Magnetic materials for three-dimensional damascene metallization: Void-free electrodeposition of Ni and Ni70Fe30 using 2-mercapto-5-benzimidazolesulfonic acid

Article abstract of DOI:10.1149/1.2907433

Superconformal filling of submicrometer trenches with electrodeposited Ni is demonstrated using a NiS O4 -Ni Cl2 -FeS O4 electrolyte containing 2-mercapto-5-benzimidazolesulfonic acid (MBIS). The process offers the ability to build three-dimensional magnetically active structures that may be easily integrated with other state-of-the-art metallization schemes. MBIS acts to inhibit Ni(Fe) electrodeposition, although under certain conditions rapid, autocatalytic breakdown accompanies the onset of Ni deposition. Optimal trench filling is associated with the positive feedback process, moderated by electrolyte internal-resistance losses, and manifests as a hysteretic voltammetric response on planar electrodes for an MBIS concentration ～100 μmolL. On freshly immersed substrates, trench filling is characterized by an initial period of uniform growth followed by the development of a v-notch geometry associated with transient depletion of MBIS within the recessed feature. Interestingly, the finest submicrometer features are filled while only minimal deposition occurs on the neighboring free surface. Continued growth of the MBIS derived v-notch geometry also results in void-free filling of the larger features by geometrical leveling. The same behavior is observed for soft magnetic alloys, such as Ni-rich Ni-Fe alloys. Preliminary experiments indicated that MBIS does not significantly perturb the low coercivity of Ni-Fe alloys, an important attribute for prospective technical applications.

Full text of DOI:10.1149/1.2907433

Journal of The Electrochemical Society, 155 ͑7͒ D499-D507 ͑2008͒
D499
0013-4651/2008/155͑7͒/D499/9/$23.00 © The Electrochemical Society
Magnetic Materials for Three-Dimensional Damascene
Metallization: Void-Free Electrodeposition of Ni and Ni70Fe30
Using 2-Mercapto-5-benzimidazolesulfonic Acid
*
**,z
Chang Hwa Lee, John E. Bonevich, Joseph E. Davies, and Thomas P. Moffat
Material Science and Engineering Laboratory, National Institute of Standards and Technology,
Gaithersburg, Maryland 20899, USA
Superconformal filling of submicrometer trenches with electrodeposited Ni is demonstrated using a NiSO4–NiCl2–FeSO4 elec-
trolyte containing 2-mercapto-5-benzimidazolesulfonic acid ͑MBIS͒. The process offers the ability to build three-dimensional
magnetically active structures that may be easily integrated with other state-of-the-art metallization schemes. MBIS acts to inhibit
Ni͑Fe͒ electrodeposition, although under certain conditions rapid, autocatalytic breakdown accompanies the onset of Ni deposi-
tion. Optimal trench filling is associated with the positive feedback process, moderated by electrolyte internal-resistance losses,
and manifests as a hysteretic voltammetric response on planar electrodes for an MBIS concentration ϳ100 ␮mol/L. On freshly
immersed substrates, trench filling is characterized by an initial period of uniform growth followed by the development of a
v-notch geometry associated with transient depletion of MBIS within the recessed feature. Interestingly, the finest submicrometer
features are filled while only minimal deposition occurs on the neighboring free surface. Continued growth of the MBIS derived
v-notch geometry also results in void-free filling of the larger features by geometrical leveling. The same behavior is observed for
soft magnetic alloys, such as Ni-rich Ni–Fe alloys. Preliminary experiments indicated that MBIS does not significantly perturb the
low coercivity of Ni–Fe alloys, an important attribute for prospective technical applications.
© 2008 The Electrochemical Society. ͓DOI: 10.1149/1.2907433͔ All rights reserved.
Manuscript submitted February 7, 2008; revised manuscript received March 18, 2008. Available electronically May 2, 2008.
Electrodeposition of ferromagnetic materials, such as Co, Ni, Fe,
and related alloys has been widely used in the fabrication of record-
ing heads and media for hard disk drives,1-4 magnetic sensors,5,6
inductors, and actuators.7 At the same time, new applications of
ferromagnetic materials in magnetoresistive random access memory
the case of copper,15,16 silver,17 and gold,18 void-free superconfor-
mal feature filling has been demonstrated by the use of electrolyte
additives that locally modify the rate of growth leading to bottom-up
filling of recessed features, such as trenches and vias. A quantitative
understanding of feature filling in the above systems is provided by
the curvature-enhanced accelerator coverage ͑CEAC͒ model, which
details the competitive adsorption between species that inhibit and
accelerate the metal deposition rate.15-18 A key element of the
mechanism is the sharp change in additive surface coverage that
accompanies rapid changes in the electroactive area, which occur
during the filling of submicrometer features. In this case, adsorbate
evolution due to the area change effect dominates alterations asso-
ciated with adsorption from the electrolyte or consumption into the
solid. The CEAC formalism also represents an analytical description
of additive-induced “brightening” phenomena.
2,8
and even biomedical systems9 have drawn the attention of re-
searchers to the possibility of electrochemical fabrication of such
structures. Indeed, the recent emergence of the concept of magnetic
race track memories that build upon the new physics of current-
induced domain wall motion portend new opportunities for the in-
tegration of ferromagnetic materials as active devices within elec-
tronic circuitry.10,11 Furthermore, the utility of ferromagnetic
materials as contacts for semiconductors and related devices is a
subject of expanding interest that is closely coupled with the explo-
ration of new semiconductor materials, as well as electronic and
spintronic device architectures.
In the case of ferromagnetic materials, the effect of additives on
the morphological evolution and physical properties of the deposits
have been examined in some detail. The influence of dilute concen-
trations of species, such as saccharin,19-22 quinoline,23,24
thiourea,25-27 coumarin28,29 on improving film properties such as in-
ternal stress,30,31 corrosion resistance,32,33 and leveling the micro-
roughness, have been widely reported.25,29,34-36 Among them, thio-
urea and coumarin are well-known levelers that reduce the
difference in height between protruding and recessed surface fea-
tures. The leveling process is understood in terms of diffusion-
controlled adsorption and incorporation of the inhibiting additive
whereby the decreased flux to recessed surface features naturally
gives rise to a superconformal growth mode. The strength of the
leveling phenomenon is expected to be near its maximum when the
depth of the feature is slightly less than the dimensions of the hy-
drodynamic boundary layer. The understanding of this process was
developed by examining the leveling of micromachined grooves,
typically 10 ␮m deep. The filling process usually occurs under con-
ditions of diffusion limited transport and adsorption of the leveling
agents over a 10–100 ␮m wide hydrodynamic boundary layer.
