity can be calculated. mðCÞ=m0 According to its definition, bubble’s capture efficiency Ecapt of a
is found to be less than 1.0002. The modification of the viscosity due to given particle size range centered on dp is defined as the ratio of the
the presence of the particles is insignificant. number of particles in the size range dp captured by one bubble
np(dp) over the number of particles np0(dp) initially located in the
volume swept by the bubble. By assuming that the particle
4.2. Particle–bubble capture efficiency concentration C(dp) remains uniform in the whole capture zone,
capture efficiency Ecapt can be written as
4.2.1. Method
Let us consider how to deduce the bubble particle capture np ðdp Þ npt ðdp Þ=nb
Ecapt ðdp Þ ¼ ¼ ð15Þ
efficiency by simply measuring bubble’s size and its rise velocity np0 ðdp Þ Cðdp Þ  H  pd2b =4
992 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

where H is the height of the capture zone, nb is the number of contamination level ðyclean Þ due to the capture of the particles
bubbles and npt is the total number of particles collected during ðdp ¼ 25 mm, rp ¼ 2363 kg m3 Þ has been calculated by using
each experiment. Eqs. (16)–(20) and is presented on Fig. 15. Both the effects of surface
Eq. (15) is valid under the condition that bubbles trajectory are contamination and additional mass have been considered. This figure
not disturbed by the particles and they rise along a vertical straight. shows that for yclean A ½165,180 and [0, 60], the variations of Ub are,
For an air bubble in water, its diameter should be less than 1.5 mm respectively, less than 8 mm s  1 (about 2.6% of Uby ¼ 180 ) and 3 mm
and dp/db should be much smaller than unity in order to keep s  1 (about 2.6% of Uby ¼ 60 ). It can be noted that within this ranges
moving straightly. In all the single bubble flotation experiments bubble’s rise velocity is insensitive to the variation of yclean . In another
presented, the size of the bubbles was in the range of 0.22–1.16 mm word, it is difficult to estimate Ecapt by measuring DUb , since the
and dp =db o 0:25. measurement accuracy on Ub is of the same order of magnitude
In the classical counting methods, npt can be obtained by the ð  2%Þ. By comparing Tables 7 and 8 to Fig. 15, we can see that for all
particle counter, which gives the number and the distribution of the experiments performed, yclean varies between 941 and 1651, where
particles captured by a series of identical bubbles during each Ub drops quickly as yclean decreases. As a consequence, the measure-
experiment (Anfruns and Kitchener, 1977; Hewitt et al., 1994; Dai ment error is small compared to the velocity variation due to the
et al., 1998; Sarrot et al., 2007). Here, a new method to determine capture and the evaluation of Ecapt becomes possible and accurate.
the capture efficiency is proposed. Bubble’s rise velocity Ub(h) was
measured at different levels h in the flotation column. The bubble
4.2.3. Capture efficiency vs. bubble rise velocity
size is determined in the bubbling zone and is supposed to remain
Figs. 16 and 17 show two examples of the evolution of the
constant. Note that the change in diameter of a db ¼1 mm bubble
bubble rise velocity and its capture efficiency in the flotation
due to the pressure variation along h ¼300 mm is only 3%. The drag
column. It concerns bubbles of diameter db ¼0.80 and 1:07 mm. All
coefficient Cd ðh, yclean Þ can be calculated by Eq. (8). By using the
the results for different bubble sizes are given in Table 7. Bubble
‘‘Stagnant Cap Model’’ (Eqs. (4) and (5)), we can obtain the angle
surface is supposed to be covered by a monolayer of solid particles
yclean and then calculate the coverage surface from Eq. (3). Based on
ðdp ¼ 25 mmÞ with a packing level f ¼ 0:8. It is shown that bubble’s
the monosize and monolayer hypothesis, the number of captured
rise velocity Ub is reduced along its trajectory due to the capture as
particles can be deducted by np ðhÞ ¼ 4Sim ðhÞf=pd2p . Bubble’s appar-
we have explained previously. Bubble’s capture efficiency Ecapt are
ent density rb ðhÞ is obtained by using Eq. (11). An iterative
also observed to decrease (see Figs. 16b and 17b), since its surface
procedure is used to solve the following system of equations and
contamination level increases. It is in accordance with the numer-
to determine the total number of particles np captured at each
ical simulation results obtained by Sarrot et al. (2005) and Legendre
instant t ¼tj:
 et al. (2009): the particle–bubble collision efficiency Ecoll goes down
1 1 1 as the yclean decreases.
Cd ðhÞ ¼ ½pyclean ðhÞ þ sin½yclean ðhÞ þ sin½2yclean ðhÞ
p 2p 2p

