Atomic layer deposition of Ru thin films using 2,4-(dimethylpentadienyl) (ethylcyclopentadienyl)Ru by a liquid injection system

Article abstract of DOI:10.1149/1.2403081

Ru thin films were grown on Si, Si O2, Ti O2, and TiN substrates by atomic-layer deposition using 2,4-(dimethylpentadienyl)(ethylcyclopentadienyl)Ru and O2 as Ru precursor and reactant, respectively, at temperatures ranging from 230 to 280°C. A saturated growth rate of 0.04 nmcycle and a low oxygen concentration (below the detection limit of Auger electron spectroscopy) were obtained at 250°C. The Ru film showed a negligible incubation period on all the different types of substrates, and an active nucleation behavior which resulted in a very smooth film surface morphology. The dimethylpentadienyl ligand enhanced the active nucleation of Ru and retarded the oxidation of the grown Ru layer. Good step coverage (90%) was obtained from the Ru film grown on a capacitor hole with an aspect ratio of 17 and an opening diameter of 150 nm by a proper control of the Ar carrier gas flow rate.

Full text of DOI:10.1149/1.2403081

Journal of The Electrochemical Society, 154 ͑2͒ D95-D101 ͑2007͒
D95
0013-4651/2007/154͑2͒/D95/7/$20.00 © The Electrochemical Society
Atomic Layer Deposition of Ru Thin Films Using
2,4-(Dimethylpentadienyl)(ethylcyclopentadienyl)Ru
by a Liquid Injection System
Seong Keun Kim,a Sang Young Lee,a Sang Woon Lee,a Gyu Weon Hwang,a
b
b
Cheol Seong Hwang,a,z, Jin Wook Lee, and Jaehack Jeong
*
aSchool of Materials Science and Engineering and Interuniversity Semiconductor Research Center, Seoul
National University, Seoul 151-744, Korea
bQuros Company, Sungnam, Kyunggi-do 462-120, Korea
Ru thin films were grown on Si, SiO2, TiO2, and TiN substrates by atomic-layer deposition using 2,4-
͑dimethylpentadienyl͒͑ethylcyclopentadienyl͒Ru and O2 as Ru precursor and reactant, respectively, at temperatures ranging from
230 to 280°C. A saturated growth rate of 0.04 nm/cycle and a low oxygen concentration ͑below the detection limit of Auger
electron spectroscopy͒ were obtained at 250°C. The Ru film showed a negligible incubation period on all the different types of
substrates, and an active nucleation behavior which resulted in a very smooth film surface morphology. The dimethylpentadienyl
ligand enhanced the active nucleation of Ru and retarded the oxidation of the grown Ru layer. Good step coverage ͑Ͼ90%͒ was
obtained from the Ru film grown on a capacitor hole with an aspect ratio of 17 and an opening diameter of 150 nm by a proper
control of the Ar carrier gas flow rate.
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2403081͔ All rights reserved.
Manuscript submitted July 3, 2006; revised manuscript received September 25, 2006. Available electronically January 4, 2007.
The deposition of Ru thin films either by metallorganic chemical
vapor deposition ͑MOCVD͒ or atomic-layer deposition ͑ALD͒ has
been extensively studied in recent years using various types of pre-
cursors including carbonyl, ␤-diketonates, and cyclopentadienyl
compounds of Ru for its application in capacitors of dynamic ran-
dom access memory ͑DRAM͒ or as diffusion barrier in
Cu-metallization.1-10 For both applications, the conformal deposition
over severe three-dimensional structures with thin and uniform mi-
crostructures is the essential factor. MOCVD and ALD have been
considered as the proper process for Ru thin-film deposition for
these applications. As DRAM devices scale down, a better step cov-
erage over severe three-dimensional contact hole structure ͑aspect
ratio Ͼ15͒ and smaller thickness of the Ru film ͑Ͻ10 nm͒ are re-
quired. Although MOCVD appears to have several merits such as
higher growth rate, ALD becomes crucial for satisfying these re-
quirements. ALD has the merits of a better step coverage and thick-
ness controllability, although it has its own demerits such as a lower
growth rate. However, as the necessary thickness of the Ru film
becomes smaller, Ͻ10 nm, and faster processing technology, such
as semibatch type deposition systems, is developed, ALD appears to
be more proper for the Ru film deposition process.
et al. used NH3 plasma instead of oxygen as reactant in their
plasma-enhanced ALD ͑PEALD͒ process of Ru film to avoid this
problem. They showed that Ru films with a very low impurity con-
centration and a reasonable saturated growth rate ͑0.038 nm/cycle͒
can be obtained.15 However, the PEALD required a complex hard-
ware for the plasma power application, and step coverage was not
reported.
