Frequency-dependent pulsed direct current magnetron sputtering

of titanium oxide films

J.-Y. Kim,a),b) E. Barnat, E. J. Rymaszewski, and T.-M. Lu
Center for Integrated Electronics and Electronic Manufacturing, Rensselaer Polytechnic Institute,
Troy, New York 12180

Received 16 October 2000; accepted 8 January 2001

Dramatic dependence of dielectric properties on the pulsing frequency was found for titanium oxide
films deposited using the pulsed dc magnetron sputtering technique at room temperature. The
frequency range studied was between 50 and 250 kHz by varying the oxygen pressure. A minimum
leakage current density of 0.22 A/cm2 at 0.5 MV/cm electric-field strength for a film with
dielectric constant of 26 was achieved for relative oxygen pressure P60% P(%) P O2 /( P O2
P Ar and frequency f 200 kHz. 2001 American Vacuum Society.
DOI: 10.1116/1.1351064

I. INTRODUCTION
Titanium dioxide (TiO2 ) has been widely investigated for
optical application because it is mechanically hard and
chemically resistant and has a high refractive index (n
2.5). 17 Also, it has been studied as a dielectric material
capacitor and gate material for integrated circuit application because of its high dielectric constant.815 Many techniques have been used to prepare titanium oxide thin films.
Examples are: reactive evaporation,1,24 reactive direct current
dc or radio-frequency rf sputtering,26,9,14,23 chemicalvapor deposition CVD,13 plasma-enhanced chemical-vapor
deposition PECVD,16 and the solgel process.17
Depending on the deposition techniques and deposition
conditions, the dielectric constant of the titanium oxide films
can vary from 10 to 150.8,9,1215,17 The variations in the dielectric constant of titanium oxide are related to the presence
of low-permittivity interfacial layers between titanium oxide
and the substrate and on the crystalline phase of the titanium
oxide films. For example, Rausch and Burte15 showed that
the dielectric constant of titanium oxide is 28 as deposited,
and increased up to 70 after the postannealing process due to
development of a new crystalline phase. From a technical
viewpoint, rutile is the most favorable crystalline phase because it has a high dielectric constant and refractive index.
However, in many applications, one cannot allow the hightemperature annealing that is required to obtain the rutile
phase. In such cases, one has to research a deposition technique that can provide high-quality amorphous films with a
reasonable dielectric constant.
In reactive sputtering, a metal target is used and a partial
pressure of reactive gas (O2 ) is added to the Ar sputtering
atmosphere to deposit an oxide film. One advantage of reactive sputtering is that the stoichiometry of the deposited film
can be controlled. Also, a higher deposition rate can be obtained when compared to the sputtering of an oxide coma

Electronic mail: j-kim12@ti.com

Now with Texas Instruments, Inc., 13560 North Central Expressway,
MS3737, Dallas, TX 75243.

429

J. Vac. Sci. Technol. A 192, MarApr 2001

pound target because metal targets generally have a higher

sputtering yield than that of the compound target. However,
target poisoning, which is a build up of an insulating layer
reaction product on the target surface during deposition,
has a detrimental effect: arcing. Arcing results in the generation of particulates, which can degrade the properties of the
deposited oxide films. Rf sputtering is an alternative method
for sputtering of oxide dielectric films. However, its drawbacks includes deposition rate, excessive substrate heating,
and lack of uniformity in ion energy distribution in the
plasma during deposition.18
Pulsed reactive dc magnetron sputtering has been developed to improve the quality of oxide films by avoiding
arcing.1821 For a pulsed reactive dc magnetron sputtering,
the polarity of the power supplied to the target oscillates
between negative and positive potential. During the positive
period, the insulating layer on the target surface is discharged and arcing is prevented. From the defect density
found in the dielectric oxide film and the deposition rate, it is
estimated that the typical operating frequencies for pulsed dc
magnetron sputtering of oxides should be in the 10250 kHz
range.1821 Belkind et al.18 suggested that arcing could be
avoided when pulsing frequency is higher than a critical frequency that depends on the cathode current and reverse time.
So far, very few attempts have been made to study the pulsing effect on the electrical properties of oxide films.
In the present work, the dependence of the properties of
the titanium oxide dielectric films on the pulsing frequency
was investigated at low deposition temperature 200 C.
The properties of the films prepared by reactive sputtering
also depend on other deposition conditions. For example,
oxygen partial pressure is an important factor for the electrical properties of tantalum oxide films deposited by reactive
dc sputtering.20 In this work, the physical and electrical properties of the titanium oxide films were studied in terms of
dielectric constant, leakage current density, and breakdown
field strength using different pulsing frequencies and oxygen
partial pressure in the sputtering atmosphere.

