Via filling by electrodeposition superconformal silver and copper and conformal nickel

Article abstract of DOI:10.1149/1.1517583

Superconformal deposition of silver in vias was studied. The observed experimental fill behavior is compared with predictions from a model based on the curvature-enhanced accelerator coverage mechanism of superconformal deposition. Superconformal copper deposition and conformal nickel deposition results are also modeled. The previously published model predicts via filling behavior using the dependence of deposition rate kinetics on the coverage of adsorbed catalyst. The requisite kinetic parameters are obtained from independent current-voltage and current-time transient studies conducted on planar substrates.

Full text of DOI:10.1149/1.1517583

Journal of The Electrochemical Society, 149 ͑12͒ C637-C641 ͑2002͒
C637
0013-4651/2002/149͑12͒/C637/5/$7.00 © The Electrochemical Society, Inc.
Via Filling by Electrodeposition
Superconformal Silver and Copper and Conformal Nickel
b
a
a,
D. Josell,a,z B. Baker,a, C. Witt, D. Wheeler, and T. P. Moffat
*
*
aNational Institute of Standards and Technology, Metallurgy Division, Gaithersburg, Maryland 20899, USA
bInternational Sematech, Austin, Texas, USA
Superconformal deposition of silver in vias was studied. The observed experimental fill behavior is compared with predictions
from a model based on the curvature-enhanced accelerator coverage mechanism of superconformal deposition. Superconformal
copper deposition and conformal nickel deposition results are also modeled. The previously published model predicts via filling
behavior using the dependence of deposition rate kinetics on the coverage of adsorbed catalyst. The requisite kinetic parameters
are obtained from independent current-voltage and current-time transient studies conducted on planar substrates.
© 2002 The Electrochemical Society. ͓DOI: 10.1149/1.1517583͔ All rights reserved.
Manuscript submitted January 30, 2002; revised manuscript received June 5, 2002. Available electronically October 24, 2002.
The use of electrodeposited copper in integrated circuits is due,
in large part, to the ability of the deposition process to fill high-
aspect-ratio features superconformally. This superconformal filling
process was originally modeled, with only limited success, using
traditional leveling-type theories involving depleted concentrations
of deposition rate-inhibiting additives down the filling features.1 The
more recent curvature-enhanced accelerator coverage ͑CEAC͒
model has focused on competitive adsorption of deposition rate-
inhibiting and catalyzing additives and conservation of the more
strongly adsorbed catalyst during area change associated with metal
deposition.2-8 These models have predicted the initial ‘‘incubation’’
period of conformal deposition, the superconformal bottom-to-top
filling itself, and the subsequent ‘‘momentum’’ plating bump forma-
tion over filled features that are commonly observed but which
could not be explained by the leveling models.9-11 Furthermore,
these predictions were made with no fitting parameters; all kinetic
factors were obtained from current-voltage studies with planar sub-
strates in electrolytes containing a range of catalyst concentrations.
An alternative model based on accumulation of catalyst at the bot-
toms of features has been published;12 however, it required tuning of
parameters to fit experimental fill results in order to model filling in
a given electrolyte composition.
To date, papers that utilize CEAC-based models have detailed
the agreement between predicted and experimental filling during
copper2-5 and silver6 electrodeposition in trenches only. Supercon-
formal deposition during iodine-catalyzed chemical vapor deposi-
tion of copper in vias7 has been predicted using a CEAC-based
model that describes superconformal filling of vias.8 However, the
work presented here is the first quantitative comparison of CEAC-
based predictions and experiment for superconformal filling for the
via geometry.
