metal gate and junction technologies for system leading
TRANSCRIPT
1K.Shibahara
Metal Gate and Junction Technologies for Leading Edge Devices
Kentaro Shibahara
Research Center for Nanodevices and Systems, Hiroshima Univ.
2K.Shibahara
3DCSSSystem
multi-chip information processing with wireless Interconnect
Tera-bit processing high RF performance devices
ggd
T
RCff
8max
gs
mT C
gf
2Parasitic series resistance should be low
Low resistive junction formationLaser Annealing
Gate resistance should be lowMetal gate
3K.Shibahara
Looking back the following 5 years activities for our COE Pj.
Metal Gate Technology
Shallow Junction Formation
4K.Shibahara
Recover EOT loss due to gate depletion.
Ion vs Jg trade-off
Lower resistivityRF application fmax
However, worfunctionappropriate to CMOS devices are needed.
Why metal gate?: Basic expectation
~ 0.4 nm (gate dope : 1 1020 cm-3)
5K.Shibahara
36404552596878901Metal ½ Pitch
.31.32.32.32.33.63.63.63EOT
1.31.41.41.51.61.92.02.1EOTLSTP
.32.32.32.32.32.64.65.65EOT
.9.9.91.01.11.21.31.4EOTLOP
.25.25.27.28.29.74.74.73EOT
.5.5.65.75.91.11.11.2EOTHP
20122011201020092008200720062005
Based on ITRS 2005 PIDS Table: Planar Bulk DevicesIntroduction Timing of Metal Gate
Shifted at 2006 update
Band edge workfunction is hard to obtain.6K.Shibahara
Tunable Workfunction: Single metal gate CMOS How?
by Pileup formation• Gate first process
Mo + N
• FUSI (Fully Silicided) gateNiSi and Pd2Si + P, As, Sb, B……
Our Target
7K.Shibahara
High density impurity pileup at metal/insulator interface
Dipole formation and Workfunction shift
Metal SiO2
X O SiO
O
_ +
Electron negativity difference Dipole formation
Workfunction Tuning by Pileup Formation
VG>0
Si-sub.
-----
+++++
SiO2Metal
Electric Dipole
oxFB tV /:Charge density
:Distancet
8K.Shibahara
P-18Workfunction Tuning of NiSi and Pd2Si Fully-Silicided Gates by PredopingT. Hosoi, K. Sano, M. Hino, and K. Shibahara
Evaluation of Chemical Structures and Work Function of NiSinear the Interface between NiSi and SiO2H. Murakami, H. Yoshinaga, D. Azuma, A. Ohta, Y. Munetaka, S. Higashi, S. Miyazaki, T. Aoyama, K. Hosaka, and K. Shibahara
Related PresentationFUSI
First, Mo results, then FUSI briefly
9K.Shibahara
Categollization: Processes for Metal/Metal Compound
Motolora+STM IEDM03
10K.Shibahara
Mo Fine patterning
Easy Fine PatterningHigh Selectivity against Oxide
Oxide 2 nm
Suitable for Gate First ProcessIn addition, workfunction tunable
11K.Shibahara
Nitrogen Solid-Phase Diffusion into Mo Gatep-Si (100)
Gate oxidation (5 nm)
Mo & TiN sputter (50 & 30 nm)
Nitrogen Solid-Phase Diffusion(800 C, 1min)
TiN removal
S/D implantation
Al wiring and PMA
LOCOS formation
( As : 5x1015 cm-2, 30 keV )
Gate formation
S/D activation annealing(900 C, 1min)
diode
MOSFET
Mo~ EV
Formation of N pileupMo~ -0.46eV
12K.Shibahara
Nitrogen Solid-Phase Diffusion into Mo Gate
10-14
10-12
10-10
10-8
10-6
10-4
0 1 2 3
w/o N-SPDMOSFETexpected (diode)
I d (A
)
Vg (V)
L/W : 1/10 mTox : 5 nm
Vd : 0.05 V
diodeMOSFET -0.1 eV
-0.46 eVMo
13K.Shibahara
Nitrogen Redistribution by S/D Activation Annealing
1019
1020
1021
0 10 20 30 40
Nitr
ogen
Con
cent
ratio
n (c
m-3
)
Relative Depth (nm)
Mo SiO2 Si
diode MOSFET
Reduction of N concentration in bulk Mo (N out-diffusion)
Mo-gate MOSFET
back-side SIMS
Modified Process Oxide Cover & TiN Cap
Reversible workfunction shift(-0.46 eV -0.1 eV)
Pileup reduction &
14K.Shibahara
Vth Dependence on Gate Length
1
1.2
1.4
1.6
1.8
2
0 5 10 15 20
V th (V
)
Lg ( m)
w/o N-SPD
oxide-cover
TiN-cap
Vth = -0.5V
Oxide cover: Vth ~0.15 VTiN Cap: Vth decreased as Lg decreased
Wg : 10 mTox : 5 nm
15K.Shibahara
Nitrogen Redistribution in Mo Gate (oxide-cover)
just after N-SPD
after S/D activation annealing
N pileup formation at Mo/SiO2interface
N pileup at bottom Mo/SiO2Interface redistributes to top and two side interfaces.
