Nucleation and growth of pulsed CVD ru films from tricarbonyl [η4 -cyclohexa-1,3-diene]ruthenium

Article abstract of DOI:10.1149/1.2988062

Nucleation and growth of ruthenium films have been studied in temperature ranges of 155-290°C using tricarbonyl [η⁴ -cyclohexa-1,3-diene]ruthenium, ammonia, or nitrous oxide and consecutively pulsed chemical vapor deposition (CVD) conditions. High Ru nuclei surface concentration without nucleation delay was obtained at high deposition temperatures. At low temperatures, surface nucleation was a limiting step of the deposition process. It resulted in a discontinuous film structure with large grains, high surface roughness, and high film resistivity. A shot-time high-temperature pulsed CVD nucleation step and sputtered subnanometer thick metal seed layer allowed significant improvement of ruthenium film morphology.

Full text of DOI:10.1149/1.2988062

Journal of The Electrochemical Society, 155 ͑12͒ D763-D770 ͑2008͒
D763
0
013-4651/2008/155͑12͒/D763/8/$23.00 © The Electrochemical Society
Nucleation and Growth of Pulsed CVD Ru Films
4
from Tricarbonyl[␩ -cyclohexa-1,3-diene]ruthenium
Vladislav Yu. Vasilyev, * Konstantin P. Mogilnikov, and Yong Won Song *
a, ,c,z
b,z
a, ,z
a
Department of Nano-Optical Engineering, Korea Polytechnic University, Gyeonggi-Do, 429-793,
Republic of Korea
b
Institute of Semiconductor Physics, Siberian Branch of Russian Academy of Science, 630099, Novosibirsk,
Russia
Nucleation and growth of ruthenium films have been studied in temperature ranges of 155–290°C using
4
tricarbonyl͓␩ -cyclohexa-1,3-diene͔ruthenium, ammonia, or nitrous oxide and consecutively pulsed chemical vapor deposition
͑
CVD͒ conditions. High Ru nuclei surface concentration without nucleation delay was obtained at high deposition temperatures.
At low temperatures, surface nucleation was a limiting step of the deposition process. It resulted in a discontinuous film structure
with large grains, high surface roughness, and high film resistivity. A shot-time high-temperature pulsed CVD nucleation step and
sputtered subnanometer thick metal seed layer allowed significant improvement of ruthenium film morphology.
©
2008 The Electrochemical Society. ͓DOI: 10.1149/1.2988062͔ All rights reserved.
Manuscript submitted May 7, 2008; revised manuscript received August 22, 2008. Published October 14, 2008.
Thin ruthenium metal film is proposed as one of the promising
materials for micro- and nanoscaled device applications, such as
film growth, the film structure ͑characterized by X-ray diffraction͒,
the film properties ͓characterized by scanning electron microscopy
͑SEM͒, energy-dispersive X-ray analysis ͑EDX͒, X-ray reflectivity
͑XRR͒, scanned probe microscopy ͑SPM͒, sheet resistivity͔ were
presented in Ref. 21-24. The films were deposited using a dedicated
1
electrode materials in metal–insulator–metal capacitors, gate
2
3
4
electrodes, diffusion barriers, and glue layers in copper metalliza-
5
tion technology. Ruthenium oxide ͑RuO ͒ is also conductive,
which is an important feature of Ru metal. Physical vapor deposition
2
2
1
deposition tool and consecutive pulsed mode CVD, which is nec-
essary to achieve monoatomic layer-by-layer thin-film growth. An-
other research group pursued a “classic” continual metallorganic
͑
PVD͒ processes were used to deposit Ru films for integrated circuit
6
technology, and chlorine-based plasma has been used for the film
patterning. However, PVD implementation in three-dimensional
2
5
7
CVD technology approach with the use of the same precursor and
produced encouraging results. Summarizing our observations and
literature data, we came to the conclusion that the poor nucleation
issue and granular type of film growth are similar for studied pre-
cursors. This convinced us that Ru film nucleation features and the
first steps of the film growth need to be investigated in detail. Until
recently, there was a lack of such data, and, specifically, there were
͑
3D͒ device structures with deep and narrow trenches revealed se-
8
vere step coverage limitations. Intensive efforts have been made by
many research groups to find the best conditions for manufacturable
Ru chemical vapor deposition ͑CVD͒ or atomic layer deposition
͑
ALD͒ processes. However, Ru chemistry is complicated and the
best solutions for Ru precursor, technique, and deposition conditions
are still under research.
no such data for Ru͑CO͒ ͓C H ͔ precursor. This article is focused
3
6
8
Different types of ruthenium precursors have been investigated
using thermally activated CVD and ALD methods. For instance,
on the study of Ru film nucleation features and their correlations
with the film morphology and the film resistivity.