In thermal MOCVD or ALD ͑without plasma͒ the impurity con-
centration is dependent on the precursors. The research group at
Helsinki University reported that the Ru ALD using Ru͑thd͒ and
3
Ru͑Cp͒2 and air as reactant showed a rather different growth behav-
ior and residual impurities even with the same reactors and under
almost identical deposition conditions.4,6 The saturated ALD growth
rate ͑Gsrat͒ was almost identical ͑ϳ0.04 nm/cycle͒. They also re-
ported very comprehensive works on the ALD mechanism for Ru
films using a quadruple mass spectrometer and quartz crystal
microbalance.4,10 It was concluded that oxygen was adsorbed on the
grown Ru surface which reacted with the incoming Ru precursors
during the subsequent Ru pulse step. The reaction reduced the
Ru͑O͒ film by liberating H2O and CO2. During the oxygen pulse
step the remaining ligands of the adsorbed Ru precursor reacted with
oxygen producing H2O and CO2 as reaction by-products as well. It
was clearly shown that oxygen existed at the subsurfaces of the Ru
film during ALD, although it was mostly removed during the sub-
sequent steps.10 However, it is also probable that the removal of the
oxygen is not complete depending on the process conditions ͑short
reaction cycle͒ and precursors ͑less reactive precursor͒.
MOCVD and ALD of Ru films are preceeded by oxidative de-
composition of ␤-diketonate, such as Ru͑C11H19O2͒3 ͓Ru͑thd͒ ͔,4,5
3
and cyclopentadienyl Ru precursors, such as Ru͑C5H5͒2 ͓Ru͑Cp͒ ͔
2
͑Ref. 6 and 10͒ and Ru͑C2H5C5H4͒2 ͓Ru͑EtCp͒ ͔.11-13 Although
2
2,3
carbonyl-based precursors, such as Ru3͑CO͒ , are used to grow
12
Ru films without the help of oxidant, the resulting films are usually
impure, which makes them less suitable. Oxygen plays a key role in
the Ru film deposition in both MOCVD and ALD using
␤-diketonate and cyclopentadienyl as Ru precursors. The removal of
the ligands is certainly improved by the addition of oxygen, whereas
the oxidation of deposited Ru is limited to a low degree by kinetic
reasons.5,11 Most of the supplied oxygen is consumed by oxidation
of C and H derived from the ligand because they have a higher
oxidation potential than Ru.5,11 When the precursor is delivered as a
solution form in the liquid delivery technique, as in this experiment,
the oxygen consumption by the solvent is even more serious. The
residual oxygen in Ru films induces several serious problems in the
integrated devices, especially during high temperature ͑Ͼ600°C͒
postdeposition processes: oxidation of underlying barrier layers and
deterioration of the structural stability in DRAM capacitors.14 Kwon
Another concern about the Ru MOCVD or ALD is the difficult
nucleation of the film on SiO2 and TiN, which are the two contact-
ing surfaces of Ru layers in integrated structures. This was observed
both in ALD using Ru͑Cp͒ ͑Ref. 6͒ and Ru͑thd͒ ͑Ref. 4͒ and
2
3
MOCVD using Ru͑EtCp͒2,12,16 where the retardation of film growth
was observed during the initial period of the deposition process.
This might be understood from the growth mechanism of a Ru film
suggested by the Helsinki group.4,10 The oxidative adsorption and
decomposition of Ru precursor on a predeposited Ru layer is largely
facilitated by the reaction between the ligands of incoming Ru pre-
cursor and adsorbed oxygen on the predeposited Ru layer. If the
oxygen adsorption on the substrate surface is not active, or the do-
nation of the preadsorbed oxygen to the ligands of coming Ru pre-
cursor is limited by the strong bonding between the surface and
oxygen, the Ru nucleation might be suppressed. Once Ru nuclei are
formed the Ru growth mechanism is actively working on the Ru
nuclei and a steady Ru growth rate is achieved. Therefore, it can be
*
Electrochemical Society Active Member.
z E-mail: cheolsh@snu.ac.kr
Downloaded on 2014-12-16 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D96
Journal of The Electrochemical Society, 154 ͑2͒ D95-D101 ͑2007͒
anticipated that the initial growth of Ru on a SiO2 and TiN surface is
dependent on the chemical adsorption or thermal decomposition
property of the Ru precursors. Especially in ALD, where the thermal
decomposition or oxidative decomposition of the Ru precursor is
suppressed, the active adsorption property of the precursor on such a
less-active ͑in terms of the oxygen activity͒ surface is the crucial
factor for obtaining active nucleation and uniform film growth.
