Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲 D201

0013-4651/2007/154共4兲/D201/7/$20.00 © The Electrochemical Society

Influence of Chloride Anions on the Mechanism of Copper

Electrodeposition from Acidic Sulfate Electrolytes
W. Shao,* G. Pattanaik, and G. Zangari**,z
Department of Materials Science and Engineering and Center for Electrochemical Science and
Engineering, University of Virginia, Charlottesville, Virginia 22904-4745, USA

We investigate the influence of chloride Cl− ions in a broad range of chloride concentrations on the kinetics and mechanism of
copper electrodeposition from sulfate-based acidic electrolytes. Chloride ions influence copper deposition through two competitive
effects: at low Cl− concentration 共few mM兲, chloride ions depolarize the Cu reduction process, while higher Cl− concentrations
induce complexation of copper species and cause a cathodic polarization of the deposition process. Cu reduction proceeds through
two parallel mechanisms, a direct two-step reduction and a chloride-mediated route, whose relative importance depends on the
amount of chloride present. A transition between these two mechanisms can be identified both by steady-state and impedance
methods; however, the chloride concentration at which it occurs depends on the time scale probed by the two techniques.
Impedance measurements further demonstrate that the presence of chlorides changes the double-layer structure.
© 2007 The Electrochemical Society. 关DOI: 10.1149/1.2434682兴 All rights reserved.

Manuscript submitted September 13, 2006; revised manuscript received November 24, 2006.
Available electronically February 5, 2007.

The process of copper electrodeposition has been used exten- librium potential of Reaction 4 for relatively long times6 or when
sively for copper refining, electroforming, and plating purposes. In using a rotating disk electrode at a very high rotating speed9 and is
particular, since 1997, with the replacement of aluminum by elec- usually not observed under conventional plating conditions.
troplated copper for interconnects in ultralarge-scale integrated cir- Most investigations regarding the influence of chloride ions on
cuits, this process has played an important role in the continuing copper deposition have focused on the effect of very small amounts
evolution of integrated-circuit manufacturing technology.1,2 The 共mmol/L兲 of chloride,6-8 of interest in copper damascene processes.
most common electrolytes for copper plating are acidic and use However, large chloride concentrations 共up to mol/L兲 are also of
copper sulfate as the metal source. Cathodic deposition of copper in interest, for example in copper electrowinning. In this paper we
acidic sulfate solutions is generally accepted to occur through two present a comprehensive investigation of the influence of chloride
consecutive reactions ions in a broad range of chloride concentrations on the mechanism
of copper electrodeposition. To this end, various electrochemical
Cu2+ + e = Cu+ E° = − 0.087 VSCE 关1兴 methods, including cyclic voltammetry, steady-state current-
potential curves, and electrochemical impedance spectroscopy tech-
Cu+ + e = Cu E° = 0.281 VSCE 关2兴 niques are used.
3-5
of which the first one is the rate determining step. SCE indicates
the saturated calomel electrode. In order to control deposition kinet- Experimental
ics as well as the microstructure and morphology of electrodeposited All experiments were carried out using a three-electrode pris-
films, the addition of several additives to the electrolyte is necessary. matic cell, with the counter and working electrodes placed in a
Because chloride ions are among the most common additives, un- vertical configuration. The electrolyte was never stirred. Polycrys-
derstanding their influence on the deposition process is of central talline copper 共99.99%, oxygen free兲 was used as the working elec-
importance for understanding the combined effect of all the other trode. Cu sheets were mechanically polished with silicon carbide
additives. In recent years cyclic voltammetry combined with an paper up to 1200 grit, followed by electrochemical polishing in 85%
electrochemical quartz-crystal microbalance,6 voltammetry com- phosphoric acid using a cell voltage of 1.7 V for 20 min, to achieve
bined with radiotracer adsorption,7 and in situ scanning tunnel mi- a mirror-like surface. Finally, the working electrodes were rinsed in
croscope 共STM兲 imaging8 studies have concluded that chloride has a deionized 共DI兲 water. The counter electrode was a large area plati-
catalytic action on copper deposition and a definite influence on the num mesh. The SCE was separated from the main compartment by
structure, orientation and dynamics of surface steps.6,8 These studies a Luggin capillary. The experiments were controlled and data were
concluded that the presence of Cl−, at potentials used during tech- recorded using an EG&G PAR 273 potentiostat/galvanostat.
nical plating, accelerates the reduction of copper through the forma- The electrolytes were made up with analytical-grade chemicals,
tion at the electrode surface of a CuCl monolayer, generated by the and high-purity DI water 共Milli-Q purification system from Milli-
following reaction6 pore SA 67120 Molsheim, France, resistivity ⬎18 M⍀ cm兲 was
used throughout. Copper electrodeposition was carried out 共i兲 in a
−
Cu2+ + Clads + e = CuClads E° = 0.338 VSCE 关3兴 sulfate solution containing 0.3 M CuSO4 and 0.5 M H2SO4, 共ii兲 in
which is then reduced through Reaction 4 sulfate solutions with additions of NaCl in the concentration range
1–100 mM, or 共iii兲 in an all-chloride solution containing 0.3 M
CuClads + e = Cu + Cl− E° = − 0.063 VSCE 关4兴 CuCl2 + 0.5 M HCl. The pH of all electrolytes was measured after
This chloride-mediated mechanism proceeds in parallel with the solution preparation and before performing the experiment, and their
two-stage reduction of Cu2+ 共Reactions 1 and 2兲 and results in an values were between 0.32 and 0.34 in all cases.
acceleration of the overall reduction kinetics. The adsorbed Cl− layer Cyclic voltammograms were measured in the negative direction
acts as a dynamic template which guides surface-step orientation from the open-circuit potential of 0.06 down to −0.5 VSCE with a
and copper deposition while floating on the electrode surface.8 For- scan rate of 20 mV s−1. Steady-state current-potential curves were
mation of a bulk CuCl phase is possible in principle but has been measured by applying a given potential value and holding that value
observed only when holding the electrode potential above the equi- for about 30 s before reading the corresponding steady current. The
complex impedance was recorded from 100 kHz to 0.01 Hz with
10 mV ac amplitude signals using a Solartron 1255 frequency re-
* Electrochemical Society Student Member. sponse analyzer under potential control. All potentials in the follow-
** Electrochemical Society Active Member. ing are reported with respect to the SCE 关0.242 V vs normal hydro-
z
E-mail: gz3e@virginia.edu gen electrode 共NHE兲兴. Chloride contamination due to leakage

