Al2O3 hosted attenuated phase shift mask materials for 157 nm

Rochester Institute of Technology

RIT Scholar Works

Theses

Thesis/Dissertation Collections

2005

Al2O3 hosted attenuated phase shift mask

materials for 157 nm

Yang Liu

Follow this and additional works at:

http://scholarworks.rit.edu/theses

This Thesis is brought to you for free and open access by the Thesis/Dissertation Collections at RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please [email protected].

Recommended Citation

Master of Science in Materials Science and Engineering

Al

2

0

3

Hosted Attenuated Phase Shift Mask Materials

for 157 nm

By Yang Liu

Submitted in Fulfillment of the Requirements for the Degree of

Master of Science

Approved:

Bruce W. Smith

Thesis Advisor

K.S.V. Santhanam

Department Head

Department of Materials Science and Engineering

Rochester Institute of Technology

Rochester, New York

I, Yang Liu, welcome and authorize anybody to photocopy or reproduce all or any part of

my print thesis.

Author:

Yang Liu

Date:

7i. Iff

H%>S

Table

of

Contents

Tableof contents i

Listof

figures

iii

Listoftables v

Acknowledgements

vi

Abstract 1

1. Introduction 3

1.1 Backgroundand

technology

trend 3

2.

Theory

13

2.1 ParametersofAPSM. 13

2.2 Materials screeningforAPSM. 15

2.3 ConsiderationofAPSMat1 57nm 16

2.4

Modeling

17

2.5

Modeling

ofthinfilmoptical properties 21

2.6 Photoresistpatterning:lift-offprocess 24

3. Experimental 27

3.1 Thinfilmdeposition 27

3.2 Determinationof n and k offhinfilm 31

3.2.1

Ellipsometry

31

3.3.2 Ellipsometer measuringprocess 32

3.3 Lift-offprocess 35

3.3.1 Processconsideration 35

3.3.2Imagereversal process 37

3.3.3

Processing

and analysis 38

4.1 Candidate

films

46

4.1.1

Al-NbOx

46

4.2.2

Al-TaOx

52

4.1.3 Al-CrOx 59

4.1.4

Al-MoOx

61

4.2 Lift-Off. 63

5.Conclusion 66

6. References 69

List

of

figures

Figure1.1

Moore'

slawandthefutureof

lithography

₄

Figure 1.2.

Microlithography

goestosub-wavelength 5

Figure 1.3 . Critical Dimensionvs.Numerical Aperturefor differentwavelengths 7

Figure 1.4.Depthoffocusvs.NumericalAperturefor differentwavelengths 8 Figure 1.5. Comparisonbetween

Binary

MaskandAPSM. 10

Figure1.6. The

0th,

1st

and2nddiffractionorder valuesforAPSMtransmissionbetween

0%

(binary

_mask)and20%with2%increments 12

Figure 2.1.Diagramof materials refractiveindexand extinction coefficientdiagramat

157nm ₁₅

Figure 2.2. n and k effect on reflectance ₁₇

Figure 3.1. IllustrationofPE2400ARF sputter system ₂₉

Figure32. Thefilmcomposition as afunctionof

A^

inDCreactive_sputteringofTa....30 Figure 3.3. Aschematicdiagramof a_nullingellipsometer_setup 32 Figure 3.4. J.A. Woollam WVASEspectroscopic ellipsometer 33

Figure 3.5. Dataanalysisflowchart 34

Figure 3.6 Image Reversal Lift-Offprocessflowchart 38 Figure 3.7 AZ5 21 4E coatingthicknessvs. Spinnerspeed 39 Figure 3.8 AZ5214EdilutedwithAZ1500thinner_coatingthicknessvs. Spinnerspeed.39 Figure 3.9 Crosssectionof photoresist profile

during

exposure and afterdevelopment..42 Figure.3. 1 0 Cross-sectionprofile afterPVDthermalprocess 43

Figure 4.1

NbOx

complexindex fromEMAmodel 47

Figure4.2

Al-NbOx

complexindex from EMAmodel 47

Figure 4.3 Transmissionof

Al-NbOx

thinfilm

by

simulationbasedonEMAmodel 47

Figure 4.4

Al-NbOx

complexindex frommeasurement 48

Figure 4.5 Comparison between modelingprediction andrealtransmissiondispersionof

Al-NbOx

48

Figure4.6 PhaseshiftdispersionchartforAl-NbOxfilm beforethicknessmodification.49 Figure 4.7 Phase shiftfor

Al-NbOx

filmafterthicknessmodification 50 Figure4.8Transmissiondispersion chartfor

Al-NbOx

50

Figure4.9 ComparisonbetweenEMAmodel generateddataand real data for

transmissiondispersion 51

Figure 4.10

TaOx

complex index fromEMAmodel 52

Figure 4.11

AI2O3

complexindexfromEMAmodel 53

Figure4.12 TransmissionofAl-TaOxthinfilm

by

simulationbasedonEMAmodel 53

Figure 4.13 Voltage VS. Powerfor

TaOx

54

Figure 4.14 Voltage VS. Powerfor

Al203

54

Figure 4.15

Ellipsometry

analysis softwarefitchartforfabricatedAl-TaOxthinfilm 57 Figure 4. 16Phase shiftdispersionchartforAl-NbOxfilm 57 Figure4. 17 TransmissiondispersionchartforAl-TaOx 57

Figure 4.18

Al-TaOx

complexindex frommeasurement 58

Figure4.19 AhO3 complexindex fromEMAmodel 60

Figure 4.20

CrOx

complex_{index fromEMA}model 60

Figure 4. Figure 4. Figure4.

Figure4.

Figure4.

Figure4

Figure4

Figure4

Figure4

Figure4

Figure4

22 Transmissionof

Al-CrOx

thinfilm

by

simulationbasedonEMAmodel 61

23

MoOx

complexindexfromEMAmodel ₆₂

24

Al-MoOx

complexindex fromEMAmodel ₆₃

25 Transmission dispersionchartfor

Al-MoOx

63

115CPEB/0.1 second of exposure 64

125CPEB/0.5 secondof exposure 64

120CPEB/0.3 second of exposure 64

120CPEB/0.2secondofexposure 64

115CPEB/0.3 second ofexposure 65

3 1 SEMpicture of cleaved patteredAPSMfilm

by

Imagereversal process 65

.32Microscopepicture of patteredAPSMfilm

by

Imagereversal process 65

26

Condition.

1

27 Condition.2

List

of

tables

Table 0.1 APSM filmrequirements: 2

Table 1.1. Intelprocessorsincreasespeed with moretransistorsinside 4

Table3.1 Basic depositionparameter

setup 30

Table 3.2Ellipsometer filmanalysis process 34

Table 3.3 FeaturesofAZ5214Ephotoresists 38

Table3.4. Lift-offprocedures 45

Table4.1 EMAmodel predictionfor

Al-NbOx

APSMfilm 46 Table 4.2 Al-NbOx fabricationparameter and characteristics 51 Table4.3 EMAmodelpredictionfor

Al-TaOx

APSMfilm 52

Table4.4comparison of

AI2O3

and

TaOx

depositionrate 55 Table 4.5

Al-TaOx

depositionparameter_{setup from}calculation 56

Table 4.6 Al-TaOx fabricationparameter and characteristics 59

Table 4.7 EMAmodel predictionfor

Al-CrOx

APSMfilm 61

Table 4.8 EMAmodelpredictionfor

Al-MoOx

APSMfilm 62

Table 7. 1 Materials'

