Nanoindentation investigation of HfO2 and Al2O 3 films grown by atomic layer deposition

Article abstract of DOI:10.1149/1.2919106

The challenges of reducing gate leakage current and dielectric breakdown beyond the 45 nm technology node have shifted engineers' attention from the traditional and proven dielectric SiO2 to materials of higher dielectric constant also known as high- k materials such as hafnium oxide (HfO₂) and aluminum oxide (Al₂O₃). These high- k materials are projected to replace silicon oxide (SiO2). In order to address the complex process integration and reliability issues, it is important to investigate the mechanical properties of these dielectric materials in addition to their electrical properties. In this study, HfO₂ and Al₂O ₃ have been fabricated using atomic layer deposition (ALD) on (100) p-type Si wafers. Using nanoindentation and the continuous stiffness method, we report the elastomechanical properties of HfO₂ and Al ₂O₃ on Si. ALD HfO₂ thin films were measured to have a hardness of 9.5±2 GPa and a modulus of 220±40 GPa, whereas the ALD Al₂O₃ thin films have a hardness of 10.5±2 GPa and a modulus of 220±40 GPa. The two materials are also distinguished by very different interface properties. HfO₂ forms a hafnium silicate interlayer, which influences its nanoindentation properties close to the interface with the Si substrate, while Al₂O₃ does not exhibit any interlayer.

Full text of DOI:10.1149/1.2919106

Journal of The Electrochemical Society, 155 ͑7͒ H545-H551 ͑2008͒
H545
0013-4651/2008/155͑7͒/H545/7/$23.00 © The Electrochemical Society
Nanoindentation Investigation of HfO2 and Al2O3 Films Grown
by Atomic Layer Deposition
d,e
*
*
a,c,**,z
K. Tapily,a,c, J. E. Jakes, D. S. Stone,d,e P. Shrestha,a,c, D. Gu,
b,c
a,c,**
H. Baumgart,
and A. A. Elmustafa
aDepartment of Electrical Engineering and bDepartment of Mechanical Engineering, Old Dominion
University, Norfolk, Virginia 23529, USA
cThomas Jefferson Laboratory, Applied Research Center, Newport News, Virginia 23606, USA
dDepartment of Materials Science and Engineering, University of Wisconsin-Madison, Madison,
Wisconsin 53706, USA
eUnited States Department of Agriculture Forest Products Laboratory, Madison, Wisconsin 53726, USA
The challenges of reducing gate leakage current and dielectric breakdown beyond the 45 nm technology node have shifted
engineers’ attention from the traditional and proven dielectric SiO2 to materials of higher dielectric constant also known as high-k
materials such as hafnium oxide ͑HfO2͒ and aluminum oxide ͑Al2O3͒. These high-k materials are projected to replace silicon
oxide ͑SiO2͒. In order to address the complex process integration and reliability issues, it is important to investigate the mechani-
cal properties of these dielectric materials in addition to their electrical properties. In this study, HfO2 and Al2O3 have been
fabricated using atomic layer deposition ͑ALD͒ on ͑100͒ p-type Si wafers. Using nanoindentation and the continuous stiffness
method, we report the elastomechanical properties of HfO2 and Al2O3 on Si. ALD HfO2 thin films were measured to have a
hardness of 9.5 Ϯ 2 GPa and a modulus of 220 Ϯ 40 GPa, whereas the ALD Al2O3 thin films have a hardness of 10.5 Ϯ 2 GPa
and a modulus of 220 Ϯ 40 GPa. The two materials are also distinguished by very different interface properties. HfO2 forms a
hafnium silicate interlayer, which influences its nanoindentation properties close to the interface with the Si substrate, while Al2O3
does not exhibit any interlayer.
© 2008 The Electrochemical Society. ͓DOI: 10.1149/1.2919106͔ All rights reserved.
Manuscript submitted January 14, 2008; revised manuscript received March 24, 2008. Available electronically May 20, 2008.
For the past 40 years the microelectronics industry has relied on
the scaling down of device size in order to improve the performance,
functionality, and bit density of chips, as described by Moore’s law.
