Separation and Purification Technology 184 (2017) 240–247

Contents lists available at ScienceDirect

Separation and Purification Technology

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

The effect of particle size distribution on froth stability in flotation

A. Norori-McCormac, P.R. Brito-Parada, K. Hadler ⇑, K. Cole, J.J. Cilliers
Department of Earth Science and Engineering, Imperial College London, SW7 2AZ, United Kingdom

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

Article history: Separation of particles of different surface properties using froth flotation is a widely-used industrial pro-
Received 26 July 2016 cess, particularly in the minerals industry where it is used to concentrate minerals from ore. One of the
Received in revised form 11 April 2017 key challenges in developing models to predict flotation performance is the interdependent nature of the
Accepted 18 April 2017
process variables and operating parameters, which limits the application of optimising process control
Available online 21 April 2017
strategies at industrial scale. Froth stability, which can be quantified using air recovery (the fraction of
air entering a flotation cell that overflows in the concentrate as unburst bubbles), has been shown to
Keywords:
be linked to flotation separation performance, with stable froths yielding improved mineral recoveries.
Froth flotation
Froth stability
While it is widely acknowledged that there is an optimum particle size range for collection of particles
Particle size in the pulp phase, the role of particle size on the measured air recovery and the resulting link to changes
Air recovery in flotation performance is less well understood. This is related to the difficulty in separating particle size
Flotation performance and liberation effects.
In this work, the effects of particle size distribution on air recovery are studied in a single species (sil-
ica) system using a continuous steady-state laboratory flotation cell. This allows an investigation into the
effects of particle size distribution only on froth stability, using solids content and solids recovery as indi-
cators of flotation performance. It is shown that, as the cell air rate is increased, the air recovery of the
silica system passes through a peak, exhibiting the same froth behaviour as measured industrially. The
air recovery profiles of systems with three different particle size distributions (d80 of 89.6, 103.5 and
157.1 lm) are compared. The results show that, at lower air rates, the intermediate particle size distri-
bution (103.5 lm) yields the most stable froth, while at higher air rates, the finest particles (89.6 lm)
result in higher air recoveries. This is subsequently linked to changes in flotation performance. The results
presented here highlight, for the first time, the link between particle size distribution in flotation feeds,
air recovery and flotation performance. The results demonstrate that there is an optimal air rate for each
particle size distribution, therefore changes in particle size distribution in the feed to flotation cells
require a change in air rate in order to maximise mineral recovery.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction Johansson and Pugh [15] found that, with small quartz particles
(26–44 lm), the stability of the froth was maximised at an inter-
Froth flotation is a complex and widely used separation tech- mediate degree of hydrophobicity. Rahman et al. [26] indicated
nique, which relies on differences in the surface properties of the that fine particles increase froth stability and impede coalescence
desired particles to the unwanted particles. One of the key param- through increasing the rigidity of the bubble films in a specialised
eters in flotation is the particle size distribution of the feed and it agitated column, a finding additionally demonstrated by Ata [2].
has been shown in several studies that the optimal size range for Furthermore, Aktas et al. [1] demonstrated with a platinum ore
flotation is relatively narrow, approximately 20–150 lm [9,14]. in a lab based study that particle size had a significant effect on
Particle hydrophobicity and contact angle (in addition to hydrody- froth stability with fine particles of approximately 30–40 lm
namic factors) determine the attachment of particles to bubbles in increasing froth stability at low air rates. Conversely, Tao et al.
the pulp, which in turn affects the froth stability [17,15]. There [30] observed that coal particles of less than 30 lm diameter
remains much debate, however, on the effect of particle size on destabilised the froth in a laboratory based flotation column
froth stability. Similarly, there is no general agreement on the effect of coarse
particles on froth stability. Livshits and Dudenkov [20] suggested
that their presence retarded film coalescence in the froth phase,
⇑ Corresponding author.
improving stability, whereas Moudgil [22] indicated that there is
E-mail address: [email protected] (K. Hadler).

