This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 140.113.38.11
This content was downloaded on 25/04/2014 at 17:45
Please note that terms and conditions apply.
Thermal Stability and Electrical Characteristics of Tungsten Nitride Gates in
Metal–Oxide–Semiconductor Devices
View the table of contents for this issue, or go to the journal homepage for more 2008 Jpn. J. Appl. Phys. 47 872
(http://iopscience.iop.org/1347-4065/47/2R/872)
Thermal Stability and Electrical Characteristics of Tungsten Nitride Gates
in Metal–Oxide–Semiconductor Devices
Chih-Feng HUANG, Bing-Yue TSUI, and Chih-Hsun LU
Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, ED630, No. 1001, Ta-Hsueh Road, Hsinchu 300, Taiwan, R.O.C.
(Received June 17, 2007; accepted October 29, 2007; published online February 15, 2008)
Tungsten nitride (WNx) was investigated to be used as a metal gate of metal–oxide–semiconductor field effect transistors
(MOSFETs). WNxfilms with various N/W atomic ratios were deposited on SiO2and HfO2by reactive sputter deposition at
different N2/Ar ratio flows. Nitrogen concentration in WNx films increases rapidly with the N2/Ar gas ratio and tends to
saturate. WNxfilms with nitrogen atomic ratio higher than 44% have a main phase of WN, and the WN phase is stable up to
800C. The higher-order WN
xphase does not form even if the nitrogen concentration is as high as 61%. Many of the excess
nitrogen atoms in WNx films are desorbed at temperatures below 766C. The excess nitrogen in WNx films can cause
effective work function lowering. A weak Fermi-level pinning effect is observed on the HfO2film. In this case, WNxis not
suitable to be metal gate of bulk p-channel MOSFETs. Fully depleted silicon-on-insulator (FD SOI) devices require a work function of 0.2 eV from the midgap of the Si energy band. Therefore, a WNx/HfO2gate stack can be applied to p-channel FD
SOI devices. The good integrity of the WNx/HfO2gate stack also suggests that WNxis a promising gate material.
[DOI:10.1143/JJAP.47.872]
KEYWORDS: metal gate, WN, work function, MOSFET
1. Introduction
Refractory metals and their nitrides have wide spread applications in various fields including microelectronics due to their excellent hardness, high melting point, and good chemical stability. Among refractory metals, tungsten has attracted considerable attention in modern very large scale integrated circuits (VLSI) because of its low resistivity, easy patterning, and compatibility to conventional complemen-tary metal–oxide–semiconductor (CMOS) technology.1,2)
Tungsten plugs have been used to electrically connect the multilevel interconnects in mass production. Tungsten nitride has also emerged as a promising candidate for the Cu diffusion barrier.3)Metal gates are expected to replace polycrystalline silicon (poly-Si) gates beyond the 45-nm technology node to continue device scaling. Tungsten and its nitride (WNx) were again considered as possible materials of
metal gates since metal gates have been studied to replace poly-Si gates in the late 1970s because of their low resistivity.1,2) Tungsten with a midgap work function
was considered as a single metal gate of fully depleted silicon-on-insulator (FD SOI) CMOS devices to provide a symmetric threshold voltage for both n- and p-channel devices.1,4) WN
x has a higher work function and was
suggested as the metal gate of p-channel bulk MOS field effect transistors (MOSFETs).5,6)There have been numerous
studies on the basic material properties of WNx films,
including the crystalline structure, effects of nitrogen composition, resistivity, patterning, residual stress, and thermal stability.7–10)Moreover, there are also some reports on the electrical characteristics and reliability of MOS devices with tungsten nitride gates.11–14) An obvious
phe-nomenon of nitrogen desorption from tungsten nitride films was noticed. However, electrical effects of nitrogen desorp-tion in previous studies were not thoroughly explored.
In this work, nitrogen-rich tungsten nitride films deposited by reactive sputtering were used as electrodes of MOS
capacitors with SiO2 and HfO2/SiO2 stack gate dielectrics.
