(Norbert H. Nickel, Evgenii Terukov) Zinc Oxide

Zinc Oxide -- A Material for
Micro- and Optoelectronic Applications
NATO Science Series

A Series presenting the results of scientific meetings supported under the NATO Science
Programme.
The Series is published by IOS Press, Amsterdam, and Springer (formerly Kluwer Academic
Publishers) in conjunction with the NATO Public Diplomacy Division.
Sub-Series
I.
II.
III.
IV.
Life and Behavioural Sciences

Mathematics, Physics and Chemistry
Computer and Systems Science
Earth and Environmental Sciences
IOS Press
Springer (formerly Kluwer Academic Publishers)
IOS Press
Springer (formerly Kluwer Academic Publishers)
The NATO Science Series continues the series of books published formerly as the NATO ASI Series.
The NATO Science Programme offers support for collaboration in civil science between scientists of
countries of the Euro-Atlantic Partnership Council. The types of scientific meeting generally supported
are Advanced Study Institutes and Advanced Research Workshops, and the NATO Science Series
collects together the results of these meetings. The meetings are co-organized by scientists from
,
NATO countries and scientists from NATO s Partner countries countries of the CIS and Central and
Eastern Europe.
Advanced Study Institutes are high-level tutorial courses offering in-depth study of latest advances
in a field.
Advanced Research Workshops are expert meetings aimed at critical assessment of a field, and
identification of directions for future action.
As a consequence of the restructuring of the NATO Science Programme in 1999, the NATO Science
Series was re-organized to the four sub-series noted above. Please consult the following web sites for
information on previous volumes published in the Series.
http://www.nato.int/science
http://www.springeronline.com
http://www.iospress.nl
Series II: Mathematics, Physics and Chemistry Vol. 194
Zinc Oxide -- A Material for

Micro- and Optoelectronic
Applications
edited by
Norbert H. Nickel
Hahn-Meitner-Institut Berlin,
Germany
and
Evgenii Terukov
A.F. Ioffe Physico-Technical Institute,
St. Petersburg, Russia
Published in cooperation with NATO Public Diplomacy Division
Proceedings of the NATO Advanced Research Workshop on

Zinc Oxide as a Material for Micro- and Optoelectronic Applications
St. Petersburg, Russia
23 25 June 2004
A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN-10
ISBN-13
ISBN-10
ISBN-10
ISBN-13
ISBN-13
1-4020-3474-1 (PB)
978-1-4020-3474-9 (PB)
1-4020-3473-3 (HB)
1-4020-3475-X (e-book)
978-1-4020-3473-2 (HB)
978-1-4020-3475-6 (e-book)
Published by Springer,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.
www.springeronline.com
Printed on acid-free paper
All Rights Reserved

2005 Springer
No part of this work may be reproduced, stored in a retrieval system, or transmitted in
any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception of any
material supplied specifically for the purpose of being entered and executed on a
computer system, for exclusive use by the purchaser of the work.
Printed in the Netherlands.
Contents
Contributing Authors
ix
Preface
xv
PART I: ZnO Bulk and Layer Growth
The Scope of Zinc Oxide Bulk Growth

ROBERT TRIBOULET, VICENTE MUNOZ-SANJOSE,
RAMON TENA-ZAERA, MARI CARMEN MARTINEZ-TOMAS,
AND SAD HASSANI
Growth Mechanism of ZnO Layers

A. KH. ABDUEV, A. SH. ASVAROV, A. K. AKHMEDOV,
I. K. KAMILOV, AND S. N. SULYANOV
Kinetics of High-Temperature Defect Formation in ZnO in the
Stream of Oxygen Radicals
M. B. KOTLYAREVSKY, I. V. ROGOZIN, AND
A. V. MARAKHOVSKY
15
25
PART II: Electrical, Optical, and Structural Properties
35
Electrical Properties of ZnO

DAVID C. LOOK, BRUCE B. CLAFLIN, GENE CANTWELL,
SEONG-JU PARK, AND GARY M. RENLUND
37
vi
Zinc oxide - a material for micro- and optoelectronic applications
Electrical Properties of ZnO Thinn Films and Single Crystals

M. GRUNDMANN, H. VON WENCKSTERN, R. PICKENHAIN,
S. WEINHOLD, B. CHENGNUI, AND O. BREITENSTEIN
47
Structure, Morphology, and Photoluminescence of ZnO Films

V. A. KARPINA, V. D. KHRANOVSKYY, V. I. LAZORENKO,
G. V. LASHKAREV, I. V. BLONSKY, AND V. A. BATURIN
59
Optics and Spectroscopy of Point Defects in ZnO

VLADIMIR NIKITENKO
69
Whispering Gallery Modes in Hexagonal Zinc Oxide Micro- and

Nanocrystals
THOMAS NOBIS, EVGENI M. KAIDASHEV, ANDREAS RAHM,
MICHAEL LORENZ, AND MARIUS GRUNDMANN
Properties of Dislocations in Epitaxial ZnO Layers Analyzed by
Transmission Electron Microscopy
E. MLLER, D. LIVINOV, D. GERTHSEN, C. KIRCHNER,
A. WAAG, N. OLEYNIK, A. DADGAR, AND A. KROST
83
99
PART III: Role of Hydrogen
113
Muon Spin Rotation Measurements on Zinc Oxide

E. A. DAVIS
115
Hydrogen Donors in Zinc Oxide

M. D. MCCLUSKEY AND S. J. JOKELA
125
Hydrogen-Related Defects in Zno Studied by IR Absorption

Spectroscopy
E.V. LAVROV, F. BRRNERT, AND J. WEBER
133
Influence of the Hydrogen Concentration on H Bonding in Zinc Oxide 145

N. H. NICKEL
PART IV: Fundamental Properties
Valence Band Ordering and Magneto-Optical Properties of Free
and Bound Excitons in ZnO
A.V. RODINA, M. STRASSBURG, M. DWORZAK,
U. HABOECK, A. HOFFMANN, H. R. ALVES, A. ZEUNER,
D. M. HOFMANN, AND B. K. MEYER
157
159
vii
Fundamental Optical Spectra and Electronic Structure of Zno Crystals 171

V. VAL. SOBOLEV AND V. V. SOBOLEV
Photo-Induced Localized Lattice Vibrations in ZnO Doped with 3d
Transition Metal Impurities
ALEXEY KISLOV
183
PART V: Device Applications
195
ZnO Window Layers for Solar Cells

WALTHER FUHS
197
ZnO/AlGaN Ultraviolet Light Emitting Diodes

E. V. KALININA, A. E. CHERENKOV, G. A. ONUSHKIN,
YA. I. ALIVOV, D. C. LOOK, B. M. ATAEV, A. K. OMAEV,
AND C. M. CHUKICHEV
211
ZnO Transparent Thin-Film Transistor Device Physics

J. F. WAGER
217
Zinc Oxide Thin-Film Transistors

E. FORTUNATO, P. BARQUINHA, A. PIMENTEL,
A. GONALVES, A. MARQUES, L. PEREIRA, AND R. MARTINS
225
Index
239
Contributing Authors
A. Kh. Abduev
Russian Academy of Sciences, Makhachkala, Russia
A. K. Akhmedov
Ya. I. Alivov
Institute of Microel. Technology, Chernogolovka, Russia
H. R. Alves
Justus-Liebig-Universitt Giessen, Giessen, Germany
A. Sh.Asvarov
B. M. Ataev
Daghestan Scientific Centre of RAS, Makhachkala, Russia
P. Barquinha
New University of Lisbon, Caparica, Portugal
V. A. Baturin
National Academy of Sciences of Ukraine, Kyiv, Ukraine
I. V. Blonsky
ix

F. Brrnert
Technical University Dresden, Dresden, Germany
O. Breitenstein
Max-Planck-Institut fr Mikrostrukturphysik, Halle / Saale, Germany
Gene Cantwell
ZN Technology, Brea, California, USA
B. Chengnui
Universitt Leipzig, Leipzig, Germany
A. E. Cherenkov
A. F. Ioffe Physico-Technical Institute, St. Petersburg, Russia
C. M. Chukichev
Moscow State University, Moscow, Russia
Bruce B. Claflin
Wright State University, Dayton, Ohio, USA
A. Dadgar
Otto von Guericke Universitt, Magdeburg, Germany
E. A. Davis
University of Cambridge, Cambridge, United Kingdom
M. Dworzak
Technical University of Berlin, Berlin, Germany
E. Fortunato
Walther Fuhs
Hahn-Meitner-Institut Berlin, Berlin, Germany
D. Gerthsen
Universitt Karlsruhe (TH), Karlsruhe, Germany
A. Gonalves

Marius Grundmann
U. Haboeck
Sad Hassani
CNRS, Meudon, France
A. Hoffmann
D. M. Hofmann
S. J. Jokela
Washington State University, Pullman, Washington, USA
Evgeni M. Kaidashev
E. V. Kalinina
I. K. Kamilov
V. A. Karpina
V. D. Khranovskyy
C. Kirchner
Universitt Ulm, Ulm, Germany
Alexey Kislov
Ural State Technical University, Ekaterinburg, Russia
M. B. Kotlyarevsky
Academy of Management and Information Techn., Berdyansk, Ukraine
xi
xii

A. Krost
G. V. Lashkarev
E. V. Lavrov
V. I. Lazorenko
D. Livinov
David C. Look
Wright State University, Dayton, Ohio, USA
Michael Lorenz
A. V. Marakhovsky
Berdyansk Teachers Training University, Berdyansk, Ukraine
A. Marques
R. Martins
Mari Carmen Martinez-Tomas
Universitat de Valncia, Burjassot, Spain
M. D. McCluskey
Washington State University, Pullman, Washington, USA
B. K. Meyer
E. Mller

Vicente Munoz-Sanjos
Norbert H. Nickel
Hahn-Meitner-Institut Berlin, Berlin, Germany
Vladimir Nikitenko
Moscow State Transport University, Moscow, Russia
Thomas Nobis
N. Oleynik
A. K. Omaev
Daghestan Scientific Centre of RAS, Makhachkala, Russia
G. A. Onushkin
Seong-Ju Park
K-JIST, Kwangju, Korea
L. Pereira
R. Pickenhain
A. Pimentel
Andreas Rahm
Gary M. Renlund
ON International, Salt Lake City, Utah, USA
A.V. Rodina
xiii
xiv

I. V. Rogozin
Berdyansk Teachers Training University, Berdyansk, Ukraine
M. Strassburg
Ramon Tena-Zaera
Robert Triboulet
CNRS, Meudon, France
V. Val. Sobolev
Udmurt State University, Izhevsk, Russia
V. V. Sobolev
Udmurt State University, Izhevsk, Russia
S. N. Sulyanov
Russian Academy of Sciences, Moscow, Russia
A. Waag
Universitt Braunschweig, Braunschweig, Germany
J. F. Wager
Oregon State University, Corvallis, Oregon, USA
H. von Wenckstern
J. Weber
S. Weinhold
A. Zeuner
Preface
Recently, a significant effort has been devoted to the investigation of

