interconnect opportunities a sematech perspective 5... · the interconnect space ... •all options...
TRANSCRIPT
Accelerating the next technology revolution
Copyright ©2012
SEMATECH, Inc. SEMATECH, and the SEMATECH logo are registered servicemarks of SEMATECH, Inc. International SEMATECH Manufacturing Initiative, ISMI, Advanced Materials Research Center
and AMRC are servicemarks of SEMATECH, Inc. All other servicemarks and trademarks are the property of their respective owners.
Interconnect Opportunities –
A SEMATECH Perspective
Sitaram Arkalgud
Sr. Director– Interconnect/3D IC
8th Annual Japan Symposium
June 26, 2012
Tokyo, Japan
Portions of this document are SEMATECH confidential, as marked
The Interconnect Space
• TSVs are just entering mainstream – Considerable issues around scaling (TSV and C2C interconnects), cost, thermal
management and thin wafer/die handling
• Cu will continue to be interconnect mainstay for many years – Resistance will rise exponentially with scaling
– Considerable work on addressing issues at member companies, universities, consortia
– Need to address weakest link in interconnect scheme • Chip to chip (medium term driven by BW) or within chip (near term)
• Power and signal mismatches between different chips will gate 3D stacking
• ~Terabit/s data handling are contemplated for 2nd/3rd gen wide I/O DRAMs
• Mixed electrical/optical approaches may be the only option
• Next generation options – Optical: nanophotonics
– Electrical: graphene, nanowires and CNT
Product functionality and power efficient performance at the right cost are the interconnect drivers
22 June 2012 2
2013
3D Roadmap
2012
memory memory Logic
Si TSV interposer
memory memory
Heat Sink and
TIM
3D Interposer
Common Industry Needs - Materials and Process Evaluations
- Thermal Management
- Materials, EDA
- Understanding Costs
- Reference flows, BOM to CoO
- Technology Maturity & Standards
- Supply Chain
- Interfaces and handoffs
- Pathfinding Tools
- Performance, power, floor planning
2015 2020
3D+
memory memory Logic
Si TSV interposer
memory
memory
Heat Sink and TIM
NAND Analog
NGI?
Production
Development
Pre-Development
22 June 2012 3
2013
3D Interconnect Roadmap
2012
memory memory Logic
Si TSV interposer
memory memory
Heat Sink and
TIM
3D Interposer
2015 2020
3D+
memory memory Logic
Si TSV interposer
memory
memory
Heat Sink and TIM
NAND Analog
NGI?
Production
Development
Pre-Development
Courtesy: WWW.Intel.com
First gen products underway Common Industry Needs - Materials and Process Evaluations
- Thermal Management
- Materials, EDA
- Understanding Costs
- Reference flows, BOM to CoO
- Technology Maturity & Standards
- Supply Chain
- Interfaces and handoffs
- Pathfinding Tools
- Performance, power, floor planning
22 June 2012 4
Mission of Materials and Emerging Technologies
Develop infrastructure for 3D transition in semiconductor industry
• Infrastructure development:
• Materials: Identify and down select
• Tools/Processes: Develop with universities/suppliers and harden
• Mechanisms: Fundamental understanding for manufacturability & scalability
• Models: Develop approaches to aid characterization and reliability assessment
• Key modules that enable acceleration of new technologies for
manufacturing
• Core competencies
• Narrow options: data driven consensus
• Standardize methods: data driven benchmarking
• Minimize cost, risk and avoid duplication
• Leverage resources from industry eco-system
• Address problems that are difficult for any individual company to tackle
• Nurture partnerships with suppliers and academia to execute mission
22 June 2012 5
• Demonstrate manufacturability of key 3D processes
needed for future generations • Low temperature Cu direct bonding process
• Underfill material development and assessment methodology
for thermal management
• 20: 1 aspect ratio TSV test structures and associated
challenges
• Materials and process development for thin wafer handling,
temporary bonding and room temperature debonding
• Test structure development for package interactions and
reliability monitoring
• Enable and assess next generation 3D stacking
implementation needs by driving
inspection/metrology, standards and specifications
through 3D EC
• Gap assessment for feasible Next Generation
Interconnect options
Key Activities
22 June 2012 6
Cu-Cu Direct Bond – Throughput
Improvements
• Conventional process (400ºC)
has a throughput of ~0.5 WPH – Vacuum requirement and high bond
temperature negatively influence tool
throughput
• Feasibility of low temperature
(~200ºC, 5 min) process
demonstrated on blankets – Process below Pb-free reflow
temperatures introduces compatibility
with all modern packaging materials
• New tool concept proposed to
increase throughput to 30 WPH
– 88% reduction in COO
POR
Low T
Process
New Tool
Concept
Cu-Cu bond
Interface
200ºC/10min
High Resolution TEM
195 C / 5 minute
bond E. Stinzianni
R Edgeworth 22 June 2012 7
TSV Module Roadmap Via Middle Gap Identification by TSV Node
TSV Scaling Issues
• Increasing Aspect Ratio
• Dielectric Liner Coverage and Continuity
• Barrier-Seed Coverage and Continuity
• Void-Free Cu Fill
• Increasing RC Delay
• Dielectric Liner Thickness
• Cu Volume
• Thermo-mechanical Performance
• Improves with scaling, but remains a
concern for device performance (KOZ)
and reliability in increasingly fragile ULK
dielectric layers
• Bond and Thin Unit Process Development
• Thin Wafer/Die Handling
Module Process TSV Node
5µm × 50µm 2µm × 40µm
TSV
LTH
RIE
WETS
MTR
INS
MTL
PLT
FRN
CMP
Ready for HVM
Ready for Initial Ramp-Up
Not Ready
Identify tool, material, and
process gaps in the
fabrication of high aspect
ratio (20:1) TSVs; use
cost-modeling to identify
the HVM-capable
solutions that best
address these gaps.
