Effect of additives on copper outplating onto silicon surface from dilute HF solutions

Article abstract of DOI:10.1149/1.1688340

The effects of additives such as acids and surfactants on copper outplating onto silicon surfaces from dilute HF solution were studied. It was found that some additives could significantly reduce copper outplating. Results from potentiommetry, total-refle

Full text of DOI:10.1149/1.1688340

G360
Journal of The Electrochemical Society, 151 ͑5͒ G360-G367 ͑2004͒
0
013-4651/2004/151͑5͒/G360/8/$7.00 © The Electrochemical Society, Inc.
Effect of Additives on Copper Outplating onto Silicon Surface
from Dilute HF Solutions
a
z
Zhan Chen and Rajiv K. Singh
Department of Materials Science and Engineering, University of Florida, Gainesville, Florida 32611, USA
The effects of additives such as acids and surfactants on copper outplating onto silicon surfaces from dilute HF solution were
studied. It was found that some additives could significantly reduce copper outplating. Results from potentiommetry, total-
reflectance X-ray fluorescence, time-of-flight–secondary ion mass spectroscopy, dynamic light scattering, and electron microscopy
suggested that the anionic surfactant had strong interaction with cupric ions in solution, and their complex was adsorbed onto
silicon surfaces, causing nonredox-type copper contamination. Dynamic light-scattering results also revealed strong interaction
between the anionic surfactant and nickel ions. The roles of surfactants in copper outplating are discussed in this paper.
©
2004 The Electrochemical Society. ͓DOI: 10.1149/1.1688340͔ All rights reserved.
Manuscript submitted April 24, 2003; revised manuscript received October 18, 2003. Available electronically April 5, 2004.
With semiconductor devices continuously moving to smaller and
ric sensor for the detection of trace metallic contaminants in HF
1
2
smaller sizes, it is becoming more critical to keep silicon surfaces
contamination free in order to improve device functionality, yield,
and reliability. The RCA-based wet chemical clean is still widely
solution. They found that metal ions that can oxidize silicon shift
the open circuit potential ͑OCP͒ of a silicon electrode positively, and
the amount of shift is proportional to the logarithm of oxidizing ion
concentration in solution. However, the linearity between the OCP
shift and the ion concentration has not yet been discussed. Bertagna
et al. also studied the mechanism of copper outplating on silicon
surfaces in DHF solutions by monitoring the OCP change of a sili-
con electrode as a function of copper contamination time and bulk
1
,2
used in semiconductor device fabrication processes. After SC-1
and SC-2 treatments the silicon surface possesses a chemical oxide
layer about 1 nm thick. For pregate clean, the low-quality chemical
oxide layer is ideally to be removed before growing a high-quality
thermal gate oxide layer. This can be done by dilute HF ͑DHF͒
treatment. DHF-based clean is also widely used in other steps in
semiconductor processing. However, during DHF cleans particles
tend to redeposit onto the silicon surface due to large dispersion
1
3
contamination level.
In this study we combined the potentiometry technique, TXRF,
and ICP-MS to determine the effects of different additives on sup-
pressing copper outplating on silicon surfaces in DHF solutions.
Surfactant—cupric ion interaction in solutions was studied with
light scattering and transmission electron microscopy ͑TEM͒. The
surfactant effect on reducing copper nucleation on silicon surfaces
was studied with scanning electron microscopy ͑SEM͒. The ab-
sorbed surfactant layer on the silicon surface was verified with an
atomic force microscopy ͑AFM͒. Based on these results, the surfac-
tant effect on copper outplating in DHF clean was summarized. The
mechanism of OCP change of a silicon electrode as a function of
surface metal contamination is discussed in the Appendix.
3
interactions between the bare silicon surface and particles, and
heavy metal ions such as Cu2 can outplate from DHF baths, leav-
ing pitting and metal particles on silicon surfaces through redox
reactions4
ϩ
Si ϩ 2Cu2 ϩ 6HF ϭ H SiF ϩ 2Cu ϩ 4H
ϩ
ϩ
͓1͔
2
6
In order to prevent particle redeposition in DHF cleans, surfac-
tants are frequently used to modify electric charges on both silicon
and particle surfaces so that a repulsive electrostatic force over-
3
,5
comes attractive dispersion force between the two surfaces. Sur-
factants are also reported to have a different effect on copper out-
plating from HF solutions onto silicon surfaces. Ohmi et al. have
shown that copper deposition on silicon wafers from BHF63 solu-
tion was decreased if a certain type of hydrocarbon or fluorocarbon
Experimental
Different additives such as HCl, H O , HNO , a cationic sur-
2
2
3
factant, and an anionic surfactant were selected to study their effects
on copper outplating. The cationic surfactant we used is an alkyltet-
ramethylammonium bromide ͑CTAB͒, and the anionic surfactant is
a sulfur-containing surfactant. They have similar chain structures
and molecular weights. Our previous results showed that both of
them are effective in preventing particle redeposition during dilute
6
surfactant was used. Obeng demonstrated that the addition of an-
ionic perfluorocarbon surfactants to a 5% HF solution reduced sili-
7
con surface copper concentration fourfold. Jeon et al. reported a
tenfold decrease in copper outplating by using surfactant OHS,
8
which is an alkylphenol polyglycidol nonionic surfactant. However,
3
HF cleaning of silicon surfaces.