5
some examples include triruthenium dodecacarbonyl Ru ͑CO͒ ,
3
12
9
Experimental
␤
-diketonates such as Ru͑C H O ͒ , cyclopentadienyl-based pre-
11 19 2 3
10
21
3
cursors such as Ru͑C H ͒ , and their derivatives. The primary
A deposition tool included a 5 dm cold-wall deposition cham-
5
5 2
concern of previous methods and precursors is that poor Ru nucle-
ation causes a nonsteady film deposition rate. Attempts to improve
Ru nucleation by using plasma enhanced ͑PE͒ ALD conditions
ber ͑DC͒ with a 200 mm wafer heater and heated multichannel
showerhead. The vacuum system of the tool included an Edwards
iQDP 40 dry vacuum pump, shut and throttle valves, and a Baratron
gauge. The vacuum system parameters were set: a base pressure
11,12
have been considered successful, but limited in application. In fact,
rapid recombination of plasma-generated radicals on the substrate
surfaces at PE ALD is a serious issue for the conformal film step
coverage in complicated 3D structures. Thus, thermally activated
deposition processes remain attractive for Ru film deposition for
advanced device applications. In fact, attempts have been made to
−
3
ϳ3 ϫ 10 Torr, the leakage value from ambient ϳ5
ϫ 10− dm ϫ Torr ϫ s , and the fixed DC time constant ϳ1.3 s.
4
3
−1
Details about the ruthenium precursor Ru͑CO͒ ͓C H ͔ can be
3
6
8
found in Ref. 17-21. Argon gas was used for Ru͑CO͒ ͓C H ͔ de-
3
6
8
livery and purge. Standard semiconductor grade ammonia and ni-
trous oxide gases were used as the second reactants. The total depo-
sition process cycle duration was 20 s. This computer-managed
cycle included an 8-step pulse sequence with the following corre-
sponded timing: Step ͑1͒ Ar + Ru͑CO͒ ͓C H ͔ pulse, 1 s; Step ͑2͒
1
3
improve the nucleation by surface pretreatment in plasma, UV and
1
4
15
vapor TiCl4,
O ,
as well as searches for new complicated
advanced properties, such as
bis͑N,NЈ-di-tert-butylacetamidinato͒ruthenium͑II͒ dicarbonyl.
3
precursors
with
1
6
3
6
8
Among the other types of studied precursors, some Ru carbonyl
derivatives are an interesting option. Synthesis of some carbonyl
complexes and successful Ru CVD with cyclohexadiene derivatives
were reported in Ref. 17-20, respectively. We studied Ru films de-
DC evacuation, varied from 1 to 3 s; Step ͑3͒ Ar purge pulse, varied
from 1 to 3 s; Step ͑4͒ DC evacuation, varied from 1 to 3 s; Step
͑
5͒ second reactant pulse, varied from 0 to 3 s; Step ͑6͒ DC evacu-
ation, varied from 1 to 3 s; Step ͑7͒ Ar purge pulse, varied from
4
posited using a liquid-phase tricarbonyl͓␩ -cyclohexa-1,3-diene͔
1
to 3 s; Step ͑8͒ DC evacuation, varied from 1 to 3 s.
To measure the precursor consumption per process cycle, the
ruthenium ͑below Ru͑CO͒ ͓C H ͔͒ and NH or N O as the second
3
6
8
3
2
reactants. The advantage of using Ru͑CO͒ ͓C H ͔ precursor is a
3
6
8
weight loss of the precursor canister was periodically controlled. In
the canister temperature range of 35–55°C, the precursor consump-
tion was 0.7–1.9 mg per pulse ͑for the total cycle duration 20 s,
Ar + Ru͑CO͒ ͓C H ͔ pulse duration 1 s, Ar flow rate 20 sccm͒.
high deposition rate at a low deposition temperature. Ruthenium
3
6
8
*
Electrochemical Society Active Member.
Permanent address: Novosibirsk State Technical University, 630092, Novosibirsk,
This precursor consumption range corresponded to the precursor
c
18
18
dose range of ϳ1.6 ϫ 10 to ϳ4.3 ϫ 10 molecules per one
Russia.
z
Ru͑CO͒ ͓C H ͔ pulse. Substrates with silicon and thermally oxi-
dized silicon surfaces were used without any additional surface
E-mail: vladislavគvគ2000@yahoo.com; mogkonst@thermo.isp.nsc.ru;
3
6
8
ywsong@kpu.ac.kr
Downloaded on 2015-06-25 to IP 134.129.182.74 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D764
Journal of The Electrochemical Society, 155 ͑12͒ D763-D770 ͑2008͒
treatment. The experiments were conducted in a range of substrate
temperatures ͑Tdep͒ of 155–290°C measured at reduced pressures
using a dedicated silicon wafer with nine thermocouples ͑Prime
Technology Co, Ltd., Korea͒. The working pressure of the gas pulse
maximum was ϳ0.8 Torr. The total deposition process duration var-
ied from 1 to 300 cycles.
