Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Статьи основные / % 1999 Base science and TCAD

.pdf
Скачиваний:
9
Добавлен:
11.03.2016
Размер:
328.24 Кб
Скачать

Advances in Solid State Physics 38, 565−582 (1999)

JBasic. Dabrowski,Germany;ScienceSemiconductoH.-andJ. Mussig,ChallengesM. Duane,() in bProcessS. T. DunhamSimulationc, R. Goossensd,

 

H.-H. Vuong

 

 

 

r

 

 

 

Walter-Korsing-Stra e 2, D-15230 Frankfurt

Institute fo

 

 

 

 

 

 

 

 

 

 

 

and Comp

ter Engine ring

 

Dept.Physics,Boston UniversityAustin,Boston, MA

 

02215, USA;

(O er),

 

 

 

 

 

 

 

b

Advanced Micro Devices, Inc.,

 

 

Texas, USA;

c

 

 

Semiconducto

Research Corporation, National Semiconductor Corporation;ElectricalBel

Laboratories, Lucent T chnologies, Murray Hill, NJ 07944, USA

w

 

 

 

CMOS

 

 

Summary: We intend to turn the reader's attention

 

 

 

 

 

 

(Comp ementary Metal-Oxide-

 

 

 

r), the

mainstreamardsemicon-

 

 

may fail. Th

 

 

calls for

 

 

 

Semiconducttrol of at mistic

processes. W

r

 

 

 

 

 

ductor technology. W thin a decade, silicon CMOS devices will be 200

 

 

atoms

long and 50

toms deep. Of 50

billiony f

devices on a

 

 

hip only 64

 

 

needs of CMOS. It is

 

otedextremehat

 

 

which adopt meth-

 

 

researc

top cs in b

sic

a er als s ience which are likely to ma c

view

 

 

schemes to easily

solvable continuum equationsmodelsu t be constructed.

 

 

 

1

ods ranging from

quantum mechanihierarcthroughs

clas ical and Mon

 

Carlo

 

Introduction

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ts of ma

Atomic-scale studies of solid state systems mark major achi v

 

 

 

terials science. The

 

 

 

ream microchip technology, Compl

 

emen

 

 

itiv

ss of

 

 

 

(Fig.mains1), tom stic

 

h and modeling

 

 

 

 

nhance compet-

 

 

 

 

 

 

 

 

by

 

 

ng techno progic t

ideas on funda

 

 

 

 

al lev Metal.

Oxide-Semicon

 

ctor (CMOS), may also

 

from them. Stimula edtaryb

CMOS

A

tec

nologistproductionfte nds

 

 

researcat cro sroad. Ph

 

 

 

insigh

 

can

 

one

 

h

 

h

ice of

directiontesto head, he ping in early devysical

 

 

 

t phasesguide. Th

miniaturizationbe bene cial

f

 

ilure and senhimselfitivitsupplyanalysis within

the exis

 

 

ng technologies.

same insi h

 

 

 

 

 

 

 

a scienti to

e cien simulatiopmenmodels. This can

Starting

frompositionsverview of CMOS, w

discuss

simulati

 

 

en

 

 

 

ocused

 

 

di crystalu ion-relat d problems viewed

om the perspectiv

 

of the whole

CAD tools,

 

 

 

 

 

 

growth, wafer processing, impl

tation,

 

di usion,

 

xid t,

This

ummarizes the 182nd H raeus-Seminar \Interna

 

 

W

 

 

 

 

vironmenChal-

silicides,lenge in Predict vprocesses,Proc

Simulation", ChiPPS'97 [1, 2], whicorkshopwasprimari y

fabrication process.

These issues are posted for rankingtionalthe

Internet [2].

 

 

 

back end

 

 

 

 

 

 

 

g tt ring, and prospectiv

metrologies and theories.

 

 

Figure 1

 

 

 

 

 

 

 

drives

chnology into the atomic

 

 

. In this STM

 

 

 

image of the clean S (001), atomic details as dimer rows, defects, and monatomic

 

 

steps are

clearlyMiniaturizationvis ble. The

length of the scanned area is 0regime.1 . Within one

2

decade, his will be the minimum feature

size of state-of-the-art CMOS devices.

MOS technology

 

 

 

is the Field E ect Transistor (MOSFET)

The workhorse of MOS tec

 

 

 

 

Fig. 2). Basically, it is a swhnologytccontrolled

y the gate potential. MOSFETs re

named after the type of carriers which ow b

 

 

w en the source (S) and the

 

 

 

 

D): nMOSFETs are built on p-doped material, with n

+

-doped source

 

 

drain

(S/D), while pMOSFETs are built

 

 

-doped

 

 

 

 

 

 

 

 

with p

+

-dopeand S/D.

