Experimental Validation of a Computational Fluid Dynamics

Model of Copper Electrowinning

MARTIN J. LEAHY and M. PHILIP SCHWARZ

The hydrodynamics that occur in the space between the electrode plates in copper electrowin-
ning (EW) are simulated using a computational fluid dynamics model (CFD). The model solves
for the phases of gas oxygen bubbles and electrolyte using the Navier–Stokes equations in a
CFD framework. An oxygen source is added to the anode, which sets up a recirculation pattern.
The gradients in copper near the cathode lead to buoyancy forces, which result in an uplift in the
electrolyte close to the cathode. This study investigates the experimental validation of the CFD
model using a small/medium-scale real EW system. The predicted fluid velocity profiles are
compared with the experimental values, which have been measured along various cross sections
of the gap between the anode and the cathode. The results show that the CFD model accurately
predicts the velocity profile at several heights in the plate pair. The CFD model prediction of the
gas hold-up and the recirculation pattern is compared with visualizations from the experiment.
The CFD model prediction is shown to be good across several different operating conditions
and geometries, showing that the fundamental underlying equations used in the CFD model
transfer to these cases without adjusting the model parameters.

DOI: 10.1007/s11663-010-9432-y
 The Minerals, Metals & Materials Society and ASM International 2010

I. INTRODUCTION the hydrodynamics are locally such that the mixing is

COPPER electrowinning (EW) is the process of
winning copper from an electrolyte to a solid form on
a cathode by passing an electric current through the
electrolyte to attract copper ions to the cathode. Copper
EW takes place in a rectangular geometry, with two
plate electrodes opposing each other; the current moves
between the electrodes and depletes copper ions at the
cathode, whereas oxygen bubbles are generated on the
anode. It is well known that the oxygen bubbles cause a
large recirculation zone to develop in the space between
the electrodes and that this recirculation has a strong
effect on the mass transfer to the cathode because of the
mixing nature of the recirculation.[1] Copper EW oper-
ators run at a current density well below the limiting
current density, and one reason for this is that the
copper deposits become rough and have a poor quality,
which is evidenced by the nodules that form along parts
of the cathode. It is widely known that one of the main
reason these nodules form is because of the limiting
mass-transfer rate of copper to the cathode surface,[2]
whereby copper cannot be supplied locally to the
cathode surface at the rate it is being depleted from
the boundary layer solution by plating. That is to say,
MARTIN J. LEAHY, Postdoctoral Fellow, is with CSIRO Earth
Sciences & Resource Engineering and with CSIRO Minerals Down
Under National Research Flagship, Clayton, Victoria, 3168 Australia.
Contact e-mail: martin.leahy@csiro.au. M. PHILIP SCHWARZ,
Program Leader, is with CSIRO Mathematics, Informatics and
Statistics, Clayton, Victoria, 3168 Australia, and with CSIRO
Minerals Down Under National Research Flagship.
Manuscript submitted May 7, 2010.
Article published online September 14, 2010.
too poor to provide sufficient fresh copper to the
boundary layer where copper is being depleted. A
computational fluid dynamics (CFD) model of the
bubble-generated recirculation can allow insights into
the hydrodynamic behavior in the plate pair to under-
stand the detail of the mass transfer to different parts of
the cathode. In particular, a CFD model can elucidate
details of the mass transfer of copper to the cathode,
which is the limiting factor in the plating of copper to
the cathode. Detailed experimental data of the hydro-
dynamics are needed in both the bulk of the plate pair as
well as reasonably close to the cathode and anode. The
CFD model prediction of the velocity profile then can be
validated and used with confidence to predict the fluid
dynamics occurring in the EW process.
In the copper EW literature, only a limited number
of CFD modeling studies of EW have taken place, and
indeed, none has compared the model comprehensively
with experimental data of the velocity profiles in the
plate pair. Ziegler[3] and Ziegler and Evans[4] collected
limited data of the velocity profile in a large system
and compared the velocity profiles with a simple fluid
dynamics model with some success. Filzwieser’s[5,6]
studies are among the only efforts to obtain the
experimental data of the velocity profiles in an EW
plate pair. Filzwieser[6] does not compare the CFD
model developed with the experimental data but does
discuss some aspects of the CFD model, including the
basic recirculation zone that develops in the plate pair.
Filzwieser[6,7] discusses the variation in the local
copper concentration close to the cathode, as a
function of height along the cathode, but this discus-
sion is brief.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 41B, DECEMBER 2010—1247

