Physical and electrical characterization of ZrO2 gate insulators deposited on Si(100) using Zr(Oi-Pr)2(thd)2 and O2

Article abstract of DOI:10.1149/1.1471891

The characteristics of ultrathin ZrO₂ films deposited using molecular oxygen and the zirconium precursor Zr(Oⁱ-Pr)₂(thd)₂ [where Oⁱ-Pr is isopropoxide and the is 2,2,6,6-tetramethyl-3,5-heptanedionate] were investigated. The organometallic was dissolved as a 0.15 M solution in octane and introduced into the deposition chamber using a liquid injection system. The deposition rate was insensitive to molecular oxygen flow but changed with liquid injection rate and was thermally activated in the range 390-550°C. No evidence of Zr-C and Zr-Si bonds were found in the X-ray photoelectron spectroscopy (XPS), spectra, and carbon concentrations, <0.1 atom %, the detection limit of the XPS depth profiling measurements, were obtained at the lowest deposition temperatures and deposition rates. High-resolution transmission electron microscopy showed the ZrO₂ films to be polycrystalline as deposited, with an amorphous zirconium silicate interfacial layer. The effects of postdeposition annealing were also demonstrated. After proper annealing treatments, promising capacitance-voltage and current voltage characteristics were achieved. A film with an equivalent oxide thickness of 2.3 nm showed current reductions of approximately two orders of magnitude when compared to SiO₂, but some improvements are required if these films are to be used as a gate-insulator beyond the 100 nm CMOS (complementary metal oxide semiconductor) technology node.

Full text of DOI:10.1149/1.1471891

Journal of The Electrochemical Society, 149 ͑6͒ F49-F55 ͑2002͒
F49
0
013-4651/2002/149͑6͒/F49/7/$7.00 © The Electrochemical Society, Inc.
Physical and Electrical Characterization of ZrO Gate
2
Insulators Deposited on Si„100… Using
i
Zr„O -Pr… „thd… and O
2
2
2
a,z
a,b
c,
c
c
c
H.-W. Chen, T.-Y. Huang, D. Landheer, * X. Wu, S. Moisa, G. I. Sproule,
and T.-S. Chao
b,d
a
Institute of Electronics Engineering, National Chiao-Tung University, Hsinchu 300, Taiwan
National Nano Device Laboratories, Hsinchu 300, Taiwan
b
c
Institute for Microstructural Sciences, National Research Council of Canada, Ottawa,
Ontario K1A 0R6, Canada
d
Department of Electrophysics, National Chiao-Tung University, Hsinchu 300, Taiwan
i
The characteristics of ultrathin ZrO films deposited using molecular oxygen and the zirconium precursor Zr͑O -Pr) ͑thd͒ ͓where
2
2
2
i
O -Pr is isopropoxide and thd is 2,2,6,6-tetramethyl-3,5-heptanedionate͔ were investigated. The organometallic was dissolved as
a 0.15 M solution in octane and introduced into the deposition chamber using a liquid injection system. The deposition rate was
insensitive to molecular oxygen flow but changed with liquid injection rate and was thermally activated in the range 390-550°C.
No evidence of Zr-C and Zr-Si bonds were found in the X-ray photoelectron spectroscopy ͑XPS͒, spectra, and carbon concentra-
tions, Ͻ0.1 atom %, the detection limit of the XPS depth profiling measurements, were obtained at the lowest deposition
temperatures and deposition rates. High-resolution transmission electron microscopy showed the ZrO films to be polycrystalline
2
as deposited, with an amorphous zirconium silicate interfacial layer. The effects of postdeposition annealing were also demon-
strated. After proper annealing treatments, promising capacitance-voltage and current-voltage characteristics were achieved. A film
with an equivalent oxide thickness of 2.3 nm showed current reductions of approximately two orders of magnitude when compared
to SiO , but some improvements are required if these films are to be used as a gate-insulator beyond the 100 nm CMOS
2
͑complementary metal oxide semiconductor͒ technology node.
©
2002 The Electrochemical Society. ͓DOI: 10.1149/1.1471891͔ All rights reserved.
Manuscript submitted August 31, 2001; revised manuscript received December 2, 2001. Available electronically April 12, 2002.
