Semiconductors A

NEW RESEARCH
ON SEMICONDUCTORS
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
NEW RESEARCH
ON SEMICONDUCTORS
THOMAS B. ELLIOT
EDITOR
Nova Science Publishers, Inc.

New York
Copyright 2006 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
NOTICE TO THE READER
The Publisher has taken reasonable care in the preparation of this book, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or
omissions. No liability is assumed for incidental or consequential damages in connection with
or arising out of information contained in this book. The Publisher shall not be liable for any
special, consequential, or exemplary damages resulting, in whole or in part, from the readers
use of, or reliance upon, this material.
This publication is designed to provide accurate and authoritative information with regard to
the subject matter covered herein. It is sold with the clear understanding that the Publisher is
not engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF
THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

New research on semiconductors / Thomas B. Elliot (editor).
p. cm.
Includes bibliographical references and index.
ISBN 978-1-60876-510-2 (E-Book)
1. Semiconductors--Research. 2. Semiconductor nanocrystals--Research. 3. Semiconductor
films--Research. I. Elliot, Thomas B.
QC611.26.N497 2006
537.6'22072--dc22 2005034761
Published by Nova Science Publishers, Inc. New York

CONTENTS
Preface vii
Chapter 1 Crystal Growth of Ternary and Quaternary Alloy 1
Semiconductors by Rotational Bridgman Method
Yasuhiro Hayakawa, M.Haris, Masashi Kumagawa
and Tetsuo Ozawa
Chapter 2 Formation of Grown-in Microdefects in Dislocation-free 31
Silicon Monocrystals
V.I. Talanin and I.E. Talanin
Chapter 3 Laser-induced Phase Transformation in Nanocrystalline 69
Silicon Thin Films
R.H. Buitrago and S.B. Conari
Chapter 4 Monte Carlo Simulation of Hot-Phonon Effects in Biased 95
Nitride Channels
M. Ramonas and A. Matulionis
Chapter 5 Researches of Growth and Device Based on ZnO by Plasma 123
Assisted Molecular Beam Epitaxy
Y. M. Lu, D. Z. Shen, J. Y. Zhang, Y. C. Liu and X. W. Fan
Chapter 6 Structural Disorder in Cds-x-se-1-x Thin Film Probed by 143
Microdiffraction and Optical Spectroscopes
V. Capozzi, G. Perna, S. Pagliara and L. Sangaletti
Chapter 7 Electrodeposition of CuInSe2 Photovoltaic Cell Application 159
Shigeyuki Nakamura
Index 209
PREFACE
This book includes within its scope studies of the structural, electrical, optical and
acoustical properties of bulk, low-dimensional and amorphous semiconductors;
computational semiconductor physics; interface properties, including the physics and
chemistry of heterojunctions, metal-semiconductor and insulator-semiconductor junctions; all
multi-layered structures involving semiconductor components. Dopant incorporation. Growth
and preparation of materials, including both epitaxial (e.g. molecular beam and chemical
vapour methods) and bulk techniques; in situ monitoring of epitaxial growth processes, also
included are appropriate aspects of surface science such as the influence of growth kinetics
and chemical processing on layer and device properties. The physics of semiconductor
electronic and optoelectronic devices are examined , including theoretical modelling and
experimental demonstration; all aspects of the technology of semiconductor device and circuit
fabrication. Relevant areas of 'molecular electronics' and semiconductor structures
incorporating Langmuir- Blodgett films; resists, lithography and metallization where they are
concerned with the definition of small geometry structure. The structural, electrical and
optical characterization of materials and device structures are also included. The scope
encompasses materials and device reliability: reliability evaluation of technologies; failure
analysis and advanced analysis techniques such as SEM, E-beam, optical emission
microscopy, acoustic microscopy techniques; liquid crystal techniques; noise measurement,
reliability prediction and simulation; reliability indicators; failure mechanisms, including
charge migration, trapping, oxide breakdown, hot carrier effects, electro-migration, stress
migration; package- related failure mechanisms; effects of operational and environmental
stresses on reliability.
As outlined in Chapter 1, ternary and quaternary alloy crystals of semiconductors are
important materials for optoelectronic devices because fundamental properties such as lattice
constant and band gap can be controlled by adjusting composition ratio. In order to grow
high quality ternary bulk crystals, various techniques have been carried out. Nakajima et al.
succeeded in growing compositionaly graded InxGa1-xAs (x=0.5-3.30) single crystals by the
Brigman method. Hayakawa et. al. introduced ultrasonic vibration into the melt from the
bottom of the crucible in the Czochralski method to prevent the occurrence of constitutional
supercooling. This method was applied for the growth of InxGa1-xSb single crystals, and it
was found that the growth thickness in the single crystalline state was increased with an
increase of the output power of ultrasonic vibrations. We have modified the Czochralski
method in order to improve the homogeneous distribution of impurities by making a seed
viii Thomas B. Elliot
crystal rotate alternately clockwise and counterclockwise within a narrow range of angle. We
have also modified the Brigman method to stir the melt. In Section 2, the growth of InSb1-
xBix ternary bulk crystals and the results analysed are presented.
Chapter 2 presents a general scheme that depicts the formation and transformation
mechanisms of the grown-in microdefects in dislocation-free silicon single crystals grown by
the floating-zone (FZ-Si) and Czochralski (CZ-Si) methods as a function of a crystal growth
rate. The distribution patterns and physical nature (the sign of the lattice strain) of grown-in
microdefects FZ-Si and CZ-Si crystals were studied by selective etching and transmission
electron microscopy. A heterogeneous mechanism of formation and transformation of grown-
in microdefects are proposed.
It is experimentally established and confirmed by thermodynamic analysis that in single
FZ-Si and CZ-Si crystals close to the crystallization front the recombination process of
intrinsic point defects is hampered by a recombinational barrier. ooling-induced
decomposition of the oversaturated solid solution of point defects in silicon follows two
mechanisms: vacancy-type and interstitial-type. Therefore vacancies and self-interstitials find
drains in the form of atoms of background impurities like oxygen and carbon. Depending on
growth conditions (e.g. parameter V/G, where V is a crystal growth rate; G is an axial
temperature gradient) the silicon crystals grow in vacancy-interstitial and (or) interstitial
regimes. The parameter V/G is determined experimentally with allowance for crystal
diameter, cooling rate and radial temperature gradient.
In vacancy-interstitial growth regime of FZ-Si and CZ-Si crystals during disintegration of
the supersaturated vacancies solid solution the oxygen-vacancy aggregate occurs causing
SiO2 presipitates to be formed. During disintegration of the supersaturated self-interstitials
solid solution the interstitials-carbon aggregate occurs causing the aggregates, which consist
of self-interstitials, atoms of carbon and oxygen, to be formed. During the formation of such
aggregates the carbon atoms act as catalysts.
The point defects aggregation in vacancy-interstitial and interstitial growth regimes under
certain thermal growth conditions during the crystal cooling may result in the formation of
secondary defects near the primary oxygen-vacancy and carbon-interstitial aggregates,
namely octahedral vacancy microvoids and interstitial dislocation loops accordingly. It is
experimentally established that A-microdefects may occur in the course of two different
formation mechanisms: condensations of self-interstitials and generation of interstitial
dislocation loops, which is called by a field of elastic deformation all around the presipitates.
As is established, the nature and behaviour of processes of the grown-in microdefects
formation in FZ-Si and CZ-Si crystals are identical. The uniform classification of grown-in
microdefects is offered and the heterogeneous mechanism of formation and transformation of
grown-in microdefects in dislocation-free silicon single crystals is confirmed experimentally.
On the ground of experimental results the physical classification of grown-in
microdefects is offered. Such a classification follows from the heterogeneous mechanism of
formation and transformation of grown-in microdefects.
A comprehensive knowledge of the so-called micro or nanocrystalline silicon has not
been reached till now. Nanocrystalline silicon structure network is quite complex. It consists
of crystallites of around 10 to 30 nm grain size. Because of the little crystallites size, a large
grain boundary tissue as well as amorphous phase is present, providing particular optical and
electrical properties. Being that the nanocrystalline silicon is thermodynamically more stable
Preface ix
and with less light-induced degradation than the amorphous one, this material presents a great
interest to scientists and technologists in thin films solar cells.
The technology used to prepare silicon solar cells is at present well known. The
deposition techniques take place at high temperature and because of the low thickness of the
films, they present very high residual tensions. In the other side, the deposition rate
determines the crystallite grain size. As light differences in crystalline volume fraction and
grain size result in quite different optical and electrical properties, it appears of considerable
interest to study phase transformation processes in nanocrystalline silicon, to know more
about the transition from amorphous to microcrystalline silicon under compression tension
and temperature, which is still not really understood.
In Chapter 3 an up-to-date arts state of pressure-induced phase transformation processes
in silicon are made. Also results of laser-induced phase transformation processes in intrinsic
and boron-doped silicon thin films prepared by Plasma Enhanced Chemical Vapor Deposition
technique (PECVD) are presented and analyzed.
Through Raman spectroscopy, it is possible to determine that the induced phase
transformation is interrelated with the silane flow rate, and boron concentration in the silane
diluted with hydrogen used as a reactive gasses in the preparation of the films, on the grain
size, and the crystalline volume fraction of them. Measurements of dark conductivity,
Transmission Electron Microscopy (TEM), X-ray and Atomic-Force Microscopy (AFM) help
to corroborate the Raman measurements.
The results show that the laser-induced phase transformation processes in nanocrystalline
silicon thin films present different characteristics from the well known ones produced in
single-crystal silicon by hydrostatic experiments. A model for hot-spot phase transformation
is discussed to explain the laser-induced phase transformation processes in the silicon thin
films. The little volume exposed to the laser beam reaches high temperature and is under high
compress tensions due to the cooler surrounding material. In these conditions, phase Si-XII
emerges and becomes more stable than Si-I. These two phases are present in equilibrium with
the amorphous phase in the silicon thin films with no presence of other crystalline phases.
Recent progress in Monte Carlo simulation of hot electron transport and fluctuations in
nominally undoped AlGaN/GaN and AlGaN/AlN/GaN heterostructures with degenerate two-
dimensional electron gas channels is reviewed in Chapter 4. Input scattering probabilities of
the electrons are calculated in a semiclassical approach from the confined-electron envelope
wavefunctions obtained through a self-consistent Poisson Schrodinger method. Additional
scattering due to accumulation of nonequilibrium longitudinal optical phonons, termed hot
phonons, is treated together with lattice heating and other scattering mechanisms of
importance for electron transport at high electric fields applied in the plane of electron
confinement. Possible ways to treat electron gas degeneracy and hot-phonon effects through
Monte Carlo procedures are described. Hot-electron and hot-phonon distribution functions are
presented for indepth discussion of the results.
Complementary information on hot phonons is extracted from electron energy dissipation
and fluctuations. In nitride channels, the hot phonons are found to slow down electron energy
dissipation and establish the hot-electron distribution controlled by the electron temperature.
The hot electron temperatures are evaluated from the electron distribution functions
calculated at different electric field strengths. The obtained dependence of hot-electron
temperature on supplied power is in a good agreement with the available experimental data on
microwave noise.
x Thomas B. Elliot
In Chapter 5, the ZnO based materials, such as ZnO thin film and its heterostructure,
were prepared on c-plans sapphire (Al2O3) substrates by plasma-assisted molecular beam
epitaxy (P-MBE). The influence of growth temperature on ZnO film quality was investigated.
The result reveals a change of the growth mode from three-dimensional (3D) nucleation to
two-dimensional (2D) nucleation. The photoluminescence (PL) spectra exhibit a strong
ultraviolet (UV) emission and a weaker visible emission for all samples. By the
measurements of PL spectra at different temperature, the origin of UV emission peak at RT is
considered to be from free exciton emission and the shift of UV emission peak position is
attributed to quantum confinement effect due to different crystal grain sizes. In addition, the
carrier concentration of ZnO thin films decreases with increasing growth temperature. In the
optimum growth condition, the carrier concentration is N=7.66 1016 / cm3 which is closed to
that of bulk ZnO.
On the other hands, N-doped p-type ZnO thin films were grown by P-MBE on c-plane
Al2O3 using radical NO as oxygen source and nitrogen dopant. The reproducing ZnO thin
films have maximum hole concentration of 1.21019cm-3and lower resistivity of 9.95cm. In
absorption spectra, the subbandgap related to N in the ZnO band gap was observed.
Comparing with optical emission spectra (OES) of radical nitrogen, a difference between
radical NO and N2 was found. The OES of radical nitrogen show strong ultraviolet emission
related to radical nitrogen (N2*), which is a shallow double donor in the ZnO films. While the
radical atoms (N*) are dominant in the OES of radical NO. The experiment result indicates
that p-type ZnO is realized by employing radical NO as N dopant.
A n-ZnO/p-GaN heterostructure and its light emitting diode were fabricated by P-MBE.
And the optoelectrical properties of the heterojunction are investigated. The device exhibited
rectifying I-V characteristics to the diode. Under forward bias voltage, electroluminescence
spectrum shows that a blue emission from GaN layer at the room temperature. To improve the
luminescence properties of the device and make emission from ZnO layer, we designed the
device structure and fabricated a p-GaN/i-ZnO/n-ZnO p-i-n heterojunction. Comparing with
p-n heterojunction, the emission peaks shifts to high-energy side in the EL spectrum of p-i-n
heterojunction. The origin is attributed to the emissions from i-ZnO layer of p-i-n
heterojunction.
As presented in Chapter 6, X-ray diffraction with a large area detector from CdSxSe1-x
alloy thin films, deposited on Si (111)-oriented substrates by laser ablation technique, allowed
us to point out the effects of structural disorder which can not be observed by conventional -
2 diffractometers. The X-ray spectra measured by using conventional -2 diffractometer
have shown that the CdSxSe1-x films grow oriented along the (002)-direction in the hexagonal
phase with the c-axis perpendicular to the film layers. The width of the 002 reflection is
probably related to substitutional disorder arisng from alloying. However, further disorder
effects are detected when one resorts to microdiffraction. Indeed, the broadening along the
Debye rings reveals that the films are not perfectly epitaxially grown. Moreover, only two
families of planes are present in the microdiffraction spectra of all samples. This finding is
discussed by considering a structural disorder not related to alloying. An attempt to simulate
this structural disorder has been carried out on the basis of symmetry considerations and
discussed. A correlation is found with the results of luminescence and Raman spectroscopies
that give evidence of disorder effects in terms of excitation localization in photoluminescence
and a broad, disorder activated, band in the Raman spectra.
Preface xi
Chapter 7 describes preparation of a ternary chalcopyrite semiconductor, copper indium

diselenide (CuInSe2:CIS), thin films as a light absorption layer for thin film solar cells by
electrodeposition.
This semiconductor is very attractive for an absorber material in the thin film solar cell
because of its suitable bandgap and a large absorption coefficient. CuInSe2 has the bandgap
of about 1.0 eV and the absorption coefficient of ~ 105 cm1. The large absorption coefficient
enables us to realize the thin film solar cell. Although the highest theoretical efficiency is
obtained with the semiconductor whose bandgap is ranging from 1.4 to 1.5 eV, the bandgap
of CIS can be enlarged up to 1.68 eV by adding Ga to form solid solution of Cu(InxGa1x)Se2
(CIGS).
Electrodeposition as a method for thin film semiconductor preparation is a good approach
with respect to economic consideration. An important advantage of electrodeposition as a
method for thin film preparation is that films with a large area can be prepared without a
vacuum, using simple and low-cost equipments. However, conversion efficiencies ever
reported for the cells fabricated by this method are considerably lower than those reported for
the cells fabricated by other methods. One of the important problems for development of this
technique is to control sample compositions. Understanding of deposition mechanisms of
each species is essential in order to achieve higher controllability and reproducibility of film
composition and then to improve performance of the photovoltaic cell prepared by
electrodeposition. Moreover, an excellent morphology is also essential to achieve higher
conversion efficiency. A poor morphology causes short-circuiting between the front and back
electrodes. Voids and cracks in thin films degrade the conversion efficiency. On the basis of
above mentioned, following subjects were studied.
Electrodeposition of Cu-In-Se films has been studied with an aqueous solution containing
CuCl2, InCl3 and SeO2, in terms of composition control of deposited films for the preparation
of CuInSe2. When a Si wafer is employed as a substrate, both the electrode potential
dependence of In/Cu ratio in the film and a stirring effect on film composition are found to
become small, compared with Mo substrate.
This is explained by taking account of the existence of a space charge layer at the
semiconductor surface. From the relationship between In/Cu or Se/Cu ratio in the bath and
that in the deposited films, ratios of mass-transfer coefficients for In and Cu, k(In)/k(Cu) or
for Se and Cu, k(Se)/k(Cu), have been obtained and their dependencies on deposition current
density or stirring of the solution have been studied. The rate-determining step in the
deposition process for each ion, reaction-limited or diffusion-limited, has been also
discussed. By employing the stirring, a remarkable improvement, by a factor of 3, is attained
for run-to-run and position-to-position fluctuation of the In/Cu ratio in the films. X-ray
diffraction patterns show that CuInSe2 is contained in as-deposited films.
In order to improve surface morphology, a revers bias is applied after electrodeposition in
the same electrolyte. Pluse-plated electrodeposition was demonstrated for the same purpose.
In order to improve crystallinity, as-deposited films were annealed in an inert
atomosphere. Annealing effects on film properties, such as crystallinity, morphology and
chemical composition, were investigated. Effects of Se concentration were also investigated.
Excess Se is found to degrade controllability of composition and crystallinity.
The conversion efficiency of 0.866 % (Voc=0.47 V, Isc=1.9 mA/cm2, FF=0.533) is
obtained with a chemical bath deposited CdS buffer layer on the electrodeposited CIS layer
without any transparent conductive oxide layers. Preparation of CuInSe2 thin films with a
xii Thomas B. Elliot
bilayer structure by electrodeposition were studied. Change in a Cu/In ratio with depth is
essential to obtain higher conversion efficiency. Thus, a new electrodeposition technique for
preparing CuInSe2 thin films with controlled depth profile were developed. The CuInSe2 thin
films with both Cu-rich and In-rich layers are deposited by changing substrate potential
during electrodeposition.
In: New Research on Semiconductors ISBN: 1-59454-920-6
Editor: Thomas B. Elliot, pp. 1-30 2006 Nova Science Publishers, Inc.
Chapter 1
CRYSTAL GROWTH OF TERNARY AND QUATERNARY

ALLOY SEMICONDUCTORS BY ROTATIONAL
BRIDGMAN METHOD
Yasuhiro Hayakawa*, M.Haris, Masashi Kumagawa

Research Institute of Electronics, Shizuoka University, 3-5-1 Johoku,
Hamamatsu, Shizuoka 432-8011, Japan
Tetsuo Ozawa
Department of Electrical Engineering, Shizuoka Institute of Science and Technology,
2200-2 Toyosawa, Fukuroi, Shizuoka 437-8555, Japan
Abstract
Ternary and quaternary alloy crystals of semiconductors are important materials for
optoelectronic devices because fundamental properties such as lattice constant and band gap
can be controlled by adjusting composition ratio. In order to grow high quality ternary bulk
crystals, various techniques have been carried out. Nakajima et al. succeeded in growing
compositionaly graded InxGa1-xAs (x=0.5-3.30) single crystals by the Brigman method.
Hayakawa et. al. introduced ultrasonic vibration into the melt from the bottom of the crucible
in the Czochralski method to prevent the occurrence of constitutional supercooling. This
method was applied for the growth of InxGa1-xSb single crystals, and it was found that the
growth thickness in the single crystalline state was increased with an increase of the output
power of ultrasonic vibrations. We have modified the Czochralski method in order to
improve the homogeneous distribution of impurities by making a seed crystal rotate
alternately clockwise and counterclockwise within a narrow range of angle. We have also
modified the Brigman method to stir the melt. In Section 2, the growth of InSb1-xBix ternary
bulk crystals and the results analysed are presented.
*
E-mail address: royhaya@ipc.shizuoka.ac.jp, Tel: 81-53-478-1310
2 Yasuhiro Hayakawa, Tetsuo Ozawa, M. Haris et al.
1 Introduction
Ternary and quaternary alloy crystals of semiconductors are important materials for opto-
electronic devices because fundamental properties such as lattice constant and band gap can
be controlled by adjusting composition ratio. The alloy crystals offer the possibility to reduce
the problem of lattice mismatch at the interface between a substrate and an epilayer. Ternary
alloy crystals such as InGaAs [1], GaAsP [2], InGaSb [3], InAsP [4] and quaternary alloy
crystals such as GaInAsSb [5], InAsSb [6] have been grown until now. However, it is very
difficult to grow large single ternary alloy crystals of high quality, because there are three
major problems which must be overcome. The first is the constitutional supercooling which
appears in the source solution ahead of the growth interface. Since the separation between
liquidus and solidus lines is wide, the degree of constitutional supercooling becomes large
which brings about the growth of polycrystals [7,8]. The second is the occurrence of
composition changes both in the grown crystal and in the solution during growth. This means
that the composition ratio in the alloy crystal continues to change because the segregation
coefficient of each component is not unity. The third relates to the flows caused by heat and
mass transfer. The quality of alloy crystals is strongly affected by these flow patterns.
InSb1-xBix is a very interesting material because InBi has a semi-metallic character and
the band gap energy is controlled and can be reduced by replacing Bi into Sb sites. This
crystal is useful for detectors in the middle infrared region (7.4-16m) [9]. InxGa1-xSb is the
ternary alloy between InSb and GaSb binary alloys which can modulate the wavelength
region uniquely from 1.7 to 6.8m by adjusting the indium to gallium composition ratio; it
can be used as a device material in infrared region and further well suitable for thermo-
photovoltaic cells. InxGa1-xAs is an important material to fabricate thermo-photovoltaic cells
in the near infrared region, because the energy gap and the lattice constant can be tuned in the
range of 0.87-3.5m and 5.6533-6.0584 by adjusting the indium composition. InGaSbBi
quaternary bulk single crystals, which have not been grown up to the present is very
attractive, because it is composed of semimetal (InBi) and semiconductors (InSb and GaSb).
The band-gap and lattice constant can be separately varied in the wide range.
In order to grow high quality ternary bulk crystals, various techniques have been carried
out. Nakajima et al. succeeded in growing compositionally graded InxGa1-xAs (x=0.05-0.30)
single crystals by the Bridgman method. These graded crystals were used as an InGaAs seed
on which In0.25Ga0.75As bulk crystal was grown under lattice-matched condition [10]. Kodama
et al. grew a 28 mm long single-crystalline InxGa1-xAs (x=0.3) ternary bulk crystal on GaAs
seed crystal using the two-step multi component zone melting method. The furnace
temperature decreased to increase InAs content in growing crystal and kept constant to grow
homogeneous crystal [11]. Sheibani et al. improved liquid phase electro epitaxy system by the
application of an external static magnetic field to grow, flat, thick and uniform In0.04Ga0.96As
single crystals [12].
Hayakawa et.al introduced ultrasonic vibration into the melt from the bottom of the
crucible in the Czochralski method to prevent the occurrence of constitutional supercooling
[13-18]. The facet region located in the centre of the InSb single crystal shifted toward the
periphery of the crystal and decreased in size by introducing ultrasonic vibrations.
Furthermore, the difference in impurity concentration between the facet and the off-facet
regions became smaller. This method was applied for the growth of InxGa1xSb single
Crystal Growth of Ternary and Quaternary Alloy Semiconductors by Rotational 3
crystals, and it was found that the growth thickness in the single crystalline state was
increased with an increase of the output power of ultrasonic vibrations [19-22]. We have
modified the Czochralski method in order to improve the homogeneous distribution of
impurities by making a seed crystal rotate alternately clockwise and counterclockwise within
a narrow range of angle [23,24].
We have also modified the Bridgman method to stir the melt. The method (which we
called "rotational Bridgman method") is similar to the tipping method developed by Nelson
[25]. However, to increase the stirring effect, a relative motion between the growing surface
and the solution can be given by rotating the growth ampoule at high speed, where the
growing surface covered about 6090% with the solution. We adopted this method to grow
InSb1xBix, InxGa1xSb, In1xGaxSb1-yBiy and Ga1xInxAsySb1y alloy crystals [26-31]. The
growth of crystals was controlled by decreasing the temperature or by moving the ampoule to
the lower temperature region along with the ampoule rotation. The effect of ampoule-moving
rate and rotation rate on the growth rate and the indium composition profile in the InxGa1-xAs
(x=0.02-0.03) crystals was investigated using the impurity striations intentionally introduced
into the growing crystal by thermal pulse technique [32].
To investigate the solution convection in the rotational Bridgman method, both flow
patterns and temperature distribution were calculated by solving Navier-Stokes, continuity
and energy equations in 3-dimensional model. By increasing the ampoule rotation rate from 0
rpm to 100 rpm, the flow velocity in the solution increased and the temperature distribution
tended to be uniform [33,34].
In Section 2, the growth of InSb1-xBix ternary bulk crystals and the results analysed are
presented. The growth of InxGa1-xAs ternary bulk crystals and the determination of growth
rate with respect to the ampoule moving rates and rotation rates are investigated, and are
described in Section 3. The mechanism of the diffusion of both Ga and Bi into InSb seeds
during the growth of InGaSbBi and the formation of domains in the InSb seed crystal is
analysed. The results have been compared with the InSb1-yBiy and InxGa1-xSb crystals and are
described in Section 4. The numerical simulation results of the effect of ampoule rotation are
explained in Section 5. The important results and the conclusions are summarized in
Section 6.
2 Growth of InSb1-XBiX Bulk Crystals

InSb1-xBix ternary crystals were grown by the following two kinds of improved Bridgman
method (1) with high speed ampoule rotation and (2) with both high speed rotation and
continuous source feeding (continuous feeding rotationary Bridgman method (CFRBM)). The
growth procedure of each method and the results are as follows.
2.1 Rotationary Bridgman Method (RBM)
Fig. 1 shows schematic diagram of growth arrangement used for the growth of InSb1-xBix.
The gold furnace, which is semi-transparent, was fixed on one side of a seesaw plate. The
plate could be inclined by driving a motor (C). Prior to crystal growth, a silk hat shaped InSb
seed crystal and In-Sb-Bi source materials were charged in the quartz ampoule. After
alternative evacuation and filling with nitrogen gas several times, the ampoule was sealed to
about 510-4 Pa and set to a connecting rod of motor (A). By inclining the seesaw plate to the
opposite direction, the seeding procedure was carried out by connecting the seed with the
solution. Crystal growth was initiated by lowering the furnace temperature at the rate of 0.6 to
1.8C/h. The gold coated furnace was very useful to make the seeding procedure easy,
because inside of a growth ampoule can be seen during growth.
Gold furnace
In-Sb-Bi solution
In-Sb-Bi grown crystal
InSb seed
Carbon block
Mov
eme
nt o f amMotor (A)
poul
e
Motor (B)
Motor (C)
Seesaw plate
Fig. 1 Schematic diagram of rotationary growth arrangement used for the growth of InSb1-x Bix.
Fig. 2(a) A roughly polished surface of InSb1-xBix. This crystal was grown from the solution of 1 part
InSb and 1 part InBi. (b) The profile of bismuth composition ratio (x) along the growth direction.
Fig. 2(a) shows a roughly polished surface of InSb1-xBix. This crystal was grown from the
solution of 1 part InSb and 1 part InBi. The crystal was grown at a cooling rate of 1.2C/h
from 480C to 250C with the ampoule rotation rate of 100rpm. InSb1-xBix grew in the single
crystalline state on the InSb seed crystal of 13 mm in diameter. The length of the single
crystal was about 17mm. High speed ampoule rotation enhances the relative motion between
the crystal and the solution, and further it supports the solution move back and forth in
parallel to the ampoule axis just like the tide. This motion suppresses the occurrence of
contitutional supercooling. As a result, the single crystal was grown.
The profile of bismuth component ratio (x) along the growth direction is shown in Fig.
2(b). The value of x was 0.015 at first, and increased to about 0.03 at 17mm. Since the
segregation coefficient of bismuth was smaller than unity, the Bismuth composition in the
crystal was smaller than that in the solution. When the crystal was growing, the Bismuth
component was rejected and starts accumulating in the solution. As a result, the Bismuth
concentration increased as the growth was progressed. The result agrees well with the solidus
curve in the InSb-InBi phase diagram [35] which is shown in Fig. 3.
Fig. 3 InSb-InBi phase diagram.
2.2 Continuous Rotationary Feeding Bridgman Method (CFRBM)
Figs. 4(a) and (b) show a schematic representation of growth ampoule and the corresponding
temperature profile in the furnace, respectively. InSb was used as a seed crystal and a feeding
source. The inclination of the ampoule is adjusted so that the solution can contact both seed
and feed crystals. The seed crystal was set in a slightly lower temperature but higher
temperature gradient region than the feeding source. Grown crystals were cut along the
growth direction. They were polished roughly at first and then mirror polished with alumina
abrasive. They were etched with a KMnO4:HF:CH3COOH etchant solution. Fig. 5(a) shows a
polished surface of an InSb1-xBix crystal grown using CFRBM. A starting temperature of
490C, a cooling rate of 1.8C/h and an ampoule rotation rate of 100 rpm was applied in this
case. The upper part of the crystal was polycrystalline. However, the length of the single
crystal region in a part of the grown crystal was about 15 mm. This result suggests that long-
sized single crystals can be grown if a large amount of the source is fed in the solution.
Profiles of bismuth component ratio along the growth direction and the diameter far 2mm
from the seed-grown crystal interface are given in Figs. 5(b) and 5(c), respectively. The
bismuth comcentration was nearly constant through the grown crystal. It indicated that the
InSb component was supplied from the InSb feeding source into the solution during growth.
So CFRBM is an adequate technique to grow homogeneous InSb1-xBix ternary crystals.
Fig. 4(a) A schematic representation of growth ampoule, (b) the corresponding temperature profile in
the furnace.
Fig. 5 Profiles of bismuth composition ratio along (a) the growth direction and (b) the diameter far
2mm from the seed-grown crystal interface.
3 Growth of InXGa1XAs Ternary Bulk Crystals

Fig. 6 shows schematic diagram of growth arrangement used for the growth of InxGa1-xAs.
The gold furnace, which is semi-transparent, was fixed on one side of a seesaw plate. Prior to
crystal growth, a silk hat shaped GaAs seed crystal, a polycrystalline GaAs feed and In-Ga-As
source materials were charged in the quartz ampoule. GaAs (100) single and poly crystals
were used to prepare the seed and feed crystals, respectively. The ampoule was sealed under
about 510-4 Pa and set to a connecting rod of motor (A). Feeding from polycrystalline source
to the solution was done by inclining the seesaw plate (using motor(C)) until the solution was
saturated. By inclining the seesaw plate to the opposite direction, the seeding procedure was
carried out by connecting both seed and feed crystals with the solution. Crystal growth was
initiated by moving the ampoule toward the low temperature region of the furnace using a
motor (B). For maintaining the temperature profile in the furnace, the seed was set at higher
temperature gradient region than the feeding source as shown in Fig. 7.
Fig. 6 Schematic diagram of growth arrangement used for the growth of InxGa1-xAs.
Fig. 7 Temperature profile in the furnace where the seed was set at higher temperature gradient region
than the feeding source.
3.1 Direct Growth Method
Fig. 8 shows a roughly polished surface of an InxGa1xAs crystal directly grown on GaAs
seed, where the nominal indium composition is 0.10. The growth was carried out with an
ampoule-moving rate was 0.5mm/h with the seed-rotation rate of 75 rpm. The length of
InxGa1xAs single crystal was 6.5mm. The shape of the growth interface was convex towards
the seed. Fig. 9 shows the indium composition profiles along the growth direction of the
crystals. When the indium composition ratio (x) was less than 0.12, the indium profile was
nearly uniform. This indicates that the GaAs component was fed continuously into the In-Ga-
As solution during growth. When the x value is 0.20, the indium composition increased
slightly along the growth direction. This means the supply of GaAs is not enough in this case.
Fig. 8 Roughly polished surfaces of InxGa1xAs crystal directly grown on GaAs seed, where the
nominal indium composition is 0.10.
Fig. 9 Indium composition profiles along the growth direction of the InxGa1-xAs crystals.
The relationship between indium composition ratio x and normalized FWHM of XRD in
the InxGa1xAs crystals is shown in Fig. 10. FWHM of the grown crystals was normalized by
that of GaAs seed crystal. The normalized FWHM increased with an increase of indium
composition ratio. This indicated that the quality of the crystals became poor by increasing
indium composition because the lattice mismatch between the grown crystal and the seed was
increased.
Fig. 10 Relationship between indium composition ratio x and normalized FWHM of XRD in the
InxGa1xAs crystals.
3.2 Multi-step Growth Method
Fig. 11 shows the temperature-time profile of multi-step growth. InxGa1-xAs ternary crystal
was grown by adopting three growth steps. InxGa1-xAs (x=0.03) was grown on a GaAs seed,
and then InxGa1-xAs (x=0.05) was grown on InxGa1-xAs (x=0.03). Finally InxGa1-xAs (x=0.07)
was grown on InxGa1-xAs (x=0.05). To increase the indium composition step by step, the
temperature of the solution at each step were adjusted at 700C, 800C and 900C,
respectively. Figs. 12(a) and (b) show feeding process and growth processes, respectively. In
(a) GaAs component was fed from GaAs feed source to the In-Ga-As solution until the
solution was saturated. In (b) crystal growth was started by moving the ampoule toward the
low temperature region of the furnace with a moving rate of 0.75mm/day. The main feature of
the multi-step growth was that feeding and growth processes were repeated and the indium
composition in the InxGa1xAs crystal increased step by step.
Fig. 11 Temperature-time profile of multi-step growth.

Fig. 12 (a) Feeding process and (b) growth processes.
Fig. 13 shows indium composition profile in an InxGa1xAs crystal grown by the multi-
step growth method. The lengths of the single crystalline portion were 1mm, 1.5mm and
6.5mm at each step region. The indium composition in the first and second step regions was
nearly constant throughout the grown crystals. In the third step region, the indium
composition was homogeneous in the wide region. It proved that the GaAs component was
fed continuously into the In-Ga-As solution at each step.
Fig. 13 Indium composition profiles in the InxGa1xAs crystals grown by the multi-step growth method.
Fig. 14 XRD FWHM profiles along the growth direction in the InxGa1xAs (x=0.10) grown by (a) the
direct growth method and (b) the multi-step growth method.
Fig. 14 shows the XRD FWHM profiles along the growth direction in the InxGa1xAs
(x=0.10) grown by the direct and the multi-step growth methods. In the InxGa1xAs crystal
grown directly on GaAs, XRD-FWHM was in the range of 140 arcsec, whereas the FWHM
of the seed crystal was 35 arcsec. In the InxGa1xAs grown by the multi-step growth method,
FWHM at the interface of the seed and the first step region was 120 arcsec. This was similar
to the value of the crystal directly grown on the seed. However, FWHM in the second step
region decreased to 70 arcsec. Furthermore, FWHM in the third step region was nearly equal
to that of the seed crystal. FWHM decreased to one fourth the value of the direct growth. This
indicates that the influence of lattice mismatch decreased by using the multi-step growth
method.
Fig. 15(a) Interface between GaAs seed and InxGa1-xAs(x=0.07) crystal of the direct growth, (b)
interface between InxGa1-xAs(x=0.05) and InxGa1-xAs(x=0.07) of the multi-step growth.
Figs. 15(a) and (b) show the interface between GaAs seed and InxGa1-xAs(x=0.07) crystal
grown using direct growth process, and the interface between InxGa1-xAs(x=0.05) and InxGa1-
xAs(x=0.07) crystal regions grown using multi-step growth process, respectively. The
interface was irregular and many inclusions were incorporated in the crystal grown using
direct growth. On the contrary, the interface became smooth and the number of inclusions
decreased in the crystal grown using multi-step growth process. When InxGa1-xAs (x=0.07)
was grown directly on a GaAs seed crystal, the lattice mismatch between them was 0.5%. On
the other hand, when InxGa1-xAs (x=0.03) was grown on GaAs seed and then InxGa1-xAs
(x=0.05) and InxGa1-xAs (x=0.07) were grown, the lattice mismatch at each interface could be
decreased to 0.21%, 0.14% and 0.14%, respectively. Therefore, the quality of the crystals was
improved by the multi-step growth process.
3.3 Effect of Ampoule- moving Rate
To investigate the effect of ampoule-moving rate on the growth rate and the indium
composition profile, the moving rates were changed step by step as 20, 30, and 40 m/h under
the constant rotation rate of 40 rpm. The thermal pulse technique was used to investigate the
real growth rate [36,37]. A small amount of tellurium having a concentration of about 1.0 x
1019cm-3 was added to the In-Ga-As solution as an impurity. While crystal growing, the
temperature was raised to 5C above the real temperature (i.e.from 800C to 805C) in one-
minute duration, hold it for 1 min and again lowered to 800C in one-minute duration. This
procedure was repeated for every 10 hours. Impurity striations were revealed by etching the
polished surface using chemical etchant (HF: H2O2: H2O = 10:1:1 (vol%)) under light
irradiation for 90 seconds. The real growth rates were calculated from the thermal pulse
striations observed using optical microscope.
Fig. 16(a) (010) cut surface of the InGaAs grown crystal. Ampoule-moving rates were changed step by
step as 20, 30, and 40 m/h under the constant rotation rate of 40 rpm. (b) An enlarged photograph of
the region near the GaAs seed and the InGaAs grown crystal. Impurity striations caused by thermal
pulses were clearly seen in the grown region
Fig. 17(a) Growth rate and (b) indium composition profile measured along the line ST shown in Fig.
16(a).
Fig. 16(a) shows a (010) cut surface of the crystal. About 8 mm thick InGaAs crystal with
diameter of 12 mm was grown. Several cracks were seen but InGaAs crystal directly grown
on GaAs seed was single crystalline. The shape of the growth interface was convex towards
the seed. Fig.16 (b) shows an enlarged photograph of the region near the GaAs seed and the
InGaAs grown crystal as indicated by N in Fig.16 (a). There was no striation in the seed,
while the impurity striations caused by thermal pulses were clearly seen in the grown crystal.
Since the thermal pulse modulated Te impurity concentration, these striations were formed.
The distance between the striations was short near the seed, and it gradually increased as the
crystal was grown. It indicated that the growth rate was not constant during growth.
Fig.17 (a) indicates the growth rate calculated by measuring the distance of the striations.
The growth did not follow the ampoule-moving rate. The growth rate gradually increased
from 5m/h to 20m/h. As inferred from the phase diagram of In-Ga-As at the temperature
800C, the atomic concentrations of indium, gallium and arsenic are 41.5, 53.2, 5.3 (at%),
respectively. Compared with the indium and gallium concentrations, the arsenic concentration
was very small. When the InxGa1-xAs crystal was grown, 50 at% arsenic was incorporated
into the crystal. Therefore, the arsenic in the solution was deficient. It suggested that the
growth rate was mainly determined by the amount of arsenic. The growth rate was small at
the initial growth stage, but it increased gradually because the arsenic was supplied by
dissolving the GaAs feed in the solution during growth. When the ampoule-moving rate
increased to 30 m/h, the growth rate decreased from 20 m/h to 10 m/h, and then
increased. The change of the moving rate brought instability of mass transfer. The growth rate
followed the same trend for all the ampoule moving rates.
Fig.17(b) shows the indium composition profile measured along the line ST shown in
Fig.16(a). The average indium value was 0.022 with a fluctuation of 0.001. The indium
composition was nearly uniform even if the growth rate was not constant. Since the growth
rate was less than the ampoule-moving rate, the growth temperature should decrease during
growth and the indium composition should increase gradually. Two reasons were considered.
One was that the change of the solidus composition at about 0.02 was not sensitive to the
temperature at nearly 800C as shown from the In-Ga-As ternary phase diagram [38]. The
second was that the gallium was supplied by dissolving the GaAs feed in the solution and was
transported towards the growth interface mainly by convection. Therefore, the sufficient
gallium was supplied during growth to compensate the increase of indium.
3.4 Effect of Ampoule Rotation Rate
In order to investigate the ampoule rotation rate on the growth rate and the indium
composition, the rotation rates was changed as 8, 40, 80 rpm during growth under the
constant ampoule-moving rate of 20m/h. Fig. 18 shows a (010) cut surface of the crystal.
About 8.5 mm thick InGaAs crystal with a diameter of 12 mm was grown. Twin structures of
{111} plane were formed at the middle region of the crystal. They started to grow from the
ampoule wall. However, InGaAs crystal directly grown on GaAs seed was monocrystalline.
Also the shape of the interface becomes convex towards the grown crystal thereby indicating
the influence of rotation rate on the shape of the interface.
Fig. 18. (010) cut surface of the InGaAs grown crystal. Ampoule rotation rate was changed as 8, 40, 80
rpm during growth under the constant ampoule moving rate of 20m/h.
Fig. 19 (a) Growth rate and (b) indium composition profile measured along the line PQ shown in
Fig. 18.
Fig. 19(a) indicates the growth rate along the line shown by PQ in Fig. 18. The growth
rate vastly increased from 8 to 47 m/hr by increasing the rotation rate from 8 rpm to 40 rpm.
The length of the liquid zone is approximately about 16mm. It should be noted that the
liquidus zone would cover only half cylindrical part of the ampoule. The inner diameter of the
ampoule is 12 mm. Hence the volume of the In-Ga-As solution is approximately 0.9 cm3.
After that the growth rate was decreasing and approaching to the value of the ampoule-
moving rate. It was because the solute transport from the feed to the growth interface was
enhanced due to convection. Furthermore, the temperature decreased by 4q since dissipation
of heat increased by increasing the rotation rate as the quartz ampoule was surrounded by air.
It indicated that the change of ampoule rotation from 8 to 40 rpm brought about the instability
of mass and heat transfers. On the other hand, the change of growth rate was small when the
rotation rate increased from 40 to 80 rpm. It suggested that the effect of ampoule rotation on
mass and heat transfers was not significant compared with the rotation change from 8 to 40
rpm. Fig. 19(b) shows the indium composition profile of the crystal along the PQ line. The
indium composition ranged between 0.02 and 0.03 and it did not follow the change of
growth rate.
4 Growth and Analysis of In1-xGaxSb1-yBiy Quaternary Crystals

4.1 Growth Morphology
Fig. 20 shows a roughly polished surface of an In1-xGaxSb1-yBiy crystal. The initial

composition ratio of a source solution was In : Ga : Sb : Bi = 48 : 2 : 25 : 25 (at%). Crystal
growth was performed by lowering the solution temperature from 480qC at a speed of 0.6qC/h
with the ampoule rotation rate of 100 rpm by using motor (A)(refer Fig.1). A seed crystal of
silk-hat shape at the left-hand side was InSb, and an InGaSbBi crystal grew to the right-hand-
side direction. From the observations by X-ray Laue patterns and growth morphology, the
orientation of crystal was different from at the end of the grown crystal, but about 13 mm
thickness of crystal was single. As seen clearly in Fig. 20, white and small areas were present
in the InSb seed crystal. To confirm the growth morphology more clearly, an X-ray topograph
of the seed region was taken using the KD1 line from an Ag target as given in Fig.21. After
the crystal was polished to 400Pm thick, an X-ray of (220) diffraction was detected. A broken
line indicates the boundary between InSb and InGaSbBi. Many black areas were clearly seen
and corresponded to the white and small areas in Fig. 20. These specular areas are domains.
They were about 2 mm long and 100 - 200Pm wide near the interface, but they reduced
gradually in size towards the left-hand side from the growth interface. At a distance of about
6.5 mm from the interface, the domains disappeared suddenly. Although many vertical lines
were seen on the X-ray photograph, they were wrinkles of emulsion on the film.
Fig. 20 Roughly polished surface of an InGaSbBi crystal. The initial composition ratio of a source
solution was In : Ga : Sb : Bi = 48 : 2 : 25 : 25 (at%).
Fig. 21 X-ray topograph of the seed region taken using the KD1 line from an Ag target.
Fig. 22 InGaSbBi quaternary crystal grown from the solution of different ratio of In : Ga : Sb : Bi = 46 :
4 : 25 : 25 (at. %).
Fig. 22 shows an InGaSbBi quaternary crystal grown from the solution of different ratio
of In : Ga : Sb : Bi = 46 : 4 : 25 : 25 (at. %). Compared with the grown crystal shown in Fig.
20, the gallium composition ratio in the solution was larger by 2 at%, the starting growth
temperature (495qC) was higher by 15qC, and the period of growth (133h) was longer by 32
h. The length of single crystal reached about 12.5 mm. The rotational Bridgman method
served as a good way of growing InGaSbBi quaternary crystals with higher Ga ratio. In this
crystal, many domains were seen not only in the InSb seed but also in a part of the grown
crystals near the interface.
4.2 Composition Profiles
Fig. 23 indicates composition ratios of Ga(x) and Bi(y) in the InGaSbBi crystal shown in Fig.
20. Solid and opened circles denote x and y values, respectively. The measurement was
carried out so as not to cross the domains along the line AB. In the grown InGaSbBi crystal,
the gallium composition ratio had the maximum value of 0.13 at the interface and gradually
decreased to 0.037 at a distance of 11mm. This was because the segregation coefficient of
gallium was larger than unity. On the contrary, the ratio of bismuth was very low, and there
was a slight increase owing to the segregation coefficient less than unity. Therefore, it was
found that the bismuth composition in the solution played a role as a solvent and further made
the melting point of solution reduce below 525qC of that of InSb. Both gallium and bismuth
compositions were incorporated in the InSb seed,. The gallium composition ratio decreased
gradually from the interface toward the end of the seed, and finally became zero at the
distance of -6.5 mm. This distance corresponds to the position where the domains disappeared
suddenly.
Fig. 23 Composition ratios of Ga(x) and Bi(y) in the InGaSbBi crystal shown in Fig.20.
Fig. 24 Composition ratios of Ga(x) and Bi(y) in the InGaSbBi crystal shown in Fig.22.
Fig. 24 shows composition profile of the sample grown from the source solution
contained 4 at.% gallium, as seen in Fig. 22. The measurement was carried out so as not to
cross the domains along the line CD. The gallium value at the interface was 0.21, which was
higher than that of the crystals grown from the 2 at.% gallium solution as indicated in Fig. 23
The difference in gallium ratio at the interface resulted from the gallium composition ratio in
the solution. In the inside of grown InGaSbBi, the gallium ratio decreased steadily to 0.033
with growing. On the contrary, the gallium composition profile in the InSb seed crystal was
very different from the corresponding one in Fig. 23 The gallium value reduced abruptly to
about 0.15 in composition ratio near the interface, and it was almost constant until the seed
end. It was found that the incorporation of gallium was promoted by both higher growth
temperature and longer growth period.
4.3 Lattice Constant
The XRD FWHM was measured by four-crystal X-ray diffractometry (model SLX-1, Rigaku
Co., Ltd.). The X-ray intensity profile with respect to 2T of the crystal grown from 2 at.%
gallium solutions is illustrated in Figs. 25(a) and (b). The Cu target was used as an X-ray
source. At a place of -7.0 mm, there was only one peak corresponding to the (220) diffraction
of InSb; however, two peaks appeared at -5 mm. The higher angle peak indicated with B
reduced in intensity and broadened in width than the peak of lower angle of A. The peak B
corresponded to InGaSbBi, because the diffraction peak shifted towards the higher angle with
an increase of the gallium composition ratio and vice versa. The peak from InSb was sharper
than that from the InGaSbBi grown crystal. It indicated that the quality of the InGaSbBi-
grown crystals was inferior to that of InSb. Fig. 26 represents lattice constant calculated from
Figs. 25(a) and (b). The lattice constant increased from the interface (distance = 0mm)
towards both the InGaSbBi-grown crystals and the seed. Since the lattice constant of InSb is
larger than that of GaSb, this suggested that the gallium atoms were substituted for indium
atoms in the InSb seed.
Fig. 25 The X-ray intensity profile with respect to 2T of the crystal grown from 2 at.% gallium
solutions. (a) seed, (b) grown crystal.
Fig. 26. Lattice constant calculated from Figs. 25(a) and (b).
4.4 Formation of Domains in the InSb Seed Crystal
On the formation of domains formed in the InSb seed crystal, their compositions were
measured. Respective area profiles of In, Ga, Sb and Bi measured by Energy Dispersive
Spectroscopy (EDS) (model: JED-2000, Nippon-denshi Co., Ltd.) are illustrated in Figs.
27(a)-28(d), respectively. White points denote the existence of each composition and the
intensity is proportional to the weight ratio of the composition. The intensity of bismuth was
strong inside a domain. The indium composition had roughly the same tendency as antimony,
although it was not completely zero inside of a domain. In the inside and the outside of the
domain, respective depth profiles measured by secondary-ion-mass spectroscopy (SIMS)
(model: ims4f, Cameca Co., Ltd.) are illustrated in Figs. 28(a) and (b). Although quantitative
values can not be decided due to the lack of standard samples, SIMS measurement has higher
sensitivity than EDS. Cs+ ions were implanted for 6000s, but the dipped depth was quite
different inside and outside of the domain. The domain was about 20 times easier to be
dipped than the outside of the domain. The atomic bond energy of Bi-Bi is 25.0 kcal/mol and
that of Sb-Sb is 30.2 kcal/mol [39]. Although the values of other atomic bonds such as In-In,
Ga-Ga, In-Ga, Ga-Sb were not known, the above results suggest that the bismuth-rich domain
had a weaker atomic bond than that that of InGaSbBi crystal. The counting rate of bismuth in
the domain reached to nearly 5x103, though it was extremely low in the outside of the
domain. On the other hand, the value of indium was not zero in the domain and it decreased
to about 1/10 compared with that of the outside. This corresponded to the result of Fig. 27(a)
which showed that a small amount of indium was incorporated in the domain. The
compositions of antimony and gallium were below the detection limit of the SIMS analysis.
This indicated that the domain was mainly composed of bismuth and indium. EPMA
measurement was performed in order to determine the atomic ratio of the compositions in the
domains. Fig. 29 shows composition ratios of Ga(x) and Bi(y) along the distance in the
domains in the crystal shown in Fig. 20. The gallium and antimony atomic ratios were almost
zero, and the composition ratios of indium and bismuth were nearly unity. This indicated that
the InBi compound was formed in every domain.
Fig. 27 Respective area profiles of (a) indium, (b) gallium, (c) antimony and (d) bismuth measured by
EDS.
Fig. 28 Composition profiles in the crystal shown in Fig. 22 analyzed by SIMS: (a) inside, (b) outside
of the domain.
Fig. 29 Composition ratios of gallium(x) and bismuth(y) along the distance in the domains in the crystal
shown in Fig. 20.
4.5 Growth of InSb1-yBiy and In1-xGaxSb Ternary Crystals
In order to investigate the incorporation of gallium and bismuth into the InSb seed more
clearly, InSb1-yBiy and In1-xGaxSb ternary crystals were grown and compared with the
composition profiles of InGaSbBi. The value of bismuth composition ratio y in the InSb1-yBiy
ternary crystal is measured along the growth direction by EPMA and is shown in Fig. 30. The
initial composition ratio in the solution was In : Ga : Sb : Bi = 50 : 0 : 25 : 25(at%). The
starting growth temperature and the growth period were 474qC and 182 h, respectively. Since
the equilibrium segregation coefficient of bismuth in InSb was less than unity, the value of y
increased gradually from 0.002 to about 0.005 in mole fraction with crystal growth. On the
contrary, the bismuth composition was not present in the InSb seed. These results indicated
that the bismuth atoms could not diffuse into the InSb seed if there was no gallium.
Fig. 30 Bismuth composition ratio y along the growth direction of the InSb1-yBiy ternary crystal
measured by EPMA. The initial composition ratio in the solution was In : Ga : Sb : Bi = 50 : 0 : 25 : 25.
Fig. 31 Gallium composition ratio x along the growth direction of the InxGa1-xSb ternary crystal
measured by EPMA. The initial composition ratio in the solution was In : Ga : Sb : Bi = 62.81 : 3.75 :
33.44 : 0(at.%).
The profile of gallium in the In1-xGaxSb crystal measured by EPMA is given in Fig. 31.
The initial composition ratio in the solution was In : Ga : Sb : Bi = 62.81 : 3.75 : 33.44 : 0
(at.%). The starting growth temperature and the growth period were 506qC and 86 h,
respectively. Since the seed contacted with the solution for 118 h before the start of growth,
the total period was 204h. The gallium ratio at the interface was 0.16, and it gradually
decreased toward the end of the InGaSb-grown crystal. The gallium was incorporated in the
InSb seed, and the value of gallium became zero at the distance of -7.8mm. From the well-
known Einstein`s equation L = (Dt)1/2 and the Boltzmann-Matano analysis [40], the value of
D was estimated to be on the order of 10-8 - 10-7 cm2/s, where L is diffusion length, D
diffusion coefficient of gallium, and t the time. This value was extremely high in contrast to
self-diffusion coefficients of indium and antimony in InSb [41] which ranged from the order
of 10-16 to 10-14 cm2/s in the temperature range from 475qC to 517qC. This result suggested
that gallium atoms were easily diffusing in the InSb seed crystal.
As seen in Fig. 30, bismuth atoms were not able to diffuse into the InSb seed during the
growth of InSbBi, and there were no domains in the InSb seed. In the growth of InGaSb,
gallium atoms were easy to diffuse in the seed, and domains of indium were formed. In the
case of InGaSbBi crystals, both bismuth and gallium were incorporated in the InSb seed, and
domains of InBi compound were precipitated in the seed. From these results, the contribution
of gallium was necessary for the diffusion of bismuth into the InSb seed. one of the
explanations for the mechanism of these processes is as follows. The incorportation of
gallium atoms in the InSb produces the excess indium atoms. This brings about the
precipitation of indium and the formation of dislocations. Therefore, the diffusion process
takes place quickly and the diffusion of bismuth becomes possible. Finally, InBi compound
are formed in the InSb seed crystal.
5 Numerical Simulation of Effect of Ampoule Rotation

for the Growth of Ingasb by Rotational Bridgman Method
5.1 Numerical Analysis
Fig. 32(a) shows a schematic RBM growth system and temperature profile in the furnace. The
temperature profile in the furnace is symmetric for the central axis. The temperature is the
highest at the center of the solution (TH) and the lowest at the seed and feed (TL). In-Ga-Sb
solution was put between GaSb seed and feed crystals, where the seed and the feed crystals
were cylindrical in shape and the In-Ga-Sb solution were semi-cylindrical. The ampoule
rotational rates were changed in the range 0-100 rpm. Crystal growth was initiated by moving
the ampoule toward the low-temperature region of the furnace. Fig. 32(b) shows a three-
dimensional model, which simulates In-Ga-Sb solution region in the RBM experimental
system. The GaSb seed and feed crystals formed in the semi-circle shape are put at both ends
of the semi-cylinder. The flat top plate of the semi-cylinder means the free surface of In-Ga-
Sb solution. The sidewall of the semi-cylinder is made of quartz glass. The In-Ga-Sb solution
region consists of 181520 segments. R and Z are lengths of radial and axial directions of
the growth solution reservoir. The symbol T is the angle of rotational direction. The
governing equations are given as follows:
NavierStokes equations
wu wu w wu wu 1 wP w 2u 1 wu 1 w 2u w 2u
v u v 2 (1)
wt wr r wT wz U wz wr r wr r 2 wT 2 wz 2
wv wv w wv wv w2 1 wP w 2 v 1 wv 1 w 2 v w 2 v v 2 ww (2)
v u v 2 g E 'T ,
wt wr r wT wz r U wr wr r wr r 2 wT 2 wz 2 r r 2 wT
ww ww w ww ww vw 1 wP w 2 w 1 ww 1 w 2 w 2 w 2v w
v u v 2 (3)
wt wr r wT wz r U r wT wr r wr r 2 wT 2 r 2 wT 2 r 2
Continuity equation:
1 w (vr ) 1 ww wu
0 (4)
r wr r wT wz
Energy equation:
wT wT w wT wT w 2 1 wT 1 w 2T w 2T
v u N 2 (5)
wt wr r wT wz wr r wr r 2 wT 2 wz 2
Fig. 32 (a) Schematic of rotational Bridgman method growth system and temperature profile in the
furnace. (b) three-dimensional model, which simulated the experimental system.
Here r, z and T are the respective coordinate axes in the radial, axial and angular
directions respectively. Symbols of u, v, w are axial, radial, and angular flow components of
the fluid velocity, respectively. P is the pressure, g the gravitational acceleration, the
volume expansion coefficient, T the temperature, Q the kinematic viscosity and is the
thermal conductivity.
The following boundary conditions for velocity and temperature were employed. The
non-slip condition was applied at the quartz glass wall and the solidliquid interface.
At the interface between the seed and the solution:
u = 0, v = 0, w = (rpm), T = 660(qC) (6)
At the quartz glass wall:
u = 0, v = 0, w =(rpm), T(z) = - az2 + 660 + (TH-TL) (7)
where a is a constant.
At the interface between the feed and the solution:
u = 0, v = 0, w = (rpm), T = 660(qC) (8)

In the numerical analysis, the rotational rates of the growth ampoule, the temperature
differences between the highest temperature TH and the lowest temperature TL in the furnace,
and the aspect ratios (axial length/radius length) were changed in the ranges 0-100(rpm), 10-
140C, and 1-4, respectively. The flow vector and temperature were calculated under a
steady-state condition. The numerical analysis was performed by the finite differential
method.
5.2 Analysis Results
Fig. 33 shows flow patterns of zr cross section in the In-Ga-Sb solution for the aspect ratios
of 1, 2 and 4 when the ampoule rotation rates are 0 and 50 (rpm). The temperature difference
(TH-TL) in the furnace was 30C. At 0rpm, natural convection generated at the bottom of the
In-Ga-Sb solution, because the flux heated at the middle of the ampoule wall moved upward
the free surface.The flow patterns separated into two large flows. One of them, in the left
region, moved up the center and down along the GaSb seed wall. The right one flowed
symmetrically in the opposite direction. With the increase of the aspect ratio, the flow
velocity increased slightly. But, flow pattern was the same as that of the other aspect ratios.
On the contrary, the ampoule rotation brought about the large effect on flow patterns. At
50rpm, centrifugal force due to the amoule rotation led to a change in the fluid flow. The flow
velocity became stronger near the interface of seed and feed crystals. The fluid was
concentrated at the center of the solution by centrifugal force, and moved up to the free
surface. The flow velocity increased with the increase of the ampoule rotation rate. In
comparison with the result of 0rpm, average flow velocity of 50rpm increased by about 100
times. As a result, the forced convection brought by the ampoule rotation was dominant in the
solution. At an aspect ratio of 4, flow pattern was different from that of 1 and 2, because the
flow generated by the centrifugal force near the solid-liquid interface was spiral flow and
became weak at the center of the solution. The flow direction was opposite to that of the
aspect ratio of 1 and 2.
Fig. 33 Flow patterns of zr cross section in the In-Ga-Sb solution for the aspect ratios of 1, 2 and 4,
where each aspect ratio was calculated under ampoule rotation rates of 0 and 50 (rpm), respectively.
Fig. 34 Isothermal curves of zr cross section in the InGaSb solution, where the temperature interval
was 2C.
Fig. 34 illustrates the isothermal curves of zr cross section in the In-Ga-Sb solution,
where the temperature interval was 2C. The temperature profiles changed dramatically by
increasing ampoule rotation rates. At 0rpm, the temperature was low at the GaSb seed and
feed sides, and high at the center of the ampoule wall. The spacings between isothermal
curves became narrow near the interface of the seed and feed crystals. On the contrary, the
temperature distribution became homogeneous in spite of the size of growth reservoir by
increasing the ampoule rotation rate.
Fig. 35 Temperature difference (T) at the solidliquid interface divided by (TH - TL) as a function of
aspect ratio when (TH - TL) was 30C.
Fig. 35 shows the temperature difference (T) at the solid-liquid interface divided by
(TH - TL) as a function of aspect ratio when (TH - TL) was 30C. At 0rpm, the temperature
difference increased with an increase of the aspect ratio. At 10rpm, however, the temperature
difference decreased at the aspect ratio of 2 and increase at 4. It may be because the flow
pattern was changed as shown in Fig. 33. At 50rpm, the temperature distribution in the
solution became homogeneous. It was due to the centrifugal force generated near solid-liquid
interface.
Fig. 36 T normalized by T0rpm as a function of rotation rate. Here, T and T0rpm indicate the
temperature differences when the ampoule rotation rates are and 0rpm, respectively.
Fig. 36 shows T normalized by T0rpm as a function of rotation rate. Here, T and

