PEALD of a ruthenium adhesion layer for copper interconnects

Article abstract of DOI:10.1149/1.1809576

Ruthenium thin films were produced by plasma-enhanced atomic layer deposition (PEALD) using an alternating supply of bis(ethylcyclopentadienyl)ruthenium [Ru(EtCp)₂] and NH₃ plasma at a deposition temperature of 270°C. The film thickness per cycle was self-limited at 0.038 nm/cycle, which was thinner than the thickness obtained from the conventional ALD using oxygen instead of NH₃ plasma. The ruthenium thin film prepared with PEALD had a preferential orientation toward (002), and it was progressively promoted with NH₃ plasma power. The PEALD of ruthenium shows a merit in controlling ultrathin film thickness with less than 2 nm more precisely and more easily than the conventional ALD, due to the reduced transient period at the initial film growth stage. Also, ruthenium thin film improved the interfacial adhesion of metallorganic chemical vapor deposited copper to diffusion barrier metals by forming Cu-Ru chemical bonds at the interface without degrading the film resistivity of copper.

Full text of DOI:10.1149/1.1809576

Journal of The Electrochemical Society, 151 ͑12͒ C753-C756 ͑2004͒
C753
0013-4651/2004/151͑12͒/C753/4/$7.00 © The Electrochemical Society, Inc.
PEALD of a Ruthenium Adhesion Layer for Copper
Interconnects
,z
Oh-Kyum Kwon,a Se-Hun Kwon,a Hyoung-Sang Park,b and Sang-Won Kanga,
*
aDepartment of Materials Science and Engineering, Korea Advanced Institute of Science and Technology,
Yusong-gu, Taejon 305-701, Korea
bGenitech, Incorporated, Taedok-gu, Taejon 306-230, Korea
Ruthenium thin films were produced by plasma-enhanced atomic layer deposition ͑PEALD͒ using an alternating supply of
bis͑ethylcyclopentadienyl͒ruthenium Ru͑EtCp) ] and NH plasma at a deposition temperature of 270°C. The film thickness per
͓
2
3
cycle was self-limited at 0.038 nm/cycle, which was thinner than the thickness obtained from the conventional ALD using oxygen
instead of NH3 plasma. The ruthenium thin film prepared with PEALD had a preferential orientation toward ͑002͒, and it was
progressively promoted with NH3 plasma power. The PEALD of ruthenium shows a merit in controlling ultrathin film thickness
with less than 2 nm more precisely and more easily than the conventional ALD, due to the reduced transient period at the initial
film growth stage. Also, ruthenium thin film improved the interfacial adhesion of metallorganic chemical vapor deposited copper
to diffusion barrier metals by forming Cu-Ru chemical bonds at the interface without degrading the film resistivity of copper.
© 2004 The Electrochemical Society. ͓DOI: 10.1149/1.1809576͔ All rights reserved.
Manuscript submitted March 17, 2004; revised manuscript received April 29, 2004. Available electronically October 28, 2004.
With the increasing performance and packing density required in
microelectronic devices, copper has been used as an interconnection
metal to enhance operating speed and reliability.1 Electrochemical
plating ͑ECP͒ has now become a routine process for forming copper
damascene interconnects, due to its super-filling ability, i.e., it fills
trenches and via holes in a bottom-up fashion without any seams or
voids.2 To realize the bottom-up filling processes, a thin and confor-
mal copper seed-layer is required. Currently, physical vapor deposi-
tion ͑PVD͒ technique is used to deposit the seed layer. However, one
challenge is to find a way to deposit a uniform seed-layer in high
aspect ratio via holes and trenches of damascene structures as the
feature size decreases. To overcome this technological barrier, stud-
ies have examined seedless copper damascene filling with ECP or
metallorganic chemical vapor deposition ͑MOCVD͒, in which cop-
per is deposited directly on copper diffusion barrier metals, and
copper seed-layer formation using MOCVD or atomic layer deposi-
tion ͑ALD͒ have been studied.3-5 However, the copper films fabri-
cated using ECP, MOCVD, and ALD techniques have poor interfa-
cial adhesion to common copper diffusion barrier metals, such as
TiN, TaN, and WN, which causes failure of the copper interconnect
system.6,7 Therefore, an adequate adhesion layer for reliable adhe-
sion is required in utilizing the seedless copper bottom-up filling
techniques or preparing the copper seed-layers with MOCVD or
ALD.