 In other EW applications, such as water electrolysis,
to produce hydrogen, a reasonable amount of work has
been done to develop a CFD model of the two-phase
flow that occurs between the electrodes.[8–12] The fluid
dynamics in these systems is similar to that in copper
EW, in regard to the generation of oxygen bubbles on
the anode. However, hydrogen bubbles also are gener-
ated on the cathode so that in those systems the
electrolyte is forced through the plate pair at a very
fast rate (compared with copper EW); this speed
dominates the overall flow, and no recirculation zone
resulting from bubbles is set up. Therefore, although
these systems are useful for validating the underlying
fundamentals of the CFD model, they do not have the
same recirculation hydrodynamics that determine the
mass transfer of copper to the cathode. Therefore, they
are only of partial relevance to copper EW.
This study describes the validation of a CFD model of
the hydrodynamics occurring in a single plate pair of a
copper EW cell. The work represents the first time a
CFD model of the copper EW process has been
compared with experimental data of the hydrodynamics
in a single plate pair. In addition, this is the first time a
comprehensive understanding of the underlying fluid
dynamics in copper EW cells has been provided. The
CFD model has been validated successfully in an
electrorefining context[13] in single phase, thus providing
confidence in the component of the model that deals
with buoyancy-driven flows related to the copper
depletion.
The article has the following structure: a description
of the experimental setup taken from the literature is
given followed by a description of the CFD model and
the comparison between the CFD results and the
experimental data, and finally, conclusions are given.
II. EW EXPERIMENTAL SETUP
A. General Experimental Description
Filzwieser[6] presents both CFD modeling and the
experimental work in a copper sulfate/acid EW system.
A small laboratory EW cell is used, and measurements
of the cross-sectional vertical velocity at several heights
are taken by means of laser Doppler anemometry
(LDA). The scaled-down cell is transparent to allow
LDA measurements to be made and to allow photos of
the internal bubble distribution to be taken.
The cell is shown schematically in Figure 1, indicating
that the electrodes cover most of the cell wall. The
pregnant electrolyte is fed into the cell at the base
through glass bulbs to realize laminar flow, as shown in
Figure 1. The electrolyte is removed through two off-
pipes at the top of the cell, with the pipes located in the
middle of the cell and at a depth below the liquid-free
surface. It is not known to what extent the pipes affect
the fluid flow, but with the low superficial velocity used
(0.04 mm s1), they must have little effect compared
with the velocity of the gas-bubble-driven flow (magni-
tude 1 mm s1 to 100 mm s1). The exact dimension
and position of the in and out pipes are unknown, and
1248—VOLUME 41B, DECEMBER 2010
consequently, the CFD model simplifies the through-
flow and ignores the pipes. As shown in Figure 1 (right),
the cell is somewhat irregular in shape, with indents at
the top and bottom of the electrodes. Because of the
large electrode thickness of 10 mm (5 mm each) out of a
total (cross section) of 40 mm, a box-like shape is
apparent above the electrodes, which may affect the
fluid flow significantly.
A summary of the operating conditions is given as
follows:
(a) The electrolyte consisted of 0.8M copper sulfate
(50 kg m3) and 1.7M acid (165 kg m3) and is thus
a close representation of industrial cell conditions.
(b) The cell is fed continuously with electrolyte at a low
velocity (with a cell turnover time of 2 hours or a
superficial velocity of 4 9 105 m s1) to keep the
copper concentration at reasonable levels and to
maintain constant temperature.
(c) Small amounts of Cu(OH)2 are added according to
the current density to ensure the cell did not become
copper depleted.
(d) Forty parts per million Cl– is used as an inhibitor.
(e) The cell is constructed so that the cell gap could be
changed between 15 mm and 30 mm.
(f) Cell dimensions include an electrode gap of 15 mm
and 30 mm, an electrode height of 190 mm, a total
cell height of 291 mm, and electrodes both 5-mm
thick.
The details are summarized in Table I.
B. Experimental Results
Filzwieser[6] provided flow visualization plots that are
reproduced in Figure 2, which shows photos of the
experiment at steady state. These photos are provided at
three increasing current densities. Figure 2 indicates that
a fog of bubbles has developed for each current density,
as shown by the darker region, with the lighter white
region corresponding to low or close-to-zero bubble
concentrations in the electrolyte. In Figure 2, for
200 A m2 (on the left), bubbles are present in the
upper part of the cell, and a line separating the bubbly
region is clearly visible. The bubble separation line
moves lower down the cell as the current density
increases from 200 to 400 A m2, to the point where
almost all of the cell has a fog of bubbles in the
400 A m2 case. The main difference between the three
current densities is that the higher the current density,
the more extensive the bubble fog is, which occurs
because of the following reasons: first, a higher current
density means a higher volume of bubbles is produced,
and second, a higher down-flow velocity is present, thus
the reentrainment of more bubbles occurs, which are
dragged further down in the bulk. The photos for an
anode-to-cathode gap spacing of 15 mm are given in
Figure 3. When comparing Figure 2 (left) and Figure 3
(left), it is shown that a more extensive bubble fog
develops when the anode–cathode gap is 15 mm, com-
pared with a 30-mm gap for the same current density
(200 A m2). This outcome is a result of the smaller
distance between plates causing a faster down-flow of
METALLURGICAL AND MATERIALS TRANSACTIONS B
Fig. 1—Schematic experimental setup of Filzwieser,[5,6] with the anode on the right and the cathode on the left side of each schematic. Red lines
indicate LDA measurement cross section at height (H) shown. Each cross section includes approximately 70 point measurements. Reproduced
with permission from Filzwieser.[5]
electrolyte because of the smaller space for down-flow
(i.e., effective higher superficial velocity).
Filzwieser[5,6] notes the following physical aspects:
(a) The departure bubble diameter was around 50 lm.
It also was noted that the largest bubble size (near
the top of the cell) was 100 lm.
(b) The bubble-driven recirculation dominates the flow,
except in a thin layer near the cathode where natural
convection is dominant.
(c) Turbulent eddies form close to the cathode, where
downward recirculation flow driven by gas bubbles
meets upward flow as a result of natural convection
(metal depletion).
(d) A hydrodynamic boundary layer forms close to the
cathode, where convection is low and diffusion
METALLURGICAL AND MATERIALS TRANSACTIONS B
limits copper mass transport; it is expected that