Film thickness was determined by using a routine SEM cross-
section analysis technique ͑Hitachi microscope, S-4700͒. In addi-
tion, optical characteristics of the Ru films and ruthenium film thick-
ness were determined by a spectroscopic ellipsometer, developed
and produced at the Institute of Semiconductor Physics in Novosi-
birsk, Russia. The spectral range of the ellipsometer was
250–900 nm with wavelength resolution up to 1 nm. The angle of
incidence was 70° and the beam diameter was 3 mm. The typical
precisions of measurement of the ellipsometric angles ␺ and ␦ were
0.02 and 0.05°, respectively. The calculating program allowed us to
use up to six layers to model a sample of investigation. Different
effective media approximation ͑EMA͒ models were used to calcu-
late the optical parameters with this program. An FEI Nova
NanoSEM 200 and a Seiko SPA500 SPM tools were used to evalu-
ate the surface morphology of the films. The sheet resistance ͑RS͒
was collected by a four-point-probe measurement system CMT-
SR2000N from Advanced Instrument Technology, Inc. Fifteen data
points on half of each 200 mm substrate were measured.
Result and Discussion
Basic ruthenium film growth features.— The surface images of
typical Ru films grown at low and high deposition temperatures are
presented in Fig. 1a and b. The granular structure of the films can be
clearly seen. The grain size was smaller with the increase of Tdep and
Ru precursor consumption. Similar grained films were obtained with
the use of N O, NH , as well as without any second reactant ͑i.e.,
2
3
under pulsed metallorganic CVD conditions͒. The film surface
roughness increased at Tdep Ͻ ϳ200°C and it was twice as high for
the samples deposited with N O compared to the samples grown
2
with NH . Also, N O-deposited Ru films showed rounder and
3
2
clearer grain boundaries in SEM images compared to NH -grown
3
samples. More details about Ru films and related XRD, EDX, XRR,
and SPM data can be found in Ref. 22-24 and 26.
Assuming a steady Ru film deposition rate through the pulsed
CVD process, it can be evaluated by dividing the film thickness
value by the number of the deposition cycles. The grains in the Ru
films reduced the accuracy of SEM cross-sectional thickness results.
We proposed that the Ru film thickness measured by SEM cross-
sectional technique was a certain “apparent thickness.” Generally, it
was higher than the real film thickness, especially for films depos-
ited at low temperatures. Logically, the use of apparent thickness
value should produce the “apparent” derivative parameters, such as
an “apparent deposition rate” ͑DRAP͒ and an “apparent film resis-
tivity” ͑␳AP͒. As shown in Fig. 2a, the DRAP values measured by
SEM are similar to the samples deposited with N O or NH . How-
2
3
ever, as shown in Fig. 2b, Ru film resistivity was different depend-
ing on Tdep. Namely, ␳AP was high at low deposition temperatures,
but at Tdep Ͼ 220°C N O-deposited samples revealed a lower ␳
2
AP
compared to NH -grown samples. These results agree with data for
3
Figure 1. SEM photos of ruthenium films deposited at ͑a͒ 180°C, 300 cycles
and ͑b͒ 265°C, 30 cycles. Precursor consumption ϳ1.9 mg/pulse, pulse du-
ration ratio N O/͕Ar + Ru͑CO͒ ͓C H ͔͖ = 1.
2
0
continual CVD. According to this reference, the use of N O at
2
temperatures above ϳ210°C reduces the Ru film resistivity to
2
3
6
8
ϳ138 ␮⍀ cm, which was about 20% less than that obtained under
metallorganic CVD conditions. High enough Ru film ␳ values
AP
compared to bulk Ru resistivity ͑ϳ7 ␮⍀ cm͒ are usually explained
as high impurity concentration in the films, for instance, by carbon
residues. This opinion seemed accurate, but we could not explain the
reason for the abrupt increase of the Ru film resistivity at low tem-
peratures only by carbon residues. To find the possible causes of this
phenomenon, we performed a comparative analysis of the Ru films
deposited at different cycle numbers.
