PowerPoint Presentation by PvU21amT



           Alessandro Paccagnella

        DEI, Università di Padova, Padova, Italy
        e-mail: alessandro.paccagnella@unipd.it

       EWRHE, Villlard de Lans, 31 March 2004

        Floating Gate (FG) memories: Flash
            SEE and data retention
            TID and endurance
        Stressing the gate oxides
        Plasma damage
        Final considerations

A. Paccagnella 2/67         31 March 2004    EWRHE


              Electrical                          Radiation
               stresses                           stresses


A. Paccagnella 3/67             31 March 2004                  EWRHE
Floating Gate memories
                          Floating Gate memories overview

                                           Electrons stored
                       (CG)                in the FG
                                           VTH grows
                                           Writing (Flash):
                                              Channel Hot
                                           Erasing (Flash):
  Main reliability issues:                    Fowler-
  - Endurance (W/R/E cycles)                  Nordheim
  - Data retention                            tunnelling

A. Paccagnella 5/67        31 March 2004           EWRHE

                                        Decrease of
                                      the programming
                                      window as the
                                      number of P/E
                                      cycles increase
                                      of defects in the
                                      tunnel oxide

A. Paccagnella 6/67   31 March 2004          EWRHE

                                      Increase of
                                      the writing
                                      time for
                                      number of

A. Paccagnella 7/67   31 March 2004        EWRHE
                                         Data retention

  Log                                   After the
  (bit)                               burn-in (bake)
                                      test, one bit has
                                      lost charge and
                                      VT decreased

A. Paccagnella 8/67   31 March 2004         EWRHE
                                  Rad effects: literature review

  Test of off-the-shelf components:
      No information on internal circuitry
      No information on actual threshold voltage of cells (Direct
      Memory Access - DMA not available: only 0/1 outputs)
  Many different failure modes:
      Standby current exceeding limits
      Stuck bits (cannot be reprogrammed)
      Failing in writing/erasing
      Functional interruption (Single Event Functional Interrupt,
      SEFI), etc.
      Destructive events (Single Event Latchup, SEL), etc.

A. Paccagnella 9/67         31 March 2004               EWRHE
                                      Literature review – TID/1

  Time to erasure [s]
                                                  Nguyen, et al,
                                                  IEEE Rad.
                                                  Eff. Data
                                                  Intel Flash,
                                                  Charge pump
                                                  is the weak

                        Total Dose [krad(SiO2)]

A. Paccagnella 10/67      31 March 2004                EWRHE
                                 Literature review – TID/2

                                          Nguyen, et al,
                                          IEEE T-NS,
                                          64Mbit, 2

  Irradiated in Write/Read/Erase mode
  Unbiased device better than biased one  circuitry!
A. Paccagnella 11/67   31 March 2004            EWRHE
                                  Literature review – SEE/1

    Schwartz et al,
    IEEE T-NS, 1997
    Several different
    failure modes

A. Paccagnella 12/67    31 March 2004            EWRHE
                                         Literature review – SEE/2

              Nguyen, et al, IEEE T-NS, 1999
              Cross section for buffer errors

                Threshold LET
                                                    cross section

A. Paccagnella 13/67            31 March 2004            EWRHE
                                            Literature review – SEE/3

                                                      Roth, et al, IEEE
                                                      Rad. Eff. Data
                                                      Workshop, 2000
                                                      Toshiba 256
                                                      Cross section for
                                                      page buffer and
                                                      address buffer
                                                      Different failures
                                                      are possible!
particle       Silicon   Nickel        Iodine

A. Paccagnella 14/67              31 March 2004            EWRHE
                Radiation effects on Floating Gate memories

   Most of published studies:
       Both SEE and TID tests
       Failure happens in control systems (state machines,
       buffers, charge pump)
       Stored information not directly accessible in commercial
       devices (no DMA available)

   Recent results have shown that [1][2]:
       Bit flips are not likely but may happen in FG arrays
       The main charge loss is due to a transient mechanism

                                      [1] Cellere, et al., T-NS, Dec.2001
                                      [2] Cellere, et al., T-NS, Dec.2002

