Sunteți pe pagina 1din 10

This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Electrochimica Acta 89 (2013) 717725
Contents lists available at SciVerse ScienceDirect
Electrochimica Acta
j our nal homepage: www. el sevi er . com/ l ocat e/ el ect act a
Kinetic and hydrodynamic implications of 1-D and 2-D models for copper
electrodeposition under mixed kinetic-mass transfer control
Jorge Vazquez-Arenas

, Mark Pritzker, Michael Fowler


Chemical Engineering Department, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada
a r t i c l e i n f o
Article history:
Received 7 September 2012
Received in revised form6 November 2012
Accepted 7 November 2012
Available online 15 November 2012
Keywords:
Electrodeposition
Fluid mechanics
Diffusion
Voltammogram
Rotating disk electrode
a b s t r a c t
Numerical analysis of the hydrodynamics and kinetics of copper deposition on a copper rotating disk
electrode during voltammetry at a scan rate of 100 mV s
1
is conducted to determine the impact of t-
ting for the system parameters using a 1-dimensional model when the process operates under transition
(1000 rpm) and turbulent (1500 rpm) velocity ow conditions. The 1-dimensional model is based on
solution of the transient mass transport equation for Cu(II) utilizing the well-known series solution by
von Karman and Cochran for the uid velocity. The suitability of using this model is assessed by com-
parison to an axially symmetric 2-dimensional model of the system. In the 2-dimensional model, the
Reynolds-averaged NavierStokes (RANS) equations are solved rst for the pressure and the uid veloc-
ity components which are then used in the transient mass transport equation. The turbulent viscosity is
estimated using the k model. In both models, transport of Cu(II) by diffusion and convection is cou-
pled to the ux associated with its reduction at the cathode surface. The velocity obtained from the
1-dimensional model is found to be accurate only at locations within a distance of 1.4 10
5
m from
the electrode. Although no signicant differences are found for the tted kinetic parameters obtained
using the two models, the diffusion coefcient of Cu(II) tted using the 1-dimensional model exceeds
that obtained from the 2-dimensional model by approximately 44%. The model with the tted parame-
ters can accurately predict the experimental behavior at a different CuSO
4
concentration, although larger
deviations arise when the scan rate is reduced to 5mV s
1
.
2012 Elsevier Ltd. All rights reserved.
1. Introduction
Newadvances in computer architecture (CPUs and memory) as
well as the development of specialized software (i.e. COMSOL

,
Fluent

, Matlab

) have enabled models for electrochemical pro-


cesses involving the effects of uid motion and mass transfer
to be developed and solved numerically with fewer restrictive
assumptions [14]. Several experimental and modeling studies of
copper deposition have been conducted under different conditions
of electrolyte composition, pH, temperature, etc [512]. Although
the electrode reaction mechanism for this system has been ana-
lyzed using physicochemical models incorporating mass-transport
effects [69], relatively few analyses of copper deposition involv-
ing the hydrodynamics in a comprehensive or rigorous manner
have been reported in the literature [13]. In the case of deposition
onto a rotating disk electrode (RDE), the system is usually treated
as a 1-dimensional problem and the NavierStokes equations are
solved for the uid velocity in terms of an innite series expansion