Ruthenium film nucleation and growth features.— For detailed
studies, we used the following experimental conditions: Tdep
198°C as “low” and Tdep = 290°C as “high” temperature points,
=
N2O gas as the second reactant gas, Ru͑CO͒3͓C6H8͔ consumption
ϳ0.7 mg per pulse, and the pulse time duration ratio N2O/͕Ar
+ Ru͑CO͒ ͓C H ͔͖ = 3. Experiments were performed using silicon
3
6
8
Downloaded on 2015-06-25 to IP 134.129.182.74 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, 155 ͑12͒ D763-D770 ͑2008͒
D765
1
1
0
0
,5
,0
,5
,0
1000
wafers with 100 nm thick thermal oxide. Consecutive SEM images
presented in Fig. 3a-d and Fig. 4a-d show that in experimental con-
ditions the Ru films are granular. These images clearly show that in
the experiment’s temperature region the film growth began with the
formation of Ru nuclei on the surface.
(
b)
(
a)
Results of the SEM cross-sectional thickness analysis of these
samples are presented in Fig. 5. In the case of steady Ru film
growth, all data points should follow a linear dependence. Our
analysis showed that the only data points with high Ru film thick-
nesses can be approximated with the straight lines ͑dashed lines in
Fig. 5͒. However, data points obtained for thinner Ru films showed
similar ͑around 20 nm͒ SEM thickness values far from the drawn
linear dependencies. We believe these SEM measurements repre-
sented an average height of the grains in the Ru films. Obviously,
the use of SEM analysis for granular films is meaningless without
taking into account the large empty spaces between the grains. From
the data in Fig. 5, we estimated that the lowest granular Ru film
thickness reasonable for the SEM evaluation was at least 35–40 nm.
The same Ru film samples were analyzed using the optical ellip-
sometry technique. In this case, resulting data are presented also as
an apparent thickness, but the meaning of the term apparent is dif-
100
NH3
N O
2
2
-step process
10
150
225
TDEP( C)
300
150
225
300
o
o
TDEP( C)
Figure 2. Dependencies of ͑a͒ DRAP and ͑b͒ ␳AP on the deposition tempera-
ture. Sample descriptions are similar in both figures. Film thickness was
measured using SEM.
Figure 3. SEM images of Ru films depos-
ited at Tdep = 198°C, and deposition pro-
cess cycle numbers: ͑a͒ 50, ͑b͒ 100, ͑c͒
2
00, and ͑d͒ 300.
Downloaded on 2015-06-25 to IP 134.129.182.74 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D766
Journal of The Electrochemical Society, 155 ͑12͒ D763-D770 ͑2008͒
Figure 4. SEM images of Ru films depos-
ited at Tdep = 290°C and deposition pro-
cess cycle numbers: ͑a͒ 1, ͑b͒ 10, ͑c͒ 50,
and ͑d͒ 200.
ferent from that mentioned above. Optical indexes of the metallic
Ru were obtained from the ellipsometry measurements of the thick-
est film deposited at 290°C and 200 cycles. This film was proposed
to serve as a reference material and its optical indexes were used for
modeling the optical properties of the thinner Ru films. A Brugge-
man model for EMA consisting of Ru and void was used to fit the
experimental and modeling data that gave thickness and Ru content
in the film. The product of this content and thickness equaled the
quantity of Ru in the film, and it was further recalculated into an
apparent thickness of a dense Ru film. The fitting was good enough
for the Ru films deposited in the range of 1–20 cycles and for
8
6
4
2
0
0
0
0
0
o
290 C
1
100 cycles of film deposition. It was quite good for the films
o
98 C
deposited from 30 to 50 cycles. The modeling showed that the
ellipsometry measurements were very sensitive to the morphology
of the Ru film. The grained and continuous films with the same
content of ruthenium revealed drastically different ellipsometry
spectra. The discrepancy between the experimental and calculated
ellipsometry data is explained by the presence on the substrate sur-
face of some film regions with different connectivity of the grains.
The results of ellipsometry thickness measurements of the
samples in Fig. 3 and 4 are shown in Fig. 6. At a high deposition
temperature, the film started to grow without any nucleation delay.
0
100
200
300
Cycle number
Figure 5. SEM measured Ru film thickness vs deposition process cycle
number at different deposition temperatures.
Downloaded on 2015-06-25 to IP 134.129.182.74 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, 155 ͑12͒ D763-D770 ͑2008͒
D767
3
2
1
0
0
0
0
290°C
surface roughness root-mean-square ͑rms͒ values for the films de-
posited at low temperatures were generally two times higher com-
pared to high-temperature deposited films. The dependence of rms
vs cycle number for samples deposited at 290°C revealed two
maxima. The first sharp rms increase in the range from
1 to 10 cycles we can explain by the predominate enlargement of
the separated Ru nuclei. The subsequent rms decrease at
1
1
98°C
55°C
2
0–25 cycles was possibly the result of the initial grain coalescence.
