The Ballistic MOSFET - nanoHUB · nanoHUB.org online simulations and more Network for Computational...

43
nanoHUB.org online simulations and more Network for Computational Nanotechnology Mark Lundstrom Purdue University Network for Computational Nanoechnology Simple Theory of the Ballistic Nanotransistor

Transcript of The Ballistic MOSFET - nanoHUB · nanoHUB.org online simulations and more Network for Computational...

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

Mark LundstromPurdue University

Network for Computational Nanoechnology

Simple Theoryof the

Ballistic Nanotransistor

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

outline

I) Traditional MOS theoryII) A “bottom-up” approachIII) The ballistic nanotransistorIV) DiscussionV) Summary

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

nanoscale MOSFETs

Intel Technical J., Vol. 6, May 16, 2002.

130 nm technology (LG = 60 nm)

Low VT

I DS

(mA

/µm

)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

MOSFET IV: low VDS

VG VD0

!

ID

=W Qix( )" x

(x) =W Qi0( )" x

(0)

!

ID =W Cox VGS "VT( )µeffE x

!

Ex

=VDS

L

!

Qix( ) = "C

oxVGS"V

T"V (x)( )

ID

VDS

VGS

!

ID =W

LµeffCox VGS "VT( )VDS

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

MOSFET IV: high VDS

VG VD0

!

ID

=W Qix( )" x

(x) =W Qi0( )" x

(0)

!

ID =W Cox VGS "VT( )µeffE x

!

V x( ) = VGS"V

T( )

VGS

ID

VDS

!

ID =W

LµeffCox VGS "VT( )

2

2GS T

x

V V

L

!"E

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

velocity saturation

electric field V/cm --->

velo

city

cm

/s --

->

107

104

!

" = µE!

" ="sat

41.5V25 10 V/cm

60nm

DSV

L= ! "

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

MOSFET IV: velocity saturation

VG VD0

!

ID

=W Qix( )" x

(x) =W Qi0( )" x

(0)

!

ID

=W CoxVGS"V

T( )# sat

( )D ox sat GS TI WC V V!= "

410

x>>E

0 0.4 0.8 1.2 1.4

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

MOSFET IV: velocity overshoot

Position along Channel (µm) Position along Channel (mm)

Frank, Laux, and Fischetti, IEDM Tech. Dig., p. 553, 1992

µ

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the MOSFET as a BJT

S DG

ID

VDS

VGS

electron energy vs. position

VD≈ 0V

VD= VDD

E.O. Johnson, RCA Review, 34, 80, 1973

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the MOSFET as a BJT

JC = qDn

WB

ni2

NA

eqVBE /kBT ! e

qVBC /kBT( ) = qDn

WB

ni2

NA

eqVBE /kBT 1! e

qVCE /kBT( )

VBE!"

S

VCE!V

SD

WB! L

Dn !kBT

qµeff

IDS!W t

inv

tinv !kBT q

E S

E S =qNA

(m !1)Cox

ni

NA

!

"#$

%&

2

= e'q2( B /kBT = e

'qVT /mkBT

!S=

Cox

Cox+ C

D

VG=VG

m

BJT Theory:

( ) ( ) ( )2

/ /1 1

GS T B DS Bq V V mk T qV k Tb

D eff ox

k TWI C m e e

L qµ

!" #" #== ! !$ %$ %

& ' & '

MOSFET Theory:

eqn. (3.36) on p. 128 of Fundamentals of Modern VLSI Devices,Yuan Taur and Tak Ning, Cambridge Univ. Press, 1998.

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the MOSFET as a BJT

VGS

log 1

0 ID

S

( )/~

GS T Bq V V mk Te

!

( )~GS TV V!

above threshold:

( )i ox GS TQ C V V= !

/ 2~ S Bq k T

iQ e!

E.O. Johnson, RCA Review, 34, 80, 1973

( )~ lnS GS T

V V! "

( )/~ ~S B

k T

D GS TI e V V

!"