Corresponding author. Tel.: +1 519 888 4567x31622; fax: +1 519 888 4347.
E-mail addresses: jgvazque@uwaterloo.ca, jorge gva@hotmail.com
(J. Vazquez-Arenas).
[13,14]. The validity of this series solutionrequires that the Schmidt
number for this system be large, a condition that is easily met
for transport in aqueous solutions [15]. However, the experimen-
tal conditions used in virtually all RDE studies are inconsistent
with several other assumptions or conditions on which this solu-
tion for the NavierStokes and mass transport equations is based.
First, the various phenomena such as uid ow and reaction rates
do not always occur uniformly over the electrode surface [16]. In
the model proposed by von Krmn [13], the disk electrode has
an innitely large radius and zero thickness and rotates slowly
enough that laminar owconditions prevail in the electrolyte dur-
ing deposition [16]. However, in practice, higher rotation speeds
in the turbulent region are usually applied to ensure uniformmass
transfer over the electrode surface [17]. Additionally, although the
owin the boundary layer remains laminar for Reynolds numbers
(Re =r
2
/) up to about 210
5
, the owat larger radial positions
becomes turbulent [16]. Consequently, a model describing uid
owover theentirecell shouldbebasedontheNavierStokes equa-
tions for turbulent conditions. To avoid solving the NavierStokes
equations for turbulent regimes, approaches that have been used
are to develop empirical expressions in terms of the Sherwood
number and eddy diffusivities to account for the mass transfer
rates under transition and turbulent ow conditions. Although
0013-4686/$ see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electacta.2012.11.024
Author's personal copy
718 J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725
these expressions were shown to be successful for electrochem-
ical systems [18,19], they have not been widely used since their
implementation for data analysis can be cumbersome. In general,
most studies concerned with the analysis of experimental data rely
on faster and simpler methods using the Levich equation or simple
1-dimensional models.
The von Krmn model [13] assumes the electrode to be
immersed face upward in an electrolyte located above it that can
be modeledas a semi-innite domain. Consequently, the inuences
of specic cell geometry and orientation and events occurring far
from the electrode on uid ow and electrode reactions are con-
sidered to be negligible [16]. The commonly used series solution
for the uid velocity obtained by Cochran [14] is valid only in close
proximity to the electrode surface. The usual practice of truncat-
ing this series to the rst termfurther restricts its validity to small
distances from the electrode and low rotation speeds and poten-
tially increases the error in applying this solution to more realistic
systems.
Given these uncertainties regarding the 1-dimensional mod-
els, a number of research groups have recently presented 2- and
3-dimensional models for electrochemical cells with a RDE and
solved them using computational uid dynamics (CFD) to more
thoroughly describe the behavior in these cells and to assess
the accuracy of the 1-dimensional models [1,4,2023]. However,
one aspect that has not been explored is the impact of tting
1-dimensional models to experimental data typically used for
the purpose of obtaining the numerical values of the system
parameters (i.e., rate constants, transfer coefcients and diffu-
sion coefcients) and the identication of the rate-controlling
phenomena. In fact, relatively fewstudies have focused on quanti-
tatively and rigorously tting any type of physicochemical model
to experimental data such as voltammograms or electrochemical
impedance spectra [6,7,9,24]. Also, although computation of uid
motion through use of the NavierStokes equations has received
attention, its coupling with mass transport to model electrodepo-
sitionprocesses under transient conditions has not beencommonly
done.
The objectives of the present study are to assess the impact
of estimating kinetic parameters by the conventional approach
of tting a 1-dimensional model to experimental data on cop-
per deposition and to use simulations of an axially symmetric
2-dimensional model to gain further insight into phenomena
occurring during this process. Both versions of the model describe
the conditions for electrodeposition of copper onto a rotating
disk electrode and relate the detailed electrode reaction kinetics
(i.e., accounting for elementary steps) to mass transport of elec-
troactive species by diffusion and convection. The 2-dimensional
model makes use of the Reynolds-averaged NavierStokes (RANS)
equations where the turbulent viscosity is estimated using the
k model to account for the hydrodynamic effects. This model
should also help close the gap existing between mesoscopic
and macroscopic analyses of electrodeposition. Electrode kinet-
ics are described by radial-dependent ButlerVolmer expressions
accounting for Cu(II) reduction by two consecutive single electron
transfer steps involving the adsorption of the intermediate Cu(I)
ads
on the electrode surface. Special emphasis is also placed on the sit-
uation arising when the hydrodynamics change from laminar to
turbulent ow. These conditions are particularly relevant since the
rotation speeds used during typical RDE experiments on electrode-
positionusuallyfall inthis range of uidow. Obviously, the quality
of the deposits is strongly affected by the hydrodynamics of the
electrolyte and the uniformity of the owconditions in the vicinity
of the electrode surface. Smooth, dense and compact deposits are
desirable for some applications, whereas the formation of granular
and loosely adherent deposits for easy removal is the objective for
others.
2. Experimental
Linear sweep voltammetry (LSV) was conducted at a scan rate
of 100mV s
1
at room temperature in a supporting electrolyte
of 1800mol m
3
H
2
SO
4
(98%, Fisher Scientic) and 200mol m
3
CuSO
4
5H
2
O(99.5%, VWR). All solutions werefreshlypreparedwith
deionized water (pH 6.8) and deoxygenated with N
2
(Praxair,
grade 4.8) for 30min prior to the electrochemical experiments. No
additives or complexing agents were used in these plating baths to
avoid masking the kinetics of copper deposition. Compact copper
deposits were obtained under these experimental conditions sim-
ilar to those reported by Vazquez-Arenas et al. [5,6]. The measured
electrode potentials were corrected for the ohmic drop utilizing
the R
s
value reported in Tables 2 and 3. Since R
s
turns out to have
a strong effect on the computed LSVs, it was treated as a tting
parameter, as discussed later.
A three-electrode cell was utilized to performthe LSV. A copper
electrode with an exposed area of 1.96m10
5
m to the solu-
tion was used as the working electrode (RDE). A pure graphite rod
(6.15mmdia 152mmlong, AlfaAESAR, 99.999%) andHgHg
2
SO
4
electrode (Radiometer Analytical) were used as the counter and
reference electrodes, respectively. All potentials referred herein
correspond to the SSE scale.
Prior to each experiment, the working electrode surface was
mechanically polished using SiC-type abrasive paper (1200 grade)
and polished to a mirror nish using Buehler alumina powder
(nal grain size 0.05m). In order to remove any dust or alumina
particles on the electrode surface, the electrode was then rinsed
with ultrapure water and placed in an ultrasonic bath for 5min.
The working electrode was mounted on a Teon shaft (EDI101
Radiometer, Copenhagen) and rotated at 1000 and 1500rpm
using a speed controller unit (CTV101 Radiometer, Copenhagen).
The three electrodes were connected to an EPP-4000 poten-
tiostat/galvanostat (Princeton Applied Research) controlled by a
personal computer using the EC-Lab software.
3. Model development
The 1-dimensional model used in this work was previously pre-
sented as Model 1 in reference [6] and is presented in section 3.5.
Most of the development that follows focuses on the axially sym-
metric 2-dimensional model. Fig. 1 depicts the domain considered
for this model. The domain is a cross-section of the entire cell and
is axially symmetric to conform to the geometry associated with
Fig. 1. Schematic diagram showing the cross-section of the cell utilized for cop-
per electrodeposition onto a copper RDE. R
d
=radius of the active copper disk,
R
d
+R
i
=outer radius of the insulating material (nylon).
Author's personal copy
J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725 719
the electrodeposition onto a RDE [25]. The dimensions correspond
to that of an actual laboratory-scale cell used to electrochemically
deposit copper from an aqueous solution containing 200mol m
3
CuSO
4
5H
2
O and 1800mol m
3
H
2
SO
4
at 25

C [6].
The NavierStokes equations and mass transport equation
accounting for the diffusion and convection of the dissolved elec-
troactive species Cu(II) are utilized to obtain the velocity and
concentration proles, respectively, throughout the 2-dimensional
domain. Due to the axially symmetric nature of the system, none of
the quantities in the NavierStokes and mass transport equations
andassociatedboundary conditions exhibit any dependence onthe
angular dimension 0. However, the model accounts for the velocity
components (u, v, w) in all three directions.
3.1. Assumptions of 2-dimensional model
The following assumptions are made in developing the 2-
dimensional model:

An axially symmetric 2-dimensional domain is considered.

All three electrolyte velocity components (u, v, w) in the radial,


axial andangular directions, respectively, are consideredover the
domain.

Electrolyte is incompressible and Newtonian.

Electrolyte transport properties (i.e., diffusion coefcient, den-


sity) and kinetic parameters are uniformthroughout the system.

Process occurs at 298K under isothermal conditions.

Ideally dilute behavior is assumed in the solution (i.e. activity


coefcients are equal to 1.0).

Transport of the electroactive species in the electrolyte obeys


dilute solution behavior.

Since the solution contains a highly concentrated supporting


electrolyte, transport of electroactive species by migration is
neglected in comparison to that by diffusion and convection. This
is supported by the results of a previous study on copper depo-
sition showing that the contribution of migration is negligibly
small and can be safely ignored in a solution with composition
close to that considered in this work [7]. Also, no homogeneous
reactions are considered to occur.

Moving boundary due to the growth of the copper deposit is not


accounted for in the model.