On the basis of the rapid development and diverse applications of
passive and/or active submicrometer-to-nanometer scale metalliza-
tions, it is only natural to explore the prospect of superconformal
deposition of ferromagnetic materials. Recently, a demonstration of
superconformal feature filling of Ni in submicrometer trenches was
published.37 A selection of cationic surfactants, including polyethyl-
eneimine ͑PEI͒, cetyltrimethylammonium chloride, and 4-picoline,
were shown to exert significant inhibition on Ni electrodeposition at
specific ranges of concentration and overpotential.37 In particular,
Conventionally, thin-film ferromagnetic microstructures are pro-
duced by through-mask electroplating ͑also known as the LIGA͒,
which combines template production ͑e.g., lithography͒ with elec-
trodeposition whereby metal deposition proceeds on the exposed
sections of the underlying conductive seed layer. This conceptually
simple process yields microstructures that are the negative image of
the template structure.1,12 The implementation of the process is ef-
fectively two-dimensional in nature. Developing complex multilay-
ered three-dimensional ͑3D͒ structures is much more challenging,
particularly with regard to electrical addressability as required for
the electrodeposition process as well as subsequent isolation of the
final intricate structures. In this respect, the Damascene process,
which is widely used for producing multilevel Cu interconnection in
ultralarge-scale integration,13,14 offers an alternative approach to
building 3D magnetic nanostructures. A topographically patterned
dielectric is first metallized with a thin seed layer followed by metal
electrodeposition across the entire surface. A subsequent planariza-
tion step removes the overburden, leaving the desired metal struc-
tures embedded within the dielectric. The process can be repeated as
needed to produce multilevel interconnected structures that, in cur-
rent implementations, often exceed ten layers. A central challenge in
the Damascene metallization scheme is to ensure the void-free fill-
ing of recessed surface features during the electrodeposition step. In
*
Electrochemical Society Student Member.
Electrochemical Society Active Member.
**
z E-mail: thomas.moffat@nist.gov
Downloaded on 2015-04-09 to IP 130.159.70.209 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D500
Journal of The Electrochemical Society, 155 ͑7͒ D499-D507 ͑2008͒
the polyelectrolyte PEI was shown to give rise to an unusual super-
conformal growth mode whereby preferential deposition occurs at
the bottom corners of the trenches with almost negligible deposition
occurring on the neighboring free surface area, at least during the
initial stages of trench filling. Furthermore, the deposition process
was found to be highly heterogeneous at the pattern length scale
whereby dense arrays of narrower trenches would be filled followed
by lateral propagation of the growth front onto neighboring planar
areas. This growth process, although interesting and rich with pos-
sibilities, is not easily inserted into a conventional Damascene pro-
cessing scheme that demands a more uniform growth profile at the
pattern length scale of a given wafer ͑please note this is not to be
confused with the desired nonuniform superconformal growth mode
operating at the feature length scale͒. In order to further explore and
identify a Ni electroplating process that is congruent with conven-
tional Damascene processing, the ability of heterocyclic benzimida-
zole derivatives to induce void-free feature filling of submicrometer
trenches with Ni is being examined. This paper will focus on the
utility of 2-mercapto-5-benzimidazolesulfonic acid ͑MBIS͒ addi-
tions to a conventional NiSO4–NiCl2–͑FeSO4͒ Watts plating bath.
Figure 1. ͑Color online͒ CV for Ni deposition from an electrolyte comprised
of 1.0 mol/L NiSO4·6H2O, 0.2 mol/L NiCl2·6H2O, and 0.5 mol/L H3BO3
and a postexperiment iR corrected curve.
Experimental
A Watts bath consisting of 1.0 mol/L NiSO4·6H2O, 0.2 mol/L
NiCl2·6H2O, and 0.5 mol/L H3BO3, with pH 2.5, was used as the
base electrolyte. The concentration of MBIS was varied from
0 to 400 ␮mol/L. The sodium salt dihydrate of MBIS
͑C7H5N2NaO3S2·2H2O, Aldricha͒ was used. Other benzimidazole
derivatives have also been examined and will be reported on else-
where.
electrodes into the electrolyte. Both single and multiple voltammet-
ric sweep experiments were performed under quiescent conditions
using a slow scan rate of 1 mV/s, without active iR correction.
Chronoamperometric experiments were performed using both planar
and patterned wafer fragments. Specimens were immersed at a pre-
set growth potential. Field-emission scanning electron microscopy
͓͑FESEM͒, Hitachi S-4700͔ was used to examine cross sectional
profiles of the trench filling. Transmission electron microscopy
͑TEM͒ was used to more closely examine the microstructure of an
array of Ni-filled trenches. The specimen was prepared by focused
ion-beam ͑FIB͒ sectioning. X-ray diffraction ͑XRD͒ and atomic
force microscopy ͓͑AFM͒, Asylum Research MFP-3D͔ were used to
examine the crystallographic texture and the surface morphology of
the Ni films grown on planar substrates. The magnetic properties of
planar Ni and Ni–Fe films were examined by vibrating sample mag-
netometer ͑VSM͒, and the saturated magnetic moment, coercivity
and squareness were evaluated as a function of MBIS concentration.
Specimens were prepared by cleaving to obtain square-shaped
samples. The specimen area and thickness were measured by an
optical scanner and FESEM, respectively.
Both planar and patterned Cu seeded Si wafers were used as
working electrodes for electrochemical analyses and film growth
experiments. The planar specimens were prepared by physical vapor
deposition of a thin Ti adhesion layer followed by 100 nm of Cu.
The trenches including the barrier layer were 670 nm deep, and the
widths varied from 590 to 90 nm at the bottom and from
720 to 120 nm at the top, corresponding to sloping sidewall angle
that ranged from 6 to 2 deg. On the basis of the average width, the
features range from 6.4 to 1.0 in aspect ratio. In this paper, features
will be specified by the respective bottom width.