blocks were separated by an unpatterned region less than 10 ␮m
wide. The pitch ranged from approximately 0.5 to 4 ␮m. The vias
were patterned in 1 ␮m thick dielectric with a tantalum barrier layer
and 100 nm thick sputter-deposited copper seed layer ͑thickness on
the field outside the vias͒. The wafer was sectioned, in accordance
with the periodic pattern, to obtain test specimens that were approxi-
mately 20 ϫ 20 mm. Electrodeposition experiments at constant
overpotential were conducted using these specimens in one of three
electrolytes in order to obtain superconformal copper, superconfor-
mal silver, or conformal nickel deposits. The electrolyte composi-
tions are detailed in Table I. Specimens were immersed in the elec-
trolyte at the deposition voltage in order to avoid damage to the seed
layer on the sidewalls of the vias. Standard three-electrode cells
were used for all filling experiments. For the copper depositions, a
copper anode and a saturated calumel reference electrode ͑SCE͒
were used. For the silver depositions, a platinum anode and a silver
reference electrode were used. For the nickel depositions, a nickel
anode was used ͑SCE͒. In all cases, the specimen was the cathode.
Specimens were inserted, and held during deposition, in a verti-
cal position such that buoyancy-driven convection arising from the
concentration gradient near the surface would be expected to induce
formation of a boundary layer during deposition. The specimens
were removed from the electrolyte after predetermined deposition
times, rinsed in deionized water, and dried. They were then cross-
sectioned for viewing by scanning electron microscope using stan-
dard techniques.
Filling vias.—Figure 1 shows a series of cross-sectioned vias,
obtained by scanning electron microscope, that detail the filling pro-
cess during copper electrodeposition in the largest features studied.
The images, reflecting the fill profiles obtained at successive 10 s
intervals, were obtained by interrupting electrodeposition on each
specimen after the indicated deposition time. The copper deposition
is conformal during the first 60 s. Deposition on the bottom surface
accelerates at ϳ70 s and superconformal bottom-to-top filling oc-
curs. Figure 2 shows a similar set of images obtained during silver
electrodeposition. The silver deposition is also initially conformal,
with rapid bottom-to-top filling starting at ϳ40 s. A series of images
obtained during the period when the rapid bottom-to-top filling oc-
curs is shown in Fig. 3; the significant difference in the deposition
͑growth͒ rates on the bottom and sidewall surfaces that characterizes
the superconformal deposition process is clear. The time dependence
of the feature filling obtained through conformal nickel deposition in
the same patterned vias is detailed in Fig. 4. Growth is conformal
from start to finish, resulting in a v-notch surface profile whose
internal angle reflects the tilt of the original sidewalls.
The model developed in Ref. 8 is used to describe superconfor-
mal filling of vias during both silver and copper electrodeposition in
superfilling electrolytes. Because the vias being filled, with side-
walls canted by ϳ5°, are far from the idealized cylindrical structure
envisioned in the model, nickel is also deposited conformally for
comparison.
Experimental
Specimen fabrication.—A wafer, patterned with a repeating test
structure, and including a copper seed, was provided by Interna-
tional Sematech. The test structure included rectangular blocks of
vias, each containing vias of a single diameter between approxi-
mately 0.5 and 0.15 ␮m. The blocks were more than 3 mm long but
ranged from 240 ␮m down to less than 10 ␮m wide, depending on
the via diameter and spacing ͑pitch͒ within the block. Adjacent
Quantifying the deposition.—The time-dependent thicknesses of
the metal deposits on the via base, obtained from cross-sectioned
specimens including those shown in Fig. 1-3, are summarized in
Fig. 5 for the copper and Fig. 6 for the silver deposits. In both cases
* Electrochemical Society Active Member.
z E-mail: daniel.josell@nist.gov
Downloaded on 2015-05-09 to IP 61.129.42.30 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C638
Journal of The Electrochemical Society, 149 ͑12͒ C637-C641 ͑2002͒
Table I. Constituent concentrations of the silver,6 copper,5 and nickel electrolytes.a
Metal
Electrolyte compositions
0.34 mol/L KAg(CN)2 ϩ 2.3 mol/L KCN ϩ proprietarya
Silver
Copper
0.25 mol/L CuSO4"5H2O ϩ 1.8 mol/L H2SO4 ϩ 1 ϫ 10Ϫ3 mol/L NaCl
ϩ 88.2 ϫ 10Ϫ6 mol/LH (OCH2CH2)55OH ϩ 6.4 ϫ 10Ϫ6 mol/L Na2(SO3(CH2)3S)2
1.5 mol/L Ni(SO3NH2)2 ϩ 0.5 mol/L H3BO3
Nickel
a Additives in the ‘‘Techni-Silver E’’ electrolyte from Technic, Inc., are not disclosed. Note that corporate and product names are provided for experimental
accuracy. They do not imply NIST endorsement.
