Atomic Layer Deposition of Ruthenium Thin Films for Copper Glue Layer

Article abstract of DOI:10.1149/1.1640633

Ruthenium thin films were produced by atomic layer deposition (ALD) using an alternating supply of bis(ethylcyclopentadienyl)ruthenium [Ru(EtCp) ₂] and oxygen at a deposition temperature of 270°C. The relative ratio of the Ru(EtCp)₂ adsorbed on the film surface to the oxygen partial pressure in the following oxygen pulse determines whether Ru or RuO ₂ film was obtained. At the range with higher relative ratio the film was composed of ruthenium, but the film deposited at the lower range was revealed to be ruthenium oxide. In case of the ruthenium thin film, the film thickness per cycle was saturated at 0.15 nm/cycle, and its resistivity was about 15 μω cm. The impurities of carbon and oxygen were incorporated into the film with less than 2 atom %. It was also demonstrated that the ruthenium thin films prepared by ALD can be used as an excellent glue layer to improve the interfacial adhesion of metallorganic chemical vapor deposited copper to TiN. Secondary ion mass spectroscopy analysis showed that the ruthenium glue layer suppressed the interfacial contaminants, such as carbon and fluorine, which originated from the metallorganic precursors of copper.

Full text of DOI:10.1149/1.1640633

Journal of The Electrochemical Society, 151 ͑2͒ G109-G112 ͑2004͒
G109
0
013-4651/2004/151͑2͒/G109/4/$7.00 © The Electrochemical Society, Inc.
Atomic Layer Deposition of Ruthenium Thin Films for Copper
Glue Layer
a
a
b
a, ,z
Oh-Kyum Kwon, Jae-Hoon Kim, Hyoung-Sang Park, and Sang-Won Kang *
aDepartment of Materials Science and Engineering, Korea Advanced Institute of Science and Technology,
Yusong-gu, Taejon 305-701, Korea
b
Genitech, Incorporation, Taedok-gu, Taejon 306-230, Korea
Ruthenium thin films were produced by atomic layer deposition ͑ALD͒ using an alternating supply of bis͑ethylcyclopentadi-
enyl͒ruthenium ͓Ru(EtCp) ͔ and oxygen at a deposition temperature of 270°C. The relative ratio of the Ru(EtCp)2 adsorbed on
2
the film surface to the oxygen partial pressure in the following oxygen pulse determines whether Ru or RuO film was obtained.
2
At the range with higher relative ratio the film was composed of ruthenium, but the film deposited at the lower range was revealed
to be ruthenium oxide. In case of the ruthenium thin film, the film thickness per cycle was saturated at 0.15 nm/cycle, and its
resistivity was about 15 ␮⍀ cm. The impurities of carbon and oxygen were incorporated into the film with less than 2 atom %. It
was also demonstrated that the ruthenium thin films prepared by ALD can be used as an excellent glue layer to improve the
interfacial adhesion of metallorganic chemical vapor deposited copper to TiN. Secondary ion mass spectroscopy analysis showed
that the ruthenium glue layer suppressed the interfacial contaminants, such as carbon and fluorine, which originated from the
metallorganic precursors of copper.
©
2004 The Electrochemical Society. ͓DOI: 10.1149/1.1640633͔ All rights reserved.
Manuscript submitted May 8, 2003; revised manuscript received September 8, 2003. Available electronically January 8, 2004.
With increasing performance and packing density in microelec-
tronic devices, copper is being used as an interconnection metal to
enhance operating speed and reliability.1,2 The damascene process,
which consists of copper deposition into the preformed holes and
trenches in interlayer dielectrics and followed by chemical mechani-
cal polishing ͑CMP͒ of copper, is the standard method for forming
copper interconnects.3 Electroplating has been widely considered
the main damascene fill process due to its superfilling ability, i.e., it
fills trenches and via holes in a bottom-up fashion without any
prepared ruthenium thin film acts as an excellent glue layer for
MOCVD copper by suppressing the formation of interfacial layer
with the residues of carbon and fluorine.