 

Circuits must be fa ter, cost less per bit,

 

 

 

less power. T ey are c eap

er

with

 

terconn cts

 

 

hip. Hence

 

 

 

 

 

 

 

 

 

 

[4]: shor er channels,

 

 

oxides, shallo

 

e

 

 

 

 

 

. Gate oxideminiaturizationthicknesssubstrates,now 8 nm, will be thinn3

by 2001, wi

 

 

roug

ness approaching the

 

 

 

 

ox

 

 

 

 

t [3]. Estim

 

fro

 

 

 

 

 

 

step

 

 

 

 

1997 indicated

 

junctionscan be scaled downatomic2

 

 

[5]; inheigh same year

 

0.06 m

MOSFET

 

 

 

 

ox

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ed [6]. Suc

 

trantesistors

 

 

thatof only 1.2 nm (12 A) was denm

 

 

 

 

 

 

 

 

work fast butwith

ox

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tun eling through the oxide.

v high standby power losses d

 

 

 

 

 

 

Maintaining an increas

 

perfo mance of

 

ub-

 

 

MOSFETs may be expen

siv

[7]. Also costs of

billion

 

lines escal

 

te:

currenstrat

thegenerationrev

\f

b" (fac-

 

ry)

 

sts

around

 

dolla

. Only

a

fraction

of

 

 

y

th

 

 

c me from

utting-edgeproductionhnologie

 

(today 0.25

 

 

linewidth, or the

mate

xi

 

 

length, whi h

 

 

 

 

 

 

 

 

 

the

hannel length). Even 95% of the productus

shipped

1997somewhatby

 

 

 

microprocessor

 

ompany y havuebeen made

with

 

 

 

0.5 and 0.8 proc sses. Inde d,

 

 

 

 

 

 

 

 

 

 

 

 

ow has its c allenges,

 

A

 

 

 

which makeprosperingyestxceedsday's technologies cheaptomorrmay e welcome today.

 

 

ypical in

 

 

 

 

circuit needs severalthoughundred

 

 

 

 

 

teps whic

 

pply\old"position,

 

 

 

 

 

 

xidation, implantation, etching, polishing. The

wafer

 

 

 

 

 

 

by ocal doping and

 

 

 

 

.

 

Thenfabricationgate

 

 

 

are grown

innovationsnd co e

 

ytegratedgalithographysilicon,po

 

 

tacts are made, inte conn xidesare add d. It

remainis to

break the wafer into conhips and seal them

for protection. The steps

needed tostructuredmak the devices are na

ed FronisolationEnd Of the Line (FEOL) processes,

while the steps needed to wire them are dubbed Back End Of the Line (BEOL)

 

 

 

 

 

 

source

 

 

 

 

 

 

 

 

gate

 

 

 

 

 

 

drain

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cross

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ga

xide

FEOL. termi

 

 

 

 

 

 

But the road from physics to tec

 

 

 

 

 

 

 

 

 

 

 

 

Most\channel"S/Dare

 

 

inv

 

 

on channel

 

 

 

 

 

 

 

 

 

sub trate

 

 

gat(Dthek.,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

wn)

 

 

 

also.heavilyAtheymbolizeprocessgate2highlytheisolationbetwAdpotgateschematiceendopedsteps\bird'stialpolySixideacerand.greateareThebeaks",(DSiO\stin. rThesubstrateeldip;-BEOLthan2sectionorsidetheelectrodesxide"Sithe-3othandNofbth4ectsor)a wesholdMtheyFelearehicOofX)nSFETtrodes,LOCOScontactedare.encompasvolandriangtheage,[8]sourceD.primaryBlackxidationVblarTes,siliciope(S)theareasnsd.andyieldgateAlsoppositelyati(notonconducting.drain-Thenslimiterstheshoofblsides

 

FmarkofOXigureshownthe

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

is

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I plan ed

 

atoms

 

 

re activated

 

 

to

ubstitutional sites b

 

anneali g

Standard dopants are B, P, and As; new designshortery emplo

In and Sb. Implants

are s ot

hroughdopanpro

 

 

e SiOhnologyla er, someti

es

vered b

 

Si N . This may

th n used for

junctions and

 

 

 

 

 

 

ts of

h

 

l pro les. Gates

 

 

be de

CMOS combi

es nMOS and pMOS on the same waften, usually pre-

 

 

 

with

reduce chann l

g. T makectiv shallower pro les, B is co

 

imp anted as BF .

B. The n-

ype

 

ells (\tubs") for pMOS are

 

 

 

. Low-energy impl nts

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3

 

4

 

 

 

2

 

posited already doped, but they are usually implantedbecause device dopedarameters

are optimal when the gate andadjustmenth channel have the same

ork

 

 

 

 

 

.

 

 

grownbeaky high-temperature dryOxidation after striping the rem

 

 

ing padproducesxide.

containing gas. This LOCal

 

 

 

 

 

 

 

of

Silicon

(LOCOS) sc

 

 

 

 

[7]

 

 

 

 

deposition steps, may be

eeded in

 

 

 

future [10]. High-

 

 

y functiongate xides are

The activ

areas are iso ated by

 

 

 

 

 

 

 

x dation of the pad

 

xide in O-

\bird

 

" (Fig. 2), so Shallow Trenc

Isolation (STI),

whicheme

 

 

etch and

joined by vertic

W \plugs". T

 

thermalfutu material

nesCu [11]. Glasses (BPSGels")

(salicidation)

 

Ti, in the future

 

ybe Co.

The metalqualit

 

ks S , but not

After gate deposition and etch, con

 

cts are ma

by se

-alignaddsed licidation

SiO . The

 

 

 

lm is removed, leaving well-d

 

 

 

contacts.