Experimental
Ruthenium thin films were deposited on 40 nm thick TiN/100 nm
thick SiO2 /Si wafers at a deposition temperature and pressure of
270°C and 3 Torr, respectively. Ru͑EtCp)2 and NH3 plasma were
used as the ruthenium precursor and reducing agent, respectively.
Ru͑EtCp)2 is a liquid compound with a relatively high vapor pres-
sure of 0.18 Torr at 80°C and good thermal stability. The Ru͑EtCp)2
was contained in a bubbler, which was heated at 80°C, and carried
by argon gas at a flow rate of 50 sccm. To prevent precursor con-
densation in the feeding line, the stainless steel feeding line was
heated to 90°C. One deposition cycle of ruthenium PEALD included
four consecutive pulses: a pulse of Ru͑EtCp)2 vapor, a purge pulse
with argon gas at a flow rate of 50 sccm, a pulse for NH3 plasma
exposure, and a second purge pulse with 50 sccm argon. Each purge
pulse lasted for 10 s. During the NH3 plasma pulse, NH3 gas was
introduced at a flow rate of 100 sccm mixed with 100 sccm argon.
To ignite and sustain the NH3 plasma synchronized with the depo-
sition cycle, a rectangular shaped electrical power was applied be-
tween the upper and lower electrodes. The electrodes were capaci-
tively coupled with a radio frequency ͑rf, 13.56 MHz͒ plasma
source, operated at powers from 20 to 100 W. The lower electrode
was grounded and the showerhead used to distribute the reactant
gases in the reactor uniformly was used as the upper electrode.
Figure 1 shows a schematic diagram of the ruthenium PEALD sys-
tem. Film thickness was analyzed using field-emission scanning
electron microscopy ͑FESEM͒. Film resistivity was calculated from
the sheet resistance measured with a four-point probe and the film
thickness. X-ray diffraction ͑XRD͒ analysis using Cu K␣ radiation
(␭ ϭ 1.5405 Å͒ was used to determine the crystal structure of the
deposited films. The film density was measured using 2.4 MeV
He2ϩ Rutherford backscattering spectroscopy ͑RBS͒. To evaluate
the feasibility of PEALD ruthenium film as an adhesion layer for
MOCVD copper films, a peel-off test using 3M Scotch tape was
performed on a sample ͑ϳ1 ␮m thick MOCVD Cu/Ru adhesion
layer/40 nm thick TiN/100 nm thick thermal SiO2 stacked on Si
substrate͒. The MOCVD of copper was carried out for 10 min at a
temperature of 180°C and a pressure of 5 Torr using hexafluoro-
acetylacetonate-copper-vinyltrimethylsilane (hfac͒CuI͑vtms) as
A ruthenium thin film prepared by ALD using bis͑ethylcyclopen-
tadienyl͒ruthenium Ru͑EtCp) ] and oxygen gas was reported to
͓
2
have excellent performance as an adhesion layer to improve the
interfacial adhesion of MOCVD copper to TiN.8 However, the oxy-
gen gas used as a reactant can form an interfacial metal oxide film
with under-layered diffusion barrier metals, such as Ta and TaN.9-11
This causes interconnection failure or increases via-hole contact re-
sistance. To address this issue, we have been investigating a plasma-
enhanced atomic layer deposition ͑PEALD͒ of ruthenium thin films
with NH3 plasma as a reducing agent, instead of oxygen gas. Our
initial results showing that the PEALD technique enabled ruthenium
film deposition, while avoiding the formation of an interfacial metal
oxide, were reported recently.12 In this paper, we present further
results, including the difference in the saturated film thickness/cycle
and the film growth in the transient region between the PEALD
and the conventional ALD. In addition, we report that the ultrathin
͑ϳ2 nm͒ ruthenium adhesion layer prepared by PEALD greatly im-
proves the adhesion strength of MOCVD copper to diffusion barrier
metal by forming Cu-Ru bonds.
