Electrodeposition of Ni in Submicrometer Trenches

Article abstract of DOI:10.1149/1.2749188

A survey of the effect of cationic, anionic, and nonionic surfactants on the rate and morphological evolution of nickel electrodeposition is presented. Attention is given to the prospect for void-free filling of submicrometer trenches. Cationic species such as polyethyleneimine (PEI) and cetyl-trimethyl-ammonium (CTA+) yield significant inhibition of nickel deposition. For a range of concentrations the single cationic surfactant systems exhibit hysteretic voltammetric curves that, when corrected for ohmic electrolyte losses, reveal an S-shaped negative differential resistance. Void-free bottom-up superconformal feature filling is observed when operating at potentials within the hysteretic regime whereby metal deposition begins preferentially in the most densely patterned regions of the wafer followed by propagation of the growth front laterally across the wafer surface. In contrast, at low overpotentials and concentrations, sulfur-bearing additives such as thiourea (TU) exert a depolarizing effect on nickel deposition and negligible hysteresis. When PEI and TU are both present, the suppression provided by PEI is diminished and feature filling leads to uniform deposition on the wafer scale. Suitable combinations of PEI and TU enable near void-free filling of 230 nm wide trenches with sloping (～3.5°) sidewalls. Initial conformal growth is followed by geometric leveling once the deposits on the sloping sidewalls meet.

Full text of DOI:10.1149/1.2749188

Journal of The Electrochemical Society, 154 ͑9͒ D443-D451 ͑2007͒
D443
0013-4651/2007/154͑9͒/D443/9/$20.00 © The Electrochemical Society
Electrodeposition of Ni in Submicrometer Trenches
a
,z
*
S.-K. Kim, J. E. Bonevich, D. Josell, and T. P. Moffat
Materials Science and Engineering Laboratory, National Institute of Standards and Technology,
Gaithersburg, Maryland 20899, USA
A survey of the effect of cationic, anionic, and nonionic surfactants on the rate and morphological evolution of nickel electrodepo-
sition is presented. Attention is given to the prospect for void-free filling of submicrometer trenches. Cationic species such as
polyethyleneimine ͑PEI͒ and cetyl-trimethyl-ammonium ͑CTA + ͒ yield significant inhibition of nickel deposition. For a range of
concentrations the single cationic surfactant systems exhibit hysteretic voltammetric curves that, when corrected for ohmic
electrolyte losses, reveal an S-shaped negative differential resistance. Void-free bottom-up superconformal feature filling is ob-
served when operating at potentials within the hysteretic regime whereby metal deposition begins preferentially in the most
densely patterned regions of the wafer followed by propagation of the growth front laterally across the wafer surface. In contrast,
at low overpotentials and concentrations, sulfur-bearing additives such as thiourea ͑TU͒ exert a depolarizing effect on nickel
deposition and negligible hysteresis. When PEI and TU are both present, the suppression provided by PEI is diminished and
feature filling leads to uniform deposition on the wafer scale. Suitable combinations of PEI and TU enable near void-free filling
of Ն230 nm wide trenches with sloping ͑ϳ3.5°͒ sidewalls. Initial conformal growth is followed by geometric leveling once the
deposits on the sloping sidewalls meet.
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2749188͔ All rights reserved.
Manuscript submitted March 13, 2007; revised manuscript received April 25, 2007. Available electronically July 2, 2007.
One of the challenges in manufacturing three-dimensional ͑3D͒
structures associated with ultralarge-scale integration ͑ULSI͒, micro-
electromechanical systems ͑MEMS͒, and 3D packaging is achieving
void-free filling of trenches and vias. Two different processing
nickel cations occurs, simplifying the analyses. Likewise, the studies
were performed at pH 3, which minimized the parasitic effects of
hydrogen evolution on the voltammetric measurements and avoided
potential complications associated with the H2 bubbles that form
readily in more acidic electrolytes. As a point of departure the ef-
fects of prototypical cationic, anionic, and nonionic surfactants on
the nickel deposition were examined. The survey is far from exhaus-
tive, and the molecules examined were largely selected from prior
reports of conventional Ni leveling11-23 as well as Cu electroplating
and superfilling.23-30
1
schemes known as through-mask plating ͑or LIGA͒ and Dama-
scene processing,2 respectively, have emerged. In through-mask
plating, metal deposition occurs on the exposed areas of an
otherwise-blocked electrode. Two manifestations of this are metal
plating in nanopores of
a metal-backed anodized alumina
membrane3 and selective plating on a metallized substrate patterned
with an overlying insulating photoresist.2 In contrast, the demands
of producing multilevel wiring for microelectronic interconnects has
spawned the Damascene process whereby a 3D patterned surface is
first metallized to ensure conductivity across the entire surface,2 and
void-free bottom-up superconformal filling of the trenches and vias
is then accomplished through the effect of competitive adsorption of
accelerating and inhibiting electrolyte additives combined with the
consequences of area change.4 The resulting superconformal depo-
sition mode, called “superfilling,” has been quantitatively described
in terms of a curvature enhanced adsorbate coverage model.
Through-mask deposition has been applied to a wide variety of ma-
terials ranging from metals to semiconductors, including applica-
tions in both passive and active devices. In contrast, the Damascene
electrodeposition filling process has been limited to passive metal
conductors such as copper,4 silver5,6 and gold.7,8 At this juncture it is
interesting to see if the bottom-up superfilling process can be ex-
tended to ferromagnetic materials that might be useful as active
device elements in ULSI and MEMS. A variety of device geometries
can be imagined from partially filled trenches, i.e., horseshoe mag-
nets, or vias, i.e., magnetic cylinders, to completely filled trench or
via structures that are embedded in a magnetically isolated, nonfer-
rous environment, etc. In this paper, nickel deposition was examined
as a prototypical case of magnetic material with full recognition that
the results are likely to also be applicable to the deposition of cobalt,
iron, and alloys thereof.9,10
Experimental
Ni electrodeposition was studied using a variant of a Watts bath
composed of 1 mol/L NiSO4·6H2O, 0.2 mol/L NiCl2·6H2O, and
0.5 mol/L H3BO3 dissolved in 18 M⍀ cm deionized water. The
working electrode for the electrochemical analysis was an oxygen-
free, high-conductivity Cu plate sequentially polished from 350 to
2400 grade SiC papers. The working electrode was covered with 3M
plater’s tape that had been punched to define a circular working area
of 2.62 cm2. A nickel plate and a saturated calomel electrode ͑SCE͒
were used as the counter and reference electrodes, respectively. The
cell for the electrochemical experiments was a Teflon cylinder of
5 cm diameter and 8.5 cm height with parallel, vertically oriented
working and counter electrodes separated by a distance of 1.3 cm.