Acknowledgements

Iwishtoexpress_my sincereacknowledgementto the

following

peoplefortheirgenerous

support and great contributionto_mythesiswork: _myadvisorDr. BruceW

Smith,

Anatoli

Bourov,

Yongfa

Fan,

Professor Carl Hirshman. I also would like to extend_my gratitude

and appreciation to the Microelectronics

Engineering

Fabrication

Laboratory

of

Rochester Institute of

Technology

and

Corning

Tropel for use of equipments and

Abstract

The Semiconductor

Industry

Association

Roadmap

2003 has put 157 nm optical

lithography

as the next generation

lithography

wavelength for the node of 70 nm

integrated circuits and below. The small departure from 193 nm puts more challenge on

imaging

tools and processes. One of the crucial areas is the exploration of thin film

optical materials formask coatings at vacuum ultraviolet

(VUV)

wavelength. Attenuated

PhaseShift Mask

(APSM)

isone oftheoptical enhancementtechnologies_pushingcritical

dimension of

lithography

to diffraction limits. As no _existing single material satisfies

those

demanding

requirements of

APSM,

non-stoichiometric composite materials will be

explored to

keep

optical

lithography

from demise. In_{this case, the} individual dielectric

response ofthose constituents must be combined in an effective model to reproduce

composite film dielectric response. The relationships between thin film microstructure

and optical properties are _extremely useful for the development and characterization of

thin films. Effective Media Approximation

(EMA)

theory

will bridge thin film

composition and/or microstructuretooptical properties.

APSMcan improveboth resolutionand depth offocus with _{complexity in}fabrication. It

has become an attractive candidatetoreplace thoseofmultilevelmask processes such as

aligning,

rim,

or sub-resolution schemes. The APSM should have transmission between

4% and 15%with an phase shift [1]. Alsothe thinfilm mustbe feasibleto pattern. This

thesis worked on several potential

AI2O3

hosted composite materials.

By

_carefully

combining the absorbing and _{non-absorbing} components based on the database ofRIT

properties ofthe composite thinfilm canbetailored toachievethe desired requirements

by

adjustingthe depositionconditions,whichinclude reactive gas partial_pressure,power

and time. Four _promising

AI2O3

hosted composite materials were proposed and two of

themwerefabricatedandtestedwith_satisfactoryresults.

Animage reversallift-offprocess was adapted and modified for_patterning oftheAPSM

film. The lift-off process was able to pattern the APSM film with 2 microns critical

dimension.

AI2O3

hosted oxide non-stoichiometric composite materials _satisfy all the

APSMfilm requirements and canbepatterned

by

lift-offprocesswithreasonable critical

dimension.

Criteria Target Range

Transmissionat157nm 4-15%

Transmissionat 193 nm below 50%

Reflectanceat 157nm Below 15%

PhaseShiftat 157nm 180 degree

Critical dimensionof patternedfilm 2micons

1.

Introduction

1.1 Background

and

technology

trend

Semiconductors,

sometimes referredto as computer chips or Integrated Circuits

(ICs),

contain numerous electrical _pathways, which connect thousands or even

millions oftransistors. These circuits

buildup

connections and _networks, which

fictionalize various chips to work for us. The microchip is everywhere in our

modernsociety.

Microchip

has been _getting smaller, smarter, denser and much _cheaper, which

are all driven

by

the advancement of the _{manufacturing} methods of ICs.

Microlithography

is one of the most _enabling technologies that make this

possible.

Microlithography

is the

technology

used to image a pattern from a

photomask onto a silicon wafer coated with a light sensitive material called

photoresist. Radiation is passed _{through photomask, exposing}photoresist on the

wafer and_creatingtheblueprintforthemetalthatwill be depositedonit. Wafers

MOORE'S LAW Intel" ttamum*2_Procoosor^

Intel'

ILaniumC_Procooaor^

IntelH Penbumn AProcoswfj

Intel8Pentiums ill Processor_j

tntet" Pentium** M_Processorj Intel Pentiums

transistors

1.000.000.00C

100,000.000

10,000,000

1.000.000

100,000

10,000

1.000

[image:13.546.79.426.305.577.2]

1970 1975 1980 1985 1990 1995 2O0O 2005

Figure 1.1 Moore's lawandthefutureoflithography.InpresentincarnationofMoore'slaw,the

storage

density

of_memorychipsdoubles_{every 18}months[4].

YearofIntroduction Transistors

4004 1971 2,250

8008 1972 2,500

8080 1974 5,000

8086 1978 29,000

286 1982 120,000

InteI386

processor 1985 275,000

Intel486

processor 1989 1,180,000

IntelPentiumprocessor 1993 3,100,000

Intel Pentium IIprocessor 1997 7,500,000

Intel Pentium IIIprocessor 1999 24,000,000

Intel Pentium 4processor 2000 42,000,000

Intel Itaniumprocessor 2002 220,000,000

Intel Itanium 2processor 2003 410,000,000

Table1.1 Intelprocessorsincreasespeed with moretransistorsinside.

More than a quarter _{century ago,} when Intel was

developing

the first

oftransistors on amicroprocessor woulddouble_{approximately every 18} months.

To

date,

Moore'slawhasproven_remarkablyright.

Some peoplehave been_predictingthedemise ofMoore's lawwhile some others

continually proved them _wrong for_years, no matter which side is

right,

it's for

sure that we must

keep

_pushing our ICs to be

faster,

denser and even cheaper

thaneverbefore.

Above Wavelength

Near Wavelength

&Oul*l SuiaWavetongth

l-Ot

OJSu/n

~i

[image:14.546.82.472.243.473.2]

r-i i 3 1 1 i 1 i 1 a i 1 i 1 1 i ! 1 i 1 i 1 i r

Figure 1.2.

Microlithography

goestosub-wavelength[23].

Microlithography

is the most important _step in semiconductor fabrication

process. Advances in _{microlithography} have been the

key

for _achieving the

constant _{scaling down in}line width ofIntegrated

Circuits,

imperative for speed

increase and

lowering

power consumption. Optical

lithography

has been

struggling for improvements because the critical challenges from _every aspects

like photomask materials, radiation _{source, photoresist,} as we push to next

193 nm are _routinelyusedtofabricate devicesat dimensions of_{sub-wavelength,}

and recent developments in 157 nm

lithography

has suggested that optical

lithography

may be used even further. Combined with _chemically amplified

photoresists, routine adoption of anti-reflective coatings and resolution

enhancement techniques

(RET),

we are ableto

keep

_using optical

lithography

to

patterneven smallerfeatures.

Generally,

there are three techniques ofRET.

First,

radiation wave direction is

controlled

by designing

special illuminators (off-axis

illumination,

OAI)

to

combines angle andpitchinformationof maskfeaturestoenhance certain spatial

frequencies inthe image.

Second,

wavefront amplitudeiscontrolled

by

_changing

aperture sizes and shapes (optical process _correction,

OPC),

to improve overall

fidelity

and reduce linewidth variation _according to appropriate adjustments of

certainportions offeatures on photomask layout.

Lastly,

local wavefront-phase

is controlled

by

_changing materials properties or _etching structures into the

surface ofthemask(phase-shiftmasks,PSM).

The smallest printable feature is called critical dimension or resolution.

Generally,

the smallerthe critical dimension of a

device,

the faster the speed of

the device.

Also,

shrinking the linewidth ofa device allows more devices to be

put on each wafer unit _area,

leading

to _{increased productivity} and profit. A

general form of optical

lithography

resolution _according to Rayleigh's criterion

R =

k1

NA

(1)

Where R is resolution (resolvable

image),

k^

is a process-dependent factor that

incorporates _{everything in} a

lithography

process that is not wavelength

(A,)

or

numerical aperture

(NA),

which is determined

by

the size and acceptance angle

ofthelens.