the deposits are initially conformal, as reflected in the equal depo-
sition rates on the bottoms and sidewalls of the vias. At later times,
the silver and copper deposit more rapidly on the bottom surfaces
than on the sidewalls, resulting in bottom-to-top filling characteristic
of superconformal deposition. For the silver, the uncertainty ͑2␴͒ of
measurements during the conformal growth period is Ϯ0.03 ␮m.
This value reflects surface roughness and measurement resolution.
Once bottom-up filling begins, the data scatter increases to Ϯ0.25
␮m for all partially filled vias. This value represents measured varia-
tions across individual specimens. For the ϳ10 s period of super-
conformal growth, a Ϯ2 s variation in when superconformal filling
begins ͑i.e., the duration of the incubation period͒ could provide the
observed variation. Such a variation in the ϳ40 s incubation period
would result from a Ϯ5% variation in boundary layer thickness, for
example.
Modeling
Predicted curves are superimposed on Fig. 5 and 6 for vias with
aspect ratios ͑height/diameter͒ of 1.6 and 2.2. These aspect ratios
correspond to those of vias that are 1 ␮m high with diameters equal
to those at the tops and bottoms, respectively, of the experimental
vias. The predictions were obtained using the CEAC-based model of
superconformal deposition in vias published in Ref. 8. All the ki-
netic parameters used in the modeling are shown in Table II; they
come from Ref. 6 for the silver and Ref. 5 for the copper. They were
obtained entirely from studies of the deposition behavior on planar
substrates in both cases. There are no fitting parameters. The elec-
trolytes were comprised of aquo ϪAgϩ and Cu2ϩ ions; thus, two
electrons of charge transfer correspond to electrodeposition of two
silver atoms or one copper atom, respectively. The simulations pre-
dict the conformal growth period, the rapid change to superconfor-
Figure 1. Vias cross-sectioned after copper deposition for the indicated pe-
riods of time at Ϫ0.2 V overpotential ͑SCE͒. The process of feature filling is
captured at 10 s intervals. Note that filling accelerates after approximately 70
s, with the bottom surface moving up more than half the height of the via
during the following 10 s interval. The thickness of the deposit outside the
vias includes a copper seed over the ͑bright͒ tantalum layer. Specimen prepa-
ration resulted in smearing of the tantalum layer in specimens with 20 and 60
s of deposit. These vias were in square arrays with via spacings of ϳ4 ␮m.
Figure 2. Vias cross-sectioned after silver deposition for the indicated peri-
ods of time at Ϫ0.485 V overpotential vs. silver. The process of feature
filling is captured at 10 s intervals. Note that filling accelerates at approxi-
mately 40 s, with the bottom surface moving up more than half the height of
the via during the following 10 s interval. The copper seed is visible between
the ͑bright͒ tantalum layer and the silver deposit outside each via. These vias
were in square arrays with via spacings of ϳ2 ␮m.
Downloaded on 2015-05-09 to IP 61.129.42.30 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, 149 ͑12͒ C637-C641 ͑2002͒
C639
Figure 5. The time-dependent thickness of the copper deposit on the bot-
toms of the filling vias. ͑ࡗ, ᭿, ᭡͒ Data for vias spaced 1, 2, or 4 ␮m apart,
respectively. The sidewall deposits thicken, throughout via filling, at essen-
tially the same rate as the deposits on the bottom surface during the first 60
s. Deposit thicknesses predicted by the model8 for isolated vias, using the
parameters in Table II, are shown for two aspect ratios that span the dimen-
sions of the nonideal vias. ͑ ͒ The predicted thicknesses on the bottom
surface. ͑ ͒ Predicted thickness on the sidewalls. The aspect ratio of
the via being modeled is indicated next to the corresponding solid curve; the
dashed curves are distinguishable only from where they terminate ͑i.e., when
the bottom surface escapes the via͒.