1
 sin½3yclean ðhÞ ½Cdim ðhÞCdm ðhÞ þ Cdm ðhÞ, ð16Þ 4.2.4. Influence of particle to bubble size ratio
3
In the literature, several collision models have been proposed
for the inertial free particles. All these collision models are very
" !#1=2 similar and show that Ecoll varies with the particle to bubble size
4 rb ðhÞ
Ubexp ðhÞ ¼ 1 , ð17Þ ratio rp/rb: for a clean bubble Ecoll  ðrp =rb Þ while for a totally
3Cd ðhÞ rf contaminated bubble Ecoll  ðrp2 =rb2 Þ. We have plotted the capture

Sim ðhÞ ¼ pðrp þrb Þ2 f1 þ cos½yclean ðhÞ g, ð18Þ

np ðhÞ ¼ 4fSim ðhÞ=pd2p , ð19Þ

 3 0.3
dp
rb ðhÞ ¼ rb0 þ rp np ðhÞ : ð20Þ
db
So along Dh, one air bubble captures Dnp ¼ np ðhþ DhÞnp ðhÞ 0.25
particles. Using Eq. (15), the capture efficiency Ecapt is then
deduced:
  0.2
Ub [mm /s]

Dh Dnp
Ecapt h þ ¼ 2 : ð21Þ
2 prb C Dh
0.15

4.2.2. Measurement accuracy

From the previous demonstration, we can see that the reliability 0.1
of the results depends on the accuracy of the measurement of
bubble size db and its rise velocity Uexpb . Knowing the precision of
digital images, which is about 0.01–0.05 mm pixel  1 (and varies 0.05
from one experiment to another), we can estimate measurement
errors of Ub and db (see Table 7). The propagation of uncertainty of
yclean ¼ f ðDUb Þ as well as Ecapt ¼ f ðDyclean Þ due to measurement 0
0 20 40 60 80 100 120 140 160 180
limitations can be calculated by applying Taylor expansion to
clean
Eqs. (16) and (17) and to Eqs. (18), (19), (21), respectively.
However, it should be noted that the evaluation of Ecapt relies Fig. 15. Theoretical evolution of Ub as a function of yclean of a spherical bubble
directly on the variation of Ub. The theoretical evolution of Ub of a (db ¼1 mm) covered by the spherical particles (dp ¼ 25 mm, rp ¼ 2363 kg m3 ) in
spherical air bubble (db ¼ 1 mm) with the increase of surface pure water obtained by using Eqs. (16)–(20).
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 993

Table 8
Summary of experimental results.

db (mm) Remax (–) Remin (–) Reave (–) ymax

clean (1) ymin yave
clean (1)
Ecapt
clean (1)
b b b

Max Min Ave

1 0.22 8 7 7 156 133 145 0.1970 0.0840 0.1314

2 0.25 9 8 9 161 145 153 0.1307 0.0894 0.1067
3 0.39 30 26 28 144 128 136 0.1087 0.0597 0.0816
4 0.53 58 40 43 137 94 116 0.0773 0.0434 0.0693
5 0.57 84 66 70 168 133 151 0.1039 0.0612 0.0741
6 0.59 87 65 70 150 121 136 0.0901 0.0480 0.0611
7 0.68 105 90 94 139 126 133 0.0560 0.0418 0.0518
8 0.80 200 158 168 158 136 147 0.0770 0.0412 0.0534
9 0.82 195 159 170 155 137 146 0.1042 0.0308 0.0501
10 0.97 299 246 266 154 139 147 0.0579 0.0453 0.0534
11 0.98 251 206 219 144 129 137 0.0599 0.0254 0.0444
12 1.04 378 352 358 160 152 156 0.0650 0.0268 0.0474
13 1.07 345 258 277 165 137 151 0.0721 0.0331 0.0435
14 1.08 366 292 309 163 135 149 0.0669 0.0247 0.0437
15 1.17 456 386 404 161 147 154 0.0930 0.0435 0.0563