T0rpm indicate the temperature differences when the ampoule rotation rates are and 0rpm,
respectively. The normalized value at 100rpm was one-fifth smaller than that at 0rpm
irrespective of the value of (TH - TL). It indicated that the temperature distribution became
uniform with increasing the ampoule rotation rate. The stirring effect of ampoule rotation
enhanced with an increase of the rotation rates as the forced convection became dominant in
the solution.
6 Conclusion
Rotational Bridgman method was carried out to grow InSb1xBix, InxGa1xSb and InxGa1xSb1-
yBiy alloy crystals. Both flow patterns and temperature distributions were calculated to
investigate the solution convection in the rotational Bridgman method .
(1) InSb1-xBix crystals were grown on InSb seeds in the single crystalline state by the
rotational Bridgman method. The bismuth composition ratio changes from 0.015 at the initial
stage to 0.03 at the final stage due to segregation. By supplying the InSb component into the
solution during growth, homogeneous crystals were grown.
(2) InxGa1xAs single crystals with x values up to 0.12 were grown by the direct growth
on GaAs seed. The maximum thickness of the InxGa1xAs single crystals was 6.5 mm. The
indium composition ratios in the crystals were homogeneous due to the feeding GaAs sources
continuously into the In-Ga-As solution. However, the XRD FWHM increased with
increasing of indium composition. InxGa1xAs crystals were grown by the multi-step method.
As a result, the XRD FWHM decreased to the value of seed crystal. These results suggest that
the multi-step growth is useful to improve the quality of ternary crystals.
(3) InxGa1-xAs ternary alloy bulk crystals were grown on a GaAs (100) seed by the
rotational Bridgman method. Experiments have been conducted on various ampoule-moving
rate and rotation rate in order to find out the growth rates of InGaAs crystals. The thermal
pulse technique was employed for introducing the impurity striations, which was helpful in
measuring the growth rates for different moving and rotation rates. It has been found that
different moving rate and rotation rates did not affect the composition but had affected largely
on the growth rate. Since the growth rate is weakly connected with the pulling rate, it can be
concluded that heat transfer did not control the process, but it was controlled by mass transfer
and was similar to growth from solution.
(4) In1-xGaxSb1-yBiy (0<xd0.21,0<yd0.005) single crystals more than 10 mm in length

were grown by the rotational Bridgman method. Owing to the segregation phenomena, the
bismuth composition ratio increased and the gallium ratio decreased as the crystal grew. The
lattice constant changed according to the gallium composition profile. In the InSb seed, both
gallium and bismuth components were incorporated and InBi domains were formed. For
comparison, InSb1-yBiy (0<yd0.05) and In1-xGaxSb (0<xd0.16) ternary crystals were grown.
Although gallium diffused into the InSb seed during the growth of InGaSb, bismuth was not
incorporated during the InSbBi growth. The contribution of gallium was necessary for the
diffusion of bismuth into the InSb seed crystal. The incorportation of gallium produced the
excess indium and as a result InBi domains were formed.
(5) To investigate the solution convection for the rotational Bridgman method, both flow
patterns and temperature distributions were calculated by solving the three equations in the
three-dimensional analysis: NavierStokes, continuity and energy. By increasing the ampoule
rotation rate from 0 to 100rpm, the flow velocity in the In-Ga-Sb solution increased and the
temperature distribution became uniform in spite of changing the temperature difference in
the furnace and the size of the growth reservoir. The stirring effect of ampoule rotation
enhanced with an increase of the rotation rates as the forced convection became dominant in
the solution.
Acknowledgment
The authors wish to thank Profs T.Sukegawa and A. Tanaka for the four-crystal X-ray
diffractometry, Prof.S.Fuke and Mr.K.Kuwabara for SIMS, Prof..K.Murakami for scanning
electron microscopy, Mr.T.Koyama and Mr.K.Katuno for EPMA. We also thank
Mr.Y.Momose for preparing the growth ampoules. This research was supported in part by a
Grant-in-Aid for International Scientific Program and by a part of Grand Research for Space
Utilization promoted by NASDA and Japan Space Forum.
References
[1] K.Nakajima, T.Kusunoki and C.Takenaka, J. Crystal Growth 113 (1991) 485.
[2] T.Hibiya, H.Watanabe, H.Matsumoto and N.Iwata, J. Electrochem. Soc. 134
(1987) 981.
[3] K.J.Bachmann, F.A.Thiel, H.Schreiber, Jr., and J.J.Rubin, J.Electron.Mater. 9
(1980) 445.
[4] M.Ohmukai, H.Naito, M.Okuda and K.Kurosawa, J. Crystal Growth 152 (1995) 247.
[5] K.J.Bachmann, F.A.Thiel and S.Ferris. J. Crystal Growth 43 (1978) 752.
[6] V.K.Dixit, Bhavtosh Bansal, V.Venkataraman, H.L.Bhat and G.N.Subbanna,
Appl.Phys.Lett 81(2002)1630.
[7] R.Singh, A.F.Witt and H.C.Gatos, J.Electrochem. Soc. 121 (1974) 380.
[8] K.M.Kim, J. Crystal Growth 44 (1978) 403.
[9] A.M.Jean-Louis, B.Ayrault and J.Vargas, Phys. Status Solidi 34 (1969) 341.
[10] K.Nakajima, T.Kusunoki and K.Otsubo, J. Crystal Growth 173 (1997) 42.
[11] S.Kodama, Y.Furumura, K.Kinoshita, H.Kato and S.Yoda, J.Crystal Growth 208
(2000) 165.
[12] H.Sheibani, S.Dost, S.Sakai and B.Lent, J.Crystal Growth 258 (2003) 283.
[13] M.Kumagawa, H.Oka, N.V.Quang and Y.Hayakawa, Jpn. J. Appl. Phys. 19 (1980) 753.
[14] Y.Hayakawa, Y.Sone, K.Tatsumi and M.Kumagawa, Jpn. J. Appl. Phys. 21
(1982) 1273.
[15] Y. Hayakawa and M.Kumagawa, Jpn. J. Appl. Phys. 22 (1983) 206.
[16] Y.Hayakawa, Y. Sone, F. Ishino and M.Kumagawa, Jpn. J. Appl. Phys. 22 (1983) 1069.
[17] Y.Hayakawa and M.Kumagawa, Crystal Res. Technol. 20 (1985) 3.
[18] Y.Hayakawa and M.Kumagawa, Jpn. J. Appl. Phys. 24 Suppl. 24-1 (1985) 166.
[19] T.Tsuruta,Y.Hayakawa and M.Kumagawa, Jpn.J.Appl.Phys. 27 Suppl.27-1 (1988) 47.
[20] T.Tsuruta,Y.Hayakawa and M.Kumagawa, Jpn.J.Appl.Phys.28 Suppl.28-1 (1989) 36.
[21] T.Tsuruta,Y.Hayakawa and M.Kumagawa, Jpn.J.Appl.Phys. 31 Suppl.31-1 (1992) 23.
[22] M.Kumagawa, T.Tsuruta, N.Nishida, J.Ohtsuki, K.Takahashi, S.Adachi and
Y.Hayakawa, Cryst.Res.Technol. 29 (1994) 1037.
[23] M.Kumagawa, M.Tamaki and Y.Hayakawa, Jpn.J.Appl.Phys. 26 (1987) 180.
[24] Y.Hayakawa, M.Nagura and M.Kumagawa, Semicond.Sci.Technol. 3 (1988) 372.
[25] M.Nelson. RCA Rev. 24 (1963) 603.
[26] M.Kumagawa, T.Ozawa and Y.Hayakawa, Appl. Surface Sci. 33/34 (1988) 611.
[27] T.Ozawa, Y.Hayakawa and M.Kumagawa, J.Crystal Growth 109 (1991) 212.
[28] T.Ozawa, Y.Hayakawa and M.Kumagawa, J. Crystal Growth 115 (1991) 728.
[29] T.Ozawa, Y.Hayakawa, K.Balakrishnan, F.Ohonishi, T.Koyama and M.Kumagawa, J.
Crystal Growth 229 (2001) 124.
[30] Y.Hayakawa, M.Ando, T.Matsuyama, E.Hamakawa, T.Koyama, S.Adachi,
K.Takahashi, V.G.Lifshits and M.Kumagawa, J.Appl.Phys. 76 (1994) 858.
[31] T.Ozawa, Y.Hayakawa, K.Balakrishnan, F.Ohonishi, T.Koyama and M.Kumagawa,
J.Crystal Growth 229 (2001) 124.
[32] Y.Hayakawa, T.Ozawa, M.Haris and M.Kumagawa, J.Crystal. Growth (2005) in press.
[33] T.Ozawa, N.Ishigami,Y.Hayakawa, K.Koyama and M.Kumagawa, Computer Modeling
in Engineering & Sciences,10 (2000) 1.
[34] T.Ozawa, Y.Hayakawa, K. Balakrishnan and M.Kumagawa, J.Crystal Growth 237-239

(2002) 1692.
[35] B.Joukoff and A.M.Jean-Louis, J.Crystal Growth 12 (1972) 169.
[36] P.Grnert, Crystal Research and Technology 15 (1980) 627.
[37] A.N.Danilewsky and K.W.Benz, J. Crystal Growth 106 (1990) 273.
[38] K.Nakajima, J.Crystal Growth 144 (1991) 633.
[39] L.Pauling, The Nature of the Chemical Bond (Cornell University Press, Ithaca,
NY,1960), p.72
[40] P.D.Shewmon, Diffusion in Solids (McGraw-Hill, New York, 1963) 29.
[41] D.L. Kendall and R.A.Huggins, J. Apply.Phys. 40 (1969) 2750.
Chapter 2
FORMATION OF GROWN-IN MICRODEFECTS

IN DISLOCATION-FREE SILICON MONOCRYSTALS
V.I. Talanin*, I.E. Talanin

1
Institute of State & Municipal Government, Zhukovskii str. 70B,
69002 Zaporozhye, Ukraine
Abstract
The paper presents a general scheme that depicts the formation and transformation
mechanisms of the grown-in microdefects in dislocation-free silicon single crystals grown by
the floating-zone (FZ-Si) and Czochralski (CZ-Si) methods as a function of a crystal growth
rate. The distribution patterns and physical nature (the sign of the lattice strain) of grown-in
microdefects FZ-Si and CZ-Si crystals were studied by selective etching and transmission
electron microscopy. A heterogeneous mechanism of formation and transformation of grown-
in microdefects are proposed.
It is experimentally established and confirmed by thermodynamic analysis that in single
FZ-Si and CZ-Si crystals close to the crystallization front the recombination process of
intrinsic point defects is hampered by a recombinational barrier. ooling-induced
decomposition of the oversaturated solid solution of point defects in silicon follows two
mechanisms: vacancy-type and interstitial-type. Therefore vacancies and self-interstitials find
drains in the form of atoms of background impurities like oxygen and carbon. Depending on
growth conditions (e.g. parameter V/G, where V is a crystal growth rate; G is an axial
temperature gradient) the silicon crystals grow in vacancy-interstitial and (or) interstitial
regimes. The parameter V/G is determined experimentally with allowance for crystal
diameter, cooling rate and radial temperature gradient.
In vacancy-interstitial growth regime of FZ-Si and CZ-Si crystals during disintegration of
the supersaturated vacancies solid solution the oxygen-vacancy aggregate occurs causing SiO2
presipitates to be formed. During disintegration of the supersaturated self-interstitials solid
solution the interstitials-carbon aggregate occurs causing the aggregates, which consist of self-
interstitials, atoms of carbon and oxygen, to be formed. During the formation of such
aggregates the carbon atoms act as catalysts.
The point defects aggregation in vacancy-interstitial and interstitial growth regimes under
certain thermal growth conditions during the crystal cooling may result in the formation of
secondary defects near the primary oxygen-vacancy and carbon-interstitial aggregates, namely
*
E-mail address: v.i.talanin@mail.ru
32 V.I. Talanin and I.E. Talanin
octahedral vacancy microvoids and interstitial dislocation loops accordingly. It is

experimentally established that A-microdefects may occur in the course of two different
formation mechanisms: condensations of self-interstitials and generation of interstitial
dislocation loops, which is called by a field of elastic deformation all around the presipitates.
As is established, the nature and behaviour of processes of the grown-in microdefects
formation in FZ-Si and CZ-Si crystals are identical. The uniform classification of grown-in
microdefects is offered and the heterogeneous mechanism of formation and transformation of
grown-in microdefects in dislocation-free silicon single crystals is confirmed experimentally.
On the ground of experimental results the physical classification of grown-in
microdefects is offered. Such a classification follows from the heterogeneous mechanism of
formation and transformation of grown-in microdefects.
Keywords: grown-in microdefects, mechanism of formation, intrinsic point defects, oxygen,

carbon, oxygen-vacancy aggregates, carbon-interstitial aggregates, physical
classification.
Introduction
In the modern world the magnification of industrial production is impossible without
electronics engineering. Material basis of electronics engineering it is dislocation-free silicon
monocrystals. Search of other materials and assimilation of their production its only adding
to silicon.
Achievements reached in the area of hyper pure silicon single crystals growing
highlighted the further problem to study the peculiarities of the defect formation in such a
perfect structure medium. It becomes of great importance now as microelectronics turned to
megabit and gigabit integrated circuits fabricated by submicron-level technologies. Successful
fabrication of high-quality single crystals and device structures with controlled and
predictable parameters depends substantially on how a point defect ensemble is controlled.
The structure imperfections named as grown-in microdefects are formed during high-
temperature growing and further cooling of dislocation-free single Si crystals. Presently, any
local lattice imperfection from a few tens angstroms to few micrometers is deemed as grown-
in microdefect. According to the generally accepted geometry (size) classification the grown-
in microdefects belong to the transient class of defects between point and linear ones.
Complex of problems which are decided during modern researches in the field of physics,
chemistry, material, both technology of semiconducting silicon and structures on it basis is
huge. The problem of grown-in microdefects in dislocation-free silicon monocrystals is
central in this complex. In particular from our view point the solution of a grown-in
microdefects problem is a key to a solution of a problem of thermal donors and thermal
acceptors formation in monocrystalline silicon.
First data of etching pits in dislocation-free single Si crystals refer to the late sixties of
last century [1]. A.J.R. de Kock introduced fist classification for empty etching pits, having
identified them as clusters. According to size of etching pits he divided clusters into two
types: -clusters (-microdefects) and -clusters (-microdefects) [2, 3]. It should be noted
that till the late seventies the grown-in microdefects have been studied in small-scale FZ-Si
crystals (26 to 30 mm in diameter). It was established that -microdefects observed in FZ-Si
single crystals of 30 mm in diameter had a striated distribution at growth rate V = 1 to 3.5
mm/min and B-microdefects had a striated distribution at V d 4.5 mm/min [4]. When -
Formation of Grown-in Microdefects in Dislocation-Free Silicon Monocrystals 33
microdefects size increased up to a certain limit (about 40 Pm at V = 1 mm/min), dislocations

appear at single silicon crystals [5]. It was shown that fluctuation in temperature and presence
of impurities (especially, carbon [6, 7] and oxygen [8, 9]) are basic factors responsible for the
mechanism of - and -microdefects formation.
When crystals are grown at sufficiently high rates, a new type of microdefects was
observed as uniform distribution and could be identified as dim regions after preferential
etching and decoration [10]. These defects were classified as C- or D-microdefects depending
on the their macrodistribution images. Both types were visualized at the regions with uniform
defect distribution of high density and, according to the authors of the paper [10], differed
from each other by a distribution geometry. The uniform D-microdefects distribution is
focused in the form of a channel in the central part of the crystal. Currently C-
microdefects often referred to as V-distribution (or V-band) and W-distribution (or
W-band). However C-microdefects are also observed as rings or circles of irregular shape
located at the most extensively cooled parts of the crystal. Therefore the term C-
microdefects has a general meaning.
In small-scale silicon crystals grown by Czochralski method (CZ-Si) A-, B- and D-
microdefects, identical to defects in FZ-Si, were also observed using preferential etching
technique [11, 12]. Undoped or low-doped single crystals of 50 mm in diameter, which were
grown at V d 1 mm/min, contain - -microdefects in a striated distribution. Furthermore,
at V ~ 1 mm/min - and -microdefects are formed in concentrations comparable to their
concentrations in silicon single crystals grown by floating-zone melting method. At rate of V
t 2 mm/min the - and -microdefects generation is fully suppressed.
It should be noted that a discrepancy in classifying D-microdefects emerged. As stated
above, first the term D-microdefects was introduced in 1977 by Sheikhet et all [10] for
grown-in microdefects in FZ-Si (V = 6 mm/min), which have uniform distribution in the form
of a channel in the central part of the crystal. Quite possible, Roksnoer did not know of the
paper [10] and in 1981 determined as D-microdefects such defects, which are generated also
as uniform distribution in the form of a channel in the central part of the crystal FZ-Si (but
at V = 7 to 8 mm/min) [13]. On the basis of X-ray topography results coupled with copper
decoration, researchers in [13] made an assumption that the D-microdefects in FZ-Si are
vacancy-type. It must be allowed for that in the theoretical model of the formation mechanism
of microdefects in FZ-Si [14] Voronkov defines D-microdefects in accordance with
classification [13]. The fact that uniform distribution of D-microdefects in the form of a
channel in the central part of the crystal (in accordance with classification [10]) is not
observed in CZ-Si makes the confusion in classification worse. Therefore, most authors still
use a definition D-microdefects in accordance with [13]. Nevertheless, we believe that in
the context of history and experimental results as for grown-in microdefects origin, such a
definition of D-microdefects is erroneous.
Fig.1. Grown-in microdefects in FZ-Si and CZ-Si crystals, selective etching: a) A-microdefects (FZ-Si,
V = 2 mm/min), plane (112); b) B-microdefects (FZ-Si, V = 4 mm/min), plane (111); c) D-microdefects
(FZ-Si, V = 6 mm/min), plane (111); d) A-microdefects and (I+V)-microdefects (CZ-Si, V = 2.5
mm/min), plane (112); e) (I+V)-microdefects (FZ-Si, V = 8 mm/min), plane (111); f) C-microdefects
(FZ-Si), plane (112).
What grounds are for such a conclusion? Later, using the transmission electron
microscopy (TEM) we established that in FZ-Si crystals at V = 6 mm/min only interstitial-
type defects are generated, while at V = 7 to 8 mm/min in FZ-Si crystals the interstitial-type
and vacancy-type defects co-exist simultaneously [4, 15, 16]. Furthermore, we showed
experimentally that at Vcrit a 6.5 mm/min in accordance with ratio V/G (where V is a crystal
growth rate; G is an axial temperature gradient) the mode of FZ-Si crystal growth is changing
(from interstitial-type into vacancy-interstitial type and vise versa) [4, 15]. Thus, two quite
different types of grown-in microdefects having uniform distribution were referred the same
as to D-microdefects. Here we use the classification, which historically was first and more
comprehensive [10]. Based on the classification [10] and TEM research results we define in
this paper the grown-in microdefects in FZ-Si (V = 7 to 8 mm/min) having uniform
distribution in the form of a channel in the central part of the crystal as (I+V)-
microdefects [4, 15, 16]. On Fig.1 the macrodistribution pictures of main kinds grown-in
microdefects in FZ-S and CZ-Si crystals are represented.
New types of grown-in microdefects are observed in large-scale silicon crystals (over 100
mm in diameter). In particular, presently thermal oxidation typically leads to a ring-shaped
distribution in plane (111) of so-called oxidation-induced stacking faults, OSF-ring [17,
18]. Researches in a growth rate influence on a distribution of grown-in microdefects in
single CZ-Si crystals of 150 mm in diameter revealed that OSF-ring is observed in crystals
grown at a moderate growth rate (0.7 to 0.8 mm/min), and absent in crystals grown at growth
rates 0.4 mm/min and 1.1 mm/min [19]. As is shown, stacking faults are formed around the
plate-like oxygen precipitates [18]. When evaluating future trends in growing crystals of 300
mm and 400 mm in diameter [20], it should be noted that required growth rate decreasing (up
to 0.55 mm/min for the crystal of 300 mm in diameter, which is twice less comparing to the
crystal of 200 mm in diameter) below the critical rate in order to diminish the thermal stress
in the crystal makes the task to fabricate ingots without OSF-ring rather complicated.
Outside OSF-ring the interstitial dislocation loops (A-microdefects) are observed [21,
22]. Inside OSF-ring there are defects, which according to observation techniques were
named as FP-defects (flow pattern defects), COP-defects (crystal-originated particles) and
LST-defects (light-scattering tomography defects) [23]. As is shown, all these defects are
different forms of the same defect presentation obtained by various techniques, i.e. vacancy
complexes, which are formed inside OSF-ring [24]. With the use of TEM the form and size of
LST-defects were determined (100 to 300 nm) [25, 26]. Although TEM observation fails to
determine the sign of the defect-induced lattice imperfection, on the basis of a smaller content
of oxygen in the center of the defect as compared to the periphery and defect contrast image,
was assumed that LST-defects constitute microvoids, i.e. they are vacancy-type. This enabled
researchers to interpret all the defect types, identified by different techniques, as a one type of
defects identical to D-microdefects in FZ-Si and CZ-Si, formed as a result of homogenous
nucleation [25-28]. Unfortunately, these results strengthened the erroneous opinion among
researchers that D-microdefects in FZ-Si and CZ-Si crystals are microvoids. It is generally
assumed that depending on V/G ratio only interstitial-type dislocation loops (A-microdefects)
or only microvoids (D-microdefects) are formed in large-diameter crystals. It is clear and
simple approaches as in particular large microdefects either interstitial-type (A-microdefects)
or vacancy-type (microvoids) considerably affect both electrophysical and mechanical
properties of single-crystal silicon and discrete devices and integrated circuits. However, we
believe such an approach is exceptionally technological because it does not allow for the
physical nature of defect formation in the dislocation-free single crystal silicon.
There are several models of grown-in microdefects formation differing from each other
both initial physical concept and level of validity. Every model is based on types and origin of
dominating point defects in the growing crystal, which then are transformed into microdefects
during cooling. Basic models of grown-in microdefects formation include the following:
equilibrium interstitial-type model [7], non-equilibrium interstitial-type model [29], local
melting model (drop model) [30], vacancy-type model [31], vacancy-interstitial type model
[32, 33] and recombination-diffusion model [14, 27]. At present the recombination-diffusion
model is the most well-grounded and optimal from the point of its quality. It is distinguished
by compiling basics of other models. Furthermore, as many researchers acknowledged, it is
convenient and plain to divide the grown-in microdefects simply to A-microdefects and
microvoids on the basis of V/G ratio. The point is that V/G ratio allows correlating the lattice
imperfection with the thermal growth conditions. Adjusting thermal growth conditions one
can calculate the crystal defect structure. Voronkovs common model is based on the
following: 1) there is a fast recombination between intrinsic point defects at temperatures
close to melting point; 2) there are areas with only interstitial-type (A- and B-microdefects) or
only vacancy-type microdefects (microvoids or D-microdefects); 3) there is a critical
parameter Ccrit = V/G. When V/G < Ccrit, self-interstitial atoms dominate in the crystal
(interstitial-type crystal growth) and when V/G > Ccrit, vacancies prevail in the crystal
(vacancy-type crystal growth). The effect of V/G ratio is reversible and repeatable: reduce
V/G and find A-microdefects and no microvoids; increase V/G and find microvoids and
no A-microdefects. To apply this criterion one should measure or calculate the value of G.
However this model does not allow for the affect of impurities (except for oxygen) and does
not reflect a real physical nature of formation processes of the grown-in microdefects in
silicon single crystals. Recently, Voronkov reported he took into consideration the other
impurities in addition to oxygen, which effect the formation of microdefects [34]. He
suggested that V/G ratio is shifted by impurities. Some impurities, such as oxygen, nitrogen,
and hydrogen, trap vacancies and cause a downward shift in the critical V/G ratio (and also a
fast increase in the fraction of trapped vacancies). Other impurities, like carbon, trap self-
interstitial, and cause an upward shift in the critical V/G ratio (and also an increase in the
fraction of impurity interstitials) [34].
The drawbacks of the existing classification (and the drawbacks of the theoretical models
based upon it) are in that it does not provide any information about the sign of lattice
imperfection, which depends on the microdefect type. The sign of defect-induced lattice
imperfection (often defined as the microdefect origin) bears an implicit information of the
defect chemical composition. At the same time it allows to determine the mechanisms,
through which the grown-in microdefects form and transform. It was exactly the absence of
this parameter that caused confusion in identification and classification of D-microdefects in
FZ-Si and CZ-Si.
1 Experimental
The comprehensive study aiming to reveal the mechanism of the grown-in microdefects
formation and transformation is required to solve the problem of the defect formation in
dislocation-free single crystal silicon. Following this way, we have been researching the
distribution images and images of individual properties (size, form, sign of lattice
imperfection) of grown-in microdefects in FZ-Si and CZ-Si crystals, which were grown at
different thermal conditions. Special attention was drawn to study the FZ-Si crystals, which
are hyper pure crystals (most of them have contents of oxygen and carbon less than
51015 cm-3). The principal objective of study was to distinguish the operative body (basis) of
the mechanism of the grown-in microdefects formation, effecting which (i.e. changing in
thermal growth conditions, doping, thermal processing) one can control the process of defect
formation in the dislocation-free single crystal silicon.
We have conducted a series of experiments with the FZ-Si and CZ-Si using preferential
etching and transmission electron microscopy. The silicon crystals of 30 mm (FZ-Si) and 50
mm (CZ-Si) in diameter were studied. These diameters were chosen because all the known
types of the grown-in microdefects are formed in crystals of such diameters when thermal
conditions of crystal growth vary. FZ-Si crystals were not doped, oxygen and carbon
concentrations were less than 51015 cm-3, the number of passes varied from 2 to 10, U =
(24)103 :cm. Concentration of the electrically active impurities were less than 1012 cm-3.
CZ-Si n-type crystals were with U = 1050 :cm, oxygen concentration ~ (47)1017 m-3,
carbon ~ (57)1016 m-3. The crystals were grown at constant growth rates ranging from 1
to 9 mm/min (for FZ-Si) and 0.5 to 3 mm/min (for CZ-Si). Several crystals were obtained
under specially programmed growth rate change. To freeze-in the initial stages of formation
of grown-in microdefects we conducted quenching of crystals. To study the reasons of
formation of microdefects the crystals were subjected to special thermal processing.
Furthermore, to confirm the influence of oxygen on formation of microdefects, the
experiments with intentional doping of crystals by oxygen were conducted. The oxygen
doping was carried out from a gas phase at adding the steams of H2O2 to the growth chamber.
Sirtl etch was used to identify microdefects of different types and their distribution
images. The inside-outside contrast method, black-white contrast method and defocused dark-
field image method (2.5D) were used to determine the sign of lattice imperfection. Electron
microscope with 100 kV acceleration voltage was employed that excludes radiation defect
infusion [35, 36]. To avoid contamination of the surface by alkali atoms and hydrogen (which
are known to penetrate while etching to a depth of several micrometers), the samples for the
TEM study were not subjected to that procedure. Knowing the etch pattern for a certain
silicon wafer, one can rather exactly cut the samples for the TEM study from a selected place
of a near-by wafer, thus the identity of microdefects distribution over the cross-section of
both wafers being ensured.
2 Results
2.1 Scheme Formation and Transformation Mechanism of Grown-in
Microdefects in Silicon Single Crystals
Large number of crystals studied (more 60 ingots FZ-Si and more 20 ingots CZ-Si), which
were obtained under various thermal growth conditions, enabled us to make a schematic
representation of the microdefect distribution in FZ-Si and CZ-Si under the changing crystal
growth rate (Fig.2).
Fig 2. Schematic formation and transformation mechanism of microdefects in silicon single crystals,
plane (112): a FZ-Si crystal 30 mm; b CZ-Si crystal 50 mm; Vcrit, Vcrit 2 crystal growth rates
defined experimentally, at which the vacancy-type microdefects appear (disappear), mm/min; Vr
crystal growth rate, at which the remelting phenomenon is suppressed; Vcrit 1 crystal growth rate
calculated theoretically, at which vacancy-type microdefects appear while the interstitial-type defects
disappear (according to Voronkovs model), mm/min.
As one can see in Fig.2a, with gradual increase of growth rate, one type of microdefects
in FZ-Si crystals is replaced progressively with the other type. A-microdefect large swirls are
replaced with smaller swirls of B-microdefects. After the rate Vr reaches a certain value (at
which the remelting phenomenon is suppressed) B-microdefects are replaced with uniform D-
microdefects distribution. For FZ-Si we determined the adjusted value of rate Vr to be ~ 4.5
mm/min. Our study results showed that above a certain critical growth rate Vcrit the (I+V)-
microdefects (defect-free region on Fig.2a) begin to appear in crystals, while before Vcrit only
D-microdefects could be detected. If V > Vr, D-microdefects usually converge in a channel,
which afterwards at V t Vcrit begins to diverge towards the periphery of the crystal (C-
microdefects in accordance with classification [10]). If the crystal growth rate continues to
rise, the ring size diminishes (in plane (111)), and at V ~ 8 mm/min the ring of D()-
microdefects disappears. Fig.1c and Fig.1e shows the difference between the images of FZ-Si
crystals selective etching at V = 6 mm/min (D-microdefects) and at V = 8 mm/min ((I+V)-
microdefects). Thus, as the crystal growth rate decreases, the grown-in microdefects are
continuously transformed.
As for SZ-Si crystals growing, the growth rates are moved to smaller values. The
appearance of uniform defect distribution is typical for the values: 1.5 mm/min < V < 2
mm/min. Though, according to Voronkovs model, interstitial-type A- and B-swirls should
appear only at V < Vcrit 1, where Vcrit 1 is defined as a theoretical value characterised by the
vacancy-type microdefects appearance and the interstitial-type microdefects disappearance
[14, 27]. Data in Fig.2b makes it clear that D-microdefects are not detected in CZ-Si in
channel distribution. This may be concerned with the following. The change in thermal
growth conditions causes the suppression of remelting phenomenon (this phenomenon is
responsible for striated distribution of A- and B-microdefects) that appears in FZ-Si at
interstitial-type growth (i.e. at Vr < Vcrit) and in CZ-Si at vacancy-interstitial-type growth (i.e.
at Vr > Vcrit). Furthermore, oxygen and carbon are present in CZ-Si crystal in much greater
concentrations than in FZ-Si. Therefore, oxygen precipitation is in close interrelation with the
CZ-Si defect structure and leads to interstitial-type dislocation loops (A-microdefects)
appearance outside the D(C)-microdefects distribution (i.e. outside the D(C)-microdefects
ring in plane (111)).
Consequently, as the crystal growth rate increases, the crystallisation front curvature
decreases, thus conducing to that the plane (111) emerges to the crystallisation front and the
channel and ring distribution of D-microdefects are formed. This process proves the
heterogeneous nature of microdefect generation. Decreasing the crystallisation front curvature
results in the temperature gradient reducing. However, in its turn, the axial temperature
gradient is not constant through the entire crystal diameter. Thus, crit = V/G(r) (where G(r)
generally increases from the crystal center to its edge) [37], that leads to generating V-shaped
distribution of D()-microdefects in plane (112).
Growth rate values, at which the ring-shaped distribution in plane (111) of D(C)-
microdefects is observed, range within 1 to 2 mm/min for CZ-Si crystals of 80 mm in
diameter and within a 0.5 to 0.8 mm/min for CZ-Si crystals of 300 mm in diameter [20]. The
ring-shaped distribution of D(C)-microdefects can be defined as ring OSF-defects after
thermal processing. Ravi suggested [38], that microprecipitates SiO2 as well as
microprecipitates SiC are responsible for generation of stacking faults.
We regard the defects within the area of ring-shaped distribution as ones similar to D(C)-
microdefects in FZ-Si crystals. Therefore, varying the thermal growth conditions results in
that D(C)-microdefects in CZ-Si crystals are observed only in the ring-shaped distribution.
The ring-shaped uniform distribution of D(C)-microdefects serves as the boundary between
the interstitial microdefects area in the striated distribution (A-microdefects) and (I+V)-
microdefects. Ring-shaped distribution of D(C)-microdefects is formed as a result of Si
crystal thermal growth conditions and after the plane (111) is transposed to the crystallization
front (edge effect). V-type distribution of D(C)-microdefects in plane (112) is the
experimental representation of V/G ratio. Obtained experimental results allow suggesting that
the defect formation mechanism (and classification of the grown-in microdefects) in FZ-Si is
identical to that in CZ-Si.
2.2 Quenched Crystals Study
It is essential to study initial stages of defect formation in order to determine the mechanism
of grown-in microdefects formation and transformation. With this purpose the FZ-Si single
crystals grown at various growth rates (2.0, 3.0, 6.0, 9.0 mm/min) were subjected to
quenching. One of most effective crystal quenching methods was applied, namely melting
zone decantation, when at a certain moment a zone is blown with the directed argon flow. The
temperature of grown-in microdefects was determined experimentally and by theoretical
calculation. Neimark et al [39] measured thermal fields, which take place during growth of
silicon single crystals, by the thermocouple. They suggested the following empirical formula
to describe the dependence of an axial temperature gradient on the growth rate of crystal:
dT/dL = 10 + (L - 16)2exp(-61.2V 0.28), (1)
where L is a distance from crystallization front, cm; V is a crystal growth rate, cm/s. The
experimental values well coincide with values calculated from (1). The error between the
calculations from (1) and experimental values does not exceed r2%. Our experiments
allowed defining the temperatures of grown-in microdefects formation.
When studying a detachment surface, we established that microdefects have been not
detected directly at the crystallization front because of introducing dislocations due to a
thermal impact [4]. Therefore it is hard to define the exact starting moment of the formation
process of B- and (I+V)-microdefects. Although we can assume that at the given growth
conditions the - and (I+V)-microdefects are formed as soon as cooling starts due to
complexation of silicon interstitial atoms, vacancies and impurity atoms. Distance from the
front of crystallization, where the dislocations are observed, is not more 1 to 3 mm and
theoretically tends to zero (Fig.3).
Table 1. The temperature of grown-in microdefects formation
Growth rate, Conditions Type of grown- Distance from the front Temperature of
mm/min of treatment in microdefects of crystallization, mm formation, r 20
2.0 Quenching 23 = 1373
3.0 Quenching - = 1653
6.0 Quenching D 26 D = 1423
6.0, 9.0 Quenching I+V - TI+V=1653
Fig.3. Schematic distribution grown-in microdefects in FZ-Si crystals after quenching: a) V = 2

mm/min; b) V = 3 mm/min; c) V = 6 mm/min; d) V = 9 mm/min.
Experiments in crystal quenching demonstrate that the (I+V)-microdefects are first to

appear near the crystallization front and then during the cooling the D()-microdefects, B-
microdefects and A-microdefects are formed. It is worth to note that Z-Si and FZ-Si crystals
thermal processing also gives rise to transformation of interstitial-type microdefects in
direction as follows: D()-microdefects o B-microdefects o A-microdefects [16, 40].
When quenching the FZ-Si crystals grown at V = 6 mm/min, the so-called defect-free
area between the crystallyzation front and D-microdefects area is generated. The TEM study
imaged this defect-free area as black-white spots (similar to Fig.4d). Such a contrast is known
to be caused by dynamic conditions of electron reflection from crystallographic planes near
the defect. In the thin parts of samples these defects are imaged as black spots for kinematic
conditions of electron reflection. It should be noted that kinematic images of defects are more
useful for the evaluation of the size of these defects. For the majority of the observed defects
it is 3 to 7 nm. The concentration of defects found by electron microscopy is a 4.51013 m-3.
The view of defect images did not yet allow defining the geometric shape of these defects. To
state the physical nature of the observed microdefects (the sign of the imperfection around it),
their black-white contrast with all the peculiarities of their behaviour has been investigated in
dependence on the depth of the microdefect position in thin crystals [41]. For an
unambiguous identification of observed defects the method 2.5D has been applied
(observation of stereocouples of underfocused and overfocused images) [42]. These methods
reveal that the defect-free area contains interstitial-type and vacancy-type defects. Defects
have similar image contrast, but vacancy defects on the whole are somewhat greater in size
than defects of interstitial type. It should be noted that area with D-microdefects contains only
interstitial-type defects a half as large again that size of (I+V)-microdefects, while the
concentration of D-microdefects is three times less than the concentration of (I+V)-
microdefects.
In addition, with use of study made for FZ-Si crystals grown at V = 9 mm/min and
then quenched, the interstitial-type and vacancy-type microdefects have been revealed near
the crystallization front in concentrations comparable with each other. Thus, defect-free areas
near the crystallization front indeed contain defects of both deformation types, which should
be interpreted as quite new and rather substantial fact. Furthermore, it makes clear from these
experiments why we introduce the term (I+V)-microdefects into the classification of
grown-in microdefects.
Experimental results obtained in the quenched crystals study demonstrate that the
recombination process of intrinsic point defects near the crystallization front is hindered due
to the recombination barrier. Microscopic model of such a barrier was developed in detail in
papers by Gosele [43-45]. The root of model lies in that the configuration of intrinsic point
defects at high temperatures defines the barrier's dependence on temperature. As is assumed
within the models framework, at high temperatures the self-interstitials and vacancies are
extended through several atomic volumes (11 atoms occupies 10 cells), i.e. there is a
disordered area around the point defect, which is isotropic-extended up to the atoms of the
second coordination sphere. Recombination only occurs if both defects are simultaneously
contracted all-around the same atomic volume. As the extended defect configurations have
more microstates than the point defect, such a contraction reduces entropy and, consequently,
an entropy barrier 'S 0 exists. As temperature is lowering, the barrier decreases and
disappears at low temperatures at all and defects recombine easily. This is connected with
changing in configuration of intrinsic point defects, which are extended at high temperatures
and have a point-like dumbbell shaped configuration at low temperatures, as shown in
[45, 46]. It should be emphasized that a theory of extended defect configurations as well as
the theory of recombination barrier has been acknowledged in a number of up-to date papers
[47-50].
Taking into consideration the available dynamic equilibrium between the vacancies and
intrinsic points of silicon atoms as a result of competition between the recombination and
continuous thermal generation of Frenkel pairs and considering the local equilibrium
CI CV CIeqvCVeqv , (2)
which is established regardless the generating technique of the concentration ratio of

eqv
vacancies (CV) to intrinsic interstitial atoms (CI), where C I is an equilibrium concentration
eqv
of intrinsic interstitial atoms, CV is an equilibrium concentration of vacancies, accordingly,
then the time required to reach equilibrium provided the there is no recombination barrier
may be determined by:
:
Wd
4SD S r0 , (3)
where : is a volume cell; DS is a self-diffusion coefficient; r0 is an effective recombination

cross-section [14]. Assuming that the recombination is defined not only by diffusion but also
by height of the recombination barrier, which exceeds free diffusion energy on the value of
'G 'H T'S , (4)
as well as that the recombination barrier is defined by the entropy component of the
expression (4) and applying the concept of extended defect configurations for entropy barrier
microscopic model [45], we obtained the barrier value at = 1685 (the crystallization
temperature), which is 'G = 1.674 eV [4]. Then, we can estimate W value from (3) allowing
for existing recombination barrier:
:
W
4SD S r0 exp( 'G )
kT (5)
With regard to 'G value, we obtain W | 5.3 min. Therefore the recombination does not
get through when standard silicon ingots and standard time of its growth are taken. Such
discrepancies may be explained as follows. The Voronkovs model, firstly, assumes the
theory of enthalpy barrier and, secondly, denies the fact that intrinsic point defects of both
types co-exist simultaneously. Besides, according to Voronkovs model W is determined
individually for interstitial-type and vacancy-type growth modes proposed by him, without
taking into consideration the coefficient of vacancy and intrinsic interstitials joint self-
diffusion. Experimentally obtained results, proving the co-existence of vacancy-type and
interstitial-type microdefects and, consequently, occurrence of both types of intrinsic point
defects near the crystallization front, may be only explained within the framework of theory,
which allows for such a fact and theory of entropy barrier.
Since at temperatures close to the melting temperature the equilibrium concentrations of
vacancies and self-interstitial atoms co-exist simultaneously in dislocation-free silicon single-
crystals, the oversaturated solid-state solution of intrinsic point defects has been decomposing
concurrently at two mechanisms: vacancy-type and interstitial-type. Depending on growth
conditions (in particular, on the V/G ratio) the silicon crystals grow under the vacancy-
interstitial and interstitial growth modes. This definition of crystal growth modes is concerned
with the nature of grown-in microdefects.
2.3 Physical Nature of the Grown-in Microdefects
TEM study of FZ-Si and Z-Si crystals grown at accelerated growth rates gave us a
possibility to define the physical nature of (I+V)-microdefects, D-microdefects and C-
microdefects (see Fig.2). We were first who observed these defects in FZ-Si [15]. Fig.4 shows
the typical TEM images of (I+V)-microdefects, C-microdefects and D-microdefects.
Fig.4. TEM images grown-in microdefects in FZ-Si: a) D-microdefects, g ( 220) , dark field, s 0;
b) D-microdefects, g ( 220) , light field, s ! 0 ; c) C-microdefects, g (202) , dark field, s 0 ; d)
(I+V)-microdefects (1 V-defects, 2 I-defects), TEM, g ( 220) , dark field, s 0.
TEM study was performed similar to above mentioned in section 2.2. We investigated
more than 1000 specimens to establish the statistical relevance of our results. We carried out
the TEM study on electronic microscope with accelerating voltage 100 kV, thus excluding the
introduction of radiation defects. As the strain field around the defect exceeds the defect size,
the shape of the defect is not visible. Main results for FZ-Si [4, 15]:
D-microdefects constitute clusters of interstitial-type point defects with size of 4 to

10 nm; they may be considered as uniform B-microdefect distribution;
C-microdefects are entirely identical to D-microdefects as contrast of TEM images
and the sign of lattice imperfection is concerned; they only differ by the distribution
geometry. Therefore, there is no need to separate them into a distinct type;
At high growth rates (exceeding 6 mm/min), the interstitial-type microdefects occur
simultaneously with the vacancy-type microdefects and localize themselves in the
same areas (I+V)-microdefects.
Evaluation of quantitative ratio for vacancy to interstitial-type defects gave us a value

a1:4 (at V = 7.5 mm/min). When accelerating the crystal growth, the share of vacancy-type
microdefects increases in total number of defects. It is established in crystals with D-
microdefects that the growth rate accelerating gives diminishing in defect sizes.
When CZ-Si 50-mm crystals are grown at a changing growth rate, an area of uniform
defect distribution are formed at V > 2 mm/min with its increase in diameter as the growth
rate is raising (Fig.5a).
Fig.5. D()-microdefects and (I+V)-microdefects in Z-Si (V = 1.82.8 mm/min): a) selective

etching, plane (112): 1 defect-free area with (I+V)-microdefects, 2 ring D(C)-microdefects, 3
(A+B)-microdefects; b) D(C)-microdefects, TEM, g ( 220) , dark field, s 0 ; c) (I+V)-microdefects
(1 V-defects, 2 I-defects), TEM, g ( 220) , dark field, s 0.
Quite often preferential etching does not reveal the defects in the center of the crystal.
However, in this case, the central region is surrounded by high-density uniform defect
distribution. According to classification [10] and researches presented in [11, 12], we
identified these defects as D()-microdefects that correspond to D-microdefects in FZ-Si. In
plane (111) one can observe a ring of D()-microdefects containing so-called "defect-free"
area inside. It will be observed that this ring of defects is often called OSF-ring in
publications. But "oxidation-induced stacking faults" do not pertain to grown-in microdefects,
because they occur after different thermal treatment procedures [17]. As the growth rate
increases, the ring of D-microdefects moves to the crystal periphery while the "defect-free"
area inside the ring increases in diameter. At V = 3 mm/min the ring of D-microdefects
disappears completely.
We conducted TEM directly within the ring of D(C)-microdefects and inside this ring (in
the "defect-free" area). Within the ring of D(C)-microdefects we observed black-white
contrast defects. Their concentration was 1013 to 1014 cm-3, and their size was 4 to 12 nm. A.
Bourret [8] was first who observed these defects in CZ-Si, however the sign of lattice
imperfection has not been determined yet. By black-white and 2.5D methods we determined
that these defects only induce interstitial-type CZ-Si lattice strain. In defect-free area inside
the ring of D(C)-microdefects we observed defects of the same size and concentration. The
contrast analysis of TEM-images showed both vacancy-type and interstitial-type defects. To
establish the defect nature by the black-white contrast method we determined the sign of the
product g l on images obtained in dynamic conditions and also determined the defect depth
to avoid ambiguities in the interpretation of the black-white contrast. We applied the
stereomicrophotography of crystals to determine the defect depth. We investigated 500
specimens to establish the statistical relevance of our results. TEM-observation of CZ-Si
grown at V = 3 mm/min also prove the coexistence of both vacancy-type and interstitial-type
microdefects.
Obtained experimental results are consistent with experimental results for FZ-Si [4]. We
analyzed the white-black contrast of D(C)-microdefects image at dynamic imaging
conditions. Such analysis helps to define the crystallographic properties of these defects [51].
Inclusions, dislocations dipoles, dislocation loops may be characterized as small-size lattice
imperfections with localized strain fields [41, 52]. The contrast analysis proved that D(C)-
microdefects may be theoretically considered as dislocation loops with Burgers vector b =
1/2[100] and b = 1/2[110]. Planes {100}, {110}, {111} can be possible planes of dislocation
loops occurrence [4].
These experimental results were obtained with use of TEM for the case of amplitude
contrast by means of analyzing the diffraction image with black-white imaging contrast and
2.5D techniques. However, these techniques have certain weaknesses and restrictions, related
both to uncertain depth of defect occurrence and to especially small size of defects being
investigated. That is why D-microdefects study using the independent TEM technique by the
lattice direct resolution [15] is of great value. Being armed with this method based on phase
contrast and giving the lattice image of a 0.2 nm resolution, we obtained detailed information
of D-microdefects fine structure. We have been studying n-type undoped FZ-Si single-
crystals with U = 2000 cm, which were vacuum-grown at V = 6 mm/min. Oxygen and
carbon concentration was determined by infrared absorption spectrum and amounted to: <
11016 cm-3 for oxygen and 1.61016 cm-3 for carbon. Authors [15, 53] observed the D-
microdefects electron-microscopic images of two types: having regular (periodic) and
irregular (amorphous) structure (Fig. 6). Microprecipitates of SiO2 amorphous phase produce
the amorphous image [53]. From our point of view, the regular structure belongs to SiC
microprecipitates. Both types of D-microdefects revealed are of interstitial-type. Hence, two
independent TEM study techniques (amplitude contrast and direct lattice resolution)
established the interstitial nature of D-microdefects while being of two types in the crystal.
TEM studies of crystals with B-microdefects (FZ-Si, V = 4 mm/min) showed the same
defects of white-black contrast of image as for the crystals with D-microdefects. However, B-
microdefects size ranges within 15 to 50 nm, having concentration a1010cm-3. Zero contrast
line is often displayed as broken line. The white-black contrast imaging and 2.5D methods
were used to define the physical nature of B-microdefects. We established that all theses
defects induce the compressive strain in the crystal, i.e. they are of interstitial type [4, 16].
Geometry of a certain part of B-microdefects emerged under conditions deviating from
dynamic image. In this case B-microdefects look like plates having at planes {111} projection
a square and diamond shapes (Fig.7d). Either diamond diagonal lines or square sides are
parallel to the diffraction vector g ( 220) . Hence, such defects have cuts in directions [100]
or [110] and lie in planes {100}.
Fig.6. D-microdefects in FZ-Si (V = 6 mm/min), TEM (direct solution of crystalline lattice): a)

recurring structure; b) amorphous structure.
It should be noted that the agglomerations in planes {100} with cuts along the directions
[110] have been observed in CZ-Si crystals and identified as SiO2 precipitates [8, 54]. This
agglomerates has an amorphous structure and they represent SiO2 amorphous phase [8, 15].
They may be considered as a model of edge dislocation interstitial loops with Burgers vector
b 1 / 2[100] [8]. The generation and growth of such agglomerations caused by the oxygen
precipitation induce the strong compression of silicon lattice. Relieving from compression
results in the prismatic extruding of the interstitial-type dislocation loops (A-microdefects) [4,
55]. Furthermore, SiO2 precipitates growing causes the oversaturation of solid-state solution
of silicon intrinsic interstitials and leads to A-microdefects occurrence [4, 7, 56, 57]. Hence,
A-microdefects generation is associated with the generation and growth of precipitates (i.e.
primary grown-in microdefects). Our TEM studies prove that (I+V)-microdefects, D(C)-
microdefects and B-microdefects should be referred to as the primary grown-in microdefects.

Besides, two forms of D(C)-microdefects existence and size nearness and similar contrast of
electron-microscopic images enable us to presume they are uniform distributed small-size B-
microdefects.
TEM studies of crystals with -microdefects (FZ-Si, V = 2 mm/min) were carried out
with the use of electronic microscope JEM-1000 [58]. A-microdefects physical nature has
been analyzed by the inside-outside contrast method [59]. The dislocation interstitial loops
with Burgers vector b 1 / 2[110] were revealed lying in planes {110} and {111} (Fig.7a,b).
The loop size increases as the crystal growth rate slows down. In some cases dislocation lines
of loops were decorated with residual oxygen and carbon impurities. These experiments quite
well match the earlier obtained results of -microdefects study [29, 57]. It should be pointed
to that it was recently shown the -microdefects in CZ-Si crystals are also interstitial type
dislocation loops [21, 60]. As is reported, the vacancy microvoids were not observed in small-
scale FZ-Si and CZ-Si crystals.
Fig. 7. TEM imaged A-microdefects and B-microdefects in FZ-Si: a) A-microdefects, g (022) ,

light field, s 0; b) A-microdefects, g (022) , light field, s 0; c) B-microdefects,
g ( 2 20) , dark field, s | 0 ; d) B-microdefects, g ( 202) , dark field, s ! 0 .

Our experimental results in studying the physical nature of grown-in microdefects

illustrate the entire interrelation of their generation and further growth with the presence of
carbon and oxygen impurities in the silicon crystal. It is worth to emphasize, this interrelation
is revealed experimentally in hyper pure undoped silicon single crystals, where carbon and
oxygen concentration is less than 51015 cm-3. So, the study of so-called oxygen-free FZ-Si
crystals is of much importance. A background oxygen impurity is an important factor
influencing the formation mechanism of the grown-in microdefects. And a carbon
background impurity is as important as above, especially for FZ-Si [4, 16]. For instance, we
showed in CZ-Si that the points with maximum carbon concentration and the points with
maximum density of microdefects coincide [40]. However, if microdefects concentration is
considered as a function of carbon concentration in different CZ-Si crystals, the relationship
will be a bit different. Then, as carbon content increases to 51016 m-3, the concentration of
the grown-in microdefects oscillates. When carbon concentration exceeds 51016 cm-3, the
concentration of microdefects tends to rise as carbon concentration rises. Apparently, when
carbon concentration is below 51016 -3, oxygen impurities are more important in the
formation of microdefects. After doping by oxygen from a gas phase, D-microdefects are
formed and concentration of B-microdefects is increased [16].
Comparing FZ-Si crystals with CZ-Si crystals one can see the changes in distribution
images of grown-in microdefects. This is related to both the change in thermal growth
conditions of the crystal and a much greater concentration of carbon and oxygen impurities
(with the latter being especially important). However the formation mechanism of the grown-
in microdefects in CZ-Si will allow two independent directions in which the supersaturated
solid solution of the point defects is relaxed. The study of the nature of the grown-in
microdefects in CZ-Si proves that the processes responsible for their formation are similar to
those in FZ-Si.
So, the recombination process of intrinsic point defects in FZ-Si and CZ-Si single-
crystals near the crystallization front is hindered due to the recombination barrier. Vacancies
and self-interstitials find out drains in the form of background oxygen and carbon impurities.
Consequently, in the context of physical nature of defect formation in real small-scale crystals
the Voronkovs theoretical model fails to be consistent. Considering a concept of entropy
barrier and a factor of simultaneous self-diffusion of vacancies and self-interstitials [45], we
may reason in this way contrary to the theory of Voronkov that there is a recombination
barrier [4]. Experimental results of TEM studies and heterogeneous formation mechanism of
the grown-in microdefects prove the theoretical considerations of Hu, Sirtl [32, 33] and
Gosele [45] to be true.
2.4 Interaction of Grown-in Microdefects with Radiation Defects
On efficiency of interaction of vacancies or self-interstitials entered by J-rays with

microdefects of a concrete type is possible also to judge about a nature of microdefects. We
investigated of ingot of FZ-Si with U | 100 :cm grown in vacuum with variable growth rate
(V = 1.0; 3.0; 6.0 mm/min). From this crystals were cut of samples which contained
approximately identical amounts of technological impurities but various concentration and
types of grown-in microdefects. Depending on a kind of macrodistribution pictures of
microdefects these crystals were saw as three groups. The measurement of W was conducted
by a phase method and the informations about concentration of carriers n0 are obtained from
measurements of Hall factor. The performances of these crystals are indicated in table 2.
Table 2 The crystal performances
Type of Growth rate, Size, Concentration, W, n0,

Group
defect mm/min Pm cm-3 Ps cm-3
I D 6 0,01 11013 90130 5,51013
II + 3 0,11 1,7105 50120 5,11013
III 1 up to 20 5102 90120 6,01013
The informations about sizes and concentration of grown-in microdefects are obtained
with the help of transmission electronic microscopy. The form of defects and sign of a strain
of a silicon crystalline lattice correspond to experimental results of paragraph 2.3.
In crystals, investigated by us, the authors of Ref. [61] analyzed the dependence of life-
time of unmain carriers (W) and concentration of main carriers (n0) from temperature
( = 100400 ) and also dependences of W on a level of excitation of unmain carriers
(J = 'n/n0 =10-4100) which were obtained before and after an exposure by various streams
J-quantum 60. The experimental results have shown that in all crystals at exposure the E-
centres will be effectively formed and through their levels the recombination of unmain
carriers happens. However, degree of a radiation modification of life-time was unequal in
crystals which were irradiated with the same stream J-quantum. From the analysis of doses
temperature dependences (Wf) the factors W of a radiation modification of life-time (WF = 1/Wf
1/W0, where F - fluens of J-quantum, W0 - life-time before exposure) were determined. These
factors are shown in Tab. 3. In Tab. 3 the calculated rates of introduction ((K = dM/dF, where
M - concentration of radiation defects) of E-centres (K) and centres Ci-Cs (K) are
shown also.
Table 3 The value of factors W and rates of introduction K and K [61].
Group Type of defect W, cm2s-1 K, cm-1 K, cm-1

I D 1,210-10 4,010-4 5,210-4
II + (46)10-11 (14)10-4 (85)10-4
III (513)10-11 (14)10-4 (85)10-4
The modification of W can be stipulated only by decreasing K, as a nature and parameters

of defects (for example, the factors of grab of electrons and holes on levels of radiation
defects) were identical in crystals of all three groups. The analysis of temperature
dependences of Hall factor has shown that the rate of introduction of E-centres (K) correlates
with value W [61].
From our view point, the reason of a decreasing of a rate of E-centers introduction in
crystals II and III groups is a presence of drains for generated vacancies. Such defects can be
agglomerates or precipitates of self-interstitials both background impurities of oxygen and
carbon which the compression of matrix. Due to this they effectively seize free vacancies and
it results in a decrease of a rate of introduction of secondary radiation defects vacancy-type
(including E-centres and, therefore, factors W). Under our judgement, difference of factors t
from a sample to sample (their sizes was 1,5u2u13 mm3) is a corollary of striated distribution
of grown-in microdefects on crystals volume.
The intensive grab of vacancies results in increasing of life-time of self-interstitials and,
therefore, increasing of a rate of introduction of interstitial-type defects. Really, the
experiment has shown that simultaneously with a decrease of K was observed incresing of a
introduction rate of complexes Ci-Cs (K). These complexes poses a level 2 = 0.17 eV.
In difference in crystals of I groups happened a minor decrease of K and K [61]. It can mean
that in crystals of I groups are present the drains for vacancies and for self-interstitials.
2.5 Transformation of Grown-in Microdefects
During production of semiconducting devices the initial monocrystals or the silicon structures
are subjected to sequential technological operations. Their thermal processing happens which
results in origin of additional defects and also to transformation of initial defects. Therefore,
from our view point, it was necessary to establish the correlation of grown-in microdefects in
initial substrates both defects of epitaxy structures and mechanisms with the help of which the
initial microdefects are transformed to defects of actual area of structures.
We investigated semiconducting structures obtained by a method conversion epitaxy. The
essence of a method consists in a sedimentation on high-resistance (as a rule dislocation-free)
silicon substrate of high-doped (defective) epitaxy stratum with consequent two-sided
chemical-machining processing of wafers up to a necessary thickness. In a conversion
structure a working (instrument) stratum is the substrate and basic (contact) - epitaxy film. It
is necessary to mark, that long effect of high temperatures and origin of thermal-mechanical
stresses stipulate a sharp decrease of structural perfection of a substrate and the denseness of
dislocations in her reaches of values ~ 103106 -2. Therefore, greatest interest for us is
study of the factors which determine formation of dislocations in a substrate in epitaxy
process.
For researches we used substrates made from FZ-Si grown with rates 2.0; 3.0; 6.0; 8.0
mm/min. A part of substrates subjected to heat treatment in hydrogen atmosphere at # 1453
K within 3 hours. Such imitation of a temperature mode of epitaxy growth allows precisely to
reveal transformation of an initial defective structure of a substrate. On other part of
substrates in separate process were formed the epitaxial stratums.
TEM-research of thermal-treatment substrates on base of monocrystals obtained with
rates 28 mm/min reveal defects of black-white contrast in dynamic conditions. Under our
judgement, it testifies to continuously going process of defect formation at thermal effect on
crystal. In result TEM-researches of thermal-treatment substrates with initial D-microdefects
(at V = 6 mm/min) are fixed increasing of size of microdefects up to 1014 nm and also
availability of defects with force contrast of a complicated type (Fig. 8a). The concentration
of such large microdefects at centre of crystal is unsignificantly reduced with increase of
crystal growth rate. In substrates from monocrystals obtained at V = 8 mm/min the
dislocations in a central part of substrates practically are absent. The processes of direct
epitaxial growth and thermal imitation of epitaxy result in formation at centre of a wafer of
the 600-s dislocations with a Burgers vector b 1 / 2[110] are in planes {111} (Fig. 8b).
Fig.8. Defects in substrata after epitaxial growth: a) FZ-Si (V = 6 mm/min), TEM, light field,
g ( 220) ; b) FZ-Si (V = 2 mm/min), TEM, light field, g ( 220) .
The processes of transformation of initial grown-in microdefects and defects which arise
during epitaxial growth are similar. The main mechanism of formation of dislocations in a
working stratum of conversion structures is stipulated by process of transformation of initial
grown-in microdefects. We see the following sequence of stages of interstitial microdefects
transformation: D-microdefects o B-microdefects o A-microdefects. The growth of A-
microdefects stimulates formation of dislocations.
Hence, in process of epitaxial growth the initial defective structure essentially varies. The
growth of A-microdefects results in formation of dislocations in volume of crystals. At the
same time B-microdefects are transformed in dislocation loops (A-microdefects) with
consequent formation of dislocations. D-microdefects grow and are transformed in B-
microdefects. The formation of dislocations in samples with D-microdefects is hampered as
the process of transformation is limited by a diffusion of intrinsic point defects and
uncontrollable impurities. Obtained during epitaxial experiments the datas about
transformation of small-sized microdefects in larger allow to assume that D-microdefects are
an initial stage of formation A-microdefects, i.e. interstitial dislocation loops.
For clarification of this position we have conducted experiments on heat treatment of
crystals [4, 16, 40]. Two monocrystals FZ-Si were obtained: one from them grown at V = 6
mm/min and in the certain instant the growth stayed during 60 mines and other crystal grown
at V = 6 mm/min and then was cutted on four parts (cylinders). Each from four parts was
bisected on a plane {112}. One half of each cylinder was subjected to etching for revealing
distribution of microdefects and appropriate second halves was subjected thermal processing
during 60 mines in vacuum at temperatures 1073, 1173, 1273, 1373 K accordingly.
During a stop of growth in crystal all known types of interstitial microdefects were
formed. Each from these types is characteristic usually of crystals obtained at the certain
growth rates (Fig.9a). Here they are present at one crystal and this fact allows to explain their
emerging only by thermal processing. The stop of crystal growth can be considered as a
modification of actual temperature conditions of crystal growth. The crystall in points located
below place of a stop during the certain time receives additional processing at various
temperatures.
Fig.9. Defects in FZ-Si crystal with intermediate stop of growth in the course 60 min: a) selective
etching, plane (112); b) A-microdefects, g (02 2) , light field, s 0 ; c) defects of complicated
form, TEM, g ( 2 20) , light field, s ! 0 .
TEM-research of crystal areas with A-microdefects (near to a place of a stop of growth)

is shown availability of interstitial dislocation loops with size 220 nm. The dislocation line
of loops hardly decored by background impurities because of additional heat treatment (Fig.
9b). In the field of crystal with (A + B)-microdefects are detected defects of complicated form
(Fig. 9c). Such defects represent the intermediate form between A-microdefects and B-
microdefects. In the field of crystal with B-microdefects the interstitial-type defects with sizes
2060 nm emerge. In the field of crystal with D-microdefects the interstitial-type defects
with black-white contrast with sizes up to 15 nm are detected.
TEM-research of the second crystal shown that the heat treatment is accompanied by
redistribution of defects and increasing of their sizes. Heat treatment at 1173 K increases the
sizes of D-microdefects up to 1012 nm. In crystals treated at 1373 K were observed defects
such as "sockets" with force contrast and sizes 50100 nm (Fig. 10a).
Fig.10. Defects in crystals after of thermal treatment: a) FZ-Si (V = 6 mm/min, T =1373 K, t = 60 min,
vacuum), TEM, g ( 2 20) , light field; b) FZ-Si (V = 6 mm/min, T = 1373 K, t = 180 min,
hydrogen), TEM, g ( 2 20) , light field; c) defects in the ring D(C)-microdefects CZ-Si (V = 2.5
mm/min, T = 1373 K, t = 600 min, vacuum), TEM, g (22 0) , light field; d) defects inside the ring
D(C)-microdefects CZ-Si (V = 2.5 mm/min, T = 1373 K, t = 600 min, vacuum), TEM, g (22 0) ,
light field;
Then the FZ-Si grown with various rates were investigated (2.0; 3.0; 6.0; 8.0 mm/min)
and subjected to heat treatment within three hours at the temperature of ~ 1373 K in
atmosphere of hydrogen. After heat treatment everywhere there are defects of black-white
contrast in dynamic conditions. In a comparison with grown-in D-microdefects the size of
defects in treatment crystals is increased up to 1215 nm. Furthermore, the defects with
force contrast as "sockets" (Fig. 10a) or more complicated form (Fig. 10b) with sizes up to
300700 nm will be formed.
Thus, during thermal processings the sizes of defects are increased and defects with force
contrast are emerged. It testifies to a continuity of defect formation process and
transformation of initial microdefects at thermal effect on crystal. The defects with force
contrast can be sources of interstitial dislocation loops. It is necessary to mark, that the same
defects were revealed in FZ-Si crystals (V = 3 mm/min) and were interpreted as B-
microdefects [29]. From our view point, these defects are an intermediate stage of
development of defects between B- and A-microdefects.
Processing procedures lead to the transformation of the grown-in microdefects. As a

result, secondary defects are formed. The transformation of the grown-in microdefects in
course of processing (e.g. thermal treatment or ion-implantation) provides information about
the relaxation process of the supersaturated solid solution of point defects. Single crystal of
CZ-Si, 50 mm in diameter, grown at V = 2.5 mm/min, was subjected to vacuum thermal
treatment at 1373 K for 600 min. In the region of the OSF-ring (in non-processed CZ-Si it
was the ring of D-microdefects), using TEM, we identified black-white contrast defects with
sizes of 10 to 20 nm and defects having a strong strain contrast (precipitates) with sizes of
300 to 600 nm (Fig.10c). Inside OSF-ring the complex-shaped defects we are observed with
sizes of 700 to 900 nm (Fig.10d).
Further, two phosphorus-doped CZ-Si crystals of 80 mm in diameter were grown at
growth rates V = 0.6 mm/min (U = 1830 :cm) and V = 2 mm/min (U = 710 :cm). The
crystals were subjected to thermal treatment in the growing vessel at 923 K for 60 min in the
ambient air medium. We studied the distribution of microdefects in three areas located at the
"top", "middle" and "bottom" of the crystals before and after thermal treatment. Table 4
summarizes the measurement results of defect density and carbon and oxygen concentrations.
Table 4 Density of microdefects and concentration of background impurities in crystals

before and after thermal treatment ( = 923 K, t = 60 min)
Oxygen Carbon
Growth rate, Microdefects density in plane (111),
concentration, concentration,
mm/min cm-2
cm-3 cm-3
before thermal after thermal
centre rim centre rim
processing processing
centre rim centre rim
2.0 ("top") 3.71017 3.21017 4.81016 51016 2.3104 2.8104 3.9104 4.9104
2.0 ("middle") 3.41017 31017 7.91016 7.61016 2.8104 2.7104 4104 3.2104
2.0 ("bottom") 31017 2.81017 1.51017 1.61017 2.310 4
2.9104 2.510 5
4104
0.6 ("top") 3.81017 1.91017 4.61016 4.51016 2.610 4
1.9104 4.510 4
1.1105
0.6 ("middle") 3.51017 2.91017 5.91016 5.31016 1.810 4
2.8104 4.410 4
5104
0.6 ("bottom") 3.71017 1.81017 8.61016 8.91016 2.610 4
2.8104 310 5
9.6104
Carbon concentration increases towards the bottom part of the crystal. After thermal
treatment the etching pit sizes become smaller and the swirl distribution in the bottom part of
the crystal disappears. In CZ-Si crystals grown at V = 0.6 mm/min we identified big etching
pits (~ 103 m-2) after thermal treatment, which were absent in crystals grown at V = 2
mm/min. Consequently, the view of microdefects distribution changes due to thermal
treatment: the concentration of microdefects and their size increase steeply and the swirl
distribution disappears in the bottom part of the crystal where carbon concentration is higher.
CZ-Si initial defect structure change due to ion-implantation was the subject of second
series of experiments. CZ-Si crystal had the grown-in microdefects of a certain type. CZ-Si
crystal of 50 mm in diameter had striated distribution of A- and B-microdefects in the region
with a growth rate V = 0.5 mm/min; striated distribution of B-microdefects in the region with
a growth rate V = 1 mm/min. In the region with a growth rate V = 2.5 mm/min we used the
area with both interstitial-type and vacancy-type microdefects.
It needs to be noted that first two crystal growth regions indicate no substantial difference
(crystals with A- and B-microdefects and crystals with B-microdefects), so we will denote
them as the crystal with (A+B)-microdefects. Let us denote the crystal growth region with
V = 2.5 mm/min as the crystal with (I+V)-microdefects. We may compile experimental
results obtained by TEM observation into three groups. The first group results cover samples
not subjected to annealing after ion-implantation (I); the second group covers the samples
subjected to annealing at 923 K (II); the third group covers the samples subjected to annealing
at 1123 K (III).
The study of group (I) samples with (A+B)-microdefects revealed the following
(Fig.11a):
a) small-scale defects with sizes ranging 6 to 20 nm at concentration of 5u(108109) cm-3;

b) precipitates of regular (square-shaped, rhomb-shaped) and irregular shape with sizes
ranging 30 to 100 nm;
c) dislocation loops with sizes ranging 30 to 150 nm.
The concentration of precipitates amounted to 3u(108109) cm-3 and the concentration

of dislocation loops was 4u(108109) cm-3. Imperfections of the similar type were found in
samples with (I+V)-microdefects, but their sizes were two-three times smaller. At the same
time, the concentration of precipitates and concentration of dislocation loops were one order
of magnitude smaller than those in the samples with (A+B)-microdefects.
Fig.11. Defects in CZ-Si after of ion implantation: a) arsenic implantation without annealing, TEM,
dark field, g ( 2 20) ; b) arsenic implantation and annealing at 923 K, TEM, dark field,
g ( 2 20) ; c) antimony implantation and annealing at 923 K, TEM, dark field, g ( 2 20) ; d)
arsenic implantation and annealing at 1123 K, TEM, dark field, g ( 2 20) .