The thermal stability and electrical stability were inves-tigated after prolonged (30 min) post metal annealing at 400 – 800C in N
2 ambient. The effective work function of
WNxgates on SiO2and HfO2was first studied in the context
of incorporated nitrogen concentration and interface bonds between WNxelectrodes and gate dielectrics. The electrical
characteristics of MOS capacitors with a pure tungsten gate were also presented as a reference.
2. Experiments
Simple MOS capacitors were used as the test devices. The starting material were 6-in., (100)-oriented, and boron-doped p-type silicon wafers with a resistivity of 15 – 25 cm. After wafer cleaning, oxide layers with various thicknesses of 40, 70, and 100 nm were thermally grown in dry O2 ambient.
Some wafers were then deposited with a 5-nm-thick HfO2
film in a reactively sputtered deposition system with an Ar/O2 gas ratio of 24 : 3. During the deposition, the
chamber pressure was 7.6 mTorr and the DC power was 100 W. Neither substrate bias nor substrate heating was intentionally applied. Because the bottom oxide was thick enough, further oxidation during HfO2 deposition did not
occur.15)After the preparation of gate dielectrics with a SiO 2
single layer and a HfO2/SiO2 stack, all of the wafers were
annealed in N2ambient at 900C for 30 s by a rapid thermal
annealing (RTA) to reduce the density of oxide charges and to form a stable microstructure of HfO2, that will not
changed again during the post metal annealing. A lift-off process was used to define the metal gate electrodes. The WNx films with various N/W atomic ratios were reactively
sputtered at 4.5 mTorr in a DC sputtering system to a thickness of 60 nm. The N2/Ar gas flow ratios used in this
work were 0 : 20, 5 : 20, 10 : 20, and 20 : 20. During deposition, the DC bias was set to 25 W and neither substrate bias nor substrate heating was applied. After the lift-off process, wafers were diced and annealed at 400, 500, 600, 700, and 800C in a N
2 ambient furnace for 30 min.
Finally, the aluminum film was deposited on the wafer back E-mail address: jeff[email protected]
Vol. 47, No. 2, 2008, pp. 872–878
#2008 The Japan Society of Applied Physics
surface by thermal evaporation to establish good ohmic contact. Table I lists the sample conditions.
The atomic compositions of WNxsamples were identified
by Rutherford backscattering spectroscopy (RBS). The RBS spectra were taken using a van de Graaff accelerator with 2 MeV particles and calibrated with bulk samples of gold and silicon. A graphite substrate was used instead of a silicon substrate to avoid the weak nitrogen signal of the WNx layer overlapped with the strong Si signal of the
substrate. A 30-nm-thick SiO2 layer was deposited in a
plasma-enhanced chemical vapor deposition (PECVD) sys-tem on the graphite substrate before WNxfilm deposition to
provide surface conditions similar to those of real SiO2/Si
samples. The crystalline phase of WNx films was examined
by the X-ray diffraction (XRD) technique using a MAC Science MXP18 XRD system with Cu K radiation. Thermal desorption spectroscopy (TDS) was used to detect the nitrogen desorption, and the nitrogen depth profile in the WNxfilms was examined by the Auger electron
spectrosco-py (AES) using a Physical Electronics Auger 670 PHI Xi system. After removing the WNx films, chemical bonds at
the surface of HfO2 were analyzed by X-ray photoelectron
spectroscopy (XPS). The XPS spectra with 0.8 eV exper-imental resolution were recorded in a Physical Electronics PHI 1600 system, using an Mg K source operating at 250 W and 15 KeV.
The capacitance–voltage (C–V) curves of capacitors were measured at 100 kHz using an Agilent 4284A precision impedance meter, and the flatband voltage (Vfb) was
extracted from the C–V curves. The effective work function was extracted by the typical method of extrapolating Vfb at
zero equivalent-oxide-thickness (EOT) from the Vfb versus
EOT plot.15) Constant-voltage stress application was
per-formed using an Agilent 4156C semiconductor parameter analyzer.
