Gate Insulator Process Dependent NBTI in SiON p-MOSFETs
S. Mahapatra and V. D. Maheta
Department of Electrical Engineering, lIT Bombay, Mumbai 400076, India
Email: [email protected]
Abstract
2. Device details
The material dependence ofNBTI in SiON p-MOSFETs
is studied using the UF-OTF IOLIN method. It is shown
that the N density at the Si/SiON interface plays a very
crucial role in determining the magnitude as well as the
time, temperature and field dependence of NBTI. The
relative contribution of interface trap generation and hole
trapping to overall degradation is qualitatively discussed.
Table-I shows the process details of devices used in this
work. ToXB is starting base oxide thickness, atomic N%
and EOT were determined respectively by XPS and CV
measurements followed by QM corrections.
1. Introduction
Negative Bias Temperature Instability (NBTI) is a very
serious reliability concern in p-MOSFETs having Silicon
Oxynitride (SiON) gate insulators [1-4]. It is well known
that NBTI induced shift in parameters (e.g., linear drain
current, IOLIN; threshold voltage, VT; etc.) is governed by
oxide field (E ox ) and not by stress gate voltage (VG), and
proper choice of VG is crucial for reliable stress test [5].
It is also well known that NBTI parametric shift recovers
after removal of stress for conventional Measure-StressMeasure (or MSM) sequence, which results in incorrect
degradation magnitude and time dependence [2,4,6,7].
Therefore, On-The-Fly (OTF) IOLIN [6] or fast MSM [8]
methods are used for NBTI characterization. Recently,
an Ultra-Fast On-The-Fly (UF-OTF) IOLIN technique has
also been proposed [9] and employed to study the SiON
process dependence of NBTI.
It is important to understand and model NBTI physical
mechanism to extrapolate short time stress data to end of
life at operating condition for reliable determination of
device lifetime. It is generally accepted that NBTI results
in generation of interface traps (.~NIT) and hole trapping
(~Nh) in N-related bulk SiON traps, though the relative
dominance of ~NIT or ~Nh is debated [2,4,7-10]. While
~Nh is a faster process that saturates quickly in time and
shows weaker T dependence, ~NIT gradually builds up in
time, shows power law (tn) time dependence at longer
stress time and shows stronger T dependence.
In this paper, a review is done of our recently published
results on SiON process dependence of NBTI [4,9,11,
12]. By using Plasma Nitrided Oxides (PNO) with and
without Post Nitridation Anneal (PNA) and Thermal
Nitrided Oxide (INO), it is shown that N density at the
Si/SiON interface governs the relative ~NIT and ~Nh
contribution; and as a result the magnitude, as well as the
time, E ox and temperature (T) dependence of NBTI. It is
shown that NBTI physical process in PNO devices with
proper PNA is dominated by interface trap generation.
W#
Toxa
Type
NOOSE
EOT
N%
W3
15
PNO
2.8
14.0
22
W4
15
PNO
5.8
12.3
41
49
W5
15
PNO
7.8
11.9
W6
20
PNO
2.9
17.7
19
W7
20
PNO
5.3
15.6
34
W8
20
PNO
6.8
14.6
42
WI0
25
PNO
3.1
23.5
16
WII
25
PNO
5.5
21.4
29
WI2
25
PNO
6.8
19.9
35
W23
20
INO
0.8
18.5
06
W24
W25
20
20
PNO*
PNO**
2.0
2.7
22.2
20.2
12
17
Table-I. Process details of devices used in this work
(TOXB and EOT in A 0, NOOSE in cm-2). All PNO have
proper 2-step PNA, except PNO** (moderate PNA)
and PNO* (improper PNA).
3. Time and temperature dependence
Fig.l shows the measured IOLIN transients during NBTI
stress for W6 and W23. The reduction in IOLIN is due to
defect generation during stress, which is more for W23
in spite of having lower N dose when compared to W6.