For all experiments, the working electrode was sealed with a
Teflon holder having an exposed area of 1.28 cm2, facing up to the
electrolyte in order to facilitate removal of H2 associated with para-
sitic proton reduction. The Pt counter electrode was formed as a
circular wire of similar radius to, and suspended above, the working
electrode. A saturated calomel electrode reference electrode was
firmly positioned with respect to the working and counter electrode
in a Teflon jig, thereby fixing the electrical distribution and resulting
internal-resistance ͑iR͒ effects between different experiments. Im-
pedance measurements for this configuration indicate that the Watts
electrolyte provides a resistance between the working electrode and
a reference point of ϳ7.66 ⍀ cm2. Voltammetric measurements
were performed to examine the effect of MBIS on Ni deposition
kinetics on planar Cu seeded wafer fragments. The voltammetric
scan was initiated within a fraction of a second after immersing the
Results and Discussion
Voltammetry.— Ni electrodeposition.— Ni deposition from the
additive-free Watts bath was examined by cyclic voltammetry ͑CV͒.
The onset of metal deposition is evident near −0.7 V, followed by a
rapid increase in the current that tends toward a linear response at
higher current densities as shown in Fig. 1. The linear behavior at
higher overpotentials arises from the ohmic drop associated with the
electrolyte resistance, as verified by impedance measurements. Al-
most negligible hysteresis is evident upon cycling the potential al-
though the onset potential for the negative-going scan of the first
cycle is slightly more negative compared to the latter cycles, reveal-
ing a measurable barrier exists to nucleation of Ni on the Cu surface.
A postexperiment iR-corrected voltammogram for the first cycle is
shown in Fig. 1. The correction does not account for the fact that the
actual scan rate deviates below the 1 mV/s nominal rate at higher
current densities. However, in the present case the deposition rate is
under kinetic control; thus, the distortion of the perturbing wave-
form is not expected to have a significant effect on the measured
kinetics parameters.
Effect of MBIS in Ni electrodeposition.— The addition of MBIS
to the electrolyte results in inhibition of the Ni deposition reaction
that scales with MBIS concentration, as shown in Fig. 2. On the
negative-going sweep, the onset of metal deposition is progressively
shifted toward more negative potentials by the formation of an
a The names of companies and products are included for completeness of descrip-
tion. They do not imply NIST endorsement.
Downloaded on 2015-04-09 to IP 130.159.70.209 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, 155 ͑7͒ D499-D507 ͑2008͒
D501
Figure 2. ͑Color online͒ CV for Ni deposition obtained from an electrolyte
comprised of 1.0 mol/L NiSO4·6H2O, 0.2 mol/L NiCl2·6H2O, and
0.5 mol/L H3BO3 and varying MBIS concentrations ranging from
0 to 400 ␮mol/L.
MBIS-derived blocking layer. For MBIS-concentrations of
Ͻ100 ␮mol/L, the blocking layer breaks down at negative poten-
tials and the voltammetry evolves toward that of additive-free Ni
deposition. A minor degree of residual inhibition of the deposition
reaction, relative to additive-free case, is still evident on the return
sweep, although the dominant voltammetric characteristic is the hys-
teretic or positive feedback response of the system. For MBIS con-
centrations of Ͼ150 ␮mol/L, another process is evident, because as
the initial breakdown is followed by the onset of a second inhibition
process that increases in strength during the remainder of the sweep.
The strengthening inhibition gives rise to hysteretic voltammetry
that, in this instance, corresponds to a negative feedback process.
The voltammetric response for 150 ␮mol/L MBIS delineates the
transition between the two different forms of feedback. The differ-
ence is even more sharply defined when the voltammetric curves are
corrected for the iR drop associated with the electrolyte resistance.
The iR-corrected curves are shown in Fig. 3a, and the breakdown
process for MBIS of Ͻ100 ␮mol/L is seen to occur over a very
narrower potential range characteristic of a critical phenomenon.
The correction for 150 ␮mol/L MBIS gives a negative differential
resistance ͑NDR͒ response, which also convolves the unresolved
iR-induced distortion of the perturbing triangular potential wave-
form. For MBIS concentrations of Ͼ150 ␮mol/L, the critical break-
down process is attenuated by negative feedback associated with a
secondary inhibition process.
Closer analysis of the voltammetry reveals an additional process,
namely, a 0.5 mA/cm2 current peak near −0.8 V. This is most
clearly evident in the voltammogram for 400 ␮mol/L MBIS, al-
though the same process is present for all MBIS concentrations
shown in Fig. 3b. The magnitude of the current is low but still of the
order expected for parasitic proton reduction and becomes diffusion
limited. The same process may also occur in the additive-free elec-
trolyte, but the contribution is obscured in this instance by the un-
inhibited metal deposition reaction. It is also possible that acid-base
chemistry of adsorbed MBIS or its decomposition products may also
play a role in this process. A quartz crystal microbalance study
should be helpful in sorting out the relative contributions of the
various processes and their effect on the current efficiency of the
metal deposition reaction.
Figure 3. ͑Color online͒ ͑a͒ Postexperiment iR-corrected voltammograms
for different MBIS concentrations. ͑b͒ Expanded scale reveals a current peak
associated with a second reaction related to proton reduction and/or some
aspect of MBIS chemistry.
model is available that couples mass-transport limited adsorption of
a blocking additive with current density-dependent removal of the
blocking species from the surface by its incorporation into the grow-
ing deposit.38 The deactivation process can be modeled as a linear
combination of the current density, additive coverage, and incorpo-
ration rate constant ͑or partition coefficient͒. The current density-
dependent incorporation, or perhaps deactivation process, provides
positive feedback on the metal deposition reaction and thus plays the
role of activator in the bifurcation process. The iR drop that accom-
panies deposition in the present system acts to decrease the electrode
overpotential. The decrease in driving force ͑overpotential͒ with in-
creasing growth velocity ͑current͒ provides an element of negative
feedback and thereby contributes inertia, or attenuation, to the rate
of inhibition breakdown. The two competing current-dependent ef-
fects, namely, current-induced breakdown of the blocking layer and
the decrease in overpotential due to iR effects, must be considered in
constructing a full model for the effect of additives on Ni deposition
in this potential regime. Likewise, parasitic proton and water reduc-
tion along with the associated interfacial pH shift and related effects
should also be considered.