ZnO as a suitable semiconductor for UV light-emitting diodes, lasers, and
detectors and hetero-substrates for GaN. Research is driven not only by the
technological requirements of state-of-the-art applications but also by the
lack of a fundamental understanding of growth processes, the role of
intrinsic defects and dopants, and the properties of hydrogen. The NATO
Advanced Research Workshop on Zinc oxide as a material for micro- and
optoelectronic applications, held from June 23 to June 25 2004 in St.
Petersburg, Russia, was organized accordingly and started with the growth
of ZnO. A variety of growth methods for bulk and layer growth were
discussed. These techniques comprised growth methods such as closed space
vapor transport (CSVT), metal-organic chemical vapor deposition, reactive
ion sputtering, and pulsed laser deposition. From a structural point of view
using these growth techniques ZnO can be fabricated ranging from single
crystalline bulk material to polycrystalline ZnO and nanowhiskers. A major
aspect of the ZnO growth is doping. n-type doping is relatively easy to
accomplish with elements such al Al or Ga. At room temperature single
crystal ZnO exhibits a resistivity of about 0.3 :-cm, an electron mobility of
225 cm2/Vs, and a carrier concentration of |1017 cm-3. In n-type ZnO two
shallow donors are observable with activation energies of 30 40 meV and
60 70 meV. It has been established that the shallower donor is interstitial
hydrogen while it is believed that the deeper donor is a group IV element
such as Al. On the other hand, reliable p-type doping is very difficult to
achieve. Usually group I and group V elements are used as acceptors. So far
the most common acceptor has been nitrogen because it should readily
substitute for O and not distort the lattice significantly. A major problem is
xv
xvi
the fact that most growth techniques either use hydrogen as a transport gas or
H containing precursor gases such as NH3. Therefore, in order to achieve ptype conductivity ZnO has to be subjected to a post-growth thermal treatment
at temperatures above 650 C. The purpose is two-fold: (i) interstitial
hydrogen donors are eliminated and (ii) H passivated nitrogen acceptors are
activated. The properties of H in ZnO expand well beyond the formation of
donors and the neutralization of impurities and deep defects. The presence of
hidden H has been suggested, which might be molecular H2 that can be
activated to form IR active O-H centers. Moreover, in sputtered ZnO the
total H concentration can be as high as 1021 cm-3. In addition, with increasing
H concentration the average H binding energy increases; H becomes more
strongly bound. Besides N, phosphorous and arsenic have been successfully
used to fabricate p-type ZnO. Since the fabrication of p-type ZnO becomes
more and more reliable the fabrication of p-n homojunction LEDs is not far
off.
Recently, a new and rapidly developing area emerged, namely the field of
transparent electronics based on ZnO transparent thin-film transistors. These
devices exhibit an on-off ratio of 6 orders of magnitude, a channel mobility
of 1 3 cm2/Vs and can be fabricated at room temperature. Hence, already in
the near future invisible large area electronics based on ZnO will
revolutionize the electronics market. Although first small steps towards ZnO
based optoelectronic applications are currently being made there is still a vast
lack of fundamental knowledge regarding doping mechanisms, influence of
impurities, formation of point and line defects, and their interactions with
hydrogen. Some controversies and questions that arose during the workshop
are based upon the fact that a materials basis for state-of-the-arte ZnO has
not been established yet. Depending on the fabrication process the materials
qualities such as impurities, their concentration, H content, number of point
defects and grain boundaries can vary significantly.
This book reviews the latest experimental and theoretical studies in the
field of ZnO. The chapters reflect the topics that were discussed during the
NATO advanced research workshop, which brought together researchers
from NATO countries, Russia, Ukraine, and Kazakhstan to encourage cross
fertilization and wider dissemination of the advances in this important
research field.
We would like to thank NATO for funding the Advanced Research
Workshop. Above all, we would like to thank all participants for their
interest in this workshop and active work in the vibrant field of ZnO.
Norbert H. Nickel
Evgenii Terukov
PART I: ZnO BULK AND LAYER GROWTH
Chapter 1
THE SCOPE OF ZINC OXIDE BULK GROWTH
Robert Triboulet1, Vicente Munoz-Sanjos2, Ramon Tena-Zaera2, Mari
Carmen Martinez-Tomas2 and Sad Hassani1
1
CNRS, Laboratoire de Physique des Solides et de Cristallogense, 1 Place Aristide Briand,

F-92195 Meudon Cedex, France ; 2Dept. F
Fsica Aplicada i Electromagnetisme, Universitat de
Valncia, C/ Dr. Moliner 50, 46100 Burjassot, Spain
Abstract:
The techniques classically used for the growth of bulk ZnO crystals, meltgrowth, vapour growth, solution-growth and hydrothermal growth, are
reviewed. In the case of vapor growth, numerical simulations suggest that
such species as H2 + H2O, CH4, Zn, C or Fe could act as sublimation
activators. The electrical and structural properties of crystals obtained by these
various techniques are compared.
Key words:
ZnO synthesis, melt-growth, vapor-growth, solution-growth, hydrothermalgrowth, electrical and structural properties
1.
INTRODUCTION
ZnO can be considered as an 'old' semiconductor which has been

compelling research attention for a long time because of its applications in
many scientific and industrials areas such as piezoelectric transducers,
optical waveguides, acoustooptic media, conductive gas sensors, transparent
conductive electrodes, varistors.1 It has now received increasing attention
and recognized as a promising candidate for applications related to its
optoelectronic possibilities in the UV range. Its piezoelectric properties could
also allow developing SAW filters to be integrated in future analog circuits
for portable electronic for which there is a strong need. Furthermore, ZnO
transparent thin-film transistors (TTFTs) are a recent and important
development in the emerging field of transparent electronics. These potential
applications have boosted research related to the growth of high quality ZnO
thin films by a lot of different techniques that need high quality substrates.
3
N.H. Nickel and E. Terukov (eds.),
Zinc Oxide A Material for Micro- and Optoelectronic Applications, 314.
2005 Springer. Printed in the Netherlands.
R. Triboulet et al.
That is why research related to the ZnO bulk growth has received a
considerable interest during these last years.
As a result of the small ionic radii of its constituting elements, mainly the
oxygen one, and of its subsequently very short and energetic chemical bond,
ZnO has a very high melting point of about 1900C. Furthermore, because
of the high electronegativity of its oxygen and zinc and because of its usual
off-stoichiometry, ZnO shows an extremely high reactivity with any
surrounding material at high temperature. This intrinsic property will be
shown to affect strongly its growth, not only its melt-growth but also its
vapour-growth.
In this chapter, the various techniques classically used for the growth
of bulk ZnO crystals will be reviewed.
2.
ZINC OXIDE SYNTHESIS
It is worth noting that ZnO occupies already an enviable place in the

industrial market. Tens of thousands tons of ZnO powder are industrially
produced each year which are used in the rubber industry as vulcanisation
activator (~36 %), in the industry of ceramics as a flux (~26 %), in the
chemical industry (desulphuration of gases, fabrication of stearates,
phosphates. etc) (~20 %), as trace elements in the animal food (~12 %), in
the paints (~3 %; 50 % in 1961!). The last ~3 % are used for different
applications, in electronics (ferrites, varistors), ends of matches,
pharmaceutic industry (fungicidal properties of ZnO for skin-problems, trace
elements ...etc.).
ZnO powder is produced from the combustion of vapors coming from the
distillation of metallic zinc according to the so-called French process or dry
process. Using vapors coming from fractionated distillation, extra-pure oxide
powders are produced, as illustrated from the chemical analysis data of two
kinds of powder produced by two different companies:
pharmacy B grade, produced by Union Minire: Pb<20 ppm,
Cd<10ppm, Fe<5 ppm, Cu<1 ppm.
colloidal grade, produced by Silar SA: Pb<20 ppm, Cd<30 ppm, Fe<5
ppm, Cu<2 ppm, Mn<2 ppm.
The grain size can change according to the variety considered.
The scope of zinc oxide growth
3.
ZINC OXIDE MELT-GROWTH
In spite of its very high melting-point (~ 1900C) and very high

reactivity with any surrounding material at high temperature, large ZnO bulk
are grown from the melt at Cermet, Inc., Atlanta, GA, using a patented
pressurized melt growth process.2,3 The technique uses a high-pressure
induction melting apparatus, wherein the ZnO melt is contained in a watercooled crucible. The whole melting and containment process is carried out in
under a controlled oxygen overpressure close to 50 atm preventing the
decomposition of the compound into atomic species and establishing a
thermodynamic equilibrium between the liquid ZnO and the oxygen
pressure, thereby preventing reduction of the lattice. The molten charge is
somewhat contained in a ZnO liner, because the cooled material in contact
with the cold-wall crucible constitutes a solid thermal barrier. This prevents
the molten material from coming into direct contact with the cooled surface
and solves the problem of the extreme reactivity of ZnO with any crucible
material at high temperature. This allows furthermore to get rid from any
contamination from the crucible material. 5.5-inch-diameter, kilogram sized
ingots have been reported to be grown.3 But the drawback of the technique
lies in the concave liquid/solid growth interface, as a result of the cold-wall
crucible, which makes it difficult to get a large grain size.
4.
ZINC OXIDE VAPOR GROWTH
Because of the high melting-point of ZnO, there have been many

attempts of growth of ZnO single crystals in the vapor phase.
Small crystals have been obtained by chemical vapour transport (CVT) in
closed tubes using such chemical transport agents as Zn, ZnCl2, ammonium
halides (NH4X), HCl, Cl2, NH3, NH4Cl, HgCl2, H2, Br2, at source
temperatures ranging from 800 to 1150C and 'T
T from 20 to 200C.4-8
Crystals of small size have been grown as well in open tube systems,
either by oxidation of ZnI2,9 ZnS, ZnSe,10 ZnBr211 and Zn12-14 or by
hydrolysis of ZnF2,15 ZnCl2,16,17 ZnI2.18 The oxidation or hydrolysis character
of the reaction can depend on the temperature range used for a same source.
The temperature of the growth region ranges generally from 900 to 1350C.
This field of research on the vapor-growth of ZnO has been reactivated
from ~ 1998 by the need of ZnO substrates for growing epitaxial structures
in view of photonic applications.
Large diameter (2-inch diameter) boules have been reported to be grown
at Eagle Picher at 1000-1200 C by seeded chemical vapor transport (SCVT)
in a nearly closed horizontal tube using H2 as a carrier gas and a small
R. Triboulet et al.
amount of water to maintain the proper stoichiometry:19 The crystals are

