2012 Unexpected failures due to dynamic avalanching caused by bipolar ESD stress

Conference paper Unexpected failures due to dynamic
avalanching caused by bipolar ESD stress
Taiwan ESD and reliability conference 2012
The bipolar nature of ESD pulses such as MM introduces failure
mechanisms that cannot be reproduced by TLP/HBM. A lowered
breakdown voltage due to dynamic avalanching was observed. The
key issue is that carriers injected during the first swing remain in the
device after the current switches polarity. A case study for high-
voltage diodes is presented.

Unexpected failures due to dynamic avalanching
caused by bipolar ESD stress
Bart Sorgeloos, Ilse Backers, Benjamin Van Camp, Olivier Marichal
Sofics BVBA, Groendreef 31, 9880 Aalter, Belgium,
phone: +32-9-21-68-333; fax: +32-9-37-46-846; e-mail: bsorgeloos@sofics.com
Abstract - The bipolar nature of ESD pulses such as MM introduces failure mechanisms that cannot be
reproduced by TLP/HBM. A lowered breakdown voltage due to dynamic avalanching was observed. The key
issue is that carriers injected during the first swing remain in the device after the current switches polarity. A
case study for high-voltage diodes is presented.
I. Introduction
Transmission Line Pulsing (TLP) is an indispensable
tool in analyzing ESD protection designs and making
accurate predictions about their corresponding ESD
robustness. Failure current (It2) measurements
obtained from a 100ns TLP tester are typically used to
define a technology-specific correlation factor
between TLP failure current and HBM or MM failure
voltage. On the other hand, for CDM, this is not so
straightforward because of the very short duration of
the pulse (only a few nanoseconds) and the fact that
CDM is a one-pin test as opposed to the two-pin TLP
set-up. Even the correlation with Very Fast TLP (VF-
TLP has yet to be proven [1]. For HBM and MM it is
more common to have a good correlation, although
this is not always true. In the past various correlation
issues between TLP and HBM have been published. A
well determined and widely applicable correlation
factor offers many advantages. If a linear scaling of
the perimeter of a protection device gives a linear
scaling in TLP failure current, such a correlation
factor will also induce the same linear increase in
HBM and MM failure voltage. This enables the ESD
designer to reach any targeted HBM or MM
specification. However, this assumption is not always
valid. In some cases, the MM failure voltage is much
lower than the calculated failure voltage by using the
fixed HBM/MM ratio and empirical HBM test results.
Furthermore the MM failure voltage doesn’t scale in
function of the perimeter of the device, even though
HBM does. In this case, it is impossible to use one
unique ratio between MM and TLP or HBM for that
technology. This lower failure level is in this case
only observed with MM tests and never with HBM or
TLP tests.
In general there are three good reasons to justify the
existence of a fixed ratio between HBM and MM.
First of all, although the peak current and induced
energies differ between the HBM and MM model,
these can be correlated and taken into account in the
correlation factor (Figure 1). Secondly, the rise time
of both models is comparable. Finally, the bipolar
nature of MM (this means that the current/voltage has
both polarities (positive and negative) during the
course of the pulse) is rather slow so that for most
cases the influence of the previous positive or
negative swing can be neglected, reducing the
physical effects of MM to the unipolar case as HBM.
All these three facts explain the theoretical validity of
the fixed correlation factor used in the industry
between HBM and MM.
Figure 1: MM and HBM pulse, note the bipolar characteristic of
the MM pulse
However, this is not always true. Important exceptions
can arise, because the bipolar nature of MM will have
an important influence on the behavior of the ESD
circuits, which can no longer be neglected. Note that
the bipolar nature of ESD pulses is not just a
characteristic of a MM pulse alone, but also of other
ESD models like system level (IEC tests) as presented
in previous work [4] and that the phenomena
described in this paper could also occur in these cases.
In the following section, this paper introduces one of