The study of group (II) samples with (A+B)-microdefects revealed the following
(Fig.11b,c):
a) small-scale defects with sizes ranging 4 to 20 nm;

b) precipitates with sizes ranging 30 to 100 nm;
c) dislocation loops with sizes 20 to 40 nm;
d) straight dislocation segments being angled to surface (111);
e) hexagonal formations with dimensions 160u80 nm to 600u240 nm.
Group (II) samples contain higher concentration of small-scale defects, precipitates and
dislocation loops by one to two orders of magnitude as compared to those of group (I). The
concentrations of dislocations and hexagonal formations are ~ 107 cm-3. Apparently, these
formations represent hexagonal stacking faults [38]. In group (II) samples with (I+V)-
microdefects the concentrations of all the imperfections observed were smaller by one to two
orders of magnitude, while the straight dislocation segments were not found.
The study of group (III) samples with (A+B)-microdefects revealed the following
(Fig.11d):
a) small-scale defects with sizes ranging 4 to 20 nm;

b) precipitates with sizes ranging 30 to 130 nm;
c) dislocation loops with sizes ranging 40 to 100 nm;
d) straight dislocation segments;
e) hexagonal formations.
The concentration of small-scale defects, precipitates and dislocation loops in samples of

group (III) decreases by one to two orders of magnitude as compared to samples of group (II).
Precipitates often occur as plate-like formations in the shape of square, rhombus or rectangle.
In samples of group (III) with (I+V)-microdefects the concentrations of all the imperfections
observed were one to two orders of magnitude smaller, while the straight dislocation
segments were not found.
Obtained experimental results analysis shows that in course of ion-implantation and
further technological annealing, CZ-Si grown-in microdefects growth and transformation
occur, following by relaxation of the supersaturated solid solution of point defects. The
structural imperfections transformation is as follows: grown-in microdefects + point defects
o dislocation loops precipitates o plate-like formations + dislocations.
Processing (ion-implantation, thermal treatment) leads to generating the supersaturated
solid solution of point defects. The initial grown-in microdefects facilitate the subsequent
decomposing of this solution. This process leads to the transformation of the initial
microdefects and the appearance of new imperfections. Low-temperature annealing of
implanted CZ-Si at 923 K induces dislocations and hexagonal stacking faults. However, such
defects have not been found in implanted CZ-Si not subjected to thermal treatment.
Seemingly, low-temperature annealing entails a process when some impurity atoms transit
from lattice sites into interstitial ones [62]. The decrease in the defects concentration during
annealing at 1123 K may be explained both by the decomposition of the oversaturated
silicon-impurity solid solution and by the diffusion process of the intrinsic point defects.
The prismatic punch out mechanism induces the formation of dislocation loops, while their
growth is connected with self-interstitials emitted by the growing precipitates.
Ion-implantation to CZ-Si obtained by vacancy-interstitial-type growth indicates the
general direction of the defect formation mechanism. The supersaturated solid solution of
point defects is forming in course of implantation and consists of arsenic atoms, native point
defects and background impurity atoms of carbon and oxygen. Other researchers in [63] point
to that in such a case the SiAs complex generation is more energy favorable, while the VAs2
complex generation is not proved experimentally [62]. SiAs complex generation is
accompanied by the process when free vacancies are absorbed by the interstitial-type
microdefects. Thus, both interstitial-type microdefects grow emitting self-interstitial atoms of
silicon that results in their excess number. Self-interstitials interact with the vacancy-type
microdefects, which during transformation change their sign of imperfection from the
vacancy-type to the interstitial-type.
Therefore, the mechanisms and processes of formation, growth and transformation of
grown-in microdefects in FZ-Si and CZ-Si crystals are identical. During growth of crystals
the continuous process of transformation of an initial defective structure happens.
3 Conclusion
3.1 Formation Mechanism of the Grown-in Microdefects
From our point of view the experiments with hyper pure FZ-Si are of especial importance. It
was these experiments that allowed us to suggest the heterogeneous formation mechanism of
the grown-in microdefects [4]. Its basic principles can be outlined as follows:
recombination of intrinsic point defects at crystallization temperature is hindered

because of recombination barrier;
background impurities of oxygen and carbon participate in the process of defect
formation as nucleation centers, and consequently participate in the processes of
further growth and transformation of grown-in microdefects;
cooling-induced decomposition of the oversaturated solid solution of point defects in
silicon follows two mechanisms: vacancy-type and interstitial-type.
For the vacancy mechanism theoretically only vacancy aggregation and joint vacancy-
impurity aggregation are possible. Vacancy-impurity aggregation begins earlier than a
vacancy aggregation. Interstitial atoms of oxygen are very mobile, and therefore the
formation of complexes is stimulated by leaving interstitial oxygen in positions of
substitution Os:
Oi + VSi o Os o 2Os + 3Os ++ nOs.
At lower temperature, Os can be formation centers of microprecipitates of oxygen. When

microprecipitates are formed, there is a surplus of volume and one vacancy will be consumed
by a growing precipitate for each two oxygen atoms precipitated. The absorption of vacancies
and impurities by growing microdefects results in a decreased concentration of vacancies in
comparison with oxygen concentration. As a result precipitates begin to absorb oxygen

without participation of vacancies, their sizes are increased and then the type of a strain
around them varies from vacancy (tensile) to interstitial (compressive). The full transition
boundary can be determined from the relation V/G. Thus, the parameter Ccrit does not
describe a change in condition of growth mode (interstitial or vacancy). This parameter
describes conditions of the vanishing (emerging) of microdefects of a vacancy type as a result
of diffusion and interaction of point defects during cooling of the crystal.
The centers of nucleation in a base of interstitial atoms of oxygen exist also for the
interstitial mechanism. However, here a catalytic role is played by carbon atoms. The
oversaturation of interstitial atoms of silicon results in the appearance of complexes [CsISi]:
Cs + ISi l [CsISi].
The decrease of the critical radius [CsISi] nuclei and the acceleration of a diffusion Cs
occur here. Thus, agglomerates [CsISi] are formed. Furthermore, during supersaturation of ISi
co-precipitation of Oi and Cs may occur. Thus, for the formation of B-microdefects:
ISi + Cs o [CI] + Oi o -microdefects.
The growth of interstitial microdefects results in a significant decrease of the

concentration of interstitial atoms of silicon. It creates conditions for precipitations of
impurities. In this case the formation of particles of impurity phase is accompanied by the
generation of self-interstitial atoms in positions between knots of the lattice. Thus two types
of interstitial microdefects are formed: as interstitial congestion (drains for interstitial atoms
of silicon) and as impurity precipitates (sources of these atoms).
Therefore, the vacancy-oxygen aggregation occurs in the vacancy-interstitial mode of
crystal growth when the oversaturated solid-state solution of vacancies is decomposed and the
vacancy-carbon aggregation occurs when the oversaturated solid-state solution of intrinsic
silicon interstitials is decomposed. At a certain critical value of V/G ratio both independent
mechanisms (vacancy-type and interstitial-type) result in generation of D-microdefects. These
D-microdefects are uniform distributed small-size -microdefects, which are further
transformed into -microdefects. At the interstitial-type crystal growth mode the growth of
precipitates facilitate the oversaturation in self-interstitials and induce the homogeneous
nucleation of interstitial clusters (interstitial-type dislocation loops, i.e. A-microdefects). In
general, A-microdefects are formed during B-microdefect transformation both through the
mechanism of prismatic extruding or through the condensation of silicon intrinsic interstitials.
For vacancy-type mechanism:
1) nOi + VSi o n(VO2) o vacancy-type microdefects.

2) n(VO2) + Oi ++ nOi o n[(VnOn) + ISi] o D-microdefects.
For interstitial-type mechanism:
1) s + ISi o (CsISi) o D-microdefects.

2) (CsISi) + Oi o n[(CsISi) + Oi] o B-microdefects.
3) B-microdefects + ISi o A-microdefects.
Thus, follows that regardless of the growth method of dislocation-free silicon single-
crystals, the interstitial-type and vacancy-type microdefects are arisen during the crystals
cooling below the temperature of crystallization, and then transformed into interstitial-type
microdefects. Further transformation of interstitial-type microdefects follows the scheme: D-
microdefects o -microdefects o -microdefects. When the oversaturated solid-state
solution of intrinsic point defects is decomposed the background oxygen and carbon
impurities present themselves as nucleation centers of grown-in microdefects and participate
in their further growth and transformation.
We consider the suggested mechanism as the core of formation mechanism of defects.
To change the core (i.e. to control decomposition of the oversaturated solid solutions of
point defects) is possible as follows:
- to vary thermal growth conditions (growth rate, temperature gradients at the

phase boundary, cooling rate);
- to adjust the condition of the native point defects ensemble in the crystal by
employing various external factors (doping, radiation, etc.);
- to subject the grown single crystals to different kinds of thermal processing.
The heterogeneous mechanism of grown-in microdefects formation describes in the

whole the defect structure of small-scale FZ-Si and CZ-Si crystals. From our viewpoint this
heterogeneous mechanism is works in case of large-scale crystals FZ-Si and CZ-Si. Although
new type of grown-in microdefects, namely vacancy microvoids, appear in such crystals. We
believe the vacancy microvoids are as secondary defects as A-microdefects. They appear due
to change in growth conditions for large-scale FZ-Si and CZ-Si crystals.
3.2 Grown-in Microdefects in Large-Diameter Silicon Crystals
In small-scale crystals under changing thermal growth conditions the grown-in microdefects
of any type except for microvoids are appeared. Adjusting the thermal growth conditions in
large-scale crystals shows that dislocation-free growth is available within the limited range of
growth rate. From Fig.12 one can see that above is existence conditions for ring of D(C)-
microdefects (often called as ring OSF-defects), for A-microdefects outside the ring of D(C)-
microdefects and for (I+V)-microdefects inside the ring of D(C)-microdefects.
Recently the authors [64] have used the X-ray diffuse scattering method to detect both
interstitial-type and vacancy-type defects coexisting at CZ-Si vacancy-interstitial-type
growth. This experimental results reveal that in large-scale CZ-Si the cooling-induced
decomposition of the oversaturated solid solution of point defects follows two mechanisms:
vacancy-type and interstitial-type. Then why do microvoids appear in large-scale crystals, but
not in small-scale crystals?
Fig.12. Schematic large-diameter silicon crystals, plane (112).
It was defined that microvoids are forming and growing at the temperature range ~ 1373
to 1343 K and the oxide film on their side-walls is growing at the temperature range a
13231173 K [65-67]. Here, microvoids have been forming in the large-diameter crystals
under conditions of vacancy-interstitial-type growth, which is characterized by relatively fast
growth rates, lesser curvature of crystallization front and a small temperature gradient. Under
such growth conditions, at a certain cooling stage as thermal conditions change it is quite
possible that the pure vacancy agglomeration will prevail over oxygen-vacancy one. Vacancy
agglomeration will result in forming single or dual-type microvoids and their thermodynamic
equilibrium shape will be octahedral [14]. As is well-known, single microvoids are formed
under a relatively rapid cooling rate and a small oxygen concentration, and dual-type
microvoids are formed under a relatively slow cooling rate and a large oxygen concentration
while the oxide film is growing on side-walls of microvoids and hinders further growth of
dual microvoids [67].
As established by [65], short-time thermal processing (during from 15 minutes at =
1373 K) induces microvoids to diminish their size steeply and to change the structure and,
what is more important, precipitates are revealed. It is determined, that small voids are first to
diminish and disappear and then large voids of dual microvoids [65]. Thorough observations
make it clear that the process of microvoid size diminishing begins in the area between two
voids [67]. Hence, the process of the microvoid formation runs at the expense of the process
of vacancy agglomeration on the nucleation centers during crystal cooling. Nucleation centers
may be either agglomerates of oxygen-vacancy type or agglomerates of carbon-interstitial
type that are formed at higher temperatures. Under certain thermal growth conditions of large-
scale silicon crystals the vacancy oversaturation becomes a prevailing factor to precipitate
vacancies on oxygen-vacancy and carbon-interstitial aggregates. Above assertion is verified
by the experimental fact that there is an oxygen in certain microvoids and a carbon in other
microvoids [68]. Paper [67] is a direct support to the heterogeneous formation mechanism of
grown-in microdefects in large-scale crystals [4]. Therefore, we may establish a fact that
octahedral microvoids and interstitial-type A-microdefects are secondary defects.
3.3 Physical Classification of the Grown-in Microdefects
Silicon single crystals contain a noticeable number of growth impurities (doping or

background) that engender the complex systems like semiconductor-impurity to be created
in crystals. Under certain conditions (temperature, impurity concentration) such systems
become the decomposing solid solutions. As their intrinsic point defects not having drains,
the impurities are actively involved in the decomposition of solid solution of the point
defects. As is seen from experimental results of our study the main and determining factor
when the decomposition of solid solution of the point defects in semiconductor silicon
concerned is the formation of primary defects either in the form of oxygen-vacancy
aggregates (identified here as SiO2 microprecipitates) or carbon-interstitial aggregates
(identified here as SiC micro-precipitates). Heterogeneous nature of defects nucleation
determines in substance the further defect formation mechanism in semiconductor silicon.
As effective drains in the form of dislocations are not available, the decomposition of
solid solution of the point defects entails the intensive formation of structure from secondary
strains, which are agglomerates of intrinsic point defects. Nature of secondary defects is
established by low equilibrium density of the intrinsic point defects as well as by the value
and sign of the misfit volume, which accompany the formation of new phase. Consequently,
the decomposition of solid solution of the point defects is first and foremost distinguished for
the generation of secondary defects when the new phase particle is growing. The generation is
defined by the atom volume difference between the matrix and precipitate. At that, the
secondary defects are mostly localized as agglomerates close to precipitates.
As our study of hyper pure CZ-Si and FZ-Si crystals shows the secondary defects
revealed in them are interstitial-type -microdefects (interstitial dislocation loops) and
vacancy-type octahedral microvoids. Consequently, the lattice imperfection can be defined by
not only new phase particles having insignificant total volume due to low impurity solubility
in silicon, but by secondary defects. They diffuse to a distance of few micrometers from the
new phase particles that generated them, spread over the total volume of crystal and make a
problem for its use in future.
Primary and secondary grown-in microdefects existing in silicon single crystals gives us
an opportunity to introduce a new classification. Such a classification must include all known
grown-in microdefects and proceed from the kinematics of defect formation which is defined
by the heterogeneous formation mechanism of the microdefects. From our point of view, such
classification may avoid discrepancy in letter designations and concepts that constitute a
problem in understanding the grown-in microdefects in dislocation-free silicon, which
originates from existing practice to assign a letter designation to a microdefect of a certain
type. At the same time a classification must reveal the physical nature of defect formation and
be simple and coherent.
So, on the basis of heterogeneous formation mechanism of the grown-in microdefects and
introduced concepts of primary and secondary grown-in microdefects we may develop a new
physical classification (Fig.13). It follows from the assumption that the primary oxygen-
vacancy and carbon-interstitial agglomerates formed on impurity centers are the motive force
of defect formation. Under certain thermal conditions of crystal growth the process of defect
formation entails the generation of secondary defects (agglomerates of intrinsic point defects).
In this case the oxygen-vacancy and carbon-interstitial agglomerates (B-, D- and I+V-
microdefects) turn to be the primary grown-in microdefects and the agglomerates of intrinsic
point defects turn to be the secondary grown-in microdefects (-microdefects and vacancy
microvoids). To form some type of grown-in microdefects or other it depends on adjusting
the thermal conditions of crystal growth. We think this classification is simple and reflects the
physical nature of the defect formation mechanism in dislocation-free silicon single crystals.
Fig.13 outlines the physical classification posed as described above.
Fig.13. Schematic illustration (physical classification) of the grown-in microdefects based on the
heterogeneous mechanism of their formation and transformation: I interstitial-type agglomerates, V
vacancy-type agglomerates; C carbon, O oxygen, i self-interstitials, v - vacancies
In addition to the physical classification, Fig.13 depicts the universal heterogeneous

formation mechanism of the grown-in microdefects in semiconductor silicon crystals of
different diameters. Recombination of intrinsic point defects at about crystallization
temperature is hindered because of recombination barrier. Cooling-induced decomposition of
the oversaturated solid solution of point defects in silicon follows two mechanisms: vacancy-
type and interstitial-type. Background impurities of oxygen and carbon are involved in the
process of defect formation as nucleation centers, and consequently participate in the
processes of further growth and transformation of the grown-in microdefects. As nucleation
centers for oxygen-vacancy aggregates serve the areas of contraction close to oxygen
interstitial atoms the redundant vacancies and other oxygen interstitial atoms are directed
towards them. As nucleation centers for carbon-interstitial aggregates serve the areas of strain
around of carbon substitutions, the redundant intrinsic interstitial atoms and oxygen
interstitial atoms are directed towards them. Such agglomerates growth results in the
formation of (I+V)- microdefects. The latter are primary microdefects, which are formed near
the crystallization front. However, V-microdefects are not referred to microvoids. They are
oxygen-vacancy agglomerates (SiO2 microprecipitates). Growth and transformation of (I+V)-
microdefects gives rise to formation of the secondary defects (A-microdefects) [16]. In real
small-scale crystals the microvoids are not formed.
Increasing the crystal diameter leads to an abrupt change of thermal growth conditions as
compared to small-scale crystals. Adjusting thermal growth conditions as well as crystal
doping allow controlling the formation mechanism and growth of vacancy-type and
interstitial-type defects under conditions of vacancy-interstitial-type growth. This results in
the defect size and shape change and give an advantage to vacancy agglomeration over
oxygen-vacancy agglomeration. The process leads to the formation of the secondary defects,
namely microvoids. Hence, there are secondary defects of two types in large-scale crystals
(A-microdefects and microvoids). However, the basic principles (i.e. core) of the
heterogeneous mechanism suggested for the grown-in microdefects formation in FZ-Si and
CZ-Si remain unchanged.
Above mechanism is alternative to assumptions of Voronkov while not denying the

importance of V/G ratio and mathematical tool of Voronkovs model. V/G ratio characterizes
the conditions when vacancy microdefects appear (disappear). But non-critical approach to
the Voronkovs model leads to that the elephants (A-microdefects and microvoids) can hide
the genuine cause and distort the physical nature of the defect formation in dislocation-free
silicon single crystals.
3.4 Formation Mechanisms of the Secondary Grown-in Microdefects
By the formation mechanism the defects can be divided into two groups: strain-type and
coagulation-type. To the strain-type secondary defects we can refer the dislocation loops.
They are formed by the prismatic punching-out that is a rather common mechanism of defect
generation, which activates under a large difference between the precipitate atomic volume
and matrix atomic volume and is independent of its sign. To the coagulation-type secondary
defects we can refer the defects of two types. The defects of first type in the form of complete
interstitial dislocation loops appear by way of combining the redundant silicon interstitial
atoms located near the precipitates into clusters predominantly in directions {111} and {110}.
We can suggest that the coagulation mechanism of the secondary defects generation activates
at relatively small strain levels near the precipitates when the value of elastic strain is
insufficient to induce the shear dislocation loop. The motive force for such secondary defects
to grow may be undersaturation with vacancies that gone to relieve the elastic strain near the
precipitates. Contrary to above defects the motive force for secondary defects of other type is
oversaturation with vacancies. The oversaturation results in generating the single and dual-
type octahedral voids near the precipitates.
Low growth rates for a crystal of certain diameter combined with a large curvature of
crystallization front and a large axial temperature gradient are crucial parameters affecting the
formation of interstitial dislocation loops (A-microdefects) by employing the strain and
coagulation mechanisms of the secondary defects formation. At the same time high growth
rates for a crystal of certain diameter combined with a small curvature of crystallization front
and a small axial temperature gradient bring out to the formation of octahedral microvoids by
employing the coagulation formation mechanism of the secondary defects. In view of that
during further high-temperature processing the dislocation loops tend to continue to grow, the
silicon crystals should be grown under vacancy-interstitial mode because the defect structure
at this growth stage is very much controlled either by adjusting the thermal growth conditions
or by doping with special dopant impurity (nitrogen, hydrogen) and additional thermal
processing. So, as the formation of the microdefect of any type is connected with carbon and
oxygen impurities, which are present in the crystal, and also with varying the crystal growth
conditions, it is worthwhile to notice that the defect formation may be affected both by
selecting the optimal thermal conditions of growth and by improving the quality of raw
material for silicon single crystals.
To conclude it should be denoted that the heterogeneous formation mechanism of the
grown-in microdefects make it possible to explain the experimental results observed during
the microdefect study of dislocation-free silicon single crystals, in particular, how the striated
distribution of A- and B-microdefect is generated and suppressed, how impurities affect the
nucleation and growth processes of interstitial and vacancy-type microdefects, and thermal
growth conditions of the microdefects formation [69, 70]. It is arguable that such a
mechanism reflects as complete as possible the processes that take place in dislocation-free
silicon single crystals during the crystal growth and cooling. Results of numerous
experiments prove this to be correct.
The special importance of the heterogeneous mechanism of formation and transformation
of grown-in microdefects consists in following: he allows to understand the defect formation
processes in small-scale and large-scale crystals of FZ-Si and CZ-Si. On base of
heterogeneous mechanism of grown-in microdefects formation it is possible to create a data
base on defects in wafers of monocrystalline silicon. In a basis of the heterogeneous
mechanism of grown-in microdefects formation there are processes of disintegration of solid
solutions of intrinsic point defects on impurity centres. Therefore, this mechanism assumes
the participation in defect formation process other impurities (for example, nitrogen, iron,
dopant impurities etc.) and impurity centres.
4 Summary
1. It is established and proved experimentally the heterogeneous formation and transformation
mechanism of the grown-in microdefects in dislocation-free silicon single crystals, which use
completely new approaches:
concentrations of vacancies and intrinsic interstitial silicon atoms at the

crystallization front near the melting point are comparable, recombination of intrinsic
defects are hindered at high temperatures;
the decomposition of the oversaturated solid-state solution of point defects during the
silicon cooling below the crystallization temperature follows two independent
mechanisms: vacancy-type and interstitial-type;
the driving force of the defect formation is initial oxygen-vacancy agglomerates and
carbon-interstitial agglomerates formed on impurities centers;
the reigning feature of decomposition of the oversaturated solid-state solution of
point defects is the generation of secondary defects (agglomerates of intrinsic point
defects), which accompanies a new phase growth (the primary defects).
2. Depending on growth conditions (in particular V/G ratio) dislocation-free silicon

monocrystals grow in vacancy-interstitial and interstitial modes of growth.
3. The technological effects result in a modification of volumetric distribution and

transformation of grown-in microdefects. In particular, irrespective of a crystal growth
method the transformation of interstitial microdefects happens under the following scheme:
D-microdefects o B-microdefects o A-microdefects.
4. At a certain thermal growth conditions the aggregation of point defects under vacancy-
interstitial-type or interstitial type growth mode in the course of crystal cooling may cause the
secondary defects occurrence around primary oxygen-vacancy and carbon-interstitial
aggregates, namely vacancy octahedral microvoids and interstitial-type dislocation loops,
accordingly.
5. The physical classification of grown-in microdefects based on the heterogeneous

mechanism of grown-in microdefects formation in semiconductor silicon is proposed.
References
[1] Abe, T.; Samizo, T.; Marujama S. Jpn. J. Appl. Phys. 1966, vol, 5, p. 458.
[2] Kock de, A.J.R. J. Electrochem. Soc. 1971, vol, 118, p. 1851.
[3] Kock de, A.J.R.; Roksnoer, P.J.; Boonen, P.G.T. J. Cryst. Growth 1974, vol, 22, p. 311.
[4] Talanin, V.I.; Talanin, I.E.; Levinzon, D.I. Cryst. Res. & Technol. 2002, vol, 37, p. 983.
[5] Bernewitz, L.J.; Kolbesen, B.O.; Mayer, K.R.; Shuh, G.E. Appl. Phys. Lett. 1974,
vol, 25, p. 277.
[6] Kolbesen, B.O.; Muhlbauer, A. Sol. State Electr. 1977, vol, 25, p. 1561.
[7] Foll, H.; Gosele, U.; Kolbesen, B.O. J. Cryst. Growth 1977, vol, 40, p. 90.
[8] Bourret, A.; Thibault-Desseaux, J.; Seidman, D.N. J. Appl. Phys. 1984, vol, 55, p. 825.
[9] Voronkov, V.V.; Milvidskii, M.G. Kristallographya 1988, Vol, 33, p. 471.
[10] Veselovskaya, N.V.; Sheikhet, E.G.; Neimark, K.N.; Falkevich, E.S. In Rost i
legirovanie polyprovodnikovych kristallov i plenok; Ed. Alexandrov, L.N.; (Nauka:
Novosibirsk, Russia), 1977; Vol. 2, pp. 284-287.
[11] Kock de, A.J.R.; Stacy, W.T.; Wijgert van de, W.M. Appl. Phys. Lett. 1979,
vol, 34, p. 611.
[12] Kock de, A.J.R.; Wijgert van de, W.M. J. Cryst. Growth 1980, Vol, 49, p. 718.
[13] Roksnoer, P.J.; Boom van den, M.M.B. J. Cryst. Growth 1981, vol, 53, p. 563.
[14] Voronkov, V.V. J. Cryst. Growth 1982, vol, 59, p. 625.
[15] Sitnikova, A.A.; Sorokin, L.M.; Talanin, I.E.; Sheikhet, E.G.; Falkevich, E.S. Phys. Stat.
Sol. (a). 1985, vol, 90, p. K31.
[16] Talanin, V.I.; Talanin, I.E.; Levinzon, D.I. Semicond. Sci. & Technol. 2002,
vol, 17, p. 104.
[17] Shimizu, H.; Munakata, C.; Honma, N.; Aoki S.; Kosaka, Y. Jpn. J. Appl. Phys. 1992,
vol, 31, p. 1817.
[18] Harada, K.; Tanaka, H.; Watanabe T.; Furuya, H. Jpn. J. Appl. Phys. 1998, vol, 37,
p. 3194.
[19] Sadamitsu, S.; Umeno, S.; Koike, Y.; Hourai, M.; Sumita S.; Shigematsu, T. Jpn. J.
Appl. Phys. 1993, vol, 32, p. 3675.
[20] Ammon von, W.; Dornberger E.; Hansson, P.O. J. Cryst. Growth 1999, vol, 198-199,
p. 390.
[21] Furukawa, J.; Tanaka, H.; Nakada, Y.; Ono N.; Shiraki, S. J. Cryst. Growth 2000, vol,
210, p. 26.
[22] Puzanov N.I.; Eidenzon, A.M. J. Cryst. Growth 1994, vol, 137, p. 642.
[23] Umeno, S.; Yanase, Y.; Hourai, M.; Sano, M.; Shida Y.; Tsuya, H. Jpn. J. Appl. Phys.
1999, vol, 38, p. 5725.
[24] Umeno, S.; Okui, M.; Hourai, M.; Sano M.; Tsuya, H. Jpn. J. Appl. Phys. 1997, vol, 36,
p. L591.
[25] Kato, M.; Yoshida, T.; Ikeda Y.; Kitagawara, Y. Jpn. J. Appl. Phys. 1996, vol, 35,
p. 5597.
[26] Nishimura, M.; Yoshino, S.; Motoura, H.; Shimura, S.; Mchedlidze T.; Hikone, T. J.
Electrochem. Soc. 1996, vol, 143, p. L243.
[27] Voronkov V.V.; Falster, R. J. Cryst. Growth 1998, vol, 194, p. 76.
[28] Voronkov V.V.; Falster, R. J. Cryst. Growth 1999, vol, 198-199, p. 399.
[29] Petroff P.M.; Kock de, A.J.R. J. Cryst. Growth 1975, vol, 30, p. 117.
[30] Chikawa J.; Shirai, S. J. Cryst. Growth 1977, vol, 39, p. 328.
[31] Vechten van, J.A. Phys. Rev. B 1978, vol, 17, p. 3197.
[32] Hu, S.M. J. Vac. Sci. and Technol. 1977, vol, 14, p. 17.
[33] Sirtl, E. In Semiconductor silicon; Ed. Huff, R.; Burgess, R.R.; (Electrochem. Soc.:
Princeton), 1977; pp. 4-17.
[34] Voronkov V.V.; Falster, R. J. Electrochem. Soc. 2002, vol, 149, p. G167.
[35] Pasemann, M.; Werner, P. Phys. Stat. Sol. (a) 1979, vol, 54, p. 179.
[36] Pasemann, M.; Werner, P. Phys. Stat. Sol. (a) 1980, vol, 58, p. K1.
[37] Okui, M.; Nishimoto, M. J. Cryst. Growth 2002, vol, 237239, p. 1651.
[38] Ravi, K.V.; Varker, C.J. J.Appl. Phys. 1974, vol, 45, p. 272.
[39] Neimark, K.N.; Sakharov, B.A.; Chulitskii, B.F.; Osovskii, M.I. In Kremnii i Germanii;
Eds: Falkevich, E.S.; Levinzon, D.I.; (Metallurgia: Moskow, Russia), 1970; Vol. 2,
pp. 32-37.
[40] Talanin, V.I.; Talanin, I.E. Phys. Stat. Sol. (a) 2003, vol, 200, p. 297.
[41] Wilkens, M.; Foll, H. Phys. Stat. Sol. (a) 1978, vol, 49, p. 555.
[42] Mitchell, J.B.; Bell, W.L. Acta Met. 1976, vol, 24, p. 147.
[43] Seeger, A.; Frank, W.; Gosele, U. In Radiation effects in semiconductors; Eds: Urli,
N.B.; Corbett, J.W.; Conference series N 46; (Inst. of Phys.: Bristol, London), 1979;
pp. 148-157.
[44] Gosele, U.; Frank, W.; Seeger, A. J. Appl. Phys. 1980, vol, 23, p. 361.
[45] Gosele, U.; Frank, W.; Seeger, A. Solid State Commun. 1983, vol, 45, p. 31.
[46] Dzelme, J.; Ertsinsh, I.; Zapol, B.; Misiuk, A. Phys. Stat. Sol. (a) 1999, vol, 171, p. 197.
[47] Fedina, L.; Gutakovskii, A.; Aseev, A.; Landujt van, J.; Vanhellemont, J. Phys. Stat. Sol.
(a) 1999, vol, 171, p. 147.
[48] Fedina, L.; Gutakovskii, A.; Aseev, A. Cryst. Res. & Technol. 2000, vol, 35, p. 775.
[49] Tognato, R. Phys. Stat. Sol. (a) 1986, vol, 98, p. K133.
[50] Pizzini, S. Phys. Stat. Sol. (a) 1999, vol, 171, p. 123.
[51] Katerbau, K.H. Phys. Stat. Sol. (a) 1976, vol, 38, p. 463.
[52] Ohr, S.M. Phys. Stat. Sol. (a) 1976, vol, 38, p. 553.
[53] Sitnikova, A.A.; Sorokin L.M.; Sheikhet, E.G. Fizika Tverdogo Tela 1987, vol, 29,
p 2623.
[54] Tan, T.Y.; Tice, W.K. Phil. Mag. 1976, vol, 34, p. 615.
[55] Bender, H. Phys. Stat. Sol. (a) 1984, vol, 86, p. 245.
[56] Matsushita, Y.; Kishino, S.; Kanamori, M. Jpn. J. Appl. Phys. 1980, vol, 19, p. L101.
[57] Foll, H.; Kolbesen, B.O. J. Appl. Phys. 1975, vol, 8, p. 319.
[58] Sitnikova, A.A.; Sorokin, L.M.; Talanin, I.E.; Malyshev, K.L.; Sheikhet, E.G.;
Falkevich, E.S. Fizika Tverdogo Tela 1986, vol, 28, p 1829.
[59] Foll, H.; Wilkens, M. Phys. Stat. Sol. (a) 1975, vol, 31, p. 519.
[60] Iida, S.; Aoki, Y.; Sugita, Y.; Abe, T.; Kawata, H. Jpn. J. Appl. Phys. 2000, vol, 39,
p. 6130.
[61] Kolkovskii, I.I.; Latyshenko, V.F.; Sheikhet, E.G.; Shusha, V.V. Fizika i technika
polyprovodnikov 1986, vol, 21, p. 1821.
[62] Popov, V.P.; Dvurechenskii, A.V.; Kashnikov, B.P.; Popov, A.I. Phys. Stat. Sol. (a) 1986,
vol, 94, p. 569.
[63] Terrasi, A.; Rimini, E.; Raineri, V. Appl. Phys. Lett. 1998, vol, 73, p. 2633.
[64] Byblik, V.T.; Zotov, N.M. Kristallographya 1997, vol, 42, p. 1103.
[65] Itsumi, M. J. Cryst. Growth 2002, vol, 237-239, p. 1773.
[66] Yanase, Y.; Nishihata, H.; Ochiai, T.; Tsuya, H. Jpn. J. Appl. Phys. 1998, vol, 37, p. 1.
[67] Ueki, T.; Itsumi, M.; Takeda, T.; Yoshida, K.; Takaoka, A.; Nakajima, S. Jpn. J. Appl.
Phys. 1998, vol, 37, p. L771.
[68] Ueki, T.; Itsumi, M.; Takeda, T. Jpn. J. Appl. Phys. 1999, vol, 38, p. 5695.
[69] Talanin, V.I.; Talanin, I.E.; Levinzon, D.I. Crystallography Reports 2004, vol, 49,
p. 188.
[70] Talanin,V.I.; Talanin, I.E. Defect & Diffusion Forum 2004, vol, 230-232, p.177.
Chapter 3
LASER-INDUCED PHASE TRANSFORMATION

IN NANOCRYSTALLINE SILICON THIN FILMS
R. H. Buitrago1, 2* and S. B. Concari1**

1
Departamento de Fsica, Facultad de Ingeniera Qumica (UNL), Stgo. del Estero 2829
(3000), Santa Fe, Argentina
2
Instituto de Desarrollo Tecnolgico para la Industria Qumica (CONICET-UNL),
Gemes 3450 (3000), Santa Fe, Argentina
Abstract
A comprehensive knowledge of the so-called micro or nanocrystalline silicon has not
been reached till now. Nanocrystalline silicon structure network is quite complex. It consists
of crystallites of around 10 to 30 nm grain size. Because of the little crystallites size, a large
grain boundary tissue as well as amorphous phase is present, providing particular optical and
electrical properties. Being that the nanocrystalline silicon is thermodynamically more stable
and with less light-induced degradation than the amorphous one, this material presents a great
interest to scientists and technologists in thin films solar cells.
The technology used to prepare silicon solar cells is at present well known. The
deposition techniques take place at high temperature and because of the low thickness of the
films, they present very high residual tensions. In the other side, the deposition rate
determines the crystallite grain size. As light differences in crystalline volume fraction and
grain size result in quite different optical and electrical properties, it appears of considerable
interest to study phase transformation processes in nanocrystalline silicon, to know more
about the transition from amorphous to microcrystalline silicon under compression tension
and temperature, which is still not really understood.
In this chapter an up-to-date arts state of pressure-induced phase transformation
processes in silicon are made. Also results of laser-induced phase transformation processes in
intrinsic and boron-doped silicon thin films prepared by Plasma Enhanced Chemical Vapor
Deposition technique (PECVD) are presented and analyzed.
Through Raman spectroscopy, it is possible to determine that the induced phase
transformation is interrelated with the silane flow rate, and boron concentration in the silane
diluted with hydrogen used as a reactive gasses in the preparation of the films, on the grain
*
E-mail address: rbuitre@intec.unl.edu.ar
**
E-mail address: sconcari@fiqus.unl.edu.ar
70 R. H. Buitrago and S. B. Concari
size, and the crystalline volume fraction of them. Measurements of dark conductivity,
Transmission Electron Microscopy (TEM), X-ray and Atomic-Force Microscopy (AFM) help
to corroborate the Raman measurements.
The results show that the laser-induced phase transformation processes in nanocrystalline
silicon thin films present different characteristics from the well known ones produced in
single-crystal silicon by hydrostatic experiments. A model for hot-spot phase transformation is
discussed to explain the laser-induced phase transformation processes in the silicon thin films.
The little volume exposed to the laser beam reaches high temperature and is under high
compress tensions due to the cooler surrounding material. In these conditions, phase Si-XII
emerges and becomes more stable than Si-I. These two phases are present in equilibrium with
the amorphous phase in the silicon thin films with no presence of other crystalline phases.
1 Introduction
In spite of the fact that this chapter deals with nanocrystalline silicon thin films, some features
of crystalline silicon used as a basic photovoltaic material are discussed first.
The composition, flow rate and pressure of the reactant gases, the substrate temperature
and doping concentration are the most important variables in the PECVD, the preparation
technique used in the present study to deposit the films. For selected values of those variables
two set of samples of films were chosen to study the laser-induced phase transformation
processes in nanocrystalline silicon.
Phase transformation processes of the films are better understood if the crystalline
structure, composition and other morphological, optical and electrical properties of them are
well known. Being that these properties are strongly dependent on the preparation conditions,
as we said before, for this reason, a complete characterization of these properties for
nanocrystalline silicon thin films are also presented in this chapter.
2 Silicon as a Photovoltaic Material

Photovoltaic research and market have been first focused on crystalline silicon, since the first
solar cell was developed at Bell Laboratories in 1954. Although others materials have also
been explored, there is still great interest in crystalline silicon, because it has the most
developed technology as well as its abundance on the earth surface. Relative cost of
photovoltaic system compare to hydrocarbon energies sources is the actual challenger.
During the 80th a great research effort was dedicated to develop the amorphous silicon
thin film solar cells, due to it potential low relative cost. The goals prices were reached and
companies started to produce amorphous modules, but latter the modules showed a
degradation process when were exposed to sun. This problem known as Staebler-Wronski
effect (Staebler and Wronski, 1977) is controlled through a multijuncture system, which turns
the photovoltaic module price up, equivalent or more than the single-crystal silicon one.
Since the middle of 90th decade, the group of Neuchtel at Swisserland (Meier et al.,
1994) proved that microcrystalline silicon could be used as a absorber layer in a pin solar cell,
obtaining efficiencies higher than 7 %, and presenting two very important properties: a) it was
stable under light soaking and b) it could be produce using the same equipment used for
amorphous silicon thin films. After that microcrystalline or nanocrystalline silicon thin film
have converted in the most promising new material to make thin film solar cells.
Laser-Induced Phase Transformation in Nanocrystalline Silicon Thin Films 71
The optimization of the photovoltaic materials is essential for improving the solar cells
conversion efficiency. To this aim, the knowledge of structural and physical properties of the
materials plays a relevant role, and many groups of researchers are advocated to this task. In
last years silicon prepared in thin films with micro or nano-structure (grain size between 3
and 50 nm) has been investigated by He et al., 1994; Meier et al., 1998; Tzolov et al, 1997
and Houben et al., 1998, among others. Not only for applications on sensors and transistors
(TFT), but also for certain quantum effects such as the band gap blue shift, photoluminiscence
and visible electroluminiscence and resonant tunnelling. These last properties have possible
application in opto-electronic devices (Chen et al., 1996).
The present interest on silicon thin films is evident from the great amount of research
reports on silicon thin films (amorphous, polymorphous and micro or nanocrystalline)
presented at the 20th International Conference on Amorphous and Microcrystalline
Semiconductors (ICAMS) held at Campos do Jordao (Brasil) at August, 20031. The
improvements in the deposition techniques and conditions as well as the optimization of
structural and other physical properties are the scope of great research in this area.
3 Silicon Phases
Besides crystalline (Si-I) and amorphous (a-Si) faces present at normal temperature and
pressure, there are other ten known Silicon faces: Si-II, with metallic structure (-Sn), Si-III,
with eight atoms in the primitive cell in a center cubic structure (bcc or bc8) and Si-IV, with
four atoms in the primitive cell in an hexagonal diamond structure (hd, wurtzita-type) (Needs
and Mujica, 1995; Crain et al., 1994); Si-V, with simple hexagonal structure (sh) (Hu and
Spain, 1984); Si-VI, with a non yet known structure (initially assigned a double hexagonal
packed structure: dhcp); Si-VII, with hexagonal packed structure (hcp) (Needs and Mujica,
1995); Si-VIII y Si-IX, with tetragonal structure (Zhao et al., 1986); Si-X, face center cubic
structure (fcc) and Si-XI, with orthorhombic center cubic structure named as Imma
(McMahon and Nelmes, 1993); SiXII, with rombohedric structure (r8) (Crain et al., 1994).
These structural faces are present in Silicon when it is under high pressure. Faces II, V,
VI y XI are stables faces at high pressure, while the rest of the faces are meta-stable ones.
In general all the preparation techniques use Hydrogen-diluted silane as a reactant gas to
deposit hydrogenated nanocrystalline Silicon (nc-Si:H) thin films. The amount of Hydrogen
bounded to silicon in the amorphous tissue as well as in the grain boundaries of the
nanocrystallites, depends on the dilution grade and both influence the growth rate, the grain
size and the residual stress in the films.
Usually nanocrystalline Silicon thin films deposited by PECVD have great residual
compression stress because the differences in thermal expansion coefficient of both Silicon
and substrate. So many stable and meta-stable crystalline faces may be present in these films.
That it follows is a brief description of the faces involved in the transformation faces
processes of intrinsic and p-doped nanocrystalline Silicon thin films.
Si-III is a crystalline form with distortion tetrahedral bonds and primitive cell length a =
6.64 . This Silicon face is very well known as a meta-stable one, present in Silicon under
1
A summary of the ICAMS 2003 made by Christian Godet (2003) is at the web site: http://www.u-
picardie.fr/labo/lpmc/ GDRCarbone/documents/ICAMS_2003_ReportCG.rtf
de-pressurization processes in the range 10 13 GPa (100 130 kbar). Kobliska and Solin
(1972) have studied this face in polycrystalline Silicon obtained by hydrostatic pressure of
face I up to 12 GPa and then relaxing to normal pressure. They have calculated the following
Raman scattering lines for this face at 478 cm-1, 433 cm-1, 403 cm-1, 397 cm-1 and 181 cm-1
and identified experimentally the 463 cm-1, 433 cm-1, 416 cm-1 and 384 cm-1 ones with a
maximum error of 3 cm-1. Such lines are appreciably wide (mean height wide, MHW 15
cm-1). This fact is attributed by the authors to the very high residual tensions in the surface.
Others researchers have assigned to this face wave number of scattering Raman at 432 cm-1,
412 cm-1, 382 cm-1 and 163 cm-1 (Kailer et al., 1997).
Si-IV is a meta-stable face with wurtzita type in layers structure. This face has been
detected in re-cristalization processes of Si-III face (Wentorf and Kasper, 1963) and in de-
pressurization processes of Si-I (Kailer et al., 1997; Piltz et al., 1995; Gogotsi et al., 1999).
This face presents a Raman scattering signal at 511 cm-1 (Kailer et al., 1997). The hexagonal
diamond structure of this face have a bonded energy slightly greater than the diamond
structure (~ 0,01 eV per atom).
Si-XII is a meta-stable phase that has been first reported by Crain et al. (1994). It consists
of a rhombohedral diamond structure (r8) with eight atoms in the primitive cell. The density
of this phase is higher than any of the tetrahedral forms of Silicon due to non-bonded
neighbors separation in r8, which is substantially shorter than those of the other forms. This
reduction in next-near neighbors distance is achieved through important local deviations from
tetrahedral bonding, making the r8 phase the most structurally distorted crystalline tetrahedral
form of silicon.
As Piltz et al. (1995) have pointed out, the importance of the r8 structure is further
increased by the presence of add fold bonded rings, which are known to greatly affect
electronic properties.
Distance between atoms in Si-XII is 3.32 in (111) direction with a primitive cell length
a = 5.609 (Crain et al., 1994). This face is not a distorted bc8 face but a different one.
Besides the six atoms, it has rings of five atoms and it is the tetrahedral crystalline more
structurally distorted form of Silicon. This face has been detected in de-pressurization
processes of Silicon at pressure of 8.2 GPa and it predominates over face III at pressure
superior than 7 GPa (Crain et al., 1994). Gogotsi et al. (1999) have identified positions of TO
scattering Raman for Si-XII (r8) at 514 cm-1, 440 cm-1,350 cm-1, 337 cm-1,234 cm-1,183 cm-
1
,179 cm-1. This values are in great agreement with those calculated by Piltz et al. (1995).
The phase-transformation processes induced by compressive stress in bulk Silicon have
been studied through microindentation using diamond tools (Lucazeau and Abello, 1999;
Kailer et al., 1997), by micro-cutting (Gogotsi et al., 1999; Tanikella et al., 1996), and by
hydrostatic tests (Crain et al., 1994; Piltz et al., 1995). Such crystallographic studies have
established that, for pressures up to 10 GPa, single-crystal silicon structures (c-Si) identified
as a phase I (dc) are transformed totally into metallic phase II (-Sn). As pressure is being
slowly reduced to 9.3 GPa, phase XII (r8) starts to form, maintaining itself, down to a
pressure of 2.6 GPa. Reducing the pressure even more, phase XII is partially transformed into
phase III (bc8), which at pressures smaller than 1.6 Gpa transforms to phase I. If the process
of depressurization is slow, phases XII, III and I can coexist at normal pressure. Similar
studies of phase-transformation processes in silicon thin films have not been reported.
Intermediate order in tetrahedrally coordinated silicon thin films has been recently
investigated by Tsu et al. (2003), and Sundae-Meya et al. (2001), but in other experiments of
crystallization of silicon thin films with laser excimer beam, no phases other than I and
amorphous have been detected. In addition, Sundae-Meya et al. found that higher irradiation
power and higher initial hydrogen concentration both lead to films with decreased crystalline
size. It is generally accepted that structural disorder of the Silicon network is influenced by
the Hydrogen concentration. It is also well known that the correlations between the Hydrogen
content and the degree of disorder depend strongly on the preparation method. In addition to
the neutralization of Si dangling bonds, the hydrogenation induces rearrangement of the SiSi
bonding, so differences in phase transformation are expected for different contents of
Hydrogen in the films.
4 Nanocrystalline Silicon Thin Films

4.1 Preparation Technologies
There are available many different techniques to deposit microcrystalline silicon. As was
mentioned before the deposition is achieved by diluting silane in Hydrogen, issues on this
research include mainly how to break the Hydrogen bonds, in both silane and Hydrogen,
obtaining more free radicals and less ions. The most developed technique is the Plasma
Enhanced Chemical Vapor Deposition (PECVD), using excitation frequencies within 13 to
120 MHz (Kroll et al., 1996), the use of high frequency have the goal of increasing the
deposition rates. Between the others techniques that compete with PECVD may be mentioned
the Hot Wire or Catalytic Chemical Vapor Deposition (HW-CVD) (Koh et al., 1995), the
High Pressure Depletion (HPD) (Kondo et al., 2001), and RF Hybrid Thermal Plasma
Chemical Vapor Deposition (TPCVD) (Eguchi et al., 1994).
By the Plasma Enhanced Chemical Vapor Deposition (PECVD) technique has been
possible to obtain polycrystalline material formed by grains of dimensions smaller than 20
nm, imbibed in an amorphous matrix that has been denominated polymorphous as well as
microcrystalline or nanocrystalline silicon (Roca i Cabarrocas et al., 1998; Matsuda, 1983).
The reactive gas concentration to prepare the intrinsic material as well as the doping
gases are variables of the deposition process that influence greatly on properties of the
material. One of the most used gas is silane (SiH4) highly diluted in Hydrogen, Helium and/or
Argon. By using the PECVD technique, crystalline fractions close to 100 % have been
reported in intrinsic nc-Si:H (Siebke et al., 1998; Zhou et al., 1997), p-type (Rubino et al.,
1993), and n type (Nakahata et al., 1998; Vprek et al., 1991).
The plasma contains much different ions and neutral radicals generated by collisions
between electrons and molecules. These ones, in turn, produce secondary reactions while they
move to the substrate. All the species contained in typical reactors plasma reach the substrate.
Most part of the film in formation is finished with Hydrogen, and no SiH3 radicals are taken.
So these radicals present great superficial mobility, to find sites energetically favorable, a
necessary condition for the growing of a high quality material (Guha et al., 2000). Others
radicals such as SiH2 can be incorporated directly at the surface with a high sticking
coefficient. Presence of these and others heavy radicals in the plasma leads to a poor
microcrystalline structure of the film.
Nanocrystalline Silicon thin films have some advantage related to amorphous silicon and
the use of its technology, yet at the industry level. It presents greater photon optic absorption
in the range 1 to 3 eV, compared with crystalline silicon. Such property makes this material a
promising one for factoring devices based on the photoemission effect of silicon. When
crystallites size is reduced to a few nanometers, some convenient properties are obtained for
applications in optics and electrical thin films devices, multijunctions and p-i-n solar cells.
The presence of the B2H5 group and the Hydrogen bounded to silicon in the bulk as well
as in the grain boundaries of the nanocrystallites, depends on the dilution grade and both
influence the growth rate, the grain size and the residual stress in the films.
4.2 Characterization Techniques
To characterize nanocrystalline thin films many physical methods can be used. We choose the
most standard ones. Lateral dark conductivity, measured in a cryostat in the temperature
range of 25 to 160 C, X-ray diffraction of low angle at room temperature, Infra Red
spectrometry on a single crystal silicon substrate, Atomic-Force Microscopy (AFM) to
examine the surface structure, Transmission Electron Microscopy (TEM) which gives
information on crystalline phases present in the films, and finally Raman scattering
spectroscopy.
The latter is a straightforward and very sensitive method of examining crystalline
structures (Lucazeau and Abello, 1997; Campbell and Fauchet, 1986; Lengsfeld et al., 2000)
and phase transformations (Gogotsi et al., 1999; Tanikella et al., 1996) of thin solid films, as
well as, a non-destructive way of monitoring stress in those films (Lucazeau and Abello,
1997; Gogotsi et al., 1999; Paillard et al., 2001). However, compressive or tensile stress, grain
size, phonon confinement and temperature effects can lead to a misinterpretation of the
Raman spectra of these microcrystalline materials (Viera et al., 2001).
4.3 Preparation Conditions
Nanocrystalline silicon thin films study in this chapter were deposited on Corning 7059 glass
by PECVD in a capacitive-type radio frequency reactor working at 50 MHz. The 20 cm
diameter electrodes were 1.35 cm apart from each other. The reactant gas used was SiH4
containing B2H6 highly diluted in H2. Silane concentration is defined as:
Q( SiH 4 )
C ( SiH 4 ) x 100 [%] (1)
Q( SiH 4 ) Q( H 2 )
with Q(SiH4) and Q(H2) being the flow of silane and Hydrogen gases, respectively.
Diborane concentration in the reactant gas was established by partial pressure
relationship as a mixture prepared previously, as:
p( B2 H 6 )
C(B2 H 6 ) x 10 6 [ppm] (2)
p(B2 H 6 ) p(SiH 4 )
with p(B2H6) and p(SiH4) being the partial pressure of diborane and silane gases, respectively.
We analyze here experimental results of two series of samples of intrinsic and doped
films respectively. The samples were prepared with a different dilution of Hydrogen in the
reactant gas. The applied RF power, the total mixture gas flow, the gas pressure, and the
substrate temperature were 110 mW cm-2, 40 sccm, 60 Pa, and 150 qC, respectively. For the
doped films the total mixture gas flow was 6 %, and the substrate temperature was 140 qC,
with the same other experimental conditions. C(B2H6) was varied in the range 25 ppm to
10000 ppm.
The nomenclature used to identify the intrinsic samples is the following: IN being N =
2, 3, 4, 5, 6, according to the value of silane concentration used in the deposition. Doped
samples are identified as PM x 10 5 being M the diborane flow related to the total gas flow
evaluated as M = C(B2H6) x 10 -6 x Q(SiH4) [%], that was in the range of 1.5 x 10 -4 % to 6.0
x 10 -2 %.
To analyse the structure of as-deposited thin films, we collected micro Raman spectra at
low power density (<1 kWcm2) with a triple Jobin-Yvon T64000 spectrometer equipped
with a liquid-N2-cooled charged coupled device camera. We used the 514.5 nm line of an Ar-
ion laser for excitation in a backscattering geometry. A 10X lens provided an apparent spot
size of about 1025 m. Spectral resolution was 0.5 cm1.
A multichannel Ar laser TRS-600SZ-P with an excitation wavelength of 514.5 nm of a
photometric Raman spectrometer was used to induce phase transformation. Power density of
the laser beam was ~100 kW cm2, penetration length ~0.5 m and the spatial resolution was
1 m. All spectra were obtained using a 0.5 cm1 step.
5 Characterization of the Nanocrystalline Silicon Thin Films

In this section a brief report of the results of electrical measurements, AFM, TEM, X-ray and
IR studies as well as optical measurements of the nanocrystalline silicon thin films prepared
by PECVD are presented. Thickness of the films measured from transmittance spectra was in
the range 400 nm to 660 nm and 360 nm to 820 nm for intrinsic and doped films respectively.
5.1 Electrical Measurements
All the intrinsic films studied present an activated dark conductivity (d) as a function of the
inverse of temperature. The obtained values of curves of d versus T1 in the temperature
range of 300450 K, are shown in Figure 1. The great difference in the d value of samples I6
and I5 shows an onset from amorphous to microcrystalline material. The values of dark
conductivity at room temperature are close to those obtained by Asano (1990), Siebke et al.
(1998) and by the Neuchtel group (Meier et al., 1997).
-2
10
I3
-3 I2
10 I4
-4
10 I5
Vd (: cm )
-1
-5
10
-6
10
-7
10
-8
10 a
-9
10 I6
2,2 2,4 2,6 2,8 3,0 3,2 3,4

