Transcript SEE Mechanisms in Technology Nodes Below 100nm

SEE Scaling Effects
Lloyd Massengill
10 May 2005
Hierarchical Multi-Scale Analysis
of Radiation Effects
Institute for Space and Defense Electronics
Vanderbilt Engineering
Materials
Device Structure
IC Design
Energy
Deposition
Defect Models
Device Simulation
MURI 05 Kickoff
10 May 2005
Circuit Response
2
Scaling and SEUs
Institute for Space and Defense Electronics
Vanderbilt Engineering
“Unconventional” SEE Mechanisms in Technology Nodes
Below 100nm:
 Charge Sharing
• Distributed effects
 Secondary (nuclear) Reactions
• Probabilistic effects
 Track Size as Large as the Critical Region
• Spatial effects
 Circuit Speed on the order of Collection Dynamics
• Temporal effects
MURI 05 Kickoff
10 May 2005
3
Scaling and SEUs
Institute for Space and Defense Electronics
Vanderbilt Engineering
“Unconventional” SEE Mechanisms in Technology Nodes
Below 100nm:
 Charge Sharing
• Distributed effects
 Secondary (nuclear) Reactions
• Probabilistic effects
 Track Size as Large as the Critical Region
• Spatial effects
 Circuit Speed on the order of Collection Dynamics
• Temporal effects
MURI 05 Kickoff
10 May 2005
4
Charge Sharing *
Institute for Space and Defense Electronics
Vanderbilt Engineering
Issue: At the 250nm CMOS technology node, we have observed
two “unusual” effects


Heavy Ion – Very low LETth
Laser (2) – N-Well Edge
NFETs
*
McMorrow et al, ”Single-Event Upset in Flip-Chip SRAM Induced by ThroughWafer, Two-Photon Absorption”, accepted to the 2005 NSREC
Warren et al., “The Contribution of Nuclear Reactions to Single Event Upset
Cross-Section Measurements in a High-Density SEU Hardened SRAM
Technology”, accepted to the 2005 NSREC
MURI 05 Kickoff
10 May 2005
PFETs
5
Charge Sharing
Background *
Institute for Space and Defense Electronics
Vanderbilt Engineering
Distributed-storage hardened
SRAM cell:
 “Single Node” events
cannot produce an upset
 Charge collection required
at nodes on both legs
 Event sequence
H
L
• Sufficient charge must be
collected on one leg to
float the opposite leg
• The floating leg must then
collected enough charge
to lose its state
*
Warren et al., “The Contribution of Nuclear Reactions to Single Event Upset
Cross-Section Measurements in a High-Density SEU Hardened SRAM
Technology”, accepted to the 2005 NSREC
MURI 05 Kickoff
10 May 2005
6
Charge Sharing
Simulations *
Institute for Space and Defense Electronics
Vanderbilt Engineering
 Mixed-mode simulations
(3-D TCAD + compact model)
useful for efficiency
 We constructed calibrated
base structure for SEE
mapping
 VAMPIRE simulations:
4G Opteron nodes
550,000 elements, 1.5 wks per
TCAD
* Olson et al, “Simultaneous SE Charge Sharing and Parasitic Bipolar Conduction in
a Highly-Scaled SRAM Design,“ accepted to NSREC 2005
MURI 05 Kickoff
10 May 2005
SPICE
7
Charge Sharing
Preliminary Findings *
Institute for Space and Defense Electronics
Vanderbilt Engineering
 Functionally disjoint, but
H
physically adjacent nodes
share charge
 In addition, parasitic
conduction affects PFET of
opposite rail
M7
M8
L
M2
M0
M9
M4
M10
M5
* Olson et al, “Enhanced Single-Event Charge Collection in Submicron
CMOS Due to a Diffusion-Triggered Parasitic LPNP” presented
at HEART 05
MURI 05 Kickoff
10 May 2005
8
Scaling and SEUs
Institute for Space and Defense Electronics
Vanderbilt Engineering
“Unconventional” SEE Mechanisms in Technology Nodes
Below 100nm:
 Charge Sharing
• Distributed effects
 Secondary (nuclear) Reactions
• Probabilistic effects
 Track Size as Large as the Critical Region
• Spatial effects
 Circuit Speed on the order of Collection Dynamics
• Temporal effects
MURI 05 Kickoff
10 May 2005
9
Secondary (Nuclear) Effects
Institute for Space and Defense Electronics