copper is depleted in the boundary layer, although
this was not measured experimentally by Filzwieser.
(e) Filzwieser[5,6] notes that the hydrodynamic bound-
ary layer diminishes further up the cell as a result of
turbulence created by gas bubbles.
The data presented in Figure 4 show the vertical
velocity profiles measured by LDA from Filzwieser[5,6]
at one of four heights (95 mm). Velocity also was
measured at heights of 20 mm, 132.5 mm, and 170 mm.
The data in Figure 4 show high-velocity up-flow near
the anode as well as slow down-flow in the bulk of
electrolyte. Some up-flow occurs near the cathode. The
data show little evidence of eddies, except perhaps in the
slight nonuniformity of the down-flow; such features
VOLUME 41B, DECEMBER 2010—1249
 Table I. Table of Parameters in EW Model*
Parameter
2
Current density i (A m )
Temperature (K) (oC)
Superficial flow velocity (m s1)
Liquid density q1 (kg m3)
Oxygen density q2 (kg m3)
Liquid laminar viscosity l1 (kg m1 s1)
Oxygen viscosity l2 (kg m1 s1)
Diffusion coefficient D (m2 s1)
Transport number t+ (–)
Coefficient of Expansion b (m3 kg1)
Reference concentration Cref (kg m3)
Acid concentration (kg m3)
Dimensions of cell (height 9 width 9 depth)
Molecular weight MCu (g mol1)
*Note: viscosity, density, and diffusion coefficient of expansion values are based on a source in the literature or on the tables of data from Zaytsev
and Aseyev.[14]
Fig. 2—Photo of gas hold-up (dark areas indicate bubbles, and light
areas represent bubble-free regions) for an anode-to-cathode gap of
30 mm for three current densities of (left) 200 A m2, (middle)
300 A m2, and (right) 400 A m2. Reproduced with permission
from Filzwieser.[5]
may not be picked up by the measurement technique.
LDA averages over turbulence, so unsteady eddies will
not show up. The data in Figure 4 is a profile from the
base case (case I in Table II) and has the most
comprehensive data available, with velocity profiles
taken at each end of the electrode.
As shown in Table II, Filzwieser[5,6] also presents data
for higher current density at 400 A m2 (case II),
including the velocity profiles as in Figure 4, but only
at a height of 95 mm. Filzwieser[5,6] also presents data
for a cell gap of 15 mm at a current density of
200 A m2 (case III). For case III (200 A m2), data
were taken at heights of 20 mm and 95 mm. The
conditions for which data and photos were provided
by Filzwieser[5,6] are summarized in Table II.
1250—VOLUME 41B, DECEMBER 2010
Parameter
200, 400
323.15 K (50 C)
2.61 9 105
1200
1.2
0.835 9 103
2.18 9 105
8.62 9 1010
0.0849
0.0019 (copper)
50 (0.8M copper)
165 (1.7M acid)
291 mm 9 30 mm 9 11 cm and 291 mm 9 15 mm 9 11 cm
63.546 (copper)
Fig. 3—Photo (dark areas indicate bubbles, and light areas represent
bubble-free regions) for an anode-to-cathode gap of 15 mm for three
current densities of (left) 200 A m2, (middle) 400 A m2, and
(right) 600 A m2. Reproduced with permission from Filzwieser.[5]
III. EW CFD MODEL
A CFD model has been developed for the aforemen-
tioned experimental configuration, accounting for two-
phase bubbly flow in the EW cell. The CFD EW model
is two-dimensional in the Y-Z plane and is applied to a
cross section of the cell, as shown in Figure 5, with the
assumption that the flow is uniform in the third
(horizontal) dimension (X direction) parallel to the
electrodes. The geometry modeled is split into discrete
cells, which are used to solve the equations. In this work,
the geometry generally has the form shown in Figure 5.
A standard two-phase gas–liquid CFD model is
employed in ANSYS CFX (ANSYS, Canonsburg,
PA),[15] which treats the liquid phase (electrolyte) as
the continuous phase and the oxygen bubbles as the
METALLURGICAL AND MATERIALS TRANSACTIONS B
dispersed phase. The two phases can be envisaged as a
continuous liquid phase, with many gas bubbles dis-
persed among the continuous phase.
A mass balance equation is solved, which ensures the
conservation of mass of each phase. This process is
known as the equation of continuity, and for each phase
is given by the following:
r  ðai qi vi Þ ¼ Si
where for phase i (i = 1 is liquid and i = 2 is gas), Si
is the mass source/sink term (e.g., at the anode and
free surface where gas enters and leaves), qi is the
phase density, and vi is the velocity. A momentum
equation known as the Navier–Stokes equation is
solved for each phase, which balances the forces pres-
ent in the two-phase flow. In steady state, the Navier-
Stokes equation in vector form is expressed as follows:
r  ðai qi vi  vi Þ ¼  ai rp0 þ r  ½ai ðlL;i þ lT;i Þ
 ðrvi þ ðrvi ÞT Þ þ Mi þ Si vi
where for phase i (i = 1 is liquid and i = 2 is gas),
p’ is the (modified) pressure, and Mi is the sum of
the body forces, described subsequently. The laminar
viscosity is denoted lL,i (kg m1 s1) and lT,i
(kg m1 s1) is the turbulent viscosity, which is
described subsequently. The sum of the body forces is
given as follows:
Mi ¼ Bi þ Fi þ Ti þ Ai
Fig. 4—Example of velocity profile experimental data from Filzwie-
ser[5,6] that shows the three main parts of the velocity profile, includ-
ing the up-flow near the anode, down-flow in bulk, and up-flow near
the cathode.
where Bi, Fi, Ti, and Ai are the buoyancy force, drag
force, turbulent dispersion force, and concentration-re-
lated buoyancy force, respectively. The forces are given
by the following:
B2 ¼ B1
¼ ai ðqi  qref Þg Phase-related buoyancy force ½4
½1
3 CD
F2 ¼ F1 ¼  q a2 jv2  v1 jðv2  v1 Þ Drag force
4 d 1
½5
T2 ¼ T1 ¼ Ctd qi kra2 Turbulent dispersion force
½6
A1 ¼ a1 ½q1 gbðC  Cref Þ; A2 ¼ 0
Concentration-related buoyancy force liquid ði ¼ 1Þ
½2
½7
2 1
where g (m s ) is the gravity vector, v2 – v1 (m s ) is
the slip velocity, |v2 – v1| is the size (modulus) of the slip
velocity, qref (kg m3) is the reference density taken as
that of the electrolyte (i = 1), CD is the drag coefficient,
Ctd is the turbulent dispersion coefficient (taken as 1),
k (m2 s2) is the turbulence kinetic energy, C (kg m3) is
the concentration of copper, Cref (kg m3) is the
½3 reference concentration of copper (initial and inlet
concentration), and b (m3 kg1) is the coefficient of
expansion for the copper species.
The turbulent viscosity in Eq. [2] is determined by
solving turbulence transport equations using the well-
known k–x model.[16] This model originally was derived
for single-phase flows, but it can be used for multiphase
flows by solving the k–x model for the continuous liquid
phase and by using the same turbulence quantities (k, x)
for the dispersed gas phase. Typically, the gas turbulent
viscosity (lT;2 ) is multiplied by the ratio of the phases’
density difference and then by dividing by the turbulent
Prandtl number r (taken as 1 when the particle
relaxation time is short compared with the turbulence
dissipation time scales, such as with small dispersed
particles, as in this work). We adopt this method[15] in
this work; in particular, we use the following relation-
ship to define the turbulent eddy viscosity for the gas
phase:
q2 lT;1
lT;2 ¼ ½8
q1 r

