A New Technique to Extract the S/D Series Resistance of Sub-100nm MOSFETs
A New Technique to Extract the S/D Series Resistance of Sub-100nm MOSFETs
Dominique FLEURY Ph.D. Student
STMicroelectronics / IMEP-LAHC lab.
Dominique FLEURY Ph.D. Student
STMicroelectronics / IMEP-LAHC lab.
Tuesday, April 28
Introduction: some definitions Introduction: some definitions
Series resistance: Rsd = Rs + Rd (from the testing pad to the beginning of the channel)
Channel resistance: Rch
Total resistance: Rtot = Rch + Rsd
2009 VLSI-TSA Symposium 3
Introduction: the I
d(V
gs) model Introduction: the I
d(V
gs) model
current factor
returns how many current flows in the channel
gate overdrive
bias above the threshold voltage: Vgt=Vgs–Vth
Mobility reduction factors
- model the mobility reduction due to phonon and surface roughness scatterings
Introduction: the I
d(V
gs) model Introduction: the I
d(V
gs) model
The current gain factor can be extracted (ξξξξ- function[1] / Y-function[2] / Hamer[3]).
It contains Leff and µ0 so it may be used to make an extraction insensitive to these parameters.
low field mobility
effective channel length
2009 VLSI-TSA Symposium 5
Introduction: why measuring R
sd? Introduction: why measuring R
sd?
Essential parameter for performance improvement:
A lower Rsd improves the MOSFET performances
To extract transport parameters:
Effective mobility (ex: split C-V technique) Drift carrier velocity
Ballistic rate
0 10 20 30 40
0.0 0.5 1.0 1.5 2.0
gate bias (V) drain current (µA/µm) no change (Rsd)
Rsd + 100Ohm.µm Rsd + 200Ohm.µm
L = 40nm Vds = 10mV
Impact of Rsd on the linear drain current
Rsd≈ 65Ω.µm ΩΩ
How to extract R
sd? How to extract R
sd?
The Rtot(L) technique[4]
Provides erroneous values when mobility changes from a long to a short channel.
Mobility changes may be due to:
• pocket implants [5]
• strain [6]
• implantation defects close to S/D junctions [7]
2009 VLSI-TSA Symposium 7
How to extract R
sd? How to extract R
sd?
The Y-function technique[2]
Insensitive to mobility variations
High dispersion on Rsd due to the ΘΘΘΘ1111-parameter
0 0.5 1 1.5 2
0 5 10
β β β
β (mA/V²) ΘΘΘΘ1 (V-1 )
Rsd≈≈≈≈115Ω.µm Θ
Θ Θ
Θ1,01,01,01,0 ≈ 0.34 ≈ 0.34 ≈ 0.34 ≈ 0.34V-1 72 dice, 300mm wafer 552 points, R²=0.62
app βΘΘ ⋅+= sdapp R11
How to extract R
sd? How to extract R
sd?
What can we expect from an extraction methodology in an industrial context ?
RobustnessRobustness
Insensitivity towards µeff/Leff variations, no technology
dependence
Fast measurement Fast measurement capability capability
Only requires Id(Vgs) measurements
Accuracy Accuracy
Accuracy is limited by
dispersion in the Θ1 parameter
Can be improved
Easy portability Easy portability
No computation algorithms simple test setup
2009 VLSI-TSA Symposium 9
The new technique: R
tot- ββββ The new technique: R
tot- ββββ
Rtot can be computed as a function of 1/ββββ
When Rtot is plotted versus 1/ββββ, Rsd can be read at the intercept with the y-axis.
assumed constant for a given Vgt
parameter result
(Vgs)
The new technique: R
tot- ββββ The new technique: R
tot- ββββ
Rtot(1/ββββ) linear regression returns Rsd for the technology (at a given Vgt)
Rsd is read at the intercept (at the given Vgt)
The mobility attenuation at the given Vgt is read from the slope
dispersion is very low and accuracy is improved.
2009 VLSI-TSA Symposium 11
The new technique: R
tot- ββββ The new technique: R
tot- ββββ
Rsd can be extracted as a function of Vgs
approximation: Vgs ≈≈≈≈ Vgt + <Vth>
consistent with Rov variations of the literature[8]
Rsd(Vgs) trend is extracted.
data from [8]
The new technique: R
tot- ββββ The new technique: R
tot- ββββ
Mobility attenuation is extracted from the slope, for the given technology
2009 VLSI-TSA Symposium 13
The new technique: robustness The new technique: robustness
Results are consistent with the Y-fuction extraction
Accuracy is improved !
High dispersion using the Y-
function technique
The new technique: accuracy The new technique: accuracy
Extracted values of Rsd follow the change if an additional resistance ∆∆∆∆Rsd is added
y = 1.8929x + 164.12
y = 1.8935x + 65.229
y = 1.886x + 261.19
500 1000 1500 2000 2500 3000
Rtot(Ω.µm)
no change (Rsd) Rsd + 100Ohm.µm Rsd + 200Ohm.µm
2009 VLSI-TSA Symposium 15
The new technique: discussion The new technique: discussion
Vth may change as a function of L
short channel effects (SCE) and S/D halos (RSCE)
Vgs ≈ Vgt + <Vth>
Rtot = Rch(Vgt) + Rsd(Vgs)
accuracy on Rsd(Vgs)
?
The new technique: discussion The new technique: discussion
Is the Vgs = Vgt + <Vth> approximation still valid for bulk devices ? YES
<Vth(L)> approximation still enable extraction within a 3% accuracy
2009 VLSI-TSA Symposium 17
Conclusion Conclusion
Is the Rtot-ββββ suitable to an industrial context ?
RobustnessRobustness
good agreements on various technologies (bulk, FDSOI, SON)
Fast measurement Fast measurement capability capability
require only Id(Vgs) measurements, can be deployed in automatic testing
AccuracyAccuracy
improved compare to the other methodologies (Y-function, Rtot- L). Up to 3%.
Easy portabilityEasy portability
No specific computation algorithms. No specific test setup.
Rsd extraction
References References
[1] D. Fleury et al., ICMTS, 2008
[2] G. Ghibaudo et al., Electronics Letters, 24, 9, 1988
[3] M. F. Hamer, IEE Proceedings, 133, 2, 1986
[4] Y. Taur et al., EDL, 13, 5, 1992
[5] K. M. Cao et al., IEDM, 1999
[6] F. Andrieu et al., EDL, 26, 10, 2005
[7] A. Cros et al., IEDM, 2006
[8] S.-D. Kim et al., TED, 49, 3, 2002
Thank you for your attention !