seeded at full diameter; the source/substrate distance is of about 3 inches; the
stoichiometry is controlled from the presence of a residual water pressure;
the growth rates are of about 40 Pm per hour. 2-inch diameter crystals of
about 1-cm-thickness are produced after 150-175 h; the etch pit (dislocation)
density is of about 104 cm-2; the X-ray double diffraction rocking curve
FWHM is of about 40 arcsec, with some scattering indicating the presence of
residual strains. GDMS analysis reveals the purest material ever done by
Eagle Picher. The crystals are of n-type with n ~ 8u1016 cm-3 and P ~150-350
cm2/Vs @ RT. The homoepitaxial growth of ZnO on such substrates has
been achieved on <0001> Zn faces chemo-mechanically polished. The future
goal is the growth of larger diameter substrates, up to 3 inch.
Centimeter-size single crystals with rocking curve FWHM ~ 28 arcsec
have been obtained using C or CH4 as a transporting chemical agent.20, 21
ZnO, ZnO:Mn and Zn1-xMnxO (x<0.05) crystals were grown by CVT as
well using hydrogen or nitrogen or chlorine and a little of water and carbon
as a transporting agent.22
We had observed that a weak sublimation only occurred in a classical
Piper and Polich configuration when using ampoules sealed under vacuum
or filled with pure dry hydrogen with the material charge at 1000-1030C
and the top of the ampoule at 950C. When such gases and gas mixtures as
H2+H2O or Ar are introduced in the ampoules, fast transport to the cold point
was observed, demonstrating the activator role of theses gases and gas
mixtures.23
Given the fact that the free enthalpy of the reactions: ZnO + H2 l Zn +
H2O and ZnO + H2O l Zn + H2 + O2 are positive and that such reactions
cannot consequently occur towards the right-hand side, we have tried to
investigate the actual mechanisms of the CVT transport with hydrogen and
other species both from close-spaced vapor transport (CSVT) experiments
and from a numerical modeling.
4.1
CSVT experiments
CSVT allows short experiments and easy determination of the growth

rate. In this technique, a source and a substrate are placed at very short
distance (~ 1mm). The experiments were achieved under a hydrogen flow of
1 l/min at two source temperatures, 550 and 600C. In Fig. 1 is pictured the
instantaneous growth rate as a function of the inverse substrate temperature
for these two source temperatures. The solid and dashed lines correspond to a
fit using a theoretical model.23, 24
From this fit, the enthalpy of formation has been calculated and found
abnormally small, with a value of 0.42 eV, more than three times lower than
the one found for ZnSe. Furthermore, the variation of the instantaneous
growth rate as a function of time has been found to be non-linear, as shown
in Figure 2.
Growth rate ln(A/min)
14
Ts = 600 C
C
12
10
6
Ts = 550 C
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
-1
1000//4 (K )
Figure 1. ZnO instantaneous growth rate as the function of the inverse source temperature for
two source temperatures, 550C and 600C.
Figure 2. Instantaneous growth rate as a function of time.
R. Triboulet et al.
From this variation of the growth rate with time and from the small H
value, compared to the ZnSe one, we suggested that the vapor growth of ZnO
with hydrogen occurred according to a complex chemically assisted
sublimation mechanism involving a competition between two different
mechanisms, one associated with ZnO, the other with Zn.
4.2
Numerical simulation
In addition to this chemically assisted sublimation suggested with such

gaseous mixtures like H2 + H2O, acting as sublimation activators, several
experimental observations have led us to undertake a numerical simulation of
the transport.25 While no sublimation has been found to occur in closed
ampoules under vacuum conditions, high vapor transport rates are observed
not only with gases like H2 + H2O or CH4 but also if some particular
additional species like Zn, C or Fe are introduced in the ampoule.
A simple experimental set-up consisting in a horizontal ampoule of
diameter D and length L with the source material at one end and the growing
crystal at the other one has been taken into account for the simulation.25 The
source and crystal temperatures are TS and TC respectively and the ampoule
wall being subjected to a given temperature profile. We consider a PVT
process in which, after ZnO decomposition, the O2 (g) and Zn (g) are treated
as two species transported in a residual gas considered as inert. These species
combine at the crystallization interface giving ZnO. The involved
mechanism of decomposition and crystallization being unclear, we focus
only on transport phenomena. Following previous numerical studies of the
mass transport process on II-VI materials,26 surface reactions are taken into
account via the partial pressures of compounds at interfaces by using the
equilibrium constant and experimental data, if known, for the stoichiometric
coefficient. The governing equations for the vapor transport are the
conservation equations for momentum, heat and mass transfer. For the
numerical resolution of the conservation equations, the computational code
FLUENT has been used. The simulation has shown that if equilibrium partial
pressures at interfaces are considered, the growth can only occur for a
reduced interval of residual gas pressure (~10-3 atm). Fig. 3 shows that mass
fluxes through both interfaces are only similar for a narrow subinterval of
total pressures around 10-3 atm. Partial pressures of species, mainly Zn,
higher than those corresponding to the congruent sublimation (equilibrium
pressures) are required to obtain an effective mass transport in a wide
interval of total pressures. The generation of such pressures higher than the
equilibrium ones becomes a limiting step in the transport of ZnO by vapor
transport. This behaviour suggests the presence of an activated mechanism
in the ZnO decomposition, as proposed in the CSVT approach. Some kind of

autocatalytic process is induced from the reaction of ZnO with such
species as H2 + H2O, H2O, Zn, C, CO, CH4 or Fe resulting in the generation
of an excess of Zn.
Searching for a mechanism that could justify these high pressures, we
have suggested that:
A plausible mechanism could be the autocatalytic role of Zn that, by
means of a surface reaction, would reduce the activation energy of the ZnO
decomposition and would generate partial pressures higher than those at
equilibrium.
The role of some additional species as Zn, Fe, Cu, C and H2, in the ZnO
growth by vapor transport, can be interpreted in this framework. These
species would produce a Zn excess, either directly or from an O2
consumption.
This autocatalytic mechanism paves the way to new processes to
promote the growth of ZnO by vapor transport using the generation of an
additional Zn vapor pressure and limiting the off-stoichiometric effects by
controlling it.
1 .6
F(ZnO)-sub /F(ZnO)-crys
1 .4
1 .2
1 .0
0 .8
0 .6
1 e -5
1 e -4
1 e -3
1 e -2
1 e -1
T o ta l p re s s u re (a tm )
Figure 3. Calculated ratio between ZnO mass fluxes at both interfaces as a function of total
pressure under equilibrium conditions.
10
5.
R. Triboulet et al.
ZINC OXIDE SOLUTION GROWTH
Such solvents as PbF227 and V2O5/P2O5 mixtures28 have been used for the
flux growth of ZnO crystals. Using PbF2 as the solvent in sealed Pt
crucibles, ZnO crystals have been grown by THM.29 PbCl2 has been found to
be a very good solvent of ZnO as well, but showing also great reactivity with
any surrounding material but platinum.30 The same authors have found the
Zn-In alloys to be good solvents of ZnO without reactivity with silica. NaCl
has been found recently to be an excellent solvent of ZnO allowing solutiongrowth in silica at temperatures of about 900C.31 Figure 4 pictures a ZnO
crystal grown using NaCl as a solvent.
Figure 4. ZnO crystal grown using NaCl as a solvent.
6.
ZINC OXIDE HYDROTHERMAL GROWTH
Hydrothermal-growth is usually defined as the use of an aqueous solvent

at elevated temperature and pressure to dissolve a solute which would be
insoluble under ordinary conditions. The advantages of the hydrothermal
technique are a low growth temperature, a 'T
T close to 0 at liquid/solid
interface, an easily scalable technique, the reduction of most of the
impurities from source. The disadvantages are the presence of intermediate
products, the lithium or sodium or potassium incorporation when such
solvents as LiOH, NaOH or KOH are used, the slow growth rates (~10 mils
per day), the inert liner needed, the occasional incorporation of OH and H2O.
The first experiments on ZnO crystal growth by the hydrothermal method
began in the early sixties, both in Bell Telephone Laboratories and in the
Institute of Crystallography of the USSR Academy of Sciences. The
11
hydrothermal method has been shown suitable for the growth of large ZnO
crystals from (OH)- solutions at temperatures <500C under high pressure
(15 to 150 MPa) with a temperature difference 'T
T ~ 3 - 80K and a growth
rate in the range 0.05-0.3 mm/day.32-37 More recently, high quality ZnO
crystals, as demonstrated by X-ray rocking curves in the 40 arcsec range and
sharp PL peaks, have been grown hydrothermally at 355 C with a 'T of
10C from NaOH/KOH solutions as the solvent, pressures of 18,000 to
22,000 psi (maximum pressure 1500 atm).38 The crystals present a
dislocation density < 500 cm-2, a X-ray double diffraction FWHM ~130
arcsec and carrier mobilities ~175 cm2/Vs @ RT. The O and Zn surfaces of
(0001) planes have been found to behave differently. GaN layers deposited at
750 C on such substrates show DDX FWHM of 735 arcsec. Using a
KOH/LiOH solvent, ZnO bulk single crystals about 10 mm in dimension
have been grown at temperatures less than 400 C at pressures ranging from
830 to 1110 Mpa.39
Hydrothermal growth of ZnO remains very active in the Institute of
Crystallography of the Russian Academy of Sciences.40 Typical growth
conditions are40: temperature of the dissolution zone 270-350C, temperature
difference between the growth and the dissolution zones, 'T ~ 20-80C,
pressure 15-140 MPa, solvent KOH + LiOH, silver liner of capacity 500 cm3
put in autoclaves of capacity 1000 cm3. The growth rate is shown to depend
on the crystal orientation on the basis of the polar character of the crystal
structure along the c-axis and the presence of negatively charged species in
hydrothermal solutions.40 Typical crystals have a cross section equal to about
3 cm and a height along the hexagonal axis equal to 1-1.5 cm.41 They contain
concentrations 10-2-10-4 wt. % of Li and also concentrations 10-3-10-4 wt. %
of Cu, Mg, Si, Fe, Mn and Ag.
Very spectacular results in the hydrothermal growth of ZnO have been
recently reported by Oshshima et al.42 from Fukudas laboratory in Senda.
2-in ZnO single crystals (size up to 505015 mm3) have been grown using
KOH/LiOH @ T
T= 300-400C and P = 80-100 MPa. Rocking-curve FWHM
of 8 arc.sec for (002) reflections and 25 arc.sec for the (101) reflections are
reported (to be compared with 42 and 370 arc.sec respectively for a CVTgrown one). But incorporation of Li and K originated from LiOH and KOH
was observed at concentrations of 0.5-12 ppm for Li and 0.04-0.2 ppm for K.
In addition, incorporation of Fe and Al was also observed. Note that the
complex equipment used includes a 3 m length and 200 mm inner diameter
Pt liner preventing impurity incorporation from the autoclave inner surface.
12
7.
R. Triboulet et al.
CHARACTERIZATION OF THE CRYSTALS
Some electrical and structural properties of crystals grown by various

techniques, melt-growth, hydrothermal growth, SCVT and CVT are
summarized in the following table.
Table 1. Electrical and structural characteristics of the crystals.
n
technique
growth conditions
size
FWHM
(arcsec) (cm-3)
Melt-growth
CERMETs
1 inch
125
51017
Process
5.5 inch 40-100 8.41016