An additional effect observed in the presence of 100 ␮mol/L
MBIS is the progressive decrease of the onset potential for proton
reduction and/or metal deposition during multicycle voltammetry, as
shown in Fig. 4. The onset potential shifts from about −0.800 to
near −0.700 V. This shift is probably related to sulfiding of the Ni
surface because a similar effect was previously noted for other
sulfur-bearing additives, such as short chain alkyl-thiols, disulfides,
and thiourea.37 These surface species might hinder the formation of
the blocking layer between MBIS and the Ni electrode.
Similar positive-feedback, hysteretic voltammetry has been pre-
viously reported for Ni deposition in the presence of cationic
N-bearing polymers ͑i.e., polyethyleneimine͒ and surfactant mol-
ecules ͑i.e., dodecyltetraalkylammonium salts͒.37 As noted therein,
such bifurcation processes have a significant history in electrochem-
istry. More specifically, in the case of additive plating, a steady-state
Downloaded on 2015-04-09 to IP 130.159.70.209 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D502
Journal of The Electrochemical Society, 155 ͑7͒ D499-D507 ͑2008͒
Figure 4. ͑Color online͒ Multicycle ͑three cycles͒ voltammetry for Ni depo-
sition in 1.0 mol/L NiSO4·6H2O, 0.2 mol/L NiCl2·6H2O, 0.5 mol/L H3BO3,
and 100 ␮mol/L MBIS.
Figure 6. ͑Color online͒ Chronoamperometry of Ni deposition on a trench
patterned Cu electrode for various concentrations of MBIS at a fixed poten-
tial of −0.925 V.
Chronoamperometry.— The influence of MBIS on the kinetics of
Ni deposition during deposition at fixed potential was examined.
Measurements were performed on both planar and trench-patterned
substrates as a function of MBIS concentration and potential. Tran-
sients for Ni deposition on planar Cu surfaces immersed under po-
tential control at −0.925 V are summarized in Fig. 5 for a range of
MBIS concentration. The transient exhibits a systematic increase in
inhibition with MBIS concentration. The initial decay of the tran-
sient in the presence of MBIS includes double-layer charging and
accumulation of MBIS. A subsequent increase in the growth rate is
associated with the breakdown of adsorbed MBIS on the surface.
The onset of breakdown measured after the initial rise of the tran-
sient corresponds to about 6, 18, 24, and 120 s for 50, 80, 100, and
150 ␮mol/L MBIS, respectively. Although the time it takes to reach
the steady-state currents is dependent on the MBIS concentration,
the corresponding values are congruent with the extracted values
from voltammetric experiments ͑i.e., positive-going scans in Fig. 2͒.
A detailed evaluation of the chronoamperometric data is challenging
due to variation in the supersaturation that accompany changes in
the current as a result of iR electrolyte losses. Possible changes in
the current efficiency provides an additional complication.
ments. The transients for films grown at −0.925 V as a function of
MBIS concentration are shown in Fig. 6. For the MBIS-free elec-
trolyte, a current peak is apparent at 25 s, followed by steady-state
current of 6 mA/cm2 for times beyond 100 s. The initial increase in
current, relative to a planar surface, must be partly due to nucleation
and growth on the greater surface area of the patterned substrates,
while by 100 s the trenches are nominally filled and a planar growth
front remains. A similar situation is evident for 50 ␮mol/L MBIS
case. In this instance, the growth potential of −0.925 V lies beyond
the hysteretic voltammetric regime noted earlier. In contrast, for
larger MBIS concentrations, the deposition transients lie within the
NDR regime and the current transients exhibit distinctly different
behavior. For MBIS concentrations between 80 and 150 ␮mol/L,
the transients exhibit similar shape. The initial current decay is fol-
lowed by an acceleration of inhibition that begins around 25 s. A
minimum is subsequently observed at 50 s, and then the current
begins to increase to a local maximum followed by a slight decay
that converges to values consistent with measurements on planar
electrodes. Interpretation of the transients is challenging in light of
the significant changes in electrode area that accompany trench fill-
ing combined with the distributed nature of the feature size across
the wafer fragment. The overall trend of inhibition is clearly evident
although the minima observed at 50 s must be coupled with some
aspect of feature filling. At longer times, beyond 200 s, the tran-
sients for the patterned substrate converge with the current values
observed on planar substrates, indicating most of the trenches have
been filled and the original patterned substrates are now effectively
planar. Similar trends were observed for transients collected at dif-
ferent potentials. The results for a fixed MBIS concentration of
100 ␮mol/L are summarized in Fig. 7. A further discussion of the
transients for patterned specimens will accompany the examination
of the corresponding feature filling experiments.
Chronoamperometric experiments were also performed on pat-
terned substrates to allow for correlation with feature filling experi-
Feature filling in the presence of MBIS.— A series of feature
filling experiments were performed as a function of MBIS concen-
tration. Ni was deposited at −0.925 V for 3 min, and then the speci-
mens were cross sectioned for FESEM analysis. Images of trench
arrays with four different dimensions, 520, 260, 150, and 90 nm
wide bottom widths, are presented as representative examples of the
filling behavior. The dark contrast at the bottom of the trench and on
the free surface between the trenches is associated with the copper
seed layer. In the absence of MBIS, a large seam is apparent in the
narrowest trenches while voids are evident at the top of all but the
widest features shown in Fig. 8a. The free surface is also notably
rough relative to the dimension of the trenches and overburden.
Figure 5. ͑Color online͒ Chronoamperometry of Ni deposition on a planar
Cu electrode for various concentrations of MBIS at a fixed potential of
−0.925 V.
Downloaded on 2015-04-09 to IP 130.159.70.209 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, 155 ͑7͒ D499-D507 ͑2008͒
D503
Figure 7. ͑Color online͒ Chronoamperometry of Ni deposition on a trench
patterned Cu electrode at various potentials ranging from −0.850 to
− 0.975 V. The MBIS concentration was 100 ␮mol/L.
Figure 9. Time evolution of Ni trench filling in the presence of 100 ␮M
MBIS. Four different trench widths are shown: ͑a͒ 590, ͑b͒ 520, ͑c͒ 305, ͑d͒
160, and ͑e͒ 130 nm.
Thus, conformal deposition coupled with high surface roughness
leads to the formation of voids when the rough opposing side walls
impinge.
