Negative Bias Temperature Instability NBTI in pMOSFETs: Characterization, Modeling

1
Negative Bias Temperature Instability (NBTI) in
p-MOSFETs Characterization, Modeling Material
Dependence
Souvik Mahapatra Department of Electrical
Engineering IIT Bombay, Mumbai, India Presently
(June-Nov 2006) Applied Materials, Santa Clara,
USA
2
Acknowledgement
IIT Bombay D. Varghese, G. Gupta, L. Madhav, D.
Saha J. Vasi Purdue University M. A. Alam A.
E. Islam Applied Materials K. Ahmed F. Nouri
(Financial support) Renesas Technologies,
Tokyo E. Murakami H. Aono (Financial
support) Lucent Technologies / Agere
Systems Steven Hillenius
3
Outline
Introduction background
Impact of measurement delay, time dependence
High voltage impact on time dependence
Reaction Diffusion Model, Temperature dependence
Bias N2 process dependence
NBTI recovery
Conclusion
4
What is NBTI?
Issue for p-MOSFET in inversion Gate negative to
Source, Drain Bulk, aggravated at more negative
bias
Device parameters (VT, Gm, IDLIN, IDSAT, etc.)
shift with time Instability More pronounced at
higher bias and temperature
Aggravated for SiON films Most serious
reliability concern
5
Device degradation after NBTI stress
Reduction in ID
Positive charge generation
Yamamoto, TED 1999
6
Motivation
Choice of stress conditions Accurate measurement
of experimental data
Stress
DVT
time
7
NBTI Issue Impact of Nitrogen
Increased NBTI with higher N2 content
Mitani, IEDM 2002
8
NBTI Recovery
NBTI recovers after stress
Chen, IRPS 2003
9
Why do we need to understand NBTI?
Accurate lifetime projection for each technology
node
Proper choice of burn-in and stress (WLR)
conditions
Develop TCAD and SPICE models for predictive
simulation
Identify process parameters for control and
mitigation
10
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
11
Impact of Measurement Delay
Delay recovery, higher fraction at lower time
12
Impact of Measurement Time
IDLIN measurements More recovery higher slope
for larger measurement delay
Varghese, IEDM 2005
13
Impact of Measurement Bias
Higher recovery higher slope for lower
(absolute) measurement bias
Varghese, IEDM 2005
14
Uncertain Power Law Exponents
Higher slope at higher T
Varghese, IEDM 2005
15
Charge Pumping Measurements
Gate pulsed from Inversion to Accumulation
Electron-hole recombination at interface DC
current
Direct measure of interface traps
Delay inherent, higher slope with higher delay
time, T
Varghese, IEDM 2005
CP Groeseneken, TED 1984
16
CP IDLIN Comparison (Slope)
IDLIN
Higher recovery higher slope for larger delay,
higher T
Varghese, TED (submitted, 2006)
17
No Delay Measurements (DC On-the-fly IDLIN)
Universal power law time dependence with n
0.14-0.16
Stress VG and T
DC On-the-fly Rangan, IEDM 2003
Varghese, IEDM 2005
18
Ultra Fast (ms delay) VT Measurements
Oxynitride, EOT2nm Long time slope 1/6
Reisinger, IRPS 2006
19
Long Time Bake Experiments
Plasma N2, EOT1.2nm
Slope n 1/6 No saturation
Chen, IRPS 2005
20
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
21
Power Law Slope Voltage Dependence
Plasma N2, EOT1.2nm Increased power-law slope at
higher VG
Chen, IRPS 2005
22
Bias dependent physical processes
Low VG, VB0 HH absent
Mahapatra, IEDM 2002
23
Enhanced degradation at high VG
Low VG Single (lower) degradation rate High
VG Increased degradation rate (long time)
Mahapatra, IEDM 2002
24
High VG SILC
Bulk trap assisted tunneling Increased current
tracks trap generation
Takagi, IEDM 1999 Ghetti, TED 2000 Alam, TED
2002
25
Enhanced Degradation at High VB
Low VB Single (lower) degradation rate High
VB Increased degradation rate (longer time)
Mahapatra, IEDM 2002
26
Hot Hole Effect