As microelectronics is transitioning into deep nanotechnology, the
drawback of the increasing miniaturization of devices is the increase
of gate leakage current and oxide breakdown.1 To reduce the gate
leakage current and breakdown field across the gate insulator, re-
searchers are looking into high-k dielectric materials. High-k mate-
rials such as HfO2 and Al2O3 will increase the transistor drive cur-
rent and the transistor switching speed.2 HfO2 is predicted to replace
SiO2, SiOxNy, and Si3N4 as the gate dielectric of complementary
metal oxide semiconductor ͑CMOS͒ devices at the 45 nm technol-
ogy node and beyond. HfO2 and Al2O3 have dielectric constants of
approximately k = 25 and 8, respectively,3 which compare favorably
with k = 3.9 for SiO2. Various deposition techniques have been used
to deposit high-k materials. Among these growth techniques are
metallorganic chemical vapor deposition ͑MOCVD͒,4-6 pulsed laser
deposition ͑PLD͒,7 and atomic layer deposition ͑ALD͒.4-6,8
MOCVD and PLD require a high temperature during processing and
film fabrication.9 For example, a minimum temperature of 600°C is
required to deposit HfO2 with MOCVD, whereas HfO2 crystallizes
once the temperature reaches 600°C.10 ALD is a chemical reaction-
based deposition technique that requires only relatively low tem-
peratures. ALD provides absolute film deposition uniformity ͑atomic
layer by atomic layer͒, precise composition control, high conformal-
ity, and completely self-limiting surface reactions, which makes
ALD the most suitable low-temperature high-k dielectric materials’
deposition technique for coating of complex surface topographies in
nanoelectronic applications.
technique to deposit high-k dielectric nanoelectronic materials is the
high aspect ratio of the films deposited, which is important for to-
day’s complex surface topographies. In fact, ALD deposits an accu-
rate film thickness and offers a large area uniformity. The final thick-
ness of ALD films depends on the number of ALD deposition cycles
used. In an ALD cycle, chemical precursors are pulsed and purged
consecutively until all precursors are reacted and deposited.
The electrical properties of high-k dielectric materials such as
HfO2 and Al2O3 have been widely studied and investigated. How-
ever, little is known about their mechanical properties. The nanome-
chanical properties of high-k dielectrics are of great technological
importance because the elastomechanical response to thermal cy-
cling and process-induced stress has an effect on the process inte-
gration compatibility and long-term reliability. Nanoindentation is
widely used as a testing mechanism for hardness, modulus, and
fracture toughness of thin films.11-13 In this paper, we use nanoin-
dentation testing techniques and atomic force microscopy ͑AFM͒
imaging of the indentation impressions to investigate the mechanical
properties such as modulus and hardness of HfO2 and Al2O3 thin
films.
Sample Fabrication
We deposited 60, 30, and 10 nm films of HfO2 by ALD on ͑100͒
Si substrates using identical deposition conditions for each film.
Similar film thicknesses of Al2O3 were also deposited. Tetrakisdim-
ethylamidohafnium IV ͑TDMAH͒ and trimethylaluminum ͑TMA͒
were used as chemical precursors for HfO2 and Al2O3, respectively,
for these reactions. H2O vapor was the oxidation source for the
reactions. For HfO2 deposition, the TDMAH precursor tank was
heated at 75°C prior to deposition. The deposition cycle consisted of
pulsing TDMAH, purging N2, and pulsing H2O vapor. Film deposi-
tion parameters such as flow rate, pulse time, pump time, exposure,
and delay time were maintained fixed. For example, the HfO2 films
were deposited at 250°C whereas Al2O3 films were deposited on the
Si substrates at 300°C. The chamber pressure was 2.1
ϫ 10−1 Torr for both HfO2 and Al2O3. For each ALD growth cycle
the oxidizing agent in the form of water vapor ͑H2O͒ was pulsed for
25 ms and the TDMAH and TMA precursors were each pulsed for 1
s duration. Al2O3 was deposited under similar conditions as in
HfO2. The cycle consisted of pulsing TMA, purging N2, and pulsing
H2O. However, in this case the TMA precursor tank was not heated.