0734-210120011924296$18.00

2001 American Vacuum Society

429

430

Kim et al.: Frequency-dependent pulsed dc magnetron sputtering

TABLE I. Pulsing frequency and duty cycle used in the deposition of titanium oxide films for 3 kW sputtering power.
Pulsing frequency
kHz

Reverse time
ns

Reverse bias duty cycle

%

50
150
200
250

4016
1296
976
816

20.08
19.44
19.52
20.40

II. EXPERIMENT
A CVC model AST-601 sputter system, which had a three
20.3 cm circular target assembly installed in the bottom base
plate, was used to deposit the titanium oxide films. The target was a 20.3-cm-diam, 0.63-cm-thick Ti plate of 99.99%
purity. The target was located at the bottom of the chamber
in a sputter-up configuration. The circular rotary substrate
holder was 5 cm above the Ti target. Two types of substrates
were loaded into the chamber: 5 in. p-type silicon wafers
with or without Al metal. Bare silicon wafers were cleaned
by the standard RCA procedure just before being loaded into
the chamber and used as substrates to deposit titanium oxide
films for thickness and refractive index measurements. Films
deposited on the metallized silicon wafers were used for
electrical evaluation.
The system was evacuated by a compressed helium cryogenic high vacuum pump below 2107 Torr prior to deposition. Different gas inlets with mass flow controllers to the
deposition chamber were used to regulate the oxygen partial
pressure of the sputtering atmosphere, which was a mixture
of argon and oxygen gas. Before the deposition, the titanium
target was presputtered with 100% Ar gas for 5 min. The
total pressure was kept at 6 mTorr during the deposition. No
intentional substrate heating was used during the depositions.
A reactive plasma generator, ENI RPG-50, was used to
deposit the films. The pulsing frequency can be varied from
50 to 250 kHz with a 40% maximum reverse bias duty cycle.
The pulsing frequency and revere bias duty cycle used in this
experiment are shown in Table I. The reverse bias duty cycle
is defined as the reverse time divided by the total time, which
is the inverse of the pulsing frequency. The deposition power
and time were kept constant, 3 kW and 50 min., respectively.
The deposition rates of the titanium oxide films were 2023
/min.
Single-wavelength ellipsometry 633 nm, Rudolphs
AL-II at an incident angle of 70 from normal was used to
determine the film thickness, deposition rate, and refractive
index. Metalinsulatormetal MIM structures were fabricated to evaluate the electrical properties of the titanium oxide films. Al electrodes were deposited by an e-beam technique through a metal shadow mask on the top of the
titanium oxide film to make the MIM structure. Two
and 17.67
different-sized capacitors (4.32103
103 cm2 ) were used for electrical characterization. A
HP4192A rf impedance analyzer was used for the capacitance measurement with 1.1 V sweeping signal at a frequency of 10 kHz or 1 MHz and a HP4140B pA meter was
J. Vac. Sci. Technol. A, Vol. 19, No. 2, MarApr 2001

430

used for leakage current and breakdown voltage measurements. The breakdown voltage was defined as the current
compliance of the instrument, which was 1102 A.
The structural characterization of the deposited films on
the silicon wafer was analyzed using x-ray diffraction with
Cu K radiation 40 kV and 30 mA and a nickel filter. The
measured diffraction angle was from 20 to 70 grazing
with 0.03 increments. The stoichiometry of the titanium oxide films was analyzed by both Rutherford backscattering
spectroscopy RBS using a 2 MeV He beam and x-ray
photoelectron spectroscopy XPS.