Critical Dimension VS. Numberical Aperature

e fl , 436nm

3

8

-4,

365nm

k-^.

^"~~>*~-.2 0.3 -(0

^^"^*

~~~~* -~-248nm ~~^~~~-_. " ?. C v c 02 ~"~^-~_ J

~~"~-~--^i

5 <

^^^r^^^iizr^

.

___193nm

u 0 . 157nm

0.5 0.6 0.7 0.8 0.9

Numerical Aperature

(NA)

Figure1.3. Critical Dimension_(CD)vs.NumericalAperture_(NA)for differentwavelengths.

We can see from Figure 1.3 the critical dimension decreases proportionally as

numericalaperture increases and wavelengthsdecreases. Forthesame numerical

aperture value, the smaller the illumination wavelength, the smaller the critical

dimension. Depthoffocus

(DOF)

needs to be considered _along with resolution

criteria for optical lithography. Depth of focus is defined as the distance _along

[image:16.546.82.444.262.484.2]

DOF =

k

X

2 NA2

(2)

Where

k2

is also aprocess-dependent factor. For a resist material of_reasonably

high _contrast, k2is 0.5. DOF decreases

linearly

with wavelength and as the

square of numerical aperture.

So,

we can see in order to get smaller resolution

while

keeping

the same level ofdepth of

focus,

it is more desirable to lower

wavelengththan toincreasenumerical aperture.

Depth ofFocus VS. Numberical

Aperature

436nm

365nm

248nm

193nm

157nm

0.5 0.6 0.7 0.8 0.9

[image:17.546.82.456.256.501.2]

NumericalAperature

(NA)

Figure 1.4. Depthoffocus(DOF)vs.Numerical Aperture(NA)for differentwavelengths

It can be see from Figure 1.4., depth offocus decreases as we go to shorter

wavelength or/and larger numerical aperture, but wavelength does less to

decrease depth offocus than numerical aperture

does,

this is because depth of

focus is

inversely

proportional to the square of numerical aperture while

Anoptical lithographicprojection systemis designedto transfer

image

_correctly

and accurately.Themaskisdesignedtorepresenttheidealpatternthat thecircuit

designer intends to put on the surface ofthe wafer.

Any

lack of

fidelity

in the

imagetransfercaused

by

themask will leadto an errorintheprinted circuits. To

deal with smaller

images,

tighter process latitude _tolerances, and greater data

volumes of modern semiconductormanufacturing, masksneedtobe designedto

minimize undesirable effects of optical diffraction. Also new materials and

structures are required for the shorter wavelengths lithography. The mask

industry

has resorted to sophisticated APSM in order to extend optical

lithography.

The concept of

increasing

resolution of lithographic image

by

_modifying the

optical phase ofthe mask transmissionwas proposed

by

Levenson in 1982 [2]. Several distinct types ofphase masks have been invented.

They

all share the

common feature that some non-transparent areas ofthe mask are given a 180

shift in optical phase relative to _nearby transparent areas. The interaction

between the aerial images of two features with a relative phase difference of

Binary-_{Mask APSM.}

Pbiu A W> B

I

Quarts

Chrome

ElectricFieldon Reticle

*r~i

n

n

JlpsorptivePhase Shifter

O

rt

ElectricField-_{ok Wafer}

H=F

Intensify

on wafer

iJ' i^* V^i* *i^

[image:19.546.76.461.51.266.2]

AAA/

II

Figure 1.5. Comparisonbetween_BinaryMaskand^PSM.

APSMreplaces the dark (less than 0.1%

transmission)

chrome with absorptive

phase _shifter, which can let 5-10% transmission pass through this area to get

sharper image at the edge between dark and transparent areas on mask while

those transmittedradiationthroughabsorptive phase shifter isnot _strong enough

to exceedthe thresholdvalue of currenthigh-contrastresistto exposuretheresist

ordegradethedesired image.

We can see this from Figure

1.5,

for the conventional

binary

_{mask, the} mask

function as a

filter,

the opaque area intentto get zero radiation while the open

area intent to get 100% transmission. Due to diffraction of

light,

the

intensity

under the opaque area is not _zero, which will degrade image. For

APSM,

the

opaque area is replaced

by

_partially

transmitting

_areas, which changes thephase

of light with 180 relative to the clear regions. The total opposite phase of

Owing

to destructiveopticalinterferenceattheedges of circuit

features,

a

phase-shiftingmask

(PSM)

improvesdepthoffocus and resolution. Becauseattenuated

phase-shiftingmask

(APSM)

can overcome phase conflict problemsfor _arbitrary

mask patterns and canbe more _easily fabricated than the othertypes of

PSM,

it

becomes the most _promising choice _{among PSMs. APSM} uses a _partially

transmitting

180phase-shiftingfilmtoreplacethechromium_absorbinglayeron

the mask. Interference between light from the transparent regions ofthe mask

and phase-shifted_{light passing}through the_partially

transmitting

absorber gives a

steeper slope ofthe aerial image offeature edges.

Usually,

transmission ofthe

phase-shifting absorber ranges from 5 to _{10% [21]. The partially}

transmitting

absorber allows afewof undesirable radiationto fall on_{the photoresist,} butthe

amount ofthoseradiationisnot enoughtoexposethephotoresist.

The common metric usedfor_measuringcontrastistheimage

log

slope,whichis

the slope ofthe aerial image formed at the interface ofthe clear and the dark

regions.The improvement intheimage

log

slope

by

_usinganAPSMascompared

toa conventionalmaskis illustratedin Figure 1.5.

From Figure

1.6,

we can see how

Ot, 1st,

2nd

diffraction orders are influenced

when anAPSMis employed _comparing with

binary

mask. The 1st and 2nd

order

diffractions of radiationarethemostcritical part to givethe informationto form

the latentimage inthephotoresist [1]. These plotsdescribe comparisonbetween

APSM and

binary

by

_changing transmission ofAPSMunder different feature

duty

ratios. As we can _see, APSM can decrease 0th

amplitudefor

1st

and

2nd

order ofdiffraction light_especiallyaround

duty

ratio of

1:1 (space to line). The advantage ofAPSMcompare with

binary

mask can be

seen

by

tailoring

the transmission ofAPSMand

determining

a suitable solution

for_everyspecific

duty

ratio featureto getthemost optimizedlatent image.

0.2 0.3 0.4 0.5 O.B 0.7 0.8 0.9

Fractionalspace width

20%APSM

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.

Fractionalspace width

0.3 -, 20% APSM

^= 0.2

-cr> 0.1

-Jp^ _Binary

a> "o

0 -0.1

-T3 O -0.2

-a> 00

-0.3 ^

O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 O.8 0.9

[image:21.546.90.368.186.598.2]

Fractionalspace width

Figure 1.6. The 0th, 1st and 2nd diffraction order values for APSMtransmission between 0%

(binarymask) and20%with2% increments. Fractional space widthvalue istheportion ofthe pitch value that consists of a space. Where 0.5 means 1:1 space/line ratio. As APSM

transmission

increases,

thevalue of0thorderdecreaseswhile 1st

and

2nd

ordersincrease.

2.

Theory

2.1

Parameters

of

APSM

From basic optical

theory,

phase shift ofincident radiation_passing atransparent

material is dependent on a films thickness

(f),

real refractive index of the

material

(

n

),

andthewavelength of radiation(X).

1.71

+

=(n-l)t

(3)

where <j> isthephase shift_value,real refractiveindexof airis 1.

Or the thicknessof materials

(/)

_giving exact 180 phase-shift canbe expressed likethis:

X

r=

2(12-1)

(4)

According

toBeer's law of

transmission,

absorption

(a)

oflight in_anymaterial is _mainly affected

by

the extinction coefficient

(k)

in the formula

(5)

and the relationshipbetween a andtransmission

(T)

is from formula (6):

4nk a =

(5)

T

= e~"

(6)

The _reflectivity of mask film is also _very important. It will cause _stray light or

flare in the optical system. And high level ofreflection will _severely degrade

incidenceangle s expressed as:

R=

[(n-l)2+k2]

[(+

D2+*2]

(7)

So the total phase change of light through APSMcompare with light through

