45
Disclaimer: This course was prepared, in its entirety, by Adam Teman. Many materials were copied from sources freely available on the internet. When possible, these sources have been cited; however, some references may have been cited incorrectly or overlooked. If you feel that a picture, graph, or code example has been copied from you and either needs to be cited or removed, please feel free to email adam [email protected] and I will address this as soon as possible. Digital Integrated Circuits (83-313) Lecture 12: The NanoScaled MOSFET Semester B, 2016-17 Lecturer : Dr. Adam Teman TAs : Itamar Levi, Robert Giterman 19 June 2017

Digital Integrated Circuits · 2017. 6. 20. · Leakage Summary •Subthreshold Leakage: •I DS >0 when V GS

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

  • Disclaimer: This course was prepared, in its entirety, by Adam Teman. Many materials were copied from sources freely available on the internet. When possible, these sources have been cited;

    however, some references may have been cited incorrectly or overlooked. If you feel that a picture, graph, or code example has been copied from you and either needs to be cited or removed,

    please feel free to email [email protected] and I will address this as soon as possible.

    Digital Integrated Circuits(83-313)

    Lecture 12:

    The NanoScaled MOSFETSemester B, 2016-17

    Lecturer: Dr. Adam Teman

    TAs: Itamar Levi, Robert Giterman

    19 June 2017

    mailto:[email protected]

  • Lecture Content

    2

  • Leakage in NanoScaled Transistors

    3

  • Leakage in Nanoscaled Transistors

    • Transistors that are supposed to be off actually leak!

    4

  • Main Types of Leakage

    5

    P-substrate

    n+n+

    Source Drain

    Polysilicon Gate

    Oxide

    subthreshold

    Ga

    te

    Punchthrough

    GS TV V

    subI e

    1,gate GC

    ox

    I Vt

    GIDL DGI V

    DSVDIBL e

  • Subthreshold Leakage

    • When VGS

  • Subthreshold Leakage

    • Let’s make things easier:

    • Remember that:

    • And we now defined the threshold voltage according to current:

    • So the edge condition requires:

    • And we can now calculate subthreshold current as:

    7

    subGS T

    T

    V Vn

    DS

    WI Const e

    L

    100GS T

    T

    V Vn

    sub

    WI nA e

    L

    100nADS GS TW

    I V VL

    0

    sub 100nATGS T

    n

    DS V V

    W W WI Const e Const

    L L L

    Subthreshold Leakage

  • • An even easier way is to define the subthreshold slope:

    • And Ioff, which is defined as the current when VGS=0 is:

    Subthreshold Leakage

    8

    100

    100 10

    GS T

    T

    GS T

    V Vn

    sub

    V VS

    WI nA e

    L

    W

    L

    100 10TV

    Soff

    WI nA

    L

    ln 10 1 2.3dep Tox

    CkTS n

    q C

    Subthreshold Leakage

  • Subthreshold Leakage

    • So subthreshold leakage is:

    • Exponentially dependent on VGS.

    • Exponentially dependent on VT.

    • S is the subthreshold swing coefficient.

    • Optimally, (Sopt)-1=60 mV/dec

    • Realistically (S)-1≈100 mV/dec

    9

    ln10 0.06TS n

    1dep

    oxe

    Cn

    C

    Subthreshold Leakage

  • Subthreshold Leakage

    Example:

    • We want to design a transistor with:

    • What is the minimum VT?

    10

    410on

    off

    I

    I

    mV60

    decS

    Subthreshold Leakage

  • Impact of DIBL

    • DIBL causes an additional exponential increase

    in subthreshold leakage with VDS.

    11

    90nm technology.

    Gate length: 45nm

    Intel, T. Ghani et al., IEDM 2003

    0 (1 )

    GS T DS DS

    t t t

    V V V V

    n n

    subI I e e e

    Subthreshold Leakage

  • Subthreshold Dependence on Temperature

    • This is rather complex, as mobility degrades with temperature and

    other device values such as flatband voltage are temperature

    dependent.

    • Altogether, subthreshold leakage rises

    exponentially with temperature.

    12

    GS T

    T

    V V

    subI e

    TkT

    q

    Subthreshold Leakage

  • Temperature Inversion

    • Standard approach to temperature effect on delay:

    • BUT!