͓
͔
the copper precursor. The interaction between the MOCVD copper
film and the PEALD ruthenium film was analyzed using X-ray pho-
toelectron spectroscopy ͑XPS͒.
Results and Discussion
Until now, most of the MOCVD or ALD in forming ruthenium or
ruthenium oxide thin films has been utilized oxygen gas, and the
* Electrochemical Society Active Member.
z E-mail: swkang@kaist.ac.kr
Downloaded on 2014-10-14 to IP 155.97.178.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C754
Journal of The Electrochemical Society, 151 ͑12͒ C753-C756 ͑2004͒
Figure 1. A schematic diagram of the deposition apparatus used in the ru-
thenium PEALD system.
Figure 3. The number of ruthenium atoms/cycle, which calculated from the
thickness/cycle and the film density measured by RBS analysis, in the
PEALD and conventional ALD.
film to be deposited was determined depending on the partial pres-
sure of oxygen.8,13,14 However, as mentioned before, oxygen is not a
proper choice in forming ruthenium thin films which is to be applied
for an adhesion layer, since oxygen has a potential to make an in-
terfacial metal oxide layer, such as titanium or tantalum oxides, by
oxidizing copper diffusion barrier metals. To avoid this problem,
NH3 plasma was used instead of oxygen. Figure 2 plots the film
thickness deposited for a cycle ͑the film thickness/cycle͒ of ruthe-
nium thin films grown by PEALD using NH3 plasma and by con-
ventional ALD using oxygen depending on the Ru͑EtCp)2 pulse
time at a deposition temperature of 270°C. The pulse time of
Ru͑EtCp)2 was varied from 2 to 10 s. The NH3 plasma exposure
time in the PEALD and the oxygen pulse time in the ALD were
fixed at 10 s, respectively. In the PEALD, the film thickness/cycle
increased with the Ru͑EtCp)2 pulse time and became saturated at
about 0.038 nm/cycle when the Ru͑EtCp)2 pulse time exceeded 5 s,
which is one of the typical characteristics of ALD, so-called a self-
limited reaction. The resistivity of the ruthenium film was constant
at about 12 ␮⍀ cm, and was independent on the pulse time of
Ru͑EtCp)2 . These results show that NH3 plasma successfully re-
duces the metal-organic precursor of Ru͑EtCp)2 into elementary ru-
thenium. It is reasonable to postulate that the lone pair of electrons
in the NH3 interacts with the ruthenium in the Ru͑EtCp)2 and weak-
ens the Ru-͑EtCp͒ bond, and that the excited species in NH3 plasma
activate the reducing reaction by making volatile compounds with
the ligands in Ru͑EtCp)2 . On the other hand, the saturated film
thickness/cycle with the PEALD, 0.038 nm/cycle, was much thinner
than that of the conventional ALD using oxygen, 0.15 nm/cycle. To
elucidate the cause of the difference in the saturated film thickness/
cycle between the PEALD and the conventional ALD, the actual
number of ruthenium atoms was calculated from the film thickness/
cycle and the film density measured from RBS analysis. The film
density increased from 8.7 g/cm3 in the conventional ALD to 12.03
g/cm3 in the PEALD. The actual number of ruthenium atoms of the
PEALD was less than 40% as compared to those of the conventional
ALD. The calculated results are plotted in Fig. 3. The difference can
be understood by adopting the concept of secondary adsorption.15,16
During the oxygen pulse in conventional ALD, oxygen will react
with Ru͑EtCp)2 adsorbed on the film surface and forms a new ru-
thenium film, and at the same time the supplementary oxygen is
likely to adsorb on the newly formed ruthenium film surface. The
adsorption of the supplementary oxygen is termed the secondary
adsorption. As a result, the secondary adsorbed oxygen will make an
additional ruthenium film to react with Ru͑EtCp)2 molecules sup-
plied by the following pulse of Ru͑EtCp)2 . The effect of secondary
adsorption of oxygen was verified by removing the secondary ad-
sorbed oxygen with hydrogen plasma. A pulse for hydrogen
plasma was added after every oxygen pulses during the film depo-
sition. In this case, the film thickness/cycle reduced drastically as
