Content uploaded by V. Axelrad
Author content
All content in this area was uploaded by V. Axelrad
Content may be subject to copyright.
The Effects of Substrate Coupling on Triggering Uniformity and ESD
Failure Threshold of Fully Silicided NMOS Transistors
Yoon J. Huh, Valery Axerad*, Jau-Wen Chen and Peter Bendix
Device Technology Gr., LSI Logic Corp., 1551 McCarthy Blvd. J-100 Milpitas CA 95035, (408) 954-3244
*Sequoia Design Systems, 137 Chapman Rd. Woodside, CA 94062, (650) 529 1704
Abstract
We present a multi-finger turn-on model incorporating
substrate coupling effects in multi-finger NMOS transistors
during ESD events. It is demonstrated that the substrate
coupling enables uniform triggering in a multi-finger
structure. In addition, we show that fully silicided transistors
can be used successfully as an ESD protection device without
any design/process options if the effective epi thickness is
larger than 1.5µm or bulk wafer is used.
Introduction
To ensure electrostatic discharge (ESD) robustness of
silicided transistors, numerous papers have analyzed the cause
of low voltage ESD failure of silicided transistors and
suggested solutions to overcome the silicidation effects [1]-
[3]. All efforts to improve the ESD performance of silicided
transistors are based on the standard thinking that V
t2
(the
voltage for triggering into second breakdown) should be
greater than V
t1
(the voltage at which snapback occurs) in
order for all fingers in a multi-finger structure to turn on prior
to failure. Our investigations show that the above standard
thinking is misleading and does not represent the major effect
due to ignoring substrate coupling effects in previous works.
Based on our multi-finger turn-on model, we demonstrate that
substrate coupling enables uniform triggering in multi-finger
devices even with V
t2
< V
t1
. The impact of epi thickness (t
epi
)
on uniform triggering in multi-finger devices is also studied.
ESD performance was evaluated from test structures and real
products using 0.13µm and 0.18µm technologies. The
interpretation of these results is owing largely to novel
simulation software [4] which we use for characterizing ESD
events.
Multi-Finger Turn-on Model for NMOS Transistors
Fig. 1 shows the cross section of a multi-finger structure for
both bulk and epi wafers. Note that only two fingers are
depicted in Fig. 1. To understand the triggering mechanism in
this multi-finger structure, a novel mixed-mode ESD
simulation was performed. The simulation structure for bulk
wafers is shown in Fig. 1 (a). It is assumed that the left
transistor has a 10% lower channel doping to consider
process-induced non-uniformity. The source current is
measured separately to monitor when each transistor triggers.
Fig. 2 shows the current waveforms along with the drain
voltage. The drain voltage drops by a volt from its peak right
after the left finger triggers. It is interesting to note that the
right finger triggers as well after a delay time of 1ns even at a
lower drain voltage than its trigger voltage. This implies that
the triggering of the second transistor doesn’t rely on the drain
voltage alone. Fig. 3 shows how potential contours spread out
from the channel of the first finger and reach the second and
third finger to trigger them. This result clearly shows that the
uniform triggering of multi-finger structures is achieved by
substrate coupling effect. The result for epi wafers is shown in
Fig. 4. A t
epi
of 0.75µm was used. Since p+ dopants in the p+
substrate move up to the p- epi layer, the effective p- epi
thickness (t
epi
) is reduced with heat cycle for a given process.
The effective epi thickness is defined by the distance from the
silicon surface to the p- epi with a dopant concentration of
1E18/cm
3
as shown in Fig. 1 (b). We observe that the left
transistor with a lower channel doping absorbs the entire ESD
current along with the substrate current (I
sub
) since the
neighboring NMOS (F2) doesn’t trigger. This result is not
consistent with the bulk wafer result that neighboring fingers
are turned on by substrate coupling effects. The potential
contour shown in Fig. 5 confirms that substrate coupling is not
sufficient for uniform triggering in this case. The impact of t
epi
on substrate-coupling-induced multi-finger triggering is shown
in Fig. 6. Three fingers of NMOS were used in this
simulation. We see that triggering uniformity in multi-finger
structures is a strong function of t
epi
. In particular, the delay
time for triggering between fingers varies with t
epi
, which
determines current distribution among the fingers. Note that
the third finger (F3) shows a larger delay time than that of the
second finger since it triggers after the second finger.
Fig. 7 represents a multi-finger turn-on model incorporating
substrate coupling effects. For epi-type substrates the heavily
doped bulk can be considered as one electrical node and only
the resistance of the p- epi layer has to be considered as
spreading resistance. As indicated in the potential contours of
Fig. 5, the substrate potential floats up under the left gate but
cannot spread laterally to raise the potential under the second
poly finger. Because of the low resistivity and thickness of
the p+ bulk, the injected hole current flows almost directly
down through the p- epi layer into the p+ bulk. Therefore, the
model equation for the case of weak finger to finger substrate
coupling (poor triggering uniformity) may be written as
G
vertical
> G
lateral
(1)
The parameters G
vertical
and G
lateral
represent the conductivity
of the vertical (finger to p+ bulk) and lateral (finger to finger)
direction, respectively. For the bulk wafer and epi-type
substrate with t
epi
larger than 1.5um cases, we would expect
that the multi-finger triggering occurs with
G
vertical
< G
lateral
(2)
In Fig. 7, increasing the source resistance (Rs) improves the
coupling effects since Rs*Is1 enhances the propagation of the
substrate potential (Φ1) to the neighboring finger (Φ2). This
phenomenon can be understood by comparing the source
current of the right finger in Fig. 1 (b) with different Rs values
as shown in Fig. 8. The same reasoning can also be applied to
non-silicided (silicide blocked) transistors that have higher Rs
than that of fully silicided transistors.