Table II. Cases Where Data From Filzwieser[6] is Used

Electrode Current No. Velocity Height from Base Photo and Streak
Gap (mm) Density (A m2) Profiles of Electrode (mm) Vectors Provided?
Case I (base case) 30 200 4 20, 95, 132.5, 170 Yes
Case II 30 400 1 95 Yes
Case III 15 200 2 20, 95 Yes

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 41B, DECEMBER 2010—1251

Fig. 5—Schematic CFD geometry at the side and cross section and the cross-section mesh view. High aspect ratio of the geometry causes a high
cell aspect ratio. A much larger number of cells than shown actually is used in the model.

The drag coefficient is dependent on the bubble size, where SCu (kg m3 s1) is the source term, which
and for small bubbles, this dependency is described describes the flux of copper at the cathode or the source
well by the following Schiller–Nauman equation: or sink of copper at the inlet and outlet, respectively,
D (m2 s1) is the diffusion coefficient of copper ions, and
24

CD ¼ 1 þ 0:15Re0:687 ½9 ScT (–) is the turbulent Schmidt number, which typically
b
Reb is given a value of 0.9, as in this work.
where Reb (–) is the bubble Reynolds number given by
the following:
A. Boundary Conditions on the Anode
db q1 jv2  v1 j
Reb ¼ ½10 The geometry shown in Figure 5 is useful to describe
l1 the boundary conditions used. On the anode side where
oxygen is produced, the following boundary condition
where db (m) is the characteristic length scale (bubble
for the superficial gas velocity in the Y direction v2,Y
diameter). For 50-lm diameter bubbles, as noted in the
(m s1) is based on Faraday’s Law. If we assume all
experiment by Filzwieser,[5,6] the bubble Reynolds num-
current is converted to oxygen gas at the anode wall
ber is Reb~0.1 causing CD to be large, which means that
surface, then the superficial gas production rate is given
the Stokes drag regime is prevalent, where the drag is so
by the following:
high the two phases have almost the same velocity.
A closure equation is required for the volume fraction
1 iRT
equations, and is given by the following: v2;Y ¼ ½13
4 Patm F
a1 þ a2 ¼ 1 ½11
where R (J K1mol1) is the gas constant, T is the
The additional transport equation for the copper spe- temperature (K), Patm (Pa) is the atmospheric pressure,
cies (Cu2+) in the liquid phase is expressed in steady F (A s mol1) is Faraday’s constant, and i (A m2) is
state as follows: the current density, which is assumed constant at all
     points in the cell and along the electrodes; in future
lT;1 C
r  ða1 Cv1 Þ ¼ r  a1 q1 D þ r þ SCu ½12 work, this assumption could be addressed to include a
ScT q1 variable current density.