Vanderbilt Engineering
Issue:
 Unable to explain upsets at very low LET in TCAD
 Upset cross section too small to be intra-cell variation
Quantum 4M SRAM SEU Data
TCAD Prediction
Sample 581 Versus Operating Conditions
1.0E-07
TCAD with
average LET
ion does not
explain low
LET upsets
Upset Xsection (cm2/bit)
1.0E-08
1.0E-09
Approximate Ion
Track Area (intra-cell
Limit)
1.0E-10
1.0E-11
1.0E-12
1.0E-13
0
10
20
Ultra Low s
(follow-on at BNL)
No latchup after 1E7 particles @ LETeff
of 120 MeV/mg/cm2. (3.6V, 125 C)
MURI 05 Kickoff
30
40
50
60
70
80
LETeff (MeV/mg/cm2)
Dyn, 3.14V, 125C
Static, 3.14V, 125C
10 May 2005
90
100
110
120
130
Brookhaven Labs, 3/27/02
rdb
Dyn, 3.3V, 32C
10
Secondary (Nuclear) Effects
Hypotheses
Institute for Space and Defense Electronics
Vanderbilt Engineering
 Nuclear reaction events can increase the effective LET to +/- 8x
the incident LET for this simulation
 In scaled technologies, the low probability of occurrence offset
by:
- high number of sensitive volumes (4 Mbit SRAM)
- elimination of lower-LET sensitivity via hardening
-Electrons
MURI 05 Kickoff
LET
10 May 2005
Nuclear
Reaction
Events
523 MeV Neon
LET = 1.79 MeV/mg/cm2
11
Secondary (Nuclear) Effects
Preliminary Analysis *
Institute for Space and Defense Electronics
Vanderbilt Engineering
MRED used for preliminary investigation of potential event
sources:
 Conceptually simple relative energy deposition experiment
 1x108 Monte-Carlo type simulations
 Reactions in the interconnect materials result in charge
deposition from secondary species in the sensitive volume
Oxide Only
Oxide and Metallization
W
*
Warren et al., “The Contribution of Nuclear Reactions to Single Event Upset Cross-Section Measurements in a High-Density
SEU Hardened SRAM Technology” accepted to the 2005 NSREC
MURI 05 Kickoff
10 May 2005
12
Secondary (Nuclear) Effects
Preliminary MRED Results
Institute for Space and Defense Electronics
Vanderbilt Engineering
*Nuclear events only (in 1x108 simulations)
1.E+05
Primary LET
High Energy
Events
# of Events
1.E+04
1.E+03
SiO2/Si
1.E+02
W/Ti/TiN/SiO2/Si
1.E+01
1.E+00
1.00E+02
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
Energy (eV)
 Nuclear Reactions can produce products greater than 8x in
energy deposition than the primary LET
 Effect is exacerbated with inter-connect materials (tungsten,
aluminum, etc…)
MURI 05 Kickoff
10 May 2005
13
Secondary (Nuclear) Effects
On-Going Work
Institute for Space and Defense Electronics
Vanderbilt Engineering
 Integration of multiple tools
• Layout
• TCAD modules
• MRED (nuclear)
 Accurate spatial
relationship between
sensitive regions and
materials
 Interconnect material
affects the probability of
extreme value events
 Events must occur near the
sensitive region to upset the
cell
 A step closer to developing
a predictive tool
MURI 05 Kickoff
10 May 2005
14
Scaling and SEUs
Institute for Space and Defense Electronics
Vanderbilt Engineering
“Unconventional” SEE Mechanisms in Technology Nodes
Below 100nm:
 Charge Sharing
• Distributed effects
 Secondary (nuclear) Reactions
• Probabilistic effects
 Track Size as Large as the Critical Region
• Spatial effects
 Circuit Speed on the order of Collection Dynamics
• Temporal effects
MURI 05 Kickoff
10 May 2005
15
Track Size
Institute for Space and Defense Electronics
Vanderbilt Engineering
Issue:
 Device dimensions becoming
smaller than
• Commonly assumed radial
dimensions for SE charge
generation tracks
• Minority carrier diffusion
lengths (even in Drain/Source
regions)
(1)
 Device topology has
implications for charge
deposition
(1)
(2)
(2)
M. L. Alles, et. al. , “Considerations for Single Event Effects in Non-Planar Multi-Gate SOI FETs”, submitted for
presentation at the 2005 IEEE International SOI Conference.
