Substrate maturity and readiness in large volume to support mass adoption of ULP FDSOI platforms
Christophe MalevilleChristophe Maleville
January 21st, 2016SOI Consortium Conference – Tokyo 2016
Substrate readiness – 3 lenses view
C1 - Restricted 2January 21st, 2016SOI Consortium Conference – Tokyo 2016
SOI maturity curve
2005PC
2015Mobile
• NXP/Philips reliability thruthick isolation
• IBM and AMD leveragedPerformanceBoost of SOI
1992Satellite
Boost of SOI
• RFSOI bringsPerformance/costbreakthru
• FDSOI enableslow power devices
• ~2 millions wafer/year HVM
3January 21st, 2016SOI Consortium Conference – Tokyo 2016
4January 21st, 2016SOI Consortium Conference – Tokyo 2016
Planar FDSOI: Undoped Channel
Thin Silicon Channel
0.6
Thre
shold
Voltage (
mV)
V
(V)
LG=25nm - multiple sources contributing to VT variation
• FDSOI Substrate: Thin Si & Thin buried oxide
• Ultra thin BOX option �Back bias control
• Silicon thickness variations contribution to VT variations - VT sensitivity ~ 25mV/nm
• “All transistors should have Si thickness within ±5Å”
5January 21st, 2016SOI Consortium Conference – Tokyo 2016
6.5 7.0 7.5 8.0 8.50.3
0.4
0.5
VTsat
Thre
shold
Voltage (
mV)
Silicon Thickness (nm)
VTlin(V
)
Challenges and Opportunities of Extremely Thin SOI (ETSOI) CMOS Technology, A. Khakifirooz, VLSI-TSA 2010
Source: J. Hartmann, GSA Apr. 2012
FD-SOI: long term collaboration
ResearchInstitute
Advanced R&D
IndustrialisationDonor substrate
2005 20082005 2010
• Key achievements
Smart Cut & device experts
2005
FDSOI ∆Vt Yield BBias• Key achievements FDSOI ∆Vt Yield BBias
● ●
● ●
• Advanced crystal for 6nm active layer
• Wafering quality for thin Box bonding
● ● ●• ±5 Å Thickness uniformity
• All spatial frequency thickness control ● ●
• HVM ready ● ● ● ●
6January 21st, 2016SOI Consortium Conference – Tokyo 2016
Smart-Cut 1.0 – On-wafer uniformity
• Oxide uniformity • Post-split Silicon uniformity
Box WiW thickness post oxidation (Å) Post split WiW Si uniformity (Å)
0 5 10 15 20
7January 21st, 2016SOI Consortium Conference – Tokyo 2016
0 5 10 15 20
FD-SOI Uniformity – Bandwith for improvements
• Current ellipsometry accurately describes FD-SOI wafer
95% of wafer surface compliant @ +/-4A
90% of wafer surface compliant @ +/-3A
8January 21st, 2016SOI Consortium Conference – Tokyo 2016
SmartCut 2.0 for FDSOI
Implant
Cleaning
Start Material
OxTop
Adapted to Oxide Film Thickness
Splitting
Final sorting
Finishing
Bonding
Adapted to SOI Film Thickness
9January 21st, 2016SOI Consortium Conference – Tokyo 2016
Smart-Cut 2.0
• Ultrathin Box Capability • Uniformity vs roughness
rela
ted
defe
cts
FDSOIPDSOI
Bon
ding
rela
ted
10January 21st, 2016SOI Consortium Conference – Tokyo 2016
Thermal smoothing principle
• Silicon surface smoothing at high temperature (RTA, BA)• Material transport mechanism
�Bulk diffusion
�Evaporation / Condensation
�Surface diffusion�Surface diffusion
Silicon
Gas
Bulk diffusion
Evaporation / condensation
Surface diffusion
Reaction withcontaminant
F.De Crecy – CEA/LETI
Simulation of silicon smoothing under high temp anneal
11January 21st, 2016SOI Consortium Conference – Tokyo 2016
Layer Total Thickness Variation (Å)
“Conventional” control plan – Thickness + Roughness
BoxHandle
SOI
µm-1
10-6 10-2 1WtW
Ellipso
WiW
EllipsoDRM
0AFM
Performance +/-5 Å 6-10 Å P-V 0.8Å RMS
Evidence
12January 21st, 2016SOI Consortium Conference – Tokyo 2016
• Reflectivity of a FD-SOI stack (SOI + BOX) is function of wavelengths
• By filtering specific wavelength, reflectivity becomes only sensitive to SOI layer thickness variations
Differential Reflective Microscopy (DRM) for complete SOI thickness monitoring
Microscope calibration
100
105
110
115
120
125
130
135
6000 7000 8000 9000 10000 11000 12000
Grey scale
Elli
pso
thk,
A
ElipsoTheory, 540nm
Grey scale map
Signal treatmentCalibration wafers
Grey scale calibration curve Thickness profile
HSEB Baldur tool
Spatial wavelength range: 0.5µm – 80µm
13January 21st, 2016SOI Consortium Conference – Tokyo 2016
Layer Total Thickness Variation (Å)
All included SOI thickness control – (all wafers, all points, all frequencies)
BoxHandle
SOI
µm-1
10-6 10-2 1WtW
Ellipso
WiW
EllipsoDRM
0AFM
Performance +/-5 Å 6-10 Å P-V 0.8Å RMS
Evidence
14January 21st, 2016SOI Consortium Conference – Tokyo 2016
FD28nm: 30 000 Wafers: ± 1 Atomic Layer !