A. Paccagnella 15/67        31 March 2004                        EWRHE
                                    DVTH spatial distribution

     After 2x107 Iodine ions/cm2

  All cells programmed to “1”            Arrows: |DVTH|>2V
  Multiple Hits on a cell (~1.5% of      3 bits with VTH<4V
  hit cells)

A. Paccagnella 16/67    31 March 2004             EWRHE
            Retention properties of Floating Gate memories

   FG cells should ensure a data retention time of 10
   yrs at room temperature: low leakage from FG
   Accelerated electrical stresses are customarily
   performed in order to verify that Write/Erase
   electrical operations at high fields do not endanger
   the long time retention properties
   Appearance of a single failing bit is a big concern:
   the memory chip fails
   Data retention properties of irradiated FG cells: an
   open question

A. Paccagnella 17/67      31 March 2004         EWRHE
                                                  VTH statistical distribution/1

   0.15mm2 FG                                100000
   devices                                                After 2x107 I

                           Number of cells
   I ion irradiation                                      ions /cm2
   A secondary peak
   appears at ~6V                              100
   The very low VTH                             10
   tail is probably due
   to double hits                                1
                                                      4        5        6        7     8
   (1.5% of hit cells at                                             VTH (V)
   this fluence)

A. Paccagnella 18/67       31 March 2004                                       EWRHE
                                               On the charge loss mechanism

  I ion irradiation                            100000
                                                            Fresh           DQ
      7,000 e/h pairs                          10000        After 2•107 I

                             Number of cells
      generated                                             ions /cm2
      1% surviving                              1000
      recombination                              100
      70 holes injected in
      FG                                          10

  DQ=3,500                                         1
  electrons                                             4        5        6        7     8
                                                                       VTH (V)

  No known mechanism can explain these data!

A. Paccagnella 19/67         31 March 2004                                       EWRHE
                                                     VTH statistical distribution/2
  0.15mm2      FG

                         Cumulative probability
                                                              After 2·106 cm-2
  devices                              63.2%                  After 107 cm-2
                                       12.6%                  After 2·107 cm-2
  Si ion irradiation
  Large statistical                     0.2%
  tails, linearly                      0.03%
  growing with                        0.004%                                     134 MeV Si
                                                         6     6.5        7        7.5        8
  Tail bits can result                                                  VTH(V)
  in read error                                     6%

                                                         0                  1                 2
                                                                            7            2
                                                             Fluence (10 ions/cm )
A. Paccagnella 20/67      31 March 2004                                           EWRHE
                               ID-VG characteristics of hit cells

  EPROM, Ag irradiation
  Hit cells exhibited a shift (but no clear deformation) of the ID-VG
  FG charge loss is the
  only immediately
  detectable effect of
  irradiation (FG MOSFET
  is still working)

   …but is this effect
    the only one?

A. Paccagnella 21/67        31 March 2004                EWRHE
                       Radiation Effects on VTH distributions

  Large tails after               0.04mm2 (small) devices

  irradiation            99.99%
  Number of bits in       93.4%

                       Cumulative probability
  tail does not           30.7%
  depend on ion
  LET (it depends
  on fluence)             0.67%
  DVTH strongly           0.09%

  depends on ion         0.012%
                                             Ag       Ni
  LET                    0.002%
  These cells (hit) only        3    4      5       6    7              8
  are considered in the                      VTH (V)
  next experiments
A. Paccagnella 22/67                            31 March 2004   EWRHE
                                                    Data retention in hit FG cells - 1

                                                                  0.04mm2 (small) devices
  Hit devices                                   99.99%

                       Cumulative probability
  only were re-                                  93.4%
                                                                  After 20
  programmed                                     30.7%
  After only                                     0.67%
  30min a clear                                  0.09%
  tail appears…                                 0.012%
  …which                                        0.002%
  increases                                              3          4       5          6    7      8
  more and                                                       After 30    VTH (V)       After
  more with                                                      min                       program
  time                            After 20 days, DVTH~4V!!!
A. Paccagnella 23/67                              31 March 2004                            EWRHE
                                                   Data retention in hit FG cells - 2

                                                            0.15mm2 (large) devices
    If the same                                 99.99%
                                                             After program