Hydrothermal
335-400C, T=10C
T
KOH/NaOH or /LiOH
~ 100
growth
Pmax ~ 800-1500 MPa
KOH/LiOH
2-in
8
Pmax ~ 80-100 MPa
Growth rate ~ 8 m/h
Source 1000-1200C
SCVT
dist. source-substrate 3 inch
40
81016
res. atm. H2+H2O; seeds
40 mm
Growth rate: 40 m/h
CVT
C as an activator
cm-size
28
cm2/Vs
130
Ref.
150
3
3335
175
38
13
20
Whatever the method used, the electrical and structural properties of the
crystals are very close except the record structural quality recently reported
for hydrothermally grown crystals by Ohshima et al.38 Note the low growth
rates of the vapor grown crystals and more again of the hydrothermally
grown ones which could be a handicap in view of industrial production.
8.
CONCLUSIONS
In spite of the numerous efforts dedicated to the ZnO bulk growth for
several years, high quality large and cheap ZnO substrates are not yet
available. The best crystals obtained so far as far as the crystallographic
quality is concerned have been grown by the hydrothermal method, whose
possibilities are limited by an extremely weak growth rate and the
unavoidable incorporation of impurities coming from the solvent, in addition
to complex equipment. The better knowledge and understanding of the
mechanisms of vapor transport can let expect significant progresses in this
field in a near future.
13
REFERENCES
1. R. Triboulet, Proceedings SPIE 4412, 1 (2000).
2. G. Agarwal, J. E. Nause, and D. N. Hill, Mat. Res. Soc. Symp. Proc. 512, 41 (1998).
3. D. C. Reynolds, C. W. Litton, D. C. Look, J. E. Hoeschler, B. Claflin, T. C. Collins, J.
Nause, and B. Nemeth, J. Appl. Phys. 95, 4802 (2004).
4. M. Shiloh and J. Gutman, J. Crystal Growth 11, 105 (1971).
5. W. Piekarczyk, S. Gazda, and T. Niemyski, J. Crystal Growth 12, 272 (1972).
6. K. Matsumoto, K. Konemura, and G. Shimaoka, J. Crystal Growth 71, 99 (1985).
7. K. Matsumoto and G. Shimaoka, J. Crystal Growth 86, 410 (1988).
8. K. Matsumoto and K. Noda, J. Crystal Growth 102, 137 (1990).
9. M. Hirose and I. Kubo, Jpn. J. Appl. Phys. 8, 402 (1969).
10. Y. S. Park and D. C. Reynolds, J. Appl. Phys. 38, 756 (1967).
11. M. Hirose, Y. Furuya, and I. Kubo, Jpn. J. Appl. Phys. 9, 726 (1970).
12. E. Scharowsky, Z. Phys. 135, 318 (1953).
13. K. F. Nielsen, J. Crystal Growth 3/4, 141 (1968).
14. K. J. Fischer, J. Crystal Growth 34, 139 (1976).
15. I. Kubo, J. Phys. Soc. Jap. 16, 2358 (1961).
16. T. Takahasi, A. Ebina, and A. Kamigawa, Jpn. J. Appl. Phys. 5, 560 (1966).
17. E. A. Weaver, J. Crystal Growth 1, 320 (1967).
18. M. Hirose and Y. Furuya, Jpn. J. Appl. Phys. 9, 423 (1970).
19. D. C. Look, D. C. Reynolds, J. R. Sizelove, R. L. Jones, C. W. Litton, G. Cantwell, and W.
C. Harsch, Solid State Comm. 105, 399 (1998).
20. J.-M. Ntep, S. Said Hassani, A. Lusson, A. Tromson-Carli, D. Ballutaud, G. Didier, and R.
Triboulet, J. Crystal Growth, 207, 30 (1999).
21. S. Hassani, A. Tromson-Carli, A. Lusson, G. Didier, and R. Triboulet, Phys. Stat. Sol. 229,
835 (2002).
22. A. Mycielski, A. Szadkowski, L. Kowalczyk, B. Witkowska, W. Kaliszek, B. Chwalisz, A.
Wysmolek, R. Stepniewski, J. M. Baranowski, M. Potemski, A. Witowski, R. Jakiela, A.
Barsz, P. Aleshkevych, M. Jouanne, W. Szuszkiewicz, A. Suchocki, E. Lusakowska, E.
Kaminska, and W. Dobrowsolski, Proc. 11th International Conference on II-VI
Compounds, to be published.
23. J.-M. Ntep, M. Barb, G. Cohen-Solal, F. Bailly, A. Lusson, and R. Triboulet, J. Crystal
Growth 184/185, 1026 (1998).
24. R. Triboulet, J.-M. Ntep, M. Barb, P. Lemasson, I. Mora-Sero, and V. Munoz, J. Crystal
Growth 198/199, 968 (1999).
25. R.Tena-Zaera, M. C. Martnez-Toms, S. Hassani, R. Triboulet, and V. Muoz-Sanjos, J.
Crystal Growth, to be published.
26. N. Ramachandran, C. Su, and S. L. Lehoczky, J. Crystal Growth 208, 269 (2000).
27. K. F. Nielsen and E. F. Dearborn, J. Phys. Chem. 64, 1762, (1960).
28. B. M. Wanklyn, J. Crystal Growth 7, 107 (1970).
29. G. A. Wolf and H. E. LaBelle Jr., J. Am. Ceram. Soc., 48, 441 (1965).
30. J.-M. Ntep, PhD thesis, Paris 2000.
31. S. Hassani, G. Didier, P. Galtier, and R. Triboulet, to be published.
32. R. A. Laudise and A. Ballman, J. Phys. Chem. 64, 688 (1960).
33. E. D Kolb and R. A. Laudise, J. Am. Ceram. Soc.49, 302 (1966).
34. R. A. Laudise, E. D. Kolb, and A.J. Caporaso, J. Am. Ceram. Soc. 47, 9 (1964).
35. E. F. Venger, A. V. Melnichuk, L. Yu. Melnichuk, and Yu. A. Pasechnik, phys. stat. sol.
(b) 188, 823 (1995).
14
R. Triboulet et al.
36. I. P. Kuzmina and V. F. Antonova, M. Nauka 4, 151 (1964).

37. I. P. Kuzmina, Kristallografiya 13, 920 (1968).
38. M. Suscavage, M. Harris, D. Bliss, P. Yip, S.-Q. Wang, D. Schwall, L. Bouthillette, J.
Bailey, M. Callahan, D. C. Look, D. C. Reynolds, R. L. Jones, and C. W. Litton, Mat. Res.
Soc. Symp. Proc. 537, 1999 to be published.
39. T. Sekiguchi, S. Miyashita, K. Obara, T. Shishido, and N. Sakagami, J. Crystal Growth
214/215, 72 (2000).
40. L. N. Demianets and D. V. Kostomarov, Ann. Chim. Sci. 26, 193 (2001).
41. V. A. Nikitenko, J. Appl. Spectr. 57, 783 (1994).
42. E. Ohshima, H. Ogino, I. Niikura, K. Maeda, M. Sato, M. Ito, and T. Fukuda, J. Crystal
Growth 260, 166 (2004).
Chapter 2
GROWTH MECHANISM OF ZnO LAYERS
A. Kh. Abduev1, A. Sh.Asvarov1, A. K. Akhmedov1, I. K. Kamilov1 and S.
N. Sulyanov2
1
IInstitute of physics, DSC of RAS, 367003, Yaragscogo str., 94 Makhachkala, Russia;

Institute of Crystalografy of RAS, 119333, Leninski pr-t 59, Moscow, Russia
Abstract:
The formation processes of clusters in gas phase in the Zn - O2 system were

studied. It was shown that Zn12O clusters are formed in a surplus of zinc vapor.
We studied dc-magnetron sputtering of ZnO layers by co-sputtering
with ZnO:Ga ceramic and Zn metal targets method. We observe that at surplus
of zinc vapor in the reagent stream leads to synthesis of homogeneous layers
without columnar structure. We propose that growth of ZnO layers at surplus
of zinc vapor is governed by the vapor-liquid-solid mechanism.
Key words:
Cluster, whisker, columnar structure, vapor-liquid-solid mechanism
1.
INTRODUCTION
Development of new technologies of synthesis of ZnO layers is

stimulated by potential applications in opto- and nanoelectronics. In the past
decade, blue-UV generation has been obtained in ZnO thin films1 and ZnO
whiskers2 at room temperature; p-type ZnO thin films,3 thin-film diode
structures4. Also transparent transistors5 were synthesized. The interest in
ZnO layers is also motivated by the search of alternatives to expensive
transparent electrodes on the basis of In2O3-SnO2 system.6 Unique surface
properties of zinc oxide materials stimulate research toward development of
gas-sensing devices.7 Thin layers of ZnO have been grown by various
techniques, including chemical vapor deposition (CVD),8,9 pulsed laser
deposition (PLD), molecular beam epitaxy (MBE),10,11 dc- and rf- magnetron
sputtering.12-16 From a commercial viewpoint, the method of dc-magnetron
sputtering is the most attractive method for the synthesis of ZnO layers on
15
16
A. Kh. Abduev et al.
large-area substrates. Meanwhile there are several essential drawbacks in this

method including:
the formation of columnar structures into ZnO layers, which results in
decrease of carrier mobility and dispersion of light on grain boundaries;
the presence of irradiated defects caused by the method itself;
the poor structural perfection of ZnO layers due to far from equilibrium
growth conditions of ZnO layers.
In the present work, the influence of reagents stream composition on the
structure of ZnO layers has been investigated for dc-magnetron sputtering.
Particular attention has been dedicated to the study of processes of reagent
interactions in the gas phase. And, for a deeper understanding of this process,
the process of formation of clusters in the gas phase of the simple Zn-O2
system was investigated.
2.
EXPERIMENT
To examine the processes of cluster formation in the gas phase a Knudsen