The SCE reference electrode was placed midway between the work-
ing and counter electrode but laterally positioned so as not to inter-
fere with the current distribution between the other electrodes. A
distance of 1.8 cm separated the working and reference electrode
and impedance measurements revealed an uncompensated ohmic re-
sistance of 9.35 ⍀ cm2.
The effect of a variety of organic additives on the nickel
deposition rate was assessed using slow sweep rate ͑1 mV/s͒
voltammetry. The species examined were drawn from previous re-
ports of additive-induced inhibition and/or acceleration in nickel9-23
and copper23-30 plating systems. Potential suppressor or rate-
inhibiting species include dodecyltrimethylammonium chloride
͓DTAC, CH3͑CH2͒11N+͑CH3͒3Cl−; Flukab͔, dodecyltrimethylam-
monium bromide ͓DTAB, CH3͑CH2͒11N+͑CH3͒3Br−͔; sodium
A survey of the effect of various electrolyte additives on the
nickel deposition kinetics, as revealed by voltammetric and chrono-
amperometric measurements, is presented. This is complemented by
studies of feature filling that reveal a range of interesting effects on
morphological evolution. Filling studies were performed at modest
potentials and current density where negligible depletion of the
dodecyl sulfate ͓SDS, CH3͑CH2͒11OSO−3Na+; Fisher͔, 2-butyne-1,4-
diol ͑2-B-1,4-D, HOCH2CwCCH2OH, Aldrich͒, saccharin
͑C7H5NO3S,
Aldrich͒,
Aldrich͒,
sodium
cetyltrimethylammonium
benzenesulfonate
͑SBS,
chloride
C6H5SO−3Na+,
*
Electrochemical Society Active Member.
a Present address: Center for Fuel Cell Research, Korea Institute of Science and
Technology, Cheongryang, Seoul 130-650, Korea.
b The names of companies and products are included for completeness of descrip-
tion. They do not imply NIST endorsement.
z E-mail: thomas.moffat@nist.gov
Downloaded on 2015-03-14 to IP 134.71.135.144 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D444
Journal of The Electrochemical Society, 154 ͑9͒ D443-D451 ͑2007͒
͓CTAC, CH3͑CH2͒15N+͑CH3͒3Cl−; Alfa Aesar͔, polyethyleneimine
͕PEI,-͓NH+2CH2CH2NH+͑CH2CH2NH3+͒CH2CH−2͔n,
1800 Mw;
branched, Alfa Aesar͖, 4-picoline ͑CH3C5H4N, Alfa Aesar͒, polyeth-
ylene glycol ͓PEG, ͑-CH2CH2O-͒n, 3400 Mw; Aldrich͔. Some
sulfur-containing organic additives, such as thiourea ͑TU,
H2NCSNH2, Alfa Aesar͒, 3-mercapto-1-propane sulfonic acid, so-
dium salt ͓MPS, HS͑CH2͒3SO−3Na+, Raschig͔, bis͑3-sulfopropyl͒ di-
sulfide, sodium salt ͓SPS, Na+2͓SO3−͑CH2͒3S͔2, Raschig͔, and 3-
N,N-dimethylaminodithiocarbamoyl-1-propane sulfonic acid, so-
dium salt ͓DPS, Na+SO3−͑CH2͒3SCSN͑CH3͒2, Raschig͔ were inves-
tigated as potential accelerating or depolarizing additives. For the
survey, the concentrations of the suppressors were fixed at
100 ␮mol/L with the exception of PEI that exhibited similar inhibi-
tion at much lower concentrations.28 In order to more fully charac-
terize the effect of PEI the concentration was varied from 2 to
200 ␮mol/L. For comparison to the rate-suppressing additives, ac-
celerating sulfur-bearing species were surveyed using a fixed con-
centration of 100 ␮mol/L. This concentration is within the reported
depolarizing regime for TU, a bifunctional additive that exhibits an
accelerating effect at low concentrations and an inhibiting effect at
higher concentrations.11,12,20 Similar behavior has been reported for
DPS on Cu deposition.27
Feature filling was examined using Cu-seeded trenches approxi-
mately 770 nm deep and 5 ␮m to 210 nm wide to probe the effect
of the suppressing and accelerating additives as well as combina-
tions thereof. Depositions were conducted at −0.9 V vs SCE for
3 min in the base electrolyte with designated concentrations of the
additives. In order to minimize seed-layer corrosion the specimens
were immersed into the electrolyte with the potential applied. Fea-
ture filling was also examined as a function of deposition time.
Specimen cross sections obtained by mechanical polishing followed
by ion milling were examined by field emission scanning electron
microscopy ͑FESEM͒. A subset of samples was also prepared by a
single-step focus ion beam milling and examined by SEM. Filled
features were also examined by transmission electron microscopy of
cross sections prepared using traditional dimpling and ion-milling
methods.
Results and Discussion
Figure 1. ͑Color online͒ Slow sweep voltammetry survey of the effect of
cationic, anionic, and nonionic surfactants on Ni deposition from an electro-
lyte of 1 mol/L NiSO4·6H2O + 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3.
The sweep rate was 1 mV/s. ͑a͒ Results for additives having no, or minimal,
polarizing effect. ͑b͒ Results for additives that significantly inhibit the nickel
deposition.
Inhibitors.— A voltammetric survey of the influence of rate sup-
pressing additives on the kinetics of Ni deposition is presented in
Fig. 1. The results were separated into two groups on the basis of the
observed inhibition: ineffective ͑Fig. 1a͒ and effective ͑Fig. 1b͒. The
voltammetric curve for the additive-free Ni electrodeposition shows
an onset of deposition between −0.65 and −0.70 V followed by a
linear characteristic at higher overpotentials, i.e., where the current
density is greater than 20 mA/cm2, ͑not shown͒ that reflects the
considerable electrolyte resistance associated with the system.
Evaluation of the reciprocal of the slope for the additive-free system
between 60 mA/cm2 at −1.45 V and 20 mA/cm2 at −1.1 V yields
an electrode area normalized solution resistance of 8.75 ⍀ cm2 that
is consistent with the value of 9.35 ⍀ cm2 determined by high-
frequency impedance measurements. The additives examined in Fig.
