MOSFET Modeling & Simulation with MATLAB Dr.

IsmailSaad

Lecturer Electrical&ElectronicsEngineeringProgram SchoolofEngineering&IT Universiti MalaysiaSabah(UMS)

Email:[email protected]

The MOS capacitor

(a) Physicalstructureofann+ Si/SiO2/pSiMOScapacitor (b) crosssection (c) theenergybanddiagramunder chargeneutrality (d) theenergybanddiagramat equilibrium(notethatthesurface oftheptypesubstratenearthe oxideinterfacehasbecomeweakly inverted).

EnergybanddiagramsfortheMOScapacitorwith thechargedistributionsfor3 biasconditions.

(a) Equilibrium:Electronsfromthe n+ gatetransfertothepSi substrate,resultinginapositive gateandanegativedepletion regioninthesubstrate. (b) Accumulation:Anegativevoltage isappliedtothegatewithrespect tothesubstratesuchthatholes accumulateatthesiliconto silicondioxideinterface. (c) Applied+ve 2V (d) Withtime,electronsgeneratedin thetransitionregionaretrapped inthepotentialwellatthe interfaceuntilsteadystateis reach

iac = 2fC vac

(a)Circuitformeasuringthe capacitanceofaMOScapacitor.

1 1 1 = + C Cox Cs
(b)Capacitancevoltage(CV) characteristicforaMOScapacitoratlow andhighfrequencies.

Schematicdiagramofnchannel siliconbasedMOSFETstructure
ThechannelwidthW,lengthL,and oxidethicknesstox areshown ThesymbolsS,G,D,andB represent thesource,gate,drain,andsubstrate (body)respectively.

(a) CrosssectionofannchannelFET (b)Energybanddiagramatequilibrium

Accumulation:Thechannelchargeaccumulatesinthebulk neartheoxideinterface.Inthiscasethechannelcharge consistsofelectrons.

Inversion:Theband bendingatthesurface ofthesemiconductor. Atthreshold,theFermi levelisasfarabove theintrinsiclevelat thesurface(lefthand edge)asitisbelowthe intrinsiclevelinthe bulk.

TheIDVDS characteristicsofatypical MOSFET.Thethresholdvoltageis0.5V

ParallelPlateCapacitorVs. MOSCapacitor

Q = CV C =

s A
d

o A
d

Q C

= C

(V GS

V T V ch

C s o QA = C AV C A = = = A d d

ox o ox C = = = A t ox t ox
s

qn

= =

ox
t ox

(V GS (V GT

V T V ch V ( y))

ox
t ox

QuantumChannel
(a) Squarewellvs MOSFET triangularchannel (b) Classicalvs quantum channel (c) Liftingofconduction bandbyquantum energy (d) Thewavefunction peaksatadistance awayfromthe interface (e) Classicalandquantum electronconcentration profile(ns=1012 cm2)

EffectiveCapacitancedueto QuantumEffect(QE) Si 2 Eo CQ = tQ xo = tQ 3 qET

0.9

1 = Cox CG = Cox 1+

Cox 1 + C Q 1 ox tQ

RELATIVE CAPACITANCE

1 1 1 = + CG Cox CQ

10 nm 0.85

0.8 5 nm 0.75

Si tox

0.7

0.65 10

15

20

25

30

35

40

45

50

ELECTRIC FIELD (MV/m)

GateFieldInducedMobility Degradation DuetoQE

lf =
1 + (VGT Vch )

--

o e (VGT Vch )

lf =

1 + (VGS VT Vch )

GateFieldInducedMobility Degradation DuetoQE

Example : = 0 . 13 V
1

t ox = 5 nm V T = 1 V 0
12

cm 2 = 480 V .s

C ox =

ox
t ox

3 . 9 8 . 85 10 5 10
9

F m = 6 . 9 10

F m2

V GS = V T

V GT

V GS = 3 V

V GT

cm 2 = 0 l f = 0 = 480 V .s cm 2 480 cm 2 V . s = 2V lf = = 380 1 + 0 . 13 V 1 ( 2 V ) V .s