compared to the ALD with oxygen only, and the actual number of
ruthenium atoms formed for a cycle decreases to a similar level of
PEALD, as shown by a dashed line in Fig. 3. In the ruthenium
PEALD, even though NH3 remains on the film surface after the
pulse of NH3 plasma as a result of secondary adsorption, the re-
mained NH3 cannot contribute to forming an additional ruthenium
film, because NH3 itself cannot decompose Ru͑EtCp)2 at a deposi-
tion temperature of 270°C without using plasma. These results lead
to a conclusion that the absence of the effect caused from the sec-
ondary adsorption in PEALD leads to decreasing the actual number
of Ru͑EtCp)2 molecules consumed effectively to constitute ruthe-
nium film in a cycle. As a consequence, the saturated film thickness/
cycle of PEALD reduces as compared to that of the conventional
ALD of ruthenium.
Figure 2. The dependence of the thickness/cycle of ruthenium thin films
grown by PEALD using NH3 plasma and conventional ALD using oxygen
on the Ru͑EtCp)2 pulse time. The pulse time of Ru͑EtCp)2 was varied from
2 to 10 s. The NH3 plasma exposure time in the PEALD and the oxygen
pulse time in ALD time were fixed at 10 s, respectively.
Figure 4a depicts the dependence of the film thickness/cycle on
plasma pulse time and power. At a plasma power of 100 W, the film
Downloaded on 2014-10-14 to IP 155.97.178.73 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 ͑12͒ C753-C756 ͑2004͒
C755
Figure 4. ͑a͒ Dependence of the film thickness/cycle on plasma pulse time
for plasma power of 100 and 20 W. ͑b͒ XRD patterns obtained from the
ruthenium films grown by PEALD and conventional ALD.
Figure 5. Thickness of ruthenium films as a function of deposition cycles in
͑a͒ the PEALD and ͑b͒ the conventional ALD.
ventional ALD. This means there exits an extended period having
lower growth rate in the early stage of the film growth in the con-
ventional ALD. Generally, this early stage is called as transient re-
gion. Aaltonen reported a similar behavior with a long transient
region in the ruthenium ALD using oxygen,14 and insists that the
film growth retards until abundant ruthenium nuclei are formed,
since the dissociative adsorption of oxygen, which leads the oxida-
tive decomposition of adsorbed Ru͑EtCp)2 , is probably facilitated
on ruthenium nuclei than on the original substrate surface. On the
contrary, in ruthenium PEALD, the energetic reducing agent of NH3
plasma can directly decompose the adsorbed ruthenium precursor
without having a dependence on the substrate surface. Thus, a
longer transient region is appeared on the ALD using oxygen than
on the PEALD using NH3 plasma. In a previous paper, we reported
that the ruthenium film deposited by the PEALD has an improved
surface roughness as compared with that prepared by the conven-
tional ALD.12 From the above analysis, it is expected that the ab-
sence of transient region in the PEALD makes the ruthenium film
have a smoother surface than the film prepared by the ALD. Also,
the film smoothness will be helpful to form a continuous and ultra-
thin ͑Ͻ2 nm͒ ruthenium film.
thickness/cycle was saturated when the plasma pulse time exceeded
5 s. That is, a 5 s NH3 plasma pulse is sufficient to completely
achieve the reduction of adsorbed Ru͑EtCp)2 . When the plasma
power was decreased to 20 W, the plasma pulse time to complete the
reduction reaction elongated to 20 s. Figure 4b shows the XRD
patterns obtained from ruthenium films prepared with PEALD and
ALD. In the ruthenium film prepared by the ALD with oxygen, it
was preferentially oriented to ͑101͒. But, the preferred orientation of
the film prepared by the PEALD with NH3 plasma was changed
from ͑101͒ to ͑002͒ with increasing NH3 plasma power. This implies
that the additional energy supplied by NH3 plasma enhances the
rearrangement of ruthenium atoms to the orientation of ͑002͒, the
most thermodynamically stable face in a hexagonal close-packed
͑hcp͒ crystal structure.