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 D202 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲

Table I. Reactions and equilibrium constants for copper in aque-

ous solutions with chloride.

Equilibrium
Reaction Eq. constant
Cu2+ + Cu = 2 Cu+ 5 1.58 ⫻ 10−6
Cu+ + Cl− = CuClaq 6 5.01 ⫻ 102
Cu+ + Cl− = CuCls 7 1.72 ⫻ 10−7
Cu+ + 2Cl− = CuCl−2 8 1.15 ⫻ 106
Cu2+ + Cl− = CuCl+ 9 2.31
Cu2+ + 2Cl− = CuCl2 10 0.65

chloride ion can thus inhibit copper reduction; this effect is further
enhanced in the all-chloride solution, where the current-potential
curve is shifted to an even more negative potential. The correspond-
ing increase observed in the reduction peak is not an intrinsic effect
but is due to the larger electrode roughness obtained when plating
from electrolytes with high chloride concentration.
The depolarization in the onset of copper deposition in the solu-
tion containing chloride 1–10 mM with respect to the sulfate solu-
tion is due to Cl−. Cl− adsorbs on the electrode and binds with Cu2+
ions from the electrolyte to form an adsorbed complex Cu–Cl, a
reaction intermediate which catalyzes the chloride mediated Cu re-
duction route.6,10 This mechanism occurs in parallel with the estab-
lished Cu reduction path in sulfate solutions 共Reactions 1 and 2兲,
resulting in an overall acceleration of the reduction kinetics.
The above discussion, however, does not explain why at high
Figure 1. 共a兲 CV curves at a copper electrode immersed in 0.3 M CuSO4 chloride concentration the reduction process is strongly cathodically
+ 0.5 M H2SO4 with 0, 1, 4, and 10 mM chloride as an additive; 共b兲 CV polarized at all potentials. Inhibition at potentials positive of
curves at a copper electrode in 0.3 M CuSO4 + 0.5 M H2SO4 without or −0.06 V may be attributed to the transient formation of a solid pre-
with 0.1 M chloride and in 0.3 M CuCl2 + 0.5 M HCl. SO2−
4 indicate the Cu cipitate of CuCls whose growth kinetics may now be faster as a
sulfate electrolyte. Scan rate: 20 mV/s. consequence of the higher Cl− concentration; such a layer should
dissolve however at potentials negative than −0.06 V. On the other
hand, an excess of chlorides is known to stabilize the cuprous ion,
through the Vycor frit of the SCE was calculated to be below 2 ␮M not only at the electrode by the formation of a CuCl layer but also in
during the maximum duration of our experiments 共20 min兲. This solution via the formation of chloride complexes. The concentration
contamination level is below the concentration of chloride caused by of the various copper complexes in copper chloride solutions was
the impurities in the chemicals we used. calculated at equilibrium by Kekesi et al. as a function of the free
The surface roughness of selected Cu electrodeposits was quan- chloride concentration.11 These results, however, cannot be used di-
tified by ex situ atomic force microscopy 共AFM兲, using a Molecular rectly to analyze our experiments because free chloride concentra-
Imaging Pico Plus instrument. tion is not a directly controllable experimental parameter. The con-
centration of the various copper complexes in the solution was thus
Results and Discussion calculated as a function of the total chloride concentration assuming
Figures 1a and b show the cyclic voltammograms 共CVs兲 for cop- the solution and the copper electrode at equilibrium.
per deposition at a copper electrode in the sulfate solution, com- The chemical equilibria 5–10 共Table I兲 were considered in order
pared to the CVs acquired in the same solution with the addition of to calculate the concentrations of Cu2+, Cu+, and their
1, 4, 10, or 100 mM NaCl, and in the all-chloride solution. In the complexes.7,12-16 Equations 5-10 were solved simultaneously in a
sulfate solution the forward scan exhibits a reduction peak at range of chloride concentrations by keeping the total copper ion
−0.26 V, followed by a diffusion-limited plateau. On the reverse concentration at 0.3 M. In this calculation the activity coefficients of
scan a nucleation loop is observed, probably due to partial reoxida- all species in the solution are assumed to be unity, except for the
tion of the substrate after electropolishing and before immersion in activity coefficient of the cupric ion, where the value
the electrolyte. ␥Cu2+ = 0.086 was used.17 In addition, no complex species higher
With the addition of 1–10 mM Cl−, the curve in the forward scan than the second order 共for example, CuCl2− 3 and CuCl3 兲 were con-
−