It can be anticipated that the chemisorption of Ru͑Cp͒ and
2
Ru͑EtCp͒ on these substrates is not easy considering the stable
2
bonding nature between Ru and Cp ligands due to the ␲-conjugation
of valence electrons in the Cp rings. The preadsorbed oxygen is
necessary to induce the oxidative dissociation of these Cp rings, as
mentioned previously, but the SiO2 and TiN surfaces are not favor-
able for the adsorption and subsequent supply of oxygen to the
incoming Ru͑Cp͒ or Ru͑EtCp͒2.
2
Recently, Shibutami et al. used 2,4-͑dimethylpentadienyl͒
͑ethylcyclopentadienyl͒Ru ͑DER͒ as Ru precursor with the help of
oxygen in the MOCVD process.16 They reported that the MOCVD
with DER involves negligible initial growth retardation compared to
Figure 1. The molecular structure of DER precursor.
the Ru͑EtCp͒ case. It is believed that the dimethylpentadienyl
2
changed from 30 to 500 sccm. The Ar carrier injection time was
also changed from 1 to 5 s. The total amount of precursor that was
supplied to the reactor during the Ru source feeding step was con-
trolled either by the solution injection time or Ar carrier gas injec-
tion time. Typical purge time after the Ru source feeding and O2
pulse steps was 5–8 s, which was long enough to confirm that the
film growth rate was saturated with this purge time.
The film thickness was measured by field-emission scanning
electron microscopy ͑FESEM͒ and cross-sectional high-resolution
transmission electron microscopy ͑HRTEM͒. Due to the relatively
small thickness of the Ru films deposited in this experiment
͑15–35 nm͒, thickness measurement by FESEM had a limited accu-
racy. Therefore, the area density of Ru films measured by X-ray
fluorescence spectroscopy ͑XRF͒ was used to check the saturation
behavior of the film growth. The sheet resistance of Ru thin films
was measured by a four-point probe at room temperature. The crys-
talline structure of the Ru thin films was investigated by X-ray dif-
fraction ͑XRD͒ using Cu K␣ radiation. The impurity concentration,
including oxygen, in the films was investigated by Auger electron
spectroscopy ͑AES͒.
͑DMPD͒ ligand of the DER facilitates the initial adsorption of the
Ru precursor onto the SiO2 and TiN surface which enhances the
subsequent Ru film growth. They compared the thermal decompo-
sition behavior of the two precursors ͓DER and Ru͑EtCp͒ ͔ using
2
differential scanning calorimetry under a N2 atmosphere, and found
that the DER is more vulnerable to thermal decomposition com-
pared to Ru͑EtCp͒ ͑thermal decomposition temperature was lower
2
by ϳ100°C͒. This might be due to the less stable bonding between
the DMPD ligand and Ru ions. It is also probable that once the
DMPD ligand dissociates from the precursor molecule, the bonding
between the remaining EtCp ligand and Ru ions also becomes less
stable and finally the thermal decomposition occurs at a lower tem-
perature. Therefore, it is reasonable to assume that the DER precur-
sor also results in a better deposition property by ALD by reducing
the incubation period.
In this paper, the ALD behavior of Ru films using DER precursor
is reported in detail. The step coverage of the Ru film on a severe
three-dimensional structure ͑hole structure with a diameter of
150 nm and an aspect ratio of 17͒ is investigated considering the
next-generation DRAM application.
Results and discussion
Experimental
Figure 2a and b shows the variations in the area density of a Ru
film deposited on a Si substrate as a function of the Ru source
injection time and Ar carrier gas injection time, respectively, at a
growth temperature ͑Tg͒ of 250°C. Here, the O2 flow rate and pulse
time were 1000 sccm and 3 s, respectively, and the number of depo-
sition cycles ͑ncy͒ was 300. In a and b, the Ar carrier gas injection
time and Ru source injection time were 3 s and 6 ms, respectively.
The area density was independent of the source and carrier gas input
time, suggesting that the ALD process worked under these condi-
tions. Although the Ar carrier gas injection time had a negligible
Ru films were deposited using an 8 in. diameter wafer scale
͑shower head diameter of 12 in.͒ shower-head type ALD reactor
͑Quros. Co., Plus-200͒ at growth temperatures ranging from
230 to 280°C. The ALD system was equipped with a liquid injec-
tion system for feeding Ru precursor into the reactor. DER ͑Tosoh.
