A study of the redistribution of iron, chromium and molybdenum transition metal impurities in SIMOX structures during hi...


Material Information

A study of the redistribution of iron, chromium and molybdenum transition metal impurities in SIMOX structures during high-temperature furnace annealing
Physical Description:
Puga, Maria Margarida Sancho, 1955-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 24458733
oclc - 23725721
System ID:

Full Text








I would like to express my deepest appreciation to my

committee chairman, Professor R.E. Hummel, for his valuable

guidance and constant encouragement throughout this research

work. His many comments and suggestions were greatly

appreciated. I would also like to acknowledge my gratitude

to my committee cochairman, Dr. D.E. Burk, for her constant

support and valuable technical assistance during the course

of this investigation. Her extensive interaction and many

special meetings and discussions were greatly appreciated.

I would like to extend my gratitude to my committee members,

Dr. P.H. Holloway, Dr. R.T. DeHoff and Dr. S.S. Li, for

their assistance and valuable time and patience. Thanks are

especially extended to Dr. J.J. Hren for his guidance in

selecting the topic of this investigation.

I gratefully acknowledge that financial support for

this work was provided by the High Technological Council of

Florida and the department of Materials Science and


Many other people have contributed to the realization

of this work, and for that I am very grateful. I would like

to express my thanks to Mr. James Chamblee and Mr. Jim Hales

from the Microelectronics Laboratory for their valuable

technical assistance and constant help during the course of

my experimental work in that laboratory. Special thanks are

directed to Mr. Jim Hales for his patience and time during

the spreading resistance measurements. I would like to

extend my appreciation to Mrs. Carolynn McGilvray from the

Integrated Circuit Group in the Electrical Engineering

Department for her valuable help in dealing with all the

paperwork and bureaucratic details necessary to the success

of this research work.

Many thanks are directed to the SIMS group at North

Carolina State University in Raleigh, N.C., for their

valuable technical assistance with the SIMS measurements,

and to Dr. J.J. Hren and Mrs. Susan Hofmeister for their

friendly hospitality during my numerous trips up there.

I would also like to express my appreciation to

Dr. Augusto Morrone and Mr. Eric Lambers from the Major

Analytical Instrumentation Center for their valuable

technical assistance and help with the TEM and the Auger

analysis, respectively. I would like to extend my gratitude

to Ed Solley for his help in the optimization of the cross-

sectional TEM technique. Many thanks are extended to my co-

workers Wei-Xi and Dave Malone for their encouragement and

good humor during the course of writing this dissertation.

Special thanks are directed to my friends from the

"Portuguese Community" in Gainesville, in particular the

Principe family, for their valuable friendship and good

humor all these years. To my friends, Eric Lambers, Teresa


Abrantes, Rui DaCosta and Tony Buonaquisti, I would like to

express my deepest gratitude for their continued moral

support and encouragement all these years. And most of all,

I would like to thank my parents, Josd and Maria de Lourdes

Puga, for their immense patience, understanding and

encouragement throughout the years, and their belief in me

that made it possible to reach this far.


ACKNOWLEDGMENTS ............................................ ii

ABSTRACT .................................................. vii


1 INTRODUCTION ......................................... 1

2 LITERATURE REVIEW ................................... 10

2.1. Historical Background ........................... 10
2.2. Fundamentals of SIMOX Technology ............... 15
2.2.1. Ion Implantation Process ................ 15
2.2.2. Defect Formation in SIMOX ............... 24
2.2.3. Post-Implantation Annealing ............. 27
2.2.4. Effect of SIMOX Processing on Device
Characterization ...................... 32
2.3. Transition Metal Impurities in SIMOX ........... 33

3 EXPERIMENTAL PROCEDURE ............................... 38

3.1. Introduction ................................... 38
3.2. Sample Preparation .............................. 39
3.3. Materials Characterization ..................... 40
3.3.1. Secondary Ion Mass Spectroscopy ......... 42
3.3.2. Transmission Electron Microscopy ........ 50
3.3.3. Auger Electron Spectroscopy ............. 56
3.3.4. Differential Reflectometry .............. 58
3.4. Electrical Characterization .................... 61

4 RESULTS AND DISCUSSION ............................... 66

4.1. Introduction ................................... 66
4.2. SIMOX Co-Implanted with Transition Metal
Impurities ................................... 67
4.2.1. Analysis of Metal Impurity
Distributions .......................... 67
4.2.2. Microstructural Analyses ............... 111
4.2.3. Evidence for Metal Precipitates from
TEM Analysis ......................... 153
4.2.4. Oxygen Depth Profiles .................. 159

4.3. Correlation of Transition Metal Redistribution
SIMS Profiles with Damage Recovery .......... 174
4.4. SIMOX Implanted without Transition Metal
Impurities .................................. 188
4.4.1. Microstuctural Analyses ................. 188
4.4.2. Oxygen Depth Profiles ................... 204
4.5. Effect of Transition Metal Impurities
on the Electronic Structure of SIMOX ........ 209
4.6. Resistivity Analysis .......................... 219

5 CONCLUSIONS AND FUTURE WORK ........................ 226

APPENDIX ................................................. 232

BIBLIOGRAPHY ............................................. 239

BIOGRAPHICAL SKETCH ...................................... 250

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





Chairman: Rolf E. Hummel
Cochairman: Dorothea E. Burk
Major Department: Materials Science and Engineering

The formation of subsurface dielectric Si02 layers by

ion implantation of high doses of oxygen into single crystal

silicon (SIMOX) is one of the most promising silicon-on-

insulator (SOI) technologies for applications in VLSI

circuits. However, ion implantation results inadvertantly

in structural defects and undesired transition metal

impurities which may degrade the quality of the top silicon

layer and its suitability for device fabrication.

The effects of post-implantation annealing on the

density and depth distribution of Fe, Cr and Mo introduced

in the SIMOX structure during implantation were investigated

by secondary ion mass spectroscopy (SIMS). SIMOX structures

were formed by implanting high doses of oxygen ions


(2.25x1018 0+/cm2, 150 KeY, 5100C) into (100) n-type silicon

single crystal substrates using a medium current implanter

(10 mA)o Furnace annealing was carried out between 9000 and

13000C for 2 and 5 hours in Ar+l.5%02. SIMOX structures

were characterized by transmission electron microscopy

(TEM). Auger electron spectroscopy (AES) was used to study

the oxygen depth profiles. These results were related to

the impurity distribution profiles.

It was found that before annealing most of the Fe, Cr

and Mo are located in the top silicon layer. After

annealing, the distribution of Fe, Cr and Mo is dominated by

the tendency to segregate in various regions of the SIMOX

structure. The factors or combination of factors that

affect the structure is as follows: (1) supersaturation of

silicon interstitials, (2) growth of a surface SiO2 due to

annealing in a partially oxidizing argon atmosphere, (3)

dislocation trapping, and (4) metal precipitation. The

dislocation density in the top silicon layer remains

constant (109/cm2) with annealing. It was found that Fe is

gettered in the form of FeSi2 precipitates. But no evidence

of Cr or Mo precipitates was detected.

A comparitive study of Fe, Cr and Mo contaminated

structures with non-contaminated SIMOX structures implanted

and annealed under the same conditions did not reveal any


substantial structural differences. Furthermore, an

investigation performed by differential reflectometry (DR)

showed that transition metal impurities to the extent used

in this work do not change the electronic band structure

significantly. Finally, spreading resistance profiles (SRP)

were performed to investigate the resistivity of the top

silicon layer.


The electronics industry has been characterized by the

fast evolution of semiconductor technology over the last two

decades. The cost performance of most microelectronic

products has improved substantially due to a greater ability

to integrate more and more circuit elements on a single

silicon chip. But this new era of very large scale

integrated (VLSI) circuits with continuing reduction in

feature size to values less than 1 gm has created many

problems for bulk silicon technology such as parasitic

coupling between devices or adjacent circuits through the

silicon substrate. As a result, considerable interest has

been directed towards the development of new technologies

such as silicon-on-insulator (SOI), which have potential

application to the production of very large scale integrated

circuits (VLSIC's), high voltage IC's, large area IC's, and

vertical IC's (1).

A typical silicon-on-insulator structure is illustrated

in Figure 1.1. It consists of 3 major layers, (a) a thin

defect-free silicon layer about 0.1-0.2 pm thick (top Si

layer), (b) an insulating layer of silicon oxide with a

0 .1 -0 .2 g~m

0.5 pgm

R'............. ..

5 ..........

Figure 1.1. Schematic cross-section of a typical
silicon-on-insulator structure.


thickness of 0.3-0.5 Am (buried oxide layer) sandwiched

between the top silicon layer and (c) the substrate, which

is single crystal silicon and has the same crystallographic

orientation as the top silicon layer. Abrupt interfaces

separate the buried Si02 dielectric from the upper and lower

silicon layers. The top silicon layer can be used as seed

for epitaxial growth thus increasing the thickness of the

single-crystal silicon film available for device fabrication


The buried silicon oxide layer dielectrically separates

the active device layers from the substrate thus offering

considerable advantages over conventional bulk silicon

technologies (3). It increases operating speed and lowers

power dissipations by reducing parasitic capacitances.

It achieves complete device isolation which can minimize

capacitive coupling between various circuit elements over

the entire IC chip and eliminates latchup in complementary

metal oxide semiconductor (CMOS) circuits (a disruptive

process in which current flows between neighboring

transistors). This permits greater packing densities.

Also, SOI IC's are less susceptible to radiation induced

failure because the charge carriers formed in the substrate

by ionizing radiation cannot diffuse through the insulator

to the electrically active layers. Only the charge produced

in the thin surface silicon layer is detrimental to the

device performance. This volume is very small compared to

the bulk volume and therefore makes SOI a powerful

technology for space and defense applications as well.

The first attempts to synthesize dielectrically

isolated (D.I.) Si islands to achieve complete isolation of

the elements of an IC consisted of single crystal silicon

islands supported by a thick deposited polysilicon substrate

which were electrically insulated from each other by SiO2

(4). But due to the complexity of this process and the

limited size of the Si islands, new alternatives have been

pursued (5, 6).

Extensive efforts have been directed towards the

development of SOI technologies. The main goal is to obtain

a high quality silicon crystal in a thin layer electrically

isolated from the substrate, which provides mechanical

support. Silicon-on-sapphire (SOS) substrates formed by the

heteroepitaxial growth of silicon on single crystalline

A1203 have generated considerable interest (7). However,

SOS has not had the expected commercial success because of

the high cost and mechanical fragility of the sapphire

substrate, Al outdiffusion into the Si and the large defect

density (microtwins and stacking faults) of the SOS films at

the Si/Sapphire interface. These planar defects limit the

carrier mobilities to values smaller than those obtained by

bulk single crystalline silicon. IC's fabricated on SOS

usually show mobility degradation and an increase in leakage

current (8). Low lifetimes hamper the use of SOS substrates

for circuits where the lifetime is important such as dynamic

random access memory (DRAM) (9).