ALD also provides the user better control over the deposition
parameters.8 Each chemical reaction that takes place in an ALD
reactor is self-limiting, meaning a given reactant will not react fur-
ther than surface saturation in a given pulse, even if the exposure
with the chemical precursor is continued for a long time. The reac-
tion by-products are purged out with an inert gas, typically N2 or Ar.
Another trait that uniquely defines ALD as the most appropriate
*
Electrochemical Society Student Member.
Electrochemical Society Active Member.
**
z E-mail: dgu@odu.edu
Downloaded on 2015-07-06 to IP 142.103.160.110 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H546
Journal of The Electrochemical Society, 155 ͑7͒ H545-H551 ͑2008͒
Table I. Summary of the film thicknesses obtained by spectro-
scopic ellipsometry.
Thin film
Desired thickness ͑Å͒
Measured thickness ͑Å͒
HfO2
HfO2
HfO2
Al2O3
Al2O3
Al2O3
600
300
100
600
300
100
583
296
104
607
300
103
Figure 1. ͑Color online͒ Linearity of HfO2 ͑black square͒ and Al2O3 ͑circle͒
film thickness with ALD cycles at 250°C. The film thicknesses were mea-
sured using the spectroscopic ellipsometer ͑Woollam, VASE͒.
The thickness vs the number of cycles for HfO2 and Al2O3 is shown
in Fig. 1. From the graph, one can see that the deposition rate is
linear, which allowed us to predict the number of cycles necessary to
deposit a desired film thickness. The thickness of the films was
measured by a spectroscopic ellipsometer ͑Woollam, VASE͒. Table I
summarizes the actual films’ thickness subsequent to ALD deposi-
tion.
Transmission electron microscope ͑TEM͒ and AFM analysis
were performed on a 4 nm HfO2 sample to illustrate the uniformity
and roughness of the films. A high-resolution TEM micrograph of
HfO2 film on bulk silicon is shown in Fig. 2. Note that the HfO2
films deposited at 250°C are primarily amorphous, as depicted by
the high-magnification TEM cross section in Fig. 2. These HfO2
films also contain a number of crystallites, which are not shown in
this TEM micrograph, but which are clearly evident from our AFM
characterization. AFM analysis was performed in the tapping mode
on the samples to examine the surface morphology in a scan area of
Figure 2. High-resolution TEM cross-section micrograph of a 4 nm HfO2
film deposited by ALD on a Si substrate.
1 ϫ 1 ␮m. The surface roughness of the Al2O3 samples was about
0.12 nm and almost constant for various thicknesses. In contrast, the
surface roughness of HfO2 increases as a function of the film thick-
ness. A final root-mean-square ͑rms͒ surface roughness of 3.3 nm
was observed for the 60 nm films. Such an increased surface rough-
ness affects the nanoindentation measurements. ALD HfO2 films
deposited at 250°C were almost 30 times rougher than the Al2O3
films deposited at 300°C. A three-dimensional ͑3D͒ AFM image in
Fig. 3 shows the 60 nm Al2O3 and the 60 nm HfO2 films side by
side.
Figure 3. ͑Color online͒ 3D AFM images of 60 nm ALD Al2O3 and HfO2 films. The rms roughness of 60 nm Al2O3 films is 30 times smoother than the 60
nm HfO2 films.
Downloaded on 2015-07-06 to IP 142.103.160.110 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 ͑7͒ H545-H551 ͑2008͒
H547
Figure 5. ͑Color online͒ Comparison between the 1000 and 50 nm indents
Figure 4. ͑Color online͒ Graph showing the 1000 nm indents for 60 nm film
for the 60 nm film thickness of Al2O3 and HfO2.
thickness of Al2O3 and HfO2.
