Production and characterization of high quality solid phase crystallized silicon thin films on insulators

MISSING IMAGE

Material Information

Title:
Production and characterization of high quality solid phase crystallized silicon thin films on insulators
Physical Description:
vi, 195 leaves : ill. ; 29 cm.
Language:
English
Creator:
Jung, Soon-Moon, 1961-
Publication Date:

Subjects

Subjects / Keywords:
Materials Science and Engineering thesis, Ph. D
Dissertations, Academic -- Materials Science and Engineering -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 182-188).
Statement of Responsibility:
by Soon-Moon Jung.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 023176358
oclc - 35001741
System ID:
AA00017668:00001


This item is only available as the following downloads:


Full Text









PRODUCTION AND CHARACTERIZATION OF HIGH QUALITY SOLID
PHASE CRYSTALLIZED SILICON THIN FILMS ON INSULATORS














By

SOON-MOON JUNG


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1996








TABLE OF CONTENTS


Page

ABSTRACT v

CHAPTERS

1. INTRODUCTION 1

1-1. Silicon On Insulator Technology 1--
1-2. Synthesis of Polycrystalline Silicon 3
1-3. Present Approach --------- 5

2. REVIEW 6

2-1. Solid Phase Random Nucleation and Growth in a-Si/c-Si Phase
Transformation 6

2-1-1. Thermodynamic Parameters in a-Si/c-Si Phase Transformation 6
2-1-2. Classical Nucleation Theory for the a-Si/c-Si Phase
Transformation --- 10
2-1-2-1. Energetics of nucleation in a-Si/c-Si Phase
Transformation 10
2-1-2-2. Kinetics of nucleation 12
2-1-2-3. Crystalline fraction 15
2-1-3. Origin of Crystallization in Amorphous Si Thin Films on
Insulators ---- -----15

2-2. Solid Phase Growth of Crystalline Si in Amorphous Si 18

2-2-1. Solid Phase Growth Rate 18
2-2-2. Grain Growth in Crystallization of Amorphous Si on Insulating
substrates 21

2-3. Crystallization Techniques 23

2-3-1. Furnace Anneal -- ------24
2-3-2. Rapid Thermal Anneal 25
2-3-3. Laser Anneal -----26

2-4. Nondestructive Optical Characterization Techniques for a-Si
Crystallization 27









3. EXPERIMENTAL PROCEDURE


3-1. Sample Preparation ----- 30

3-1-1. Deposition of Amorphous Si 30
3-1-2. Solid Phase Crystallization of a-Si 30

3-2. Characterization 31

3-2-1. Differential Reflectometer ----- 31
3-2-2. X-Ray Diffraction --31
3-2-3. Transmission Electron Microscopy -- 37
3-2-4. Hall Measurement 38

4. RESULTS 40

4-1. Differential Reflectivity of As-deposited Amorphous Si 40

4-1-1. Thickness Effect on the Differential Reflectivity ------ 40
4-1-2. Annealing Time Dependence of Differential Reflectivity 44
4-1-3. Reference Sample Effects on Differential Reflectivity 51

4-2. Random Nucleation and Growth of Amorphous Si on Insulator without
Seed 54

4-2-1. Determination of Volume Fraction of the Crystallized Si in the
Amorphous Si layer from Differential Reflectivity ---- 54
4-2-2. Incubation Time --- --- 65
4-2-3. Volume Fraction of Crystallized Si Thin Films 68
4-2-4. Nucleation Rate ------ 73
4-2-5. Growth Rate 75
4-2-6. Crystallization Time 75
4-2-7. Growth Mode of Crystallized Si Particle 75
4-2-8. Final Grain Size 81

4-3. Top Seed Induced Crystallization 86

4-3-1. Top Seeding Effect 86
4-3-2. Hall Mobility 109
4-3-3. AFM analysis 111
4-3-4 Surface Roughened Seed 124
4-3-5. Si Self Ion Implantation Effect on Suppression of Crystallization
of Amorphous Si 127
4-3-6. Post Annealing Effect -128


_ __ __I I








4-3-7. Time Dependency ---- 41
4-3-8. Top Seed Crystallographic Orientation Dependency 141

5. DISCUSSION 148

5-1. Estimation of the Interfacial Energy between Amorphous Si and
Crystalline Si -148

5-2. Estimation of Size of a Critical Nucleus in Solid Phase Crystallization
of a-Si ------154

5-3. Examination of the Origin of Crystallization in Random Nucleation and
Growth ------------.--... 155

5-4. Si Self Ion Implantation Effect on Suppression of Crystallization
of Amorphous Si 158

5-5 Origin of Top Seeding Induced Crystallization of a-Si Thin Films on
Insulators --------162

5. CONCLUSION 178

REFERENCE LIST 182

APPENDICES 189

A. Derivation of the Volume Fraction Formula 189

B. Derivation of the Final Grain Size Formula -------- ----- 192

BIOGRAPHICAL SKETCH ---- 195








Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



PRODUCTION AND CHARACTERIZATION OF HIGH QUALITY SOLID
PHASE CRYSTALLIZED SILICON THIN FILMS ON INSULATORS

By

Soon-Moon Jung

May 1996


Chairman: RolfE. Hummel
Co- chair: Rajiv K. Singh
Major Department : Materials Science and Engineering


Future active matrix liquid crystal displays ( AMLCDs ) require high

performance thin film transistors ( TFTs ) on glass substrates in order to achieve high

resolution, and fast response time. In this dissertation, a novel technique to produce

large grained Si films on insulators at relatively low temperatures ( below 6000C )

has been developed. This method relies on the use of a single crystal Si wafer to

initiate controlled nucleation and growth of crystalline silicon films from the

amorphous phase. Detailed materials characterization has been conducted to

understand the orientation, microstructure ( grain sizes, defects, etc. ), surface

roughness, and nature of the nucleation phenomena.

For fabrication of crystalline Si films on insulators, amorphous Si was

deposited on a Si02 / Si substrate using LPCVD ( low pressure chemical vapor








deposition) at 540C. The thickness ofa-Si was varied from 300 to 1000A. The

samples were furnace annealed at temperatures varying from 525 to 675 OC. A single

crystal Si wafer was used as a top seed to induce crystallization from the top surface

of the a-Si film. The seeded and unseeded polycrystalline Si films were characterized

using the techniques of differential reflectometry, transmission electron microscopy,

X-ray diffraction, atomic force microscopy, and Hall measurements.

The origin of unseeded crystallization of a-Si film on insulators does not

involve growth of pre-existing microcrystallites, but instead a random nucleation and

growth process. The preferred orientation normal to the surface of grains was found

to be <111> and the average grain size as found to be less than 0.2pm. The

activation energies for nucleation and growth were determined to be 5.6 eV, and 3.7

eV, respectively, while the size of the critical nucleus was estimated to be 6.1~7.1 A.

The top seeding technique produced very large <110> grain Si thin films on

insulators when annealed at low temperatures and substantial short times. The

transmission electron microscopy analysis showed that the crystallization of the films

initiated from the top surface. However, these films showed also high density of twins

and stacking faults. The preferred orientation of the films was found to be

independent of the orientation of the top seed crystal. The Si self-ion implantation

was found to increase the crystallization temperature. Hall mobility from these Si

films (258 cm2 / Vsec) were found to be considerably larger than that of typical

polycrystalline Si (- 10cm2/Vsec).













CHAPTER 1
INTRODUCTION

1-1. Silicon On Insulator Technology

Silicon On Insulator ( SOI ) technology has been of considerable interest in

microelectronics. As the size of transistors become smaller and the access time for

data is required to become faster, the need of the SOI technology grows fast in the

silicon microelectronics industry because of their inherent advantages such as low

leakage current, perfect isolation, low parasitic capacitance, and high radiation

immunity.' In SOI technology, the active Si layer on which the devices are

fabricated lies on an insulator rather than on a bulk Si substrate. For best

functionality, the top active Si layer on the insulator has to have good crystallinity.

Recently, standard techniques to achieve a SOI structure rely on SIMOX

(Separation by Implantation of Oxygen ) or BESOI ( Bond and Etch back SOI)

techniques in the Si microelectronic industry. The SIMOX wafer is made by oxygen

implantation deep into a Si substrate followed by high temperature anneal in order to

form a buried oxide layer underneath of a top Si layer.2 In case of BESOI, two

wafers are bonded together by a high temperature anneal after oxidation of the

surface of the wafer.3 Subsequently, the bonded wafers are etched or polished to a

thin top bonded wafer. Unfortunately, these techniques utilize a single crystal Si

substrate and high annealing temperatures( > 10000C ). Therefore, these techniques

are not suitable for LCD( Liquid Crystal Display ) process in which a low melting







temperature glass is used as a substrate. Another SOI technology, which utilizes

polycrystalline silicon thin films on insulators, is used to make TFTs ( Thin Film

Transistors ) for LCDs and SRAM (Static Random Access Memory ).4'5

The rapidly growing LCD industry demands high quality silicon films on a glass

substrate. However, inexpensive glass substrates have a low glass transition

temperature. In order to avoid the degradation of a glass substrate during processing,

the process temperature should be limited to below 600 OC. Therefore, hydrogenated

amorphous silicon ( a-Si:H) thin films have been widely used to create an active

matrix liquid crystal display ( AMLCD ) panel.6 Even though amorphous silicon

films have been successfully implemented to make an AMLCD panel, the a-Si:H

thin film has very poor electrical properties due to an abundance of dangling bonds.

For example, the typical electron and hole mobilities of an a-Si:H thin film transistor

are less than Icm2N sec.7

For future AMLCD panels, there are growing demands, such as higher

resolution, faster charging times, and integration of the driver circuits onto the same

glass substrate because they are required for high definition displays, low cost, and

better reliability. In order to accomplish these, better quality substrate materials are

required. Transistors made from polycrystalline silicon thin films on insulators,

which are formed by furnace anneal, usually have a carrier higher mobility ( about

10-50 cm2N sec ) due to better crystallinity compared to a-Si:H TFT.8 However, the

mobility is still much lower than that of single crystalline silicon ( electron mobility

:1500cm/Vsec, hole mobility: 450 cm2/sec ) because of grain boundaries and

various kinds of defects in the polycrystalline silicon film. In order to fabricate a








high performance transistor, the carrier mobility and other transistor parameters,

such as low threshold voltage, low leakage current, and high gate oxide integrity,

needs to be further improved.


One of the prime reasons for the low carrier mobility in polycrystalline and

amorphous silicon is the presence of large defects, such as dangling bonds, or grain

boundaries. The typical grain sizes of furnace annealed polycrystalline Si are of the

order of 1 pm.9 Grain boundaries substantially degrade electrical properties in a

polycrystalline silicon TFTs.o1 The boundaries act electrically as extra charged states

and trapping sites which can result in high threshold voltage, recombination centers

which cause a high leakage current, and scattering centers which degrade the

mobility. In addition to these, surface roughness is also affected by grain size. A

rough surface degrades the gate oxide integrity and surface mobility in a MOS TFT

(Metal Oxide Semiconductor Thin Film Transistor ). Therefore, a larger grain size

polycrystalline silicon thin film can provide better performance for thin film

transistors. It is the goal of the present work to provide large grain sized

polycrystalline Si on an insulating substrate.