MBIS experiments, performed for the same duration and potential,
reveal progressively less overburden above the filled features, which
reflects the inhibition provided by MBIS. This effect is most notable
over the widest features. A further increase of MBIS concentration
to 150 ␮mol/L leads to a sharp decrease in the overall deposition
rate. Very limited deposition has occurred on the free surface while
substantial filling of the trenches has occurred. Indeed, the four
smallest trenches are completely filled while the growth front in the
widest features has a well-defined v-notch shape. The latter provides
direct evidence of a superconformal film growth mode; whereby the
effect of a linear gradient of the sidewall growth velocity with trench
depth is evident and deposition at the bottom of the trench is sub-
stantially progressed compared to the neighboring free surface.
In order to gain a more detailed view of shape evolution during
feature filling, a series of specimens were examined as a function of
deposition time at −0.925 V in the 100 ␮mol/L MBIS electrolyte.
The results for five different trench widths are shown in Fig. 9. For
reference purposes, the Cu-seeded substrate ͑t = 0͒ is shown prior to
Ni deposition. The buildup of Cu on the top and bottom surfaces,
relative to the sidewalls, reflects the nature and limitations of the
physical vapor deposition or sputtering process used to form the
seed layer. A representative current transient corresponding to 300 s
of growth on the patterned substrates is shown in Fig. 10. Examina-
tion of the filling images after 25 s of deposition reveal an essen-
tially conformal Ni layer along the entire surface profile, where the
height of the bottom surface and the sidewall thickness are almost
the same. The extent of Ni deposition in the recessed regions can be
compared to the integrated deposition charge derived from the chro-
noamperometry transient. Because the surface area changes signifi-
cantly during feature filling, a comparison of nominal thickness
value is only considered for the first 25 s and the last 100 s from
200 to 300 s, assuming that there is minor area change occurring for
the respective cases. The film thicknesses calculated from the tran-
sients, based on 100% current efficient metal deposition, are in rea-
sonable agreement with those measured from scanning electron mi-
croscope ͑SEM͒ images, as summarized in Table I.
Deposition in the presence of 50 ␮mol/L MBIS results in clearly
smoother overburden above the widest trench; however, close ex-
amination reveals centerline voids near the top of the smallest
trenches. Increasing the MBIS concentration to 80 ␮mol/L yields a
smoother free surface; however, voids are still evident at the top of
finest trenches although they are smaller than for the 50 ␮mol/L
MBIS case. A further increase in the MBIS concentration to
100 ␮mol/L leads to void-free filling for all features examined.
Comparison between the additive-free, 50, 80, and 100 ␮mol/L
Between 25 and 50 s, the two finest features are filled while
remarkably almost negligible Ni deposition has occurred on the free
surface. The geometry of the feature filling is such that electrode
area change effects will be felt most strongly during the filling of the
smallest and highest aspect ratio features. In the three wider
trenches, a gradient in the deposition rate on the sidewalls has
clearly developed as a function of depth similar to that noted earlier
in Fig. 8e. The shape transition to sloping sidewalls is a common
characteristic of traditional steady-state inhibitor-based leveling
Figure 8. Electrodeposition of Ni in trenches with various bottom widths
from 520 to 90 nm as a function of MBIS concentration ͑a͒ 0, ͑b͒ 50, ͑c͒ 80,
͑d͒ 100, ͑e͒ 150 ␮mol/L MBIS. Deposition was performed at −925 mV for
300 s.
Downloaded on 2015-04-09 to IP 130.159.70.209 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D504
Journal of The Electrochemical Society, 155 ͑7͒ D499-D507 ͑2008͒
Figure 11. Cross sections of an array of trenches filled with Ni electrode-
posited from an electrolyte consisting of ͑a͒ 10 ␮mol/L PEI at −850 mV for
600 s, and ͑b͒ 100 ␮mol/L MBIS at −925 mV for 300 s. An unusual later-
ally heterogeneous deposition process is induced by PEI in contrast to the
more uniform characteristics associated with MBIS.
of sloping sidewalls, and the subsequent onset of impingement of
opposing sidewalls. The remaining cusp or v-notch shape is then
filled by geometric leveling. The key role of MBIS appears to be the
development of sloping sidewalls, which is most likely associated
with transient depletion of the MBIS flux within the trench. The
process is conceptually analogous to traditional leveling, although in
the present case the effect is transient and associated with the system
following a trajectory across the NDR regime. A limited degree of
stabilization of the evolving system is provided by decreases in the
overpotential that accompany any increase in the current, due to the
finite electrolyte resistance ͑iR͒. This coupling provides a paradigm
that is derived, yet distinct from traditional steady-state leveling
models. A more detailed view of this coupling will be presented in
another paper.
Similar trench-filling dynamics were recently reported in a study
of the effect of cationic PEI on Ni deposition.37 Although the dy-
namics at the trench length scale were similar, surprising heteroge-
neities were noted at larger dimensions. Namely, a significant com-
ponent of lateral propagation of the growth front across trench
arrays was observed and found to be closely correlated with the
pattern density, as shown in Fig. 11. Note there is negligible depo-
sition in the open field on the left-hand side, while the array of
narrow trenches is almost completely filled. In contrast, deposition
in the presence of MBIS is essentially uniform on the pattern density
length scale, making the MBIS process a more viable manufacturing
solution in the context of conventional Damascene processing.
The potential dependence of trench filling was also examined for
a fixed MBIS concentration of 100 ␮mol/L and a fixed deposition
time of 300 s. At −0.850 V, the deposition overpotential is so low
that the transition to the activated surface and the MBIS passivated
one is effectively negligible, as indicated in the voltammogram
shown in Fig. 4. Deposition on the trench pattern, albeit limited,
proceeds conformally, and the roughness that develops on the side-
walls of the finest feature leads to incipient occlusion of voids along
the centerline of the finest features, as shown in Fig. 12. Increasing
the overpotential by 0.025 V to − 0.875 V places the system within
the positive feedback hysteretic voltammetric regime. Void-free fill-
ing of the finest features is evident along with the development of
sloping sidewalls in the wider trenches, although this is somewhat
less well developed in the largest features. When the overpotential is
increased further by 0.025 V, sloping sidewalls are readily evident
in the widest trench while the remaining trenches are filled and
voidfree. For growth at −0.925 and −0.950 V, all features are void-
free and the overburden thickness increased sharply as expected.