                       Cumulative probability
    experiment is                               93.4%        After 12h
    repeated on                                 30.7%        After 36h
                                                             After 3 days
    large FG                                     4.8%        After 2 weeks
    devices: same                               0.67%        After 3 months
    type of results,                            0.09%
    but smaller                                 0.01%
    Strong                                           4           5      6       7      8
    dependence on                                                    VTH (V)
    FG area

A. Paccagnella 24/67                             31 March 2004                 EWRHE

  In order to evaluate quantitatively the leakage current,
  an original model has been developed (Larcher et al.,
  IEEE-TNS, 2003)
  Modeling has been possible only by knowing in detail
  the cell structure and layout
  Close collaboration with foundry is needed to develop
  quantitative models of complex devices

A. Paccagnella 25/67    31 March 2004           EWRHE
                       Data retention in hit FG cells: model

  Why are devices with smaller FG area more sensitive?
      Q  charge stored in the FG
      CPP FG/CG capacitance
      VTH VTH after UV erasure (no excess charge FG)
            dVTH d  UV Q  I tunnel   I path
                 VTH            
             dt  dt     CPP  CPP     AFG C
  If Itunnel is due to a small path (independent on FG
  area), charge loss over time increases if FG area
  Charge loss is mediated by a localized conductive
             path through tunnel oxide
A. Paccagnella 26/67      31 March 2004            EWRHE
                                                                Model – 1

  Ion generates a                          Floating gate
  plasma of electrons                      (electrons stored)
  and holes

                            Tunnel Oxide
                                                   Ion track
  Part of these will
  recombine (99% for                               holes
  Iodine in our                                   electrons
  conditions)…                                Oxide defects

  …generating some
  defect                                   Substrate

A. Paccagnella 27/67    31 March 2004                           EWRHE
                                                                    Model – 2

  Electrons will be
  quickly swept away…                          Floating gate
                                               (electrons stored)
  ..while holes will take

                                Tunnel Oxide
  more time
  During their motion,
                                               Electron tunneling
  holes generate other
    Finally, we obtain a
     number of defects                         Substrate
 (depending on ion LET)
aligned along the ion track
A. Paccagnella 28/67        31 March 2004                           EWRHE
                                                                  Model - 3

    How can we evaluate the current along this path?

    Consider the ion track [1]               Floating gate
                                             (electrons stored)
    Randomly generate a

                              Tunnel Oxide
    Gaussian distribution

    of defects
                                               I=    n
    Evaluate the current
    through each possible
    Then sum all the
    currents            [1] Oldham, T-NS, Dec.1985           8nm
A. Paccagnella 29/67     31 March 2004                            EWRHE
                                                      Model – 4

   The whole procedure is repeated 10,000 times, to
   obtain (Monte Carlo approach):
       Average value of the current
       Statistical variations
   The number of defects is used as a running
       Considers all possible percolation paths
       Considers barrier deformation induced by positively
       charged traps
       Couples electrons to oxide phonons
       Calculates capture and emission rate of each trap, based
       on its energetic and physical position

A. Paccagnella 30/67        31 March 2004             EWRHE
                                             Model vs. Experimental

   Experimental data        10-19
   obtained from            10-20
            2 device        10-21

                            IG (A)
   irradiated with I
   previously shown         10-24
   Bars  variance          10-25                    Experimental
   (spread) of              10-26
                                0.8 1   1.2     1.4     1.6     1.8
   experimental data                     FOX (MV/cm)
   Lines            With 20 defects, good agreement on:
   calculations        Mean value
                        Extreme values (spread)

A. Paccagnella 31/67        31 March 2004                EWRHE
                                                Scaling oxide thickness

  The model can be                 10-17
  use to predict                   10                                       +
  retention properties             10
  of device with                   10-20

                         IG (A)
  thinner tunnel oxide                                                    mean
  This simulations                   -23
  done for                           -24
                                   10          EOX=1.8MV/cm                  -
  EOX=1.8MV/cm                     10-25
  Number of defects                10-26
  scaled accordingly                       7         8              9             10
                                                         TOX (nm)
  to oxide thickness
   Gate leakage increases by more than one order of
          magnitude if TOX decreases by 30%
A. Paccagnella 32/67              31 March 2004                         EWRHE
                           Cosiderations on FG cells/SEE

     Large tails in cumulative probability plots of Flash
     device after irradiation with heavy ions
     Hit cell (and only hit ones) show retention
     A path of defects is generated by the impinging ion
     Electrons can tunnel through these defects
     discharging the FG
     Proposed a model to describe results, featuring
     statistical approach and phonon-assisted tunneling
     Device with smaller FG area and thinner tunnel
     oxide are more sensitive (NAND-based?)