Experimental
The hardness measurement critically depends on the area of in-
dentation and the indenter tip calibration. To verify that the areas of
indentation were accurately measured and that the indenter tip was
Nanoindentation analysis was used to investigate the mechanical
properties of ALD Al2O3 and HfO2 thin films. We obtained data
from the HfO2 and Al2O3 films in three different steps. In the first
step, we only performed indents up to 80% of the top nanolayer film
of HfO2 and Al2O3. During the second indentation step, we pen-
etrated through the entire film to the interface, which provides data
up to about 100% of the film thickness. Finally, we performed in-
dents up to 1000 nm deep into the bulk Si. For example, for the 60
nm films, a set of shallow indents was made up to 50 nm, which is
80% of the film thickness. Then, another set of indents was made at
60 nm, which is 100% of the film thickness. Finally, a set of indents
was made at 1000 nm. The indents were made with a three-sided
pyramidal Berkovich tip made of diamond using the continuous
stiffness measurement ͑CSM͒. The CSM method consists of con-
tinuously applying and recording the displacement of the indenter as
a function of the applied force during a complete cycle of loading
and unloading. Despite the fact that it is redundant to make indents
at different depths of indentation because all the desired information
i.e., the mechanical properties, can be obtained from one indent
performed at a deep depth of indentation with the CSM engaged, we
still obtained data at different depths of indentation for comparison
and demonstration purposes. In fact, as can be detected from Fig. 4,
the 1000 nm indents alone would have been sufficient with the CSM
engaged because the hardness can be determined for the surface
layers as well as the bulk silicon. From Fig. 5, it is evident that for
the 60 nm films of Al2O3 and HfO2 the hardness data from the 1000
and 50 nm indents overlap. In this study, the remainder of the analy-
sis and simulations was performed based on the 1000 nm CSM deep
indents. Typical load–depth curves are shown in Fig. 6 and 7. The
plots show loading and unloading of an indentation cycle. Figure 6
depicts the 1000 nm indents whereas Fig. 7 demonstrates how the
1000 nm indents overlap the 50 nm indents for both films. During
loading, typically the material undergoes elastic and plastic defor-
mation. The peak load during the loading cycle is used to define the
hardness. Nanoindentation hardness is defined as the maximum in-
dention load divided by the projected contact area of the indenter tip
appropriately calibrated, we plotͱin Fig. 8 the contact depth vs the
square root of the contact area A for Al2O3 and HfO2. Data for
fused silica calibration standards are shown as well. It is evident that
the data of Al2O3 and HfO2 correlate well with each other and with
the fused calibration standards. Therefore, we conclude that the ar-
eas of indentation measurements are accurate and that the hardness
calculations are accurate as well.
The unloading cycle was mainly dominated by elastic displace-
ment. Therefore, the modulus can be obtained from the unloading
curve. The elastic modulus was obtained by dividing the slope of the
load vs displacement curve at the maximum load data point by the
projected contact area of the indenter tip
Pmax
H =
͓1͔
A
where H is the hardness, Pmax is the max load, and A is the area of
contact.
Figure 6. ͑Color online͒ Load vs depth showing the loading and unloading
mode for 1000 nm deep indents.
Downloaded on 2015-07-06 to IP 142.103.160.110 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H548
Journal of The Electrochemical Society, 155 ͑7͒ H545-H551 ͑2008͒
Figure 7. ͑Color online͒ Load vs depth for 50 and 1000 nm indents. The
1000 nm profile overlaps the 50 nm one.
1
dP
dh
E =
͓2͔
ء
␤ ͱA
where E is the reduced modulus of the specimen, A is the area of
contact, dP/dh is the slope of the load–depth curve, and ␤ is a
constant depending on both the indenter geometry and Poisson’s
ratio.