In Fig. 5, the ruthenium film thickness is plotted against the
number of deposition cycles for the PEALD and ALD. As expected,
the film thickness increased linearly with the number of deposition
cycles, regardless of the PEALD and ALD. However, the extrapo-
lated line did not pass through the origin, and it intercepted on x-axis
with a perceptible positive value especially for the case of the con-
Downloaded on 2014-10-14 to IP 155.97.178.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C756
Journal of The Electrochemical Society, 151 ͑12͒ C753-C756 ͑2004͒
such as by increasing the resistance of the copper layer, despite
Table I. Results of adhesion test of 1 ␮m thick MOCVD copper.
exhibiting excellent adhesion. Therefore, in addition to its adhesion
properties, stability with the copper layer is important for utilizing
ruthenium thin films as an adhesion layer. To verify the stability of
ruthenium with a copper layer, the sheet resistance of a copper layer
deposited on a ruthenium adhesion layer has been investigated be-
fore and after a heat treatment. Heat-treatment was carried out using
a rapid thermal process ͑RTP͒ at a temperature of 450°C for 1 h in
argon ambient. The sheet resistance of the copper layer is not
changed after the heat-treatment, as ruthenium is immiscible with
copper thermodynamically. Therefore, ruthenium films grown by
PEALD can be used as an effective adhesion layer between
MOCVD copper and copper diffusion barrier metals.
Cycles
None
—
20 cycle
0.73 nm
40 cycle
1.11 nm
60 cycle
1.88 nm
80 cycle
2.64 nm
Thickness of
Ru adhesion
layer
Results of
adhesion test
Fail
Fail
76%
Pass
100%
Pass
100%
Pass
To evaluate the feasibility of PEALD ruthenium thin films as an
adhesion layer to improve the interfacial adhesion of MOCVD cop-
per to copper diffusion barrier metals, a peel-off adhesion test using
Scotch tape was performed after MOCVD of about 1 ␮m thick
copper on a ruthenium adhesion layer. Before the scotch tape peel-
off test, the crossed grid with 2 mm intervals was scratched on the
specimen with a cutter to quantitatively analyze the adhesion prop-
erty, so that the results of the adhesion test are expressed as the ratio
of number of the remained grids after adhesion test to the total
number of grids. Because the adhesion layer should be as thin as
possible, we carried out adhesion tests while progressively decreas-
ing the thickness of ruthenium layer. A 2 nm thick ruthenium adhe-
sion layer is sufficient to improve the interfacial adhesion of
MOCVD copper to the diffusion barrier metals without any failure
in the peel-off adhesion test, as summarized in Table I. To determine
the mechanism responsible for the increased adhesion strength, XPS
analysis was performed. As shown in Fig. 6, the copper atoms at the
interface between copper and ruthenium are electronically perturbed
with respect to the bulk copper atoms, and shift towards a lower
binding energy in the Cu 2p3/2 XPS spectra, while Cu 2p3/2 XPS
spectra of interface copper on TiN is not shifted. These results indi-
cate that a chemical interaction between Cu and Ru is anticipated at
the interface and, therefore, the interfacial adhesion of MOCVD
copper to the diffusion barrier metal is improved.