Exponent ‘‘max’’ means that measurements are taken at the entrance of flotation column, while ‘‘min’’ means that measurements are taken at the exit.

190
0.14

185 0.12

0.1
Ubexp [mm /s]

180
0.08
Ecapt

175 0.06

0.04
170
0.02

165 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
h [mm] h [mm]

Fig. 16. Evolution of bubble’s rise velocity Uexp

b and corresponding capture efficiency Ecapt – Test No. 6: db ¼ 0.80 mm. (a) Uexp
b vs. h and (b) Ecapt vs. h.

250
0.14

245 0.12

240 0.1
Ubexp [mm /s]

0.08
Ecapt

235
0.06
230
0.04
225 0.02

220 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
h [mm] h [mm]

Fig. 17. Evolution of bubble’s rise velocity Uexp

b and corresponding capture efficiency Ecapt – Test No. 13: db ¼ 1.07 mm. (a) Uexp
b vs. h and (b) Ecapt vs. h.

  " #
efficiency vs. rp/rb in Fig. 18. On the same log–log curve, the collision 3 rp 2 ð3=16ÞReb
efficiencies of a clean bubble Em Eim ¼ 1þ : ð23Þ
coll and a completely contaminated coll
2 rb 1 þ 0:249Re0:56
b
bubble Eim coll calculated by using the Sarrot et al. (2005) model are
plotted for comparison: The solid lines and the dash lines represent, respectively, Ecapt for a
  1=2
! clean bubble and for a totally contaminated bubble at Reynolds
rp rp 16 þ 3:315Reb þ 3Reb
Em
coll ¼ 1 þ 1=2
ð22Þ number corresponding to the maximum (Remax b ¼456) and the
rb rb 16 þ 3:315Re þReb b minimum (Reminb ¼8) considered in this study. It is clear that all our
994 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

100 0.2

0.18
Reb = 456
0.16
Reb = 8

0.14

0.12

Ecapt
Ecapt

10−1 0.1

0.08

0.06

0.04

0.02

10−2 0
10−2 10−1 100 0 0.5 1 1.5
rp / rb db [mm]

Fig. 18. Capture efficiency as a function of the particle to bubble size ratio. 3: Fig. 19. Experimental results of capture efficiency vs. bubble diameter. 3: this study,
experimental points, —: Eq. (22), –––: Eq. (23).  : experimental results of Yoon and Luttrell (1989) dp ¼ 31 mm, : experimental
results of Nguyen (1994) dp ¼ 15 mm, : experimental results of Dai et al. (1998).