V GS = 5 V

V GT = 4 V

lf

cm 2 480 cm 2 V . s = = 316 1 1 + 0 . 13 V ( 4 V ) V .s

Channel electron mobility and velocity (v = E) as a function of lateral field for VGS = 1.42 V

v=

EL 1+ Ec v sat

lf EL
eff =
EL 1+ Ec

lf

Ec =

dV EL dy

ModelingnanoMOSFET IDVD Characteristics

CrosssectionviewedofatypicalNchannelMOSFETwhichshowsthe transverseET andlongitudinalEL electricfieldandoxidethicknesstOX along thechannellengthL

In a nanoscale channel the validity of Ohms law that predict a linear drift velocity response to the applied electric field, = 0E is revealed to lost its supremacy as the applied electric field is now much higher than the channel critical electric field Ec. The drift velocity is now can be represented as below:

wherelf isthelowfieldOhmic mobility,theEc wasrelatedtothesaturation velocitysat andlf andthelongitudinalelectricfieldEL canbeapproximateasthe gradientofappliedvoltageValongthenegativeydirectionofthechannel:

L = lf L 1+ c

(1)

sat c = lf

ET

y=0 V=0
EL

y=L V=VD

dV ( y ) L dy

(2)

Thecurrentinthenanoscale channelisafunctionofcarrierconcentrationn, the chargeq andthevelocity ofcarrierintheinversionchannelperunitareaA:

I = nqvA

(3)

In a 2D channel, the area A of the inversion channel is a function of the channel width W and its thickness xi and the n is replaced by carrier sheet concentration ns that are the number of carrier per meter square given by:

ns = nxi (# / m 2 )
Equation(3)isnowwrittenas:

(4) (5)

I D = ns qvW
Qi = ns q = CG (VGT V ( y ))

Thedensityofinversionchargeofacurrenttoflowinthe2Dchannelisgivenby:

(6)

whereVGT =VGS VT thedifferenceofanappliedgatevoltageVGS andthreshold voltageVT ,CG isthegatecapacitanceandV(y) isthechannelvoltageiny direction. Byinvokingeq.(1)and(6)intoeq.(5),thedraincurrentinthe2Dinverted channelisrepresentedas:

Thisdraincurrenthastobeintegratealongydirectionfromy=0 toy=L or fromappliedvoltageV=0 toV=VDS thedrainvoltageintheinvertedchannel asshowninfig.below:

L I D = CG (VGT V ( y ) )lf W L 1+ c

(7)

ET

y=0 V=0
EL

y=L V=VD

Byrearrangeeq.(7)andintegrateit:
L L I D (1 + )dy = CG (VGT V ( y ) )lf LWdy c 0 0 L

(8)

ByreplacingEL =dV/dy asineq.(2)intoeq.(8):

1 dV dV I D (1 + )dy = CG (VGT V ( y ) )lf W dy c dy dy 0 0

Thus,eq.(9)simplifyto:

(9)

I D (1 +
0
L / VD

ID

1 1 2 I D .L(1 + VD ) = CG lf W VGTVD VD c .L 2

1 dV )dy = CG lf W (VGT V ( y ) )dV c dy 0 V 1 dV (1 + )dy = CG lf W (VGT V ( y ) )dV c dy 0 2 1 VD I D ( L + VD ) = CG lf W VGTVD 2 c

D

VD

(10)

Rearrangeeq.(10)anddefinedthecriticalvoltageVc as:

sat Vc = c .L = L lf
1 2 VGTVD VD CG lf W 2 ID = VD L 1+ Vc

(11)

Therefore,thedraincurrentequationofananoscale channelwhenVDS <VDsat is givenas:

VD VDsat

(12)

ThedrainsaturationcurrentIDsat ontheonsetofthevelocitysaturationsat whenVDS =VDsat isnowgivenby:

I Dsat = CG (VGT VDsat ) satW

VD VDsat

(13)

Forlongchannel(LC)thedrainvoltageVD issmallerthanthecriticalvoltageVc (VD <<Vc),thereforethedraincurrentequationisgivenby:

ID =

CG lf W L

1 2 (VGTVD VD ) 2

(14)