small Vth shift16K.Shibahara
Mechanism of Vth shift decrease (TiN-cap)
Nitrogen atoms distributed to bottom interface reduces as Lg reduced.
N pileups are also formed at side interfaces of Mo.
17K.Shibahara
Thus, combining Mo+N process to MOSFET fabrication process maintaining expected workfucntion value is a tough issue.
How about other metals?
Control of Workfunction
18K.Shibahara
04EDL Vol25 p337 C.Ren et al. Thermal Stability of Metal Workfunction
Important Issue for Gate First Devices.Integration scheme of dual metal is also ….
19K.Shibahara
Fabrication Flow for FUSI Gate
20K.Shibahara
FUSI Gate MOS Diode Fabrication with Predoping
p-Si(100)Gate oxidationPoly-Si deposition
P, As, Sb, B …. Predoping by Ion Implantation
Ni or Pd sputter depositionFull-Silicidation
Gate patterning
Activation annealing
Impurities
Impurities Solubility: Silicide << poly-SiImpurity CondensationPileup formation at Oxide Interface
21K.Shibahara
Workfunction Tunable Range of FUSI Gate
• Pd2Si that can be processed at lower temperature provided wide workfunction tunable range.• What is the origin of different results for NiSi?
22K.Shibahara
Characterization of NiSi FUSI MOS Interface
Source: Dr. Murakami’s presentationSpecimen: Made in Fujitsu
23K.Shibahara
Sb-O signal was disappeared after Ar+ sputteringnecessary for workfunction evaluation.
Oxidized Sb in SiO2 is essential for NiSi workfunction shift.Specimen: Made in Hiroshima Univ.
Quite different conclusion. Why?
Another Characterization of NiSi FUSI MOS Interface
24K.Shibahara
Influcence of Process Condition on Pileup Formation
450 C : M = -0.32eV500 C : no M shift
Source: Dr Hosoi’s Poster
Pileup peak concentration : 450 C > 500 CFUSI Workfunction Tuning: Sensitive to process condition
25K.Shibahara
Hot-Plate 250oC
Void
Needle Growth
metal
poly-SiSiO2
Silicide
Stage Heating
HighPd
LowSi
Silicidation TemperatureDiffusion Specie
Hot-Plate 300oC Undoped,P,Sb
UndopedP,As,Sb
As
Undoped, P,As,Sb
Heating Apparatus
Influence of Temperature and Doping: Pd2Si
Slow Heat up
Silicidation is sensitive to temperature and doping
26K.Shibahara
Jan. 27, 2007http://www-03.ibm.com/press/us/en/pressrelease/20980.wss
IBM Advancement to Spawn New Generation of Chips: “First fundamental change to basic transistor in forty years”
http://www.intel.com/pressroom/archive/releases/20070128comp.htm
Intel Transistor Technology Breakthrough Represents Biggest Change to Computer Chips In 40 Years
Press Release From IBM and Intel: Metal+high-k in 2008
Intel’s new gate stack???
Even if they really ship products in 2008, research on metal gate should be continued. We have not understand well nature of the new gate stack yet.