A. Paccagnella 33/67     31 March 2004          EWRHE
                                        TID effects on FG cells

     Tests on           test structures with Co60
     gamma rays
     FG cells feature a selection MOSFET (2
     MOSFETs per memory cell)
     TID studies have been coupled to usual
     endurance stresses to evaluate the impact of
     ionizing radiation on the W/R/E characteristics

A. Paccagnella 34/67    31 March 2004                EWRHE
                                                                          Charge loss from charged FG

     Reduction of stored charge                                                  Errors in array:
     change of FG VTH, moving
                                                                                    lower part of the
     toward the “intrinsic” VTH                                                     “0” distribution
                                 Fresh                                              approaches the
  Cumulative probability

                                 9kra d (S 2)
                           99.9%                                                    1.7V limit after
                                 27kra d (S 2)
                                 90kra d (S 2)
                                                                                    90krad (SiO2)
                                 270kra d (S 2)
                                              i0                                    failures in the
                                 900kra d (S 2)                                     array are possible
                                                                                    (even if not sure)
                            0.1%               "1"                  "0"             after this dose has
                              0%                                                    been exceeded
                                -2      -1         0     1      2         3   4

A. Paccagnella 35/67                                         31 March 2004                 EWRHE
                       Basic models for charge loss from FG

                 A) electron/ion interaction


Ionizing                               FG
radiation                            B) Holes trapped in
                                     tunnel oxide
                                     C) Holes injected in FG
      Si-sub      Tunnel Oxide
                          Snyder, et al, IEEE T-NS 1989

A. Paccagnella 36/67         31 March 2004                EWRHE
                              Modeling charge loss from FG/1

        Photoemission (direct interaction between
        stored electrons and radiation)
        Charge trapping in the oxide:
            Neglected in modeling
            (Relatively) thin oxides
            Can be evaluated after irradiation and electrical

     In our model, the key role is played by charge generation
                    and recombination in oxides

A. Paccagnella 37/67         31 March 2004               EWRHE
                                Modeling charge loss from FG/2

               Control Gate                    contributions:
                 “top” SiO2
                                                 Gate oxide (Qtunnel)
               “bottom” SiO2                     “Bottom” SiO2 layer
               Floating Gate
                                                 Nitride layer
                Gate oxide
                 Substrate                       (QONO2)
     STI                           STI
                                                 “Top” SiO2 layer

    Qtunnel                  QONO1

A. Paccagnella 38/67           31 March 2004                EWRHE
                                       Modeling charge loss from FG/3

                                                Charge generation by
              Control Gate
                “top” SiO2                      radiation:
              “bottom” SiO2
                                                      18eV per e/h pair (Benedetto
                                                      et al, IEEE T-NS, 1986)
              Floating Gate
               Gate oxide                       Recombination:
    STI                         STI                   function of electric field
                                                      Evaluated based on
          Qtunnel           QONO1                     existing data (Ausman, HDLR,
 Floating Gate                Si3N4                   1986, Oldham, et al, IEEE T-NS,

                                                Remaining e/h are
                                                injected in FG
   Substrate             Floating Gate                electrons             holes

A. Paccagnella 39/67                  31 March 2004                      EWRHE
                                   Modeling charge loss from FG/4

                                                         Control Gate
            Control Gate                     QONO2
              “top” SiO2
                 Si3N4                                 SiO2
            “bottom” SiO2
            Floating Gate
             Gate oxide                    electrons
    STI                     STI            holes
  Nitride layer:
                                                         Floating Gate
      No charge generation
      100% charge trapping
  Why DVTH?
      Charge injected in nitride from “top” and “bottom”  sheet
      charge layer  DVTH
A. Paccagnella 40/67              31 March 2004                EWRHE
                                 Modeling charge loss from FG/5