Torcheux et al. reported that FC-98, a perfluoroalkylcyclohexylsul-
The potentiometry measurement set up was as follows. A 2 in.
9
fonate clearly increased silicon surface copper contamination.
͑100͒ n-type silicon wafer with 1-10 ⍀ cm resistivity ͑Silicon Quest
It has been postulated that the reasons for surfactants to be ef-
fective in decreasing copper outplating are ͑i͒ surfactant films ad-
sorb on and passivate substrate surfaces, and (ii) the micelles im-
mobilize cupric ions, which reduces the efficiency of electron
transfer kinetics and free cupric ion concentration, and thus the driv-
International͒ was subjected to SC-1 cleaning, DHF clean, and DI
water rinse to remove particles and chemical oxide before being
mounted on a custom-made Teflon electrochemical cell. Only the
polished side of a wafer was contacted with solution and the back
side was contacted to a stainless steel base that had a stainless steel
connection to the outside. A thin film of eutectic Ga-In alloy ͑Alfa
Aesar͒ was employed between the back side of the wafer and the
metal base to ensure ohmic contact. A HF-resistant Ag/AgCl refer-
ence electrode ͑Fisher Scientific͒ was positioned near the wafer. 50
mL 0.5 wt % solutions with/without additives were used in the cell.
A 25 W Leica incandescent lamp was employed and the light di-
rectly shined onto the wafer. Copper was added by using 1000 ppm
atomic absorption spectroscopy ͑AAS͒ standard ͑Fisher Scientific͒
to different final bulk concentrations in the solutions. Each time the
Cu was added, the solution was gently shaken until it appeared well
mixed. OCP change with time was recorded using a CHI 660 Elec-
trochemical workstation. After 15 min more Cu was added to the
7
ing force for copper outplating. Nevertheless, the role of surfactants
in increasing copper contamination is still unclear.
Because the amount of copper contamination on silicon surface
is much less than 1 ϫ 1014 atom/cm , the order of magnitude of a
monolayer, the characterization of surface copper contamination is
usually conducted by employing total-reflectance X-ray fluorescence
2
͑
͑
TXRF͒ or inductively coupled plasma mass spectrometry
1
0,11
ICP-MS͒.
Recently, Chyan et al. reported a novel potentiomet-
a
Present address: Intel Corporation, Santa Clara, California 95052, USA.
E-mail: rsing@mse.ufl.edu
z
Downloaded on 2015-01-07 to IP 193.0.65.67 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, 151 ͑5͒ G360-G367 ͑2004͒
G361
Figure 1. OCP change with copper addition ͑fresh 0.5% HF is used at the
final stage͒.
Figure 2. OCP change with copper addition ͑the surface copper is stripped
and fresh 0.5% HF is used at the final stage͒.
solution and the processes were repeated. To regenerate the silicon
per ions in the solution. This statement is further verified by the
results plotted in Fig. 2 in which the stripping solution is used to
dissolve the surface copper, and the OCP in fresh 0.5% HF goes
back to about the original value. This confirms that the surface cop-
per, not copper in the solution, is responsible for the OCP change.
Figure 3 plots the OCP change as a function of the logarithm of
bulk copper concentration. The least-squares regression result shows
that the relationship between OCP change and log͓Cu/ppb͔ is linear
wafer electrode, a 5 mL 5% HF ϩ 5% HNO ϩ 5% H O ͑wt͒ so-
3
2
2
lution was employed to remove surface Cu.
Surface contaminations were examined by TXRF and time-of-
flight secondary ion mass spectroscopy ͑TOF-SIMS͒. Silicon wafer
samples were dipped into 0.5% HF ϩ 100 ppb copper baths with/
without surfactant for 10 min under illumination. For all the experi-
ments, the surfactant concentrations were 1% ͑wt͒. Particle size
analyses were carried out in 1% anionic surfactant ϩ500 ppm cop-
per contaminant water solution by dynamic light scattering. TEM
analyses on particles in 1% anionic surfactant ϩ500 ppm copper
contaminant were also conducted. The TEM sample was prepared
by dipping a copper grid covered with an amorphous carbon thin
film ͑Ted Pella͒ into the solution to capture particles.
