Front End Processes - PowerPoint PPT Presentation

1 / 23
About This Presentation
Title:

Front End Processes

Description:

DRAFT Work In Progress - NOT FOR PUBLICATION 13 July 2005. Front End Processes ... Allowable defect densities for HP MPU consider new chip size scenario in 2005 ... – PowerPoint PPT presentation

Number of Views:70
Avg rating:3.0/5.0
Slides: 24
Provided by: danfe7
Category:

less

Transcript and Presenter's Notes

Title: Front End Processes


1
Front End Processes
Starting Emerging Materials Surface
Preparation Thermal Thin Films Doping Etching St
acked-Capacitor DRAM Trench-Capacitor DRAM Flash
Memory FeRAM
Carl Osburn (osburn_at_ncsu.edu) Jeff Butterbaugh
(jeff.butterbaugh_at_fsi-intl.com)
  • ITRS Public Conference
  • July 13, 2005
  • San Francisco, CA

2
Starting Materials Defect Densities Issues and
Challenges
  • HP MPU chip size
  • Previous ITRS defect densities were based on C/P
    chip sizes (140 mm2)
  • Allowable defect densities for HP MPU consider
    new chip size scenario in 2005
  • Site Flatness and Nanotopography
  • SFQR remains the most commonly used metric
  • Lack of consensus regarding requirement
  • some IDMs favor values greater than the
    technology generation (TG) while others state
    values equal to or less than TG
  • Nanotopography impact is not universal, seemingly
    critical for some IDMs but of little interest to
    others

3
Starting Materials-Edge Exclusion
  • Set at 1.5 mm in 2007 (from 2 mm in 2003) by
    Factory Integration TWG, Where it will be Red
  • Implementation Poses a Significant, Immediate
    Challenge in FEP
  • Edge Rolloff
  • Definition currently being addressed by SEMI /
    JEITA
  • Assessment of table entry for future ITRS
    publications (post 2005) to be implemented once
    agreed upon metric is obtained
  • Defects
  • Leading, Current SOI Manufacturing Processes
    Produce a Spatial Region that is not Usable Real
    Estate for IC Production Hence Coloration
    Starts with Red

4
Emerging Materials
  • ? Ultra-Thin Body, Fully-Depleted SOI
  • Layer Thicknesses Reduced Based Upon More
    Thorough PIDS Designs
  • 0.35 x Lgate 8 nm /-10 in 2008
  • ? Strained Silicon and Strained Silicon on
    Insulator (sSOI)
  • Treated in Starting Materials Sub-chapter Text
  • ? Orientation Modification for Optimization of
    CMOS Electron and Hole Mobility Values
  • Treated in Starting Materials Sub-chapter Text

5
450 mm Wafers
  • 450 mm Wafers Required in 2012 (per ORTC
    direction)
  • Introduction of 450 mm presents unprecedented
    challenges
  • Technical (meeting specs over larger areas)
  • Economic (especially for wafer, equipment, and
    metrology suppliers)
  • Critical path definition - We are already late to
    meet this development cycle
  • Standardization
  • Wafer specification (type, thickness, diameter
    tolerance, etc.)
  • Factory automation (load lock, transportation
    method, etc.)
  • Wafer package (FOSB, FOUP, door configuration,
    etc.)
  • 450 mm Wafer Issues Will be Highlighted in 2005
    ITRS
  • FEP subchapter on this topic
  • Supplemental position paper
  • Inclusion of material in executive summary

6
Yield Model and Surface Particles
  • Killer defect density value for Surface Prep is
    calculated based on Poissons model for yield
    99
  • Values changes from 2004 are small, due to die
    size changes
  • 97 (in 2004 ITRS) to 94.2 (in 2005 ITRS) for 2005
    critical particle count per wafer
  • Currently different models are used for Yield
    Enhancement and Starting Materials/ Surface Prep
  • Final particle density values are approximately
    the same for all models, however use of different
    equations is because goals of groups are
    different
  • YE and FEP will have a workshop to determine the
    appropriate Starting Materials defect levels and
    establish specific call outs for critical cleans
    (specifically pre-gate)
  • Concern is ability to measure defects at critical
    particle diameter that contributes to yield loss