1a include a negatively charged C12 surfactant ͑SDS͒, a diol with an
unsaturated hydrocarbon chain ͑2-B-1,4-D͒, a negatively charged
aromatic compound ͑SBS͒, and heterocyclic aromatic compound
͑saccharin͒, all of which have previously been studied for Ni
electrodeposition.11-23 At first glance, none of these additives exert a
significant effect on the voltammetric behavior of Ni deposition for
the concentration examined. In contrast, the cationic surfactants
DTAB, DTAC, 4-picoline, CTAC, and PEI all exhibit some degree
of inhibition, as shown Fig. 1a and b. The voltammetric curves have
a hysteretic character whereby the inhibition evident at small over-
potentials is followed by abrupt breakdown and reversion to the
response of the additive-free system for the remainder of the volta-
mmetric sweep ͑including the return scan͒. The position of the
breakdown or critical potential is a function of the surfactant spe-
cies: −0.87 V for 4-picoline and ϳ−0.92 V for CTAC and PEI, as
shown in Fig. 1b, while weaker inhibition is provided by DTAC and
DTAB with the critical potential being near −0.81 V, as shown in
Fig. 1a.
Based on the literature correlations31 between the point of zero
charge ͑pzc͒ and work function values, it is anticipated that the Ni
surface is negatively charged under the given deposition conditions.
Thus, one plausible origin of the inhibition of the nickel deposition
reaction may be an ion-pairing interaction between the surface and
the cationic surfactant that blocks access of nickel cations to the
electrode. The observation that CTAC is more effective compared to
DTAC or DTAB indicates that some contribution related to hydro-
phobicity ͑C16 vs C12͒ is also important. As the pyridine derivative
4-picoline has suppressor characteristics that lie between that of
DTAC, DTAB, and PEI/CTAC, it would appear that the strength of
the hydrophobic interaction associated with the aromatic rings lies
between that provided by the C16 vs C12 alkane chains. As shown in
Fig. 1b, PEI is a particularly effective suppressor. The nitrogen-
bearing weak polyelectrolyte has a pKa ͑pH͒ of ϳ10.4,32 such that
the numerous ͑ϳ40͒ imine groups within a given molecule can
acquire a positive charge via protonation in the acidic Watts bath.
Downloaded on 2015-03-14 to IP 134.71.135.144 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 ͑9͒ D443-D451 ͑2007͒
D445
Figure 2. ͑Color online͒ PEI concentration strongly influences the slow
sweep voltammetry of Ni deposition from mol/L NiSO4·6H2O
Figure 3. ͑Color online͒ S-NDR is revealed in the voltammograms of Fig.
1b once they are corrected for the IR voltage drop associated with the resis-
tive electrolyte. The multiplicity of current values for a given potential gives
rise to strong nonlinear active-passive behavior. The three different regimes
of behavior are indicated in the insert.
1
+ 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3. The sweep rate was 1 mV/s.
Inhibition increases monotonically as PEI concentration is increased from 0
to 200 ␮mol/L. Inhibition breaks down at increasingly more negative poten-
tials as PEI concentrations increases to 5 ␮mol/L and subsequently, the
additive-free deposition kinetics prevail.
couples mass-transport-limited adsorption of a blocking additive and
current-density-dependent removal of the blocking additive from the
surface by incorporation into the growing deposit, the latter repre-
senting a positive feedback term.35 In the absence of significant
ohmic losses in the electrolyte such bistable NDR systems move
rapidly between the two states depending on the experimental
conditions.34,35 In the presence of significant ohmic losses in the
electrolyte the NDR is obscured and the voltammetric response ex-
hibits only a monotonic slope, as evident in Fig. 1b and 2. Impor-
tantly, in related activated-inhibitor coupled systems, the combina-
tion of NDR with a distributed electrical potential has been shown to
give rise to standing Turing-type patterns on the electrode.34,36 In the
present case the patterned substrate provides a perturbation on the
mass-transport conditions as a function of position while the finite
seed-layer resistance combines with the electrolyte resistance to per-
turb the electrical distribution during metal deposition. Preliminary
feature-filling experiments, as detailed in a later section, provide
some evidence for such patterning effects in this system.
The multiple cationic sites per molecule also account for the low
͑5 ␮mol/L͒ PEI concentration required to achieve substantial inhi-
bition.
The nonionic polymer PEG also exerts strong inhibition that is
sustained over the entire potential range shown in Fig. 1b. Negli-
gible hysteretic character is evident and an identical response is
observed for a lower, e.g., 50 ␮mol/L, PEG concentration ͑not
shown͒. This is similar to the saturated suppression behavior ob-
served for a PEG-Cl blocking layer in the copper system and can be
described in terms of a simple area blockage model that includes a
persistent “leakage current” resulting in spatially uniform deposition
across the work piece.26,33
In contrast to PEG, inhibition derived from PEI exhibits a strong
dependence on concentration. As shown in Fig. 2, the critical poten-
tial shifts systematically to more negative potentials with increasing
PEI concentration. For concentrations up to 10 ␮mol/L PEI, the
kinetic response reverts to additive-free behavior beyond the critical
potential; this gives rise to a well-defined hysteretic response. For
concentrations beyond 10 ␮mol/L, inhibition continues to increase;
however, the hysteretic response is attenuated and the additive-free
kinetic response is no longer observed. By 20 ␮mol/L, PEI sup-
presses deposition as effectively as PEG over the full potential
range, although the rough character of the curve suggests unstable
behavior. For PEI concentrations of 100 ␮mol/L and above, the
inhibition saturates and negligible hysteresis is seen.