Fig.2 shows the time evolution of degradation calculated
as ~ V = -~IOLIN/loLiNO * (VG-VTO), where IOLINO is the
value of IOLIN measured at time to (=IJ.ls or Ims) after the
application of stress VG, ~IOLIN refers to the degradation
in IOLIN from IOLINO for a given stress time and VTO is the
pre-stress VT. The choice of to influences the magnitude
and time evolution of ~ V, more for W23 as compared to
W6. This is due to the larger difference for W23 in IOLINO
when obtained at t o=lJ.ls and t o=lms (Fig. 1). Note that a
to of Ims is standard for conventional OTF [2,4,6,7, I 0,
12], and the necessity of UF-OTF [9] having to of IJ.ls is
obvious from Fig.2. For a given to, ~V for W23 is much
larger when compared to W6, which is consistent with
Fig.l. For t o= IJ.lS, W23 shows very high degradation at
short stress time (--30mV in Ims), which is not observed
978-1-4244-2186-2/08/$25.00 ©2008 IEEE
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 31, 2008 at 01:15 from IEEE Xplore. Restrictions apply.
for W6. Note that conventional OTF fails to capture this
large difference in the initial degradation between PNO
(W6) and TNO (W23) films.
1.00
,
o
..........
~ 0.98
.~
(ij
E 0.96
a
0.92
Eox
1
10-
-
8.5 MV I em
T (DC)
o
o
55
125
W23 "
Sz
_~0.94
[14,15]. Therefore, the difference between W6 and W23
can be attributed to the difference in Si/SiON N density
between the films, as discussed in [9,12].
T=125°C
Eox
-
,
8.5 MV I em
10-7 10-5 10-3 10-1 10 1
o
3x1 0-3L.....1..11.-..................-......-............-.._............L...U.LIl-.........--.....~
103
stress time (s)
Fig.I. Time evolution of IOLIN degradation for W6
and W23 at identical stress Eox and T. For details
of measurement setup, see [9].
10-3
10-6
100
103
stress time (s)
Fig.3. Time evolution of L\V for W6 and W23 at
identical stress Eox but different stress T.
1
10-
T (DC)
o 125
o
55
3
1010-3
10°
stress time (s)
10
3
Fig.2. Calculated time evolution of L\ V (see text)
from IOLIN transients of Fig. 1 for different to for
W6 and W23 at identical stress E ox and T.
Fig.3 shows the time evolution of L\V at different stress
T for W6 and W23. Note that clear T dependence for the
entire stress duration is observed for W6. This is not the
case for W23, which shows T independent degradation
at short stress duration and small overall T activation at
longer stress time. It is important to note that the large
overall degradation observed for W23 (compared to W6)
is almost accounted for by this T insensitive component
observed at short stress time.
Figs.I-3 show that irrespective of higher total N dose.
W6 shows much smaller NBTI compared to W23. It is
well known that the N density at the Si/SiON interface
for PNO films (having proper, 2-step PNA [13]) is much
smaller compared to TNO films, as most of the N during
PNO is incorporated close to the SiON/poly-Si interface
10 0
10 3
stress time (s)
Fig.4. Time evolution of L\V for W24 and W25
at identical stress Eox but different stress T.
Fig.4 shows L\ V time evolution at different T for W24
and W25. Unlike W6 that underwent proper 2-step PNA
[13], W24 saw incorrect PNA, while W25 was subjected
to moderately correct PNA. Though NOOSE is highest for
W6 followed by W25 and W24, NBTI is highest for
W24, followed by W25 and W26. Unlike W6, W24 and
to some extent W25 show high, T independent NBTI at
short stress time similar to W23. It is evident that PNO
films having moderate N dose show very high NBTI, as
also shown in [8], unless subjected to proper PNA [13].
4. Time exponents
Though not true at short stress time, long time stress data
(t-stress > Is) shows power-law time dependence. The
time exponent (n) is a very important metric, as it is used
to extrapolate short time stress data to failure limit to
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 31, 2008 at 01:15 from IEEE Xplore. Restrictions apply.
determine safe operating condition. Moreover, the value
of n was used to determine the underlying physical
mechanism [1-4,7-12,16,1 7]. It is therefore important to
study the SiON process dependence of n.
Fig.5 shows measured n (linear fit of 1Os-1 ODDs stress
data plotted in a log-log scale) as a function of to delay
for a variety of SiON processes. First, it is important to
highlight the strong impact of to delay on n observed for
all devices, with conventional OTF yielding somewhat
higher n than UF-OTF. It is also important to note that n
saturates (within error) for to < 10J.ls for all SiON films.