7
Back Surface and Edge Particles
  • Goal for 2005 is to relate back surface particles
    to yield
  • Currently little public data available to set
    metrics
  • The values in the 2004 and 2005 are based on
    survey results
  • 2005 ITRS value are the same as the 2004 update
  • Wide range of data obtained in the survey
  • Survey results are not consistent Yield Model
  • Contribution from edge particles will be examined
    in 2005/6
  • Contribution from cluster of back surface
    particles vs. single particles are known to be
    important
  • Especially to Lithography depth of focus

8
Other Surface Prep Changes
  • Watermark criteria re-established at table metric
  • Watermark 0 for all years
  • Silicon and oxide loss numbers continually
    decrease, down to 0.4Å at 57 nm node, and 0.2Å at
    32nm node
  • Inclusion of discussion of high-k and the surface
    preparation requirements
  • Gate oxide dielectric and high-k have same low
    contamination requirements
  • Values With No Changes in 2005
  • Critical GOI metals and mobile ions
  • Surface Roughness
  • Surface carbon contamination

9
Parallel Lines to Capture Alternative Device
Scenarios
  • Extended Bulk and PD-SOI Devices to Overlap 4
    Years with FD-SOI / Multi-Gate
  • Recognizes Different Approaches in Different
    Companies
  • Roadmap will Illustrate Requirements for
    Different Scenarios
  • Increases the Number of Lines in the Table
  • FD-SOI and multi-gate will be placed on separate,
    over-lapping lines.

2005 2006 2007 2008 2009 2010
2011 2012 2013 2014 2015 2016
Bulk/PD-SOI
FD-SOI
Multi-Gate
The Baseline Scenario Envisions a Nominal Switch
from Bulk to FD-SOI Around 2008, and a Switch to
Multi-Gate Around 2012
10
Deferred Implementation of High k
Year 2003 ITRS 2005 ITRS
2004 2005 2006 2007 2008
Enhanced Mobility
High k for Low Power
High K for MPU Metal Gate FD-SOI
Enhanced Mobility High k for Low Power
High K for MPU Metal Gate FD-SOI
  • Puts Multiple, Major Changes into One Year
    (2008)
  • Causes Other Parameters, e.g., Xj, to Scale More
    Aggressively

11
Metal Gate Electrode Work Functions
  • Constraint was Added to PIDS Device Design
    Scenario for
  • Bulk, FDSOI, and Multi-Gate, Namely to Minimize
    the Number
  • of Different Work Function Metals Needed Over
    Time
  • ? Three-Four Systems Seen to Meet All Needs
    Through 2020
  • Band Edge (100 meV inside band) for Bulk
  • 150 meV Above/Below Midgap for HP FDSOI
  • Midgap for HP Multi-Gate, LOP FDSOI and
    Multi-Gate
  • 100 meV Below/Above Midgap for LSTP FDSOI and
    Multi-Gate
  • Improvements Can be Made by Allowing
    Workfunction to
  • Vary Yearly, but fms Control is a Vt Tolerance
    Issue
  • Band Edge Metal Gates Viewed as Limiting Factor
    in High k
  • Deployment

NMOS/PMOS
12
Doping and Junctions
  • EOT Values to be Given for Different Poly-Si
    Doping Levels
  • 1e20/cm3 (lightly-doped) tdepletion 0.5 nm
  • 1.5e20/cm3 (baseline doping) tdepletion 0.4 nm
  • 3e20/cm3 (advanced, e.g., laser annealed)
    tdepletion 0.2nm
  • Junction Depths are Scaled More Aggressively in
    2005 to
  • Accommodate Slower EOT Scaling
  • Junction Implant Parameters to meet Xj-Rs
    Requirements to
  • be Given in Supplemental Tables
  • Maximum Implant Energy
  • Dose
  • Series Resistance Becomes a Increasingly
    Important
  • Limitation (Ominous)
  • Scenario, Rather than Requirements, Given for
    Elevated
  • Junctions in FDSOI and Multi-Gate

13
Repartitioning of Lithography and Etch
Contributions to Physical Gate Length
  • Issue Neither Etch nor Lithography Could Meet
    Tolerance Budgets to Achieve /- 10 Control of
    Physical Gate Length
  • 2005 Partial Solution
  • Maintain Final (Etched) Physical Gate Length at
    2003 Values
  • Relax (increase) Printed Gate Length Dimension
    in Resist
  • Increase the Amount of Resist Trim
  • Increase the Total Tolerance on Physical Gate
    Length to 12
  • Repartition Tolerance Budgets from 80 Litho/20
    Etch to
  • 75 Litho/25 Etch