Co.͒ precursor dissolved in ethylcyclohexane ͑ECH͒ with a concen-
tration of 0.2 M was used as Ru source. The molecular structure of
DER is shown in Fig. 1. O2 was used as reactant and the O2 gas flow
rate was varied from 300 to 1000 sccm. Ar gas with a flow rate
ranging from 30 to 500 sccm was used as carrier gas, and Ar gas
with a flow rate of 1000 sccm was used as purge gas. ALD TiO2,
sputtered TiN, thermal SiO2 films on Si wafers and bare Si͑100͒
wafers were used as substrates in order to investigate substrate-
dependent deposition properties of metallic Ru films. Also, Ru films
were deposited on a capacitor hole structure formed in SiO2 sub-
strates with an aspect ratio ϳ17 to investigate the step coverage of
Ru films.
A deposition cycle for Ru films consisted of four steps: Ru
source feeding-purge-oxygen feeding-purge. Because a liquid deliv-
ery method was adopted for the Ru solution source, the precursor
solution was pressurized by Ar gas and was fed into the hot vapor-
izer ͑maintained at 200°C͒ through an electrically controlled nozzle.
The amount of precursor solution supply was controlled by chang-
ing the opening time of the nozzle from 6 to 15 ms. The precursor
solution was flash evaporated in the vaporizer and carried into the
reactor with the help of Ar carrier gas, the flow rate of which was
Figure 2. The variations in the area density of a Ru film grown on a Si
substrate as a function of ͑a͒ the Ru source injection time, ͑b͒ Ar carrier gas
injection time.
Downloaded on 2014-12-16 to IP 138.251.14.35 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, 154 ͑2͒ D95-D101 ͑2007͒
D97
Figure 3. The variations in the area density and resistivity of a Ru film
grown on a Si substrate as a function of the O2 gas injection time.
influence on the growth rate ͑Gr͒ of the Ru films on Si substrates,
the flow rate of the Ar carrier gas had a profound effect on the step
coverage on the hole structure, as shown later.
Figure 4. AES depth profile of Ru films grown on Si at a growth tempera-
ture of 250°C with an O2 injection time of ͑a͒ 2, ͑b͒ 3, ͑c͒ 5 s, and ͑d͒
collected oxygen data from the three samples.
Figure 3 shows the variations in the area density of a Ru film
deposited on a Si substrate as a function of the O2 gas injection time,
where the O2 flow rate, Ru source injection time, and Ru carrier gas
injection time were 1000 sccm, 6 ms, and 3 s, respectively. The area
density was almost independent of the O2 gas injection time, rang-
ing from 1 to 5 s, suggesting that the saturated growth behavior of
the ALD Ru process was also confirmed with respect to the O2 flow
rate. The variation of the resistivity of the film calculated from the
sheet resistance and film thickness is also shown in Fig. 3. The
resistivity was saturated to ϳ19 ␮⍀ cm with increasing O2 injec-
tion time. These values are rather high compared to the bulk value of
Ru ͑7.2 ␮⍀ cm͒. However, this is comparable to other reports on
Ru thin films grown by MOCVD or ALD.6,8,17 Considering the
small thickness of the Ru films grown here ͑ϳ10 nm͒, a resistivity
of 19 ␮⍀ cm is a reasonable value. The density of the film, calcu-
lated using the area density from XRF and the thickness gained from
SEM, grown on SiO2 is ϳ9.2 g cm−3, which is ϳ75% of the bulk
value of Ru ͑12.2 g cm−3͒. This relatively low density may consti-
tute one of the reasons for the relatively high resistivity of Ru thin
films. A density of 9.2 g cm−3 is also comparable to other reports.9
The resistivity did not show any abrupt variation with increasing
O2 flow rate. It has been reported that RuOx ͑x ϳ 2͒ films are
formed when the Ru-source/O2 ratio decreases to a certain value in
Although the direct comparison of the present experiment with
other reports where a RuOx phase was observed for a low Ru-source
͓Ru͑EtCp͒ ͔/O2 ratio8 is difficult due to the different geometries of
2
the reactors and other conditions, it is quite certain that the O2 flow
rate and time in this experiment were excessive ͑1000 sccm with
2–5 s vs 10–60 sccm with 10–25 s͒. Considering the similar Gsrat of
both experiments ͑ϳ0.04 nm/cycle͒, it might be reasonable to com-
pare the oxidation behavior of the deposited Ru films based on the
Ru ALD using Ru͑EtCp͒ and O2.8 No increase in the resistivity
2
with increasing O2 injection time up to 5 s is shown in Fig. 3, and
this suggests that RuOx was not formed under these ALD conditions.