1252—VOLUME 41B, DECEMBER 2010 METALLURGICAL AND MATERIALS TRANSACTIONS B

 The oxygen mass flow rate m_ oxygen can be calculated
based on the superficial velocity from Eq. [13] as
follows:
m_ oxygen ¼ v2;Y Aan q2 ½14
where Aan (m2) is the area of the anode.
B. Boundary Conditions on the Cathode for Copper
The flux of copper at the anode and cathode walls
m_ Cu (kg m2 s1) based on Faraday’s Law is expressed
as follows:
i MCu
m_ Cu ¼  ½15
zF 1000
where i (A m2) is the current density, F (A s mol1)
is Faraday’s constant, z (–) is the valency, and MCu
(g mol1) is the molecular weight of copper. The
boundary condition for Eq. [15] at the cathode wall
(Figure 5) is essentially the diffusional flux of copper
ions, which for a binary electrolyte or a dilute solution
of copper in a sulfuric-acid-supporting electrolyte, can
be expressed approximately as follows[17]:
ið1  tþ Þ MCu
m_ Cu ¼ ½16
zF 1000
where t+ () is the transference number (t+ ~8.49 pct in
Table I), which is defined as the proportion of current
carried by copper ions in a uniform solution without
concentration gradients.[17] At all walls, no slip bound-
ary conditions are applied, whereas at the top free
surface, a wall boundary with a free slip (no friction)
boundary condition is applied to simulate a quiescent
free surface. At the walls, the grid resolution is such that
Y+ = 1, and therefore, in the k–x formulation, inte-
gration in the CFD solver is carried out to the wall.
C. Wall and Free Surface Boundary Conditions
At all walls, no slip boundary conditions are applied
for the liquid phase, whereas free slip boundary
conditions are applied for the gas phase. At the free
surface, a free slip (no friction) boundary condition is
applied to the liquid phase, whereas for the gas phase,
a degassing boundary condition is used. The degassing
boundary condition allows gas bubbles to leave the
liquid through the surface at the rate at which they
arrive at the surface.
D. Inlet and Outlet Boundary Conditions
An inlet and outlet is added near the base to the
geometry to allow through-flow and for the introduc-
tion of copper to avoid depletion. This is not exactly
the same through-flow configuration as in the exper-
iment (Figure 1), but the details of the experimental
arrangement are difficult to ascertain, and as men-
tioned previously, through-flow velocities are much
smaller than those generated by the bubbles, so the
exact through-flow configuration is unlikely to have a
significant effect on the simulation results. At a current
METALLURGICAL AND MATERIALS TRANSACTIONS B
density of 400 A m2, a higher through-flow velocity
(increase by a factor of two) is used in the CFD model
to avoid copper depletion. However, in the experiment,
this is not the case; instead, small amounts of Cu(OH)2
are added according to the current density. This
discrepancy in the experimental flow rate and the
CFD flow rate for this high current density case has
insignificant effects on the flow because of the low flow
rates compared with the gas-generated flow rates,
which are several orders of magnitude higher; this
setup merely simulates the CFD-predicted copper
topped up, which could have been done in another
nominal way. The parameters used are shown in
Table I.
IV. VALIDATION OF CFD EW MODEL
The experimental data of Filzwieser[5,6] were discussed
in Section II and now are used to compare with the CFD
model. The computational mesh used for the simulation
has a cell spacing that is finer near the walls than in the
middle to resolve the higher velocity gradients near the
walls. The operating conditions are the same as those
used in the experiment, as described in Section II. The
parameters used in the CFD model, including boundary
conditions, are given in Table I.
A. Comparison with Case I LDA Data
The experimental data in case I (the base case) were
the most comprehensive (with four velocity profiles).
Case I used a current density of 200 A m2 and an
electrode gap of 30 mm. In this section, we show the
comparison of CFD results with case I data, as given in
Figures 6(a) through (d), for the heights (from the base
of the electrodes) of 20 mm, 95 mm, 132.5 mm, and
170 mm, respectively. These positions are shown clearly
in Figure 1. The results at the height of 20 mm (from the
base of the electrodes) in Figure 6(a) show a fairly good
agreement between the experimental data and the CFD
prediction, with most parts of the profile in good
agreement, except for an overprediction of the maxi-
mum velocity near the anode where the oxygen gener-
ated rises quickly and drags liquid upward. In the
middle of the cell, the data indicates slow downward
flow, resulting from the electrolyte recirculation zone
present. This flow is predicted well by the CFD model.
Near the cathode, an upward flow is caused by copper
depletion and by the associated natural convection
buoyancy, which also is predicted closely by the CFD
model, with the shape and maximum velocity in close
agreement.
At the higher positions of 95 mm, 132.5 mm, and
170 mm from the electrode base (Figures 6(b) through
(d)), the agreement between CFD and the experiment is
also good, with close comparison at all parts of the cross
section, including the correct maximum velocity near the
anode. In the middle section, good agreement occurs in
all cases. The maximum velocity near the cathode is in
close agreement in most cases.
VOLUME 41B, DECEMBER 2010—1253
Fig. 6—Comparison of CFD results with case I experimental data from Filzwieser[6] for 200 A m2 and a 30-mm gap. Vertical velocity compo-
nent (mm s1) vs distance from cathode (mm) at a height of (a) 20 mm (from base of electrode), (b) 95 mm, (c) 132.5 mm, and (d) 170 mm. In-
set figure shows a close-up near the cathode.
The highest positions of 132.5 mm and 170 mm from
the electrode base (Figures 6(c) and (d)) experience some
wavy behavior in the CFD prediction in the bulk, which
is a result of unsteady eddies that form in the bulk,
whereas the data indicate that the velocity profile is flat.
The data points represent an average of the eddies (and
the associated fluctuations in the velocity) that may be
present in the bulk. Future work is required to establish
whether the CFD eddies can be averaged over time to
give the same overall flat velocity profile.
Mostly, a relative difference of around 20 pct is noted
between the data and the CFD values of the maximum
velocity near the cathode and the anode, although the
relative difference often was much less than this. This
result is considered satisfactory given the limited knowl-
edge of the operating conditions, assumptions required,
possible errors in liquid velocity measurements (partic-
ularly in the bubbly region), and complexity of modeling
multiphase flow.
1254—VOLUME 41B, DECEMBER 2010
B. Comparison between CFD and Experimental
Visualization for Case I
The streak vector plot taken from the observations of
case I is shown in Figure 7(a) together with the
associated CFD prediction of the vector field for case
I. The CFD vector plot in Figure 7(b) (normalized)
shows that a fairly good qualitative agreement exists
between the CFD and the schematic streak plot of the
experimental vector field in regard to the behavior and
flow pattern, including a large recirculation zone and
up-flow near the anode caused by the bubbles as well as
an up-flow near the cathode as a result of copper
concentration gradients. Agreement also was found in
the down-flow in the bulk. The shape of the velocity
profile (as shown in Figure 6 from LDA measurements)
is reflected in the schematic streak plot of the experi-
mental vector field and the CFD prediction of the vector
field in Figures 7 and 8(a) in regard to the characteristics
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 7—Case I (a) experimental schematic sketch of the vector field
from the experiment (reproduced with permission from Filzwieser,[5,6])
and (b) the CFD vector field (with vector renormalization for clarity
of all vector scales), with a close-up near the cathode (top inset,
no vector renormalization) and anode (bottom inset, no vector
renormalization).
of fast up-flow near the anode, low level of up-flow near
the cathode, and low level of down-flow in the bulk.
Looking at the CFD results in Figure 9, a close up of
the anode wall is shown where oxygen bubbles are
produced (see oxygen volume fraction in Figure 9(a))
and remain extremely close to the anode, predominantly
in the first computational cell. The gas remains in the
tight region close to the anode because there, only
advection is transporting this phase, despite a turbulent
dispersion force being added to the gas phase. Gas rises
upward (see superficial velocity in Figure 9(b)), which
drags the electrolyte upward (see electrolyte velocity in
Figure 9(a)). As shown in Figure 9(a), after leaving the
top of the anode, the oxygen moves into the top region
and is dispersed by advection. Figure 8(a) shows on a
larger scale view that the gas then is carried upward to