Hence, a faster OTF implementation is unlikely to show
significantly different value of n than that reported in
this work. In the remaining discussion, the value of n at
t o=lJ.ls will be referred.
en
0.20,..-------------,
0 W3
o
g
0 W6
~
~ W10
o
o
~0.15~ V W23
&
....c
.........
~
(1)
c
8. 0.1 0 ~
V
x
(1)
(1)
V
E
~0.05
V
"I
Eox- 8.5 MV/cm
" ..I
10.6 10-5 10.4 10-3 10. 2
g
~
~ 0.15~
~
~c
0
~
~
1)
&.0.10~ 0
Q)
.~
~
o
o
0.14
~ 0.10
W4
W7
o
8.
0.08
x
0
o
E
o
0
~
O
M
0.16~----------.,
,,&6
V W24
0 W25
........,
en
W3
0.20,......---------,
0 W6
0 W7
~ W8
Fig.6 shows n as a function of TOXB for PNO with proper
PNA devices for various NOOSE. Note, n is independent
of stress E ox and T [9,12], and hence the variation seen
is due to variations in TOXB and NOOSE. The N density for
PNO films peaks near SiON/poly-Si and exponentially
falls towards the Si/SiON interface [15]. Hence, increase
in N density with increase in NOOSE (fixed TOXB) is lower
at Si/SiON compared to SiON/poly-Si interface, while N
density increases only at Si/SiON as TOXB is reduced for
a given NOOSE. Si/SiON N density slightly increases but
n remains independent as TOXB is reduced for moderate
NOOSE. Large increase in Si/SiON N density and large
reduction in n is seen with TOXB scaling at higher NOOSE.
For thicker TOXB, n shows slight reduction with increase
in NOOSE due to smaller increase in Si/SiON N density,
while for thin TOXB the increase in Si/SiON N density
and reduction in n with increase in NOOSE is significant.
It is clear that n is governed by N density at Si/SiON and
not at SiON/poly-Si interface, which is also consistent
with difference seen for PNO and TNO films. It is also
clear that proper 2-step PNA reduces Si/SiON N density
and results in lower NBTI and higher n for PNO films.
V
to delay(s)
(i)
the necessity to treat PNO films with proper PNA. Fig.5
also shows the importance of using UF-OTF for reliable
determination of power law time exponent n.
~
~
.........
~
0.12
c
(1)
(1)
0.06
+:i
0.04
10
T = 125°C
Eox - 8.5 MV/cm
..... 0.05 . . . . .,... . .I-......I-&...I..LI..I.LI.I&-II...........LLI~.uL......I.IL...L..L.L.LLLI"II-....L.I..&...L..L.IJ..IIIL,,"I--......
10-6 10-5 10-4 10.3 10-2
to delay (s)
Fig.5. Time exponent (n) as a function of to delay
for different process splits.
Note, PNO (proper PNA) films having moderate NOOSE
(W3, W6, WID) show identical n irrespective of TOXB '
and is much higher as compared to TNO (W23) device.
Moreover, n reduces with increase in PNO dose (W6,
W7, W8), while PNO with inadequate PNA (W24, W25)
show very low n, similar to TNO device. Such very low
n (and large degradation) for PNO films was also shown
in [8] for moderately dosed PNO films, and highlights
0
0
0
0
W8
W5
15
W10
BW11
W12
to = 1~s
20
o
T BASE (A )
25
30
Fig.6. Time exponent (n) for PNO as a function
of base oxide thickness for various N dose.
5. Field acceleration
As mentioned before, degradation during NBTI is a field
driven process [5]. Like time exponent that is used to
extrapolate short time stress data to end of life, the E ox
dependence is also important to extrapolate from stress
(high E ox ) to operating (low Eox ) condition to determine
device lifetime. Fig.7 shows ~V (normalized to EOT) at
identical stress time and T as a function of stress E ox for
various SiON films. The E ox dependent slope (f) is also
mentioned. PNO (proper PNA) films with moderate N
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 31, 2008 at 01:15 from IEEE Xplore. Restrictions apply.
dose (W3, W6, WIO) show identical ~V that is much
lower as compared to TNO (W23) for all Eox . Unlike n
dependence (Fig. 5)~ r reduces slightly as TOXB is scaled,
although r is much higher for PNO compared to TNO
films. Moreover, ~ V increases and r reduces as N nosE is
increased for PNO with proper PNA devices (W6, W7,
W8), and improper PNA (W24, W25) devices show high
degradation and low r, same as TNO (W23).
saturates fast which causes low overall T activation
(when added to ~NIT component that gradually builds up
in time) seen in these films at longer stress time [4].