open areas is expressed as formula

(9)

when the contributions ofinterface are

considered.

A0

=At/, ₊

H

(air-apsm) T^T^apsm-sub) '^T(sub-air)

(8)

Theseadditional interfacephase shifttermsare non-negligible as k

increases,

so

relativelower k value materials are preferred.

Mask_reflectivity should be lowerthan 15% as a rule ofthumb [20].

Excluding

the light reflectance offthe

top

and bottom

interfaces,

the transmittedradiation

throughphase shifteris

T

= ~

2 2

1-

n~nmr

. 1-

n~nsub

n+

nair

n+

nsub

At*.

(9)

A deviation within 5 is acceptable for anAPSMthin film [21].

Any

deviation

from 180

will reducethe image

log

slope. Eventheimage _qualitywillbeworse

forAPSMthan

binary

mask ifthephase erroristoobig.

So,

theAPSMthin film

should be

designed,

fabricated _correctly and _accurately to ensure the _matching

2.2 Materials

Screening

for APSM

n&kchart@1 57nm

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 ?a-Si Mo Nb

?_CrN VIoN

MoOx Ti ^Ta" MgF2_(b) ?Ta:>05 TiN ?TiO ? _TaN CrOx 2r ? NbN ?_TaN

0.65 0.85 1.05 1.25 1.45

Refractive index(n)

a^2(b)

1.65 ?Si02 ZrN Nb2C5(n) AI203(n) AI203(r) [image:24.546.86.465.83.331.2]

1.85 2.05 2.25

Figure 2.1.Diagram of materials refractive index and extinction coefficient diagram at 157nm

[22].

The Refractiveindex and extinction coefficient diagramis constructedbased on

the database of material evaluation forAPSM[22]. For

157nm,

where a narrow

wedgehas been identified formaterialsthatcan produce a 180phase shift with

transmissionbetween 4%and 15%atappropriatethickness. Thiswedgehas been

created based on first order phase calculation, neglecting interface effects as

describedinaboveFigure 2.1.

As suggested from the

data,

it is not

likely

that elemental thin films or

compounds exist which_{satisfy APSM}requirements at 157 nm. Inspection ofthe

optical properties of these materials does

indicate,

however,

that

A wider range of material properties are possible

by

_relaxing stoichiometric

requirements.

By

_controlling the ratios of material components

during

deposition,

optical behavior can be modified.

Furthermore,

by

_{co-sputtering} or

by

forming

graded

films,

some candidates are possible.

Throughuse ofthe datain Fig.

2.1,

several material combinationshave potential

forapplication asAPSMat 157 nm canbeproposed. Theinterest inthis thesisis

to research several groups ofthin films hosted

by

AI2O3

combined with other

oxide materials asAPSMfor 157nm.

2.3 Consideration

of

APSM

at

157nm

The basic requirement for APSM film is to offer a 180 phase shift of

transmitting

157nm radiation with transmission around 10%. At the same

time,

reflectivity shouldbeas lowas possiblebecause_anyunwanted_stray oflightwill

deteriorate image _quality and resolution. Andmostofthereflectance is fromthe

APSMfilm surface. From Figure

2.1,

wecan concludethat inorder toget small

Complex refractive index effect on reflectance

~. 1.25

"SS 1 a> u

1

0.75

o u

U.b o

Z 0.25

S

Lkl ₀

1.25 1.5 1.75

[image:26.546.83.409.45.260.2]

Refractive index(n) 30%reflection 20%reflection 10%reflection

Figure 2.2nandkeffect on reflectance.From_toptobottom isa set of complex indexchart of

threeseries of materials which can offer_30%,20%and _{10% respectively}ofinterfacereflectance betweenfilmand air when radiation reflect uponthefilm.Calculation is basedonformula(7).

Also the transmission ofalignment _radiation, which is needed to position this

reticle _correctly inthe stepper needs to be lowerthan 40% [21].

Normally

the

inspection or alignment wavelength is longerthan the exposure _wavelength, so

probably 193 or248nmisusedfor 157nmstepper.Allthese above requirements

place a _very narrow range forus to select ourAPSM_materials, which is almost

impossible to find in _{any naturally occurring} elemental or composite materials.

So we need to move our sight from stoichiometric materials to

non-stoichiometricmaterials.

2.4

Modeling

In order to characterize the film accurately, a realistic model of complex

refractive index of for the thin film materials is needed. It must be able to

non-uniformity, grain

boundaries,

voids and other disordered regions. Besides the

homogeneous _materials, it is _{usually necessary} to determine the optical

properties ofheterogeneous or composite systems and superlattice structuresthat

are more _accurately described as mixtures of separate regions oftwo or more

materials [5].

Generally,

heterogeneousmaterials consist of microscopic regions

that are orders of magnitude smaller than the wavelength of

light,

yet large

enough to retain their own dielectric properties.

So,

we need to take into

considerationthose individual components ofthe composite materials and their

relationship for our model. Grain

boundaries,

voids and other structural

inhomogeneities affect the uv/vuv optical properties ofmaterials. The dielectric

response of an individual homogeneous region depends on its composition and

presence of

long

range order (such as grain size). Optical constants of materials

define how light interacts with amaterial.The real partor index of_{refraction, n,}

definesthephase_velocityoflightinmaterial:

n=c/v

(10)

Where v is the speed of light in the material and c is the speed of light in

vacuum.The

imaginary

part or extinction _coefficient,

k,

is

directly

related to the

absorption of material.lt is related to the absorption coefficient a_, X is the

wavelengthoflight. Opticalconstants_varywithwavelengthandtemperature. To

predict and interpret thin film optical

behavior,

appropriate models based on

constituent materialproperties andcertainassumptions arerequired.

M=aE'

(11)

P = _Na E'= _{E for}

gases but not solids.The electric displacement vector D is

related to E

by

D=e E = _{E+4np, where s} is _{the dielectric constant or}

function,

which_maybea scalar ortensor.

So,

we can_say

s=1+4tiP

(12)

The local electric field _usually differs from the applied electric field (from the

light _source) because of the field produced

by

the local polarization. For

isotropic media, thesefields are related

by

E'=

E+4tzP/3,

thelastterm

being

the

Lorenz fieldproduced

by

local dipoles.

The microscopic polarizability<2 is related to the dielectric constant

by

the

Lorentz-Lorenzformula

a =

(

)

4ttN s + 2

(13)

Optical constants

(n,

k)

are _routinely related to electrical properties

(,

_") of

materials

by

_neglecting material structure and_considering macroscopic material

quantities only. The realand

imaginary

parts ofthedielectric constant andindex

ofrefractionarerelated

by

t?=n2-k2

(14)

^=1nk

(15)

(16),

(17)

Inordertoaccountformaterial _structure, Drudeanalysis of optical and electrical

constants _{describes free} electron or metallic behaviorof materials quite well in

the visible and IR region. To account for optical properties at shorter

wavelengths, Bound Electron

Theory

needs to be utilized. For dielectric

materials, no intraband transitions exist because of filled valence bands.

Interband transitions are also limited inIR and visible regions because oflarge

band_gap energies. Bound electron

theory

aloneissufficienttodescribe classical

dielectric material behavior. Characterization of metallic and _{non-insulating}

materials in UVand visible regions requires use ofboth freeelectron and bound

electron theory.

By

_assuming a given number of free electrons and a given

number of harmonic _oscillators, optical properties of materials over a wide

wavelength range canbe described.Modelsdescribed hereprovide a quantitative

descriptionofthemacroscopicdielectric functionof aheterogeneousmaterialin

terms ofits composition and microstructure [4]. Since precise microstructure is