B Sapp 22 June 2012 8
Scaling of Chip to Chip Interconnects
Bonding Method
C4 FC (Contolled
Collapse Chip
Connect)
C2 FC
(Chip Connect)
TC/LR (Local
Reflow) FC TC FC
Schematic Diagram
Major Bump Pitch Range at
Application > 130 um 140 um ~ 60 um 80 um ~ 20 um < 30 um
Bonding Method
Conventional
Reflow
Reflow with Cu
pillar
Thermal
Compression
with Cu pillar
Thermal
Compression
Bump Metallurgy Solder (SnAg or
SnAgCu)
Cu + Solder
(SnAg or Sn)
Cu + Solder
(SnAg or Sn)
Cap
Cu
Bump Collapse Yes No No No
Underfill Method - Capillary
- No flow
- Capillary
- No flow
- Wafer Level
- No flow
- Wafer Level
- No flow
- Wafer Level
M-S Suh 22 June 2012 9
Scaling of Chip to Chip Interconnects
Bonding Method
C4 FC (Contolled
Collapse Chip
Connect)
C2 FC
(Chip Connect)
TC/LR (Local
Reflow) FC TC FC
Schematic Diagram
Major Bump Pitch Range at
Application > 130 um 140 um ~ 60 um 80 um ~ 20 um < 30 um
Bonding Method
Conventional
Reflow
Reflow with Cu
pillar
Thermal
Compression
with Cu pillar
Thermal
Compression
Bump Metallurgy Solder (SnAg or
SnAgCu)
Cu + Solder
(SnAg or Sn)
Cu + Solder
(SnAg or Sn)
Cap
Cu
Bump Collapse Yes No No No
Underfill Method - Capillary
- No flow
- Capillary
- No flow
- Wafer Level
- No flow
- Wafer Level
- No flow
- Wafer Level
2012 TSV
2014/15 TSV
M-S Suh 22 June 2012 10
Next Generation Interconnect
• All options will leverage 3D IC in some form
• We will use our 3D IC capability to narrow down options for manufacturability
22 June 2012 11
University
Research
Member Company
Implementation
2012 2017
SEMATECH testbed for materials, processes
and tool evaluations
Doped Copper
Silver
Gold
Graphene
Carbon Nanotubes
Nanowires
Superconductors
2D Conductors
Optical Interposer
Optical Onchip
Better evaluated at MC
Futuristic
Difficult to implement in fab
Still in research mode
Futuristic
Futuristic
Narr
ow
to
1-2
“m
ost lik
ely
” o
ptio
ns
Attributes of CNT and Graphene
• Both graphene and carbon nanotubes promise superior properties compared to Cu
– Virtually ballistic transport
– Higher reliability
– Higher thermal conductivity
Si Cu Graphene CNT
Resistivity (u.ohm.cm) 1.68 1 1
Current density (A/cm2) 10E6 10E8 10E9
Carrier Mobility (cm2/Vs) 450 200,000 100,000
Thermal Conductivity (W/mK) 400 5000 3000
22 June 2012 12
GNR – Modeling and Status
• Naeemi plot vs Murali plot
A Naeemi and J Meindl, IEEE TED (56) 9, 2009 R Murali et al EDL(30) 96 2009
• Graphene shows considerable promise with large MFP (virtually ballistic), mobility,
thermal conductivity and EM resistance
• Still relatively immature
• Minimizing edge effects and demonstrating low resistivity is key for interconnects
Attractive
alternatives to Cu
Resistivity shows
increase with scaling!