The abrupt breakdown of inhibition characterized by the critical
potential most clearly evident for PEI, CTAC, and 4-picoline de-
serves further analysis. When the experimental voltammetric results
are corrected for the uncompensated solution resistance, a negative
differential resistance ͑NDR͒ in the interface kinetics is revealed as
shown in Fig. 3. ͑This postexperiment correction does not consider
the nonlinear distortions of the actual sweep rate associated with the
ohmic drop, the magnitude of which scales with current density.͒
The S-shaped NDR ͑S-NDR͒ region corresponds to the existence of
two stable states for a given potential. Such bifurcation processes
have a rich history in electrochemistry, ranging from studies of
active-passive transitions of corroding metals to electrocatalytic pro-
cesses, all of which have been connected with pattern formation via
the coupling of interface kinetics and transport processes.34 For the
case of additive plating, a steady-state model is available that
Accelerators.— Voltammetry was also used to identify depolar-
izing additives for Ni electrodeposition; the results are summarized
in Fig. 4. Four sulfur-containing additives were examined: TU, SPS,
MPS, and DPS. TU, one of the most investigated additives in Ni
electrodeposition, can either accelerate or inhibit the deposition re-
action depending on its concentration in the electrolyte. The shift
from depolarizing to polarizing influence has been reported to occur
at concentrations between several hundreds ␮mol/L and several
mmol/L.11,12,20 Successful leveling during Ni electrodeposition us-
ing the suppression effects associated with higher concentrations of
TU has been reported.11,19 There is also one report of TU-
derivatized electrodes becoming bright during nickel plating in the
absence of TU.37 This report, along with similar observations re-
ported in the case of copper plating, motivated the investigation
detailed herein.4,24,25,38,39 The depolarization associated with low
TU concentrations in the electrolyte has been ascribed to the sulfide
generated during electrochemical reduction of sulfur associated with
the decomposition of TU at the cathode surface during the Ni
deposition.20 The characteristic reduction potential for such TU de-
composition to sulfide ion has been attributed to a voltammetric
wave near −0.7 V vs SCE.17 A similar reduction wave is evident in
Fig. 4 for the S-containing depolarizing additives with the wave
peak approximately at −0.75 V for TU, −0.72 V for SPS, −0.71 V
Downloaded on 2015-03-14 to IP 134.71.135.144 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D446
Journal of The Electrochemical Society, 154 ͑9͒ D443-D451 ͑2007͒
Figure 6. ͑Color online͒ Slow sweep voltammetry survey of the combined
effect of CTAC and depolarizing sulfur-bearing additives on the rate of
Figure 4. ͑Color online͒ Slow sweep voltammetry survey of the effect of
sulfur-bearing additives on the rate of Ni deposition from mol/L
LNiSO4·6H2O + 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3. The sweep
rate was 1 mV/s.
1
Ni deposition. The base electrolyte was
1
mol/L NiSO4·6H2O
+ 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3 and the sweep rate was
1 mV/s.
for MPS, and −0.69 V for DPS. The magnitude of the depolarizing
effects at −0.8 V increases in order of TU Ͼ SPS Ͼ DPS Ͼ MPS.
As the overpotential increases the depolarizing effects of the addi-
tives diminish; TU and SPS actually begin to inhibit the nickel depo-
sition rate at potentials below −0.92 V. A similar potential-
dependent transition in TU behavior from acceleration to inhibition
was also recently reported.17
Nonetheless, the accelerators induce depassivation of the electrode
at smaller overpotentials than that in the presence of PEI alone, as
evident by the shift in the onset potential of abrupt depassivation
toward more positive potentials. The strength of the depolarizing
action increased in an order TU Ͼ DPS ꢀ SPS Ͼ MPS. The most
effective PEI depassivating agents, DPS and TU, have amine and
dimethyl amine groups ͓-NH2, -N͑CH3͒2͔, respectively, conjugated
with the neighboring CvS bond in their molecular structures. No
such functionality exists for the less effective depassivating addi-
tives SPS and MPS. It is possible that the lone pair electrons avail-
able in the amines for adsorption or protonation, similar to PEI,
facilitate access and interaction of the S-bearing group of TU and
DPS with the nickel surface beneath the PEI layer and thereby en-
able disruption of the PEI layer at smaller overpotentials. Similar
behavior was observed for combinations of 100 ␮mol/L CTAC sup-
pressor and 100 ␮mol/L accelerator as shown in Fig. 6; in this case,
however, the order of depassivation activity was MPS/SPS
Ͼ DPS Ͼ TU. Combinations of 100 ␮mol/L PEG and the
100 ␮mol/L S-bearing additives exhibited only a very slight accel-
eration at small overpotentials, as shown in Fig. 7. Surprisingly, all
the systems, with the exception of TU-PEG, lead to slightly greater
inhibition at more negative potentials then that provided by PEG
alone.
Combination of inhibitors and accelerators.— Voltammetry
was used to examine the interactions between the inhibiting cationic
and nonionic surfactants, PEI, CTAC, and PEG, and the S-bearing
accelerator species. The voltammetric behavior of Ni deposition in
the presence of 5 ␮mol/L PEI with 100 ␮mol/L of the various de-
polarizing additives is shown in Fig. 5. The reduction wave near
−0.7 V in the voltammetric response of the “accelerators-only” elec-
trolyte ͑i.e., Fig. 4͒ is sharply suppressed in the presence of PEI.
Feature filling.— Based on the above survey, PEI, TU, and com-
binations thereof were chosen for further study of their effect on
submicrometer feature filling. For the experiments shown here, Cu-
seeded Damascene trench structures were examined; for consis-
tency, the voltammetric results presented in the previous section
were also obtained using copper substrates. The copper seed layers
were used because the surface oxide can be easily reduced; the
reduction of oxidized nickel substrates is more challenging, particu-
larly for a pH greater than 2.5. A small number of Ni-seeded speci-
mens was also examined to ascertain if any significant observations
were associated with idiosyncrasies of nickel deposition on the cop-
per seeds; none were.
Figure 8 shows a cross-sectional image of a Ni electrodeposit
grown in additive-free solution for 3 min at −0.9 V. The Ni deposit
on the 5 ␮m wide trench is conformal. The slightly smaller thick-
ness on the sidewalls reflects the difference between the seed-layer
thickness on the sidewalls vs the free surface and bottom due to
line-of-sight constraints during physical vapor deposition ͑PVD͒
preparation. The growth front exhibits noticeable roughness. The
Figure 5. ͑Color online͒ Slow sweep voltammetry survey of the combined
effect of PEI and depolarizing sulfur-bearing additives on the rate of Ni
deposition. The electrolytes were composed of 1 mol/L NiSO4·6H2O
+ 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3 and the sweep rate was
1 mV/s.
Downloaded on 2015-03-14 to IP 134.71.135.144 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 ͑9͒ D443-D451 ͑2007͒
D447
Figure 7. ͑Color online͒ Slow sweep voltammetry survey of the combined
effect of PEG and depolarizing sulfur-bearing additives on the rate of
Ni deposition. The base electrolyte was mol/L NiSO4·6H2O
1
+ 0.2 mol/L NiCl2·6H2O + 0.5 mol/L H3BO3 and the sweep rate was
1 mV/s.
two widest, low-aspect-ratio trenches appear to be void-free while
voids are evident ͑as marked͒ in the narrower high-aspect-ratio fea-
tures. In light of the modest overpotential, negligible depletion of
the Ni2+ accompanies the deposition, and thus the voids likely result
from the roughness of the impinging sidewall surfaces.