unavailable and would be unpractical for use even if absolute limits on the

dielectric function are obtained. Assumption of macroscopic

homogeneity

has

been made, which in practice means that _any

inhomogeneity

considered has a

structure size smallerthan 0.1X. Optical constants

(n,

k)

are _routinelyrelated to

electrical properties

(s,

s"

considering macroscopic material quantities only. In order to account for

material structure, Drude analysis of optical and electrical constants

describes

free electron or metallic behavior of materials quite well in the visible and IR

region[4].

2.5

Modeling

of

thin

film

optical properties

Grain

boundaries,

voids and other structuralinhomogeneitieswill affect UV/VUV

optical properties of materials. The dielectric response of an individual

homogeneous region depends on its composition and presence of

long

range

order (such as grain size).

Screening

charge develops on grain

boundaries,

causing difference between local and applied field. The effectiveness of this

screening effectdepends ontheshape andsize ofthe grains. Sincethedielectric

function is definedasthe dipolemomentper unit_{volume, the}

density

ofthefilm

should also affect the dielectric response. Models should be able to provide a

quantitative description of the macroscopic dielectric function of a

heterogeneous material in terms ofits composition and microstructure. In order

to account for all the above requirements, some assumption are _{necessary, the}

thin film material is macroscopic

homogeneous,

which means _any

inhomogeneity

shouldhasa structure size smallerthan0.1 A.

Ifaphaseis

homogeneous,

itsdielectric functioncanbeobtainedfrom literature.

Microscopically

inhomogeneous phases are often modeled _using the effective

medium approximation

(EMA)

(Aspnes,

1982and

Chang,

1989"

Jellison,

1993).

When the dielectric functions of composite constituent materials (a and

b)

are

available, the dielectric constant ofthe composite material can be expressed as

the Lorenz-Lorenz effective media approximation below. Where

fa

is the

volumefactionof one material a.

g-1 = f

ea-\

f eb-l

s +1 ^a+2 A+2

(18)

Inthemost_widely used

EMA,

the effective dielectric constant s is given

by

the

Bruggerman expression. This assumes two _randomly mixed phases with

dielectric constants -0and _eb with fractions

fa

and

fb

respectively. The

dielectric constant ofthe composite material can be expressed

by

the

Lorentz-Lorenz effective approximation and also can be extended for more than two

phases. The Maxwell-Garnet EMA is an alternative approach that assumes

layeredphases.

Ageneralexpressionofthedielectric functioncanbeexpressedas:

c=eaeb+e{faa+fbeb) -=(l~gk

e +

ifaeb+fbea)

_q

eh is dielectric function of the host material

(typically

material with

higher

composition) and_{q is}thedepolarizationfactor (microstructureparameter). _{If q}is

0,

it means all the boundaries inside materials parallel to the field applied and

q=\ means alltheboundariesperpendicularto the fieldapplied. When qbetween

0 and 1 represents practical situation. For example, #=1/3 means an atomistic

level mixed material [7]. EMA is based on a solution of the Lorentz-Lorenz

problem and provides the connection between the microstructure of a

heterogeneousthin filmandits macroscopicdielectricresponse s, where

fa

and

fb

are the volume percentage of each constituent materials in the composite

materials.

Optical properties of a composite thin film _{may be} explained in terms of its

anisotropic electrical properties on a scale largerthan themolecule.

Namely,

the

ordered arrangement of similar particles of_{optically isotropic} material's size is

large compare withthe dimensionsof_molecules, but is lessthan thewavelength

oflight. Let'sconsidera simplified case of regular_assemblyof alaminate ofthin

layers of component separated

by

thinlayers ofcomponents underthe condition

that thelinear dimensions ofthefaces oftheplates are large butthe thickness _tj

and _t2small comparedto thewavelength.

The models have beenexpanded to include more_{than two constituents,} whichis

useful _especially for composite _masking films. In _general, if el and _s2 are

essentially independent of microstructure. If _sx and _e2 are _{substantially}

different,

thenmicrostructureisthedominantvariable[20].

2.6

Photoresist

patterning:

lift-off

process

Generally,

positive photoresist is a three-component mixture of an alkaline

soluble novolac_resin, a photosensitive dissolution

inhibitor,

often called Photo

Active Compound

(PAC)

and a carrier solvent. PAC serves to retard the

dissolution rate of the novolac resin in alkaline solutions. When exposed to

proper wavelength and sufficient _{energy, the} PAC will undergo photochemical

decomposition,

whichrendersthepolymer soluble inanalkalinedeveloper. The

PAC is a

diazoketone,

which upon exposureto ultravioletradiation generates a

highly

reactive intermediate ketene and liberates nitrogen. The ketene will react

with available waterto form an indene carboxylic _acid,which is now soluble in

thebasic developer. This is calledpositive photoresistbecausethe clear areas of

the mask remain as clear areas in the photoresist. The molecular weight and

exact percentage of solids is

tightly

controlled to assure exceptional film

thickness_{uniformity [15].}

Image reversal of positive resist is capable of

improving

_resolution, sidewall

angles, and process latitude. Based on novolac resists, new _chemistry with

'blocked'

reactive agents has been developed. The reactive agent can be

thermally

liberated fromthe resin, orbe_{photochemically}generated. Itwillreact

insoluble after

being

generated. The flood exposure converts the _remaining

photoactive compound in the resist layer after postbake.

Finally,

development

gives negative image ofthe mask. _{The crosslinking} betweenreactive agent and

indene carboxylic acids groups

during

_processing improves the photoresist

contrast because both ofthem are photogenerated. _{The crosslinking} is a second

order reactionbased ontwo photogenerated_{species, the} concentration gradients

in the film are sharper than a single PAC _system, which follows first order

kinetics [19]. This featurewill improve_resolution, contrast and process latitude.

Anothergreatbenefitofimage reversal isgoodthermal _stabilitycomes with the

formation of crosslinked negative

images,

which is _{very helpful} for PVD

deposited APSM. Ourfilm is fabricated ina relativehotter process

(>100C),

so

the thermal _stability ofthe photoresistis a major concernand it's related to the

pattern ofthefilmdirectly.

The kinetics ofthe image reversal is dependent on the initial exposure _energy,

temperature and duration of the post exposure bake and various ambient

conditions, which all can be explained

by

the _crosslinking mechanism. Upon

exposure, the photoactive compound generates an acid. Because ofthe unique

structure ofthe photoactive compound used inthe

AZ5214E,

this acid is much

strongerthanis

typically

generatedinmost_commercially available photoresists.

Following

the _{relatively high} temperature ofthe post exposure

bake,

this acid

diffuses through the resin system of the film and cause acid catalyzed

the acid through control ofthe quantity of the acid photogenerated and the

temperatureanddurationofthepost exposurebake.

AZ5214E is intended for lift-off

techniques,

which calls for a negative wall profile. Although

they

are positive photoresists comprised of a novolac resin and

naphthoquinone diazide as photoactive compound.

They

are capable of image

reversal _resulting in a negative pattern ofthe mask. In fact AZ5214E is almost exclusively used in the IR-mode.