This can be further confirmed from the AES results shown in Fig. 4.
Figure 4a-c shows the AES depth profile results of Ru films
grown on Si at a wafer temperature of 250°C with an O2 injection
time of 2, 3, and 5 s, respectively, and d shows the collected O data
from the three samples. The residual oxygen concentration in the Ru
layer was negligible up to an O2 injection time of 3 s. However, it
increased to ϳ2–3% when the O2 injection time increased to 5 s.
The relatively high O concentration at the Ru/Si interface was due
to the presence of native SiO2.
Figure 5a-c shows the AES depth profile results of the Ru films
grown on Si at a wafer temperature of 280°C with an O2 flow rate of
300, 500, and 1000 sccm, respectively, and d shows the collected O
data from the three samples. Here, the O2 injection time was 3 s.
Different from the deposition results at 250°C, all the films con-
tained 1–2% of oxygen. This might be due to the slightly increased
oxidation rate of the grown Ru film. Therefore, the optimum depo-
sition temperature was ϳ250°C.
Figure 5. AES depth profile of Ru films grown on Si at a growth tempera-
ture of 280°C with an O2 flow rate of ͑a͒ 300, ͑b͒ 500, ͑c͒ 1000 sccm, and
͑d͒ collected oxygen data from the three samples.
Downloaded on 2014-12-16 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D98
Journal of The Electrochemical Society, 154 ͑2͒ D95-D101 ͑2007͒
Figure 6. The variations in the growth rate and resistivity of a Ru film as a
function of the growth temperature.
different structures of the Ru precursors ͓Ru͑EtCp͒ vs DER͔. This
2
is discussed later in conjunction with the negligible incubation pe-
riod of the present experiment.
Figure 6 shows the variations in Gr and the resistivity of a Ru
film as a function of the growth temperature. Here, the Ru source,
Ar carrier, and O2 injection times were 6 ms, 3 s, and 3 s, respec-
tively. The growth rate slightly increased with temperature from
0.036 nm/cycle at 230°C to 0.05 nm/cycle at 280°C. The slight in-
crease in Gr might be due to the increase in the activity of the ALD
reaction with increasing temperature. In a report on the MOCVD of
Ru using the same precursor and O2, Shibutami et al. showed that
Gr increases with increasing growth temperature from 260 to 300°C
by almost 10 times.16 Although the investigated growth temperature
region was slightly different, the relatively small change in Gr of the
present experiment suggests that the ALD window was maintained
in this temperature region.
Figure 7 shows the variations in the film thickness ͑measured by
cross-sectional SEM͒ and resistivity as a function of ncy on SiO2,
TiO2, and TiN substrates. Here, the growth temperature, Ru source,
Ar carrier, and O2 gas injection times were 250°C, 6 ms, 3 s, and
3 s, respectively. The resistivity was measured for a film grown on a
SiO2 surface. It can be observed that the film thickness increases
linearly with increasing ncy with an almost identical slope indepen-
dent of the substrates types. The slope corresponds to a growth rate
Figure 8. SEM images of Ru films grown on ͑a͒ SiO2 and ͑b͒ TiO2 sub-
strates.
of 0.04 nm/cycle. There is a negligible incubation period at the ini-
tial stage of the film growth ͑almost zero x-axis intercept of the
linear extrapolated graph͒ on the three substrates. This is a quite
different behavior from the Ru film deposition using Ru͑thd͒3,
Ru͑Cp͒2, or Ru͑EtCp͒ either by ALD or MOCVD where rather
2
long incubation periods were observed on Si, SiO2, and
Al2O3/borosilicate glass substrates.4,6,9,12 Kang et al. reported that in
MOCVD of Ru using Ru͑EtCp͒ and O2 the incubation period of
2
the Ru film growth was rather long on a Si and SiO2 surface but it
was negligible on a TiO2 surface.18 When the incubation period was
long, the nuclei density was low and a rough Ru film was obtained.
They ascribed the substrate dependent film growth to the different
interface energy.18
Figure 8a and b shows the SEM surface morphologies of Ru
films grown on SiO2 and TiO2 substrates, respectively, with varying
ncy. The surface morphology of a Ru film grown on a TiN substrate
was almost identical to that of the film on a TiO2 substrate ͑data not
shown͒. Films as thin as 4 nm ͑100 cycles͒ showed uniform and
very smooth morphologies, suggesting that the nucleation of Ru on
these substrates was quite active. It appeared that the Ru film on
TiO2 was slightly rougher than that on SiO2. This might be due to
the slightly rougher TiO2 compared to SiO2, as shown by the cross-
sectional TEM in Fig. 9.