The relatively fluent rms growth to the second maximum seems to
reflect a gradual increase of the grain size. The decrease of rms after
100 cycles can be explained by the film smoothening at high film
thickness.
As-deposited Ru film sheet resistivity values are presented as a
function of cycle number by the solid lines in Fig. 8. No conductiv-
ity was found in the films deposited at less than 200 cycles and
Tdep = 198°C and less than 20 cycles at Tdep = 290°C. According to
the images presented in Fig. 3 and 4, those Ru films were discon-
tinuous with clearly separated grains. The film samples deposited at
Tdep = 290°C and 25–30 cycles revealed a high resistivity disper-
sion. Note also that the ellipsometry thickness dispersion and the
decrease of the first surface roughness maximum both occurred at
this cycle interval. We believe that under these conditions some
neighbor grains in the films come into contact, therefore changing
the film properties. As a result, even a small dispersion of the grain
density within the substrate surface can play a significant role in the
measurements. At a higher cycle number, when the films become
continuous, they revealed quite uniform resistivity values within the
substrate, as well as a low thickness dispersion. We evaluated the
level of the Ru film resistivity for thickest as-deposited Ru films as
0
100
Cycle number
200
300
Figure 6. Ellipsometry measured film thickness vs deposition process cycle
number at different deposition temperatures.
In this case, the apparent deposition rate was steady along with the
process cycle number; a slope of the straight line gave the steady
DRAP ϳ 0.25 nm per cycle. The high thickness dispersion for the
samples deposited at 30 and 50 cycles can be explained using the
presented SEM images in Fig. 4. In fact, the grains in the film
deposited at 50 cycles grow closer and the film morphology be-
comes almost continuous. Thus, even small grain density dispersion
within the substrate ͑and therefore also within the ellipsometer laser
beam͒ seems to cause the high deviation of the optical parameters
and, therefore, the calculated apparent thickness. At Tdep = 198°C
there was a clear nucleation delay for ϳ50 cycles followed by a
low as ϳ19.5 ␮⍀ cm at T = 290°C and ϳ25.9 ␮⍀ cm at Tdep
dep
=
198°C.
The anneal of Ru films using rapid thermal processing ͑RTP͒ at
700°C
for 60 s in argon ambient under dynamic vacuum
23
conditions allowed us to reduce the sheet resistivity and its non-
uniformity within the wafer ͑see the dashed lines in Fig. 8͒. Thick
Ru films revealed the film resistivity as low as ϳ14.6 ␮⍀ cm for
linear film growth along the process time with the steady DR
AP
ϳ 0.13 nm per cycle. At a very low deposition temperature, 155°C,
the nucleation delay was more than 300 cycles, which was the long-
est experimental process time duration. Comparison of data pre-
sented in Fig. 5 and 6 clearly shows that the SEM cross-sectional
analysis provided significantly higher film thicknesses than ex-
pected.
T
= 290°C and ϳ15.8 ␮⍀ cm T = 198°C, respectively. RTP
dep
dep
caused the appearance of the sheet resistivity in the Ru films depos-
ited at Tdep = 290°C and 20 cycles, with severe reduction of the
sheet resistivity in the film deposited at 25 cycles. To explain this
effect, we analyzed the morphology of these films after RTP ͑see the
SEM images in Fig. 9͒. Images in Fig. 9a and d show either no or
little visible changes in the morphology of Ru films deposited at the
low and the high cycle numbers ͑see Fig. 4͒. In contrast, the images
of RTP-annealed Ru films deposited in the range of 20–50 cycles
showed severe changes in the film morphology. There was a coales-
Ruthenium film morphology and resistivity.— Figure 7 presents
SPM data obtained for the films partly presented in Fig. 3 and 4. The
o
290 C
o
8
6
4
2
0
1
98 C
6
5
4
3
2
1
0
o
2
-step process
10
290 C, as-deposited
o
1
2
1
98 C, as-deposited
o
10
10
10
10
10
10
90 C, RTP
o
98 C, RTP
0
100
200
300
1
0
100
Cycle number
1000
Cycle number
Figure 7. As-deposited ruthenium film rms vs deposition process cycle num-
ber for different deposition conditions.
Figure 8. Film sheet resistivity of as-deposited and RTP annealed films vs
deposition process cycle number.