Photosensitivity

of a resist has its origin on photoactive compound

(PAC),

which is consumed

during

exposure. Once

fully

exposed, the resist volume will no longer contain unreacted PAC. In

AZ5214E,

catalyst is formed _{simultaneously} inthe volume of exposed resistthat promotes polymer _{cross-linking}

during

_any heat treatment _{later. The crosslinking} agent

becomes active at temperatures above 110

C

[13].

They

are _only active in exposed areas while the unexposed areas still behave like a normal unexposed positive photoresist. Polymer cross-linkingwill makethe exposed resist volume

more resistant against developer attack. After

baking

the exposed _{sample, the}

image reversal process is completed

by

_exposing the whole sample area. The purpose of a flood exposure is to turn those resist areas that were _originally shielded

by

a mask soluble to

developer,

which means

they

still function as positive resist. Cross-linked areas contain little or no

PAC,

so flood exposure will not reduce _{solubility considerably} there. After

development,

the pattern of

3. Experimental

3.1

Thin Film Deposition

Thinfilm depositionwas usedto_applythincoatings ofAPSMfilms.Thin_{films generally}

have thicknesslessthanafewmicrons.

They

canbe asthin as a_{few hundred}angstroms.

Thin film depositiontechniquesrequire a vacuum environment.

Physical Vapor Deposition is a

technology

in which the material is released from the

source at much lower temperature thanevaporation. The substrate isplaced in a vacuum

chamber with the source _material, named a

target,

and an inert gas (such as _argon) is

introduced at low pressure. Radio

Frequency

power source is used to ignite _plasma,

causing the gas to become ionized. With

RF-sputtering,

the _sputtering plasma is

generated

by

an_oscillating power source. It can be used to depositboth conductors and

insulators,

hencecompound materials.

During

_sputtering,ions are acceleratedtowards the

surface ofthe_{target, causing} atomsofthe source materialto breakofffromthe targetin

vaporformand condense on all surfaces ontheirpaths. A schematic diagramof atypical

RF sputteringsystem isshownintheFigure3.1.

Thin film process

by

sputter deposition has _many advantages. The low-temperature

processing is a unique feature that _{distinguishes sputtering} from evaporation. The

temperature is lowenough notto degrade thephotoresistI used forthe definition of our

APSM films

during

next image reversal lift-off process. The sputtering deposition is

implemented in a plasma environment rich in charged particles.

Therefore,

electric and

Through

_{low energy}ionirradiation

during

filmgrowth, themicro structural control ofthe

film

might be realized. At the

time,

_many species in the plasma are more _chemically

reactive because

they

are at their excited states. The deposition of a compound film is

therefore possible

by

reactive _{sputtering instead} of _sputtering a _compound, which

requires a harsh high voltage condition or a CVD process at high temperatures. The

sputter rate of aluminum and other candidate materials will be studied as a function of

power. In _{addition, the} basics of _reactively sputtered deposition thin films will be

introduced.

Specifically,

the effect ofthe composition ofthe sputter gases ondeposition

rates,and optical propertieswillbestudied.

Asimplified

depositing

systemis illustrated inFig.3.1. Insidethe_{workingchamber, there}

is a pair ofparallel metal electrodes. The cathode is the target to be

deposited,

andthe

anode isthe substrate. The anode is _usually grounded while the cathode is connectedto

the negativeterminalof aRFpower_supplywith atypicalvoltage of several kilovolts. A

working gas,

typically

argon is introducedto the chamber to serve as the plasma media

afterthechamberisevacuated. The pressure ofthe_working gas_mayrange fromafewto

ahundredmil-torr. An appropriate voltage andpowerare _necessarytomaintain a visible

glow discharge between the electrodes. A very small portion ofthe _{working gas,} e.g.

argonisionizedintheglowdischarge:

+

Ar+

(21)

The electric field drives the positive ions toward the cathode. Given sufficient _energy,

they

can_physicallysputtertargetatomsthroughmomentumtransferwhen_strikingonthe

target. If surviving collisions with molecular particles and electrons in the media, the

Figure 3.1.IllustrationsofPE2400ARFsputter system configuration and outside view.

Ourvacuum system uses one mechanic_pumpto_pumpfromatmospheric pressureto 10"3

Torr,

and another _cryo-pump to achieve highvacuum from

10"3

to 10"7 Torr. The

cryo-pump onlywork undervacuumoffered

by

themechanical_pump, bothofthepumps must

beon whenthePVDsystemisworking.

In the presence of a reactive gas

during

the process of _sputtering, thin films of

compounds are deposited on the substrates. The most

frequently

employed gases are

oxygen to form oxides, nitrogen to form nitrides.

Depending

on the

depositing

conditions, the resultant film could be a solid _alloy solution ofthe target metal doped

with the reactive element, a _compound, or some mixture of the two. Exemplified in

Figure 3.2 is the reactive _sputtering deposition of TaNfilm at DC voltages of 3000 to

5000 voltages andpressure of30 mTorr. As

illustrated,

the composition ofthe film is a

500

asoo

100-u

-t \M I I_I ' Mill r (

i ii fagM^TttW ItaN

10--rr-ilfiOO It E vzoo _S TON RESISTIVITY k-J-800 c 4x , t-wc f, 10^ O0 K)^ 5^^0, Sxlff4 PARTIAL PRESSURE OF MITROGENttorr] 3

Figure32. Thefilmcompositionas afunctionof_N2in DCreactive_sputteringofTa[9],

By

controllingthepartial pressure and_voltage,thecompositionofthedeposited filmcanbeadjusted.

Reactive _sputtering has advantages over

directly

_sputtering compoundtargets. Note that

pure metallic targets are cheaper and easier to

buy

than compounds targets.

Also, they

allow to make thin films with diffe ent compound or _elementary thin film just

by

controlling process conditions. Our APSMthin films were fabricated

by

Perkin Elmer

2400A PVD system, whichis aRFmagnetron-enhanced_sputteringsystem. The filmsare

reactive sputtered and somefabricationparameters arelistedin Table3.1.

Power 500-1500w

BasePressure 2E-6 Torr

Deposition Pressure 8-12mil Torr

Pre-sputter Time 15 min

Deposition Time 50-90min

Rotation Speed 2rpm

ReflectedPower 0microAmpere

Reactivegases

Argon,

_oxygen,nitrogen

Targetsize _{8 inches}

[image:39.546.73.265.53.186.2]

Substrate _{High quality}quartz andSiwafer

[image:39.546.68.479.447.659.2]

Priorto

deposition,

all substrates are cleaned anddehydrated.Filmsaredepositedwithout

additional substrateheating.Plasmaisignitedtowarm_upatleast 15minutestoeliminate

any oxidized or reacted materials onthe target so pure targetmaterialisexposed laterto

plasmabombardmentto depositpurer films. _{After pre-sputtering,} shieldbetweenplasma

andtargetsare removed

by

shield rotation. Target isstrike

by

plasma_particles, so that the

wanted materials are stroke offto substrate onthe deposition table.

During deposition,

the substrate table is_rotating at speed of2 rpm, which gives more evendeposited film.

The target is heatedwhenthehighvoltage is applied and alsothe plasmais

hot,

a water

coolingsystem is hooked_upto the backsideof allthe targets to getridoftheexcess heat

generated

by

thePVDprocess.

By

_{setting up}thegas ratiobetweenargon and reactive gas

suchlike _{oxygen, the} decisiveparameterofthe composition ofdeposited film is _onlythe

voltage appliedto each _target, which means we have easier control and_{less complexity}

during

depositionprocess.

3.2 Determination

of n and

k

of

thin

film

3.2.1 Ellipsometry:

Ellipsometry

is a sensitive optical analysis technique for

determining

properties of