Semi-empirical wall functions are usedto approximate near-wall


turbulent viscosities using the k model [26,27].
3.2. Kinetics
The electrode kinetics of Cu(II) reduction from acidic sulphate
solutions have been extensively studied. The most widely accepted
mechanism involves consecutive 1-electron transfer elementary
steps [68,10]:
Cu(II) + e

Cu(I)
ads
(1)
Cu(I)
ads
+e

Cu
0
(2)
that entails the formation of an adsorbed intermediate species
Cu(I)
ads
on the surface of the electrode. Reaction (1) has been
shown to be the rate-controlling step in this mechanismfor typical
electroplating conditions. Consequently, we ignore the backward
direction of this step in this analysis. Reactions (1) and (2) occur
over the surface of the electrode and are assumed to obey Lang-
muir adsorption behavior. Consequently, the expressions for the
net rates of reactions (1) and (2) over the interval 0r R
d
can be
written as:
r
1Cu
(r, t) = k
1Cu
exp

Cu1
F

(t)
R1

C
s
Cu
(r, t)(1 0
Cu
(r, t)) (3)
r
2Cu
(r, t) = k
2Cu
exp

Cu2
F

(t)
R1

0
Cu
(r, t) k
2Cu
exp

(1
Cu2
)
F

(t)
R1

(1 0
Cu
(r, t)) (4)
Inthese expressions, C
s
Cu
(r, t) and0
Cu
(r, t) correspondto the sur-
face concentration of Cu(II) and the fractional coverage of Cu(I)
ads
species adsorbed on the electrode, respectively, at radial position r
and time t. k
1Cu
and k
2Cu
are rate constants for the forward direc-
tions of reactions (1) and (2) and k
2Cu
is the rate constant for
the backward direction of reaction (2), while
Cu1
and
Cu2
are
the corresponding charge transfer coefcients for these steps. The
rate constants and transfer coefcients are assumed to be uniform
over the surface of the electrode and independent of the electrolyte
concentration and potential.
A material balance on the electrode surface sites occupied by
Cu(I)
ads
yields:
1
Cu
0
Cu
(r, t)
t
= r
1Cu
(r, t) r
2Cu
(r, t) (5)
1
Cu
is the maximumadsorption density of Cu(I)
ads
required to
completely cover a monolayer.
The local faradaic current due to copper deposition varies with
radial position and is related to the electrode reaction rates by
Faradays law, i.e.,
i
Cu
(r, t) = F[r
1Cu
(r, t) +r
2Cu
(r, t)] (6)
The measurable current density is obtained by integrating the
local current density values over the entire electrode surface and
dividing by the total exposed electrode area to yield:
i
f
(t) =

R
d
0
2r i
Cu
(r, t) dr
R
2
d
(7)
The total current density i(t) measured in the system is com-
posed of a faradaic and a capacitive component:
i(t) = i
f
(t) +C
dl
d

(t)
dt
(8)
where C
dl
is the double layer capacity and E

(t) is the electrode


potential corrected for the ohmic resistance R
s
in the solution as
follows:

(t) = (t) i(t)R


s
(9)
The model is used to describe the response of the RDE when its
potential is perturbed during linear sweep voltammetry. For the
particular experiments of this study, the potential of the RDE is
varied linearly with time in the negative direction fromthe open-
circuit potential E
OCP
:
(t) =
OCP
s
rate
t (10)
where s
rate
is the scan rate in Vs
1
.
3.3. Fluid mechanics
The hydrodynamics of the system is described by the steady-
state NavierStokes equations (i.e., momentum balance and
continuity equations) since the uid motion induced by the elec-
trode rotation has already reached steady when electrodeposition
begins. Newtonian and incompressible ow of the electrolyte is
considered in order to relate the shear stress between uid layers
to the velocity gradient in the direction perpendicular to these lay-
ers. However, under turbulent conditions, the uid motion in the
cell uctuates andthe velocity proles become difcult toestimate.
Thus, for stationary asymmetric ow under turbulent conditions,
an averaged representation of the turbulent ows can be obtained
Author's personal copy
720 J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725
through the Reynolds-averaged NavierStokes (RANS) equations
[16,2528]:

u
u
r

w
2
r
+v
u
z

+
p
r
=

q +
q
T
o
k

1
r

r
u
r

u
r
2
+

2
u
z
2

(11)

u
w
r

uw
r
+v
w
z

q +
q
T
o
k

1
r

r
w
r

w
r
2
+

2
w
z
2

(12)

u
v
r
+v
v
z

+
p
z
=

q +
q
T
o
k

1
r

r
v
r

+

2
v
z
2

+F
z
(13)
The Reynolds stresses are described in terms of the turbulent
viscosity, i.e., q
T
=C

k
2
/, using the k model [26,27]. The conti-
nuity equation is written as:
1
r

r
(ru) +
v
z
= 0 (14)
In Eqs. (11)(14), is the density (kgm
3
), q is the dynamic vis-
cosity (Pa s), P is the pressure (Pa), u the radial velocity (ms
1
), v
the axial velocity, w the rotational velocity and F
z
the volumetric
force due to the specic weight of water (Pa m
1
). k is the turbu-
lent kinetic energy(m
2
s
2
) and is the dissipationrate of turbulent
energy (m
2
s
3
) determined according to the k model, while C

and o
k
are experimentally determined constants [27]. The numer-
ical values of these constants are listed in Table 1.
3.3.1. Boundary conditions
The no-slip condition is applied to the electrolyte in direct con-
tact with the electrode and insulator surfaces (boundaries IIII in
Fig. 1), i.e., the electrolyte at these boundaries has zero velocity rel-
ative to the surfaces. Thus, the velocity components u and v at the
boundaries are zero, while the angular component wis equal to the
product of the angular velocity and the radial coordinate. Conse-
quently, the following conditions can be applied to each boundary
I, II and III:
n u = 0, v = 0, t w = r (15)
where is the rotational speed of the electrode (rads
1
), n the
normal unit vector to the boundary, t is the tangential unit vector
to the boundary and u=(u, v, w) is the vector velocity (ms
1
).
A symmetry condition (i.e., slip) is utilized to describe the free
surface alongthe tophorizontal boundaryIV. Thus, owoccurs only
Table 1
Parameters held xed during the tting procedure.
Parameter Value Reference
1000kgm
3
[33]
q 8.9110
4
Pa s [27]
qT Ck
2
/ [27]
C
+
5.5 [27]
k 0.42 [27]