Downloaded on 2015-06-25 to IP 134.129.182.74 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D768
Journal of The Electrochemical Society, 155 ͑12͒ D763-D770 ͑2008͒
Figure 9. SEM images of films deposited
at ͑a͒ 5, ͑b͒ 30, ͑c͒ 50, and ͑d͒ 200 cycles
and Tdep = 290°C after RTP anneal at
00°C in argon gas for 60 s.
7
cence of the small grains either in the larger grains ͑see Fig. 9b͒ or
into the grain chains, as shown in Fig. 9c. We believe these images
can clearly explain the cause of Ru agglomeration effect mentioned
of gas-phase processes at high deposition temperature. Obviously,
this high temperature was not applicable in obtaining conformal step
coverage of Ru films.
With low-temperature deposition we expected better film step
coverage. However, the image in Fig. 10b shows that due to the
nucleation delay there were no Ru grains inside the holes at Tdep
2
7
earlier.
Ru film growth on 3D structures.— From the first look, data
show that Ru deposition at high temperatures is preferable. How-
ever, from a general point of view, the high temperature is known to
activate undesirable gas-phase processes during CVD, for instance,
the formation of reaction intermediates in the gas phase. A well-
known consequence of the gas-phase film growth mechanisms is the
worsening of the film step coverage ͑and the gap-fill͒ in complicated
=
155°C ͑300 cycles͒. At temperatures up to 200°C, very rare
grains were found in the holes confirming poor Ru nucleation on the
surfaces of 3D structures. To improve the nucleation, we proposed
that a short-time high-temperature pulsed CVD step can be used for
Ru film nucleation. According to our logic, this first CVD process
step could provide the sufficient Ru nuclei density without the
nucleation delay. Therefore, the second subsequent low-temperature
deposition step should begin on the already existent nuclei. Eventu-
ally, we expected a continuous Ru film with a much better step
coverage. We examined this two-step pulsed CVD process idea with
the same deposition conditions as mentioned above. The first nucle-
ation step was performed at Tdep = 285°C for five cycles followed
by the second film growth step at Tdep = 155°C for 295 cycles. Data
clearly showed the improvement of Ru film parameters, namely, a
higher deposition rate ͑see Fig. 2a͒ and the lower surface roughness
2
8,29
3D structures.
Below, we present the results of Ru film deposi-
tion on the test structure with 0.35 ␮m diameter contact holes
etched into 0.75 ␮m thick high-density plasma CVD oxide. The
experiments performed using Ru precursor consumption of
ϳ1.1 mg per pulse, NH ͑gas͒, and the pulse time duration ratio
3
NH /͕Ru͑CO͒ ͓C H ͔͖ = 1. The image of Ru film deposited at
3
3
6
8
Tdep = 285°C ͑100 cycles͒ in Fig. 10a shows the grains on the
sample surface. A gradual decrease of the Ru grain density from the
top to the bottom of the hole can also be seen. We believe this grain
nonuniformity confirms the suggestion about the undesirable impact
Downloaded on 2015-06-25 to IP 134.129.182.74 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, 155 ͑12͒ D763-D770 ͑2008͒
D769
Figure 10. Comparison of the film mor-
phology on 3D structures at one-step pro-
cesses at ͑a͒ 275°C, 100 cycles and at ͑b͒
55°C, 300 cycles, ͑c͒ two-step process at
85°C, 5 cycles followed by 155°C,
95 cycles, and at ͑d͒ 1-step process at
55°C, 300 cycles, with the use of subna-
nometer metal seed layer.
1
2
2
1
value ͑see Fig. 7͒. Uniform Ru grain density on the walls and bot-
tom of the holes can be seen in Fig. 10c. The films grown in a
two-step pulsed CVD process showed much lower Ru film resistiv-
ity, as shown in Fig. 2b. This observation clearly shows that the film
nucleation and the grained film structure have a valuable impact on
the increase of the Ru film resistivity under low-temperature depo-
sition conditions.
the size and the surface concentration of Ru nuclei have also been
2
4,26
found to depend on many other process parameters.
For in-
stance, Ru nucleation also depends on surface type, the second re-
actant, process pressure, and surface treatment. These effects will be
the focus of separate publications.