-1
1000/T (K )
0
10 P3000
P4500
-1
10 P750
-2
10
P180
-3 P100
10
-1
Vd (: cm)
-4
10
-1
-5
10 P60
-6
10
-7
10 P6000
b
-8
P30
10
P15
2,2 2,4 2,6 2,8 3,0 3,2 3,4
-1
1000/T (K )
Figure 1. Dark conductivity vs T 1 for (a) intrinsic (from Concari and Buitrago, 2003) and (b) p-doped
(from Concari, 2004) nc-Si:H films
Light incident slightly enhances the conductivity of the intrinsic nc-Si-H thin films.
Photoconductivity of these samples is very low but stable under standard illuminating
conditions (ph > 10 -3 :-1 cm-1). Photoconductivity of sample I6 is quite similar to one of an
amorphous material. It is 104 times the dark conductivity (ph = 7 x 10 -5 :-1 cm-1). Values of
d at room temperature and activation energy (Ea) are displayed en Figure 2. Both d and Ea
values of intrinsic films are in a broad range as C(SiH4) varies from 6 % to 2 %.
10
-2 a 0.7
-3
10
0.6
-4
10
0.5
EA (eV)
-1
-5
Vd (: cm)
10
-6 0.4
10
-7
10 0.3
-8
10 0.2
-9
10
0.1
2 3 4 5 6
Silane flow (%)
0.8
10
0 b
-1
0.7
10
-2
0.6
10
-3 0.5
10
-1
Ea (eV)
Vd (: cm)
-4 0.4
10
-5 0.3
10
-6 0.2
10
-7
10 0.1
-8
10 0.0
0.00 0.01 0.02 0.03 0.04 0.05 0.06
Diborane flow (%)
Figure 2. Dark conductivity at room temperature () and activation energy measured at 300450 K (o)
for (a) intrinsic and (b) p-doped nc-Si:H films (from Concari et al., 2003)
Also dark conductivity of doped films is strongly dependent on doping. Lightly doped
nanocrystalline silicon gives a low activation energies corresponding to n-type semiconductor
material (see Figure 2). As doping increases, the Fermi level moves down to the center of the
gap, corresponding to compensate nanocrystalline films, which have the lowest d and the
largest Ea. Increasing further more the diborane flow, the activation energy decreases and the
films become to a p-type semiconductor, reaching the maximum d and the minimum Ea for
the sample P3000. The electric transport behavior of this film corresponds to a crystalline
material, with large grain size, so the influence of the presence of a grain boundary barrier in
the sample P3000 is neglectable. For diborane concentration larger than 5000 ppm, the
amount of boron is such that induces a higher disorder instead of a better crystallization
process as before, moving back d and Ea to lower and higher values respectively.
5.2 AFM and TEM Studies
The influence of silane dilution in hydrogen on grain size could be studied by AFM. Figure 3
shows the AFM images of the samples I2 and I6, taken in the same conditions and having the
same scales. When the silane is diluted in Hydrogen the resulting material has a better defined
structure, with increasing coalescence for decreasing silane flow. The bilayer morphology
difference may be stated by the value of the root mean square (RMS) roughness: 6.0 nm for
I6 and 8.5 nm for I2. As the thickness of all the samples is around 400650 nm, the difference
of the RMS values may be attributed to the difference in the grain size.
b
2 nm
m
Figure 3. 2d and 3d AFM pictures of the intrinsic films surface (a) I2 (from Concari and Buitrago,
2003) and (b) I5
The effect of the diborane flow on the coalescence of p-doped nc-Si:H:B films can be
seen in the AFM images of Figure 4. The bilayer morphology difference may be stated by the
value of the root mean square (RMS) roughness presented in Table I. For doping up to 5000
ppm the grain size increases with boron doping (see Table I) but for a greater diborane
concentration it results in films with small crystallites.
Figure 4. AFM pictures of the p-doped films surface (a) P15, (b) P3000 and (c) P6000 (from Concari
and Buitrago, 2004)
TEM measurements indicate that nc-Si:H films are a heterogeneous mixture of an

amorphous matrix and crystallites, the grain size of the later growing with Hydrogen dilution.
Micrograph and its corresponding transmission electron diffractogram of sample I3 are
shown in Figure 5 as example. Select area electron diffraction showed the presence of sharp
spots, including (111), (220) and (311) reflections. The spots are place on rings corresponding
to the amorphous matrix. The presence of the spots clearly indicated that the nacrocrystallites
had the long-range order needed to establish such crystalline diffraction. The first ring
indicates preferential growth in the (111) direction and the small size of the crystallites.
Micrograph shows nanocrystallites of ~ 10 to 20 nm. Similar TEM diffractograms result for
the intrinsic samples I2, I4 and I5 (Concari and Buitrago, 2003).
Figure 5. Bright field planar TEM and select area diffraction pattern of sample I3 (adapted from
Concari and Buitrago, 2003)
Table I. Root mean square (RMS) roughness, mean height (h), and crystalline volume fraction
(Xc) of p-doped nc-Si:H films (From Concari and Buitrago, 2003).
Sample C(B2H6) (ppm) RMS (nm) h (nm) Xc

P15 25 4.6 12 0.52
P60 104 4.3 13 0.37
P180 312 4.9 11 0.40
P750 1250 5.2 16 0.56
P3000 5000 9.1 26 0.82
P4500 7500 5.5 17 0.70
P6000 10000 1.7 7 0.12
5.3 X-ray and IR Studies
The IR spectra were not well defined due to the thickness of the samples. Peaks at 2100 and
2200 cm1 were not resolved, and the 650 cm1 peak presents an interference with oxygen
signal which was probably adsorbed at the surface. In Figure 6 are plotted the absorption IR
spectra of the intrinsic sample. In spite of this difficulty, from absorption peak at 2200 cm-1
the Hydrogen content can be obtained. I6 sample has a higher concentration of Hydrogen
estimated to 8 %. The rest of the intrinsic samples oscillate around 5 %. Doped samples have
Hydrogen content less than 10 %.
I2
I3
Absorbance (a.u.)
I5
I6
500 1000 1500 2000 2500 3000

-1
Wavenumber (cm )
P0
P180
P3000
Absorbance (a.u.)
P4500
P6000
500 1000 1500 2000 2500 3000

-1
Wavenumber (cm )
Figure 6. Infrared spectra of the (a) intrinsic and (b) p-doped nc-Si:H films
X-ray spectrum of the I2 surface sample shown in Figure 7 presents a peak at 28, as well
as the less resolute spectrum of sample I6. The peak at this angle accounts for a preferential
growing in a (111) plane.
Intensity (a.u.)
I2
I6
22 24 26 28 30 32 34 36 38
o
2T( )
Figure 7. X-ray diffraction spectra obtained by subtracting the spectrum of substrate material for
samples I2 and I6
5.4 Raman Studies
Figure 8 shows the micro-Raman scattering spectra of the as-deposited intrinsic films at room
temperature. A semi-quantitative measurement of the crystalline volume fraction XC can be
obtained from the deconvolution of the Raman TO phonon band into components of
amorphous, nanocrocrystalline and crystalline. These peaks are fitted with Lorentzian curves
centered at ~480 cm1, 510514 cm1 and ~520 cm1, respectively.
Intensity (a. u.)
I2
I3
I4
I5
I6
c-Si
350 400 450 500 550 600 650

-1
Raman shift (cm )
Figure 8. Micro-Raman scattering spectra obtained at ~1kW/cm2 and three successive expositions of 20
s from intrinsic nc-Si:H films (from Concari and Buitrago, 2003). Spectrum of c-Si is included at the
bottom as reference
In order to obtain XC, the differences in the Raman cross sections should also be taken
into account. The following equation is used:
Ic
XC (3)
Ic SIa
where Ic is the total microcrystalline and crystalline peak area, Ia is the amorphous peak area
and S is the relative integrated Raman scattering cross-section for c-Si to a-Si. We have used
the data S = 1, 13 obtained at 2.4 eV by Tsu et al. (1982). Values of XC calculated with
Equation 3 are shown in Table I and Table II for doped and intrinsic films respectively. These
results are in agreement with dark conductivity, TEM, IR and AFM data reported above.
Equation 3 is used by other authors taking S not as the integrated Raman scattering cross-
section for c-Si to a-Si, but as a correction to the crystallite size (Schropp and Zeman, 1998;
Bustarret et al., 1988). For the estimated grain size of our samples, volume fraction calculated
with both methods differs by 25 %.
Table II. Crystalline volume fraction (Xc) and grain size (d) of intrinsic nc-Si:H films
Sample C(SiH4) (%) Xc d (nm)

I2 2 0.95 15
I3 3 0.95 15
I4 4 0.90 12
I5 5 0.18 10
I6 6 0.04 9
In Figure 8 it can also be appreciated that, while the concentration of silane in the mixture
of reactive gases decreases, the peak at 520 cm1 (phase Si-I) grows. The broad one at 480
cm1, due to the amorphous phase, shifts to higher frequencies and decreases until
disappearing at high Hydrogen dilution. Structural disorder of the silicon network is
influenced by the Hydrogen concentration of the reactant gases because the neutralization of
the dangling bonds as well as the induced rearrangement of the Si-Si bonds while deposition
occurs. So the sample I2 is expected to be the most crystalline and the I6 the less one. The
values of Xc shown in Table I and II evidence these expectations were right. The low
deposition rate for the samples I2, I3 and I4 favors the coalescent growing of the
nanocrystallites leading to a predominating crystalline phase.
The grain size (d) was obtained from the low frequency shift ('Z) of the TO signal of c-
Si from the value 520 cm-1 by the relation
1/ 2
B
d 2S (4)
'Z
where B = 2 cm-1 nm2 (He et al., 1994). Values of d for the intrinsic films are shown in Table
II. As Sundae-Meya et al. (2001) have stated, the greater Hydrogen concentration, the minor
grain size of the films.
The empiric relation of Dey et al. (2000) between angle deviation (') of Silicon atom
2
bonds and the middle height width () of the dispersion Raman amorphous phase peak can
be used too. With a Lorentzian fitting of amorphous Raman peak the value ' | 6q was
obtained for simple I6 and ' < 3q for the others intrinsic samples confirming the amorphous
and nanocrystalline structure respectively.
In spite of the low deposition temperature, high diborane concentration in the reactant
gases induces a high value of crystalline fraction (Xc > 80 %) in the doped films. Low
diborane concentrations lead to crystalline fraction superior to 40 %. However, when
increasing diborane concentration above 5000 ppm, crystalline fraction decreases.
The peak position of TO Raman signal is 520 cm -1 for crystalline phase I, but it shifts to
lower frequencies (510 - 514 cm -1) as microcrystallites grain size decreases (Tsu et al., 2003).
5.5 Optical Measurements
Hydrogen induces changes in SiSi bondings so differences in optical and electrical

properties as well as in the structure of thin silicon films with different hydrogen content are
expected.
In this section the results of refraction index, absorption coefficient, optical gap and
activation energy are presented.
Table III shows the values obtained for refraction index (n) derived by Swanepoels
method (extrapolated to wavelength O o f) from transmittance spectra (Swanepoel, 1983).
These values agree within the 3 % with the values obtained from the method of Hernndez-
Rojas et al. (1992). Deposition rate was ~ 1 /s.
Table III. Refraction index of intrinsic nc-Si:H films.
Sample I2 I3 I4 I5 I6
n 2.8 3.5 2.8 3.1 3.0
Table IV. Refraction index of doped nc-Si:H films.
Sample P0 P7 P15 P30 P60 P100 P750 P3000 P4500 P6000
n 3.3 3.1 3.0 3.0 3.1 3.0 3.0 3.1 3.2 3.0
Table V shows values of absorption coefficient (D) obtained for all samples in the energy
range of 2.0 - 2.6 eV. It can be seen that D decreases with hydrogen dilution in this range.
Vprek et al. (1981) have attributed this fact to the improvement of optical path due to
internal dispersion by grain boundary or tensioned regions, or due to superficial roughness.
But mean roughness of samples measured by AFM (Figures 3 and 4) of around 6 - 8 nm is
very much smaller than the incident photon wavelength in that range. So the absorption
2
Deys relation for silicon is: 0.5 (cm-2) =7.5 + 3 ' ( ). A value' > 6 indicates presence of great amorphous
complex density in the material. ' 3.5 indicates transition amorphous-crystalline and ' < 3 corresponds
to a crystalline material.
increment of these films would not be explained in terms of the light scattering on the rough
surface.
The optic transition in intrinsic films prepared at dilution higher than 5 % are dominantly
indirect with an approximate gap of 1.4 eV, we interpret, as Kalkan and Fonash (1997) have
proposed, that the small grain size of nc-Si:H films produces phonon confinement, which
induce a change in the phonon distribution density, increasing the relative amount of
acoustical phonon. This effect favors indirect optic transitions and so the increment of light
absorption. This correlation of D increasing with small grain size has already been reported by
Scholten and Akimov (1993).
This increasing of absorption has also been correlated with broadening and asymmetry of
Raman peak at ~520 cm-1 (see Figure 8). Both effects are attributed in this model to spatial
confinement of phonons in microcrystalline grains as well as others authors have done
(Richter et al., 1981; Campbell and Fauchet, 1986 and Dey et al., 2000). This confinement
has been also observed in amorphous silicon (Diehl et al., 1998).
Table V. Absorption coefficient at 2 eV and optical gap for doped nc-Si:H films
Sample P0 P7 P15 P30 P60 P100 P750 P3000 P4500 P6000
D (cm-1) 9.5x103 9.2x103 7.9x103 7.6x103 8.9x103 9.0x103 9.4x103 9.0x103 7.5x103 7.6x103
Eg (eV) 0.99 1.04 1.44 1.30 1.35 1.00 1.03 1.43 1.50 0.90
6 Laser-Induced Phase Transformation

In this section the results of several Raman scattering experiments are presented. Silicon thin
films were exposed to a power laser beam of 40 mW. In-situ spectra were registered as are
shown in Figure 9. In the spectra of samples I5 and I6 it can be observed three peaks. A
slightly asymmetric one at 520 cm-1 (phase Si-I), another quite wide at ~ 480 cm-1
(amorphous phase) and a wide and no symmetric one at ~ 353 cm-1. This last peak does not
appear in the micro-Raman spectra of Figure 8.
Presence of peak at ~ 353 cm-1 has not been informed in the bibliography of Silicon thin
films. Spectra of Raman dispersion generally are taken in the frequency range of 400 - 650
cm-1, and usually low power laser is used to neglect temperature effects. For that reasons it is
possible that signal at ~ 353 cm-1 has not been detected until now in Silicon thin films.
Effects of the Raman equipment were discharged by obtaining spectra of the samples at
other frequency ranges as well as spectra of other materials in the range of interest. No
spectra showed any extra peak as the one at ~ 353 cm-1.
As was described in section 3, phase Si-XII (r8) has a principal TO Raman scattering
signal at 350-353 cm-1 (Piltz et al., 1995; Kailer et al., 1997; Gogotsi et al., 1999) and it is
present in phase transformation processes in bulk single-crystal Silicon. We have identified
the peak at ~ 353 cm-1 detected in the Raman spectra of the intrinsic samples as the principal
TO signal of crystalline phase Si-XII (r8) (Piltz et al., 1995; Kailer et al., 1997; Gogotsi et al.,
1999). As it follows, experiments that corroborate this statement are analyzed.
a
I6
Intensity (a.u.)
I5
I4
I3
I2
325 350 375 400 425 450 475 500 525

-1
Raman shift (cm )
P15
P60
Intensity (a.u.)
P3000
P4500
P6000
325 350 375 400 425 450 475 500 525 550 575
-1
Raman shift (cm )
Figure 9. Raman scattering spectra obtained at ~100 kW/cm2 and one scan of 40 s from (a) intrinsic
(from Concari et al., 2003) and (b) p-doped nc-Si:H films (from Concari et al., 2004). Spectra are
normalized to the peak of 520 cm-1
In the spectra of samples I4, I3 and I2 in Figure 9 the peak from the amorphous tissue
disappears while it is present in the spectra of samples I6 and with low height in spectrum of
sample I5. Also the peak at 520 cm-1 becomes wider and more asymmetric and the one at 353
cm-1 diminishes its height in the spectra of samples I4, I3 and I2 with respect to those of
samples I5 and I6. Comparing these spectra with the micro-Raman spectra of Figure 8, we
interpret that both amorphous Silicon fraction and nanocrystalline in phase I present in the
films have been transformed and phase XII in now present too.
Being the films very thin the substrate effect could be important. The substrate effects in
the laser-induced phase transformation as well as the laser power and exposition time were
investigated. If local temperature has an important effect, ultra-thin films deposited on
substrate with different thermal conductivity should present differences in the laser-induced
phase transformation processes.
Films of variable thickness from 10 to 60 nm were simultaneously deposited on 7059
Corning glass, Tin Oxide coating layer (TCO) and stainless steel (SS) using 6% of diluted
silane and 5000 ppm. The registered Raman spectra of these films are shown in Figure 10. In
the spectra of the film deposited on glass, two peaks are observed at ~520 cm1 and ~353
cm1. They appear even on the films with the minimum thickness after exposing for 40 s to
the laser power of 40 mW. On the other hand, a thickness greater than 40 nm was necessary
to observe the formation of crystalline phases on a dominant amorphous base in films
deposited on stainless steel.
The crystals growth is evidenced by the existence of the scattering Raman peak at 520
cm-1, witch is present in the first 100 on glass while the first layers on TCO and SS are
amorphous. Thickness of amorphous material is greater than 230 on SS and 150 on
TCO. These minimum incubation thickness of nc-Si would be associated to the selective
interaction between growing layer and the substrate as has been explained by Roca i
Cabarrocas et al. (1995). Also the peak at 353 cm-1 is present in layers of 15 nm on glass
while it takes a double thickness of films deposited on TCO and SS showing a laser-induced
phase transformation substrate dependent (Concari, 2004).
a
60 nm
35 nm
Intensity (a.u.)
17 nm
12 nm*
6 nm*
350 400 450 500 550
-1
Raman shift (cm )
b
Intensity (a.u.)
35 nm
17 nm
12 nm
6 nm*
350 400 450 500 550
-1
Raman shift (cm )
Figure 10. Raman scattering spectra from p-doped nc-Si films registered at 40 mW and 100 s deposited
on (a) TCO and (b) 7059 Corning glass. Spectra are normalized to the peak at 520 cm-1. The film
thickness is indicated for each spectrum. The substrate spectrum has been subtracted for sample
indicated with (*)
It has been reported that in phase transformation processes in single-crystal Silicon the
presence of phase III (bc8) (Crain et al., 1994; Piltz et al., 1995; Lucazeau and Abello, 1997;
Kailer et al., 1997; Tanikella, 1996), but phase III (bc8) is not detected in the films as there is
no differentiated peak at ~ 430 cm -1 in the Raman scattering spectra of Figure 9. The absence
of a 430 cm -1 peak and other peaks indicates that nc-Si:H and nc-Si:H:B phase
transformations present different characteristics from those which are produced in the single-
crystal Silicon.
These differences could lie in the fact that the temperature of the sample in this study is
much higher than room temperature. For the power and diameter of the laser used, the mean
temperature could be approximately 260 C (Lucazeau and Abello, 1997), but it could even
reach ~ 700 C at the center spot (Viera and Boufendi, 2001).
In order to study heating of the illuminated sample area effect a third experiment was
carried on. In-situ Raman spectra of intrinsic samples I6 (amorphous) and I2 (most
crystalline) were registered at different temperatures (TR) in an inert atmosphere of Nitrogen.
TR was varied from 20 C to 220 C. In all the spectra (see Figure 11 for sample I2) only the
two peaks at 520 cm-1 and 353 cm-1 were observed. As was expected, the peaks height
diminishes with the increasing temperature due to the growing phonon dispersion in the
nanocrystallites. The peak at 520 cm1 being more sensitive disappeared at 160 C in the
spectrum of sample I2. This experiment proves that the peak at 353 cm1 corresponds to a
quite stable crystalline phase at high temperature.
Phase assigned to signal at 353 cm-1 decreases when the difference between the
temperature of the exposed volume and the surrounding material decreases.
o
T ( C)
20
60
100
140
160
300 350 400 450 500 550 600
Figure 11. Raman scattering spectra from sample I2 registered at increasing temperature which is
indicated for each spectrum (from Concari et al., 2003)
From Figure 9a we can observe that the relative height of the peak at ~ 353 cm -1
decreases with Hydrogen dilution. As Hydrogen bonded in any way to Si atoms is believed to
contribute to the relief of stress in the samples (Street, 1991), sample I6 is expected to be the
most stressed one.
Figure 9b also shows that the relative amount of phase Si-XII decreases with boron
concentration. The amount of the Si-XII phase in the sample P3000 is the smallest one. This
film has the greater crystalline fraction and the larger grain size. As the presence of the Si-XII
phase is assigned to compression tensions in the films, we can conclude that residual tensions
in the nc-Si:H:B decrease with increasing crystalline fraction. As we have analyzed above,
volume crystal fraction of the films depends on the diborane doping.
The presence of phase Si-XII occurs in stressed material, so we think that the films with
high crystalline fraction have higher thermal conductivity and lower residual stress than those
films with low crystal fraction, that contain amorphous phase, such as P6000 and I6.
The presence of crystalline phase Si-XII in both nc-Si:H and nc-Si:H:B films could be
explained as a structure transformation originated in a small volume corresponding to the
zone exposed to the laser beam. As the spot is at a higher temperature than the rest of the film
due to its low thermal conductivity, the sampling volume is compressed by the cooler
surrounding material, thus inducing a higher compression state. As exposition time increases,
local temperature also rises. So we can assume higher compression given an increase of the
intensity of the phase Si-XII peak related to the phase I peak. It has been proved that in
intrinsic nc-Si:H films the phase XII is a crystalline phase more stable at higher temperature
than phase I (Concari and Buitrago, 2003).
On the other hand, the sampling volume at a high temperature could present a different
pattern of structure phase transformation than at room temperature. The Si-XII phase, more
stable and may be requiring less compression, would prevail over the Si-III phase.
An interesting fact to note in Figure 11 it is that the peak of phase I and XII remain in the
same position, within the experimental error. However, increasing temperature should make
the TO Raman peak shift toward lower frequencies (Viera et al., 2001; Cerqueira and
Ferreira, 1999). This can be explained as follows: as it has been well established for both
crystalline and amorphous silicon that the increase of the compressive stress makes the TO
peak shift to higher frequencies (Hishikawa, 1987; Danesh et al., 1996), the effects of
temperature could then be compensate with the effects of the compressive stress over the
sample volume in our films.
Another experiment to study the laser power effect on the induced-phase transformation
processes was developed. Raman spectra of a thin piece of a single-crystal registered for
increasing laser power and exposures are plotted in Figure 12. It can be observed that the
peak at 353 cm-1 appears and grows progressively more than the 520 cm-1signal while other
peak is detected. The Raman spectra obtained at low power (40 mW) and low exposures (15
and 20 s) do not present the peak at ~ 350 cm-1. This peak appears for exposure greater than
50 s. A simple calculus leads to a minimum power of 2 W for the laser used to induce the
phase XII in crystalline Silicon. For a height of peak at 353 cm-1 be comparable to that of the
peak at 520 cm-1 minimum power of 400 mW and a exposure of 1,100 seconds are required.
Peak at 353 cm-1 decreases when locate heating is suspended. Si-XII is a meta-stable
phase which can co-exist with phase I at normal pressure and room temperature but is
reversible with a simple heating at 100 C during two hours. The same results was obtained by
Piltz et al. (1995) in bulk single-crystal Si.
Figure 12. Raman scattering spectra from single c-Si registered for 200 s and normalized to the peak at
520 cm-1. Laser power and exposition time are indicated.
Peak at ~ 353 cm-1 is asymmetric and can be fitted with two Lorenzian curves centered at
353 cm-1 and 349 cm-1 respectively with a area rate of A(353) / A(349) = 0.45 (see
Figure 13). The asymmetry could be due to the superposition of two effects, one the locate
compression that originates the transformation to phase XII and second, a phonon
confinement in the small volume of Si-XII. By increasing the laser exposure the area rate
becomes equal to the unit. This is a evidence of a greater material volume of phase XII witch
is surrounding by Si-I. The confinement effects that produces the asymmetry of the peak
decreases with the laser exposure.
Figure 13. Fitting of the Si-XII peak (...) with Lorentzians curves at 348 cm-1 (-.-) and 352 cm-1 (- -)
The spectra of Figure 9 were registered with a laser power of 40 mW and a exposure
compatible with base noise of 40 s. By increasing the exposure time up to 100 s (Figure 14) it
is observed as happens in the single-crystal Si, the growing of phase XII. In the case of nc-
Si:H this induced-phase transformation is more important as can be seen for sample I6, where
height of Si-XII peak is close to the Si-I one.
Figure 14. Raman scattering spectra from intrinsic samples registered at 40 mW and 100 s. Spectra are
normalized to the signal at 520 cm-1
Measurements accomplished on totally amorphous material show that both Si-I and Si-
XII phases grow simultaneously. So, the crystallization induced by the laser beam in
nanocrystalline silicon films is summarized in the transformation of the amorphous phase into
phase I and then by compression tensions, to phase Si-XII. Reinterpretation of the asymmetry
of the peak at 520 cm -1, may be done in terms of a non-homogenous distribution of grain
size, instead of the presence of phase Si-IV. Also Kobliska and Solin (1972) have found no
evidence of a transformation from Si-III to this wurzite phase caused by sample heating in the
focused laser. They did observe however, a weak line at 355 cm -1. At that time they were not
able to identify it with any known Si phase.
As Tsu et. al. (2003) state, rather than offering evidence for of intermediate order, like
phase Si-IV, the 510 514 cm-1 Raman scattering can be considered due to the
nanocrystallites, where small particle effects lead to observing more of the phonon spectrum,
thus causing the apparent shift to lower frequencies as shown by Richter et al. (1981).
The differences in phase Si-XII growth in the films deposited on glass and stainless steel
can also be explained in terms of a thermal effect. Thermal conductivity and dissipation of
stainless steel are greater than those of glass. Therefore, the temperature of the sampling
volume and compression tensions of films deposited on stainless steel would be smaller than
those deposited on glass.
On the other hand, the sampling volume at a high temperature could present a different
pattern of structure phase transformation than at room temperature. The Si-XII phase, more
stable and may be requiring less compression, would prevail over the Si-III phase. However,
at present, there are factors that are not known about these transformations. Gupta and Ruoff
(1980) have shown that for single-crystals in the (111) direction, semiconductingmetallic
transition is activated at a lower compressive stress than along the (100) direction on single
crystals. As X-ray diffraction spectra showed a preferential growth on the (111) direction of
our microcrystalline films, no presence of phase III could be explained by this fact.
The latter reveals that the relative amount of the Si-XII phase is smallest for the sample
I2, corresponding to the greater crystalline fraction and presumably the larger grain size. As
the presence of the Si-XII phase is assigned to compression tensions in the films, we can
conclude that residual tensions in the nc-Si decrease with crystalline fraction.
7 Synthesis and Conclusions

The characterization methods used point out that the sample series under study are
nanocrystalline with different degrees of volume crystal fraction. Results of AFM and
conductivity measurements are consistent with the Raman spectra. Increasing the Raman spot
laser power produces a change in the film structure. Both Si-I and a-Si transform to an
intermediate phase Si-XII.
The presence of crystalline phase Si-XII in nc-Si:H films could be explained as a
structure transformation originated in a small volume corresponding to the zone exposed to
the laser beam. As the spot is at a higher temperature than the rest of the film due to its low
thermal conductivity, the sampling volume is compressed by the cooler surrounding material,
thus inducing a higher compression state. As exposition time increases, local temperature also
rises. So we can assume higher compression given an increase of the intensity of the phase Si-
XII peak related to the phase I peak.
From what has been reported here, we can conclude that crystalline phase Si-III, present
in single crystals under compression tensions at room temperature and in equilibrium with
phase Si-XII and Si-I, is not detected in the intrinsic and p-doped nc-Si:H thin films. In the
same way, other effects overlap peak shift from 520 cm -1 toward 510 cm -1, but we do not
think it is due to the presence of phase Si-IV. TO Raman peak can be found at lower
frequencies than 520 cm -1 (even as low as 508 cm-1) due to small particle effects (Tsu et al.,
2003).
Therefore, we propose a model of phase-transformation process for the silicon in the
form of amorphous or nanocrystalline thin films as the following: the amorphous phase
crystallizes directly in Si-I phase by the effect of the temperature. This one arrives at phase
Si-XII without going through phase Si-III due to the effect of high temperature and
compression stress in the films.
The Si-XII phase is presented with other phases in single-crystal silicon submitted to
pressures in the range 1.6 2.6 GPa (Piltz et al., 1995). The residual tensions in the
amorphous silicon, always present surrounding the nanocrystals and measured by different
methods (de Lima et al., 1999; Spanakis et al., 2001; Danesh et al., 2001) are in the range 0.7
1.0 GPa. The occurrence of a Raman peak at ~ 350 cm-1, attributed to the Si-XII phase, is
suggesting that the focalized high power laser beam should increase the residual tensions of
the thin films to values higher than 1.6 GPa.
Acknowledgements
Authors would like to thank the American Institute of Physics (AIP), the Institute of Physics
Publishing (IOP) and Elsevier for the permissions to reproduce material from the papers
published in the Journal of Applied Physics (Concari et al., 2003), Semiconductor Science
and Technology (Concari and Buitrago, 2003) and the Journal of Non Crystalline Solids
(Concari and Buitrago, 2004), respectively.
References
Asano, A. Appl. Phys. Lett. 1990, 56 (6) 533
Bustarret, E.; Hichicha, H. A.; Bunei, M. Appl. Phys. Lett.1988, 52 (20) 1675
Campbell, I. H.; Fauchet, P. M. Sol. Stat. Comm. 1986, 58 (10) 739
Cerqueira, M. F.; Ferreira, J. A. J. Mater. Process. Tech. 1999, 923, 235
Chen K.; Qin, H.; Huang, X.; Ikuta, K.; Matsuda, A.; Tanaka, K. J. 0f Non-Crystalline Solids.
1996, 198-200, 891
Cicala, G.; Capezzuto P.; Bruno, G. Thin Solid Films. 1999, 337, 59
Concari, S. B. Buitrago, R. H.; Gutirrez, M. T.; Ganda, J. J. J. Appl. Phys. 2003, 94 (5)
2417-2422,
Concari, S. B.; Buitrago, R. H. J. Non-Cryst. Solids. 2004, 338-340, 192
Concari, S. B.; Buitrago, R. H. Semicond. Sci. Technol. 2003, 18, 1
Concari, S. B. Doctoral Thesis. Universidad Nacional de Rosario: Argentina. 2004
Crain, J.; Ackland, G. J.; Maclean, J. R.; Piltz, R. O.; Hatton, P. D.; Pawley, G. S. Phys. Rev.
B 1994-I, 50 (17) 13043
Danesh, P.; Pantchev, B.; Grambole, D.; Schmidt, B. J. Appl. Phys. 2001, 90 (6) 3065
Danesh, P.; Savatinova, I.; Toneva, A.; Liarokapis, E. J. Non-Cryst. Solids. 1996, 204, 265
Dey, S.; Roy, Ch.; Pradhan, A.; Varma, S. J. Appl. Phys. 2000, 87 (3) 1110
Diehl, F.; Scheib, M.; Schrder B.; Oechsner, H. J. Non-Cryst. Solids. 1998, 227-230, 973
Eguchi, E.; Sata, S.; Tosida, T. Appl. Phys. Lett. 1994, 64, 58
Godet, C. (2003). ICAMS 2003. Compte-rendu tabli pour le GDR 2449. http://www.u-
picardie.fr/labo/lpmc/GDRCarbone/documents/ICAMS_2003_ReportCG.rtf
Gogotsi, Y.; Baek, C.; Kirscht, F. Semic. Sci. Technol., 1999, 14 (10) 936. Erratum. Semic.
Sci. Technol. 1999, 14 (11) 1019
Guha, S.; Yang, J.; Banerjee, A. Prog. Photovolt. Res. Appl. 2000, 8, 141
He, Y.; Yin, Ch.; Cheng, G.; Wang, L.; Liu, X. Appl. Phys. 1994, 75, 797
Hernndez-Rojas, J. L.; Lucia, M. L.; Martil, I.; Gonzlez-Daz, G.; Santamara, J.; Snchez-
Quesada, F. Appl. Optics. 1992, 31, 1606
Hishikawa, Y. J. Appl. Phys. 1987, 62, 3150
Houben, L.; Luysberg, M.; Hapke, P.; Vetterl, O.; Finger, F.; Carius, R.; Wagner, H. J. Non-
Cryst. Solids. 1998, 227-230, 896
Hu, J. Z.; Spain, I. L. Solid State Commun. 1984, 51, 263
Kailer, A.; Gogotsi, Y. G.; Nickel, K. G. J. Appl. Phys. 1997, 81 (7) 3057
Kalkan, A. K.; Fonash, S. Mat. Res. Soc. Symp. Proc. 1997, 467, 319
Kobliska, R. J.; Solin, S. A. Phys. Rev. Lett. 1972, 29 (11) 725
Koh, J.; Ferlautto, A. S.; Rovira, P. J.; Wronski, C. R.; Collins, R. W. Appl. Phys. Lett. 1995,
75, 2295
Kondo, M.; Susuki, S.; Nasuno, Y.; Matsuda. A. Mat. Res. Symp. Proc. 2001, l664, A4.3.1
Kroll, U.; Meier, J.; Shah, A.; Mikhailov, J.; Weber, K. J. Appl. Phys. 1996, 80, 4971
Lengsfeld, P.; Nickel, N. H.; Fuhs, W. Appl. Phys. Lett. 2000, 76 (13) 1680
Lucazeau, G.; Abello, L. J. Mater. Res. 1997, 12 (9) 2262
Matsuda, A. J. Non-Cryst. Solids. 1983, 50-60, 767

McMahon, M. I.; Nelmes, R. J. Phys. Rev B 1993, 47, 8337
Meier, J.; Dubail, S.; Cuperus, J.; Kroll, U.; Platz, R.; Torres, P.; Anna Selvan, J. A; Permnet,
P.; Beck, N.; Pellaton Vaucher, N.; Hof, Ch.; Fischer, D.; Keppner H.; Shah, A. J. Non-
Cryst. Solids. 1998, 227-230, 1250
Meier, J.; Fluckiger, R.; Keppner, H.; Shah, A. Appl. Phys. Lett. 1994, 65 (7) 860
Nakahata, K.; Niida, A.; Kamiza, T.; Fortmann, C. A.; Shimitzu, I. Thin Solid Films. 1998,
337, 45
Needs, R.J.; Mujica, A. Phys. Rev. B 1995, 51 (15) 9652
Paillard, P.; Puech, P.; Sirvin, R. J. Appl. Phys. 2001, 90 (7) 3276
Piltz, R. O.; Maclean, J. R.; Clark, S. J.; Ackland, G. J.; Hatton, P. D.; Crain, J. Phys. Rev. B
1995, 52 (6) 4072
Richter, H.; Wang, Z. P.; Ley, L. Solid State Commun. 1981, 39 625
Roca i Cabarrocas, P.; Hamma, S.; Sharma, S. N.; Viera, G.; Bertrn, E.; Costa, J. J. Non-
Cryst. Solids. 1998, 227-230, 871
Rubino, A.; Addonizio, M. L.; Conte, G.; Nobile, G.; Terzini, E.; Madan, A. Mat. Res. Soc.
Symp. Proc. 1993, 297, 509
Scholten, A. J.; Akimov, J. I. J. Non-Cryst. Solids. 1993, 164-166, 923
Schropp, R. E. I.; Zeman, M. Amorphous and microcrystalline silicon solar cells. Modelling,
Materials and Device Technology; Kluwer Academic Publishers: London, U.K., 1998; 49
Siebke, F.; Yata, S.; Hishikawa, Y.; Tanaka, M. Jpn. J. Appl. Phys. 1998, 37, 1730
Spanakis, E.; Stratakis, E.; Tzanetakis, P.; Fritzsche, H.; Guha, S.; Yang, J. ICAMS 19. 2001
Staebler, D. L.; Wronski, C. R. Appl. Phys. Lett. 1977, 31, 292
Street, R. A. Hydrogenated Amorphous Silicon; Cambridge University Press: Cambridge,
U.K., 1991
Sundae-Meya, A.; Gracin, D.; Dutta, J.; Vlahovic, B.; Nemanich, R. J. Mat. Res. Soc. Symp.
Proc. 2001, 664, 691
Swanepoel, R. J. Phys. E.: Sci. Instrum. 1983, 16, 1214
Tanikella, B. V.; Somasekhar, A. H.; Sowers, A. T.; Nemanich, R. J.; Scattergood, R. O.
Appl. Phys.Lett. 1996, 69 (19) 2870
Tsu, D. V.; Chao, B. S.; Jones, S. J. Sol. En. Mat. Sol. Cel. 2003, 78 (1-4) 115
Tzolov, M.; Finger, F.; Carius, R; Hapke, P. J. J. Appl. Phys. 1997, 81 (11) 7376
Vprek, S.; Sarott, F. J. Non Cryst. Solids. 1991, 137-138, 733
Viera, G.; Huet, S.; Boufendi, L. J. Appl. Phys. 2001, 90 (8) 4175
Wentorf, R. H.; Kasper, J. S. Science. 1963, 139, 338
Zhao, Y. X.; Buheler, F.; Sites, J. R.; Spain, I. L. Solid State Commun. 1986, 59, 679
Zhou, J. H.; Ikuta, K.; Yasuda, T.; Umeda, T.; Yamasaki, S. Appl. Phys. Lett., 1997, 71
(11) 1534
In: New Research on Semiconductors ISBN 1-59454-920-6
Editor: Thomas B. Elliot, pp. 95-121
c 2006 Nova Science Publishers, Inc.
Chapter 4
M ONTE C ARLO S IMULATION OF H OT-P HONON

E FFECTS IN B IASED N ITRIDE C HANNELS
M. Ramonas and A. Matulionis
Semiconductor Physics Institute, A. Gostauto 11,
Vilnius 01108, Lithuania
Abstract
Recent progress in Monte Carlo simulation of hot electron transport and fluc-
tuations in nominally undoped AlGaN/GaN and AlGaN/AlN/GaN heterostructures
with degenerate two-dimensional electron gas channels is reviewed. Input scatter-
ing probabilities of the electrons are calculated in a semiclassical approach from the
confined-electron envelope wavefunctions obtained through a self-consistent Poisson
Schrodinger method. Additional scattering due to accumulation of nonequilibrium
longitudinal optical phonons, termed hot phonons, is treated together with lattice
heating and other scattering mechanisms of importance for electron transport at high
electric fields applied in the plane of electron confinement. Possible ways to treat
electron gas degeneracy and hot-phonon effects through Monte Carlo procedures are
described. Hot-electron and hot-phonon distribution functions are presented for in-
depth discussion of the results.
Complementary information on hot phonons is extracted from electron energy dis-
sipation and fluctuations. In nitride channels, the hot phonons are found to slow down
electron energy dissipation and establish the hot-electron distribution controlled by the
electron temperature. The hot electron temperatures are evaluated from the electron
distribution functions calculated at different electric field strengths. The obtained de-
pendence of hot-electron temperature on supplied power is in a good agreement with
the available experimental data on microwave noise.
Keywords: Monte Carlo simulation, electron transport, nitride heterostructures, hot-

phonons, degeneracy, energy dissipation.

E-mail address: ramonas@pfi.lt
96 M. Ramonas and A. Matulionis
1 Introduction
Semiconductor heterostructure channels with two-dimensional electron gas (2DEG) are un-
der intense investigation due to excellent frequency and noise performance of field-effect
transistors. The 2DEG channels for microwave applications contain a high electron density,
they operate at high electric fields and, consequently, must withstand dissipated powers of a
high density. An adequate understanding of energy dissipation and electron transport under
these specific conditions is vital for device engineering [1, 2].
Theoretic investigation of electron transport at a microscopic level requires analysis of
the Boltzmann kinetic equation. Since analytical solutions are possible only for a limited
number of simple models, device engineering prefers approaches based on coupled energy
and momentum balance equations. Yet, the validity of these equations under conditions
of small volumes subjected to high electric fields is questionable. Monte Carlo technique
seems to be the most appropriate way for studying electronic processes in biased two- and
three-dimensional channels and devices [3, 4, 5].
Emission of longitudinal optical phonons (LO phonons) is the main energy dissipation
mechanism for high-energy electrons [6, 7]. Because of low group velocity, the emitted
LO phonons stay in the channel until they either decay into other phonon modes or are
reabsorbed by the electrons. The accumulated nonequilibrium LO phonons are termed hot
phonons. Possible effects of hot phonons on electron energy dissipation and transport have
been treated theoretically through Monte Carlo simulation [8]. The simulation is equivalent
to solution of coupled kinetic equations for hot electrons and hot phonons. The latter are
found to bottleneck the energy relaxation of hot electrons. Recent interest in hot-phonon
problem is stimulated by GaN-based high-power devices. Extremely strong electron cou-
pling with LO phonons and a high electric power supplied to the electrons necessitates
reconsidering the hot-phonon effects on electron transport and energy dissipation with a
special emphasize on nitride 2DEG channels.
Gallium nitride and related heterostructures are of increasing interest for use in ad-
vanced semiconductor devices [1]. In nitride heterostructures, a high-density 2DEG forms
without an intentional doping [9]. A wide band gap of GaN, together with a high thermal
conductivity bodes well for transistor high-power operation. The measured electron drift
velocity reaches high values at high electric fields [10, 11]. These features predict, for
nitride field-effect transistors, an excellent performance at microwave frequencies.
Indeed, an AlGaN/GaN high electron mobility transistor (HEMT) is the best device for
generation of microwave power at 10 GHz microwave frequency [12, 13, 14, 15]. The cut-
off frequency of AlGaN/GaN HEMTs is determined by electron drift velocity [16] unless
parasitic capacitances and resistances decide the frequency range. The drift velocity, ex-
tracted from the cutoff frequency of HEMTs is 1.2 107 cm/s [16, 17]. There is some
room for improvement of the frequency performance, since, in a GaN p-i-n diode, the mea-
sured drift velocity peaks at 7 107 cm/s [10]. However, the electron transport conditions
differ in p-i-n diodes and HEMTs, and an in-depth investigation of electronic processes is
needed for a better understanding of electron transport limitations in 2DEG channels at high
electric fields. Several sources of such discrepancy can be pointed out.
Since a nitride 2DEG channel usually contains a high-density electron gas, a high-
energy electron does not emit an LO phonon if the final low-energy state is occupied. Thus,
the electron gas degeneracy plays its role [18].
Monte Carlo Simulation of Hot-Phonon Effects in Biased Nitride Channels 97
Under a high bias and a small volume, the supplied electric power density is extremely
high, and channel self-heating takes place. The Joule heat is known to reduce electron
drift velocity. Short voltage pulses have been used to avoid the self-heating effect during
experimental investigation of electron drift velocity [19, 20, 21].
In nitride 2DEG channels, the LO-phonon decay time, termed hot-phonon lifetime,
exceeds considerably the time of LO-phonon emission by a hot electron. Moreover, the
maximum dissipated power per electron 10-100 times exceeds that in arsenide 2DEG chan-
nels [14]. As a result, the mentioned accumulation of hot phonons is the most essential
feature of biased nitride channels [18, 22, 23, 24, 25]. In particular, an almost immediate
absorption of a hot LO phonon after emission of another LO phonon changes direction of
the electron motion with the resultant negative contribution to the drift velocity.
The present chapter starts with the section on self-consistent ShrodingerPoisson cal-
culations of the electron wavefunctions for nitride heterostructures. The electron scattering
mechanisms included into Monte Carlo model are outlined in the sect. 3. The Monte Carlo
model is shortly introduced in sec. 4. Section 5 deals with the hot-phonon effect. The
electron drift velocity in nitride channels in presence of channel self-heating, hot-phonons,
and electron gas degeneracy is discussed in sec. 6. The Monte Carlo results for the elec-
tron temperature and the noise temperature are presented in sec. 7. The electron power
dissipation is discussed in sec. 8. The chapter ends with the conclusions.
2 Conduction Band Profile and Electron Wavefunctions

Electron wavefunctions enter expressions for matrix elements of electron scattering rates.
In a simple approach, the electron wavefunctions are calculated in the low electron density
approximation, without the charge of free electrons taken into account. Analytical expres-
sions for the electron wavefunctions are available for rectangular potential profiles (double
heterojunction structures) and triangular potential barriers (single heterojunction quantum
wells).
As mentioned, nitride channels for high power applications usually contain a high den-
sity 2DEG, and one cannot ignore the charge of conduction electrons. The charge distri-
bution is needed in order to solve Poisson equation. However, the charge density is not
available before the energy bands are obtained from Shrodinger equation where the solu-
tion of the Poisson equation is a prerequisite. Thus, for the 2DEG confined in a quan-
tum well, the electron envelope wavefunctions follow through a self-consistent solution of
the Schrodinger and Poisson equations, for example within the numerical finite-difference
method [26].
Let the electrons be confined by the potential V (z) defined by:
V (z) = ee (z) + Ec (z), (1)
where z denotes the direction normal to the 2DEG plane, e is the electronic charge, e (z)
is the electrostatic potential and Ec (z) is the step function of the interface barrier. The
electrostatic potential is related to the charge distribution by the Poisson equation:

d d e (Nf (z) n(z))
s (z) e (z) = , (2)
dz dz 0
where s (z) is the static dielectric constant, 0 is the permittivity of free space, Nf (z) is
the space-dependent fixed charge that includes the charges induced by piezoelectric and
spontaneous polarization, the surface charge, and that of ionized donors and acceptors. The
charge distribution of the quantum-confined electrons is:
m (z)kB T0
Ef i

n(z) = ln 1 + exp |i (z)|2 , (3)
2 kB T0
i
where kB is the Boltzmann constant, T0 is the absolute temperature, Ef is the Fermi energy,
i is the eigenenergy for the ith subband and i (z) is the normalized to unity envelope
wavefunction of an electron in the ith subband.
It is important to use Fermi statistics instead of Boltzmann statistics, as the electronic
gas in the subbands can be degenerate. In this approach, the charge of two-dimensionally
confined electrons is determined through solution of the one-dimensional Schrodinger
equation within the effective mass approximation for the subband envelope functions and
eigenenergies:

2 d 1 d
i (z) + V (z)i (z) = i i (z). (4)
2 dz m (z) dz
A conventional approach to the solution of the Shrodinger equation is the finite-

difference method. Real space is divided into discrete mesh points and the wavefunction
is solved within those discrete spacings. The RungeKutta method is used to solve the
discretized Shrodinger equation. The electron wavefunctions should vanish far away from
the well inside the confining barriers, and a two-point boundary value problem have to be
solved. The electron wavefunction is chosen to be zero inside one barrier far away from the
quantum well. The electron energy eigenvalue i in the Shrodinger equation is adjusted to
receive a vanishing electron wavefunction in the other barrier.
The Poisson equation is discretized the same way as the Shrodinger equation. The
derivative of the electrostatic potential is taken to be zero in the barrier layers far away from
the quantum well.
An iteration procedure is used to obtain solutions for equations (2) and (4). Starting with
a trial potential V (z), the wavefunctions and their eigenenergies are obtained from equation
(4). They are used to calculate the electron density distribution n(z) from equation (3) and
further e (z) from equation (2). Now, equation (1) provides with the new potential V (z).
The subsequent iterations yield the final self-consistent solutions for e (z) and i (z) with
the required accuracy.
The wurtzite group III nitrides, GaN and AlN, are tetrahedrally coordinated semicon-
ductors; a hexagonal Bravais lattice contains four atoms per unit cell. For binary com-
pounds with wurtzite structure, the sequence of atomic layers of the two constituents is
reversed along the c-axis. In the case of GaN, a basal surface is either Ga- or N- faced.
In the following, we shall deal with AlGaN/AlN/GaN heterostructure that consists of a
12 nm Al0.33 Ga0.67 N layer, 1.5 nm AlN layer and a thick GaN layer. The investigated Al-
GaN/GaN heterostructure consists of a 25 nm Al0.15 Ga0.85 N layer and a thick GaN layer.
The structures are chosen to be Ga-face with c-axis taken to be perpendicular to the het-
erointerfaces.
The heterostructures are intentionally undoped. In nitride structures, a high density

2DEG forms without doping due to polarization difference in the layers [9]. Spontaneous
polarization of AlN, GaN and AlGaN exists at zero strain. It is a convention that the positive
direction of a polarization goes from a gallium atom (cation) to the nearest nitrogen (anion)
atom.
The spontaneous polarization PSP is negative, and its value differs in AlN, GaN and
AlGaN [9]:
PSP (Alx Ga1x N) = (0.052x 0.029)C/m2 . (5)
Under strain, a piezoelectric polarization is induced in addition to the spontaneous po-
larization. The values of the piezoelectric constants in GaN, InN and AlN exceed those
in GaAs-based crystals up to ten times. The piezoelectric polarization is negative for ten-
sile and positive for compressive strained layers, respectively. In the heterostructures under
consideration, thin AlGaN and AlN layers grown onto a thick GaN buffer layer are under
tensile strain. The piezoelectric polarization in the AlGaN alloy layer grown on the GaN,
in the direction of the c axis, can be determined by:
PPE (Alx Ga1x N) =

a(0) a(x) C13 (x)
2 e31 (x) e33 (x) , (6)
a(x) C33 (x)
where a(0) and a(x) are the lattice constants in GaN and AlGaN, respectively, PPE is the
piezoelectric polarization, eij are the piezoelectric coefficients, Cij are the elastic constants.
The piezoelectric and spontaneous polarizations point in the same direction; this increases
the difference in the values of the overall polarization of the heterostructure layers. The
spatial gradient of the polarization at an abrupt interface between the top and the bottom
layers induces the fixed charge density given by:
= [PPE (bottom) + PSP (bottom)]

[PPE (top) + PSP (top)] . (7)
For the AlGaN/AlN/GaN heterostructure, the charge density /e is negative at the Al-
GaN/AlN interface (4.5 1013 cm2 ), and the charge density is positive between AlN
and GaN layers (6.4 1013 cm2 ). The resultant fixed polarization-induced charge is
positive, and free electrons tend to compensate it. For the AlGaN/GaN heterostructure, the
positive piezoelectric charge 8.15 1012 cm2 is generated at the AlGaN/GaN interface.
The polarization-induced charge causes formation of the 2DEG in the GaN layer near
the interface. The 2DEG density is assumed to be 1.4 1013 cm2 for the AlGaN/AlN/GaN
heterostructure and 6 1012 cm2 for the AlGaN/GaN heterostructure. An additional neg-
ative charge on the AlGaN surface takes into account surface states occupied by electrons.
The negative residual acceptor charge is introduced in the GaN buffer for a better 2DEG
confinement.
The self-consistent results for the potential profile and the first three envelope functions
at zero electric field and 300 K temperature are shown in Fig. 1. For the AlGaN/GaN
heterostructure, due to the low interface barrier between AlGaN and GaN, the envelope
functions penetrate into AlGaN considerably: the higher-subband electrons are shared by
1.0 3.5
(a) (b)
3.0
0.8 AlGaN/GaN
AlGaN/AlN/GaN
300K
2.5 300K
Energy (eV)
Energy (eV) 0.6
2.0
0.4 1.5
1.0
0.2
0.5
0.0 0.0
0 10 20 30 40 50 0 10 20 30
Distance (nm) Distance (nm)
Figure 1: The potential profile (solid line) and the first three confined-electron wavefunc-
tions, for AlGaN/GaN (a) and AlGaN/AlN/GaN (b) heterostructures. The dashed, doted
and short-doted lines indicate the envelope functions of the first, second and third sub-
bands respectively. The envelope functions are plotted in arbitrary units, and the zero of
each wavefunction is the corresponding eigenenergy. Dash-doted line is the Fermi energy.
Reprinted with permission from [25]. Copyright (2005) by the American Physical Society.
AlGaN and GaN layers (Fig. 1 (a)). The effective mass of a two-dimensional electron
moving in parallel to the quantum well plane will be composed of the electron effective
mass in the AlGaN barrier and the electron effective mass in the GaN proportionally to the
probabilities of finding an electron in the well and in the barrier. So, the electron penetration
into the barrier material will change the effective mass and the scattering rates.
The electron penetration into the barrier material is avoided when a thin AlN layer is
inserted between AlGaN barrier and GaN channel. The high AlN/GaN interface barrier
confines the electrons in the GaN layer (Fig. 1 (b)) and induces a high-density 2DEG.
Because of the lattice mismatch, only a thin strained (pseudomorphic) layer of AlN can be
grown on a thick GaN layer without strain relaxation and cracking.
Under bias, the electron distribution in the subbands changes. Consequently, the elec-
tron wavefunctions and the electron scattering rates should be reevaluated during the sim-
ulation of the electron transport. Thus the ShrodingerPoisson equations should be solved
and the electron scattering rates should be calculated periodically during the Monte Carlo
simulation until the self-consistency is achieved for the steady state. This iterative scheme
requires enormous computational time. As a rule, the equilibrium FermiDirac distribution
function is used in the PoissonShrodinger solver (3), and the electron wavefunctions are
calculated once at the beginning of the simulation.
At high electron densities, the electron temperature approximation is known to work,
and Fermi distribution function can be used in Poisson-Shrodinger solver (3) with the elec-
tron temperature Te as a parameter. Figure 2 shows the first and second subband eigenen-
ergy and the Fermi level for the AlGaN/AlN/GaN heterostructure. The electron population
redistribution alters the potential given by the electrons and changes the electron wavefunc-
tions and eigenenergies. The effect of electron heating on the eigenenergy of the first sub-
band is almost negligible in the investigated range of electron temperatures. The eigenen-
ergy of the second subband rises as the electron temperature increases, but the change is
0.65 AlGaN/AlN/GaN
300K
0.60
0.55
Energy (eV)
0.50
0.45
0.40
0.35
0 500 1000 1500 2000
Electron temperature (K)
Figure 2: The first (dotted line) and the second (dashed line) subband eigenenergies and the
Fermi level (dash-dotted line) as functions of the electron temperature.
below 20 %. Therefore, the subband energies and the wavefunctions can be assumed to
remain constant and equal to the equilibrium values in order to save computation time.
3 Electron Scattering Rates

Scattering rates are essential input data for Monte Carlo transport simulation [3]. The scat-
tering rates are calculated starting from the Fermi Golden Rule. The electron scattering
probability from the state i to the state f is:
2
Wif = |Hi,f |2 (f i ), (8)

where Hi,f is the interaction matrix element calculated for self-consistent electron wave-
functions, i and f are energies of the initial and the final states. In this approach, the
dynamics of electron interactions is assumed to be independent of the applied field, and the
collisions are assumed to occur instantaneously in the framework of the first-order approx-
imation, thus only two-body interactions are included.
The scattering mechanisms included in our Monte Carlo simulator for 2DEG nitride
channel are acoustic phonon scattering and LO-phonon scattering. As a first step, the model
neglects ionized impurity scattering, interelectron collisions and electron scattering by the
interface roughness, dislocations and other structural defects.
The electron scattering by acoustic phonons at room temperature is often treated as an
elastic process because the energies of the involved acoustic phonons are small. However,
this approximation would mean that electrons could not dissipate energy unless they were
accelerated to the energies greater than the LO-phonon energy. For GaN, the LO-phonon
energy is quite large, and the low-field results in the elastic approximation may lead to an
inaccurate evaluation of the electron energy dissipation. In our simulation, the electron
scattering by acoustic phonons is treated as an inelastic process. In the two-dimensional
case, the integration over the final electron states is complicated by the fuzziness of the
conservation of transverse momentum. To simplify integration the approximation qz q
can be adopted, where q is the in-plane component of the acoustic phonon wave vector q,
and qz is the transverse one [27].
The electrons interact with acoustic phonons through deformation potential and electro-
static polarization associated with atom vibrations. The nitrides exhibit strong piezoelectric
effects, and the piezoelectric scattering is comparable to the deformation potential scatter-
ing at 300 K.
In wurtzite structures, the deformation potential in the central valley is a diagonal sec-
ond rank tensor. The value of Dzz is generally expected to be different from Dxx = Dyy . To
our knowledge, no experimental value is available for GaN to date. Usually equal diagonal
elements are assumed, and the deformation potential tensor is treated as a scalar quantity
[28].
The electron transition probability per unit time from the state (k, i) in ith subband to
the state (k , f ) in f th subband for acoustic deformation potential scattering is:
D2 q 1 1
Wif (k, k) = Mif (qz ){Nac (q ) + }
V vl 2 2

(f (k ) i (k) vl q ), (9)
where the upper and the lower symbols refer to emission and absorption, respectively, i and
f are the indexes of the initial and final subbands, is the crystal mass density, V is the
crystal volume, vl is the longitudinal sound velocity, D is the deformation potential, i (k)
is the electron energy, and 2 is the Planck constant. The overlap integrals Mif (qz ) are
calculated using the self-consistent electron wavefunctions i (z):
2
Mif (qz ) = | i (z)ejqz z f (z)dz| . (10)
Nac (q ) stands for the average acoustic phonon number, in equilibrium:
1
Nac (q ) = v q
. (11)
exp( kBlT0 ) 1
The strength of electron scattering with acoustic phonons via piezoelectric interaction
is determined by the dimensionless electromechanical coupling coefficient. This quantity
contains contributions both for longitudinal (LA) and transverse (TA) acoustical phonons.
After angular averaging the electromechanical coupling coefficient acquires the following
form[29]:
e2 e2
K2 = LA
+ TA
, (12)
0 s cLA 0 s cTA
where cLA , cTA , eLA and eTA are the angular averages of the elastic and piezoelectric
constants, 0 is the permittivity of free space and s is the static dielectric constant. The
electron transition rate for acoustic piezoelectric interaction is:
e2 K 2 vl q2
Wif (k, k) = Mif (qz )
0 s V q (q + q0 )2
1 1
{Nac (q ) + }
2 2
(f (k ) i (k) vl q ), (13)
where q0 is the inverse screening length.

The electronLO-phonon coupling in wurtzite crystals is different from the well-known
cubic case. The electrons interact both with LO-like and TO-like modes, rather than with
a single LO mode as in the cubic case. But recently it has been shown [30] that the TO-
like scattering rate is more than two orders of magnitude lower than the LO-like scattering
rate. Moreover, the LO scattering rate in the cubic approximation is valid regardless of the
chosen point in the Brillouin zone [30]. Thus, the conventional cubic approximation and
Prices formulation for the scattering probabilities can be used, assuming that the phonons
are three-dimensional [31]. The electron transition rate for polar optical interaction is:
e2 (LO 2 2 ) q2
Wif (k, k) = TO
0 V LO q 2 (q + q0 )2
1 1
Mif (qz ){Nopt (q) + }
2 2
(f (k ) i (k) LO ), (14)
where is the high frequency dielectric constant, LO is the LO-phonon frequency, TO

is the TO phonon frequency and Nopt (q) is the LO-phonon number (occupancy).
The electron screening is taken into account for the polar optical phonon scattering and
the piezoelectric scattering. We have assumed long-wavelength diagonal screening derived
from the matrix-random phase approximation [32]:
me2
q0 = fi (0), (15)
20 s 2
i
where i is the subband index and fi (0) is the electron distribution function at the bottom of
each subband.
Integration of Eq. (9, 13, 14) over all possible final states k yields the integrated prob-
ability for the electron in the ith subband with wave vector k to be scattered into subband
f per unit time. The integration is performed numerically, and the total scattering rates are
tabulated for use in the Monte Carlo algorithm.
Figure 3 shows the calculated scattering rates for the electrons in the first subband of
valley of AlGaN/GaN heterostructure. For the low energy electrons, the only energy dissi-
pation channel is acoustic (deformation potential or piezoelectric) phonon emission. At a
higher electron energy (higher than the LO-phonon energy), the LO-phonon emission pre-
vails, and the main part of electron energy will by dissipated through LO-phonon emission.
Small steps indicate intersubband transfer.
4 Monte Carlo Algorithm

Monte Carlo methods are the numerical methods based on random quantities [3]. Electron
transport in semiconductors is treated in terms of realistic motion of one or a number of
electrons in the external electric and magnetic fields. The motion is interrupted by scatter-
ing events. The duration of electron free flight between two successive collisions and the
scattering mechanism responsible for the end of the free flight are selected stochastically
14
10
polar optical emission
13
10
Scattering rate (s )
acoustic deformation potential
-1
12
10 polar optical absorption
acoustic piezoelectric
11
10
10
10
0.0 0.1 0.2 0.3
Electron energy (eV)
Figure 3: Scattering rates for the valley first subband electrons in the AlGaN/GaN het-
erostructure at 300K. Scattering rates for acoustic deformation potential, acoustic piezo-
electric, LO-phonon absorption and emission as functions of electron energy.
in accordance with the scattering probabilities associated with the microscopic process of
interest.
It is sufficient to simulate the motion of one electron when steady state homogeneous
transport is under investigation. From the ergodicity one can assume that a long enough
trajectory of the electron will provide with information about behavior of the entire elec-
tron gas. However, one cannot rely on the ergodicity of the system if the transport is not
homogeneous or is not stationary. Thus, more general Ensemble Monte Carlo techniques
were developed for considering nonhomogeneous systems or calculating the response to
time-dependent electric and magnetic fields.
For wurtzite-phase GaN, the conduction band minimum is located at the point (1 ).
The lowest satellite valleys of the conduction band are at the U point that is two thirds on
the way between the L- and M - symmetry points. The higher conduction band valleys are
located at the point (3 ), at the M point, and at the K point. In the range of electric
fields where the electron scattering into the upper valleys is negligible, a one-valley (1 )
many-subband spherical parabolic model is employed for the scattering rate calculation
and electron transport.
When the physical system is defined, the next step is generation of the initial condi-
tions for each electron under simulation. In the case of electron transport in a homoge-
neous material, only wave vector for each electron should be defined. Usually equilibrium
FermiDirac distribution function can be used to generate the initial electron velocities. If
the nonequilibrium steady-state situation is simulated, the simulation time should be long
enough to avoid the influence of initial conditions. Figure 4 shows the time dependence of
electron drift velocity and LO-phonon population for the initial 5 ps. In order to avoid the
undesirable effect of the transient on the average results, the transient part of the simulation
is excluded from the statistics for a better convergence.
2300
2.5 Wph=1 ps
LO - phonon population (a.u.)