experimental results are located in the zone limited by the two

solid lines predicted by the collision model for a bubble with a clean
interface. These results agree with the results in Table 7 concerning As the Ecapt behavior is closer to that for a clean bubble, we
the respective contamination angle yclean . For most of bubbles compare our experimental results to the collision model proposed
generated, the cap angle yclean is greater than 1201. According to the by Sarrot et al. (2005) (Eq. (22)) and by Nguyen (1994) (Eq. (24)) for
numerical analyses of Sarrot (2006), particles capture efficiency a clean bubble and inertial free particles, in which the effect of the
behaves like clean bubble in such condition of low contamination bubble’s Reynolds number is under consideration:
level. However, for a given size ratio rp/rb, a significant decrease of "    #
the capture efficiency is observed due to the surface recovering. m 2 rp rp 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ecoll ¼  X 2 þ 3Y 2 X X 2 þ 3Y 2 þ2X
27Y 2 rb rb
ð24Þ
4.2.5. Influence of bubble size and its Reynolds number
with
The integration of the capture efficiency Ecapt(h) along the height
of the capture zone studied gives the average capture efficiency 0:0637Reb 0:0537Reb
X ¼ 1þ Y¼ : ð25Þ
Ecapt. This average value allows us to compare our results with the 1 þ0:0438Re0:976
b 1þ 0:0318Re1:309
b
capture efficiency obtained by classical counting method. In the
In Fig. 20, the average capture efficiency Ecapt divided by (rp/rb) is
following, Ecapt refers to this average value along the bubble path.
reported as function of bubble’s average Reynolds number Reb. The
The results of all the experiments are summarized in Table 8.
functions Ecoll/(rp/rb) are represented by the solid and dash lines for
In fact, most of the previous experiments are carried out in liquid
the models of Sarrot et al. (2005) and that of Nguyen (1994),
containing surface active agents or other impurities, so the bubble’s
respectively. It is seen that Ecapt increases with Reb as predicted by
surface is totally retarded (Reay and Ratcliff, 1973; Anfruns and
the model. All the experimental values are slightly lower than that
Kitchener, 1977; Yoon and Luttrell, 1989; Nguyen, 1994). There are
predicted by the modeling, except for one point. Note that capture
very few experimental data for particle capture by clean bubbles,
efficiency Ecapt ¼ Ecoll  Eatt  Estab , should be lower than the collision
because of the difficulties to generate the bubble with a size wanted
efficiency since Eatt and Estab are in general less than unity.
and to control the bubbles emission frequency. Concerning the recent
researches on air bubbles with mobile surface, the experimental
works of Dai et al. (1998), Ralston and Dukhin (1999) and the 4.2.6. Influence of bubble surface contamination
numerical simulations of Nguyen (1998), Sarrot et al. (2005), The value of the maximum and minimum values of the interface
max min
Legendre et al. (2009) can be referred to. The experimental results contamination yclean (at entrance) and yclean (at exit) are reported in
of Dai et al. (1998) are plotted in Fig. 19, represented by ðÞ, as well as Table 8. It can be noted that for most cases, the interface
our results represented by the symbol ð3Þ. It is observed that the contamination level characterized by yclean varies between 180
max min
capture efficiency decreases as the bubble size increases. The same and 90, with yclean ¼ 1683 and yclean ¼ 943 . As discussed in the
evolution is observed for these two experimental studies. precedent paragraph, bubble’s surface contamination may greatly
For comparison, Ecapt for fully contaminated bubbles have also affect bubble–particle heterocoagulation. In Fig. 21, we try to
been plotted in Fig. 19 represented by the symbols (  ) and ðÞ compare the experimental results with the collision model pro-
extracted from the experiments of Yoon and Luttrell (1989) and posed by Legendre et al. (2009), where the effect of the surface
Nguyen (1994), respectively. It is evident that Ecapt for a clean contamination is accounted for. The symbols 3 represents the
bubble is much higher than that for a totally retarded bubble as average Ecapt and the error bar represents Emax min
capt and Ecapt calculated
already reported by collision models. from the value of yclean at entrance and at exit, respectively.
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 995

ub 180 np
4 uG ¼ f ðReb Þ if yclean 4 ð27Þ
2 2
with n¼(2+ 0.2Re0.5 0.5
b )/(1+0.2Reb ) and yc is the collision angle
3.5 defined by the grazing trajectory:
2þ 0:16Re0:5
b p
3 yc ¼ yclean if 0 o yclean r
3þ 0:16Re0:5
b
2
f = Ecapt / (rp/rb)