III. RESULTS AND DISCUSSION

A. Film deposition

For dc reactive sputtering, there are three oxidation modes

of the target as a function of reactive gas concentration (O2 ):
metallic mode at low gas flow, reactive mode at high gas
flow, and transition mode in which the target changes from
metallic to reactive mode. The different modes represent the
degree of oxide coverage on the target; hence, it would influence the I V characteristics of the plasma discharge. For
example, at a constant sputtering power, the potential increased with O2 partial pressure in the sputtering atmosphere, reached a maximum at a particular O2 partial pressure, then decreased again. It is also known that the degree of
oxide coverage on the target depends on the plasma power
density.22 For example, the transition region was increased
from 10% to 20% O2 when a sputtering power increased
from 1.1 to 2.5 kW for the Ta2O5 films.20 Schiller et al.22
proposed that the cathode voltage could be used as a reference to deposit high-quality films at constant power and
pressure. According to Schillers criterion, the optimum
deposition condition to give a high deposition rate and a
good film stoichiometry was a point at which the target potential reached the maximum value. However, we found that
the best dielectric properties high dielectric constant, low
leakage current density, and high dielectric field strength
could be obtained from a higher oxygen partial pressure than
that of the maximum cathode voltage for the Ta2O5 films.20
The cathode voltage as a function of oxygen partial pressure, P(%) P O2 /( P O2 P Ar), in the plasma were recorded
at a fixed sputtering power and different pulsing frequencies.
The film deposition was performed at a power regulation 3
kW mode and the total pressure was kept at 6 mTorr for all
depositions. Oxygen was fed into the chamber starting from
P0% Ar only and increased up to P100% (O2 only.
The plasma was initiated at each oxygen partial pressure and
allowed for stabilization, then the cathode voltage and current were recorded.
Figure 1 shows the cathode voltage as a function of P in
the sputtering atmosphere. There was an abrupt transition in
the cathode voltage between 6% and 10% O2 for all frequencies. The cathode voltage slightly increased with pulsing frequency and oxygen partial pressure, but the difference was
less than 2% at the same P. From Fig. 1, it is seen that the

431

Kim et al.: Frequency-dependent pulsed dc magnetron sputtering

431

by x-ray diffraction because no additional heating was applied during sputtering. No diffraction peaks except that
from the Si substrate were found when the titanium oxide
films on the silicon wafer were scanned from 20 to 70 in
the x-ray diffraction. The stoichiometry of the titanium oxide
film was obtained using RBS and XPS. The results showed
that the O/Ti atomic ratio of the films ranged from 2.2 to 2.4
as a function of oxygen partial pressure and there was not
much difference for different pulsing conditions.
B. Electrical properties

FIG. 1. Cathode voltage variation as a function of oxygen partial pressure

and pulsing frequency. The sputtering power was 3 kW.

transition region does not depend significantly on the pulsing

frequency.
The refractive index n of the films as a function of oxygen partial pressure and pulsing frequency is shown in Fig.
2. The refractive indices ranged from 2.25 to 2.40. It can be
seen that the refractive index increased, reached a maximum,
and then decreased. The trend appears to be quite similar for
different pulsing conditions. It is known that the refractive
index is related to the densification of the films.7 For example, Meng and co-workers showed that the density of the
film and the refractive index in the dc magnetron sputtering
process decrease as the oxygen partial pressure increases.5 It
was also found that the refractive index of the deposited film
increases with an increase in the substrate temperature.2 In
this work, the deposited films were amorphous as identified