3. Results and Discussion
3.1 Physical properties of N-rich WNxfilms
The WNxfilms were deposited by reactive sputtering with
a mixture of Ar and N2 gases, and the increase in N2 gas
flow rate formed high-N-ratio WNx films. Figure 1 shows
the N concentration and corresponding resistivity of the as-deposited WNx films. The N concentrations analyzed by
RBS are 0, 44, 56, and 61 at. % for WN-0, WN-1, WN-2, and WN-3, respectively. The N concentration increases rapidly and then tends to saturate. The corresponding resistivities are 94.09, 1322, 1735, and 2438 m cm, respectively. The resisitivities of the WNxfilms are much higher than that of
W2N (220 m cm) and are considered to originate from a
high-order WNx phase.7,10) The typical tungsten nitride
phases exist as W2N or WN with 33.3 and 50% nitrogen
stoichiometry, respectively. Because high-order WNxphases
are unstable, any further incorporation of N in WNxfilms is
difficult.7,10)It is believed that the stability of the WNxphase
is dependent on its heat of formation. The heat of formation for W2N is 5:3 kcal/mol and is slightly more exothermic
than that for WN, 3:6 kcal/mol.16)W2N is the most stable
tungsten nitride phase with a cubic crystal lattice.
To investigate the thermal properties of the N-rich WNx
films, high-temperature annealing was performed. After RTA at a temperature increase rate of 90C/s, the WN-2
and WN-3 films had bubbles or cracks. As shown in Fig. 2, the WN-2 film had bubbles at 500C and cracks at 800C.
The phenomenon seems to involve gas expansion to create bubbles and bubble blowout to generate cracks in WNx
Table I. WNxsample conditions.
WN-0 WN-1 WN-2 WN-3
N2/Ar flow ratio 0/20 5/20 10/20 20/20
N (at. %) (as-deposited) 0 44 56 61 Resistivity ( cm) (as-deposited) 94.09 1322 1735 2438 Main phase W WN WN WN 0.0 0.2 0.4 0.6 0.8 1.0 0 500 1000 1500 2000 2500
N2/Ar gas flow ratio
Resistivity ( µΩµΩ cm) 0 20 40 60 80 100 N concentration in WNx films (%)
Fig. 1. Nitrogen concentration and relative resistivity of the WNxfilms as
a function of the N2/Ar gas flow ratio during sputtering deposition.
(a)
(b)
Fig. 2. Plane-view SEM micrographs of the WN-2 films after annealing at (a) 500C and (b) 800C in a rapid thermal annealing system. The
temperature increase rate is 90C/s.
Jpn. J. Appl. Phys., Vol. 47, No. 2 (2008) C.-F. HUANGet al.
films. TDS analysis was used to investigate the temperature-dependent thermal desorption of nitrogen. Figure 3 shows the background signal and sample signal of the WN-2/SiO2
sample. The temperature increase rate was set at 20C/min. The intensity of the background signal is at least one order of magnitude lower than that of the sample signal and can be neglected. Nitrogen gas concentration increases rapidly with the temperature up to T ¼ 600C and slowly reaches a
maximum at T ¼ 766C. After annealing at 800C for
10 min, the signal becomes weak. Most of the nitrogen not strongly bound with tungsten is desorbed at temperatures below 766C and the WN phase of WN-2 was identified by
XRD. The sample after TDS analysis did not have bubbles and cracks. Therefore, the temperature increase rate of 20C/min during the TDS analysis was hence used for the
furnace annealing to prevent film cracking.