1-2. Synthesis of Polvcrystalline Si film

Presently, there are two general approaches to form polycrystalline silicon thin

films on insulators. One is a direct deposition method in which polycrystalline Si

films are deposited at relatively low temperatures, (below 600C), with the use of

disilane (Si2H6 ) as a gas precursor in replacement of SiH4 or plasma enhanced








chemical vapor deposition (PECVD).,"12 In this direct deposition method, the grain

sizes are very small (--O.2pm), which detracts from the advantage of a low thermal

budget. Also, the surface is very rough, thus further hampering the properties of the

transistors.

A second method for fabrication of polycrystalline Si films is by crystallization of

amorphous Si films deposited at low temperature by CVD methods as outlined

earlier. The crystallization is conducted by furnace heating, RTA (Rapid Thermal

Annealing), or pulsed laser annealing etc. Generally, these techniques provide larger

grain sizes and better surface morphology compared to the low temperature direct

deposition technique.13 Among them, conventional furnace annealing is the simplest

and most widely used technique. It can provide grains having sizes of l-2tm by

applying long time ( about I day ) annealing at 550-600 C. The small grain size and

substantial number of defects lead to poor electrical properties. Further, the long

processing times are not acceptable with respect to throughput and thermal budget.

There have been other attempts to increase the grain size utilizing Si self- ion

implantation into the deposited amorphous Si film on insulator. This method can

increase the grain size by suppressing the nucleation rate of Si crystallites in the

film. Unfortunately, the technique increases the crystallization temperature and

process time significantly. Another technique, RTA (Rapid Thermal Annealing ), can

shorten the processing time and reduces defects because of the high temperatures

used.14 However, the grain size is very small ( < 0.5pom) due to very rapid

crystallization at high temperatures. In contrast to furnace and rapid thermal

annealing, the silicon film is melted for a very short time by laser annealing. This











technique also provides a low thermal budget and low defect density inside the grains

due to melting and rapid regrowth. Such rapid solidification results in very fine grains

in the range of 0.1- 0.2pm.'5 However, the low defect density inside the grains have

led to give rise to high carrier mobilities ( 100~ 300cm 2 / V- sec ).16 Despite the

large mobilities produced by this method, there are several disadvantages. The

formation of very small grains gives rise to surface nonuniformity which leads to non-

uniform oxide growth. In addition, uniformity in processing is difficult due to small

size of the laser beam.

1-3. Present Approach

In order to fabricate very large grain sized polycrystalline Si thin films on

insulators at low temperatures and with short processing times, we have investigated

a novel technique in this study. This technique is based on the application of a

surface seed to induce crystallographic orientation in thin amorphous films. The

application of a surface seed leads to oriented growth of the silicon film resulting in

the formation of large grains. Grain sizes greater than 10 Im and low angle grain

boundaries have been obtained. Studies to understand the nature of the growth

process have been investigated in this dissertation. In addition, a new analysis

technique for characterizing the Si crystallization utilizing differential reflectometry

will be discussed.













CHAPTER 2
REVIEW

The formation of polycrystalline Si films on insulators is controlled by nucleation

and grain growth processes. If there are seeds available for epitaxial growth, then

solid phase epitaxial growth process is important in controlling the microstructure.

This chapter reviews the literature for nucleation and growth mechanisms as well as

crystallization methods.


2.1 Solid Phase Random Nucleation and Growth in
a- Si/c-SiPhase Transformations

Solid phase crystallization of Si from amorphous phases on insulators has been

investigated intensively for many years. 1723 Generally, the kinetics of solid phase

crystallization of Si has been treated and explained by classical nucleation and

growth theory. When amorphous Si thin films on insulators are heated to a certain

temperature, the amorphous Si is crystallized through a nucleation and growth

process. In this process, small silicon microcrystals are formed and grow with time in

the amorphous Si. In this section, the classical theory of solid phase nucleation and

growth in amorphous Si will be reviewed. Also, the thermodynamic parameters of

nucleation and growth will be discussed.







2-1-1. Thermodynamic Parameters in a-Si /c-Si Phase Transformations

The crystallization of amorphous Si layers on insulators occurs through a

nucleation and growth process.23 This phenomena has been explained by the

classical nucleation and growth theory.7-23 According to classical theory, the

nucleation of a microcrystal occurs because this process reduces the total free energy

of the system. The Gibbs free energy change for this nucleation process, AGn, is

determined by the difference in volume free energy change for atoms in the

crystalline phase relative to that of the amorphous phase and the amount of interface

energy which is generated by the a-Si / c-Si interface ,if the strain energy is small

enough to be neglected.

AGn.,= AG,,,,V + yS (1)

where AGv is the free energy difference per unit volume due to the phase

transformation, V is the volume of an n-mer of the nucleating crystalline phase, y is

the interfacial free energy between the crystalline and amorphous phase, S,c is the

surface area of the nucleated crystalline phase. In order to understand this, it is quite

important to determine the thermodynamic data of crystallization.

Donovan et al. 24 has measured the heat of crystallization of a-Si to be AH, =

11.9 0.7 KJ/mol, utilizing differential scanning calorimetry in ion implanted

amorphous Si. In order to calculate the Gibbs free energy for the a-Si /c-Si phase

transformation, the values of entropy and heat capacity are required. The AG.a is

calculated from the formula as shown below:

T
AHI= AH( Ti)+ JACp dT (2)
Ti








where AH, is the enthalpy of crystallization ,and ACp is differences of heat capacity

between a-Si and c-Si.

T
AS, = ASoac + J (ACp / T) dT (3)
0

where AS~ is the entropy of crystallization, and T is the temperature of

crystallization.

AG,. = AHw T-AS, (4)

where AGQ, is Gibbs free energy for the a-Si /c-Si phase transformation.

Spaepen25 calculated AS, = 0.2 R, where R is ideal gas constant, from the

excess configurational entropy of the ideal four coordinated random network of Si.

In Fig. 1, an estimate of the Gibbs free energy of amorphous Si relative to crystalline

Si is shown. In this figure, the difference of heat capacity ( ACp ) between a-Si and c-

Si is assumed to be ACp= 0, or ACp = -0.024 + 4.8 ( T / 1685 ) J / mol K by Spaepen.

According to these calculations the melting temperature of amorphous Si becomes

1295(K for AC= 0 or 14200K for ACp= -0.024 + 4.8 ( T / 1685 ).

The interfacial energy between amorphous and crystalline Si has been investigated

by several researchers.26-2s From the a-Si /(111) c-Si interface building model,

Spaepen26 estimated the interfacial energy as 310 ergs / cm2. Roorda et al.27 had

fitted the classical nucleation theory to their experimental results to find the

interfacial energy, to be 430 ergs / cm2. Morgiel et al28 used direct in-situ TEM

measurements of nucleus sizes, and then inserted those data into the formula of

classical nucleation theory to calculate the interfacial energy, to be 600 ergs / cm2.





































T(K) *


Fig. 1 The estimated Gibbs free energies of amorphous Si and liquid Si relative to
crystalline Si as a function of temperature : line a(1) assuming ACp=O line a(2)
assuming ACp= 0.0224 + 4.8 ( T / 1685 ) J/mol K ( From (24), Donovan et al,
Appl. Phys. Lett. 42 ,p698 (1983))







However, all these values are no direct measurements but instead estimates based on

the models or theoretical equations of classical nucleation theory.

There have been few reports on the interfacial energy for various crystallographic

planes. These values are particularly important for this dissertation because it

controls the orientation of the growing crystal. Based on the values for free silicon

surfaces, we can only speculate about their trend. Jaccodine29 had measured the

(111) plane surface energy utilizing a cleavage technique, and then estimated the

surface energies of the (110) and (100) plane with the value of(11 1). These values

are shown in Fig.2. This figure shows that the values correspond to 2130 ergs / cm2,

1510 ergs / cm2, 1230 ergs / cm2 for (100), (110), and (111) planes respectively.

These numbers show that the a-Si / (100) interface should possess the highest

interfacial energy, while the a-Si/(l 11) interface should possess the lowest interfacial

energy

2-1-2. Classical Nucleation Theory for the a-Si /c-Si Phase Transformations

Theoretical formulas for the time dependent crystallization allow us to extract

kinetic parameters and activation energies of the crystallization from experimentally

measured values, such as crystallized volume fraction, and nucleation rate.

Especially, it is important to assess the stability of an amorphous phase at a given

temperature and time, which can be used to control the random nucleation and

growth process in amorphous Si thin films on insulators. Therefore, many

experimental and theoretical studies for crystallization phenomena have been









2500




2000 -


1500 -




1000 -


500 -


0






0


0


-I


(100)


(110)


I
(111)


(h k I) plane



Fig.2 Experimentally measured free surface energies of planes of silicon
( Data from (29),J.Jaccodine, J.Electrochem. Soc. 110 ,p524 (1963))


E

2o
a)


C



=3
u)
C.)



Co







conducted by a number of researchers."-' '" Each of these factors are discussed in

detail.

2-1-2-1. Energetics of nucleation in a-Si /c-Si phase transformation

The energy change due to formation of an n-mer of a nucleating phase is given

by34

AGn = AG,-V+ y-S + AG8-V (5)

where AGy is the free energy difference per unit volume due to the phase

transformation, V is the volume of an n-mer of the nucleating phase, y is the

interfacial free energy between the cluster and the parent phase, S is the surface area

of the cluster, AG, is the strain energy per unit volume associated with the formation

of the cluster. For simplicity, we assumed that the strain energy effect is very small so

that it can be ignored. In order to describe the cluster of arbitrary shape and

configuration, the atomic volume, 0, and the shape factor, g (0) = ( 2 + cos 0 ) (1 -

cos2 ) / 4, are introduced. The volume of an n-mer cluster and surface area can be

expressed by V= n-* and S= g(0)-02*n3.. Equation (5) can be simplified as follows:

AGn = AG,-f-n + y-g(0).O-KI.n2 (6)

When the first derivative of (6) equals zero, the cluster reaches the critical size

which can be a nuclei. From this, we can derive the critical free energy, AG*k and

the number of atoms in the critical cluster, k.


k= (8/27).(1/Q).[ y.g(0) / AGv ]3 (7)


AG*k = 1/2.-k--AGv







2-1-2-2. Kinetics of nucleation


The following derivation of the nucleation rate and the incubation time is based on

the work done by many researchers.339 The nucleation rate can be derived by

taking considering the likelihood of growth and decomposition of all sizes of clusters.

Frenke35 suggested the flux of clusters in size space as follows:

Jn = Cn, C n+, a, (9)

where Cn and C,+i are the numbers of clusters of size n and n+1 per unit volume, B,

is the frequency with which the n-mer is promoted to an n+l-mer, and an+l is the

frequency with which an n+1- mer decomposes to an n-mer. Frenkel applied the

detailed balancing that the net flux between any two size classes at equilibrium is

zero.

Pn-Cn an+rCn+l =0 (10)

where Cn, Con+i correspond to the metastable equilibrium concentration.