Determination of VDsat and IDsat

TodeterminedVDsat ,thepointonwhichthevelocitysaturate,eq.(12)and (13)mustbereconcilebymakingID thesameandmakeVD =Vdsat:

1 2 VGTVDsat VDsat CG lf W 2 = C (V V ) G GT Dsat satW V L 1 + Dsat Vc

(15)

Rearrangeeq.(15)andsolvingforVDsat:

1 2 sat VDsat ) VGTVDsat VDsat = (VGT VDsat )( L)(1 + 2 lf Vc VDsat 1 2 ) VGT VDsat VDsat = (VGT VDsat )(Vc )(1 + Vc 2
VGT VDsat 1 2 VDsat ) VDsat = (VGTVc VDsatVc )(1 + 2 Vc

VGTVDsat

1 2 2 VDsat = VGTVc + VGTVDsat VDsatVc VDsat 2

1 2 2 VDsat + VDsat + VDsatVc VGT Vc = 0 2

2 Dsat

+ 2VcVDsat 2VGT Vc = 0

(16)

Equation(16)isaquadraticandsolvingitwillgivetheVDsat as:

VDsat =
VDsat

2Vc

(2Vc )2 + 4 2VGTVc
2

2VGT = Vc 1 + 1 + Vc

VDsat

2VGT = Vc 1 + 1 Vc

(17)

TheIDsat canbedeterminedbyusingeq.(16)andsolvingitusingeq.(13):
2 VDsat + 2VcVDsat 2VGT Vc = 0

2Vc (VGT VDsat ) = V

2 Dsat

(18)

Thenbyinsertingeq.(18)intoeq.(13):

I Dsat = CG (VGT VDsat ) satW

I Dsat
2 1 VDsat = CG satW 2 Vc

I Dsat

1 V = CG satW sat 2 L

2 Dsat

lf

ThustheIDsat issimplifyingas:

I Dsat

1 CG lf W 2 = VDsat 2 L

(19)

TheIDVDS characteristicsoftheNFET:resultsfromthesimplemodel.Forthisdevice W/L=5,tox =4nm,CG =8.63 103 F/m2,andmn =500cm2/Vs.

ComparisonofIDVDS characteristicscomputedbyusingtheconstant mobilitymodel(q =0,dashedlines)andtakingintoaccounttheeffect ofthetransversefield(solidline)forq =0.13V1.Thetransversefield tendstoreducethecurrents

ThecalculatedIV characteristicsforthesimplemodelandfor NMOSandPMOSwithcarriervelocitysaturationaccountedfor.

Thesaturationvoltageasafunctionofchannel lengthfor|(VGS VT)|=2.6V.

Saturationcurrent|IDsat|asafunctionofL forn andp channelsiliconMOSFETs.Here|(VGS VT)|=2.6V,the widthtolengthratioisW/L =10,andWp =2.5Wn.

Comparisonofthesimplelongchannelmodel,themodel includingvelocitysaturation,themodelincluding bothvelocitysaturationandtheseriesresistancesRS andRD,and theactualmeasureddata.ForthisNFET device,L =0.25mm,W =9.9mm,andVT =0.3V.Thegatesource voltageis1.8V.

Forshortchanneldevices,thereductionofthe effectivechannellengthwithincreasingVDS resultsinanincreaseinID

(a)Forlongchannelsthedrainvoltagehasnegligibleeffecton thebarrieratthesourcechannelinterface.(b)Forshort channeldevicesthedrainvoltagetendstoreducethebarrierat thesourceend.(c)Theresultisthatthethresholdvoltageis decreased.Theeffectismorepronouncedasthechannelsget shorter.

Thedifferencebetweenlong andshortchannelbehavior dependsontheoxidethicknesstox,thesourceanddrain junctiondepthxj,andthedepletionregionthicknessatthe source,wS,anddrain,wD.

IllustrationofMOSFETscaling.Toreduceshort channeleffects,whenthechannellengthisreduced, thedimensionsanddopinglevelsofthedevice(a)are adjustedtoreducetheoxidethicknessandjunction depth(b).