Figure 10. ͑Color online͒ A representative chronoamperometric transient of
Ni electrodeposition on the patterned trench arrays shown in Fig. 9. The
highlighted points correspond to the filling time indicated on the respective
images.
models,39-41 whereby depletion of the inhibitor flux into recessed
features leads to enhanced deposition therein. Further examination
reveals that the amount of Ni deposited on the bottom surface is
identical to that on the sidewalls immediately adjacent to the bot-
tom. Notably, the amount of Ni deposited on the sidewalls of the
larger features exceeds half of the width of the two narrowest
trenches consistent with those features having already been filled by
sidewall collision, followed by a geometric zipping process. The
combination of decrease in electroactive area ͑i.e., the filling of the
finest features͒ and MBIS accumulation accounts for the accelerated
decay of the current transient ͑Fig. 10͒ observed between 25 and
50 s. The transient depletion of MBIS within the trench that results
in the sloping sidewalls or v-notch shape is critical to the overall
void-free filling process.
Between 50 and 100 s growth continues on the bottom and side-
walls of the 305 nm wide trench. Interestingly, the slope associated
with sidewalls does not change substantially between 50 and 100 s.
This indicates a transition back to a conformal growth mode as
reflected by the constant sidewall growth velocity during this time
increment. Evolution of the growth front in the widest trench also
supports this notion. In the absence of any compositional gradients,
continued growth on such a v-notched surface geometry will result
in a planar surface by an effect known as geometrical leveling.40-42
Indeed, following trench filling, significant growth begins on the
adjacent free surface and the remaining cusp shape in the larger
features is filled congruent with geometric leveling.
Summarizing, trench filling is associated with three stages, a
brief period of near conformal growth, followed by the development
Table I. Comparison between film thickness measured by SEM
and the thickness calculated from the amount of deposition
charge at 100 ␮M MBIS. (Although the surface area of patterned
substrate, which is 25% higher than the planar substrate, was
assumed to be a constant during the first 25 s, that of patterned
substrate was assumed to be the same as the planer substrate
after 200 s deposition, where the features are completely filled.)
TEM.— FESEM imaging indicates that the addition of MBIS
enables void-free trench filling. The distinctive v-notch growth con-
tours associated with deposition in the recessed features raise inter-
esting questions concerning possible correlations with the resulting
microstructure. An array of 160 nm wide ͑bottom width͒ Ni-filled
trenches was examined using TEM, and a bright-field image of the
microstructure is shown in Fig. 13a. As noted earlier, the seed-layer
0–25 s
200–300 s
Thickness calculated from Fig. 6
Total charge ͑mA s͒
Thickness ͑nm͒
142.0
31.2
430.6
126.13
Thickness measured from SEM
Thickness ͑nm͒
36.7 Ϯ 5.2
144.6 Ϯ 51.9
Downloaded on 2015-04-09 to IP 130.159.70.209 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, 155 ͑7͒ D499-D507 ͑2008͒
D505
Figure 13. ͑a͒ FIB-TEM cross section of 160 nm ͑excluding the barrier͒
width trench filled by Ni electrodeposition in the presence of 100 ␮M MBIS.
Deposition was performed at −0.925 V for 150 s. ͑b͒ Secondary electron
contrast from a FESEM cross section of a related specimen reveals the
buildup of the Cu seed layer on the free surface adjacent to the trench.
Comparison with the TEM image indicates that very limited Ni deposition
has occurred in this region compared to the trench.
Figure 12. The influence of deposition potential on trench filling in
100 ␮mol/L MBIS. Deposition was performed for 300 s at various poten-
tials: ͑a͒ −850, ͑b͒ −875, ͑c͒ −900, ͑d͒ −925, and ͑e͒ −950 mV.
deposition process results in significantly thicker Cu layer on the
free surface and trench bottom, vs the trench sidewalls, as shown in
Fig. 9 ͑t = 0͒. Indeed, many of the FESEM images shown in Fig. 8,
9, and 12 have enough secondary electron contrast to allow the
sputtered Cu seed layer to be distinguished from the electrodepos-
ited Ni. For reference purposes, a magnified FESEM image is shown
in Fig. 13b. Comparison to the TEM image reveals that the volume
of materials associated with the free surface between the trenches is
mostly Cu, which is only covered with a thin 25 nm thick Ni layer.
In contrast, the 160 nm wide trench is entirely filled with Ni and free
of any obvious voids. As noted in the FESEM images shown in Fig.
8 and 9, the inhibition evident on the free surface as compared to the
trench is quite remarkable given the submicron scale of the feature
and the 150 s long Ni deposition time. Further inspection of the
trenches reveals evidence of a microstructural seam along the top
two-thirds of the centerline, which is consistent with expectations
from the study of time evolution shown in Fig. 9. This consolidated
boundary forms during deposition in a manner that is analogous to
grain boundary generation associated with coalescence during
Volmer–Weber film growth. Bright-field contrast suggests a grain
size on the order of 30–40 nm, although certain grains are clearly
larger with well-defined twins or stacking faults evident.
tion of 100 ␮mol/L MBIS leads to further suppression of the alloy
deposition rate, which is comparable to the MBIS-induced shift ob-
served for pure Ni. Energy dispersive X-ray analysis of the planar
film deposited at −0.950 V indicates the alloy composition is
Ni70Fe30. Cross-sectional FESEM images of trenches that were
filled by deposition at −0.950 V for 300 s are shown in Fig. 15.
Void-free feature filling is clearly evident while the overburden is
substantially less than Ni films grown under the same condition ͑see
Fig. 12e͒. The latter demonstrates additional inhibition is provided
Ni–Fe alloys.— One of the most interesting aspects of iron
group metals, and alloys thereof, are the soft magnetic properties. A
prototypical example is Permalloy, a Ni80Fe20 alloy that has a rich
history in the development of magnetic storage devices. The elec-
trodeposition of the Ni–Fe system is often classified as anomalous
due to the observation of the enhanced deposition of Fe relative to
Ni that represents an inversion of the thermodynamic tendency of
the respective elemental forms. Herein, the extension and efficacy of
MBIS for inducing superconformal deposition of Ni–Fe alloys is
examined. The effect of adding 0.05 mol/L FeSO4 to the Ni Watts
bath on the deposition kinetics is shown in Fig. 14, where the iron
partial reaction leads to inhibition of the deposition rate at a given
potential relative to the deposition of pure nickel. Subsequent addi-
Figure 14. ͑Color online͒ Voltammetry reveals the effect of adding
0.05 mol/L FeSO4 to the Ni plating bath ͑1.0 mol/L NiSO4·6H2O,
0.2 mol/L NiCl2·6H2O, and 0.5 mol/L H3BO3͒. MBIS provides additional
inhibition.