  Qtunnel: generated in gate
  QONO1: generated in
  “bottom” SiO2 layer
  QONO2: resulting from
  charge injection and
  trapping in nitride
  Good agreement with
  experimental data
  Underestimation at high doses:
      Electric field decreases when FG loses electrons (or holes)  enhanced
  “Bottom” SiO2 layer is the key for device reliability
A. Paccagnella 41/67           31 March 2004                   EWRHE
                                                  Interaction of electrical stress and irradiation

                           Electrical stress  irradiation (2.7 Mrad(SiO2))
                                                                                Electrical stress
                                 Before                       10Kc yc les       generates E’
Cumulative probability

                                 rad "1"
                         99.9%                                                  centers
                         84.1%                                                  These are
                                                           100Kc yc les         occupied by
                         0.1%                                                   generated by
                                    20c yc les                 rad"0"
                            -1.5           0      1.5         3           4.5   Negative charge
                                                 VTH (V)                        trapping
  A. Paccagnella 42/67                                     31 March 2004                 EWRHE
                                                                                       FG leakage
    Leakage Current [a.u.]

                                       100 P/E                                   No increase of
                                       kcycles                                 the gate leakage
                                                                               current observed
                                               10 P/E              20 P/E      after 3 Mrad (red
                                               kcycles             cycles      symbols)
                             3   3,5       4     4,5     5   5,5   6     6,5

                                         Electric Field [MV/cm]

A. Paccagnella 43/67                                     31 March 2004                 EWRHE
                            Considerations on FG memories

  Errors may take place in the FG array 
  Less probable than errors in the control circuitry (higher doses
  required) 
  More insidious: the event itself often cannot be immediately
  detected, but it can lead to:
     Reduced noise margin  increased error rate 
     Retention problems 
  Designing a rad-hard control circuitry does not automatically
  qualify FG memories for harsh environments 
  For TID and SEE:
     Prompt effects
     Long-term effects
  Device scaling and design issues still open

A. Paccagnella 44/67        31 March 2004              EWRHE
Stressing the gate oxides
                                 Assessing the gate oxide quality

        Gate oxide Breakdown has been taken as the life
        time end marker for a long time, when a single
        breakdown mode was possible (tox>7 nm)
        Different electrical tests are available to test the gate
        oxide quality and reliability in CMOS components,
        which is typically measured by looking at the
        statistical distributions of parameters such:

            Breakdown field


        …obtained through accelerated stresses at fields
        higher than during the device operation

A. Paccagnella 46/67            31 March 2004              EWRHE
                           Accelerated life tests: Electrical Stresses

     Electrical Stresses:                              Fresh
        Constant Current Stress (CCS)
        Constant Voltage Stress (CVS)
                                                   Electrical stress
        Pulsed Voltage Stress (PVS)
        Ramped Stresses                              Post-stress
   CCS          Device Under Test CVS


A. Paccagnella 47/67         31 March 2004                 EWRHE
                                        Gate oxide leakage currents

   Due to defect generation several leakage currents
   can appear in thin gate oxides after stress:
          Stress Induced leakage
          Current (SILC)



          Soft Breakdown (SB)

          Hard Breakdown (HB)

A. Paccagnella 48/67         31 March 2004                    EWRHE
                                                  Problems associated to scaling Tox

               10-6                                                High electric fields
                                   HB                              applied to the gate oxide
                                                                   may produce anomalous
                               SB                                  gate leakage:
  Jg (A/cm²)

                                                                          SILC (Stress Induced Leakage
               10-9                                                       Current)

               10-10                                                      SB (Soft Breakdown)
                                                                          HB (Hard Breakdown)
                                                       fresh       What is the end-of-life marker?
                       0   1   2        3     4    5     6     7          MOSFET may survive SB [1]
                                         Vg (V)                           and SILC, not HB
                                                                   [1]   B.E. Weir et al., IEDM 1997