The thermal desorption of nitrogen was detected by TDS analysis and then XPS analysis was used to detect if nitrogen diffuses into the under layer dielectric. After the tungsten nitride film was removed using H2O2 solution, the exposed
surface of the dielectric was analyzed. The nitrogen distribution in the WNx/SiO2 structure has been analyzed
by Yamada et al.14) The nitrogen was incorporated at the
interface between WNx and SiO2 during the reactive
sputtering deposition of WNx films and its concentration
was the same after annealing for 30 min at 750C in N 2
ambient. The binding energy of N core level reflected major NSi3 bonds and minor Si–N=O2 bonds.14) It was thus
concluded that nitrogen does not diffuse into SiO2 during
annealing. In this work, we focus on the WNx/HfO2
structure. Figure 4 shows the core-level binding energies of O, Hf, and N for the 500C-annealed WN-1 sample. The
binding energy of Hf core level only reflects Hf–O bonds. The signal of O binding energy is composed of two signals, one for N–O bonds and one for Hf–O bonds. The signal for the binding energy of N core level is very weak and represents N–O bonds. After 800C annealing, the binding
energies of O and Hf are all the same as those of the 500
C-annealed sample, and the area ratio of N–O and Hf–O is 1/4, which indicates that high-temperature annealing does not enhance the incorporation of N into the HfO2 film.
The phases of the WNx films were analyzed on the basis
of the XRD patterns. Figure 5 shows the diffraction patterns of WNx films annealed at different temperatures for 30
min in N2 ambient. The as-deposited WNx films are all
amorphous. After 600C annealing, the WN-1 film
obvi-ously reveals a WN phase. The WN-2 film is crystallized at 400C, and the main phase is still WN. The WN-3 sample is
also crystallized at 400C and peak intensity increases with
annealing temperature. The increase in nitrogen content to above 50% does not produce the higher-order WNx phase,
which confirms that the stable highest-order WNx phase is
WN after 800C annealing.7,10)The broaden peak in WN-2
and WN-3 XRD spectra implies that the excess nitrogen can suppress grain growth. The relatively weak WO3 phase
is presumably due to the slight surface oxidation during annealing. The WN phase is stable up to 800C while it was
reported that the WN and W2N phases can change into the
W phase at 900C in N
2 ambient, and the temperature for 00:00 00:14 00:28 00:43 00:57
1x10-10 1x10-9 1x10-8
Time
Signal Intensity (arb.
unit) 0 200 400 600 800 m/e=28 (N2) Background T emperature ( °C)
Fig. 3. Thermal desorption spectrum of the as-deposited WN-2 film.
542 540 538 536 534 532 530 528 526 524
(a) Hf-O
N-O O 1s
Binding Energy (eV)
30 28 26 24 22 20 18 16 14 12 10
(b)
Hf-O Hf-O Hf 4f
Binding energy (eV)
390 395 400 405 410
(c) N-O
N 1s
Binding Energy (eV)
Fig. 4. XPS spectra of the (a) O 1s, (b) Hf 4f, and (c) N 1s of the WN-1 film after annealing at 500C for 30 min in N
2 ambient. The
de-convolution of O 1s is also given.
phase transition is lower in H2 and N2 gas mixture
ambient.17) Since the phase of the annealed WN
x film is
WN, unmerous N atoms in the as-deposited WNxfilms are
not strongly bounded to W. Figure 6 shows the nitrogen depth profile of WN-2 detected by AES. The intensity of the nitrogen signal for the 800C annealed WN-2/SiO2sample
is much lower than that for the as-deposited sample, and the intensities of the W signal for both samples are the same. These physical analyses clearly confirm that nitrogen desorption occurred in the whole film after annealing. 3.2 Electrical characteristic of MOS capacitor
with WNx gate
The effective work function was extrapolated from the Vfb–EOT plot at zero EOT. For example, the Vfb–EOT plots
of 600C-annealed capacitors with SiO
2 and a HfO2/SiO2
stack are shown in Fig. 7. The good linearity reflects the validity of the effective work function extraction. For the single-layer SiO2 samples, the y-axis intersection in the Vfb–
EOT plot represents the work function difference between the metal gate and the Si substrate (ms¼ms). It
should be noted that the HfO2sample has a HfO2/SiO2stack
that affect the y-axis intersection point including the work function difference (ms) and the HfO2/SiO2 interface
charges (Qhigh-k/SiO2). With the Si substrate concentration known, the effective work function of metal gate on SiO2
can be extracted. However, on the HfO2, an offset due to the
Qhigh-k/SiO2 is inevitable. The evaluated work function of the HfO2 sample will be denoted as 0m,eff accounting for the
offset. The detailed extraction procedure for effective work function has been discussed in literature.15)The slopes of the SiO2 curves are almost the same and reflect the positive
oxide charges of about 1:8 1011cm2 density. The slopes of HfO2 curves indicate that the positive oxide charges
decrease from 3:6 1011 to 1:5 1011cm2. Compared
with the pure W gate, all of the WNx gates have few
effective oxide charges. The reduction of the density of oxide charges is suspected to be due to the nitrogen plasma during reactive sputtering.