http://dx.doi.org/10.1016/j.seppur.2017.04.022
1383-5866/Ó 2017 Elsevier B.V. All rights reserved.
 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247 241

a tendency for coarse particles to rupture the thin films of bubbles, the effect of individual operating parameters can be investigated.
resulting in particles detaching and draining back into the froth However, operating such a system over a long period of time with
phase. Yianatos et al. [36] demonstrated that coarse particle a real ore presents a challenge. Ores are heterogeneous and min-
entrainment is insignificant using radioactive tracers in an indus- eral liberation varies, which can result in a lack of consistency in
trial flotation cell, implying attachment is the dominant mecha- operational performance within and between experiments.
nism of recovery. Separately, Zheng et al. [37] indicated that Several authors have previously developed semi-continuous
coarse particles recovered at low air rates may be recovered systems based on batch flotation cells. Kaya and Laplante [16]
through entrapment, their presence under specific operating con- modified a Denver cell, introducing a continuous feed of milled
ditions otherwise unexplained as they should settle out of the froth copper ore to allow for extended operation at steady-state to
phase. investigate entrainment at a range of air rates, although the
Recently, Liang et al. [19] showed that varying the hydrophobic- total operating time of the system is not specified. Cornell
ity and proportion of fine coal particles affects their ability to sta- et al. [6] continuously fed circuits of batch test cells with feed
bilise the froth. An intermediate particle size of 74–250 lm containing cobalt in order to optimise recovery. Vera et al.
resulted in the deepest and most stable froths. When a mixture [33] made use of a bench-scale Leeds cell in both batch and
of particle sizes was considered, a combination of approximately continuous modes, with continuous operation of up to 5 min,
55% fines
74 lm with 45% coarse particles at 250–500 lm however the system was difficult to control, was not represen-
resulted in the greatest recovery of coal to the concentrate. tative of true flotation cell design, and did not eliminate the
While the effect of particle size on froth stability and flotation issues relating to control of particle size and mineral liberation,
performance has been the focus of much research, the literature making comparisons between tests more difficult. An alternative
is contradictory on whether fine particles stabilise or destabilise approach is that of Welsby et al. [34], where a continuous flota-
flotation froths from a physics-based perspective (i.e. negating lib- tion system was developed using 5, 20 and 40 L cells, using feed
eration and surface chemistry effects). Many tests have been car- from a plant, which was held in three 230 L sumps. While this
ried out using static froth column experimental systems to reduces the effects of feed variability, it does not allow the sys-
measure froth stability (e.g. [8,30,4]), however there are few tematic investigation of particle size effects explicitly. Other lab-
reported studies into particle size effects on froth stability for a oratory flotation systems that operate continuously have been
flowing flotation froth. This is due, in part, to the difficulty in developed by Brito-Parada and Cilliers [5] and Shean et al.
decoupling particle size and liberation effects. [27]. The former used a 50 L flotation tank with continuous
For continuous flotation systems, dynamic froth stability can be recycle to test different overflowing and launder configurations
quantified using air recovery, a, as defined in Eq. (1): while, more recently, Shean et al. [27] developed a continuous
70 L laboratory flotation cell with the objective of developing
v f hw L and optimising a flotation control system based on peak air
a¼ ð1Þ
Qa recovery; however these systems were two phase (surfactant
solution and air) only.
where Q a is the air flowrate into the cell, v f the froth velocity as it While previous studies have investigated the effect of particle
overflows the lip, hw the height of the froth above the lip and L is the size distribution for specific mineral systems, or at constant air
lip length of the flotation cell. Hadler and Cilliers [10] noted that rate, in this study the objective is to develop an understanding of
increasing air rate leads to a peak in air recovery (PAR), and this cor- the interrelationship between particle size distribution, cell air rate
responds with a peak in the recovery of mineral to the concentrate. and the resulting effect on froth stability (measured using air
Furthermore, Hadler et al. [12] demonstrated that operating at PAR recovery). This will have important implications for the optimisa-
resulted in optimal recovery in banks of flotation cells, describing tion of flotation separation performance, particularly with regards
the changes as being related to changes in particle loading on bub- to process control. To this end, a continuous steady state laboratory
bles, bubble stability and froth mobility. This was further demon- system, exhibiting particle-stabilised froth behaviour that is typi-
strated by Smith et al. [28]. cal of industrial flotation cells has been developed. A single species
Air recovery and peak air recovery (PAR) were also affected by (silica) system is used in order to remove liberation and/or oxida-
variations in operating parameters such as froth depth. Hadler tion effects and to enable continuous stable operation over a period
et al. [11] showed that air recovery passed through a peak with of time.
increasing froth depth at a constant air rate, or conversely that
the PAR air rate for shallow froths is lower than that for deep
froths. 2. Methodology
There have been few studies considering particle size effects on
froth stability in industrial flotation cells. One study by Tasdemir 2.1. Experimental system
et al. [31] considered the effect of variation in operating parame-
ters on particle size in a Jameson cell, with increasing concentrate The experimental system is a 4 L closed loop, continuously
flowrate associated with increased coarse particle recovery, and an overflowing bench-scale flotation cell. The continuous recycle of
increased jet length and velocity resulting in the increase of fine concentrate as feed to the cell allows for steady state conditions
particle recovery at the expense of coarse particles. One of the chal- to be established. The design of the cell is based upon the standard
lenges of investigating the effect of particle size at industrial scale mixing tank by Costes and Couderc [7].
is that there is typically little control over the feed particle size, The cell is cylindrical and has an internal diameter of 180 mm. It
with fluctuations being driven by changes in the milling circuit. is of equal width and height, stirred by a six bladed Rushton impel-
To this end, in this study, a continuous steady-state bench-scale ler connected via a 2:1 rpm ratio stepped gearbox (supplied by
flotation system has been developed. Automation International Ltd) to the motor, supplied by Marelli
Bench-scale experiments are critical to developing an under- Motori. Four baffles are installed to a height of 120 mm at 90°
standing of the fundamental processes that govern froth flotation intervals. At the base is an air reservoir, separated from the cell
at the micro and macro scale. Standard laboratory batch flotation by a frit (manufactured by Carbis Filtration Ltd), 110 mm in diam-
tests are limited, however, by the lack of steady-state operation. eter and 1.3 mm in thickness, centred below the impeller. The frit
The benefits of a continuous laboratory flotation system are that has a mesh hole-size of 20 lm, fabricated from multiple sintered
242 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247