One of the most promising techniques of synthesizing

SOI structures is the implantation of high doses of oxygen

into silicon (10). This process has been designated SIMOX,

an acronym for "Separation by IMplanted Oxygen" (11). In

this technique, oxygen is implanted with a dose of the order

of 1018/cm2 to achieve stoichiometric Si02, at an energy

high enough such that a buried oxide layer is formed below

the surface and a surface single crystal silicon film

remains above the buried oxide layer, which can be used as

seed for epitaxial growth. Nitrogen has also been used to

form buried dielectric nitride layers (12). Although both

silicon oxide and silicon nitride have widespread

applications as insulators in integrated circuits, stronger

emphasis has been placed on the use of oxygen ions due to

the higher stability of the Si/SiO2 interface (13). When

the amount of implanted oxygen exceeds the amount needed to

synthesize a stoichiometric SiO2 layer thicker than the ion

straggle distance, oxygen redistributes and a thicker oxide

layer with abrupt interfaces is formed. Nitrogen

implantations, on the other hand, are not self-limiting

(14). At high dose levels, larger than those required for

the formation of stoichiometric Si3N4, the excess N2 remains

embedded within the buried nitride layer. This excess

nitrogen may affect the long term stability of the material

since it can diffuse through the substrate (14). The

resistance to radiation of a nitride insulating layer in the

SOI structure is more susceptible to degradation than an

oxide layer because of storage of electric charge at the

interface with the substrate.

It has been reported that SIMOX structures are superior

to SOS both electrically and microstructurally (8, 15).

Devices built on SIMOX substrates have shown uniformity in

threshold voltage and mobility, and a leakage current of

2x10-12 A, which is about three orders of magnitude less

than the same circuits fabricated on SOS material (16).

Other SOI techniques being developed involve epitaxial

regrowth from the melt, chemical vapor deposition (CVD)

and/or solid phase epitaxy, wafer bonding and oxidation of

porous silicon (FIPOS) (5, 6, 9, 17, 18). Constraints of

lattice mismatch and substrate access, high defect

densities, and geometrical constraints on the growth of the

SOI films are some of the limitations characterizing these

SOI approaches. SIMOX offers the advantage of producing SOI

substrates with good reproducibility with fewer

crystallographic defects and promises high yields for large-

component-number devices as well as fine lithography for

small devices and large flexibility for circuit design (19).

Therefore, the formation of buried dielectric layers by the

high-dose implantation of oxygen into silicon and subsequent

thermal annealing has emerged as the most potential

candidate for advanced VLSI circuit processes and defense


However, the potential benefits to VLSI performance

afforded by the buried oxide silicon-on-insulator technique

may be compromised by the presence of defects and undesired

metallic impurities in the top silicon layer and in the

buried oxide. The defects, formed during the oxygen

implant, are mainly oxide precipitates and dislocations.
The density of dislocations ranges from 106-1010/cm2 (20).

Oxygen precipitation results from supersaturation of the

Si/SiO2 interface with excess oxygen above its solid

solubility limit in silicon. Defects affect device

performance by (a) reducing carrier mobility, (b) enhancing

diffusion of dopants along defects which causes high leakage

current, and (c) degradation of the gate oxide thus

affecting threshold voltage uniformity and gate oxide


The metal impurities arise inadvertently from the

secondary sputtering of the implanter housing onto the wafer

surface during the high-dose oxygen implant. Mass filtering

of the ion beam during implantation would be expected to

lead to a contamination-free implantation. However,

sputtering of apertures and other components in the

implanter by the ion beam can unintentionally introduce

metallic contamination into the implanted material. Many

ion implantation parts are manufactured from commercial

grades of stainless steel.

Integrated circuit device performance is strongly

influenced by the distribution of impurities resulting from

implantation and subsequent annealing processes. These

metallic impurities may form deep level traps, precipitate

out in the gate oxide surface, remain interstitially or

substitutionally in the top silicon layer, and/or decorate

dislocations thus adversely affecting device yield and

performance (21-24). The importance of a deep trap to the

electrical characteristics of a circuit depend on the

position of the energy level within the band gap and the

capture and emission rates when the trap is empty or full.

Metallic impurities that form deep traps near the middle of

the band gap cause excess junction leakage by reducing the

carrier generation lifetime. This can be detrimental to

high density circuits because of excess power dissipation.

Small precipitates of metallic impurities may act as nuclei

for oxygen induced stacking faults and the surrounding

partial dislocations may be decorated by precipitates of

metallic impurities. This can adversely affect the current-

voltage characteristics of a pn junction (24).

The purpose of this research is to study the effect of

different post-implantation annealing treatments on the

behavior of transition metal impurities inadvertently

introduced in the SIMOX structure during oxygen implantation

in a medium current implanter (10 mA). With the advent of

high current implanters (100 mA), impurity contamination has

been significantly reduced by lining the implanter beam line

and wafer station with silicon. However, metal impurity

contamination continues to be a problem in the semiconductor

industry. Control of the metal impurities distribution

during annealing is essential since several annealing steps

are performed during manufacturing of devices. Therefore,

the primary objective of this investigation is to provide a

better understanding of the factors involved in the behavior

of transition metals in SIMOX during annealing. The SIMOX

wafers used in this study were provided by the semiconductor

industry. The distribution of the transition metal

impurities in the structure will be correlated to the

microstructural changes that occur upon annealing while

maintaining the implantation conditions constant. The

effect of radiation-induced damage and annealing and the

presence of transition metal impurities on the electrical

response of the material will be investigated. A comparison

with SIMOX wafers implanted under the same conditions but

without metal co-implantation will be presented.

This dissertation will commence with a literature

survey in Chapter 2 that describes the previous work done on

the SIMOX technology and pertinent aspects of the resulting

structure. The experimental procedure is described in

Chapter 3 with respect to the characterization techniques

used in this study and sample preparation. Chapter 4

contains the experimental results and the discussion of the

results obtained both on SIMOX wafers co-implanted with and

without transition metal impurities. Conclusions and

suggestions for future work will be presented in Chapter 5.


2.1. Historical Background

The idea of converting certain elemental surfaces to

oxides and nitrides using high doses of oxygen and nitrogen

ions dates back to 1956 (25). But it was only several years

later that the formation of a silicon oxide layer in silicon

by oxygen ion implantation into single crystal silicon was

first reported (26). Since then several workers have

attempted to demonstrate the feasibility of preparing

implanted oxide and nitride layers, which would be

electrically equivalent to those thermally grown. Ion

implantation offers a very precise and reproducible way of

introducing ionized-projectile atoms into targets at

specific depths, which can be selected for each particular

application. The major drawback of ion implantation,

however, is the damage that needs to be removed by a post-

implantation annealing step.

Homogeneous silicon nitride films about 1.5 gm below

the surface were the first buried films to be formed by ion

implantation of 1 MeV nitrogen into silicon (27). Dexter et

al. (28) proposed that buried nitride insulating layers

could be used for dielectric isolation in integrated

circuits (IC) devices. Doses of 5x1017 N+/cm2 were

implanted at 150 keV into silicon to obtain peak

concentrations of 5xl021/cm3. The samples were annealed at

12000C in a dry nitrogen atmosphere to form a buried nitride

layer. Epitaxial growth of silicon layers on the implanted

surface showed that the silicon surface remains crystalline

and is appropriate for building devices in this structure

(28). The breakdown field measured in the nitride layer was

7x105 V/cm, slightly lower than that of thermally deposited

silicon nitride (28). Similar results were later reported

by Bourguet et al. (12). These authors determined that for

a 180 keV nitrogen implant, a dose of 2x1018 N+/cm2 was

necessary to produce a continuous layer of silicon nitride

and to achieve dielectric isolation.

With the basic steps to form buried insulators already

laid out, most workers started to concentrate on the

formation of buried oxides. Dielectric isolation in IC's

requires that the characteristics of the interface between

the top silicon layer and the dielectric layer be well

understood. The higher stability of the Si/Si02 interface

and the better knowledge of its characteristics have made

the formation of SiO2 more attractive as an insulator

material (13).

Early SiO2 layers formed by implantation were used for

protection of pn junctions (29). Several studies were then

reported on the formation of low energy (30 key) surface

oxides (30-32). These oxide layers were mainly assessed by

infrared absorption and electrical methods and showed poor

quality. Buried silicon oxide layers produced with

implantation of 1 MeV oxygen ions at a dose of 1x1016/cm2

followed by annealing at temperatures greater than 10000C

were successfully obtained (33). The microstructure of the

annealed specimens showed a defect-free silicon surface

followed by a highly twinned transition region and an

implanted oxide area which contained an ordered Si-O phase

(33). A similar structure was reported later by Hayashi et

al. (34). Electrical characterization of implantation grown

silicon oxide films by capacitance/voltage (C/V) and

current/voltage (I/V) measurements showed that they were

equivalent to those of thermally grown silicon oxide after

annealing in N2 for 1 hour at temperatures that ranged from

7500 to 11000C (35). The chemical composition of the films

was stoichiometric SiO2. It was then suggested that buried

oxides could be used to achieve complete isolation of the

elements in IC's. However, it was only in 1978 that Izumi

et al. (10) first reported the application of the process of

isolation by ion implantation for fabrication of devices and

circuits. The authors above showed that CMOS devices built

on an silicon layer epitaxially grown on an oxide implanted

silicon substrate operate at twice the speed of the same

circuits fabricated on conventional bulk silicon. This

increase in speed was attributed to the reduction of

parasitic capacitances. The substrates were formed by

implantation of 1.2x1018 0+/cm2 at 150 keV and annealed at

11500C for 2 hours in N2.

Devices can be built directly above the top silicon

layer without additional epitaxial growth (36). Metal oxide

semiconductor (MOS) circuits fabricated directly on buried

oxide SOI have shown good current-voltage characteristics.

The electron and hole mobilities obtained for n- and p-

channel devices were 600 and 300 cm2/sec, respectively.

Successful MOSFETS with submicron channel lengths were also

fabricated directly on buried oxide SOI (37). The SOI

devices showed a smaller threshold voltage shift as a

function of channel length than devices fabricated in bulk

silicon. This result was attributed to the elimination of

the "punch through" effect by the SOI structure.

The idea of implantation and epitaxial growth was then

extended to a three dimensional SOI structure where each

Sio2 layer was stoichiometric and each silicon layer was a

single crystal with the same orientation as the substrate

(38). This process consisted in growing a 1.2 Am thick

epitaxial layer on the SIMOX surface previously annealed at

11500C for 2 hours, followed by implantation of 5xl017/cm2

molecular oxygen ions at 160 keY to form a buried oxide

within the epitaxial layer.

Successful fabrication of a large SIMOX circuit, a 256k

CMOS SRAM with an access time of 12 nsec demonstrated that

the buried oxide is production worthy and is the most

promising of all the SOI techniques (11).

However, acceptance of this technique viability has

been very slow due to the time required for processing

industrial wafers and the need for high flux implanters. As

mentioned before, this technique requires the implantation

of very large doses of oxygen (1018 0+/cm2) to obtain a

stoichiometric SiO2. Also, implantation must be carried out

at elevated temperatures (500-6000C) to mantain the

crystallinity of the surface layer as it will be discussed

in the following section.