Experimental Results for Fully Silicided Transistors
Fig. 9 shows the width dependence of fully silicided multi-
finger structures for a 0.18µm technology using bulk wafers.
As expected from our multi-finger turn-on model, the failure
current (i
t2
) shows a strong width dependence like non-
silicided transistors. We have used fully silicided NMOS
transistors as a primary protection device and/or self-
protecting output drivers in real products without any
design/process options, and pass 3000V HBM and 700V
CDM. A high current I-V curve measured from a real IO of
our 0.13µm technology is shown in Fig. 10. These results are
consistent with the simulation results in that they also indicate
that, for bulk wafers with high substrate resistivity, fully
silicided multi-finger structures will show uniform triggering
due to substrate coupling effects, even with V
t2
< V
t1
.
Conclusion
In this paper, we have presented a multi-finger turn-on
model incorporating substrate coupling effects in a multi-
finger structure. We have demonstrated the importance of the
starting material for triggering uniformity and thus current
handling capability of multi-finger transistors. The
experimental and simulation results provide insight into the
nature of uniform triggering, and show the possibility of using
fully silicided transistors as an ESD protection device without
any process/design modifications.
References
[1] Thomas L. Polgreen, et al., IEEE TED-39,
pp. 379, 1992.
[2]
Markus P. J. Mergens, et al.,
ESD/EOS Symposium 2001, pp. 1.
[3] Ajith Amerasekera, et al., IEEE TED-38,
pp. 2161, 1991.
[4]
SEQUOIA Device Designer User’s Guide,
SEQUOIA Design Systems, 1998-2001.
Fig. 1 – Cross section of multi-finge
r
N
MOS transistors. Note that F1 has a 10%
lower channel doping than that of F2.
Fig. 2 – The current waveforms along with
the drain voltage for bulk wafer case.
Fig. 3 – Potential contours of multi-finger structure on bulk wafer. Potential
contours spread out from the first finger (F1) and reach the second finger to
trigger it. The third finger turns on with the same mechanism (domino-effects).
S1 D1 S2 D2
S1 D1 S2 D2
F1 F2 F3
F1 F2 F3
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1E-09 2E-09 3E-09 4E-09 5E-09
TIME (SEC)
Iesd (A)
0
1
2
3
4
5
6
7
DRAIN VOLTAGE (V)
Is2 Is1 Id Vd
source current of F2
source current of F1
drain current
drain voltage
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1E-08 2E-08 3E-08 4E-08 5E-08
TIME (SEC)
Iesd (A)
0
1
2
3
4
5
6
7
DRAIN VOLTAGE (V)
I
d
V
d
I
s1
I
sub
I
s2
source
drain
p- substrate
source
V
ESD
A
A
I
S2
+
-
I
S1
+
-
(a) bulk wafer
F1
F2
Fig. 4 – The current waveforms along
with the drain voltage for epi wafer.
N
ote that F2 (I
s2
) doesn’t trigger.
Fig. 5 – Potential contours at the
peak of ESD current (t=10ns) for two
finger structure. The t
epi
is 0.75µm.
N
o substrate coupling occurs.
Fig. 6 – Triggering delay as a
function t
epi
. Three fingers of NMOS
were used in this simulation.
0.0E+00
1.0E-01
2.0E-01
3.0E-01
4.0E-01
5.0E-01
6.0E-01
0.0E+00 1.0E-08 2.0E-08 3.0E-08 4.0E-08
TIME (SEC)
Iesd (A)
Rs=2ohm
Rs=1ohm
Rs=0.5ohm
Rs=0.25ohm
Rs=0.1ohm
Rs
G
lateral
G
vertical
G
vertical
Rs
F1
F2
V
ESD
φ
1
φ
2
I
S1
Fig. 7 – Representation of multi-
finger turn-on model. The multi-
finger triggering occurs when lateral
conductance is greater than vertical
conductance (G
lateral
> G
vertical
).
Fig. 8 – Simulation results for multi-
finger triggering as a function o
f
source resistance (Rs) shown in Fig. 7.
The source current (I
s2
) of the second
finger was measured.
Fig. 10 - TLP I-V characteristics measured
from real IO with silicided NMOS as a sel
f
protecting output driver. The width is
360um and length is 0.24µm. Note that V
t1
is larger than V
t2
.
Fig. 9 – TLP I-V characteristics for
multi-finger structures having 2, 4, 6,
8,and 14 poly fingers which are 25µm
wide with channel length of 0.18µm.
I
t2
level of 8mA/µm was obtained.
(a) right after F1 trigger (at t=1ns) (b) at the peak of ESD current (at t=10ns)
source
drain
p+ substrate
source
V
ESD
A
A
I
S2
+
-
I
S1
+
-
F1 F2
t
epi
1E18/cm
3
(b) epi wafer
p- epi
0
0.5
1
1.5
2
2.5
3
0246810
VPAD (V)
Iesd (A)
1.0E-09 1.0E-06 1.0E-0
3
Ileak (A)
I
esd
I
leak
V
t2
V
t1
0.0E+00
1.0E-09
2.0E-09
3.0E-09
4.0E-09
5.0E-09
6.0E-09
7.0E-09
8.0E-09
9.0E-09
0.511.522.533.544.5
Epi Thickness (um)
Dela
y
Time
(
SEC
)
dela
y
time between F1 and F2
(t
F2
- t
F1
)
No triggering of 3'rd finger
left of this point
No triggering of
2'nd finger left
of this point
dela
y
time between F1 and F3
(t
F3
- t
F1
)
0
0.5
1
1.5
2
2.5
3
3456
VPAD (V)
Iesd (A)
F=2
F=4
F=6
F=8
F=14