Voltage Pulse Stress Effect on Gate stack TDDB distributions
at nanometric scale: Consequence on Aging by ESD
R. Foissac, S. Blonkowski, M. Gros-Jean
STMicroelectronics
850 rue Jean Monnet, 38926, Crolles, France
+33 4 38 78 03 97, romain.foissac@cea.fr
M. Kogelschatz,
Univ. Grenoble Alpes, Lab. LTM-UMR CNRS (CEA-LETI/Minatec),
17 avenue des martyrs, 38054 Grenoble, France
Abstract— The effects of a voltage pulse stress on time to
breakdown and breakdown voltage distributions of a 1.4nm thick
SiON layer are investigated at nanometric scale using an atomic
force microscope in conduction mode under ultra-high vacuum.
Thanks to the reduced tip/sample contact area, a large window of
voltage pulse amplitude and length is allowed. It is found that the
pre-stress pulse does impact the slope of the cumulative failure
distribution but has no effect on the voltage acceleration factor.
Using this assumption an extrapolation formula is given to
predict the decrease of the time to breakdown as a function of the
pre-stress pulse parameters.
ESD; C-AFM; Dielectric Breakdown
I. INTRODUCTION
With the downscaling of microelectronics devices, new
questions have been raised concerning the impact of
electrostatic discharge (ESD) on ultra-thin gate oxide lifetime.
Effects of ESD on CMOS devices has been widely studied [1 -
5]. It has been shown that the main statistical properties i.e.
Weibull slope and acceleration factor observed for usual Time
Dependent Dielectric Breakdown (TDDB) remain unchanged
[4, 5] at the time range of ESD. However the impact of a pre-
stress pulse on TDDB distributions seems more pertinent to
study the effect of an isolated ESD on the Time to Breakdown
(t
BD
). According to [6, 7] the impact of such a pre-stress pulse
seems to slightly affect the Weibull slope.
At nanometric scale, thanks to the area scaling effect, the
Breakdown Voltage (V
BD
) is sufficiently high [8] to measure
the impact of larger preset voltage pulse than at a device scale.
Atomic Force Microscopy used in conductive mode (C-AFM)
is a powerful technique to study such a local phenomenon as
the dielectric breakdown. Due to the reduced tip/sample
contact area, the tunneling current is decreased providing new
insights in pre breakdown current characteristics [9, 10]. The
nanoscale surface of the C-AFM tip allows a large range of
pulse amplitude and width without being destructive for the
oxide layer. In this study C-AFM in ultra-high vacuum is used
to study the effect of the pre-stress pulse height and duration on
the TDDB and V
BD
statistic distributions. An extrapolation
formula for the the time for which 63% of devices have failed,
t
63%
evolution as a function of the pulse parameters, i.e. ESD
aging as a function of the pre-stress pulse characteristics will be
deduced.
II. EXPERIMENTAL DETAILS
In the present study, C-AFM electrical tests have been
carried out on a 1.4nm thick SiON layer on Silicon substrate
(5.07 x 10
15
cm
-3
, p-type). The silicon oxinitride layer has been
prepared in the following way: after a HF-SC1 clean, a
thin SiON interfacial layer was formed by oxidation of the
silicon substrate with a 800°C RTO, followed by Inductively
Coupled Plasma (ICP) nitridation and a post nitridation anneal
at 1000°C.
C-AFM measurements were performed at room
temperature with an Omicron AFM/scanning tunneling
microscopy system under UHV (<10
9
Torr) with conductive
diamond tips (B doped). The AFM tip was used as a grounded
top electrode (tip area = 10 nm² [8]), and a negative voltage
was applied to the substrate. The current was recorded using a
Keithley 6430 equipped with a sub-femtoampere sourcemeter.
Electrical contact between the sample and the stage was
assured with indium solder, and all of the samples were
outgassed at 150°C for 3 h at 4x10
8
Torr. The electrical
measurements were done in contact mode (normal force = 20
nN) while applying a constant voltage to the stack.
Topographic images were performed in contact mode using the
diamond tip. Current mapping was performed on the dielectric
with a constant tip voltage while the tip is scanning the layer
surface, the current was recorded by the AFM software Iscala.
III. EXPERIMENTAL RESULTS
In this paper the effect of a pre-stress pulse is characterized
by measuring either the breakdown voltage by applying a
Ramp Voltage Stress (RVS) or the time to breakdown by
applying a Constant Voltage Stress (CVS). For each point of
the distributions the tip is stopped on the oxide, a pre-stress
pulse is triggered followed by RVS or CVS until the dielectric
breakdown is detected. The pulsed voltage pre-stress is
supplied by a DS345 function generator from Stanford
Research Systems and the CVS and RVS characteristics are
obtained by using a Keithley 6430 electrometer (Fig.1).
French Network Renatech
978-1-4799-3317-4/14/$31.00 ©2014 IEEE GD.5.1