    • VT decreases by as much as -3mV/°C

    13

    1

    I

    T

    1pdt T

    1T

    pd T

    V T

    t V

    pdt T

    Subthreshold Leakage

  • Gate Leakage

    • Two mechanisms:• Direct tunneling (dominant)

    • Fowler Nordheim tunneling

    • Exponentially Dependent on:• Gate Voltage (VG)

    • Oxide Thickness (tox).

    • Non-dependent on temperature.

    • Much stronger in nMOS than pMOS (higher barrier for holes)

    • Minimum tox=1.2nm!!!

    14

    Subthreshold Leakage

    Gate Leakage

  • Gate Induced Drain Leakage

    • GIDL current flows from the drain to the substrate.

    • Caused by high electric field under the gate/drain

    overlap, causing e-h pair creation.

    • Main phenomena is Band-to-Band Tunneling

    15

    Subthreshold Leakage

    Gate Leakage

    GIDL

  • Diode Leakage

    • JS=10-100pA/μm2 @25°C for 0.25μm CMOS.

    • JS doubles for every 9°C

    • Much smaller than other leakages in deep sub-micron.

    16

    DL SI J A

    Subthreshold Leakage

    Gate Leakage

    GIDL

    Diode Leakage

  • Punchthrough

    • As VDS grows, so does the drain depletion region,

    and the channel length decreases.

    • In severe cases, the source and drain are connected

    causing non-controllable leakage current.

    17

    rjWs

    xorjWD

    L +VD

    rjWs

    xorjWD

    L++VD

    e-

    Subthreshold Leakage

    Gate Leakage

    GIDL

    Diode Leakage

    Punchthrough

  • Leakage Summary• Subthreshold Leakage:

    • IDS>0 when VGS0 due to direct tunneling through the oxide.

    • Grows with VGB

    • Gate Induced Drain Lowering (GIDL):• IDB>0 due to high electric field in the GD overlap region.

    • Band to Band tunneling.

    • Reverse Biased Diode Leakage• ISB, IDB Due to diffusion and thermal generation.

    • Punchthrough:• IDS due to drain and source depletion layers touching.

    18

    Subthreshold Leakage

    Gate Leakage

    GIDL

    Diode Leakage

    Punchthrough

  • Latchup

    • Latchup is more of a design problem than a leakage current.

    • “Thyristors” created by parasitic BJT transistors

    can turn on and short VDD and GND.

    • This requires power down at the least, and

    sometimes causes chip destruction.

    19

    Subthreshold Leakage

    Gate Leakage

    GIDL

    Diode Leakage

    Punchthrough

    Latchup

  • Secondary Effects and Solutions

    20

  • Problem: Mobility Degradation

    • Reminder: we have degraded mobility due to:

    • Velocity Saturation

    • Surface Scattering

    • …and in general, we want more speed!

    • There are a number of solutions that are currently

    used or are being developed:

    • Strained Silicon

    • Miller Index

    • Different Materials

    21

    Mobility Degradation

  • Problem: Surface Scattering

    • The mobility at the surface is vertical field (VG) dependent.

    • The stronger the field, the more carriers

    “hit” the interface and scatter. 0

    1 GS TV V

    Mobility Degradation

  • Solution: Strained Silicon

    23

    Chang, IEDM 2005

    Enhanced Channel Mobility

    Mobility Degradation

  • Solution: Improved Channel Materials

    24

    Mobility Degradation

  • Solution: Silicon Orientation

    25

    Mobility Degradation

  • Problem: Serial Resistance

    • The resistance of the Source and Drain areas, especially

    with Lightly Doped Drain (LDD),

    can have a large impact (>15%)

    on transistor conductance

    )(1 0

    0

    tgs

    sdsat

    dsatdsat

    VV

    RI

    II

    0 0dsat dsat dsat s dV V I R R

    K. Kuhn, IEDM 2008

    Serial Resistance

    Mobility Degradation

  • Solution: Salicides

    • Salicide = Self Aligned Silicide

    • A low resistance contact is formed

    on the diffusion and polysilicon

    surface through annealing.