Characterization of adsorbed surfactant layers on a silicon sur-
face was done by using a surface force measurement technique with
an atomic force microscope ͑Digital Instruments SPM nanoscope
III͒. The liquid cell and the silicon nitride tips were all from Digital
Instruments. The detailed method and experimental setup can be
found elsewhere.3 Solutions for the measurement were prepared
with DI water ϩ1% surfactant and the pH values were adjusted with
HCl to 1.87-1.90. The solutions contained 0.1% HF in order to
prevent oxide formation on the silicon surface.
2
with r ϭ 0.99. Figure 4 summarizes the OCP change vs. time on
different p-type wafers from the same batch and Fig. 5 shows the
same parameter obtained from different n-type wafers from the same
batch. It shows that n-type wafers have higher slope than p-type
wafers. Norga et al. reported that n-type wafers are more susceptible
1
4
to copper outplating than p-type wafers. Thus, the slope change is
qualitatively correlated to the susceptibility of copper outplating.
From Fig. 4 and 5 we can see that the slopes of the regression lines
are much more repeatable than the absolute OCP values. This is
because the OCP of a silicon electrode is also dependent on its
surface conditions. It is very difficult to have identical surface con-
ditions on different wafers or on the same wafer in different times
͑the mechanism of the OCP change of a silicon electrode in a DHF
solution is discussed in the Appendix͒. Nevertheless, the repeatable
slope of regression line can provide a qualitative measure of the
degree of copper outplating and thus the efficiency of an additive on
The surfactant effect on protecting damaged silicon surfaces
from copper outplating was also studied. Surface damages were in-
tentionally introduced by scratching the silicon surface with a dia-
mond scribe. Scratched wafer pieces 1 ϫ 1 cm in area were dipped
into 0.5% HF ϩ 1 ppm Cu2 solutions with and without surfactant
for 10 min under 25 W incandescent light illumination. Then the
wafer pieces were DI water rinsed and subject to SEM observation.
Secondary electron images and characteristic Cu and Si K␣ X-ray
images were taken.
ϩ
Results and Discussion
Potentiometric method to characterize additive efficiency on
copper outplating.—Copper ions in a DHF solution cause the
change of the OCP between a silicon electrode and a reference
electrode.12 The OCP change of a silicon electrode with copper ad-
dition is plotted in Fig. 1. OCP change correlates well with the bulk
copper concentration. Each time right after copper is added, the OCP
increases suddenly and then gradually changes to a relative stable
value after 15 min. With the addition of more copper contaminants,
similar change in the OCP value is observed. After a DI water rinse
and a fresh 0.5% HF solution without copper contamination is used
on the same electrode, the OCP does not go back to its original
value, thus suggesting that the OCP response is not caused by cop-
Figure 3. OCP change vs. the logarithm of copper bulk concentration.
Downloaded on 2015-01-07 to IP 193.0.65.67 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G362
Journal of The Electrochemical Society, 151 ͑5͒ G360-G367 ͑2004͒
Figure 4. p-Type wafer OCP change vs. the logarithm of copper bulk con-
centration ͑different wafers from the same batch͒. The slope is repeatable but
the absolute OCP values are not repeatable. Each type of symbol represents
data from the same run of experiment.
Figure 6. Effects of H O , HCl, and HNO on copper outplating ͑5%
2
2
3
H O , 0.5 M HCl, and 0.5 M HNO in 0.5% HF, respectively͒. Plots of
2
2
3
0.5% HF only are also graphed. These additives clearly decrease the slopes,
indicating decreased copper outplating. Each type of symbol represents data
from the same run of experiment.
copper outplating. We used three additives, H O , HCl, and HNO ,
2
2
3
to further examine this correlation, whose effects on minimizing
copper outplating are well documented.1
are shown in Fig. 6. When 5% H O , 0.5 M HCl, or 0.5 M HNO
3
5-18
The effects of additives
2
2
high copper contamination level ͑Ͼ50 ppb͒, the slopes with and
without surfactant are about the same ͑44.5 mV with surfactant vs.
is added the slope is much smaller than that obtained in the refer-
ence sample. Therefore the slope of OCP change can be used to
characterize the surfactant effect on copper outplating in dilute HF
solutions.
47.5 mV without surfactant͒. This suggests that the cationic surfac-
tant can reduce copper outplating effectively at low copper contami-
nation level. The TXRF surface copper concentration results in Fig.