The effective modulus of the specimen is discussed later and the
definition of Eeff is provided in Eq. 3. The nanoindentation stiffness
of the composite HfO2/Si and Al2O3/Si systems was modeled using
elasticity theory for indentation against a layered specimen, which
will be discussed later.14 However, during the indentation of some
materials, fracture events or debonding at the film–substrate inter-
face may occur and can be observed as discontinuities in the load vs
displacement curves. In Fig. 9, AFM micrographs of an indent made
by a Berkovich diamond tip on the 104 and 296 Å HfO2 films show
evidence of cracks. We have also noticed that, in the immediate
vicinity of the indent, the material that was displaced by the indenter
Figure 9. AFM picture of a Berkovich indent that was done on ͑top͒ 296 Å
HfO2 and ͑bottom͒ 104 Å HfO2 thin films. The increase in roughness with
film thickness can be seen.
is pushed up along the sides of the indenter, similar to a snow
plough effect. An example of that phenomenon can be seen in the
104 Å HfO2 in Fig. 9.
Results and Discussion
In this section, a detailed discussion pertaining to the 1000 nm
CSM indents will be presented, which includes all the information
of interest from the film surface to the bulk Si underneath. Also,
values calculated from individual indents performed by a Hysitron
͑Minneapolis, MN͒ Triboindenter with areas measured from AFM
images were compared to the values obtained from CSM data. In
nanoindentation, the presence of the substrate introduces biases in
the measurement of the modulus and hardness of thin films. As the
indentation depth gets closer to the interface between the thin film
and the substrate, the effect of the substrate becomes more pro-
nounced. One way of reducing the effect of the substrate on the
mechanical properties of the high-k dielectric thin films is to model
the whole system and perform some data fitting when the substrate
properties are known.14 To take into account the substrate effect,
simulations, and modeling based on areas from the AFM images, the
substrate properties and the specimen compliance were performed.
Figure 10 shows the plot of the normalized compliance and the area
Figure 8. ͑Color online͒ Contact depth of indentation vs ͱA. Data for fused
silica calibration standards are shown as well.
Downloaded on 2015-07-06 to IP 142.103.160.110 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 ͑7͒ H545-H551 ͑2008͒
H549
Figure 11. Graph shows the modulus vs normalized square root of the area
of contact by film thickness. Modulus of Al2O3 is about 220 Ϯ 40 GPa.
Eeff = film modulus, hf = film thickness, and A = area of the indent. The
error bars are based on the uncertainty of 10 nm of ͱA.
10.5 Ϯ 2 GPa. From the individual indent data for both modulus
and hardness in Fig. 11 and 12, a decrease in hardness and modulus
is observed at a normalized squared root of the area value of 4–5.
This corresponds to an observation that all the indents performed on
this 60 nm Al2O3 film had a discontinuity in the load vs displace-
ment curves, likely a fracture event, which corresponds to a load of
1 mN. This fracture event most likely influenced the calculated val-
ues of modulus and hardness. Figure 13 shows a high-resolution
AFM micrograph of the Berkovich indenter tip into the 10 and 30
nm films of Al2O3 film, demonstrating the smooth surface morphol-
ogy and absence of cracks.
Similar results and analysis were obtained for the 60 nm HfO2
film. The modulus of HfO2 films is 220 Ϯ 40 GPa and the hardness
is 9.5 Ϯ 2 GPa, respectively. Hardness and modulus data of the
Figure 10. Compliance vs normalized area plot for ͑top͒ Al2O3 and ͑bottom͒
HfO2 films. H = film thickness, A = area, and C = compliance.
by the film thickness. Comparing the slope of the compliance vs
area plot to the simulations allows us to determine the modulus of
the film.14 Figure 11 shows the modulus of Al2O3 films vs normal-
ized square root of the area by the film thickness. The data represent
the CSM measurement, with individual indents from the Hysitron
indenter, and the simulated moduli based on the area measured by
AFM.14 The effective modulus Eeff from Fig. 11 was obtained from
the simulations and is modeled as follows
2
s
2
i
−1
1 − ␯
1 − ␯
S
ͱA
Eeff
=
=
␤
+
͓3͔
ͩ
ͪ
ͬ
ͫ
Es
Ei
for a monolithic specimen, where ␤ is a constant and for this simu-
lation ␤ = 1.22, ␯ = substrate Poisson’s ratio, ␯ = indenter Pois-
s
i
son’s ratio, Es = substrate Young’s modulus, and Ei = indenter
Young’s modulus ͑in this case the Berkovich indenter tip is made of
diamond͒.