A. Paccagnella 49/67                                   31 March 2004                            EWRHE
                                                                              SILC and SB vs. tox

             10-5                                             10-6
                         2.8 nm        4 nm                          2.8 nm
             10-6                                             10-7
Jg [A/cm2]

             10-7                                             10-8

                                                          Ig [A]
             10-8                                             10-9                4 nm
                                               5.2 nm         10-10
             10-9                                                                            5.2 nm
                                               SILC           10-11
             10-10                             Fresh                                         SB
                                                              10-12                          Fresh
                     0    1   2    3       4    5    6        10-13 0    1    2    3     4    5    6
                                  Vg [V]                                          Vg [V]
                     • SILC tends to become negligible in comparison with
                       the pre-stress current as tox decreases
                     • SB current is typically >> SILC
                     • HB is unlikely in ultrathin oxides due to the low VDD
 A. Paccagnella 50/67                               31 March 2004                        EWRHE
                                 Radiation effects on gate oxide

      TID: positive charge accumulates near the
      Si/SiO2 interface producing a negative VT shift,
      that tends to disappear in ultra-thin gate oxides
      proper of the last CMOS generations
      TID: Interface states are generated, stretching
      the subthreshold characteristics, increasing the
      Ids,off and reducing gm; even this effects tends to
      be ruduced in contemporary CMOS

A. Paccagnella 51/67     31 March 2004              EWRHE
                               Radiation effects and electrical stresses

        Dielectric quality and long-term reliability of gate
        oxides are tested by accelerated electrical stresses at
        high oxide fields applied to MOS capacitors: Time-
        To-Breakdown (TTB), Charge-To-Breakdown (QBD),
        Breakdown field (EBD)
        No synergetic effects between radiation damage and
        accelerated electrical stresses have been observed,
        for X-rays, g-rays, heavy ions, down to tox=7 nm and
        up to 20 Mrad(Si)
        D.M. Fleetwood et al., IEEE-TNS, 2000

A. Paccagnella 52/67             31 March 2004              EWRHE
                                                                  g radiation and electrical stresses

                      10 kGy (Si)
                                                                           Radiation damage
                                                                           from g rays has no
                      150 kGy (Si)
                      200 kGy (Si)
                                                                           impact on the oxide

               0      300 kGy (Si)
                                                                           reliability, as in thick
                                                                           Time-to-Breakdown is
                                                                           left unchanged even
                                          Co Gamma Irradiation
                                        tox = 3.2 nm, Vstress= - 5.0 V     by high radiation
               -3                                                          doses
                            10                              100

                           Time-to-Breakdown (s)                           The g induced point
                                                                           defects are not
                                                                           effective in promoting
                    J.H. Suehle et al., APL, 2002
                                                                           the oxide BD

A. Paccagnella 53/67                                   31 March 2004                     EWRHE
                                                                               Heavy ions and electrical stresses

                             Xe Irradiation, tox = 3.0 nm, Vstress = - 4.9 V
                  1                                                                               Heavy ions impact
                                                                                                  the oxide reliability!

                                                                                                  breakdown is
                                                                               Pre-irradiation    strongly reduced
                                                                               1 x 105 ions/cm2
                                                                               1 x 106 ions/cm2
                                                                                                  by the pre-existing
                                                                               1 x 107 ions/cm2   radiation damage
                       1         10           100         1000         10000                      BD takes
                                      Time-to-Breakdown (s)                                       advantage of
                                                                                                  defect clusters
                                                                                                  linked to a
                            J.H. Suehle et al., APL, 2002
                                                                                                  conductive path

A. Paccagnella 54/67                                                 31 March 2004                        EWRHE
                                                    Radiation induced Wear-Out: SB

                                                                       • tox = 2.8 nm
    10-3                                                               • Gate Area = 10-3 cm2
    10-4                                                               • Ion Fluence = 107 I ions/cm2
                                                                       SB after irradiation +
                                                                       4000s CVS @ Vg=-2V
Ig [A]

    10-8              A                                    20
                                  After                    15                  A