7
1
b indicate that the surfactant indeed decreases copper outplating. At
00 ppb bulk contamination level as in TXRF experiments, the sur-
Surfactant effect on copper outplating from dilute HF solution
on the silicon surfaces.—In this study, we combined the potentio-
metric method and direct measurement methods ͑TXRF and TOF-
MS͒ to characterize copper outplating onto silicon surfaces. The
plots of OCP change with bulk copper concentration with and with-
out the cationic surfactant are shown in Fig. 7a. The open symbols
represent the case without surfactant, the solid symbols represent the
case with surfactant, and each type of symbol represents data from
the same run of experiment. At low copper contamination level
factant starts losing its capability to reduce copper outplating sug-
gested by Fig. 7a. Nevertheless, the overall surface copper contami-
nation is still lowered. It could be understood by drawing a
regression line passing only the first four points of the same experi-
mental run with surfactant, i.e., the points corresponding to the bulk
concentration up to 100 ppb. The slope of this line is lower than that
of the control sample.
The effect of the selected anionic surfactant on copper outplating
is shown in Fig. 8. The OCP change vs. log͓Cu͔ plots in Fig. 8a
indicate that the surfactant is also effective in decreasing copper
outplating, because it gives significantly smaller slopes ͑7.8 mV
with surfactant vs. 50.0 mV without surfactant͒. However, TXRF
results in Fig. 8b show only about twofold decreases in copper out-
plating, contradicting our prediction. In order to investigate the cop-
per outplating mechanism with the presence of this surfactant, sur-
face copper change with dipping time was measured. The bulk
copper concentration was still 100 ppb. The results are given in Fig.
͑
Ͻ50 ppb͒, the slope is remarkably smaller when surfactant is
present ͑13.4 mV with surfactant vs. 47.5 mV without surfactant͒; at
9. The most important observation in this experiment is that the
surface copper concentrations do not change with dipping time
when the surfactant is present, implying that the surface copper may
not be due to redox reaction but some kind of adsorption-type reac-
tion. In the solution some free cupric ions associate with surfactant
molecules on the anionic group, and the charges are neutralized.
Thus, surfactants lose their hydrophilicity, and they have more ten-
dency to adsorb onto hydrophobic bare silicon surfaces. Further-
more, the final rinse by DI water cannot remove adsorbed
surfactant-copper complexes totally from silicon surfaces. There-
fore, copper surface contamination is high. The OCP method re-
sponds only to metallic copper ͑see Appendix͒. That is why the OCP
method suggests low copper contamination while TXRF results
show high copper contamination.
Figure 5. n-Type wafer OCP change vs. the logarithm of copper bulk con-
centration ͑different wafers from the same batch͒. The slope is repeatable but
the absolute OCP values are not. Each type of symbol represents data from
the same run of experiment.
In order to support our hypothesis on copper-surfactant associa-
tion, we increased the copper concentration to 500 ppm in a
0.5% HF ϩ 1% anionic surfactant solution, and the solution was
Downloaded on 2015-01-07 to IP 193.0.65.67 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, 151 ͑5͒ G360-G367 ͑2004͒
G363
Figure 8. The effect of the selected anionic surfactant on copper outplating
from DHF solution onto silicon surfaces. ͑a͒ The plots of OCP change vs.
log͓Cu͔. The anionic surfactant also decreases the slope significantly. ͑ᮀ,᭺͒
The case without surfactant; ͑᭿,᭹͒ the case with surfactant. Each type of
symbol represents data from the same run of experiment. ͑b͒ TXRF results of
surface copper concentration after 10 min dipping into 100 ppb copper-
contaminated DHF solution. The surfactant decreases copper outplating only
about twofold.
Figure 7. Effect of the selected cationic surfactant on copper outplating
from DHF solution onto silicon surfaces. ͑a͒ Plots of OCP change vs.
log͓Cu͔. The cationic surfactant decreases the slope. ͑᭞,᭺,᭝,ᮀ͒ The case
without surfactant; ͑᭢,᭿͒ the case with surfactant. Each type of symbol
represents data from the same run of experiment. ͑b͒ TXRF results of surface
copper concentration after 10 min dipping into 100 ppb copper-contaminated
DHF solution. The cationic surfactant decreases surface copper contamina-
tion.
subject to dynamic light-scattering particle size analysis. The results
are graphed in Fig. 10. They show that after the addition of cupric
ions to achieve bulk concentration of 500 ppm, the measured par-
ticle distribution clearly shifted to larger sizes compared with the
case without copper contaminants. Those particles were subjected to
TEM analysis. The bright-field TEM in Fig. 11 shows that the pre-
cipitation formed an oily film with localized droplets on amorphous
thin carbon film of copper grid. The selected area diffraction ͑SAD͒
was done both on the oily film and the droplets. The SAD pattern
indicates that they are amorphous.