As the dimensions of complementary metal oxide semiconductor
CMOS͒ devices are scaled in the deep submicrometer regime, the
In order to combine the advantages of alkoxides and
␤-diketonates, the mixed ligand zirconium precursor was developed
to lower the thermal stability but with the higher ambient stability
and higher volatility. This paper reports the first investigation of the
physical and electrical properties of ultrathin ZrO2 films deposited
using molecular oxygen and a recently developed six-coordinate
͑
thickness of the required gate dielectric will steadily decrease to
1
dimensions Ͻ1 nm. For SiO , however, the large resulting direct
2
tunneling currents result in large power consumption and reduced
2
,3
reliability. With higher dielectric constant, ␬, the gate insulator
thickness can increase, while maintaining the equivalent oxide
thickness ͑EOT͒, significantly reducing the tunneling leakage cur-
rent. Recently, high-␬ materials predicted to be thermodynamically
i
i
mixed alkoxide/␤-diketonate compound, Zr͑O -Pr) (thd) ͑O -Pr is
2
2
18
isopropoxide and thd is 2,2,6,6-tetramethyl-3,5-heptanedionate͒,
introduced into the chamber with a liquid injection system. The
films were analyzed by high-resolution transmission electron mi-
croscopy ͑HRTEM͒ and X-ray photoelectron spectroscopy ͑XPS͒.
Pure, stoichiometric ZrO2 films were deposited and the effects of
postdeposition annealing investigated. Low leakage current density
vs. voltage ͑J-V͒ characteristics and capacitance-voltage ͑C-V͒
characteristics with small hysteresis are demonstrated.
4
stable with Si, such as ZrO , HfO , La O , Al O , and their sili-
2
2
2
3
2
3
cates, have been considered as candidates for alternative gate
5
dielectrics. ZrO has a high dielectric constant ͑ϳ25͒, high bandgap
2
5,6
energy ͑5.8-7.8 eV͒, and thermal compatibility with contemporary
CMOS processes.
7
-9
Although electron-beam evaporation or sputtering are useful
techniques for depositing new materials for evaluation, they can
damage electrical devices, have difficulties with uniformity and pu-
rity, and are hard to maintain in production, resulting in low
throughput. Thus, chemical vapor deposition ͑CVD͒ is usually pre-
ferred and, in this paper, organometallic CVD ͑MOCVD͒ was used
Experimental
Si͑100͒ substrates, 100 mm diam, n-type ͑␳ ϭ 0.02-0.06 ⍀ cm͒
were given an HF-last RCA clean prior to film deposition. The CVD
chamber was equipped with a 360 L/s turbomolecular pump and a
liquid injection system ͑LDS-300B produced by ATMI͒. The latter
consisted of a liquid pump to pump the precursor, a 0.15 M solution
1
0
to deposit ZrO . Since zirconium chloride ͑ZrCl ͒ and alkoxide
2
4
t
i
9,11,12
͓
Zr͑O -Bu) and Zr͑O -Pr) ͔
precursors contain an unsaturated
four coordinate Zr center, they are air and moisture sensitive. For
4
4
IV
i
of Zr͑O -Pr) (thd) in octane, through a hot glass frit at a rate of 0.2
2
2
zirconium ␤-diketonates ͓e.g., Zr(acac) , Zr(thd) , and Zr(tfac) ͔,
mL/min. The vapors were carried with a 50 sccm flow of Ar to a gas
distribution ring 10 cm from the substrate. The glass frit, the com-
ponents of the vaporizer, the gas ring, and the connecting tube were
maintained at a temperature of 190°C with heating tapes and blan-
kets, while the substrate temperature was controlled in the range
390-550°C with quartz-halogen lamps and a thermocouple. Oxygen
was introduced into the chamber at flow rates of 0-150 sccm through
a separate gas distribution ring 30 cm from the substrate. Just prior
to deposition the wafers were heated for 10 min at 500°C in 10
4
4
4
the chelating ligand ͑␤-diketonate͒ can increase the coordination
IV
number of Zr center to eight to achieve the coordination satura-
tion. Although it makes the Zr coordination compound less ambient
sensitive, the compound has higher thermal stability, requiring
higher deposition temperatures, and films exhibit severe carbon or
fluorine contamination.1
2,13
Thus, a mature MOCVD process re-
quires precursors, stable upon exposure to atmosphere with higher
vapor pressure, to replace the chloride or alkoxide precursors usu-
ally employed. Most recently, the anhydrous zirconium nitrate
mTorr O to replace the surface hydrogen termination with oxygen.
2
14-17
Zr͑NO ) was also investigated for gate dielectrics.
Films were analyzed by X-ray photoelectron spectroscopy ͑XPS͒
using a PHI 5500 system with a monochromatic Al K␣ X-ray source
in a standard 90° geometry ͑X-ray source and electron spectrometer
3
4
4
5° off normal͒ and a bandpass energy of 58.7 eV. Depth profiling
*
Electrochemical Society Active Member.