                                                     Ig [nA]
    10-10                         Irradiation
    10-11                                                  10
                                     fresh                                              SB onset
         0    0.5    1     1.5   2      2.5     3              5
                                                                   0        1      2      3        4
                          Vg [V]
                                                                            Time [x1000 s]
             A. Cester et al., IEEE-TNS, 2003

A. Paccagnella 55/67                     31 March 2004                                  EWRHE
                                                Radiation induced Wear-Out: HB

                                                                        • tox = 2.8 nm
    10-3                                                                • Gate Area = 10-3 cm2
    10-4                                                                • Ion Fluence = 107 I ions/cm2
                                                                    HB after irradiation + 360s
                                                                    CVS @ Vg=-3.5V
Ig [A]

    10-8               A                                      500
    10-9                                                                                 HB onset
                                  After                       400

                                                    Ig [nA]
    10-10                         Irradiation                                                        B
    10-11                             fresh                   300
         0   0.5       1    1.5   2      2.5    3             200
                                                                    0      100     200 300          400
                           Vg [V]
                                                                                 Time [s]

A. Paccagnella 56/67                      31 March 2004                                   EWRHE
                                                        Light Emission Microscopy observations

                                   CVS       CVS                              10-3
                                  Vg=-2V   Vg=-3.5V                           10-4
 Immediatelly after irradiation


                                                                     Ig [A]
                                                                              10-8                  A
                                                                                                    After Irradiation
                                                                A                     0   0.5   1    1.5 2      2.5     3
                                                                                                    Vg [V]

                                  After CVS the gate current flows across preferential paths
                                  corresponding to localized regions previously weakened by

A. Paccagnella 57/67                                      31 March 2004                                 EWRHE
                                                           LET impact on Wear-Out

                                 Vg = 4V
                                                                  Several SB
          7                 Br                                    spots are
Ig [mA]

                                                                  during the
                                                                  depending on
          4                                    Fresh              the ion LET
                                                                  (and fluence)
          6000       6200    6400        6600      6800    7000
                                 Tstress [s]

                 High LET ions produce weakened oxide regions more prone
                   to develop SB/HB under subsequent electrical stresses

A. Paccagnella 58/67                       31 March 2004                EWRHE
                                            Gate oxides: comments

        Electrical stresses enhance the radiation effects on
        ultra-thin gate oxide conductance
        Electrical/radiation stresses are typically performed
        on large area MOS capacitors, not on many small
        size MOSFETs with equivalent gate area
        Statistical tests to assess oxide quality
        Standardized test methods
        Size effects are not well reproduced by capacitors:
        perimeter/area ratio, lateral fields, side walls, LDD
        Peculiar results when reliability tests are performed
        directly on MOSFETs (such as hot electron injection),
        in particular for single ion effects, but also vs total

A. Paccagnella 59/67        31 March 2004               EWRHE
Plasma damage
                                               Plasma treatments

        Plasma-based treatments are widely used in
        fabrication processes of ULSI integrated circuits,
        being well suited to comply with the continuing
        CMOS dimensional scaling, lowering of the thermal
        budget, and increasing of interconnects levels
        Plasma etching is used to pattern many layers of
        ULSI circuits through selective and anisotropic
        During plasma treatments some damage can be
        accumulated in the dielectric layers of CMOS circuits
        Part of the plasma induced defects is annealed by
        successive thermal treatments, but part of them could
        be only passivated, reappearing after subsequent
        electrical or radiation stresses

A. Paccagnella 61/67       31 March 2004              EWRHE
                             Plasma process schematics

A. Paccagnella 62/67   31 March 2004         EWRHE
                                            Plasma damage

        In both plasma etching and deposition the
        electromagnetic field generates energetic
        atoms, chemical radicals, ions, and electrons,
        which hit the Si wafer under treatment. In
        addition, plasma-to-substrate voltage
        differences can be produced
        Plasma damage (PD) in the Si wafer can be
        exacerbated or mitigated by subsequent
        processing and by circuit and device layout

A. Paccagnella 63/67     31 March 2004          EWRHE
                                                      Gate antenna

Dense finger
(DF) antenna                                  Antenna Ratio AR:
                                         Antenna Area/Gate Area