Enhanced degradation correlates well with Quantum
Yield (ISD/IG) of hot hole generation
Mahapatra, IEDM 2002
27
Hot Holes Impact on Interface Traps
VB0 Negligible HH, lower rate, no SILC VBgt0 HH
present, higher rate, SILC, additional amount
show similar rate as SILC
Charge pumping measurement
Varghese, EDL 2005
28
Different types of traps
29
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
30
Reaction Diffusion Model (Interface Traps)
Poly
Reaction Si-H bond breaks into Si and H
Precursor
Jeppson, JAP 1977 Alam, IEDM 2003
31
Universal Scaling Scheme T Dependence
Assume similar EA for kR and kF T dependence
through D
R-D model solution DNIT kF.NO/kRm (Dt)n
EA (NBTI) EA (D) n
Alam, IWGI 2001 Mahapatra, IEDM 2004 Alam, IRPS
(T) 2005
32
Universal T Activation of Diffusion
X-axis scaling
Identical EA for DPNO Control
Identical EA for Idlin C-P measurements
EA(D) consistent with power law slope (n) from
R-D model
Varghese, IEDM 2005
P1 1.2nm (14), P2 1.2nm (21) P3 1.7nm (28),
P4 2.2nm (29)
33
Temperature Scaling Scheme
EOT1.2nm, T1(11), T2(17), P1(14), P2(21)
Varghese, IEDM 2005 Gupta, IRPS 2006
34
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
35
Dependence on Gate Leakage / Stress VG
No correlation with IG
Mahapatra, IEDM 2004
36
Dependence on Inversion Hole Density
Good correlation
Mahapatra, IEDM 2004
37
Voltage / Field Dependence
Field (chemical) not Voltage (energy) driven
Mahapatra, IEDM 2004
38
NBTI Bias Dependence
Hole-assisted thermal dissociation of Si-H bonds
Bias dependence Hole density, Tunneling
probability, Capture cross section
N2 near interface enhance Si-H dissociation
mechanism
39
Bias Dependence Effect of N2 Dose, Type
R-D solution (H2) DVT A Eox2/3 exp(BEOX)
Reduced reaction barrier for higher N2 Lower
EOX dependent slope
Lower field dependence for TNO, higher N2 density
at Si-SiO2 interface
Separately plotted for clarity, does not reflect
difference in actual magnitude between PNO and TNO
Gupta, IRPS 2006
40
Field Dependence of IDLIN and CP
Identical EOX dependent slope for DVT (IDLIN) and
DNIT (CP)
Varghese, TED (Submitted, 2006)
41
Impact of Nitridation Method
Strong dependence on method of N2 incorporation
NISS N2O RT-N2O RPN
Liu, IEDM 2001
42
Impact of Nitrogen Profile
Higher N2 density at Si-SiO2 interface More NBTI
Sasaki, EDL 2003
43
Poor Films (Hole Trapping), Good Films
Reduced hole trapping for properly optimized
films Reduced DVT magnitude, increased slope (n)
Huard, IRPS 2006
44
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
45
Fast Slow Recovery Components
Rapid initial recovery followed by a slower
component
Gupta, IRPS 2006
46
Fast Slow Recovery Mechanics
Fast recovery due to hydrogen diffusion in oxide
Slow recovery due to hydrogen diffusion
back from poly
FR (fast) 2TOX / (DPOLY tSTRESS)½
Krishnan, IEDM 2005
47
Fast (Fractional) Recovery Signatures
Larger FR for thicker EOT
Reduced FR for higher stress time
Gupta, IRPS 2006
48
CP IDLIN Comparison (Recovery)
Identical T dependent generation vs. recovery
correlation
Varghese, TED (Submitted, 2006)
49
Outline
Introduction background Impact of measurement
delay, time dependence High voltage impact on
time dependence Reaction Diffusion Model,
Temperature dependence Bias N2 process
dependence NBTI recovery Conclusion
50
Conclusion
NBTI due to Interface traps R-D model (diffusion
of molecular H2) can explain many observed
features
No conclusive evidence for hole trapping for
properly optimized ultrathin oxynitride films
Must eliminate measurement delay for proper
estimation of degradation magnitude power-law
time exponent
Careful choice of stress bias to avoid unwanted
bulk trap generation