If copper is indeed precipitated onto silicon surfaces with surfac-
tants, then the silicon surfaces should have high sulfur contamina-
tion as well, because the anionic surfactant we studied contains sul-
fur. Unfortunately, TXRF did not detect a sulfur signal at this
contamination level. In order to prove our hypothesis, another an-
ionic surfactant with two sulfur-containing groups was briefly stud-
ied on its effect on copper outplating by OCP monitoring and TOF-
SIMS surface analysis. Because this surfactant has two sulfur atoms
on one molecule, it is expected to give a higher sulfur signal even if
the adsorption amount is about the same as the previous surfactant.
OCP change with bulk copper concentration plots in Fig. 12a sug-
gest it is effective in reducing copper contamination. However, the
TOF-SIMS results in Fig. 12b reveal increased copper contamina-
tion. This time the sulfur signal was clearly observed to be higher
than the control sample. These results are in agreement with
Figure 9. TXRF results of surface copper concentrations as a function of
dipping time. The copper contamination with the presence of the anionic
surfactant does not change with time. The solution is 0.5% HF ϩ 100 ppb
copper with/without surfactant.
Downloaded on 2015-01-07 to IP 193.0.65.67 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G364
Journal of The Electrochemical Society, 151 ͑5͒ G360-G367 ͑2004͒
Figure 10. Particle size distributions in 0.5% HF ϩ 500 ppm copper with/
without the anionic surfactant. Addition of the surfactant clearly shifts par-
ticle size distribution.
our hypothesis. Our experiment also explained the observation by
Torcheuxet al., who reported that a perfluoroalkylcyclohexylsul-
9
fonate clearly increased silicon surface copper contamination.
Generally speaking, in DHF clean only noble metals are con-
cerned for metal contamination, because the mechanism for metal
outplating from HF solution is redox reaction. However, if anionic
surfactant is used, the adsorption mechanism should also be consid-
ered. In this mechanism, not only cupric ions but also other metal
ions are possible to form precipitation with the surfactant and are
adsorbed onto silicon wafer surfaces. Therefore, we also briefly
Figure 12. The effect of an anionic surfactant with two sulfur groups on
copper outplating. ͑a͒ Plots of OCP change vs. log͓Cu͔, also providing a
smaller slope. ͑b͒ TOF-SIMS surface analysis results. Copper contamination
and sulfur contamination are higher.
3
ϩ
studied the interactions between the selected surfactant and Fe
2
ϩ
and Ni . Figure 13a and b shows the dynamic light scattering
3
ϩ
2ϩ
results on 1% surfactant ϩ500 ppm Fe and Ni , respectively,
3
ϩ
illustrating that Fe does not form precipitation with the surfactant,
while Ni2 indeed forms precipitation with this surfactant. There-
fore, it is also possible to increase nickel contamination if this sur-
factant is used in DHF clean.
ϩ
amorphous region19 due to preferential nucleation in the amorphous
region because of higher dangling bond density. We also reported
that the preferential deposition of Cu onto a silicon surface could be
induced by employing an AFM scratch on the surface.20 Here we
studied the effect of surfactant on preferential nucleation of copper
on a silicon surface by intentionally introducing surface scratches on
the samples with a diamond scribe and then performing copper con-
tamination experiments. Figure 14 is the result with no surfactant
present in the contaminating solution. It shows that copper only
outplates on the scratch-damaged region. The X-ray mapping ͑Fig.
Preferential surface copper nucleation and the effect of
surfactant on elimination of such preferential nucleation.—We re-
ported that for a silicon surface with both an amorphous region and
a crystalline region, copper was preferentially deposited on the
14b, Cu K␣ mapping; Fig. 14c, Si K␣ mapping͒ verified that the
deposited particles are copper. Amazingly, some surfactants can
eliminate the structural-defect-enhanced copper outplating. We
added 1% CTAB into the copper outplating solution and dipped a
silicon sample with intentional scratches on its surface into this so-
lution for 10 min under the same conditions as the control sample
͑Fig. 14͒. The micrographs of this sample are presented in Fig. 15.
The SEM image ͑Fig. 15a͒ shows that the scratches are intact and
even the sharp features are preserved, which are most prone to cop-
per outplating. The Cu K␣ map ͑Fig. 15b͒ and Si K␣ map show that
the copper contamination level is lower than electron-dispersive
spectroscopy detection limits; the random white dots in Fig. 15b are
due to noise.
Summary on surfactant effects on copper outplating during DHF
cleans.—According to the results presented and those published pre-
viously, the roles a surfactant plays in copper contamination during
DHF cleans can be summarized as follows:
1. Surfactants form a protecting layer on silicon surfaces. The
Figure 11. TEM bright-field image of surfactant-copper particles, amor-
phous in nature ͑SAD pattern is not shown͒.
adsorbed surfactant molecular layer on silicon hinders electron
transfer between a silicon surface and a cupric ion. Because the
Downloaded on 2015-01-07 to IP 193.0.65.67 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, 151 ͑5͒ G360-G367 ͑2004͒
G365
Figure 14. Micrographs of a scratched silicon surface after copper contami-
nation in a 0.5% HF solution: ͑a͒ SEM image, ͑b͒ Cu K␣ X-ray image, and
͑
c͒ Si K␣ X-ray image. Copper deposits only in the scratch-damaged region.
if surfactants are used in DHF cleans, which is proposed to mini-
mize particle redeposition, it is critical to study its effect on metal
contamination and also other metals with low reducing potentials.