the free surface; the electrolyte cannot leave the system,
but gas can leave the system depending on the local
hydrodynamics. However, some oxygen is carried back
down into the bulk, dragged along with the electrolyte,
which occurs fairly often, and results in a gas hold-up in
the bulk of the top square block region of up to 5 pct.
The gas hold-up in the bulk in the middle of the top of
the electrode region is up to 2 pct, which decreases
downward because of the down-flow velocity decreasing
downward and also because the bubbles tend to rise as a
result of their buoyancy.
The experimental visualization in Figure 8(c) shows
the distribution of oxygen gas hold-up (the dark regions
correspond to high gas hold-up, and the light regions
correspond to low gas hold-up) compared with the
CFD-predicted oxygen gas hold-up (Figures 8(a) and
METALLURGICAL AND MATERIALS TRANSACTIONS B
(b)); broadly, good agreement is found in that the hold-
up is high at the top and decreases down in the bulk.
CFD and experimental agreement also is noted for a
sharp decrease in the gas hold-up near the cathode (i.e.,
a decrease occurs in the oxygen volume fraction close to
the cathode extending upward, as indicated in Figure 8).
This finding is true in three cases (cases I, II, and III), as
indicated on the CFD contour plot of volume fraction
(Figure 10 [bottom]) vs the experimental (Figure 10
[top]). Also note the line plot inset in Figure 10(a) and
the close-up in Figure 8(b). This region of low gas hold-
up near the cathode is a result of the following: a closed
recirculation vortex coincides with this elongated bub-
ble-free region close to the cathode, which is cut-off
from the main large recirculation coming from the
anode where bubbles originate. The recirculation in this
elongated bubble-free region close to the cathode is
driven by copper gradients and by the associated natural
convection recirculation pattern (Figure 8(a)), which are
included in the CFD model and predict the correct type
of behavior in approximately the correct position.
C. Comparison between CFD and Experimental
Visualization for Cases I Through III
The photos taken during the experiment (which
indicates gas hold-up) are shown in Figure 9 in addition
to the CFD prediction of the velocity and oxygen
volume fraction for cases I through III. This allows a
comparison between the CFD-predicted oxygen volume
fraction (gas hold-up) and the visual images taken from
the experiment. It is shown that a reasonable agreement
exists in the oxygen volume fraction (or gas hold-up) in
regard to the distance the bubbly region extends toward
the bottom of the cell.
The value of the oxygen volume fraction predicted by
the CFD model is higher for case II compared with case
I (as in Figure 10) because of the higher down-flow
velocity and more bubble entrainment into the bulk,
which is caused by the factor of two increase in the
current density that provides double the oxygen flow
rate up the anode (because of Faraday’s Law in
Eq. [13]), and thus, more reentrainment of bubbles
occurs. This is a result of the following factors: (1) a
higher gas flow increases the up-flow velocity as well as
the down-flow velocity, thus enabling more reentrain-
ment of bubbles caught by the down-flow at the free
surface; and (2) a higher gas flow increases the proportion
of bubbles, which then can be reentrained into the bulk.
For case II, the maximum oxygen volume fraction in
the bulk predicted by the CFD model in the bulk is 0.05
compared with 0.02 for case I. Qualitatively, agreement
occurs in CFD and the experiment between cases I and
II in the lower penetration of the gas bubbles.
Comparing cases I and III in Figures 9(a) and (c), we
observe that for the same current density but a smaller
gap width, greater gas hold-up occurs; this finding is
confirmed in the photos that show the gas hold-up taken
during the experiment and also in the CFD prediction of
the oxygen volume fraction. The higher gas hold-up is a
result of the smaller gap width leading to faster down-
flow, thus dragging more bubbles downward.
VOLUME 41B, DECEMBER 2010—1255
Fig. 8—Case I (a) streamlines of electrolyte flow, electrolyte vector plot, and volume fraction showing recirculation patterns as well as a
(b) close-up of the box in (a), and an (c) experimental visualization for comparison. Bold arrows indicate similar bubble-free regions near the
cathode, indicating qualitative agreement between the experiment and CFD. Streamlines show that a closed recirculation region coincides with
this bubble-free region.
In summary, the overall trend of the CFD model is in
agreement with the experimental photos. However, the
penetration of the gas bubbles toward the bottom of the
bulk as predicted by the CFD is not as extensive as
the photos for the three cases in Figure 9 indicate. This
shortcoming may be a result of the CFD model for the
following reasons:
(a) The assumption of a single bubble size and/or
overestimating the average size chosen (50 micron)
(b) Underpredicting the bubble dispersion into the
bulk, which does not occur significantly along the
anode but mainly near the top of the free surface
(c) Ignoring three-dimensional side-wall effects that
were present in the experiment, as this may be
affecting the fluid flow significantly.
D. Comparison of Case II LDA Data
We now show the comparison between the CFD ver-
tical velocity and the experimental LDA measurements
1256—VOLUME 41B, DECEMBER 2010
for case II (summarized in Table II) in which the anode–
cathode gap is 30 mm and a current density of
400 A m2 is used. Figure 11 shows the comparison
between the CFD and the experimental data, indicating
a fairly good agreement between the CFD and the
experiment in the maximum value in the velocity at the
anode and cathode sides as well as in the down-flow
velocity in the middle section. A higher velocity is
observed near the anode and in the bulk in both the
experiment and the CFD compared with case I for
200 A m2 (compare the velocity profile for case I in
Figure 6(b) with case II in Figure 11). This follows
because of the doubled current density (and doubled gas
flow rate) for the same gap width, which causes a faster
velocity near the anode and in the bulk.
E. Comparison of Case III LDA Data
We now show the comparison for case III (summa-
rized Table II) in which the electrode gap is 15 mm and
the current density is 200 A m2. Figures 12(a) and (b),
METALLURGICAL AND MATERIALS TRANSACTIONS B
Fig. 9—Close-up near the anode (near top), depicting the (a) fill contours of the oxygen volume fraction (–), electrolyte vector plot, and stream-
lines of electrolyte and (b) the filled contours of the oxygen superficial velocity (mm s1), oxygen superficial velocity vectors (mm s1), and a
streamline of the oxygen phase.
show the comparison at positions of 20 mm and 95 mm
from the base of the electrode, respectively. The CFD
results for both heights show a close comparison with
the experimental data at all parts of the profile, except
for an underprediction close to the anode. The up-flow
near the cathode is predicted well by the CFD model.
V. CONCLUSIONS
A CFD model has been developed to simulate flow
and copper distribution in an EW cell consisting of a
single anode–cathode pair. The model incorporates the
transport of oxygen bubbles generated on the anode
and copper in the electrolyte, including the deple-
tion of copper at the cathode and density-related
buoyancy forces resulting from metal concentration
gradients.
METALLURGICAL AND MATERIALS TRANSACTIONS B
The experimental setup used by Filzwieser[5,6]
involves a copper sulfate and acid mixture in a medium
sized cell with through-flow; thus, the experiment is a
good representation of a scaled-down real EW cell.
Data from three cases with different current densities
and interelectrode gaps compared well with the CFD
model.
The conclusions of the study are as follows:
1. The CFD model closely can predict the behavior in
the scaled EW cell across a range of operating con-
ditions and electrode spacings.
2. The CFD-predicted velocity profile is close to
experimental data in most parts of the cross section,
with flow near the anode and cathode in good
agreement. For the base case, a relative difference of
around 20 pct was noted between the data and the
CFD values in the maximum velocity near the cath-
ode and anode, although often, the relative difference
VOLUME 41B, DECEMBER 2010—1257
Fig. 10—Comparison of a photo indicating gas hold-up (top) (Reproduced with permission from Filzwieser[5,6]) and CFD-predicted gas hold-up
(or oxygen volume fraction) (bottom) for the following cases: (a) case I (CD200 EA30) with an inset of a line plot of the oxygen volume fraction
at a height of 170 mm, (b) case II (CD400 EA30), and (c) case III (CD200 EA15). Th figure indicates that CFD predicts bubble behavior
reasonably well across three operating conditions. In CFD, the scale is shown with a maximum of 0.15, although the maximum oxygen volume
fraction is 0.88 at the top of anode.