Many efforts have been directed towards the

optimization of the oxygen implantation process and

application of high current implanters to reduce the

implantation time. Several improvements have been achieved

over the years. For example, modifications performed on a

standard Nova NV 10 mA-160 keV medium current implanter

permit the use of very large oxygen dose implants at

elevated temperatures (400-6000C) (39). Ten wafers, 4"

diameter are scanned at angles between 00 and 150. This

implanter allows the operator to specify and maintain a

preset temperature for the entire implant. The operation is

automatic and the temperature is continually monitored and

controlled during the implantation. The development of a

high-current oxygen implanter which operates at 200 key with

an oxygen ion beam source of 100 mA was reported recently

(18, 40). Wafers of 100, 125 and 50 mm size are preheated

at 5000C such that the final wafer temperature (due to

additional beam heating) is less than 9000C. The ion source

lifetime is greater than 40 hours and the doping uniformity

is about 5%.

The need to obtain a SIMOX structure suitable for

commercial device fabrication has lead numerous researchers

to perform detailed material analysis of the SIMOX layers in

order to acquire a thorough understanding of the effect of

ion implantation processing and subsequent thermal annealing

on the materials properties and electrical characteristics.

Although there is considerable knowledge about the structure

of the SIMOX layers, no systematic investigation on the

behavior of transition metal impurities co-implanted during

SIMOX processing has been performed to this date.

In the next section, a review of the process of

implantation and formation of buried oxide layers will be

presented. The effects of implantation parameters and

annealing conditions on the quality of the SIMOX

microstructure will be considered. This review will be

followed by a review on the properties of transition metals

in silicon and its effects in SIMOX.

2.2. Fundamentals of SIMOX Technology

2.2.1. Ion Implantation Process

Separation of silicon layers through implantation of

oxygen consists of implanting high doses of oxygen in the

order of l-2x1018 0+/cm2 into a single crystal silicon wafer

to depth ranges of 0.3 to 0.5 gm. A schematic illustration

of the SIMOX technique is presented in Figure 2.1. The top

surface layer of the wafer suffers less ion induced damage

than the silicon near the peak oxygen concentration, which

is completely amorphized.

As the oxygen ions enter the silicon wafer by virtue of

their kinetic energy, some of the ions may pass freely down

the open channels in the silicon lattice, but many displace

silicon atoms from their lattice sites, which in turn can

displace other atoms thus creating a cascade of collisions.

The channeling effect depends on the direction of the beam

relative to the crystallographic orientation in silicon.

The main loss mechanisms are electronic stopping and nuclear

collisions. Upon hitting the sample the energetic oxygen

ions loose energy through electronic interactions mostly.

This energy is ultimately dissipated as heat and by

ionization. Initially, very little energy is lost through

nuclear interactions. The momentum transfer is small and

the silicon crystal lattice suffers little ion induced

damage. But as the incident oxygen ion slows down, the

electronic interaction cross-section decreases and the

nuclear interaction cross-section increases. When enough

momentum is transferred from the incident ion to a nucleus

in the silicon target, displacement of the nucleus from its

lattice position occurs. If this nucleus has sufficient

momentum, it can cause secondary nuclear displacements, and

thus generate displacement cascades. As the process


Top Si (100)

Bured xideSiO2

si Substrate (100)


Figure 2.1.

Top Si (100)

Buried Oxide SiO21

Si Substrate (100)


Schematic representation of SIMOX process.
(a) Distribution of implanted oxygen
in silicon
(b) Damage distribution
(c) Annealing of buried SiO2
(d) Epitaxial Si growth.

continues, a larger number of atoms are displaced from their

lattice sites in the target. Eventually, the atoms have

less energy and both incident ions and displaced atoms come

to rest.

The displacement or ion implantation damage is created

as Frenkel pairs. The number of these pairs is given by the

modified Kinchin and Pease (41) expression kED/2Ed where k

(-0.8) is the displacement efficiency, ED is the energy

necessary to produce displacements and Ed is the

displacement threshold energy, which is approximately 15 eV

at room temperature for silicon (42). Amorphization of the

silicon structure occurs via a first order transition from

crystalline to amorphous phase when the energy deposited

into collisions exceeds 12 eV/atom, or 6.0x1023 eV/cm3 (43).

If the surface damage level remains below this critical

value, the surface layer remains a single crystal after

implantation. Amorphization first occurs at the depth of

the maximum collision energy deposition which is slightly

less than the projected range Rp, and spreads laterally

towards both the surface and greater depths in the crystal.

It has been reported that heavily damaged silicon relaxes to

an amorphous state when about 10% of the atoms have been

displaced from their lattice sites at a temperature low

enough to prevent long range migration of elementary point

defects (44).

Post-implantation annealing at high temperature

homogenizes the buried oxide due to further redistribution

of oxygen with incorporation into the buried oxide layer,

and allows recrystallization of the top silicon layer.

After annealing the SiO2 layer previously deposited on the

surface of the top silicon layer or grown during annealing

is removed. The top silicon layer can then be used as a

single crystal seed for epitaxial growth.

The range of oxygen implantation and associated damage

distribution can be obtained using Monte Carlo simulations

programs such as the TRIMM computer program (45). The TRIMM

computer program estimates the ion and deposited energy

distribution of incoming energetic ions into solid targets.

The ions may be of any atomic number at energies up to

2 GeV/amu. The targets may be made up of compound materials

with up to three layers made up of different materials.

The degree of implantation damage is strongly dependent

on parameters such as ion dose, ion beam energy, substrate

temperature and subsequent annealing conditions. As a

consequence, the microstructure of the top silicon layer is

determined by the implantation and annealing process.

Ion Dose and Ion Beam Energy Dependence. The

concentration profile of the oxygen atoms in silicon follows

a Gaussian distribution up to doses required to provide the

stoichiometric concentration for the formation of SiO2 at

the peak of the distribution (46). As the dose is further

increased, the Si/O atomic ratio exceeds the stoichiometric

ratio of Si:O, 1:2, at the distribution peak. The oxygen

distribution changes from a Gaussian profile (unsaturated

profile) to a flat topped profile when the buried oxide

layer is formed, and then develops into a square

distribution (saturated profile). The oxygen distribution

has been measured using x-ray photoelectron spectroscopy

(XPS), Rutherford backscattering (RBS), Auger electron

spectroscopy (AES) and secondary ion mass spectroscopy

(SIMS) techniques (34, 46-53).

The concentration of implanted oxygen never exceeds the

concentration of oxygen in SiO2. The excess oxygen diffuses

rapidly to the wings of the distribution and reacts with the

silicon atoms at the two interfaces leading to the formation

of abrupt Si/Sio2 interfaces (15, 53, 54). These sharp

interfaces have charge densities near that of thermal oxide

(16). The distribution of implanted oxygen tends to be

asymmetrical (55). Oxygen diffuses preferentially to the

top surface due to radiation-enhanced diffusion (56). The

continuous formation of new silicon interstitials in the top

silicon layer causes an asymmetry in the chemical driving

force. SIMS experiments with 018 tracer and nitride buried

markers have demonstrated that the damage region facilitates

the rapid growth of SiC2 at the upper interface of the

buried layer by providing a source of "energetic" silicon

i.e, a region of broken silicon bonds (57, 58).

The effect of oxygen ion dose on the microstructure of

SIMOX has been extensively studied (15, 34, 53, 59). As an

example, Figure 2.2 shows a schematic correlation of the

oxygen depth distribution with the microstructure in the top



Depth (Am)

i.2*1018 O+/cm2

i.8*1018 O+/cm2

2.4*1018 O+/cm2


Si Single-Crystal


SiO2 Precipitates


Buried Oxide SiO2
I l/Ilm IdIuried

Si Substrate

Figure 2.2. Schematic illustration of the effects of
ion dose on SIMOX.
(a) Oxygen depth profiles
(b) Top silicon layer microstructure.

silicon layer as a function of oxygen ion dose. The

formation of microtwins and a polysilicon region between the

top silicon and buried oxide layers was observed for the

implantation of 1.2x1018 0+/cm2 at 150 KeV. With increasing

ion doses from 1.8 to 2.4xl018 0+/cm2 at the same

implantation energy, only a single crystal Si layer was

observed without the microtwins and polysilicon intermediate

region (53, 59).

For a given peak, the depth and thickness of the

implanted silicon oxide increases with increasing energy

(46). Typical energies used are in the order of 80-200 keY.

High energies spread the impact of the oxygen ions into a

large volume of silicon, and an increase of the implantation

dose is required to form the stoichiometric buried oxide

layer. Lower implant doses can be used with lower energy

but result in a thinner oxide, closer to the surface and in

a more defective silicon overlayer. Mayema and Kajiyama

(14) have shown that the implantation of 70 keV oxygen ions

into silicon requires a dose of 1.5x1018 0+/cm2 to achieve

stoichiometry. Although these implantation conditions

resulted in a smaller thickness of the top silicon layer

(-0.05 Am), the crystalline quality of the film was reported

to be good enough to support epitaxial growth. For very

high dose levels an oxide extending to the surface can be

formed. Beyond a critical dose level the silicon oxide

layer thickness decreases with increasing implanted dose

(46). This effect was attributed to sputtering combined

with an increase in the stopping power once SiO2 is formed.

Analytical models to describe the evolution of the

oxygen profile were developed by Maydell-Ondrusz and Wilson

(52) as well as Jager et al. (60). These models specify the

effect of processing parameters on the position and

thickness of the buried oxide.

Substrate Temperature Dependence. Substrate heating

during implantation is a critical parameter in determining

the as-implanted SIMOX microstructure. In order to minimize

build-up of radiation induced damage, implantation is

carried out at elevated temperatures typically within the

range 400-6000C. High temperature during implantation

ensures that the damage at the front surface remains below

the critical value for the formation of amorphous silicon.

The influence of substrate temperature on the SIMOX

microstructure has been widely investigated (50, 61-64).

In-situ annealing can be accomplished either by using a

high current density ion beam (direct beam heating) or a

resistive heater. However, one of the problems of direct

beam heating is that during the initial stage of

implantation, the sample is at room temperature. Therefore,

the damage morphology that develops before the sample

reaches its equilibrium temperature will determine the

evolution of the as-implanted structure. For example,

the formation of a polycrystalline top silicon layer was

reported for implantation at temperatures of 1000C or below

(65). Substrate temperatures above 1000C but lower than

4000C result in the formation of a polysilicon layer between

the buried oxide and the top silicon layers (49). This

intermediate polysilicon layer disappears if higher

implantation temperatures are used (49, 66). During

irradiation at low C temperatures, clustering of non-

equilibrium concentrations of point defects occurs and

dislocation loops are generated. Once nucleated, their

removal may be difficult even at higher temperatures, and

they provide nucleation sites for oxygen precipitation. The

use of a resistive heater sets a lower limit on the sample

temperature and allows a constant temperature to be

maintained during the implant (66, 67). Thermal diffusion

permits the motion of defects and their annihilation either

by recombination or escape from the wafer surface, which is

a sink for these defects. However, it has been shown that a

compromise between in-situ annealing and enhanced oxygen

diffusion at higher temperatures must be observed to produce

a buried stoichiometric SiO2 and optimum crystallinity on

the top silicon layer (66).