12nm
Handle
SOI (12nm)
Box (25nm)
Wafer
FDSOIPDSOI
∆SiStack
Micro
Local
15January 21st, 2016SOI Consortium Conference – Tokyo 2016
1 sigma (Å)
Improved thickness control for 22nm node
20 nm
Stack
• Polished Bulk• 20FD• 28FD
∆SiWafer
MicroLocal
16January 21st, 2016SOI Consortium Conference – Tokyo 2016
1 sigma (Å)
17January 21st, 2016SOI Consortium Conference – Tokyo 2016
Soitec – SOI adoption via Partnerships and CollaborationsFrom technology development to manufacturing
2005 2009 2011 2012 2015
RF-SOI 300mm ramp
Soitec RFeSisubstrate ramp
RF-SOI switch mainstream
1st RF-SOI switch
TR-SOI UCL & Soitec IP
RF-
SOI
>50%
FD-S
OI
Advanced R&D
IndustrialPartner
Materials Research28FD
Foundry offer
2005 2008 2010 2014 2015
…
22FDFoundry offer
18January 21st, 2016SOI Consortium Conference – Tokyo 2016
Antenna Switch Module
(SP9T ) Cost
RF-SOI value propositionEnabling best performance, integration and cost-efficiency
Integration
¢
Cost efficientPerformance
4G/LTE-A & beyond
performance enabler
High integration of RF Front-End module
Cost effectivetechnology
0
5
10
15
20
25
30
GaAS BSOS RFSOI
(SP9T ) Cost¢
Cheaper process cost
SOI Bulk GaAs
Switch
PA output stage
PA first stages
Integration on-goingin all 4G smartphones
x2die size reduction
19January 21st, 2016SOI Consortium Conference – Tokyo 2016
More expensive Substrates for cheaper dies
COST
1. FDSOI allows up to 6 mask levelsreduction
2. From 28nm, FDSOI processed wafer isless expensive than equivalent HKMG on bulk silicon
Processed wafer cost, FDSOI vs bulk
Source: GF, Semicon West 2013COST
IBS, 2014
20January 21st, 2016SOI Consortium Conference – Tokyo 2016
FD-SOI value propositionEnabling best performance, power and cost-efficiency
Energy EfficientEnergy EfficientUnique PerformanceUnique Performance Cost effectiveCost effective
+60% Faster than 28 nm Bulk SLP
5x more battery life than current generation
50% lower mask cost than FinFET
FD-SOI14 FinFET
Nor
rmal
ized
Per
form
ance
1.8
2
>20% Lower die cost
FD-SOI extends Moore’s Law beyond 28 nm enabling co st sensitive applications
Sources: GlobalFoundries FD-SOI technology webinar June 2015.EETimes article: Freescale, Cisco, Ciena Give Nod to FD-SOI | EE Times, March 1,2015
Used in Networking ASICs for datacenters
Cisco
40 bulk 28 bulk FinFET FD-SOI
19.9
4.5 6.411.6
Day
sof
bat
tery
life
$$$$
FD-SOI14 FinFET
20 nm Bulk
28 nm Bulk HPP
28 nm Bulk SLP
Norrmalized Cost/die (Mature Yields, Q1’19)
Nor
rmal
ized
Per
form
ance
0.9 1 1.1 1.2 1.30.8
1
1.2
1.4
1.6
Source: GlobalFoundries FD-SOI technology webinar, June 2015
Sources: GlobalFoundries FD-SOI technology webinar, June 2015
21January 21st, 2016SOI Consortium Conference – Tokyo 2016
22January 21st, 2016SOI Consortium Conference – Tokyo 2016
Multi-sourcing in place for HVM SOI platforms
200mm
300mm
23January 21st, 2016SOI Consortium Conference – Tokyo 2016
Take-aways
• FDSOI substrate is active part of the transistor
• SmartCut technology allows to meet FDSOI requirement s
• Metrology defined to predict/protect variability on device
• Die cost reduction can be supported thru right balance
between cost and simplification, delivering a pre-proc essed
wafer.
• Substrate availability /maturity is ready to suppor t FDSOI
mass adoption.
24January 21st, 2016SOI Consortium Conference – Tokyo 2016
SOI end customers in 2025
25January 21st, 2016SOI Consortium Conference – Tokyo 2016