1258—VOLUME 41B, DECEMBER 2010 METALLURGICAL AND MATERIALS TRANSACTIONS B

 was much less than this. This outcome is consid- because of a lack of information about the exact
ered satisfactory given the unknowns, assumptions experimental configuration.
required, possible errors in the data, and complexity
4. Photos from the experiments of the voidage were
of modeling multiphase flow.
compared with the CFD model, which established
3. Some differences in the overlay of CFD results and
the following:
experimental data of the velocity are attributed to
complicating issues including, (a) Trends in the CFD-predicted gas hold-up are in
agreement with the experiments, with the higher
(a) Wall effects that were present in the experiment,
current densities and a smaller gap width lead-
which may be affecting the fluid flow significantly
ing to greater gas hold-up
(b) The CFD model exhibits some unsteady behavior,
(b) Good agreement was noted in the existence of a
with eddies present in the bulk
vertically elongated bubble-free region close to the
(c) The through-flow conditions in the CFD model
cathode, resulting from a closed recirculation
are not set up exactly the same as in the experiment
vortex that is cut off from the main large recir-
culation coming from the anode where bubbles
originate. The recirculation in this elongated
bubble-free region close to the cathode is driven
by copper gradients and the associated natural
convection recirculation pattern.
5. Streak vector plots from the experiments were com-
pared with the CFD model, with general agreement
observed in the following features:
(a) Large recirculation, especially in the region
above the electrodes,
(b) Up-flow near the anode and cathode
(c) Down-flow in the bulk
6. Generally, the CFD model underpredicted the gas
hold-up in the bulk, which may be caused by a
combination of the following:
(a) The assumption of a single bubble size and/or
overestimating the average size chosen (50
Fig. 11—Comparison of CFD results with the case II experimental
data from Filzwieser.[5,6] The vertical velocity component (mm s1)
micron)
vs the distance from the cathode (mm) at a height of 95 mm from (b) Underestimating bubble dispersion along the
the base of the electrode at 400 A m2 and a 30-mm gap. anode