 

 

 

 

 

 

The

rcuit withred with planes of ho izontal Al

 

(\metallization v

 

wires. New dielectrics

will be needed and even Cu mafacilitatesy hav to be replaced.

are used fornreactedisolion,

but their

h gh dielectric conste nedts d lay the

 

 

 

 

 

 

of electric signals (paras tic capacitances,

RC) and

 

 

 

cross-talkpropagationb wee

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

DRAM SIZE (BYTES)

 

 

 

 

 

 

 

 

 

 

 

 

PRODUCT/TECHNOLOGY SPECIFICATIONS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1M

1M

4M

16M

64M

256M

1G

4G

16G

64G

1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LOGIC DESIGN

 

 

 

 

CHANGE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.1

 

BIT PER COST DRAM

FAN OUT

 

 

 

 

 

 

 

ARCHITECTURE

 

 

 

 

 

 

 

COST PER LOT ($)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CLOCK SPEED

 

 

 

 

 

 

 

LOGIC GATES

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CLOCK SKEW

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CHANGE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PHYSICAL DESIGN

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SHEET RES.

 

 

 

 

 

 

 

TRANISTOR PLACEMENT

 

 

 

 

 

100k

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

THRESHOLD V.

 

 

 

 

 

 

 

GATE LENGTH

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.01

 

TARNSIT FREQ.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CHANGE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PROCESS DESIGN

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

OXIDE THICK.

 

 

 

 

 

 

 

TEMPERATURE

 

 

 

 

 

 

 

DRAM

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cents)-(milli

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

DOPANT DISTR.

 

 

 

 

 

 

IMPLANT DOSE

 

 

CHANGE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PROCESS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10 − 3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NO

 

 

 

 

10k

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PRODUCT

 

 

 

 

 

MET?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10 − 4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1985

 

 

1990

1995

 

2000

2005

2010

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

YES

 

 

 

 

 

 

 

 

 

 

 

 

 

 

YEAR

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SHIP

 

 

 

 

 

F

 

 

 

 

3

Left: Pr

 

 

per bit (circl

 

, righ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ordinate) and cost per lot of 25 wafers

 

 

R

 

ht:

 

s, left

 

 

 

 

ce) for DRAM

 

 

 

 

 

v rsus the y

 

 

of introduction [9].

 

 

 

 

 

 

ordinatalization. Logic design trans ates the

productk

sp ci cations into

 

 

ates,re gisters,

tc., assuming

 

 

 

parameters as

 

ar

wit

 

in

ranges

 

 

squa

teed

by

 

the

 

physical

layou . T

latter

 

 

out ines

th

 

physical

 

 

 

 

 

 

 

(transistor placement, linewidths) which should ful clocl the

peearquests. Thestructureproces

3

 

design Productsts the physical

design

generationsto process

conditions.

 

 

 

 

 

 

 

 

 

 

 

ulation

Environmentargetsand TCAD tools

 

to the

 

 

 

 

.

 

 

 

 

 

of technological processes is incr asingly attractiv

 

 

wafers.

First,Simt can sav

time.

 

 

 

 

it can reduce processing costs of

 

 

 

 

 

 

N w generat ons are

 

veloped b

 

 

scaling the

 

 

 

 

 

ones; problems

are

Modelsolv

b

matrix experimSecond,ts. A tec

 

 

 

 

 

is of en

 

uned during

 

 

 

 

 

 

by varying

the

process

for some wafer lots (\splitexistinglo s"). If thesilicproduction,hange

 

 

 

 

wafers can be lost. By the year 2010, a lo

 

of 50 wafers

will cost nearly

 

million dollars (Fig. 3) and this w

 

 

ma

prov

 

too expensiv

 

or too slow.

 

 

 

A semi onductor fab acts as a

 

 

 

 

 

 

 

h may need

 

 

few mindustrythstoo

drastic,run batc

 

(Fig. 3). Simulation ccomputerhnologyv

 

 

 

 

within a few

 

 

ys,

so virtual

 

 

 

 

 

ts w uld be optimal. But th s maywhicrk

 

ly if all

design

tools are

experimpr dic ive. This ma

 

be di cult (Tab. 1), but any progress is certainly

 

elcome.

 

Simulation for

CMOS includ

equipmenresults

 

 

 

 

 

crystal growth,

CVD,

rapid th rmal and plasma processing), pr

 

ss

 

 

 

 

 

 

 

(di usion,

 

 

inter-

 

 

 

 

 

 

 

 

epitaxy

etching), and device

 

 

 

deling(physi s,

 

 

 

 

ac ion of devices, in

 

 

 

 

 

 

 

 

 

on, burnut). These

 

 

tivities

eed

 

tion,ex en

v

 

 

 

 

se coupled to physically soundmoddels.

Process modedesign,mplantag y

be assixidation,ste atab

tomistterconnects,the retical tools employing quantum mechanics

 

 

molecular

dynamics. Di usion or in

 

rfacial physics ap

ear

particu arly

suiteand

for them. But in many cases directdegradatim asurements or empirical models will dom-

Applicationedictivanced TCAD control

Accuracyvery highneeded

proCommentsvidebably an elusivodelsgoal: : :

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

process cen

 

 

ring

 

 

 

 

high

 

optimizemacrofor

 

mature product

adverse m processling

 

 

 

 

 

use to extract coe cients

 

 

early exploration

 

 

 

medium

l

 

splits in proces ing of lots

failure analys

 

 

 

 

 

 

 

probable cause

 

 

 

t

 

 

learning and insight

 

 

 

 

 

low

 

high return on inves

 

 

 

 

Table 1

 

 

 

TCAD applications. The top

testitems get most attentiotmen, while

 

 

much useful work is done in the bottom thre

classes. See also [3, 12, 13, 15].

inate. As examples, one may

 

 

 

 

 

reacti n rates

nd

 

 

king

 

 

 

 

ts.