Figure 9a and b shows the low-magnification and high-resolution
TEM, respectively, of a 20 nm thick Ru ͑ncy = 500͒ film grown on
an 8 nm thick TiO2 film. The TiO2 film was amorphous and Ru was
polycrystalline, with a columnar grain structure. The average grain
diameter appeared to be 10–15 nm. It appeared that there was a
negligible interfacial reaction between the Ru and TiO2 layers. The
Ru film adhesion was quite stable, whereas that on SiO2 was worse
than that on TiO2.
The present results are also consistent with the report on Ru
MOCVD by Shibutami et al., in which they reported the Ru depo-
Figure 7. The variations in the film thickness and resistivity as a function of
ncy on SiO2, TiO2, TiN substrates.
sition behavior on TiN substrates using DER and Ru͑EtCp͒ at 275
2
Downloaded on 2014-12-16 to IP 138.251.14.35 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, 154 ͑2͒ D95-D101 ͑2007͒
D99
the Ru͑EtCp͒ molecules adsorbed on the predeposited Ru surface
2
which actually contained oxygen, they reacted with the oxygen that
was supplied from the predeposited Ru layer. The EtCp ligands be-
come very unstable after dissociation from Ru ions by the disruption
of the ␲-conjugated electron structure. Then, they may instanta-
neously react with oxygen, resulting in CO2 and H2O formation as
shown by the research group at Helsinki University.10 During the O2
supply step a similar reaction may occur between the remaining
EtCp ligand on the film surface and the incoming O2 leaving the
oxygen containing Ru surface again. If the oxygen supply was ex-
cessive an oxidation of the Ru film might be possible under these
circumstances.
When DER precursor was used instead of Ru͑EtCp͒2, the reac-
tion may have proceeded in the following way: During the oxygen
injection step, oxygen reacts with the EtCp ligand because the
DMPD ligand has already been removed in the previous step due to
its less stable bonding with Ru ion in DER. As indicated by Aal-
tonen et al.,10 the adsorbed oxygen starts to penetrate into the sub-
surface after surface saturation has been achieved. The degree of the
oxygen penetration is dependent on the surface oxygen concentra-
tion. In this experiment, the O2 flow rate is excessive ͑1000 sccm͒
so that the amount of adsorbed oxygen may be relatively large.
Then, DMPD ligand may react with the adsorbed oxygen prior to
the EtCp ligand during the subsequent DER injection. The adsorbed
oxygen is consumed during this step. This means that the oxygen
consumption during the ALD using the DER precursor could be
higher than the ALD using the Ru͑EtCp͒2. It was found that RuO2
film growth by MOCVD using DER and O2 ͑up to the O2 flow rate
of 1000 sccm͒ at temperatures ranging from 250 to 350°C was not
possible from the authors’ own experiment. Only Ru films were
obtained. However, RuO2 films were easily deposited using
Ru͑EtCp͒ and O2 with the similar O2 flow rate in the same tem-
2
perature range ͑data not reported͒. This also supports that the oxi-
dation of Ru metal atoms derived from DER is relatively difficult in
ALD.
The Ru film growth itself still might be controlled by the re-
moval of the EtCp ligand if the removal of the DMPD ligand is
easier than that of the EtCp ligand. This can be confirmed by the
similar Gsrat ͑0.04 nm/cycle͒ of ALD Ru films with either DER or
Figure 9. ͑a͒ Low magnification and ͑b͒ high-resolution TEM images of a
20 nm thick Ru film grown on an 8 nm thick TiO2 film.
Ru͑Cp͒2,6,10 and comparable activation energies in the ALD win-
dow ͓0.83 eV for DER in this experiment vs 1.06 eV for
and 350°C. They showed that the DER precursor does not involve
any incubation period but the Ru͑EtCp͒ precursor results in a quite
2
Ru͑Cp͒ ͔.6,10
extensive incubation period at both temperatures.
2
As discussed in detail previously, the ALD of Ru films using
Ru͑thd͒3, Ru͑Cp͒2, and Ru͑EtCp͒2, and O2 proceeds with the oxi-
dative dissociation of the ligands. Because the oxidative dissociation
also proceeds during the Ru-precursor injection periods,10 the sup-
ply of oxygen to the incoming Ru precursor from the predeposited
Ru surface was the key factor. Becasue the Si, SiO2, and TiN sur-
faces are generally not favorable surfaces for oxygen donation, the
ALD of Ru films suffered from poor nucleation at the initial stage of
the film growth.