Figure 13. Particle size distributions in 0.5% HF ϩ 500 ppm metal ions
with/without surfactant: ͑a͒ Fe͑III͒ and ͑b͒ Ni͑II͒ ions. Ni͑II͒ also forms
particles with this surfactant.
copper reduction on a silicon surface is no longer diffusion con-
trolled ͑Fig. 9͒, it is deduced that the rate-limiting step is electron
transfer. Direct observance of the lack of preferential nucleation
͑
Fig. 15͒ also supports this conclusion. The evidences of surfactant
3
layer formation on the silicon surface are the zeta-potential changes
and the results from direct force measurement. Figure 16 is a plot of
interaction force between a silicon nitride AFM tip and a silicon
surface in a 0.1% HF ϩ 1% CTAB solution as a function of sepa-
ration distance. The pH of the solution was adjusted with HCl to
about 2. The force is normalized by the effective tip radius, R. The
force curve in Fig. 16 is discontinuous at about 2.2 nm separation,
indicating a surface film structure. This film is broken by the con-
tinuous movement of the tip toward the silicon surface.
2. Surfactants interact with cupric ions. Surfactants, especially
anionic surfactants, may have strong interaction with cupric ions.
Our light-scattering experimental results can serve as evidence ͑Fig.
10͒. Surfactants have two effects on copper outplating. One is sur-
factants decrease free cupric ion concentration, resulting in de-
creased driving force for copper outplating ͑Fig. 8͒. The other is the
surfactant-cupric ion complex may be adsorbed onto silicon sur-
faces. If this complex cannot be rinsed totally by following DI water
rinse, then the copper contamination increases ͑Fig. 12͒.
The roles of surfactants in copper outplating are not isolated.
Hence, we observed different effects of surfactants on copper out-
plating. The contamination mechanism via adsorption suggests that
Figure 15. Micrographs of a scratched silicon surface after copper contami-
nation in a 0.5% HF solution with the presence of 1% CTAB: ͑a͒ SEM
image, ͑b͒ Cu K␣ X-ray image ͑white dots due to noise͒, and ͑c͒ Si K␣
X-ray image. No copper was observed even in the scratch-damaged region.
Downloaded on 2015-01-07 to IP 193.0.65.67 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G366
Journal of The Electrochemical Society, 151 ͑5͒ G360-G367 ͑2004͒
ions and can both decrease metal outplating driving force and in-
crease the tendency of physical adsorption of metal ion—surfactant
complexes.
Acknowledgments
We acknowledge Ashland Chemical Company, who funded this
project. We also give special thanks to Rita Vos at IMEC and Chris-
topher Hobbs at Motorola for their generous help on TXRF and
TOF-SIMS measurements.
The University of Florida assisted in meeting the publication costs of this
article.
Appendix
Mechanism of silicon electrode OCP change as
contamination level in DHF solutions
a function of bulk metal
The mechanism of silicon electrode OCP change as a function of bulk cupric ion
concentration can be understood by considering catalyzed mixed potential theory. In our
system if there are no copper ions, the silicon is oxidized and the proton is reduced, as
described by following equations
Figure 16. Plot of interaction force as a function of separation distance
between a silicon nitride tip and a silicon surface in a pH ϳ2 solution with
ϩ
Si Ϫ 4e ϩ 6HF ϭ H2SiF6 ϩ 4H ͑Oxidation͒
͓A-1͔
͓A-2͔
1% CTAB. The force is normalized by effective tip radius, R. Discontinuity
ϩ
at about 2.2 nm suggests the presence of a surfactant film structure.