 

 

The key

 

 

omistic issues in

 

 

 

 

 

 

f r crystal growth and w fer pro-

 

 

 

 

about

efects

 

quoted h mogeneity. The mainstichallenges in the front

nd are are ted

to

opan

pro lesdelingshallow junc ions and short

 

 

a nel , an

cessingxcellent pl

yground for atomistic

 

physics. It is int iguing thatcoecombicien tion of

models (e.g., for h

h-concentrati

 

 

 

 

regime and

 

for defect-mediat

 

 

 

di usion)

remains

problemat c. Promising

for

 

 

-scale physics in back

 

nd are sur-

face processes du ing deposition and

tomiching. The sim

 

 

on methods

 

but

key paramet rs are di cult to get and little physics is

 

 

ilt into BEOL models.

Technology

 

 

 

 

Aided Design (TCAD) can beulatsed on man

 

problems

(Tab. 1). It

 

 

ust be e cient; physics is just one of the means to ac

 

 

exist,v this.

In early

dev

 

 

 

 

 

sim

 

gives

in igh

into technology direc-

 

and elucidatesComputerelopmen eractions

 

 

etween optional olu ions. A high return

inveonstm

 

 

can be

 

 

by getulationing appr

 

 

 

but

arly answers to the

right

ques

 

 

 

 

 

. The earlierphases,problem is ca ght, the

 

 

 

it to correct. onI

the late dev

 

 

 

 

 

 

obtainedphase, TCAD is useful for

 

 

 

 

optimization, sensitivity

a lysis, and diagnosis [16]. Such quantitativeoximateplicatioeasier

 

 

more demanding

(T . 1). Butelopmenv insigh

 

 

helps. E.g, und

 

standi

g of

 

 

 

 

 

processes

nables

 

pot

 

 

 

of processalone

 

bility and of p ocess-inducared v

 

 

 

.

 

 

A

 

 

 

 

tial for modeling is in new materials.

 

 

 

 

 

a lot of data on

existigreattechnologies, so extrapolations

usually work. Thereb

 

questi

 

 

 

with new materials, as Cu instead

 

 

 

f Al, or

 

 

 

ts for SiO ariations. An ther key

ea wh

 

 

basic physics may help is in

 

 

 

 

 

 

 

 

2

 

 

 

comean

 

 

 

. Any insightexistingth

provide

arly in the

 

 

 

of

 

 

ew

chnoloreplacemen

 

have

 

large impact.

Practical TCAD uses

 

 

 

 

 

 

nallyterfacesheap,

 

tinuum

 

 

 

. This tradi-

tional methodpredictionsis

achingcomputatioin defect-mediat d pro le

 

olution,

hemical

 

hysical rea tion modeling, pred

tivteequipmgiescont and

 

 

 

 

 

physics

 

 

and practical non-equilibriumdevelopmentransport. It is unclear when it widelsbreakmodeling,down

but technology advances fast and th

may happen before wpographev alize

 

. Discus-

sions between technologists and physicists are needed to smoothen theitransition

 

 

fabprocess/deviceSpeequipmenialization

 

 

KnoTm inwledgeimplantatineededcompletefora ectsTCADdi usion;ow,

ocess/deviceTo in RTAphysics

 

 

 

analytical

 

 

 

 

SIMSknock-on; SRPprocessbe pressure

 

 

 

 

 

 

 

 

 

 

metrology

 

 

 

 

unce

ta

y in poly

 

 

 

 

 

calibrationxid thickness

 

 

 

 

 

 

 

 

electrical tests

 

 

eleust

 

versus optical

xide thickness

 

 

variation

 

 

 

fusion later [17]tation. T

during RTA is

di cult to measu

The[18], but 1 v

 

 

simulation

 

 

 

 

m

 

icalmitati

ns; gr

depend

.

 

 

fer tempera ure T

 

 

 

 

T

 

 

2

Knowledge needed

f

TCAD

 

 

 

 

 

 

 

 

 

 

 

 

 

implan

 

 

 

 

can a ect the

 

 

 

lengtho defect geence

 

 

nd thereby

he dif

 

 

 

 

duringable220 V AC line c

 

 

T

 

y 6

 

 

[19].calibrationThe beam

nergy can

 

 

SIMS

 

 

 

pro les [20]. SRP dep nds

 

probe pressure [21]. Poly linewidth andoltxide thick-

 

 

 

ness are di cult to

hangedasure;

 

 

1% change a ects the saturation currena ectby 1%.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

which would reliablydimensionalex rap late 1D data re needed. When adopted

 

Considiscreteer t

 

exampltomistbe . First,

electrical

characteristic of tiny tr

nsistors

 

 

 

 

 

 

 

 

 

 

 

 

. A

 

 

c phy

 

 

 

 

s not easily integrable into the exi ting

a ected by tmodelsw - and three-

 

 

 

 

 

 

 

 

(2D and 3D) structural details [14].

Thare

tools and will hav

 

 

to

 

rst tran la ed to continuum models.

 

 

 

 

 

.