layers. Attached to the exterior of the cell is a launder, angled at

10° from back to front that has a width of 60 mm. The experimen-
tal system is shown in Fig. 1.
A single species system is described in this paper; this can be
used to determine the effects of changing operating variables, for
example air rate, into the flotation cell. Spherical silica ballotini
for the experiments were supplied by Sigmund Lindner. Pre-sized
distributions were sourced with ranges from 40–70 lm, 70–
110 lm and 90–150 lm. Three particle sets were created by mix-
ing the pre-sized material, obtaining the particle size distributions
presented in Fig. 2. Representative diameters for the three particle
sets are shown in Table 1, where it can be seen that particle set A
has the coarsest particles and the broadest size distribution while
particle set C is the finest of the three. This allows investigation
into the effects of relatively small variation in particle size distribu-
tion on air recovery and flotation performance.
For each experiment, 900 g of ballotini were used. The ballotini
were initially prepared by mixing with 2 L of 5  10
4 M sodium
hydroxide solution, based on a similar preparation procedure as
set out by Ata et al. [3]. This was undertaken to strip the ballotini
Fig. 2. Cumulative particle size distributions for the particle sets considered in this
surfaces of impurities. The ballotini were washed with deionized paper.
water and then mixed with 2.9 L of deionized water to form a
slurry with a pulp pH of 7.9 in suspension. The collector, Tetrade-
cyltrimethylammonium bromide (TTAB), was added at a concen-
tration of 278.4 ppb. The slurry was then allowed to mix at Table 1
915 Hz for 30 min. Methyl isobutyl carbinol (MIBC) was then Particle size for the particle sets considered in this paper.
added at 352 ppm, and the slurry allowed to mix for a further Characteristic diameter (lm) Particle set
5 min.
A B C
Air rates for the experiments were selected from a range
d10 73.3 57.0 50.7
between 10 and 30 lpm (litres per minute) as shown in Table 2.
d50 116.5 81.6 71.4
This range corresponds to superficial gas velocities of 0.66– d80 157.1 103.5 89.6
1.97 cm s
1. A further 15 min of conditioning was then allowed
with the froth overflowing before the first air rate was selected.
At this time, dosing started with additional MIBC at a rate of
4.2 ml h
1 of 102 ppm MIBC solution to maintain froth stability.
Table 2
Froth depth (from cell lip to pulp-froth interface) was maintained
Air rates and the equivalent super-
at 7.5 cm by adjusting the recycle pump as required. ficial gas velocities (Jg) investigated
in this paper.

2.2. Measurement Air rate (lpm) Jg (cm s
1)

10 0.66
For each air rate, 5 min of conditioning was allowed, before 15 0.98
5 min of on-line data collection. The height of the overflowing froth 20 1.31
25 1.64
above the lip of the cell was measured via an on-line laser (0D1300,
30 1.97
supplied by ifm electronic) and the data recorded through

Fig. 1. Schematic of the bench-scale flotation cell.

