1. 2009 NCN@Purdue‐Intel Summer School. Notes on PercolaKon and
Reliability Theory. Lecture A4a. On Reliability and Randomness in Electronic
Devices.
2009
NCN@Purdue‐Intel
Summer
School
Notes
on
Percola9on
and
Reliability
Theory
Lecture
A4a
On
Reliability
and
Randomness
in
Electronic
Devices Muhammad A. Alam Electrical and Computer Engineering Purdue University West Lafayette, IN USA NCN
www.nanohub.org
1
outline of lecture 7 1. Background
informa5on
2. Principles
of
reliability
physics
3. Classifica5on
of
Electronic
Reliability
4. Structure
Defects
in
Electronic
Materials
5. Conclusions
2
process, reliability, and design Reliability
Process plus
t=0
-15%
-10%
-5%
ID
ID,nom
+15%
Design
We do not have too much margin
3
complexity of the reliability problem Mobility
Improvement
TI
Ge
channel
Strained
silicon
uniaxial
bi‐axial
III‐V
materials
Improvement
in
Electrosta5cs
Berkeley
FINFET
IBM
Silicon
on
insulator
(SOI)
Double‐gated
devices
Tri‐gate
or
FINFET
devices
Surround
gate
devices
(VRG,
Si‐NW)
4
outline of lecture 7 1. Background
informa5on
2. Principles
of
reliability
physics
3. Classifica5on
of
Electronic
Reliability
4. Structure
Defects
in
Electronic
Materials
5. Conclusions
5
Equivalence between Spatial/Temporal Fluctuation Process
Reliability plus
Side view (TDDB)
top view (RDF)
Spa5al
and
temporal
fluctua5on
should
be
considered
with
same
framework
…
model
6
Equivalence of spatial/temporal fluctuations Process
Device
Process
Reliability
Same
theore5cal
tools
Device
•
Time‐ordered
(N~t)
•
Spa9al
Correla9on
Temp
Manufacture
E‐field
Store
Home/office
Time
7
Gate
Current
reliability: stochastic process terminated by a threshold
Pey,
IRPS02
HBD
ln
(5me)
8
reliability: physics of how things break A
child
drops
a
glass
Bridges
(Tacoma
Narrows)
ShuLle
(Challenger)
Ligh9ng
in
a
rain‐soaked
night
Volcano,
landslides
&
forest
fire
Check‐out
queues,
scheduling
A stochastic process terminated by threshold 9
theory of accelerated testing Nonlinear
projec5on
for
processes
with
threshold
V3>
V2
>
V1
Days
Vop
Years
ln
(TBD)
Gate
Current
years
Std.
Days
V G
Empirical projection is very difficult, if not impossible … 10
Nonlinear Projection: an illustrative example Stochas9c
Process
Stochas9c
Process
with
a
threshold
Absorption site
v
How far away from the trap-site do I need to inject the particles so that after TBD sec of diffusion no more than y percent of particles is lost? 11
accelerated testing: empirical approach m
v
Absorption site q
p
accelerate
p~1
ln
Tav
years
decreasing
p
vsafe
p
or
velocity
5me
12
average arrival time distribution q
x=0
p
x=xm
x‐d
x
x+d
Time
to
absorp9on
a\er
being
injected
at
x
….
13
average arrival time distribution x=0
q
p
x=xm
x‐d
x
x+d
Average lifetime diverges at small velocity, v or p -> 0 14
physical vs. empirical projection p
q
x=0
x=xm
Ln
T
x
physical
years
vsafe
vsafe
Empirical
meas. & comp. simulation would not do
Prediction is possible because t=0- is smooth
15
outline of lecture 7 1. Background
informa5on
2. Principles
of
reliability
physics
3. Classifica5on
of
Electronic
Reliability
4. Structure
Defects
in
Electronic
Materials
5. Conclusions
16
front and backend reliability
Transistor
Reliability
Issues
The
Oregonian,
2008
Gate
Dielectric
Breakdown
Nega5ve
Bias
Temperature
Instability
Hot
Carrier
Degrada5on
Radia5on
Induced
damage
Interconnect
Reliability
Issues
Electro‐migra5on/Stress‐migra5on.
Inter‐level
dielectric
breakdown
17
Initially ….