surfaces andthin films. Polarizationstate ofincidentlightwillbechangedwhenreflected

fromasurface.Thechange oftheshapeandorientationoftheellipsedependontheangle

of

incidence,

the direction of the polarization of the incident

light,

and the reflection

properties ofthe surface. We can measure the polarization ofthe reflected light with a

quarter-_{wave plate} followed

by

an _{analyzer; the} orientations ofthe quarter-wave plate

orientations and the direction of polarization ofthe incident light we can calculate the

relativephase _{change, A,} andthe relative amplitude _{change, \\i,}

introduced

by

reflection

fromthe surface. Anellipsometer measures the changes inthepolarization state oflight

whenit isreflected from a sample. Ifthesample undergoes a_change, forexample athin

film onthe surface changes its

thickness,

thenitsreflection properties will also change.

Measuring

these changes in the reflection properties can allow us to deduce the actual

change inthe film's thickness and or other properties. The _sensitivity of an ellipsometer

allowsthe detection of afew Angstroms change ofthin filmthickness. The ellipsometer

is used to measure vj/-and _A, and thus the film thickness and refractive index can be

calculated. Sincetheequation contains complex quantities,itssolutionis_usually done

by

a computer or programmed calculator. _{Because ellipsometry} measures the ratio oftwo

values,itcanbe

highly

accurateand_veryreproducible.

3.2.2 Ellipsometer measuringprocess

Figure 3.3. A schematic diagramofa_nullingellipsometerwith the quarter-_{wave plate placedbefore the}

light is reflectedfromthe sample. L: the light source; P:the _{polarizing prism; Q:}the quarter-wave plate

It shows abasic schematic diagramof a _nulling ellipsometer with a quarter-wave plate

positionedbeforereflection. Lrepresents an unpolarized monochromatic lightsourcethat

is collimated with a small beam divergence. _{A mercury}

lamp

with a green filter and

collimator serves this purpose. The _{mercury light is}

initially

linearly

polarized. When a

polarizedlight source is _{used, the} light is_circularlypolarizedjustafteritemerges froma

quarter-wave plate. The _polarizer,

P,

follows the light source followed

by

the

compensator, Q. The compensator is mounted in a holder so that its plane is

perpendicularto the light beamandthe angle measuredfromtheplane ofincidenceto its

fastaxis,canbemeasured. Theangleismeasuredthesame asP.

Following

thesample is

theanalyzer. The analyzer_angle,

A,

isagain measuredfromtheplane ofincidenceto the

analyzertransmissionaxis

by

an observer

looking

into the beam. I used a J.A.Woollam

WVASESpectroscopic

Ellipsometer,

whichhas a spectroscopic rangefrom 140 to 1400

nm. 157 nm is our interest of research. Also this ellipsometer use Variable Angle

Spectroscopic

Ellipsometry

(VASE). It means it can change the incident angles of

measurements

during

each set of_measurement,which can offer more informationabout

thefilm_propertyandbuildmore reliable and realisticmodeltoforellipsometric analysis.

Figure 3.4. J.A.Woollam WVASEspectroscopic ellipsometer.Thissystemhas lightrangefrom 150nmto

Mcasuremen

T

Model

T

Fit

1

Results

*, Exp, Data

njk

l/\

Compare

- *?

*

Fit Parameters

_ .

Thickness

Unilbnnil> '^S-Roujlmiss

[image:43.546.111.382.45.253.2]

k

Gen.

Data

Figure 3.5. Dataanalysisflow-chart [11].

1 Acquired experimental data through measurement. Data was

typically

acquired

versus wavelength andangle ofincidence.

2 Built an optical model upon our film structure and process _using as much

informationabout the sample as possible. Itis importantto accountforall layers in

thesample structure.

3 Generated theoretical data from the optical model that corresponds to the

experimentaldata basedonthedatabase.

4 Compared generateddatato experimentaldata

by

fitting

_{using WVASE}software. In

this process, those unknown parameters in the optical _model, such as thin film

thickness and/or optical _constants, were variedto

try

and produce a "bestfit" to the

experimental data. Regressionalgorithms are usedto_varyunknownparameters and

minimizethe difference betweenthegeneratedandexperimentaldata.

5 Physical parameters of the sample such as film _thickness, optical _constants,

composition, surface _roughness, etc. were obtained once a good "fit" to the

experimentaldata isachieved

[image:43.546.53.497.309.628.2]

3.3 Lift-off

process

3.3.1 Process consideration

The lift-offpartwas done atRIT micro-electronics _{engineering fabrication cleanroom, I}

conducted a series of experiments to_verify the _validityofimage reversal as atechnique

to produce the repeatable and controllable photoresist profiles needed for a composite

thin film "lift-off' process. One ofthe most basic mistakes one can commit

during

the

preparation of an experiment of

lithography

is to rush a proven-to-be good recipe that

originally developed

by

other testers. I modifiedDr. Hirshman's lift-offrecipe, which is

designedtopattern pure evaporated metalfilm (above 2000 angstroms). For_evaporation,

the depositionprocess is more ofline of sight process, which does not require negative

profile of photoresistbefore

they

put downthefilm as our process because forphysical

vapor deposition process plasma permeate _every space ofthe PVD chamber. Also we

needto make surethe _{higher PVD working}temperature doesnot "burn" or"reflow"the

photoresistwe usedtolift-offthefilm.

"Lift-off"

is a _{simple, easy} method to pattern deposited films. A pattern is defined on a

substrate _using photoresist. A film is blanket-deposited all over _{the substrate, covering}

thephotoresistandareasinwhichthephotoresisthas beencleared.

During

the actuallift

off, the photoresist underthe film is removed with solvent,

taking

the film with

it,

and

leaving

onlythe

film,

which wasdeposited

directly

onthesubstrate. Themotiveforalift

offprocess is the need topattern metal lines on semiconductor substrateswhere the use

ofchemical orplasma _etching is either undesirable or incompatible with theprocess or

Etching

these materials would require _very harsh chemicals that would _severely attack

themask substrate and decreasetheprocess latitude. Lift-off ismore desirable since the

solvents used to remove the insulator in lift-off cause less damage to the _underlying

substrate than do the etch processes.

Secondly,

_{lift-off is} applied when tight linewidth

control is required. For 157nm

lithography,

tight control of APSM mask _{linewidth is}

essentiallyto thecorrectiontransferofimage. Errorfromvariation of etch process willbe

eliminated sincelift-offprocessdepends_onlyonthecontrol ofthephotoresist.

Depending

on the type oflift-offprocess _used, patterns can be defined with _extremely

high

fidelity

and for_very fine geometries. For example, it's the choice for_patterning

e-beam written metal lines. To achieve good

lift-off, firstly,

_{it's necessary}to have aclear

separationbetweenthedepositedAPSMfilmontheresistlayerandthe^4PSMfilmonthe

open substrate areas

directly,

which meansthe depositedAPSMfilmis notcontinuous at

the edge of the resist _profile, so we can _actually "lift"

the unwanted APSM film

by

removingthephotoresistlayerunderneathit.

So,

the_controlling oftheresist edge profile

iscrucial,which means a negative slopeispreferredornecessary.

Image reversalprocess canturn a positive slope into negative

by

_{reversing 'everything'.}

It was first adopted

by

IBM and since then numerous ways ofimage reversal methods

have been tried [12]. The objective of the

following

experiments was to _verify and

modify a image reversal process for _{patterning my APSM} film. The process must be

capable of

being

used on a routine

basis,