E-mail: hwchen.ee86g@nctu.edu.tw
z
ϩ
was accomplished with intermittent Ar sputtering at 1 keV, 50
Downloaded on 2015-04-12 to IP 128.122.253.228 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F50
Journal of The Electrochemical Society, 149 ͑6͒ F49-F55 ͑2002͒
Figure 1. The deposition rate vs. oxygen flow for films deposited in con-
tinuous mode at ͑-᭹-͒ 500 and ͑-᭿-͒ 650°C. The precursor was injected at
0
.05 mL/min with 50 sccm Ar.
Figure 3. Logarithmic plot of the deposition rate vs. inverse temperature.
The activation energy determined from the slope is 1.27 eV.
nA, and 45° incidence. Thick films were typically stoichiometric
ZrO when analyzed by Rutherford backscattering. These were used
2
time from the known optical constants for Si and assuming the film
was uniform, nonabsorbing, and had a refractive index of 2. This
results in an approximate thickness measurement that must be cali-
brated by ex situ analysis.
to calibrate the sensitivity factors for the XPS analysis of thinner
films using the Zr 3d and O 1s peaks. The peak positions were
referenced to the main surface carbon peak at 284.8 eV.
Since the liquid pump was not sufficiently stable at flow rates
below 0.2 mL/min, to reduce deposition rates and further reduce the
carbon contamination, most depositions were done in ‘‘pulse mode’’
in which the precursor and oxygen were introduced separately with
intervening pumping periods of 15 s. Nitrogen at a flow rate of 100
sccm was introduced into the oxygen gas ring and flowed during the
complete deposition cycle. Thus, each cycle of deposition consisted
Atomic force microscopy ͑AFM͒ measurements were made on
the as-deposited and annealed samples using a Digital Instruments
Nanoscope III operating in tapping mode with 5 nm diam silicon
probes. Cross-sectional samples were made using a low-angle ion-
milling technique. The structure of the films and the effect of an-
nealing was observed directly in HRTEM micrographs made on a
Philips EM-430T instrument operating at 250 keV.
of four stages: 100 sccm N , 150 sccm O ϩ 100 sccm N , 100
2
2
2
Samples were annealed in a Heatpulse 610 ͑Steag RTP Systems͒
sccm N , and 50 sccm Ar ϩ 100 sccm N ϩ Zr precursor ͑0.2 m
2
2
rapid thermal processor. For spike anneals in O , the temperature
2
L/min͒. Pressures at various stages of the deposition cycle were in
the 8-12 mTorr range.
The depositions were monitored in situ with a Rudolph i1000
ellipsometer operating at a wavelength of 633 nm. The ellipsometric
angles were used to calculate the deposited film thickness in real-
was ramped at 125°C/s to 800, 850, 900, or 950°C and held for 1 s
before the quartz lamps were turned off. Forming gas anneals
͑
FGAs͒ were done in a 4% H in N mixture for 10-30 min in the
2 2
temperature range 380-600°C.
For the electrical measurements, Al-gated capacitors were made
by evaporating aluminum through a shadow mask and back contacts
were made with In-Ga eutectic. The FGAs were done before metal
deposition to avoid reaction of the Al with the films. The area of the
capacitors was measured to an accuracy of Ϯ2% with a calibrated
digitizing camera. Electrical measurements were made by probing
the Al gates in a probe-station attached to two instruments, a multi-
frequency LCR meter ͑HP model 4275A͒ for C-V characteristics,
and a picoammeter/dc voltage source ͑HP model 4140B͒ for current-
voltage characteristics. High-frequency C-V measurements were
made by stepping with 0.1 V steps each second, first from Ϫ2.5 to
ϩ2.5 V and then in the reverse direction to look for hysteresis. The
equivalent oxide thickness EOT was obtained from the 100 kHz
C-V characteristics by using the NCSU C-V fitting routine,19 which
includes quantum effects in the channel.
Results and Discussion
Physical characteristics.—Initial films were deposited in the
continuous mode in which the oxygen and precursor were intro-
duced simultaneously. Figure 1 shows the deposition rate as a func-
tion of oxygen flow rate for depositions at 500 and 650°C using a
continuous flow of precursor of 0.05 mL/min. The deposition rate is
insensitive to molecular oxygen flow at 500°C; however, turning off
the oxygen does significantly reduce the deposition rate at both tem-
peratures. With an oxygen flow of 100 sccm the deposition rate was
proportional to precursor flow rate over the range of 0.05-0.20 mL/
Figure 2. Film thickness as a function of time obtained from the in situ
ellipsometer during one cycle of deposition for substrate temperatures 470,
5
10, and 550°C.