MATLABCODING
1.DefineConstantData

hbar=1.0545887e-34; kb=1.380662e-23; q=1.6021892e-19; T=300; eox=3.9*8.8541878e-12; esi=11.9*8.8541878e-12; m3=0.98*9.109534e-31; m1=0.19*9.109534e-31;

tox=1.59*10^-9; L=80*10^-9; W=1.3*10^-6; VT=0.3542; VGS=0.7;

1.VelocitySaturationModel vth2=sqrt(pi*kb*T/(2*m1)); %non-degenerate 2D velocity vth=sqrt(2*kb*T/m1); % general thermal velocity

m2 =
vth =

k BT
2m *
2k BT mt*

%The following component calculates the vsat value for ElecFDrain infinity
13 2 for k=1:6 VGS + VT h 2 q 2 Eteff Eteff Eo = VGS(k)=0.7+(k-1)*0.1; 2m* 6t0 x 3 VGT(k)=VGS(k)-VT; Et(k)=(VGS(k)+VT)/(6*tox); Eo Eo(k)=(hbar^2/(2*m3))^(1/3)*(9*pi*q*Et(k)/8)^(2/3); zo = qEteff zo(k)=Eo(k)/(q*Et(k)); 2 (2.338 Eo ) zQM = zQM(k)=2*zo(k)/3; 3 qE teff ox toxeff(k)=tox+(eox/esi)*zQM(k); toxeff = tox + zQM si CG(k)=eox/toxeff(k); uo(k)=0.0695*exp(-0.7568*VGS(k)); % exponential fit to exp

lf = 0.07.e (V

GS

1.33 )

QuantumConfinement&MobilityModel
CG =

ox
toxeff

toxeff
= 1+ Cox 1 zQM 3 tox

ox 1 = tox + zQM tox + zQM 3 Si

2 (2.338 Eo ) = 3 qE teff
2 2 2 teff * 3

zQM

Eo zo = qEteff

h q E Eo = 2m

13

Eteff

VGS + VT 6t0 x

lf = 0.07.e

(VGS 1.33 )

Continue.Iterativesolutionforvsat when=1
V vsat2(k)=vth2; Vc = sat L VDsat1 = Vc 1 + 2 GT 1 lf Vc for j=1:10 Vc(k)=vsat2(k)*L/uo(k); VDsat1(k)=Vc(k)*(sqrt(1+(2*VGT(k)/Vc(k)))-1); n2(k)=CG(k)*(VGT(k)-VDsat1(k))/q; % 2D Carrier sheet concentration u(k)=n2(k)/NcT; %find fermi(0)eta2 eta(k)= log(exp(u(k))-1); %eta2 t1=fermi(eta(k),0.5); %find fermi 0.5 eta2 vi2(k)=vth2*(t1/u(k)); 1/ 2 ( F 2 ) end

vi 2 = vth 2

n2cm(k)=n2(k)/1e4; K(k)=CG(k)*uo(k)*W/L; IDsat1(k)=K(k)*VDsat1(k)^2/2; end

o ( F 2 )

% Idsat for full saturation

I Dsat1 =

1 CG lf W 2 VDsat1 2 L

BallisticSaturationVelocityModel
i
E=0

i
vi 2 = vth 2
2 i

vth 2 = vth

(3 / 2) (1)

vth =

k BT
2mt*

1/ 2 ( F 2 ) o ( F 2 )

vth =

2k B T mt*

E = LARGE

1 j ( ) = ( j + 1)
NC 2 = m * k BT h 2

x j dx 1 + e x

vi 2 =

.k B .T
2.m

NC 2 F1 ( F 2 ) n2 2

n2 = Nc 2F0 (2 )
F =
2( EF EC ) m*

vi 2 ND =

k BT
2m

vi 2 D =

2h
*

2n2

3m

IEEE International Conference on Semiconductor Electronics (ICSE2008) 25th 27th November 2008 (Johor Bahru)

NanoMOSFET :I VCharacteristics
1 2 V V VD GT D CG lf W 2 ID = V L 1+ D Vc

0 VD VDsat
VD VDsat

I Dsat = CG (VGT VDsat ) satW

Vc = c .L =

sat L lf

= D sat

E
T

y=0 V=0

y=L EL V=VD

V VDsat1 = Vc 1 + 2 GT 1 Vc

=1
(Full Saturation)