From eqn. (10 ) one obtains

anc+ = Bn-CO CWn+ (11)

After substituting eqn.( 11) in eqn.( 9) and replacing the difference equation

by a differential, eqn.( 9 ) becomes

Jn = n-Con [8( C,/ C,) / n ] (12)

In order to obtain a differential that describe the time dependent nucleation process,

the continuity is applied to eqn. (12).

acn / at = divJ n =- aC (J )/Oan

= 8 {( N-Cn,[O(C ) / Con)/ at ]} / n (13)








For steady state nucleation, dCn / Ot = 0, eqn.( 12 ) is rearranged and then

integrated to obtain the steady state nucleation rate, J".

o0
I (dn / n-C*) = 1 / k (14)
1


Then,

Ak = Z-'k-Ck (15)

where Z is the Zeldovich factor.


Z =[(-1/2lklT)-( 2AGo / n 2 )]i2 (16)


The time dependent nucleation rate was given by Turbull38 as



J* (t ) = JSk exp ( / t) (17)


where t is the time and r is an incubation time.

The incubation time can be derived by the assumption of random walk in the

critical region.34 In this approximation, r, the incubation time, is defined as the time

for a cluster to move from the subcritical size, k 5/2, to a supercritical size, k+ 8/2

along the Gibbs free energy barrier with the random walk distance, 8. The value of 5

is computed from the Gibbs free energy change equations by setting

AG AGOn = kBT (18)


k- n-= /2


(19)









Thus,

8= [-8ksT/( c2AGO./a n2 )k1, (20)


In the random walk theory, the time which is necessary for a particle to displace the

distance, 6, in one dimension is given by

T= 62/ k (21)

Therefore, the incubation time, T, can be expressed as follows:


= 8 kBT / [ O3k( cAG*n / an2 )k ] (22)

2-1-2-3. Crystalline volume fraction

Generally, the time dependence of the crystalline fraction can be expressed by

Avrami's equation.40-42

X= 1- exp[-(t-T)I/t]m+' (23)


where t is an incubation time, t, is a characteristic time m is a dimension of

the growth mode.

For example, in two dimensional growth ( m=2 ), the crystallized volume fraction is

expressed by

X= 1-exp[-(t -)/(3/i-v.-Js.d)i3]3 (24)


where vg is growth rate, J is the steady state nucleation rate, and d is the film

thickness. Therefore, we can extract the unmeasured kinetic parameters by fitting the

measured crystalline volume fraction and the measured parameters into eqn. (24).







2-1-3 Origin of Crystallization in Amorphous Si Thin Films on Insulators

There have been many studies about the mechanism of the crystallization ofa-Si

thin films on insulator.4352 Specifically, there has been some controversy about the

origin of the crystallization. By now, there are two conflicting models to explain the

solid phase crystallization ofa-Si. One of the models is based on the existence of a

pre-existing microcrystals in the amorphous Si. The other model is based on the

nucleation and growth of crystalline Si through atomic arrangement in the

amorphous Si phase.

Saito et al.43 found microcrystals in as-deposited amorphous Si thin films on a Si

substrate utilizing a high resolution TEM. Hayashi et al. 4 suggested that

crystallization might occur by the growth of pre-existing microcrystals based on

their TEM observations of the crystallization process. Bisaro et al.45 found that the

initial crystallization occurs from the interface between amorphous Si thin films and

the quartz substrate utilizing Raman scattering measurements taken on the front and

back side of the substrate. They proposed that the crystallization commences from

submicrocrystals at the interface between the amorphous Si thin films and the

insulator.33 However, the model involving pre-existing microcrystal has some

difficulty to explain the incubation time of the crystallization. If the crystallization

resulted from the growth of the pre-existing microcrystals, no time delay for

crystallization is expected because there is no energy barrier.

TEM had been used by several researchers4-50 to directly observe the initial

nucleation sites in amorphous Si thin films on insulators. Adachi et al.46 showed that







the nucleation initiated at the interface between the a-Si film and substrate with TEM

cross sectional observations. In Morgiel's in-situ HRTEM observations, initial

nucleation occurred at the interface between a-Si and SiO2 .4 They also found pre-

existing microcrysts at the interface. However, according to their report, the pre-

existing microcrystals did not grow during annealing, instead, randomly nucleated

grains grew. Generally, it is well recognized that the initial nucleation occurs at the

interface between a-Si and the insulating substrate. In order to explain this

phenomenon, relaxation mechanism was suggested by Adachi et al. and Bisaro et

al..45"46 The difference in thermal coefficients leads to release of tensile stress at the

interface between the films and the substrate. On the other hand, there have been

further work that supports the pre-existing microcrystal growth mechanism

Wu et al. 47 has reported that Si self-ion implantation could suppress the nucleation.

Two models have been suggested for this suppression of the nucleation.in implanted

Si thin films on insulators. The first model is based on that oxygen atoms recoiling

from the underlying silicon dioxide can hamper the nucleation.47 In the other

model, the presence of the pre-existing microcrystal at the interface can be destroyed

by ion implantation.52 Although several models have been suggested, these seem to

be no definitive work which can explain the nucleation and crystallization of

amorphous silicon.

As we reviewed, the solid phase crystallization of a-Si thin films on insulators has

been thought of as the phase transformation from a metastable amorphous phase to a

stable crystalline phase in order to minimize the Gibbs free energy during annealing.







In order to analyze the kinetics of crystallization the thermodynamic data related to

the crystallization are necessary. Therefore, the Gibbs free energy for the

crystallization has been determined based on the direct measurement of the heat of

crystallization, the calculation of entropy from the atomic model of Si network, and

the assumption of the difference of heat capacities. The interfacial energy ofa-Si/c-

Si could not be measured directly. It has been determined by indirect methods, such

as a-Si/c-Si interface building model, fitting the classical nucleation theory formula

into the measured data, and measuring the nucleus sizes utilizing TEM. Therefore,

there are large differences among the estimated values of the interfacial energy.

In addition to that, the basic formula of the classical nucleation theory has been

reviewed in order to use those formula to extract kinetic parameters and activation

energies of the crystallization from experimental data later. Our analysis on the solid

phase crystallization will be based on this reviewed classical nucleation theory.

There are two conflicting models, which are the pre-existing microcrystal model

and the growing of the nucleated Si crystal, to explain the crystallization of a-Si.

Therefore, those thermodynamic data and the kinetic data, which will be acquired by

applying the formula of the classical nucleation theory, will be intensively used to

clarify the origin of crystallization in a-Si thin films on insulator.

2-2. Solid Phase Growth of Crystalline Si in Amorphous Si

2-2-1 .Solid Phase Growth Rate

Solid phase growth of crystalline Si from amorphous Si have been studied by

many researchers. s536 Particularly, the epitaxial regrowth of crystalline Si layers was








250




200 -



E
150 -
0<








50
00
0)


I 50-



50 -





(100) (110) ( 111

(h k I) plane

Fig.3 Solid phase epitaxial regrowth rate for implanted amorphous Si at 573.50 C
( Data from (56), L.Csepregi et al J.Appl. Phys. 49,p3906 (1978))







investigated intensively in 1970's and 1980's because ion implantation of a dopant

into a Si substrate produces amorphous Si layers near the surface. The ion implanted

amorphous layer can be recrystallized by low temperature annealing. Csepregi et al 3

had studied the crystallographic dependency of the epitaxial regrowth rate of

amorphous Si on Si substrates. The results are shown in Fig.3. The regrowth rate in

the <100> direction is the fastest followed by <110> and then by <111>. According

to their research, the activation energy for the epitaxial regrowth is 2.35 eV. Csepregi

et al. 5455 reported about the effects of impurities on the epitaxial regrowth. In their

research, electrically active dopants, such as As, B, and P, increase the regrowth rate

by a factors of 6-20.5 On the other hand, inactive impurities, such as O, C, N, and

noble gases, retard the growth rate significantly.55

In order to explain the orientation dependency and impurity effects several

models have been suggested."' Csepregi et al.56 have proposed that, as geometrical

criteria for recrystallizaion at the interface between a-Si and c-Si, Si atoms can be

transferred from a-Si to c-Si at positions where at least two nearest neighboring atoms

are already in crystalline Si positions. According to that model, (111) can not grow

epitaxially. Therefore, in the (111) plane nucleation and growth of twins is necessary

for the epitaxial regrowth. However, this model cannot explain the regrowth in the

(110) plane.

Spaepen et al 57 proposed an atomistic model of amorphous / crystalline interface in

the diamond structure. It was demonstrated that the interface can be constructed

without any broken bond. In the model, the crystalline phase is made of six fold rings,







and the amorphous phase is made of mixtures of five fold to eight fold rings.

Therefore, the crystallization process at the interface is breaking the bonds of rings in

the amorphous phase and rearranging the bonds to the sixfold ring of the crystalline

phase.

The broken bonds need to propagate through ledges to rearrange. The orientation

dependency is explained by the concentration of ledges in a growing plane.

Another model was suggested by Drosd et al.59 In Drosd's model, the criteria for

the a-Si / c-Si interface migration is that the atoms in a-Si must make at least two

undistorted bonds with the c-Si. For example, the (100) plane needs a single atom to

satisfy the criteria. On the other hand, the (111) plane needs three atoms. Therefore,

the (100) grows fastest, and the (111) slowest. Also, the effect of micro twins on a

(111) plane regrowth was explained by this model. If there is a twin boundary on

(I 111), only two atoms are required to grow instead of three atoms, and the regrowth

rate is accelerated compared to that of the original (111) plane.

Washburn et al.60 proposed that the epitaxial regrowth occurs by nucleation and

growth of disc- shaped growth steps on a-Si /c-Si interface. The orientation

dependency is explained by an increase in the surface energy due to the disc- shaped

growth step. For example, the increase of the surface energy is the largest in the (111)

plane. Therefore, the growth rate of the (111) plane becomes the slowest.

Among the proposed models, the Spaepen's model seems to be the most plausible

because this is the only model that can explain constant activation energy for various

recrystallization directions. In case of Spaepen's model, the activation energy of







regrowth is related to the breaking of bonds and the rearranging of bonds. On the

other hand, in another model, the activation energy is different for different

directions. For example, in Drosd's model, the number of atoms which is needed for

regrowth are a function of the crystallographic direction. Thus, these models predicts

different activation energies. The different surface energies in Washburn's model

also lead to different activation energies in various crystallographic direction.

2- 2-2. Grain Growth in Crystallization of Amorphous Si on Insulating substrates

The grain growth mechanism in solid phase crystallization of amorphous Si on

insulating substrate is the same as that of solid phase epitaxial growth. However, this

crystallization needs a nucleation step before the growth process occurs. Due to this

complication, an accurate measurement of the growth rate becomes very difficult.