Conclusions
Ruthenium thin films were successfully deposited by PEALD
using Ru͑EtCp)2 and NH3 plasma at 270°C. NH3 plasma acted as an
effective reducing agent for Ru͑EtCp)2 . The ruthenium film formed
during one deposition cycle was saturated at 0.038 nm/cycle, and its
resistivity was 12 ␮⍀ cm. The decrease in the film thickness/cycle
compared with conventional ALD was attributed to film densifica-
tion and the decrease in the number of atoms deposited per cycle
due to the absence of the effect caused from the secondary adsorp-
tion in PEALD. The preferred orientation of the PEALD ruthenium
film was changed from the ͑101͒ peak to the ͑002͒ peak with in-
creasing NH3 plasma power. The transient region at the early stage
of film growth was even shorter in PEALD than in conventional
ALD, which facilitates the precise thickness control of ultrathin
films. A peel-off adhesion test confirmed that an ultrathin ruthenium
adhesion layer prepared by PEALD greatly improved the interfacial
adhesion of MOCVD copper to TiN barrier metal. The formation of
localized Cu-Ru chemical bonds at the interface enhances adhesion.
Therefore, a PEALD ruthenium thin film can be effectively utilized
as an adhesion layer.
Acknowledgments
For a given adhesion layer, there is a trade-off between stability
and adhesion performance with copper. That is, if a thin film layer
does not react with copper at all, it may exhibit excellent stability,
but poor adhesion. Conversely, if the layer reacts too easily with
copper, it may degrade the performance of the copper interconnect,
This work was supported by the project of National Research
Laboratory ͑NRL͒.
The Korea Advanced Institute of Science and Technology assisted in
meeting the publication costs of this article.
References
1. S. P. Murarka, Mater. Sci. Eng., R., 19, 87 ͑1997͒.
2. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
Res. Dev., 42, 567 ͑1998͒.
3. O. K. Kwon, J. H. Kim, H. S. Park, and S. W. Kang, Electrochem. Solid-State Lett.,
6, C109 ͑2003͒.
¨
4. R. Kroger, M. Eizenberg, D. Cong, N. Yoshida, L. Y. Chen, S. Ranaswami, and D.
Carl, J. Electrochem. Soc., 146, 3248 ͑1999͒.
5. R. Solanki and B. Pathangey, Electrochem. Solid-State Lett., 3, 479 ͑2000͒.
6. A. Furuyama, N. Hosoi, and Y. Ohsihita, J. Appl. Phys., 78, 5989 ͑1995͒.
7. T. Hara and K. Sakata, Electrochem. Solid-State Lett., 4, G77 ͑2001͒.
8. O. K. Kwon, J. H. Kim, H. S. Park, and S. W. Kang, J. Electrochem. Soc., 151,
G109 ͑2004͒.
9. Z. Wang, H. Sakaue, S. Shingubara, and T. Takahagi, Jpn. J. Appl. Phys., Part 1,
42, 1843 ͑2003͒.
10. H. J. Sun, K. M. Kim, Y. Kim, K. J. Cho, K. S. Park, J. M. Lee, and J. S. Roh, Jpn.
J. Appl. Phys., Part 1, 42, 582 ͑2003͒.
11. H. G. Tompkins and J. A. Sellers, J. Vac. Sci. Technol. A, 12, 2446 ͑1994͒.
12. O. K. Kwon, S. H. Kwon, H. S. Park, and S. W. Kang, Electrochem. Solid-State
Lett., 7, C46 ͑2004͒.
13. S. E. Park, H. M. Kim, K. B. Kim, and S. H. Min, Thin Solid Films, 341, 52
͑1999͒.
´
¨
14. T. Aaltonen, P. Alen, M. Ritala, and M. Leskela, Chem. Vap. Deposition, 9, 45
͑2003͒.
15. W. S. Jeon, S. Yang, C. S. Lee, and S. W. Kang, J. Electrochem. Soc., 149, C306
͑2002͒.
Figure 6. XPS analysis of MOCVD copper on ruthenium and TiN.
16. J. W. Lim, H. S. Park, and S. W. Kang, J. Appl. Phys., 88, 6327 ͑2000͒.
Downloaded on 2014-10-14 to IP 155.97.178.73 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).