2.5 3 p 2p
yc ¼ y if o yclean o
4 clean 2 3
2 h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i 2p
yc ¼ arccos X 2 þ3Y 2 X =3Y if r yclean o p ð28Þ
3
1.5
where X and Y are given by Eq. (25).
It can be noted that the experimental results are in good agreement
1 with the theoretical curves: quite all the experimental points are
located between the theoretical curve for yclean ¼ 1803 and 903 . Table 8
0.5 shows that the average yclean is about 135–1401, while the correspond-
ing values Ecapt are observed closer to the value of the model for
yclean ¼ 1203 , lower than expected. This may be caused by the inequal-
0 ity between the collision efficiency Ecoll and the capture efficiency Ecapt,
0 100 200 300 400 500
as explained in the previous Section 4.2.5. While for large bubble, for
Reb
example db ¼1.17 mm, the capture efficiency is surprisingly greater
Fig. 20. Capture efficiency divided by the size ratio rp/rb vs. bubble’s Reynolds number
than its collision efficiency. By referring to Table 7, it can be found that
Reb. — : Eq. (22) (Sarrot et al., 2005), –––: Eq. (24), 3: experimental points (this study). the concentration of the particle suspension is very low and the
incertitude of the measurement is more important as the number of
particles captured by one bubble remains too small to be accurate.
0.25 Note that in the experiments of Hewitt et al. (1994) and Ralston et al.
(1999), the initial particle concentration in the capture zone is about
0.05–0.20 wt%. From Table 7, it is seen that for db  0:2 mm (tests no.
n°1

1 and no. 2), the dimensionless sedimentation velocity us ¼Up/Ub is

0.2 about 0.01, ten times greater than in the other cases. Since the
sedimentation of particles and the rise of the bubble are in opposite
directions, the effect of gravity may be beneficial for collision.
According to Nguyen (1998), the contribution made by gravitational
0.15 sedimentation can be evaluated by us sin2 yc =ð1þ us Þ. In this case, the
n°2

increase of the collision efficiency is about 5–10%.

n°3
Ecapt

Since interface coverage angle varies as the bubble passes through

n°5

n°9

the suspension and is different from one experiment to another, each

n°15
n°6
n°4

0.1 experimental value is compared to the value given by Eqs. (26) and
n°12
n°13
n°14
n°7
n°8

n°11
n°10

0.16
0.05

0.14

0
0.12
0.2 0.4 0.6 0.8 1 1.2
db [mm]
0.1
Ecapt
exp

Fig. 21. Bubble–particles ðdp ¼ 20 mmÞ capture efficiency vs. bubble size. 3: Experi-
mental points (ave. Ecapt in Table 8), —: yclean ¼ 1801, –––: yclean ¼ 1351,   :
yclean ¼ 1201,   : yclean ¼ 901 (Legendre et al., 2009).
0.08

The collision efficiency were calculated and reported in Fig. 21 by

using Eqs. (26) and (28) for yclean ¼ 1803 , 1351, 1201 and 901: 0.06

 
rp rp uG
Ecoll ¼ 2sinyc 1þ ð26Þ 0.04
rb rb u b

where uG is maximum tangential velocity at the bubble surface 0.02

given by 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
  Ecoll
ub 180 y np
uG ¼ f ðReb Þsinn clean if yclean r
2 n 2 Fig. 22. Theoretical collision efficiency in comparison with experimental results.
996 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997

(28) for the same conditions (Fig. 22). For complete accordance Greek symbols
between the experimental results and the model, plotted points have
to lie on straightly diagonal line, represented by ðÞ, with a slope of w bubble deformation (–)
451. We can observe that the collision model is rather valid for inertial f particles packing level (–)
free particles. More experiments for inertial particles (typically m dynamic viscosity (Pa s)
us Z 0:1) are now required for a comparison with model from the yc collision angle (1)
literature. Furthermore, it is shown that in Figs. 19–22, the error of yclean clean angle (1)
Ecapt due to the measure accuracy for several experiences is more than r density (kg m  3)
50%. Comparing to Table 8, we can note that it concerns the points Subscripts
max
with yclean A ½160,168 . As we have discussed in Section 4.2, in this
region, Ub is insensitive to yclean . As a result, the discrepancy may be b bubble
enlarged even for a small measurement error. l liquid
p particle