Observation At the 90 nm Node, Most
Manufacturers are using Gate Lengths Longer than
ITRS Values. - If Trend Persists, ITRS
Values May Need to be Adjusted Upwards
14
Gate Etching
  • Gate sizing variation allowance
  • Surveys in 2003 and 2004 indicated that the
    industry
  • was not processing etched gates at lt10 3
    sigma
  • Feedback from device modeling indicated that
    relaxing
  • to 12 3 sigma would not degrade performance
  • Requirements for gate etch were relaxed as
    follows

(Litho is given 75 of the allowable gate length
variance)
15
Gate Etching
Increase of overall gate sizing tolerance from
10-12 and allocation of variance to etch from
20-25 - Increases Litho allowance by 18 -
Increases Etch allowance by 35 - Allows
improved coloration of the table entries
2001 ITRS 2003 ITRS 2005 ITRS
16
Stacked-Capacitor DRAM Changes for ITRS 2005
  • Cell size factor, a, is projected to remain at 8
    through 2007not scaling as rapidly as projected
    in 2003-4
  • Capacitor dielectric material and Dielectric
    constant are modified based on survey of DRAM
    manufacturers

17
Modified Parameters for Stacked-Capacitor DRAM
Year of Production
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
8.0
7.5
7
7
6
6
6
6
6
6
6
6
6
5
5
Cell size factor a
8
8
8
6
6
6
6
6
6
6
6
6
6
6
6
6
Proposal for 2005
- Discussions still on-going regarding capacitor
dielectric material
18
Trench-Capacitor DRAM Roadmap
Bottle
Geometry
HSG
Aspect Ratio
?601
gt1001
Effective Oxide Thickness
5.2nm
1.2nm
Materials
High-k Dielectric
Metal Electrode
New materials
19
Trench-Capacitor DRAM RoadmapChanges in 2005
versus 2004 Update
Solutions exist down to 57nm node (WAS
70nm)Red brick wall moved out to 28nmHigher
trench aspect ratios anticipated for 45nm MIM
option for 50nm (WAS 45nm)MIM only option
for 35nm (WAS epi-high-k as 2nd option)
20
ITRS 2005 Trench DRAM Roadmap
Known Solutions exist through 57nm node Red
brick wall moved to 28nmMIM option for 50nm
only option for 35nm
21
Flash Memory
  • A sharp decrease in the interpoly dielectric
    (IPD) EOT is needed at 45-50 nm, for NOR, and
    maybe for NAND, to enable further scaling of the
    Bit Line pitch and the Floating Gate spacing and
    to maintain an acceptable Coupling Ratio (CR)

10-12 nm EOT
4-6 nm EOT
  • ONO (Oxide-Nitride-Oxide) technology is no longer
    feasible and new materials/architecture are
    necessary (e.g., high-k,)
  • Alternatively new floating gate design (e.g., SiN
    charge storage)
  • Since STI may be formed post floating gate and
    because it is deeper, its aspect ratio is higher
    than in logic ? trench fill technology is more
    critical (MLD or SOD)

22
FeRAM Roadmap
  • Japan PIDS Surveyed the Commercial Production
    Status of FeRAM
  • 2005 Technology Node Feature Sizes will be for
    Stand-alone memory at 10k/mo volume by 2
    companies (as with DRAM)
  • 2003-4 ITRS Used Both Stand-alone and Embedded
    Memory
  • Scaling Rate of FeRAM is Subject of Ongoing
    FEP/PIDS Discussion

23
FeRAM Roadmap
  • Access and Cycle Time are now Expected to Scale
    More Slowly than Projected in 2003 ITRS
  • Assuming a Minimum Switching Charge Density of
    30mC/cm2 and a Bit Line Voltage Swing of 140mV,
  • 3 Dimensional (3D) Capacitors will be Needed
    in 2010
  • Definition of Minimum Switching Charge Density
    after 3D Introduction will be Determined with PIDS
Write a Comment
User Comments (0)
About PowerShow.com