+
w
=
w
l

100 [27]
D 1.510
6
m
2
s
1
[33]
Fz 9810Pa m
1
[33]
C 0.09 [27]
o
k
1.0 [27]
R
d
0.002m
I
Cu
5.7410
5
mol m
2
[6]
in the axial and rotational directions, whereas the radial velocity u
becomes 0. Also, no stress in the tangential direction is generated:
n u = 0, t

pI +

q +
q
1
o
k

= 0 (16)
where I is the identity matrix. The boundaries V and VI located at
the stationary walls of the cell offer a resistance to the uid ow
along each direction, leading to:
n u = 0 (17)
In addition, semi-empirical wall functions which only apply in
regions close to the different boundaries are used to approximate
the near-wall turbulent velocities (RANS), as described elsewhere
[28,29]. This approach is used since the velocities close to the walls
have much steeper proles than the regular patterns obtained by
solution of the NavierStokes equations.
The axial symmetry boundary condition applies at all locations
along r =0 (boundary VII). Flow can occur in the axial direction
z along this boundary, but not in the normal (r) or rotational (0)
directions. This leads to the following boundary conditions:
n u = 0
dv
dr
= 0 t w = 0 (18)
The domain contains no outowboundary which would x the
pressure at that location to any particular value. This introduces a
degree of freedomin dening the pressure everywhere in the sys-
tem. Thus, the pressure is arbitrarily set to be zero at the top right
corner of the cell [28]. Since the model is solved for turbulent con-
ditions (i.e. high Reynolds number), the velocities drop rapidly as
one moves awayfromthe walls. Therefore, empirical wall functions
relating the velocities and wall friction are used to replace the thin
boundary layer near the wall [2628]. These functions assume that
the owis parallel to the wall and the velocity within a distance
w
fromthe wall can be expressed by:
U
+
=
1
k
ln

w
l

+C
+
(19)
where k is the Karman constant, C
+
is the universal constant for
smooth walls and l
*
is the viscous length scale determined by the
k model. More information about the implementation of these
functions can be found in refs. [2628].
Since the system is assumed to follow dilute solution behav-
ior, transport of Cu(II) depends on the velocity of the electrolyte,
whereas the electrolyte velocity is not affected by Cu(II) transport.
The electrochemical experiments are carried out by rst adjusting
the speed of the RDE to the desired level and allowing the uid ow
to reach steady conditions before adjusting the electrode potential
to a value where Cu(II) reduction begins. Thus, steady state uid
ow conditions can be assumed throughout the deposition pro-
cess. This allows us to obtain the velocity elds rst by solving the
NavierStokes equations for incompressible Newtonianowunder
steady-state conditions independently of the mass transport equa-
tion for Cu(II). Once obtained, these velocities can be substituted
into the transient mass balance equation for Cu(II) which are then
solved to determine its concentration prole and the local current
density distribution over the electrode surface.
3.4. Mass transfer
A 2-dimensional transient mass balance for Cu(II) including
contributions from diffusion and convection yields the following
equation:
C
Cu
(r, z, t)
t
= (D
Cu
C
Cu
) u C
Cu
(20)
Author's personal copy
J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725 721
D
Cu
and C
Cu
(r, z, t) are the diffusion coefcient (m
2
s
1
) and
concentration (mol m
3
) of Cu(II), respectively.
3.4.1. Boundary and initial conditions
At the electrode surface (z =0, boundary I), the mass transfer
rate of Cu(II) to the electrode surface matches the net rate at which
Cu(II) is consumed by reaction (1), i.e.,
n (D
Cu
C
Cu
) = r
1Cu
(r, t) (21)
Axial symmetry applies along the vertical rotation axis (r =0,
boundary VII), while no mass transfer occurs at the remaining
boundaries (IIVI). For all of these boundaries, this leads to the
following condition:
n (D
Cu
C
Cu
) = 0 (22)
The initial condition required to solve for C
Cu
(r, z, t) is dened
as:
C
Cu
(r, z, 0) = C
b
Cu
(23)
where C
b
Cu
is the Cu(II) concentration in the bulk solution. Finally,
an additional initial condition is specied to determine 0(r, t) along
the electrode surface, i.e.,
0(r, 0) = 0 (24)
3.5. 1-dimensional model
In the 1-dimensional model, the domain is separated into a
boundary layer region of thickness and a well-mixed bulk region
where no depletion of any species takes place. It is assumed that
the use of the rotating disk electrode provides well-dened hydro-
dynamic conditions in the boundary layer. The bulk region is
consideredtobefullyestablishedat adistance3fromtheelectrode
surface where is dened as follows [30]:
= 1.61D
1]3
i

1]2
D
1]6
(25)
Thetransient mass balancefor Cu(II) presentedinEq. (20) occurs
only in the z-direction:
C
Cu
(z, t)
t
= D
Cu

2
C
Cu
(z, t)
z
2
v
C
Cu
(z, t)
z
(26)
over the region 0<z <3. The uid velocity component v normal to
the RDE is given by the following 3-termseries expansion [30]:
v = 0.51023
3]2
D
1]2
z
2
+0.33334
2
D
1
z
3
0.10267
5]2
D
3]2
z
4
(27)
where D is the kinematic viscosity. The initial and boundary condi-
tions expressedby Eqs. (21)(24) also apply inthis case but without
a radial dependence.
3.6. Computational details
The analysis of the model presented in the previous section
involves both numerical solution and then tting to experimental
data to estimate values of the kinetic parameters, D
Cu
, C
dl
and R
s
.
The parameters held xed throughout the modeling are listed in
Table 1. A two-step procedure was adopted to improve the conver-
gence of the numerical solution of the 2-dimensional model. In the
rst step, the steady state NavierStokes equations (section 3.3)
alone were solved to yield the velocities. Once obtained, the solu-
tions for these velocities were used without any further change.
In the second step, these velocities were substituted into the mass
transport equations (section 3.4) in order to complete the problem.
The system of PDEs, algebraic expressions, boundary conditions
and initial conditions given by Eqs. (3)(24) was solved using the
nite element method in the COMSOL Multiphysics

3.5a software
package [28]. Initially, quadratic Lagrange polynomials with fourth
order integration error were used as the shape functions for the
nite elements. However, subsequent analysis showed that the
accuracyof the numerical solutions couldbe signicantlyimproved
if quartic Lagrange polynomials (8th order integration error) were
used instead for this purpose. For this reason, the numerical results
presented here were obtained using quartic Lagrange polynomials
as the shape functions. An interactive-smart mesh with maximum
size of 510
5
mwas utilized for boundaries I, II and VII (refer to
Fig. 1), while a size of 610
4
mwas considered for the elements
next to the other boundaries. This produced a ner mesh close
to the reactive zone of the electrode and near the walls in order
to obtain more accurate solutions for the concentrations and
velocities in these regions. A total of 56,244 triangular-shaped
elements were used to discretize the entire domain. The statistical
tting of the model to the experimental data was carried out
by linking the solutions from the COMSOL Multiphysics

solver
(PARDISO routine) to the Matlab

R2010b Toolbox that uses a


trustregionreective algorithm to minimize the least-square
error [31]. During experiments, the average current density over
the electrode surface was measurable, but not the local current
densities at individual radial positions. Consequently, the average
current densities given by Eq. (7) in the 2-dimensional model were
t to the measured current densities collected during the LSV
experiments at particular electrode potentials. An Extreme Intel