The negligible incubation period of the present experiment on
the three substrate types suggests that the removal of DMPD ligand
from the DER molecule does not require such an amount of oxygen
supply due to less stable bonding between the DMPD ligand and Ru
ions. However, it is expected that the removal of DMPD ligand does
not proceed by pure thermal decomposition because proper film
growth was not achieved without oxygen. The occurrence of a satu-
rated growth with Ru precursor and O2 supply ͑Fig. 2 and 3͒ also
suggests that the growth is driven by the ALD mechanism and not
by a thermal decomposition mechanism. This suggests that the ALD
of Ru films using DER requires less oxygen compared to
Ru͑EtCp͒2, although the direct comparison was not made using the
same deposition system. This will be the next topic of research.
The use of DER as precursor did not induce RuOx formation and
a lower concentration of residual oxygen even at very high oxygen
doses ͑1000 sccm, 5 s injection time͒. This can be understood from
the different oxygen reactivities of DMPD and EtCp ligands. When
Figure 10a-c shows the XRD patterns of 20 nm thick Ru films
grown on Si, SiO2, TiN, and TiO2 substrates at 230, 250, and
280°C, respectively. The films grown on Si, TiN, and TiO2 sub-
strates show a smaller diffraction peak intensity compared to that on
SiO2, suggesting a higher nucleation density and smaller grain size
on these substrates, although this was not clearly observed by SEM
in Fig. 8. The increase in growth temperature slightly improved the
crystalline quality of the Ru films on TiO2, but this could not be
observed on Si and TiN. On SiO2, c-axis preferential growth be-
come obvious with increasing growth temperature due to the
densely packed structure of the ͑001͒ plane. This is also consistent
with other reports.4,6,9
Finally, the step coverage on the capacitor hole structure with an
aspect ratio of 17 and a top opening diameter of 150 nm formed in
SiO2 was investigated. Figure 11a shows the SEM cross section of
the hole with the Ru layer grown on top under the following growth
conditions; deposition temperature 250°C, Ru source injection time
6 ms, Ar carrier gas flow rate of 30 sccm with an injection time of
3 s, O2 flow rate of 1000 sccm with an injection time of 3 s. A Ru
film step coverage ͑defined by bottom thickness/top thickness͒ of
Ͼ90% was obtained, which is the best value reported to the best of
the authors’ knowledge. Figure 11b shows the magnified images of
top and bottom portion of the hole.
However, the step coverage was not necessarily guaranteed un-
der other process conditions. For the otherwise given conditions, the
Ar carrier gas flow rate had a profound influence on the step cover-
Downloaded on 2014-12-16 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D100
Journal of The Electrochemical Society, 154 ͑2͒ D95-D101 ͑2007͒
Figure 11. ͑a͒ Cross-sectional SEM images of the hole structure with the Ru
layer. The magnified images of the top and bottom portion of the hole with
Ar carrier gas flow rates of ͑b͒ 30, ͑c͒ 100, ͑d͒ 300, ͑e͒ 500 sccm, respec-
tively. ͑f͒ The variation in the step coverage as a function of the Ar carrier
gas flow rate.
coverage severely degraded with an increasing Ar carrier gas flow
rate, although the Ru film thickness on the upper part of the hole
structure was almost identical. At 500 sccm, almost no Ru film was
grown on the bottom area. The negligible change in the film thick-
ness on top and the decreasing film thickness on the bottom with
increasing carrier gas flow rate suggests that the degradation in step
coverage was not due to the change in the sticking property of the
precursor. For the given precursor pulse injection time an increased
Ar carrier gas flow resulted in a decrease in the Ru precursor con-
centration in the gas phase inside the ALD reactor. When the pre-
cursor concentration decreased to a low value it could be lower than
the minimum gas phase concentration that guaranteed Gsrat deep in-
side the hole. Then, Gr decreased on the deep hole region which
resulted in the poor step coverage.