2.0 2200
Velocity (10 cm/s)

7
1.5
2100
Wvdr=0.3 ps
1.0
AlGaN/AlN/GaN 2000
0.5 300 K
0.0 1900
0 1 2 3 4 5
Time (ps)
Figure 4: The time dependence of electron drift velocity (solid line, left axis) and LO-
phonon population (solid line, right axis) for the initial 5 ps of the electron motion in
30 kV/cm field. The dashed lines are the exponential function approximations with
ph = 1 ps and vdr = 0.3 ps. Reprinted with permission from [25]. Copyright (2005)
by the American Physical Society.
In transport simulation, the uniform electric field is applied along the 2DEG channel,
and the electron continuously changes its wave vector according to the classical relations of
motion:
k = eE. (16)
In order to find the electron wave vector just before a scattering event, one should
know the electron free-flight time. The free-flight time is evaluated using the Rees self-
scattering technique [3]. A fictitious self-scattering is introduced, such that the total
scattering probability (the sum of all real scattering probabilities plus self-scattering) is
constant and equal to 01 . If the carrier undergoes a self-scattering, its state after the
scattering event k is taken equal to its state before the event; the electron motion continues
unperturbed as if no scattering at all have occurred.
A random evenly distributed number r is used to generate the stochastic electron free
flight ttr :
ttr = 0 ln(1 r). (17)
During its free flight, the electron is moving according to the equation of motion (16).
Thus, the electron wave vector k is known at the end of the free flight, and the scattering
rates Wj (k) can be evaluated for each mechanism j. The probability of self-scattering
will be a complement to the sum of all real scattering mechanisms represented by . The
scattering mechanism responsible for the end of the free flight must now be chosen among
all those possible, according to the relative strength of different scattering mechanisms. An
evenly distributed random number r is generated, and the product y = r is compared with
the successive sums of the Wj (k). The jth scattering mechanism is chosen if j is such that
the first of the partial sums W1 , W1 + W2 , W1 + W2 + W3 , . . . is larger than y:
y < W 1 + W2 + + W j . (18)
Once the scattering mechanism responsible for termination of the free flight is selected,
the new electron state after the scattering event must be determined. For a true scattering,
the electron state after the scattering event is generated stochastically, according to the
differential cross-section of the selected jth scattering mechanism Wif (k, k).
Since the differential cross-sections are complicated for electrons in a 2DEG, the rejec-
tion technique helps to find the electron state after the scattering event. The state after the
scattering event is the initial state for the next free flight.
Supposing that the electrons are considered independent, the motion of every electron
is recorded, the history of each ith electron is subdivided into equal time intervals and all
required data are collected at the end of each interval. The average value of quantity A(t)
is defined as the ensemble average at time t over N electrons of the system [3] :
1
A(t) = Ai (t). (19)
N
i
The electron mean energy and drift velocity can be calculated, after Eq. (19).
The spectral intensity of current fluctuations is related to a drift velocity autocorrelation
function according to Wiener-Khintchine theorem:

4e2 n
Sj () = v (t) cos(t)dt, (20)
V 0
where:
v (t) = v (t1 )v (t1 + t) v 2 , (21)
can be evaluated from the Monte Carlo simulation [33, 34].
The equivalent noise temperature or simply the noise temperature is defined as [2]:
V Sj ()
kB Tn () = , (22)
4 ()
where Sj () is the spectral intensity of current fluctuations, means or , and and

are the diagonal tensorial components of the small-signal conductances under bias in the
directions transverse and longitudinal in respect to the bias.
The required conductances are approximated by:
= I/U, (23)
and
= dI/dU, (24)
can be obtained from the calculated drift velocity dependence on the electric field.
The Monte Carlo method is a semiclassical method: the electrons move according to
classical laws, the scattering events are treated according to quantum mechanical transition
probabilities. The electrons are fermions and must obey the Pauli exclusion principle: each
k-space state can be occupied by two electrons at best, their spin quantum number must
differ. The Fermi level in the heterostructure 2DEG channels for HEMTs is located above
the bottom of the lowest subband, and, as a rule, the electron gas is degenerate.
In general, electron scattering probability from a state k to a state k can be expressed

as:
P (k, k ) = W (k, k )f (k)[1 f (k )], (25)
where W (k, k ) is the electron transition probability per unit time, and f (k) is the electron
distribution function. The scattering probability is proportional to the probability f (k) that
initial state is occupied and the probability [1 f (k )] that the final state is unoccupied.
The traditional Monte Carlo algorithm uses approximation f (k ) = 0. That is, it as-
sumes that the electron gas is nondegenerate, and all final states are available. In general,
the Pauli exclusion principle should be incorporated into Monte Carlo algorithm [35].
For calculation of the electron distribution in the momentum space, the plane of two-
dimensional electron wave vectors k is subdivided into cells of fixed area k. Let us cal-
culate either the time spent by the electrons in each cell or the number of electrons present
in each cell at a given moment of time. The mean steady-state electron distribution func-
tion is proportional to the dwell time. However, the time-dependent electron distribution
is available through counting the number of electrons n(k)k found at time t in the cell
k around k. The Ensemble Monte Carlo technique generates the electron distribution
function that is known at each time step of the simulation and evolves with time.
The algorithm proposed by Lugli and Ferry [36] easily includes the Pauli exclusion
principle into the Ensemble Monte Carlo procedure. According to the standard Monte
Carlo procedure no final state of the electron is needed in order to determine the duration of
a free flight. Thus, the exclusion principle does not interfere the selection of the scattering
mechanism responsible for the free flight termination. Once the final state is selected, f (k )
becomes known, and a random number between 0 and 1 can be used to accept or reject the
transition.
In the Ensemble Monte Carlo technique, the electron distribution function is expressed
as a number of simulated electrons in each cell of the k-plane grid. The distribution function
should be normalized to unity for the use in the proposed procedure. If the number of
simulated electrons is N and n is the real electron density, the effective real area S of the
simulated heterostructure is S = N/n. The density of allowed wave vectors of one spin in
k-space is (2)2 /S. Every cell in the k-space grid can accommodate at most Nc electrons:
2SSc
Nc = , (26)
(2)2
where Sc is the area of the cell in the k-plane, and 2 accounts for the electron spin. The
distribution function is normalized to unity by dividing the number of simulated electrons
in each cell by Nc . The Nc must be sufficiently large.
After the scattering event, the cell of the electron distribution function, corresponding
to the final state of the electron is found. The normalized distribution function in the cell fc
is compared with the random number r evenly distributed between 0 and 1. If r > fc the
transition is accepted. If, instead, r < fc , the scattering event is treated as the self-scattering
(it does not change electron wave vector). The distribution function is updated at the regular
intervals, and the procedure is iterated until the probability of electron transitions into k
state becomes proportional to the occupancy of that state.
5 Hot Phonons
The Ensemble Monte Carlo technique have been proposed to follow the time evolution of
LO-phonon distribution function and evaluate modifications to the electron transport due to
accumulation of hot phonons [37, 38]. For the Monte Carlo results on biased nitride 2DEG
channels see [18, 22, 21, 25, 39].
In [18, 25], the time-dependent LO-phonon distribution, Nopt (q), is calculated by set-
ting up a histogram hph (t) defined over the grid in the phonon wave-vector space q. In
biased 2DEG channels, the applied electric field breaks the symmetry. As a consequence,
a direction-independent integrated distribution hph (q) is not sufficient, and the full phonon
distribution function hph (q) is considered.
0.6
on i
stribut
0.4
n di
phono
0.2
0.0
-2
2
qy
0
(10
0 -1 )
m
9
9
(1 0
m
-1
2 -2 qx
)
Figure 5: The LO-phonon distribution function for AlGaN/AlN/GaN heterostructure at

room temperature and 20 kV/cm electric field (applied in the x direction). Reprinted with
permission from [25]. Copyright (2005) by the American Physical Society.
At the beginning of the simulation, the mesh Nopt (t = 0) and the histogram hph (t = 0)
are set to the equilibrium value given by Bose distribution function. Under the reasonable
assumption of dispersionless LO phonons, Nopt (t = 0) is independent of the phonon wave
vector. During the simulation, after each event of LO-phonon emission (absorption), hph
is added to (subtracted from) the corresponding cell of the histogram hph . The actual elec-
tron density ne and the number of simulated particles Nsim is taken into account as follows:

2 2 2 ne
hph = , (27)
qx qy qz Nsim Le
where qx qy qz is the volume of the cell in the q space and Le is the effective
channel width. Due to low group velocity of the LO phonons, the launched phonons are
assumed to remain in the 2DEG channel.
The excess LO phonons decay via anharmonic interaction into the zone-boundary
phonons which weakly interact with electrons. Those phonons, in turn, decay into acoustic
modes of the thermal bath. This complex phonon process, acts as an efficient sink of energy
and momentum for the electronLO-phonon subsystem. For a demonstration of hot-phonon
effect on drift velocity the decay of nonequilibrium LO-phonon is treated in relaxation-time
approximation. At fixed times iT , i = 1, . . . , M during the simulation (with T shorter
than the average LO-phonon scattering time), hph is updated through:
T
hph (iT ) = hph (iT ) [hph (iT ) hph (0)] , (28)
ph
where ph is the LO-phonon relaxation time (the lifetime with respect to their decay into
other phonons). The LO-phonon lifetime can be determined experimentally or used as a
fitting parameter in the Monte Carlo simulation. In framework of the discussed nitride
channel model, the best agreement between Monte Carlo calculations and experimental
results is obtained for ph =1 ps [25]. This value for the hot-phonon lifetime is in a good
agreement with that obtained from Raman measurements [41].
The distribution Nopt (t) is refreshed at the end of each time step using the histogram
hph . The rejection technique is used to avoid recalculation of electronLO-phonon scat-
tering rate at the end of each time step. The overall scattering rates are calculated at the
beginning of the simulation. For convenience, the maximum value of the phonon distri-
bution Nmax stands for the actual q-dependent LO-phonon distribution function Nopt (q)
contained in the electronLO-phonon scattering rate, Eq. (14). The non-physical artificial
enhancement of the electron scattering rates due to Nmax is compensated through the fol-
lowing rejection technique. Once the final electron state after the scattering is known, the
wave vector of the involved LO-phonon is determined. The random number r, evenly dis-
tributed between 0 and Nmax , is generated and compared with the phonon occupancy Nc of
the cell associated with the wave vector of the involved phonon. If r < Nc , the transition is
accepted, else the scattering event is treated as a self-scattering one. The computer time for
taking care of the self-scattering events is more than compensated for by the simplification
of the procedure.
Figure 5 shows a cross-section of the simulated LO-phonon distribution function for
AlGaN/AlN/GaN heterostructure at room temperature and 20 kV/cm electric field. The
LO-phonon occupancy exceeds the equilibrium one in the limited q-plane area. Due to
energy and momentum conservation the electrons cannot emit phonons with wave-vector
values close to zero. The distribution of hot phonons is shifted in the direction of the applied
electric field. Strong electronLO-phonon interaction supports a streaming motion of low-
energy electrons. When a lucky electron reaches energy = LO , it shortly emits an
LO phonon, returns to the state near 0 and repeats the acceleration and the LO-phonon
emission many times. The emitted LO phonons are almost identical, they form a sharp peak
at the corresponding q value. The peak is clearly resolved over the washed-out hot-phonon
distribution (Fig. 5).
10
Velocity (10 cm/s)

6
6
4
AlGaN/GaN
2
0
0 5 10 15 20
Electric field (kV/cm)
Figure 6: Monte Carlo results for the electron velocityfield characteristics at 300 K am-
bient temperature without self-heating effect (line) and with self-heating effect included
(triangles) [20]. Circles stand for the experimental data [22].
6 Drift Velocity
Heterostructure channels for high-power devices contain high electron densities and operate
at high electrical fields applied along the channel. The dissipated power density exceeds
tens W/mm, and the channel temperature exceeds the ambient temperaturechannel self-
heating takes place. In AlGaN/GaN channels, the current decreases with time passed after
the step of electric field [19]. The results indicate that the electron drift velocity decreases
as the Joule heat accumulates. The lattice temperature in the channel can be evaluated from
the time-dependent noise power measurements [42].
The experimental dependence of the lattice temperature on the applied electric field for
0.5 s voltage pulses [22] has been used for an illustration of the effect of Joule heat on
the electron drift velocity through Monte Carlo simulation for AlGaN/GaN model with hot
phonons taken into account [20]. Figure 6 presents the electron velocityfield characteris-
tics at 300 K ambient temperature calculated without the self-heating effect (line) and with
the self-heating effect included (triangles).
The results indicate that the self-heating effect is weak at electric fields E <5 kV/cm.
At higher electric fields, the self-heating reduces the electron drift velocity, and the electron
velocity tends to saturate at the electric fields lower as compared with those in absence of
the self-heating. The experimental results (circles) are close to the simulation data obtained
with the self-heating taken into account.
In the following, we shall discuss the results obtained under conditions of negligi-
ble channel self-heating. The corresponding experimental currentfield characteristics are
available for voltage pulses lasting 110 ns [20, 21, 25, 43]. In particular, the results for Al-
GaN/AlN/GaN channels indicate that the self-heating effect is negligible in the field range
up to 100 kV/cm when the pulse duration is 150 ns (Fig. 7, closed triangles).
An example of Monte Carlo simulation of the electron drift velocity for AlGaN/GaN
heterostructure is illustrated in Figure 8 [18]. The simulation shows that electron gas de-
-1 AlGaN/AlN/GaN
10
300K
Current (A)
1 ns
-2
10 0.15 Ps
2 Ps
-3
10
0.1 1 10 100
Figure 7: The current dependence on applied electric field for AlGaN/AlN/GaN het-
erostructure at 300 K lattice temperature. The voltage-pulse duration: 1 ns (open squares),
0.15 s (closed triangles), 2 s (open triangles) [43]. Reprinted with permission from [25].
Copyright (2005) by the American Physical Society.
generacy and hot phonons influence the electron drift velocity. When the degeneracy is
neglected, the electron drift velocity (Fig. 8, dashes) exceeds that obtained for the complete
model (solid line). The absolute increase is the highest at the intermediate electric fields
( 4 kV/cm). Since the electron mean-energy increases with field, the degeneracy effect
tends to decrease as the electric field increases. As a result, the drift velocity curves calcu-
lated with and without considering the degeneracy tend to merge at high fields. When the
hot-phonon effect is neglected, the calculated drift velocity (dots) exceeds considerably the
experimental data (circles) and the simulation results for the complete model (solid line).
The Pauli exclusion principle can affect the drift velocity because some scattering events
are forbidden due to occupancy of the final states. This means that the electron scattering
probability is lower everywhere when the degeneracy is taken into account. Thus, it is
easier for a low-energy electron to acquire the velocity component in the drift direction.
Supposing that this effect prevailed, the drift velocity would increase. However, since more
electrons gain enough energy sufficient for optical phonon emission, the number of LO-
phonon emission events increases despite of the reduced scattering probability, and more
electrons return back into the low-energy region. The net effect on the drift velocity is dif-
ficult to predict as it depends on many circumstances (electron density, lattice temperature,
optical phonon energy, electric field).
The Monte Carlo simulation for AlGaN/GaN at 300 K lattice temperature leads to a
lower drift velocity when the degeneracy is taken into account (Fig. 8, dashes). This result
can be explained in terms of angular dependence of final electron states after events of
LO-phonon emission. The statistics for the angles between the electric field and the final
electron velocity vector at 5 kV/cm electric field are collected and shown in Figure 9. Small
angles are most frequent if the degeneracy is neglected (open circles). When the Pauli
principle is included, the small-angle scattering rate reduces dramatically (cf. closed and
open circles).
14
AlGaN/GaN
12
300 K
Velocity (10 cm/s)

10
6
8
0
0 2 4 6 8 10 12 14
Figure 8: Monte Carlo results for the electron velocityfield characteristics at 300 K. Solid
line takes into account electron gas degeneracy and hot-phonon effects (ph = 1 ps), dashed
line neglects degeneracy and includes hot phonons, and doted line includes degeneracy and
neglects hot phonons. Open circles stand for the experimental results. Reprinted with
permission from [18]. Copyright (2005) by the IOP Publishing Ltd.
Figure 10 gives a schematic reasoning for the enhanced probability for large-angle scat-
tering events. Since the final states for the small-angle scattering are occupied (k ), the
electrons are forced to be scattered at large angles (k ), and this implies a strong negative
contribution to the drift velocity.
The Monte Carlo simulation shows that accumulation of hot phonons has an effect
on the electron drift velocity at high electric fields. The explanation is as follows. The
enhanced occupancy of the involved phonon states (Fig. 5) supports a stronger scattering
of electrons. Indeed, the scattering due to LO-phonon absorption increases because of
hot-phonon reabsorption, and the scattering due to LO-phonon emission increases due to
stimulated emission. This explains the reduction in the drift velocity when hot phonons are
taken into account (cf. dots and solid line in Fig. 8). The stronger is applied electric field,
the more pronounced is the hot-phonon effect.
In response to a step of applied electric field the electron drift velocity overshoots its
steady-state value if the hot phonons are taken into account (Fig. 11, solid line). Indeed,
it takes time for the hot-phonon population to build up and reach the steady state. During
this time, the electronLO-phonon scattering rate gradually increases. As a result, the drift
velocity passes the maximum value and decreases down to its steady-state value. Without
the hot-phonon effect, the electron drift velocity overshoot is almost absent in this range of
electric fields (Fig. 11, dashes).
The transients of electron drift velocity and hot-phonon population can be approximated
by exponential functions (Fig. 4, dashed lines). The electron drift velocity reaches the
steady-state value faster (vdr = 0.3 ps) than the LO-phonon occupancy (ph = 1 ps). In
the linear situation, the relaxation times should coincide. However, the considered electron
hot-phonon processes are essentially nonlinear when the electron gas degeneracy and hot-
4.0
Scattering rate (a.u.)

3.5
3.0
2.5
-180 -120 -60 0 60 120 180

Scattering angle, (deg)
Figure 9: Monte Carlo simulation data for the angle between the electric field and the
electron velocity after a scattering event. Open circles - without Pauli exclusion principle,
full circles - with Pauli exclusion principle. Reprinted with permission from [18]. Copyright
(2005) by the IOP Publishing Ltd.
phonon effects are included. As a result, the electron drift velocity relaxes faster than the
hot-phonon population.
7 Electron Temperature and Noise Temperature

The electron temperature in the biased 2DEG channel is evaluated through simulation of the
electron energy distribution functions. In absence of hot phonons for biased semiconduc-
tors with strong electronLO-phonon interaction, the simulated electron energy distribution
function has a kink near the LO-phonon energy [40]. The electrons with energy, higher than
the LO-phonon energy, quickly emit LO-phonons and appear in the low-energy region. The
electron-temperature approximation fails.
Hot phonons support a strong energy exchange between the electron and the LO-phonon
subsystems. This, together with a weak interaction of hot phonons with the thermal bath,
forms a nearly isolated hot-electronhot-phonon subsystem [22]. Under a steady-state, the
particles of the subsystem reach a near-equilibrium state at the elevated (hot) temperature
that differs from the lattice temperature and the ambient temperature. Here the lattice tem-
perature means the temperature of acoustic phonons strongly coupled with the thermal bath.
The electron distribution functions for the first three subbands of the AlGaN/AlN/GaN
heterostructure are presented in Fig. 12 (symbols). In the first subband, the 2DEG is de-
generate (circles): the occupancy of low-energy electron states is close to unity. The reab-
sorption of hot phonons have nearly washed out the kink near the LO-phonon energy. The
simulated functions can be fitted with a FermiDirac distribution function (Fig. 12, solid
lines). The fitting is good. Two parameters can be extracted from the fitting: the electron
temperature and the Fermi energy. The same values for the electron temperature and the
Fermi energy suit each subband.
kBT
k k
Q k
Figure 10: Electron distribution and possible electron states after LO-phonon emission.
Solid line is the Fermi energy, dots stand for the possible final states. Reprinted with per-
mission from [18]. Copyright (2005) by the IOP Publishing Ltd.
This close proximity of the hot-electronhot-phonon subsystem to the equilibrium at

the elevated temperature is essential for electric fluctuations [44]. Hot-electron velocity
fluctuations cause noise in all directions in the 2DEG plane. In addition to this source, a
great variety of noise sources is agitated in the bias direction. For example, any resistance
fluctuation modulates the bias current and generates a modulation-type excess noise in the
bias direction. The modulation-type sources cause no noise in an isotropic 2DEG in the
transverse direction to the current.
In the electron-temperature approximation, the effective temperature of electron gas Te
has to be substituted for the absolute temperature T0 into the Nyquist theorem in order to
account for the chaotic motion of hot electrons enhanced by the applied electric field. For
brevity, let us call this source the hot-electron thermal noise. In an isotropic 2DEG, the
hot-electron thermal noise is the only source of noise acting in the direction transverse to
the mean current. Thus, the transverse noise temperature equals the electron temperature
[2]:
Tn = Te (29)
The modulation-type sources of the excess noise appear in the bias direction together
with the hot-electron thermal noise:
Tn = Te [1 + C(Te )] (30)
where Tn is the longitudinal noise temperature, C(Te ) is the correction term that takes
into account the noise sources acting in the bias direction only. Real-space-transfer noise,
intervalley transfer noise, impact ionization noise are examples of modulation-type sources
2.5
AlGaN/AlN/GaN
300K
2.0
without hot phonons
Velocity (10 cm/s)

1.5
7
with hot phonons
1.0
0.5
0.0
0 1 2 3 4 5
Time (ps)
Figure 11: The time-dependent electron drift velocity at 20 kV/cm electric field: with hot-
phonon effect (solid line) and without hot-phonon effect (dashed line).
important at high electric fields at microwave frequencies. In the range of subthreshold

fields for these sources, the main correction goes for electron-temperature fluctuations. The
noise temperature anisotropy due to the electron-temperature fluctuations has been recently
treated theoretically for a biased degenerate 2DEG [45]. These inevitable fluctuations cause
resistance fluctuations if the currentvoltage dependence is not linear. Thus, the associated
noise appears in the bias direction together with the hot-electron thermal noise.
Figure 13 presents some results of Monte Carlo simulation of the noise temperature
anisotropy [44]. For the chosen AlGaN/AlN/GaN model, the hot-electron thermal noise
and the electron-temperature fluctuations are the only sources of noise in the considered
range of supplied power. When the hot-phonon effect is not included into the Monte Carlo
Simulation (closed squares), the longitudinal noise temperature is considerably higher as
compared with the transverse noise temperature. The accumulation of hot phonons causes
an essential reduction of the anisotropy (open squares). These results suggest that the elec-
tron temperature can be obtained from the experimental data on the longitudinal noise tem-
perature with a better than 10 % accuracy in the range of fields where real-space transfer,
intervalley transfer and other similar sources of noise do not manifest themselves. Since
there is no anisotropy at equilibrium, the obtained very low anisotropy of the noise tem-
perature supports the idea that the mentioned strong electronhot-phonon interaction and
the weak coupling with the thermal bath are responsible for formation of the quasi-isolated
hot-electronhot-phonon subsystem where the hot electrons and the hot phonons are almost
at equilibrium at the elevated (hot-electron) temperature [44].
Figure 14 presents the noise temperature for the 2DEG channels in AlGaN/GaN and
AlGaN/AlN/GaN heterostructures as a function of the applied electric field [25]. In Al-
GaN/GaN, the main sources of noise are caused by the thermal motion of the electrons, by
the electron-temperature fluctuations, and by the electron sharing [46]. At electric fields up
to 7 kV/cm, the noise temperature is nearly the same for AlGaN/GaN and AlGaN/AlN/GaN
heterostructures. In this range of fields, the electrons are confined in the GaN in both het-
0
10
AlGaN/AlN/GaN
Electron distribution function

13 -2
1.4 10 cm
-1
300 K
10
-2
10
-3
10
0.0 0.1 0.2 0.3 0.4 0.5
Electron kinetic energy (eV)
Figure 12: The simulated hot-electron distribution functions for AlGaN/AlN/GaN het-
erostructure at 300 K lattice temperature and 10 kV/cm applied electric field. Circles,
squares, and triangles stand for the first, second, and third subband, respectively. The
solid lines are fitted FermiDirac distribution functions. Estimated electron temperature
is Te = 590 K. Reprinted with permission from [25]. Copyright (2005) by the American
Physical Society.
1.6
1.5 AlGaN/AlN/GaN
300 K
1.4
1.3
TII/TT
1.2
1.1
1.0
0.9
0.1 1 10 100
Supplied power (nW/electron)
Figure 13: Monte Carlo simulation of dependence of noise temperature anisotropy on sup-
plied electric power: hot-phonon accumulation taken into account (open squares) and hot
phonons neglected (closed squares)[44].
4000 300 K 4000

AlGaN/GaN
10 GHz
Electron temperature (K)

Noise temperature (K)
3000 3000
2000 2000
1000 AlGaN/AlN/GaN 1000

experiment
MC
0 0
0 10 20 30 40 50 60
Figure 14: The experimental results for field-dependent hot-electron longitudinal noise tem-
perature at 10 GHz frequency at 300 K temperature for AlGaN/GaN (open squares, left
axis) and AlGaN/AlN/GaN (open circles, left axis) heterostructures. The solid lines guide
the eye. The closed triangles (right axis) represent the electron temperatures evaluated from
the Monte Carlo results. The vertical bars indicate errors. Reprinted with permission from
[25]. Copyright (2005) by the American Physical Society.
erostructures. At a higher electric field, the difference appears: the electron sharing starts,
and the additional noise source due to electron sharing emerges in the AlGaN/GaN het-
erostructure (Fig. 14, open squares). The AlN barrier prevents electron penetration into
AlGaN barrier, and the noise due to electron sharing is absent at electric fields under in-
vestigation (Fig. 14, open circles). The Monte Carlo results on the electron temperature
(triangles) are close to the experimental results on the longitudinal noise temperature [25].
8 Power Dissipation
The dissipated power can be directly evaluated during the Monte Carlo simulation. After
each electron scattering event, the change of electron energy during the scattering event
is recorded. The total energy change is normalized to the simulated electron number and
the simulated time of the electron motion. The results for AlGaN/AlN/GaN are shown as
closed triangles in the Fig. 15.
Under the steady-state conditions, the power received from the applied electric field
equals the power dissipated by the electrons. The experimental results on the supplied
power are in a good agreement with the Monte Carlo results on the dissipated power. This
indicates that the main scattering mechanisms responsible for the power dissipation are
included into the Monte Carlo model. The only fitting parameter is the LO-phonon lifetime.
The best agreement is obtained for the hot-phonon lifetime ph = 1 ps.
1000/Te (1/K)
3
0 1 2 3 4 3
10 10
300 K
2 AlGaN/GaN 2
Supplied power (nW/electron)

10 10
Dissipated power (nW)

experiment
1 1
10 10
0 0
10 10
-1 -1
10 AlGaN/AlN/GaN 10
-2
experiment -2
10 MC 10
-3 -3
10 10
-4 -4
10 10
0 1 2 3 4
1000/Tn (1/K)
Figure 15: The supplied electric power per electron against inverse noise temperature (ex-
perimental results, down and left axes). The open circlesAlGaN/AlN/GaN heterostruc-
ture, the open squaresAlGaN/GaN heterostructure. The Monte Carlo results (close trian-
gles, up and right axes) stand for the dissipated power as a function of the inverse electron
temperature for AlGaN/AlN/GaN heterostructure. Reprinted with permission from [25].
Copyright (2005) by the American Physical Society.
9 Conclusion
Monte Carlo simulation of channel self-heating, hot-phonon accumulation and electron gas
degeneracy on electron drift velocity is studied for nitride 2DEG channels at room tempera-
ture. The degeneracy reduces the electron drift velocity at low and moderate electric fields,
and this effect is mainly caused by the angular dependence of the scattering rate: the elec-
trons moving at a small angle to the electric field direction prefer to be scattered at a large
angle. Hot phonons and self-heating reduce electron drift velocity at high electric fields due
to increased phonon scattering.
Under conditions of negligible self-heating, the investigation of microwave noise shows
that AlN interbarrier prevents the hot-electron penetration into the adjacent AlGaN barrier.
Simulation results indicate that hot phonons slow down the hot-electron power dissipation.
A reasonable agreement with the experimental data is obtained when electron gas degener-
acy and hot-phonon effect are included into Monte Carlo simulation. Electron temperature
and power dissipation measurements yield a more reliable estimate of the hot-phonon life-
time as compared with currentvoltage measurements.
Accumulation of hot phonons reduces the anisotropy of hot-electron noise temperature.
Therefore, the electron temperature can be obtained from the measured longitudinal noise
temperature in the electric field range where the electron sharing effect is not present.
Acknowledgements
This work has been partly funded by the US Office of Naval Research through ONR Award
No. N00014-03-1-0558 monitored by Dr. Colin Wood, the Lithuanian National Foundation
for Science and Education (Contract No. C-33/2005), and the European Commission within
the Network of Excellence SINANO (IST-506844).
References
[1] Morkoc, H. Nitride Semiconductors and Devices; Springer: Heidelberg, 1999.
[2] Hartnagel, H. L.; Katilius, R., & Matulionis A. Microwave Noise in Semiconductor
Devices; John Wiley & Sons: New York, 2001.
[3] Jacoboni, C.; Lugli, P. The Monte Carlo Method for Semiconductor Device Simula-
tion; Springer: Wien, 1989.
[4] Laux, S. E.; Fischetti, M. V. Monte Carlo Device Simulation: Full Band and Beyond;
Kluwer: Boston, 1991.
[5] Moglestue, C. Monte Carlo Simulation of Semiconductor Devices; Chapman and Hall:
London, 1993.
[6] Shah, J. In Spectroscopy of Nonequilibrium Electrons and Phonons; Shank, C. V., &
Zakharchenya, B. P.; Ed.; North Holland: Amsterdam, 1992; pp 57-112.
[7] Kash, J. A. ibid.; pp 113-168.
[8] Lugli, P. ibid.; pp 1-56.
[9] Ambacher, O.; Foutz, B.; Smart, J.; Shealy, J. R.; Weimann, N. G.; Chu, K.; Mur-
phy, M.; Sierakowski, A. J.; Schaff, W. J.; Eastman, L. F.; Dimitrov, R; Mitchell, A.,
& Stutzmann M. J. Appl. Phys. 2000, 87, 334-344.
[10] Wraback, M.; Shen, H.; Carrano, J. C.; Collins, C. J.; Campbell, J. C.; Dupuis, R. D.;
Schuman, M. J., & Ferguson, I. T. Appl. Phys. Lett. 2001, 79, 1303-1305.
[11] Wraback, M.; Shen, H.; Rudin, S.; Bellotti, E.; Goano, M.; Carrano, J. C.;
Collins, C. J.; Campbell, J. C., & Dupuis, R. D. Appl. Phys. Lett. 2003, 82,
3674-3676.
[12] Eastman, L. F.; Tilak, V.; Kaper, V.; Smart, J.; Thompson, R.; Green, B.; Shealy, J. R.,
& Prunty, T. Phys. Status Solidi A 2002, 194, 433-438.
[13] Mishra,U.K. In 62nd Device Research Conference Digest;(IEEE Cat. No. 04TH8724),
2004, pp 3.
[14] Eastman, L. F. In Proc. 29th WOCSDICE; Cardiff, Wales, 2005; pp 57-58.
[15] Mishra, U. K. ibid.; pp 107-110.
[16] Vertiatchikh, A.; Schaff, W. J., & Eastman L. F. In International Symposium on Com-
pound Semciconductors; (IEEE Cat. No. 03TH8675), 2003, pp 53-54.
[17] Oxley, C. H., & Uren, M. J IEEE Trans. Electron Dev. 2005, ED-52, 165-169.
[18] Ramonas, M.; Matulionis, A., & Rota, L. Semicond. Sci. Technol. 2003, 18, 118-124.
[19] Gaska, R.; Osinsky, A.; Wang, J. W., & Shur, M. IEEE Trans. Electron Dev. Lett.
1998, 19, 89-91.
[20] Ardaravicius, L.; Matulionis, A.; Liberis, J.; Kiprijanovic, O.; Ramonas, M.; East-
man, L. F.; Shealy, J. R., & Vertiatchikh, A. Appl. Phys. Lett. 2003, 83, 4038-4040.
[21] Barker, J. M.; Ferry, D. K.; Goodnick, S. M.; Koleske, D. D.; Allerman, A., &
Shul, R. J. J. Vac. Sci. Technol. B 2004, 22, 2045-2050.
[22] Matulionis, A.; Liberis, J.; Matulioniene, I.; Ramonas, M.; Eastman, L.F.; Shealy,
J. R.; Tilak, V., & Vertiatchikh, A. Phys. Rev. B 2003, 68, 035338.
[23] Gokden, S.; Balkan, N., & and Ridley, B. K. Semicond. Sci. Technol. 2003, 18,
206-211.
[24] Ridley, B. K.; Schaff, W. J., & Eastman L. F. J. Appl. Phys. 2004, 96, 1449-1502.
[25] Ramonas, M.; Matulionis, A.; Liberis, J.; Eastman, L. F.; Chen, X.; Sun, Y.-J. Phys.
Rev. B 2005, 71, 075324.
[26] Chu, R. M.; Zhou, Y. G.; Zheng, Y. D.; Han, P.; Shen, B., & Gu, S. L. Appl. Phys.
Lett. 2001, 79, 2270-2272.
[27] Ridley, B. K. Rep. Prog. Phys. 1991, 54, 169-256.
[28] Kolnik, J.; Oguzman, I. H.; Brennan, K. F.; Wang, R.; Ruden, P. P., & Wang, Y. J.
Appl. Phys. 1995, 78, 1033-1038.
[29] Ridley, B. K.; Foutz, B. E., & Eastman, L. F. Phys. Rev. B 2000, 61, 16862-16869.
[30] Bulutay, C.; Ridley, B. K., & Zakhleniuk, N. A. Phys. Rev. B 2000,62, 15754-15763.
[31] Price, P. J. Ann. Phys. 1981, 133, 217-239.
[32] Goodnick, S. M.; Lugli, P. Phys. Rev. B 1988, 38, 10135-10138.
[33] Reggiani, L.; Kuhn, T., & Varani, L. Appl. Phys. A 1992, 54, 411-427.
[34] Adams, J. G.; Tang, T. IEEE Electron Device Letters 1992, 13, 378-380.
[35] Bosi, S.; Jacoboni, C. J. Phys. C 1976, 9, 315-319.
[36] Lugli, P.; Ferry, D. K. IEEE Trans. Electron Dev. 1985, ED-32, 2431-2437.
[37] Lugli, P.; Jacoboni, C.; Reggiani L., & Kocevar, P. Appl. Phys. Lett. 1987, 50,
1251-1253.
[38] Mickevicius, R., & Reklaitis, A. Solid State Communications 1987, 64, 1305-1308.
[39] Ramonas, M., & Matulionis, A. Semicond. Sci. Technol. 2004, 19, S424-S426.
[40] Paulavicius. G.; Mitin, V. V., & Bannov N. A. J. Appl. Phys. 1997, 82, 5580-5588;
[41] Shi, L.; Ponce, F. A., & Menendez, J. Appl. Phys. Lett. 2004, 84, 3471-3473.
[42] Ardaravicius, L.; Liberis, J.; Matulionis, A.; Eastman, L. F.; Shealy, J. R., & Verti-
atchikh, A. Phys. Stat. Sol.(a) 2004, 201, 203-206.
[43] Ardaravicius, L.; Ramonas, M.; Kiprijanovic, O. Liberis, J.; Matulionis, A.; East-
man, L. F.; Shealy, J. R.; Chen, X., & Sun, Y. J.Phys. Stat. Sol. (a) 2005, 202,
808-811.
[44] Matulionis, A.; Liberis, J., & Ramonas, M. AIP Conf. Proc 2005, CP780, 105.
[45] Katilius, R. Phys. Rev. B 2004, 69, 245315.
[46] Matulionis, A., & Liberis, J. IEE Proc.-Circuits Devices Syst. 2004, 151, 148-154.
Chapter 5
RESEARCHES OF GROWTH AND DEVICE BASED

ON ZNO BY PLASMA ASSISTED MOLECULAR
BEAM EPITAXY
Y. M. Lu, D. Z. Shen, J. Y. Zhang, Y. C. Liu and X. W. Fan

Laboratory of Excited State Processes, Chang Chun Institute of Optics,
Fine Mechanics and Physics, Chinese Academy of Sciences, 130033,P.R.China
Abstract
In this paper, the ZnO based materials, such as ZnO thin film and its heterostructure,
were prepared on c-plans sapphire (Al2O3) substrates by plasma-assisted molecular beam
epitaxy (P-MBE). The influence of growth temperature on ZnO film quality was investigated.
The result reveals a change of the growth mode from three-dimensional (3D) nucleation to
two-dimensional (2D) nucleation. The photoluminescence (PL) spectra exhibit a strong
ultraviolet (UV) emission and a weaker visible emission for all samples. By the measurements
of PL spectra at different temperature, the origin of UV emission peak at RT is considered to
be from free exciton emission and the shift of UV emission peak position is attributed to
quantum confinement effect due to different crystal grain sizes. In addition, the carrier
concentration of ZnO thin films decreases with increasing growth temperature. In the
optimum growth condition, the carrier concentration is N=7.66 1016 / cm3 which is closed to
that of bulk ZnO.
On the other hands, N-doped p-type ZnO thin films were grown by P-MBE on c-plane
Al2O3 using radical NO as oxygen source and nitrogen dopant. The reproducing ZnO thin
films have maximum hole concentration of 1.21019cm-3and lower resistivity of 9.95cm. In
absorption spectra, the subbandgap related to N in the ZnO band gap was observed.
Comparing with optical emission spectra (OES) of radical nitrogen, a difference between
radical NO and N2 was found. The OES of radical nitrogen show strong ultraviolet emission
related to radical nitrogen (N2*), which is a shallow double donor in the ZnO films. While the
radical atoms (N*) are dominant in the OES of radical NO. The experiment result indicates
that p-type ZnO is realized by employing radical NO as N dopant.
A n-ZnO/p-GaN heterostructure and its light emitting diode were fabricated by P-MBE.
And the optoelectrical properties of the heterojunction are investigated. The device exhibited
rectifying I-V characteristics to the diode. Under forward bias voltage, electroluminescence
spectrum shows that a blue emission from GaN layer at the room temperature. To improve the
luminescence properties of the device and make emission from ZnO layer, we designed the
124 Y. M. Lu, D. Z. Shen, J. Y. Zhang et al.
device structure and fabricated a p-GaN/i-ZnO/n-ZnO p-i-n heterojunction. Comparing with

p-n heterojunction, the emission peaks shifts to high-energy side in the EL spectrum of p-i-n
heterojunction. The origin is attributed to the emissions from i-ZnO layer of p-i-n
heterojunction.
1 Introduction
For some time now there has been a great deal of attention in wide band-gap semiconductors,
in particular there is strong commercial desire to produce efficient and lasting blue-light
emitting diodes (LEDs) and short-wavelength laser diodes (LDs). World-wide research
efforts were initially focused on ZnSe and its alloys[1,2], but recently, GaN based
technologies have progressed more rapidly[3]. These developments have culminated in the
demonstration of room-temperature (RT) operating green-blue as well as blue LD structure[4-
6]. ZnO, as an oxide, is another wide band semiconductor material with a direct band gap of
3.37eV at RT[7]. It is superior over nitrides and selenides in chemical and thermal stability
and in resistance to chemical attack and oxidation. It has a high exciton binding energy of 60
meV[8], which in principle should allow efficient excitonic lasing mechanisms to operate at
RT. Ultraviolet (UV) stimulated emission and lasing have been extensively reported in ZnO
microcrystallite thin films[9] and its quantum well structure[10]. To realize these device
applications, an important issue is to fabricate both high-quality p-type and n-type ZnO films.
However, like most wide band-gap semiconductors, ZnO has the asymmetric doping
limitation, i.e., it can be an easily doped high-quality n-type, but it is difficult to dope p-type.
This is due to the existent of some problems such as its self-compensating effect, deep
acceptor level, and low solubility of the acceptor dopant. For the realization of p-type ZnO, it
is necessary to grow high crystalline quality ZnO thin film. Therefore, the extensive
researches were reported on the growth of ZnO thin films by various growth techniques
including magnetron sputtering [11], chemical vapor deposition (CVD) [12], and plasma-
molecular beam epitaxy (P-MBE) [13]. However, it has hardly been obtained for the
preparation of high quality ZnO thin films. This is due to that insufficient supply of oxygen
during growth results in the formation of many structural defects such as oxygen vacancies
(VO) or Zn interstitials (Zni) in ZnO films [14,15]. On the other hand, the 18% lattice
mismatch between ZnO and Al2O3 substrate[13,16] will strongly influence the structural,
optical and electrical properties of ZnO thin films. So far, many researches on the growth of
high-quality ZnO thin films are reported, including accommodation of II/VI ratio [14, 17],
post-deposition annealing [15,18], low-temperature buffer layers [16, 19], MgO buffer layers
[17, 20] and polarity control [18, 21]. Although high temperature growth has revealed the
improvement of ZnO crystal quality[20], the systemic study on effects of growth temperature
on optical and electrical properties of ZnO grown directly on Al2O3 substrates have been not
reported. In this paper, the influence of growth temperature on ZnO film quality grown on
Al2O3 substrates was investigated by the measurements of optical and electrical properties.
The result indicates that the crystalline quality of ZnO films was improved with increasing
growth temperature.
The unintentionally doped ZnO usually shows n-type conduction, which is related to
either presents of native donor defects, such as O vacancies (VO) or Zn interstitials (Zni) [22].
These native defects have been demonstrated that they have lower formation energy than
other acceptor defects (O interstitials or Zn vacancies VZn) whether in Zn rich condition or not
Researches of Growth and Device Based on ZnO by Plasma Assisted 125
[15]. For Compensation of these native defects, all of the common group V elements, i.e.,
N[23], P[24] and As[25], have been used to make p-type ZnO. In particular, N was
established as being the more soluble group-V impurity. The theory predicts that N is a better
candidate in current research for p-type doping of ZnO [26]. Under proper conditions, it is
expected to form nitrogen substitution in the oxygen site (NO), which is a single shallow
acceptor [27]. However, surprising behavior was uncovered: Doping with N2O or N2 source
led to n-type, not p-type conduction [28-30], while doping with an NO2 or NO source without
the use of plasma led to p-type behavior successfully by MOVCD technology [31,32]. This is
due to the difficult for breaking N2 bond, so the (N2)O was easily formed, which is a shallow
double donor. For NO molecules, it is more easily to provide single-N atoms. (N2)O centers
are difficult to be formed due to the tiny probability of the process of two N atoms arriving
simultaneously at the same site on the growth surface [27]. In this work, we successfully
prepared p-type ZnO thin films by P-MBE using NO as both O source and N dopant. The
significant difference between our experiment and reports [31] is that the N-O bond was
broken by an r.f. atomic source before the molecules arrived to the substrate. In an optimum
growth temperature, it is found that the p-type ZnO films can be prepared easily by activated
NO doping and the results have good repeatability. Additionally, compared to direct NO-
doping [31], the use of activated NO can decrease distinctly the growth temperature for p-
type ZnO thin films.
Although many groups reported to have successfully made p-type ZnO by using group V
elements, such as N[23], P[24], As[25] and so on, the hole concentration or mobility was too
low to meet the demands of device fabrication. We should also mention here that there is a p-
type ZnO with lower resistivity using NO as dopant. However, there is still no progress in
homo-junction based ZnO p-n diodes. So far, only two results of ZnO p-n homo-junction with
weak bluish-white light emission and weak white-violet emission were reported [33,34]. To
take the advantage of ZnO, many researchers tried to use ZnO heterojunction devices,
including p-n heterojunction such as n-ZnO/p-transparent oxide semiconductor (p-
SrCu2O2[35], p-ZnRu2O[36], Cu2O[37]), n-ZnO/diamond[38] and n-ZnO/p-Si[39], and MIS
diode with insulator formed by N+ ion implantation[40]. Among numerous candidates, GaN is
with the same crystal structure as that of ZnO and with similar band gap energy at room
temperature (ZnO 3.37 eV, GaN 3.4 eV). Due to the accessibility of high quality p-type GaN,
it starts to draw great attention for the fabrication of ZnO heterojunction. Stronger emission
has been observed in n-ZnO/p-GaN heterojunction, although the emission of the p-GaN still
plays dominant role in the EL spectra [41] due to the large density of electrons injected into
the p-GaN layer. Recently, Ya. I. Alivov [42] reported the use of AlGaN as the hole injection
layer in the ZnO heterojunction to effectively block the electron, resulting in the successful
observation of the exciton emission originated from the ZnO. Since the epitaxially grown
ZnO is intrinsically n-type with high electron concentration, it is difficult to make ZnO p-i-n
heterostructure with the recombination region in the ZnO layer. In this paper, we make use of
the high resistivity characteristic of the nitrogen-doped ZnO to design and fabricate p-i-n
heterojunction. By inserting the insulating N-doped ZnO layer between the n-ZnO and the p-
GaN layers, the electrons and holes are believed to be blocked inside the N-doped ZnO. In the
EL spectra, the near ultraviolet emission band was observed at room temperature for the p-i-n
structure, which is different from the blue emission for the p-n junction samples. The
recombination mechanism will also be discussed based on the comparison between the PL,
EL and absorption spectra.
2 Experiments
Al2O3 substrate was cleaned by acetone and methanol for 10 min, respectively, and then
etched by solution of H2SO4:H3PO4=3:1 for 10 min, followed by a rinse in deionized water
and dried by the high-pure(5N) nitrogen gas. The chemical cleaned substrates were degassed
at 650 in ultrahigh vacuum atmosphere (<110-6 Pa) for about 30 min. A Knudsen effusion
cell is used to evaporate elemental zinc with 99.9999% purity. Atomic oxygen is activated
from ultrapure O2 gas (99.999%) by an rf-plasma source with an electrostatic ion trap
operated at 500V. Before growth, the substrate that mounted on a molybdenum holder using
indium as solder was treated by O-plasma at 650 for 30 min, which is expected to remove
surface contaminant and produce oxygen terminated Al2O3 (0001) surface[13]. During
growth, the Zn source flux and O2 partial pressure in the growth chamber are kept at 410-5
Pa and 610-3 Pa, respectively. The growth temperature of ZnO film was varied between
350C and 700 C. For the growth of p-type ZnO films, two gas sources were used,
respectively, i.e., N2 (99.9999%) + O2 (99.9999%), and NO(99.99%). Zn beam pressure
is fixed at 410-4 Pa and the substrate temperature was kept at 400 C. The thicknesses of
all the samples are about 200-400 nm. The growth process was monitored by in-situ RHEED.
For the growth of the p-i-n heterojunction using MBE, Mg doped p-type GaN grown on
c-plane sapphire by metal organic chemical vapor deposition (MOCVD) served as the
substrate. The hole concentration and mobility, determined by Hall effect measurement, were
p=2.7 1017cm-3 and =5.1cm2 V-1S-1. The epilayers were consisted of 20nm insulating ZnO
followed by 200 nm n-type ZnO. In order to grow the insulator layer of p-i-n structure, N2
gas was introduced into the growth chamber through the same r.f. plasma as that of the
oxygen. The as-grown ZnO showed n-type conductivity (n=2 1018cm-3, U=0.6cm) in our
samples due to the intrinsic donor defects, such as Zni or VO3. Just like previous reports
[40,43], nitrogen doped ZnO showed high resistivity characteristics (U=4650cm). To make
ohmic contact, indium electrode was fabricated on the n-type ZnO film by soldering at 300 0C
for 10 min. Ohmic contact to p-type GaN was made by thermal evaporation of Ni/Au and
thermal anneal at 4000C for 15min in oxygen gas.
The quality of the grown samples was characterized by the Rigaku Company O/max-RA
X-ray system, D/Max 2400 double crystal diffractometer at 40 kV and 98 mA and VG
ESCALABMK XPS system. The surface morphology was analyzed by the measurement of
atomic force microscope (AFM). The optical properties of all samples were studied by
absorption and photoluminescence (PL) spectra. UV absorption spectra were measured using
the SHIMADZU UV-3101 PC spectrophotometer. A JY63 Micro Raman spectrometer was
employed for PL measurement. The luminescence was detected by a charged-coupled device
(CCD) detector. The 325nm line of a He-Cd laser was used as the excitation source. The
liquid-nitrogen cooling system was used in conjunction with the sample stage to cool a
sample down to 80K. The temperature was controlled by the TMS94 from 80K to room
temperature. The ZnO/GaN diodes were excited by pulse current in forward bias with the
pulse width of 0.2ms and repetition frequency of 160Hz. The EL spectra were measured using
a grating monochromater of Model Spex 1404 with an RCA-C31034 cooled photomultiplier.
The electrical properties of as-grown samples were measured by the Hall system of
LakeShore Company with Model 7707 at the field strength of 3200 Gauss.
3 Results and Discussion

A Growth of High Quality ZnO Thin Filmes
Fig.1 shows the RHEED patterns of the substrate exposed to O-plasma at 650 for 30 min
and the samples grown at 350 , 550 and 650 , respectively. After plasma treatment, the
substrate gives a streaky pattern (Fig.1a), which indicates a well-ordered and flat Al2 O3
surface. For the grown samples, RHEED results exhibit different patterns with the changing
of growth temperature, as seen in Fig.1b~Fig.1d. At growth temperature of 350 , a spotty
pattern indicates a rough surface consisting of ZnO crystallites. When the growth temperature
increases to 550 , a elongated spotty pattern was observed. While at high temperature (above
650 ), sharp streaky RHEED patterns show the formations of the smooth ZnO surfaces. Y.
Chen et al. [44] had reported a gradual change of the RHEED patterns from streak to spotty
during the growth stages for ZnO film, which was due to the transition of the epitaxial growth
mode from 2D nucleation to 3D nucleation. Compared to our results, the RHEED patterns
reveals that the growth mode changes from 3D to 2D with increasing growth temperature.
(a) (b)
(c) (d)
Fig.1. The RHEED patterns of Al2O3 substrate treated by O-plasma (a) and ZnO thin films grown at
350 (b), 550 (c) and 650 (d).
To understand the effect of the growth temperature on surface morphology, the ZnO thin
film surfaces were analyzed by AFM measurements. Fig.2 shows the 2m2m surface
roughness images of the samples grown at 350 , 450 , 550 and 650 , respectively. With
increase of the growth temperature, the root-mean-square (RMS) roughness of ZnO samples
gradually decreased from 6.5 nm to 3.4 nm. The change of the surface roughness indicates the
difference of the epitaxial growth mode, in which the formation of 2D growth is easy at high
temperature. The growth mode dependent on the growth temperature can be clearly seen by
the measurements of the mean crystal grain sizes. At the growth temperatures of 350 and
450 , the mean crystal grain sizes are in range of 20-25 nm and 25-30 nm, respectively. Be
close to growth temperature of 550 , the crystal grain size is more than 100nm. For growth
temperature of 650 , the crystal grain size becomes upwards of 200nm. The temperature
dependence of the grain size can be explained by the migration and the diffusion of the
reactants. At low temperature, the zinc and oxygen atoms can be absorbed easily on the
substrate surface and diffuse into the nucleation sites to form ZnO film. However, the lower
migration and diffusion rates will lead to 3D nucleation. In this case, ZnO epilayer has a small
crystal grain size and a rough surface. With the increasing temperature, the enhancement of
the atomic migration ability results in the increasing of the grain size and the decreasing of
the surface roughness. This evolution reveals a transition of the growth mode from the 3D
growth to 2D growth with increasing growth temperature. The 2D growth would be
preferable for the growth of high-quality ZnO films[45].
(a) (b)
(c) (d)
Fig.2 AFM images of ZnO thin films grown at (a)350 , (b)400 ,(c)550 and (d)650 .
Table 1 The crystal quality and optoelectrical properties of the ZnO samples grown on Al2O3
substrates at different temperatures.
Growth temperature FWHMs of ZnO (002) Intensity ratio of UV Carrier

of samples differaction peaks emission and visible emission concentration
350 0.84867 0.36 1.061019/cm3
400 0.79063 1.68 1.921018/cm3
450 0.77077 2.96 1.871018/cm3
500 0.61130 6.10 9.801017/cm3
550 0.47916 10.86 9.191016/cm3
600 0.44731 43.20 8.161016/cm3
650 0.29392 82.70 7.661016/cm3
X-ray diffraction spectra of the ZnO films give only diffraction peak of ZnO (00l)
orientation, indicating the formation of wurtzite structural ZnO with a preferential orientation.
In X-ray rocking curve spectra of ZnO (002) diffraction peak, the full width at half maximum
(FWHM) decreases from 088 to 0.20 with increasing growth temperature from 350C to
650 C, as is shown in Table 1. While at 700 , the FWHM becomes up to 0.78. Fig.3 shows
the PL spectra of all as-grown samples at RT. As is clearly seen in Fig.3, the main features of
the PL spectra are similar for all samples. A strong ultraviolet (UV) emission and a weak
visible emission in the PL spectra are observed. The UV emission is considered to be from
free exciton emission, which will be identified by the temperature-dependent PL spectra
discussed below. The visible emission located at around 2.500eV is usually associated with
intrinsic defects such as oxygen vacancies (VO) [15]. As the growth temperature increases
from 350 to 650 , the FWHM of UV emission band becomes narrow and relative intensity
of the visible emission decreases, indicating the improvement of the ZnO thin film quality
with increasing growth temperature. Table 1 gives the integral intensity ratio of UV emission
and visible emission. For the ZnO films grown at 350 and 400 , the energy positions of
UV emission peaks are located at 3.357 eV and 3.341 eV, respectively. The blue-shifts are
observed by comparing with other samples grown in temperature range of 450 ~650 , in
which positions of the UV peaks are located at about 3.295 eV. When the growth temperature
is 700 , the FWHM of the UV emission band becomes broad and the intensity of the visible
emission increases, as seen in Fig.3h. We note that UV emission peak at 700 shows a
noticeable red-shift (>80meV). At high temperature, because of the reevaporation of surface
atoms and the mismatch of thermal expansion coefficient lead to high density of structural
defects and misfit dislocations formed in ZnO films. It is suggested that the UV emission
peak at 3.212eV is related to the bound excitons due to these defects as deep centers. The
above results indicate that the quality of ZnO thin film at 700 is relatively poor.
(a)
(b)
(c)
PL Intensity (a.u.)
(d)
(e)
(f)
(g)
(h)
2.4 3.0 3.6

Energy(eV)
Fig.3 The PL spectra at RT of ZnO thin films grown at (a)350, (b)400, (c) 450, (d)500, (e) 550, (f) 600,
(g) 650 and (h)700 .
The room temperature electrical properties of ZnO films grown at different temperatures
were measured by the four-probe van der Pauw method. Based on these measurements, all
samples show n-type characteristics. The room temperature carrier concentration as function
of the growth temperature was shown in Table 1. As the growth temperature was increased
from 350 to 650 , the carrier concentration decreases from 1.06 1019 /cm3 to 7.66 1016
/cm3. At growth temperature of 700 , the carrier concentration increases to N=8.5 1018 /
cm3. Joshy Jose et al. reported that the grain boundary would influence carrier concentration
[46]. Because the defects can be very easily concentrated in the grain boundary, the sample of
smaller grain size has a high carrier concentration. With the growth temperature increase, the
changing of carrier concentration with that of the intensity of the visible emission related to
oxygen vacancy [12]. The oxygen vacancy is considered to be an intrinsic donor in ZnO.
However, if only oxygen vacancy density affect electrical properties, the measured carrier
concentration is quite surprising. S.B.Zhang et al.[12] had proposed that interstitial Zn (Zni)
as another donor would play an important role in the electrical properties, because the Zni
level is shallower than the level of oxygen vacancy [12]. Zn atoms absorbed on the substrate
surface need enough thermal activated energy to take the lattice sites. A low growth
temperature will lead to the formation of Zni with higher density in ZnO films. In this case,
the Zn atoms diffused into interstitial sites will act as shallow donors. This means that the
coexistence of oxygen vacancy and Zni defects will influence the carrier concentration. With
increasing temperature, these crystal defects contributing to the carrier concentration are
decreased due to the enhancement of the atomic diffusion ability. However, the formation of
thermal defects with high density will result in the increasing of carrier concentration at high
temperature. In the optimum growth condition(650 ), the carrier concentration has a
minimum of N=7.66 1016 / cm3 which is closed to that of bulk ZnO[47].
E2 3.378eV a)
E1 3.392eV
intensity(a.u)
E 3 3 .3 5 8 e V b)
in te n s ity (a .u )
3 .2 0 3 .2 5 3 .3 0 3 .3 5 3 .4 0 3 .4 5 3 .5 0
e n e r g y (e V )
Fig.4 The PL spectra of the ZnO thin films grown at 350 and 650 measured from 82K to 300K.
In order to investigate the origin of UVE peak at RT, dependence of the PL spectra on the
temperature was measured for all samples. Fig.4 gives the measured results of the samples
grown at 350 and 650 . At the temperature of 82K, PL spectrum of the sample grown at 350
shows that two UV emission peaks labeled E1 and E2 were located at 3.378eV and
3.392eV, respectively. For the sample grown at 650 , only one UVE peak E3 at 3.358eV was
observed at 82 K. With increasing temperature, the integral intensity ratio of E1 to E2

increases and the E2 peak disappears around 170K. The temperature dependence of the
integrated PL intensity can be expressed by the following equation [48]
I0
I (T ) (1)
'E
1 A exp( )
kBT
where 'E is the thermal activation energy, k B is the Boltzman constant, I 0 is the emission
intensity at 0K, T is the thermodynamic temperature and A is a contains ratio of optical-
collection efficiencies and effective degeneracy between the unbound and bound states. The
experimental data can be fitted by using the equation (1), as shown in Fig.5 with the solid
lines and broken lines, 'E values of the bands E1, E2 and E3 are 63meV, 20meV and
58meV, respectively. In Ref.49, the free exciton binding energy of bulk ZnO is about 60meV,
which is approximately equal to the 'E values of the bands E1 and E3. Therefore, the origin
of the bands E1and E3 was considered to be from the free exciton recombination. The
excitons bound to neutral donor in ZnO have been reported [49], in which the binding energy
of the bound exciton was about 20 meV. We suggest that the band E2 observed in the range
of low temperature is from donor-bound-exciton transition.
Experimental data of E2 peak