Downloaded on 2015-04-12 to IP 128.122.253.228 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, 149 ͑6͒ F49-F55 ͑2002͒
F51
Figure 5. More XPS spectra for the film of Fig. 4: ͑a͒ Zr 3d peak, and ͑b͒ Si
p peak. In ͑b͒ the peak near 103 eV shifted 3.3 eV from the substrate peak
2
is associated with a silicate layer at the Si interface. The solid curves are a
guide to the eye.
Carbon concentrations tended to be lower for the lowest pump
rates, but subsequent depositions were done at 0.20 mL/min to avoid
periodic instabilities in the pumping rate caused by back-pressure
instabilities in the liquid injection pump at the lower speeds. To
reduce carbon contamination, subsequent depositions were done in
the pulse mode.
For our flow parameters, the film thickness depends on the num-
ber of deposition cycles, substrate temperature, O injection time,
2
and precursor injection time during pulse-mode deposition. In the
range of temperature and pressure investigated here no saturation of
the precursor absorption was noted, i.e., the film deposition rate did
not saturate as the precursor injection time was increased. Further-
more, when the oxygen was turned off the deposition rate did not
decrease significantly in the pulse mode. This indicates that atomic
layer deposition ͑ALD͒ was not occurring, possibly because back-
ground oxygen from the process was sufficient to effect the oxida-
tion. Since the Zr is coordinated to six oxygens in the precursor, it is
Figure 4. XPS spectra showing data ͑-᭹-͒ and ͑---͒ fitted peaks for the ZrO2
film deposited with 75 cycles at 390°C: ͑a͒ C 1s ͑surface͒, ͑b͒ O 1s ͑surface͒,
and ͑c͒ O 1s ͑after two cycles of sputtering͒. The solid lines are a guide to the
eye in ͑a͒ and represent the sum of the fitted peaks in ͑b͒ and ͑c͒.
min. Depth profiles of the C 1s peaks showed that the films depos-
ited in the continuous mode were contaminated with carbon, with
the smallest contamination in the 1-2 atom % range, which is con-
sistent with the value reported in Ref. 12 at the lower temperature
where the deposition rates were lowest.
also possible that an oxidant is not necessary to form ZrO . How-
2
ever, without oxygen the carbon in the films was significant, so films
were typically deposited with an oxygen injection time of 18 s and
Downloaded on 2015-04-12 to IP 128.122.253.228 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F52
Journal of The Electrochemical Society, 149 ͑6͒ F49-F55 ͑2002͒
Figure 6. AFM image for the film of Fig. 4. The rms surface roughness is
.4 Å.
4
a precursor injection time of 43 s. With this ‘‘pulse mode’’ depo-
sition, O can react with the film and reduce the carbon in the film
2
effectively.
The thickness output by the in situ ellipsometer during one cycle
is shown in Fig. 2 for substrate temperatures of 470, 510, and
550°C. The film thickness increases during the precursor injection
but seems to decrease during oxygen flow. The latter decrease is
recovered during the subsequent pump-down ͑period with nitrogen
flow only͒ and may be associated with adsorption and desorption of
oxygen on the surface. A plot of the log of the average deposition
rate as a function of inverse temperature is shown in Fig. 3. The
activation energy determined from the slope of the fit to the data in
the range 390-550°C is 1.27 eV.
In Fig. 4, the XPS C 1s ͑surface͒, O 1s ͑surface͒, and O 1s ͑after
two cycles of sputtering͒ spectra for the ZrO film deposited with 75
2
cycles at 390°C are shown. The C 1s spectrum of Fig. 4a reveals
carbon peaks at binding energies of 284.8 and 288.5 eV, correspond-
ing to C-C and C-O bonds, respectively. No lower binding energy
peak characteristic of the Zr-C bond was found. After two cycles of
sputtering these peaks disappeared in the films deposited at 390 and
470°C, indicating that the carbon concentrations in these films were
less than the detection limit of our instrument, i.e., Ͻ0.1 atom %.
2
0
Guittet et al. found the binding energy, E , for the O 1s peak
b
Figure 7. HRTEM cross-sectional images of the film of Fig. 4: ͑a͒ as-
to be E ϭ 530.0 eV for ZrO2 and E ϭ 531.3 eV for ZrSiO .
b
b
4
deposited, ͑b͒ after spike anneal at 850°C for 1 s in O followed by FGA at
2
These energies are close to the fitted components of the O 1s spectra
shown in both Fig. 4b and c. In Fig. 4b, the O 1s spectrum of the
as-deposited film has a component peak at the binding energy of
380°C for 20 min.