Usually, the growth rate is determined by fitting the crystallized volume and

nucleation rate into the Avrami equation.17 Another method to measure the growth

rate is a direct observation of grains to grow with the TEM. 8 The reported

activation energies for the grain growth, (2.4-3.3 eV), are similar to those for

epitaxial regrowth, (2.4-2.9 eV).17-19

Grain growth mechanism have been studied by several researches.2022 Nakamura et

al.20 found that the grains of(110) surface plane have a preferential growth in the

<112> direction along twin boundaries. It results in the formation of dendrite shaped

grains. Noma et al.21 had investigated the grain growth process for differently

oriented grains. The growth rates for the grains have the same trend as those of the

epitaxial regrowth ( v(100) >v (110) > v(l 11)). The nucleated grains are bounded by







the slowest growing plane, (111), during the crystallization process. Therefore, the

growth rate of grains is determined by the rate of forming the (111) plane. As it is

mentioned in previous reviews of epitaxial growth, the (111) plane is difficult to

grow due to more incorporated atoms for migration of the a-Si / c-Si interface

compared to other planes. In (111) growth, microtwins are formed on the interface to

enhance the growth. An end point of twin plane provides atomic steps as nucleation

sites for the growth.22

According to Noma's model, the shapes of variously oriented grains are decided by

a location of (l11) twin planes with respect to film surface. For example, the <110>

oriented surface grains have a preferential growth in the <211> direction and branch

in other coplanar <211> directions. This results in a dendrite shape. On the other

hand, when (111) twin planes are parallel to the film surface in <111> oriented

grains, the shapes become disc type.21 Based on the work by others, we know that

the randomly nucleated grains usually form microtwins during the grain growth.

From reviews in this section, it is known that the mechanism and the trend of solid

phase growth of crystalline Si particle in a-Si thin films on insulator is the same as

those of the epitaxial growth. The order of the growth rate among different planes is

given by v(100) > v(110) > v( 111). The Spaepen's model may be the most plausible

model to explain the constant activation energy for various planes. Usually, the

dendrite shaped grains and micro twins are observed in the crystallized Si thin films

on insulators according to TEM observations.







2-3. Crystallization Techniques

Recently, Si thin film transistors ( TFTs ) have attracted attention because of their

applications to AMLCD ( Active Matrix Liquid Crystal Display ) and SRAM ( Static

Random Access Memory ). Especially, for AMLCD, Si TFT should be fabricated at

low temperature because of the softening point of low cost glass substrates which is

around 6000C. There have been two general approaches to form polycrystalline Si

thin films on insulator. The first is a direct CVD deposition method, in which

polycrystalline Si films are deposited below 600C, employing disilane ( Si2H6 ) as a

gas precursor instead of silane ( Si H4), an ultra high vacuum, a plasma .6266

Unfortunately, in spite of a low thermal budget, grain sizes are very small. Also, the

surface becomes very rough. Consequently, it is difficult to expect good electrical

properties by this method.

The other method is crystallization of amorphous Si thin films deposited at low

temperature by using various annealing techniques, such as furnace anneal, RTA

(Rapid Thermal Anneal ), and laser anneal.67-91 Many crystallization techniques are

known since several decades. In order to be compatible with glass substrates and to

achieve good electrical properties, such as a high mobility, a low threshold voltage,

and a low leakage current, the technique should provide a low thermal budget, large

grain sizes, a flat surface, and a low defect density.

2-3-1. Furnace Anneal

Furnace annealing is the simplest and the most widely used technique. Since

Brodsky et al73 reported this method in 1960, this technique has been extensively







used.67-73 Earlier studies focused on the understanding of the kinetics of

crystallization in relatively thick films with various characterizing methods. For

example, Blum et al.67 has studied the crystallization time at a range of 5500C- 700C

by observing optical transmission changes in 5000 A thickness amorphous Si films.

They found that the activation energy for crystallization time is 3.32 eV. K6ster69

introduced the classical nucleation and growth theory and found activation energies

for nucleation and growth, 4.87eV and 2.9eV, in unsupported amorphous Si films.

Although the furnace annealing process has been extensively researched, the quality

of the polycrystalline film is not very good. This process leads to fairly large grains

(1-2 pim) with long time ( Iday) annealing at 5500C-600C The mobility of TFTs

made on polycrystalline films with these grain sizes is order of 10 cm2/ V -sec. This

value is far below compared to that of single crystalline. Also, a long process time is

not compatible with high throughput and a low thermal budget.

In order to increase the grain sizes, Si self-ion implantation into deposited

amorphous Si films on insulators had been introduced. In the beginning, Iverson et

al.'7 used Si ion implantation to amorphize the as-deposited polycrystalline Si thin

films on insulating substrates. Also, this method has been used to select surviving

grains as seeds in deposited polycrystalline Si thin films by Kung et al. 74'7 Noguchi

et al.76 have demonstrated a grain size enhancement by Si implantation. This method

increases the grain sizes by suppressing the nucleation of microcrystalline Si in

amorphous Si thin films.77 Also it can be used to control nucleation sites by selective

ion implantation.78"79 Despite grain size enlargement, this technique increases the







crystallization temperature and process time significantly. For example, according to

Wu et al.77, the crystallization time for 82nm LPCVD as-deposited amorphous Si thin

films is 5 hours at 6000C. The crystallization time increased from 5 hours to 24 hours

when the samples were implanted by Si ions ( dose lx 10'5 cm-2 ). Thus, this

technique is not a commercially viable process.

2-3-2. Rapid Thermal Anneal( RTP )

The RTP process has been developed as an alternative technique to reduce a

thermal budget.8."' This technique can shorten the processing time and reduce

defects because of the high temperature used. The RTP process uses the light sources,

such as halogen lamp, to heat the sample instantly. According to Voultas et al.80, the

average grain size is less than 0.5pm with 900-1000C annealing. The crystallization

time is less than 1 minute. Some researcher have reported better electrical properties

compared to those of a conventional furnace anneal technique. 81 In general, the

mobilities of RTP crystallized Si film are less than those of the furnace anneal

because of small grain sizes which occurs due to very rapid crystallization of the

films. Additional to that, there is a great difficulty in achieving heating uniformity

over large areas.

2-3-3. Laser Annealing

Localized melt-mediated crystallization of deposited amorphous Si thin films on

insulator has been performed for decades using a variety of energy sources, such as

electron beams, strip heaters, CW -Ar lasers, and excimer lasers.8291 Among them, in

particular the excimer laser whose use has been investigated intensively because it







heats up only the top surface layers without bulk absorption.83 Further, the excimer

lasers possess a larger beam size and a higher beam energy density, this making it

possible to anneal large areas.

Typical excimer lasers have wavelength in the ultraviolet region [ KrF ( =

248nm ) XeCI ( X = 308nm ), and ArF ( 4 =193nm ) ]. The energy density is in a

range of 100 mJ/cm2-J /cm2. A pulse width of the laser beam is in a order of 25

nano sec. The short processing laser technique can provide a reasonably low thermal

budget and a low defect density inside of grains due to surface heating at

temperatures above the melting temperature for a very short time. According to Singh

et al. 82, grain sizes become larger with increasing the laser energy density.

However, if the energy density is over the optimum, the grain sizes decrease with

further increase in laser density. 89 On the other hand, for low energy densities, a

double layer structure which consist of a top large grain layer and a bottom fine grain

layer is formed during laser crystallization because only the region near the surface is

melted by a low energy density laser.83 A large grained layer is formed on the surface

followed by a fine grain layered under the large grain region. This occurs because of

the explosive crystallization process.

Basically, the laser process features very fast solidification of melted Si. Such rapid

solidification results in very fine grains which are less than I Pm. Thus, high

mobilities have been observed but the sample is plagued by small grain sizes.

Furthermore, as the microstructure is very sensitive to the energy density, it is very

difficult to have good uniformity and reproducibility in real production.









2-4. Nondestructive Optical Characterization Techniques for a-Si
Crystallization

Optical properties of amorphous Si are quite different compared to those of

crystalline Si. For example, crystalline Si has three distinctive peaks in the reflectivity

spectrum due to interband transitions, such as L'3 -+L3 ( A = 225nm ), X4 -+ X, ( =

280nm ), and 25-- 15 ( X = 365nm ). Therefore, their optical properties have been

used to investigate the surface quality of silicon layers and the degree of crystallinity

in Si. 92'-0' Blum et al.67 used optical absorption to find the characteristic time of

crystallization ofa-Si in 1972. Duffy et al.9 implemented a UV reflectance method

for characterization of heteroepitaxial silicon thin films on a sapphire substrate. They

correlated the reflectance data with yield of SOS ( Silicon On Sapphire ) circuitry for

the production line. Harbeke et al.97 characterized surface quality of SIMOX

(Separation by Implantation of Oxygen ) wafers by UV reflectance spectroscopy.

Basically, in spite of simplicity and quick measurement, the reflectance measurement

method has difficulty in separating individually affecting factors to the reflectance of

Si, such as crystallinity, surface roughness, surface contamination, and top oxide

thickness, because the reflectance is very sensitive to all those factors.

Another tool, SE ( Spectroscopic Ellipsometry ) has been developed to analyze

crystallinity of Si thin films. 9397 Kurmar et al.99 showed that SE can be used to

monitor volume fractions of crystalline Si during deposition of Si thin films as in-situ

method. Vedam et al 102 implemented S.E to analyze the thickness, and the

compositions of Si implanted Si wafers. They used simulation method to interpret







the SE data with the Bruggeman effective medium approximation method, which is

an approximation that the effective complex dielectric function, e (= s, + i-s2 ), of a

composite layer consisted of two randomly distributed constituents ( X. + Xb 1 )

can be given by

0 = ,a [( sa )/(E .+ 2s )] + Xb [( eb- e)/ b+ 2e)] (25)

where XL and Xb are the volume fractions of two constituents in the composite layer,

eand eb are the dielectric functions of two constituents, a and b. As another

application, SE has been used to obtain structural information on amorphous Si

which has been partially crystallized by thermal annealing ( Suzuki et al 9 ). They

also used an effective medium approximation simulation to interpret the SE data. The

SE method needs a simulation to interpret the SE data .This results in considerable

complexity in interpretation of the results. The dielectric function of Si thin film is

sensitively changed by crystallinity, surface roughness, surface contamination, Si film

thickness, and top oxide thickness. Therefore it is not easy to analyze accurately a

single factor effect from SE data.

On the other hand, Hummel et al.o'0 have developed a differential reflectometer

(DR) which measures the thickness of a damaged layer, for example due to ion

implantation in single crystalline wafers. This technique uses the difference in

reflectivity between two samples which are placed side by side. Therefore, if the

reference sample is in the same environment, the effect of non-essential conditions

can be minimized.














CHAPTER 3
EXPERIMENTAL PROCEDURE

3-1. Sample Preparation

This chapter reviews the experimental procedure adopted in this study. This

involves the deposition of amorphous Si, crystallization method, and process

characterization.

3-1-1. Deposition of Amorphous Si

The structure of the samples used for the present investigations consisted of three

layers, such as a-Si / 6000A SiO2 / (100) Si substrate. The silicon oxide layers were

grown thermally on top of a Si substrate up to 5000 A in thickness, and then

additional 1000A SiO2 was deposited by LPCVD (Low Pressure Chemical Vapor

Deposition ). The top amorphous Si layers were deposited by vertical type LPCVD.