Experimental data of E1 peak
Theoretical fit of E2 peak
Theoretical fit of E1 peak
P L In t e g r a t e In t e n s it y (a .u .)
(a)
E x p e r im e n ta l d a ta o f E 3 p e a k
T h e o r e t ic a l f it o f E 3 p e a k
(b)
2 4 6 8 10 12 14
-1
1 0 0 0 /T (K )
Fig.5 The fitting curve of PL intensity as a function of measured temperature of ZnO thin films grown
at 350 and 650 .
The AFM images have revealed that the mean crystal grain sizes become small with
decreasing growth temperature. It is known that small grain size will lead to the quantum
confinement effect. An expression for the shift of energy gap ( 'E g ) due to the quantum
confinement effect is given by [50]
h 2S 2 P 1.8e 2
'E g | E g 0 (2)
2d 2 Hd
1 1 1
where E g 0 is the energy gap for bulk materials, d is the particle size, ( m e and
P me mh
m h being the electron and hole effective masses, respectively), and H is the dielectric
constant. For ZnO, the effective masses of electrons and holes are 0.38 m0 and 1.80 m0[51],
respectively, and the dielectric constant is 8.75. For the ZnO, the equation (2) can be
expressed as follows:
'Eg | 75.885d 2 (nm)2 1.902d 1(nm) (3)
By the measurement of AFM, the mean crystal grain sizes of the samples grown at 350
and 400 are about 23 and 27nm, respectively. Using equation (3), the calculated Eg values
are 60 and 34meV, respectively. Because the UVE peak at RT is identified to be from the
recombination of free exciton, the energy change of band gap should be equal to that of UV
emission peak. From Fig.3, we obtain that the blue shifts of UV emission peaks for the
samples grown at 350 and 400 are 60meV and 44 meV, respectively. These results are in
agreement with the above calculated values.
B Preparation of the p-type ZnO
Table 2 gives the ZnO samples grown by using different gas source. Sample A is the N-doped
ZnO thin films using radical N2. The N2 and O2 are separated to two different plasma
chambers. The typical N-doped ZnO thin films using radical N2 has a low carrier
concentration 2.51014cm-3 and a high resistivity 900.cm. Although the radical N2 can be
incorporated into the ZnO films and compensate the native donor defects resulting in the
trends toward to p-type, the attempts to obtain p-type ZnO by employing radical N2 are failed,
as reported in Ref. [52-54]. Sample B is the ZnO thin films prepared by employing the radical
NO. Hall effect measurement results indicate that sample B shows p-type conduction with a
high net hole carrier concentration (NA-ND) of 1.21018 cm-3 and lower resistivity of
9.95.cm. The hole carrier concentration varies from 1.01015 to 1.21018 cm-3 under the
different preparing condition, such as growth temperature, flow rate of NO, for the p-type
ZnO. The sample C, undoped ZnO thin films, shows typical n-type conduction with the
carrier concentration of 2.741019cm-3 and a lower resistivity of 0.043cm. The n-type
conduction in native ZnO films is generally accepted arose from either oxygen VO or Zni
because they have low formation energy either in O-rich or Zn-rich conditions [52].
Al2O3(006)
Intensity (a.u)
ZnO(002)
C
25 30 35 40 45 50 55 60 65 70 75
2T(Deg.)
Fig.6 XRD spectra of N-doped ZnO using radical N2 (A), radical NO (B), and undoped ZnO
sample (C).
Table 2 The prepared conditions and electronic properties of N-doped using radical
N2(Sample A) ,radical NO(Sample B),and undoped ZnO (Sample C). FO2 and FNO represent
the flow rate of O2 and NO, respectively.
Sample Gas FO2* FNO* Carrier type Resistivity Mobility Carrier
(sccm) (sccm) .cm cmV-1s-1 concentration(cm-3)
A N2 0.1 1.0 n 24.60 0.89 1.431016
B NO 0 1.0 p 9.95 0. 53 1.261018
C O2 1.0 0 n 0.043 5.34 2.741019
The crystalline structures of as-grown samples were investigated by XRD spectra, as

shown in Fig.6. Besides the diffraction peak of Al2O3(006), only one peak related to wurtzite
structural ZnO(002) can be observed for all samples, which located at 34.40. Although
nitrogen as dopant source for p-type ZnO thin film was usually found to increase the tensile
force and lattice constants for the N-doped ZnO [55], however, obvious change in d value
was not observed in our work. When radical N2 gas or radical NO gas was used for growth of
N-doped ZnO, the intensity of the diffraction peak of ZnO (002) is significantly reduced, and
FWHM is relatively broad comparing with the undoped ZnO thin film (sample C). The
FWHM of ZnO(002) diffraction peak is 0.34for sample A, 0.25for sample B and 0.15for
sample C. The broadening of N-doped ZnO(002) peak should be attributed to the defects
related to incorporation of N in the films, as observed by Li et al.[31].
RT
Absorption intensity(a.u.)
Radical N 2 doped ZnO
Radical NO doped ZnO

Undoped ZnO
400 500 600

Wavelength(nm)
Fig.7 Absorption spectra of N-doped ZnO using radical N2 (A), radical NO (B), and undoped ZnO
sample (C).
Fig.7 shows the absorption spectra for N-doped ZnO and undoped ZnO thin films. It is
obvious that the absorption intensity of N-doped ZnO thin films is smaller than that of the
undoped ZnO thin films. The sample C shows a very sharp absorption edge. While the N
doped ZnO thin films show a much mild absorption edge. Similar result was also observed by
Garces et al. [56] and Li et al. [31], which is concluded that subbandgap levels associated
with N doping present in the band-gap of ZnO.
p-type ZnO thin films

undoped ZnO thin films DAP
9000
FE
DAP-1LO
D0X
Intensity(a.u.)
6000
DAP-2LO
3000
FE-1LO
FE-2LO FE
0
3.0 3.1 3.2 3.3 3.4

Energy (eV)
Fig.8 Photoluminescence spectra of N-doped and undoped ZnO samples.
Fig.8 is the photoluminescence (PL) spectra of p-type and undoped ZnO thin films
measured at 77K. The near-band-gap emission of the sample A is not detectable. The UV
emission of the undoped ZnO (sample C) consists of free exciton band at 3.377eV, neutral-
donor-bound-exciton (D0X) band at 3.360eV and replica lines of free exciton with first and
second-longitudinal optical (LO) phonon at 3.313eV and 3.240eV, respectively [57]. While in
the p-type ZnO (sample B), beside emission of free exciton at 3.377eV, three peaks located at
3.267eV, 3.197eV and 3.124eV were found. These peak positions depend on the N
concentration. Similar result has been observed by Tamura et.al in their work [58], in which
the peak at 3.267eV is assigned to donor-acceptor pair (DAP) transition. Three peaks have
equal energy spacing of about 70meV, which is in good agreement with the energy of LO
phonon of ZnO. This implies that the peaks at 3.197eV and 3.124eV are the first and second-
LO phonon replica of DAP, respectively. The energetic position (EDAP) of DAP transition can
be calculated by the following equation [23]:
e2
EDAP E gap ED E A (4)
4SHr
where Egap, ED and EA are the energy of band gap, the binding energy of donor and the binding
energy of acceptor, respectively. e is the dielectric constant, and r is the pair separation. The
pair separation r can be very roughly estimated by letting <r>~( 3 / 4 S N A ) 1/3. The NA=ND +
1.21018cm-3, and which is estimated to be in the range of 1019-1020 cm-3. The dielectric
constant of ZnO is 8.3, the donor binding energy is known to be about 60meV and the energy
gap at 77K is Egap =3. 437eV [59,60]. Thus, EA168-234meV can be obtained, which is
similar with the reported value of EA=16540meV [61].
O*
Radical NO st
nd
N 2*(2 ) N 2*(1 )
N* O*
Intensity(a.u.)
N*
Radical N 2
N*
N*
300 400 500 600 700 800

Wavelength(nm)
Fig.9 The optical emission spectra of radical NO (a) and nitrogen (b).
To explain the formation mechanism of p-type ZnO, the optical emission spectra (OES)
of the radical N2 and radical NO are investigated. Fig.9 shows the OES of radical NO and N2,
respectively. The OES spectrum of NO is dominated by the emission peaks of atomic oxygen
(O*) located at 777 nm and 841 nm originating from 3p5P-3s5S0 and 3p3P-3s3S0 transitions,
and emission of atomic nitrogen (N*) located at 744 and 821nm from 3S4 P-3P4S0 and 3S4 P-
3P4 P0 transitions, respectively [56]. Further, a very weaker emission peaks of radical
nitrogen moleculars (N2*) originating from the second positive (C3 u-B3 g) and first
positive (B3 u-A3 u+) transition were also observed in the wavelength regions of ultraviolet
and visible light [26]. While the OES of radical N2 is dominated by the emission peaks of
second positive and first positive transition (N2*). Namely, the chemical species in radical
NO and radical N2 are different, this may be the reason for the difference in electrical
propertied of N doping ZnO by the two dopants. Although the NO acceptors were formed by
radical N2 doping, however, in the same times, the (N2)O were also introduced and which
compensated the NO acceptors because the radical N2 is consisting of more N2* than radical
NO. Based on the above depictions, the produce of N* seems to be responsible for p-type
conduction of ZnO thin films in our work. It is also reported that p-type ZnO was prepared
using NO in Ref.31 and Ref.32. The success in the p-type ZnO by N-doping indicates that the
NO is a good dopant to fabricate p-type ZnO.
C Fabrication of the ZnO/GaN pn Heterojunction LED
Fig.10 shows the current-voltage curve of the n-ZnO/i-ZnO/p-GaN heterojunction by solid

line and the inset figure gives the schematic diagram of the device structure. The
heterojunction diode exhibited rectifying I-V characteristics. In Fig.8, the dash line presents
the current-voltage curve of p-GaN/n-ZnO p-n heterojunction. We found that p-i-n
heterojunction has larger resistance than p-n heterojunction from I-V curves. This maybe
come from the existence of high resistivity N-doped ZnO layer. And current-voltage curve of
p-i-n heterojunction showed turn on voltage of ~3.3eV very close to RT band gap of ZnO.
1.5 In
n-type ZnO Ni/Au
i-ZnO
1.0 p-type GaN
Current (mA)
0.5
0.0
p-n heterojunction
p-i-n heterojunction
-0.5
-1.0
-20 -15 -10 -5 0 5 10 15 20
Voltage (V)
Fig.10 RT currentvoltage characteristics of the fabricated diodes composed of n-ZnO/p-GaN and n-

ZnO/i-ZnO/p-GaN heterojunctions. The inset is the schematic diagram of the p-i-n heterojunction
device.
EL spectra of the n-ZnO/i-ZnO/p-GaN and n-ZnO/p-GaN heterojunction at room

temperature are shown in Fig.10. From the EL spectra, very broad emission band centered at
430nm was observed for p-GaN/n-ZnO heterojunction. This is consistent with the reported
result in GaN pn junctions, in which the emission was attributed to the Mg levels in p-GaN
[41]. To understand the reason why the luminescence originates from the GaN layer in n-
ZnO/p-GaN heterojunction, we have examined the band structure of the heterojunction. If we
simply take the band offset of the GaN/ZnO heterojunction given by the Anderson model[62],
the conduction band offset (Ec) and the valence band offset (Ev) are given by
'EC F ZnO FGaN (5)
'EV Eg ZnO 'E C Eg GaN (6)
in which the EgZnO and EgGaN are the band gaps of ZnO and GaN, respectively, ZnO and GaN
are the electron affinities of ZnO (4.35 eV)[63] and GaN (4.2 eV)[64], respectively. The
calculated result gives the conduction band offset of Ec=0.15 eV and similar valence band
offset of EV=0.12 eV. Namely, the electrons in ZnO and holes in GaN will overcome almost
equal barrier to realize the carrier injection. In this work, electron concentrations of n-type
ZnO are 1018 cm-3 and hole concentration of p-type GaN are 1017 cm-3, and considering of the
mobility of the electrons is much bigger than that of the holes, the electrons will be injected
much easier into p-type GaN than holes into ZnO, resulting in the observation of the
electroluminescence originating from GaN layer. However, the emission of the p-i-n
heterojunction is different from that of the p-n junction. There appears a near ultraviolet peak
at about 400 nm in the EL spectra of the forward biased LED. For the p-i-n heterojunction,
the emission peak shows obvious blue-shift and narrowing, as seen in Fig.11. This indicates
that the emission intensity near ultraviolet region increases and the blue emission at 430 nm
becomes weak. The weak tail in the blue region is possibly due to the recombination of the
excess electrons injected into the p-GaN layer, similar to the p-n junction. This implies that
the blue emission at 430 nm originating from GaN layer is strongly suppressed in the p-i-n
heterojunction.
800 700 600 500 400 300

430nm 400nm
p-i-n heterojunction
p-n heterojunction
EL intensity (a.u.)
1.5 2.0 2.5 3.0 3.5 4.0 4.5

Energy (eV)
Fig.11. The EL spectra of the heterojunction LEDs with the solid line for the n-ZnO/i-ZnO/p-GaN
junction and dashed line for the n-ZnO/p-GaN junction.
undoped ZnO
Absorption coefficient D (x10 cm )

-2
1.0
nitrogen doped ZnO
10
0.8
Normalized PL intensity (a.u.)

1.0
2
0.6
0.4
0.2
0.5
0.0
2.0 2.5 3.0 3.5

Energy (eV)
0.0
3.0 3.2 3.4
Energy (eV)
Fig.12 Square plot of the absorption coefficient as a function of photon energy for the undoped ZnO
film and ZnO: N film grown on C-plane sapphire. The inset shows the RT PL spectra of different
samples as indicated.
We attribute the ultraviolate emission to the nitrogen doped ZnO layer sandwiched
between the n-ZnO and p-GaN. In order to investigate the properties of this layer, both
nitrogen doped ZnO and undoped ZnO thin films were grown on c-Al2O3 substrate. Fig.12
shows the absorption spectra of the above two samples at room temperature. Optical band gap
(Eg) and absorption coefficient () can be expressed by the following equation[65]
D (hQ ) A (hQ Eg )1 2 (7)
where h is the photon energy, A* is a constant independent of photon energy. Following this
relationship, the band gaps of the ZnO: N and undoped ZnO thin films are subtracted from the
plot of 2 as a function of the incident photon energy by extrapolating the linear part of the
curve to intercept the energy axis. As shown in Fig.10, the optical band gap of ZnO: N
becomes narrower than that of undoped ZnO thin film with the difference of about 90 meV
due to the formation of the nitrogen impurity band. Masanobu Futshara et al have also
observed similar phenomena in their woks [66]. The room temperature PL spectra were also
shown in the inset. The emission peak for the N-doped sample shows obvious redshift
comparing with that of the undoped ZnO thin film, which is consistent with the optical
bandgap narrowing observed in the absorption spectra. When N-doped ZnO was inserted into
heterostructure, the injected carriers inclucing electrons from n-type ZnO layer and the holes
from p-type GaN layer will be blocked and recombine inside the ZnO: N layer. The energy
diagram of the p-i-n heterojunction is shown in Fig. 13. By comparing the EL spectra of the
p-i-n heterojunction and the PL spectra of the insulating ZnO thin film, we can also come to
the conclusion that the EL emission in near-ultraviolet region originating from the i-ZnO
layer in the device. We noted that the energy position of the emission peak exhibits a slight
shift between the EL spectrum of the p-i-n heterojunction and the PL spectrum of N doped
ZnO film. This is interpreted as the co-existence of the GaN related emission located at the
lower energy side (blue region) in the p-i-n heterojunction. A small quantity of electrons still
get across the i-ZnO layer and are inject into p-GaN region, resulting in weak blue emission
from the GaN layer. Further optimized growth conditions and device structure of the
heterojunction are in progress.
Vacuum level
=4.20eV
=4.35eV
Eg=3.40eV
Eg=3.37eV
n-type ZnO i-ZnO p-type GaN
Fig.13 The band diagram of the p-GaN/i-ZnO/n-ZnO heterojunction.
4 Conclusions
The optical and electrical properties of ZnO thin films grown at different temperature by P-
MBE were investigated. The influence of growth temperature on the quality of ZnO film
prepared by P-MBE is investigated. At growth temperature of 350 , ZnO crystallites with a
rough surface was formed by 3D epitaixal growth. With the growth temperature increase, the
roughness of ZnO films gradually decreases and the crystal grain size becomes large. The
growth mode changes from the 3D growth to 2D growth as the growth temperature increases
above 650 . For the samples grown below 650 , the UV emission band at RT is identified
to be from free exciton transition. The blue shift of the samples grown at lower growth
temperatures is attributed to the quantum confinement effect due to smaller crystal grain
sizes. At the growth temperature of 700 , the UV emission peak at 3.212eV is considered to
be related to the excitons bound to neutral donor. The PL spectra and Hall effect
measurements show that the crystal quality of ZnO thin films was improved with increasing
the growth temperature.
The p-type ZnO thin films were grown using radical NO as nitrogen and oxygen sources
by P-MBE technology. The absorption spectra show the presence of subbandgap levels
associated with N doping. The p-type ZnO has a minimum resistivity of 9.95 : cm and hole
concentration of 1.21019cm-3. From OES of radical NO and radical N2, It was found that the
atomic N from radical NO is responsible for p-type ZnO and the N-O complex is possible to
be a shallow donor in the ZnO thin films.. The success in the N-doped p-type ZnO
demonstrated that it is the feasibility to use radical NO source as N dopant.
We have fabricated p-GaN/i-ZnO/n-ZnO heterojunction LEDs by plasma-enhanced

molecular beam epitaxy, in which nitrogen doped ZnO was introduced as the insulator layer.
Diode like I-V characteristic has been observed under forward bias. RT EL spectrum shows
strong emission at near-ultraviolet region without the deep center related luminescence. By
comparison with the p-n heterojunction and PL spectra of ZnO, it is confirmed that the origin
of the emission comes from the insulating ZnO layer rather than usually observed to come
from the p-GaN.
Acknowledgment
This work was supported by the 863 High Technology Research Program in China, under
Grant No 2001AA31112, the Innovation Project of Chinese Academy of Sciences, the
National Natural Science Foundation of China under Grant No 69896260 and No 69977019,
and the Program of Hundred Talents of Chinese Academy of Sciences.
References
[1] V. Nurmikko, R. L. Gunshor, Proc. Int. Symp. on Blue Laser and Light Emitting Diodes.
Chiba University, 1996 p.3
[2] H. Morkoc, S. Strne, G. B. Gao, M. E. Lin, B. Sverdlov, J. Appl. Phys. 1994, 76, 1363
[3] S. Nakamura, Proc. Int. Symp. on Blue Laser and Light Emitting Diodes. Chiba
University, 1996 p.119
[4] S. Y. Myong, S. J. Baik, C. H. Lee, W. Y. Cho, K. S. Lim, Japan.J. Appl. Phys. 1997,
35, L56
[5] Yamada, W. Wenas, M. Yoshino, M. Konagai, K. Takahashi, Japan. J. Appl. Phys.
1991, 30, 1152
[6] S. Kohiki, M. Nishitani, T. Wada, T. Hirao, Appl. Phys. Lett. 1994, 64, 2876
[7] Klingshirn, Phys. Stat. Sol. (b) 1975, 71, 547
[8] G. Thomas, J. Phys. Chem.Solids 1960, 15, 86
[9] Z. K. Tang, G. K. L. Wong, P. Yu, M. Kawasaki, A. Ohtomo, H. Koinuma, Y. Segawa,
Appl. Phys. Lett. 1998, 77, 3270
[10] T. Kakino, C. H Chia, Nguan, T. Tuan, H. D. Sun, Y. Segawa, M. Kawasaki, A.
Ohtomo, K. Tamura, H. Koinuma, Appl. Phys. lett. 2000, 77, 975
[11] R. Ondo-Ndong, F. Pascal-Delannoy, A. Boyer, A. Giani, A.Foucaran, Mater. Sci. and
eng. B 2003, 97, 68
[12] K. Haga, T. Suzuki, Y. Kashiwaba, H. Watanabe, B. P. Zhang, Y. Segawa, Thin solid
films 2003, 433, 131
[13] H. J. Ko, Y. Chen, S. K. Hong, T. Yao, J. Crystal Growth 2000, 209, 816.
[14] Kobayashi, O. F. Sankey, J. D. Dow, Phys.Rev.B 1983, 28, 946
[15] S. B. Zhang, S. H. Wei, Alex Zunger, Phys.Rev.B 2001, 63, 075205
[16] Y. Chen, D. M. Bagnall, H. J. Koh, K. T. Park, K. J. Hiraga, Z. Q. Zhu, T. Yao, J. Appl.
Lett. 1998, 84, 3912
[17] H. J. Ko, T. Yao, Y. Chen, S. K. Hong, J. Appl. Phys. 2002, 92, 4354
[18] R. J. Lad, P. D. Funkenbusch, C. R. Aita, J. Vac. Sci. Technol. 1980, 17(4), 808
[19] H. J. Ko, Y. F. Chen, S. K. Hong, T. Yao, J. Cryst. Growth 2000, 209, 819
[20] Y. F. Chen, H. J. Ko, S. K. Hong, T. Yao, Appl. Phys. Lett. 2000, 76, 559
[21] S. K. Hong, T. Hanada, H. J. Ko, Y. F. Chen, T. Yao, D. Imai, K. Araki, M. Shimohara,
Appl. Phys. Lett. 2000, 77, 3571
[22] . C. H. Park, S. B. Zhang, S. H. Wei, Phys. Rev. B 2002, 66, 073202
[23] C. Look, D. C. Reynolds, C. W. Litton, R. L. Jones, D. B. Eason, G. Cantwell, Appl.
Phys. Lett. 2002, 81, 1830
[24] K. K. Kim, H. S. Kim, D. K. Hwang, J. H. Lim, S. J. Park, Appl. Phys. Lett. 2003, 83, 63
[25] Y. R. Ryu, T. S. Lee, H. W. White, Appl. Phys. Lett. 2003, 83, 87
[26] C. Lee, Y. S. Kim, Y. G. Jin, K. J. Chang, Phys. Rev. B 2001, 64, 085120
[27] Y. Yan, S. B. Zhang, Phys. Rev. Lett. 2001, 86, 5723
[28] X. L. Guo, H. Tabata, T. Kawai, J. Cryst. Growth 2002, 237239, 544
[29] X. Wang, S. Yang, J. Wang, M. Li, X. Jiang, G. Du, X. Liu, R. P. H. Chang, J. Cryst.
Growth 2001, 226, 123
[30] K. Iwata, P. Fons, A. Yamada, K. Matsubara, S. Niki, J. Cryst. Growth 2000, 209, 526
[31] X. Li, Y. Yan, T. A. Gessert, C. D. Hart, C. L. Perkins, D. Young, T. J. Coutts,
Electrochemical and Solid-State Letters, 2003, 6 (4), C56
[32] W. Z. Xu, Z. Z. Ye, T. Zhou, B. H. Zhao, L. P. Zhu, J. Y. Huang, J. Cryst. Growth 2004,
265,133
[33] X. L. Guo, J. H. Choi, H. Tabata, T. Kawai, Jpn. J. Appl. Phys. 2001, 40, L177
[34] T. Aoki, Y. Hatanaka, D. C. Look, Appl. Phys. Lett. 2000, 76, 3257
[35] H. Ohta, K. Kawamura, M. Orita, M. Hirano, N. Sarukura, H. Hosono, Appl. Phys. Lett.
2000, 77, 475
[36] H. Ohta, H. Mizoguchi, M. H. Narushima, T. Kamiya, H. Hosono, Appl. Phys. Lett.
2003, 82, 823,
[37] S. Ishizuka, K. Suzuki, Y. Okamoto, M. Yanagita, T. Sakurai, K. Akimoto, N. Fujiwara,
H. Kobayashi, K. Matsubara, S. Niki, Phys. Stat. Sol. (c) 2004, 1, 1067
[38] X. Wang G. W. Yang, H. W. Liu, Y. H. Han, J. F. Luo, C. X. Gao, G. T. Zou, Appl.
Phys. Lett 2004, 84, 2427
[39] S. Jeong, J. H. Kim, S. Im, Appl. Phys. Lett 2004, 83, 2946
[40] Ya. I. Alivov, D. C. Look, B. M. Ataev, M. V. Chukichev, V. V. Mamedov, V.
IZinenko, Yu. A. Agafonov, A. N. Pustovit, Solid-State Electronics 2004, 48, 2343
[41] Ya. I. Alivov, J. E. Van Nostrand, D. C. Look, M. V. Chukichev, B. M. Ataev, Appl.
Phys. Lett. 2003, 83, 2943
[42] Ya. I. Alivov, E. V. Kalinina, A. E. Cherenkov, D. C. Look, B. M. Ataev, A. K. Omaev,
M.V Chukichev, D. M. Bagnall, Appl. Phys. Lett. 2003, 83, 4719
[43] X. L. Guo, H. Tabata, T. Kawai, J. Cryst. Growth 2002, 237-239, 544
[44] Y. Chen, H. J. Ko, S. K. Hong, T. Yao, Appl. Phys. Lett. 2000, 76, 559.
[45] Vigu, C. Deparis, P. Venngus, S. Vzian, M. Lagt, P. Lorenzini, C. Morhain, F.
Raymond, J. Guion,, .P. Faurie, Phys. Stat. Sol.(b) 2002, 229, 931
[46] Joshy Jose, M. Abdul Khadar, Materials Science and Engineering A 2001, 304-306, 810
[47] S. Cho, J. Ma,Y. Kim, Y. Sun, G. K. L. Womg, Appl. Phys. Lett. 1999, 75, 2761
[48] Kyu-hyun Bang, D. K. Hwang, M. C. Jeong, K. S. Sohn, J. M. Myoung Solid State
Communications 2003, 126, 623
[49] L. E. Brus, J. Chem. Phys. 1984, 80, 4403
[50] K. Hmmer, Phys. Stat. Sol. 1973, 56, 249
[51] S. Jiang, H. Jung, K. Ploog, J. Appl. Phys. 1988, 64, 1371

[52] Y. G. Wang, S. P. Lau, X. H. Zhang, H. W. Lee, H. H. Hng, B. K. Tay. Journal of
Cryst.Grwoth. 2003, 252, 265
[53] C. Lee, Y. S. Kim, Y. G. Jin, K. J. Chang, Physica B 2001, 308-310, 912
[54] Y. Sato, S. Sato, Thin Solid Films 1996, 281-282, 445
[55] J. D. Ye, S. L. Gu, S. M. Zhu, T. Chen, L. Q. Hu, F. Qin, R. Zhang, Y. Shi, Y. D. Zheng,
J. Cryst. Growth 2002, 243, 151
[56] N. Y. Garces, N. C. Giles, L. E. Halliburton, G. Cantwell, D. B. Eason, D. C. Reynolds,
D. C. Look, Appl. Phys. Lett. 2002, 80, 1334
[57] J. Ko, Y. F. Chen, K. Miyajima, A. Yamamoto, T. Goto, Appl. Phys. Lett. 2000, 77, 537
[58] K. Tamura,T. Makino, A. Tsukazaki, M. Sumiya, S. Fuke, T. Furumochi, M. Lippmaa,
C. H. Chia, Y. Segawa, H. Koinuma, M. Kawasaki, Solid State Comminications 2003,
127, 265
[59] C. Look, Mater. Sci. Eng. B 2001, 80, 383
[60] D. C. Reynolds, D. C. Look, B. Jogai, C. W. Litton, T. C. Collins, W. Harsch, G.
Cantwell, Phys. Rev. B 1998, 57, 12151
[61] Zeuner, H. Alves, D. M. Hofmann, B. K. Meyer, A. Hoffmann, U. Haboeck, M.
Strassburg, M. Dworzak, Phys. Status Solidi B 2002, 234, R7
[62] G. Milnes, D. L. Feucht, Heterojunctions and metal-semiconductor Junctions(
Academic, New York, 1972)
[63] J. A. Aranovich, D. G. Golmayo, A. L. Fahrenbruch, R. H. Bube, J. Appl. Phys. 1980,
51, 4260
[64] D. Qiao, L. S. Yu, S. S. Lau, J. M. Redwing, J. Y. Lin, H. X. Jiang, J. Appl. Phys. 2000,
87, 801
[65] J. Tauce, optical properties if solid ed F Abeles( Amserdam: North-Holland) P277
[66] M. Fyutsuhara, K. Yoshioka, O. Takai. Thin Solid Films 1998, 317, 322
Chapter 6
STRUCTURAL DISORDER IN CDSXSE1-X THIN FILM

PROBED BY MICRODIFFRACTION AND OPTICAL
SPECTROSCOPIES
V. Capozzi(1), G. Perna(1), S. Pagliara(2) and L. Sangaletti(2)

(1)
Dipartimento di Scienze Biomediche, Universit di Foggia,
Via L. Pinto, 71100 Foggia, Italy
(2)
Dipartimento di Matematica e Fisica, Universit Cattolica,
Via dei Musei 41, 25121 Brescia, Italy
Abstract
X-ray diffraction with a large area detector from CdSxSe1-x alloy thin films, deposited on
Si (111)-oriented substrates by laser ablation technique, allowed us to point out the effects of
structural disorder which can not be observed by conventional T-2T diffractometers. The X-
ray spectra measured by using conventional T-2T diffractometer have shown that the CdSxSe1-
x films grow oriented along the (002)-direction in the hexagonal phase with the c-axis
perpendicular to the film layers. The width of the 002 reflection is probably related to
substitutional disorder arisng from alloying. However, further disorder effects are detected
when one resorts to microdiffraction. Indeed, the broadening along the Debye rings reveals
that the films are not perfectly epitaxially grown. Moreover, only two families of planes are
present in the microdiffraction spectra of all samples. This finding is discussed by considering
a structural disorder not related to alloying. An attempt to simulate this structural disorder has
been carried out on the basis of symmetry considerations and discussed. A correlation is found
with the results of luminescence and Raman spectroscopies that give evidence of disorder
effects in terms of excitation localization in photoluminescence and a broad, disorder
activated, band in the Raman spectra.
Introduction
Ternary compounds of II-VI semiconductors are largely used in optoelectronic devices due to
the possibility of controlling of the lattice costant and the energy gap by modulating the
relative composition of binary compounds. CdSxSe1-x ternary alloys, in particular, are mostly
144 V. Capozzi, G. Perna, S. Pagliara et al.
important in the production of the devices such as laser diodes and light emitting diodes
operating in the visible region according to the band gap modulation obtained by varying the
sulphur concentration (x) [1].
Besides these appealing applications, the semiconductor solid solution provides a good
system for the study of the effect of disorder on the optical properties of solids. In solid
solution formed by isoelectronic substitution the disorder is usually weak, permitting easy
comparison of the optical spectra of the ternary compounds with those of the binary
semiconductors. As a rule, the atoms in such solutions keep, in average, the regular positions
so that the long range order exists and samples with perfect crystal habit can be grown.
Moreover, the isoelectronic alloying does not change the nature of the chemical bonding so
that the main features of the band structure are preserved. The random distribution of
substitutional atoms over the possible crystal sites results in fluctuations of the local
electronic density, producing the corresponding fluctations in the periodic crystal
potential [2].
The most simple theorical model for the description of the solid solutions is the so-called
virtual crystal approximation [3]; it gives a satisfactory picture of compositional dependence
of the band energy positions and the lattice costant values in solid solutions, but this
approximation does not take into account the disorder at all (e.g., bond and lattice distortion
due to the different atomic size). Indeed, for the study of the broadening of the electronic
states, models with explicit consideration of disorder should be used. This is known for
optical spectroscopy studies [2], where suitable functions are considered in the deconvolution
of data to account for electronic states induced by disorder. Likewise, excitonic lines in alloys
are larger than those of binary compounds.
Disorder in solid solutions can have various forms. In ternary compounds with
substitutional disorder, because of the difference of lattice costants of the binary compounds,
the atoms can be shifted from the exact symmetry positions which will produce strong local
stresses and will give an additional contribution to the atomic potential. Microscopic stresses
due to the bond length mismatch can be partially relaxed through formation of intrinsic
structural defects such as vacancies and interstitials [2]. The maximum defect concentration is
expected for equimolar compositions.
Stacking faults may also appear; this type of disorder can lead to structural phase
transition at some intermediate composition [2].
In the present work we report X-ray diffraction, microdiffraction, microraman, and
photo-luminescence (PL) measurements of CdSxSe1-x alloy films deposited on Si (111)-
oriented substrates by laser ablation, in order to study the microscopic nature of substitutional
and structural disorder.
Experimental
The investigated CdSxSe1-x films were deposited by laser ablation technique on Si(111)-
oriented substrate, starting from stoichiometric and high purity (99,999%) cold pressed target
of CdS and CdSe powder. The substrate temperature was optimized at 400C as discussed in
[4]. A pulsed Kr-F laser operating at 248 nm, with a repetition rate of 10 Hz, and a pulse
width of 18 nsec, was used as sublimation source. Laser fluence was optimized at 10 J/cm2.
Structural Disorder in Cdsxse1-X Thin Film Probed by Microdiffraction 145
The deposition chamber was evacuated by a turbomolecular pump and the residual pressure,
before the deposition, was about 10-6 torr.
Conventional T-2T X-ray diffraction spectra were measured by using the Cu-KD radiation
(O = 1.5406 ) of a Siemens D5000 diffractometer.
Microdiffraction patterns have been collected with a Bruker D8 Discover diffractometer
equipped with an area detector (General Area Dtector Diffraction System: GADDS) and a Co
KD X-ray source (KD1=1.789007 , KD2=1.792892 ). The minimum size of the X-ray beam
is 50 Pm. For the present experiment a collimator for the X-ray beam was used, which yielded
a 100 Pm diameter X-ray spot on the sample surface.
Microraman spectra were collected by a Dilor Labram spectrograph with a He-Ne laser
(632 nm, output power=14 mW). The spectrometer was confocally coupled with a
microscope (100X objective, N.A. = 0.9) which allowed us to collect Raman spectra with a
spatial resolution of about 1 Pm.
For the PL measurements, the samples were mounted into a He closed-cycle cryostat,
whose temperature could be varied from 10 to 300 K. They were illuminated in a direction
forming a 45 angle with respect to the normal to the film surface, by using the 4579 line of
a cw Ar-ion laser. The emission light was collected along a direction perpendicular to the
laser beam and analysed by a double grating spectrometer (line dispersion of 1 meV/mm).
The signal, detectd by a GaAs cooled photomultiplier, was recorded using photon counting
techniques. The film quality was controlled by the width of the PL excitonic lines.
Structural modeling and simulations have been performed by using the CERIUS program
on an IBM RISC system 6000 workstation [5].
Results and Discussion

X-ray Diffraction Measurements
X-ray spectra of the four CdSxSe1-x films (x = 0, 0.4, 0.6, 1) deposited on Si(111)-oriented
substrates are shown in Fig. 1. All peaks refer to the (002)-orientation of the film
corresponding to the hexagonal phase, with the c-axis perpendicular to the film layers. The
diffraction peak shifts towards larger angles with increasing x, indicating that the lattice
constant decreases with the sulphur concentration. The grain dimension D, calculated by the
Debye-Scherrer relationship [6], are reported in Table 1.The lattice plane distance dhkl (hkl are
the Miller indices) can be determined by means of the Bragg equation:
dhkl = mO/2sinT (1)
where m is the diffraction order O=1.5406 and T is the Bragg diffraction angle. The
obtained values are reported in Table 1. The lattice parameter c can be calculated according to
the following equation [7]:
1 4 h 2 +k 2 +hk l2
= + 2 (2)
d2 3 a2 c
Table 1. Full width at half maximum (FWHM) of the XRD peak of the CdSxSe1-x deposited
on Si(111)-oriented substrate; average grain dimension (D), as deduced from the Debye-
Schrerrer equation [6]; lattice plains distance (d), as obtained from the Bragg equation; values
of the lattice parameter c, obtained from eq. (3).
FWHM (0) D (nm) d(002) () c ()

CdS 0.159 0.002 47.5 0.5 3.36 0.01 6.720.01
CdS0.6Se0.4 0.201 0.002 40.6 0.4 3.42 0.01 6.84 0.01
CdS0.4Se0.6 0.2820.002 29.00.2 3.45 0.02 6.90 0.01
CdSe 0.188 0.001 43.1 0.3 3.51 0.01 7.01 0.01
Fig. 1. X-ray spectra of the CdSxSe1-x films deposited on Si(111)-oriented substrates. Each peak refers
to the (002)-orientation of the crystal domains. In the inset, the lattice papameters c of the CdSxSe1-x in
the hexagonal phase are shown for each film, as calculated from eq. (2).
The obtained c values (reported in Table 1) follow quite well Vegards rule [8], which
assumes a linear dependence of c on the composition x, according to:
c(x) = xcCdS + (1-x)cCdSe (3)
The continuous straight line, in the inset of Fig. 1, is a least square fit of the c data to eq.
(3). The c values obtained from the fit (cCdS = 6.72 r 0.01 and cCdSe = 6.98 r 0.01 ) are in
good agreement with the literature values of the corresponding bulk materials [9].
Microdiffraction Measurements
The microdiffraction patterns collected from the CdSxSe1-x samples with the GADDS system
are shown in Fig.2 a-d.
Unlike conventional T-2T diffraction, the area detector allowed us to sample a large solid
angle of the diffraction cone and thus to detect all the reflections at once from the sample.
Therefore, in addition to the reflections already detected with the conventional diffractometer,
other reflections are detected by the two-dimensional detector. The point-like spots detectable
in all patterns are ascribed to the Si(111) substrate, whereas the larger spots are ascribed to
the thin film alloy. In agreement with the data of conventional T-2T diffraction, a broadening
perpendicular to the Debye ring (i.e. the FWHM of the reflection) is found. This effect,
ascribed to structural disorder, is larger in CdS0.4Se0.6 (Fig. 2-c) than in CdS0.6Se0.4 (Fig. 2-b).
Fig.2. Microdiffraction patterns of the CdSxSe1-x films deposited on Si (111)-oriented substrates. The
bright spots are ascribed to reflections from the Si substrate, whereas the broad spots are the reflections
of the CdSxSe1-x hexagonal phase.(a) CdS (x=1.0). (b) CdS0.6Se0.4 (x=0.6). (c) CdS0.4Se0.6 (x=0.4).
(d) CdSe (x=0.0)
While the broadening perpendicular to the Debye ring is important to evaluate the
structural disorder, the broadening along the Debye rings is important to achieve information
about the texture of the film. The broadening of the spots along the Debye ring, i.e. along the
azimuthal angle I, is shown in Fig. 3, where the intensity of the 103 reflection is plotted vs.
I This broadening indicates that deviations from the perfect epitaxy are present, i.e., a small
random angle may be found between the c-axis and the axis normal to the substrate. The
largest deviations occur in CdS0.4Se0.6.
These deviations indicate that a different degree of orientation of the microcrystallites
along the a and b crystallographic directions exists. It is rather important to relate this effect
to the sample thickness. In fact a degradation of the epitaxy may occur for the thicker
samples. The sample thickness was measured with a profilometer and the results are 0.747
Pm, 3.138 Pm, 0.663 Pm and 1.950 Pm for CdS, CdS0.4Se0.6, CdS0.6Se0.4 and CdSe,
respectively.
Fig.3. Intensity of microdiffraction spots of theCdSxSe1-x films measured along one of the Debye rings.
The drawing shown in the inset defines the azimuthal I angle along which the intensity of the 103
reflection has been plotted.(a) CdS (x=1.0). (b) CdS0.6Se0.4 (x=0.6). (c) CdS0.4Se0.6 (x=0.4).
(d) CdSe (x=0.0)
As can be observed, the CdS0.4Se0.6 film is thicker than CdS0.6Se0.4. Indeed, a broader
peak of the intensity profile measured along the Debye ring is observed for CdS0.4Se0.6 (Fig.3,
thicker solid line) with respect to CdS0.6Se0.4 (Fig.3, dotted line). When the binary sample is
compared with the ternary sample (e.g. CdS vs. CdS0.6Se0.4) one can realize that, in spite of
the similar thickness (0.747 Pm and 0.663 Pm, respectively), the ternary sample shows a
larger peak of the intensity profile. This indicates that not only thickness affects the deviation
from epitaxy but also the disorder induced by alloying.
Fig. 4 shows the XRD patterns obtained by performing an integration of the diffracted
intensity along the Debye rings. Several reflections of the hexagonal phase are identified,
along with some reflections of the Si substrate, indicated with asterisks.
The following reflections have been identified: (002), (004), (102), (103), (104). All of
them can be ascribed to the hexagonal phase of CdS-Se. Moreover, the broadening is
observed for each of the mentioned diffraction lines. As expected, for all the reflections an
increase of the 2T angle with x is observed, accounting for the progressive contraction of the
lattice parameters in the CdS sample with respect to that of CdSe.
In conventional T-2T diffraction, only the 00l reflections of the hexagonal phase are
expected. The area detector allows us to map a larger volume of the reciprocal space and, in
principle, all the reflections from the hexagonal phase must be present. However, it is rather
important to observe that some reflections of the hexagonal phase are missing. In particular,
only the 00l (l = 2, 4) and the 10l (l > 1) reflections are present. This is true for all samples
and not only for the ternary compounds. Therefore, the structural features to which we should
ascribe the lack of reflections are not related to alloying.
Fig.4. X-ray diffraction patterns of CdSxSe1-x obtained by integrating the intensity along the Debye
rings. The intensity is plotted in a logaritmic scale. (a) CdS (x=1.0). (b) CdS0.6Se0.4 (x=0.6).
(c) CdS0.4Se0.6 (x=0.4). (d) CdSe (x=0.0)
The XRD patterns of CdS (dashed line) and CdSe (dotted line) generated from the reference data of the ICSD
database are also shown below the corresponding experimental pattern.
Different attempts to simulate the lack of reflections in micro-diffraction spectra have

been carried out. The measured spectra can not be described in terms of stacking fault
obtained by considering a faulted cubic-hexagonal sequence, as we have tried by using the
DIFFaX program [10].
In order to explain the features of the experimental diffraction pattern, a structural model
was developed by using the CERIUS program [5]. In this model, a structural disorder was
introduced in the a-b plane of the hexagonal cell.
Fig. 5. XRD pattern of CdSe calculated on the basis of a disorder-model. The inset shows the model
disordered structure as compared with the ordered wurtzite structure.
In the wurtzite crystal structure, shown in the inset of the Fig. 5, the vertexes of the
hexagonal ring are alternatively occupied by either Cd or S,Se. The three Cd atoms lie in the
bottom A plane at positions A1, A3, and A5, whereas the three S,Se atoms lie in the top
plane, at positions B2, B4, and B6. In our model, we have assumed that because of disorder
all the six lattice sites on each plane can be occupied by Cd or S (Se). Therefore, three new
atoms are found in the bottom A plane, at positions A2, A4 and A6, whereas three new S,Se
atoms are found in the top B plane at positions B1, B3 and B5.
This structural change quenches the intensity of some reflections in the XRD pattern and
allowed us to qualitatively reproduce the experimental data.
In Fig 5 the experimental (dots) and the calculated (continuous line) XRD data for CdSe
sample are shown. Although the simulated spectra do not closely fit some of the experimental
reflections, it can be noticed that in the calculated pattern only the reflections detected in the
experimental XRD pattern are reproduced.
Microraman Spectra
Fig. 6 shows the room-temperature microraman spectra of the two investigated CdSxSe1-x
films. From the point of view of Raman spectroscopy, the CdSxSe1-x alloy shows the two
phonon modes characteristic of Cd-S and Cd-Se vibrations [11]. Thus two TO-LO pairs are
expected, whose relative intensity and frequency characteristically depend on the
composition. In the present study, only the LO modes are observed. These are the forbidden
LO modes which become allowed near resonance, when a major contribution to Raman
scattering comes from the q-dependent intraband Frohlich term. Since this term contributes
only to LO modes, only these modes are observed in such experimental conditions [12].
Fig.6. Microraman spectra of CdS0.6Se0.4 and CdS0.4Se0.6. In the inset, the LO2 CdS-like mode is shown,
along with the two-lorentian fitting, yielding the broad ZE phonon mode at low wavenumbers.
Moreover, in the present set of samples only the Raman spectra of the mixed compounds
are observed, because for these two samples only a near-resonant Raman scattering occurs,
being the energy of the laser light (1.96 eV) close to their energy gap and off-resonance with
respect to the pure CdS and CdSe gaps.
The CdSe-like LO1 mode (at 205.8 cm-1 in CdS0.4Se0.6 and 201.0 cm-1 in CdS0.6Se0.4) and
the CdS-like LO2 mode (at 284 cm-1 in CdS0.4Se0.6 and 290 cm-1 in CdS0.6Se0.4) are the most
intense lines in the first order Raman spectra. The peak at 182.8 cm-1 is a plasma line of He-
Ne laser. Higher-order modes can be observed at larger wavenumbers. In particular, we
observe the 2LO1-phonon mode (at 406.5 cm-1 in CdS0.4Se0.6 and 395 cm-1 in CdS0.6Se0.4), the
2LO2-phonon mode (at 567 cm-1 in CdS0.4Se0.6 and 576 cm-1 in CdS0.6Se0.4) the LO1+LO2-
phonon mode (at 486 cm-1 in CdS0.4Se0.6 and 488 cm-1 in CdS0.6Se0.4). Moreover, the Raman
spectrum of the CdS0.6Se0.4 film is characterized by an intense line at 520 cm-1 due to the
emission from the Si(111)-oriented substrate. This line is not observed in the CdS0.4Se0.6
sample because of the larger thickness with respect to that of the CdS0.6Se0.4 sample. The
relative weight of LO1 and LO2 bands changes with S content. In fact, the increase of S
fraction from 0.4 to 0.6 yields an increase of LO2 intensity with respect to LO1. Likewise, the
frequency of the CdSe-like LO1 mode decreases with x, whereas the frequency of the CdS-
like LO2 mode increases with x.
The lineshapes of the LO2 phonon peaks are clearly asymmetric. This asymmetry has
been observed earlier in other III-V and II-VI mixed semiconductors [12,13,14] and ascribed
either to phonon confinement or disorder effects.
The correct fitting of these peaks can be carried out if one assumes that in alloys the
cohexistance of two Raman bands is expected: the relatively narrow bands at LO phonon
frequency of the ternary compounds and a disorder activated, broad, band associated with the
phonon density of states [15]. In fact, when disorder can not be neglected, defect-activated
Raman scattering from all phonon modes is expected. The weight of this disorder-activated
band in both CdS0.6Se0.4 and CdS0.4Se0.6 samples is remarkable.
Therefore a deconvolution of the Raman bands was carried out by fitting the bands with
the sum of two Lorentzian curves (inset of fig.6, [15]). One Lorentzian is ascribed to the
zone-centre phonon contribution, whereas the other Lorentzian is ascribed to the zone-edge
(ZE) phonon contribution [12], namely:
SLO SZE
I() = + (4)
( - LO ) + ( LO /2) ( - ZE ) 2 + ( ZE /2) 2
2 2
where SLO (SZE) is an amplitude factor relative to LO (ZE) phonon peak, ZLO (ZZE) is the
frequency of the LO (ZE) mode, *LO and *ZE are the full width at half maximum of the LO
(ZE) Lorentzian peak. In the inset of Fig. 6 the least-square fit of LO2 CdS-like mode with the
two Lorentzian model (eq.4) to the experimental data is shown as continuous line, whereas
the contributions of LO and ZE phonons are shown as dashed line and dotted line,
respectively.
As expected, the frequency of the LO2 mode increases with S content, as reported by
Chang et al. [16] and S. Permagorov and A. Reznitsky [2]. The fit of CdSe-like phonon mode
was not performed because of the presence of He-Ne ghost at 182.8 cm-1.
Luminescence Spectra
The PL spectra of the CdSxSe1-x films at T=10K are shown in Fig. 7. All the samples present
a narrow emission line in the higher energy region, corresponding to excitonic
recombinations. The excitonic lines are centered at 2.540 eV, 2.110 eV, 1.993 eV and 1.820
eV for CdS, CdS0.6Se0.4, CdS0.4Se0.6 and CdSe, respectively. The emission bands at energies
lower than the excitonic lines are due to radiative recombinations involving extrinsic defects
located in the energy gap of the CdSxSe1-x films.
Fig. 7. Photoluminescence spectra of the CdSxSe1-x ablated films, measured at the lattice temperature of
(a) 10 K and (b) 300 K. The samples were photoexcited by the line 4579 of an Ar ion laser, with an
exciting intensity of 80 W/cm2. In the inset (b), the dependence of the excitonic emission line on
sulphur concentration x at room temperature is shown: the continuous line is a least square fit of
experimental data to eq. (5). For each spectrum the relative sensitivity factor is indicated. Spectral
resolution is 1 meV.
The excitonic peak shifts towards lower energies with decreasing the x fraction, due to
the decrease of the fundamental energy gap with the sulfur content. Moreover, by raising T,
the excitonic emission intensity decreases and an exponential tail appears on the high energy
side of the excitonic peak (Fig. 7b), which is due to the thermal distribution of the excitons.
The excitonic emission is even present at room temperature, where it is the most intense band
of the PL spectra. Such a behaviour suggests the possibility of emission modulation by means
of the CdSxSe1-x system, as illustrated in the inset of Fig. 7b, where the sulfur cotent
dependence of the excitonic peak at room temperature is reported. The exciton energies XPL
(dots) are well fitted (continuous line) by a quadratic relation:
XPL(x) = XPLCdSe(1-x) + XPLCdSx bx(1-x) (5)
where b is the bowing parameter. The fitting parameters XPLCdSe = 1.741r0.009 eV and XPLCdS
= 2.448r0.009 eV are in good agreement with the value of the A-type excitonic energy at
room temperature for the CdSe and CdS single crystals, respectively [9].
An interesting result in Fig. 7 is the large PL efficiency of the CdS0.6Se0.4 film at low
temperature. This result is due to the localized nature of the excitonic recombinations, related
to the fluctuations of the local concentration in the solid solution: as a result, fluctuations in
the periodic crystal potential occur and they act as potential wells for the carriers. Therefore,
at the lowest temperatures the XPL line corresponds to radiative recombinations of excitons
localized in potential wells. Consequently, the excitons have a reduced probability of
diffusion into non radiative traps and the radiative recombination path is strongly enhanced. A
similar PL enhancement is not observed in the spectrum of the CdS0.4Se0.6, because the
exciton localization occurs only if the number of localized states is larger than the number of
impurities and other structural defects [1]. It can be deduced that the CdS0.4Se0.6 film present a
large number of impurities and defects, as also confirmed by the large PL efficiency related to
intra-gap states (spectrum B of Fig. 7a). This finding is in agreement with the results of
structural investigations which indicate that in CdS0.4Se0.6 disorder effects, detected as
broadening both along and perpendicular to the Debye rings, are larger than in CdS0.6Se0.4.
The excitonic emission from CdS and CdSe is also lower that that of CdS0.6Se0.4, because
localization effects related to fluctuations of the local concentration do not occur in such
binary compounds.
The dependencies of the spectral position of the exciton line in CdSxSe1-x films as a
function of the temperature (dots) are illustrated in Fig. 8. The red-shift of the excitonic levels
when the temperature increases is related to the thermal expansion of the lattice and the
temperature dependence of the electron-phonon interaction [17]. The experimental data of
Fig. 8 for CdS, CdSe and CdS0.4Se0.6 are in good agreement with a model, introduced by Vina
et al. [17], describing the thermal red-shift of the energy gap for typical semiconductors. The
continuous lines in Fig. 8 are a fit of such model to the experimental data. The Vina model is
represented by the following equation:
X(T) = X(0) 2aBnB (6)
where X(T) is the exciton energy at the temperature T, aB is the strength of the exciton-
phonon interaction and nB = [exp(T/T)-1]-1 is the Bose-Einstein statistical factor for phonon
emission and absorption; T is a phonon temperature corresponding to an average phonon
energy.
Fig. 8. Temperature dependence of the exciton energy in CdSxSe1-x samples, as obtained from the
analysis of the luminescence data. (a) CdSe (x=0.0). (b) CdS0.4Se0.6 (x=0.4). (c) CdS0.6Se0.4 (x=0.6).
(d) CdS (x=1.0)
In contrast, the temperature dependence of the spectral position of the excitonic emission
in CdS0.6Se0.4 follows the Vina model starting from a sufficiently high temperature (about 80
K), whereas it presents an anomalous behaviour at low temperature. In fact, a red-shift of the
excitonic line is observed below 35 K, but a blue-shift is present when the temperature
increases up to 80 K. The red-shift below 35 K can be due to a hopping process of the
excitons among localized states in the low energy tail of the exciton level distribution, which
takes place in a disordered semiconductor [18]. In fact, localized exciton can termally gain
energy to relax into localized states at lower energy. Such a behaviour causes a red-shift of
the maximum of the excitonic emission, because the peak of the localized excitonic
distribution shifts to lower energy. In contrast, the blue-shift of the exciton line in the
temperature range between 35 K and 80 K is related to the delocalization of the localized
excitons as a consequence of the temperature increase. Thus, the density of the free excitons
raises with temperature. Since free excitons have smaller binding energy than localized
excitons, a blue-shift of the excitonic emission is observed. When the temperature is larger
than 80 K, all the excitonic emission is due to free excitons and its spectral position follows
the red-shift of the energy gap.
Conclusion
In this study we have shown that X-ray microdiffraction is a powerful tool to probe structural
disorder effects in thin films, otherwise missed by conventional T-2T XRD. In fact, the
possibility to detect X-rays elastically scattered in a wide diffraction cone discloses several
features that can give important information on disorder effects. For instance, it is observed
that not all reflections expected from the hexagonal phase of the parent compounds are
present in the diffraction pattern detected by the micro-diffractometer.
According to the results of structural modelling, this finding is tentatively ascribed to

disorder effects in the lattice planes parallel to the substrate. Unlike conventional T-2T X-ray
diffraction which shows that the films grow in the hexagonal phase with the c-axis
perpendicular to the substrate, microdiffraction patterns show that CdSxSe1-x layers do not
grow in a perfect epitaxial habit, as can be observed from the analysis of the Debye rings.
This effect is particularly evident in the samples closer to an equimolar concentration of Se
and S. Moreover, the different broadening of the (002) reflections of the two films indicates
that CdS0.6Se0.4 is less disordered than CdS0.4Se0.6.
The method we propose is addressed to a class of materials (semiconductor alloys) whose
structural properties are usually discussed in terms of reflection broadening of the T-2T
conventional XRD patterns, and compliance with the Vegard's law [8]. We showed that when
a larger portion of the reciprocal space is mapped, further structural features can be pointed
out, which may suggest further disorder effects. Our attempt to justify the observed micro-
diffraction pattern, is based on symmetry considerations only. Additional work should be
done on this issue, with, e.g., first-principle calculations of the total energy of our model
system. This goes far beyond the scope of the present work. Nevertheless, our model catches
the main feature of the pattern, namely the quenching of the reflections and suggest a
direction for future work. In particular, it is known that X-ray diffraction probes the average
crystal structure of the samples. A more local view of the structural features and related
defects will be provided by EXAFS experiments.
Finally, a correlation is found between microdiffraction and PL data. Also the Raman
spectra are affected by structural disorder, whereas phonon confinement effects are not
detectable. PL spectra of the films have been measured from 10 K up to 300 K. The exciton
line recombination is well resolved in each film alloy up to room temperature. Only the PL
spectrum of the CdS0.6Se0.4 film, at low temperature, shows an enhancement of PL efficiency,
due to the compositional disorder (potential fluctuations) on a microscopic scale. In
CdS0.4Se0.6 film, on the contrary, structural defects prevail over potential fluctuations which
are held responsable for the exciton localization.
The lineshape of the LO phonon structure in the microraman spectra is characterized by
significant asymmetry and broadening induced by disorder. The data were fitted by using two
Lorentzian bands. The weight of the disorder-activated band resulted to be large in both alloy
samples.
Acknowledgements
A. Iberl is greatly acknowledged for technical assistance during the microdiffraction

experiments. We thank E. Depero for helpful and stimulating discussions about
microdiffraction measurements. We are also grateful to A. Giardini and M. Ambrico for the
fruitful scientific collaborations withthe "Istituto Materiali Speciali" of the C.N.R. of Tito
Scalo (PZ), Italy. This work has been partially supported by Progetto Finalizzato MSTAII,
CNR.
References
[1] F.Firszt, Semicond. Sci. Technol. 12 (1997) 272
[2] S. Permagorov and A. Reznitsky, J. Lumin. 52, 201 (1992)
[3] J. A. Van Vechten and T. K. Bergstresser, Phys. Rev. B 1, 3351 (1970)
[4] M. Ambrico, R. Martino, D. Smaldone, V. Capozzi, G. Lorusso, G. Perna, Giardini, and
A. Mele, Mat. Res. Soc. Symp. Proc. Vol. 397, 125 (1996)
[5] Cerius2, Release 2.0; BIOSYM/Molecular Simulations
[6] H.P. Klug, L.E. Alexander: X-ray Diffraction Procedures, New York, Wiley, 1954.
[7] G. Perna, S. Pagliara, V. Capozzi, M. Ambrico, T. Ligonzo, Thin Solid Films 349,
220 (1999).
[8] L. Vegard, F. Zeit. Physik 5, 17 (1921).
[9] O. Madelung, M. Schulz, H. Wess (Eds.), Landolt-Bornstein Tables, vols. 17a and 17b,
pag. 206, Springer, Berlin, 1982.
[10] Diffax, v.1.767, is computer program for calculating diffraction intensities from faulted
crystals; M. M. J. Treacy, M. W. Deem (1992)
[11] N. Esser and J. Geurts in Optical Characterization of Epitaxial Semiconductor Layers,
G. Bauer, W. Richter (eds.), Springer, Berlin, 1996, Chapt. 4
[12] A. Ingale and K.C. Rustagi, Phys. Rev. B 58, 7197 (1998).
[13] H.C. Lin, J. Ou, C.H. Hsu, W.K. Chen and M.C. Lee, Solid State Comm. 107,
547 (1998).
[14] A. Fischer, L. Anthony, A.D. Compaan, Appl. Phys. Lett. 72, 1822 (1984).
[15] S. Pagliara, L. Sangaletti, L.E. Depero, V. Capozzi, G. Perna, Solid State Comm. 116,
115 (2000).
[16] R. K. Chang, J. M. Ralston, and D. E. Keating, Light Scattering Spectra of Solids (G. B.
Wright, ed.) p. 369, Springer Verlag, New York (1969).
[17] L. Vina, S. Logothetidis, M. Cardona, Phys. Rev. B 30, 1979 (1984).
[18] V. Capozzi, K. Maschke, Phys. Rev. B 34, 3924 (1986).
In: New Research on Semiconductors ISBN 1-59454-920-6
Editor: Thomas B. Elliot, pp. 159-207
c 2006 Nova Science Publishers, Inc.
Chapter 7
E LECTRODEPOSITION OF C U I N S E2
FOR P HOTOVOLTAIC C ELL A PPLICATION
Shigeyuki Nakamura
Department of Electrical and Electronic Engineering
Tsuyama Natinal College of Technology
Abstract
This paper describes preparation of a ternary chalcopyrite semiconductor, copper
indium diselenide (CuInSe2 :CIS), thin films as a light absorption layer for thin film
solar cells by electrodeposition.
This semiconductor is very attractive for an absorber material in the thin film solar
cell because of its suitable bandgap and a large absorption coefficient. CuInSe2 has
the bandgap of about 1.0 eV and the absorption coefficient of 105 cm1 . The
large absorption coefficient enables us to realize the thin film solar cell. Although
the highest theoretical efficiency is obtained with the semiconductor whose bandgap
is ranging from 1.4 to 1.5 eV, the bandgap of CIS can be enlarged up to 1.68 eV by
adding Ga to form solid solution of Cu(Inx Ga1x )Se2 (CIGS).
Electrodeposition as a method for thin film semiconductor preparation is a
good approach with respect to economic consideration. An important advantage of
electrodeposition as a method for thin film preparation is that films with a large
area can be prepared without a vacuum, using simple and low-cost equipments.
However, conversion efficiencies ever reported for the cells fabricated by this method
are considerably lower than those reported for the cells fabricated by other methods.
One of the important problems for development of this technique is to control sample
compositions. Understanding of deposition mechanisms of each species is essential
in order to achieve higher controllability and reproducibility of film composition and
then to improve performance of the photovoltaic cell prepared by electrodeposition.
Moreover, an excellent morphology is also essential to achieve higher conversion
efficiency. A poor morphology causes short-circuiting between the front and back
electrodes. Voids and cracks in thin films degrade the conversion efficiency. On the
basis of above mentioned, following subjects were studied.
Electrodeposition of Cu-In-Se films has been studied with an aqueous solution
containing CuCl2 , InCl3 and SeO2 , in terms of composition control of deposited
films for the preparation of CuInSe2 . When a Si wafer is employed as a substrate,
both the electrode potential dependence of In/Cu ratio in the film and a stirring
effect on film composition are found to become small, compared with Mo substrate.
160 Shigeyuki Nakamura
This is explained by taking account of the existence of a space charge layer at the
semiconductor surface. From the relationship between In/Cu or Se/Cu ratio in the
bath and that in the deposited films, ratios of mass-transfer coefficients for In and Cu,
k(In)/k(Cu) or for Se and Cu, k(Se)/k(Cu), have been obtained and their dependencies
on deposition current density or stirring of the solution have been studied. The
rate-determining step in the deposition process for each ion, reaction-limited or
diffusion-limited, has been also discussed. By employing the stirring, a remarkable
improvement, by a factor of 3, is attained for run-to-run and position-to-position
fluctuation of the In/Cu ratio in the films. X-ray diffraction patterns show that CuInSe2
is contained in as-deposited films.
In order to improve surface morphology, a revers bias is applied after electrodepo-
sition in the same electrolyte. Pluse-plated electrodeposition was demonstrated for the
same purpose.
In order to improve crystallinity, as-deposited films were annealed in an inert
atomosphere. Annealing effects on film properties, such as crystallinity, morphology
and chemical composition, were investigated. Effects of Se concentration were
also investigated. Excess Se is found to degrade controllability of composition and
crystallinity.
The conversion efficiency of 0.866 % (Voc=0.47 V, Isc=1.9 mA/cm2 , FF=0.533) is
obtained with a chemical bath deposited CdS buffer layer on the electrodeposited CIS
layer without any transparent conductive oxide layers.
Preparation of CuInSe2 thin films with a bilayer structure by electrodeposition
were studied. Change in a Cu/In ratio with depth is essential to obtain higher
conversion efficiency. Thus, a new electrodeposition technique for preparing CuInSe2
thin films with controlled depth profile were developed. The CuInSe2 thin films with
both Cu-rich and In-rich layers are deposited by changing substrate potential during
electrodeposition.
1 Introduction
1.1 Background
1.1.1 Advantages of Solar Cells
Necessity of Solar Cells Since Industrial Revolution, an amount of energy used has
increased rapidly. This makes global environmental problems more serious recently.
For example, excess CO2 results in global warming. Furthermore, conventional energy
resources, such as fossil fuels, will be exhausted in the not-too-distant future. Therefore,
we must develop and use alternative energy resources, especially our only long-term
natural resource, the sun [1]. The solar cell is considered to be an attractive candidate
for obtaining energy from the sun because it can convert sunlight directly to electricity with
high conversion efficiency, can provide nearly permanent power at low operating cost, and
is virtually free of pollution.
Comparison with the amount of generated CO2 for each kind of power generation
systems Figure 1.1 shows amount of generated CO2 for each kind of power generation
systems [2]. One can see that the amount of generated CO2 for the solar cell is less than
Electrodeposition of CuInSe2 for Photovoltaic Cell Application 161
Figure 1.1: Amount of generated CO2 for each kind of power generation systems.
that for an oil power plant by a factor of 1/14. The solar cell plants no longer generate CO2
once they have been built. This is the main motivation to develop the solar cell.
Transition of an annual yield for solar cells Figure 1.2 shows a transition of an annual
yield for solar cells from 1995 to 2003 [3]. It has increased rapidly, especially in Japan, and
such an increase is accelerated recently. This is due to the subsidy policy for residences,
which began in 1994. Most of these cells consist of silicon. As the silicon is indirect
semiconductor, an absorption coefficient is in the order of 103 cm1 . Hence, a thickness
of 300 to 500 micron is required for the crystalline silicon solar cell to absorb sun light
sufficiently. Thus, a large amount of high purity single or large-grained polycrystalline
silicon is required to fill demand. A large amount of energy is also required to fabricate the
silicon solar cell. It is well known that silicon is an abundant material. However, there will
be a possibility that we run short of high-purity silicon for solar cells because quantity of
off-glade silicone used for fabricating the Si solar cells separated from the semiconductor
glade standard silicone is restricted [4].
Transition of solar cell costs Figure 1.3 shows the transition of the costs for the system
production and electricity generation of solar cells in Japan [5]. A BOS (balance of system),
inverter and indirect costs has decreased drastically while a cell cost has not been reduced
so much. It is, therefore, important to reduce the cell cost in order to decrease an overall
solar cell system cost.
EPT The solar cell is one of clean energy resources and one of indexes for cleanness is
an energy pay-back time (EPT). The EPT is defined as the time that an amount of energy
generated by a solar cell is equal to an amount of energy used to make the cell. Table 1.1
shows the calculated EPTs for some yield assumptions [6]. The EPTs are estimated to be
Figure 1.2: The transition of an annual yield for the solar cell.
Figure 1.3: The transition of cost for the solar cell.
1.5 to 2.4 y for poly-crystalline Si cells, 1.1 to 2.2 y for amorphous Si cells and 1.1 to 1.7
y for CdTe cells. The EPT becomes shorter as the annual yield increases. An EPT for thin
film cells is expected to be lower than that for crystalline Si solar cells.
Table 1.1: The energy pay-back time (year)
Annual yield MW Poly-crystalline Si Amorphous Si CdS/CdTe

10 2.4 2.2 1.7
30 2.2 1.7 1.4
100 1.5 1.1 1.1
1.1.2 Thin Film Solar Cells

Advantages of thin film solar cells A thin film solar cell seems to be the most attractive
to overcome such problems because it requires smaller amount of materials compared with
the crystalline silicon solar cell. A thickness of the thin film solar cell is as thin as 10 micron
or less. If the thickness of the silicon cell and that of the thin film solar cell are assumed to
be 300 micron and 5 micron, respectively, the amount of materials for the thin film cell can
be reduced by a factor of 1/60. Therefore, the production energy, amount of materials and
cost for the cell can be cut down drastically by introducing the thin film cell [7].
Absorption coefficient The most important parameter for the thin film solar cell is an
absorption coefficient. If the absorption coefficient is higher, the cell with the thinner
thickness can be used. Figure 1.4 shows absorption coefficients for several semiconductor
candidates for the thin film solar cell as a function of photon energy [8]. The absorption
coefficient for c-Si is also shown in this figure for comparison. As the semiconductor
materials, such as CuInSe2 , GaAs, CdTe and a-Si, have the larger absorption coefficients,
they are suitable for an absorption layer of the thin film solar cell [8]. The film thickness
of several micron is sufficient to absorb sun light, as shown in Table 1.2, where absorption
length L = 1/ (: absorption coefficient) [9]. On the contrary, since the crystalline silicon
has the smaller absorption coefficient, the thickness of 300 micron or more is required to
absorb the sun light sufficiently.
Table 1.2: Comparison of the absorption lengths for different semiconductors materials
Materials Absorption lengths m

FeS2 0.02
CuInSe2 0.3
CuInS2 1.0
GaAs 1.5
CdTe 2.0
a-Si 2.0
c-Si 300
Figure 1.4: The absorption coefficients for several semiconductors
1.1.3 CuInSe2 and Related Material Thin Film Solar Cells

History Among semiconductors which are candidates for the thin film solar cells, a
research for copper indium diselenide (CuInSe2 ) and related materials, such as CuInS2
and Cu(Inx Ga1x )Se2 (CIGS), is the most progressive. A mono-crystalline CuInSe2 /CdS
hetero-junction solar cell with an efficiency of 5 % has been developed in 1974 at Bell Lab.
for the first time [10]. In 1975, the efficiency has risen up to 12 % [11]. Recently, the
efficiency of solar cells based on CIGS thin films, whose bandgap is more suitable than
that of CuInSe2 , has reached 19.2 % with a small area (less than 1 cm2 ) [12]. The CIGS
layer for this cell was prepared by so called three stage process with an MBE (Molecular
Beam Epitaxy) apparatus. An efficiency of 14.2 % for the CIGS solar cell with a module
aperture-area of 900 cm2 has been achieved by Showa Shell Sekiyu [13]. The CIGS layer
of this cell was prepared by selenization of a sputtered metal precursor. The basic properties
of CuInSe2 will be detailed in next chapter.
Electrodeposition as a CuInSe2 thin film preparation method For such absorber

materials, one of the most important problems is to reduce cell costs. The CIGS layers for
the high efficiency cells described above are prepared by the co-evaporation or selenization
methods. However, these methods require expensive equipment, such as ultra high vacuum
chambers. Furthermore, the co-evaporation method is hard to apply for preparing large area
cells. The selenization method needs an annealing process with a highly toxic H2 Se gas.
These weaknesses result in high manufacturing costs. On the contrary, electrodeposition
as a method for thin film semiconductor preparation is a good approach with respect to
economic consideration. An important advantage of electrodeposition as the method for
thin film preparation is that films with a large area can be prepared without a vacuum, using
simple and low-cost equipment. Electrodeposition will be detailed in chapter 3.