FIG. 2. Refractive index of the titanium oxide films as a function of oxygen

partial pressure and pulsing frequency.
JVST A - Vacuum, Surfaces, and Films

The electrical properties of the titanium oxide films were

examined as a function of P and pulsing frequency at a constant sputtering power 3 kW. Three different values of P
40%, 60%, and 80% and four different pulsing frequencies
( f 50, 150, 200, and 250 kHz were used. The electrical
characterization of the titanium oxide films was performed
using an Al/TiOx /Al parallel-plate capacitor structure.
Figure 3 summarizes the results of the electrical properties of the titanium oxide films as a function of P and pulsing
frequencies. All electrical properties of the titanium oxide
films show similar trend as a function of P. As can be seen in
Figs. 3a, 3b, and 3c, there is a minimum in the dielectric
constant and leakage current density and a maximum in the
breakdown field strength at around 60% O2 regardless of
pulsing frequency.
Figure 3a shows the measured dielectric constant of the
parallel-plate capacitor structure. They ranged from 25.3
( P60% and f 200 kHz) to 65.9 ( P40% and f
150 kHz). As expected, the dielectric constant is smaller
than that of the bulk titanium oxide 170, but is comparable to the result of thin films prepared by PECVD
60,16 electron-beam evaporation 39,24 CVD
2086,13 dc sputtering 95,9 and low-pressure CVD
70.15 The deviation between the bulk and the reported
dielectric constant in thin-film titanium oxide may be caused
by a number of reasons such as an interfacial layer, which
was formed during an annealing process or deposition
process,13,16,24 different crystalline phase formations, or the
growth of preferred orientation of the grain structure.15 The
dielectric constant measured in our experiment was mainly
from the titanium oxide layer only because there was no
postannealing process in our experiment so that the interfacial layer contribution would be small and the films were
noncrystalline.
Figure 3b shows the leakage current density of the titanium oxide films measured at 0.5 MV/cm electric field. For
the leakage current measurement, the bias was put on the top
Al electrode, the so-called cathodic polarization.20 The
leakage current density ranged from 0.2 A/cm2 ( P60%
and f 200 kHz) to 7.84 mA/cm2 ( P40% and f
150 kHz). The leakage current density at 50 kHz pulsing
frequency is not shown in Fig. 3b because the breakdown
observed was below the 0.5 MV/cm electric field. Typically,
the leakage current density of titanium oxide films deposited
at high temperature or with postannealing was less than
1010 107 A/cm2 . For example, the leakage current den-

432

Kim et al.: Frequency-dependent pulsed dc magnetron sputtering

FIG. 3. Electrical properties of the titanium oxide films as a function of

oxygen partial pressure for different pulsing frequency: a dielectric constant, b leakage current density, and c dielectric breakdown field. The
sputtering power was 3.0 kW.

J. Vac. Sci. Technol. A, Vol. 19, No. 2, MarApr 2001

432

sity at 0.5 MV/cm for the metalinsulatorsemiconductor

to
MIS
structure
changed
from
101 A/cm2
5
8
2
10 10 A/cm after annealing prepared by metal
organic CVD Ref. 25 and at higher temperature deposition
from 10 A/cm2 400 C deposition to 106 108 A/cm2
680 C deposition.24 For the metalinsulatormetal structo
ture,
it
changed
from
103 104 A/cm2
7
8
2
10 10 A/cm prepared by reactive sputtering after
annealing.9 The leakage current density (107 A/cm2 at
0.5 MV/cm obtained in our experiment was less than those
films prepared without postannealing or at the hightemperature deposition temperature reported in the literature.
The low oxygen partial pressure in the sputtering atmosphere would produce inferior electrical properties of the titanium oxide films due to insufficient oxygen bonding imperfect oxidation in the film.14,20 It is critical to supply
sufficient oxygen to grow the film during deposition. If not,
additional annealing is necessary to form a fully oxidized
film. The sample with high leakage current also gives high
loss when compared to that of the 60%, 0.315 ( P40%)
and 0.005 ( P60%) at 200 kHz pulsing frequency. The
oxidation mode in the film depends on the cathode potential
during sputtering. A reactive mode on the target in reactive
sputtering may result in an injection of titanium oxide complex from the target. In the metallic mode, elemental metal
ejected from the target resulted in an increase of leakage
current density. We recently found that there was not much
difference between 60% and 80% O2 sputtering atmosphere
for tantalum oxide films in their electrical properties.20 However, in the present case of titanium oxide films, the leakage
current density and dielectric constant both increase at
80% O2 .
The main purpose of pulsing is to discharge the insulating
layer on the target and prevent the breakdown of the insulator arcing on the target. If arcing happened, unwanted particles are gettered to the substrate. The lower limit of the
pulsing frequency was determined to prevent the electrical
breakdown arcing of the oxide layer on the target. Recently, Belkind and co-workers suggested that the pulsing
frequency should be higher than the critical frequency to
prevent arcing for Al2O3 , in which the critical frequency lies
in the range of 1250 kHz.18 In our experiment, electrical
results leakage current density, loss tangent, and dielectric
breakdown strength showed that the 200 kHz pulsing frequency can produce a better leakage current density over
other frequencies. At 200 kHz, the sputtering is stable and
would produce none or minimum arcing during deposition.
Figure 3c shows the dielectric breakdown field as a
function of P. The breakdown was defined as an abrupt current increase to 1102 A) when a voltage was supplied to
the top Al electrode. The maximum breakdown field was
1.35 MV/cm for P60 and f 200 kHz and the minimum
was 0.46 MV/cm for P40 and f 50 kHz. The typical I V
behavior of the titanium oxide films are shown in Fig. 4.
Because of different film thickness, the leakage current density was plotted as a function of an electric field instead of
the bias voltage.