Figure 3. Vias cross-sectioned after silver deposition for the indicated peri-
ods of time at Ϫ0.485 V overpotential vs. Ag. The superconformal filling of
the via is captured at 3 s intervals. These vias were in square arrays with via
spacings of ϳ2 ␮m.
mal filling, and within measurement uncertainty, the duration of the
superconformal growth period. The predicted conformal growth pe-
riods are, however, significantly shorter than the experimental val-
ues. Note that although this particular model does not address the
latter stage of deposition, other CEAC-based models predict bump
formation over filled features such as observed here ͑Fig. 3͒.
Figure 7 shows the predictions for the conformal nickel deposi-
tion. The predictions prior to sidewall impingment are based
on
a time- and location-independent growth velocity of
vo
ϭ 0.0026 ␮m/s obtained from the 0.31 ␮m sidewall thickness at
120 s ͑Fig. 4͒. The time ␶ before the sidewalls impinge is given by
i
␶ Ϸ R/ o , where R is the radius at the bottom of the via. For
v
i
Figure 4. Vias cross-sectioned after nickel deposition for the indicated pe-
riods of time at Ϫ1.42 V SCE. The conformal filling of the via is captured at
15 s intervals, with the notch manifesting geometrical leveling from the
sloping sidewalls visible in the final image ͑after a 30 s interval͒. These vias
were in a square array with feature spacing of ϳ2 ␮m. Smearing of the
nickel deposits during specimen preparation is visible on the surface outside
the via with 30 s of deposition.
Figure 6. The time-dependent thickness of the silver deposit on the bottoms
of the filling vias. ͑ࡗ, ᭿, ᭡͒ Data for the vias spaced 1, 2, or 4 ␮m apart,
respectively. Deposit thicknesses on the bottoms of isolated vias predicted by
the model8 using the parameters in Table II are shown ͑
͒ for two aspect
ratios that span the dimensions of the nonideal vias. The aspect ratio of the
via being modeled is indicated next to the corresponding solid curve.
Downloaded on 2015-05-09 to IP 61.129.42.30 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C640
Journal of The Electrochemical Society, 149 ͑12͒ C637-C641 ͑2002͒
Table II. Parameters for the modeling of superconformal deposition of silver and copper from Ref. 6 and 5, respectively.
⍀
Dcatalyst
Dmetal Ion
⌫
k(␩) ͓or k (␩)]
io
Ј
␦
␮m
Metal
(cm3/mol)
(cm2/s)
(cm2/s)
(mol/cm2)
(cm3/mol s)
␣
(mA/cm2)
2
3
4.5 ϫ 10Ϫ6
5 ϫ 10Ϫ6
7.6 ϫ 10Ϫ6
1 ϫ 10Ϫ5
7.6 ϫ 10Ϫ10
9.7 ϫ 10Ϫ10
5.2 ϫ 104 ϩ 5.1 ϫ 106␩
1.8 ϫ 105 Ϫ 2.7 ϫ 107␩
0.17 ϩ 0.11␪
0.45 ϩ 0.30␪
Silver
Copper
10.3
7.1
135
150
0.32 ϩ 0.86␪
0.069 ϩ 0.64␪
T ϭ 298 K
R
B ϭ 8.314 J/mol K
F ϭ 96485 C/mol
approximations on the period of conformal growth is unclear, as an
exact solution for this geometry does not presently exist to allow
comparison.