5. Conclusions Superscripts

cal calculated
This study presents experiments regarding the capture of small
exp experimental measurement
spherical particles by air bubbles in a well-controlled flotation
im immobile interface
column. High speed video observation of the particle–bubble
m mobile interface
interaction shows that after collision, adhering particles keep
max maximum
sliding on the bubble surface and accumulate at the bottom of
min minimum
the bubble. Particles cluster or multilayer adhesion may be present,
if the cohesive forces between particles are important. These
adhering particles not only increase apparent density of bubble,
but also immobilise the rear part of the bubble surface. As a Acknowledgement
consequence, the bubble rise velocity decreases along its motion
through the particles suspension. This work demonstrates that by The authors would like to especially thank the CNRS Federation
measuring the evolution of bubble’s rise velocity, the surface FERMAT for the financial and technical supports.
coverage can be determined, and the number of particles collected
and carried by the bubble can be deduced. Based on the relationship
between bubble’s rise velocity, surface coverage and number of References
particles collected per bubble, a new experimental method is
developed to determine the bubble–particles capture efficiency. Anfruns, J.F., Kitchener, J.A., 1977. Rate of capture of small particles in flotation.
Transactions of the American Institute of Mining and Metallurgical Engineers
It’s noted that this new technique gives satisfactory agreement 86, 9–15.
with experiments done by the counting methods. This experi- Cuenot, B., Magnaudet, J., Spennato, B., 1997. The effects of slightly soluble
mental capture efficiency has been compared to the value pre- surfactants on the flow around a spherical bubble. Journal of Fluid Mechanics
339, 25–53.
dicted by the previous models and to the experimental results in Dai, Z., Dukhin, S., Fornasiero, D., Ralston, J., 1998. The inertial hydrodynamic
the literature: rather good agreements are achieved. As the classical interaction of particles and rising bubbles with mobile surface. Journal of Colloid
counting techniques to determine the capture efficiency are far to and Interface Science 197, 275–292.
Dani, A., 2007. Transfert de masse entre une bulle et un liquide: simulations
be easy to implement, this new experimental approach is therefore
numeriques directes et fluorescence induite par nappe laser. Ph.D. Thesis,
very interesting because it is only based on the measure of the Institut National des Sciences Appliquees de Toulouse.
bubble velocity. For instance, it may be applied to several new Dani, A., Cockx, A., Guiraud, P., 2006. Direct numerical simulation of mass transfer
experimental situations like the capture of very small particles from spherical bubbles: the effect of interface contamination at low Reynolds
numbers. International Journal of Chemical Reaction Engineering 4 (A2).
(nano-particles) or micro-algae for instance, because with these Duineveld, P., 1995. The rise velocity and shape of bubbles in pure water at high
type of classical objects the use of counting methods may be Reynolds number. Journal of Fluid Mechanics 292, 325–332.
impossible. Heindel, T.J., Bloom, F., 1999. Exact and approximate expressions for bubble particle
collision. Journal of Colloid and Interface Science 213 (1), 101–111.
Hewitt, D., Fornasiero, D., Ralston, J., 1994. Bubble-particle attachment efficiency.
Minerals Engineering 7, 657–665.
Hewitt, D., Fornasiero, D., Ralston, J., 1995. Bubble–particle attachment. Journal of
Nomenclature the Chemical Society, Faraday Transactions 91 (13), 1977–2001.
Hu, Y., Qiu, G., Miller, J., 2003. Hydrodynamic interactions between particles in
Roman symbols aggregation and flotation. International Journal of Mineral Processing 70,
157–170.
C particle concentration (p ml  1)
Jameson, G., Nam, S., Moo, Y., 1977. Physical factors affection recovery rate in
Cd drag coefficient (–) flotation. Mineral Science Engineering 9, 103–118.
d diameter (mm) Legendre, D., Sarrot, V., Guiraud, P., 2009. On the particle inertia-free collision with a
partially contaminated spherical bubble. International Journal of Multiphase
Eatt attachment efficiency (–)
Flow 35, 163–170.
Ecapt capture efficiency (–) Mei, R., Klausner, J.F., Lawrence, C., 1994. A note on the history force on a spherical
Ecoll collision efficiency (–) bubble at finite Reynolds number. Physics Fluids 6, 418–420.
Estab stabilisation efficiency (–) Moore, D.W., 1965. The velocity rise of distorted gas bubbles in a liquid of small
viscosity. Journal of Fluid Mechanics 23, 749–766.
g gravitational constant (m s  1) Nguyen, A.V., 1994. The collision between fine particles and single air bubbles in
h capture zone height (mm) flotation. Journal of Colloid and Interface Science 162, 123–128.
np particle number (part) Nguyen, A.V., 1998. Particle–bubble encounter probability with mobile bubble
surface. International Journal of Mineral Processing 55, 73–86.
Re Reynolds number (–) Nguyen, A.V., Evans, G.M., 2004. Movement of fine particles on an air bubble surface
r radius (m) studied using high-speed video microscopy. Journal of Colloid and Interface
S surface (m2) Science 273, 271–277.
Nguyen, A.V., George, P., Jameson, G.J., 2006. Demonstration of a minimum in
St Stokes number (–) recovery of nanoparticles by flotation theory and experiment. Chemical
U velocity (m s  1) Engineering Science 61, 2494–2509.
 Z. Huang et al. / Chemical Engineering Science 66 (2011) 982–997 997