The deposition temperature was 5400C in order to assure a complete amorphous

phase since the transition temperature for a-Si to c-Si in LPCVD is about 6000C. The

source gas was SiH4, whose dissociation reaction is pyrolysis. The heat cycle of a-Si

deposition is shown in Fig .4 The starting temperature was 4500C. The ramping up

rate was 100C/min and the ramping down rate was 3.3C/min. The flowing rate of

SiH4 gas was 120 cc/cm3. The deposition rate was 20A / min at 5400C. Different

thickness of a-Si varying 100 to 1000A were deposited.. Therefore, the holding time

at 5400C was used to control the thickness of the a-Si layers on the SiO2 / Si





















1 540C


p


Fig.4 Heat cycle of deposition of amorphous Si thin films using LPCVD


4500C ]

gas

[N2 J


r-3.3C/min

:44500C

F ^ N21


SiH4
120 CC/min.







substrate. The raw samples were kindly provided by the Samsung Electronic Co. of

Korea.

3-1-2. Solid phase crystallization of a-Si

Annealing for solid phase crystallization was performed in a furnace. An N2

atmosphere was used to prevent oxidation of the silicon. The annealing temperature

was in the range of 5250C 675C. To ensure that the surface of silicon was clean, a

cleaning procedure was adopted which is outlined below:

(1) D.I deionizedd) water rinse for 30 sec. => (2) Soaking in H2 SO4 for min. =>

(3) D.I water rinse => (4) Oxide etching with 6:1( NH4 F ) BOE ( Buffered Oxide

Etchant ) for lmin.=> (5) D.I water rinse -=> (6) Drying by blowing

tetrafluoroethane gas.

The cleaning process with H2 SO4 was used to remove organic impurities on the

surface. BOE was used to remove native oxide and passivate the surface of a-Si layer

with Si-H bonding.

For unseeded random crystallization of the a-Si, the annealing was done in a

furnace utilizing various temperatures. In contrast, for "the top seeding induced

crystallization", the seeds were brought in contact with top of the a-Si layer right

after cleaning, this is done within Iminute before the annealing. It is known that the

thickness of native oxide is time dependent after etching at room temperature. The

schematic diagram of "the top seeding procedure" is shown in Fig.5. The top seeds

consisted of single crystalline Si wafers with specific crystallographic planes and

controlled surface roughnesses. They were also cleaned utilizing the same cleaning

procedure as for the samples before they were put on the a-Si. The seed wafer was









a-Si a-Si
Si2-substrate Si-substrate

Si-substrate Si-substrate


Sample structure


Growth


4
Pressure



Single crystal
Si

a-Si
S12

Si-substrate


SDetacI


c-Si
Si-substrate2

Si-substrate

_ _


e>

Seed Single
Crystal-Si
















Annealing 550C



Seed Single
Crystal-Si
^~


Fig.5 Schematic diagrams of the top seeding technique used in this study


I







placed in close intimate contact with the experimental wafer by a mechanical steel

clamp during the annealing for the crystallization. After the annealing process, the

top seed wafer was detached physically from the surface of the crystallized Si thin

films. In order to control the surface morphology of the top seeds before being put on

the a-Si, the surface of the seeds were ground with A1203 polishing powder. In case of

surface roughened seeds, the seeds were annealed at 12000C for 2 hours to recover

the damage after the grinding.

3-2.Process Characterization

3-2-1.Differential Reflectometer

The differential reflectometer ( DR ) is an optical instrument whose results are

easy to interpret. The technique is nondestructive, and fast. A schematic diagram of

the DR is shown in Fig.6. The DR measures the difference in reflectivity between

two samples which are placed side by side and sequentially scanned by

monochromatic light of various wavelengths. The schematic configuration of

measuring the samples is shown in Fig.7. The measured value is AR / R (= 2 ( R, -

R2) / (RI + R2 )). The incident beam is nearly normal to the surfaces of the samples

and is in the spectral range between 200~800nm. The differential reflectivity, AR / R,

is very sensitive to small changes on the Si surface, especially changing of phases.

For example, the optical properties of amorphous Si are quite different compared to

those of crystalline Si because crystalline Si has very distinctive peaks of reflectivity

due to interband transitions, such as L3 => L3 ( = 225nm ), X4 => X, ( =

280nm), and F25 => F15 ( = 365nm). The differential reflectivity was measured

after annealing in various conditions to determine the volume fraction of the






























Source


M Monochroma.or


Pholomultiplie Output
Tub* Signal


:J_-'. --_ ---- ----:-- -amples


Fig.6 Schematic diagram of the differential reflectometer













R2
I/


R1
\/


as-dep. a-Si annealed a-Si
SiO2 SiO2

Si-substrate Si-substrate


Schematic Configuration of Measuring Samples
R 2(R1_ R2)
R R1+ R2


Fig.7 Schematic diagram of the configuration for measuring sample pairs in the DR







crystallized Si.

3-2-2. X-ray Diffraction

X-ray diffraction ( XRD ) analysis was used to investigate the phase change from

amorphous to crystalline and the preferred orientation normal to surface of the

crystallized Si thin films. For XRD analysis, a Philips Powder Diffractometer was

used. The X-ray generator settings were 40kV and 20mA. The normal incident X-ray

beam was Cu KaI ( = 1.54060 A). The size of measured samples is approximately

lx 1 cm2. The range of 20 was between 20*-80*. The 20 of the plane of crystalline

Si was determined by Bragg's law, nX = 2 d sine. The 20 angles for intense

reflection from silicon at 28.4420, 47.3020, 56.1210, 69.130, and 76.3770,

correspond to the (111), (220), (311), (400), and ( 331) planes respectively.

3-2-3 Transmission Electron Microscopy (TEM)

In order to study the microstructure of the crystallized Si thin films, plan view

TEM and cross-sectional TEM methods were used. TEM experiments were

conducted utilizing a JEOL 200CX STEM ( Scanning Transmission Electron

Microscopy ), whose electron acceleration voltage is 200KV.

The sample preparation procedure of the plan view TEM was as follows:

(1) Cutting the sample into 3 mm diameter discs.

(2) Coating the surface of the crystallized Si thin films with anti-etching wax.

(3) In order to etch the samples from the back side of a Si substrate to the front, the

samples were mounted onto a Teflon cylinder in a way that the surface of the Si thin

films contacts the Teflon, which has a hole to drain the etchant after a hole in the

sample has occurred.








(4)A plastic sheet having a small hole is wrapped around the sample to etch only its

center.

(5) The etchant, consisting of a mixture of 25% HF and 75% HNO3, is directed

towards the sample until a tiny hole has been etched into the surface.

(7) After etching, the sample is soaked into heptane solution to remove the protecting

wax.

The cross-sectional TEM samples were prepared by the following method.

(1) The sample is sliced into very thin cross sectional slabs with a diamond saw.

(2) Two surfaces of the Si thin films of the sliced slab are cemented together face to

face with M bond. Dummy samples are attached to them for handling and mounting

on a Cu ring.

(3) The bonded sample is mechanically ground i.e. thinned to 100lim using a slurry of

Al2 03 in water.

(4) The polished sample is mounted onto a 3 mm diameter Cu ring.

(5) As a final step, the mounted sample is ion milled with argon ions until a tiny hole

is formed.

2.4 Hall measurement

The mobility of carriers in Si thin films is a very important electrical parameter

which is directly connected to the driving current of a TFT (Thin FilmTransistor)

made on the films. The mobility is quite dependent on defects, such as grain







boundaries, stacking faults, and dangling bonds. Therefore, these measurements are

essential to evaluate the crystallinity of the crystallized thin films.

For Hall mobility measurement, boron was used as dopant of the amorphous Si

layer before crystallization. After crystallization, the samples were annealed at

8000C. The size of the samples was 7mm x 7mm. The thickness of the Si thin films

was approximately 1000A. Indium was used to make ohmic contacts that is the

metallization on the crystallized Si thin films. The applied magnetic field was about

7.5KG. The measuring temperature was 297.3K.

After the first measurement, a second measurement was performed by changing the

electrodes to have opposite polarity with the first measurement.












CHAPTER 4
RESULTS

4-1. Differential Reflectivity of As-deposited Amorphous Si

The optical properties of amorphous silicon are quite different than those of

crystalline silicon. The differential reflectivity of as-deposited, amorphous silicon on a

6000A Si02 layer has been measured utilizing DR (Differential Reflectometry). The

differential reflectivity spectrum of amorphous Si is shown in Fig. 8. A (100) n-type

single crystalline silicon wafer was used as a reference. The spectrum shows three

distinct maxima, labelled El(X= 365nm), E2(k = 280nm), and E3( ,= 225nm) for every

sample. Among them E2( k=280nm) is generally the most pronounced. The maxima

correspond to three distinct peaks in the crystalline-silicon reflectivity, which are due to

three specific interband transitions. These are L'3 = L3 ( X=225nm), X4 =X1

(X=280nm), and F25 => F15 ( ,=365nm). As comparison, the DR spectrum for

amorphous silicon and crystalline silicon on 6000A SiO2 /single crystalline Si

underlayers were simulated by a computer program, which has been developed by

Hummel, and W.Xi.'0' This is shown in Figures 8 (experiment) and 9 ( simulation)

compare quite favorably for a-Si. Understandably, the Differential reflectogram for c-Si

shows no El, E2, or E3 peaks since test sample and reference are equal. There are another

peaks at wavelengths larger than 400nm. These are steamed from interference effects

which involves reflections from the interface between a-Si and SiO2.












LIiOOOQAS. I EO E
EEoo A siE2
SiE3












800nm 700nm 600nm 500nm 400nm 300nm 200nm




Fig.8 Measured differential reflectivity spectrum of 500A as-deposited amorphous Si thin film on 6000A SiO2 / (100) Si substrate. A
Si single crystal was used as a reference











150




100




50




0




-50


-100 I i I- '
800 750 700 650 600 550 500 450 400 350 300 250 200


Wavelength (nm)



Fig.9 Simulated differential reflectivity spectrum of a 500A as-deposited, amorphous Si thin filmon 6000A SiOz / (100) Si substrate.
A Si single crystal was used as a reference












I- .


S.__ i- __-::E3 _
o .:r---r >4 ..



1000










400nm 300nm 200nm







Fig. 10 Measured differential reflectivity spectra of 1000A, 500A,and 300A as-deposited
amorphous Si thin films on 6000A SiO2 / (100) Si substrate. A Si single crystal was used
as a reference


i i







4-1-1. Thickness Effect on the Differential Reflectivity

The differential reflectivity of amorphous Si layers has been measured for four

different thicknesses of the a-Si layer, namely for 1000A, 500A 300A and 100A. The

reference sample as before was a (100) n-type, single crystalline wafer. The spectra ofa-

Si 1000A, 500A and 300 A samples are shown in Fig.10. In the case of 1000 A and

500A ,the E2 and E3 are well matched with the theoretical predictions. Both spectra have

almost the same intensity. This means that there is no interference effect due to the

penetration and reflection of incident light to and from the bottom interface. On the other

hand, for the sample having a 300A a-Si layer, even though it shows a very similar

spectrum, the maxima are slightly shifted from the theoretical wavelength because of a

possible interference effects involving the bottom layer. The interference effects on the

differential reflection spectra were simulated for various amorphous silicon thicknesses

below 300A and for a variation of the underlying silicon dioxide thickness. In Fig 11, the

thicknesses of amorphous Si layers were varied, from 250A to 300 A. The

consequential effects on the thickness are very small, there is almost no change in

reflectivity at wavelengths below 300nm. However, when the thickness of the

underlying SiO2 layers is changed, the shapes of the spectra become quite different

(Fig. 12 ). The interference greatly affects the spectrum of a sample whose amorphous Si

layer is 100A thick ( Fig.13 ). Any characteristic maxima of the crystalline silicon cannot

be distinguished in this spectrum because of the interference peaks, which is caused by

the penetration of light into the underlayers and a reflection from the interfaces. The

respective simulated spectrum is shown in Fig. 14 for comparison.

