S
G
.... a few months later
D
Drain Current
front end transistor reliability VG2
VG1
Drain Voltage Broken
Si‐H
bonds:
Nega5ve
Bias
Temperature
Instability
(NBTI)
Hot
carrier
degrada5on
(HCI)
Broken
Si‐O
bonds:
Gate
dielectric
Breakdown
(TDDB)
Radia5on
induced
damage
Trapping:
Posi5ve
bias
temperature
instability
18
outline of lecture 7 1. Background
informa5on
2. Principles
of
reliability
physics
3. Classifica5on
of
Electronic
Reliability
4. Structure
Defects
in
Electronic
Materials
5. Conclusions
19
Crystalline, Amorphous, and Random Materials Crystalline
Amorphous
Random
N
r 20
crystalline Si and SiO2 Crystalline
Si
Crystalline
SiO2
For
a
dynamic
view
of
SiO2
see,
hdp://cst‐www.nrl.navy.mil/laece/struk.jmol/coesite.html
crystalline vs. amorphous SiO2 Fixed
bond
lengths
Sof
Si‐O‐Si
bond
angles
oxygen
Si
22
composition vs. coordination
4
Si
atoms
x
1/8
=
½
1
Si
atoms
(white)
1
oxygen
atom
(white)
=
1
4
oxygen
atom
x
½/cell
=
2
Si1/2O
=
SiO2
SiO2
…..
Oxygen
coordina5on
…
4
Hf
coordina5on
–
8
Oxygen
coordina5on
…
2
Silicon
coordina5on
…
4
23
amorphous is neither completely Random, nor it is defective Amorphous
small scatter in ring distribution … 24
geometry defines ring statistics Euler
Rela5onship
in
2D
ver5ces
edges
cell
number
Interpreted
as
edges/cell
1/2
of
a
cell:
must
have
6‐sides!
25
Ring Statistics in Random Structure Euler
Rela5onship
Prob.
of
k‐sided
ring
26
amorphous material and bandtail states
Ec
Ev
27
origin of defects: Maxwell Constraints – a Poor Man’s MD Zero
T
approxima5on
dimensionality
Points
to
be
stabilized
constraints
q1
q2 q3 Maxwell,
Phil.
Mag.
27,
1864,
294.
J.C.
Phillips,
1979.
Thorpe,
1982.
28
examples: SiO2 and Si3N4
Maxwell,
Phil.
Mag.
27,
1864,
294.
J.C.
Phillips,
1979.
Thorpe,
1982.
29
meaning of an oxide/nitride defect
30
crystalline vs. amorphous oxides
NY
Times,
2007.
31
bulk defects of missing oxygen: E’, K, and P defects Courtesy:
Prof.
Lenahan
Responsible for gate dielectric breakdown Will discuss in Lecture 8 and 9
Pb centers – interface traps
Pb1
[211]
[111]
Responsible for NBTI degradation Detailed discussion in Lecture 10 33
Interface structures by Molecular dynamics A. Bongiorno et al. PRL, 90, 2003.
34
electron spin resonance: a ‘microscope’ for defects Variable
Fixed
Microwave
B‐field
ΔE
=
geμBB0
hν
=
geμBB0
Microwaves
@
~
10
GHz
B
~
3500
G
(0.35
T)
Absorp5on
Spectra
35
ESR invisible defects Paramagne5c
materials
may
appear
diamagne5c
Phonon assisted e-e attraction (superconductivity).
36
conclusions
Reliability
is
a
very
important
technology
problem.
The
success
or
failure
of
any
technology
depends
on
its
intrinsic
reliability
(e.g.
Stone
vs.
Cu
age)
Historically
reliability
has
been
an
empirical
science;
it
has
changed
drama5cally
since
WWII
(Military
applica5ons,
AT&T
trans‐Atlan5c
cable,
IBM
mainframe,
Introduc5on
of
CMOS).
Reliability
as
a
‘Stochas5c
process
with
a
threshold’
provides
the
physical
basis
for
wide
range
of
phenomena.
There
are
many
reliability
concerns
for
electronic
devices
–
an
analysis
of
these
problems
beings
with
the
analysis
of
the
nature
of
defects.
We
will
related
bulk
oxide
E’
centers
to
TDDB
and
depassiva5on
of
SiH
bonds
at
the
interface
to
NBTI.
figure references Checkout:
hdp://vinlai.com/gallery2/v/1042‐2/heathrow_4.jpg
hdp://i111.photobucket.com/albums/n125/woodz1415/broken20glass.gif
Lightning:
hdp://www.uwec.edu/jolhm/EH3/
Group2/Pictures/lightning.jpg
Concord:
hdp://www.ahrtp.com/EG_Images2/ Concorde_crash_opt600x412_concordesst.jpg
hdp://www.theharrowgroup.com/ar5cles/20020513/TacomaBridge.gif
hdp://www.ifw‐dresden.de/ins5tutes/ikm/organisa5on/dep‐31/ research‐topics/electromigra5on/insituem_small.jpg
38