Cell has been used. A mixture of ZnO and graphite powders was positioned
inside the effusion cell. The carbothermal reaction between ZnO and graphite
was used to create zinc vapor. The effusion cell was heated up to a
temperature above 1000C.
The co-sputtering method, when ZnO:Al (2 at.%) ceramic target and Zn
metal target were sputtered simultaneously in atmosphere of pure Ar
(99.99%), was used for the formation of the ZnO-Zn flow. The targets were
discs 3 mm in thickness and 30 mm in diameter. The discharge current for
the ZnO:Ga target was maintained at 60 mA. Sputtering of the Zn target was
carried out at a discharge current in the range of 25-90 mA. The ZnO layers
were deposited on amorphous SiO2 thermally grown on Si substrate. The
thickness of amorphous SiO2 was 100 nm.
The chamber pressure was down to 5u10-4 Pa using a diffusion pump
before introducing the Ar sputtering gas. During deposition the pressure of
Ar was maintained at |1 Pa.
X-ray diffraction measurements were carried out using KARD-6
diffractometer with a 2D-area detector based on the multi-wire planar
proportional chamber (Cu K-radiation, graphite monochromator, 0.1mm
width incident beam slit).
LEO-1450 scanning electron microscope (SEM) was used to characterize
the surface and the cross-section of the samples.
Hall voltage was measured by a four-point probe method for the samples
shaped like a cross, using dc current and dc magnetic field.
Growth mechanism of ZnO layers
3.
RESULTS
3.1
Cluster formation in the gas phase
17
We took interest in the study of formation of nanoparticles in a gas phase

and their influence on the formation of columnar structures of ZnO layers.
It is known,17,18 that non-equilibrium conditions of scattering and cooling
of the plasma leads to interactions of plasma components and the formation
of clusters and nanoparticles.
Our previous studies19,20 have shown that the interaction of equal
fractions of zinc vapor and oxygen in a gas phase result in the step-by-step
formation of clusters, fractal clusters and nanoparticles according to the
Witten-Sander mechanism. Also it has been shown, that formation of
whiskers at surplus of Zn vapour in gas phase proceeds with the vapourliquid-solid (VLS) mechanism in the absence of a liquid-forming impurity
agent. We believe that the liquid-forming agent was a ZnO1-x surface phase
with relatively low melting-point. ZnO1-x was formed due to high pressure of
Zn vapor near the growing surface.
Figure 1. ZnO tetrahedral aggregate. The inset shows the model of Zn12O nucleus-cluster
In order to model the processes of cluster formation in a gas phase at a

surplus of zinc vapour we carried out the effusion of zinc vapour in the
oxygen atmosphere. There is the nucleation and the growth of the symmetric
18
tetrahedral aggregates in a gas phase (Fig. 1).

The characteristic feature of these tetrahedral aggregates was the
simultaneous growth of four (0001)-oriented ZnO whiskers from one
nucleus. Apparently the nucleus of such tetrahedral aggregates cannot be a
close-packed crystallite. We speculate that the nucleus of such aggregate has
tetragonal structure that leads to the further growth of the four hexagonal
crystallites (whiskers).
Hence, we suggest the following model of Zn12O nucleus-clusters. The
nucleus-cluster has the shape of the truncated tetrahedron with 12 atoms of
zinc and one atom of oxygen in the center (inset of Fig. 1). Such nucleuscluster is a loosely coupled system formed by an atom of oxygen and 12
equidistant atoms of zinc. Scattering in zinc vapor at increased partial
pressure of oxygen, facilitates the nucleation and growth of (0001)ZnO
whiskers on four facets of the Zn12O cluster occurs.
3.2
Magnetron co-sputtering of ZnO and Zn target
The columnar structure with dominant basic orientation is typical feature

of the ZnO layers synthesized by methods of magnetron sputtering in
conditions far from equilibrium. According to Knuyt et al.21 the formation of
columnar structures is bound to the tendency of a system to diminish its free
energy. Earlier it has been shown, that ZnO columns can be formed from
nano-sized grains.22 Thus there is an angle of several degrees of disorder of
these grains. However our results reveals that columns may grow in isolation
and thus do not interact at the initial stage of deposition. We believe that it is
possible to explain formation of columnar structure by the presence of
clusters in a flow of reagents at high chamber pressures.
At first we have investigated the structure of ZnO layers grown by dcmagnetron sputtering of ZnO:Ga ceramic target, as a function of the distance
between the target and the substrate.
ZnO layers were deposited on the substrate at room temperature. Targetto-substrate distances were 30, 60 and 100 mm. SEM micrographs of crosssections of ZnO layers that have been deposited at these target-to-substrate
distances are given in figure 2. The increase of the distance leads to a
decrease of column width. This distance dependence may be explained by
the cluster separation based on height and sizes of the clusters that reached
the growing surface. In other words, we believe that the size of clusters
diminishes due to smaller free length and gravity as target-to-substrate
distance increases. The formation of the columnar structure can be explained
by the presence of nano-sized clusters in a flow of reagents, and that the
formation of layers can be described by a model close to Witten-Sanders
one.
19
Figure 2. SEM micrograph of cross-sections of ZnO layers prepared with various target-tosubstrate distances: a - 30 mm; b - 60 mm; c - 100 mm.
In order to eliminate columnar structures it is necessary to increase the

migration ability of atoms and molecules on the surface and also to remove
the involvement of clusters in the process of growth. To prevent the
participation of clusters and nanoparticles during growth and in order to
create the conditions for highly perfect ZnO layers synthesis, we explored
the possibility of ZnO layers growth according to VLS-model.
The co-sputtering of ZnO:2at%Ga ceramic target and Zn metal target in
Ar atmosphere had resulted in the surplus of zinc in a flow of reagents and
thus created favorable conditions for the growth of ZnO layers according to
VLS-model. SEM micrographs of a surfaces and cross-section of ZnO layers
which have been deposited at various temperatures are given in figure 3.
At substrate temperatures of about 300C the ZnO layers have the
advanced surface and typical columnar structure. The increase of the
20
Figure 3. SEM micrographs of surfaces (a1, b1) and cross-sections (a2, b2) of ZnO layers
deposited by dc-magnetron co-sputtering (a - Tsub=3000C; b - Tsub=5000C).
substrate temperature to 5000C gives in the essential reduction of the

surface relief of ZnO layer and the disappearance of columnar structures.
These ZnO layers and also the control ZnO layer deposited without cosputtering (by dc-magnetron sputtering only of ZnO:Ga target) were
characterized by the X-ray diffraction method. Si-single crystal substrate
111-reflection was used as a reference for the instrumental resolution
function determination. For every exposure, the 0002-reflection from the
layer was integrated in the same small rectangular area of detector. The
method described in reference 23 was used. Results are plotted in fig. 4.
Figure 4 shows that the FWHM values of X-ray rocking are 1.80 and 1.20 for
ZnO layers deposited by co-sputtering on a substrate heated at 3000C (curve
3) and 5000C (curve 4), respectively, whereas the FWHM value of X-ray
rocking for control ZnO layer (curve 1) is much wider. From these data it is
possible to draw the conclusion, that the presence of zinc surplus in a flow of
reagents near the growing surface results in the considerable improvement of
the perfection of ZnO layers.
21
Figure 4. XRD date of: curve 1 - without co-sputtering (only sputtering of ZnO:Ga target);
curve 2 - instrumental resolution function; curve 3 - co-sputtering (Tsubstrate=3000C); curve 4 co-sputtering (Tsubstrate=5000C)
The resistivity of these ZnO layers decreased with increasing substrate

temperature, caused mainly by an increase in Hall mobility. So, the values of
Hall mobility were 18.65 cm2/Vs and 33.78 cm2/Vs for ZnO layers
deposited by dc-magnetron co-sputtering on a substrate heated at 3000C and
5000C respectively, whereas the value of Hall mobility for the control ZnO
layer (without co-sputtering) was 8.31 cm2/Vs. Herewith, carrier
concentrations for all three samples were similar and approximately 3u1020
cm-3.
Also the structure of ZnO layers as a function of the ratio of discharge
currents for ZnO:Ga and Zn targets has been investigated. Synthesis of
layers was carried out at various discharge currents for Zn target (25, 50 and
90 mA) and constant discharge current for ZnO:Ga target. The temperature
of substrates was 5000C. SEM micrographs of ZnO layers obtained under
these conditions are shown in figure 5. We observe that upon an increase of
zinc component in the flow of reagents there is a transition from a typical
columnar structure (fig. 5a) to homogeneous layers (fig. 5d). It is clearly
seen (fig. 5b), that at low discharge current of the Zn target the columnar
structure of layer is still present. But the data of the XRD-measurements
show, that the structural perfection of ZnO layers, deposited under maximum
discharge current for Zn metal target, was improved. The value of Hall
22
mobility at maximum discharge current of Zn target also decreased. This

trend has not been explained clearly. The films, prepared at the substrate
temperature above 5000C and current of Zn target 50 mA, shown best
structural and electrical properties.
The involvement of clusters during growth is excluded by the increasing
of vapor pressure of Zn mass and the creation of low-melting ZnO1-x phase at
the growing surface.
Figure 5. SEM micrographs of the cross-section of ZnO layers, deposited at various discharge
currents for the Zn target: a - 0 mA; b - 25 mA; c - 50 mA; d - 90 mA.
4.
23
CONCLUSION
Our experimental results and their subsequent analysis lead us to the

following conclusions:
1. At relatively high chamber pressure (P | 1 Pa) the partial aggregation of
reagents is observed in the plasma of magnetron discharge.
2. When zinc vapors are spurted through the gas mixture containing
oxygen, the formation of tetrahedral aggregates consisting of four
whiskers originated at the nucleus tends to occur in the gas phase.
According to our model, the growth of these aggregates occurs under
non-equilibrium conditions as a result of the formation of Zn12O nucleusclusters which subsequently grow into the tetrahedral aggregates at the
adiabatic expansion of zinc vapors and Zn12O clusters in the gas mixture
containing oxygen.
3. The mechanism of growth of columnar structure ZnO layers assumes the
presence of nanosized clusters in the reagent flow and that the formation
of ZnO layers can be described by a model close to the one of WittenSanders. Increase of target-to-substrate distance leads to the decrease of
the number of aggregates in the flow near the substrate and to the
improvement of the structural perfection of ZnO layers.
4. Ata a substrate temperature above 500 OC and at surplus of zinc in the
flow of reagents near the growing surface there is a synthesis of
structurally perfect ZnO layers. The growth mechanism of ZnO layers at
surplus of zinc vapor is governed by vapor-liquid-solid mechanism. It is
shown, that in the absence of liquid-forming impurity agent (Au) the
growth mechanism of ZnO layers and ZnO whiskers is governed by the
vapor-liquid-solid mechanism. ZnO1-x surface phase with relatively low
melting-point is formed due to high pressure of Zn vapor near the
growing surface.
ACKNOWLEDGMENTS
This research was supported by Analytical Centre of DSC RAN.
REFERENCES
1.
2.
3.
4.
Sunglae Cho, Yunki Kim, Yi Sun, George K.L., Appl. Phys. Lett. 75, 2761 (1999).
M. H. Hiang, S. Mao, H. Feick et al, Science 292, 1897 (2001).
K. Minegishi, Y. Koiwai, Y. Kikuchi et al. , Jpn. J. Appl. Phys. 36, L1454 (1997).
Y. R. Ryu, W. J. Kim, H. W. White. , J. of Cryst. Growth, 219, 419 (2000).
24
5. P. F. Carsia, R. S. McLean, M. H. Reilly et al. , Appl. Phys. Lett. 82, 1117 (2003).
6. T. Minami, MRS BULLETIN AUGUST, 38-44 (2000).
7. S. Roy, S. Basu, Bull. Mater. Sci. 25, 513 (2002).
8. C. R. Gorla, N. W. Emanetoglu, S. Liang et al., J. Appl. Phys. 85, 2595 (1999).
9. B. S. Li, Y. C. Liu, Z. Z. Zhi et al., J. Vac. Sci. Technol. A 20, 1779 (2002).
10. H. Z. Wu, K. M. He, D. J. Qiu, and D. M. Huang, J. of Crystal Growth 217, 131 (2000).
11. T. Ohgaki, N. Ohashi, H. Kakemoto et al., J. Appl. Phys. 93, 1961 (2003).
12. M. Chen, Z. L. Pei, C. Sun et al., J. Crystal Growth 220, 254 (2000).
13. K. Tominaga, T. Murayama, I. Mori et al., Thin Solid Films 386, 267 (2001).
14. Y. Igasaki, T. Naito, K. Murakami et al., Appl. Surf. Science 169-170, 512 (2001).
15. F-J. Haug, Zs. Geller, H. Zogg et al, J. Vac. Sci. Technol. A 19, 171 (2001).
16. T. Minami, T. Miyata, T. Yamamoto et al., J. Vac. Sci. Technol. A 18, 1584 (2000).
17. B. M. Smirnov, Phys. Usp. 37, 621 (1994).
18. V. N. Tsytovich, Phys. Usp., 40, 53 (1997).
19. A. Kh. Abduev, A. K. Achmedov, V. G. Baryshnikov et al., Technical Physics Letters 26,
332 (2000).
20. A. Kh. Abduev, A. Sh. Asvarov, A. K. Achmedov et al., Technical Physics Letters, 28, 952
(2002).
21. G. Knuyt, C. D. Quaeyhaegens, J. Haen et al., Thin Solid Films 258, 159 (1995).
22. I. Sieber, N. Wanderka, I. Urban et al., Thin Solid Films 330, 108 (1998).
23. S. N. Sulyanov, A. N. Popov, and D. M. Kheiker, J.Appl.Cryst. 27, 934, (1994)
Chapter 3
KINETICS OF HIGH-TEMPERATURE DEFECT
FORMATION IN ZnO IN THE STREAM OF
OXYGEN RADICALS
M. B. Kotlyarevsky1, I. V. Rogozin2, and A. V. Marakhovsky2
1
Academy of Management and Information Technologies ARIM, 3, Uritskogo Str.,