5
31.4 eV. Because the O 1s binding energies of the Zr silicate,
21
hydroxide, and carbonate are in the same range as the higher en-
ergy peak, some of this peak in Fig. 4b is due to the H-O and C-O
bonds from remnant precursor and from water vapor adsorbed dur-
ing transfer to the XPS system. After two cycles of sputtering, the O
reported in Ref. 20. Figure 5b shows the Si 2p spectrum of the same
sample shown in Fig. 5a. The peak near the binding energy of ϳ103
eV is 3.3 eV from the Si substrate peak, a chemical shift smaller
22
than the Ͼ4 eV shift observed for 1.8 nm thick SiO films. This is
2
1
s spectrum, shown in Fig. 4c, has a main peak with binding energy
characteristic of a silicate layer consistent with the TEM image and
C-V data shown later. In addition, the residual carbon contamination
for this film is less than the detection limit. The root mean square
of 530.7 eV, close to that expected for ZrO . The higher energy
2
peak appearing in Fig. 4b is diminished because there is no water
vapor and the residual carbon in the film is Ͻ0.1 atom %. Its bind-
ing energy of 531.3 eV is associated with the Zr-O-Si bonds in the
interfacial layer.
͑rms͒ surface roughness obtained from the AFM image of the same
film ͑shown in Fig. 6͒ is 4.4 Å.
In Fig. 7a and b, an amorphous interfacial layer is evident
between the substrate and the polycrystalline ZrO2 layer in the
HRTEM image of the film. This layer is identified as the silicate
layer responsible for the Si 2p feature in the XPS spectrum. The
interface between silicate and Si is atomically sharp. After spike
Figure 5a shows the Zr 3d spectrum of the same ZrO film de-
2
posited with 75 cycles at 390°C. No evidence of Zr-Si bond forma-
tion is observed from Fig. 5a. The binding energy difference E ͑O
1
b
s͒-Eb ͑Zr 3d5/2) ϭ 348.4 eV is close to the value of 348.1 eV
Downloaded on 2015-04-12 to IP 128.122.253.228 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, 149 ͑6͒ F49-F55 ͑2002͒
F53
Figure 8. High-frequency C-V characteristics of capacitors with ZrO2 gate
dielectrics obtained at 100 kHz. The samples were spike annealed for 1 s in
O2 at 850°C first, then annealed in forming gas at 380°C for 10-30 min:
͑-૽-͒ no anneals, ͑-᭢-͒ spike anneal only, ͑-᭹-͒ 10 min FGA, ͑-᭡-͒ 20 min
FGA, and ͑-᭿-͒ 30 min FGA.
anneal in O at 850°C and FGA, the thickness of the silicate layer
2
was increased by 2 Å, as shown in Fig. 7b. This is due to oxygen
penetration to the Si substrate since ZrO is a fast ion conductor
2
6
with significant oxygen ion diffusivity. The most likely mechanism
for silicate formation is that the oxygen reacted with the Si substrate
to grow SiO , and SiO reacted with ZrO to form Zr silicate. Films
2
2
2
were grown without injecting oxygen and, as stated previously, this
did not significantly reduce the film deposition rate. It also did not
reduce the thickness of the interfacial silicate layer, further indica-
tion that oxygen released from the precursor is available for surface
reactions.
Electrical characteristics.—Electrical analysis was performed on
the ZrO film formed with 75 deposition cycles at 390°C. The 100
2
kHz C-V characteristics of ZrO films deposited are shown in Fig. 8
2
for different annealing treatments. For the as-deposited sample,
there is a significant hump in the C-V curve in accumulation, a
flatband voltage VFB ϭ Ϫ0.06 V ͑scan from inversion to accumula-
tion͒, corresponding to a large negative fixed charge density ͑N
ϭ Ϫ1.1 ϫ 10 /cm ͒, and a significant clockwise hysteresis given
by ⌬VFB ϳ 0.21 V. The hysteresis is associated with an effective
f
1
2
2
density of trapped charge at the interface, ⌬N ϭ Ϫ1.7
t
1
2
2
ϫ 10 /cm , attributed to charge trapping or detrapping. It is be-
lieved that the appearance of the hump is due to P centers at the
b
interface.23 For the sample spike annealed for 1 s in O at 850°C, the
2
12
2
flatband voltage shifts to Ϫ0.76 V ͑N ϭ 6.0 ϫ 10 /cm ͒, the hys-
f
teresis becomes smaller, but the hump still exists. Thus O anneal-
2
ing reduces the charge trapping-detrapping ͑to ⌬N ϳ Ϫ1.6
t
1
1
2
ϫ 10 /cm ͒ but does not passivate the P centers effectively.