3D narro

-

 

 

nel e ct becomes

 

 

 

 

 

 

 

to the 2D

hort-channel

 

 

 

to the continuum formalism, they

canmparableused in

 

modern

ulators wh ehectal

 

Since 2D dopan

 

pro les cannot be properly me su

 

(section 10), di usion

low to de ne equations. Second, applications beyond 0.1 simwill need

 

 

[4].

 

models

 

 

of defects and

 

 

ts, with stochastic analysis of uctuatio

 

 

 

be

 

 

 

ated b

ts to experiments not

lw ys doable

und conditionsdiscreteclo

 

Optim lly, TCAD should

 

 

 

cor ect

 

 

 

haracte istics given any pro-

 

 

those inBute production. The

 

 

 

 

 

 

 

maelectricallso employ wrong as

umptions or

 

scriptionCalibra on is also

 

 

 

 

 

 

 

 

 

 

ypically andwhole

 

is simulated. If

ess ow.

 

 

the models hayieldumeopan rous, often obscure

 

 

. They

 

 

 

cto

 

 

in no physics. The predicivitmodelsy pends thus on the

parametershoic of

 

structures,

mterpretu sholdvalue,

one experimenust heck assumptions in the proce

d d vice

 

 

ula-

in

 

 

 

 

 

 

 

of

 

he

 

 

 

t, and in

 

 

 

 

experience of the user.

 

 

he th

 

 

 

 

voltage from the

 

subsequentuitiondevice analysis does not

match the

ion , the test structure

 

yout,becaused electrical m asuremenprocess. K wledge simany

 

M dels wi h physicaldi ycultmeaningful parame ers

 

 

with known windows for

understanding,

nd

 

process

c nditions

 

 

v

 

the last 20Somey importanrs. void confu-

t eir

variation are thus

favored. Calibrat on

trategies are

 

t. A

 

 

 

-

ar as is required (Tab. 2). Phys cs can help by systematizing this knowledge

 

chy of

mo

els

 

 

 

 

atomi tic to con

 

 

uum

is needed.

 

 

basic

 

 

 

ierar.g.,

for segregatio

from, ust be re-

 

 

 

 

tin light of advand es in metr logymodels,

 

 

sion, implementation of the examinedodels ust b

backward compatible; their phy ical

interpretation would assists users and vendors, adding value to TCAD tools.

 

4

 

 

 

 

 

for crystal growth and wafer processing

 

 

 

 

 

 

 

puri ies

grow

 

ust

 

 

 

 

 

Si base material with homogeneous

 

 

 

 

 

 

 

on

 

the

 

lt anproducetheir inc

rporation

 

 

 

 

to the

crystal. Givendistributincre imng

CrystModelsal ameters (a 300 mm technology fab is

 

 

 

 

 

built in Dresde

Germany),

of O

 

nd dopants. Mo

 

eling can be used to predict the

ransport of the

 

-

 

 

 

 

 

modeling

 

 

 

 

 

 

o

sid

heat

 

 

 

 

 

beingy conv ction,

 

 

 

 

 

 

 

 

 

and

radi ection. 3D gas conv ction and reactiontransfergasulation

will gain

(SiHCl H ) with the

 

sts of developme

 

will

 

 

 

 

 

and sim

 

 

 

 

 

 

 

mporta ce. But melt

improvement [22]. M dels for

 

 

dimensi

 

 

 

 

 

 

are

 

 

 

 

velopmconduction,p se.

 

conv

 

 

 

is turbulent and thre

 

 

 

l, and thus di cult to handle.

 

 

surfEpitaxyce ust be includ

 

.increaseA curacy

 

f heat

ransfer and convection

 

 

 

 

 

 

 

needs

 

Grown-in defects (shouldctahedral voids)

 

as-grownmixtureSi uce the yimodelsld (t

 

 

-

be simulated by

 

 

 

into accounreactiondykinetamics of pointhedefect

 

3

 

2

 

 

 

 

 

 

 

 

 

(self-in erstitials,

centage of working

hips)

 

 

must be contained. Their

istributions will h vperto

in Si

wafers (includitakingatoms)hermal stress), defects aextende

 

 

y in

pitaxy, and

 

Fi

cies,ally

model

 

 

 

ls for O

 

 

 

 

 

 

 

 

of

 

 

 

th

 

 

treatm

 

 

ts, includ-

vaca

 

 

 

foreign

 

 

 

 

 

and the form

 

 

 

 

 

 

 

d defects [23].

 

 

 

 

 

 

 

5

Models for ion implanprecipitation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n 9), stress

 

g n uence of

mpuri ies (e. .

 

boron),

getteriduringof metal

 

 

 

 

molecular simulation for

 

 

cleaning and

 

lishing of whomogeneitafers us(sectiobe developed.

 

A major challe ge is the formation of shallow, heavily doped

 

 

 

 

 

 

 

 

[3]. Junc-

Stand

rd TCAD yields the sp tial distrib

 

 

 

duringof dopants

yjunctionsa histicated ex-

tion depths continue usly decre

 

se: in

 

 

0.1 technology (year 2007), they

in the

 

 

 

out

 

of

the

implanted regi n

 

 

 

 

 

 

 

subsequen

 

process ng. For

 

The ions create Si in

 

rstitials

(Si and v

 

 

 

 

steam(Si ). Sincedopants di use

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lty lies in channeling

 

 

ects and

will be 150 to 450 A (15 to 45 nm). The di c

 

 

 

reviewsdi usionimplantation

 

and its relationthe

di u

on, see [33]

 

 

[31].

di usion

ra

 

modeling

provides initial

 

 

 

 

 

 

 

 

 

 

 

 

imulation ofand

 

 

 

 

 

 

 

 

of SIMS da

 

 

. This mayconditionsrun ut ofor

 

 

 

for near-0.1 m structures.

by poinolation-d fect medi

 

 

 

mec

 

 

 

 

 

(Tab. 3), the distribution of Si , Si

 

 

and

their clust rs (s eds ofedxt nded

 

 

 

 

 

isacancieded for further mo

eling of dif-

fusion.