Berdyansk, Ukraine, 17118; 2Berdyansk Teachers Training University, 4, Shmidt Str.,
Berdyansk, Ukraine, 71100
Abstract:
The kinetic analysis of defect formation made it possible to develop a new

method for obtaining single crystal layers by means of heat treating the II-VI
compound single crystal substrates in the atomic chalcogen flux. With this
method, an inversion of the conductivity in ZnO from n- to p-type has been
accomplished. This technology made it also possible to obtain heterostructures.
Key words:
Kinetic analysis, oxygen radicals, heterostructures, n-type ZnO, p-type ZnO
1.
INTRODUCTION
In this work methods of high-energy chemistry are widely employed in

the synthesis of II-VI compounds with controlled properties.1 In this
approach, a gaseous reactant is activated to produce excited neutral radicals
(atoms) using an external source (photolysis, gas discharge, and others) and
is then delivered to the reaction zone. One of the chief advantages of the
high-energy chemical methods is that the synthesis temperature can be
reduced substantially.
We have shown that this approach has also considerable potential for use
in the technology of II-VI semiconductors containing nonmetals (O, S, Se,
and Te) that exist only in the vapor phase at typical synthesis temperatures.
A drawback common to the wide-gap ((E
Eg t 2.5 eV) semiconductors in
question (ZnS, ZnO, ZnSe) is the high and, hence, difficult to control
25
26
M. B. Kotlyarevsky et al.
concentration of native defects, which are generated at a high rate during

high-temperature synthesis. This, in turn, leads to native-defect monopolar
(n-type) conduction, limiting the application of these compounds in
semiconductor optoelectronics.
Consider the interaction of II-VI crystals (ZnS, ZnSe, ZnO) with
chalcogen VI (S, Se, O) vapor. High-temperature annealing of a II-VI crystal
in chalcogen vapor under equilibrium conditions does not produce a
chalcogen excess (e.g.: due to native acceptors such as chalcogen interstitials
Bi or Zn vacancies VZn ) because the formation of an acceptor center must be
accompanied by the formation of a compensating donor, VB or Zni, for the
system to remain in thermodynamic equilibrium. Thermodynamic analysis of
the high-temperature native-defect equilibrium in ZnS, ZnSe and ZnO were
analyzed.2-4 The calculations unambiguously show that there is a broad range
of chalcogen vapor pressures (e.g.: up to p | 1001000 GPa for ZnS and
ZnSe) in which the dominant defect species in these compounds are VA and
VB, so that the native defect conduction is fully compensated.
The thermodynamic and kinetic aspects of hole conduction selfcompensation were analyzed in detail.57 It was concluded that the main
reason for the monopolar conduction in the compounds under consideration
is that the chalcogen is present in the vapor phase in molecular form. It was
also shown that the compensation of acceptor centers in II-VI compounds
can be suppressed by reducing the synthesis or doping temperature to below
a critical temperature, if the process is run in saturated chalcogen vapor.
Figure 1 shows the calculated defect densities in ZnS as functions of the
zinc and sulfur partial pressures at 1000 K. As it is visible from the figure,
there is a rather broad range of chalcogen vapor pressures (10-1 to 106 Pa) in
which the conductivity is very low and varies insignificantly. The accounts
show, that area of pressure pB2 , at which the degree of self-compensation
begins to decrease, is in the range of unattainable chalcogen vapor pressure.
These conclusions relate to ZnO as well.
Thermodynamic analysis suggests approaches to achieving native-defect
and impurity-related hole conduction in II-VI compounds. These are
primarily non-equilibrium, high-energy chemical methods, such as ion
implantation of the B component or acceptor impurity, annealing under a
layer of a substance that dissolves markedly different amounts of the A and
B components, and annealing in an atomic-chalcogen atmosphere.67 This
last method will be considered further.
Kinetics of high-temperature defect formation in ZnO
27
Figure 1. Native-defect equilibrium in ZnS at 1000 K; points A and B correspond to saturated

Zn and S2 vapors, respectively. Other designations are explained in text.
2.
KINETIC ANALYSIS
This method consists of increasing the atomic chalcogen flux JB to the above
equilibrium value for a given pressure and temperature via dissociation
of chalcogen molecules in response to external influences. As shown,10 the
fraction of dissociated oxygen molecules in an RF discharge attains 15%. At
pO2 1 Pa, the dissociation of 1% of the oxygen molecules produces atomic
oxygen pressure pO ~10-2 Pa. The same activated oxygen pressure can be
achieved via thermal dissociation of molecular oxygen at 1000 K and a
pressure pO2 pO2 K D 102 GPa where KD is the equilibrium constant of the
reaction O2 = 2O + D (D
( is the dissociation energy of molecular oxygen).
Thus, RF activation during heat treatment in the vapor phase produces an
atomic oxygen pressure that is essentially unattainable via thermal
dissociation under typical heat-treatment conditions.
28
subsurface layer
Figure 2. (a) Schematic illustrating the ADC equilibrium on the surface of an AB crystal and
(b) the corresponding electronic transitions: (1) transition resulting from the interaction of a
surface B atom with a VA vacancy ( J ddes ), (2) transition resulting from the interaction of
a surface B atom with a hole ( J ddes ), (3) transition resulting from the recombination of a
).
B adatom with a VBx vacancy ( J
rec
The non-equilibrium chalcogen concentration during annealing

corresponds to a higher effective pB2 accompanied by no condensation, which
offers the possibility of achieving the states represented by points M
and N in Fig. 1 (stoichiometric and chalcogen-enriched compositions).
Kinetic analysis of the chemistry of native defects proved very useful in the
case of crystal-activated chalcogen vapor systems. It is this approach that
revealed the possibility of substantially extending the ranges of controlled
native-defect conductivities5 and achieving a type conversion in the
compounds under consideration. Analysis of defect formation kinetics
allowed us to propose a number of physicochemical approaches to
controlling defect formation processes, which are governed by the dynamic
adsorption-desorption-crystallization (ADC) equilibrium on the crystal
surface (Fig. 2).
Based on the ADC model, we proposed a kinetic model for defect
formation in II-VI compounds. A system of rate equations describing the
interaction between a II-VI crystal and chalcogen vapor was set up and
solved under the assumption that only the atomic component of the vapor
phase reacts with the crystal surface. In calculating the ADC equilibrium on
29
the crystal surface, we took into account the interactions of neutral and
charged defects with the vapor phase.
System of rate equation describing processes of defect formation can be
written in the form:
dN B'
dW
JB
J dec
2J des 2
J cr
dVA
dW
J cr
dVA'
dW
g [VA ] h p[VA' ] J ddes
dVB
dW
J ddes
dp
dW
J rec
g [VA ] h p[
p VB VA' .
J rec
g[V ] p[V ' ] J
'
A
J rec
] J ddes
J rec
(1)
(2)
(3)
(4)
J rec
(5)
(6)
Figure 3 shows the surface concentrations of defects in ZnO as functions

of the atomic-oxygen flux. As seen, raising the temperature at a constant flux
JB increases the degree of compensation, without influencing the hole
conductivity. The flux at which the conductivity changes from compensated
to uncompensated and the charge neutrality condition changes from
VZn' V0x to VZn' p (the formation of a metal vacancy is accompanied
by the appearance of a free hole in the valence band, rather than by the
formation of a compensated chalcogen vacancy) depends on the annealing
temperature.
For a given flux, there is a critical temperature above which
uncompensated hole conduction cannot be achieved:
30
TC
GdesB
k
Eg
Ea1
Ed1
GA
Gddiifff B
(7)
Figure 4 shows the critical temperature as a function of the atomicoxygen flux. These data demonstrate that the critical temperature can only be
raised substantially by increasing JB. The calculations show that for ZnO Tc =
m2s1 in really reachable
800 K at chalcogen fluxes of orders 1016-1017 cm
experiments. Thus, it follows from the above thermodynamic and kinetic
analyses that one way of controlling the chemistry of native defects in II-VI
compounds is by increasing JB to above the equilibrium value at a given
pB2 and temperature via activation of chalcogen molecules.
3.
RADICAL-BEAM GETTERING EPITAXY
Based on the kinetic analysis of the ADC equilibrium on a crystal surface