Downloaded on 2015-04-09 to IP 130.159.70.209 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D506
Journal of The Electrochemical Society, 155 ͑7͒ D499-D507 ͑2008͒
Figure 15. Cross-sectional FESEM images of trenches filled with the elec-
trodeposited Ni70Fe30 alloy. Deposition was performed at −0.950 V for
300 s.
by MBIS during Fe deposition. Further studies are underway to
demonstrate the compositional homogeneity and generality of this
effect for cobalt and related binary and ternary iron group alloys.
Figure 16. ͑Color online͒ Symmetric X-ray diffraction results for Ni films
electrodeposited on planar Cu/Ti/SiO2/Si substrates as a function of MBIS
concentration. Film were grown for 300 s at −0.925 V. Specimen film thick-
ness and the grain size derived from the coherence length are summarized in
Table II. The black curve ͑Cu seed only͒ corresponds to the x axis while the
others are progressively offset for clarity.
XRD and AFM studies.— A brief study of the effect of MBIS on
the texture and scattering coherence of Ni films was initiated. The
thickness of the films was measured by examining cleaved cross
sections of the electrodeposited films grown on the Cu/Ti/SiO2/Si
substrates ͑Table II͒. The results of symmetric scattering are sum-
marized in Fig. 16. For the planar Cu seed layer, only a ͑111͒ peak
is visible while ͑200͒ is absent, reflecting a ͑111͒ texturing. The
seed-layer texture is apparently sustained by the electrodeposited
films. A Scherrer’s equation assessment of the coherence length sug-
gests the grain size is in the range of 41–34 nm, decreasing with
increasing MBIS concentration. Because the films varied from
150 to 210 nm in thickness, a given cross section might have on the
order of 4–6 grains. This assessment agrees favorably with the TEM
result. The root-mean-square roughness ͑Rs͒ of the films was also
examined over a 2 ␮m2 area by AFM, and the results are summa-
rized along with the X-ray data in Table II. The films become
smoother with increasing MBIS concentration, with the most no-
table improvement occurring between 100 and 150 ␮mol/L MBIS.
Conclusion
Superconformal electrodeposition of Ni and Ni–Fe alloys has
been demonstrated using a single benzimidazole derivative, MBIS,
as an additive to the respective sulfate/chloride mixed electrolyte.
The process offers a rich variety of opportunities for integration of
the ferromagnetic materials into Damascene processes, microelec-
tromechanical systems and related thin-film derived technologies.
Voltammetry and chronoamperometry reveal that inhibition of the
deposition reaction scales with MBIS concentration. For concentra-
tions below 150 ␮mol/L MBIS, a positive feedback mechanism as-
sociated with deactivation and consumption of the adsorbed mol-
ecule within the growing solid gives rise to hysteretic voltammetry.
When corrected for resistive electrolyte losses, a critical potential is
evident that may be linked with a domain of NDR. Feature filling at
potentials within the voltammetrically identified critical regime re-
sult in void-free superconformal film growth. For freshly immersed
specimens, trench filling initially proceeds with a period of confor-
mal growth followed by the development of sloping sidewalls asso-
ciated with transient depletion of MBIS within the recessed features.
The resulting v-notch growth front then evolves in a manner that is
difficult to distinguish from geometric leveling. Stabilization of the
feature filling process within the voltammetrically identified critical
domain is provided by a decrease in the overpotential for metal
deposition that accompanies the flow of current in the resistive elec-
trolyte.
Magnetic properties.— A VSM was used to briefly survey the
effect of MBIS on the magnetic properties of Ni and Ni70Fe30 films.
The normalized M–H curves are shown in Fig. 17a and b, respec-
tively. The addition of MBIS is seen to result in a slight but pro-
gressive increase in coercivity ͑Hc͒ of Ni, while little effect on the
saturated magnetization ͑Ms͒ is observed, as summarized in Table
II. A more noticeable loss in squareness is evident in the M–H loops.
In the case of the Ni70Fe30 alloy, the coercivity decreased to
0.60 mT and saturated magnetization increased to 867 Ϯ 4
͑ϫ103 A/m͒. However, 100 ␮mol/L MBIS had a negligible effect
on the coercivity while the squareness of the M–H loop was some-
what degraded.
Table II. Ni film properties deposited in the presence of various MBIS concentrations. The deviation in Ms values represents extrema of
variation in measured volume.
Grain size
MBIS
concentration
͑␮M͒
Full width at
half-maximum
͑deg͒
estimated by
Scherrer Equation
͑nm͒
Thickness
͑nm͒
Rs
͑nm͒
Rs/Thickness
Msa
␮0Hc
͑mT͒
͑%͒
͑ϫ103 A/m͒
0
50
80
100
150
163
177
196
210
180
10.8
8.6
9.5
9.8
6.7
6.6
4.8
4.8
4.6
3.7
0.207
0.227
0.244
0.241
0.248
41.3
37.8
35.2
35.6
34.6
454 Ϯ 5
508 Ϯ 10
443 Ϯ 18
520 Ϯ 103
515 Ϯ 57
11.3
11.6
12.4
12.5
14.8
a Ms of bulk Ni is 485 ϫ 103 A/m.