Experimental
Ruthenium thin films were deposited on TiN (40 nm thick)/SiO2
͑100 nm thick͒/Si wafers at a deposition temperature of 270°C under
the deposition pressure of 1 Torr. Bis͑ethylcyclopentadienyl͒ru-
thenium ͓Ru͑EtCp͒ ͔ and oxygen gas were used as the ruthenium
2
4
seams or voids. In addition, it has been recently reported that the
precursor and reactant, respectively. The TiN films were prepared by
phenomenon of bottom-up filling of copper is revealed in the met-
allorganic chemical vapor deposition ͑MOCVD͒ of copper using
iodine as a catalytic surfactant.5 However, to realize both
bottom-up filling processes, a thin and conformal copper seed layer
is positively required. Unfortunately, it is not easy to prepare con-
formal copper seed layers on high-aspect-ratio via holes and
trenches with the current seed layer process, physical vapor deposi-
plasma-enhanced atomic layer deposition ͑PEALD͒. Ru͑EtCp͒ is a
liquid precursor with a relatively high vapor pressure of 0.18 Torr at
2
,6
8
0°C and has good thermal stability. Ru͑EtCp͒ was contained in a
2
bubbler, which was heated to 80°C, and carried by an argon gas at a
flow rate of 50 sccm. To prevent precursor condensation in the feed-
ing line, the stainless steel feeding line was heated to 90°C. One
deposition cycle of ruthenium ALD included four consecutive
pulses, which consisted of a pulse with the metallorganic precursor
7
tion ͑PVD͒. To overcome this technological barrier, seedless copper
formation with electroplating or MOCVD, in which copper is de-
posited directly onto the copper diffusion barrier metals, and copper
seed layer formation using MOCVD or atomic layer deposition
Ru͑EtCp͒ , a purge pulse with argon at a flow rate of 50 sccm, a
2
pulse with 20-120 sccm oxygen gas mixed with 100 sccm argon,
and another purge pulse with argon gas at a flow rate of 50 sccm.
The periods of the purge pulse were 20 s, respectively.
8-10
͑
ALD͒ has been suggested as an alternative.
Because copper films prepared by electroplating, MOCVD, and
Film thicknesses were analyzed by field-emission scanning elec-
tron microscopy ͑FESEM͒ and transmission electron microscopy
͑TEM͒. Film resistivity was calculated from the sheet resistance
measured with a four-point probe and the film thickness. X-ray dif-
fraction ͑XRD͒ analysis using Cu K␣ radiation (␭ ϭ 1.5405 Å)
was used to determine the crystal structure of the deposited films.
Film composition was measured using Auger electron spectroscopy
͑AES͒ and elastic recoil detection time-of-flight spectroscopy ͑ERD-
TOF͒. To evaluate the feasibility of ALD ruthenium as a glue layer
for MOCVD copper films, a peel-off test using 3M Scotch tape was
performed on a sample ͑about 1 ␮m thick MOCVD Cu/Ru glue
layer/40 nm thick TiN/100 nm thick thermal SiO2 stacked on Si
substrates, Fig. 1͒. The MOCVD of copper was carried out for
10 min at a temperature of 180°C under the pressure of 5 Torr
ALD techniques have poor interfacial adhesion to common barrier
metals such as TiN, TaN, and WN, they peel easily during subse-
quent CMP processes.1
1,12
Thus, an adequate glue layer is necessary
in utilizing seedless copper bottom-up filling techniques or in pre-
paring copper seed layers with MOCVD or ALD. Generally, it is
considered that the residues of fluorine and carbon impurities lo-
cated at the interface degrades adhesion of copper films to the cop-
per diffusion barrier metals.1
3,14
Cobalt and nickel have been pro-
posed as a glue layer metal, and it has been proved that both metals
1
5,16
improve adhesion by forming an intermixing layer with copper.
However, these metals, which have high solid solubility limits with
copper, are apt to diffuse into the copper and lead to the increment
of resistivity of copper film.17
To address this issue, we proposed ruthenium as a glue layer in
this paper. Because ruthenium is immiscible with copper and has a
high melting point ͑2334°C͒, it does not degrade the resistivity of
copper film. A thin and conformal ruthenium film has been success-
fully deposited by ALD using Ru(EtCp)2 and oxygen gas, and the
using
hexafluoroacetylacetonate-copper-vinyltrimethylsilane
I
͓͑hfac͒Cu (vtms)͔ as the copper precursor. The decrease in the con-
centrations of fluorine and carbon at the interfacial layer between the
MOCVD copper film and the ALD ruthenium film was analyzed by
secondary ion mass spectroscopy ͑SIMS͒.
Results and Discussion
Figure 2 shows the dependence of film thickness deposited in
one cycle and the resistivity of the deposited films on the Ru͑EtCp͒2
*
Electrochemical Society Active Member.