Conclusion
Ru thin films were grown on Si, SiO2, TiO2, and TiN substrates
using DER and O2 as Ru precursor and reactant, respectively, at
temperatures ranging from 230 to 280°C by ALD. The best deposi-
tion temperature appears to be 250°C considering the saturated
growth rate ͑0.04 nm/cycle͒ and low oxygen concentration ͑below
the detection limit of AES͒. DER precursor showed a wider ALD
window in terms of the oxygen dose. Active film growth was
achieved even from the initial state of the deposition irrespective of
the substrate types. Negligible RuO2 was formed even with an ex-
cessive oxygen dose. The film showed a smooth surface, suggesting
that nucleation was active by virtue of the DMPD ligand. A high
enough step coverage ͑Ͼ90%͒ was obtained from the Ru film
grown on a capacitor hole with an aspect ratio of 17 and an opening
diameter of 150 nm by a proper control of the Ar carrier gas flow
rate.
Acknowledgment
The work was supported by the System IC 2010 program and the
0.1 Tbit NVM project of the Korean government. C.S.H. gratefully
acknowledges the supply of DER precursor by Tosoh Co.
Figure 10. XRD patterns of 20 nm thick Ru films grown on Si, SiO2, TiN,
TiO2 substrates at the growth temperature of ͑a͒ 230, ͑b͒ 250, ͑c͒ 280°C.
Seoul National University assisted in meeting the publication costs of this
article.
age as shown in Fig. 11c-e. In Fig. 11c-e, the top and bottom portion
of the hole with Ar carrier gas flow rates of 100, 300, and 500 sccm,
respectively, are shown. Figure 11f shows the variation in the step
coverage as a function of the Ar carrier gas flow rate. The step
References
1. H. Kim, Surf. Coat. Technol., 200, 3104 ͑2006͒.
2. M. L. Green, M. E. Gross, L. E. Papa, K. J. Schnoes, and D. Brasen, J. Electro-
Downloaded on 2014-12-16 to IP 138.251.14.35 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, 154 ͑2͒ D95-D101 ͑2007͒
D101
chem. Soc., 132, 2677 ͑1985͒.
6, C130 ͑2003͒.
3. Q. Wang, J. G. Ekerdt, D. Gay, Y.-M. Sun, and J. M. White, Appl. Phys. Lett., 84,
1380 ͑2004͒.
4. T. Aaltonen, M. Ritala, K. Arstila, J. Keinonen, and M. Leskelä, Chem. Vap. Depo-
sition, 10, 215 ͑2004͒.
5. M. Kadoshima, T. Nabatame, M. Hiratani, Y. Nakamura, I. Asano, and T. Suzuki,
Jpn. J. Appl. Phys., Part 2, 41, L347 ͑2002͒.
6. T. Aaltonen, P. Alén, M. Ritala, and M. Leskelä, Chem. Vap. Deposition, 9, 45
͑2003͒.
7. T. Aaltonen, M. Ritala, Y.-L. Tung, Y. Chi, K. Arstila, K. Meinander, and M.
Leskelä, J. Mater. Res., 19, 3353 ͑2004͒.
8. O.-K. Kwon, J.-H. Kim, H.-S. Park, and S.-W. Kang, J. Electrochem. Soc., 151,
G109 ͑2004͒.
9. O.-K. Kwon, S.-H. Kwon, H.-S. Park, and S.-W. Kang, J. Electrochem. Soc., 151,
C753 ͑2004͒.
11. S. Y. Kang, K. H. Choi, S. K. Lee, C. S. Hwang, and H. J. Kim, J. Electrochem.
Soc., 147, 1161 ͑2000͒.
12. H. Han, J. J. Kim, and D. Y. Yoon, J. Vac. Sci. Technol. A, 22, 1120 ͑2004͒.
13. Y. Matsui, M. Hiratani, T. Nabatame, Y. Shimamoto, and S. Kimura, Electrochem.
Solid-State Lett., 4, C9 ͑2001͒.
14. S. Y. Kang, H. J. Lim, C. S. Hwang, and H. J. Kim, J. Korean Phys. Soc., 44, 25
͑2000͒.
15. O.-K. Kwon, S.-H. Kwon, H.-S. Park, and S.-W. Kang, Electrochem. Solid-State
Lett., 7, C46 ͑2004͒.
16. T. Shibutami, K. Kawano, N. Oshima, S. Yokoyama, and H. Funakubo, Electro-
chem. Solid-State Lett., 6, C117 ͑2003͒.
17. S. Y. Kang, H. J. Lim, C. S. Hwang, and H. J. Kim, J. Electrochem. Soc., 149,
C317 ͑2002͒.
10. T. Aaltonen, A. Rahtu, M. Ritala, and M. Leskelä, Electrochem. Solid-State Lett.,
18. S. Y. Kang, C. S. Hwang, and H. J. Kim, J. Electrochem. Soc., 152, C15 ͑2005͒.
Downloaded on 2014-12-16 to IP 138.251.14.35 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).