Appropriate

 

atomistic simulators hav

 

 

been

presented [24, 25, 26],

but

can accoun f

 

 

 

 

 

 

 

 

and

 

defects)-mediated di nusigurations(section 6.1), they

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

V

 

 

 

and on thermal

more data is ne ded on the intehanismsdiate cluster c

 

 

 

 

 

 

 

The basic theoretichannelingmethods which

 

 

helpshalloobtain the

 

issi

 

I

 

 

V

 

 

 

 

 

 

g data are:

e ects

 

 

de ect migrat on

 

 

r com

 

 

 

 

 

 

 

 

[25, 27]. Though TCAD models

not good enough for heavily doped

 

 

 

 

 

 

 

 

 

 

 

 

wer than 100 nm.

 

 

 

 

 

 

 

 

1. Quantum mechanics. The mostjunctionsbinationex but

he most

CPU-

 

 

nsive; hence

 

only small systems can be examined. Static calculations dominate, dynamic

 

 

speciesB

 

ype

 

 

atomicmass

 

importance

researchmature

di usivity vehicled usion

 

 

 

 

 

P

 

 

 

 

 

 

15

 

 

 

 

 

 

 

 

 

 

 

high

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

In

acceptor

 

49

 

 

 

high

 

advanced

medium

 

 

 

 

, Si

 

I

 

 

 

 

 

 

As

donor

 

 

33

 

 

 

 

SiI

 

, Si

 

 

 

 

 

 

 

Sb

 

 

51

 

 

medium

 

 

olving

 

 

 

low

 

 

V

 

 

 

 

 

 

orate the dopant pro le of

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

V

 

 

[28]

 

 

 

 

 

 

submicrometer device. Ga di uses fast in SiO

 

.

 

 

 

 

 

 

 

 

 

 

 

Table 3 Shallow d

pants. Note the absence of fast-di using Al; it would deteri-

 

 

simulations

 

 

needed.

Can v

 

 

 

 

 

higher-level methods

[29].

.

 

 

2

Bette

oupling

 

MD

 

 

experimealidate/calibrateimprov mo

 

of

 

 

 

 

 

 

 

 

 

Classical

 

 

 

 

 

dynamics (MD). Su

 

for low-energy impac

 

3.

[24, 26]. Problarems with computational reso

 

cessfuland ph

 

intuition. Hits

 

ina y collisimolecularns (BC). The leadi g atomistic appr

 

ach [25, 30, 31, 32]

 

 

its limits at am

rphization and

 

fect evolution for

> 1 ns.

 

 

 

 

pro les.

 

 

Monte-Carlo based. Successful for ballistic processes, y elysicalion ran

 

 

 

 

A

 

c y

f models ranging from the lowillest (ab initio) to the hidamageh st v l

(whichierarmay adopt

Mto

 

C

rlo for self-ann

aling, di u ion, and clustering of d

 

fects [25]),

with

 

 

 

 

 

and

 

 

 

 

 

 

 

w delingul greatly

 

hance

he predictability. Self-a

nihilation and thermalparameters,e ect (dependence of

 

 

 

 

 

 

-

how species lik

N and Si a speciesecthanneling and

 

 

. Si can be

 

 

 

 

impland

ation and di usionconsistentemperature

[17])

 

 

be then properly included.

more

to

Mostly sim

 

 

 

are the

 

ransferablof T

. 3. It would be us ful to

learn

 

 

 

the crystal in order to reducecouldhanneling and sp ed up s

 

 

 

 

 

 

 

 

examples wstripesulatedimulation plays a mi or role. Sdimulationusion

 

tends

to belicida\s ep

in narrow

 

 

. N implants help

 

ontrol d

 

. Th se are

 

importan

 

 

 

hind"amorphizetec nology; the issue is how to get \in-step" with

 

ec

 

 

.

 

 

 

 

 

 

 

 

 

 

Brie y, in order

to mo el shallow junct on ,

TCAD

ust

 

interaction

main challenge is in making modules interact, not in improviincludenghnologyac module.