The effective work functions (m,eff and 0m,eff) of WNx
films on SiO2 and HfO2 after annealing are extracted and
30 40 50 60 WN-3 2Θ (deg) WN-2 WO3 (200) 800°C 800°C 600°C 600°C 400°C As-deposited As-deposited 400°C 800°C 600°C 400°C As-deposited WN (100) WN-1
Fig. 5. XRD patterns of the WNx films after annealing at different
temperatures. All of the as-deposited films are amorphous.
0 200 400 600 800 1000 1200 1400 0.0 2.0x105 4.0x105 6.0x105 8.0x105 1.0x106 W_as-deposited N_as-deposited W_800°C N_800°C WN-2 Intensity (arb. unit) Sputter time (s)
Fig. 6. AES depth profiles of the as-deposited and 800C-annealed WN-2
films. 0 20 40 60 80 100 120 -2.0 -1.6 -1.2 -0.8 -0.4 0.0 0.4 Slop=-0.0086 Slop=-0.0084 SiO2 600°C (a) WN-0 WN-1 WN-2 WN-3
Flatband V
olta
g
e
(V)
EOT (nm)
0 20 40 60 80 100 120 -2.0 -1.6 -1.2 -0.8 -0.4 0.0 0.4 Slop=-0.0064 Slop=-0.0166 HfO2 600°C (b) WN-0 WN-1 WN-2 WN-3Flatband V
olta
g
e
(V)
EOT (nm)
Fig. 7. Vfb vs EOT plots of (a) WNx/SiO2/Si structure and (b) WNx/
HfO2/SiO2/Si structure after annealing at 600C.
Jpn. J. Appl. Phys., Vol. 47, No. 2 (2008) C.-F. HUANGet al.
shown in Fig. 8. Regarding the WNx on SiO2, the m,eff of
pure W (WN-0) is stable at 4:60 eV below 800C, and the m,eff of tungsten nitrides seems to increase with annealing
temperature and tends to approach a stable value at 700C.
The WN-1, WN-2, and WN-3 samples annealed above 700C shows
m,effvalues of 5:12, 5:05, and 5:0 eV,
respectively. The effective work function of the metal gate is affected by its composition, orientation, phase, and interface interaction.18,19)To explain the extracted work functions of
WNxfilms, these factors are considered. First, WNxfilms do
not chemically react with SiO2 and HfO2. Although the
NSi3bonds and N–O bonds were observed at the interface
between WNxand the underlayer dielectric, these bonds are
stable without further interaction below 800C.14)The phase
of the W film is W(110) below 900C.10)The main phase of
N-rich WNxis WN(100) and stable up to 800C. Therefore,
nitrogen incorporation in tungsten films forms the tungsten nitride compound (WN), which is the main reason why the effective work function changes from 4:60 eV for the W phase to 5:12 eV for the WN phase. The only parameter that changes during high-temperature annealing is the N content in WNx. This suggests that the thermal instability of
effective work function for WNx comes from the excess
nitrogen contents. This result is similar to that for MoN films.20,21) The excess nitrogen atoms (piled-up nitrogen
atoms) at the interface between MoN and SiO2 reduce the
m,eff value of MoN.20,21)The nitrogen concentration at the
interface determined the reduction of m,eff.20,21)Therefore,
it is reasonable to conclude that WN-2 and WN-3 with more excess nitrogen have a relatively lower m,eff than WN-1.