433

Kim et al.: Frequency-dependent pulsed dc magnetron sputtering

433

FIG. 4. Representative I V curves of the titanium oxide films as a function of oxygen partial pressure at a 50 kHz, b 150 kHz, c 200 kHz, and d 250
kHz pulsing conditions.

A figure of merit FM, which was defined as the product of the dielectric constant and breakdown field, can be
used to determine the quality of the dielectric films.20 The
FM value represents the maximum charge storage density of
the films. The FM of the titanium oxide films ranged from 19
( P40% and f 50 kHz) to 50 ( P40% and f
150 kHz). Although the FM showed a maximum value at
P40% and f 150 kHz, the leakage current density was
still high 7.8 mA/cm2. A minimum leakage current can be
obtained at P60% and f 200 kHz FM35.
IV. CONCLUSIONS
The pulsed reactive dc magnetron sputtering technique
was used to deposit titanium oxide films with different oxygen partial pressure P in the sputtering atmosphere and
pulsing frequency f. Two different sputtering modes metallic and reactive mode were identified as a function of
oxygen partial pressure and pulsing frequency from the
change of the cathode voltage, which is related to the oxide
coverage of the target. For titanium oxide thin films, the
oxidation state of the target mainly depends on the oxygen
JVST A - Vacuum, Surfaces, and Films

partial pressure in the sputtering atmosphere, but not much

on the pulsing frequency. The transition region from the metallic to reactive mode occurs at 6%10% O2 . The O/Ti
ratio of the deposited titanium oxide film is around 2.22.4.
For the pulsing frequency ranging from 50 to 200 kHz, a
dramatic dependence of the dielectric properties was found
for the titanium oxide films. In this work, a large dielectric
constant, 65.9, was found at P40% O2 and f 150 kHz,
but the corresponding leakage current density was also very
high 7.8 mA/cm2. The minimum leakage current density
was 0.2 A/cm2 at 0.5 MV/cm electric field at P60% O2
and f 200 kHz. It also showed high breakdown strength
1.3 MV/cm. Increasing the pulsing frequency from 50 to
200 kHz improved the dielectric properties of the titanium
oxide film, but not for 250 kHz. The FM of the titanium
oxide films ranged from 19 ( P40% and f 50 kHz) to 50
( P40% and f 150 kHz). From the electrical characterization, it was found that the best titanium oxide films could be
obtained at a preferred pulsing frequency and oxygen partial
pressure ( P60% and f 200 kHz). At 200 kHz, the