R ϭ 0.25 ␮m, impingment occurs at ␶ Ϸ 96 s. After the sidewalls
impinge, the upward velocity of the bottom of the resulting notch is
i
given by o /sin(␪), where ␪ is the outward tilt of the sidewalls from
v
vertical. This is a manifestation of geometrical leveling. Because the
tilt angle is small, ␪ Ϸ 5°, the upward velocity after impingement is
comparatively high, and the resulting predictions are sensitive to
Feature density.—Feature density does impact the time before
the rapid upward filling begins ͑Fig. 5 and 6͒. That the delay in-
creases with pattern density implies that this is a manifestation of
either/both metal ion or catalyst depletion over the features. Though
the model includes composition gradients of the metal ion and cata-
lyst across the boundary layer, these gradients are determined by
balancing the fluxes across the boundary layer with those attaching
to a planar surface. Metal deposition and/or catalyst adsorption on
the additional surface area associated with the sidewalls evidently
leads to a significant increase in the magnitude of the gradient and
associated drop in the concentrations at the specimen surface. Ini-
tially, the additional surface area responsible for this effect is ex-
pressed using the factor ␺
both and ␪. A value of o that is 5% higher, or a value of tilt angle
vo
v
that is 1° smaller, is all that is required to make the prediction agree
with the measured height of the notch at 120 s. Both changes are
within measurement uncertainty.
Discussion
It appears unlikely that the CEAC mechanism is not the phenom-
enon underlying the superfilling of the vias: the model predicts the
conformal deposition followed by the superconformal deposition,
including the relatively sharp transition between the two, and within
measurement error, the duration of the superconformal growth pe-
riod, all with no fitting parameters. Several factors that could affect
the duration of the conformal growth period, for which predictions
underestimate the experimental results, are described in what fol-
lows. The CEAC-based models that have been quantitatively suc-
cessful in predicting the geometry of superconformal filling in
trenches, also with no fitting parameters, have not examined the time
dependence of filling as done here.
2␲Rh
d2
␺ ϭ 1 ϩ
͓1͔
which equals the actual surface area divided by the planar area,
given via radius R, via spacing d, and via height h. For a flat
surface (d → ϱ), ␺ ϭ 1. For the 1 ␮m deep vias, using an average
radius of 0.25 ␮m, ␺ ϭ 1.1 for the 4 ␮m via spacing, ␺ ϭ 1.4 for
the 2 ␮m via spacing, and ␺ ϭ 2.6 for the 1 ␮m via spacing.
An approximate mass balance equation representing equality of
the metal ions diffusing across the boundary layer with the rate of
metal deposition, for the nonplanar surface, is expressed as follows.
In terms of the metal ion concentration drop from the bulk electro-
lyte concentration, Cbulk , to that at the top of the feature, Ctop , for
boundary layer thickness ␦, atomic volume ⍀M , and diffusion co-
efficient DM , one can write
Approximations and Uncertainties
The model itself.—The particular CEAC-based model used here
contains several approximations, including an assumption of instan-
taneous redistribution on the bottom surface of catalyst that had
been adsorbed on eliminated sidewall area. The impact of these
Cbulk Ϫ Ctop
␺
o ϭ ⍀MDM
͓2͔
v
␦
This presumes spatially uniform deposition rate o , ignoring com-
v
position gradients within the via,3-5 and only applies prior to the
start of superconformal deposition. A similar analysis can be done
for the impact of increased surface area on the accumulation of
catalyst. When accumulation is diffusion limited, so that Ctop
Ӷ Cbulk , the right side is nearly constant and the deposition rate
vo
scales inversely to ␺. The incubation time for the inception of
bottom-up filling thus scales with ␺. Under such conditions, Eq. 2
predicts that the incubation times for the vias with 4 ␮m spacing
will be only ϳ10% longer than those for isolated vias, while those
with 2 ␮m spacing will be ϳ40% longer, and those with 1 ␮m
spacing will be ϳ160% longer. These effects are less pronounced
when interface kinetics play a significant role in limiting either
metal deposition or catalyst adsorption, such as is the case for both
the silver and copper depositions. As would be expected from these
results, the experimental data for the vias with 4 ␮m spacing lie
closest to the curves ͑Fig. 5 and 6͒, which are predictions for iso-
lated vias.
Figure 7. The time-dependent thickness of the nickel deposit on the bottoms
and sidewalls of the filling vias. The different thicknesses measured from
͑᭿͒ the bottoms and ͑᭡͒ sidewalls prior to 100 s are likely due to the
difficulty of discriminating between the copper seed and nickel electrode-
posit within the features. ͑- - - -͒ Linear growth rate determined from the
Depletion over the experimental regions due to consumption
from the global distribution of vias, rather than the local ␺, is also
likely. This is of particular concern because the distance between
patterned blocks, and the sizes of some blocks ͑noted earlier͒, are
sidewall thickness measured at 120 s. ͑
͒ The associated prediction for
the height of the bottom surface, obtained for sidewalls tilting outward 5°
from vertical.