The WNx films annealed at low temperatures have high
contents of excess nitrogen to lower m,eff. After
high-temperature annealing (T > 600C), most of the excess nitrogen is desorbed so that m,eff reaches a high and stable
value. Similarly, the results for WNx/HfO2 samples are
determined by the nitrogen contents. The 0
m,eff values of
pure W and WN-1 on HfO2 are 4:75 and 5:18 eV after
annealing at 800C, respectively.
The magnitude of work function adjustment (0 m,eff)
on HfO2 film is smaller than that on SiO2 samples. This
means that the WNx deposited by reactive sputtering will
cause the effective work function to be different if the dielectric is different. The difference may be affected by the work function offset due to interface charge (Qhigh-k/SiO2) between HfO2 and SiO2 and the Fermi-level pinning (FLP)
effect.22) Comparing pure W with WN-1 samples, the 0m,eff values on HfO2 (m,eff on SiO2) films after
annealing at 400 and 800C are 0.31 eV (0.50 eV) and 0.37 eV (0.52 eV), respectively. Moreover, comparing pure W with WN-3 samples, the 0
m,eff values on HfO2
(m,eff on SiO2) films after annealing at 400 and 800C
are 0.05 eV (0.3 eV) and 0.21 eV (0.4 eV), respectively. The difference between 0
m,eff and m,eff increases slightly
with nitrogen concentration. It is indeed necessary to consider if the interface dipole layer at the WNx/HfO2
interface induces the FLP effect. The reported FLP effects always force the work function of the metal gate to be pinned near the midgap of silicon.22)Therefore, it is believed
that FLP possibly occurs in the WNx/HfO2 stack. Because
of the FLP effect, WNx films are not suitable for bulk
p-channel MOSFETs with a HfO2 dielectric. The FLP effect
shifts the work function to less than 0.15 – 0.2 eV, and WNx
can be sued in p-channel FD SOI devices.
Figures 9(a) and 9(b) show the statistic time-to-break-down (Tbd) distributions of the WNx/SiO2 (5 nm)/Si and
WNx/HfO2 (2 nm)/SiO2 (2 nm)/Si MOS capacitors with
different post metal annealing temperatures. The capaci-tance equivalent thickness (CET) of the HfO2/SiO2 stack
is 3.5 nm. Samples were stressed at a constant voltage of Vg¼ 6 V for WNx/SiO2 capacitors and at Vg¼ 4:5 V
for WNx/HfO2/SiO2capacitors because these samples have
different gate dielectric thickness. Both are in the gate injection mode, and the electric field within the SiO2layer of
HfO2and SiO2 samples is 12 MV/cm. It is found that the
integrity of both SiO2and HfO2/SiO2dielectrics with a pure
W gate is not degraded during post metal annealing. On the other hand, the SiO2dielectric with a WN-1 gate is seriously
degraded during the 700C post metal annealing while the HfO2/SiO2 dielectric with WNx gates is not degraded.
Compared with pure-W SiO2 samples, the 500C and
600C WN-1 SiO
2 samples showed better time-dependent
dielectric breakdown (TDDB) characteristics.
To better understand the TDDB characteristics, the current–voltage (I–V) curves shown in Fig. 10 were used to evaluate the effect of work function. The work functions of pure W is lower than that of WN-1, and the same bias
400 500 600 700 800 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 (a) WN-0 WN-1 WN-2 WN-3
φ
m,eff(eV)
Temperature (
°C)
400 500 600 700 800 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 (b) WN-0 WN-1 WN-2 WN-3φ
'
m,eff(eV)
Temperature(
°C)
Fig. 8. Effective work functions (m,effand 0m,eff) of WNxon (a) SiO2
and (b) HfO2 as a function of annealing temperature. 0m,eff takes into
account the effect of interface charges between HfO2and SiO2.