is increasingly shifted towards less negative potentials 共Fig. 1a兲; sidered in the calculation; this is a good approximation at low chlo-
concurrently, the peak current increased with Cl− concentration. ride concentration.18
Both the depolarization and the increase in peak current reveal a The results are shown in Fig. 2. With increasing concentration of
catalytic effect of chloride on copper reduction.6,10 The whole ca- the chloride ion the concentrations of free Cu2+ and free Cu+ de-
thodic branches of CVs in the solutions with 关Cl−兴 up to 10 mM are crease. For example, the concentration of free Cu2+ falls from
shifted to a less negative potential compared to the chloride-free 0.29985 M at 关Cl−兴 = 1 mM to 0.2508 M when the total amount of
solution, indicating the absence of any inhibition effect at low po- chloride is 0.1 M. The cumulative amount of the free Cu+ and all the
tentials. This proves that no insoluble CuCl layer is formed during other Cu共I兲 complexes, however, increases with increasing 关Cl−兴,
the potential scan under the chosen experimental conditions, and in indicating that overall, chloride stabilizes the cuprous ion. The de-
particular that CuCl formation is kinetically hindered even above crease in the concentration of free cupric ion is consistent with the
−0.06 V, where its formation would be thermodynamically possible. observed decrease of the open-circuit potential with increasing chlo-
Upon increase of 关Cl−兴 to 100 mM 共Fig. 1b兲 the trend reverses ride concentration; this is the first cause for the negative shift of the
and the current-potential curves shift towards more negative poten- CV curve in the solutions containing 0.1 M NaCl and in the pure
tials compared to the sulfate solution. At high concentration the chloride solution. The decrease in concentration of free cupric ions