R. Chau et. al., “Advanced Depleted-Substrate Transistors: Single-Gate, Double-Gate, Tri-Gate”, 2002 International
Conference on Solid State Devices and Materials (SSDM 2002), Nagoya, Japan.
MURI 05 Kickoff
10 May 2005
16
Track Size
Institute for Space and Defense Electronics
Vanderbilt Engineering
Preliminary Simulations:
 Simulation of an ion hit using 3D TCAD
• (Alpha Particle, LET=2.4 MeV/mg-cm2)
• Normal incidence
• Three locations (Body, Drain, Source)
 Response shows much longer time profile vs. direct collection;
• Potential profile indicates bipolar action
Hit to Channel
M. L. Alles, et. al. , “Considerations for Single Event Effects in Non-Planar Multi-Gate SOI FETs”, submitted for
presentation at the 2005 IEEE International SOI Conference. Simulated Tri-gate device based on: J Choi et. al.,
IEDM Technical Digest, 647 (2004).
MURI 05 Kickoff
10 May 2005
17
Track Size
Institute for Space and Defense Electronics
Vanderbilt Engineering
Preliminary Findings:
 Hits to Drain (and Source) can
lead to notable charge collection
• Implication for sensitive area
• Drain and source contribution
found to depend on contact
placement and size
 Energy deposition process in
such small devices unclear
• Convention LET concept may
break down
• We will study this with detailed
simulations (MRED)
• Accurate TCAD modeling
(FLOODS)
M. L. Alles, et. al. , “Considerations for Single Event Effects in Non-Planar Multi-Gate SOI FETs”, submitted for
presentation at the 2005 IEEE International SOI Conference.
MURI 05 Kickoff
10 May 2005
18
Scaling and SEUs
Institute for Space and Defense Electronics
Vanderbilt Engineering
“Unconventional” SEE Mechanisms in Technology Nodes
Below 100nm:
 Charge Sharing
• Distributed effects
 Secondary (nuclear) Reactions
• Probabilistic effects
 Track Size as Large as the Critical Region
• Spatial effects
 Circuit Speed on the order of Collection Dynamics
• Temporal effects
MURI 05 Kickoff
10 May 2005
19
Circuit Speed
Institute for Space and Defense Electronics
Vanderbilt Engineering
Issue:
 GHz circuitry have response dynamics on the same order




as charge collection transient profiles
In these cases, SE transients are indistinguishable from
legitimate signals
Some hardening techniques are ineffective
Dynamic modeling required
Detailed SE charge collection profiles are needed
MURI 05 Kickoff
10 May 2005
20
Conclusions
Institute for Space and Defense Electronics
Vanderbilt Engineering
An understanding of SEE in emerging, scaled technologies
(SiGe, ultra-small CMOS and 3-d SOI) constructed with
novel materials systems (strained Si, alternative dielectrics,
new metallizations) requires:
 Improved physical models for ion interaction with materials




other than Si
Detailed SE track structure models
Improved understanding of nuclear reaction effects
Improved device physics for mixed-mode modeling
Improved understanding of parasitic charge collection
MURI 05 Kickoff
10 May 2005
21