                          Plasma damage strongly depends on gate
                       interconnect topographies, device position on
                       wafer, plasma characteristics, and processing
                          Different interconnect topographies  an
                       "antenna" (polysilicon or metal) is usually
                       connected to the gate
                          The antenna enhances the charge collected
                       by the gate during plasma processes

A. Paccagnella 64/67          31 March 2004               EWRHE
           Plasma damage: annealing, passivation and reactivation

        The gate oxide plasma damage is mostly annealed
        during subsequent thermal processing
        Part of the oxide defects are just passivated (likely by
        H atoms) and they can reappear during the device
        operating life. Passivation effects can be particularly
        insidious because they mask damage which can then
        later affect reliability
         Electrical stresses are commonly used to accelerate
        the depassivation and show up this latent damage.
        Radiation may do the same!
        The effect of plasma damage on long-time reliability
        and its impact on dielectric breakdown are still open
        questions, whose answers will becomes more and
        more crucial for ultra-thin oxides

A. Paccagnella 65/67         31 March 2004               EWRHE
                                               Radiation depassivation of PD – 12 nm

                             nMOSFET, Tox=12 nm, 8 MeV electrons 10 Mrad(SiO2)
            0.2                                                                    5.0E-09
           0.15                                                                    4.5E-09
            0.1                                                                    4.0E-09
           0.05                                                                    3.5E-09
              0                                                                    3.0E-09

                                                                     Icp,max (A)

           -0.05                                                                   2.5E-09
            -0.1                                     Min. size                     2.0E-09
                                                     Dense ¼
           -0.15                                     Dense ½                       1.5E-09
                                                                                                                Min. size
            -0.2                                     Dense fingers                 1.0E-09                      Dense ¼
           -0.25                                                                   5.0E-10                      Dense ½
                                                                                                                Dense fingers
            -0.3                                                                   0.0E+00
                     Fresh     After rad   Sweep 1    Sweep 2                                0   5    10       15               20

                   • Successive VG sweeps enhance the threshold voltage and the Si/SiO2 interface
                   defect density
                   • The positive radiation induced charge grows with the AR  depassivation and
                   positive charging of latent defects generated by the plasma treatments
                   • After the 1st VG sweep the positive charge recombines or is passivated by the
                   injected electrons

       A. Paccagnella 66/67                                 31 March 2004                                    EWRHE
                                                   Radiation depassivation of PD – 5 nm

                           pMOSFET, Tox = 5 nm, 8 MeV electrons 10 Mrad(SiO2)
             0                                                                2.50E-08

      -0.02                                                                                Minimum size
                                                                                           Dense fingers 500
                                                                                           Dense fingers 1000

                                                                Icp,max (A)

           -0.1   Minimum size
                  Dense fingers 500                                           5.00E-09
                  Dense fingers 1000
      -0.14                                                                   0.00E+00
                   Fresh               After rad      Sweep 1
                                                                                         Fresh           1              10      50
                                                                                                      Total dose [Mrad(SiO2)]

              Differences among DF and MS devices are reduced after the VG sweep, but a
            negative threshold shift is still measured  the positive charge has not been fully
            compensated, or the interface states add a positive charge masking the effects of
            the negative charge trapped in the oxide

     A. Paccagnella 67/67                                31 March 2004                                                 EWRHE
                                                   PD: a summary

        Excess oxide charge and interface states are
        observed in MOSFETs with large antenna ratio after
        irradiation with different ionizing sources
        As the oxide thickness scales down, the impact of
        depassivated latent plasma damage becomes less
        effective in modifying the MOSFET electrical
        The gate bias applied during irradiation can mitigate
        (or enhance) the effects of latent damage
        depassivation by ionizing radiation, but comparable
        effects have been observed on biased and unbiased

A. Paccagnella 68/67        31 March 2004              EWRHE
                                               Final considerations

        Synergetic effects are observed between radiation
        damage and accelerated electrical stresses on gate
        oxides, but few results are available
        The device/circuit reliability should not be assessed
        by looking separately at the electrical and radiation
        Interaction between electrical and radiation stresses
        in CMOS components must be assessed for the
        different technologies, device types, applications,…

A. Paccagnella 69/67        31 March 2004               EWRHE

To top