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲 D203

Figure 4. Tafel plots for copper deposition in 共a兲 the sulfate solution, and in
the sulfate solution with the addition of 共b兲 1 mM NaCl, 共c兲 4 mM NaCl, 共d兲
10 mM NaCl, and 共e兲 0.1 M NaCl, and 共f兲 in 0.3 M CuCl2 + 0.5 M HCl.
Figure 2. The concentration of copper-chloride species in the solution vs the
total chloride concentration. The total copper concentration is kept at 0.3 M.
The reactions used to calculate these concentrations are reported in Table I.
共*兲 Molar quantity of CuCls, which would precipitate from 1 L of solution. all-chloride curve at low overpotentials, where a potential-dependent
transition of the reduction process may have been detected. The low
currents observed in the range −0.04–0 V may in fact be interpreted
in the electrolyte bulk can further decrease the reduction current through the formation of a solid CuCl layer, which inhibits Cu re-
through Reactions 1 and 3. This could be another cause for the duction. In the solution containing 0.1 M Cl− the current increases
negative shift of the CVs in the solutions containing 0.1 M Cl− or around −0.05 V, possibly indicating the dissolution of the solid
higher. CuCl layer. Such a transition occurs instead around −0.1 V in the
According to Fig. 2 the CuCls precipitate is thermodynamically all-chloride solution, indicating faster growth kinetics of CuCls at
stable only when the total chloride concentration increases above higher Cl− concentration and a consequent longer time needed for its
2.1 mM; this is consistent with the reported experimental value.6 dissolution. This transition is not clearly visible in solutions contain-
The amount of precipitate at equilibrium increases with chloride ing 10 mM chloride or less, probably because the kinetics of forma-
concentration, but its growth rate is kinetically limited during poten- tion of CuCls is slowed down and the fraction of the electrode cov-
tial scans and thermodynamically impossible under copper deposi- ered by CuCls under such conditions is small. Furthermore, such
tion conditions, when potentials are much more negative than transition is not seen in CVs in the corresponding solutions. This is
−0.06 V 共Reaction 4兲. probably due to the limited extent of CuCls growth during the short
Trends similar to those observed in the CVs are also seen in duration of CV experiments.
current-potential curves obtained by the steady-state method, as The current-potential curve in the pure chloride solution is depo-
shown in Fig. 3. In synthesis, when 关Cl−兴 is less than or equal to larized relative to that obtained in the sulfate solution when the
10 mM, the curves are depolarized relative to the chloride-free so- potential is negative to −0.18 V, a voltage different from that ob-
lution; on the contrary, they shift to more negative potentials when tained by cyclic voltammetry. This is because CuCls layer dissolu-
关Cl−兴 ⬎ 0.1 M. Of particular interest are the latter curve and the tion is also kinetically hindered and requires some time to be com-
plete, time during which the potential in CV experiments is
decreased. The larger current observed in the chloride solution be-
low −0.2 V is due to the increase in electrode area caused by elec-
trode roughening when plating from a solution with high chloride
concentration.
Kinetic parameters of the Cu reduction reaction were quantified
through a least-squares fitting of the steady-state current-potential
curves. Fitting of the experimental data was performed using both
a Tafel approximation in the potential range from −0.10 to
+ 0.02 V 共Fig. 4兲 and a mass-transport correction to the Tafel char-
acteristics in the voltage range −0.15–0.02 V using the following
equation19
共il − i兲
␩ = a + b log 关11兴
i
in which ␩ is the overvoltage, il is the limiting current density, a is
a constant, and b is the Tafel slope. Experimentally determined val-
Figure 3. Steady-state current-potential curves for a copper electrode im- ues for il, in the range 32–40 mA/cm2, were used in the fitting
mersed in 0.3 M CuSO4 + 0.5 M H2SO4 with various concentrations of procedure. Both methods yield similar results. The Tafel slopes are
chloride 共0–100 mM兲 and in 0.3 M CuCl2 + 0.5 M HCl. about 118–140 mV/dec 共110–114 mV/dec with mass-transport cor-

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 D204 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲

rection兲 in the solutions without or with chloride up to 10 mM 共Fig.