2.2.2. Defect Formation in SIMOX

It is well known that ion implantation highly damages

the surface layers of specimens (68). Oxygen ion

implantation results primarily in a supersaturation of the

implant species and a non-equilibrium concentration of point

defects generated by nuclear collisions. As mentioned

previously (Figure 2.1), the distribution of the point

defects is determined by the damage profile, which has its

maximum closer to the surface than the oxygen peak

concentration. These processes lead to (a) nucleation of

Si02 precipitates, (b) point defect clusters, (c)

dislocation loops generated by collapse of point defects,

and (d) amorphous layers. The kinetics of these processes

are temperature dependent as evidenced by the effects of

implantation temperature on the as-implanted microstructure.

Supersaturation of interstitial oxygen in silicon above

its equilibrium concentration results in the formation of

oxygen precipitates and a volume expansion, as a Si02

molecule occupies approximately twice the volume of a Si

atom in the crystalline silicon matrix (69). The volume

expansion associated with the formation of SiO2 precipitates

is approximately 108% (70). The stress created by this

volume increase can be relieved either by (a) injection of

self-interstitials into the surrounding silicon matrix

according to the reaction

2Si + 20i -> Si02 + Sii (2.1)

where Oi and Sii are oxygen and silicon interstitials,

respectively, or by (b) punching out prismatic dislocation

loops. The first mechanism results in a supersaturation of

silicon self-interstitials in addition to the background

level already created by displacement damage. During the

initial stages of implantation, the excess interstitials

generated by both processes will tend to migrate through the

crystal. This may result in annihilation by recombination

with vacancies and/or escape towards the silicon free

surface if the diffusion length is large enough. As

described previously, this process is highly dependent on

the substrate temperature, which determines the rate at

which the defects are annihilated. However, it must be

considered that vacancies and interstitials have different

migration enthalpies (71). This means that the implantation

temperature will affect both the absolute concentration of

point defects and the relative concentration of vacancies

and interstitials. As implantation proceeds, the presence

of newly formed Sio2 precipitates hampers the movement of

interstitials since the diffusion coefficient of Si in SiO2

is very small (10-20 cm2/sec at 6000C) (72). Consequently,

the concentration of self-interstitials rises and becomes

higher near the upper interface with the buried oxide. This

supersaturation results in an increase of the surface free

energy, which the system tends to minimize by condensation

of point defects into dislocation loops.

The observation of strain-free precipitates in SIMOX

has lead to consider supersaturation of silicon self-

interstitials as the most plausible dislocation source

mechanism in SIMOX (70). This fact was also based on the

observation that the strain around a precipitate is low

(less than 0.01) (73). It has been reported that this value

is not sufficient to generate dislocations if precipitates

are less than 100 nm in size (70).

Dislocations near the buried oxide region may also

result from clustering of point defects that have escaped

from the amorphous region of the buried oxide during its

formation and from recombination with vacancies. Due to

straggling of the oxygen implant, the number density of the

dislocations in the top silicon layer does not reach the

critical value necessary for amorphization. It has been

shown that dislocations act as preferred sites for the

nucleation of SiO2 precipitates in silicon (74).

2.2.3. Post-Implantation Annealing

The microstructure of the top silicon layer is also a

strong function of the annealing conditions (73, 75, 76).

Still, microstructure studies have shown that, even after

annealing, the final structure is very dependent on the

structure formed during implantation (61, 62, 77-79).

Most of the early annealing studies on SIMOX were

performed at 11500C for 2 hours, either in N2 or Ar (80).

It was found that times shorter than 2 hours did not anneal

out the damage completely. Longer times degraded the sample

surface, particularly in an N2 ambient, which resulted in

severe pitting (59). It was also observed that the wafers

annealed in an Ar ambient showed a lower defect density than

those annealed in an N2 ambient. These effects were

interpreted in terms of (a) a possible reaction between the

N2 and the top silicon layer at 11500C, which formed a thin

silicon nitride layer on the sample surface, and (b)

variations in the concentration of point defects at the near

surface layers caused by the ambient (81).

In order to avoid surface damage, it has become common

to cap the SIMOX wafers with an SiO2 layer before the high

temperature anneal. The chemical state of the surface,

whether it is bare or covered by an oxide cap, can, however,

affect the concentration of point defects (82). The surface

can play a major role as a sink for point defects. On the

other hand, it can act as a source of additional silicon

interstitials (82, 83).

Growth and Dissolution of SiO2 Precipitates. The

density of SiO2 precipitates formed during implantation is

remarkably reduced by high temperature annealing. The

classical theory of nucleation and growth of a second phase

in a matrix, and in particular the concept of the particle

of critical radius has been used to explain the evolution

and stability of the SiO2 precipitates during annealing of

the SIMOX structures (70). Growth and coalescence of SiO2

precipitates take place during annealing. At high enough

temperatures, the precipitates dissolve. A thorough

description of the dissolution of SiO2 precipitates from a

thermodynamical point of view has been presented by Jassaud

et al. (84).

It has been shown that temperatures above 12000C are

necessary to dissolve the SiO2 precipitates and obtain sharp

Si/SiO2 interfaces (73). Complete dissolution of the oxide

precipitates in 30 minutes has also been achieved by a newly

developed technique (76). This method consists in heating

the SIMOX wafer up to a temperature of 14050C for 30 minutes

by melting the backside of the wafer in a special lamp

furnace. An alternative method to remove the SiO2

precipitates from the top silicon layer has been proposed by

Mizuno et al. (85). It consists of an additional

implantation with H2 after the oxygen implant, followed by a

11500C anneal. The authors above suggested that this method

has the advantage of avoiding possible furnace contamination

at higher temperatures.

Dislocation Density. Most of the material studies in

SIMOX have focused on the effects of implantation

temperature and annealing conditions for a given dose and

energy. As already mentioned, the morphology of the top

silicon layer is strongly determined during the implantation

process. Although complete removal of oxide precipitates

from the top silicon layer has been achieved with high

temperature annealing, the question of dislocations still

remains. This fact is of major concern. It is well

established that dislocations with energy levels close to

the middle of the band gap provide efficient generation-

recombination (GR) centers and introduce surface states.

This results in a reduced minority carrier lifetime.

Furthermore, dislocations are known to provide effective

trap centers for metallic impurities. Calculations of the

interaction energy between atoms of metal impurities such as

Fe, Ni, Cu and dislocations have shown that impurity

atmospheres or precipitates tend to form at dislocations

preferentially (86).

There has been some controversy regarding the effect of

higher annealing temperatures on the dislocation density of

SIMOX. In fact, some groups claim it remains constant (87)

while others have observed a significant decrease (75).

Table 2.1 shows dislocation densities ranging from 108 to

105/cm2, even as low as 103/cm2 that have been reported for

different implant and annealing conditions (73, 75-79, 87-

92). The improvement in dislocation density seems to be

closely related to the implantation conditions, which

determine the initial damage morphology. The most

remarkable implantation conditions recently reported in the

literature involve the following.

(a) Implantation at low dose rates (1.5 MA/cm2) into a

channeling direction (77). This leads to precipitate

ordering along the <100> directions of the silicon lattice

in which strain is mostly easily accomodated. The <100>

direction has the minimum elastic constant of the silicon

matrix. A simple cubic network of amorphous spherical

precipitates with a 2 nm diameter and a periodicity of 5 nm

prevents the dislocations from climbing to the surface.

(b) Sequential low dose implantation and annealing with

oxygen doses less than a critical value of 4xl017/cm2 at

150 keV between annealing cycles (89). Dislocation

densities of less than 103/cm2 were obtained with this

Table 2.1. Dislocation densities reported in the literature.


Dose E

(xl018cm-2) (keV)




Temp. Time

(0C) (hr)



600 1300

700 1150
i 1300

600 1350

600 1150
I 1200
of 1250
i 1295

500 1150
i 1210

500 1250

500 1405

500 1150
of 1250

580 1405

550 -
if 1300

475 1250

600 1300

540 1300




















1. 1X106*








* Dislocation densities estimated from chemical etching may
not be directly comparable with those detected by TEM.










































method. This sharp decrease in dislocation density was

explained by the migration of excess silicon interstitials

towards the silicon free surface where they can be

incorporated in the lattice (90). For a review on these

methods refer to (93).

2.2.4. Effect of SIMOX Processing on Device Characteristics

Extensive electrical measurements have been performed

on SIMOX devices. Good uniformity in individual device

parameters across the same wafer as well as from wafer to

wafer is critical for VLSI applications (94, 95). It has

been shown that device characteristics are strongly related

to the changes in implantation and annealing conditions,

which determine the microstructure of the SIMOX top silicon

layer (96-101). Some of the most important device

parameters that have been monitored are threshold voltage,

surface carrier mobility and transistor leakage. It was

found that surface carrier mobilities and leakage currents

of SIMOX devices identical to those of bulk devices could be

obtained by proper control of oxygen ion dose and transition

metal contamination during implantation, and annealing

temperature (99, 101). Improvement of the contamination

levels in SIMOX wafers has affected device performance

considerably. For example, it has been reported that

prevention of metal contamination during oxygen implantation

results in transistor leakage current values less than

0.1 pA/Am channel width with a 5 V power supply (102).

2.3. Transition Metal Impurities in SIMOX

It is well known that device yield, performance and

reliability of silicon devices are adversely affected by the

presence of transition metal impurities in particular the 3d

transition metals such as Cr, Mn, Fe, Co, Ni and Cu.

Furthermore, the interaction of impurities with physical

defects may result in defect decoration and precipitate

formation, which altogether produce electronic states within

the silicon bandgap. These sites are very effective carrier

traps or GR centers, which reduce the minority carrier

lifetime and increase the leakage current in pn junctions.

All device types are susceptible to some degradation of

properties by impurity introduction. For example, it was

reported that the generation lifetime, surface generation

velocity and dielectric breakdown strength of SiO2 in a MOS

device are strongly affected by the presence of Fe at the

surface if the Fe concentration exceeds 1x1012, 5x1012 and

ix1013 atoms/cm2, respectively (103). Prediction of the

behavior of any metallic element during processing is,

therefore, extremely valuable in assessing the electrical

properties of future SIMOX devices.

The physical properties of transition metals in silicon

have been thoroughly investigated and a detailed review of

their solubility, diffusion and electrical activity has been

given by Weber (104). The 3d transition metal group is

characterized by a dominant interstitial diffusion and

solution at high temperature in thermal equilibrium. The

diffusivity decreases from Cu to Ti due to the larger atomic

radius of atoms with smaller positive charge of the nucleus.

The largest diffusion coefficient in silicon has been

observed for Cu. The solubility varies in a similar manner.

The heavier 3d metals have larger solubility. The

solubility of 3d metals strongly decreases with temperature

to values of 1 atom/cm3 at room temperature (RT). Heavier

transition metals like the 4d and 5d metals are slower

diffusers and have a dominant substitutional diffusion

Lattice defects created by ion implantation can have a

significant effect on the redistribution and electrical

activity of co-implanted impurities after annealing. A

rapid decrease in solid solubility with decreasing

temperature and a high diffusivity within the same

temperature interval lead to heterogeneous precipitation of

metal impurities. The presence of defects such as oxide

precipitates and dislocations can provide efficient

gettering or trapping sites for transition metal impurities.

Gettering is a general term given to the process by which

harmful impurities are removed from the regions in the wafer

where devices are fabricated, and transported to sinks that

can absorb impurities (68). It has been shown that metals

tend to decorate preferentially some kinds of microstructure

(105). This effect is related to the decrease in the system

free-energy that results from trapping metallic impurities

to high energy sites. Dislocations represent favorable

trapping sites due to the formation of Cottrell atmospheres

(106). Oxide precipitates induce stress fields which create

silicon self-interstitials that attract transition metal

impurities (106).