Core i7 CPU running at 3.47GHz with RAM memory of 12.00GB


was used to carry out these calculations.
For the 1-dimensional model, Eqs. (25)(27) and the bound-
ary and initial conditions (Eqs. (21)(24)) were solved using the
nite element method in the COMSOL Multiphysics

3.5a soft-
ware package [28]. In this case, a second-order Lagrange quadratic
polynomial was used as the shape function and the domain was
discretized into 133 elements with an interactive-smart mesh. The
statistical tting of the model to the experimental data was per-
formed similarly to the 2-dimensional model, e.g. linking COMSOL
Multiphysics

to Matlab

.
4. Results and discussion
The uid velocity close to the surface of the electrode can
have a signicant effect on the uniformity of the electrodeposits
formed particularly if a vortex forms or turbulence occurs due
to the rotation of the working electrode. Thus, it is important to
determine the distribution of the uid owin the cell. Fig. 2 shows
the streamlines of the velocity elds obtained at two different
rotation speeds of the RDE: (a) 1000 (Re =3770, transition ow)
and (b) 1500rpm (Re =5655, turbulent ow). The velocity eld is
comprised of contributions from both the radial and axial veloc-
ities and is dened as (u
2
+v
2
)
1]2
. The colors of the streamlines
correspond to the velocity eld values dened in the legend on
the right hand-side of each plot. Not surprisingly, the velocity
elds are higher throughout the cell when the electrode is rotated
at 1500rpm (Fig. 2b) than when rotated at 1000rpm (Fig. 2a). In
addition, a new region appears next to the right side wall midway
along the axial direction. This stems from the shift in the uid
motion fromthe transition to turbulent owregimes.
The velocity gradients are steep in the vicinity of the electrode,
particularly near the outer radial edge. The electrolyte reaches the
highest velocities along the insulating side walls of the RDE and up
to the electrolyte/air interface. The rotation of the electrode moves
the layers of the uid fromthe bottomof the cell up to the surface
of the electrode. Since the uid cannot penetrate the electrode sur-
face, it is dragged along this surface fromthe center of the electrode
out toward the edges.
Author's personal copy
722 J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725
Fig. 2. Computed streamline plots for the electrolyte velocity elds within the cell when the disk rotates at speeds of: (a) 1000 (Re =3770, transition ow) and (b) 1500rpm
(Re =5655, turbulent ow).
The axial dependence of the axial velocitycomponent v obtained
from the 1- and 2-dimensional models is presented in Fig. 3
(1000rpm) and 4 (1500rpm). Fig. 3 shows the axial velocity cal-
culated according to the 2-dimensional model at different radial
positions (r =0.006, 0.004, 0.002 and 0.001m) when the RDE has a
rotation speed of 1000rpm. For the 1-dimensional model, the axial
velocitycanbegiveninterms of aseries solutionrst derivedbyvon
Karman and Cochran [13,14,16,17]. Up to the rst three terms, the
velocity canbe expressedas Eq. (27). AlsoincludedinFig. 3for com-
parison are plots of v obtained fromthe 1-dimensional model using
this 3-term expansion and also following the common practice of
truncating the above series to only the rst term. Fig. 4 shows the
corresponding v values for the 1- and 2-dimensional model at a
rotationspeedof 1500rpm. The velocities showninFigs. 3and4are
calculated over a range of axial positions fromzero (i.e., electrode
surface) to a distance z =110
4
mfromthe electrode surface.
A number of observations can be made concerning the results
presented in Figs. 3 and 4. Together the two models show that v
changes signicantly in both the axial and radial directions. The
2-dimensional model reveals that v increases as one moves away
fromthe electrode surface at any particular radial position. At the
same time, Fig. 3 shows that at a rotation speed of 1000rpm, v
remains unchanged as one moves radially from the center of the
electrode disk (r =0) toward the outer edge of the insulator (as
far as r =0.004m) at any given axial position within at least a dis-
tance z =110
4
m from the surface. However, close to the outer
edge of the insulator (r =0.006m), the axial velocity begins to rise
Fig. 3. Variation of the axial velocity v with axial position z computed for the 1-
and 2-dimensional models at different radial positions along the electrode surface
(r =0.001, 0.002, 0.004 and0.006m) whenthe RDE is rotatedat 1000rpm(Re =3770,
transition ow). The inset shows an expanded viewof the results obtained within a
short axial distance fromthe electrode.
signicantly (Fig. 3). An increase in rotation speed to 1500rpm
raises v everywhere, as expected, but also begins to affect its uni-
formity over the electrode surface although the effect is still small
(Fig. 4). Perhaps more interesting are comparisons of the results
obtained with the 1- and 2-dimensional models. At both rota-
tion speeds, the common approach of using a 1-term expression
for v gives good agreement with the values obtained from the 2-
dimensional model for distances up to 1210
5
m from the
electrode surface (see insets of bothgures). It is interesting to note
that this distance agrees well with the values of =1.2510
5
m
for a 1000rpmrotation speed and =1.0210
5
mfor a 1500rpm
rotation speed calculated using the well-known Levich equation
used to dene the boundary layer thickness in the 1-dimensional
model, i.e., Eq. (25). Beyond this distance, the values from the
1-dimensional model diverge widely from and greatly overesti-
mate those obtained from the 2-dimensional model. The use of
the 3-term expression for v does reduce the discrepancy from the
2-dimensional model at larger axial distances, but does not signif-
icantly enlarge the region where the 1- and 2- dimensional model
velocities coincide.
Although not shown here, the radial velocity component u at
any particular distance z from the electrode surface increases as
one moves radially from the center of the disk (r =0) outward as
far as the outer edge of the insulator (r =0.006m). However, at the
same time, the radial velocity remains unchanged as one moves in
the axial direction up to at least z =110
4
m from the electrode
0
0.002
0.004
0.006
0.008
0.0E+00 5.0E-05 1.0E-04
-
v
/

m

s
-
1
axial coordinate/ m
2-D, r = 0.004
2-D, r = 0.006 m
2-D, r = 0.002
2-D, r = 0.001
1-D, 1-term v
1-D, 3-term v
0
0.001
0.002
0.0E+00 2.5E-05 5.0E-05
-
v
/

m

axial coordinate/ m
Fig. 4. Variation of the axial velocity v with axial position z computed for the 1-
and 2-dimensional models at different radial positions along the electrode surface
(r =0.001, 0.002, 0.004 and0.006m) whenthe RDE is rotatedat 1500rpm(Re =5655,
turbulent ow). The inset shows an expanded viewof the results obtained within a
short axial distance fromthe electrode.
Author's personal copy
J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725 723
0
500
1000
1500
2000
-1.2 -1 -0.8 -0.6 -0.4
-
i