Downloaded on 2015-04-09 to IP 130.159.70.209 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, 155 ͑7͒ D499-D507 ͑2008͒
References
D507
1. L. T. Romankiw, Electrochim. Acta, 42, 2985 ͑1997͒, and references therein.
2. C. A. Ross, Annu. Rev. Mater. Res., 31, 203 ͑2001͒, and references therein.
3. E. I. Cooper, C. Bongôte, 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.
4. T. Osaka, Electrochim. Acta, 45, 3311 ͑2000͒.
5. J. E. Lenz, Proc. IEEE, 78, 973 ͑1990͒.
6. R. S. Popovic, J. A. Flanagan, and P. A. Besse, Sens. Actuators, A, 56, 39 ͑1996͒.
7. J. W. Judy and R. S. Muller, Sens. Actuators, A, 53, 392 ͑1996͒.
8. S. S. P. Parkin, K. P. Roche, M. G. Samant, P. M. Rice, R. B. Beyers, R. E.
Scheuerlein, E. J. O’Sullivan, S. L. Brown, J. Bucchigano, D. W. Abraham, et al.,
J. Appl. Phys., 85, 5828 ͑1999͒.
9. C. C. Berry, J. Mater. Chem., 15, 543 ͑2005͒.
10. S. A. Wolf, A. Y. Chtchelkanova, and D. M. Treger, IBM J. Res. Dev., 50, 101
͑2006͒.
11. S. S. P. Parkin, M. Hayashi, and L. Thomas, Science, 320, 190 ͑2008͒.
12. M. Baghbanan, U. Erb, and G. Palumbo, Phys. Status Solidi A, 203, 1259 ͑2006͒.
13. R. Rosenberg, D. C. Edelstein, C.-K. Hu, and K. P. Rodbell, Annu. Rev. Mater. Sci.,
30, 229 ͑2000͒.
14. S. P. Murarka, Mater. Sci. Eng., R., 19, 87 ͑1997͒.
15. T. P. Moffat, D. Wheeler, M. Edelstein, and D. Josell, IBM J. Res. Dev., 49, 19
͑2005͒.
16. T. P. Moffat, D. Wheeler, S.-K. Kim, and D. Josell, J. Electrochem. Soc., 153,
C127 ͑2006͒.
17. B. Baker, C. Witt, D. Wheeler, D. Josell, and T. P. Moffat, Electrochem. Solid-State
Lett., 6, C67 ͑2003͒.
18. D. Josell, D. Wheeler, and T. P. Moffat, J. Electrochem. Soc., 153, C11 ͑2006͒.
19. C. C. Roth and H. Leidheiser, J. Electrochem. Soc., 100, 553 ͑1953͒.
20. T. Osaka, T. Sawaguchi, F. Mizutani, T. Yokoshima, M. Takai, and Y. Okinaka, J.
Electrochem. Soc., 146, 3295 ͑1999͒.
21. S. R. Brankovic, R. Haislmaier, and N. Vasiljevic, Electrochem. Solid-State Lett.,
10, D67 ͑2007͒.
22. H. Kwon and A. A. Gewirth, J. Electrochem. Soc., 154, D577 ͑2007͒.
23. J. Edward and M. J. Levett, Trans. Inst. Met. Finish., 44, 27 ͑1966͒.
24. W. R. Doty and B. J. Riley, J. Electrochem. Soc., 114, 50 ͑1967͒.
25. G. T. Rogers, M. J. Ware, and R. V. Fellows, J. Electrochem. Soc., 107, 677
͑1960͒.
26. R. F. Renner and K. C. Liddell, J. Appl. Electrochem., 32, 621 ͑2002͒.
27. U. S. Mohanty, B. C. Tripathy, S. C. Das, and V. N. Misra, Metall. Mater. Trans. B,
36B, 737 ͑2005͒.
28. J. Edwards and M. J. Levett, Trans. Inst. Met. Finish., 41, 157 ͑1964͒.
29. S. S. Kruglikov, N. T. Kudryavtsev, and R. P. Sobolev, Electrochim. Acta, 12,
1263 ͑1967͒.
30. A. Bodnevas and J. Zahavi, Plat. Surf. Finish., 81, 75 ͑1994͒.
31. I. Tabakovic, V. Inturi, and S. Riemer, J. Electrochem. Soc., 149, C18 ͑2002͒.
32. T. Osaka, M. Takai, Y. Sogawa, T. Momma, K. Ohashi, M. Satto, and K. Yamada,
J. Electrochem. Soc., 146, 2092 ͑1999͒.
Figure 17. ͑Color online͒ Normalized M–H loops for ͑a͒ Ni and ͑b͒
Ni70Fe30. In the case of Ni MBIS leads to a decrease in squareness and
increase in coercivity while for Ni70Fe30 a negligible change in coercivity
͑from 0.60 to 0.58 mT͒ is observed, although the squareness is somewhat
diminished.
33. M. Saito, K. Yamada, K. Ohashi, Y. Yasue, Y. Sogawa, and T. Osaka, J. Electro-
chem. Soc., 146, 2845 ͑1999͒.
34. D. G. Foulke and O. Kardos, Proc. Am. Electroplat. Soc., 43, 172 ͑1956͒.
35. S. A. Watson and J. Edwards, Trans. Inst. Met. Finish., 34, 167 ͑1957͒.
36. A. T. Vagramyan and N. K. Baraboshkina, Plating, 54, 930 ͑1967͒.
37. S.-K. Kim, J. E. Bonevich, D. Josell, and T. P. Moffat, J. Electrochem. Soc., 154,
D443 ͑2007͒.
38. D. Roha and U. Landau, J. Electrochem. Soc., 137, 824 ͑1990͒.
39. R. Chalupa, Y. Cao, and A. C. West, J. Appl. Electrochem., 32, 135 ͑2002͒.
40. C. Madore, M. Matlosz, and D. Landolt, J. Electrochem. Soc., 143, 3927 ͑1996͒.
41. C. Madore and D. Landolt, J. Electrochem. Soc., 143, 3936 ͑1996͒.
42. O. Kardos and D. Gardner Foulke, in Advances in Electrochemistry and Electro-
chemical Engineering, Vol. 2, p. 145–233, P. Delahay and C. W. Tobias, Editors,
John Wiley & Sons, Hoboken, NJ ͑1962͒.
Acknowledgments
The authors thank Dr. Jonathan J. Mallet, Dr. Mark D. Vaudin,
and Dr. Koo-Hyun Chung for help with the XRD and AFM experi-
ments, and Sandra W. Claggett for assistance in preparation of speci-
mens for FESEM. We are also grateful to Professor Jae Jeong Kim,
at Seoul National University, for helpful discussions.
National Institute of Standards and Technology assisted in meeting the
publication costs of this article.
Downloaded on 2015-04-09 to IP 130.159.70.209 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).