Metals such as V, Cr, Mn and Fe can be quenched into

interstitial sites of tetrahedral symmetry whereas Cu, Ni

and Co precipitate readily even with fast quenching because

of their rapid decrease in solid solubility with decreasing

temperature and higher diffusivity. Microscopy studies by

TEM have revealed that most transition metals in silicon

tend to form silicon rich silicide precipitates of the type

MSi2 (104).

Those 3d metals that can be metastable as interstitials

form deep energy levels. Metal precipitates do not

generally produce deep energy levels as they are not

sufficiently coupled to the electron band structure.

However, they can be electrically conductive and decrease

carrier mobility by scattering. It has been observed that

transition metal precipitates represent a dielectric

constant discontinuity which reduces the breakdown strength

of the thin gate oxide in MOS devices (107, 108). Iron in

particular is very harmful because it tends to form long

needle-shaped precipitates that shorten the junctions.

Contamination levels in silicon resulting from ion

implantation have been reported to be in excess of 1-0.02%

of the implanted dose (109). Very few studies have been

performed on the behavior of transition metals co-implanted

in SIMOX. So far, only Cr and Cu have been investigated

(110, 111). It was observed that Cr, which was

intentionally implanted into the top silicon layer of a

SIMOX structure, tends to migrate towards the surface

SiO2/top Si interface during annealing (110). Only a very

small fraction of Cr is gettered by the damage area near the

buried oxide interfaces. On the other hand, Cu (also

intentionally implanted into the top silicon layer of SIMOX)

diffuses from the top silicon through the buried oxide layer

during annealing and tends to segregate into the substrate

silicon (110). Large colonies of Cu precipitates associated

with dislocations at the lower Si02/Si substrate interface

have been observed in SIMOX co-implanted with Cu (111).

This work will focus on Fe, Cr and Mo metal impurities.

Both Fe and Cr result from sputtering of the surface of

stainless steel implanters. They are considered to be

fast diffusers in silicon with a diffusivity in the

10-5_10-6 cm2/sec range (104). Impurity levels as low as

0.02 ppm have been reported to induce structural changes in

silicon such as precipitates and dislocation colonies (106).

Contamination with Mo results from sputtering of the

apertures. Molybdenum has not been extensively studied as a

transition metal impurity in silicon. However, it has been

observed that concentrations as low as I012-I013/cm3 can

strongly reduce minority carrier diffusion lengths and

lifetimes (112). Molybdenum is particularly detrimental for

bipolar devices because of the extremely low levels (ppt) at

which it affects recombination lifetime and the difficulty

in being gettered. On the other hand, Mo seems to be less

effective as a generation lifetime killer, which makes it

less detrimental for MOS devices. This results from having

an energy level EV+0.31 eV far from the midgap (112). So

far, Mo has been considered to be a slow diffuser in silicon

with a diffusion coefficient of the order of i0-1l cm2/sec

for temperatures less than 10000C (113). But it was

reported recently that the diffusion coefficient of Mo in

silicon at 12000C has a lower limit of 10-8 cm2/sec (112).

This value seems to be more consistent with the high

diffusivities shown by other transition metals (104).

The future of VLSI and ultra large scale integrated

(ULSI) circuits requires a very stringent control of

metallic impurities to levels less than 0.001 ppba (113).

The success of the SIMOX technology depends on being able to

reproduce defect and contamination-free structures on a

large scale.


3.1. Introduction

It is well known that the main purpose of semiconductor

materials characterization is to improve device performance

and manufacturing yield. These parameters depend on the

material properties. Establishing an effective correlation

between the starting material characteristics and the

electrical parameters of interest is thus essential to

ensure high performance and yield. As already described in

Chapter 1, this is one of the problems encountered in SIMOX

technology, as transition metal impurities are co-implanted

with oxygen. Characterization and control of metal

impurities is, therefore, essential. Characterization of

SIMOX from a materials and electrical point of view was

undertaken through a sequential study of post-implantation

annealing treatments. In this Chapter, the details of the

experimental process, sample preparation, measurement and

analytical techniques used in the characterization of the

different SIMOX specimens for this study will be presented.

3.2. Sample Preparation

The starting wafers for this investigation were three

inch, n-type, phosphorous doped, (100) oriented silicon

substrates with a resistivity in the range of 3-5 ohm-cm.

A capping layer of 250 A thermal Sio2 was grown at 9500C

before the implant. The wafers were implanted with

2.25x1018 0+ ions/cm2 at an implantation energy of 150 keV

and at a temperature of 510150C using a 10 mA class oxygen

implanter (the implantation was done at Texas Instruments

Inc.). The projected range and standard deviation for

oxygen into silicon obtained from the Lindhard, Scharff and

Schiott (LSS) calculation was 0.3870.104 Am. Some of the

wafers were unintentionally implanted with transition metal

impurities (Fe, Cr, and Mo) during the oxygen implant. The

SiO2 cap was removed after implantation with an HF etch.

Test samples were cut from the implanted wafers and were

cleaned prior to heat-treatment by boiling consecutively in

trichloroethane (TCE), acetone and methanol, rinsed in

deionized (DI) water, then dried with nitrogen. Post-

implantation annealing was performed in the temperature

range of 900-13000C for 2 and 5 hours in flowing Ar+l.5%02.

The partial oxidizing ambient was used to preserve the

surface integrity of the samples (115). Temperatures were

measured with a thermocouple inside the furnace tube.

The samples were inserted and removed relatively slowly

(-15 min.). The annealing times quoted above are the times

the samples spent in the central zone of the furnace. Some

samples were heat-treated in Ar for 2 hours in the

temperature range of 1150-13000C. Some of the wafers were

annealed in flowing N2+1.5%02 for 2 hours in the temperature

range of 1150-1250C followed by epitaxial growth of a

0.35-0.40 gm thick silicon layer at 11500C (done at Texas

Instruments Inc.). Implantation and annealing conditions

used in this work are summarized in Table 3.1.

3.3. Materials Characterization

The materials characterization procedure will be

described next. Both, as-implanted an heat-treated samples

were analyzed by secondary ion mass spectroscopy (SIMS),

cross-sectional and plan-view transmission electron

microscopy (XTEM and PTEM, respectively), and Auger electron

spectroscopy (AES). It was of particular interest to

examine the distribution of transition metal impurities

after the heat-treatment sequence necessary to anneal out

the radiation damage. The objective of this

characterization is to provide a correlation of the

compositional distribution with the microstructural

evolution during annealing. The effect of transition metal

impurities on the electronic band structure of SIMOX was

also investigated by differential reflectometry (DR). A

brief description of each one of the techniques used in this

study will be given as well.

Table 3.1. Implantation and annealing conditions.





2.25x1018 O+/cm2 1175-1300

150 keV 1150-1300

510150C 1150,1250



Ambient Furnace Tube

Ar+l.5%02 Quartz

Ar+l.5%02 SiC





* Annealing was done at Texas Instruments Inc.


3.3.1. Ion Microscopy and Secondary Ion Mass Spectroscopy

The continuous decrease of lateral and depth dimensions

in semiconductor devices requires accurate analysis of

impurity concentrations with increasing depth resolution.

SIMS is the analytical technique of choice to characterize

impurity depth distribution in semiconductors due to the

high sensitivity and good depth resolution achievable.

Elemental sensitivities vary by roughly 5 orders of

magnitude across the periodic table (Figure 3.1) but

detection limits are in the ppm and ppb range. SIMS also

has the capability of detecting all elements including

hydrogen and to distinguish isotopes.

SIMS profiles of the metal impurity distributions were

obtained with a Cameca IMS 3F ion microprobe located at

North Carolina State University. A schematic diagram of

this system is shown in Figure 3.2.

An ion source, either of cold cathode duoplasmatron

type for an oxygen primary ion beam or a thermal ionization

type for a cesium primary beam, produces a primary ion beam.

In order to maximize sensitivity, ion beams are chosen such

that they produce the highest ion yields of the species

under study. Oxygen is used to analyze electropositive

elements whereas cesium is generally used to study

electronegative species. The primary ion beam is

accelerated through a three-lens column (primary beam

column), with energies up to 20 keV and finely focused onto


(a)I I I I
(a Cr FRu in I a

6mg No
10Ga R
+. Al Nn
s e Zr

B F. Nb
F 11 H f R e

~-C pd Ta O
CDJ N1 Hfh

102 IZn I Pb
C-17 Sb B

LI, 510s
-j F

N Au
1 0 20 30 40 50 60 70 80 90 100

(b) N0a

c S1


2 AU

en Go T

I A Nb
C n A g S b T

F Si
S10~ CA GaT

Mn Z n ,Cd B Hf I Hg
102 138.0 LL -jJJ 8.0.
0 10 20 30 40 50 60 70 80 90

Figure 3.1. SIMS elemental sensitivities.
(a) Positive secondary ion yields
(b) Negative secondary ion yields.


Analyzer ,


Immersion Lens :;am s;e.,






U- LLt






aday Cup

A Entrance Slit
B Energy Slit
C Exit Slit

Figure 3.2. Schematic diagram of the
Cameca IMS 3F Ion Microprobe.


a sample held at 4500 V with respect to ground. Primary

ions undergo nuclear collisions with target atoms as they

penetrate the solid surface. Kinetic energy is transferred

to the target thus triggering a collision cascade. During

the collision cascade, atoms or cluster of atoms normal to

the sample surface and with enough kinetic energy are

sputtered from the solid into the vacuum. Some fraction of

these ejected species are neutral and the remaining are

positive or negative ions. Electrostatic optics are

employed to extract and transport the secondary ions from

the sample such that their spatial resolution is maintained.

The electrostatic optics consists of a collection stage

(immersion lens) and a focusing part called transfer optics

which optimizes the sensitivity of the mass spectrometer as

a function of the dimensions of the surface analyzed for a

given mass resolution. The mass of interest is selected by

the mass spectrometer which maintains the spatial integrity

of the secondary ions during energy and mass filtering. The

mass spectrometer is of the double focusing type with an

electrostatic analyzer and a magnetic prism. The

electrostatic analyzer disperses the secondary ions as a

function of their kinetic energy. An energy slit is used to

select the chosen energy and energy resolution. The

magnetic prism is computer controlled and provides mass

dispersion of the secondary ion beam according to their

mass-to-charge ratio. A projector lens is used to focus the

mass and energy resolved ion image onto a dual microchannel

plate image converter, and a magnified equivalent electron

image can be displayed on a fluorescent screen for

observation through a binocular microscope and recording by

photography. Ion intensity measurements can be made on the

mass separated ion beam by using an electrostatic deflector

to deflect the beam onto a faraday cup or an electron

multiplier. The entire system is operated by Hewlett-

Packard 9845B desk top computer. As of 1988, the system was

controlled by an IBM PC desk top computer.

Three major vacuum systems are used in this instrument.

A turbomolecular pump with pressures reaching the low

10-8 Torr range is used to evacuate the primary beam column.

