Folie 1 - PowerPoint PPT Presentation

1 / 8
About This Presentation
Title:

Folie 1

Description:

(i.e. absolute permittivity of oxide eox = e0 eSiO2 is pinned to SiO2 in the model) ... (i.e. set PSP TOX parameter to the calculated TOXEOT value: Cox = eox / TOX ) ... – PowerPoint PPT presentation

Number of Views:27
Avg rating:3.0/5.0
Slides: 9
Provided by: assenmache
Category:
Tags: folie | ox

less

Transcript and Presenter's Notes

Title: Folie 1


1
Proposals for PSP102 Model Improvements
3Q06 CMC Meeting
Date October 2006
Infineon
Joachim Assenmacher COM BTS TD DIF
CM Acknowledgement Jae-Eun Park, IBM
2
Outline
  • Introduction of a Dielectric Constant Parameter
    in PSP103
  • Implementation of Additional Instance Parameters
    in PSP103
  • Improved Modeling Approach for Effective Doping
    in PSP103
  • Conclusion and Summary

3
Introduction of a Dielectric Constant Parameter
in PSP103
  • The TOX parameter in PSP is defined as the
    physical oxide thickness
  • (QM-effect and poly depletion effect are
    separate).
  • All electrical quantities in PSP are based on a
    fixed dielectric constant for SiO2 of 3.9
  • (i.e. absolute permittivity of oxide eox
    e0eSiO2 is pinned to SiO2 in the model).
  • In modern MOSFET devices medium-k dielectrics are
    common to reduce the gate leakage currents, at
    same drive capability (i.e. Cox).
  • E.g. A nitrided oxide with a physical thickness
    of TOXr 2nm is related to an er of 4.8.
  • Medium/high-k dielectric can be modeled as an
    equivalent oxide with thickness, adjusted for
    SiO2 (3.9) ? TOXEOT
  • TOXEOT gives a correct scaling for
    charges/capacitances and current gain factor
    (i.e. set PSP TOX parameter to the calculated
    TOXEOT value Cox eox / TOX ).
  • But TOXEOT gives a wrong scaling for gate
    tunneling currents, since above value is too low
    (for the tunneling distance) and have to be
    compensated with gate current parameter
    coefficients (unphysical). Gate currents have to
    scale always with the real physical oxide
    thickness (here TOXr of that oxinitride)!

4
Implementation of Additional Instance Parameters
in PSP103
  • Implementation of additional instance parameters
    DELVTO (threshold voltage shift) and FACTU0
    (mobility multiplication factor) to handle
    further layout effects like well-proximity,
    stress-enablement, etc. (besides the intrinsic
    PSP STI-stress model).
  • DELVTO and FACTU0 are common in BSIM4 netlists
    for commercial circuit simulators (e.g. HSPICE,
    Spectre, etc.).
  • Even if an intrinsic well-proximity model will be
    available for PSP, DELVTO and FACTU0 would be
    beneficial in the future, for modeling of
    enhanced layout effects.
  • Infineon can provide a verified Verilog-A PSP102
    model code, which supports DELVTO and FACTU0.

5
Improved Modeling Approach for Effective Doping
in PSP103
PSP102 Actual situation
PSP103 Proposal
FOL1, FOL2 RSCE of body effect
FOL1, FOL2 RSCE of body effect
NPCK, LPCK body effect
NPCK, LPCK body effect
NSUBO body effect
NSUBO body effect
NPCK, LPCK Vt-roll-up
Vbs
DPHIBL, DPHIBLEXP Vt-roll-up
DPHIBL, DPHIBLEXP Vt-adjustment
Vbs
DPHIBO Vth0-adjustment
NSUBO sets Vth0
DPHIBO Vth0-adjustment
NSUBO sets Vth0
  • PSP102 Pocket doping NPCK and pocket length LPCK
    parameter have almost no influence on
    Vt-roll-up/off, only on body effect (unphysical).
    Threshold voltage will be mainly modeled by DPHIB
    parameters (offset voltage of FB).
  • Proposal for PSP103 Pocket doping parameters
    NPCK and LPCK have an influence on Vt-roll-up/off
    and body effect as well (as they should be).
    DPHIB parameters can be used optional for
    Vt-adjustment.

6
Improved Modeling Approach for Effective Doping
(contd)
PSP103 Proposal for modified geometrical scaling
rule of effective doping
1
1
1 Introduction of one new local parameter NSUB
(Note In PSP102 Nsub is an internal variable)
7
Improved Modeling Approach for Effective Doping
(contd)
PSP103 Proposal for modified internal model
equation of FB (for Vt-adjustment)
Without FOL1 and FOL2 ! (Note PSP102 use
ln(NEFF/ni)
PSP102/103 Internal model equation of ?0 (for
body-effect) remains unchanged
With FOL1 and FOL2, i.e. RSCE of body-effect
  • PSP103 Proposal
  • FB- and ?0-equation share the same global
    parameters NSUBO, NPCK and LPCK by the
    NSUB-scaling rule (since also part of the
    NEFF-scaling rule).
  • Additional reverse short channel effect
    parameters FOL1 and FOL2 will have an influence
    just on body effect (and will not decrease the Vt
    as in PSP102 and therefore mustnt be compensated
    by DPHIB parameters ).
    DPHIB parameters can be
    used optional for Vt-adjustment and fine-tuning.

8
Conclusion and Summary
  • Introduction of a dielectric constant parameter
    in PSP103 will be needed for the correct and
    physical oxide thickness modeling of
    medium/high-k dielectrics, used in advanced CMOS
    technologies 1.
  • Implementation of additional instance parameters
    DELVTO and FACTU0 in PSP103, will be needed for
    modeling of enhanced (extrinsic) layout effects
    2.
  • A first suggestion for an improved modeling
    approach for the effective doping formulation
    (w.r.t. the influence of halo doping parameters
    on threshold voltage) in PSP103 has been made 3.
  • For more information refer to Infineons PSP102
    evaluation report, 2Q06 CMC

1 BSIM4 also allows to specify a gate
dielectric constant (EPSROX) different from 3.9
(SiO2) for the modeling of medium/high-k
dielectrics. 2 Also part of BSIM4.
3 Simple approach, requires just one new local
parameter.
Write a Comment
User Comments (0)
About PowerShow.com