b
The samples annealed in forming gas at 380°C were first spike
Figure 9. The variation of EOT and gate current density ͑J͒ as a function of
postdeposition annealing treatments. ͑a͒ 1 s spike anneal in O at 850°C from
annealed at 850°C for 1 s in O . For these samples the P centers
2
2
b
1
0
800 to 950°C, ͑b͒ FGA from 380 to 600°C for 10 min, and ͑c͒ spike anneal
for 1 s in O2 at 850°C first, then FGA at 380°C for 10-30 min; ͑-᭹-͒ EOT,
and ͑-᭡-͒ J for V -V ϭ 1 V.
are passivated. For the 20 min FGA, ⌬N is reduced to 5.0 ϫ 10 /
cm . For the 30 min FGA the flatband voltage shifts toward the ideal
t
2
G
FB
flatband voltage ͑for this sample, ideal V ϭ Wms ϳ Ϫ0.2 V,
FB
24
where Wms is the work function difference͒. The fixed charge den-
11
2
sity is reduced to 2.6 ϫ 10 /cm . This may be due to the introduc-
tion of negative fixed charge or the annealing of the positive fixed
Figure 9 shows the variation of the EOT and the gate current
density, J, after different postdeposition annealing treatments. The
value of J was obtained at a value of gate voltage, VG , for which
V -V ϭ 1 V. The EOT of the as-deposited ZrO film is 27 Å, so
charge. However, the hysteresis is increased to ⌬N ϭ Ϫ5.5
t
1
0
2
ϫ 10 /cm , so there appears to be a trade-off between trapped and
fixed charge.
G
FB
2
Downloaded on 2015-04-12 to IP 128.122.253.228 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F54
Journal of The Electrochemical Society, 149 ͑6͒ F49-F55 ͑2002͒
leakage current for the 30 min FGA is larger than expected from the
variation in EOT, and this may again be related to hydrogen effects
on the grain boundaries.
Figure 10 shows the 10 kHz and 100 kHz C-V curves of the
ZrO capacitor treated with spike annealing at 850°C for 1 s in O2
2
followed by FGA at 380°C for 20 min. The area of the capacitor is
Ϫ4
2
6.6 ϫ 10
cm . From the 100 kHz C-V curve, the EOT is 22.7 Å
with the quantum effects correction. The effective dielectric constant
of the whole dielectric layer is ϳ9.3, which is lower than that of
ZrO because the equivalent oxide capacitance is dominated by the
2
capacitance of the silicate layer. In addition, there is only slight
frequency dispersion in the C-V curves, mainly below midgap. This
indicates that frequency-dependent interface states are passivated
during the annealing. The interface state density is estimated to be
1
1
Ϫ2
Ϫ1
5
.0 ϫ 10 cm eV by the method described in Ref. 25, which
compares the 10 kHz and ideal C-V characteristics.
As shown in Fig. 11, the J-V curve for the sample in Fig. 10 is
asymmetric. The leakage characteristics were measured under light
illumination to avoid the minority carrier generation current limiting
the leakage current at inversion. The leakage current densities are
Figure 10. Frequency dependence of the C-V characteristics of the film of
6 ϫ 10 A/cm at 1 V ͑substrate injection͒ and 4.7 ϫ 10 A/cm2
Ϫ4
2
Ϫ6
Fig. 4 which has an EOT of 2.27 nm after spike anneal at 850°C in O for 1
at Ϫ1 V ͑gate injection͒, respectively. For a pure SiO film with a
2
2
s and FGA at 380°C for 20 min: ͑-᭿-͒ 10 and ͑-᭹-͒ 100 kHz. A slight
frequency dispersion is evident below midgap. The hysteresis ͑⌬VFB Ϸ 6
mV͒ is not shown.
EOT 22.7 Å, the leakage current density is about two orders of
3
magnitude larger for a gate potential VG ϭ VFB ϩ 1 V. However,
the leakage is larger than expected from tunneling and does not fit
the equations for Frenkel-Poole emission or Schottky emission.26
This might be due to the difference of the barrier heights under
different voltage polarity or might be due to a peaked distribution of
traps in the dielectric that favors trap-assisted tunneling for substrate
either spike anneal or FGA can make the EOTs lower, i.e., the ef-
fective dielectric constant of the whole gate dielectric becomes
higher. As shown in Fig. 9a, the EOT is without significant change at
27
injection. Further investigation of the leakage mechanism is nec-
essary, but reducing the film surface roughness observed by HRTEM
and AFM should improve performance.