We have shown that the Ru film nucleation and the film mor-
phology differ depending on deposition temperature. However, at
any temperature, the Ru nuclei become the center of the following
film growth once they appear on the surface. This process seems to
occur depending on the deposition temperature. At low temperatures
Proposals for ruthenium film formation.— Believing the tem-
perature as the key CVD process parameter, we have analyzed only
the temperature effects of Ru film nucleation and growth. However,
Downloaded on 2015-06-25 to IP 134.129.182.74 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D770
Journal of The Electrochemical Society, 155 ͑12͒ D763-D770 ͑2008͒
4
͑
Tdep Ͻ 200°C͒, the process of nuclei formation on the surface is
tricarbonyl͓␩ -cyclohexa-1,3-diene͔ruthenium, ammonia, or nitrous
oxide, and consecutively pulsed CVD conditions. High Ru nuclei
surface concentration without nucleation delay was obtained at high
deposition temperatures. At low temperatures, surface nucleation
was a limiting step of the deposition process, resulting in a discon-
tinuous Ru film structure with large grains. Steady film growth rate
occurred on the ruthenium nuclei formed during the fist steps of the
deposition process. Shot-time high-temperature nucleation step and
sputtered subnanometer thick metal seed layer allowed us to signifi-
cantly improve the ruthenium film morphology at very low deposi-
tion temperatures. With the use of seed layers on Si and SiO2 sur-
faces, selective Ru film growth under studied experimental
conditions can be obtained.
slow ͑see Fig. 6͒. As a result, the surface concentration of nuclei and
their size are not uniform ͑see Fig. 3a͒. In addition, these parameters
for Ru deposition have been found to depend strongly on the surface
type of the sample. These effects together indicate that the surface
nucleation is a limiting step of the deposition process at low CVD
temperatures. As long as the formation of new Ru nuclei on the
surface is a very slow process, during the pulsed CVD the Ru
growth occurs mainly on the rare and randomly distributed nuclei.
These nuclei grow in size during the CVD process. As well, some of
the growing neighbor Ru grains can combine. In fact, double, triple,
and multinuclei groups, even nuclei chains between large empty
surface areas, can be seen in Fig. 3b. At high temperatures, a few
pulsed CVD process cycles are sufficient enough to obtain the high
͑
and uniform͒ surface concentration of the Ru nuclei. The absence
Acknowledgment
of the nucleation delay and very low sensitivity to the sample sur-
face type both indicate that the Ru nuclei formation takes place with
a valuable impact of gas-phase processes. As a result, the nuclei size
in the Ru films seems to be much more uniform compared to the
low-temperature CVD conditions. Therefore, the Ru film growth at
high temperatures seems to occur along two parallel reaction routes.
One is the formation of new Ru nuclei on the surface ͑due to gas-
phase processes͒ and the other is Ru growth on existent nuclei,
causing their relatively uniform enlargement.
This work was supported by Gyeonggi Provincial Government,
Republic of Korea. The authors thank S.T. Jung and D.B. Park from
Electronic System Assembly & Inspection Technology Research
Center, Korea Polytechnic University, for SEM and SPM sample
analysis.
Korea Polytechnic University assisted in meeting the publication costs of
this article.
When we combined these results, we proposed that the best sur-
face for creating the perfect starting conditions for the steady Ru
growth can be described as follows. An “ideal surface” should con-
sist of the extremely high Ru nuclei surface concentration. These Ru
nuclei should be extremely small and uniform in size. According to
our logic, on such an ideal surface, the Ru growth could occur even
at very low deposition temperatures with the formation of a continu-
ous film structure with small uniform grains. One can propose that
References
1. M. Nayak, S. Ezhilvalavan, and T. Y. Tseng, in Handbook of Thin Film Materials,
H. S. Nalwa, Editor, p. 121, Academic Press, New York ͑2001͒.
2. V. Mistra, H. Zhong, and H. Lazar, IEEE Electron Device Lett., 23, 354 ͑2002͒.
3. I. Goswami and R. Laxman, Semicond. Int., 27, 49 ͑2004͒.
4. O.-H. Kwon, J.-H. Kim, H.-S. Park, and S.-W. Kang, J. Electrochem. Soc., 151,
G109 ͑2004͒.
5. M. L. Green, M. E. Gross, L. E. Papa, K. J. Schnoes, and D. Brasen, J. Electro-
chem. Soc., 132, 2677 ͑1985͒.
6. J.-H. Joo, J.-M. Seon, Y.-C. Jeon, K.-Y. Oh, J.-S. Roh, J.-J. Kim, and J.-T. Choi,
13-15
the surface pretreatment cited above
or similar could improve
Jpn. J. Appl. Phys., Part 1, 37, 3396 ͑1998͒.
the nucleation. However, to prove the validity of the ideal surface
proposal, we have recently examined a subnanometer thick sput-
tered metal seed layer as “a perfect nucleation layer” for Ru
deposition. Using such a seed, we have grown continuous and
small-grained Ru films in 3D test structures at temperatures as low
as 110°C. The image in Fig. 10d shows the Ru film deposited at
7. H. W. Kim, B.-S. Ju, and C.-J. Kang, Vacuum, 71, 481 ͑2003͒.
8. D. Volger and P. Doe, Solid State Technol., 46, 35 ͑2003͒.
9
. M. Kadoshima, T. Nabatame, M. Hiratani, Y. Nakamura, I. Asano, and T. Suzuki,
2
6
Jpn. J. Appl. Phys., Part 2, 41, L347 ͑2002͒.