22 June 2012 13
CNT Status*
* Adapted from “M Nihei “CNT/Graphene Technologies for Advanced Interconnects” IITC Short Course, June 3 2012 San Jose, CA”
MIRAI-Selete ==>
AIST, LEAPSamsung ==> KAIST
VIACARBON ==>
LETI, CambridgeIMEC
TechnologyRequirement
Process Single Damascene Single Damascene Single Damascene Single Damascene
Catalyst/Bottom
Electrode
Co
particle/TiN/Ta/CuCo film/Ti/TiN Fe film/AlCu Ni film/TiN
CNT Growth
Temperature
< 400CThermal CVD C2H2
365 - 450C
Remote Plasma
CVD, CH4/H2, 600C
Thermal CVD,
C2H2/H2/He, 580C
Thermal CVD,
C2H4/H2/Ar, 400 -
470C
CNT Density
(/cm2)> 10E12 10E11 - 10E12 4*10E11 2.5*10E12 7*10E10 ~ 2*10E11
Top Electrode< 400C CMP Ti/Cu (RT) CMP Ti/Al (500C) CMP Ti/Pt (RT) CMP Ti/Au (400C)
Resistivity
(mWcm)
~10 (W), ~2
(Cu)
10 (52 W @ f
70nm, 450C)
73 (293 W@f
80nm)
390k (10 kW@f
1000nm)
28k (7.9 kW@f
300nm)
Current density
(A/cm2)> 10E7 2*10E8 ---------- ---------- ----------
• CNT has properties similar to graphene
• Still relatively immature
• Demonstrating low resistivity is key for interconnects
22 June 2012 14
Electrical Interconnect Limitations
Bandwidth Limitations – Wires are not scalable
Maximum Bandwidth:
Bandwidth is fundamentally limited by the
available area (e.g. once all area is used there is
no more capacity)
Power Limitations - Total interconnect power is
too high
~50% of Total Chip Power
Expected to rise to >80%
Cannot increase total power. Limited to ~200W.
Signal Integrity and Latency Limitations
RC Delay will increase considerably with scaling
Bµ1016 Area
Length2
D. Miller, J. Parallel and Dist.
Comp. 41, 4252 (1997)
D. Miller, Proc IEEE 97, 1166 (2009)
Sam Naffziger, AMD Fellow,
2011 VLSI Symposium Keynote
Source: IBM
SEMATECH Confidential 22 June 2012 15
Why Silicon Photonics?
All components of a Silicon Photonic Interconnect have been demonstrated:
Silicon photonics is a promising technology:
High Bandwidth (>200 THz)
Low Power (<100 fJ/bit)
Compact Devices (< 2mm3)
Compatible with Fiber Optics
Seamless CMOS Integration
Silicon Photonic Circuit
Laser Electro-Optic Modulator Photodetector
Intel/UCSB – III-V Bonded Laser
MIT – Ge Laser Cornell University – Microring Modulator
Intel – Ge Detector
K-W Ang 22 June 2012 16
SEMATECH’s Si Photonic Building
Blocks Lasers Modulators Detectors Waveguides
Interconnects Heaters Gratings Switches
SEMATECH has designed a complete Silicon Photonic device library
The device library uses a 300mm reference flow
The interconnects will be integrated into a 3D CMOS process
K-W Ang 22 June 2012 17
Stages Cu 2D Cu 3D Optical Graphene CNT
Concept X X X X X
Mechanisms X X X X X
Reference Flow X X X
Materials Selection X X X
Device Feasibility X X X
Module Development X X
Equipment Hardening X X
Infrastructure X X
Reliability X X
HVM X X
C&F
UPD
Module
Interconnect Options
Maturity of Interconnect Options
Best candidate for
Next Generation HVM
Relatively immature
22 June 2012 18
Enablement Center
• Very successful program
– Focus is on enabling standards (http://wiki.sematech.org) and ensuring chip interoperability
• Obtained considerable industry attention
– Attracted several new members since Jan 2011 start
• Analog Devices, Altera, Invensas, NIST, On Semiconductor, Qualcomm
• Increased program diversity
– Network has expanded considerably
• Strong relationship with SRC, SEMI, SI2, JEDEC
• Fabless companies, OSATs and universities
• 3D Enablement Center playbook is a template for NGI – Reference flow(s)
– Gap analysis
– Standards and specifications
– Forward looking option assessments with pathfinding tools
22 June 2012 19
3D Members and Partners
22 June 2012 20
Summary
• Interconnect Center is addressing material and tool gaps by leveraging
existing capabilities in 3D program
– TSV and advanced package interactions • Thin wafer/die handling
• Cu-Cu direct bonding
• Underfill
• TSV scaling, infrastructure gaps
– Working on gaps for next generation interconnect
• Enhancing linkage to key research institutions and universities to
translate concepts from lab to fab
• Interconnect continues to require engagement of larger section of the
supply chain
– Fabless, packaging, suppliers, universities
22 June 2012 21
22 June 2012 22