800-850°C and increases at 900-950°C. The increase of EOT at the
highest spike annealing temperatures is associated with the oxygen
penetration to the Si substrate, which increases the interfacial layer
thickness as mentioned above. The accompanying reduction of the
leakage current is due to the larger physical thickness and the im-
proved robustness of the ZrO2 films. However, the samples only
treated with FGA are quite different, as shown in Fig. 9b. The leak-
age current increases with the FGA temperature, and the EOT shows
a tendency to decline. The increasing leakage current might be re-
Conclusions
Ultrathin stoichiometric, carbon-free ZrO2 films can be deposited
on Si͑100͒ using the novel precursor Zr͑O -Pr)2(thd)2 dissolved in
octane. O introduced into the chamber is necessary to reduce the
carbon contamination. In XPS data, no evidence of Zr-C and Zr-Si
bonds were found. Meanwhile, XPS and HRTEM analysis shows
that a Zr silicate layer is formed at the interface during deposition.
Although the silicate layer reduces the effective dielectric constant
of the gate dielectrics, it may provide a better and more stable in-
terface with Si, reducing the interface state density. It could also
i
2
lated to the leaky path via the ZrO polycrystalline grain boundaries
2
created by forming gas. In Fig. 9c, the EOT is essentially unchanged
after FGAs for 10-30 min, and the changes may be due entirely to
variations in film thickness over the wafer. However, the increase in
8
improve the carrier mobility in the channel of a MOSFET ͑metal
oxide semiconductor field effect transistor͒ by providing a smoother
interface in direct contact with the Si substrate. However, the thick-
ness of this silicate layer must be reduced to meet future CMOS
technology requirements.
The continuous-mode deposition data of Fig. 1 indicate that
deposition rates do decrease when the oxygen is turned off, and this
indicates that there may be a regime where growth saturates if the
layers are not fully oxidized. Thus, ALD may be possible with this
precursor if the pumping system were changed to allow higher pres-
sures during the gas pulses with shorter pumping cycles character-
istic of classical ALD. This might also result in thinner silicate lay-
ers since the total exposure of the silicon surface to oxygen might be
reduced.
The thin ZrO films exhibit good C-V and J-V characteristics.
2
After proper annealing treatments, an EOT of 22.7 Å was achieved
with a leakage current about two orders of magnitude lower than
that of SiO with the same EOT. The P centers in the ZrO films
2
b
2
were effectively passivated by FGA. The slight remnant frequency
dispersion below midgap might be reduced if a proper FGA were
performed after, rather than before, gate electrode definition. These
electrical properties demonstrate that thin ZrO films formed by
2
i
Zr͑O -Pr) (thd) and O might be a promising gate dielectric mate-
2
2
2
Figure 11. Current density as a function of gate voltage for the sample in
Fig. 10.
rial for deep-submicrometer CMOS devices if the roughness of the
layers and the thickness of the interfacial layer can be reduced. In
Downloaded on 2015-04-12 to IP 128.122.253.228 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, 149 ͑6͒ F49-F55 ͑2002͒
F55
5
6
7
. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 ͑2001͒.
. J. Robertson, J. Vac. Sci. Technol. B, 18, 1785 ͑2000͒.
. W.-J. Qi, R. Nieh, B. H. Lee, L. Kang, Y. Jeon, K. Onishi, T. Ngai, S. Banerjee, and
addition, it might be plausible to apply the thin ZrO films in flash
2
memory as the gate dielectric over the floating gate to increase
the coupling ratio. However, further investigations of its reliability
are indispensable before using this ZrO2 film process in CMOS
technology.
J. C. Lee, Tech. Dig. Int. Electron Devices Meet., 1999, 145.
8. W.-J. Qi, R. Nieh, B. H. Lee, K. Onishi, L. Kang, Y. Jeon, J. C. Lee, V. Kaushik, B.
Y. Nguyen, L. Prabhu, K. Eisenbeiser, and J. Finder, Symposium on VLSI Technol-
ogy, IEEE, p. 40 ͑2000͒.
9
. C.-H. Lee, H. F. Luan, W. P. Bai, S. J. Lee, T. S. Jeon, Y. Senzaki, D. Roberts, and
D. L. Kwong, Tech. Dig. Int. Electron Devices Meet., 2000, 27.
Acknowledgments
The authors are grateful to Dr. Shiow-Huey Chuang at National
Nano Device Laboratories for her fruitful discussion. We also ac-
knowledge E. Estwick and T. Quance for assistance in the prepara-
tion of the samples. In addition, we thank Dr. J. R. Hauser for use of
the NCSU C-V analysis program. This work was supported by the
National Science Council of the Republic of China under contract
no. NSC89-2215-E-009-071.