E-mail: swkang@kaist.ac.kr
z
Downloaded on 2014-11-17 to IP 128.82.252.58 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G110
Journal of The Electrochemical Society, 151 ͑2͒ G109-G112 ͑2004͒
Figure 1. Schematic diagram of the specimen structure for Scotch tape peel-
off adhesion test.
Figure 3. XRD spectra obtained from deposited films for Ru͑EtCp͒2 pulse
times of 2 and 7 s at an O /(Ar ϩ O ) ratio of 44%.
2
2
pulse time at a deposition temperature of 270°C. The pulse time of
Ru͑EtCp͒ was varied from 2 to 12 s, the pulse time of oxygen was
longer Ru͑EtCp͒ pulse time of 7 s shows the typical XRD pattern
generated from ruthenium metal, and the XRD pattern measured
2
2
fixed at 10, and the O /(Ar ϩ O ) ratio during the oxygen pulse
2
2
was controlled at 44%. The deposited film thickness/cycle was satu-
from the film deposited at a 2 s pulse time of Ru͑EtCp͒ coincides
2
rated at about 0.15 nm/cycle as the pulse time of Ru͑EtCp͒ exceeds
with the RuO peaks. Interestingly, this shows that the nature of the
2
2
6
s. It means that the film deposition follows one of the inherent
film deposited by ALD using Ru͑EtCp͒ and oxygen gas depends on
2
characteristics of ALD, a so-called self-limited reaction. However,
the pulse time of Ru͑EtCp͒ , i.e., the amount of the ruthenium pre-
2
when the pulse time of Ru͑EtCp͒ was shortened to 3 s or less, an
2
cursor adsorbed on the film surface even though the oxygen partial
pressure of the following oxygen pulse remains at a constant. Be-
cause one atomic monolayer ͑ML͒ of RuO2 is almost 2.3 times
thicker than that of Ru, the deposited film thickness/cycle vs. pulse
abrupt increase in the deposited film thickness/cycle appeared. This
phenomenon deviated from the expected results in typical ALD; the
deposited film thickness/cycle continuously increases with the pulse
time of precursors until it is saturated in typical ALD. Moreover, the
resistivity of the deposited film was also abruptly changed from
about 15 ␮⍀ cm to about 70 ␮⍀ cm. In order to clarify these un-
usual characteristics in the film depositions of ALD, an XRD analy-
sis was performed on the deposited films. Figure 3 shows the XRD
patterns obtained from the films deposited at the pulse time of
Ru͑EtCp͒2 of 2 and 7 s, respectively. The film deposited at the
time of Ru͑EtCp͒ did not follow a typical trend of ALD, as shown
2
in Fig. 2.
The effect of the O /(Ar ϩ O ) ratio on the film deposition was
2
2
evaluated. Figure 4 shows the dependence of the deposited film
thickness/cycle on the O /(Ar ϩ O ) ratio in the oxygen gas pulse.
2
2
Figure 2. Dependence of film thickness deposited for one cycle and resis-
tivity of ALD ruthenium films on pulse time of Ru͑EtCp͒2 at an O /(Ar
Figure 4. Dependence of deposition rate on O /(Ar ϩ O ) ratio during
2 2
2
ϩ O ) ratio of 44%. Ru͑EtCp͒ for 2-12 s and oxygen gas mixed with 100
oxygen gas pulse in the cases of Ru͑EtCp͒ pulse time of 2 and 3 s. During
2
2
2
sccm argon for 10 s were alternatively supplied.
deposition, the oxygen pulse time was fixed at 10 s.
Downloaded on 2014-11-17 to IP 128.82.252.58 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 ͑2͒ G109-G112 ͑2004͒
G111
Table I. Results of the adhesion test of CVD copper.
͑
a͒ Without glue layer
Heat-treatment As deposited
300°C
20 min
Fail
300°C
40 min
Fail
300°C
60 min
Fail
conditions
Results of tape
peel test
Fail
͑
b͒ With ruthenium glue layer
Thickness
of Ru glue
layer
4 nm
10 nm
15 nm
20 nm
͑30 cycles͒ ͑70 cycles͒ ͑100 cycles͒ ͑130 cycles͒
Results of
Tape peel
test
Pass Pass Pass Pass
increasing O /(Ar ϩ O ) ratio, and the transition takes place sud-
2
2
denly at a particular O /(Ar ϩ O ) ratio, depending on the total
2
2
number of ruthenium precursor molecules adsorbed on the film sur-
face.