From Fig. 10 and 11, the modulus of Al2O3 films corresponds to
220 Ϯ 40 GPa. CSM data also correlate well with the individual
indents from the Hysitron indenter except for shallow indents. How-
ever, there is a weak correlation for very shallow indents. We be-
lieve this is due to error and uncertainty in the measured area of the
shallow indents. In Fig. 12, the plot of hardness vs the normalized
square root of the area is shown. The hardness of Al2O3 is
Figure 12. Graph shows the hardness vs normalized square root of the area
of contact by film thickness. Hardness of Al2O3 is about 10.5 Ϯ 1 GPa.
H = film hardness, hf = film thickness, and A = area of the indent. The error
ͱ
bars are based on the uncertainty of 10 nm of A.
Downloaded on 2015-07-06 to IP 142.103.160.110 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H550
Journal of The Electrochemical Society, 155 ͑7͒ H545-H551 ͑2008͒
HfO2 exhibit more scattering compared to the Al2O3 data. This is
due to the significantly higher surface roughness of the HfO2 film.
At the constant deposition temperature of 250°C the rms surface
roughness of HfO2 films is approximately a factor of 30 higher than
Al2O3. These surface features correspond to 5.5% of the HfO2 film
thickness. In contrast, a purely amorphous HfO2, which was grown
one atomic layer at a time by ALD at lower temperatures, would
exhibit a very smooth surface morphology. In our case the AFM
observed surface roughening of HfO2 at 250°C provides experimen-
tal evidence of the onset of nucleation and growth of crystallites.
Each ALD growth cycle produces nucleation sites in a random fash-
ion over the amorphous layers of the growing HfO2 film. Random
nucleation and thermally activated growth of small crystallites are
the primary causes of surface roughening in the initially amorphous
HfO2 films.15 Grain growth of individual crystallites in the amor-
phous matrix is temperature activated. We have verified that we can
obtain purely amorphous HfO2 films with a very smooth surface
morphology by lowering the ALD temperature below 150°C. It is a
well-established fact that grain growth occurs at different speeds for
different crystallographic orientations. The random orientation of
the HfO2 crystallites embedded in the amorphous matrix guarantees
that a sufficient number of HfO2 crystallites happen to be oriented in
such a way that their maximum growth velocity is in the vertical
direction. This explains the bumps and surface roughening in ALD
HfO2 films observed by our AFM measurements. It takes well over
600°C to achieve a complete phase change to a 100% polycrystal-
line HfO2 film. At our deposition temperature of 250°C we have a
fraction of the HfO2 film crystallized in small grains, which are
embedded in the amorphous matrix. From Fig. 9 cracks are observed
around the indent on 30 and 10 nm HfO2 thin films. We attribute the
formation of the cracks to the porous, nondensified nature of the
amorphous high-k films in this study. Because the films used were
as-deposited, they incorporate a portion of the deposition gases, for
example the carrier gas N2 in our case. Furthermore, debonding at
the interface and microcracks contributed to the growth of the major
cracks observed in the films. Microcracks are clearly delineated in
the HfO2 film displaced by the Berkovich indent as seen in Fig. 14.
Further studies will be performed in the near future to investigate
the reason for the cracks of the films after indentation, to study the
Figure 13. The top picture is an AFM micrograph of a 3.4 mN indent on a
300 Å thick Al2O3 thin film; the bottom picture is an AFM micrograph of a
103 Å thick Al2O3 thin film demonstrating very smooth surface morphology
and absence of defects and cracks. This is in direct contrast with HfO2 films,
where many microcracks were detected.
Figure 14. ͑Color online͒ The picture
shows a 1.5 mN indent in the 10 nm HfO2
film with a large amount of bubbling up
around the indent. The three diamond-
shaped objects surrounding the indent rep-
resent the inverse shape of the AFM tip,
an artifact likely caused by an extremely
sharp point present on the surface of the
thin film.