Optimized PNO films with proper PNA show lower N
density at Si/SiON interface, negligible ~Nh and ~NIT
dominated NBTI physical process.
0.7
T=12SoC
0.6~
-
v
v
V
:::J
010.3 W
IS;
~
[]~O
0
0
~
V
10-4
6
::j
~
W
~
E 0.4~
I
7
0
1:: 0.3f0.2~
W3 (r=0.60)
W6 (0.52)
W10 (0.47)
W23 (0.32)
I
I
8
9
Eox (MV/cm)
W3
0
W4
0
10
W7
OW11
0
W12
0
W8
0
W5
O. 1 ~
1
~6
to
20
25
T BASE(Ao)
I
I
=1J..1S
70--L.-----:-1~....L..----1----L.-L....:..-....L..--1
5
30
Fig.8. Field acceleration (f) as a function of base
oxide thickness for various N dose.
6. Conclusion
010.
>
~
~DO
~o
~
~
~0.5~
0
~
~
v
v
"610
3
....
~
Aq
<>
<>
0
0
0
o
o
o
W6 (0.52)
W7 (0.44)
~ W8 (0.42)
o
0
V
10-4""'"--__............
6
7
<>
W24 (0.33)
W'45 (0.33)
I
I_-a..-.~-..L-_.L..-.1~---J
8
Eox (MV/cm)
9
10
Fig.5. Field acceleration of normalized degradation
for various process splits.
Fig.8 shows r as a function ofToXB for PNO with proper
PNA devices for various NnosE . r is more sensitive than
n to changes in N density at Si/SiON interface. Note that
for moderate N nosE, r reduces (unlike n) with reduction
in TOXB, though the reduction in r is more drastic for
higher N DOSE ' The reduction in r with increase in N nosE
is much larger for thinner TOXB, once again suggesting N
density at SiiSiON interface to playa crucial role.
It is therefore apparent that SiON process conditions that
result in higher N density at the Si/SiON interface also
results in higher degradation as well as lower nand f.
Additional NBTI in these films can be mostly attributed
to the T insensitive, high degradation observed at shorter
stress time, which is likely due to large ~Nh in these
films [4,9]. Note~ ~Nh has negligible T dependence and
To summarize, process impact of SiON films on NBTI is
studied. Processes that result in higher N density at the
Si/SiON interface show high NBTI, low time exponent,
T activation and field acceleration. This can be attributed
to large relative increase in hole trapping in such films.
PNO films having moderate N dose and proper 2 step
PNA shows interface trap dominated NBTI.
Acknowledgement
E. N. Kumar and S. Purawat (lIT Bombay) for help in
UF-OTF setup development; K. Ahmed and C. Olsen
(Applied Materials) for samples; M. Alam (Purdue
University) for useful discussions; Applied Materials,
SRC and Renesas Technologies for financial support.
References
[1] Y. Mitani, p.II7, IEDM 2004; [2] A. T. Krishnan,
p.688, IEDM 2005; [3] K. Sakuma, p.454, IRPS 2006;
[4] S. Mahapatra, p.I, IRPS 2007; [5] S. Mahapatra,
p.I37I, TED 2004; [6] S. Rangan, p.34I, IEDM 2003;
[7] D. Varghese, p.684, IEDM 2005; [8] C. Shen, p.333,
IEDM 2006, [9] E. N. Kumar, p.809, IEDM 2007: [10]
V. Huard, p.797, IEDM 2007; [11] S. Mahapatra, p.35,
TDMR 2008; [12] V. Maheta, to-appear, TED 2008;
[13] C. Olsen, US Patent 017596IAl, 2004; [14] 1. R.
Shallenberger, JVST-A~ p.1086, 1999; [15] S. Rauf,
p.024305, JAP 2005; [16] M. Alam~ NBTI tutorial, IRPS
2006; [17] S. Chakravarthi, p.273, IRPS 2004.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 31, 2008 at 01:15 from IEEE Xplore. Restrictions apply.