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

Outline

I) Traditional MOS theoryII) A “bottom-up” approachIII) The ballistic nanotransistorIV) DiscussionV) Summary

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

Gate

EF1 EF1-qVD( )SCF

D E U!

1

1

!"

=h

!

"2

=h

#2

a general view of nano-devices

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

“top of the barrier model”

ε(x)

EF1EF1-qVD

(0)CE

ener

gy

position

contact 1 contact 2

1 2! !

" #= = =

h h

xL/

‘device’

LDOS

!

L

FB

SCF C SU E q!= "

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

filling states from the left contact

1

1

!"

=h

( )SCF

D E U!

0

1

1

( )( ) N E NdN E

dt !

"=

0

1 1( ) ( ) ( )SCFN E D E U f E= !

EF1

Gate

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

filling states from the right contact

!

"2

=h

#2

µ2

0

2

2

( )( ) N E NdN E

dt !

"=

0

2 2( ) ( ) ( )SCFN E D E U f E= !

Gate

EF1-qVD( )SCF

D E U!

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

steady-state

0 0

1 2

1 2

( )0

N N N NdN E

dt ! !

" "= + =

[ ]1 1 2 2( ) ( ) ( ) ( )N D E f E D E f E dE= +!

( ) ( ) ( )2 1

1

1 2 1 2

SCF SCFD E D E U D E U

! "

! ! " "# $ = $

+ +

(ballistic)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

steady-state current

( )00

21

1 2

D

N NN NI

! !

" ""= =

( )1 2

2( ) ( ) ( )D

qI M E f E f E dE

h= !"

( )1 2

1 2 1 2

( )( ) ( )

2

hD EM E D E

! !"

# # ! !$ =

+ +

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

NEGF theory

Non-equilibrium Green’s Function Approach (NEGF)

S. Datta, IEDM Tech. Dig., 2002

device

!

["1]

!

[H]

!

["2]

!

["S]

Gate

EF1 EF2

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

outline

I) Traditional MOS theoryII) A “Bottom-up” approachIII) The ballistic nanotransistorIV) DiscussionV) Summary

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

assumptionsen

ergy

positioncontact 1 contact 2

1 2

x

! !"

= =L

ε(x)

EF1 EF1-qVDLDOS

!

L

1) 2D, planar MOSFET

2) 1 subband occupied

3) parabolic E(k)( )

*

2( )

C

mD E W E E

!= " #

hL

1 2( ) ( ) ( ) 2D E D E D E= =

( )1

*2 2 /x E m

!" #

= =L L

*2( )

W m EM E

!=

h

FB

SCF C C SU E E q!" = #

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

procedureen

ergy

positioncontact 1 contact 2

1 2

x

! !"

= =L

ε(x)

EF1 EF1-qVDLDOS

!

L

1) assume a ψS(sets top of the barrier energy)

2) fill states

3) self-consistentelectrostatics

4) evaluate current

N N N+ !

= +

( )1 2

2( ) ( ) ( )D

qI M E f E f E dE

h= !"

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

filling states (ballistic)

N = D1(E) f

1(E) + D

2(E) f

2(E)[ ]! dE

[ ]*

1 22( ) ( )

2

mN W f E f E dE

!= +"

hL

[ ]20 1 0 2( ) ( )

2

D

F F

NN W ! != +L F F

*

2 2

B

D

m k TN

!=

h

!F1 = EF1 " EC

FB + q# S( ) kbT

!F2 = EF1 " qVD " EC

FB + q# S( ) kbT

k

E(k)

!

h2k2

2m*

EF1EF2

C

FB

SE q!"

N at the top of the barrierdepends on: VG (through ψS) VD (through EF2)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

filled states in equilibrium

* 2( ) /( ) / / 2

( ) /

1( )

1F C BF B B

F B

E E k TE E k T m k T

E E k Tf E e e e

e

!""