exposed to chalcogen radicals (atoms) in the vapor phase, we devised a
method for hetero- and homoepitaxial growth of II-VI films. This method
was the first to rely on the heat treatment of II-VI compounds in chalcogen
vapor consisting of radicals7 (atoms) generated using an RF discharge or
photolysis. The method was called radical-beam gettering epitaxy (RBGE).
During annealing in atomic chalcogen vapor (e.g., annealing of ZnO in O
radicals), the dominant growth mechanism is quasi-epitaxy: the growth of
crystalline layers on a crystalline matrix via sorption of chalcogen atoms
from the vapor phase and gettering of metal atoms from the crystal bulk.
This process is accompanied by the formation of Zn vacancies, which act as
acceptors.
Since Zn vacancies may be in two charge states, either of the two levels
corresponding to the singly and doubly charged Zn vacancies ( VZZn' and VZZn" )
may show up in the conductivity of p-type crystals, depending on the degree
of compensation of Zn vacancies by O vacancies and the position of the
Fermi level.
The PL of oxygen-enriched ZnO was studied8 using single-crystal ZnO
layers grown on substrates vapor-grown ZnO single crystals, both nominally
undoped and doped with Li and Na. The doping level was 1016 to 1017 cm-3,
and the conductivity of the samples was 10-5 to 10-7 S/cm. All of the samples
were heat-treated under radical-beam gettering epitaxy conditions. The
parent ZnO single crystals substrate was annealed at temperatures from 670
to 970 K in an atmosphere containing oxygen radicals (atoms). The oxygen
pressure in the growth chamber was 10010-3 Pa. Oxygen radicals were
generated by an 80-W RF discharge. The resulting ions were separated by a
-2
log ( D ef. ) ( cm )
14
13
31
'
V Zn
VO
*
VO
'
V Zn
12
h
11
V Zn
10
VO
9
8
14
VO
V Zn
15
16
17
18
19
20
-2 -1
log JS ( cm s )
Figure 3. Defect densities in the surface layer of ZnO the flux of oxygen radicals (atoms) at
1200 (solid lines) and 1000 K (broken lines).
magnetic filter. The atomic oxygen flux, JO=1016 to 1017 cm-2s-1, was
determined as described. ZnO layers grown at an oxygen radical flux JO |
1.51017 cm-2s -1 and annealing temperatures in the range of 670870 K were
p-type, as determined by Hall effect measurements. The lowest resistivity ptype ZnO layers, with U | 6102 :cm and hole mobility p ~23 cm2/(Vs),
were obtained at 710 K. Raising the annealing temperature to 810 K was
found to increase U to about 3107 :cm. The layers produced at Tann t 910 K
were n-type, with the resistivity above 109 :cm. These data indicate that the
critical annealing temperature of ZnO is about 910 K, in accordance with
theoretical estimates.
In the UV range, the PL spectrum of oxygen-enriched ZnO ((p-type)
shows a strong line at 369.5 nm with a full width at half maximum (FWHM)
of 14 meV. In the visible range, we observe a weak band centered at 400 nm.
Comparison of the luminescence spectra of n- and p-type ZnO indicates that
the 369.0 and 369.5-nm lines are due to bound excitons. As shown by
Butkhuzi et al.,9 the 369.0 and 369.5-nm emissions arise from excitons
bound to neutral donors and acceptors, respectively. Increasing the annealing
time at 710 K increases the intensity of the 400-nm band. The spectra of
samples annealed for 4 h show only the 400-nm band, which spans the entire
32
excitonic region. By analogy with the 345-nm emission6 in ZnS, we assume

that the 400-nm emission in ZnO is due to the singly positively charged
x
associate VZn VO . This assumption is supported by the fact that the
concentration of oxygen vacancies in ZnO decreases with increasing oxygen
content, and, for electro neutrality to be maintained, some of the oxygen
vacancies pass into the doubly charged state ( VOxx ).
1200
1100
TC (K)
1000
900
800
700
600
12
14
16
18
20
22
-2 -1
log JS (cm s )
Figure 4. Critical temperature as a function of the atomic oxygen flux.
4.
HETEROEPITAXY OF II-VI COMPOUNDS
The kinetics of defect formation in a heterophase system such as a singlecrystal compound of one chalcogen, B' (e.g.: selenium), exposed to a flow of
radicals of another chalcogen, B'' (e.g.: oxygen), are more difficult to
analyze. The general trends of defect formation in such systems must,
however, be the same as above. The formation of VA vacancies in an AB'
crystal is accompanied by the transition of A atoms to the surface layer,
where they combine with B'' atoms to form AB' dimers. In the system of rate
equations (1-5) one must take into account the formation of molecules
containing both chalcogens (e.g.: SeO2) and desorption of such molecules.
This process results in the formation of VB' defects in the crystal, whose
composition gradually changes to AB'' owing to B'' in-diffusion through
VB' vacancies.
33
To a first approximation, the system of rate equations describing

processes in such heterophase systems can be written in the form (A = Zn, B'
= Se, B'' = O)
dN O'
dW
JO
J des
d O
dN O
dW
>
@ JO
dN ZZn
dW
J cr
dNSSe
dW
J ddes
2J des
d O2
J rec
J rec
J des
(8)
(9)
(10)
(11)
The mass-balance constraint has the form [ Se

>Se @ const. [O] +
S ] [O]
[Se] = const. In these relations we use designations analogous to those in
Eqs. (1-5): JO is the flux of oxygen atoms (radicals) incident on a twodimensional crystal, N O' is the surface concentration of oxygen adatoms, NO
is the concentration of oxygen atoms in the surface layer, and [O] is the net
oxygen concentration in the adlayer and surface layer.
Diffusion from the surface layer to the crystal bulk plays an important
role in heterophase systems. It is the diffusion processes that determine in
large part whether or not ZnO layers will grow and whether the
ZnO/ZnSe(S,Te) heterojunction will have a sharp interface or the ZnO layer
will grow via oxygen diffusion into the crystal bulk and will have a broad
boundary. Solving Eqs. (8-11), one can determine, from the relationship
between the generation rates of A and B vacancies, the quasi-heteroepitaxy
parameters at which particular kinds of heterojunctions, conductivity types,
and defect concentrations in ZnO and ZnSe(S, Te) layers can be obtained.
Thus, the proposed method for analyzing the kinetics of defect formation,
based on the quasi-epitaxy model, clearly demonstrates the role of the
surface in the ADC equilibrium and allows one to assess the effects of
temperature and JB of the equilibration of defect concentrations. Computer
simulations of defect formation kinetics in II-VI crystal non-equilibrium
chalcogen vapor systems indicate that steady state defect concentrations in
the surface layer are reached very rapidly and, accordingly, are not
34
influenced by the diffusion of defects to the crystal bulk, as was assumed in

the model in question.
The approach in question allows one to produce a wide range of device
structures for opto- and acoustoelectronics, e.g.: ZnO/ZnSe, ZnS/ZnSe,
CdS/CdSe, and ZnO/ZnTe, with variable parameters. For example, ZnO
layers grown on ZnSe may be both n- and p-type.
Using this approach, we prepared light-emitting ZnO/ZnSe structures.10 A
typical 300-K electroluminescence (EL) spectrum (bias voltage of 46 eV)
of ZnO/ZnSe structures produced by annealing ZnSe at 870 K for 0.5 h in a
flow of atomic oxygen is shows strong bands at 467 and 525 nm and a weak
band centered around 615 nm. The following features of the EL warrant
attention: low injection voltages and the presence of an EL band near the
intrinsic edge. The external luminescence efficiency of the structures was
determined to be 0.1%.
At forward biases below 2.5 V, the curve shows a linear region. The
deviation from the exponential behavior at bias voltages above 2.5 V can be
accounted for by the volume resistance of zinc selenide. Therefore, the
barrier height at the pn junction is | 2.5 V. It follows from the I
I V data that
the pn junction lies within the zinc selenide layer. Therefore, a p-ZnO/p
/ ZnSe/n-ZnSe structure was produced. The ZnO serves as an injecting contact
to the p-ZnSe layer and as an exit window for radiation.
REFERENCES
1. A. N. Georgobiani, M. B. Kotlyarevsky, V. N. Mikhalenko, USSR Inventors Certificate
no. 684, 810 (1977).
2. N. K. Morozova and O. M. Morozova, Izv. Akad. Nauk SSSR, Neorg. Mater. 17, 1335
(1981).
3. M. T. Nyges, Cand. Sci. (Phys.Math.) Dissertation, Tallinn (1974).
4. V. A. Nikitenko, S. A. Stenli, and N. K. Morozova, Izv. Akad. Nauk SSSR, Neorg. Mater.
24, 1830 (1988).
5. V. N. Mikhalenko, B. P. Dementev, M. B. Kotlyarevsky, and A. N. Georgobiani, Izv.
Vyssh. Uchebn. Zaved., Fiz. 8, 150 (1978).
6. A. N. Georgobiani, M. B. Kotlyarevsky, and V. N. Mikhalenko, Tr. Fiz. Inst. im. P. N.
Lebedeva, Akad. Nauk. SSSR. 138, 79 (1983).
7. A. N. Georgobiani, M. B. Kotlyarevsky, and V. N. Mikhalenko, V.N., Izv. Akad. Nauk
SSSR, Neorg. Mater. 18, 12 (1982).
8. M. B. Kotlyarevsky, A. N. Georgobiani, I. V. Rogozin, and A. V. Marakhovskii, Zh. Prikl.
Spektrosk. 70, 86 (2003).
9. T. V. Butkhuzi, A. V. Bureyev, A. N. Georgobiani, N. P. Kekelidze, and T. G. Khulordava
J. Cryst. Growth 117, 366 (1992).
10. A. N. Georgobiani, M. B. Kotlyarevsky, V. V. Kidalov, and I. V. Rogozin, Neorg. Mater.
33, 232 (1997) [Inorg. Mater. (Engl. Transl.) 33, 185].
PART II: ELECTRICAL, OPTICAL, AND

STRUCTURAL PROPERTIES
Chapter 4
ELECTRICAL PROPERTIES OF ZnO
David C. Look1, Bruce B. Claflin1, Gene Cantwell2, Seong-Ju Park3, and
Gary M. Renlund4
1
Semiconductor Research Center, Wright State University, Dayton, OH 45435; 2ZN