The sample chamber is equipped with a cryopump with an

ultimate pressure of 2x10-9 Torr. The sample introduction

chamber can be rapidly pumped with a turbomolecular pump.

The analyzer section is isolated from the sample chamber and

guns area by a vacuum valve and can be pumped with two ion

pumps to a typical pressure in the low 10-8 Torr range.

Primary beam energies used in this work were in the

range 9-10 keV. The focused primary beam was rastered over

an area 250x250 gm2, with secondary ion extraction from a

60 Am diameter area in the center of the crater. The

average sputtering rate was approximately 5 A/sec.. This

value was obtained by measuring the depth of the sputtered

crater and assuming that the flux of primary ions to the

sample was constant throughout the analysis. The profiles

were measured with the 02+ primary beam. The samples were

covered with gold prior to the analysis to avoid charging

effects. The thickness of the Au deposit was on the order

of 200 A. Analysis of as-implanted samples was performed

both in the mass spectrum mode and in the depth profile

mode. Annealed samples were analyzed in the depth profile

mode. Positive secondary ions were detected at mass number

52 for Cr, mass number 56 for Fe and mass number 98 for Mo.

The samples were analyzed with the high voltage offset

technique to compensate for possible shifting of the sample

potential during sputtering of the insulating buried oxide

layer. SIMS measurements were performed under high mass

resolution M/AM=3000, by using an electron multiplier

detector to avoid mass spectral interference with 56Fe+,

namely 28Si+28Si+. The stability of the primary ion beam

and the depth to which surface effects disturb the profiles

can be assessed by monitoring a reference mass. Oxygen 16

was the reference mass for the metal impurity profiles.

Due to variations in secondary ion yields and changes

in sputtering rates in different layers and across

interfaces, SIMS measurements of concentrations are

difficult. Quantification of the profiles requires the use

of standards of the impurity in the matrix layer. Standards

for this study consisted of Fe, Cr and Mo implanted

separately in Si and SiO2 matrices. These standards were

prepared by Surface Alloys using the implant conditions

shown in Table 3.2. Since it is also difficult to obtain

precisely the same operating conditions for the samples and

Table 3.2. Implant conditions used in the preparation of Fe,
Cr and Mo standards in Si and SiO2.

Impurity Matrix Dose


Fe 4.80x1015

Cr Si 5.00x1015

Mo 3.50x1015


SiO2 3.70x1013


Energy Peak Concentration









1. 00x1021

1. 00x1021

1. 00x1021

1. 06x1019


1. 01x1019









standards, it is necessary to reference the standard

profiles to the same reference mass as the test sample to

compensate for possible differences in overall


Conversion of SIMS relative intensities to

concentration profiles involves the determination of

relative sensitivity factors. A data processing program was

used to convert intensity-time profiles (raw data) to

concentration-depth profiles (processed data). Quantitative

results were obtained by taking the ratio of the metal

impurity intensities to a matrix intensity (0 in this case)

in the SIMS profiles of both the standards and SIMOX

samples. This corrects for variations in the beam intensity

due to different ion yields and sputter rates. Relative

impurity sensitivity factors (SFe, SCr, SMo) were then

obtained from each of the single matrix standards (Si and

Si02) by integrating the ratioed implant profile over the

total area under the SIMS depth profile, and equating it to

the total implanted dose (116). The concentration of the

impurity in the matrix was then determined at any point by

applying the respective sensitivity factor to the ratioed

intensity. Both the relative sensitivity factors and

sputter rates were then applied to each one of the different

layers of the SIMOX samples. Quantification in the

interface region was obtained by interpolating the

concentration factors and sputter rates from the top Si

layer, buried oxide and substrate across each interface.

No correction was made for the surface oxide covering the

SIMOX samples annealed at temperatures lower than 12000C.

Studies performed independently on the sputter rate of Si

and SiO2 showed a ratio of 1.07 with a sligthly higher ratio

in the oxide (117). As mentioned above, the depth scale was

determined from the crater depth measured with a Tencor

Alpha Step 200 stylus profilometer and assuming a uniform

sputter rate with time.

3.3.2. Transmission Electron Microscopy

TEM is one of the best techniques for structural

characterization of materials due to the very high spatial

resolution achievable and its ability to provide both

compositional and crystallographic information from the same

area of the sample. Typical spatial resolutions are within

the 2-10 A range. The incident electron beam is accelerated

typically to 200 keV and electromagnetically focused on the

specimen. A TEM micrograph can be recorded either in the

bright field (BF) or dark field (DF) mode. A schematic

illustration of the electron path during bright field and

dark field imaging is shown in Figure 3.3. In the BF mode

(Figure 3.3a) the image is formed by intercepting the

scattered (diffracted) electron beam with a small aperture

(20-120 pm) inserted in the back focal plane of the

objective lens and allowing the transmitted beam to pass

through. A DF image (Figure 3.3b) is formed by

intercepting the transmitted electron beam and allowing the

Optic Axis

I Primary Beam

Crystal Planes

Objective Aperture

Primary Beam

Transmitted Beam

Diffracted Beam

Figure 3.3. Schematic diagram of TEM
(a) Bright-Field imaging
(b) Centered Dark-Field imaging
(c) Weak Beam Dark-Field imaging.



diffracted beam to pass through. The image contrast depends

on the intensities of the electron beam emerging from the

sample. The main advantage of DF is high contrast. The

image resolution is comparable with that of the BF. The

background of a DF image is dark with bright areas

corresponding to strong scattering. Sharper contrast for

structural defects such as dislocations can be obtained by

using the two-beam diffraction or Bragg condition (the

transmitted beam and one strong diffracted beam) (118). If

the specimen is tilted away from the Bragg diffraction

condition (weak-beam method) (Figure 3.3c) the beam

diffracted by the plane near the core of the dislocations is

allowed to pass through the objective aperture thus

increasing the sharpness of the DF images (118).

One of the essential requirements of any TEM study is

the preparation of a good specimen. The physical structure

of the microscope requires that a 3 mm diameter sample with

a thickness of the order of 1 gm be used. The very thin

layers are very delicate and must be carefully handled.

Several samples must be prepared in order to obtain reliable

results. However, careful application of a good technique

and practice can provide good results. Samples for XTEM

were prepared by mechanical polishing and subsequent ion

milling to produce thin regions approximately 1 gm in

thickness. A schematic illustration of the technique is

shown in Figure 3.4. The wafers were first diced into

rectangular slabs. Epoxy was applied to each slab. The

1. Stack Construction

silicon slabsiK

test samples [

silicon slabs

'2. Slicing (Diamond Saw)

500 w

3. Disc Cutting (Ultrasonic Disc Cutter)

Figure 3.4. Schematic illustration of
XTEM sample preparation.

4. Grinding


5. Dimpling



-1 100 Am

T, i0 wmT

6. Ion Milling

T. o. 2 pm

Figure 3.4--continued.

5- 500 gm

slabs were pressed together into a stack, and then baked for

1 hour at 1900C in a oven to cure the epoxy. The stacked

slabs were then sliced into 8 cross-sections using a diamond

saw. Each cross-section of the stack was cut into a 3 mm

disc with an ultrasonic disc cutter. After grinding and

dimpling the disc samples down to an approximate thickness

of 20-15 gm, the disc samples were ion milled with a 5 keV,

3 mA focused Ar+ beam. The samples were continuously

rotated during the milling process to minimize surface

roughness. The ion miller is equipped with a turbomolecular

pump. Typical pressures during ion milling were in the

10-5 Torr range. For a more detailed description of this

preparation technique refer to the Appendix. Plan-view

specimens were prepared by grinding, dimpling, and ion

milling 3 mm discs of the material to be studied from the

unimplanted side of the samples. TEM examinations were

performed using a JEOL 200 CX transmission electron

microscope. Most of the TEM micrographs were taken under

the two-beam condition method both in the BF and DF mode.

The specimens were tilted to the two-beam condition for a

{220) plane reflection. The cross-section micrographs

obtained correspond to the (110) plane perpendicular to the

original (100) sample surface plane.

Some of the heat-treated SIMOX samples were also

analyzed by high resolution TEM (HRTEM) at Arizona State

University. This technique provides high resolution images

(1.8 A) with spacing approximately that of the reflecting

planes. These images are obtained by recombining one, two

or several diffracted beams with the transmitted beam.

Phase interference between these beams results in a periodic

fringe image. The spacing of the fringe image corresponds

to that of the reflecting planes producing the diffracted


3.3.3. Auger Electron Spectroscopy

The use of AES for compositional analysis relies on the

emission of characteristic electrons from a sample surface

by the Auger radiation process (Figure 3.5). Vacancies

created in the inner electron shells (e.g., a K-shell) of

excited atoms can be filled by electrons from a lower energy

core level (e.g., LI). The energy emitted during this

electronic transition may in turn be transmitted to another

electron of specific energy, the Auger electron, at a level

such as LIIIII. The Auger electrons predominantly have an

energy within the range of 20 to 2000 eV. The primary

excitation source in AES is usually an electron beam.

The SIMOX specimens were analyzed with a Physical

Electronics, model 660 Scanning Auger Microprobe. The

electron beam incident angle was 300 referred to the sample

normal. The primary electron beam energy was 10 KeV. An

electron beam current of 0.05 gA was used. Auger spectra

are displayed as the number of Auger electrons with a

specific energy distribution, N(E) vs E. But due to the

large background signal that arises from electrons emitted

0 Auger Electron


Valence Band

- 0_ L1

2s 0

K electron

is 0

K. x-ray

L electron

Equilibrium Ionization

Auger Emission

Figure 3.5. Schematic diagram of the
Auger radiation process.


2p {

p p

from deeper within the sample, which have lost some of their

energy during escape, enhancement of the Auger signal is

usually obtained by taking the first derivative dN(E)/dE of

the spectra. Therefore, the Si (LVV) (92-107 eV), 0 (KLL)

(475-510 eV), and Si (KLL) (1561-1619 eV) were recorded in

the first derivative mode dN(E)/dE. Auger depth profiles

were obtained by alternating between sputtering with Ar ions

and recording the peak-to-peak height intensities of the

corresponding Auger transitions. A 3 KeY Ar+ gun with

a current density of 25 mA/mm2 and a raster size of

2.5x2.5 mm2 was used to sputter the sample surface. The

base pressure in the chamber was 2x10-10 Torr. During

sputtering the pressure was 2x10-8 Torr.

After a crater was obtained following depth profiling

down to the silicon substrate, a multiplex analysis was

performed to examine the changes in energy for Si (LVV),

0 (KLL) and Si (KLL) through the SIMOX layers.

3.3.4. Differential Reflectometry

It has been shown that lattice defects induced by ion

implantation into a silicon substrate can affect the

electronic structure of the material thus causing noticeable

changes in the interband transition energies (119). It is

based on this changes that an investigation was undertaken

by means of differential reflectometry to provide further

insight into the nature of transition metal impurity

effects. This technique consists of monitoring the

normalized difference in reflectivity AR/i (where AR is the

difference in reflectivity and T is the average reflectivity

over the scanned area) between a sample and a reference

sample as a function of wavelength (X), typically from

2000-8000 A. The energies which the electrons absorb from

photons as they are excited into higher allowed energy

states can be directly measured from a series of peaks

displayed in a differential reflectogram. A differential

reflectogram is a plot of AR/K = f(X) obtained from the

ultraviolet through visible into the infrared region. Any

shift of the characteristic energies for electronic

interband transitions of a material can be detected within a

one hundredth of an eV.