50

40

30

20

10

0

-10

-20

-30

-40

-50
400 375 350


325 300 275 250 225 200


Wavelength(nm)







Fig. 11 Simulated differential reflection spectra of amorphous Si thin films on 6000A
SiO2 / (100) Si substrates with single crystal Si as a reference sample, the thickness of the
a-Si is changed from 250A-300A.
















60


50


40


30


20


10 ,


0


-10


-20
400 375 350


325 300 275 250 225 200


Wavelength(nm)






Fig. 12 Simulated differential reflection spectra of 300A thick amorphous Si films on
various thicknesses of SiO2. A Si single crystal was used as a reference
































700nm 500nm 300nm



Fig. 13 Measured differential reflection spectra of 100A thick amorphous Si films on a 6000A SiO2 / (100) Si substrate. A Si single
crystal was used as a reference












150




100




50 -



0




-50 k




-100 i I i I i I I
800 750 700 650 600 550 500 450 400 350 300 250 200


Wavelength (nm)


Figl4 Simulated differential reflection spectra of a 100A thick amorphous Si film on a 6000A SiOz / (100) Si substrate. A Si single
crystal was assumed as a reference





49






S: --_-. "--......215 min



--- -- 6 --







~ -.140 -min


----T ------- ---.---.-
V-- 4-


-- ---.--.-.-.100 m .i







-_ .... ...- .- ... ._. I- .-






500nnm 450nm 400nm 300nm 250nm 200nm



Fig. 15 Measured differential reflection spectra of 1000A thick Si thin films on a 6000A
SiO2 / (100) Si substrate when annealed at 600C for various times as indicated. A
1000A thick a-Si sample was used as a reference






















S1- .... 140
__rm_ _-_ --/
17 rain ..-- -_.: f : -._.
-E r:::'L -'----: : ---


160 min



*- i ** ,? 4 ~ y s ** -- 'j 'v -

5
I ... A ,-




-'I80 -i -:









400nm 300nm 200nm







Figl 16 Measured differential reflection spectra of 1000A thick Si thin films on a 6000A
SiO2 / (100) Si substrate when annealed at 600'C. A Si single crystal wafer was used as a
reference.
reference.







4-1-2. Annealing Time Dependence of Differential Reflectivity

Amorphous Si thin films on a Si02 / Si substrate were annealed at various

temperatures to be crystallized. The differential reflectivity then changes. ( Fig. 15 and

Fig. 16) The annealing temperature was 6000C and the film thickness of the a-Si was

i000A. In Fig.15, an as-deposited 1000 A thick amorphous Si sample was used as

reference. On the other hand, in Fig. 16, a single crystalline Si wafer was used as

reference. In both cases, a distinct change in the spectra can be seen as the annealing

time increases. The differential reflectivity of Fig. 15 increases as the annealing time

increases since the as-deposited a-Si is used as reference. However, it is opposite in

Fig. 16 because of using single crystalline wafer reference instead. Particularly, the most

significant change occurs for the E2 peak (280 nm). The reflectivities in the range

between 280mm and 325nm are not as sensitive to surface roughness and surface native

oxide as those having wavelengths below 250nm. In Fig. 16, an additional change can be

seen in the range below 250nm, even though the changes at k=280nm were stopped for

longer annealing time. This is probably due to an increase in roughness and the thickness

of the oxide due to long annealing in the furnace. In the case of a single crystalline wafer

reference, the same trend is observed. The spectra become quite eccentric for long time

annealing even though they are supposed to be flat. It is for the reason that in the

following presentations the difference in reflectivities at 280nm, and 325nm, will be used

to calculate the volume fraction of crystalline Si in the amorphous Si layer. Further, the

penetration depth of light for both wavelengths is less than 300A, which eliminates

interference effects





















30mm





25 min


20 min

125mmin

10miim




400 nm Wave Length 200 nm




Fig. 17 Time dependency of the differential reflection spectra for 500A thick Si films on
a 6000A SiO2 / (100) Si substrate, when they were annealed at 6000C. The reference was
a 500 A thick as-deposited a-Si on a 6000A SiO2 / (100) Si substrate.







4-1-3. Reference Sample Effects on Differential Reflectivity

Two different reference samples, (100) n-type single crystalline Si substrate and

as-deposited amorphous Si, which had the same vertical structure as the measured

samples, were used to measure differential reflectivities as a function of annealing time.

In Fig. 17, the reference sample is the as-deposited 500A amorphous Si. The measured

samples had been annealed at 600C for varying times. The changes in the spectra due to

crystallization are evident even for very small annealing times. Fig. 17, thus,

demonstrates the usefulness of differential reflectometry to determine the incubation

time of crystallization, which is quite influential in controlling the microstructure of

polycrystalline Si thin films on insulator. On the other hand, when a (100) n-type Si

substrate is used as a reference sample, the changes occurring after short annealing times

are not as obvious as they are when an amorphous Si reference is used. This is shown in

Fig. 16. Furthermore, in the case of long annealing times, the shapes of the differential

reflectivity spectra in the range between 200 and 250 nm become difficult to analyze due

to the high sensitivity to other factors such as surface roughness, top native oxide

thickness and defects in polycrystalline Si films. Such a high sensitivity in the just

mentioned range of wavelengths can be observed in Fig 17. When different doping types

and orientations of single crystalline Si wafer are used as reference samples, the

difference among the spectra in the range between 200 and 250 nm is bigger compared

to the rest of the wavelengths. Therefore, in order to minimize the other factors and to

detect small changes in differential reflectivity for the characterization of crystallized

amorphous Si thin films on insulators, it is most efficient to use the as-deposited a-Si

sample as the reference sample instead of the single crystalline Si substrate.









4-2.Random Nucleation and Growth of Amorphous Si on
Insulators without Seed

4-2-1. Determination of Volume Fraction of the Crystallized Si in a-Si Thin Films from
Differential Reflectivity

As previously mentioned, the changes due to crystallization in amorphous Si

have been measured utilizing the Differential Reflectometer. In order to estimate the

volume fractions of crystallized Si from the measured differential reflectivity data, two

assumptions have been made. First, the reflectivity of a partially crystallized Si layer is

assumed to be given by the simple mixing law:

R =-R-s + ( -x)R-s (1)

where X is the volume fraction of crystallized Si, R_ is the reflectivity of crystalline Si,

and Resi is the reflectivity of amorphous Si. Second, it is assumed that the ratio of the

absolute height of E2( X=280nm) for crystallized Si to that for single crystalline Si equals

about the ratio of the relative height of the E2 maxima with respect to the minima at a

specific wavelength. Details of the derivation pertaining the volume fractions gained

from the differential reflectivity will be discussed in the Appendix A. Based on these

two assumptions, the volume fraction of crystallized Si in a-Si thin films can be

expressed as:

S= [ R-si /(R.-si R-si)] 28 x [ I- ( 1+ 0.46C)/( 1 -0.46C) ] (2)

where C is the ratio of the relative heights of E2.

The normalized heights of the E2 peaks for various thicknesses of amorphous Si

at several annealing temperatures are shown in Fig. 18, 19, and 20. From these graphs,

the volume fractions of crystallized Si in the annealed samples were determined using





















o 0




-o v
-*


m 'V

-8
Sv V 650"C
0 o 625C
Sv 600*C
v 575*C
aft ^ ^ ^ .^ . i . I 1 i 1 1 1 1


0 100


200


300


400


500


600


time[min]






Fig. 18 Normalized height of the E2 peaks in differential reflectivity for 1000A thick
annealed Si thin films at various annealing temperatures vs. time.


700


__ I



























.0


0
-0


U~r


*0


S .I ..I |


0 100


200


300


400


time[min]






Fig. 19 Normalized height of the E2 peaks in differential reflectivity for 500A thick
annealed Si thin films at various annealing temperatures vs. time.


* 650"C
o 6250C
* 600C
v 5750C
L'IIII1(1111


700


I i ,


I ... w. v


I


500


600


A0















S!-


0
AM%" 0


0


200


400


600


6500C
6250C
600 0C

1000 1100


time[min]




Fig.20 Normalized height of the E2 peaks in differential reflectivity for 300A thick
annealed Si thin films at various annealing temperatures vs. time.


1
0 0
0
















1 .0 1- ^ ^ -- -- -- 1 1 1-- -1-1--1--1 1 1 \1
1.0 ,

0.9

0.8 -

0.7

0.6 o

0.5 -
v 6500C
0.4 0o 6250C

0.3 / 5750C
V 6000C
0.2

0.1 V V V

0.0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700
time[min]





Fig.21 Volume fraction of annealed IOOOA thick Si thin films in various annealing
conditions, as determined from the normalized height of the E2 peaks in differential
reflection spectra.














1.0
S/0
0.9

0.8

0.7
c *
0
0 0.6

*0.5 v
E
-o 0.4 v v 6500C
> 0 o 6250C
0.3
v 5750C
0.2 0 6000C

0.1

0.0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700

time[min]








Fig.22 Volume fraction of annealed 500A thick Si thin films in various annealing
conditions, as determined from the normalized height of the E2 peaks in differential
reflection spectra.















1.0

0.9 -
0
0.8 -

0.7
C *
0
0.6

0.5
E
2 0.4
> 6500C
0.3 o o 6250C

0.2 6000C

0.1 o

0.0
0 200 400 600 1000 1100

time[min]





Fig.23 Volume fraction of annealed 300A thick Si thin films in various annealing
conditions, as determined from the normalized height of the E2 peaks in differential
reflection spectra.









IE3
Si (400












78 '74 70 66 62 58 54 50 46. 42 38 34 30 26

20



Fig.24 XRD spectrum for an as-deposited 1000A thick a- Si thin films on a 6000A SiO2/ (100) Si substrate.










IE3




i (i l i i 1: i
l i I [I SI I I! l j; ':,,I
Si (42 0:0







.. .. o .. ,. ..- .. o o -. i
78 74 70 66 62 58 ,54 50 46 42 38 34 30 "26 22
29



Fig.25 XRD spectrum for a 1000A thick Si thin film on a 6000A SiO2 / (100) Si substrate annealed at 600C for 3 hours.
Fig.25 XRD spectrum for a I OOOA thick Si thin film on a 6000A SiO2 0 00) Si substrate annealed at 600TC for 3 hours.











1E3

I il l;si i oo h i









-'F' 'F ^ F 11 fiifIj i lll^i'^h F!,lfli iii ll IIi li''
Jill :El i









O o A -.. O .I- I F -1- 1.1111'. I00 0i :.0
J 0



S lJ`lllii ll p.Jill ii I i I



78 74 70 66 62 58 54 50 46 42 38 '34 30 26 22 i
20




Fig.26 XRD spectrum for a 1000A thick Si thin film on a 6000A SiO2 / (100) Si substrate annealed at 600C for 20 hours.