in a repeatable and controllable manner. I

linewidth

by

SEMandspectrometer.

3.3.2 Imagereversalprocess

The AZ5200 Series positive photoresist I used for lift-offwere developed to operate at

shorter wavelengths to take advantage of the

inherently

better resolution. These

photoresists offer _{exceptionally} straight sidewalls and wide process latitude to meet the

growing demands of sub-micron lithography. It can image reverse to a negative tone

using post-exposure bake

(PEB),

_making it one of the most versatile photoresists

available. The AZ5214Ephotoresists physical and chemical properties canbe found from

"AZ 5200 Positive

Photoresists,

Hoechst Celanese Corporationdatasheet"

[16].

After analysis ofthe image reversal_{process, the} exposure dose and post-exposurebake

temperature were found the most important parameters

during

_{the process,} which are

crucialto thenegativeside-walloftheresistprofile.Alsothehard bake should notbetoo

high,

otherwise it will reflow the resist and degrade the negative edge profile. The

sputtering process ofthe APSM film is a high temperature process compare with the

lithography

process, which _{may degrade} the resist

by hardening

or

decomposing

it. We

should be able to monitor the sputter process to make sure the resist profile is not

changed

dramatically

by

PVD deposition. In addition, the PVD process is more

conformalthanevaporation process. Alltheabove areas needtobe checkedto make sure

we getthe goodlift-offresults. Its broadspectral _sensitivityinthenegativetoneallows a

1 Nearverticalprofilestogive accurate patterntransfer.

2 Wideprocesslatitude.

3 Excellentresolutioninbothmetal-ion-freeandinorganic developers.

4

Highthermal_{stability (extendedup}to 250

C)

tomaintain profile_quality

during

plasmaetching,ion implantprocesses andthinfilmdeposition.

5 Sub-micronresolution capability.

Table3.3FeaturesofAZ5214Ephotoresists[16].

3.3.3

Processing

and analysis

Based on our analysis and knowledge from experience and photoresist _properties, plus

recommendation from various literatures and

laboratory

_colleagues, I organized and

detailed the

following

_processing procedures ofAZ5214E lift-offprocess for the PVD

fabricated Attenuated Phase Shift Maskcompositethinfilms.

1 ;

1X1 XXXI

11

1X11111X111

wpp *m ssw "? i

~

T^? ''il-TT "'

.TJS.. .

JS-'- _

p~

war mm w

Substrate

should be free of organic contamination or excessive _physically absorbed

moisturetoeliminatetheadhesionproblemsandimprovethe _coatingresults [13]. Wafers

shouldbe baked for aperiod oftime

long

enough to evaporate the excess water out and

thenthe surface istreated

by

hexamefhyl disilazane

(HMDS),

which is spin coated. The

purpose ofthis _step is to introduce an intermediate layer ofHMDS molecules between

the wafer surface atomsandthe photoresist. HMDS surface layer is also ableto sealthe

sample surfacefrommoisture. The dehydrationand_primingprocess (is accomplished

by

the 4" SSI wafertrack (program #4: 200C for 5 minutes dehydrationbake and

105 C

for 45 seconds afterHMDS application, details can be checked at Table 3.4). Substrates

shouldbecoated_promptlyafter surface preparation. Take a spinnerwithavacuumchuck

to coat wafer withAZ214E.

Coatingthickness(urn)VS.spinnerspeed

(RPM)withpureAZ5214Ephotoresists[5].

1.5

0.5

2000 4000 6000 8000

Figure 3.7 AZ5214E coatingthickness_(angstrom)VS. Spinnerspeed(rotationperminute)

Coatingthickness (um)VS.spinnerspeed

(RPM)for AZ5214EwithAZ1500 thinner_(1:1)

*

I

0-6i

$

0.4

-o

o

K) ~-+

1

* _{0 2000} 4000 6000 8000

Spinnerspeed_(RPM)

Figure 3.8 AZ5214E dilutedwithAZ1500 thinner _(1:1) coatingthickness _(angstrom) VS. Spinnerspeed

The APSMthin filmthickness is about800 angstroms. The photoresist thickness forthe

lift-off

process should not be too thick or thin to function _properly to lift-off. The best

photoresist thickness is about 3-5 times ofthe deposited thin film thickness [18]. Spin

speed

between

3000and6000RPMcan give good_coatinguniformity. For4'

wafer, 6000

RPM spinner speed was usedto coat3.2 um photoresist after experiments. AZ5214EZ is

dilutedwithAZ1500thinner

(1:1)

inordertogetthethicknesswithinthedesiredrange.

Softbake can remove most ofthe_remaining solventfromthe photoresistfilm. AZ5214E

resist is not _very sensitive to light until its solvent component (propylene glycol

monomethy ether _acetate) is removed. Prebake is needed onhot plate to accelerate the

solvent diffusion process before exposure. [14].

According

to AZ5214E data _sheet,

temperatures above 100C in an oven or 120C on a hot plate will lead to reduced

sensitivity and therefore can _adversely affect process latitude. It's suggested that

90-110C and 30-60seconds are appropriate for hot plate softbaking. After

taking

into

consideration that lower softbake temperature will result in decreased adhesion and

stabilityinlater processing steps,wetook 110

C for 45 seconds as softbake

temperature,

itcan acceleratethesolventdiffusionwithout

lowering

theresist photospeed [16].

Exposure will convert diazonaphthoquinone to indene carboxylic acid _groups, which

makesthe filmbase soluble.

PEB,

flood exposure anddevelopmentwill give a negative

image becauseliberated reactive agent. The exposurehas a _strong effect onthe undercut

profilebecause theimagereversal processis _{simply reversing}theoriginalprofile, italso

positive tone is suitable fornear UV (365-405 nm) exposure. These wavelengths can be

used toachieveresolutionof one micron andbelow in 1.5 micronthickresist_{films [16].}

Actual exposure energies required will depend on film

thickness,

softbake _conditions,

spectral output ofthe exposure

tool,

and

developing

conditions. Forour process of4000

angstroms of_resist, we need to optimize the exposure dose. Different level exposure

intensity

wereperformed

by

_{changing the}exposuretimefrom0.05 secondto 1 second

by

an increment of0.05 second to see find a _clearing dose for AZ5214 as positive resist,

then that valuewill bethe valuebelow whichwe should use for image reversal process.

The doseof_energyin dimensionofjoulesper squarecentimeter[J/cm

]

canbecalculated

by

multiplyingthe illumination

intensity

that strikestheresist unitvolume [W/cm

]

with

theexposuretime [s]. Underexposurewillprovidewithlift-offprofile. Overexposurewill

givepositive profile _sidewalls, which should be eliminated.

Combining

with experience

from the RIT image reversal process and our film situation, a series ofdifferent set of

exposure time and PEB temperature combination with the 436nm stepper system were

tried togetthebestprocess conditionfor lift-off.

Normally,

a positivephotoresistprofilehasapositiveslope of75 C-85 C

depending

on

the process conditions and the performance ofthe exposure equipment. This is _mainly

dueto the absorptionofthe

PAC,

which attenuatesthelightwhen_penetratingthroughthe

resist layer (so called bulk effect) [13]. The result is ahigher dissolution rate at the

top

and a lower rate at the bottom ofthe resist. When AZ 5214E is processed, the higher

exposed areas will be crosslinked to a higher degree than those with lower dose and

Figure3.9Crosssection of photoresist profile

during

exposureand afterdevelopment.Becauseofthebulk

effect,thedissolutionrateis higheratthe_topofexposed resistlayerandloweratthebottomoftheresist.

Right:When AZ5214Eis inImage Reversal mode,toparea willbemore crosslinkedthanthebottomarea

ofresist,which means a negative profile.

Themost critical parameter oftheprocessis reversal-bake

temperature,

once optimizedit

mustbe keptconstant within

+/-1

C tomaintain a consistent process. _{In any}caseitwill

fallwithintherangefrom 1 15 to 125 C. Iftemperatureis chosentoohigh (>130

C)

the

resist will

thermally

crossli