To evaluate whether the oxygen partial pressure or the total oxy-
gen dose supplied into the reaction chamber is the dominant factor
Figure 5. Dependence of deposition rate and resistivity on oxygen gas pulse
time for O /(Ar ϩ O ) ratio of 17% when the Ru͑EtCp͒ pulse time was
2
2
2
2
s.
The oxygen pulse time was fixed at 10 s and the Ru͑EtCp͒ pulse
2
time was set at 2 and 3 s, respectively. The deposited film thickness/
cycle was abruptly increased at a certain O /(Ar ϩ O ) ratio de-
2
2
pending on the Ru͑EtCp͒ pulse time. When the Ru͑EtCp͒ pulse
2
2
time was 3 s, the deposition rate abruptly jumped from 0.08 to 0.17
nm/cycle at the O /(Ar ϩ O ) ratio of around 33%. When the
2
2
Ru͑EtCp͒2 pulse time was 2 s, the deposited film thickness/cycle
was reduced on the whole region of the O /(Ar ϩ O ) ratio as
2
2
compared to that obtained from the Ru͑EtCp͒ pulse time of 3 s due
2
to the reduced amount of the ruthenium precursor adsorbed on the
film surface. More interestingly, the abrupt increment of the deposi-
tion rate also occurred at the decreased O /(Ar ϩ O ) ratio of
2
2
around 25%. This implies that the O /(Ar ϩ O ) ratio, where the
2
2
abrupt increment of the deposition rate occurs, depends on the pulse
time of Ru͑EtCp͒ , i.e., the amount of the Ru precursor adsorbed on
2
the film surface. From XRD analysis, it has been also confirmed that
the abrupt increment of the deposition rate is attributed to the for-
mation of RuO . That is, in the ALD system using Ru͑EtCp͒ and
2
2
oxygen gas the deposited film is converted from Ru to RuO with
2
Figure 6. Cross-sectional SEM image showing excellent step coverage of an
Figure 7. The SIMS depth profiles of CVD Cu deposited ͑a͒ on TiN and ͑b͒
ALD ruthenium film at a high-aspect-ratio trench.
on ruthenium substrates.
Downloaded on 2014-11-17 to IP 128.82.252.58 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G112
Journal of The Electrochemical Society, 151 ͑2͒ G109-G112 ͑2004͒
determining the types of the deposited film ͑Ru or RuO ), the oxy-
Ru(EtCp)2 to the oxygen partial pressure in the following oxygen
2
gen pulse time was changed from 10 s to 25 s, while the O /(Ar
pulse determines which film ͑Ru or RuO ) is deposited. Increasing
2
2
ϩ O ) ratio was maintained at 17% and the pulse time for
oxygen partial pressure changes the deposited film from Ru to
2
Ru(EtCp)2 was fixed at 2 s ͑Fig. 5͒. As mentioned before, at the
oxygen pulse time of 10 s ruthenium thin film was formed under this
RuO , with the transition taking place abruptly at a certain oxygen
2
partial pressure which depends on the amount of ruthenium precur-
sor molecules adsorbed on the film surface. The resistivity of the
ruthenium and RuO2 films was about 15 and about 70 ␮⍀ cm,
respectively. The impurities in the ruthenium films were carbon ͑Ͻ2
atom %͒ and oxygen ͑Ͻ2 atom %͒. A peel-off adhesion test con-
firmed that an ALD ruthenium glue layer greatly improved interfa-
cial adhesion of MOCVD copper to barrier metal of TiN. Introduc-
ing a ruthenium glue layer enhanced the adhesion by suppressing
interfacial contamination such as fluorine and carbon. Therefore,
ruthenium films grown by ALD can be effectively used as a glue
layer between MOCVD copper and diffusion barrier metal.
O /(Ar ϩ O ) ratio. As shown in Fig. 5, no abrupt transition from
2
2
Ru to RuO was observed with increasing the oxygen pulse time,
2
indicating that the partial pressure of oxygen controls the types of
the deposited film, not oxygen dose. Based on these results, we
concluded that the types of the deposited film, whether Ru or RuO2 ,
is determined by the relative ratio of the amount of the ruthenium
precursor molecules adsorbed on the film surface to the oxygen
partial pressure supplied during the following oxygen pulse. This
result will give us an additional usefulness in conveniently fabricat-
ing nanoscaled stacked films of Ru/RuO2 .