2H ϩ 2e ϭ H2 ͑Reduction͒
Therefore, the overall reaction on a silicon electrode at OCP is the summation of these
two half reactions
Conclusions
Si ϩ 6HF ϭ H SiF ϩ 2H2 ͑Total reaction͒
͓A-3͔
2
6
Additive effects on copper outplating onto silicon surfaces dur-
ing DHF cleaning were studied. We verified through different ex-
perimentation that additives such as HCl, H O , and HNO are
The partial anodic current density, IA , of Reaction A-1 can be described by
␣AnAF
2
2
3
IA ϭ nAFaSik0,A exp
ͩ
E
ͪ
͓A-4͔
effective in reducing copper outplating during DHF cleans. We re-
ported that CTAB is effective in reducing copper outplating onto
silicon surfaces during DHF cleans. Anionic surfactants may in-
crease metal outplating, not only copper but also other metals such
as nickel, by forming complexes with metal ions and subsequent
adsorption onto silicon surfaces. Combining OCP monitoring, sur-
face metal analysis, and microscopy study, one can differentiate dif-
ferent mechanisms of copper outplating. Experimental data suggest
RT
where n is the number of electron transfer, F is the Faraday constant, aSi is the chemical
activity of silicon which can be taken as one, ␣A is the anodic transfer coefficient, R is
the ideal gas constant, T is the absolute temperature, E is the electrical potential, and k0
is the rate constant for electron transfer at the equilibrium potential. The subscript A
denotes the anodic reaction, Reaction A-1.
For the reduction half reaction, if copper contaminants are present, then we also
2
ϩ
ϩ
need to consider the reduction of Cu as well as the reduction of H . However, results
in Fig. 1 and 2 suggest that the OCP shifts are due to the surface copper, not the bulk
copper in the solution. Therefore, the effect of copper on OCP shift does not result from
copper ion reduction itself. The mechanism can be understood by considering catalyzed
hydrogen evolution ͑Reaction A-2͒ with the presence of copper metal. When copper is
outplated onto a silicon surface, the copper particles act as cathodes on which protons
are reduced to hydrogen. Then the total cathode current density is the sum of the
cathode current densities from the copper particles and the bare silicon. Therefore, for
the cathodic current density we should write
͑
i͒ that the adsorbed surfactant layer on a silicon surface can retard
electron transfer between a silicon surface and a cupric ion, and that
ii) the surfactant molecules in bulk solution can interact with metal
(
Cu
␣
CF
␣C F
Cu
ϪIC ϭ FaHϩ͑1 Ϫ ␪͒k0,C exp
ͩ
Ϫ
E
ͪ
ϩ FaHϩ␪k0 exp
ͩ
Ϫ
E
ͪ
͓A-5͔
,C
RT
RT
where aHϩ is the chemical activity of protons, the subscript C denotes the cathodic
reaction ͑Eq. A-2͒, and ␪ is the coverage of copper particles, and the superscript Cu
denotes the cathode reaction on copper particles. In Eq. A-5, the number of electron
transfer for hydrogen evolution is taken to be one.
Although the copper coverage is small, the reaction rate is much faster than that
without copper. Figure 17 shows an SEM micrograph of a silicon surface contaminated
2
ϩ
with 10 ppm Cu in 0.5% HF solutions for 30 min under illumination. The surface
copper contamination was removed by using a stripping solution. The figure shows the
formation of pits due to silicon dissolution. More importantly, one observes clear bubble
markers due to hydrogen evolution. In fact, we have observed in our experiments that
bubbles were formed on the silicon surface when the silicon was dipped into HF solu-
tion with high concentration of copper contaminants for a long time. If clean HF solu-
tion was used no bubbles were observed, suggesting that the galvanic corrosion of
silicon is much faster than HF etching. Hence, we can neglect the contribution of
cathodic current from the area not covered by copper in Eq. A-5. Furthermore, we
assume that the surface coverage is linearly proportional to copper surface concentra-
tion, i.e.,
␪
ϭ k1͓Cu͔surf
͓A-6͔
where k1 is a constant. Then Eq. A-5 can be simplified to
Figure 17. SEM micrograph of a copper-contaminated silicon surface
2
ϩ
Cu
dipped into 0.5% HF ϩ 10 ppm Cu for 30 min under illumination. The
copper particles were then stripped by a stripping solution. Note the clear
bubble marks due to hydrogen evolution. Bubbles were observed with the
naked eye during copper outplating.
␣ F
C
Cu
ϪIC ϭ FaHϩk1
ͫ
Cu͔surfk exp
ͩ
Ϫ
E
ͪ
͓A-7͔
0,C
RT
Equating Eq. A-4 and A-7, we obtain
Downloaded on 2015-01-07 to IP 193.0.65.67 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, 151 ͑5͒ G360-G367 ͑2004͒
G367
Cu
6. T. Ohmi, T. Imaoka, I. Sugiyama, and T. Kezuka, J. Electrochem. Soc., 139, 3317
͑1992͒.
7. Y. Obeng, in 1994 Semiconductor Pure Water and Chemical Conference, M. K.
Balazs, Editor, p. 110, UPW & Chemical Proceedings, Sunnyvale, CA ͑1994͒.
RT
k1k0,C
RT
OCP ϭ
ln
ͩ
ͪ
ϩ
͑ln aHϩ ϩ ln͓Cu͔surf͒
F͑nA␣A ϩ ␣ CC u
F͑nA␣A ϩ ␣ CC u
nAk0,A
͒
͒
͓
A-8͔
8
. J. Jeon, S. Raghavan, H. Parks, J. Lowell, and I. Ali, J. Electrochem. Soc., 143,
2870 ͑1996͒.