"= # = $

+2 2 2 *( ) / 2

( , ) x y Bk k m k T

x yf k k e+

!h

f0 f (kx, ky)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

filling states under bias

Increasing VDS

X (nm) --->-10 -5 0 5 10

ε1 vs. x for VGS = 0.5V

ε 1 (e

V) -

-->

f (kx, ky)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the ballistic current

[ ]1 2*

2( ) ( ) ( )D

qI M E f E f E dE

m= !"

*2( )

W m EM E

!=

h

[ ]21/ 2 1 1/ 2 2( ) ( )

2

DD T F F

qNI W! " "= #F F

*

2B

T

k T

m!

"=

D nI WQ !"

n

NQ

W=

L

alternatively:

1/ 2 2 1/ 2 1

0 2 0 1

1 ( ) / ( )

1 ( ) / ( )

F F

T

F F

! !" "

! !

# $%= & '

+( )%

F F

F F( )( )

1/ 2 1

*

0 1

2 FB

T

F

k T

m

!"

# !

$ %= & '& '

( )%

F

F

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

carrier velocity in a ballistic MOSFET

Increasing VDS

X (nm) --->

EC (e

V)

---> Increasing VDS

-10 -5 0 5 10

ε1 vs. x for VGS = 0.5V

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

velocity saturation in a ballistic MOSFET

Increasing VDS

X (nm) --->-10 -5 0 5 10

ε1 vs. x for VGS = 0.5V

!

"(0) # ˜ " T

ε 1 (e

V) -

-->

( )( )

1/ 2 1

*

0 1

2 FB

T

F

k T

m

!"

# !

$ %= & '& '

( )%

F

F“injection velocity”

VDS --->

υ(0)

υin

j (10

7 cm

/s)

--->

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

IV of a ballistic MOSFET

[ ]21/ 2 1 1/ 2 2( ) ( )

2

DD T F F

qNI W! " "= #F F

[ ]20 1 0 2( ) ( )

2

D

F F

NN W ! != +L F F

( )1 1

FB

F F C S bE E q k T! "= # +

( )2 1

FB

F F D C S bE qV E q k T! "= # # +

Key equations We must express ψSin terms of

VG (1D electrostatics)

or

VG and VD (2D electrostatics)

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

1D MOS electrostatics (above threshold)

[ ]21/ 2 1 1/ 2 2( ) ( )

2

DD T F F

qNI W! " "= #F F

[ ]20 1 0 2( ) ( )

2

D

F F

NN W ! != +L F F

Key equations

( )ox GS T

qNC V V

W! =

L

(1)

(2)

(3)

equations (1), (2), and (3) give…

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

1D MOS electrostatics (above threshold)

1/ 2 1

1/ 2 1

0 1

0 1

( / )1

( )( )

( / )1

( )

F DS B

FDS ox GS T T

F DS B

F

qV k T

I WC V VqV k T

!

!"

!

!

#$ %#& && &

= # '( )#& &+& &* +

%

F

F

F

F

( ) [ ]20 1 0 1( ) ( / )

2

Dox G T F F DS B

NC V V qV k T! !" = + "F F

for non-degenerate statistics:/

/

1( )

1

DS B

DS B

qV k T

DS ox GS T T qV k T

eI WC V V

e!

"

"

# $"= " % &

+' (

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the ballistic MOSFET

1/ 2

1/ 2

0

0

( )1

( )( )

( )1

( )

F DS

F

DS ox GS T T

F DS

F

U

I WC V VU

!

!"

!

!

#$ %#& && &

= # ' (#& &+& &) *

%

F

F

F

F

VDS

!

IDS

(on) =W Cox

˜ " TVGS#V

T( )

quantum conductance

VG! V

T( )

"

IDS

!

" M2q2

h

K. Natori, JAP, 76, 4879, 1994.

idealelectrostatics

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

electrostatics (subthreshold and 2D)

CG

VS VD

CDCS

VG

!