A detailed description of the differential

reflectometry technique has been given elsewhere (120). A

schematic diagram of the apparatus is shown in Figure 3.6.

Two adjoining samples, one of which is the reference sample,

are clamped to a mounting stage. A specific area in the

samples is selected by moving the stage both vertically and

horizontally. The samples are centered relatively to the

incident beam spot. Unpolarized light passes through a

monochromator and is focused onto an oscillating mirror

which deflects the light alternately between the two

samples. The total area scanned is 2mmx4mm. The light

reflected from the samples is measured by a photomultiplier

tube (PMT) and electronically processed to yield the

normalized difference in reflectivity (AR/g), which is fed


ator Photomulliplier output
Tub Signal




Figure 3.6. Schematic diagram of the
differential reflectometer.

Light Source

to an x-y recorder and plotted as a function of X.

A typical scan takes about 3 minutes. This technique

does not require a vacuum.

Two reference samples were used for this study, SIMOX

as-implanted without metal impurities and unimplanted

silicon. The systems analyzed were as follows.

(1) SIMOX as-implanted without metal impurities vs

SIMOX as-implanted with metal impurities, and

(2) Unimplanted silicon vs SIMOX as-implanted without

metal impurities and annealed.

3.4. Electrical Characterization

The characterization of the electrical behavior of

SIMOX is important in order to determine its fundamental

electrical properties. Electrical properties of concern are

resistivity, carrier mobility and minority carrier lifetime.

Various experimental methods have been established for

measurement of these parameters (113). The electrical

methods attempted for this study were the Hall effect

measurement and the spreading resistance profiling technique

(SRP). During the preparation of samples for Hall effect

measurements, serious difficulties were encountered in

making good ohmic contacts to the top silicon layer of the

SIMOX samples. This limited the electrical characterization

of SIMOX to the use of SRP measurements.

The SRP technique can provide resistivity and dopant

concentration profiles with a spatial resolution of

1 nm/point. Measurements can be obtained for dopant

concentrations varying from greater than 1021 cm-3 to near

intrinsic material (less than i0lo cm-3) in device and

process control research, development and production

(121, 122). The SRP technique is based on a low voltage

measurement of the contact resistance of 2 point probes

which are stepped across the surface of a sample. The

sample is specially prepared by bevelling and polishing with

a fine diamond abrasive in an oil-based slurry on a frosted

glass plate. Figure 3.7 shows a schematic illustration of

this technique. The contact probes are lowered onto the

sample surface at each point. The resistance between the

two probes (called spreading resistance) is measured and

plotted. The theoretical spreading resistance is given by

R =p/4a where R is the measured spreading resistance (ohms),

a is the radius of the flat circular contact (cm) and p is

the local resistivity (ohms-cm). In practice, however, the

measured spreading resistance values also depend on the

microstructure of the probe tips, sample conductivity type,

crystal orientation and surface finish. Therefore, probes

must be conditioned by reference to calibration curves

generated for a particular set of probes on uniform and

known resistivity samples having the same material

parameters as the test specimens. The SRP data are loaded

directly into a computer containing the calibration data and

are computer-processed to yield resistivity and dopant

concentration profiles. In addition, a multilayer

Contact Probes (W-Os)

Top SiO2

Top Si

B uried Oxide SiO2

Si Substrate

Bevel Surface

Bevel Edge

Zero Depth Position in SRP Profile

Figure 3.7. Schematic illustration of
SRP measurements.

correction factor is calculated and applied to each raw data

point to account for boundary effects (e.g. pn junction,

insulating layer) in the vicinity of the probes.

Spreading resistance measurements of the SIMOX samples

as a function of annealing temperature were obtained with a

two-point probe ASR-100 C/2. As-implanted and annealed

specimens for SRP measurements were prepared by first

cleaning in TCE, acetone, methanol, DI water rinse, N2 blow-

dried, buffered oxide (BOE) etch (6-1) dip, DI water rinse

and N2 blow-dried. A 2000-3000 A thick SiO2 layer was then

deposited on the sample surface in order to prevent the

rounded bevel edge effect caused by very thin layers. Next,

each sample was waxed to a bevelling jig (00 17') and

polished with a 0.1 gm diamond paste on a rotating glass

plate. The bevel angle was measured with an optical

microscope. The probe was stepped across the surface and

down the bevel. The distance between the probes was 60 pm.

The point to point increment in the horizontal direction was

2.5 gm. This leads to a point to point increment of 125 A

in the vertical direction.

It should be reemphasized that the SRP measurements

compare the experimental results to certain standards

(provided by the National Bureau of Standards). Since

spreading resistance is a comparison method, its accuracy

may be limited by the accuracy of the calibration material

available and the data correction procedure used to convert

spreading resistance into resistivity and carrier


concentration. Because the SIMOX layers are very thin,

there may be difficulties in determining exactly the

starting point for the profiles. In addition, uncontrolled

penetration of the probe tips into the sample surface may

smear the profiles.


4.1. Introduction

This Chapter presents and discusses the experimental

results obtained on SIMOX structures, which were co-

implanted with transition metal impurities (Fe, Cr, and Mo)

and annealed within the temperature range of 900-13000C for

2 and 5 hours. The results obtained for non-contaminated

SIMOX structures, which were implanted and annealed under

the same conditions will also be given. The results were

derived from SIMS, TEM (cross-sectional and plan-view), AES,

DR and SRP measurements. As mentioned before, the purpose

of this work was to investigate the effects of annealing

treatments on the redistribution of transition metal

impurities in SIMOX structures, and to present a fundamental

understanding of their behaviour prior to device processing

through a sequential study of the compositional evolution,

structural changes, optical data and resistivity changes.

4.2. SIMOX Co-Implanted with Transition Metal Impurities

4.2.1. Analysis of Metal Impurity Distributions

Figure 4.1 shows a SIMS mass survey performed on the

surface of as-implanted SIMOX samples. A survey performed

on non-contaminated float zone silicon is also included for

comparison. Analysis of the mass spectra is made difficult

by the presence of several molecules, which have the same

mass number as other elements (e.g., 28Si28Si and 56Fe).

The presence of transition metal impurities such as Fe, Cr,

Cu and Mo can be clearly detected in the samples, although

the intensity of the Cu signal is low. As mentioned

previously, these elements were carried along the beam line

and implanted into the silicon wafer during the oxygen

implant. The presence of several alkaline metals (Na, Al, K

and Ca) is also observed. These elements result from

handling. They were not studied in this work.

The distribution depth profiles of Fe, Cr and Mo were

monitored for as-implanted and annealed SIMOX samples, and

will be presented next. Figure 4.2 depicts the Fe, Cr and

Mo distribution depth profiles in as-implanted SIMOX

samples. Two major peaks can be observed for Fe and Cr in

the top silicon layer, one near the surface (primary peak)

and another at the upper Si/SiO2 interface (secondary peak).

Only a primary peak is observed for Mo. Iron has by far the

highest peak intensities, that is the highest concentration

of the three metal impurities. Table 4.1 summarizes the




104 -

103 k-

I ,I I ,


31 p


o 104 52cr
U) 103-



100 1



28Si 29Si
I 129Si30Si

63CU 65CU

, ,I I I

I ~

70 80

Figure 4.1.

SIMS mass survey of
(a) As-implanted SIMOX, and
(b) Float zone Si.












98 Mo



I I . . . . . . I . . I . . . I I I I I I I




10 5

j 10
(~n 14

14 OH
0 2 12 N 2
010 C C

10 1

o0 0.11LL
10 20


105 28Si

W O28Si
1-J 10 29io 28S2i
Fi- 103
z 9Si Z8Si 29S]

0 2 40Ca 29Si 30



39, 49 59 69 79

FiQure 4.1--continued.

S 1 9 o19 19O~


I~s I I 10ds 61 9' ~ S ,

005 108 1 085


051 .5 1 0 .51

DEPTH (microns)

Figure 4.2. SIMS depth distribution of Fe, Cr and Mo in as-implanted SIMOX.
The detection limit is given by a dashed line.

Table 4.1. Concentration and depth of Fe, Cr
and Mo peaks in as-implanted SIMOX.

Primary Peak

Concentration Depth

(cm-3 ) (Am)


4. 00x1019

5. 00x1018




Secondary Peak

Concentration Depth

(cm-3 ) (Am)


3. 00x1017



concentration and depth for the primary and secondary peaks

of Fe, Cr and Mo in as-implanted SIMOX. The primary peaks

correspond to the "knock-on" peak obtained during co-

implantation of the metal impurities. Calculations done

with the TRIMM 88 computer simulation program have shown

that an energy of 30, 33 and 65 keY, respectively, is

sufficient to position the primary peaks of Fe, Cr and Mo in

silicon at the depths observed in the SIMS profiles. These

peaks follow a gaussian distribution (45). But dynamic

annealing during the oxygen implantation process at a

temperature of 5100C may result in redistribution of the

metals in the as-implanted structure. Both, the diffusivity

of the metals and the implantation induced damage can affect

the impurity concentration profile. At the implantation

temperature of 5100C, the diffusion coefficients of Fe and

Cr in silicon are 1.6x10-8 and 3.7x10-9 cm2/sec,

respectively (104). This means that the motion of the

impurity atoms is not limited by diffusion since the

diffusion length is greater than the thickness of the wafer.

Therefore, gettering of the metals by the damage area near

the buried oxide would explain the formation of the Fe and

Cr secondary peaks shown in the SIMS experimental profiles.

On the other hand, the absence of a secondary peak for Mo in

the SIMS profile may be explained in terms of no significant

gettering by the damage area. This result would be expected

since Mo has a much lower diffusion coefficient in silicon

(less than 10-11 cm2/sec at 10000C) (113).

The average metal concentration in the as-implanted top

silicon layer is much larger than the equilibrium solid

solubility limit of Fe, Cr and Mo in bulk silicon at the

implantation temperature of 5100C as listed in Table 4.2.

For example, the average concentration obtained for Fe in

the as-implanted SIMOX top silicon layer is 2.04x1019/cm3

whereas the equilibrium solid solubility limit of Fe in

silicon at the implantation temperature of 5100C is

2.16x107/cm3 (104). Therefore, the implanted metal atoms

are located in a supersaturated region.

The total dose of Fe, Cr and Mo incorporated into the

top silicon layer as calculated from SIMS data is on the

order of 1014/cm2 for Fe and 1013/cm2 for Cr and Mo. These

values are more than 1 ppm of the implanted oxygen dose.

The amount of metal impurities in the buried oxide region is

significantly lower than in the top silicon layer by 2-3

orders of magnitude. An average concentration of 4.40x1016,

1.25x1016 and 2.44x1014/cm3 was obtained for Fe, Cr and Mo,

respectively, in the buried oxide region.

The effect of annealing on the redistribution of the

transition metal impurities is depicted in Figures 4.3-4.12.

It should be mentioned that the samples annealed at 9000C,

11500C and 11750C were analyzed 2 years before the others.