Feature filling in the presence of PEI.— Additions of PEI in-
duce significant changes in the feature-filling dynamics. Conformal
filling of an ϳ1 ␮m wide trench in the additive-free electrolyte is
shown in Fig. 9a. The addition of 5 ␮mol/L PEI yields supercon-
formal film growth, as revealed in Fig. 9b, by preferential deposition
at the bottom and on the deeper sections of the sidewall surfaces of
the trench, while more limited deposition occurs on the neighboring
free surface. Preferential deposition of nickel is also evident in the
bottom corners of larger features such as that shown in Fig. 9c.
Preliminary examination of finer features ͑not shown͒ suggests that
void-free filling of the trenches persists up to an aspect ratio of at
least 2.3 for a midheight trench width greater than 300 nm. The
roughness of the growing sidewall surfaces most likely prevents
void-free feature filling of the higher aspect ratio trenches. The
shape change apparent in Fig. 9b might be explained as a variation
Figure 9. Trenches following 3 min of Ni deposition at −0.9 V: ͑a͒ confor-
mal Ni deposition from an additive-free electrolyte, ͑b͒ superconformal Ni
deposition in the presence of 5 ␮mol/L PEI, and ͑c͒ another region of the
same specimen showing PEI-induced preferential deposition of Ni at the
concave corner of wider features ͑FESEM͒.
dispersion in the feature-filling dynamics was observed, e.g., Fig.
10f, that might be partly related to inhomogeneities in the seed-
layer/barrier of the wafer fragments.
of the diffusion-adsorption–consumption model for
a single-
component leveler.40 However, further work detailing the feature-
filling dynamics is required in order to establish this.
Similar experiments performed in 10 ␮mol/L PEI revealed fur-
ther evidence of preferential deposition toward the bottoms of the
finer features, as shown in Fig. 10a-d. The sloping deposits on the
sidewalls suggest a PEI depletion effect in the higher aspect ratio
features that is congruent with a consumption-driven leveler deple-
tion mechanism, as suggested for Fig. 9b.40 Negligible nickel depo-
sition occurs on the top surface between the trenches shown in Fig.
10a-d; the convex bump there is almost entirely the copper seed
layer. In the most favorable case ͑Fig. 10e͒, feature filling and pla-
narity are obtained with modest overburden. However, significant
Figure 10. ͑a–d͒ PEI-induced superconformal growth can result in void-free
feature filling. ͑e͒ Subsequent formation of a smooth planar surface. ͑f͒ An
example of experimental dispersion where deposition only occurs in the
lower half of the trenches. The Ni films were all deposited for 3 min at
−0.9 V in electrolyte containing 10 ␮mol/L PEI ͑FESEM͒.
Figure 8. Trenches after 3 min Ni deposition at −0.9 V in additive-free
electrolyte. Trench widths ͑left to right͒: 5 ␮m, 700 nm, 400 nm, and
230 nm. The white circles mark voids created when the opposing sidewall
surfaces impinged ͑FESEM͒.
Downloaded on 2015-03-14 to IP 134.71.135.144 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D448
Journal of The Electrochemical Society, 154 ͑9͒ D443-D451 ͑2007͒
Figure 11. The convolution of trench filling with pattern density effects in an
electrolyte containing 10 ␮mol/L PEI for Ni deposition at −0.9 V for 3 min.
Feature filling appears to initiate in the centers of the densely patterned
regions with the growth front moving vertically within the feature while the
nucleation and growth process simultaneously propagates laterally between
neighboring features. This unusual growth mode is associated with the NDR
system behavior ͑FESEM͒.
Figure 12. Trench filling in electrolytes containing 5 ␮mol/L PEI and
␮mol/L TU concentrations indicated. Ni deposition at −0.9 V for 3 min.
Trench widths ͑left to right͒: 5 ␮m, 700 nm, 400 nm, and 230 nm ͑FESEM͒.
In addition to the superconformal feature-filling mode, deposi-
tion from 10 ␮mol/L PEI electrolyte revealed a variety of interest-
ing pattern-density-dependent effects. For example, specimens
grown at −0.9 V SCE for 3 min revealed multiple examples of pref-
erential nucleation and bottom-up growth occurring in the finest and
most densely packed trench arrays with negligible deposition evi-
dent on neighboring planar or lower feature density regions. Four
examples of this behavior are shown in Fig. 11. The images reveal
distinct active and passive areas evidently associated with the break-
down dynamics of the S-NDR voltammetric response outlined ear-
lier; limited feature filling experimentation ͑not shown͒ reveals that
the inhomogeneous deposition process diminishes as the applied po-
tential is moved away from the S-NDR region identified in the iR-
corrected slow sweep voltammetry. The same pattern density effects
are observed when a nickel seed layer is used ͑not shown͒. The
coupling of PEI depletion effects in feature filling with the potential
distribution in the resistive electrolyte could provide a natural expla-
nation for the inhomogeneous passive-active surface dynamics.34-36
The inhomogeneity in growth, a manifestation of self-patterning,
enables selective deposition triggered by the nonuniform substrate
topography. The substrate pattern may be analogous to a perturba-
tion analysis of the stability of the NDR system.34-36 A more detailed
description of this coupling behavior will be developed in a future
publication.
Similar feature-filling results have been reported for Cu deposi-
tion using PEG and halide ions as cosuppressors.41 The phenomenon
was rationalized in terms of preferential halide consumption within
the high-density patterned field; however, under the condition stud-
ied in Ref. 40, the PEG and halide system likely also exhibits
S-NDR behavior. Therefore, coupling of halide depletion gradients
and the corresponding potential distribution might also provide a
more complete description of the unusual feature-filling process.
Likewise, the fall-off in feature filling at the transition between two
different feature sizes, like that shown in Fig. 11b and c, has also
been reported for some high accelerator/suppressor ratio, low and
medium acid ͑i.e., non-negligible solution resistance͒ Damascene
copper plating baths.42 Recent research on single-additive filling of
through-vias in printed circuit boards has also revealed active-
passive areas that are linked to the substrate topography.43 The wide-
ranging nature of these observation highlights the need for further
study of this intriguing coupled gradient deposition mode, namely,
leveler depletion effects combined with a topographically dependent
potential distribution.