This paper describes preparation of ternary chalcopyrite semiconductor, i.e. CuInSe2 ,
as a light absorption layer for the thin film solar cell by electrodeposition.
Progress in research for CuInSe2 thin film preparation by electrodeposition In 1983,

electrodeposition of CuInSe2 thin films were demonstrated by Bhattacharya for the first
time [14]. The deposition solution contained In3+ of 0.018 M, Cu+ of 0.018 and SeO2
of 0.025 M as well as triethanolamine of 0.006 vol.% and NH3 of 0.007 vol.%. The
films were deposited on SnO2 :F coated glass substrates at 0.7 V vs a saturated calomel
electrode (SCE) at a room temperature and at pH of about unity. The solution was stirred
during the deposition. Film compositions were not analyzed, but X-ray diffractograms of
the post-annealed films (1 h at 600 o C under Ar) showed many of the reflections of the
chalcopyrite phase. In 1996, his group achieved an efficiency of 8 % using electrodeposited
Cu-In precursors followed by selenization in a vacuum chamber [15]. The efficiency of 9.4
% was attained in the same year by co-deposition of Cu-Se and In-Se followed by physical
vapor deposition (PVD) of Cu, In and Se to adjust final composition [16]. Fernandez et
al. carried out co-electrodeposition of Cu,In and Se, followed by selenizing as-deposited
films by chemical vapor transport by gas method [17]. From the photoelectrochemical
charactrization, energetic characterization and AES depth profile results, they concluded
that their films have a p type bulk because of the Se-rich composition and an n type surface
becasue of the In-rich composition. This group raised up an efficiency to 9.8 % [18] and
13.7 % [19] in 1998. The 9.8 % cell was prepared by adjusting composition by In, Ga and
Se evaporation on electrodeposited Cu-In-Se precursors, while the 13.7 % cell was prepared
from electrodeposited Cu-In-Ga-Se precursors followed by In, Ga and Se evaporation for
final composition adjustment. They achieved an efficiency of 15.4 %, which is the highest
one for the electrodeposited CIGS based thin film solar cells, in 2000 [20, 21]. Depth
profile of CuInSe2 thin films grown by the electrodeposition technique was analyzed by
Calixto et al [22]. A new CIGS electrodeposition bath based on a buffer solution so that the
stability of the electrodeposition process is enhanced without metal oxides or hydroxides
precipitation were developed by Bhattacharya et al [23]. They reported an efficiency of 9.4
% in that paper. These cells above mentioned require evaporating elements to adjust final
composition. Thus, they are developing the process without the PVD [24].
Shih et al. have produced photovoltaic cells with a single absorbing layer with a
conversion efficiency of 4 % in 1987 [25] and 5.2 % in 1989 [26]. They measured
a refractive index [27] and a minority carrier diffusion length [28] of electrodeposited
CuInSe2 thin films. They have achieved an efficiency of 7 % in 1995 [28]. Guillen
et al. have investigated an electrical [29] and optical [30] properties, namely effects of
thermal and chemical treatments on a composition and a crystal structure [31], reaction
pathways to CuInSe2 from electrodeposited precursors [32] and an improved selenization
method [33]. Raffaelle et al. have characterized an electrodeposited CuInSe2 nano-scale
multilayer by scanning tunneling microscopy [34]. They have fabricated the heterojunction
by electrodeposition of CdS films on electrodeposited CuInSe2 films [35] and prepared
CIGS thin films from electrodeposited CuInSe2 /Cu-Ga multilayer precursors [36].
Other interesting studies for the electrodepositon of CuInSe2 are as follows. Seleniza-
tion of electrodeposited Cu-In precursors was studied by Kim et al[37, 38]. Kampmann
et al. have reported large area electrodeposited CIGS thin film solar cells. The best
efficiency of 4.8 % was obtained with a cell area of 80 cm2 [39]. An efficiency of 9 %
has been achieved with the cell prepared by selenization of electrodeposited Cu-In-Ga
stacked layers on a Cu substrate [40]. Ganchev et al. have studied electrodeposition
processes and structural and optical porerties [41, 42]. Kumar et al. have studied
selenization of electrodeposited Cu-In precursors [43]. A new selenization process with
Se[(EtO)2 PS2 ]2 was studied by Fritz et al [44]. Formation of CuInSe2 and Cu(In,Ga)Se2
films by electrodeposition and vacuum annealing treatment was studied by Zhang et al [45].
Kois et al. have reported a partial phase diagram for electrodeposition of CuInSe2 [46].
Another unique and interesting research has been reported by Chaure et al [47].
They have prepared p-i-n type CuInSe2 multilayer solar cells by electrodeposition. They
deposited various types of CuInSe2 thin films potentiostatically in an aqueous solution
containing 0.002M CuSO4 , 0.004M In2 (SO4 )3 and 0.004M H2 SeO3 with Cu:In:Se ratio
of 1:2:2. Small samples with an area of 100mm2 were grown at different cathodic
potentials (0.3 1.2 V vs a saturated Ag/AgCl reference electrode) from the same
electrolyte to evaluate different properties of the layers. Four-layer n-n-i-p photovoltaic
cells (glass/FTO/n-CdS/n-CIS/i-CIS/p-CIS/Au) with back contact area of 0.031 cm2 (2mm
diameter dots) were fabricated. Prior to metal contacting, the cell structures were annealed
at 450 o C for 8 10 min in selenium atmosphere and etched in KCN solution to remove
any Cu-Se phases. Compositional analysis by an X-ray fluorescent spectrometer (XRF)
revealed that at lower cathodic potential range up to 0.70 V, a Cu- and Se-rich layer
was obtained. As the cathodic potential increases, an indium content also increases
and the composition of copper and selenium decreases in the film. Near stoichiometric
CuInSe2 material is obtained at higher cathodic potentials in the range varying from
0.90 to 1.00 V. They have investigated the electrical conductivity type of the layers by
photoelectrochemical characterization using a 0.01M CuSO4 solution. The polarity of the
open-circuit voltage produced by illuminated solid/liquid junction provides the information
on electrical conductivity type. The material grown at cathodic potentials of 0.30 0.65
V showed a p-type (positive photovoltage) conductivity due to more copper and selenium
deposition. At higher cathodic potentials (0.65 0.85 V), indium incorporation into the
films increases. Also such potentials reduce the incorporation of the amount of copper and
selenium resulting in a zero open-circuit voltage for the compensated intrinsic material. As
the cathodic potential increases further (0.85 1.20 V), indium inclusion increases and the
photovoltage becomes negative due to appropriate n-type doping of the material. Further
increase of the cathodic voltage causes production of metallic layers due to the deposition of
indium. It is, therefore, possible to obtain n+ materials at high cathodic potentials (above
1.20 V) and p+ materials at low cathodic potentials (below 0.3 V). They also fabricated
photovoltaic devices using n-, i- and p-type CuInSe2 materials, which were deposited
at 1.00, 0.75 and 0.60 V, respectively, on glass/FTO/n-CdS substrates. The dark and
illuminated I-V characteristics of configuration glass/FTO/n-CdS/n-CIS/i-CIS/p-CIS/Au
shows the rectification factor at 1 V of > 104 , the diode quality factor of n=1.43 and the
potential barrier height of b = 1.10 eV. The best cell parameters measured under AM 1.5
condition were Voc of 410 mV, Jsc of 36 mA/cm2 and F.F. of 0.45.
For more than decade, Lincot et al. have studied systematically reaction mechanisms
involved in electrodeposition of CuInSe2 thin films and prepared solar cells base on them
[48-58]. This group reported the deposition mechanisms including a behavior of a Cu-Se
system in the presence of In3+ [49], a phase diagram [50], transport mechanisms of each
specimen [51], relations between physical properties and deposition conditions [52]. They
found that if In3+ in the solution is excess, the film composition is controlled by the
Se4+ /Cu2+ flux ratio () arriving at the electrode [49]. While, if In3+ concentration
is not sufficiently high, the electrodeposition process is limited by diffusion of all the
three ions. Consequently, the film composition is determined by both flux ratios and
, where is the In3+ /Cu2+ flux ratio [51]. They also attempted to prepare solar
cells based on electrodeposited CuInSe2 . An efficiency of 0.9 % was obtained in 1992
with Se annealing [48]. Introduction of high pressure Se anneal in a closed chamber
with two zone furness, in which temperatures of the substrate and Se are controlled
independently, enables them to obtain higher efficiency of the cells. Efficiencies of 5.6
% [53] and 6.5 % [54] were obtained in 1994 using high pressure Se annealing [55].
Detailed properties of 6.5 % cell such as relationship between a cell performance and film
compositions [56], photoluminescences, photoelectrochemical characteristics, sintering
mechanisms, structural properties, morphologies, transport properties and device analysis
[57] were investigated. Recently, they have reported the highest efficiency of 10.2 % for
electrodeposited CIGS based solar cell [58].
Nomura et al. have prepared CuInSe2 and CIGS thin films by DC and pulse-plated
electrodeposition [59-63]. An efficiency of 1.49 % was obtained for a DC electrodeposited
CuInSe2 based solar cell. In the case of the pulse-plated electrodeposition, they concluded
that films with a stoichiometric composition and a smooth surface has been achieved by
controlling applied pulses with a duty cycle of 33 % and a cathode potential during ON-
time of 0.7 V vs SCE with annealing temperature of 400 o C and annealing duration of 90
min.
1.2 Objective of This Study

To reduce the cost for manufacturing solar cells, a low cost manufacturing method by
which the cell with the large area can be prepared must be developed. This study aim
to establish the electrodeposition technique as a ternary chalcopyrite semiconductor thin
film preparation method. For this purpose, dominant factors which affect compositions in
the films as well as controllability, reproducibility and rate determination step for each ion
were investigated because the film compositions much influence the film properties, such as
conductivity, crystallinity and grain size, and conseqently cell performance. As-deposited
films are annealed in various atmosphere to improve film crystallinity. Morphology must
be improved to fabricate solar cells because the poor morphology causes short-circuit of a
front and back electrodes. Photovoltaic properties of the electrodeposited CuInSe2 based
thin film solar cell are also discussed. Moreover, a new electrodeposition technique to
control compositional depth profile of the film was proposed. This research is believed to
be helpful to establish electrodeposition as the thin film semiconductor preparation method.
2 Basic Properties of CuInSe2 and Solar Cell Structure

To obtain high efficiency solar cells, understanding physical properties of CuInSe2 is
basically requiered. Thus, this chapter describes basic properties of CuInSe2 , i.e. crys-
tallographic properties, a phase diagram and electric properties including defect physics.
Solar cell structure as well as band structure of CuInSe2 solar cells are also shown.
Crystallography Eighteen ternary AI BIII CVI 2 chalcopyrite semiconductors where A =

Cu, Ag, B = Al, Ga, In and C = S, Se, Te are known. The crystal structure of chalcopyrite
is shown in Fig. 2.1 as well as a zindblende structure. It can be systematically constructed
starting from a cubic face centered structure, which is a cubic close-packed structure [64].
By arranging two face centered cubic units shifted 1/4 in order along a cubical diagonal,
the diamond structure is obtained. The zincblende structure can be derived when the each
unit of the cubic face centered structure in the diamond structure is occupied regularly
with two different kinds of atoms, i.e. anion and cation, as shown in Fig. 2.1 (a). Finally,
the chalcopyrite structure can be obtained by doubling the zincblende unit cell along the
c-axis and filling the lattice sites according to the following: The anions remain at their
sites and every cation site is occupied regularly by each of the two kinds of cations,
e.g. Cu and In, as shown in Fig. 2.1 (b). Consequently, each C anion, which configure
a cubic face centered structure, is coordinated by two A and two B cations and each
cation is tetrahedrally coordinated by four anions. The observed structural features for
real chalcopyrite compounds are slightly different from those obtained theoretically from
this construction rules because of two basic chemical bonds A-C and B-C with generally
unequal bond lengths RAC = RBC . This unequality propose the following two unique
properties of chalcopyrite, compered with zincblende: First, the unit cell is tetragonally
distorted with a distortion parameter = c/2a = 1. Second, the anions are displaced from
the ideal tetrahedral site u0 = 1/4 by an amount u, where u is called u parameter and
Figure 2.1: Crystal structure of zincblende (left) and chalcopyrite (right).

expressed as
2 R2
1 RAC
u= + BC
. (2.1)
4 a2
The lattice constants and bandgap of CuInSe2 are summarized in table 2.1 [65]. Note that
the distortion parameter = c/2a is slightly larger than unity.
Table 2.1: The lattice constants and the bandgap of CuInSe2 .
a [nm] c [nm] c/2a bandgap (eV)

0.579 1.160 1.0017 1.04
Figure 2.2: Cu2 Se-In2 Se3 pseudo-binary phase diagram.
Phase diagram The presence of secondary phases in the bulk of the material critically
influence structural, optical and electrical properties of chalcopyrite thin films. In most
cases, the presence of Cu-rich and In-rich secondary phases deteriorates the device
performance and should therefore be eliminated or limited at least. In this regard, phase
diagram gives us useful information in regard to the occurrence of different secondary
phases during film deposition. The available phase information for the CuInSe2 system
is largely limited to the Cu2 Se-In2 Se3 pseudo-binary phase diagram. Fig 2.2 shows the
Cu2 Se-In2 Se3 pseudo-binary phase diagram of CuInSe2 [66]. According to this phase
diagram, various compounds (e.g. Cu2 In4 Se7 , Cu3 In5 Se9 , CuIn3 Se5 , Cu5 In5 Se8 ) can occur
in this ternary system. The homogeneity ranges deduced from X-ray diffraction studies at
room temperature are indicated below the break in the temperature axis. The , and
regions represent distinct phases associated with the compounds CuInSe2 , Cu2 In4 Se7
and CuIn3 Se5 , respectively. The sphalerite phase CuInSe2 ( phase) is stable only at
temperatures higher than 810 C, whereas the chalcopyrite structure ( phase) is stable
from room temperature up to 810 C. In other words, an order-disorder phase transition

( ) occur at 810 C [67]. The chalcopyrite single phase, CuInSe2 , extends from
the stoichiometric composition of 50 mol.% In2 Se3 to the In-rich composition of about
55 mol.% In2 Se3 . The corresponding Cu/In atomic ratio for single phase CuInSe2 lies
between 1.0 and 0.82. In case where the Cu/In atomic ratio is greater than 1.0, the materials
are expected to contain secondary phases of Cu2 Se while for the Cu/In atomic ratio of less
than 0.82 the materials are expected to contain secondary phases of the type Cu2 In4 Se7 and
CuIn3 Se5 [66].
Figure 2.3: Resistivity, type of conductivity and carrier density as a function of Cu/In ratio.
Electrical Properties and defect physics Conductivities of p-type and n-type are
obtained for Se-rich (or Cu-rich) and Se-poor (or In-rich) CuInSe2 material, respectively,
in both thin film and single crystalline forms because CuInSe2 has native defects, such
as vacancies and antisites, which dominate electrical properties, such as a conductivity
type, p or n, and a resistivity, or it contains secondaly phases as writen above. Fig. 2.3
shows a resistivity, a type of conductivity and a carrier density as a function of Cu/In ratio
Table 2.2: Summary of calculated native defect formation energies and defect levels.
point defects formation energies (eV) defect levels (eV) type

V0Cu 0.80
VCu 0.63 Ev + 0.03 acceptor
V0In 3.04
VIn 3.21 Ev + 0.17 acceptor
V2
In 3.62 Ev + 0.41 acceptor
V3
In 4.29 Ev + 0.67 acceptor
Cu0In 1.54
Cu In 1.83 Ev + 0.29 acceptor
Cu2In 2.41 Ev + 0.58 acceptor
In2+
Cu 1.85
In+Cu 2.55 Ec 0.34 donor
In0Cu 3.34 Ec 0.25 donor
Cu+ i 2.04
Cu+ i 2.88 Ec 0.20 donor
VSe 2.40 Ec 0.08 donor
[68, 69, 70]. The resistivity of CuInSe2 is drastically changed with Cu/In ratios by a factor
of 106 . Cu-rich CuInSe2 shows a low resistivity because it is a complex of CuInSe2 and
low resitivity Cu2x Se. On the other hand, In-rich CuInSe2 has a high resistivity probably
due to compensation of donors and acceptors. Stoichiometric and Cu-rich CuInSe2 show
generally p-type conductivity and In-rich n-type.
Zunger et al. and Wasim calculated native defect formation energies and defect
energy levels by first principle calculation [71, 72, 73]. These are summarized in Table
2.2, where V , and i represent vacancies of atom, atom in site and
interstitial, respectively. In Fig. 2.4, thier predicted defect transition levels are compared
with experimental data [72]. From these results, Cu vacancies, which is a shallow acceptor
whose level is 30 meV above the valence band edge, are found to be easy to form. The
results also show V In and CuIn defects act as aceptors while InCu and Cui defects donors.
Moreover, they reported that the structurel tolerance to large off-stoichiometry in CuInSe2
is explained by the unusual stability of (2V0Cu + In0Cu ) defect pair whose formation energy
is as low as 6.1 eV/pair when there is a astrong interaction between the components of
the defect pairs. This low formation energy explains the existence of the Orderd Defect
Compuounds (ODC) ,such as CuIn5 Se8 , CuIn3 Se5 , Cu2 In4 Se7 and Cu3 In5 Se9 . They also
concluded that the electrically benign nature of the structural deffects is explained in terms
of the electronic passivation of the In2+
Cu by VCu .
Structure of CuInSe2 thin film solar cells Figure 2.5 shows a cross sectional schematic
representation of a CuInSe2 -based solar cell. Cell preparation starts with the deposition of a
Mo back contact on glass, followed by fabrication of a p-type ternary chalcopyrite absorber,
a high resistivity CdS or other weakly n-type buffer layer, bi-layer with undoped ZnO and
n-type transparent conductor (usually Al doped ZnO or ITO (Indium Tin Oxide)) shown as
Figure 2.4: Defect transition energy levels from (a) their thory and (b) experiments.
Figure 2.5: Schematic representation of a cross section for the CuInSe2 solar cell.
TCO in Fig. 2.5, metal grids and a antireflection coating (not shown in Fig. 2.5) [74].
Typical band structure of the CuInSe2 -based solar cell is shown in Fig. 2.6 [75, 76, 77].
In this figure, Ec and Ev denote a conduction and valence band offsets at the CIS and
Figure 2.6: Schematic representation of a band structure for the CuInSe2 solar cell.
CdS interface, respectively. Results of simulation show that Ec which act as a barier
for photo-excited electron hardly influence cell performance if Ec is less than 0.4 eV
[78]. Nelson et al. have reported Ec and Ev are about 0.4 [79] and 0.8 eV [80],
respectively. High efficiency CuInSe2 solar cells have a Cu-poor surface, resulting in a
surface composition of CuIn3 Se5 . The n-type surface material produce type inversion at
the surface of absorber. This leads to a shift of the regime p=n into the absorber and hence
away from the defect rich CdS/CIS interface resulting in a reduced recombination rate. In
addition, the grading of the valence band close to the surface provides a transport barrier
for holes which reduce the hole density at the interface and hence decrease the interface
recombination rate [76].
3 Experimental Details
3.1 Electrodeposition
Electrodeposition is one of methods to deposit metal and semiconductor thin films. It is
almost the only one method to obtain such films at a room temperature and a nomal pressure.
It is based on an electrochemical reduction reaction of ions on a cathode substrate surface
in an electrolyte. This section describes the basics of electrodeposition [81, 82, 83].
3.1.1 Theory
Electrodeposition of metallic films (commonly known as electroplating) has long been
known and used for preparing metallic mirrors and a corrosion-resistant surface, and so on
[81]. Electrodeposition process can be divided into the following three steps: (1) Ions are
provided from a bulk of an electrolyte. (2) Adatoms are generated by discharging electrons
from a cathode to ions. (3) Adatoms diffuse on the crystal surface and incorporated into
the crystal. The first process is mainly due to diffusion of the ions from the bulk of the
electrolyte to the substrate surface. The second one is interpreted as reduction reactions of
ions at the substrate surface. The third process is the same as that of vapor phase crystal
growth, such as evaporation. The redox potential Er for a electronation reaction,
M z+ + ze = M, (3.1)
is given by
RT
Er = E0 + ln[M z+ ], (3.2)
zF
where M z+ : activity of a metal ion (az+ M /aM ), z: ionic charge, M : deposited metal,
E0 : standard electrode potential, R: gas constant, T : absorute temperature, F : Faradays
constant. If the reaction starts by applying voltage between a cathode and an anode, metal
ions, M z+ , accept electrons from the cathode and metal M deposits on the surface of
the cathode. Consequently, a concentration difference of ions between the bulk of the
electrolyte and an adjacent area of the cathode surface arises because of decreasing ion
concentration near the cathode. This difference is the driving force to provide the ions
from the bulk to the cathode surface by diffusion. An overpotential, , which is an
electrodeposition driving force, has a relationship with a difference of chemical potentials
between equilibrium and under deposition () and this relationship is given by

= E Er = , (3.3)
zF
where E is the cathode potential during the deposition.
According to Sterns model, an electrode-electrolyte interface can be divided into two
parts: One is a compact double layer known as the Helmholtz double layer adjacent to the
electrode and the other is a diffuse layer known as the Gouy-Chapman layer. Figure 3.1
shows potential and ion distributions around the cathode surface. A potential varies linearly
and exponentially with distance in the Helmholtz layer and the diffuse layer, respectively,
as shown in Fig. 3.1.
The Helmholtz double layer can be divided into an inner and outer planes. The inner
Helmholz plane (IHP) is adjacent to the electrode surface and consists of completely
oriented water dipoles and contact-adsorbed ions. These layers are shown in Fig. 3.2. In
aqueous electrolytes, the cations react rather strongly with the water molecule, and their
inner hydration sphere is retained. This limits their closest distance of approach. They are
usually separated from the cathode by approximately one or two water molecules.
On the other hand, anions interact weakly with water, and their closest distance of
approach could correspond to direct contact. The outer Helmholtz plane (OHP) consists of
solvated ions at the closest distance of approach from the electrode surface. In the diffuse
layer, there is a difference between concentration of anions and that of cations. Outside of
the diffuse layer is the bulk of electrolyte, and the distribution of anions and cations is not
changed whether potential is applied or not.
In the case of alloy, Mp Nq is deposited by a eutectoid reaction as shown in Eqs.3.4 and
3.5,
M m+ + me = M, N n+ + ne = N (3.4)
Figure 3.1: Potential and ion distributions.
Figure 3.2: Helmholz double layer.

pM + qN = Mp Nq , (3.5)
and electrodeposition potentials for each component are as follows;
RT
Er,M = EM + ln[M m+ ] (3.6)
mF
RT
Er,N = EN + ln[N n+ ] (3.7)
nF
where EM : standard electrode potentials for each element in the alloy and are
and EN
given by
RT
EM = EM
o
+ ln(aM ) (3.8)
mF
RT
EN = EN o
+ ln(aN ) (3.9)
nF
where aM and aN : activity in the alloy. E is given by
E = Er,M + M = Er,N + N (3.10)
where M and N : overpotential of each element. This indicates that a difference between
redox potentials for each element is equal to a difference between overpotentials. The
difference of redox potentials is usually too large to be compensated by difference in
concentrations for each ion. On the other hand, under electrodeposition, a concentration
of metal ion adjacent to the electrode surface, Cs , is smaller than a bulk concentration, C0 .
At Cs = 0, a velocity of electrodeposition reaches critical level and increase no more even
if a potential veries more negative. A critical diffusion current id is given by
id = nF DC0 / (3.11)
where D: diffusion constant, : diffusion layer thickness (i.e. diffusion limited). For
the current higher than id , the cathode potential should be jumped to a more negative
value. Thus, co-electrodeposition is carried out under the condition of a sufficiently
negative cathode potential to deposit the less noble metal with the concentration of the
noble metal ion lowered until the deposition current for the noble metal reaches critical.
For example, electrodeposition of ZnSe is explained as follows [83]. Figure 3.3 shows a
typical current-potential curve for electrodepositon of ZnSe. In the case of the solution
concentration of [H2 SeO3 ][Zn2+ ], elemental Se begins to deposit at the potential of
Er,Se = 0.159V. ZnSe begins to deposit at Er,Zn +EZn = 0.159V when the deposition
rate of elemental Se reaches critical and the potential is shifted more negative. However,
since the ZnSe deposition is limited by diffusion of H2 SeO3 , the current reaches critical
again when the cathode potential becomes more negative (from -0.159 to -0.962 V). At the
further negative potential of Er,Zn = 0.962V, a deposition of elemental Zn results in a
Zn composition increase in the deposited material. Thus, single phase ZnSe is deposited
between 0.159 and 0.962 V.
Figure 3.3: Theoretical current-voltage curve for electrodeposition of ZnSe.
3.1.2 Experimental Apparatus

Electrodeposition was carried out potentiostatically or galvanostatically in an aqueous
solution using a standard three electrode cell. Figure 3.4 shows a schematic representation
of the electrodeposition cell. A titanium or moribudium sheet was used as a cathode
substrate electrode, a carbon rod as an anode counter electrode and a standard calomel
electrode (SCE) or a Ag/AgCl electrode (SSE) as a reference electrode. The substrates
were ultrasonically cleaned in organic solvents. If necessary, substrates were etched in
acidic solutions, a pH value was controlled by adding dilute acidic solution and a solution
was stirred by a magnetic stirrer. Detail electrodeposition conditions will be shown in each
section.
3.2 Annealing and Evaluation Methods

The highest efficiencies of electrodeposited CuInSe2 based thin film solar cells are lower
than that of evaporated or sputtered ones because quality of electrodeposited films is
generally poor. The film resistivity is higher because of smaller grain size due to a
deposition condition of high overpotential corresponding to a highly supersaturated vapor
condition for the vapor-phase growth. For this reason, all of researchers have reported that
a thermal process such as annealing is necessary to improve film quality [83]. In general,
crystallinity, grain size and adhesion of the film is improved by the annealing. An annealing
apparatus consisting of a electric furnace and a quartz tube with gas line of N2 is used in this
work. Annealing temperatures were varied from 250 to 600 o C and annealing time from 30
min to 120 min. Detailed annealing conditions will be shown later.
Composition, crystallinity, surface morphology, cross-sectional morphology and com-
positional depth profile of as-deposited and annealed films were evaluated by an electron
Figure 3.4: The electrodeposition cell.
probe microanalyzer (EPMA), an X-ray diffraction (XRD), a scanning electron microscope

(SEM) and an X-ray photoelectron spectroscope (XPS), respectively.
4 Electrodeposition of CuInSe2 Layers and Their Solar Cell

Application
4.1 Introduction
This chapter describes electrodeposition of Cu-In-Se films prepared with an aqueous
solution containing CuCl2 , InCl3 and SeO2 in terms of composition control of deposited
films for the preparation of CuInSe2 [84, 85]. Taking account of the abscondence of Se
element from deposited layers during post-annealing, the preparation of Cu-In-Se layers
with an excess Se content has been studied using a solution with a high Se content. An
important problem for the development of this technique is to control sample compositions
because they affect film properties which influence cell performance very much, as shown
in section 2. Understanding of deposition mechanisms of each species is essential in
order to achieve higher controllability and reproducibility of film composition and then to
improve performances of CuInSe2 solar cells prepared by electrodeposition. Crystallinity
of as-deposited films has also been described.
In general, electrodeposited films must be annealed to improve their crystallinities
[86]. Thus, effects of annealing on compositions, morphologies and crystallinities were
investigated.
Solar cells based on electrodeposited CIS thin films are fabricated with a chemical-
bath-deposited n-CdS buffer layer on the CIS layer without a transparence conductive layer
such as Al-doped ZnO or ITO [86]. The best efficiency of 0.87 % is obtained with the
electrodeposited CuInSe2 thin film.
4.2 Relationship between Compositions in the Films and That in the Baths
It is reasonable for us to consider that the compositions in the films are affected by that in
the baths. For composition control of deposited films, it is preferable that there is a simple
relation between them. For example, the compositions in the films are proportional to that
in the baths and proportionality constants are not affected by other parameters. Moreover,
it is preferable that, for example, a In/Cu ratio is not affected by a Se concentration. On
the basis of above, relationships between compositions in the films and that in the baths are
investigated.
Figure 4.1 shows relationships between Se/Cu ratios in the films, , and those in the
source baths, [Se(IV)]/[Cu(II)]. Deposition conditions are summarized in Table 4.1. One
can see that for the wide range of current density (10 mA/cm2 ), a linear relationship is
held between these values. Under the assumption that all the Se(IV) and Cu(II) ions arriving
at the electrode are involved in the deposited film, a ratio of mass-transfer coefficients for
Se and Cu, k(Se)/k(Cu) is obtained as a ratio of to [Se(IV)]/[Cu(II)] [87]:

k(Se)/k(Cu) = . (4.1)
[Se(IV)]/[Cu(II)]
In this case, a constant k(Se)/k(Cu) value ( 0.65) is obtained for the wide range of
current density when no stirring is employed.
Figure 4.2 shows relationships between In/Cu ratios in the films, , and those in the
source baths, [In(III)]/[Cu(II)]. The deposition condition is the same as Table 4.1. A linear
relation is held between them only for a low current density (therefore, a noble electrode
potential) and for a solution with a low solute concentration. Using the same assumption as
the case for k(Se)/k(Cu), a ratio of mass-transfer coefficients for In and Cu, k(In)/k(Cu) is
obtained as follows:

k(In)/k(Cu) = . (4.2)
[In(III)]/[Cu(II)]
The ratio increases as the current density increases as shown later.
Table 4.1: Deposition conditions No.1
Substrate material Mo
Source materials Cu: CuCl2 : 0.05mM
and In: InCl3 : 1.0mM
their concentrations Se: SeO2 : 0 1.0mM
Composition ratio in the baths Cu:In:Se = 15:20:20
A pH value of solution 2.0
A solution temperature 20 o C
Substrate potentials 0.5 2.2 V vs SCE
Current densities 1.0 10 mA/cm2
Deposition duration 9 90 min.
Stirring 90rpm (if necessary)
15
1mA/cm2
3
10 5
10
&A
5
0 10 20
[Se(IV)]/[Cu(II)]
Figure 4.1: Relationships between Se/Cu ratio in the film, , and those in the source bath,
[Se(IV)]/[Cu(II)].
4 1mA/cm2
3
5
10
&B
0 10 20
[In(III)]/[Cu(II)]
Figure 4.2: Relationships between In/Cu ratio in the film, , and those in the source bath,
[In(III)]/[Cu(II)].
4.3 Relationship between the In/Cu Ratio in the Film and Se(IV) Concentra-
tion in the Bath
In general, deposition rates of each element (Cu, In and Se) are competitive. Thus, if
a deposition rate of one element (e.g. Se) increases, that of other elements (Cu and In)
must decrease under a constant current density. If rates of the decrease for these elements
15
[In(III)]/[Cu(II)]=20
2 mA/cm2
10 pH:1.0
pH:2.0
&B
5
0 10 20
[Se(IV)]/[Cu(III)]
Figure 4.3: Relationships between In/Cu ratio in deposited film, , and Se/Cu ratio in the
source bath, [Se(IV)]/[Cu(ll)]
100
Composition [at%]
Cu
50 In
Se
0
0 10 20
[Se(IV)]/[Cu(II)]
Figure 4.4: Dependence of the composition for each element on the Se concentration in the
source-bath
are same, there is no problem in terms of composition control. However if not, some
complications, such as an In/Cu ratio in the film depends on a Se bath concentration, arise
for composition controlling. In this electrodeposition system, the Se concentration in the
bath was found to influence the In/Cu ratios in the films as follows. Figure 4.3 shows
the relationship between the In/Cu ratios in the films, , and Se(IV) concentrations in the
baths. Deposition conditions are summarized in Tabel 4.2. As seen in Fig. 4.3, decreases
drastically and then becomes constant as [Se(IV)]/[Cu(II)] increases. This result shows
In/Cu ratio in the films

2
With stirring
Without stirring
0 1 2 3
Deposition potential vs SCE [V]
Figure 4.5: Variation in In/Cu ratio in film by stirring on Mo substrate
15
Se/Cu in the films
10
5
Without stirring
With stirring
0 1 2 3
Figure 4.6: Variation in Se/Cu ratio in film by stirring
that the Se/Cu ratio in the source bath is also one of the important factors to control the
In/Cu ratio in a deposited film. To investigate origin for this dependence, a composition
dependence of each element on Se concentrations is examined. Figure 4.4 shows the
dependence of the composition for each element on the Se concentrations in the source
bath. The figure shows that a composition of In is much effected by Se concentration while
that of Cu is hardly influenced. As a result, this dependence is found to be mainly due to a
decrease in the In deposition rate with increasing the Se concentration in the bath.
Table 4.2: Deposition conditions No.2
Substrate material Mo
Source materials Cu: CuCl2 : 0.05mM
and In: InCl3 : 1.0mM
concentration Se: SeO2 : 0 1.0mM
Composition ratio in the baths Cu:In:Se = 1:20:020
A pH value of solution 2.0
A solution temperature 20 o C
Substrate potentials 0.5 2.2 V vs SCE
Current densities 1.0 10 mA/cm2
Deposition durations 9 90 min.
Stirring 90rpm (if necessary)
0.8
k(Se)/k(Cu), k(In)/k(Cu)
k(Se)/k(Cu)NS
0.6
k(Se)/k(Cu)S
0.4
k(In)/k(Cu)NS
0.2
k(In)/k(Cu)S
0 5 10
Current Density [mA/cm2]
Figure 4.7: k(Se)/k(Cu) and k(In)/k(Cu) with (-S) and without (-NS) stirring of the solution
as a function of current density.
4.4 Rate-Determination Step and Stirring Effects

From effects of stirring of the solution during the deposition on film compositions
or ratios of mass-transfer coefficients, generally, one can get useful information about
rate-determining steps of electrodeposited species. Influences of the solution stirring on
the film compositions and its controllability are shown in Figs. 4.5 and 4.6. The stirring
of the solution during the deposition decreases in both In/Cu and Se/Cu ratios in deposited
films. This seems to be due to the enhanced deposition rate of Cu by the stirring because
a rate-determining step for Cu is diffusion-limited as described later. It is noted that a
deposition rate of a diffusion-limited species is independent of a electrode potential and,
therefore, the rate does not increase even if a total deposition current increases by applying
Without stirring
In/Cu ratio in film

4
2
With stirring
0 1 2 3
Figure 4.8: Maximum and minimum values of In/Cu ratio for nine specimens prepared with
and without stirring.
In/Cu ratio in the films
Mo
2 pSi
0 2 4
Substrate potential vs SCE [V]
Figure 4.9: Substrate potential dependences on film compositions. Comparison with a Si

and Mo substrate.
a more negative potential. On the contrary, a deposition rate of a reaction-limited species

increases with a deposition current. There is a general tendency that a rate-determining step
changes from reaction-limited to diffusion-limited when a deposition current density
increases.
Figure 4.7 shows k(Se)/k(Cu) and k(In)/k(Cu) with and without the stirring as a function
of a deposition current density. Both k(Se)/k(Cu) and k(In)/k(Cu) values in Fig. 4.7 are
4 pSi substrate
In/Cu tatio in the films

With stirring
Without stirring
2
0 2 4
Substrate potential vs SCE [V]
Figure 4.10: Variation in In/Cu ratios in the films on the Si substrate by stirring.
smaller than those reported for the sulfate-based solution [87, 88]. A remarkable reduction
of the k(Se)/k(Cu) and the k(In)/k(Cu) at a low current density by the stirring shows that the
deposition rate of Cu(II) is most influenced by the stirring, compared with other species.
Since the diffusion-limited process is mainly affected by the stirring, it is reasonable to
consider that the deposition reaction of Cu(II) is diffusion-limited in the wide range of
current density (1-10 mA/cm2 ) even under the stirring condition.
The k(Se)/k(Cu) values without stirring are independent of a current density, as shown
in Fig. 4.7. This result shows that the supply of both elements is diffusion-limited when
no stirring is employed. When the solution is stirred, k(Se)/k(Cu) values at low current
densities (<5mA/cm2 ) are decreased and show a current density dependence, while those at
high current densities are not changed by stirring and still show no significant dependence
on change in the current density. This suggests that the rate-determining step for Se(IV)
deposition at a low current density is changed from diffusion to reaction by employing
the stirring.
The increase in k(In)/k(Cu) without stirring with increasing current density shows that
the deposition of In(III) is reaction-limited at a low current density (< 5 mA/cm2 ). Even
when the stirring is employed, the deposition of In(III) at a high current density (10
mA/cm2 ) is diffusion-limited, as seen in Fig. 4.7.
From the above, the rate-determining step for each ion is summarized in Fig. 4.11 The
k(In)/k(Cu) at the high current density (< 5[mA/cm2 ]) under no-stirring condition can not
be calculated because the is not proportional to [In(III)/Cu(II)] under such condition.
Marked improvement of controllability and uniformity for film composition is achieved
by employing the stirring. Figure 4.8 shows a fluctuation of In/Cu ratios in the films with
and without stirring. Nine specimens were prepared for the EPMA analysis from three thin
film samples deposited under the same conditions. Each sample with an area of 1 cm2
was divided into three specimens. The maximum and minimum composition data for nine
specimens are plotted in Fig. 4.8 as a function of electrode potentials. As can be seen in
Fig. 4.8, reproducibility and uniformity of the film composition are improved by employing
the stirring. For a film with an In/Cu ratio of about unity, the fluctuations with and without
stirring are about 7 and 21 %, respectively. Thus, the reproducibility of the In/Cu ratios in
the film is improved by a factor of 3 by employing the stirring.
4.5 Dependences on Substrate Materials of Film Compositions

This section describes dependences on substrate materials, semiconductor or metal, of
film compositions. Figure 4.9 shows the dependences of In/Cu ratios in the films on
substrate potentials, deposited on a p-Si and Mo substrates. This figure indicates that the
substrate potentials less influence compositional variations of the films deposited on the
p-Si substrate than on the Mo. Moreover, the dependence of compositions in the films
deposited on the Si substrate with or without stirring on the substrate potentials is shown
in Fig. 4.10. Compared with that of films on the Mo substrate shown in Fig. 4.5, the
stirring less influence the compositional variation for the p-Si substrate. To investigate
causes of the difference, energy band diagrams at the interface between the p-Si substrate
and a electrolyte are considered. In general, the energy band is bent at the interface as
shown in Fig. 4.12 (a). For this bending, a space charge layer whose charge distribution is
polarized is formed at the semiconductor surface, resulting in the potential drop. When a
negative bias is applied as shown in Fig. 4.12 (b), the almost voltage is applied at the space
charge layer. Figure 4.13 shows potential distributions for the case of a Mo (a) and Si (b)
substrates. In the potential distributions, an available potential for deposition is the potential
applied at the Helmholz layer. In the case of the Mo substrate, almost substrate potential
is considered to be applied to the Helmholz layer. Hence, a variation of the potentials
applied to the Helmholz layer according to a change in the substrate potential results in a
variation of the film compositions. On the contrary, in the case of the p-Si substrate, the
potential is considered to be preferentially applied to the space charge layer rather than
the Helmholz layer. A variation of the substrate potential is, therefore, consumed as the
Figure 4.11: Summery of the rate-determining step.

Figure 4.12: Energy level diagram for an interface between p-Si and the electrolyte. No
bias (a) and negative bias (b).
variation of the potential applied to the space charge layer. Consequently, the composition
change becomes small because the variation of the potentials applied to the Helmholz layer
becomes smaller than that of the potentials applied to the substrate. Moreover, the reason
why the stirring hardly influence compositions of films for the Si substrate is considered
that a rate determination step of Cu ion is not diffusion-limited due to a noble effective
potential for the Si substrate and, consequently, Cu ions are sufficiently supplied to the
substrate surface from the bulk of the electrolyte.
4.6 Crystallinity of As-Deposited Films

Figure 4.14 shows XRD patterns of as-deposited films with different In/Cu ratios. CuInSe2
was found to be contained in as-deposited films from the figure. The copper-rich and
stoichiometry films have no secondly phases while the In-rich film contains an elemental
indium. Peaks from elemental Se was not observed although as-deposited films contain
excess Se, as described above. This indicates that excess Se in these films is of amorphous
phase. These broad peaks from the CuInSe2 show that these films have poor crystallinity.
Thus, in order to improve crystallinity, an annealing process is needed. Details of
post-annealing effects will be described later.
4.7 Surface Morphology and its Improvement

Surface morphology is one of important factors for high efficiency solar cells. In this
section, surface morphologies of Se rich samples are discussed [86]. The film and solution
compositions are summarized in Table 4.3.
Figure 4.15 shows surface morphologies of samples with different Se compositions.
The figure denotes that both samples has rough surface. However there are some differences
Figure 4.13: Potential distributions for the case of the Mo (a) and Si (b) substrates.
between them, especially, in grain sizes, i.e., the sample B has a larger grain than the
sample A. The largest grain size in sample B is about 10 m. The sample B is superior
to sample A in terms of the grain size. However, a morphological improvement is required
because even the morphology of the sample B has not sufficient quality in uniformity for
Table 4.3: Compositions in the films and those in the solutions for morphology investiga-
tions.
Sample No. Solution compositions Film compositions

Cu:In:Se In/Cu Se/(Cu+In)
A 1:20:20 0.97 4.42
B 1:5:5 0.78 1.92
the solar cell application. One can also see gaps between each grain even in the sample
B. In general, substrate potentials dominate both film compositions and morphologies in
electrodeposition. In this case, first priority is film compositions when the potentials are
decided. Thus, the potentials are not optimized in terms of the surface morphologies.
Hence, electrochemical etching is attempted to improve surface morphologies by applying
reverse biases. It is expected to uniform a film thickness and to reduce a surface roughness
because the electrochemical etching preferentially occurs from the tip of the rough surface.
Figure 4.16 shows a morphological comparison before (a) and after (b) etching. The figure
shows the surface-roughness of the etched sample is improved. It also seems that the gaps
between each grain decrease. Thus, the electrochemical etching is found to be effective to
improve the surface morphology.
To enhance effects of the reverse bias, pulse-plated electrodepositon is proposed.
Namely, each forward and reverse bias are applied periodically in sufficiently short time
compared with a deposition duration. Here, lets call a period for which the forward
bias is applied to a substrate ON-time and the reverse bias OFF-time. Influences of
substrate potentials and duty ratios on surface morphologies are investigated. The duty
ratio is defined as a time ratio between ON-time and OFF-time. Figure 4.17 shows
surface morphologies of the films deposited with various duty ratios and potentials for
ON-time and OFF-time. The deposition conditions as well as In/Cu ratios in the films
Figure 4.14: XRD patterns of the as-deposited films.

Figure 4.15: Surface morphologies of samples with different Se compositions. The bar
denotes 50 m.
Figure 4.16: Surface morphology of (a) an etched film and (b) an as-deposited film. The
bar denotes 10 m.
are summarized in Table 4.4. By comparing Figs.4.17 (a), (b) and (c), grain sizes are found
to become larger and a ratio where large grains are included increases with a decrease in the
duty ratios. However, differences in size between the largest and smallest grains become
larger with an increase in the duty ratios. Uniformity in the grain sizes is not improved
by decreasing duty ratios. The figures also show a decrease in roughness with an increase
in reverse bias potentials (for example, see Figs.4.17 (c), (e) and (g)). The results indicate
that by applying the reverse bias with appropriate duty ratios the surface morphologies are
improved.
Figure 4.17: Surface morphologies of the films deposited by pulse-plated electrodeposition

with various duty ratios and potentials. The bar denotes 50 m.
4.8 Effects of Annealing on Se Excess Films

Electrodeposited CuInSe2 films must be annealed to improve their crystallinities because
as-deposited films have poor crystallinity, as shown in previous section. Taking account
of the abscondence of Se element from deposited layers during the annealing, as-deposited
Cu-In-Se films have Se-rich compositions, as described above. Thus, effects of the excess
Se on annealed film properties, such as compositions, crystallinities and morphologies,
must be investigated. In this investigation, samples with a similar Cu/In ratio and a different
Se/(Cu+In) ratio are used. Samples with very (Se/(Cu+In)=3.07.0) and slightly (1.23.0)
excess Se are labeled A and B, respectively. Annealing temperatures are varied from 150
550 o C. The samples are annealed for 30 min in N2 flow of 200 ml/min [86]. .
Change in compositions Figure 4.18 shows a change in compositions before and after
annealing at 400 o C. In/Cu ratios of sample B hardly change with annealing while those of
sample A become larger. This suggests that Cu-Se compounds with lower vapor pressure
which are produced by a reaction between Cu and excess Se evaporate. Figure 4.19 and
4.20 show variations of Se/(Cu+In) and In/Cu ratios as a function of annealing temperature,
respectively. The Se/(Cu+In) ratios of sample A drastically decrease with an increase in the
temperature, while, those of sample B hardly change. In/Cu ratios of both sample A and B
do not change so much.
Crystallinity Figures 4.21 and 4.22 show XRD patterns of annealed sample A and B,
respectively. The figures indicate that sample B is superior to A because FWHM and
peak intensities of sample B are smaller and higher than those of sample A, respectively.
Peaks from Se are observed at 150 o C and disappeared at 250 o C for both samples. At 450
o C, In O and other unknown peaks are observed. Thus, 400 o C is found to be the most
2 3
appropriate temperature for annealing electrodeposited CuInSe2 thin films.
Morphology Figures 4.23 (a) and (b) show surface morphologies of sample A and B,
respectively, annealed at 400 o C which is the best temperature for crystallinity. Sample A
has very rough surface probably due to Se and/or Cu-Se evaporation. The surface of sample
B is smoother than that of sample A.
Table 4.4: Deposition conditions and compositions of pulse-plated electrodeposition.
Duty
OFF-time Pot. (V) 100 % (D.C.) 90 % 50 %
Fig. No. (a) (b) (c)
0 ON-time Pot. (V) 1.5 1.5 2.25
In/Cu ratios 0.78 0.60 0.81
Fig. No. (d) (e)
+0.1 ON-time Pot. (V) 2.0 2.5
In/Cu ratios 0.81 1.06
Fig. No. (f) (g)
+0.5 ON-time Pot. (V) 2.0 3.0
In/Cu ratios 0.81 0.98
These results indicate that too excess Se is undesirable for film qualities, e.g. composi-
tion, crystallinity and morphology.
4.9 Cell Performance

As shown in section 2, general CuInSe2 solar cells have a CdS and TCO layers. However,
in this study, a TCO layer is not deposited because deposition conditions for the TCO
layer is not established [86]. The cell structure fabricated is Mo/CIS/CdS/Au. The CdS
buffer layer is generally prepared by a chemical bath deposition method as follows [89, 90].
Electrodeposited CIS thin films are dipped in an aqueous solution containing CdCl2 of
1mM, NH4 Cl of 5mM and (NH2 )2 SC of 7.5 mM for 60 min at 60 o C. A pH of solution is
maintained to 9.5 by adding aqueous ammonium. A film thickness and a resistivity of CdS
layer are estimated to be 273 nm and 105 cm, respectively.
Figure 4.24 shows current-voltage curves under a dark and light conditions. A
power-voltage curve is also shown in the figure. The curves show rectification and this
indicates that a pn junction is formed. From the curve under the light condition, following
cell properties are obtained: a Voc of 0.47 V, an Isc of 1.9 mA/cm2 , a fill factor (FF) of
0.533 and a conversion efficiency of 0.866 %. However, the conversion efficiency is very
low, compared with the highest one. Therefore, further investigation and development are
required to obtain higher conversion efficiency.
4.10 Conclusion
Electrodeposition of Cu-In-Se films has been studied with an aqueous solution containing
CuCl2 , InCl3 and SeO2 , in terms of composition control of deposited films for the
preparation of CuInSe2 . Dependences of compositions in the films on bath compositions,
current density, substrate potential and presence of stirring are investigated. As a result, for
In/Cu ratios in the annealed films
4
Sample A
2
Sample B
0 1 2 3
In/Cu ratios in the asdeposited films
Figure 4.18: Change in compositions before and after annealing.

Se/(Cu+In) ratios in the annealed films

6
Sample A
4
Sample B
2
0 0 200 400
Annealing temperature (oC)
Figure 4.19: Variation of Se/(Cu+In) ratios as a function of annealing temperature.

In/Cu ratios in the annealed films
0.5
Sample A
Sample B
0 0 200 400
Annealing temperature(oC)
Figure 4.20: Variation of In/Cu ratios as a function of annealing temperature.
the wide range of current density (10 mA/cm2 ), a linear relationship is found to be held
between and [Se(IV)]/[Cu(II)] while such relation is held between and [In(III)]/[Cu(II)]
only for a low current density (therefore, a noble electrode potential) and for a solution with
a low solute concentration. Not only the In/Cu ratio in the bath but also the Se concentration
is found to be affected the In/Cu ratios in the films. Under the assumption that all the
Se(IV), In(III) and Cu(II) arriving at the electrode are involved in deposited films, the ratios
of the mass-transfer coefficients k(Se)/k(Cu) and k(In)/k(Cu) has been obtained. From their
dependencies on deposition current density and stirring of the solution, the rate-determining
step for each species in the deposition process has been discussed. The reproducibility
Figure 4.21: XRD patterns of annealed sample A as a function of annealing temperature.
Figure 4.22: XRD patterns of annealed sample B as a function of annealing temperature.
and uniformity of In/Cu ratio in the film is improved by a factor of 3 by employing

the stirring. X-ray diffraction patterns show that CuInSe2 is contained in as-deposited
films. Morphologies of as-deposited films can be improved by applying reverse bias
after deposition or by pulse-plated electrodeposition, as well as annealing. Excess Se is
undesirable for film qualities, such as composition, crystallinity and morphology. The I-V
curve shows that pn junction is formed and the efficiency of the cell is 0.87 %.
Figure 4.23: Surface morphologies of sample A and B annealed at 400 o C. The bar denotes
50 m.
Figure 4.24: I-V and P-V curves.
5 New Electrodeposition Technique for Controlling Depth

Profile of CuInSe2 Thin Films
It is known that the efficiency of CuInSe2 -based solar cells can be improved using a
composition that can be modulated to be In-rich near the pn junction interface. In this
work, a new electrodeposition technique for preparing CuInSe2 thin films with controlled
depth profile was developed. CuInSe2 thin films having a bilayer structure, that is, films
with the Cu-rich and In-rich layers, were successfully deposited from one electrolyte only
by changing the substrate potential during electrodeposition.
5.1 Introduction
By introducing the so called bi-layer method to prepare CuInSe2 layers, an efficiency
larger than 10 % was attained for the first time [91]. The bilayer structure was fabricated
by sequential vacuum evaporation, that is, a Cu-rich layer was deposited first, and then
subsequently, an In-rich layer was deposited on the Cu-rich layer, resulting in films with
two layers, i.e., the Cu-rich and In-rich layers. The reason for the better performance
of the bilayer structure is as follows. The composition of CuInSe2 thin films affects the
electric and structural properties, as described in section 2 (page 170). Cu-rich films have
a low resistivity, which causes short circuiting in the pn-junction, and a large grain size,
which leads to a large diffusion length and is desirable for solar cell application. On
the other hand, In-rich films have a high resistivity and a small grain size. Thus, the
composition near the pn-junction interface should be In-rich to prevent short circuiting,
and the rest should be Cu-rich to enhance diffusion length. That is, in order to obtain a
high-efficiency cell, the compositional depth profile is one of the most important factors.
However, there are a very few reports on the compositional depth profile of electrodeposited
CIS thin films [92, 93, 94]. Moreover, conversion efficiencies ever reported for the cells
fabricated by electrodeposition are considerably lower than that for the cells fabricated by
other methods. Although Bhattacharya et al. reported the highest efficiency of 15.4 % by
this method, a physical vapor deposition (PVD) was needed to adjust the final composition
[92]. Guimard et al. have reported a conversion efficiency of 9.1 % without PVD [95].
As described above, an important problem for the development of this technique is the
control of the composition ratio including depth profile. Some authors have investigated
the compositional depth profiles of the CuInSe2 thin films prepared by electrodeposition
[92, 93, 94]. Calixto and Sebastian have reported that the concentration of Cu is higher
than that of In at the initial, and then the Cu concentration decrease with deposition
time, resulting in CuInSe2 films with the Cu-poor surface and the Cu-rich bulk [93].
Although this profile is desirable, the properties, such as a thickness of Cu-poor layer
and compositions of each layer, are uncontrollable by their method. Friedfeld et al. have
reported the two step electrodeposition method for forming a Cu-Ga/CuInSe2 bilayer [94].
A high surface content of Ga and a low bulk content form the graded bandgap. They use
two types of electrolyte: one containing Cu and Ga ions and the other Cu, In and Se ions.
However, the electrodeposition method using two types of electrolyte seems to be a little
cumbersome because the substrates must be transferred. Chaure et al. have prepared p-i-n
type CuInSe2 multilayer solar cells by electrodeposition, as shown in section 1.1.3 [96].
However, they have not mentioned the compositional depth profile of their multilayer. One
of the causes for low efficency of electrodeposited CIS thin film solar cells is considered not
to be optimized in the compositional depth profile. Thus, new electrodeposition technique
to control the compositional depth profile must be developed.
This section describes a new technique for forming CuInSe2 films with a bilayer
structure by electrodeposition [97, 98]. The Cu-rich and In-rich layers were deposited from
one electrolyte by changing the substrate potential during the deposition. Composition
controlled bilayer CuInSe2 thin films were prepared very easily by this method.
Table 5.1: Standard redox potentials.