Downloaded on 2015-05-09 to IP 61.129.42.30 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, 149 ͑12͒ C637-C641 ͑2002͒
C641
significantly smaller than the anticipated 135-150 ␮m boundary
layer thickness. Furthermore, ␺ for some blocks is significantly
higher than the values noted previously. Indeed, the general increase
of deposition times beyond the predicted values for all via spacings
as well as the relatively small impact of via spacing on deposition
times are both consistent with such a global depletion effect.
The spatially periodic geometry does lend itself to quantitative
analysis through the calculus of partial differential equations with
boundary values. Such analysis has been done, for example, to study
the impact of spatially varying area for deposition ͑achieved by local
variation of coverage with photoresist͒ on the electrical potential
and resulting deposition.13 Variations of feature density on the wa-
fer, or specimen, scale require an additional level of calculation to
deal with concentration variations as electrolyte moves over regions
of varying pattern density.14,15
the filling of the feature. According to the CEAC model, this leads to
a decrease of the catalyst coverage on that surface and an extended
incubation period. However, this would have the greatest effect
when the bottom surface is moving up significantly, experiencing
the changing radius, as during the period of superconformal deposi-
tion. It would not be expected to substantially extend the incubation
period.
Conclusions
Filling of vias with silver, copper, and nickel was studied experi-
mentally and then modeled. Feature filling during both copper and
silver deposition exhibits the bottom-to-top deposition characteristic
of superconformal growth. Feature filling during the nickel deposi-
tion occurs through geometrical leveling with no evidence of super-
conformal filling behavior. A model based on the CEAC mechanism
of superconformal deposition was used to predict the superconfor-
mal filling with copper and silver. As with previous applications of
CEAC-based models describing superconformal deposition in
trenches, the predictions for superconformal filling of vias are based
entirely on kinetics obtained from studies of deposition on planar
specimens. The model predicts the superconformal filling behavior,
including an incubation period of conformal growth and subsequent
superconformal bottom-to-top filling. Possible explanations for the
underestimated duration of the incubation period were detailed.
Incorrect or inappropriate kinetics.—The parameters used for
the copper fill modeling are for electrolyte containing the additive
3-mercapto-1-propanosulfonate ͑MPSA͒ rather than the dimer ver-
sion bis-͑sodium sulfopropyl͒-disulfide ͑SPS͒ actually used in the
experiments. They were obtained from analysis of hysteresis in cy-
clic current-voltage experiments conducted on planar substrates in
electrolytes containing various concentrations of MPSA.5 To ac-
count for the dimer nature of SPS, an MPSA concentration of twice
the experimental SPS concentration was used in the modeling. The
relevant parameters for the electrolyte with SPS are not presently
available; it is, however, known that the kinetics for adsorption of
SPS from the electrolyte are slower than for MPSA ͑as quantified by
the factor k(␩) in Table II͒.16 Slower deposition, and associated area
reduction, would be expected to lead to a longer conformal growth
period, possibly explaining the discrepancy in the incubation period.
For the silver deposition, in light of the significant unknowns about
the composition of the proprietary electrolyte, it is possible that the
associated kinetic parameters in Table II are also inaccurate.
The National Institute of Standards and Technology assisted in meeting
the publication costs of this article.
References
1. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
Res. Dev., 42, 567 ͑1998͒.
2. T. P. Moffat, D. Wheeler, W. H. Huber, and D. Josell, Electrochem. Solid-State
Lett., 4, C26 ͑2001͒.
3. D. Josell, D. Wheeler, W. H. Huber, and T. P. Moffat, Phys. Rev. Lett., 87, 016102
͑2001͒.