I Dsat1 =

1 CG lf W 2 VDsat1 2 L

El =

VDsat1 2L

1 1 x L

IEEE International Conference on Semiconductor Electronics (ICSE2008) 25th 27th November 2008 (Johor Bahru)

Continueiterativesolution<1
%The following component calculates I-V characterisitcs and alpha for finite ElecFDrain for k=1:6 alphanew=1; for i=1:20 alpha1=alphanew; s=sqrt((alpha1+((1-alpha1)*VGT(k)/Vc(k)))^2+2*alpha1*(2*alpha1-1)*VGT(k)/Vc(k)); VDsat(k)=(1/(2*alpha1-1))*((s-alpha1)*Vc(k)-(1-alpha1)*VGT(k)); alphanew=((1/alpha1)*VDsat(k)/Vc(k))/(1+((1/alpha1)*VDsat(k)/Vc(k))); end alpha(k)=alphanew; 1 [(s )Vc (1 )VGT ] V = vdrain(k)=alpha(k)*vi2(k); Dsat
V V s = + (1 ) GT + 2 (2 1) GT Vc Vc
2

(2 1)

%alpha not equal to 1 s=sqrt((alpha(k)+((1-alpha(k))*VGT(k)/Vc(k)))^2+2*alpha(k)*(2*alpha(k)-1)*VGT(k)/Vc(k)); VDsat=(1/(2*alpha(k)-1))*((s-alpha(k))*Vc(k)-(1-alpha(k))*VGT(k)); IDsat=(alpha(k)/(2*alpha(k)-1))*K(k)*Vc(k)*(alpha(k)*VGT(k)-(s-alpha(k))*Vc(k)); VDsatP(k)=VDsat; IDsatP(k)=IDsat; CG lf W end I Dsat = Vc [VGT (s )Vc ]

(2 1)

NanoMOSFET :I VCharacteristics
VDsat = 1 [(s )Vc (1 )VGT ] (2 1)

I Dsat

CG lf W = Vc [VGT (s )Vc ] (2 1) L
2

<1

V V s = + (1 ) GT + 2 (2 1) GT Vc Vc

D = sat
Ec = sat 0

E D Ec = 1 + E D Ec
ED = VDsat L

Iterative solution

IEEE International Conference on Semiconductor Electronics (ICSE2008) 25th 27th November 2008 (Johor Bahru)

Continue..

for k=1:6 VD=0:0.01:VDsatP(k); ID=(K(k)/2)*(2*VGT(k)*VD-VD.^2)./(1+(VD./Vc(k))); % for 0 V D V Dsat VDM=VDsatP(k):0.01:1.0; % for VD VDsat alphaprime=((1/alpha(k))*VDM/Vc(k))./(1+((1/alpha(k))*VDM/Vc(k))); IDMprime=alphaprime*W*CG(k)*vsat2(k)*(VGT(k)-VDsatP(k));

1 2 V V VD GT D C W 2 I D = G lf VD L 1+ Vc

I Dsat = CG (VGT VDsat ) satW

VD1=VDsatP(k):0.01:VDsat1(k); ID1=(K(k)/2)*(2*VGT(k)*VD1-VD1.^2)./(1+(VD1./Vc(k))); VD1M=VDsat1(k):0.01:1.0; ID1M=ones(1,length(VD1M))*IDsat1(k); end

1 C G lf W 2 I Dsat1 = VDsat1 2 L

PlottingGraph
% Plotting Intercept Line x=[0 0.2667 0.3311 0.3932 0.4537 0.5131 0.5716]; %VDsat1 y=[0 0.0003 0.0005 0.0007 0.0009 0.0011 0.0013]/1e-3; %IDsat1 xi=0:0.01:0.7; yi=interp1(x,y,xi,'cubic') %For VDsatP & IDsatP (Alpha < 1) x1=[0 0.1880 0.2661 0.3029 0.3387 0.3739]; y1=[0 0.0003 0.0006 0.0008 0.0010 0.0012]/1e-3; xi1=0:0.01:0.5; yi1=interp1(x1,y1,xi1,'cubic') Figure (1) plot(xi,yi,'--b',xi1,yi1,'--r',VD,ID/1e-3,'-r',VDM,IDMprime/1e-3,':r',VD1,ID1/1e-3,'b',VD1M,ID1M/1e-3, '--b','Linewidth', 3.0) hold on