 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247 243

LabVIEW [24]. The velocity of the overflowing froth was measured 0.05 cm s
1; this is likely to be linked to the continual dosing of
using a camcorder (Panasonic Video NVGS37) through Froth- MIBC through the test.
Tracker, a block-matching image analysis program developed in- Disregarding the initial 5 min of data, over the remaining
house. The output values were then used to calculate air recovery 85 min, the 95% confidence intervals for the froth velocity and
using Eq. (1). overflowing froth height were ±0.15 cm s
1 and ±0.04 cm respec-
In order to sample the concentrate, a scoop was fitted to the lip tively. Air recovery has an average value of 43.8% with a 95% con-
of the flotation cell for 20 s at the end of each measurement period. fidence interval of ±3.1%.
The sample was weighed, and the mass of the solid and liquid
phases determined by decanting the liquid into a measuring cylin-
der. The sample was then returned to the launder. Only one sample 3.2. Air recovery characterisation
was taken for each air rate tested, as removing too much of the
concentrate leads to a destabilisation of the experimental system. The response of the measured air recovery with variation in air
rate for particle set A (d80 157.1 lm) is shown in Fig. 4 (a). Follow-
ing 5 min of stabilisation time, a further 15 min of steady state
3. Results and discussion operation was allowed at each air rate; the error bars represent
the 95% confidence interval of the air recovery over this time per-
3.1. Steady state verification iod. From Fig. 4(a) it can be seen that the laboratory flotation sys-
tem passes through a peak in air recovery as the air rate is
In order to establish the stability of the flotation system when increased. This reflects trends measured industrially, as shown in
operating continuously, a test was carried out under conditions Fig. 4(b) for a single cell, taken from the down the bank data pre-
of constant air rate (Jg = 1.31 cm s
1) over a period of 90 min. For sented in Hadler and Cilliers [10]. This demonstrates the applica-
this test, particle set A (d80 157.1 lm) was used. The variation in bility of the system for the investigation of particle size effects.
froth velocity, froth overflowing height and air recovery can be While the laboratory system comprises a single species, the
seen in Fig. 3. It should be noted that these data correspond to existence of a peak in air recovery suggests that, over the range
5 s average values. of air rates tested, the froth was particle stabilised. At low air rates,
The results show that, over the course of the experiment, the lower bubble surface area fluxes allow for higher particle loading
froth velocity and overflowing froth height remained stable. The in the pulp phase. The bubbles arriving into the froth are stabilised
overflowing froth height can be observed to drop to a stable value by the particles, however longer froth residence times and lower
over the initial 5 min of the run, whereas froth velocity increases to entrainment of water and unattached particles yield a froth with
a stable value in the first 5 min. This is linked to a period of transi- low mobility and consequently lower air recovery (i.e. more bubble
tion before steady state is established. For this reason, following an bursting). At high air rates, the opposite occurs, with higher froth
air rate change in subsequent experiments, 5 min were allowed for mobility due to an increase in liquid content, but low particle load-
steady state to be reached. It is also noted that over the course of ings resulting in more bursting. At the PAR air rate, the optimal
the test, average froth velocity increased by approximately combination of bubble loading and froth mobility occurs. This is

Fig. 3. Five second average of (a) froth velocity, (b) overflowing froth height and (c) air recovery with time for the experimental system at a Jg of 1.31 cm s
1, indicating the
system is operating at steady state.
244 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247

Fig. 4. (a) Air recovery for the bench-scale system compared with (b), air recovery as a function of Jg for an industrial flotation cell [10]. Errors given as 95% confidence
intervals.

not affected by the presence of only a single species for this particle water content of the overflowing froth in association with to very
size distribution. high air recoveries. In operating regions where the froth is not par-
ticle stabilised such as this, air recovery measurements become
unrepresentative as the assumption that the overflowing froth
3.3. The effect of particle size on air recovery has a low liquid content [35] becomes invalid. It is interesting that
this phenomena was not observed for particle set A; this suggests
The variation in air recovery for the three particle sets over a that the froth moves into an entrainment-dominated regime,
range of air rates is shown in Fig. 5. Data obtained at the higher which is more likely for finer particles.
air rates of 1.64 and 1.97 cm s
1 are not shown. For the finer size Over the range of air rates tested, the coarsest particle size dis-
distributions (particle sets B and C), at higher air rates, the froth tribution (set A) resulted in the lowest air recovery. While this may
was not particle stabilised. This was evident from the very high be linked to the destabilising effect on the froth phase through rup-
turing of the thin films [23,22,26], it can also be linked to reduced
froth mobility, with coarser particles less likely to be recovered by
entrainment. Particle sets B and C, which have similar mean sizes,
demonstrate significantly different air recoveries to each other,
particularly at the highest air rate, where particle set C, which
has the smaller d80, yields the most stable froth. Conversely, at
the lowest air rate, it is particle set B that yields the highest air
recovery. This suggests that comparatively small variations in par-
ticle size distribution can have significant effects on the magnitude
of the measured air recovery. It is helpful, however, to consider
these results together with physical measurements taken, in order
to determine the effect on flotation performance.