Downloaded on 2015-07-06 to IP 142.103.160.110 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 ͑7͒ H545-H551 ͑2008͒
H551
Table II. Summary of the measured nanomechanical properties
of HfO2 and Al2O3 ALD films.
Thin film
Modulus ͑GPa͒
Hardness ͑GPa͒
HfO2
Al2O3
Bulk Si
220 Ϯ 40
220 Ϯ 40
180 Ϯ 40
9.5 Ϯ 1
10.5 Ϯ 1
13 Ϯ 1
Literature values for comparison
a
Al2O3
Al2O3
180 Ϯ 8.2
150
177
160–180
12 Ϯ 1
9.5
b
c
Al2O3
9.6
d
Al2O3
a Al2O3: ALD Al2O3 deposited at 177°C.17
b Al2O3: Al2O3 deposited by physical vapor deposition.18
c Al2O3: Al2O3 deposited by electron cyclotron resonance plasma.19
d Al2O3: Al2O3 deposited by evaporation.20
films of HfO2 and Al2O3 deposited by ALD. The nanoindentation
method was used in conjunction with the CSM method to measure
the hardness and modulus. Finally, by combining computer simula-
tions with nanoindentation experimental results for ALD HfO2 thin
films we obtain a hardness of 9.5 Ϯ 2 GPa and a modulus of
220 Ϯ 40 GPa, whereas ALD Al2O3 thin films yield a hardness and
modulus of 10.5 Ϯ 2 and 220 Ϯ 40 GPa, respectively. Our studies
revealed the formation of a much harder 2–3 nm hafnium silicate
interlayer, which is responsible for the increase in HfO2 hardness
close to the Si substrate interface. This is in contrast to Al2O3, where
no such interlayer is found. Further studies will be performed to
investigate the root cause for observed defects such as cracks, bub-
bling of the films, and pop-ins.
Figure 15. High-resolution TEM cross section of 4 nm HfO2 film on silicon
showing an interlayer of hafnium silicate of about 2 nm.
effect of rapid thermal annealing at higher temperatures on the na-
nomechanical properties, and to compare the annealed samples to
the as-deposited.
The data obtained in this study demonstrate that the hardness
results of ALD Al2O3 and HfO2 films are roughly comparable
within the accuracy of the indenter measurements.16 According to
the plots of Fig. 4 and 5 there is hardly a difference between the
experimental hardness values for both films up to a depth of 50 nm.
Once the indenter tip approaches the films’ thickness at the interface
where the substrate effect becomes significant, the hardness values
of both films experience slight changes. HfO2 instantly becomes
harder than Al2O3 and remains harder until the indenter tip reaches
a depth of 500 nm, where the hardness of both films ultimately
converges to the hardness of the bulk Si. However, during deposi-
tion of the HfO2 film on the Si substrate a 2–3 nm thick hafnium
silicate interlayer develops at the interface due to the interdiffusion,
as depicted by Fig. 15. The interlayer grows thicker for 60 nm HfO2
film. This hafnium silicate interlayer produces a harder surface di-
rectly at the Si interface. The presence of a harder hafnium silicate
interlayer combined with poorer adhesion accounts for the transient
increase of the hardness for HfO2 films as shown in Fig. 4 and 5.
Al2O3 has not experienced this temporary increase in the hardness
due to the absence of such an interface layer and due to better
substrate bonding conditions. The modulus and hardness values of
ALD Al2O3 thin films are comparable with literature values. See
Table II for details. Little is known about HfO2 thin films. Finally,
Table II summarizes the modulus and hardness of Al2O3, HfO2, and
bulk Si from this work and literature values for hardness and modu-
lus deposited with different deposition techniques.
Old Dominion University assisted in meeting the publication costs of this
article.
References
1. N. Miller, K. Tapily, H. Baumgart, A. A. Elmustafa, G. Celler, and F. Brunier,
Mater. Res. Soc. Symp. Proc., 1021E, 5 ͑2007͒.