Feature filling with PEI and TU.— Trench filling was also ex-
amined for electrolytes with 5 ␮mol/L PEI and a range of TU con-
centration. Figure 12 summarizes the results for specimens depos-
ited at −0.9 V for 3 min. The first row depicts trenches filled in the
presence of the PEI alone. As noted in Fig. 9, the overall deposit
thickness is reduced below that of the additive-free case ͑Fig. 8͒,
particularly on the free surface, due to the inhibition of deposition
by PEI. The surface roughness of the growth front is also reduced
compared to the additive-free case, although a proper comparison
must consider its dependence on film thickness. The PEI-induced
superconformal deposition evident in the inhibited growth at the
convex corners and more extensive deposition on the recessed sur-
faces was also noted earlier.
The addition of 1–50 ␮mol/L TU leads to two general morpho-
logical changes: ͑i͒ marked weakening of the superconformal
growth mode and ͑ii͒ decreased local surface roughness. The addi-
tion of only 1 ␮mol/L TU causes film growth to shift back to the
conformal mode. This is accompanied by smoothing of the local
surface roughness that permits geometrical leveling to yield nomi-
nally void-free filling of trenches with sloping sidewalls; the finite
surface roughness evident on the free surface suggests that closer
microstructural analysis of the trench will reveal a seam that might
include distributed nanometer-sized voids. Increasing the TU con-
centration to 10, 50, and 100 ␮mol/L leads to a further decrease in
local roughness to the point that the surfaces all appear perfectly
smooth, even at these high magnifications, and differentiation is not
possible. For these additive concentrations filling approaches the
analytical limit of geometric leveling whereby the sloping sidewalls
enable “void-free” filling of trenches as narrow as 230 nm. Exami-
nation of narrower trenches ͑not shown in Fig. 12͒ revealed signifi-
cant voiding along the centerline. This most likely arises from the
combination of decreased lithographically defined sidewall angle
and residual roughness from the heterogeneous nickel nucleation
and growth on the aged copper seed layer. The condition for true
Downloaded on 2015-03-14 to IP 134.71.135.144 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 ͑9͒ D443-D451 ͑2007͒
D449
Figure 14. Filling of trenches in electrolyte containing 5 ␮mol/L PEI and
100 ␮mol/L TU as a function of deposition time at −0.90 V. Trench widths
͑left to right͒: 5 ␮m, 740 nm, 450 nm, 350 nm, 300 nm, and 250 nm
͑FESEM͒.
Figure 13. Trench filling in electrolytes containing 100 ␮mol/L TU and
␮mol/L PEI concentrations indicated. Nickel deposition at −0.9 V for
3 min. Trench widths ͑left to right͒: 5 ␮m, 700 nm, 400 nm, and 230 nm
͑FESEM͒.
fers substantially; in concave regions, nickel deposition is inhibited
relative to neighboring planar surfaces while on convex surfaces the
metal deposition rate is accelerated. These effects combine to yield
subconformal deposition, evident in the recessed notch in the bottom
corners of wide trenches, and void formation along the centerlines
of filled features. Consistent with these observations, it is here pro-
posed that the 100 ␮mol/L TU single additive system has crossed
from acceleration to suppression, a result that is consistent with the
transition in Fig. 4 near the −0.9 V deposition potential, and that the
TU-derived additive has a finite, i.e., nonzero, residence time on the
surface, a requirement for area change to impact adsorbate coverage.
Adding 1 ␮mol/L PEI to the 100 ␮mol/L TU-containing elec-
trolyte yields slightly inhibited deposition on planar surfaces as well
as reduced acceleration on convex surfaces. The decreased size of
the voids in the finer features is a direct result of these changes. PEI
concentration of 5 ␮mol/L yields deposition that is further sup-
pressed and more conformal. At PEI concentration of 5 and
10 ␮mol/L this effect is sustained, yielding smooth, conformal de-
posits. The smooth, conformal deposition coupled with the sloping
sidewalls enables void-free filling.
Based on the results for the combined PEI-TU system, feature
filling as a function of time and pattern dimension was investigated.
As shown in Fig. 14, deposition at −0.9 V SCE in the presence of
5 ␮mol/L PEI and 100 ␮mol/L TU proceeds conformally, with geo-
metric leveling beginning when the deposits on the sloped sidewalls
meet. The growth velocity deduced for the 5 ␮m wide trench com-
pares favorably with the 5 mA/cm2 current density measured at
−0.9 V in Fig. 5, confirming the high efficiency of the deposition
process. The cross sections do not exhibit uniform voiding; how-
ever, the three finest arrays, with trench widths Ͻ350 nm, do con-
tain small voids located randomly along the centerline, and heavily
voided seams were evident in trenches less than 230 nm wide ͑not
shown in Fig. 14͒.
Transmission electron microscopy ͑TEM͒ was used to further
examine cross sections of the filled trenches. The specimens shown
in Fig. 15 were prepared from patterned substrates after nickel depo-
sition for 5 min at −0.9 V in the electrolyte containing 5 ␮mol/L
PEI and 100 ␮mol/L M TU. Void-free filling is evident for the
650 nm deep, 300 nm wide trenches ͑average width determined
from bottom width 250 nm, top width 350 nm, sidewall slope 3.5°͒
with aspect ratio ͑trench height /average width͒ ϳ2.2. The image of
the 250 nm wide trenches ͑Fig. 15b͒ hints at seam formation along
the midlines of several of the trenches; this could be a seam of
low-density grain boundaries or discrete nanometer-size voids that
do not span the full 100 + nm thickness of the TEM specimens.
The ϳ150 nm wide trenches exhibit extensive void formation along
their centerlines. In addition to the higher aspect ratio of this feature,
it is also important to note the smaller sidewall slope of only 1.7°
away from the vertical. The geometric factors, width, aspect ratio,
void-free filling by geometric leveling involves minimizing surface
roughness and reactant depletion at the onset of sidewall impinge-
ment. An experimentally determined scaling relationship for film
roughening/smoothing as a function of film thickness, TU concen-
tration, etc., would permit analytical predictions of the feature size
threshold for void-free trench filling as a function of sidewall
angle.44
The images in Fig. 12 indicate that the growth rate passes
through a maximum somewhere between 1 and 10 ␮mol/L TU.