Fig. 12—Comparison of CFD results with the case III experimental data from Filzwieser,[5,6] with 200 A m2 and a 15-mm gap. The vertical
velocity component (mm s1) vs distance from the cathode (mm) at a height of (a) 20 mm from the base of the electrode and at (b) 95 mm from
the base of the electrode.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 41B, DECEMBER 2010—1259

 ACKNOWLEDGMENTS GREEK SYMBOLS
The authors gratefully acknowledge funding from ai volume fraction of phase i
the AMIRA P705 sponsors. The authors also grate- b coefficient of expansion (m3 kg1)
fully acknowledge Mike Horne, for his help with the li liquid laminar dynamic viscosity (kg m1 s1)
experimental work and for his helpful discussions, and lT,i turbulent dynamic viscosity (kg m1 s1)
Peter Witt, Darrin Stephens, and Graeme Lane for qi density of phase i (kg m3)
their helpful CFD assistance and discussions. Mike rT turbulence Schmidt number (–)
Nicol is acknowledged for his helpful discussions. x Eddy frequency (s1)

SUBSCRIPTS

NOMENCLATURE An anode
atm atmospheric
A area (m2) b bubble
Ai concentration related buoyancy force of phase Cu copper
i (N m3) i phase i (i = 1 water, i = 2 gas)
Bi natural convection buoyancy force of phase i ref reference
(N m3) T turbulent
C copper concentration (kg m3)
Cref average concentration of copper over the
SUPERSCRIPTS
cathode (kg m3)
DC difference in bulk to wall concentration of T transpose
copper (kg m3) ‘ modified pressure p’
CD drag coefficient
Ctd turbulent dispersion coefficient
D diffusion coefficient (m2 s1)
REFERENCES
F Faraday’s constant (A s mol1)
Fi drag force of phase i (N m3) 1. J. Graydon and D. Kirk: Can. J. Chem. Eng., 2001, vol. 69,
pp. 564–70.
g gravitational acceleration vector (m s2) 2. G. Rigby, P. Grazier, A. Stuart, and E. Smithson: Chem. Eng. Sci.,
H height of electrode (mm) 2001, vol. 56, pp. 6329–36.
h width of electrode (mm) 3. D. Ziegler: Ph.D. Dissertation, University of California, Berkeley,
i current density (A m2) CA, 1984.
k kinetic energy (m2 s2) 4. D. Ziegler and J. Evans: J. Electrochem. Soc., 1986, vol. 133 (3),
pp. 567–76.
m _ Cu flux of copper at the cathode walls 5. A. Filzwieser, K. Hein, G. Hanko, and H. Grogger: Proc. Copper-
(kg m2 s1) Cobre 99 Vol III. – Electrorefining and Electrowinning of Copper,
_ oxygen
m flux of oxygen at the anode and cathode walls 1999, pp. 695–709.
(kg m2 s1) 6. A. Filzwieser: Ph.D. Dissertation, Leoben University, Leoben,
Austria, 2000.
MCu molecular weight of copper (g mol1) 7. A. Filzwieser, K. Hein, and G. Mori: JOM, 2002, pp. 28–31.
Mi sum of body forces of phase i (N m3) 8. K. Aldas: Appl. Math. Comput., 2004, vol. 154, pp. 507–19.
Patm atmospheric pressure (Pa) 9. M. Mat and K. Aldas: Int. J. Hydrogen Energ., 2005, vol. 30,
p pressure (Pa) pp. 411–20.
p’ modified pressure (Pa) 10. R. Wedin, L. Davoust, A. Cartellier, and P. Byrne: Exp. Therm.
Fluid Sci., 2003, vol. 27, pp. 685–96.
R cas constant (J/K/mol) 11. P. Boissonneau and P. Byrne: J. Appl. Electrochem., 2000, vol. 30,
Reb bubble Reynolds number (–) pp. 767–75.
ScT turbulent Schmidt number (–) 12. V. Agranat, S. Zhubrin, A. Maria, J. Hinatsu, M. Stemp, and M.
t+ transference number (–) Kawaji: Proc. of FEDSM2006, 2006, pp. 1–8.
13. M. Leahy and M.P. Schwarz: 16th Australasian Fluid Mechanics
Ti turbulent dispersion force of phase i (N m3) Conf, Gold Coast, Australia, 2007.
T temperature (K) (C) 14. I. Zaytsev and G. Aseyev: Properties of Aqueous Solutions of
vi velocity vector of phase i (m s1) Electrolytes, CRC Press, Boca Raton, FL, 1992, pp. 1–1081.
X X direction coordinate (m) 15. ANSYS, Inc.: CFX-11 Solver, www.ansys.com/cfx, 2007.
Y Y direction coordinate (m) 16. D. Wilcox: Proc. of the 24th AIAA Aerospace Sciences Meeting,
American Institute of Aeronautics and Astronautics, Reno, NV,
Z valency (–) 1986, pp. 1311–20.
Z Z direction coordinate (m) 17. J. Newman and K. Thomas-Alyea: Electrical Systems, 3rd ed.,
Z’ effective height above base of electrodes (m) Wiley, Hoboken, NJ, 2004, p. 14.

1260—VOLUME 41B, DECEMBER 2010 METALLURGICAL AND MATERIALS TRANSACTIONS B