stress will cause a relatively low leakage current for the WN-1 sample due to a high potential barrier for electrons on both SiO2and HfO2samples. This means that the stress condition
at the same bias for the pure-W sample is stronger than that for the WN-1 sample, which is the reason why the SiO2
samples with higher work functions have better TDDB characteristics.23) The degradation of the 700C SiO
2
sample should be due to the creation of oxide defects, which is confirmed by the increased leakage current at low bias as shown in the I–V curves. Although the stress condition is stronger for the pure-W HfO2 samples, the
TDDB characteristics of pure-W samples are almost the same as those of WN-1 HfO2samples. It is reported that the
time-to-breakdown (or charge-to-breakdown) for thin oxide (thickness less than 5 nm) is difficult to distinguish at high levels of electrical stress.24)This is a possible reason why the TDDB characteristics are undistinguishable when comparing these pure-W samples with WN-1 HfO2samples. Moreover,
the leakage current of the 700C annealed HfO2 samples is
not degraded. The degradation of the 700C-annealed WN-1/ SiO2 sample is suspected to be due to the stress exerted by
the nitrogen desorption. Since the single SiO2 layer cannot
withstand post metal annealing, the HfO2layer of the HfO2/
SiO2 stack acts as an effective buffer layer to prevent SiO2
degradation from the WNxgate during post metal annealing.
4. Conclusions
Tungsten nitride has an effective work function near the
valance band of silicon so that it satisfies the metal-gate requirement of p-channel MOSFETs to provide a suitable threshold voltage. However, the excess nitrogen will make the work function unstable during high-temperature anneal-ing, and may cause the WNxfilm to crack during the
high-temperature annealing with a high high-temperature increase rate. The post metal thermal annealing must be carefully controlled to avoid bubbling and cracking. The excess nitrogen in WNx films can cause effective-work-function
lowering. A weak Fermi-level pinning effect is observed on the HfO2 film. In this case, WNx is not suitable to be the
metal gate of bulk p-channel MOSFETs. Fully depleted SOI devices require a work function lower than the Si valence band for p-channel MOSFETs and higher than the Si conduction band for n-channel MOSFETs. Therefore, the WNx/HfO2 gate stack can be applied to p-channel fully
depleted SOI devices. The good integrity of the WNx/HfO2
gate stack also suggests that WNx is as a promising gate
material.
1) P. L. Shah: IEEE Trans. Electron Devices 26 (1979) 631.
2) A. K. Sinha, T. E. Smith, T. T. Sheng, and N. N. Axelrod:J. Vac. Sci. Technol. 10 (1973) 436.
3) B. H. Lee and K. Yong:J. Vac. Sci. Technol. B 22 (2004) 2375. 4) H. Noda, H. Sakiyama, Y. Goto, T. Kure, and S. Kimura:Jpn. J. Appl.
Phys. 35 (1996) 807.
5) D.-G. Park, Z. J. Luo, N. Edleman, W. Zhu, P. Nguyen, K. Wong, C. Cabral, P. Jamison, B. H. Lee, A. Chou, M. Chudzik, J. Bruley, O. Gluschenkov, P. Ronsheim, A. Chakravarti, R. Mitchell, V. Ku,
-15 -10 -5 0 5 10 -4 -3 -2 -1 0 1 2
(a) WNx/SiO2/Si WN-1 WN-0 500°C 600°C 700°C Ln(-Ln(1-F)) Ln(stress time) (s) 500°C 600°C 700°C -10 -5 0 5 10 -4 -3 -2 -1 0 1 2 (b) WN-1 WN-0
WNx/HfO2/SiO2/Si 500°C 600°C 700°C 500°C 600°C 700°C Ln(-Ln(1-F)) Ln(stress time) (s) -15
Fig. 9. Cumulative failure time of the (a) WNx/SiO2(5 nm)/Si capacitors
under constant-voltage stress at Vg¼ 6 V and (b) WNx/HfO2(2 nm)/
SiO2(2 nm)/Si capacitors under constant-voltage stress at Vg¼ 4:5 V
as a function of annealing temperature.