/

A

m
-
2
E - iR
s
/ V
Experimental, 1500 RPM
Model, 1500 RPM
Experimental, 1000 RPM
Model, 1000 RPM
limiting current density
Fig. 5. Comparison of the linear sweep voltammograms obtained experimentally
(symbols) andby tting of the 2-dimensional model (continuous lines) for a solution
containing 200mol m
3
CuSO
4
at scan rate of 100mV s
1
and electrode rotation
speeds of 1000 and 1500rpm. Limiting current densities for each rotation speed
estimated using the Levich equation are represented by dashed lines.
surface at any particular radial position from r =0 to r =0.006m.
Beyond the outer radius of the insulator, this behavior changes
and u increases monotonically with increasing distance from the
electrode.
Once the steady state NavierStokes equations were solved,
the velocities were substituted into the 2-dimensional mass trans-
fer equation for Cu(II) and then the model was statistically t to
data obtained from linear sweep voltammetry experiments con-
ducted at a scan rate of 100mV s
1
and rotation speeds of 1000
and 1500rpm for the electrolyte containing 200mol m
3
CuSO
4
.
Fig. 5 shows a comparison between the experimental and best
tted electrode responses based on the 2-dimensional model at
the two rotation speeds. In both cases, excellent agreement is
observed over the entire range of potentials. The 1-dimensional
model usingthe1-termexpressionfor v was alsottedtotheexper-
imental data following a similar procedure. Although not included
here, the agreement between this model and experimental data
was comparable to that of the 2-dimensional model. Not surpris-
ingly, the numerical values of the kinetic parameters, D
Cu
, C
dl
and
R
s
obtained by tting this model differ from those correspond-
ing to the 2-dimensional model. The two sets of parameters are
given in Tables 2 and 3, respectively. The limiting current densities
(dashed lines) estimated using the Levich equation with the value
D
Cu
=4.9910
10
m
2
s
1
obtained by tting the 1-dimensional
model are also included in Fig. 5 for comparison purposes. As
expected, the experimental and model-tted current densities at
both rotation speeds rise and closely approach the corresponding
limiting values as the overpotential becomes large.
Acapacitivecurrent exists at all points throughout thescansince
theelectrodepotential is continuallyvarying. However, this current
Table 2
Parameters obtained fromtting of the 1-D model to experimental data.
Parameter Value
k
1Cu
1.2310
10
ms
1
k
2Cu
5.4710
5
mol m
2
s
1
k
2Cu
6.2310
2
mol m
2
s
1

Cu1
0.47

Cu2
0.40
D
Cu
4.9910
10
m
2
s
1
Rs 110
5
ohmm
2
C
dl
0.65F m
2
Table 3
Parameters obtained fromtting of the 2-D model to experimental data.
Parameter Value
k
1Cu
1.2210
10
ms
1
k
2Cu
5.7410
5
mol m
2
s
1
k
2Cu
6.2310
2
mol m
2
s
1

Cu1
0.47

Cu2
0.41
D
Cu
3.4510
10
m
2
s
1
Rs 110
5
ohmm
2
C
dl
0.65F m
2
density always remains extremely small, oscillating about a value
of about 0.0135Am
2
according to both models so that its contri-
bution to the total electrode response is negligible. A comparison
of the values in Tables 2 and 3 reveals no signicant differences
(less than 10%) in the values of the rate constants and charge trans-
fer coefcients for Cu(II) reduction depending on which model is
ttedtothe experimental data. Likewise, these parameters are sim-
ilar to those obtained previously for the same system by tting a
1-dimensional model to electrochemical impedance spectroscopy
and steady-state polarization data [7]. On the other hand, a more
signicant differenceis observedintheestimatedvalues of D
Cu
. The
value obtained using the 1-dimensional model is approximately
44% larger than that obtained from the 2-dimensional model. The
observation that the 1-dimensional model yields a larger D
Cu
value
appears reasonable in view of the comparison of the values of the
axial uid velocities v obtained from the two models. As shown
in Figs. 3 and 4, the values of v within an axial distance of
1210
5
mare very similar for both models. However, transport
of Cu(II) to the electrode surface obviously receives contributions
fromall three directions inthe case of the 2-dimensional model, but
only in one direction in the 1-dimensional model. Consequently, a
larger value of D
Cu
is required in the 1-dimensional model in order
to match the experimental current densities.
A sensitivity analysis has also been carried out on both models
by varying each parameter listed in Tables 2 and 3 one at a time
from its best tted value and assessing the resulting effect on the
computed electrode response. This analysis shows that both mod-
els are more sensitive to the kinetic parameters associated with
the rst step (
Cu1
and k
1Cu
) than compared to the second step
(
Cu2
, k
2Cu
and k
2Cu
). This result is consistent with that of previ-
ous studies on copper deposition and reects that the rst step of
Cu(II) reduction has slower kinetics than does the second step and
so is rate-controlling [6,7]. Not surprisingly, the value of D
Cu
has
the largest effect on the electrode responses at high overpotentials
wheremass transport of Cu(II) becomes dominant. Thedifferencein
the D
Cu
value obtained depending on whether it is based on the t-
ting of the 1- or 2-dimensional model to the experimental data has
an important implication on the use of tted parameters in simula-
tions. If the value of D
Cu
determined fromthe 1-dimensional model
is subsequently used in simulations of the 2-dimensional model,
this will lead to large errors in the computed electrode response.
Similarly, problems will arise if the value of D
Cu
determined from
the 2-dimensional model is used in the 1-dimensional model. Thus,
the model used for a simulation should be the same as the one used
to determine the parameters in the rst place.
In order to assess the predictive capabilities of the 2-
dimensional model and the tted parameters, we carried out
simulations for conditions that match those of other experi-
ments not used for tting purposes and compared the two sets
of data. Fig. 6 shows a comparison of linear sweep voltam-
mograms for a scan rate of 100mV s
1
in solutions containing
1800mol m
3
H
2
SO
4
and 400mol m
3
CuSO
4
that have been
obtained experimentally with those obtained from simulations
of the 2-dimensional model utilizing the parameters reported in
Author's personal copy
724 J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725
0
1000
2000
3000
4000
-1.2 -1 -0.8 -0.6 -0.4
-
i