Q = "qN

!S

( )SG SDS G D S

qNC CCV V V

C C C C

!!

" " " "

# $ # $ # $= + + %& ' & ' & '

( ) ( ) ( )

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

procedure

for a given VG, VD:

1) guess ψS2) fill states

3) compute improved ψS

4) iterate between (2) and (3)

5) compute current

6) select new VG, VD, and go to 1

( )SG SDS G D S

qNC CCV V V

C C C C

!!

" " " "

# $ # $ # $= + + %& ' & ' & '

( ) ( ) ( )

[ ]21/ 2 1 1/ 2 2( ) ( )

2

DD T F F

qNI W! " "= #F F

see FETToy at

www.nanohub.org

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

outline

I) Traditional MOS theoryII) A “Bottom-up” approachIII) The ballistic nanotransistorIV) DiscussionV) Summary

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

the quantum capacitance

VG

insC

S QC C=

( )Gate G TQ C V V= !

ins Q

Gate

inc Q

C CC

C C=

+

( )( )2 *

2~

S

Q D F

S

qnC q D E m

!

" #= =

"S

!

if ,Q ins Gate insC C C C>> !

ID � Q!

inj

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

bandstructure effects in nano-MOSFETs

a

(001)

(100)

(010)

(111)

(110)

-tight binding model (sp3d5s*) (Boyken, Klimeck, et al.)

-bulk, UTB, nanowire MOSFETs

-Si, Ge, SiGe, GaAs, InAs, … (strained or unstrained) (heterostructure channels)

2 2

*( )

2

kE k

m=h

( ) : tabulatedE k

Top-of-the-barrier model

analytical numerical

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

scattering in nano-MOSFETs

measured ballistic

Chau et al, IEDM Technical Digest,2000, pp 45 -48

Intel 30nm bulk MOSFET

MOSFETs operate at ≈ 50% of their ballistic limit

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

relation to traditional MOSFET theory

!

ID =W

LµeffCox VGS "VT( )VDS

( )2

2D eff ox GS T

WI C V V

Lµ= !

low VDS

high VDS (long channel)

1/ 2 1

1/ 2 1

0 1

0 1

( / )1

( )( )

( / )1

( )

F DS B

FDS ox GS T T

F DS B

F

qV k T

I WC V VqV k T

!

!"

!

!

#$ %#& && &

= # '( )#& &+& &* +

%

F

F

F

F

ballistic MOSFET

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

relation to traditional MOSFET theory

( )( )21 2

1 2

( )2( ) ( )

2

DD

hD EqI f E f E dE

h ! !

" #= $% &

+' ()

ballistic transport:( )

1 2*

2 2 /x E m

! !" #

= = =L L

2

1 2

2 effD! != =

Ldiffusive transport:

see: “The Ballistic MOSFET,” unpublished notes by M.S. Lundstrom, 2005

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

Outline

I) Traditional MOS theoryII) A “Bottom-up” approachIII) The ballistic nanotransistorIV) DiscussionV) Summary

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

summary

1) A ballistic, “top-of-the barrier” model for the MOSFETis easy to formulate.

2) The ballistic model provides new insights into thephysics of nanoscale MOSFETs.

3) Although not comprehensive, the top-of-the-barrierballistic model should prove useful in exploringnew materials and structures for ultimate CMOS.

nanoHUB.orgonline simulations and more

Network for Computational Nanotechnology

references

the ballistic model:

scattering in nanotransistors:

Mark Lundstrom, “The Ballistic Nanotransistor,” unpublished notes, 2005.

Anisur Rahman, Jing Guo, Supriyo Datta, and Mark Lundstrom, “Theory ofBallistic Nanotransistors,” IEEE Trans. Electron. Dev., 50, 1853-1864, 2003.

Mark Lundstrom and Zhibin Ren, “Essential Physics of Carrier Transport inNanoscale MOSFETs,” IEEE Trans. Electron Dev., 49, pp. 133-141, 2002.