Chemical reactions Standard redox potentials vs. SHE (V)
Cu(II) + 2e Cu 0.342
In(III) + 3e In 0.338
H2 SeO3 + H+ Se + 3H2 O 0.74
5.2 Experiments
Cu, In and Se were electrodeposited from a bath containing 1 mM CuCl2 , 5-7 mM InCl3 and
3-5 mM SeO2 using a conventional three electrode electrochemical cell and a potentiostat
at room temperature. Substrate potentials were 0.5 V vs the Ag/AgCl reference electrode
for the deposition of the Cu-rich layer and 0.8 V for the deposition of the In-rich layer.
These values were selected by considering the standard redox potentials of Cu, In and Se,
as shown in Table 5.1. Conventional D.C. electrodeposition for depositing the single-layer
films was carried out at 0.5 V and 0.8 V to determine the bath composition for the
bilayer film deposition. A 1 cm2 Ti sheet and a carbon rod were used as a substrate and
a counter electrode, respectively. Each deposition time was determined by the amount
of electric charge measured using a coulombmeter: first step 0.75 C and second step
1.0 C. As-deposited films were evaluated by an electron prove micro-analyzer (EPMA)
for overall chemical compositions and an X-ray photoelectron spectroscope (XPS) for the
depth-profile. Surface morphology was observed by a scanning electron microscope (SEM).
5.3 Results and Discussion

Preliminary experiments were carried out to determine the bath composition. Table 5.2
shows the deposition conditions and composition results measured by EPMA for the single-
layer deposition. From these results, the Cu- and In-rich films were found to be obtained
at 0.5 V and 0.8 V, respectively, from the solution containing 1 mM CuCl2 , 5 mM InCl3
and 3 mM SeO2 . Thus, these values were selected for the bilayer deposition.
Table 5.3 shows overall compositions of the bilayer films prepared at 0.5 V and 0.8 V
measured by EPMA. Slightly In-rich and/or stoichiometric films were obtained under this
condition.
Figure 5.1 shows the depth-profile of the bilayer film deposited under the same
conditions described above. The bottom of the film is found to be Cu-rich and the top
In-rich. The bilayer CuInSe2 films were obtained by this method. However, the surface
seems to be Cu-rich. This remarkably differ from the evaporated films, which have In-rich
surface [99]. The thickness of the top (In-rich) layer was about 0.3 m and the overall
thickness larger than 1 m. The differences between the compositions prepared at 0.8 V
shown in Table 5.2 (Cu/In=0.85) and Fig. 5.1 (0.1) is probably due to the Cu concentration
decrease in the electrolyte because of the deposition of the Cu-rich layer as the first step
of the bilayer deposition, which leads to a decrease in Cu/In ratio at the second step.
The insufficient calibration in the quantitative analyses using the EPMA and the XPS is
considered one of the causes of such differences.
Table 5.2: Electrodeposition conditions and film compositions of preliminary experiments.
Bath comp. mM Pot. Film comp. % Film comp.

Cu In Se V Cu In Se Cu/In Se/(Cu+In)
1 5 5 -0.5 31.41 4.23 64.36 7.42 1.81
1 5 5 -0.8 25.56 12.25 62.20 2.09 1.65
1 7 5 -0.5 26.73 10.67 62.60 2.50 1.67
1 7 5 -0.8 21.03 16.36 62.61 1.29 1.67
1 5 3 -0.5 30.46 14.77 54.77 2.06 1.21
1 5 3 -0.8 25.92 23.94 50.14 1.08 1.01
Table 5.3: Overall compositions of electrodeposited bilayer thin films.
Bath comp. mM Pot. V Film comp. % Film comp.

Cu In Se 1st 2nd Cu In Se Cu/In Se/(Cu+In)
1 5 3 -0.5 -0.8 25.20 27.22 47.58 0.93 0.91
1 5 3 -0.5 -0.8 24.45 28.83 46.72 0.85 0.88
1 5 3 -0.5 -0.8 25.38 24.67 49.94 1.03 1.00
Figure 5.2 shows the morphology of the bottom (Cu-rich) and top (In-rich) layers
prepared at 0.5 V and 0.8 V, respectively. The top layer was peeled to observe the
bottom layer. The Cu-rich layer has isolated crystallites or grains of 1 to 3 m in size,
while the In-rich layer has a smooth surface. However, grain boundaries were not observed
definitively and some cracks existed in the In-rich layer.
Figure 5.1: Composition depth profile of electrodeposited bilayer film measured by XPS.
Figure 5.2: Surface morphology of bottom (Cu-rich:left) and top (In-rich:right) layers. The
bars denote 10 m.
5.4 Summary
In summary, the bilayer CuInSe2 films is successfully deposited by changing substrate
potential during electrodeposition. Further crystallographic, electrical, optical and cross-
section morphological investigations are needed to establish this method as the CuInSe2
thin film preparation method. A post-deposition annealing may be required to improve
film property, such as crystallinity and grain size. For convenience, the ternary compound
was deposited in this study. However, as a high efficiency is generally obtained using a
quaternary chalcopyrite compound solar cell, the preparation of Ga-containing films must
be studied. On the basis of the present technique, it may also be possible to obtain a graded
composition profile, which is more effective in improving the solar cell efficiency than the
step profile.
6 Conclusion
6.1 Summary of this Work
Electrodeposition is studied as a low cost thin film semiconductor preparation method.
Ternary chalcopyrite semiconductor thin films, that is, copper indium diselenide (CuInSe2 )
is prepared by this method. This is a promising material as an absorber layer for thin
film solar cells because of its suitable bandgap and a large absorption coefficient. For
ternary materials, composition controlling is essential to obtain high efficiency thin film
solar cells because properties of this material are sensitive to its composition. Thus,
electrodeposition conditions including bath compositions, substrate potentials, substrate
materials, current densities, pH of solution and presence of stirring on film properties
such as film compositions, crystallinity, surface and cross-sectional morphologies and cell
performances are investigated. The results of this work are summarized as follows.
For the wide range of current density, Se/Cu ratios in the films are proportional to
those in the source baths. On the other hand, a linear relation is held between In/Cu
ratios in the films and those in the source baths only for low current densities and for
solutions with low solute concentrations.
The In/Cu ratios in the films decrease drastically and then become constant as the
Se(IV) concentrations in the baths increase, and consequently, it is found that the
Se/Cu ratio in the source bath is also one of the important factors to control the In/Cu
ratio in a deposited film.
The rate-determining step for Cu is diffusion-limited under all condition in this

work.
The rate-determining step for Se deposition at low current densities is changed from
diffusion-limited to reaction-limited by employing stirring.
The deposition of In(III) is reaction-limited at low current densities.
When a solution is stirred, the deposition of In(III) changes from reaction-limited

to diffusion-limited with an increase in the current density.
Reproducibility of the In/Cu ratios in the films is improved by a factor of 3 by stirring.
Difference in potential distribution between metal and semiconductor substrates

causes difference in composition of the films.
A reverse bias and pulse-plated electrodeposition are effective to improve a surface

morphology of the films.
Excess Se degrades the controllability of composition and the crystallinity.
The conversion efficiency of 0.866 % (Voc=0.47 V, Isc=1.9 mA/cm2 , FF=0.533) is

obtained.
CuInSe2 thin films with the bilayer structure, i.e. the Cu-rich and In-lich layers, are
successfully deposited by changing substrate potential during electrodeposition.
6.2 Suggestion for Future Works

In this work, ternary chalcopyrite semiconductor thin films were successfully deposited
by electrodeposition. However, the efficiency of the cell is very low (< 1%). For the cell
fabrication, an n-type semiconductor, such as CdS or ZnS, as a buffer layer and a transparent
conductive layer, such as ITO (indium tin oxide) or FTO (F doped SnO2 ), must be deposited
on a electrodeposited CuInSe2 layer. Hence, deposition conditions for these layers must be
optimized. Moreover, one of the causes for the low efficiency seems to be existence of
cracks and voids in the films. Thus, to obtain crack and void free thin films, surface and
cross-sectional morphology must be investigated in more detail. Improvement of surface
properties as well as that of bulk crystal is also important.
The bi-layer CuInSe2 films were successfully deposited by changing substrate
potential during electrodeposition. However, further crystallographic, electrical, optical
and cross-section morphological investigations are needed to establish this method as the
CuInSe2 thin film preparation method. A post-deposition annealing may be required
to improve film properties, such as crystallinity and grain size. For convenience, the
ternary compound was deposited in this study. However, as the higher efficiency is
generally obtained with quaternary chalcopyrite compound Cu(Inx Ga1x )Se2 (CIGS), the
preparation of Ga containing film must be studied. Furthermore, it is also important to
attempt to obtain graded composition profile by gradually changing potential profile.
References
[1] S. M. Sze: Physics of Semiconductor Devices 2nd edition: John Wiley & Sons (1981)
790.
[2] Central Research Institute of Electric Power Industry News No.338 (2000)
http://criepi.denken.or.jp/jp/pub/news/pdf/den338.pdf [in Japanese].
[3] Solar Systems No.96 (2004) 46 [in Japanese].
[4] M. Konagai Ed. Basic and application of thin film solar cells Ohmsha (2001) 3 [in
Japanese].
[5] Photovoltaic Power Generation Technology Research Association Homepage

http://www.pvtec.or.jp/PVTEC [in Japanese].
[6] New Energy and Industrial Technology Development Organization (NEDO) report
http://www.nedo.go.jp/nedata/14fy/01/h/0001h008.htm (2000) [in Japanese].
[7] NEDO Database http://www.nedo.go.jp/nedata/14fy/01/g/0001g002.htm [in

Japanese].
[8] M. Konagai Ed. Basic and application of thin film solar cells Ohmsha (2001) 6 [in
Japanese].
[9] A. Ennaoui, S. Fiechter, Ch. Pettenkofer, N. Alonso-Vante, K. Buker, M. Bronold, Ch.

Hopfner, H. Tributsch: Solar Energy Materials and Solar Cells 29 (1993) 289-370.
[10] S. Wagner, J. L. Shay, P. Migliorato and H. M. Kasper: Appl. Phys. Lett. 24 (1974)
434.
[11] J. L. Shay, S. Wagner and H. M. Kasper: Appl. Phys. Lett. 27 (1975) 89.
[12] H. S. Ullal: Tech. Dig. 14th International Photovoltaic Science and Engineering
Conference, Bangkok, Thailand (2004) 417.
[13] M. Konagai: Tech. Dig. 14th International Photovoltaic Science and Engineering
Conference, Bangkok, Thailand (2004) 657.
[14] R. N. Bhattacharya J. Electrochem. Soc. 130 (1983) 2040.
[15] A. M. Fernandez, P. J. Sebastian, R. N. Bhattacharya, R. Noufi, M. Contreras and A.

M. Hermann: Semicond. Sci. Technol. 11 (1996) 964-967.
[16] R. N. Bhattacharya, A. M. Fernandez, M. A. Contreras, J. Keane, A. L. Tennant, K.

Ramanathan, J. R. Tuttle, R. Noufi and A. M. Hermann: J. Electrochm. Soc, 143
(1996) 854-858.
[17] A. M. Fernandez, P. J. Sebastian, M. E. Calixto, S. A. Gamboa and O. Solorza: Thin

Solid Films 298(1997)92-97.
[18] M. E. Calixto, R. N. Bhattacharya, P. J. Sebastian, A. M. Fernandez, G. A. Gamboa

and R. Noufi: Solar Energy Materials and Solar Cells 55 (1998) 23-29.
[19] R. N. Bhattacharya, W. Batchelor, J. E. Granata, F. Hasoon, H. Wiesner, K.

Ramanathan, F. Keane and R. Noufi: Solar Energy Materials and Solar Cells 55
(1998) 83-94.
[20] R. N. Bhattacharya, W. Batchelor, K. Ramanathan, M. A. Contreras, T. Moriarty Solar

Energy Materials and Solar Cells 63 (2000) 367-374.
[21] R. N. Bhattacharya, J. F. Hiltnerb, W. Batchelora, M. A. Contrerasa, R. N. Noufia and

J. R. Sitesb: Thin Solid Films 361-362 (2000) 396-399.
[22] M. E. Calixto and P. J. Sebastian: Solar Energy Materials and Solar Cells 63 (2000)
335-345.
[23] R. N. Bhattacharya and M. A. Fernandez: Solar Energy Materials and Solar Cells 76
(2003) 331-337.
[24] M. E. Calixto, P. J. Sebastian, R. N. Bhattacharya and R. Noufi: Solar Energy

Materials and Solar Cells 59 (1999) 75-84.
[25] I. Shih and C. S. Qiu: Conf. Rec. 19th PVSC New Orleans, LA (1987) 1291.
[26] C. X. Qiu and I. Shin: Can. J. Phys 67 (1989) 444.
[27] S. N. Qiu, C. X. Qui and I. Shih: Materials Letters 17 (1993) 190-191.
[28] S. N. Qiu, L. Li, C. X. Qui, I. Shih and C. H. Champness: Solar Energy Materials and
Solar Cells 37(1995)389-393.
[29] C. Guillen and J. Herrero: Proc. 11th EC Photovol. Solar Energy Conf. (1992)
815-817.
[30] C. Guillen and J. Herrero: Solar Energy Materials and Solar Cells 43 (1996) 47-57.
[31] C. Guillen and J. Herrero: J. Electrochm. Soc. 141 (1994) 225-230.

[32] C. Guillen and J. Herrero: J. Electrochm. Soc, 142 (1995) 1834-1838.
[33] C. Guillen and J. Herrero: J. Electrochm. Soc, 143 (1996) 493-498.
[34] R. P. Raffaelle, J. G. Mantovani and R. Friedfeld: Solar Energy Materials and Solar
Cells 46(1997) 201-208.
[35] R. P. Raffaelle, H. Forsell, T. Potdevin, R. Friedfeld, J. G. Mantovani, S. G. Bailey, S.

M. Hubbard, E. M. Gordon and A. F. Fepp: Solar Energy Materials and Solar Cells
57(1999) 167-178
[36] R. Friedfeld, R. P. Raffaelle and J. G. Mantovani: Solar Energy Materials and Solar
Cells 58 (1999) 375-385.
[37] Y. H. Kim, Y. E. Lee, H. J. Kim and J. Song: Conf. Rec. 1st World Conf. Photovol.
Energy Conv. (1994) 202-205.
[38] Y. E. Lee, H. J. Kim, Y. J. Kim, K. K. Lee, B. H. Choi, K. H. Yoon and J. S. Song: J.

Electrochm. Soc, 141 (1994) 558-561.
[39] A. Kampmann, V. Sittinger, J. Rechid and R. Reineke-Koch: Thin Solid Films 361-362
(2000) 309-313.
[40] J. Rechid, R. Thyen, A. Raitzig, S. Wulff, M. Mihhailova, K. Kalberlah and

A. Kampmann: Abst. Tech. Prog. 3rd World Conference on Photovoltaic Energy
Conversion Osaka, Japan (2003) 2LN-C-02.
[41] M. G. Ganchev and K. D. Kochev: Solar Energy Materials and Solar Cells 31 (1993)
163-170.
[42] N. Stratieva, E. Tzvetkova, M. Ganchev, K. Kochev and I. Tomov: Solar Energy

Materials and Solar Cells 45(1997) 87-96.
[43] S. R. Kumar, R. B. Gore and R. K. Pandey: Solar Energy Materials and Solar Cells
26 (1992) 149-158.
[44] H. P. Fritz and P. Chatziagorastou: Thin Solid Films, 247(1994) 129-133.
[45] L. Zhang, F. D. Jiang and J. Y. Feng: Solar Energy Materials and Solar Cells 80
(2003) 483-490.
[46] Julia Kois, Sergei Bereznev, Mare Altosaar and Enn Mellikov: Abst. Tech. Prog.
3rd World Conference on Photovoltaic Energy Conversion Osaka , Japan (2003)
4P-B4-29.
[47] N. B. Chaure, J. Young, A. P. Samantilleke and I. M. Dharmadasa: Solar Energy

[48] L. Thouin, J. F. Guillemoles, S. Massaccesi, R. Ortega, P. Cowache, S. Rouquette-

Sachez, D. Lincot and J. Vedel: Proc. 11th EC Photovol. Solar Energy Conf. Montreux
Switzerland (1992) 866.
[49] L. Thouin, S. Massaccesi, S. Sanchez and J. Vedel: J. Electroanal. Chem. 374 (1994)
81-88.
[50] L. Thouin, J. F. Guillemoles, S. Massaccesi, R. Ortega. Borges, S. Sachez, P. Cowache,

D. Lincot and J. Vedel: Solid State Phenomena 37-38 (1994) 527-532.
[51] L. Thouin and J. Vedel: J. Electrochem. Soc. 142 (1995) 2996-3001.
[52] J. Vedel, L. Thouin and D. Lincot: J. Electrocham. Soc. 143 (1996) 2173-2180.
[53] J. F. Guillemoles, P. Cowache, S. Massaccesi, L. Thouin, S. Sachez, D. Lincot and J.

Vedel: Adv. Mater. 6 (1994) 379-381.
[54] D. Lincot, J. F. Guillemoles, P. Cowache, S. Massaccesi, L. Thouin, K. Fezzaa, F.

Boisivon and J. Vedel: Conf. Rec. 1st World Conf. Photovol. Energy Conv. (1994)
136-139.
[55] J. F. Guillemoles, A. Lusson, P. Cowache, S. Massaccesi, J. Vedel and D. Lincot: Adv.

Mater. 6 (1994) 376-379.
[56] J. F. Guillemoles, R. Diaz, P. Cowache, L. Thouin, D. Lincot and J. Vedel: Preprint of

the 13th European Photovol. Solar Energy Conf. Nice, France (1995)
[57] J. F. Guillemoles, P. Cowache, A. Lusson, S. Massaccesi, K. Fezzaa, F. Boisivon, J.

Vedel and D. Lincot: J. Appl. Phys. 79 (1996) 7293-7302.
[58] D. Guimard, N. Bodereau, J. Kurdi, J. F. Guillemoles, D. Lincot, P. P. Grand, M.

BenFarrah, S. Taunier, O. Kerrec and P. Mogensen: Abst. Tech. Prog. 3rd World
Conference on Photovoltaic Energy Conversion Osaka, Japan (2003) 2P-D3-58.
[59] Y. Sudo, S. Endo and T. Irie: Jpn. J. Appl. Phys. Vol. 32 (1993) 1562-1567.
[60] T. Matsuoka, Y. Nagahori and S. Endo: Jpn. J. Appl. Phys. 33 (1994) 6105-6110.
[61] S. Endo, Y. Nagahori and S. Nomura: Jpn. J. Appl. Phys. 35 (1996) L1101-L1103.
[62] S. Nomura, K. Nishiyama, K. Tanaka, M. Sakakibara, M. Ohtsubo, N. Furutani and

S. Endo :Jpn. J. Appl. Phys. 37 (1998) 3232-3237.
[63] S. Nomura and S. Endo Ternary and Multinary Compound in the 21st Century IPAP
Books 1 (2001) 131-136.
[64] T. Riedle: Ph.D Thesis, Technical University Berlin (2002) 4
[65] N. Yamamoto Ed.: Novel Functional Semiconductors IPC (1989) 19 [in Japanese].
[66] M. L. Feartheiley: Solar Cells 16 (1986) 91.
[67] N. Yamamoto Ed. : Novel Functional Semiconductors IPC (1989) 157 [in Japanese].
[68] J. Bougnot, S. Duchemin and M. Savelli: Solar Cells 16 (1986) 221.

[69] V. K. Kapur, P. Singh, M. V. Choudary, F. M. Auno, L. Elyash and S. Meised: Proc.

17th IEEE Photovol. Specealists Conf. (1984) 777.
[70] R. Noufi, R. Axton, C. Heffington and S. K. Deb: Appl. Phys. Lett. 45 (1984) 668.
[71] A. Zunger, S. B. Zhang and S-H. Wei: Proc. 16th IEEE photoboltaic Specialists Conf.
(1997) 313-318.
[72] S. B. Zhang, S-H. Wei and A. Zunger: Physical Review B, 57 (1998) 9642-9656.
[73] S. M. Wasim: Solar Cells 16 (1986) 289-316.
[74] M. Kemell: Dr Thesis, University of Helsinki (2003) 20.
[75] U. Rau and H. W. Shock: Appl. Phys. A69 (1999) 131-147.
[76] T. Dullweber, G. Hanna, U. Rau and H. W. Shock: Tech. Dig. 11th International
Photovoltaic Science & Engineering Conference (1999) 85-86.
[77] M. Konagai Ed.: Basic and application of thin film solar cells Ohmsha (2001) 187 [in
Japanese].
[78] M.Ruckh, D.Schmid and H.W.Schock, J. Appl. Phys. 76 (1994) 5945.
[79] S-H. Wei and A. Zunger: Appl.Phys, Lett. 63 (1993) 2549.
[80] A. J. Nelson, C. T. Schwerdtfeger, S-H. Wei, A. Zunger, D. Rioux, R.Patel and

H.Hochst: Appl. Phys. Lett. 62 (1993) 2557.
[81] P. K. Pandey, S. N. Sahu, S. Chandra: Handbook of Semiconductor Electrodeposition

Marcel Dekker, Inc. (1996).
[82] I. Ohno and S. Haruyama, Jap. J. Materia Japan (1991) 735-742 [in Japanese].
[83] I Ohno: Oyo Buturi 64 (1995) 798-802 [in Japanese].
[84] S. Nakamura, S. Sugawara, A. Hashimoto and A. Yamamoto: Tech. Dig. 9th

International Photovoltaic Science & Engineering Conference, Miyazaki, Japan
P-I-D-51 (1996) 387-388.
[85] S. Nakamura, S. Sugawara, A. Hashimoto and A. Yamamoto: Solar Energy Materials

and Solar Cells, 50 (1998) 25-30.
[86] S. Sugawara: Master Thesis, Fukui University (1997).
[87] L. Thouin and J. Vedel: J. Electrochem. Soc., 142 (1995) 2996.
[88] L. Thouin, S. Massaccesi, S. Sanchez, and J. Vedel. J. Electroanal. Chem., 374

(1994) 81.
[89] T. Nakanishi and K.Ito: Solar Energy Materials and Solar Cells 35 (1994) 171-178.
[90] S. Kuranouch, T. Nakazawa, A. Ashida and N.Yamamoto: Solar Energy Materials

and Solar Cells 35 (1994) 185-191.
[91] R. A. Mickelsen, W. S. Chen, Y. R. Hsiao and V. E. Lowe: IEEE Trans. Electron

Devices, 31 (1984) 542.
[92] R. N. Bhattacharya, W. Batchelor, K. Ramanathan, M. A. Contreras and T. Moriarty:

Solar Energy Materials and Solar Cells 63 (2000) 367-374.
[93] M. E. Calixto and P. J. Sebastian, Solar Energy Materials and Solar Cells 63 (2000)
335-345.
[94] R. Friedfeld, R. P. Raffaelle and J. G. Mantovani, Solar Energy Materials and Solar
Cells 58 (1999) 375-385.
[95] D. Guimard, D. Lincot, J. Guillemoles, S. Taunier, P. Mogensen, N. Bodereau, J.

Kurdi, P. P. Grand O.Kerrec and M. Benfarrah: Abst. Tech. Prog. 3rd World Conf.
Photovol. Energy Conv. Osaka (2003) 34.
[96] N. B. Chaure, J. Young, A. P. Samantilleke and I. M. Dharmadasa: Solar Energy

[97] S. Nakamura: Tech. Dig. 14th International Photovoltaic Science and Engineering
Conference, Bangkok, Thailand, P-138 (2004) 611-612.
[98] S. Nakamura: Jpn. J. Appl. Phys. 44 (2005) 1939-1940.
[99] D. Schmid, M. Ruckh, F. Grunwald and H. W. Schock: J. Appl. Phys. 73 (1993) 2902.
INDEX
availability, 50, 52
A averaging, 102
accommodation, 124
accounting, 148 B
accumulation, ix, 95, 97, 108, 112, 115, 116, 118
accuracy, 98, 115 backscattering, 75
acetone, 126 bandgap, xi, 138, 159, 164, 169, 197, 200
acoustic microscopy, vii barriers, 97, 98
activation, 76, 77, 84 baths, 179, 181, 183, 201
activation energy, 76, 77, 84 behavior, 77, 104, 125, 167
adhesion, 177 bending, 186
adjustment, 165 benign, 171
affect, 28, 35, 36, 63, 72, 111, 130, 167, 178 bias, x, xi, 97, 100, 106, 114, 115, 123, 126, 140,
aggregates, viii, 31, 32, 60, 61, 62, 64 160, 186, 187, 189, 190, 195, 201
aggregation, viii, 31, 57, 58, 64 binding, 124, 131, 135, 155
algorithm, 103, 107 binding energy, 124, 131, 135, 155
alloys, 2, 124, 143, 144, 152, 156 bismuth, 4, 5, 6, 17, 19, 20, 21, 22, 27, 28
alternative, 4, 63, 160 body, 37, 101
alternative energy, 160 Boltzmann constant, 98, 131
alters, 100 bonding, 72, 73, 144
ambient air, 54 bonds, 19, 71, 73, 83, 84
ammonium, 193 boundary value problem, 98
amplitude, 45, 46, 152 breakdown, vii
anisotropy, 115, 116, 118
annealing, 55, 56, 124, 164, 166, 167, 177, 178, 187,
192, 193, 194, 195, 200, 202 C
Argentina, 69, 93
calculus, 89
argon, 39
calibration, 198
ash, 101
candidates, 125, 163, 164
assimilation, 32
carbon, viii, 31, 32, 33, 36, 37, 39, 45, 46, 47, 48, 49,
assumptions, 63, 161
54, 57, 58, 59, 60, 61, 62, 63, 64, 177, 198
asymmetry, 85, 90, 91, 152, 156
carbon atoms, viii, 31, 58
atomic force, 126
carrier, x, 105, 123, 129, 130, 132, 137, 165, 170
atomic force microscope, 126
cation, 99, 168
atoms, viii, x, 18, 21, 22, 31, 36, 37, 40, 41, 42, 43,
cell, xi, 42, 70, 71, 72, 98, 107, 108, 109, 126, 149,
56, 57, 58, 62, 63, 64, 71, 72, 88, 98, 123, 125,
159, 160, 161, 162, 163, 164, 165, 166, 167, 168,
128, 129, 130, 144, 150, 168
171, 172, 173, 177, 178, 189, 193, 195, 197, 198,
attention, 36, 124, 125
200, 201
210 Index
CH3COOH, 5 cooling, viii, 5, 31, 32, 36, 40, 41, 57, 58, 59, 60, 64,
channels, ix, 95, 96, 97, 106, 108, 110, 115, 118 126
charge migration, vii copper, xi, 33, 159, 164, 166, 187, 200
chemical bonds, 168 correlation, x, 50, 85, 143, 156
chemical vapor deposition, 124 corrosion, 173
China, 123, 140 costs, 161, 164
classification, viii, 32, 33, 35, 36, 38, 39, 41, 45, 61, coupling, 96, 102, 103, 115
62, 65 crack, 202
clean energy, 161 critical value, 58
clusters, 32, 44, 58, 63 crystal growth, viii, 3, 7, 9, 21, 31, 35, 36, 37, 38, 39,
CO2, 160, 161 40, 43, 44, 47, 50, 51, 55, 58, 61, 63, 64
coagulation, 63 crystallinity, xi, 160, 167, 177, 187, 191, 192, 193,
collisions, 73, 101, 103 195, 200, 201, 202
compensation, 171 crystallisation, 39
competition, 42 crystallites, viii, 69, 74, 78, 79, 127, 139, 199
complement, 105 crystallization, viii, 31, 39, 40, 41, 42, 43, 48, 57, 59,
compliance, 156 60, 62, 63, 64, 73, 77, 91
complications, 181 crystals, vii, viii, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 16, 17,
components, vii, 24, 28, 82, 106, 171 18, 21, 22, 23, 25, 26, 27, 28, 31, 32, 33, 34, 35,
composition, vii, xi, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 36, 37, 38, 39, 40, 41, 43, 44, 45, 46, 47, 48, 49,
13, 14, 15, 16, 17, 18, 19, 21, 22, 27, 28, 36, 70, 50, 51, 52, 53, 54, 55, 57, 59, 60, 61, 62, 63, 64,
143, 144, 146, 150, 159, 160, 165, 166, 167, 170, 87, 91, 99, 103, 157
173, 176, 178, 179, 181, 182, 185, 186, 187, 193, CVD, 73, 124
195, 196, 197, 198, 200, 201, 202
compounds, 98, 143, 144, 149, 151, 152, 154, 155,
168, 169, 192 D
computation, 101
database, 149
concentration, ix, x, xi, 2, 5, 11, 13, 37, 41, 42, 45,
decay, 96, 97, 109
46, 48, 49, 50, 54, 55, 56, 57, 58, 60, 69, 70, 73,
decomposition, viii, 31, 56, 57, 59, 60, 61, 62, 64
74, 75, 77, 78, 81, 83, 84, 89, 123, 125, 126, 128,
defects, viii, 31, 32, 33, 34, 35, 38, 39, 41, 43, 44,
129, 130, 132, 133, 135, 137, 139, 144, 145, 153,
45, 46, 49, 50, 51, 52, 53, 54, 55, 56, 57, 59, 60,
154, 156, 160, 167, 174, 176, 179, 181, 182, 183,
61, 62, 63, 64, 124, 125, 126, 129, 130, 132, 133,
194, 197, 198
152, 154, 156, 170, 171
concrete, 48
definition, vii, 33, 43
condensation, 58
deformation, 41, 102, 103, 104
conduction, 97, 104, 124, 125, 132, 136, 137, 172
degenerate, ix, 95, 98, 106, 113, 115
conductivity, ix, 24, 70, 74, 75, 76, 77, 83, 86, 89,
degradation, ix, 69, 70, 147
91, 92, 96, 126, 166, 167, 170, 171
degradation process, 70
conductor, 171
demand, 161
configuration, 41, 42, 166
density, xi, 33, 45, 48, 54, 61, 72, 75, 84, 85, 96, 97,
confinement, ix, 74, 85, 90, 95, 99, 132, 152, 156
99, 100, 102, 107, 108, 110, 111, 125, 129, 130,
confusion, 33, 36
144, 152, 155, 160, 170, 173, 179, 180, 183, 184,
conservation, 101, 109
185, 193, 194, 201
construction, 168
deposition, ix, xi, 69, 71, 73, 75, 83, 84, 124, 145,
contaminant, 126
159, 160, 165, 166, 167, 169, 171, 174, 176, 177,
contamination, 37
178, 179, 180, 182, 183, 184, 185, 186, 189, 193,
context, 33, 48
194, 195, 197, 198, 200, 201, 202
continuity, 3, 28, 53
deposition rate, ix, 69, 73, 83, 180, 182, 183, 184,
control, xi, 28, 37, 59, 124, 159, 167, 178, 179, 181,
185
182, 193, 197, 201
deposits, 174
convergence, 104
desire, 124
conversion, xi, xii, 50, 51, 71, 159, 160, 165, 193,
detachment, 40
197, 201
detection, 19
Index 211
deviation, 84, 148 electrons, ix, 49, 73, 95, 96, 97, 98, 99, 100, 101,
dielectric constant, 98, 102, 103, 132, 135 102, 103, 104, 106, 107, 109, 111, 112, 113, 114,
diffraction, xi, 15, 18, 45, 46, 79, 80, 129, 133, 145, 115, 117, 118, 125, 132, 137, 138, 139, 173, 174
147, 148, 149, 155, 156, 157 electroplating, 173
diffusion, xi, 3, 22, 28, 36, 42, 43, 48, 51, 56, 58, emission, vii, x, 97, 102, 103, 104, 108, 109, 111,
127, 128, 130, 154, 160, 165, 167, 174, 176, 183, 112, 114, 123, 124, 125, 128, 129, 130, 131, 132,
184, 185, 187, 197, 201 134, 135, 136, 137, 138, 139, 140, 145, 152, 153,
diffusion process, 22, 56 154, 155
diffusion rates, 128 entropy, 41, 42, 43, 48
diodes, 96, 124, 125, 126, 136, 144 epitaxial growth, vii, 50, 51, 127
dislocation, viii, 31, 32, 35, 36, 37, 39, 43, 45, 46, equilibrium, ix, 21, 36, 42, 43, 60, 61, 70, 92, 100,
47, 50, 51, 52, 53, 55, 56, 57, 58, 59, 61, 63, 64 101, 102, 104, 108, 109, 113, 114, 115, 174
disorder, x, 73, 77, 83, 143, 144, 147, 148, 149, 150, equipment, 70, 85, 164, 165
152, 154, 155, 156, 170 etching, viii, 11, 31, 32, 33, 34, 37, 38, 44, 45, 51,
dispersion, 84, 85, 88, 145 52, 54, 189
distribution, vii, viii, ix, 1, 3, 26, 27, 28, 31, 32, 33, European Commission, 119
35, 36, 37, 38, 39, 40, 44, 45, 48, 50, 51, 54, 63, evacuation, 4
64, 85, 91, 95, 97, 98, 100, 103, 104, 107, 108, evaporation, 164, 165, 174, 192, 197
109, 113, 114, 116, 144, 153, 155, 174, 186, 201 evidence, x, 83, 90, 91, 143
distribution function, ix, 95, 100, 103, 104, 107, 108, evolution, 108, 128
109, 113, 116 EXAFS, 156
domain, 19, 20 excitation, x, 49, 73, 75, 126, 143
donors, 32, 98, 130, 171 exciton, x, 123, 124, 125, 129, 131, 132, 134, 135,
dopants, 136 139, 154, 155, 156
doping, 37, 48, 59, 60, 62, 63, 70, 73, 77, 78, 89, 96, exclusion, 106, 107, 111, 113
99, 124, 125, 134, 136, 139, 166 experimental condition, 75, 151
duration, 11, 103, 107, 110, 111, 167, 179, 189 exponential functions, 112
exposure, 49, 89, 90
expression, 42, 132
E
earth, 70 F
effusion, 126
elastic deformation, viii, 32 fabrication, vii, 32, 125, 171, 201
electric charge, 198 failure, vii
electric field, ix, 95, 96, 99, 105, 106, 108, 109, 110, Fermi level, 77, 100, 101, 106
111, 112, 113, 114, 115, 116, 117, 118 fermions, 106
electrical conductivity, 166 film thickness, 163, 189, 193
electrical fields, 110 films, vii, ix, x, xi, xii, 69, 70, 71, 73, 74, 75, 76, 77,
electrical properties, viii, ix, 69, 70, 124, 126, 129, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 91,
130, 139, 169, 170 92, 123, 124, 125, 126, 128, 129, 130, 132, 133,
electricity, 160, 161 139, 140, 143, 144, 145, 146, 147, 148, 150, 152,
electrodeposition, xi, xii, 159, 160, 164, 165, 166, 153, 154, 156, 159, 160, 164, 165, 166, 167, 173,
167, 173, 174, 176, 177, 178, 181, 189, 191, 192, 177, 178, 179, 181, 182, 183, 185, 186, 187, 189,
195, 196, 197, 198, 200, 201, 202 191, 192, 193, 194, 195, 196, 197, 198, 200, 201,
electrodes, xi, 74, 159, 167 202
electroluminescence, x, 123, 137 flight, 103, 105, 106, 107
electrolyte, 166, 174, 186, 196, 197, 198 fluctuations, ix, 95, 106, 114, 115, 144, 154, 156,
electro-migration, vii 186
electron density distribution, 98 fluid, 24, 25
electron diffraction, 79 fossil, 160
electron microscopy, viii, 28, 31, 41 France, 205
electron state, 101, 106, 109, 111, 113, 114 free radicals, 73
212 Index
influence, vii, x, 11, 13, 35, 37, 71, 73, 74, 77, 78,
G 104, 111, 123, 124, 130, 139, 167, 169, 173, 178,
181, 186, 187
gallium, 2, 13, 16, 17, 18, 19, 20, 21, 22, 28, 99
initial state, 106, 107
generation, viii, 32, 33, 39, 42, 46, 48, 57, 58, 61, 63,
input, 101
64, 96, 104, 160, 161
instability, 13, 14
goals, 70
integrated circuits, 32, 35
gold, 3, 4, 7
integration, 101, 103, 148
grading, 173
intensity, 18, 19, 89, 92, 106, 129, 130, 131, 133,
grain boundaries, 71, 74, 199
134, 137, 138, 147, 148, 149, 150, 152, 153
grains, 73, 85, 190, 199
interaction, 48, 58, 87, 101, 102, 103, 109, 113, 115,
grids, 172
154, 171
groups, 48, 49, 50, 55, 63, 71, 125
interactions, 101
growth, vii, viii, x, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12,
interest, ix, 50, 69, 70, 71, 85, 96, 104
13, 14, 15, 16, 18, 21, 22, 23, 24, 25, 26, 27, 28,
interface, vii, 2, 5, 6, 8, 11, 13, 14, 15, 16, 17, 18, 22,
31, 32, 35, 36, 37, 38, 39, 40, 42, 43, 44, 45, 46,
24, 25, 26, 27, 97, 99, 100, 101, 173, 174, 186,
48, 50, 51, 52, 54, 56, 57, 58, 59, 60, 62, 63, 64,
187, 196, 197
71, 74, 79, 87, 91, 123, 124, 125, 126, 127, 128,
interference, 81
129, 130, 132, 133, 139, 174, 177
interpretation, 45
growth modes, 42, 43
interval, 26, 106
growth rate, 3, 11, 12, 13, 14, 15, 28, 32, 35, 37, 38,
inversion, 173
39, 40, 43, 44, 45, 48, 51, 54, 59, 60, 63, 71, 74
ionization, 114
growth temperature, x, 13, 21, 22, 123, 124, 125,
ions, 19, 73, 167, 173, 174, 179, 187, 197
126, 127, 128, 129, 130, 132, 139
IR spectra, 81
iron, 64
H irradiation, 12, 73
Italy, 143, 156
heat, 2, 14, 15, 28, 50, 51, 52, 53, 97, 110 iteration, 98
heat transfer, 15, 28 I-V curves, 136
heating, ix, 88, 89, 91, 95, 97, 100, 110, 118
height, 42, 72, 80, 84, 86, 88, 89, 91, 166
histogram, 108, 109 J
homogeneity, 169
Japan, 1, 28, 140, 161, 204, 205, 206
hot carrier effects, vii
hydrogen, ix, 36, 37, 50, 53, 63, 69, 73, 78, 84
hydrogenation, 73 K
kinetic equations, 96
I kinetics, vii
knots, 58
identification, 36, 41
knowledge, viii, 69, 71, 102
identity, 37
imitation, 50
impurities, vii, viii, 1, 3, 31, 33, 36, 37, 47, 48, 49, L
51, 52, 54, 57, 58, 59, 60, 62, 63, 64, 154
inclusion, 166 laser ablation, x, 143, 144
indicators, vii lattice parameters, 148
indices, 145 laws, 106
indium, xi, 2, 3, 8, 9, 10, 11, 12, 13, 14, 15, 18, 19, lead, 54, 73, 74, 84, 91, 101, 128, 129, 130, 132, 144
20, 22, 27, 28, 126, 159, 164, 166, 187, 200, 201 LED, 136
industry, 73 lens, 75
inelastic, 101 lifetime, 49, 97, 109, 117, 118
light conditions, 193
light emitting diode, x, 123
Index 213
light scattering, 85 motivation, 161

limitation, 124
linear dependence, 146
liquid phase, 2 N
Lithuania, 95
nanocrystals, 92
localization, x, 143, 154, 156
needs, 55, 164
low temperatures, 41, 42
neglect, 85
luminescence, x, 123, 126, 137, 140, 143, 144, 155
network, viii, 69, 73, 83
lying, 47
nitrides, 98, 102, 124
nitrogen, x, 4, 36, 63, 64, 99, 123, 125, 126, 133,
M 135, 136, 138, 139, 140
nitrogen gas, 4, 126
magnetic field, 2, 103, 104 noise, vii, ix, 90, 95, 96, 97, 106, 110, 114, 115, 116,
manufacturing, 164, 167 117, 118
market, 70 nonequilibrium, ix, 95, 96, 104, 109
mass, xi, 2, 13, 15, 19, 28, 98, 100, 102, 160, 179, nucleation, x, 35, 57, 58, 59, 60, 61, 62, 63, 123,
183, 194 127, 128
matrix, 49, 61, 63, 73, 79, 97, 101, 103 nuclei, 58
measurement, vii, 16, 17, 19, 48, 54, 82, 126, 132 numerical analysis, 25
melt, vii, viii, 1, 2, 3
melting, 2, 17, 33, 36, 39, 43, 64
melting temperature, 43 O
metal organic chemical vapor deposition, 126
observations, 15, 60
methanol, 126
oil, 161
microdefects, viii, 31, 32, 33, 34, 35, 36, 37, 38, 39,
optical properties, 126, 142, 144
40, 41, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53,
optimization, 71
54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65
organic solvents, 177
microelectronics, 32
orientation, 15, 129, 145, 146, 147
microscope, 12, 37, 44, 47, 145, 178, 198
output, vii, 1, 3, 145
microvoid, 60
oxides, 165
migration, vii, 127, 128
oxygen, viii, x, 31, 32, 33, 35, 36, 37, 39, 46, 47, 48,
minority, 165
49, 54, 57, 58, 59, 60, 61, 62, 63, 64, 81, 123, 124,
misfit dislocations, 129
125, 126, 128, 129, 130, 132, 135, 139
mobility, 73, 96, 125, 126, 137
MOCVD, 126
mode, x, 35, 50, 58, 63, 64, 103, 123, 127, 128, 139, P
151, 152
modeling, 145 parameter, viii, 31, 36, 58, 100, 109, 117, 145, 146,
models, 35, 36, 96, 144 154, 163, 168, 169
modules, 70 particles, 35, 58, 61, 108, 113
mole, 21 passivation, 171
molecular beam, vii, x, 123, 124, 140 pathways, 165
molecular beam epitaxy, 124 permittivity, 98, 102
molecules, 73, 125, 174 pH, 165, 177, 179, 181, 183, 193, 201
molybdenum, 126 phase diagram, 5, 13, 166, 167, 168, 169
momentum, 96, 101, 107, 109 phase transformation, ix, 69, 70, 73, 74, 75, 85, 86,
monitoring, vii, 74 87, 88, 89, 91
Monte Carlo method, 103, 106 phonons, ix, 85, 95, 96, 97, 101, 102, 103, 108, 109,
morphology, xi, 15, 78, 126, 127, 159, 160, 167, 110, 111, 112, 113, 115, 116, 118, 152
177, 187, 188, 189, 190, 193, 195, 198, 199, 200, phosphorus, 54
201, 202 photoemission, 74
motion, 3, 5, 97, 103, 104, 105, 106, 109, 114, 115, photoluminescence, x, 123, 126, 134, 143
117 photovoltaic cells, 2, 165
214 Index
physical properties, 71, 167, 168 118, 127, 129, 131, 135, 144, 155, 166, 179, 185,
physics, vii, 32, 168, 170 194, 201
PL spectrum, 130, 138 reactant, 70, 71, 74, 75, 83, 84
Planck constant, 102 reaction mechanism, 166
plants, 161 reasoning, 112
plasma, x, 73, 123, 124, 125, 126, 127, 132, 140, 151 recombination, viii, 31, 36, 41, 42, 48, 49, 57, 62, 64,
PM, 75 125, 131, 132, 137, 154, 156, 173
point defects, viii, 31, 32, 35, 36, 41, 42, 43, 44, 48, rectification, 166, 193
51, 54, 56, 57, 58, 59, 60, 61, 62, 64, 171 redistribution, 52, 100
Poisson equation, 97, 98, 100 redshift, 138
polarity, 124, 166 reduction, 72, 112, 115, 173, 174, 185
polarization, 98, 99, 102 reflection, x, 41, 143, 147, 148, 156
pollution, 160 refraction index, 84
poor, xi, 8, 73, 129, 159, 167, 170, 173, 177, 187, refractive index, 165
191, 197 relationship, xi, 8, 48, 74, 138, 145, 160, 167, 174,
population, 100, 104, 105, 112, 113 179, 181, 194
power, vii, ix, 1, 3, 73, 75, 85, 86, 87, 88, 89, 90, 92, relationships, 179
95, 96, 97, 110, 115, 116, 117, 118, 145, 160, 161, relaxation, 54, 56, 96, 100, 109, 112
193 relaxation process, 54
precipitation, 22, 39, 46, 58, 165 relaxation times, 112
prediction, vii relevance, 44, 45
preparation, vii, ix, xi, 69, 70, 71, 73, 124, 159, 164, reliability, vii
165, 167, 171, 178, 193, 200, 202 resistance, 50, 114, 115, 124, 136
pressure, ix, 24, 69, 70, 71, 72, 74, 75, 89, 126, 145, resolution, 45, 46, 75, 145, 153
167, 173, 192 resources, 160, 161
prices, 70 returns, 109
principle, 106, 107, 111, 113, 124, 148, 156, 171 room temperature, x, 74, 75, 76, 77, 88, 89, 91, 92,
probability, 101, 102, 103, 105, 107, 111, 112, 125, 101, 108, 109, 118, 123, 125, 129, 138, 153, 154,
154 156, 165, 169, 170, 173, 198
probe, 129, 155, 178 root-mean-square, 127
production, 32, 50, 144, 161, 163, 166 roughness, 78, 80, 84, 101, 127, 128, 139, 189, 190
program, 145, 149, 157 Russia, 65, 66
proportionality, 179
pulse, 3, 11, 12, 13, 28, 110, 111, 126, 144
S
Q sample, xi, 17, 50, 76, 77, 79, 80, 81, 82, 83, 86, 87,
88, 89, 91, 92, 126, 130, 132, 133, 134, 135, 138,
quantum confinement, x, 123, 139 145, 147, 148, 150, 152, 159, 178, 185, 188, 189,
quantum well, 97, 98, 100, 124 192, 195, 196
quartz, 3, 7, 14, 23, 24, 177 sampling, 89, 91, 92
satellite, 104
scattering, ix, 35, 59, 72, 74, 82, 83, 85, 86, 87, 88,
R 90, 91, 95, 97, 100, 101, 102, 103, 104, 105, 106,
107, 109, 111, 112, 113, 117, 118, 151, 152
radiation, 37, 44, 49, 145
secondary radiation, 49
radio, 74
sedimentation, 50
radius, 25, 58
seed, vii, 1, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 13, 15, 16,
Raman spectra, x, 74, 75, 85, 86, 88, 89, 92, 143,
17, 18, 19, 21, 22, 23, 24, 25, 26, 27, 28
145, 151
seeding, 4, 7
Raman spectroscopy, ix, 69, 150
segregation, 2, 5, 17, 21, 27, 28
range, viii, 1, 2, 3, 11, 22, 23, 39, 59, 60, 72, 74, 75,
selecting, 63
76, 79, 84, 85, 92, 96, 100, 104, 110, 112, 115,
selenium, 166
Index 215
self, viii, ix, 22, 31, 32, 36, 41, 42, 43, 48, 49, 50, 57, substrates, x, 50, 123, 124, 126, 128, 143, 144, 145,
58, 62, 95, 97, 98, 99, 100, 101, 102, 105, 107, 146, 147, 165, 166, 177, 186, 188, 197, 201
109, 110, 118, 124 sulfur, 153, 154
self-consistency, 100 sulphur, 144, 145, 153
semiconductor, vii, xi, 60, 61, 62, 65, 77, 96, 124, Sun, 120, 121, 140, 141
125, 142, 144, 155, 156, 159, 160, 161, 163, 164, supply, 8, 124, 185
165, 167, 173, 186, 200, 201 suppression, 39
semiconductors, vii, 1, 2, 66, 98, 103, 113, 124, 143, surface structure, 74
144, 152, 154, 163, 164, 168 surplus, 57
sensitivity, 19, 153 Switzerland, 204
sensors, 71 symbols, 102, 113
separation, 2, 72, 135 symmetry, x, 104, 108, 143, 144, 156
series, 37, 54, 66, 75, 92 systems, 60, 104, 160, 161
shape, 8, 13, 15, 23, 33, 41, 44, 55, 56, 60, 62
sharing, 115, 117, 118
shear, 63 T
sign, viii, 31, 35, 36, 37, 41, 44, 45, 49, 57, 61
technical assistance, 156
silane, ix, 69, 71, 73, 74, 75, 78, 83, 86
technology, vii, ix, 32, 69, 70, 73, 125, 139
silicon, viii, ix, 31, 32, 33, 35, 36, 37, 38, 40, 42, 43,
tellurium, 11
46, 48, 49, 50, 56, 57, 58, 59, 60, 61, 62, 63, 64,
temperature, viii, ix, x, 2, 3, 4, 5, 6, 7, 9, 11, 13, 14,
65, 66, 69, 70, 71, 72, 73, 74, 75, 77, 83, 84, 85,
15, 16, 18, 22, 23, 24, 25, 26, 27, 28, 31, 32, 33,
89, 91, 92, 94, 161, 163
35, 39, 40, 41, 42, 49, 50, 51, 53, 56, 57, 59, 60,
simulation, vii, ix, 3, 95, 96, 100, 101, 104, 105, 106,
62, 63, 64, 69, 70, 71, 74, 75, 82, 84, 85, 86, 88,
107, 108, 109, 110, 111, 112, 113, 115, 116, 117,
89, 91, 92, 95, 97, 98, 99, 100, 101, 106, 110, 111,
118, 173
113, 114, 115, 116, 117, 118, 123, 124, 125, 126,
single crystals, vii, viii, 1, 2, 5, 27, 28, 31, 32, 33, 36,
127, 128, 129, 130, 131, 136, 139, 144, 145, 150,
38, 40, 48, 59, 60, 61, 63, 64, 92, 154
153, 154, 155, 156, 167, 169, 174, 179, 183, 192,
SiO2, viii, 31, 39, 46, 61, 62
194, 195
sites, 2, 56, 73, 128, 130, 144, 150, 168
temperature annealing, 56
solar cells, ix, xi, 69, 70, 71, 74, 94, 159, 161, 162,
temperature dependence, 49, 131, 154, 155
163, 164, 165, 166, 167, 168, 171, 173, 177, 178,
tension, ix, 69
187, 193, 196, 197, 200, 201, 202, 206
Thailand, 202, 203, 207
solid solutions, 59, 60, 144
theory, 42, 43, 48, 125
solubility, 61, 124
thermal activation, 131
Spain, 71, 93, 94
thermal activation energy, 131
species, xi, 73, 136, 159, 178, 183, 184, 185, 194
thermal evaporation, 126
spectroscopy, 19, 74, 144
thermal oxidation, 35
spectrum, x, 45, 82, 86, 87, 88, 91, 123, 124, 135,
thermal stability, 124
138, 140, 152, 153, 154, 156
thermal treatment, 45, 53, 54, 56
speed, 3, 5, 15
thin films, ix, x, xi, xii, 69, 70, 71, 72, 73, 74, 75, 76,
spin, 106, 107
85, 86, 92, 123, 124, 125, 127, 128, 129, 130, 131,
stability, 165, 171
132, 134, 136, 138, 139, 143, 155, 159, 160, 164,
stages, 37, 39, 51, 127
165, 166, 167, 169, 173, 178, 192, 193, 196, 197,
statistics, 98, 104, 111
199, 200, 201, 202
steel, 86, 87, 91
time, 9, 22, 36, 42, 49, 50, 51, 55, 60, 61, 63, 86, 89,
stoichiometry, 171, 187
90, 91, 92, 97, 100, 101, 102, 103, 104, 105, 106,
strain, 44, 45, 46, 49, 54, 58, 62, 63, 100
107, 108, 109, 110, 112, 115, 117, 124, 161, 163,
streams, 49
164, 165, 177, 189, 192, 197, 198
strength, 102, 105, 126, 154
tin, 201
stress, vii, 35, 71, 72, 74, 89, 91, 92
tin oxide, 201
structural defects, 101, 124, 144, 154, 156
tissue, viii, 69, 71, 86
subsidy, 161
titanium, 177
substitution, 57, 125, 144
total energy, 117, 156
216 Index
trajectory, 104 variable(s), 48, 70, 73, 86

transformation, viii, ix, 31, 32, 36, 38, 39, 41, 50, 51, variation, 186, 187
53, 54, 56, 57, 58, 59, 62, 64, 69, 70, 71, 72, 89, vector, 25, 45, 46, 47, 50, 108, 109, 111
90, 91, 92 velocity, 3, 24, 25, 28, 96, 97, 102, 104, 105, 106,
transformation processes, 70, 72 108, 109, 110, 111, 112, 113, 114, 115, 118, 176
transformations, 88, 91 vibration, vii, 1, 2
transistor, 96 viscosity, 24
transition, ix, 58, 69, 84, 85, 91, 102, 103, 106, 107,
109, 127, 128, 131, 135, 136, 139, 144, 161, 162,
170, 171, 172 W
transition rate, 102, 103
water, 126, 174
transitions, 85, 107, 135
wave number, 72
transmission, viii, 31, 34, 37, 49, 79
wave vector, 102, 103, 104, 105, 107, 109
transmission electron microscopy, 37
weak interaction, 113
transmittance spectra, 75, 84
web, 71
transport, ix, 14, 77, 95, 96, 100, 101, 103, 104, 105,
weight ratio, 19
108, 165, 167, 173
wide band gap, 96
trend, 13
work, 100, 118, 125, 133, 135, 136, 137, 140, 144,
trial, 98
156, 177, 196, 201
tunneling, 165
workstation, 145
U
X
Ukraine, 31
XPS, 126, 178, 198, 199
ultrasonic vibrations, vii, 1, 2, 3
X-ray diffraction (XRD), x, 8, 9, 10, 11, 18, 27, 28,
uniform, viii, 2, 3, 8, 13, 27, 28, 32, 33, 35, 38, 39,
74, 82, 91, 129, 133, 143, 144, 145, 146, 148, 149,
44, 45, 47, 58, 105, 189
150, 155, 156, 160, 169, 178, 187, 189, 192, 195
UV, x, 123, 124, 126, 128, 129, 130, 132, 134, 139
UV absorption spectra, 126
Y
V yield, 98, 118, 161, 162, 163
vacancies, viii, 31, 36, 40, 41, 42, 43, 48, 49, 50, 57,
58, 60, 62, 63, 64, 124, 129, 144, 170, 171 Z
vacuum, xi, 45, 48, 51, 53, 54, 126, 159, 164, 165,
166, 197 zinc, 126, 128
valence, 137, 171, 172, 173 ZnO, x, 123, 124, 125, 126, 127, 128, 129, 130, 131,
validity, 35, 96 132, 133, 134, 135, 136, 137, 138, 139, 140, 141,
values, 16, 19, 27, 38, 39, 40, 50, 70, 72, 75, 76, 77, 171, 178
78, 83, 84, 92, 96, 99, 101, 109, 113, 131, 132,
144, 145, 146, 179, 184, 185, 198
vapor, 165, 174, 177, 192, 197

Semiconductors A

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Semiconductors A

Uploaded by

Copyright:

Available Formats

NEW RESEARCH

Nova Science Publishers, Inc.

NOTICE TO THE READER

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Published by Nova Science Publishers, Inc.  New York

Chapter 7 describes preparation of a ternary chalcopyrite semiconductor, copper indium

CRYSTAL GROWTH OF TERNARY AND QUATERNARY

Yasuhiro Hayakawa*, M.Haris, Masashi Kumagawa

2 Growth of InSb1-XBiX Bulk Crystals

2.1 Rotationary Bridgman Method (RBM)

Fig. 3 InSb-InBi phase diagram.

2.2 Continuous Rotationary Feeding Bridgman Method (CFRBM)

3 Growth of InXGa1XAs Ternary Bulk Crystals

3.1 Direct Growth Method

3.2 Multi-step Growth Method

Fig. 11 Temperature-time profile of multi-step growth.

Fig. 12 (a) Feeding process and (b) growth processes.

3.3 Effect of Ampoule- moving Rate

3.4 Effect of Ampoule Rotation Rate

4 Growth and Analysis of In1-xGaxSb1-yBiy Quaternary Crystals

Fig. 20 shows a roughly polished surface of an In1-xGaxSb1-yBiy crystal. The initial

4.2 Composition Profiles

4.3 Lattice Constant

4.4 Formation of Domains in the InSb Seed Crystal

4.5 Growth of InSb1-yBiy and In1-xGaxSb Ternary Crystals

5 Numerical Simulation of Effect of Ampoule Rotation

u = 0, v = 0, w = (rpm), T = 660(qC) (6)

At the quartz glass wall:

u = 0, v = 0, w =(rpm), T(z) = - az2 + 660 + (TH-TL) (7)

u = 0, v = 0, w = (rpm), T = 660(qC) (8)

5.2 Analysis Results

Fig. 36 shows T normalized by T0rpm as a function of rotation rate. Here, T and

(4) In1-xGaxSb1-yBiy (0<xd0.21,0<yd0.005) single crystals more than 10 mm in length

[34] T.Ozawa, Y.Hayakawa, K. Balakrishnan and M.Kumagawa, J.Crystal Growth 237-239

FORMATION OF GROWN-IN MICRODEFECTS

V.I. Talanin*, I.E. Talanin

octahedral vacancy microvoids and interstitial dislocation loops accordingly. It is

Keywords: grown-in microdefects, mechanism of formation, intrinsic point defects, oxygen,

microdefects size increased up to a certain limit (about 40 Pm at V = 1 mm/min), dislocations

2.2 Quenched Crystals Study

dT/dL = 10 + (L - 16)2exp(-61.2V 0.28), (1)

Table 1. The temperature of grown-in microdefects formation

Fig.3. Schematic distribution grown-in microdefects in FZ-Si crystals after quenching: a) V = 2

Experiments in crystal quenching demonstrate that the (I+V)-microdefects are first to

which is established regardless the generating technique of the concentration ratio of

where : is a volume cell; DS is a self-diffusion coefficient; r0 is an effective recombination

'G 'H  T'S , (4)

2.3 Physical Nature of the Grown-in Microdefects

D-microdefects constitute clusters of interstitial-type point defects with size of 4 to

Evaluation of quantitative ratio for vacancy to interstitial-type defects gave us a value

Fig.5. D()-microdefects and (I+V)-microdefects in Z-Si (V = 1.82.8 mm/min): a) selective

Fig.6. D-microdefects in FZ-Si (V = 6 mm/min), TEM (direct solution of crystalline lattice): a)

microdefects and B-microdefects should be referred to as the primary grown-in microdefects.

Fig. 7. TEM imaged A-microdefects and B-microdefects in FZ-Si: a) A-microdefects, g (022) ,

g ( 2 20) , dark field, s | 0 ; d) B-microdefects, g ( 202) , dark field, s ! 0 .

Our experimental results in studying the physical nature of grown-in microdefects

2.4 Interaction of Grown-in Microdefects with Radiation Defects

On efficiency of interaction of vacancies or self-interstitials entered by J-rays with

Table 2 The crystal performances

Type of Growth rate, Size, Concentration, W, n0,

Table 3 The value of factors W and rates of introduction K and K [61].

Group Type of defect W, cm2s-1 K, cm-1 K, cm-1

The modification of W can be stipulated only by decreasing K, as a nature and parameters

Published by Nova Science Publishers, Inc. New York

'G 'H T'S , (4)