4a-d兲, close to the reported value of 120–135 mV/dec.13 This result
is consistent with the description of copper reduction mainly occur-
ring through two consecutive electrochemical charge-transfer steps,
Cu2+ /Cu+ and Cu+ /Cu, in which the first step 共Cu2+ /Cu+兲 is
rate-determining.3-5 This mechanism is seemingly unaltered in the
presence of adsorbed chloride ions. However, if the Tafel slope is
derived for the solution containing 0.1 M NaCl, the result is about
78 mV/dec without 共Fig. 4e兲 and 79 mV/dec with mass-transport
correction, values which are significantly different from those ob-
tained previously. The data in Fig. 4e also suggest a crossover be-
havior around −0.05 V, which is close to the equilibrium potential
of Reaction 4; this is only visible in the solutions with chloride
concentration larger than 10 mM. This may indicate that the Cu
reduction path through Reactions 3 and 4 becomes predominant in
the solution with high chloride concentration, a conclusion substan-
tiated by the Tafel slope obtained in the pure chloride solution, Figure 5. 共a兲 Equivalent electrical circuit model of the copper reduction
which is also about 78 mV/dec 共Fig. 4f兲. This value is higher than process. Rs is the solution resistance, CDL is the double-layer capacitance,
the reported value of 50–60 mV/dec, determined in pure chloride RCT is the charge-transfer resistance associated with the Cu2+ + e = Cu+ pro-
cess, and Zw is the finite-length Warburg impedance; R is the resistance
solutions with higher concentration.20
associated with the Cu+ + e = Cu step and C is the corresponding pseudoca-
The theoretical Tafel slope for the chloride-mediated reduction pacitance. 共b兲 Impedance plane plot calculated for frequencies between
process was calculated by assuming Reaction 4 to be the rate- 100 kHz and 0.01 Hz according to the electrical equivalent circuit model
determining step in the reaction mechanism described by Eq. 3 and shown in 共a兲, using the parameters reported.
4. The value of 40 mV/dec was obtained, showing a large discrep-
ancy between the experimental and theoretical value. This discrep-
ancy is probably due to the fact that even in pure chloride solution
the copper reduction by the reaction path 1 and 2, the theoretical that subject.9,22 Rather, we apply the results of these models to our
Tafel slope of which is 120 mV/dec, cannot be excluded. The coex- experimental data to provide a clear but simple physico-chemical
istence of the two reaction paths 1 + 2 and 3 + 4 would result in picture of the electrode processes occurring under widely varying
overall non-Tafel characteristics that could, however, be approxi- chloride concentrations. A typical impedance plane plot calculated
mated by an effective Tafel slope always larger than that calculated for frequencies between 100 kHz and 0.01 Hz according to this
for the chloride-only route. The discrepancy observed between our electrical equivalent circuit and using the parameters indicated is
results and those of Ref. 20 could thus be understood in view of the shown in Fig. 5b.
fact that the chloride concentration in our experiments was lower. Figure 6a shows the impedance measured in the solutions con-
The slight decrease in Tafel slope with increasing Cl− concentration taining chloride 0–4 mM at a voltage of −0.06 V, which is in the
共Fig. 4a-d兲 could similarly be understood as consequence of an in- Tafel range. In the high-frequency range 共100 kHz–50 Hz兲, a domi-
creasing influence of the chloride-mediated route. nant capacitive loop corresponding to the charge-transfer resistance
In summary, an analysis of the kinetics of the reduction process in parallel with a double-layer capacitance is observed. This is well
in copper electrolytes with variable chloride concentrations indicates defined in each solution and is attributed to the Cu2+ /Cu+ charge-
that chloride ions alter the copper deposition mechanism by intro- transfer reaction.21 The diameter of this loop in theory represents
ducing a parallel reduction route and also by complexing the copper RCT and decreases with increasing 关Cl−兴. The characteristic fre-
ion at high concentrations; the former effect is most relevant at low quency, i.e., the frequency corresponding to the largest imaginary
chloride concentration, the latter one at high 关Cl−兴; a transition from component of the capacitive loop at high frequency, increases with
an overall acceleration to an inhibition of the Cu reduction process chloride concentration. The same result was also reported by
is thus observed. Chloride may also form a solid CuCl precipitate; others,9 who related this increase to the enhanced rate constant of
its transient formation may be observed during some measurements, Reaction 3 at higher chloride concentration. The intermediate ca-
but it is usually of no importance under technical plating conditions. pacitive loop, which is typically in the range of 50–0.25 Hz, is
In order to get a further insight into the mechanism of copper thought to reflect the fast Cu+ /Cu Reaction,9 and in fact this loop is
reduction, we performed electrochemical impedance spectroscopy better resolved with increasing chloride concentration. The lower
共EIS兲 measurements. A physical model suitable for the description frequency loop 共0.25–0.01 Hz兲 reflects the Cu2+ diffusion in the
of Cu reduction from chloride electrolytes should consider, accord- deposition process.9,21 Diffusion phenomena usually result in a lin-
ing to the above discussion, an inhomogeneous electrode partly cov- ear variation of the complex impedance, the Warburg impedance.
ered with chloride ions, there forming a CuCl monolayer, and partly However, such impedance is derived under the condition of diffu-
covered with sulfate ions. Fractional coverages by Cl− on Cu single sion into or from a semi-infinite medium. When the Nernst-
crystals have been reported.8 The two regions would represent the diffusion-layer thickness is comparable to the distance covered by
locations where the two reaction pathways, direct and chloride- diffusing species under low-frequency perturbations, the Warburg
mediated, occur. Each region could be described by an equivalent impedance shows a decreasing imaginary component, resulting in
circuit, including, in the simplest case, a double-layer capacitance in the observed trend.
parallel with the series of a charge-transfer resistance and a Warburg Figure 6b shows the impedance measured at the same potential,
impedance; the whole electrode could then be modeled as the par- in the solutions containing 10 or 100 mM NaCl. The diameter of the
allel of these two circuits. However, these two formally identical high-frequency loop increases and the characteristic frequency de-
circuits could not be distinguished by the mathematical methods creases compared to the solutions at lower Cl− concentration. The
used to fit the impedance spectra; the only feasible solution is to intermediate loop, however, which is related to the Cu+ /Cu
reduce the parallel of the two circuits to one equivalent circuit. reaction,21 becomes much more pronounced in the 0.1 M Cl− solu-
Consequently, we use the equivalent circuit suggested by Reid and tion, to the extent that it obscures the low-frequency loop. This
David21 共Fig. 5a兲 to model the copper-reduction process and monitor effect may be seen as marking a transition in the reaction mecha-
how the chloride ions change the parameters of such a circuit. Our nism, which is now dominated by the chloride-mediated route 3
aim is not to develop a mathematically rigorous model and fit its + 4. The low-frequency loop is related to Cu2+ diffusion, which is
parameters to the experimental data; several papers are available on relatively unimportant at low overpotential. The solution resistance,