3.4. Indicative flotation performance: Proxies for flotation performance

for a single species system

Typical indicators of flotation separation performance are con-

centrate grade and mineral recovery. As the solid particles in this
system comprise a single species (silica), it is not possible to estab-
lish the performance of any separation, instead it is necessary to
Fig. 5. Air recovery as a function of Jg for each particle size in the feed, errors given determine proxies of performance quantifiers. For concentrate
as 95% confidence intervals. grade (the proportion of solids reporting to the concentrate that
 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247 245

are the mineral or species of interest), it is proposed that solids

content of the concentrate is a suitable proxy. A comparison
between concentrate grade and solids content for a bank of cells
for the same industrial data as given in Fig. 4(b) is given in Fig. 6.
Here it can be seen that, in general, the grade and solids content
are positively correlated down the bank. Conditions that yield high
concentrate grades are typically obtained where recovery of
entrained gangue is lower, which corresponds with lower water
recoveries and higher solids contents in the concentrate. Down
the bank, it can also be seen that there is a feed grade effect, with
lower feed grades yielding lower concentrate grades and solids
contents, suggesting there is an additional liberation effect. Note
also that froth depth is not taken into account in these data. While
subtleties in the data, such as trends in individual cells and/or with
different operating conditions, are lost in Fig. 6, the positive corre-
lation suggests that for the single species system, the use of con-
centrate solids content as a proxy for concentrate grade is
appropriate for the comparison of different operating conditions.
The flowrate of solids to the concentrate (mass pull) has been
used as an indicator of recovery at operating concentrators
Fig. 7. The relationship between solids flowrate and recovery in a bank of four
[29,13], on the basis that operating conditions that yield higher
industrial flotation cells.
mass pull also yield higher mineral recovery. This is shown in
Fig. 7 for the same industrial experimental data as shown previ-
ously, where the two measures can be seen to be positively corre- (e.g. [12,21]), while the solids flowrate initially increases, before
lated. As with the solids content and concentrate grade, it should dropping again at higher air rates. The air rate that yields the high-
be noted that this generalised approach for a bank of cells does est solids flowrate (‘recovery’) is the same as yields the peak in air
neglect subtleties in the trends. Hadler et al. [12], for example, recovery (Fig. 4(a)). In comparison, Fig. 8(b) and (c) show, respec-
showed that the mineral recovery increases together with higher tively, the normalised grade-recovery curve and the solids
mass pull with increasing air rate up to the PAR air rate. At air rates content-solids flowrate curve for the industrial cell in Fig. 4(b).
higher than the PAR air rate, the mass pull continued to increase There is good agreement between the shape of the curves from
with no corresponding increase in recovery. Similarly, the effects the industrial and laboratory data, however there is a key differ-
of feed grade and other operating parameters, such as froth depth, ence between mineral recovery in Fig. 8(b) and solids flowrate
are not considered in Fig. 7. For the single species system, where Fig. 8(c). Mineral recovery decreases at air rates above PAR, while
selective and non-selective recovery cannot be distinguished, the solids flowrate continues to increase. At high air rates, the recovery
solids flowrate will give a suitable indicator of recovery. of entrained material and water yields poor quality (i.e. low grade
The use of concentrate solids content and solids flowrate as and low solids content) concentrates. This does not lead to an
proxies for grade and recovery for the laboratory system is shown improvement in mineral recovery [12].
in Fig. 8, where Fig. 8(a) shows the indicative ‘grade-recovery’
curve for particle set A over the range of air rates given in Fig. 4
(a). It should be noted that air rate increases from the point at 3.5. The effect of particle size on indicated flotation performance
the highest solids content to the lowest. As air rate increases, solids
content decreases, which is in line with typical flotation responses The variation in solids content with air rate for the three parti-
cle sets is shown in Fig. 9. The results show that for all particle sets,
the solids content decreases with increasing air rate, a trend
observed in flotation systems in which a separation is being carried
out. This is linked to the increase in the recovery of entrained
material as air rate increases. The large error bars, particularly at
the lowest air rate, are due to manual sampling.
Although there is little difference between the concentrate
solids contents at the lowest air rate, for the higher air rates, the
coarsest particle size distribution (A) results in the highest solids
content, which can be linked to the lower air recovery and lower
rates of entrainment for coarser particles [26]. On the other hand,
there is little difference between the solids content for particle sets
B and C over any of the air rates, in spite of the higher air recoveries
obtained for the finer size distribution (C). The link between parti-
cle size and entrainment has been well-documented [32,18,25].
From Table 1, the comparison of the particle size distributions of
the three particle sets shows that sets B and C have similar distri-
butions at the fine end (d10 and d50), while set B shows a signifi-
cantly larger d80. This similarity at the fine end of the
distributions suggests that the similarity in ‘grade’ between the
two distributions is caused by similar rates of entrainment of these
fine particles.
Fig. 6. The relationship between concentrate solids content and grade for a bank of The variation in solids flowrate to concentrate with increasing
four industrial flotation cells. air rate is given in Fig. 10. Here, the solids flowrate, used as an indi-
246 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247