Higher TU concentrations result in a decreased overall deposition
rate that is most conspicuous in the filling of the 700 nm wide
trenches. TU concentrations of 200 ␮mol/L, and higher, result in a
shift from conformal to subconformal growth. In this regime the
coupling of the deposition-rate-suppressing characteristics of high
TU concentration with the effect of area change during filling be-
comes dominant. Specifically, area reduction of advancing concave
surfaces leads to enrichment of adsorbed TU with the associated
suppression of the deposition reaction evident in the bottom trench
corners. Conversely, the expanding surface area accompanying
metal deposition at the convex upper corners dilutes the coverage of
the adsorbed TU there, challenging the inhibition that it otherwise
provides. Eventually, the more rapidly growing hemicylindrical
fronts from the opposing upper corners impinge, leaving a large void
in the trench. This type of additive-enhanced, subconformal growth
was previously observed in the case of copper plating in the pres-
ence of PEG-Cl,45 whereby much larger voids were created through
the development of the convex protuberances for conditions where
little depletion of the metal cations had occurred. The demonstration
provided here ͑i.e., Fig. 12: 200 ␮mol/L TU͒ is perhaps even less
ambiguous as there is truly negligible depletion of nickel cations at
the free surface for these deposition conditions. The results reflect
the nonmonotonic coverage or concentration-dependent effect of
TU; namely, the deposition rate is accelerated at low concentrations
but inhibited at higher concentrations. Thus, when the coverage of a
saturated surface first decreases, the deposition rate passes through a
maximum before decreasing monotonically toward the additive-free
behavior.
The effect of PEI concentration on feature filling in the presence
of 100 ␮mol/L TU was also examined; the results are summarized
in Fig. 13. In the absence of PEI ͑0 ␮mol/L images͒ the thickness of
Ni deposit on planar sections, such as the bottom of a 5 ␮m wide
trench and the surrounding field, is nearly the same as that for the
additive-free case ͑Fig. 8͒. This is congruent with the voltammetric
results ͑Fig. 4͒, indicating TU has little effect on the current density
on planar electrodes. However, deposition on nonplanar regions dif-
Downloaded on 2015-03-14 to IP 134.71.135.144 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D450
Journal of The Electrochemical Society, 154 ͑9͒ D443-D451 ͑2007͒
Figure 16. FIB-SEM images of trenches filled at −0.9 V for 3 min in an
electrolyte containing 5 ␮mol/L PEI and 100 ␮mol/L TU. Sporadic voiding
at the top of ͑a͒ 350 nm and ͑b͒ 300 nm wide trenches. The trenches appear
foreshortened due to an abnormal imaging angle.
Figure 15. TEM cross sections of trenches filled by Ni deposition at −0.9 V
for 5 min in an electrolyte containing 5 ␮mol/L PEI and 100 ␮mol/L TU.
Trench widths and sidewall angles: ͑a͒ 250 nm, 3.5°; ͑b͒ 200 nm, 3.5°; and
͑c͒ 120 nm, 1.7°. The widest trenches are void-free while multiple voids are
clearly evident along the centerline of the narrowest trenches. Some of the
intermediate width trenches exhibit a low-density seam along their center-
lines.
ohmic electrolyte losses reveals an S-shaped NDR that correlates
with a bifurcation reaction between an active and passive state.
In contrast to the cationic surfactants, sulfur-bearing additives
such as TU exert a depolarizing effect on nickel deposition, at least
at low overpotentials and concentrations. When PEI and TU are both
present the suppression and the associated voltammetric hysteresis
provided by PEI is diminished, leading to uniform deposition on the
wafer scale.
Two very different schemes for feature filling were identified. In
the first instance a single additive electrolyte, based on cationic sur-
factant species such as PEI and CTAC, was shown to exhibit a
nonlinear, potential-driven, S-shaped NDR bifurcation process be-
tween an active and passive state. For PEI this phenomenon was
and sidewall angle, conspire with the finite surface roughness to
generate a significant centerline defect density in these finest fea-
tures.
According to the above results, geometrical leveling appears to
be capable of yielding void-free filling of 650 nm deep, 300 nm
wide trenches with sidewall angle of 3.5°. To explore the void-free
filling threshold further, trenches of 300 and 350 nm width were
sectioned and examined in a dual-beam focused ion beam ͑FIB͒-
SEM ͑Fig. 16͒; the imaging electron beam was inclined 52° from
the substrate normal so trench heights are foreshortened. The irregu-
lar center top voids are discontinuous, appearing and disappearing
with continued Ni removal ͑the lines beneath the voids are an arti-
fact of the FIB milling process͒. The grooved trench profile on the
top surface of a partially filled 500 nm wide trench in a similar
FIB-SEM image ͑Fig. 17͒ reveals significant surface roughness that
elucidates the origin of the void formation; the roughness and inter-
mittent contact or coalescence along the groove demarcating the
interface between the colliding sidewalls indicates how random
voids arise through “pinch-off” connected with the impingement of
the opposing sidewalls whose surface roughness is largely uncorre-
lated prior to contact.
Conclusions
Figure 17. FIB-SEM image of a partially filled trench after deposition at
−0.9 V for 3 min from an electrolyte containing 5 ␮mol/L PEI and
100 ␮mol/L TU. A thin layer of Pt was used to protect the surface of the
Ga-milled cross section; the surface details of the grooved trenches are vis-
ible in the upper, uncoated half of the image. The grooved trench profile in
the upper field reveals significant surface roughness that underlies the for-
mation of random voids during sidewall impingement that accompanies fea-
ture filling.
The impact of a selection of prototypical cationic, anionic, and
nonionic additives on the rate and morphological evolution of nickel
electrodeposition was examined. Cationic species such as proto-
nated, nitrogen-bearing PEI and CTA+ yield significant inhibition of
nickel deposition. The inhibition is subject to abrupt breakdown, as
revealed by the hysteretic voltammetric response. Correcting for
Downloaded on 2015-03-14 to IP 134.71.135.144 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 ͑9͒ D443-D451 ͑2007͒
D451
147, 219 ͑2000͒.
shown to yield unusual bottom-up superconformal feature-filling be-
havior that also included pattern scale effects whereby feature filling
began preferentially in the most densely patterned ͑high-surface-
area͒ regions and was followed by lateral propagation of the metal
nucleation and growth fronts. Many questions remain as to the de-
tailed mechanism of feature filling that evidently derives from the
close coupling of additive gradient and potential distribution.
Addition of TU to the PEI-containing electrolyte moved the sys-
tem into a noncritical regime where feature filling occurred by ho-
mogenous, conformal deposition followed by geometrical leveling.
Smoothing of the surface in the presence of TU played an important
role in minimizing void formation during geometrical leveling. The
likely extendability of these observations to the deposition of other
iron group metals and alloys is a subject for future study.
16. F. Lallemand, C. Comte, L. Ricq, P. Renaux, J. Pagetti, C. Dieppedale, and P. Gaud,
Appl. Surf. Sci., 225, 59 ͑2004͒.
17. U. S. Mohanty, B. C. Tripathy, S. C. Das, and V. N. Misra, Metall. Mater. Trans. B,
36B, 737 ͑2005͒.