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲 D205

Figure 7. 共a兲 Impedance plane plots measured at −0.18 V in

0.3 M CuSO4 + 0.5 M H2SO4 with 0, 1, 2, and 4 mM NaCl. 共b兲 Impedance
plane plots measured at −0.18 V in 0.3 M CuSO4 + 0.5 M H2SO4 with 10
Figure 6. 共a兲 Impedance plane plots measured at −0.06 V in
or 100 mM NaCl. The area of the electrode is about 0.20 cm2.
0.3 M CuSO4 + 0.5 M H2SO4 with the addition of 0, 1, and 4 mM NaCl.
共b兲 Impedance plane plots measured at −0.06 V in 0.3 M CuSO4
+ 0.5 M H2SO4 with the addition of 10 or 100 mM NaCl. The area of the
electrode in both cases is about 0.20 cm2. Fig. 5a did not yield meaningful results. Consequently, in order to
quantitatively investigate the dependence of double-layer capaci-
tance and charge-transfer resistance on deposition potential and
which is the intercept of the impedance plot with the Re共Z兲 axis at chloride concentration, the equivalent circuit shown in Fig. 8, com-
the highest frequency, is much smaller in the high-chloride- posed of a solution resistance in series with the parallel of a charge-
concentration solutions. transfer resistance and a double-layer capacitance, was used to fit
Figures 7a and b display the impedance measured in the various the high-frequency capacitive loop 共about 100 kHz–50 Hz兲. The
solutions at −0.18 V. In this voltage range the reduction process is model-parameter values are assessed by nonlinear curve fitting
under mixed control, as deduced from the steady-state characteris- based on the least-mean-squares method using the software package
tics in Fig. 3. This is directly reflected by the fact that the feature at ZView with Calc-Modulus data weighing 共Scribner Associates,
low frequency, related to a Warburg impedance, is now much more Inc.兲. We use a general and very simple model with respect to others
visible than the high-frequency loop in the overall spectrum. With proposed in the literature9,22 because our aim is to investigate the
regard to the chloride content, the diameter of the high-frequency effects of chloride additions in a wide range of concentrations, con-
loop decreases and the characteristic frequency increases when the ditions for which any single detailed model may not be appropriate.
concentration of chloride is increased up to 4 mM. These trends, Figure 9 compares the dependence of double-layer capacitance
however, are inverted when the chloride concentration is 10 or CDL on potential in the sulfate solution and in the same solution with
100 mM. This nonmonotonic behavior observed for the characteris- the addition of 1 mM chloride. CDL exhibits in both electrolytes a
tic frequency in the high-frequency loop is an indication that the shallow minimum at the same potential of −0.12 V. Capacitance
precise mechanism and deposition kinetics is the result of a balance values are smaller for the solution containing chlorides23 but in both
between two competing processes, just as observed when examining cases are much lower than those predicted by the Gouy–Chapman
the current-potential curves. The observed trend for the characteris-
tic frequency may reflect the dynamics of formation and dissolution
of the CuCl layer at high frequency, which in its turn affects the rate
of the chloride-mediated reaction mechanism. The characteristic fre-
quency is higher at −0.18 V than at −0.06 V at the corresponding
chloride concentration, indicating an increased importance of the
chloride-mediated route at large overpotential. In addition, the low-
frequency loop, which reflects the Cu2+ diffusion in the deposition
process, is more pronounced than that at more positive potentials.
This indicates that the deposition process is more diffusion-
controlled at more negative potentials, as expected. The intermediate Figure 8. Equivalent circuit used to fit the high-frequency capacitive loop on
frequency loop is obscured at this potential because diffusion is a the impedance plane plot; RS is the solution resistance, CDL is the double-
much more important process under these conditions. layer capacitance, and RCT is the charge-transfer resistance associated with
Quantitative fitting of the entire EIS spectra with the model of the Cu2+ + e = Cu+ process.

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 D206 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲

Figure 9. Dependence of the double-layer capacitance on potential in

0.3 M CuSO4 + 0.5 M H2SO4, with or without 1 mM NaCl.

theory. This is expected because adsorption of both sulfates and

chlorides is known to occur on metals.24-26 A compact layer is thus
probably present at the copper electrode, whose structure is mostly
unchanged in the range of potentials investigated.
Figure 10 compares the dependence of the charge-transfer resis-
tance RCT on potential in the sulfate solution and the solution con-
taining 1 mM chloride. RCT decreases with decreasing deposition
potential, similarly to reported experiments,21 indicating that the en-
ergy barrier for the reduction process is decreased with increasing
overpotential. In addition, RCT is smaller in solutions containing
1 mM Cl−, confirming once more the catalytic effect of chloride Figure 11. 共a兲 Dependence of double-layer capacitance on chloride concen-
ions on the Cu reduction process. tration in 0.3 M CuSO4 + 0.5 M H2SO4 + x mM NaCl at −0.18 V. 共b兲 De-
Figures 11a and b display the dependence of CDL and RCT on the pendence of charge-transfer resistance on chloride concentration in
concentration of chloride in the mixed control region 共−0.18 V兲. 0.3 M CuSO4 + 0.5 M H2SO4 + x mM NaCl at −0.18 V. The area of the
The capacitance slightly decreases from the chloride-free value for electrode in both cases is about 0.20 cm2.
关Cl−兴 = 1–4 mM, then strongly increases for concentrations equal to
or higher than 10 mM. RCT exhibits an analogue behavior, with its
value almost doubling for a chloride concentration of 10 mM or copper reduction through the chloride-mediated route when its con-
higher. The decrease of RCT at 关Cl−兴 less than 4 mM is consistent centration is low, while chloride inhibits copper reduction when its
with the depolarization of CVs and steady-state current-potential concentration is high due to its complexation with copper ions.
curves. There is a discrepancy, however, between the trend observed The decrease in CDL upon addition of chloride has been observed
with current-potential curves and the trend of RCT at higher chloride by others when adding even smaller amounts of chloride.23 This is a
content. In the former case the transition from catalysis to inhibition possible indication that the structure of the double layer changes
of the reduction process occurs above 10 mM Cl−; in the latter this upon chloride addition. Both sulfate and chloride ions adsorb on
transition occurs above 4 mM Cl−. This discrepancy may be caused metal electrodes, but the adsorption of chloride ions is stronger,24 as
by the different response time of the various processes in solutions supported by various in situ studies. In situ STM investigations on
and their different impact on ac and dc measurements. EIS data in Cu共111兲26 showed that one sulfate ion coadsorbs with water and
any case confirm a transition in the mechanism of Cu reduction H3O+, forming an adsorbate structure which is not in registry with
which depends on applied potential and chloride concentration. At the substrate. Chloride adsorption on copper is so strong that it
negative enough potentials 共less than −0.06 V兲 chloride enhances changes the surface structure and step dynamics of the copper
surface.8 Moreover, in situ X-ray absorption studies27 showed that
copper atoms were packed in registry with the top layer of chloride
ions when they were deposited on Au共111兲 in the presence of chlo-
ride, while they were packed in registry with the Au 共111兲 in the
presence of sulfate ions. To explain the dependence of CDL on the
chloride concentration, a model is introduced which takes into ac-
count the stronger adsorption of chloride ions. In the chloride-free
solution, hydrated adsorbed sulfate ions position more distant than
chloride from the electrode surface, i.e. probably at the outer Helm-
holtz plane 共OHP兲. The double-layer capacitance under these condi-
tions can be written as
CDL = COHP 关12兴
The diffuse 共Gouy–Chapman兲 layer capacitance is neglected here
because the solution is highly concentrated.28 Adsorbed chloride
ions can position closer to the electrode surface, i.e., at the inner
Helmholtz plane 共IHP兲. In the solutions containing chloride ions
Figure 10. Dependence of charge-transfer resistance on potential in CDL can thus be expressed by the series of two capacitances, the
0.3 M CuSO4 + 0.5 M H2SO4 with or without 1 mM NaCl. The area of the capacitance of the IHP, CIHP, and the capacitance of the OHP, COHP,
electrode is about 0.20 cm2. i.e.

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 154 共4兲 D201-D207 共2007兲 D207