Nguyen, A.V., Kmet, S., 1992. Collision efficiency for fine mineral particles with Sarrot, V., Guiraud, P., Legendre, D., 2005. Determination of the collision frequency
single bubble in a countercurrent flow regime. International Journal of Mineral between bubbles and particles in flotation. Chemical Engineering Science 60,
Processing 35, 205–223. 6107–6117.
Ralston, J., 1992. The influence of particle size and contact angle in flotation. Sarrot, V., Huang, Z., Legendre, D., Guiraud, P., 2007. Experimental determination
Developments in mineral processing 12, 203–224. of particles capture efficiency in flotation. Chemical Engineering Science 62,
Ralston, J., Dukhin, S.S., 1999. The interaction between particles and bubbles. 7359–7369.
Colloids and Surfaces A: Physicochemical and Engineering Aspects 151 (1–2), Schiller, L., Nauman, A., 1935. A drag coefficient correlation. V.D.I Zeitung 77, 318.
3–14. Schulze, H.J., 1977. New theoretical and experimental investigations on stability of
Ralston, J., Dukhin, S.S., Mishchuk, N.A., 2002. Wetting film stability and flotation bubble/particle aggregates in flotation: a theory on the upper particle size of
kinetics. Advances in Colloid and Interface Science 95, 145–256. floatability. International Journal of Mineral Processing 4, 241–259.
Ralston, J., Fornaseiro, D., Hayes, R., 1999. Bubble–particle attachment and detach- Schulze, H.J., 1989. Hydrodynamics of bubble-mineral particle collisions. Mineral
ment in flotation. International Journal of Mineral Processing 56, 133–164. Processing and Extractive Metallurgy Review 5, 43–76.
Reay, D., Ratcliff, G.A., 1973. Removal of fine particles from water by dispersed air Vinke, H., Hamersma, P.J., Fortuin, J.M.H., 1991. Particle-to-bubble adhesion in gas/
flotation: effects of bubble size and particle size on collection efficiency. The liquid/solid slurries. AICHE Journal 37, 1801–1809.
Canadian Journal of Chemical Engineering 51, 178–185. Whelan, P.F., Brown, D.J., 1955. Particle–bubble attachment in froth flotation.
Sadhal, S.S., Johnson, R.E., 1983. Stokes flow past bubbles and drops partially coated Transactions of the American Institute of Mining and Metallurgical Engineers
with thin films. Journal of Fluid Mechanics 126, 237–247. 65, 181.
Sarrot, V., 2006. Capture de fines particules par des inclusions fluides. Ph.D. Thesis, Yoon, R.H., Luttrell, G.H., 1989. The effect of bubble size on fine particle flotation.
Institut National des Sciences Appliquees de Toulouse, France. Mineral Processing and Extractive Metallurgy Review 5, 101–122.