I /






;
-SC C~CC. ~--~-. ., ,_ ___ __ __ __ __ ___ __











27.5 28.0 28.5 29.0 29.5
20

Fig.27 Step scanned XRD spectra for a 1000A thick Si thin film on a 6000A SiO2 /(100) Si substrate annealed at 6000C for various
annealing times, a) as- deposited, b) 3 hrs., and c) 20 hrs.








equation ( 2 ). They are shown in Figs. 21, 22, and 23. For a comparison, as-deposited

and annealed samples heated at 6000C were characterized by XRD. Their x-ray spectra

are shown in Figs. 24, 25, and 26. For the annealed samples, the (111) peak and (400)

peak are present. Additionally, a very low intensity (220) and (311) peak is evident.

(400) peaks originate from the (100) type Si substrate. In order to obtain the crystalline

volume fraction of annealed samples from the (111) peak intensity data, step scan XRD

analysis were used in the range 2 0= 27.5 29.50. This is shown in Fig. 27. It can be

reasonably assumed that the amorphous Si is fullly crystallized after annealing for 20

hours at 6000. Therefore, the ratio of the intensity respect to that of the sample annealed

for 20 hours was used to estimate the volume fraction of crystallized Si which has been

annealed at 6000C for 5 hours. The volume fraction was estimated to be 0.733, compared

to a value of 0.65, from the reflectivity data. The discrepancy of about 15% still can be

considered to be a fair agreement considering the completely different methods applied.

The 500A thick sample ,which had been annealed at 6250C for 40 min, was etched with

Wright etchant to eliminate the remaining a-Si selectively for studies applying the SEM (

see Fig. 28 ). In the estimation from the reflectivity data, the volume fraction of this

sample is 0.64.

4-2-2. Incubation Time

The Incubation time, T, is defined to be the time lag before nucleation sets in. It is

quite important for controlling the microstructure in thin films of crystallized

polycrystalline silicon on insulators. It is a measure of the stability of amorphous Si

under a given annealing condition. The incubation time is estimated using the changes

in differential reflectivity at the E2 peak. It is assumed that no crystallization takes place















































15KV~~ X1,0 r 4m


b)
Fig.28 SEM pictures of a etched 500A thick Si thin film on a 6000A SiO/ (100) Si substrate
annealed at 625C for 40 min. utilizing the Wright etchant a) x 35000
b)x 12000










105




104


1.04 1.06 1.08 1.10


1.12 1.14 1.16


1.18 1.20 1.22


1000/T ( 1/K)




Fig.29 Temperature dependency of the measured incubation time for unseeded
crystallization samples


__ 1 1 -i_ 1_ )0 A
__ I __ ___~ZIZ___00A
] __






________ _I \ _I -- _I ~



----__ __ ______ _____-
---___ __________ __ _I


103


102


101








in differential reflectivity at the E2 peak. It is assumed that no crystallization takes place

during the incubation time. Fig.29 shows the experimental incubation time as a function

of temperature. As the thickness of the amorphous Si layer increases, the incubation time

becomes longer. For example, at 6000C, the incubation time is 60 minutes for a 500A

thick a-Si layer. It increases to 120 minutes for a 1000A thick film. The incubation

times decreases when the annealing temperatures increases. The incubation times

depends on an interplay between nucleation and growth processes. If the annealing

temperature increases, the nucleation and growth process becomes fast. The temperature

dependency can be expressed by an Arrhenius relationship:

S= 0 -exp(Qr/kT) (3)

where to is a pre-exponential factor and QT is the activation energy for the incubation

time. From the slope of the Arrhenius curve in Fig. 29, an activation energy of 3.97

eVcan be calculated. Later, these incubation time data will be used to estimate the

maximum annealing time required to induce crystallization by top seeding. The

activation energy is also a useful thermodynamic parameter which can be used to

estimate the interface energy y., of a-Si/c-Si and the Gibbs free energy, AGk*, for

forming a critical nucleus.

4-2-3. Volume Fraction of Crystallized Si Thin Films

The crystalline Si volume fraction in annealed amorphous Si layers on SiO2 was

estimated utilizing D.R.data Figs. 21,22, and 23 show the annealing time dependence of

the volume fraction at various annealing temperatures for different thicknesses ofa-Si,

such as 1000 A, 500A, and 300 A. The crystallization begins after a certain time lag, the

incubation time. The time lag becomes longer at lower annealing temperatures. Longer







incubation time. The time lag becomes longer at lower annealing temperatures. Longer

annealing times are taken to fully crystallize the films as the annealing temperature is

decreased. Generally, the annealing time dependence of the volume fraction of

crystallized Si can be expressed by Avrami's equation 24-26

X(t) = 1-exp[-(t-.)/Tr]" (4)

where n is m+1, m is a dimension of the growth mode, r is the incubation time, and T c

is a characteristic time which contains nucleation growth rates, and the growth geometry.

For a two dimensional growth mode, a constant nucleation rate, and an interface

limited growth, n becomes 3 and -t is then given by2426

T= [( -.d.-l.vg2.) / 3 ]-1 (5)

where d is the thickness of the films, IL is the steady state nucleation rate, and Vg is the

growth rate. In order to extract the nucleation rate and the growth rate from the

experimental volume fraction data, equations ( 4 ) and ( 5 ) were used to fit the

experimental points in Figs. 21, 22, and 23. For fitting, the least square method was used

utilizing the software program, Sigma Plot. The exponential indexes, n, were directly

determined from the slopes of the experimental data in Figs. 30, 31, and 32. Equation (4)

can be linearized, as follows:

1-X (t)= exp [-(t-t)/ ]n

log In[ 1 /( -(t)] ) -n log +n- log(t- ) (6)

The exponetial indexes, n, can be acquired from the slope of Eqn. ( 6 ), and Vg and L are

which are obtained by the curve fitting. The solid lines in Figs. 21, 22, and 23 are the

fitted data. They fit fairly well to the experiments for some data. Therefore, we may































100


10-1


10-2


1 n-3


= = -- -i- =




Ift 6,







- -- --- = -


10 1
101


102


105


106


time (t-r) (sec)


Fig.30 In [ 1/( 1- ) ] vs. (t ) for 1000A thick Si thin films, where x is the volume
fraction of crystalline Si and T is the incubation time for crystallization.




















103


102



10'



S100
SI I Ii 1 I I 0I ill I I I II l







10-2
: = ::E ::: : E =:: : EE; i l E-l II















10-3
10-3.- 1 .... ...-. _
10' 102 103 104 106 106





Fig.31 In [ I/( 1- ) ] vs. (t ) for 500A thick Si thin films, where x is the volume
fraction of crystalline Si and t is the incubation time for crystallization




















103


102 :






-- o fil


10-
10-1


10-2


10-3
101 102 103 104 105 106

time (t-r) [sec.]



Fig.32 In [ 1/( 1- x ) ] vs. (t r ) for 300A thick Si thin films, where x is the volume
fraction of crystalline Si and T is the incubation time for crystallization.







conclude that the classical nucleation and growth theory may be applied to analyze the

kinetics of solid phase crystallization in amorphous Si.

4-2-4. Nucleation Rate

The nucleation rates and growth rates were determined by fitting the measured

volume fractions of crystallized Si with the classical nucleation and growth theory. The

nucleation rate as a function of reciprocal temperature is shown in Fig. 33. The

nucleation rate increases exponentially as the annealing temperature is increased. For

500A thick a-Si films, the nucleation rates are larger than those for 1000 A thick a-Si.

The slopes of both experimental data agree quite well. The activation energy, Qn, of

the nucleation rate, l ,was determined to be 5.6 eV from the slopes of the Arrhenius

curve ( It a exp[ Q./kT]).

This activation energy value will be used below to estimate the Gibbs free energy

change, AGk*, involved in forming a critical cluster. In a rough approximation, the

steady state nucleation rate, I,, is expressed in terms of the simple product of the number

of critical clusters unit volume, Ck which is proportional to exp ( -AGk*/ kT), and the

rate of attachment of atoms to the critical nucleus which is the same as growth rate, vg,

which is proportional to exp ( Q / kT).

Thus,

I = A- exp (-Qn~ /kT ) B-Ckg =C-exp [-(Qg + AGk *)/kT (7)

Q. Qg + AGk Q

AGk* =Qnu -Qg

where A, B, and C are constants, and Qg is the activation energy of growth rate.




















1014


1013




1012




1011


1010


1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 1.22

1000/T (1 / K)


Fig.33 Temperature dependency of the steady state nucleation rate, which is determined
by fitting the volume fraction with Avrami's equation.


T -t50A





V1
-- N-
\= =,,.........


=^::=",^== = ........= =





EEE=E==:E==^


0
E

u

L.

C-
3


G)


C








Therefore, AGk can be determined from Qn and Qg.

4-2-5. Growth Rate

The growth rates were extracted from the measured volume fractions similarly as

the nucleation rates. Fig. 34 shows some growth rates as a function of reciprocal

temperature. From Fig. 34, the activation energy of growth was acquired based on the

Arrhenius equation. It is 3.7 eV. This value is very important, along with the activation

energy for nucleation to determine the interfacial energy of the a-Si / c-Si interface. The

a-Si thickness dependence of the growth rate is shown in Fig. 35. The growth rates

increase as the thickness decreases. This may result from a larger contribution of surface

to total volume in thinner films compared to thicker ones. Generally, the mobility of

atoms at the surface is higher than in the bulk. The values of growth rates are very close

to the solid phase epitaxial regrowth rates (1~100A) which were measured at the same

temperatures by Olson et al. '00

4-2-6. Crystallization Time

The crystallization is a combination of nucleation and growth. Therefore, the

median crystallization time, ts5, in which 50% of a-Si have crystallized, was determined

from the fitted volume fraction curve. The temperature dependence of t0 is shown in

Fig. 36. The activation energy for the crystallization is 3.78 eV. It is very close to the

activation energy for the incubation time, r, which is 3.97eV. Actually, the incubation

time is related not only to the nucleation process, but also to the growth process.

Therefore, both activation energies are supposed to be very close in value.















10-5


10-6




10-7


10-8


1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 1.22
1000/T (1 / K)







Fig.34 Temperature dependency of the growth rate, which is determined by fitting the
volume fraction with the Avrami equation.


-a-i 500 A
a-5 1000A
*i
.... ... .... ,,


















180

160
600C
140 o 62
14 65(O



S100 -

S 80 -





20 -
20

0 i *-1 0 1 1--
0 200 400 600 800 1000 1200 1400

Thickness (A)



Fig.35 Thickness dependency of the growth rate. It was determined by fitting the volume
fractions ( see Figs. 21,22, and 23 ) with the Avrami equation.










104




103


102


101




100


500 525 550 575 600


625 650 675 700


Annealing temp. (C)







Fig.36 Temperature dependency of the median crystallization time, tso, for various
thicknesses of Si films.


A a-si 1000A
a-Si 500 A
N a-Si 300A






,,,







4-2-7. Grain Growth Mode of Crystallized Si Particles

The grain growth mode for nucleated crystalline Si in amorphous Si determines

the shapes of the final grains with the preferred nucleation sites and nucleation rates.