2. K. Cherkaoui, A. Negara, S. McDonnell, G. Hughes, M. Modreanu and P. K.
Hurley, in Proceedings of the 25th International Conference on Microelectronics,
Belgrade, Serbia, and Montenegro, p. 351 ͑2006͒.
3. S. K. Dey, A. Das, M. Tsai, D. Gu, M. Floyd, R. W. Carpenter, H. D. Waard, C.
Werkhoven, and S. Marcus, J. Appl. Phys., 95, 5042 ͑2004͒.
4. C. Dubourdieu, H. Roussel, C. Jimenez, M. Audier, J. P. Senateur, S. Lhostis, L.
Auvray, F. Ducroquet, B. J. O’Sullivan, P. K. Hurley, et al. Mater. Sci. Eng., B,
118, 105 ͑2005͒.
5. A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, K. Kukli, A. Rahtu, M.
Ritala, and M. Leskela, Mater. Sci. Eng., B, 118, 97 ͑2005͒.
6. D. H. Triyoso, M. Ramon, R. I. Hegde, D. Roan, R. Garcia, J. Baker, X. D. Wang,
P. Fejes, B. E. White, and P. J. Tobin, J. Electrochem. Soc., 152, G203 ͑2005͒.
7. H. Ikeda, S. Goto, K. Honda, M. Sakashita, A. Sakai, S. Zaima, and Y. Yasuda, Jpn.
J. Appl. Phys., Part 1, 41, 2476 ͑2002͒.
8. J. S. Becker, Ph.D. Thesis, Harvard University, Cambridge, MA ͑2002͒.
9. W. Deweerd, A. Delabie, S. V. Elshocht, S. D. Gendt, M. Caymax, and M. Heyns,
Future Fab Intl., 20, 93 ͑2006͒.
10. S. J. Lee, Y. M. Jung, S. J. Lim, K. H. Lee, S. K. Lee, and T. W. Seo, J. Appl. Phys.,
92, 2807 ͑2002͒.
11. M. F. Doerner and W. D. Nix, J. Mater. Res., 1, 601 ͑1986͒.
12. A. C. Fischer-Cripps, Nanoindentation, p. 21, Springer-Verlag, Berlin ͑2002͒.
13. W. C. Oliver and G. M. Pharr, J. Mater. Res., 7, 1564 ͑1992͒.
14. D. S. Stone, J. Mater. Res., 13, 3207 ͑1998͒.
15. D. M. Hausmann and R. G. Gordon, J. Cryst. Growth, 249, 251 ͑2003͒.
16. K. Tapily, J. Jakes, D. S. Stone, P. Shrestha, D. Gu, H. Baumgart, and A. A.
Elmustafa, ECS Trans., 11͑7͒, 123 ͑2007͒.
17. M. K. Tripp, C. Stampfer, D. C. Miller, T. Helbling, C. F. Herrmann, C. Hierold, K.
Gall, S. M. George, and V. M. Bright, Sens. Actuators, A A130–131, 419 ͑2006͒.
18. T. C. Chou, T. G. Neih, S. D. McAdams, and G. M. Pharr, Scr. Metall. Mater., 25,
2203 ͑1991͒.
19. J. C. Barbour, J. A. Knapp, D. M. Follsteadt, T. M. Mayer, K. G. Minor, and D. L.
Linam, Nucl. Instrum. Methods Phys. Res. B 166–167, 140 ͑2000͒.
20. N. G. Chechenin, J. Bottiger, and J. P. Krog, Thin Solid Films, 304, 70 ͑1997͒.
Conclusion
High-k dielectrics are expected to replace SiO2, SiOxNy, and
Si3N4 as metal-oxide-semiconductor field-effect transistor gate di-
electrics or dynamic random access memory ͑DRAM͒ memory ca-
pacitor dielectrics at the 45 nm technology node and beyond. HfO2
and Al2O3 have attracted attention as potential candidates to find
applications as CMOS gate or DRAM capacitor dielectrics. The goal
of this study was to focus on the nanomechanical properties of thin
Downloaded on 2015-07-06 to IP 142.103.160.110 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).