27K.Shibahara
Metal Gate Technology
Shallow Junction Formation
28K.Shibahara
Optimized Spike Annealing
(sub-seconds RTA by halogen lamp heating)
Ohuchi et al., 2002 IEDM(Toshiba)
Shallow Junction: Key Technology for MOSFET Scaling
"Diffusion Less"
"Diffusion Less" is the key word that represents a demand for post-spike annealing technologies.
We need this range!
Xj (Junction Depth) increase by diffusion
29K.Shibahara
Low temperature & Long durationSolid phase regrowth: hr-min
High temperature & Short durationFlash lamp: ms – sCW Laser Annealing: ms – sPulse Laser Annealing: ns
Candidates of Post-Spike Annealing
What is the best scheme for pulse laser?Our Target
LSA
30K.Shibahara
Summary of our activities on LA (Laser Annealing)
KrF Excimer Laser
HALA:Heat-assited LA
PMLA: Partial Melt LA
Solid State Green Laser
LA with light absorbing layer
Analysis of LA with phase switch
31K.Shibahara
Merits of Excimer Laser
Penetration DepthGreen (532 nm): 927 nmKrF Excimer Laser 5.5 nm
LA KrF Eximer Laser ( : 248 nm)Pulse Width: 38 ns
200 W Laser(1 J/pulse, 200 Hz)
32K.Shibahara
Schemes for Conventional Laser Annealing
a-Si
c-Si
liq.-Si
c-Si
c-SiFast re-crystallizationnon-equiribrium
Activation
solid-phaseepitaxy
c-SiMelt
non-Melt
Laser Irradiation= Heating Up
IonImaplantation
Cooling Down
Melting pointa-Si: 1100oC, c-Si: 1400oC
33K.Shibahara
HALA & PMLA Time Sequence
Substrate Temperature (Tsub): 250 – 525oC• Laser Energy Density (EL) : 200 – 600 mJ/cm2
• FWHM of Laser Pulse: 38 ns, Pulse Number: 1 Pulse
Temperature Profile
10 5 10
Tsub
Time [min]
Laser Irradiations (EL)Heating Up Cooling Down
34K.Shibahara
102
103
104
200 300 400 500 600 700
450oC
250oC
RT
She
et R
esis
tanc
e R
s [sq
.]
Laser Energy Density EL [mJ/cm2]
Tsub
Xj=21nm Xj>21nmA B C
Region B(300-400 mJ/cm2)Good activationwithout increasing Xj
Region Anonmelt
Region CXj > as Implantedbecause of c-Si melt
Sheet Resistance and Energy Density for HALA
TSUB: 450oC
HALA: Reduction of EL
35K.Shibahara
15352510.2450
RegrowthThickness
[nm]
RegrowthRate
[nm/min]
Tsub
[oC]
a-Si thickness ~ 11 nm
450oC: Partial Melt
525oC: non-Melt
SIMS Profile: 300 mJ/cm2
0 10 200
0.2
0.4
0.6
0.8
1
450oC525oC
Depth [nm]
SbCo
ncen
tratio
n[x
1021
cm
-3] EL: 300mJ/cm2SiO2 Si
36K.Shibahara
Basic Ideas of Schemes
PMLA: melt(Good activation) + non-melt(Diffusion less)
37K.Shibahara
Xj~ta, tEOR
melt LA
ta: Amorphizedlayer thickness
tEOR:Extendeddefects depth
tatEOR
SPE
Xj
PMLA
PMLA: Amorphization depth is free from junction depth
Advantage of PMLA
38K.Shibahara
XTEM before Laser Irradiation
39K.Shibahara
1017
1018
1019
1020
1021
-2 0 2 4 6 8 10 12 14XJ: 10.5 nm
2.5 nm/dec
No irradiation360 mJ/cm2
600 /sq.
Depth [nm]
Sb
Con
cent
ratio
n [c
m-3
]
10 nm Junction Formation with Sb and B
40K.Shibahara
Conclusions
• Achievements on metal gate and LA obtained during our COE Pj. term were reviewd.
• Metal gate and LA technologies are still exotic ones for Si LSIs. However, introduction to products is approaching.
• We will continue research to support industries by supplying better understandings on process, materials and so on related to these fields.
41K.Shibahara
Acknolowdgements
This work was a result of effort of many students and researchers at RCNS and adsm of Hiroshima Univ.