According to Avrami's model 42, the grain growth mode can be extracted from the

variation of In [ 1 / ( 1 X )] with annealing time as shown in Figs. 30, 31, and 32. They

have the relationship as follow:

log { In[ 1/( -x)]} = -n-logT +n-log(t -) (6)

where X is the volume fraction, n is the exponential index which represents the time

dependence of crystallization, T is the incubation time, and to is the characteristic

crystallization time. For example, when the grains grow three dimensionally, n is 3 4.

For two-dimensional and one-dimensional grain growth it is 2 3 and I 2

respectively. Therefore, a slope, n, of each line, was determined from Figs. 30, 31, and

32. The results are shown in Fig. 37 as a function of annealing temperature. From Fig.

37 one can see that grain growth mode below 6000C is two dimensional because the

values of n lie in the range between 2 and 3. On the other hand, the grain growth mode

becomes close to one-dimensional as the annealing temperatures increase above 6000C,

since n is in the range of n = 1 2. In this experiment, the thicknesses of the amorphous

Si thin films are less than 1000A.

The preferred nucleation site is known as the interface between a-Si and the SiO2 layer.

Therefore, when an isotropic rate of growth and only nucleations on the interface are

assumed, the grain growth mode will depend on the density of nuclei at the interface and

on the ratio between the film thickness and the mean distance, D, of the nuclei at the

interface which is the same as the average grain size in columnar or disc shape grains.













































nA


1001
500
S300




I -


500
500


0
A
A


525 550 575 600 625 650 675


700


ANNEALING TEMP. ("C)




Fig.37 Temperature dependency of the slope of In [ 1 ( 1-x ) ], n, for various
thicknesses of Si films.


t







interface which is the same as the average grain size in columnar or disc shape grains.

For example, if the nucleation rate becomes larger at high annealing temperatures, the

mean distance, D, becomes smaller due to a larger density of the nuclei. This results in

an impingements of the growing Si crystallites to each other in the lateral direction.

Therefore, the Si cyrstallites can grow only one dimensionally along a direction

perpendicular to the a-Si / SiO2 interface. On the other hand, at low annealing

temperature, the density of nuclei is lower. Therefore, if the mean distance, D, is larger

than the film thickness, they will grow in a two-dimensional mode. As the annealing

temperature increases, the grain growth mode changes from two-dimensional growth to

one-dimensional as shown in Fig. 37.

4-2-8. Final Grain Size

The final grain size can be extrapolated from the nucleation rate and growth rate

in a way which was proposed by Iverson et al.17 If the grain shape is disc like or

columnar, the area of the final grain, Ag, is expressed as:

Ag =lim[N,(t)]-' (8)
t -> o0
where Ns is the density of grains, which is the number of grains per unit area.The density

of grains as a function of time is


N,= It(/d).[ ( t') ].dt' (9)
0
where IL is the steady state nucleation rate, d is the thickness of the a-Si film, t is the total

annealing time, and x (t') is the volume fraction of crystalline Si at a given instant time,

t'. In other words, 1-X( t' ) is the volume fraction of uncrystallized a-Si at a given instant







time. x(t' ) is given by equation( 4 ). When eqn. (9) is inserted into (8), and then

integrated, it yields:

Ag = [ n-( -d-I-vg2 / 3 )1~ / I-d ]- [ ( 1/n) ]-' (10)

where n is the exponential index in Avrami's equation, Vg is the growth rate, and F( 1/n)

is a gamma function. If the grain shape is a disc, the final grain size, dg, is the diameter,

thus;

dg = ( 4-A / )2

= {(4-nv,2M /3L ) [ 1/.( xI -d) ]i }0" [ ( I/n)]-'2 (11)

Details of the appropriate derivation are shown in Appendix B. Based on ( 11 ), the final

grain sizes have been calculated. The results are shown in Fig. 38. As expected, the final

grain size decreases as the annealing temperature is increased because of a higher

nucleation rate. However, the grain sizes decreases slightly at 5750C compared to those

of the samples annealed at 6000C. The reason for that trend is speculated that the growth

rate at 5750C may be so slow that finally, more nuclei can be nucleated compared to the

numbers of nuclei at 6000C in spite of lower nucleation rates. The activation energy for

the final grain size obtained from the slope of the fitted line is 1.15ev. The estimated

grain size is in the range of 351A- 1789A. In order to examine these estimations, a

cross-sectional TEM was performed on a 500A thick sample, which had been annealed

at 6000C for 20 hours. In Fig. 39, the average grain sizes is estimated to be

approximately 1000A. The optical method yields for this sample about 1410A. Thus,

TEM and DR provide similar results with the advantage of DR that the the results are

obtained much quicker and less expensive and tedious. The results of this section are

summarized in table 1.















1000


1.04 1.06 1.08 1.10 1.12


1.14 1.16 1.18 1.20 1.22


1000/T( 1/K)



Fig.38 Calculated final grain sizes obtained from the kinetic data for unseeded
crystallization.


A 300A
S* 500A
1000A





A


100











































0.1 m



Fig.39 Cross- sectional TEM of a 500A thick Si thin film on a 6000A thick SiO2/ (100) Si
substrate annealed at 6000C for 20 hrs..









Table I Summary of the results in section 4-2.


Item Results


Volume x = [ R-si /(R-si R-si) x= 2son x [ 1 -( 1+ 0.46C )/
fraction of c- ( 1 0.46C) ] from
Si
DR


Incubation T = 2 200 min. at
Q, = 3.97 eV 575-
time 6500C


Nucleation Lt= 3.69 x 10'1 2.51 x 10' / cm3
rate Qnu = 5.6 eV


Growth rate v = 1 ~ 169 A/ sec
Q, = 3.7 eV


Crystallizatio tso = 9 850 min.
n time Qr = 3.78 eV


Growth mode 2- dimentional ( n = 2.3-2.44 ) for Temp. < 6000C
of grains 1- dimentional ( n = 1.3-1.7) for Temp. > 600C


Final grain
size d,= 351- 1789 A

________________________________







4-3. Top Seed Induced Crystallization

A new technique was implemented with the goal to improve the crystallinity of Si

thin films on insulators as explained in the chapter 3. In short, single crystalline bulk Si

wafers was played on as-deposited amorphous Si thin films ( on a 6000A SiO2 layer ),

and then annealed at 5500C. In order to avoid random nucleation and growth of the

crystalline Si, they were annealed shorter than the incubation time of random nucleation.

This way, a new mode of crystallization is expected to occur. The crystallization is

induced from the top interface between the a-Si and the seed wafer. Specifically, (100)

single crystal bulk Si was put on the a-Si layer, and then annealed at 5500C for five hours

in an N2 atmosphere

4-3-1. Top Seeding Effect

According to previous experiments involving solid phase crystallization without

any seed, the grain sizes are less than 2000A Further, the preferred plane of the grains

is (111). In order to fully crystallize the sample, it takes almost 20 hours at 6000C.

These process conditions and properties are not quite satisfactory for industrial

production, e.g. for thin film transistors ( TFT ) on glass substrate in AMLCD. For a

comparison, as-deposited amorphous Si before annealing, and unseeded crystallized Si

were analyzed utilizing XRD and TEM. Fig. 40 shows the XRD spectrum of as-

deposited a-Si. A strong Si (400) peak at 20 = -69, which stems from the (100) Si

substrate, is obtained, because normally incident CuIKa (X=1.5405A ) X-rays were used

for the analysis. As a reference, the CuKa, (X=1.5405A) X-ray penetration into Si is

shown in Fig. 41. G(x) is the accumulated fraction of diffracted intensity from a given

depth of Si. For example, if G(x) is near 1, the contributed depth of Si to the diffracted






xl03 1.80


Si (400)


1.60

1.40

1.20

1.00

0.80


0.60

0.40


0.20


20.0
100.0
80.0:
60.0
40.0
20.0

20.0


30.0 40.0 50.0 60.0 70.0


80.0


Si SILICON. SYN
27-1402


30.0 40.0 50.0 60.0


70.0


80.0


Fig.40 XRD spectrum of an as-deposited 1000A a-Si thin film on a 6000A SiO2/(100) Si substrate


















0.8





0.6 -


G(x)


0.4





0.2




0 --1I I
0 20 40 60 80 100 120

x ( pm)

Penetration Depth
G(x) = 1 exp -2-152.1956- x
sin 34.55- )
180))







Fig.41 Simulated G(x), accumulated fraction of diffracted intensity versus depth in Si for
Cu K,, ( k = 1.5405A ) radiation





x103
2.00


1.80

1.60

1.40

1.20


1.00


0.80

0.60

0.40

0.20


Si (ll1)


20.0


Si (400)


30.0 40.0 50.0 60.0


70.0


Fig.42 XRD spectra for annealed. 1000A Si thin films on a 6000A SiO2/(100) Si substrate at 5500C : a) for 5 hrs.and b) 28 hrs.








depth of Si. For example, if G(x) is near 1, the contributed depth of Si to the diffracted

intensity is 100 upm. In this experiment, the thickness of the amorphous Si layer is

1000A, and the thickness of the Si02 is 6000A. Therefore, the (100) Si substrate

contributes greatly to the diffracted X-ray intensity. One can see a Kp X-ray diffraction

peak for Si (400) at 20 = -620 because Kp was not filtered completely. However, there

are no other Si peaks except (400). This means there is no crystalline Si in these

amorphous Si films. The amorphous Si films were annealed at 5500C for 5 and 28

hours without any seed ( Fig. 42 ). For the five hours spectrum, there are no Si peaks

except (400). However, after 28 hours annealing, one observes a Si(l 11) type peak, like

in the previous XRD spectra for 6000C annealed samples. This suggests that (111) is

the preferred nucleation plane for random nucleation. In order to analyze the

microstructure of these samples, plan view TEM observations were made. The

microstructure and electron beam diffraction of as-deposited amorphous Si is shown in

Fig. 43. In the bright field image there is no contrast due to any existing crystalline phase

at x50,000 magnification. Further, electron diffraction shows a typical amorphous

diffusive halo ring pattern. Fig. 44 depicts the microstructure and electron diffraction

pattern for a (100) Si substrate. A clear (100) zone axis diffraction pattern is seen. If

amorphous Si is annealed without a seed, it becomes a typical polycrystalline Si structure

due to random nucleation. Its polycrystalline microstructure and electron diffraction are

shown in Figs. 45 and 46. Its electron diffraction is a typical ring pattern due to the

contribution from many grains. In Fig. 46, small grains are shown. In dark field image,

white spots are the growing microcrystalline Si particles. Their sizes are quite small (<





























1 4m
a)





















b)

Fig.43 Plan view TEM pictures of as-deposited 1000A a-Si, x50,000 a) bright field b)
diffraction pattern.


I 1




























rntm
a)




















b)


Fig.44 Plan view TEM pictures of a (100) single crystal Si substrate, x50,000 a) bright
field b) diffraction pattern.





























a)




















b)


Fig.45 Plan view TEM picture ofa 500A Si thin film annealed without a seed at 550C
for 10 hrs. x7,300 a) bright field b) diffraction pattern.













a)













b)














c)




Fig.46 Plan view TEM pictures of a 500A Si thin film annealed without a seed at 550OC
for 10 hrs. x20,000: a) bright field b) dark field, and c) electron diffraction pattern