 

between implantation

 

d di usion. This illustra es the key ssue

 

FEOL: the

6

 

Models for dopant

 

 

 

many deta

data, including

 

 

 

 

 

 

 

 

Research on di usion

 

Si has provi

 

 

pro les

models and the in uence of various

oc ss conditions

[34]. But dopan

 

for new

 

 

can hardly be re

icted

without short loop experiments for

 

 

 

 

. \Ramping"

 

 

 

 

 

dec

asing of the waf

 

 

 

atomistic

 

 

The traditional approach todi usionis to use

coupled

di erential equations

round 100 K/s) is an extra

omplication. Dopants may di use during ramping

calibrationlready abovdesigns600

o

C. The(increasinglo al temperature is di c lt to measure[18temperature,19].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

alloweq[34]ilibrium. Dynamicsto accoun. Recenofforditoolsinhomogenusion vvideehiclesouson(Sidistribut-equI, SilibrV [35,umofdy36, am37],csTabfdefectspoin. 3) isdefeclimwhichtedandareto

sources/ nks of Si

. The key technical details to be extendedwork out are:

n barriers,

1. Di usi

 

para I

ters for Si

 

and Si

 

 

: fo

ation energ es, migra

 

 

pre-exponentials, dependence

on the F rm -level,

 

hem cal composi ion,

 

 

[29]) data exist, but are often con

V

 

 

 

 

.

 

i usivities

 

of Si

f om

metal

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

(ab in

 

 

 

[39], tight-binding

 

 

strain. Experime tal [27, 38]

 

 

 

 

theoretical

 

 

 

 

 

 

 

 

 

 

and from the

 

 

 

depeand

 

 

 

nce of TED

(section

6.1) diI er by orders

2 di3D usionlvers and statistical analysis ftradictorystandard TCAD. Beyond 0.1 m, statis

 

 

of magnitude. Trapping of Si

 

has been suggested as an explanation [40].

 

 

tics

 

f

 

efect and dopandepthdistribution will play a role [4].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

accuracy ( ection 10).

3. Metho to measure 2D and 3D pro les with su cien

 

A

umber of

 

basic prob ems

 

 

 

 

 

 

 

un olved. The

reaction

paths of defect

ediated di usion

are

 

still

 

 

 

 

 

 

. This includes

 

 

i usion in heavily doped

ditions

(notably, at

 

 

 

 

/Si(001)doubtful rf

 

ces)completeustbe properly de ned. Di usion

in polySi is di cult to handle. Thesremain(and other) issues are addressed below.

materials. Transient di usion e ects need a

 

 

 

 

 

description. Boundary con-

6.1

 

 

 

 

 

 

 

 

 

 

 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Di usion in SiObulk

 

 

 

 

 

 

 

 

 

 

 

 

pair

 

 

 

 

 

 

 

dopants and point de-

 

 

 

 

di usion is

 

 

 

 

only described b

 

 

 

 

 

 

 

fects are ass med

comform mobile pairs.

They

can explain why dop nts di use

Dopanwith lower ac ivation energy than nativ

poin

defects (Si , Si

 

). But the errors

resulting from the simp

 

about

 

of in

 

 

 

 

 

 

 

models:v t been estim ted.

 

 

con isten

 

 

 

wledge

 

 

he

 

 

 

 

 

 

analysisb which

point

defects

mediate the

 

It

 

 

 

 

 

 

to dev

 

l p

 

consistenteractio

 

 

of experim

 

tal

data

 

without

cr ss sectionsknof

 

 

 

-dopani cationreactipathsdependequations the re ction

 

 

th. Advanced

atdi mistusion

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

V

 

 

 

 

, 42, 43]

 

 

eoriesdefectfor he interstitial [36, 41] and vacancy mediated

 

 

 

 

 

 

. Not

nly the f rm of di usion

 

 

 

 

 

but also such parameters as

mechan sms hav

 

been

 

 

 

 

 

 

 

 

 

they must be v

 

 

 

 

 

and expanded.

 

 

out

 

 

dirighcultexplanation,

 

 

 

tum-mechanical y eriv

 

 

 

 

 

 

20

 

 

?[36

 

 

 

 

 

 

 

ed potentials for defect-

 

In prac ice, only atomsattempted;injectisol d substitutionincludingtesrigorouslyact

cm

w dopantrations.

 

Dopan

 

 

 

 

 

 

is

 

 

 

 

 

 

eded.

con entrations abov

 

2 10

 

. T

sort

dopanthe in era

ions [42]enhancare

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

the kinetics of

One m

 

 

di usion

 

 

 

 

their

 

 

 

 

 

ation,Models ctivation

 

 

 

 

 

 

 

 

 

of atomicize

ext nded defect

 

 

or Si

 

 

 

 

quan

 

 

SiO /Si are mis

 

ng

 

 

such concen

 

 

 

such transientheye

 

 

ectsmany enhance or

 

 

 

 

 

 

 

 

 

 

 

.(formationF example, Si intersti-

 

Essential to thconsiderfabrication of shallow ju

c ionsdi usionthat di usion

 

a ected by

clu

 

ers o /and

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

2

 

 

and

 

 

 

 

 

 

. Useful TCAD

 

 

 

 

 

 

v

 

 

 

 

 

 

activsi icide),

 

 

 

 

 

 

 

 

ex

ist, but

nsic

 

 

 

 

 

 

 

y limitations and

lot of room

 

 

reseshallorc remmodelsin .

ess ustin

poinprecipitationdefects. Dep nding on the migration mech nism (Tab. 3),

tials enhance di usion of B and P andretarddi usion of Sb.