1 1 1 sulfate to the chloride-mediated route. Steady-state and impedance

= + 关13兴 measurements identify different values of chloride concentration for
CDL CIHP COHP
such a transition, probably due to the dynamics of formation and
The double-layer capacitance consequently decreases when dissolution of CuCls and the different time scales probed by the two
1 mM NaCl is added to the solution. CIHP increases as the specific techniques.
adsorption of chloride ions increases with higher chloride
concentration.29 CDL correspondingly increases slightly but still re- Acknowledgment
mains smaller than the value in chloride-free solutions up to 4 mM. This work was supported by National Science Foundation grant
In order to understand the large increase in CDL above 10 mM Cl−, no. DMR-0314233.
we measured the roughness of copper films electrodeposited at
References
−0.18 V for 15 min using AFM. We find that the increase in surface
area between a sample plated from the sulfate electrolyte and one 1. J. W. Dini, Electrodeposition of Copper in Modern Electroplating, 4th ed., M.
Schlesinger and M. Paunovic, Editors, p. 61, Wiley, New York 共2000兲.
plated from the 4 mM Cl− solution is ⬃6%. We could not measure 2. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
the surface area of samples grown from electrolytes containing Res. Dev., 42, 567 共1998兲; J. D. Reid, Jpn. J. Appl. Phys., Part 1, 40, 2650 共2001兲.
10 mM Cl− or higher due to their pronounced topography. We thus 3. J. O’M. Bockris and E. Mattson, Trans. Faraday Soc., 55, 1586 共1959兲.
4. J. O’M. Bockris and M. Enyo, Trans. Faraday Soc., 58, 1187 共1962兲.
attribute the large increase in CDL in the solution containing 10 and 5. O. R. Brown and H. R. Thirsk, Electrochim. Acta, 10, 383 共1965兲.
100 mM Cl− to the increase of the actual surface area of the elec- 6. D. M. Soares, S. Wasle, K. G. Weil, and K. Doblhofer, J. Electroanal. Chem., 532,
trode. 353 共2002兲.
7. G. G. Lang, M. Ujvari, and G. Horanyi, J. Electroanal. Chem., 522, 179 共2002兲.
Conclusion 8. T. P. Moffat, Mater. Res. Soc. Symp. Proc., 451, 75 共1997兲.
9. C. Gabrielli, P. Moçotéguy, H. Perrot, and R. Wiart, J. Electroanal. Chem., 572,
The kinetics and mechanisms of copper reduction from sulfate, 367 共2004兲.
chloride, and mixed electrolytes were studied by various electro- 10. Z. Nagy, J. B. Blaudeau, N. C. Hung, L. A. Curtiss, and D. J. Zurawski, J. Elec-
chemical techniques in a broad range of chloride concentrations. trochem. Soc., 142, L87 共1995兲.
11. T. Kekesi and M. Isshiki, J. Appl. Electrochem., 27, 982 共1997兲.
The presence of chloride ions causes two competing effects: at low 12. G. W. Tindall and S. Bruckenstein, Anal. Chem., 40, 1402 共1968兲; U. Bertocci,
chloride concentration 共few mM兲 the formation of an adsorbed chlo- Electrochim. Acta, 11, 1261 共1966兲.
ride layer mediates copper reduction and introduces an additional 13. K. Doblhofer, S. Wasle, D. M. Soares, K. G. Weil, G. Weinberg, and G. Ertl, Z.
reaction pathway, resulting in an overall depolarization of the reduc- Phys. Chem., 217, 479 共2003兲.
14. R. M. Smith and A. E. Martell, Critical Stability Constants, Inorganic Complexes,
tion process with no significant change of the Tafel slope. At high Vol. 4, Plenum, New York 共1976兲.
chloride concentration, complexation phenomena quantified by elec- 15. R. W. Ramette and G. Fan, Inorg. Chem., 22, 3323 共1983兲.
trochemical equilibria calculations become relevant and cause a net 16. R. W. Ramette, Inorg. Chem., 25, 2481 共1986兲.
decrease of the concentration of free cupric ions. This induces a 17. R. A. Robinson and R. H. Stokes, Faraday Trans., 45, 612 共1949兲.
18. M. Wang, Y. Zhang, and M. Muhammed, Hydrometallurgy, 45, 53 共1997兲.
cathodic polarization of the deposition process. Correspondingly, a 19. A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and Ap-
definite transition in the Tafel slope from about plications, John Wiley & Sons, New York 共1980兲.
120–140 to 78 mV/dec is observed. At high chloride concentration, 20. R. K. Astakhova and B. S. Krasilov, Zh. Prikl. Khim. (S.-Peterburg), 44, 363
low overvoltages, and under steady-state conditions there is indirect 共1971兲.
21. J. D. Reid and A. P. David, J. Electrochem. Soc., 134, 1389 共1987兲.
evidence for the formation of a solid CuCl phase, which causes 22. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3477 共1998兲.
further inhibition of the deposition process; however, this film does 23. E. Chassaing and R. Wiart, Electrochim. Acta, 29, 649 共1984兲.
not form under practical plating conditions. 24. G. M. Brown and G. A. Hope, J. Electroanal. Chem., 405, 211 共1996兲.
Impedance measurements confirm the presence of these two 25. S. Smolinski and J. Sobkowsi, J. Electroanal. Chem., 463, 1 共1999兲.
26. P. Broekmann, M. Wilms, M. Kruft, C. Stuhlmann, and K. Wandelt, J. Electroanal.
competing effects. In particular, with increasing Cl− concentration, Chem., 467, 307 共1999兲.
共i兲 the charge-transfer resistance and the double-layer capacitance 27. S. Wu, J. Lipkowski, T. Tyliszczak, and A.P Hitchcock, Prog. Surf. Sci., 50, 227
decrease and then increase and 共ii兲 the characteristic frequency in- 共1995兲.
28. C. H. Hamann, A. Hamnett, and W. Vielstich, Electrochemistry, p. 103, Wiley-
creases and then decreases. The decrease in Tafel slope corresponds VCH, New York 共1998兲.
to a large increase of the impedance loop related to the Cu+ /Cu step, 29. J. O’M. Bockris, B. E. Conway, and E. Yeager, Comprehensive Treatise of Elec-
indicating a transition in the main reduction mechanism from the trochemistry, Vol. 1, p. 143, Plenum Press, New York 共1980兲.

Downloaded on 2016-03-05 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).