the cases where we have seen the influence of the
bipolar nature of an ESD pulse. Afterwards the
concept of “dynamic avalanching” which causes this
correlation issue will be introduced and a deeper
insight in the physical background of this
phenomenon will be offered. Finally, with knowing
the theoretical background a solution is explained in
the last section. We have found this problem only in
high voltage processes.
II. Case study
One of the technologies where we have seen the
problem of the bipolar nature of the MM pulse is a
0.35 um 13.5V CMOS process. The product targeted
the high volume LCD driver market for big flat-screen
TVs. The protection scheme of one of the outputs is
shown in Figure 2. A small output driver is protected
with a dual diode protection scheme and a 200 Ohm
resistor connected to the drains of the driver. The
diode up is formed in an NWell and the diode down in
the P-substrate. The power clamp is a SGPMOS
(source gated PMOS) triggered SCR (SGPSCR). This
SGPSCR is optimized to prevent latch up.[3]
Figure 2: ESD Output protection for a 0.35 um CMOS process
All possible pin combinations reached both the 2kV
HBM and 250V MM specification, except the stress
case where the output pin was stressed positively with
respect to the VDD power pin (Table 1). For HBM,
this stress case translates into current flowing from the
output pin to the power line (VDD) through the diode.
The designed behavior of the ESD protection during
MM can be explained as follows: during the first
positive MM pulse the diode up will conduct current.
When the MM pulse becomes negative the diode will
be in reverse mode. As the trigger voltage of the
clamp is slightly higher than the avalanche voltage of
the diode, the diode will conduct some avalanching
current in reverse. The voltage across the diode stays
well below its failure voltage. When the trigger
voltage of the clamp is reached the SGPSCR will turn
on and take all the ESD current (due to the low
holding voltage of SGPSCR). All the clamps are well
designed to take the appropriate amount of current.
The failure voltage obtained from a HBM test for this
stress case equaled 5.3 kV. For this technology the
experimentally extracted HBM/MM ratio is
determined to be 20. This factor is minimum observed
correlation factor between HBM and MM from
several hundred correlation measurements over
various device types processed in this technology.
Therefore, the calculated MM value should at least
equal 265V in this case.
Table 1: HBM/MM test results
Figure 3: Failure analysis of diode up
Measurements, however, indicate a failure voltage of
150 V MM, which is significantly lower than the
expected value given above. After testing, failure
analysis showed that the diode up was destroyed with
the worst damage occurring at the edge of the diode. (
Figure 3).
The static avalanching voltage of the diode is 18V and
the failure voltage 63V. The power clamp triggers at a
voltage (24V – Figure 4) which is lower than the
failure voltage of the diode in reverse.
Assuming the bipolar nature of MM can be neglected,
this data can not explain why the diode up fails during MM.
Another striking observation is that, when the same stress
is applied to the PWell diode (stress between VSS and output) the
results are as expected based on TLP and HBM results: no early
MM failure occurs in this case.
VDD
VSS
Output
200 Ohm
HBM: 2kV VSS-GND IO-GND VDD-GND
VSS-ZAP pass pass
IO-ZAP pass pass
VDD-ZAP pass pass
MM: 250V VSS-GND IO-GND VDD-GND
VSS-ZAP pass pass
IO-ZAP pass fail
VDD-ZAP pass pass
Diode UP
(Nwell diode)
Diode DOWN
(Pwell Diode)
VSS
VDD
Output
destruction