This time gap may have lead to differences in the SIMS

intensities obtained. Also, the SIMS profiles are plotted

using different scales. It should be pointed out that due

to time and financial limitations, only one SIMS profile

Table 4.2. Average concentration of Fe, Cr and Mo in the
top silicon layer of as-implanted SIMOX as
deduced from SIMS measurements and equilibrium
solubility limit of Fe, Cr and Mo in bulk silicon
at 5100C (104).

Average Concentration



3. 69x1018

Mo 7.46xi017

Solubility Limit



7. 47x106

<4. 00x1012

was performed per sample condition. As a consequence, a

large variability in the results may occur, which limits the

reliability of the SIMS data.

Analysis of Fe Distribution. Figure 4.3 shows that

significant redistribution of Fe takes place during

annealing at temperatures within 900-11750C. Both the

primary and secondary peaks previously observed in the as-

implanted top silicon layer (seen in Figure 4.2) have merged

into one broad peak. The average concentration of Fe in the

buried oxide region diminishes substantially, approximately

1 to 2 orders of magnitude after annealing at 11500C as

illustrated in Figure 4.3. It is also observed that

extensive pile-up of Fe has occured at the lower SiO2/Si

interface. Annealing for longer times at these same

temperatures does not cause significant changes in the metal

distribution as illustrated in Figure 4.3.

Figure 4.4 depicts the distribution of Fe after

annealing at higher temperatures (1200-13000C) for 2 and 5

hours. It can be seen that after annealing for 2 hours

(Figure 4.4a) substantial accumulation of Fe takes place in

the top silicon layer. But annealing at 13000C results in a

sharper Fe plateau in the top silicon layer. This result

was also observed after annealing for 5 hours at 12000C and

12500C as illustrated in Figure 4.4b.

From the Fe SIMS profiles, it can be seen that the top

silicon layer oxidizes during annealing in Ar+l.5%02 and

forms a surface SiO2 layer with a thickness that increases

ff ff
1021 900"C, 5hrs 1 021 1150"C, 2hrs 1 021 Surface 1175*C, 2hrs
-Surface Top Buried Substrate
1020 Top Buried Substrate 120 Top Buried Substrate 1020
Si S102 SS12

1091019 1019

0 io 1018 1015

W 10 I 105_ 101

01--------- 1 0 -- -- -- -1 0 .5 1- -
8I Ida
0L .5 17 1 0

DEPTH (microns)

Figure 4.3. SIMS distribution profile of Fe in SIMOX annealed within
900-11750C. The dashed line represents the detection limit.




o co L
04I -
0 0 0 0 0 0


1111 l 1 1 1 1 lift I I I I I.,.I



SI. I. I.






I -

L I, ,

0 .......,: I I I Jilll I I I I I Jill I I I 1 11111 1 1 1 1

(C.u~) NOI.IV.ILN30NO0








i E .... . .









9 d

hum I I I Ittll I I I 1111111 I I IIYN.., I I I 1,11113 I I 1111111 I I Jo
- 30 0 0 N 10 30
Cd Cd -
o 0 0 0 0 0 0
- i







.,-I 0



0 C0 0 0 0 0
-= ,r- ,- "rT-









o cmco,
N. -7 -
CD CD 00



Co If)
o 0 0
9- 9~

E) o

o y:



with increasing temperature and time. It is also observed

that the Fe peak in the top silicon layer tends to pile-up

near the silicon side of the interface with the surface

oxide. This result is consistent with observations

previously reported (123, 124).

Figure 4.5 shows the average concentration obtained

from the SIMS profiles for Fe in the surface SiO2 and top

silicon layers as a function of annealing temperature and

time. It is observed that the concentration of Fe in the

top silicon layer does not change significantly during

annealing at 9000C. The average concentration of Fe in the

as-implanted top silicon layer (not shown in Figure 4.5) is

of the order of 2.04x1019/cm3. After the 2 hours anneal

between 10500C and 11500C, the concentration of Fe in the

top silicon layer tends to decrease. This result seems to

be coincident with the formation and growth of a thermal

oxide at the surface, which might attract some of the Fe

metal impurities. From 11750C to 12500C, the Fe

concentration in the top silicon layer increases by 1 order

of magnitude. But after annealing at 13000C, it drops

again. The results obtained for the 5 hours anneals show a

decrease in the Fe concentration both in the top silicon

layer and surface oxide with increasing temperature.

In conclusion, the changes observed in these results

suggest that several competing processes must be considered

with respect to the motion of Fe metal impurities in the

5 Flours

*0 Top Si
(*1 ) Surface Si02

(10 hours)

f I I I

950 1050 1150 1250
Temperature (C)

Figure 4.5.

Average concentration of Fe in the surface Si02
and top Si layers of SIMOX as a function of
annealing temperature and time.


10 E










L__ __


Q) 8-










SIMOX structure during annealing. These processes are as


(1) Diffusion of Fe impurity atoms in the structure.

(2) Trapping of Fe impurity atoms in the implantation

damage area and SiO2 precipitates in the SIMOX structure.

(3) Segregation of Fe at the outer Sio2/Si interface

resulting from thermal oxidation.

Iron is a fast diffuser and has a low solubility in

silicon (104). Table 4.3 shows the diffusivity, solubility,

and diffusion length of Fe in silicon for each annealing

temperature and time. It can be seen that diffusion is not

a limiting factor for the redistribution of Fe impurity

atoms since the diffusion length, L = 2V(Dt), at each

temperature is a significant fraction of the wafer

thickness. Therefore, the tendency to segregate at

different regions in the SIMOX structure dominates the

distribution of the Fe metal impurities. So, it is likely

that Fe accumulates in regions of stress and high defect

density to relieve the strain in the lattice. The peaks at

the upper and lower buried oxide interfaces (seen in Figure

4.3) can be attributed to metal segregation at regions of

substantial crystal damage within the temperature range of

900-11750C. The changes in crystal damage during annealing

were observed by TEM and will be described in the following

section (4.2.2). The damaged regions extend throughout the

silicon layer above the implanted oxygen and several tenths

of a gm beyond the buried oxide into the substrate. The

Table 4.3. Diffusivity, solubility and diffusion length
of Fe in silicon (104).

Temperature Time

(C) (hr)










(x10-6 cm2/s)










(xl015 /cm3
























presence of SiO2 precipitates and dislocations in the SIMOX

structure as will be shown by TEM, creates potential

gettering sites for the migrating Fe metal impurities (106).

At higher annealing temperatures (1200-13000C) where

recovery of implantation damage is more effective, it is

expected that the mobility of Fe metal impurities through

the structure will increase as they dissociate from the

annealed defects. However, Figure 4.4 shows that a

significant amount of Fe remains in the top silicon layer.

Comparison of the values for the solubility of Fe in

silicon with its concentration in the top silicon layer of

SIMOX shows that the Fe metal impurities are in a state of

supersaturation at each annealing temperature. Therefore,

it may be expected that Fe precipitates nucleate in the top

silicon layer. However, these precipitates may be too small

(<100 A) to be resolved by TEM. Precipitation tends to

reduce the Fe diffusivity. Furthermore, it has been

reported that the solubility of Fe in silicon decreases

significantly in the presence of SiO2 precipitates due to

redissolution of the oxygen precipitates with annealing and

formation of an iron oxide compound (125).

The surface oxide that forms on the top silicon layer

becomes an effective factor on the redistribution of Fe at

higher annealing temperatures. It is believed that this

growing surface silicon oxide will tend to getter the Fe

impurity atoms co-implanted in the silicon. As mentioned

above, Fe impurities tend to accumulate in the top silicon

layer near the interface with the thermal oxide during high

temperature annealing. The driving force for the

precipitation of Fe at this surface SiO2/top Si interface

could be the lowering of free energy caused by the strain

field present near the interface. This strain field results

from thermal expansion mismatch between SiO2 and Si.

On the other hand, a marked broadening of the Fe SIMS

distribution profiles towards greater depths in the

structure is also shown. This end tail is indicative of

enhanced diffusion towards the substrate. This end tail is

very similar to the ones typically shown by phosphourous

implanted silicon (126). In fact, extensive pile-up of Fe

at the lower Si02/Si substrate interface was observed during

annealing in the lower temperature range (seen in Figure

4.3). Some of this Fe may have been initially trapped

during the buried oxide formation. A decrease in the

concentration of the Fe peak in the top silicon layer after

annealing in the 1200-1300C range, (seen in Figure 4.4)

leads to the conclusion that Fe continues to diffuse towards

the substrate at these temperatures. Evidence of Fe

accumulation (on the order of 2x1018/cm3) at the lower

buried Si02/substrate interface during high temperature

annealing is clearly shown in Figure 4.6, which depicts a

complete SIMS depth profile of SIMOX sample annealed 2 hours

at 12500C.
It is known that a high P concentration produces an

excess of interstitials, and the high diffusivity tail is

1019 -

o y~ Fe
z 1018
O | I

W 10' I

z I I
o _- /
1016 21!

0 .2 .4 .6 .8 1
DEPTH (microns)

Figure 4.6. SIMS depth distribution profile of Fe in
SIMOX annealed at 1250C for 2 hours. The
oxygen concentration has arbitrary units.

correlated to this point defect supersaturation (126).

Therefore, diffusion of Fe into the top silicon layer may be

assisted by an excess of silicon self-interstitials.

Supersaturation of silicon self-interstitials may arise

either from growth of Si02 precipitates in the top silicon

layer or during the growth of the surface thermal Sio2.

Analysis of Cr Distribution. Figure 4.7 depicts the

distribution of Cr during annealing within the temperature

range of 900-11750C. As compared to the as-implanted Cr

profile (seen in Figure 4.2) a substantial decrease in the

concentration of Cr in the top silicon layer occurs during

annealing at 9000C as illustrated in Figure 4.7a. The Cr

concentration drops by a factor of 2-3 orders of magnitude.

Most of the Cr is located near the surface after annealing.

A higher surface pile-up is observed after the 10 hours

anneal. These results clearly suggest that Cr is not as

easily gettered as Fe by the damage area in the top silicon

layer. It tends to segregate out of the top silicon layer.

This tendency is consistent with previous observations of Cr

implantation in SIMOX and bulk Si wafers (110, 124).

Annealing at 10500C, 11000C and 11750C for 2 hours as

depicted in Figure 4.7b shows both accumulation at the

surface and the beginning of Cr motion towards greater

depths into the structure (11750C). The profile obtained at

11500C, 2 hours is not included here. It should be pointed

out that analysis of these profiles is difficult due to the

low concentration levels of Cr. The Cr concentration


log9_ 9og

9000C, lOhrs

118 1018 Top Buried Substrate
C? S1 S102
Z 9000C, 5hrs
p. lo17
1 'op Buried Substrate 10 7
ccS1 S102

ia 1'8 10 l16

015 I__ _ _ __ _ _ 1 15
0 .5 1 0 .5 1

DEPTH (microns)
Figure 4.7. SIMS distribution profile of Cr in SIMOX annealed within 900-11750C.
(a) 9000C, 5 and 10 hours; (b) 1050-11750C, 2 hours and (c) 11500C and
i1750C, 5 hours. The dashed line represents the detection limit.



I "
.0 I

E 0


o Io

m ty



h I I I I f

1o I


: 0 0 0
1,l ,-, j,, ,, ,

( .to) N IJ.U.LNONw