/

A

m
-
2
E - iR
s
/ V
Experimental, 1500 RPM
Model, 1500 RPM
Experimental, 1000 RPM
Model, 1000 RPM
Fig. 6. Comparison of the linear sweep voltammograms obtained experimentally
(symbols) and by simulation of the 2-dimensional model (continuous lines) for a
solution containing 400mol m
3
CuSO
4
at scan rate of 100mV s
1
and electrode
rotation speeds of 1000 and 1500rpm. The simulations were carried out using the
parameters reported in Tables 1 and 3.
Tables 1 and 3. It should be emphasized that no parameters were
adjusted in carrying out the simulations. As observed, the model
is able to provide good predictions of the experimental data. A
contributing factor to the small deviations between the experi-
mental and simulated curves could be the variation of transport
properties (e.g. diffusion coefcients) due to the difference in ionic
strength of this second solution that is included in the model [32].
A comparison of experimental scans obtained in a solution con-
taining 400mol m
3
CuSO
4
at a much lower scan rate of 5mV s
1
with the corresponding simulated curves is presented in Fig. 7. A
somewhat larger deviationbetweentheexperimental dataandpre-
dictions than in the previous case is evident. The model tends to
overestimate the steepness of the rapid current rise portion of the
scans, suggesting that the discrepancy is related to the kinetics of
Cu(II) reduction. Despite this, the model and the kinetic parame-
ters appear capable of accounting for the effects of a relatively wide
range of experimental conditions on the linear potential scans for
Cu(II) reduction.
Once all the parameters are determined, simulations of the
2-dimensional model can also be carried out to compute those
0
1000
2000
3000
4000
-1.2 -1 -0.8 -0.6 -0.4
-
i