Fig. 8. (a) Solids flowrate with percentage of solids in concentrate for the bench-scale system, compared with an industrial flotation cell grade and recovery (b), and its
percentage of solids in concentrate with solids flowrate (c) [10].

Fig. 9. Percentage of solids in the concentrate by mass with air rate. Errors given as Fig. 10. Solids flowrate in the concentrate with particle size in the feed, for each
95% confidence intervals. particle set. Errors given as 95% confidence intervals.

cator of recovery, increases as cell air rate increases, in line with with findings from industrial cells. Furthermore, the coarsest par-
increases in air recovery shown in Fig. 5. Furthermore, the particle ticles resulted in the lowest air recovery, the highest grade and
set that yields the highest air recovery for a given air rate (i.e. set B the lowest recovery. As the proportion of coarse particles in the
for the lowest two air rates and set C for the highest) also yield the feed decreased, the air recovery did not increase monotonically,
highest solids flowrate. This result agrees with the findings of rather a peak in air recovery was observed at lower air rates;
Hadler and Cilliers [10], Hadler et al. [12] and Smith et al. [28]. It higher recovery was also obtained under these conditions. At
is also noted that under select operating conditions (i.e. lower air higher air rates, the finest particle size distribution (i.e. smallest
rates), feeds containing coarser particles can result in higher recov- d80) resulted in the highest air recovery and the highest recovery.
eries than those containing finer particles (liberation effects These findings show that fine particles can yield more stable
notwithstanding). overflowing froths, but this is dependent on the air rate. A lower
In general, the flotation results for this system show that ‘grade’ proportion of fine particles (i.e. particle set A) yielded low air
drops as air rate increases, while ‘recovery’ increases. This is in line recoveries, high grades and low recoveries over all air rates. For a
 A. Norori-McCormac et al. / Separation and Purification Technology 184 (2017) 240–247 247