    27

    Serial Resistance

    Mobility Degradation

    Source: Wikipedia

  • Problem: Hot Carrier Effects

    • Reminder:• Carriers in strong electric fields jump over the gate energy barrier

    and get stuck in the oxide.

    • This causes VT to change over time.

    • Happens mainly close to the drain

    where strong fields exist.

    • Solution:• To solve this, Lightly Doped Drains (LDD) are used.

    • The N- area reduces the field

    gradient near the drain.

    • N+ is needed for ohmic contacts.

    28

    N gradient

    n-n+

    Serial Resistance

    Mobility Degradation

    Hot Carriers

  • Problem: Subthreshold Leakage

    • Reminder:• There is a finite number of free carriers in the channel when VGS

  • Solution: MTCMOS

    • A very common solution, already available in most PDKs for

    many generations is multi-threshold voltage devices.• High-VT (HVT) devices are slow but have lower leakage

    • Low-VT (LVT) devices are fast but very leaky

    • Nominal/Standard/Regular-VT (NVT/SVT/RVT) transistors

    are best for most operations.

    • This approach provides an easy way to trade off

    high-performance and low leakage.

    • Standard cells are often designed with identical footprints in

    order to exchange various VT options seamlessly.

    30

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Body Biasing

    • As we have previously seen, the body voltage of the transistor

    affects the threshold voltage of the transistor.

    • The body voltage of PMOS devices, residing in a shared NWELL

    can be set without the need for special process steps.

    • Changing the body voltage of NMOS devices requires the use of an

    isolated PWELL (IPW) inside a Deep NWELL (DNW) area.

    • The effectiveness of body biasing has thoroughly degraded with process

    scaling.

    • New technologies, such as Fully-depleted Ultra-thin body and buried oxide

    Silicon-on-Insulator (FD-SOI or UTBB-SOI) have given this technique “new life”.

    31

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Vertical Dimensions

    • A better model of VT, taking into account both DIBL and Roll-Off shows a strong dependence on the transistor’s vertical dimensions:

    • tox – gate thickness• Wdep – depletion width• Xj – S/D junction depth

    • Reduction of each of these provides a smaller Lmin and other advantages.

    • smaller tox means better transconductance.

    • smaller Xj means lower S/D capacitance

    • smaller Wdep means better drain-channel isolation32

    0.4 dL

    l

    t t long dsV V V e

    3d ox dep jl t W X

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Vertical Dimensions

    • Thinner oxide is the most important

    aspect of scaling, as it allows better

    gate control.

    • Shallow junctions are achieved by:• High Substrate Doping (not wanted)

    • Lightly Doped Drains (LDD)

    • Metal S/D

    • What about Wdep?33

    3d ox dep jl t W X

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Vertical Dimensions

    • Wdep can be reduced by

    “Retrograde Doping”• Reduces impurity scattering

    (improves mobility).

    • Wdep does not vary with VSB.• Causes a linear body effect.

    • Reduces punchthrough

    34

    data model

    -2 -1 0 1 2 Vsb (V)

    NFET

    PFET

    Vt 0

    Vt0

    0.6

    -0.2

    -0.6

    0.4

    -0.4

    Vt (V)

    0.2

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

    3d ox dep jl t W X

  • Solution: Multi-gate Transistors

    • To gain better control of the channel

    (better subthreshold slope) use multiple gates:

    35

    Double-gate MOSFETFinFET

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Multi-gate Transistors

    36

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot CarriersA. Planar bulk MOSFET

    B. SOI MOSFET

    C.Tri-gate SOI nanowire

    MOSFET

    D.Tri-gate Bulk FinFET

    E. Tri-gate SOI FinFET

    F. Pi-gate SOI nanowire MOSFET

    G.Omega-gate SOI nanowire

    MOSFET

    H.Horizontal Gate-all-around (GAA)

    nanowire MOSFET

    I. Vertical GAA nanowire MOSFET

    Ferain, Nature 2011

  • Solution: Silicon-On-Insulator

    • Silicon on Insulator (SOI) is a better

    (though more expensive) way to make transistors:• Fixed depletion widths (Wdep, Xj) – lower Roll-Off/DIBL

    • No body effect.

    • No Punchthrough.