 

 

 

 

 

 

 

 

 

The

 

lassic Transien

 

 

 

 

 

 

 

 

 

 

 

Di usion (TED) is

 

 

 

 

by Si

 

 

 

 

 

 

 

 

 

 

 

fimplantation,113g defects afterstorednoin-amorphizingEnhancedOf Rangeimplants,(EOR) defectsre-causedi tedamorphiz. Its maiI producedgfea

 

are

d scribed by the

 

 

 

 

 

 

 

 

of

 

 

 

 

ev

 

 

 

[27],afterbu the initial condi

ion

 

unclear

 

ion 5). Description of v

 

ical and lateral TED

 

 

extensionsuresin

not

 

reliable, particultheoryly due clusterto certainaporation par

 

 

 

 

for Si

 

 

dynamics

TED at low (seccentr

tions

(n ar to steady-state conditions) is well understood,I

Models built on

 

at high

 

of

 

few easy to measure

 

metersers are

 

 

 

 

 

 

.

 

 

 

but

 

its modeli

 

 

 

 

 

 

trations (far from

 

 

 

-

 

 

 

is probl

 

 

 

 

TED is

 

suppressthedbasisy substitutional

C [27]; rst

 

 

 

 

ate)mod ls

 

 

rematicvail-

able. The

 

 

 

 

 

questi

concenis the alleg d enhancemsteadyof junction lneededakag

 

y C.

Some

 

xidizingpressingterface

jects Si . The

 

also be

enhanc

 

 

i u ion (OED

[34])

 

 

 

 

 

 

 

 

 

 

 

 

th

 

 

 

 

 

F,

v

 

 

reparameort

to

 

 

 

 

 

 

TED [44 .

is under con .

 

 

 

 

 

 

 

xidiz Inghambienxidatsigoni cantlyredictivucea sect

 

 

 

 

umber of

ejected Si ,

impurities,robably duenotablycreation of Si

 

when the silicon is etched.

 

 

 

 

 

limi

 

at tempera ures at whic As is mobile. In consequence, As

 

 

 

 

 

msolubilittenddone

Anothersource of Si

 

 

is

 

deactivation

 

f As [45]. Activation is e ciently

 

 

 

 

lus er and

 

 

 

Annealing. TheEv 50% of

opants can be clustered. Small

As clusters

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

V

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

by Rapid ThermCl

 

 

 

 

 

 

 

 

 

(RTA) [46]: infr red lamps brie y h at the wafer

may be ejectedprecipitatef om the clusters, or the clusters

 

yIconsume Si

 

 

 

while Si

 

a

I

to a high

 

 

 

 

 

 

ture.

 

 

 

 

 

 

 

concentration of the activated As exce ds

discovery

contain Si

 

 

, and their

forma ion

l ads

to emission of Si . This

recen

 

 

[45]

 

calls

 

for mo

 

 

 

 

 

 

 

 

.

 

The

m

 

 

 

 

of the Si

 

 

 

 

 

 

is

 

not

 

clear. Si

produc(rati

 

of pointhe

defects to

 

 

 

deactiv mpurities)dopanen rgetics

 

 

 

 

 

 

 

 

 

sizeas,

 

 

 

 

 

 

 

 

V

 

 

 

 

 

 

 

 

 

 

 

 

 

is

 

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

 

Si

 

 

storage bins.

well. B clusters with Si

 

 

 

 

 

 

 

 

 

 

 

 

 

 

d;Other mplexeemissionact

 

 

M deling of TED,studiesactandv

ion,

 

 

 

d activatico

 

eeds

data on the cluster

 

 

 

 

 

 

 

 

 

 

 

 

 

ra es) of

 

 

able andhanismermedia

 

 

 

 

 

. Even groups uctureof w

 

 

 

 

 

 

 

 

 

 

 

volum

 

 

 

 

 

Frenkel p irs.

 

 

 

 

 

 

 

 

 

 

 

V

 

 

 

 

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

, notably B,

 

 

 

 

 

 

 

emission/capturedopan atoms [41] are

 

 

 

 

 

 

 

 

 

 

 

. On

the other extremclusters, dopan(clusterin ractions

Also

 

 

 

 

 

 

 

 

 

 

 

I

 

 

 

 

poinclusteredefects (Si

in the case of Ti

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

injec

 

 

 

nd Co). There

with

 

 

 

 

 

 

 

defects

mu

 

 

 

be studied. The case of polySi is

 

 

 

 

 

 

 

 

 

 

in 6.3.

both extendedsilicidationTi Co silicidation

are

measured [47]Whentoiject

SiV between 0.5 and 1

are ind

 

 

tions that th

 

 

importanis ot

 

 

jor issue.

 

 

SIMS artif

cts

 

 

 

 

absent,

times the equilibrium Si

 

 

concentra

 

at

 

V

 

 

 

betw addresseden 800 nd 890oC

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lm.

 

 

 

 

 

 

 

 

V

formatFinallycidationin Si volume due to thdi usivitiesstrain induced by the Si N

 

 

 

 

 

 

 

 

 

anism

 

 

s unclear: it

 

trationy be a

dir ctinsertiontemperaturesthe interface, or enhanced Si

 

Sil

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

V

 

 

 

 

 

 

 

 

 

 

 

 

 

f B and P by

 

fact

 

 

of two or less.

 

 

 

 

 

 

 

 

 

should thus reduce

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

, the concen

 

 

 

 

 

 

 

 

of Si

V

is enhanced by nitrid tion [48]

 

The mech-

6.2

 

 

Boundary conditio

 

s

 

 

 

 

 

 

 

 

 

 

 

 

 

3 4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A dopan

 

 

 

pro le is determined by the initial distribution of dopants and point

defects

 

(section

5) and by the physical borders (the bulk and an interface).