4. D. Wheeler, D. Josell, and T. P. Moffat, Submitted for publication.
5. D. Josell, D. Wheeler, W. H. Huber, J. E. Bonevich, and T. P. Moffat, J. Electro-
chem. Soc., 148, C767 ͑2001͒.
6. T. P. Moffat, B. Baker, D. Wheeler, J. E. Bonevich, M. Edelstein, D. R. Kelly, L.
Gan, G. R. Stafford, P. J. Chen, W. F. Egelhoff, and D. Josell, J. Electrochem. Soc.,
149, C423 ͑2002͒.
7. D. Josell, D. Wheeler, and T. P. Moffat, Electrochem. Solid-State Lett., 5, C44
͑2002͒.
8. D. Josell, D. Wheeler, and T. P. Moffat, Electrochem. Solid-State Lett., 5, C49
͑2002͒.
9. J. Reid and S. Mayer, in Advanced Metallization Conference 1999, p. 53, M. E.
Gross, T. Gessner, N. Kobayashi, and Y. Yasuda, Editors, MRS, Warrendale, PA
͑2000͒.
10. T. Ritzdorf, D. Fulton, and L. Chen, in Advanced Metallization Conference 1999, p.
101, M. E. Gross, T. Gessner, N. Kobayashi, and Y. Yasuda, Editors, MRS, War-
rendale, PA ͑2000͒.
11. E. Richard, I. Vervoort, S. H. Brongersma, H. Bender, G. Beyer, R Palmas, S
Lagrange, and K. Maex, in Advanced Metallization Conference 1999, p. 149, M. E.
Gross, T. Gessner, N. Kobayashi, and Y. Yasuda, Editors, MRS, Warrendale, PA
͑2000͒.
12. A. C. West, S. Mayer, and J. Reid, Electrochem. Solid-State Lett., 4, C50 ͑2001͒.
13. J. O. Dukovic, in Advances in Electrochemical Science and Engineering, Vol. 3, H.
Gerischer and C. W. Tobias, Editors, p. 117, VCH Publishers, New York ͑1994͒.
14. M. O. Bloomfield, S. Sen, K. E. Jansen, and T. S. Cale, in The 18th International
VLSI Multilevel Interconnection Conference Proceedings, p. 397, Santa Clara, CA,
Sept 25-26, 2001.
15. M. O. Bloomfield, S. Soukane, K. E. Jansen, and T. S. Cale, in Semiconductor
Technology, M. Yang, Editor, PV 2001-17, p. 94, The Electrochemical Society
Proceedings Series, Pennington, NJ ͑2001͒.
Incorrect boundary layer thickness.—Spatial and temporal varia-
tion of boundary layer thickness can be significant for a vertical
specimen in an unstirred solution like that used in these experi-
ments, as well as in the studies on planar substrates used to obtain
the kinetic parameters. Indeed, for an electrolyte of similar compo-
sition to that for copper in Table I ͑without additives͒, the local
current decreases monotonically with increasing height on the speci-
men, with a fractional decrease of ϳ40% going up the first 1 cm
from the specimen bottom and an additional ϳ20% going up the
next cm.17 Specimen-to-specimen variation arising from this effect
was reduced in these experiments through the study of only those
features located at and less than 3 mm above the specimen midplane
͑i.e., 1-1.3 cm above the specimen bottom͒. However, accurate as-
sessment and application of the kinetics with a CEAC-based model,
as with any other model, requires control of the boundary layer
thickness, e.g., through a rotating disk geometry. There is no signifi-
cant spatial variation of the potential for the specimen geometry and
deposition conditions used here.
Nonideal via geometry.—The nonvertical sidewalls of the pat-
terned via make the model in Ref. 8, with its vertical sidewalls, an
approximation of the experiments actually being modeled. Due to
the sloping sidewalls, the area of the bottom surface increases as it
moves upward. This offsets the area decrease that is associated with
16. Authors’ preliminary, unpublished results.
17. C. R. Wilke, M. Eisenbert, and C. W. Tobias, J. Electrochem. Soc., 100, 513 ͑1953͒.
Downloaded on 2015-05-09 to IP 61.129.42.30 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).