In the domain of discrete power devices, there are
some studies about a phenomenon called “dynamic
avalanching”[3-9]. This effect can explain the early
failure of the diode in reverse.
Figure 4: TLP measurement of the power clamp (in product) TLP
IV curve on left side. Leakage plot on right side. The power clamp
fails at a current above 2.5A. The Vt1 trigger voltage is about
24V.
III. Dynamic Avalanching
There are fundamental differences between a quasi-
static current swing and a dynamic one. In some cases
a MM pulse can not be considered quasi-static and
then dynamic avalanching can occur (Figure 5) [3-9].
In total, three degrees of dynamic avalanching exist.
Considering a simple PN-junction diode, during the
first positive MM swing, the diode is in forward
mode, and excess to carriers are injected into the bulk
of the diode (High Injection Mode, Figure 5.1). As the
current pulse switches polarity, these excess carriers
will still be present in the diode bulk (Figure 5.2) with
a reversed biased junction. And the carriers will even
contribute in the reverse avalanching, such that the
avalanching is more severe and occurs earlier in the
dynamic case for the same electrical field as
compared to the static case (Figure 5.3). This is called
first degree of dynamic avalanching, which is
typically non-destructive. The failure voltage of the
diode is nearly the same as with static avalanching.
For higher forward bias current levels, a feedback
mechanism is present that lowers the avalanche
breakdown voltage: the excess carriers can be
annihilated by the carriers created by the avalanching.
This creates a small negative resistance regime, which
is the signature of the second degree of dynamic
avalanching. Although this causes some current
filamentation, this is most often still not destructive,
as it is counteracted by two other mechanisms:
 The current filamentation can rapidly and locally
deplete the bulk of the diode from excess carriers,
such that the cause for the dynamic avalanching
locally disappears.
 With the current filamentation, a hot spot is
created. The avalanche voltage increases with
temperature, overcoming the negative resistance
regime.
Figure 5: Schematic representation, comparing the dynamic case
(1a-3a) to the static case (1b-3b) for three different moments in
time (1: Strong forward bias, 2: zero bias, 3: negative bias)
Destruction is only likely when the third degree of
dynamic avalanching is reached. The large current
flow through the diode causes the formation of a
strong electric field at the N+/Nwell junction.
Eventually this field will reach a value high enough to
cause this junction to start avalanching. The generated
carriers increase the excess carriers in the bulk of the
diode, thereby causing a positive feedback. The
electrical field associated with this behavior is shown
in Figure 6a. Instead of a linear field (dotted line),
two peaks are observed at the P+/Nwell and N+/Pwell
junctions. This is called an “Egawa” field. A
consequence is that due to the higher fields the diode
fails at a much lower voltage. Although this figure is
taken for a discrete power diode, the physics are
similar for CMOS technologies.
Dynamic Avalanching is one of the physical effects
that show that the fundamental differences between
uni-polar stress such as TLP and HBM and bipolar
stress, such as MM, cannot be neglected. It makes it
clear that using a unique correlation between MM and
TLP is not always possible. Not only diodes are
influenced by the bipolar nature of MM. Another
example related to this subject is the lowering of the
0.00
0.50
1.00
1.50
2.00
2.50
3.00
0.0 5.0 10.0 15.0 20.0 25.0 30.0
Voltage [V]
Current[A]
0.00
0.50
1.00
1.50
2.00
2.50
3.00
1.0E-10 1.0E-09 1.0E-08 1.0E-07 1.0E-06 1.0E-05 1.0E-04
Current [A]
Current[A]

trigger voltage due to carriers in the substrate of a
ESD clamp device. This becomes important when the
trigger voltage of a parasitic path becomes lower than
this of an ESD clamp. These effects cannot be
explained with uni-polar stress methods such as HBM
or TLP. This is also observed in [1].
Figure 6: a)Schematic representation of the electrical field (Solid
line) during the third degree of dynamic avalanching. b)
Formation of the depletion regions at different times [9]
In order to have a comprehensive understanding, an
additional explanation is needed for why the same
failure does not occur for the diode down. The
relevant stress case is now a negative MM zap
between IO and Vss. Considering Figure 6.b, three
regions can be designated in the diode: a depletion
region at the P+/NWell and N+/NWell region and an
intrinsic region between. In the depletion regions,
current is handled by carriers drifting through the
large electric field. The main current flow mechanism
in the second region is diffusion of the carriers from
the high concentration intrinsic regions towards the
depleted region. It can easily be seen that the
depletion region grows much faster at the P+ side than
at the N+ side. This is due to the difference in
diffusion coefficient between holes and electrons. At
the left side the velocity is determined by the electron
mobility of the electrons diffusing towards the
N+/NWell junction and at the right side due to hole
mobility of the holes flowing to the P+/Pwell
junction. Thus, the dynamic avalanching behavior is
the same for Nwell as for Pwell diodes. The electrical
field however is built up over the PN junction. The
voltage can change more rapidly in the depletion
region than in the intrinsic region (i.e. the region with
excess carriers). This gives rise to different behavior
of both diode types: for the Nwell diode, the voltage
built up is fast, as it is located at the rapidly growing
P+/Nwell depletion region, such that the excess
carriers will be swept through a large electrical field.
For Pwell diodes, the voltage change is only
moderately fast, because the N+/Pwell depletion
region extends slower [10]. Therefore, there is more
time for the excess carriers to leave the bulk of the
diode, without having to go through a large electric
field.
Figure 7: New protection scheme
IV. Improvements
Due to the Egawa fields the failure voltage of the
diode will be much lower compared to the static case
(63V). One solution to improve the results of the
reflected case is to lower the trigger voltage of the
power clamp. This voltage must be lower than the
dynamic avalanching and failure voltage of the diode
in reverse. This is solved for this product by using an
RC triggering scheme connected to the gate of the
PMOS of the SGPSCR. The MOS does no longer
operate in bipolar mode but uses MOS current to turn
on the SCR. The width of the PMOS, R and C are
determined with a SPICE simulation. The lower
trigger voltage of 10V gives the expected HBM and
MM specifications. (Figure 7 and Figure 8).
VDD
VSS
Output
200 Ohm
C
R