0. T. Aaltonen, P. Alén, M. Ritala, and M. Leskelä, Chem. Vap. Deposition, 9, 45
͑2003͒.
11. O. K. Kwon, S. H. Kwon, H. S. Park, and S. W. Kang, Electrochem. Solid-State
Lett., 7, C46 ͑2004͒.
1
155°C under the same process conditions as mentioned above. At
1
2. O. K. Kwon, S. H. Kwon, H. S. Park, and S. W. Kang, J. Electrochem. Soc., 151,
C753 ͑2004͒.
temperatures below 155°C, no Ru films or separate grains were
found on Si, SiO , Al O sample surfaces without seed layers. These
2
2
3
results suggest the possibility of selective Ru film growth under
studied experimental conditions. This matter will be discussed in a
separate paper.
Ruthenium film resistivity at low deposition temperatures re-
mains high enough, even for the Ru films with an improved film
morphology. As we mentioned above, this is possibly due to the
presence in the film of some carbon, nitrogen, or hydrogen rem-
nants. Reduction of these element concentrations is a complicated
issue. We believe that the optimized Ru film devoted to use in 3D
structure should be a trade-off between the film resistivity and the
film step coverage. The first one is more preferable at high deposi-
tion temperatures, the second one is always much better at low
temperatures. We suggest that the Ru film step coverage is a more
serious target for advanced device applications. Therefore, we
should focus on the approaches to improve the resistivity of Ru
films deposited at low temperatures. For immediate practical use,
16. H. Li, D. B. Farmer, R. G. Gordon, Y. Lin, and J. Vlassak, J. Electrochem. Soc.,
54, D642 ͑2007͒.
7. B. F. Johnson, R. D. Johnson, P. L. Josty, J. Lewis, and I. G. Williams, Nature
London), 213, 901 ͑1967͒.
8. S. L. Ingham and S. W. Magennis, J. Organomet. Chem., 574, 302 ͑1999͒.
1
1
1
(
19. S. Uhlenbrock and B. A. Vaartstra, U.S. Pat. 6576778 ͑2003͒.
2
2
2
0. M. R. Visokay, U.S. Pat. 6380080 ͑2002͒.
1. J. Lee, Y. W. Song, K. Lee, Y. Lee, and H. K. Jang, ECS Trans., 2͑4͒, 1 ͑2006͒.
2. S.-H. Chung, V. Vasilyev, E. Gorokhov, Y. W. Song, and H.-K. Jang, Mater. Res.
Soc. Symp. Proc., 0990-B08-01 ͑2007͒.
23. S. H. Chung, V. Yu. Vasilyev, S. T. Jung, D. B. Park, H. K. Jang, and Y. W. Song,
Paper PMA049, in Proceedings of 5th International Symposium NANO KOREA
2
007.
4. V. Yu. Vasilyev, S. H. Chung, and Y. W. Song, Paper PMA047, in Proceedings of
th International Symposium NANO KOREA 2007.
2
5
25. T. S. Lazars, Y. Yang, N. Kumar, D. Y. Kim, W. Noh, G. S. Girolami, and J. R.
23
Abelson, Mater. Res. Soc. Symp. Proc., 0990-B09-06 ͑2007͒.
postdeposition Ru film RTP anneal up to 700°C in inert ambient
for a few seconds seems to be an effective solution. Generally, it
reduces the resistivity of low-temperature deposited Ru films by
2
6. V. Yu. Vasilyev, S. H. Chung, and Y. W. Song, in Proceedings of the 7th ECS
International Semiconductor Technology Conference, M. Yang, Editor, PV 2008-
0
1, p. 667, The Electrochemical Society Proceedings Series, Pennington, NJ
2
2-24,26
5–10 times.
͑2008͒.
2
2
7. B. S. Kim and H. J. Kim, Abstract 1412, The Electrochemical Society Meeting
Abstracts, Vol. 2007-2, Washington, DC, October 7-12, 2007.
8. V. Yu. Vassiliev, J. L. Sudijono, and A. Cuthbertson, Solid State Technol., 44, 129
Conclusion
We have studied the nucleation and growth feature of ruthenium
films deposited in the temperature range of 155–290°C using
͑2001͒.
29. V. Yu. Vassiliev, J. Electrochem. Soc., 150, F211 ͑2003͒.
Downloaded on 2015-06-25 to IP 134.129.182.74 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).