434

Kim et al.: Frequency-dependent pulsed dc magnetron sputtering

sputtering process would produce none or a minimum of

arcing during deposition, resulting in the best electrical property.
ACKNOWLEDGMENTS
This work is supported in part by Microcoating Technology MCT, Microelectronics Advanced Research Corporation MARCO, and ENI. The authors would like to thank
Dr. Atul Kumar for the RBS and Dr. G.-R. Yang for the XPS
measurement.
H. K. Pulker, G. Paesold, and E. Ritter, Appl. Opt. 15, 2986 1976.
M. H. Suhail, G. Mohan Rao, and S. Mohan, J. Appl. Phys. 71, 1421
1992.
3
P. Lobl, M. Huppertz, and D. Mergel, Thin Solid Films 251, 72 1994.
4
N. Martin, D. Baretti, C. Rousselot, and J.-Y. Rauch, Thin Solid Films
107, 172 1998.
5
L.-J. Meng, M. Andritschky, and M. P. dos Santos, Thin Solid Films 223,
242 1993; L.-J. Meng and M. P. dos Santos, ibid. 226, 22 1993; L.-J.
Meng and M. P. dos Santos, Appl. Surf. Sci. 68, 319 1993.
6
W. T. Pawlewicz and R. Busch, Thin Solid Films 63, 251 1979.
7
K. Bange, C. R. Ottermann, O. Anderson, U. Jeschkowski, M. Laube, and
R. Feile, Thin Solid Films 197, 279 1991.
8
H.-K. Ha, M. Yoshimoto, H. Koinuma, B.-K. Moon, and H. Ishiwara,
Appl. Phys. Lett. 68, 2965 1996.
9
B. H. Lee, Y. Jeon, K. Zawadzki, W.-J. Qi, and J. Lee, Appl. Phys. Lett.
74, 3143 1999.
10
S. A. Campbell, D. C. Gilmer, X.-C. Wang, M.-T. Hsieh, H.-S. Kim, W.
1
2

J. Vac. Sci. Technol. A, Vol. 19, No. 2, MarApr 2001

434

L. Gladfelter, and I. Yan, IEEE Trans. Electron Devices 44, 104 1997;
H.-S. Kim, S. A. Campbell, and D. C. Gilmer, IEEE Electron Device Lett.
18, 465 1997.
11
M. D. Wiggins, M. C. Nelsen, and C. R. Aita, J. Vac. Sci. Technol. A 14,
772 1996.
12
K. Yokota, T. Yamada, T. Sasagawa, K. Nakamura, and F. Miyashita,
Thin Solid Films 343-344, 138 1999.
13
T. Fuyuki and H. Matsunami, Jpn. J. Appl. Phys. Part 1 25, 1288 1986.
14
M. Takeuchi, T. Itoh, and H. Nagasaka, Thin Solid Films 51, 83 1978.
15
N. Rausch and E. P. Burte, J. Electrochem. Soc. 140, 145 1993.
16
Y. H. Lee, K. K. Chan, and M. J. Brady, J. Vac. Sci. Technol. A 13, 596
1995.
17
K. A. Vorotilov, E. V. Orlova, and V. I. Ptrovsky, Thin Solid Films 207,
180 1992.
18
W. D. Sproul, M. E. Graham, M. S. Wong, S. Lopez, D. Li, and R. A.
Scholl, J. Vac. Sci. Technol. A 13, 1188 1995; A. Belkind, A. Freilich,
and R. Scholl, ibid. 17, 1934 1999.
19
P. J. Kelly and R. D. Arnell, J. Vac. Sci. Technol. A 17, 945 1999.
20
J.-Y. Kim, M. Nielsen, E. Rymaszewski, and T.-M. Lu, J. Appl. Phys. 87,
1448 2000; M. Nielsen, J.-Y. Kim, E. Rymaszewski, and T.-M. Lu,
IEEE Control Syst. Mag. 21, 274 1998.
21
W. M. Posadowski, Thin Solid Films 343-344, 85 1999.
22
S. Schiller, U. Heisig, K. Steinfelder, and J. Strumfel, Thin Solid Films
63, 369 1979.
23
M. Radecka, K. Zakrezewska, H. Czternastek, T. Stapinski, and S. Debrus, Appl. Surf. Sci. 6566, 227 1993.
24
J. V. Grahn, M. Linder, and E. Fredriksson, J. Vac. Sci. Technol. A 16,
2495 1998.
25
H.-S. Kim, S. A. Campbell, D. C. Gilmer, V. Kaushik, J. Conner, L.
Prabhu, and A. Anderson, J. Appl. Phys. 85, 3278 1999.