similar mean particle size (d50), a greater proportion of coarser par- [8] M. Falutsu, Column flotation froth characteristics — stability of the bubble-
particle system, Int. J. Miner. Process 40 (1994) 225–243, http://dx.doi.org/
ticles (i.e. d80 for particle set B) resulted in higher froth stability at
10.1016/0301-7516(94)90045-0.
lower air rates. This demonstrates that small changes in the parti- [9] A.M. Gaudin, R. Schuhmann, A.W. Schlechten, Flotation kinetics. II. The effect of
cle size range can have significant effects on the froth stability, size on the behavior of Galena particles, J. Phys. Chem. 46 (1942) 902–910,
which affects the final flotation performance. http://dx.doi.org/10.1021/j150422a013.
[10] K. Hadler, J.J. Cilliers, The relationship between the peak in air recovery and
flotation bank performance, Miner. Eng. 22 (2009) 451–455, http://dx.doi.org/
10.1016/j.mineng.2008.12.004.
4. Conclusions [11] K. Hadler, M. Greyling, N. Plint, J.J. Cilliers, The effect of froth depth on air
recovery and flotation performance, Miner. Eng. 36–38 (2012) 248–253,
Flotation experiments carried out using a continuous bench http://dx.doi.org/10.1016/j.mineng.2012.04.003.
[12] K. Hadler, C.D. Smith, J.J. Cilliers, Recovery vs. mass pull: the link to air
scale flotation system operating at steady state have provided a recovery, Miner. Eng. 23 (2010) 994–1002, http://dx.doi.org/10.1016/j.
better understanding of the effect of particle size distribution on mineng.2010.04.007.
the stability of flotation froths. In particular, the experimental sys- [13] G.J. Jameson, New directions in flotation machine design, Miner. Eng. 23
(2010) 835–841, http://dx.doi.org/10.1016/j.mineng.2010.04.001.
tem used in this work has allowed the effects of mineral liberation
[14] G.J. Jameson, A.V. Nguyen, S. Ata, The flotation of fine and coarse particles,
to be negated by considering a silica-only system. (2007) 339–372.
This study has shown that the effect of the feed particle size on [15] G. Johansson, R.J. Pugh, The influence of particle size and hydrophobicity on
the stability of flotation froths is not only dependent on the air rate the stability of mineralized froths, Int. J. Miner. Process 34 (1992) 1–21, http://
dx.doi.org/10.1016/0301-7516(92)90012-L.
but also highly sensitive to small changes in the particle size distri- [16] M. Kaya, A.R. Laplante, Investigation of batch and continuous flotation kinetics
bution. The behaviour of the froth in the laboratory system was in a modified Denver laboratory cell, Can. Metall. Q. 25 (1985) 1–8.
shown to be comparable to that of industrial froths, allowing the [17] V.I. Klassen, V.A. Mokrousov, An Introduction to the Theory of Flotation,
Butterworth, London, 1963.
determination of the effects of particle size distribution, which is [18] H.T. Lee, S.J. Neethling, J.J. Cilliers, Particle and liquid dispersion in foams,
otherwise difficult to study at industrial scale. The solids content Colloids Surfaces A Physicochem. Eng. Asp. 263 (2005) 320–329, http://dx.doi.
in the concentrate and the solids recovery in this system were org/10.1016/j.colsurfa.2004.12.064.
[19] L. Liang, Z. Li, Y. Peng, J. Tan, G. Xie, Influence of coal particles on froth stability
shown to act as adequate indicators of grade and recovery, and flotation performance, Miner. Eng. 81 (2015) 96–102, http://dx.doi.org/
respectively. 10.1016/j.mineng.2015.07.004.
This work has also highlighted the importance of considering [20] A. Livshits, S.V. Dudenkov, Some factors in flotation froth stability, Int. J. Miner.
Process 1 (1965) 367–371.
the whole size distribution of particles when assessing the effect [21] M. Maldonado, R. Araya, J. Finch, Optimizing flotation bank performance by
that its variation has on performance, as even small changes in recovery profiling, Miner. Eng. 24 (2011) 939–943, http://dx.doi.org/10.1016/j.
the particle size distribution produce important changes in air mineng.2011.04.014.
[22] B.M. Moudgil, Enhanced Recovery of Coarse Particles During Phosphate
recovery. Among the three particle size classes used, the highest
Flotation, Final Report to Florida Institute of Phosphate Research (1992).
air recovery was obtained for the finest distribution but only at [23] B.M. Moudgil, D. Gupta, Flotation of coarse particles, in: Advances in Coal and
the higher air rate. At lower air rates, the intermediate distribution Mineral Processing Using Flotation, 1989, pp. 164–168.
resulted in the highest air recovery. This dependence on air rate [24] National Instruments, 2016. labVIEW [WWW Document]. URL http://www.
ni.com/labview/.
shows that particle size cannot be linked directly to changes in [25] S.J. Neethling, J.J. Cilliers, The entrainment factor in froth flotation: model for