/

A

m
-
2
E - iR
s
/ V
Experimental, 1500 RPM
Model, 1500 RPM
Experimental, 1000 RPM
Model, 1000 RPM
Fig. 7. Comparison of the linear sweep voltammograms obtained experimentally
(symbols) and by simulation of the 2-dimensional model (continuous lines) for
a solution containing 400mol m
3
CuSO
4
at scan rate of 5mV s
1
and electrode
rotation speeds of 1000 and 1500rpm. The simulations were carried out using the
parameters reported in Tables 1 and 3.
quantities that cannot be directly measured. As an example, plots
showing the variation in the surface concentration C
s
Cu
(r, t) and
the surface coverage 0
Cu
(r, t) of the adsorbed Cu(I) intermediate
as a function of the radial coordinate and the electrode potential
at a rotation speed of 1500rpm are presented in Fig. 8 a and b,
respectively. Fig. 8a reects the typical behavior for the surface con-
centration of an electroactive species subject to mass transport by
convection and diffusion [6]. A decay in the Cu(II) concentration is
observed as the potential becomes more negative and the system
moves into a mixed kinetic-transport control regime. This trend
continues as the scan reaches more negative potentials and the
concentration approaches zero as the systembecomes completely
mass-transport controlled. On the other hand, the surface coverage
of Cu(I)
ads
rst rises as the scan proceeds before it reaches a max-
imumand begins to drop as the potential becomes more negative.
Such behavior is typical of what would be expected as the driving
force (i.e., electrode potential) is raised for a reaction mechanism
involving consecutive reactions in series in which an intermedi-
ate is formed by the rst step and consumed by the second step.
The very low values of 0
Cu
(r, t) reect the much slower kinetics of
Fig. 8. a) Surface concentration C
s
Cu
(r, t) of Cu(II) and (b) surface coverage 0
Cu
(r, t) of Cu(I)
ads
as a function of the radial coordinate r and potential corrected for ohmic drop E
- iRs obtained at an electrode rotation speed of 1000rpm.
Author's personal copy
J. Vazquez-Arenas et al. / Electrochimica Acta 89 (2013) 717725 725
the rst step of Cu(II) reduction generating Cu(I)
ads
relative to that
of the second step which consumes them. As observed fromthese
plots, both C
s
Cu
(r, t) and 0
Cu
(r, t) are essentially uniformacross the
surface of the electrode under the prevailing conditions. Although
not included here, the Cu(II) surface concentrations obtained from
the 1-dimensional model are virtually identical to the C
s
Cu
(r, t) val-
ues from the 2-dimensional model at the potentials and rotation
speeds considered in this study. The Cu(I)
ads
surface coverages
from the 1-dimensional model and the 0
Cu
(r, t) values from the
2-dimensional model differ somewhat more, remaining apart by a
factor of 2 or so, although it should be noted that the surface cov-
erages according to either model always remain below0.01 under
any circumstance.
5. Conclusions
The impact of estimating model parameters for an electrochem-
ical process using the common approach of tting a 1-dimensional
boundary layer model to experimental data has been assessed by
comparing the values so obtained using an axially symmetric 2-
dimensional model that includes the NavierStokes equations for
uid ow. The case examined in this analysis is copper deposition
onto a rotating disk electrode from acidic copper sulphate solu-
tions. Rotation speeds of 1000 and 1500rpm are considered with
the intention of analyzing uid owunder transition and turbulent
regimes. Both models account for copper deposition via a two-
step mechanism and mass transfer of Cu(II) in the electrolyte by
convection and diffusion.
Solution of the NavierStokes equations shows that the veloci-
ties close to the surface of the electrode increase as one moves from
the center of the electrode to its edge and that the uid along the
insulating side walls of the RDE and up to the electrolyte/air inter-
face exhibits the highest velocities. The calculations of the velocity
utilizing the 1-dimensional model are only accurate when they
are restricted to a distance within 1.410
5
m from the elec-
trode surface. The velocity estimated by the 1-dimensional model
is improved by increasing the number of terms considered in the
von Karman and Cochran series solution. Although no signicant
differences are found between the kinetic parameters determined
by tting of the two models to experimental data, the diffusion
coefcient of Cu(II) estimated using the 1-dimensional model is
found to exceed the value obtained from2-dimensional model by
approximately 44%.
Although the model and the kinetic parameters are able to
predict the behavior at different rotations of the electrode and dif-
ferent CuSO
4
concentrations, further renements are required to
add the effects of the ionic strength on the transport properties.
Transport properties (i.e. diffusion coefcients) estimated with the
1-dimensional model should be adequate for analyses of electrode
responses when an accurate description of the uid motion via the
NavierStokes equations under transition and turbulent regimes is
not critical. In cases where such uid motion affects the behavior
of the systemsignicantly and cannot be overlooked, it may be sat-
isfactory to use 1-dimensional models that account for turbulence
by incorporating empirical expressions in terms of the Sherwood
numbers and eddy diffusivities. However, in general, such analyses
will requiresolutionof morecomprehensive2-dimensional models
involving the RANS equations for best results.
Acknowledgments
The authors express gratitude to the Natural Sciences and
Engineering Research Council of Canada (NSERC) for the nancial
support to carry out this work. JVA expresses gratitude to the Mex-
ican Council of Science and Technology (CONACyT) for the stipends
received through SNI.
References
[1] P. Mandin, C. Fabian, D. Lincot, Importance of the density gradient effects in
modelling electro deposition process at a rotating cylinder electrode, Elec-
trochimica Acta 51 (2006) 4067.
[2] G. Nelissen, G. Weyns, P. Maciel, J. Deconinck, O.V. Vyver, H. Deconinck, Numer-
ical study of the inuence of the anode position and the electrolyte owon the
deposition of copper on a wire, Electrochimica Acta 52 (2007) 6584.
[3] O.V. Vyver, G. Nelissen, G. Weyns, J. Deconinck, M. Degrez, S. Godet, Mass trans-
fer and current distribution on a metallic wire, Electrochimica Acta 53 (2008)
6452.
[4] Q. Dong, S. Santhanagopalan, R.E. White, A comparison of numerical solutions
for the uid motion generated by a rotating disk electrode, Journal of the Elec-
trochemical Society 155 (2008) B963.
[5] J. Vazquez-Arenas, G. Vazquez, A.M. Melendez, I. Gonzalez, The effect of the
Cu
2+
/Cu
+
step on copper electrocrystallization in acid noncomplexing elec-
trolytes, Journal of the Electrochemical Society 154 (2007) D473.
[6] J. Vazquez-Arenas, Experimental and modeling analysis of the formation of
cuprous intermediate species formed during the copper deposition on a rotat-
ing disk electrode, Electrochimica Acta 55 (2010) 3550.
[7] M.E. Huerta Garrido, M. Pritzker, EISandstatistical analysis of copper electrode-
position accounting for multi-component transport and reactions, Journal of
Electroanalytical Chemistry 594 (2006) 118.
[8] J.J. Kelly, A.C. West, Copper deposition in the presence of polyethylene gly-
col II. Electrochemical impedance spectroscopy, Journal of the Electrochemical
Society 145 (1998) 3477.
[9] A. Pohjoranta, R. Tenno, A computational multi-reaction model of a Cu elec-
trolysis cell, Electrochimica Acta 54 (2009) 5949.
[10] R. Caban, T.W. Chapman, Statistical analysis of electrode kinetics
measurements-copper deposition from CuSO
4
-H
2
SO
4
solutions, Journal
of the Electrochemical Society 124 (1977) 1371.
[11] T.P. Moffat, D. Wheeler, W.H. Huber, D. Josell, Superconformal electrodeposi-
tion of copper, Electrochemical and Solid-State Letters 4 (2001) C26.
[12] E. Mattsson, J.O.M. Bockris, Galvanostatic studies of the kinetics of deposition
and dissolution in the copper +copper sulphate system, Transactions of the
Faraday Society 55 (1959) 1586.
[13] T.v. Karman, ber laminaire und turbulente Reibung, Zeitschrift fr Ange-
wandte Mathematik und Mechanik 1 (1921) 486.
[14] W.G. Cochran, The ow due to a rotating disk, Proceedings of the Cambridge
Philosophical Society 30 (1934) 365.
[15] M.E. Orazem, Application of mathematical models for interpretation of
impedance spectra, in: R.F. Savinell, A.C. West, J.M. Renton, J. Weidner
(Eds.), The Electrochemical Society Proceedings Series, Pennington, NJ, 1999,
p. 68.
[16] J. Newman, K.E. Thomas-Alyea, Electrochemical Systems, 3rd ed., John Wiley &
Sons, NewJersey, 2004.
[17] D. Pletcher, A First Course in Electrode Processes, 1st ed., The Electrochemical
Consultancy, NewYork, 1991.
[18] C.G. Law, J.P. Pierini, J. Newman, Mass transfer to rotating disks and rotating
rings in laminar, transition and fully-developed turbulent ow, International
Journal of Heat and Mass Transfer 24 (1981) 909.
[19] C. Deslouis, B. Tribollet, L. Viet, Local andoverall mass transfer rates toa rotating
disk in turbulent and transition ows, Electrochimica Acta 25 (1980) 1027.
[20] P. Mandin, T. Pauport, P. Fanouillere, D. Lincot, Modelling and numerical
simulation of hydrodynamical processes in a conned rotating electrode con-
guration, Journal of Electroanalytical Chemistry 565 (2004) 159.
[21] Q. Dong, S. Santhanagopalan, R.E. White, Simulation of the oxygen reduction
reaction at an RDE in 0.5MH
2
SO
4
including an adsorption mechanism, Journal
of the Electrochemical Society 154 (2007) A888.
[22] J. Gonzalez, C. Real, L. Hoyos, R. Miranda, F. Cervantes, Characterization of the
hydrodynamics inside a practical cell with a rotating disk electrode, Journal of
Electroanalytical Chemistry 651 (2011) 150.
[23] A. Alexiadis, A. Cornell, M.P. Dudukovic, Comparison between CFDcalculations
of the owin a rotating disk cell and the Cochran/Levich equations, Journal of
Electroanalytical Chemistry 669 (2012) 55.
[24] J. Vazquez-Arenas, M. Pritzker, Comprehensive impedance model of cobalt
deposition in sulfate solutions accounting for homogeneous reactions and
adsorptive effects, Electrochimica Acta 56 (2011) 8023.
[25] P.M. Gresho, R.L. Sani, Incompressible Flow and the Finite Element Method,
John Wiley & Sons Ltd., 1998.
[26] 1st NAFEMS Workbook of CFD Examples, Laminar and Turbulent Two-
dimensional Internal Flows, NAFEMS, 2000.
[27] D.C. Wilcox, Turbulence Modeling for CFD, DCWIndustries Inc., La Canada, CA,
1998.
[28] COMSOL
TM
3.5a, Chemical Engineering Module, Model Library in COMSOL AB
USA, 2008.
[29] B.E. Launder, D.B. Spalding, Computer Methods in Applied Mechanics and Engi-
neering 3 (1974) 269.
[30] V.G. Levich, Physicochemical Hydrodynamics, Prentice-Hall, Englewood Cliffs,
NJ, 1962.
[31] http://www.mathworks.com/products/statistics/ (September 2012).
[32] J. Vazquez-Arenas, M. Pritzker, Steady-state model for anomalous CoNi elec-
trodeposition in sulfate solutions, Electrochimica Acta 66 (2012) 139.
[33] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, Taylor & Francis, Boca
Raton, 2004.

S-ar putea să vă placă și