The correlation between bulk copper concentration and silicon surface concentra-
tion after HF dipping has been extensively studied.
9
,14,17,18,21-24
However, the results are
characterized with large within-run and between-run variations during surface copper
9. L. Torcheux, A. Mayeux, and M. Chemla, J. Electrochem. Soc., 142, 2037 ͑1995͒.
10. R. S. Hockett, Surface and Interface Analysis of Microelectronic Materials Pro-
cessing and Growth, 1186, 10, SPIE ͑1989͒.
measurements. Nevertheless, on a linear scale the plot of surface vs. bulk concentration
9
,14,21
at relatively narrow range seems parabolic,
and on a log-log scale the plot is quite
linear.18,21-24 Therefore, the relation between bulk copper concentration and surface
copper concentration can be described as
11. R. Henry, R&D Magazine, 1999, 6LS.
1
2. O. M. R. Chyan, J. Chen, H. Chien, J. Wu, M. Liu, J. A. Sees, and L. Hall, J.
Electrochem. Soc., 143, L235 ͑1996͒.
k3
bulk
͓
Cu͔surf ϭ k2͓Cu͔
͓A-9͔
13. V. Bertagna, F. Rouelle, G. Revel, and M. Chemla, J. Electrochem. Soc., 144, 4175
1997͒.
͑
Thus
14. G. Norga, M. Platero, K. Black, A. Reddy, J. Michel, and L. Kimerling, J. Elec-
trochem. Soc., 144, 2801 ͑1997͒.
ln͓Cu͔surf ϭ ln k2 ϩ k3 ln͓Cu͔bulk
͓A-10͔
15. B. C. Chung, G. A. Marshall, C. W. Pearce, and K. P. Yanders, J. Electrochem.
Soc., 144, 652 ͑1997͒.
where k2 and k3 are constants which characterize the outplating of copper. Combining
Eq. A-8 and A-10, we finally obtain
1
1
1
6. I. Teerlinck, P. W. Mertens, M. Meuris, and M. M. Heyns, in Digest of Technical
Papers-Symposium on VLSI Technology 1996, IEEE, p. 206 ͑1996͒.
7. T. Ohmi, T. Imaoka, T. Kezuka, J. Takano, and M. Kogure, J. Electrochem. Soc.,
OCP ϭ const. ϩ k log͓Cu͔bulk
where k is a constant. Thus, OCP vs. log͓Cu͔ gives a straight line.
References
͓A-11͔
1
40, 811 ͑1993͒.
8. A. L. P. Rotondaro, T. Q. Hurd, H. F. Schmidt, I. Teerlinck, M. M. Hetns, and C.
Claeys, Mater. Res. Soc. Symp. Proc., 386, 183 ͑1995͒.
19. Z. Chen, S. Lee, and R. Singh, J. Electrochem. Soc., 147, 3889 ͑2000͒.
2
2
0. J. Choi, Z. Chen, and R. Singh, J. Electrochem. Soc., 150, C563 ͑2003͒.
1. E. Hsu, H. Parks, R. Craigin, S. Tomooka, J. Ramberg, and R. Lowry, J. Electro-
chem. Soc., 139, 3659 ͑1992͒.
1
2
3
4
. W. Kern and D. Puotien, RCA Rev., 31, 187 ͑1970͒.
. W. Kern, J. Electrochem. Soc., 137, 1887 ͑1990͒.
. Z. Chen and R. Singh, J. Electrochem. Soc., 150, G667 ͑2003͒.
. F. W. Kern, Jr., M. Itano, I. Kawanabe, M. Miyashita, and T. Ohmi, in Proceeding
of the 11th Workshop on ULSI Ultra Clean Technology, Tokyo, p. 23 ͑1991͒.
. T. Ohmi, in Cleaning Technology in Semiconductor Device Manufacturing, J.
Ruzyllo and R. E. Novak, Editors, PV 94-7, p. 3, The Electrochemical Society
Proceedings Series, Pennington, NJ ͑1994͒.
2
2
2. S. Bedge, B. Barron, and H. Lamb, Mater. Res. Soc. Symp. Proc., 315, 467 ͑1993͒.
3. K. Yoneshige, H. Parks, S. Raghavan, J. Hiskey, and P. Resnick, J. Electrochem.
Soc., 142, 671 ͑1995͒.
5
24. I. Teerlink, P. Mertens, H. Schmidt, M. Meuris, and M. Heyns, J. Electrochem.
Soc., 143, 3323 ͑1996͒.
Downloaded on 2015-01-07 to IP 193.0.65.67 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).