20 nm-thick ruthenium films were deposited on a 0.2 ␮m wide
oxide trench pattern with an aspect ratio of 8:1 and its cross-
sectional SEM image is shown in Fig. 6. The ruthenium film had
excellent step coverage on the trench pattern ͑Fig. 6͒, and the impu-
rities incorporated in the film, measured with AES and ERD-TOF,
were mainly carbon and oxygen with less than 2 atom %. To evalu-
ate the feasibility of using ruthenium thin film as a glue layer to
improve the interfacial adhesion of MOCVD copper to TiN, a peel-
off adhesion test using Scotch tape was performed after MOCVD of
about 1 ␮m thick copper on ruthenium glue layer. As a controlled
experiment, an adhesion test of MOCVD copper was performed on
TiN substrate without a ruthenium glue layer. In this case, the
MOCVD copper films readily peeled off on TiN regardless of
postheat-treatment used, as shown in Table Ia. However, as shown in
Table Ib, when the ALD ruthenium glue layer was introduced, the
adhesion of MOCVD copper to TiN was greatly enhanced without
any postheat-treatment. Because the glue layer should be as thin as
possible, we carried out the adhesion tests of MOCVD copper with
decreasing the thickness of the ruthenium glue layer. We found that
a 4 nm thick ruthenium glue layer was sufficient to improve the
interfacial adhesion of MOCVD copper to the diffusion barrier
metal without any failure in the peel-off adhesion test, as shown in
Table Ib. To determine the mechanism responsible for increased
adhesion strength with introducing a ruthenium glue layer, SIMS
analysis was performed. As shown in Fig. 7, the intensities of the
fluorine and carbon in the SIMS spectrum were about two orders of
magnitude lower at the interface of CVD copper with a ruthenium
glue layer compared with directly on TiN. Thus, the ALD ruthenium
glue layer enhances adhesion by suppressing interfacial contami-
nants such as fluorine and carbon.
Acknowledgments
This work was supported by Samsung Electronics and the
projects of System IC 2010 and National Research Laboratory.
Korea Advanced Institute of Science and Technology assisted in meeting
the publication costs of this article.
References
1. S. P. Murarka, Mater. Sci. Eng., R19, 87 ͑1997͒.
. A. Jain, K. M. Chi, T. T. Kodas, and M. J. Hampden-Smith, J. Electrochem. Soc.,
40, 1434 ͑1993͒.
. F. B. Kaufman, D. B. Thompson, R. E. Broadie, M. A. Jaso, W. L. Guthrie, D. J.
Pearson, and M. B. Small, J. Electrochem. Soc., 138, 3460 ͑1991͒.
. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
Res. Dev., 42, 567 ͑1998͒.
. K. C. Shim, H. B. Lee, O. K. Kwon, H. S. Park, W. Koh, and S. W. Kang, J.
Electrochem. Soc., 149, G109 ͑2002͒.
6. O. K. Kwon, H. B. Lee, H. S. Park, and S. W. Kang, J. Vac. Sci. Technol. A, 20͑2͒,
408 ͑2002͒.
. S. Q. Wang, MRS Bull., 19, 15 ͑1994͒.
. O. K. Kwon, J. H. Kim, H. S. Park, and S. W. Kang, Unpublished results.
. R. Kr o¨ ger, M. Eizenberg, D. Cong, N. Yoshida, L. Y. Chen, S. Ramaswami, and D.
Carl, J. Electrochem. Soc., 146, 3248 ͑1999͒.
2
3
4
5
1
7
8
9
1
1
1
0. R. Solanki and B. Pathangey, Electrochem. Solid-State Lett., 3, 479 ͑2000͒.
1. A. Furuyama, N. Hosoi, and Y. Ohsihita, J. Appl. Phys., 78, 5989 ͑1995͒.
2. T. Hara and K. Sakata, Electrochem. Solid-State Lett., 4, G77 ͑2001͒.
13. Y. S. Kim and Y. Shimogaki, J. Vac. Sci. Technol. A, 19, 2642 ͑2001͒.
14. S. Gandikota, S. Voss, R. Tao, A. Duboust, D. Cong, L. Chen, S. Ramaswami, and
D. Carl, Microelectron. Eng., 50, 547 ͑2000͒.
Conclusions
1
1
1
5. L. Peters, Semicond. Int., 23„7…, 108 ͑2001͒.
6. J. Chae, H. S. Park, and S. W. Kang, Electrochem. Solid-State Lett., 5, C64 ͑2002͒.
7. J. A. Cunningham, Semicond. Int., April, 95 ͑2000͒.
Ruthenium thin films were successfully deposited by ALD using
Ru(EtCp)2 and oxygen at 270°C. The relative ratio of adsorbed
Downloaded on 2014-11-17 to IP 128.82.252.58 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).