%VDsatP %IDsatP

Continue..
%This block plots the experimental results on the same plot as previous %plot has "hold on" VDexp=0.1:0.1:1.0; IDexp12=[0.000487727 0.000874322 0.00114401 0.00132208 0.00144098... 0.00152452 0.0015872 0.00163731 0.00167961 0.00171686 ]/1e-3; IDexp11=[0.000455394, 0.000800693, 0.00102774, 0.00117086, 0.00126455... 0.00133073,0.00138145,0.00140317, 0.00144182, 0.00147613,]/1e-3; IDexp10=[4.1575E-04 7.1329E-04 8.9562E-04 1.0055E-03 1.0770E-03... 1.1286E-03 1.1695E-03 1.2044E-03 1.2358E-03 1.2649E-03 ] /1e-3; IDexp09=[ 3.6741E-04 6.1091E-04 7.4837E-04 8.2842E-04 8.8132E-04... 9.2118E-04 9.5439E-04 9.8396E-04 1.0114E-03 1.0376E-03]/1e-3; IDexp08=[ 3.0908E-04 4.9363E-04 5.8896E-04 6.4404E-04 6.8234E-04... 7.1317E-04 7.4037E-04 7.6569E-04 7.8995E-04 8.1361E-04] /1e-3; IDexp07=[ 2.648E-04 4.065E-04 4.604E-04 4.843E-04 5.013E-04... 5.148E-04 5.258E-04 5.352E-04 5.431E-04 5.500E-04 ]/1e-3; plot(VDexp,IDexp07, '.g',VDexp,IDexp08, '+g',VDexp,IDexp09, '*g',VDexp,IDexp10, 'xg', ... VDexp,IDexp11, 'sg' , VDexp,IDexp12, 'dg', 'MarkerSize',5.0,'Linewidth', 2.0) ylabel('\itI_{D} (mA)') xlabel('\it V_{D} (V)) hold off

I-V characteristics of 80-nm MOSFET for Gate Voltage, VGS =0.7 - 1.2. Solid lines are for VD < VDsat. The dotted lines are for VD > VDsat . The dashed lines are for =1.

Reference
1. Vijay K. Arora, Michael L. P. Tan, Ismail Saad, Razali Ismail. 20070906. ballistic Quantum Transport In A Nanoscale Metaloxidesemiconductor Field Effect Transistor. Jil. 91. American Institute Of Physics: Applied Physics Letter (apl 2007). 2. Michael L. P. Tan, Ismail Saad, Razali Ismail, Vijay K. Arora. 20070718. enhancement In Nanorc Switching Delay Due To The Resistance Blowup In Ingaas. Jil. 2. 4. World Scientific.com: World Scientific Publishing Company (nano 2007). 3. Ismail Saad , Michael L.p Tan, Ing Hui Hii, Razali Ismail And Vijay K. Arora. 20080726. Ballistic Mobility And Saturation Velocity In Lowdimensional Nanostructures. Vol. 40, No. 3 (2009) pp 540542. www.elsevier.com: Microelectronics Journal. 4. Ismail Saad , Michael L. P. Tan, Aaron Chii Enn Lee, Razali Ismail And Vijay K. Arora. 20080725. Scatteringlimited And Ballistic Transport In Nanocmos Transistors. Vol. 40, No. 3 (2009) pp 581583. Www.elsevier.com: Microelectronics Journal. 5. Michael L. P. Tan, Vijay K. Arora, Ismail Saad, Mohammad Taghi Ahmadi And Razali Ismail. 20090402. The Drain Velocity Overshoot In An 80 Nm Metaloxide semiconductor. American Institute Of Physics: Journal of Applied Physics (2009).