0.0 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0 -3.5 -4.0 10-14 10-13 10-12 10-11 1x10-10 1x10-9 1x10-8 (a) SiO2 sampels 700°C600°C 500°C 700°C 600°C 500°C WN-0 WN-1 I g (-A) V g(V) 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 10-14 10-13 10-12 10-11 1x10-10 1x10-9 1x10-8 (b) HfO2 samples 700°C 600°C 500°C 700°C 600°C 500°C WN-1 WN-0 I g (-A) Vg(V)
Fig. 10. I–V characteristics of (a) WNx/SiO2(5 nm)/Si capacitor and
(b) WNx/HfO2(2 nm)/SiO2(2 nm)/Si capacitors.
Jpn. J. Appl. Phys., Vol. 47, No. 2 (2008) C.-F. HUANGet al.
H. Kim, E. Duch, P. Kozlowski, C. D. Emic, V. Narayanan, A. Steegen, R. Wise, R. Rengarajan, H. Ng, A. Sekiguchi, and C. H. Wann: Symp. VLSI Techology Tech. Dig., 2004, p. 186.
6) B. Claflin, M. Binger, and G. Lucovsky:J. Vac. Sci. Technol. A 16 (1998) 1757.
7) J. Lin, A. Tsukune, T. Suzuki, and M. Yamada:J. Vac. Sci. Technol. A 17(1999) 936.
8) H.-T. Chiu and S.-H. Chuang:J. Mater. Res. 8 (1993) 1353. 9) Y. G. Shen and Y. W. Mai:J. Appl. Phys. 88 (2000) 1380. 10) C. C. Baker and S. Ismat Shah:J. Vac. Sci. Technol. A 20 (2002)
1699.
11) P.-C. Jiang and J. S. Chen:J. Electrochem. Soc. 151 (2004) G751. 12) H.-J. Cho, T.-H. Cha, K.-Y. Lim, D.-G. Park, J.-Y. Kim, J.-J. Kim,
S. Heo, I.-S. Yeo, and J. W. Park:J. Electrochem. Soc. 149 (2002) G403.
13) J. W. Lee, C. H. Han, J.-S. Park, and J. W. Park:J. Electrochem. Soc. 148(2001) G95.
14) T. Yamada, M. Moriwaki, Y. Harada, S. Fujii, and K. Eriguchi:
Microelectron. Reliab. 41 (2001) 697.
15) B. Y. Tsui, C. F. Huang, and C. H. Lu: J. Electrochem. Soc. 153 (2006) G197.
16) J. W. Klaus, S. J. Ferro, and S. M. George:Appl. Surf. Sci. 162 –163 (2000) 479.
17) Y. Z. Hu and S.-P. Tay: 10th IEEE Int. Conf. Advantage Thermal Processing of Semiconductors, 2002, p. 125.
18) K. Nakajima, Y. Akasaka, M. Kaneko, M. Tamaoki, Y. Yamada, T. Shimizu, Y. Ozawa, and K. Suguro: Symp. VLSI Technology Tech. Dig., 1999, p. 95
19) H. Wakabayashi, Y. Saito, K. Takeuchi, T. Mogami, and T. Kunio: IEDM Tech. Dig., 1999, p. 253.
20) P. Ranade, Y.-K. Choi, D. Ha, A. Agarwal, M. Ameen, and T.-J. King: IEDM Tech. Dig., 2002, p. 363.
21) P. Ranade, H. Takeuchi, T.-J. King, and C. Hu:Electrochem. Solid-State Lett. 4 (2001) G85.
22) Y.-C. Yeo, P. Ranade, T.-J. King, and C. Hu:IEEE Electron Device Lett. 23 (2002) 342.
23) D. J. DiMaria:J. Appl. Phys. 81 (1997) 3220.
24) I. C. Chen, S. Holland, and C. Hu: IEDM Tech. Dig., 1986, p. 660.