The diode inherent in the PMOS is protected by the
200 Ohm resistor.
Figure 8: Simulation of the behavior of the improved clamp
V. Conclusion
It’s crucial to be aware that the bipolar nature of an
ESD pulse can become important in some cases and
cannot always be neglected. Bipolar pulses, such as
MM, cannot always be considered as a sequence of
unipolar pulses. Effects as dynamic avalanching,
never seen in HBM or TLP pulses, become important.
Using a fixed correlation between HBM and MM can
therefore sometimes result in unexpected failures in
products. It is important that appropriate test set-ups
are used to accurately analyze ESD protection designs
in light of this phenomenon and to mitigate the risk of
unexpected product failures to the MM.
References
[1] Horst Gieser et al.,“Very-fast Tranmission line pulsing of integrated
structures and the charge device Model”, EOS/ESD symp. 1996
[2] Whitfield J. et al, “ESD MM Failures resulting from transient reverse
currents”, IRPS 2006
[3] Bart Keppens et al., “ESD Protection Solutions for HV
technologies”, EOS/ESD symp. 2004
[4] Ming-Dou Ker et al., “Component-Level Measurements for
Transient-Induced Latch-up in CMOS ICs Under System-Level ESD
considerations”, IEEE Transactions on Device and Materials
Reliability, Vol. 6, No. 3,, September 2006
[5] Benda V., “Design Considerations for Fast Soft Reverse Recovery
Diodes”, 1967
[6] M. Mori et al., “6.5 kV Ultra Soft & Fast Recovery Diode (U-SFD)
with High Reverse Recovery Capability”, ISPSD 2000
[7] Y. Wang et al., “Prediction of PIN Diode Reverse Recovery”, 2004
35th Annual IEEE Power Electronics Specialists Conference
[8] J. Lutz, “Fast Recovery diodes – reverse Recovery Behavior and
Dynamic Avalanche”, Proceedings inter-national conference on
Microelectronics 2004, Vol 1
[9] J. Lutz et al., “Dynamic Avalanche and reliability of high voltage
diodes”, Microelectronics Reliability 43, 2003, p529-536
[10] Michael J. Cudobiak, “New Approaches For Designing High
Voltage, High Current Silicon Step Recovery Diodes for Pulse
Sharpening Applications” PhD Thesis, Chapter 2
(http://www.avtechpulse.com/papers/thesis/2)

2012 Unexpected failures due to dynamic avalanching caused by bipolar ESD stress

Sofics Proprietary – ©2012
About Sofics
Sofics (www.sofics.com) is the world leader in on-chip ESD protection. Its
patented technology is proven in more than a thousand IC designs across all
major foundries and process nodes. IC companies of all sizes rely on Sofics for off-
the-shelf or custom-crafted solutions to protect overvoltage I/Os, other non-
standard I/Os, and high-voltage ICs, including those that require system-level
protection on the chip. Sofics technology produces smaller I/Os than any generic
ESD configuration. It also permits twice the IC performance in high-frequency and
high-speed applications. Sofics ESD solutions and service begin where the foundry
design manual ends.
Our service and support
Our business models include
 Single-use, multi-use or royalty bearing license for ESD clamps
 Services to customize ESD protection
o Enable unique requirements for Latch-up, ESD, EOS
o Layout, metallization and aspect ratio customization
o Area, capacitance, leakage optimization
 Transfer of individual clamps to another target technology
 Develop custom ESD clamps for foundry or proprietary process
 Debugging and correcting an existing IC or IO
 ESD testing and analysis
Notes
As is the case with many published ESD design solutions, the techniques and
protection solutions described in this data sheet are protected by patents and
patents pending and cannot be copied freely. PowerQubic, TakeCharge, and
Sofics are trademarks of Sofics BVBA.
Version
December 2012
ESD SOLUTIONS AT YOUR FINGERTIPS
Sofics BVBA
Groendreef 31
B-9880 Aalter, Belgium
(tel) +32-9-21-68-333
(fax) +32-9-37-46-846
bd@sofics.com
RPR 0472.687.037

2012 Unexpected failures due to dynamic avalanching caused by bipolar ESD stress

More Related Content

What's hot (17)

Similar to 2012 Unexpected failures due to dynamic avalanching caused by bipolar ESD stress (20)

More from Sofics (20)

Recently uploaded (20)

2012 Unexpected failures due to dynamic avalanching caused by bipolar ESD stress