18. U. S. Mohanty, B. C. Tripathy, P. Singh, and S. C. Das, J. Appl. Electrochem., 31,
969 ͑2001͒.
19. S. S. Kruglikov, N. T. Kudriavtsev, G. F. Vorobiova, and A. Ya. Antonov, Electro-
chim. Acta, 10, 253 ͑1965͒.
20. J. Edwards, Trans. Inst. Met. Finish., 41, 169 ͑1964͒.
21. G. T. Rogers, M. J. Ware, and R. V. Fellows, J. Electrochem. Soc., 107, 677
͑1960͒.
22. T. Osaka, T. Sawaguchi, F. Mizutani, T. Yokoshima, M. Takai, and Y. Okinaka, J.
Electrochem. Soc., 146, 3295 ͑1999͒.
23. Modern Electroplating, M. Schlesinger and M. Paunovic, Editors, John Wiley &
Sons, New York ͑2000͒.
24. P. L. Schilardi, O. Azzaroni, and R. C. Salvarezza, Phys. Rev. B, 62, 13098 ͑2000͒.
25. Y. M. Loshkarov and E. M. Govorova, Prot. Met., 34, 399 ͑1998͒.
26. T. P. Moffat, D. Wheeler, and D. Josell, in Superconformal Film Growth, Advances
in Electrochemical Science and Engineering, R. C. Alkire, D. Kolb, P. Ross, and J.
Lipkowski, Editors, In press ͑2007͒, and references therein.
27. S. K. Cho, S.-K. Kim, and J. J. Kim, J. Electrochem. Soc., 152, C330 ͑2005͒.
28. S.-K. Kim, D. Josell, and T. P. Moffat, J. Electrochem. Soc., 153, C616 ͑2006͒.
29. S.-K. Kim, D. Josell, and T. P. Moffat, J. Electrochem. Soc., 153, C826 ͑2006͒.
30. N. Tantavichet and M. P. Pritzker, Electrochim. Acta, 50, 1849 ͑2005͒.
31. S. Trasatti, Electrified Interfaces in Physics, Chemistry, and Biology, p. 945, R.
Guidelli, Editor, NATO ASI Series, Kluwer Academic Publishers, Boston, MA
͑1992͒.
32. R. Meszaros, L. Thompson, M. Bos, and P. De Groot, Langmuir, 18, 6164 ͑2002͒.
33. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3472 ͑1998͒.
34. K. Krischer, in Advances in Electrochemical Science and Engineering, R. C. Alkire
and D. M. Kolb, Editors, Vol. 8, p. 142, Wiley-VCH Verlag GmbH & Co, Wein-
heim, Germany ͑2003͒.
35. D. Roha and U. Landau, J. Electrochem. Soc., 137, 824 ͑1990͒.
36. Y.-J. Li, J. Oslonovitch, N. Mazouz, F. Plenge, K. Krischer, and G. Ertl, Science,
291, 2395 ͑2001͒.
37. A. T. Vagrayan and N. K. Baraboshkina, Plating, 1967, 930 ͑August͒.
38. T. P. Moffat, D. Wheeler, C. Witt, and D. Josell, Electrochem. Solid-State Lett., 5,
C110 ͑2002͒.
39. M. S. Kang, S.-K. Kim, and J. J. Kim, Jpn. J. Appl. Phys., Part 1, 44, 8107 ͑2005͒.
40. R. Chalupa, Y. Cao, and A. C. West, J. Appl. Electrochem., 32, 135 ͑2002͒.
41. M. Hayase, M. Taketani, T. Hatsuzawa, and K. Hayabusa, Electrochem. Solid-State
Lett., 6, C92 ͑2003͒.
Acknowledgments
The authors thank S. Claggett for assistance in preparation of
specimens for FESEM.
The National Institute of Standards and Technology assisted in meeting
the publication costs of this article.
References
1. L. T. Romankiw, Electrochim. Acta, 42, 2985 ͑1997͒; J. O. Dukovic, IBM J. Res.
Dev., 37, 125 ͑1993͒, and references therein.
2. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
Res. Dev., 42, 567 ͑1998͒.
3. H. Masuda and K. Fukuda, Science, 268, 1466 ͑1995͒.
4. T. P. Moffat, D. Wheeler, M. Edelstein, and D. Josell, IBM J. Res. Dev., 49, 19
͑2005͒.
5. B. Baker, C. Witt, D. Wheeler, D. Josell, and T. P. Moffat, Electrochem. Solid-State
Lett., 6, C67 ͑2003͒.
6. E. J. Ahn and J. J. Kim, Electrochem. Solid-State Lett., 7, C118 ͑2004͒.
7. D. Josell, D. Wheeler, and T. P. Moffat, J. Electrochem. Soc., 153, C11 ͑2006͒.
8. Z. M. Hu and T. Ritzdorf, J. Electrochem. Soc., 153, C467 ͑2006͒.
9. P. C. Andricacos and N. Robertson, IBM J. Res. Dev., 42, 671 ͑1998͒.
10. E. I. Cooper, C. Bonhote, J. Heidmann, Y. Hsu, P. Kern, J. W. Lam, M. Rama-
subramanian, N. Robertson, L. T. Romankiw, and H. Xu, IBM J. Res. Dev., 49,
103 ͑2005͒.
42. C. Witt, J. Srinivasan, X. Lin, and R. Carpio, ECS. Trans. 2͑6͒, 107 ͑2007͒.
43. W.-P. Dow, H.-H. Chen, M.-Y. Yen, and C.-W. Liu, ECS. Trans. 2͑6͒, 259 ͑2007͒.
44. A. De Virgiliis, O. Azzaroni, R. C. Salvarezza, and E. V. Albano, Appl. Phys. Lett.,
82, 1953 ͑2003͒.
11. S. A. Watson and J. Edwards, Trans. Inst. Met. Finish., 34, 167 ͑1957͒.
12. D. G. Foulke and O. Kardos, Proc. Amer. Electroplaters’ Soc., 43, 172 ͑1956͒.
13. O. Kardos, Proc. Amer. Electroplaters’ Soc., 43, 181 ͑1956͒.
14. J. Amblard, I. Epelboin, M. Froment, and G. Maurin, J. Appl. Electrochem., 9, 233
͑1979͒.
45. T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G. R.
15. I. Tabakovic, S. Riemer, V. Inturi, P. Jallen, and A. Thayer, J. Electrochem. Soc.,
Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 ͑2000͒.
Downloaded on 2015-03-14 to IP 134.71.135.144 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).