10. M. Copel, M. Gribelyuk, and E. Gusev, Appl. Phys. Lett., 76, 436 ͑2000͒.
1. Q. Lu, R. Lin, P. Ranade, Y.-C. Yeo, X. Meng, H. Takeuchi, T.-J. King, C. Hu, H.
Luan, S. Lee, W. Bai, C.-H. Lee, D.-L. Kwong, X. Guo, X. Wang, and T.-P. Ma,
Tech. Dig. - Int. Electron Devices Meet., 2000, 641.
12. A. C. Jones, Chem. Vap. Deposition, 4, 169 ͑1998͒.
1
13. S. B. Desu, T. Shi, and C. K. Kowk, Mater. Res. Soc. Symp. Proc., 168, 349 ͑1990͒.
1
4. D. G. Colombo, D. C. Glimer, V. G. Young, Jr., S. A. Campbell, and W. L. Glad-
felter, Chem. Vap. Deposition, 4, 220 ͑1998͒.
1
5. R. C. Smith, N. Hoilien, C. J. Taylor, T. Ma, S. A. Campbell, J. T. Roberts, M.
Copel, D. A. Buchanan, M. Gribelyuk, and W. L. Galdfelter, J. Electrochem. Soc.,
The National Chiao-Tung University assisted in meeting the publication
costs of this article.
1
47, 3472 ͑2000͒.
1
1
6. R. C. Smith, T. Ma, N. Hoillien, L. Y. Tsung, M. J. Bevan, L. Colombo, J. Roberts,
S. A. Campbell, and W. L. Gladfelter, Adv. Mater. Opt. Electron., 10, 105 ͑2000͒.
7. T. Ma, S. A. Campbell, R. Smith, N. Hoilien, B. He, W. L. Gladfelter, C. Hobbs, D.
Buchanan, C. Taylor, M. Gribelyuk, M. Tiner, M. Copel, and J. J. Lee, IEEE Trans.
Electron Devices, 48, 2348 ͑2001͒.
List of Symbols
Eb binding energy
J
gate current density
1
8. A. C. Jones, T. J. Leedham, P. J. Wright, M. J. Crosbie, P. A. Lane, D. J. Williams,
K. A. Fleeting, D. J. Otway, and P. O’Brien, Chem. Vap. Deposition, 4, 46 ͑1998͒.
Nf fixed charge density
VFB flatband voltage
VG gate voltage
19. J. R. Hauser and K. Ahmed, in Characterization and Metrology for ULSI Technol-
ogy, 1998 International Conference, D. G. Seiler, A. C. Diebold, W. M. Bullis, T. J.
Shaffner, R. McDonald, and E. J. Walters, Editors, pp. 235-239, The American
Institute of Physics, Woodbury, NY ͑1998͒.
Wms work function difference
2
0. J. Guittet, J. P. Crocombette, and M. Gautier-Soyer, Phys. Rev. B, 63, 125117
2001͒.
Greek
͑
2
1. M. J. F. Moulder, W. F. Stickle, P. E. Sobol, K. D. Bomben, and J. Chastain,
Handbook of X-Ray Photoelectron Spectroscopy, p. 45, Perkin-Elmer Corporation,
Eden Prairie, MN ͑1992͒.
⌬
⌬
Nt effective density of trapped charge
VFB flatband voltage shift
␬
dielectric constant of film
22. S. Iwata and I. Akitoshi, J. Appl. Phys., 79, 6653 ͑1996͒.
23. P. Lundgren and M. O. Andersson, J. Appl. Phys., 74, 4780 ͑1993͒.
24. E. H. Nicollian and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and
Technology, p. 465, Wiley, New York ͑1982͒.
References
1
. The International Technology Roadmap for Semiconductors, Semiconductor Indus-
try Association, San Jose, CA ͑2000͒.
. S.-H. Lo, D. A. Buchanan, Y. Taur, and W. Wang, IEEE Electron Device Lett., 18,
25. L.-Å Ragnarsson and P. Lundgren, J. Appl. Phys., 88, 938 ͑2000͒.
26. S. M. Sze, Physics of Semiconductor Devices, 2nd ed., p. 403, Wiley, New York
͑1981͒.
2
2
09 ͑1997͒.
3
4
. B. Brar, G. D. Wilk, and A. C. Seabaugh, Appl. Phys. Lett., 69, 2728 ͑1996͒.
. D. J. Hubbard and D. G. Schlom, J. Mater. Res., 11, 2757 ͑1996͒.
27. M. Houssa, M. Tuominen, M. Naili, V. Afanas’ev, A. Stesmans, S. Haukka, and M.
M. Heyns, J. Appl. Phys., 87, 8615 ͑2000͒.
Downloaded on 2015-04-12 to IP 128.122.253.228 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).