    • Junction to substrate parasitic capacitance small.

    • No latch-up between NMOS and PMOS (no substrate)

    37

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: Silicon-On-Insulator

    • Lower diffusion capacitance:

    • Raised Source/Drain for reduced resistivity.

    38

    Serial Resistance

    Subthreshold Leakage

    Mobility Degradation

    Hot Carriers

  • Problem: Gate Leakage

    • Reminder:

    • Gate leakage is exponentially dependent on tox.

    • A thin oxide reduces Roll-Off.

    • A thin oxide provides higher transconductance (Cox)

    • A thin oxide reduces the subthreshold swing:

    • Major Problem:

    • Oxide thickness reached 1.2nm (5 atomic layers) in 65nm.

    • Gate leakage is intolerable under tox=1.5nm

    • The end of scaling???

    39

    1

    ln10TS

    n1

    dep

    oxe

    Cn

    C

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: High-K Dielectrics

    • Higher K (ε) means Cox is increased, therefore:• Transconductance is higher.

    • n=1+CD/Cox is lower Better gate control.

    • So we can use thicker gates to eliminate tunneling

    without losing control or drive strength.

    40

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Mobility Degradation

    Hot Carriers

  • Solution: High-K Dielectrics

    • HfO2 has a relative dielectric constant (k) of ~24, six times large than that of SiO2. • For the same EOT, the HfO2 film presents a much thicker

    (albeit a lower) tunneling barrier to the electrons and holes.

    • Toxe can be further reduced by introducing metal-gate technology since the poly-depletion effect is eliminated.

    • The difficulties of high-k dielectrics:• Chemical reactions between them and the silicon

    substrate and gate

    • Lower surface mobility than the Si/SiO2 system

    • Too low a VT for pMOS(as if there is positive charge in the high-k dielectric).

    • Long-term reliability41

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Mobility Degradation

    Hot Carriers

  • Problem: Punchthrough Currents

    • Reminder:

    • Drain and Source depletion regions “connect”

    to each other deep underneath the channel.

    • Solutions:

    • Halo Implants

    • LDD

    • Retrograde Doping

    • SOI

    42

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Punchthrough

    Mobility Degradation

    Hot Carriers

  • Solution: Halo Implants

    • p+ implants suppress the

    source/drain depletion regions.

    • However, the also cause the

    Reverse Short Channel Effect.

    43

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Punchthrough

    Mobility Degradation

    Hot Carriers

  • Problem: Latch Up

    • Reminder:

    • A Thyristor is created because of voltage drop

    over the substrate (or well).

    • In general, cases of forward biased diodes

    are often called latchup.

    • Solutions:

    • Body Taps.

    • SOI.

    44

    Serial Resistance

    Subthreshold Leakage

    Gate Leakage

    Punchthrough

    Latchup

    Mobility Degradation

    Hot Carriers

  • Further Reading• J. Rabaey, “Digital Integrated Circuits” 2003, Chapters 2.5, 3.3-3.5

    • Weste, Harris “CMOS VLSI Design”, Chapter 7

    • C. Hu, “Modern Semiconductor Devices for Integrated Circuits”, 2010, Chapters 4-7 http://www.eecs.berkeley.edu/~hu/Book-Chapters-and-Lecture-Slides-download.html

    • Tzividis, et al. “Operation and Modeling of MOS Transistor” New York Press 2011. Chapters 1-5

    • E. Alon, Berkeley EE-141, Lecture 9 (Fall 2009) http://bwrc.eecs.berkeley.edu/classes/icdesign/ee141_f09/

    • M. Alam, Purdue ECE-606 – lectures 32-38 (2009) http://nanohub.org/resources/5749

    • A. B. Bhattacharyya “Compact MOSFET models for VLSI design”, 2009,

    • T. Sakurai, “Alpha Power-Law MOS Model” – JSSC Newsletter Oct 2004

    • Managing Process Variation in Intel’s 45nm CMOS Technology, Intel Technology Journal, Volume 12, Issue 2, 2008

    • Berkeley “BSIM 4.6.4 User’s Manual” http://www-device.eecs.berkeley.edu/~bsim/BSIM4/BSIM464/BSIM464_Manual.pdf

    45