froth stability without also considering operating parameters, such particle size and other operating parameter effects, Int. J. Miner. Process 93
as air rate. (2009) 141–148, http://dx.doi.org/10.1016/j.minpro.2009.07.004.
[26] R.M. Rahman, S. Ata, G.J. Jameson, The effect of flotation variables on the
Further, the particle size distribution that yielded the highest recovery of different particle size fractions in the froth and the pulp, Int. J.
air recovery at a given air rate also resulted in the highest recovery. Miner. Process 106–109 (2012) 70–77, http://dx.doi.org/10.1016/j.
This is an important finding, which supports the use of air recovery minpro.2012.03.001.
[27] B. Shean, K. Hadler, J.J. CIlliers, A flotation control system to optimise
for the optimisation and control of industrial flotation cells. performance using peak air recovery, Chem. Eng. Res. Des. 117 (2017) 57–
65, http://dx.doi.org/10.1016/j.cherd.2016.10.021.
[28] C.D. Smith, K. Hadler, J.J. Cilliers, Flotation bank air addition and distribution
Acknowledgements for optimal performance, Miner. Eng. 23 (2010) 1023–1029, http://dx.doi.org/
10.1016/j.mineng.2010.05.003.
The financial support of Rio Tinto through the Rio Tinto Centre [29] A. Supomo, E. Yap, X. Zheng, G. Banini, J. Mosher, A. Partanen, PT Freeport
Indonesia’s mass-pull control strategy for rougher flotation, Miner. Eng. 21
for Advanced Mineral Recovery in this research is gratefully
(2008) 808–816, http://dx.doi.org/10.1016/j.mineng.2008.07.002.
acknowledged. [30] D. Tao, G.H. Luttrell, R.-H. Yoon, A parametric study of froth stability and its
effect on column flotation of fine particles, Int. J. Miner. Process 59 (2000) 25–
43, http://dx.doi.org/10.1016/S0301-7516(99)00033-2.
References [31] A. Tasdemir, T. Tasdemir, B. Oteyaka, A. Tasßdemir, T. Tasßdemir, B. Öteyaka, The
effect of particle size and some operating parameters in the separation tank
[1] Z. Aktas, J.J. Cilliers, A.W. Banford, Dynamic froth stability: particle size, airflow and the downcomer on the Jameson cell recovery, Miner. Eng. 20 (2007) 1331–
rate and conditioning time effects, Int. J. Miner. Process 87 (2008) 65–71, 1336, http://dx.doi.org/10.1016/j.mineng.2007.08.007.
http://dx.doi.org/10.1016/j.minpro.2008.02.001. [32] W.J. Trahar, A rational interpretation of the role of particle size in flotation, Int.
[2] S. Ata, The role of frother on the detachment of particles from bubbles, Miner. J. Miner. Process 8 (1981) 289–327, http://dx.doi.org/10.1016/0301-7516(81)
Eng. 24 (2011) 476–478, http://dx.doi.org/10.1016/j.mineng.2010.12.008. 90019-3.
[3] S. Ata, N. Ahmed, G.J. Jameson, Collection of hydrophobic particles in the froth [33] M. Vera, Z. Mathe, J.-P. Franzidis, M. Harris, E. Manlapig, C. O’Connor, The
phase, Int. J. Miner. Process 64 (2002) 101–122, http://dx.doi.org/10.1016/ modelling of froth zone recovery in batch and continuously operated
S0301-7516(01)00066-7. laboratory flotation cells, Int. J. Miner. Process 64 (2002) 135–151, http://dx.
[4] N. Barbian, K. Hadler, J.J. Cilliers, The froth stability column: measuring froth doi.org/10.1016/S0301-7516(01)00068-0.
stability at an industrial scale, Miner. Eng. 19 (2006) 713–718, http://dx.doi. [34] S.D.D. Welsby, S.M.S.M. Vianna, J.-P. Franzidis, A continuous pilot-scale
org/10.1016/j.mineng.2005.09.021. flotation rig for the systematic study of flotation variables, Miner. Eng. 23
[5] P.R. Brito-Parada, J.J. Cilliers, Experimental and numerical studies of laudner (2010) 826–829, http://dx.doi.org/10.1016/j.mineng.2010.05.014.
configurations in a two-phase flotation system, Miner. Eng. 36–38 (2012) 119– [35] E.T. Woodburn, L.G. Austin, J.B. Stockton, Froth based flotation kinetic model,
125, http://dx.doi.org/10.1016/j.mineng.2012.03.009. Chem. Eng. Res. Des. 72 (1994) 211–226.
[6] W.L. Cornell, A.M. Wethington, D.C. Holtgrefe, F.H. Sharp, Continuous Flotation [36] J. Yianatos, F. Contreras, F. Díaz, A. Villanueva, Direct measurement of
Testing to Recover Cobalt from Missouri Lead Ores, 1987. entrainment in large flotation cells, Powder Technol. 189 (2009) 42–47,
[7] J. Costes, J.P. Couderc, Study by laser doppler anemometry of the turbulent http://dx.doi.org/10.1016/j.powtec.2008.05.013.
flow induced by a Rushton turbine in a stirred tank: influence of the size of the [37] X. Zheng, N.W. Johnson, J.-P. Franzidis, Modelling of entrainment in industrial
units – I. Mean flow and turbulence, Chem. Eng. Sci. 43 (1988) 2765–2772, flotation cells: water recovery and degree of entrainment, Miner. Eng. 19
http://dx.doi.org/10.1016/0009-2509(88)80018-6. (2006) 1191–1203, http://dx.doi.org/10.1016/j.mineng.2005.11.005.