Technology, 910 Columbia St., Brea, CA 92821; 3Dept. of Mat. Sci. and Tech., K-JIST,
J
Kwangju, 500-712, Korea; 4ON International, 418 W. Winchester St., Salt Lake City, U
UT
84107
Abstract:
Temperature-dependent Hall-effect measurements have been used to

determine donor and acceptor concentrations and energies in n-type and p-type
ZnO. Commonly observed donor energies in n-type material grown from the
vapor phase (VP) are 35 and 65 meV, with the 35-meV donor identified as H,
and the 65-meV donor, probably Al. Total donor concentrations in VP-grown
material are as low as mid-1016 cm-3. The only acceptors unambiguously
identified so far in this material are the Zn vacancy, at the low-1015-cm-3 level,
and substitutional NO, at a higher level but evidently passivated with H.
Acceptors that have been successfully used to create p-type ZnO by doping
include N, P, and As, and resistivities as low as 0.1 :-cm have been obtained.
The acceptor energy for N is about 90 meV for [N] ~ 1019 cm-3, and can be
estimated to be about 130 - 150 meV at low concentrations of N.
Key words:
ZnO, Hall-effect, donor, acceptor, n-type, p-type
1.
INTRODUCTION
ZnO has long been touted as a potential candidate for UV emitters, but
the lack of a good, reproducible p-type technology has hindered its
development in the past.1 Fortunately, in the last few years, much progress
has been made in this area and a few p-n homojunctions and heterojunctions
have even shown light emission.2 In spite of this recent success, much better
performance will be needed if ZnO is to compete with GaN-based lightemitting diodes (LEDs) in the blue and UV regions. One of the key required
elements will be the growth of bulk and epitaxial ZnO with low background
37
38
David C. Look et al.
concentrations of donors, so that p-type material can be produced with only

minimum initial compensation. The most important background donors that
must be considered are H, the O vacancy VO, the Zn interstitial ZnI, and the
group III elements, especially Al. Then, a way must be found to limit self
compensation during the acceptor doping process, so that the added
acceptors can efficiently produce free holes. The most suitable acceptor
dopants that have been discovered so far are the group V elements, N, P, As,
and Sb. To control the donor and acceptor species we must be able to
measure their concentrations and energies, and the only known method to
accomplish this task is temperature-dependent Hall-effect (TDH)
measurements.3 We will first discuss how such measurements are carried out
and analyzed, and then how they are applied to the problem at hand.
2.
HALL-EFFECT MEASUREMENTS AND

ANALYSIS
Few semiconductor laboratories in existence today are without a Halleffect system capable of room-temperature operation, at least. The only basic
requirements are a current source, a voltmeter, and a small magnet, of
typical field strength 0.1 1.0 T. If the current is imposed perpendicular to
the magnetic field, a voltage will be generated in the mutually perpendicular
direction, and this is the classical Hall voltage. The current, magnetic field
strength, and Hall voltage are all easily measurable quantities, and their
combined values immediately give the Hall coefficient R, where R is related
to the carrier concentration, n or p, by n = r/eR. Here e is the electronic
charge and r is the so-called Hall factor, which is usually close to unity but
also may be calculated by fitting the temperature-dependent Hall mobility
data. Indeed, the Hall mobility PH is simply defined as PH = RV, where V is
the conductivity, measured from the voltage drop parallel to the current. To
relate n and PH to physical parameters, such as the donor concentration, we
must solve the Boltzmann transport equation, which describes the relaxation
of carrier momentum via carrier scattering. If the various scattering
mechanisms are elastic, then a relaxation time W can be defined, and the
Boltzmann equation can be easily solved to give
PH
e W
m* W
(1)
Electrical properties of ZnO
39
where m* is the effective mass and Wn! is defined by
W ( )
wff 0
dE
wE
E
o
f
3/ 2 wff 0
E
dE
0
wE
E
W n ( E ) E 3/2
W n ( ) E 3/ 2 e- E / kT dE
E 3/ 2 e- E / kT dE
(2)
Here, E is the carrier energy and the second expression in Eq. 2 is valid for
Boltzmann (nondegenerate) statistics. We must note that W is the total
relaxation time, made up of contributions from various phonon and impurity
scattering mechanisms. The additive quantity is the relaxation rate, W-1, so
that the total relaxation rate is given by
W-1(E
( ) = Wac-1(E) + Wpe-1(E) + Wpo-1(E
(E ) + Wii-1(E) + Wdis-1(E)
(3)
where ac and po denote scattering by acoustical-mode and optical-mode

lattice vibrations, respectively, pe, the piezoelectric potential, ii, ionizedimpurities, and dis, dislocations. As an example
W ac
S ! 4 U d s 2 E -1/2
2
21/2 E1
3/ 2
kT
(4)
where Ud is the density, s is the speed of sound, and E1 is the deformation

potential. The other formulas can be found in Ref. 3, including an expression
for Wpo, which is only approximate because polar-optical scattering is not
really elastic. A more accurate treatment requires a full numerical solution of
the Boltzmann equation.3,4
Equations 1 4 can easily be set up on a PC and used to simulate or fit
temperature-dependent mobility data. In an n-type sample, the only
undetermined parameter is the acceptor concentration NA, so usually NA is
varied to give the best fit to the data. (For this fit, the approximate carrier
concentration, nH = 1/Re, can be used when n is required in the various
scattering formulas.) Then, the Hall factor r = W2!/W!2 can be calculated at
each temperature, and the true carrier concentration determined from the
expression n = r/eR. Next, n vs. T is fitted by use of the charge-balance
equation
n NA
N Di
1 n / I
i
(5)
Di
40
where i denotes a particular donor NDi and

I Di
g 0i
e
g 1i
D Di
k
N C 'T 3 / 2 e
ED 0i
kT
(6).
For each donor, g0/g1 is a degeneracy factor, NCc = 2(2Smn*k)3/2/h3 is the

effective conduction-band density of states at 1K, h is Plancks constant, ED
is the donor energy, and ED0 and DD are defined by ED = ED0 - DDT. The
above equation describes the simplest type of charge balance, in which each
of the one or more donors has only one charge-state transition within a few
kT of the Fermi energy. An example of such a donor is Ga on a Zn site in
ZnO. If there are double or triple donors, or more than one acceptor, proper
variations of Eq. 5 can be found in the literature3.
An example of a mobility fit for bulk ZnO grown by the VP method is
presented in Fig. 1. Here, the fitted acceptor concentration is NA = 2u1015
cm-3.
This value is then used in the charge-balance equation, Eq. 5, to fit the
carrier concentration data, as shown in Fig. 2. As seen, two donors are
2000
NA = 2 x 10
15
5
cm
-3
3
P (cm /V s)
1500
1000
500
0
100
200
300
T (K)
Figure 1. Temperature-dependent mobility of bulk ZnO.
400
10
41
15
Theo
ory: ND1=9
9 x 10
17
17
-3
n (cm )
16
10
15
-3
ND22=1 x 10
cm , ED2=61 meV;
NA= 2 x 10
cm , fro
om mo
obility
15
10
-3
cm , ED1=31 meV;
10
3
-3
15
20
-1
10 /T (K )
Figure 2. Temperature-dependent carrier concentration of bulk ZnO.
found, of concentrations 9u1015 and 1u1017 cm-3, and energies 31 and 61

meV, respectively. Their identities will be discussed in more detail below.
3.
BACKGROUND DONORS AND ACCEPTORS IN

ZnO
We first will discuss the identities of the acceptor and two donors found
in the VP-grown ZnO represented by Figs. 1 and 2. The acceptor was found
by positron annihilation to be the Zn vacancy VZn.5 This is somewhat
surprising in light of secondary-ion mass spectroscopic (SIMS) data on
similar material which shows that N is present at the 1017-cm-3 level.6
However, we believe that the N is passivated by H that is present during the
VP growth process,7,8 and that a further supply of H is responsible for the
shallow 31-meV donor, ND1. This assertion is supported by annealing
results, shown in Figs. 3 and 4. The annealing temperature, 715 qC, is high
enough to get rid of H, and indeed, the donor of shallow slope in Fig. 4 has
disappeared. Furthermore, the acceptor concentration has increased from
about 2u1015 cm-3 to about 1u1016 cm-3, as illustrated by the strong decrease
in low-temperature mobility shown in Fig. 3. Since H-N complexes have
42
been tentatively identified by infrared absorption experiments8, it is likely

that they exist in the as-grown material, and that the anneal breaks them
apart, thus activating the N acceptors. Further evidence for the annihilation
of the shallow donor comes from photoluminescence (PL) data, Fig. 5, which
shows the disappearance of a donor-bound exciton line at 3.363 eV. Recent
evidence has confirmed that this donor is related to H. Two-electron
transitions, appearing near 3.33 eV in Fig. 5, lead to a donor transition
energy of about 4/3(3.63 3.33) ~ 40 meV for H, close enough to the 31
meV measured by Hall effect measurements, especially considering that Hall
measurements in various samples give between 30 and 40 meV.
The deeper donor, having an energy of 61 meV in this sample but ranging
up to 70 meV in other samples, is probably AlZn. This identification is based
on the fact that Al is a common background impurity in ZnO, and that it
should be a simple, hydrogenic donor, i.e., with an energy of about 55 - 65
meV. Its D0X PL line has been shown, by doping experiments, to have an
energy of about 3.360 eV9, and such a line is clearly visible in Fig. 5.
2500
unannealed
anneal: 715 qC, 30 min
1500
P (cm /V s)
2000
1000
500
0
100
200
300
T (K)
Figure 3. Effects of annealing on mobility of bulk ZnO. The fits (not shown) give acceptor
concentrations of NA = 1.8u1015 cm-3 for the unannealed sample, and 1.2u1016 cm-3 for the
annealed sample.
43
10
16
10
15
10
14
-3
n (cm )
unannealed
annealed: 715 qC, 30 min
10
20
30
3
40
50
-1
10 /T (K )
Figure 4. Effect of annealing on carrier concentration of bulk ZnO. The fits (not shown) give
donor concentrations of 2.8u1016 and 3.7u1016 cm-3, and energies of 37 and 66 meV,
respectively, for the unannealed sample; and donor concentrations of 2.1u1016 and 2.8u1016
cm-3, and energies of 44 and 77 meV, respectively, for the annealed sample.
PL intensity (arb. units)
10
8
unannealed
annealed 715 qC
3.3628 eV
3.3329
6
4
x10
3.3596
3.3484
3.3296
3.3566
3.33
3.34
3.35
3.36
E (eV)
Figure 5. Effect of annealing on 4-K photoluminescence of bulk ZnO.
44
4.
P-TYPE ZnO
The difficulty in producing p-type ZnO has hindered the development of

ZnO-based p-n junctions, and thus ZnO UV LEDs and LDs. It wasnt until
1997 that the first p-type ZnO was reported10, and further reports on this
subject were sparse until 2003. (A recent review on this subject has been
given in Ref. 2). But now (2004), several groups have obtained p-type ZnO
with N doping, and a few, with P, As, and Sb doping. As an example, we
show temperatrure-dependent p and Pp curves for As-doped ZnO in Fig. 6.
20
10
5
theory
P (cm /V s), p (10
17
-3
cm )
50
Pp
2
1
100
200
300
T (K)
Figure 6. Hole concentration p and mobility Pp for an As-doped ZnO sample. The solid line is
a theoretical fit to the mobility.
Here, the mobility Pp is fitted (solid line) with a very simple, singlevalence-band model, assuming an effective mass of 0.64 m0. The only fitting
parameter is the donor concentration, which turns out to be ND = 8u1019 cm3
. Although the temperature dependence of the hole concentration p cannot
be easily fitted, because it is partially degenerate, we can estimate that p |
NA ND | 1019 cm-3 at high temperatures, giving NA | 9u1019 cm-3. Indeed,
SIMS results show an As concentration of mid-1019, consistent with As
being the dominant acceptor in this material. The fact that As can be doped
to such a high concentration is quite surprising, because, unlike the NO case,
As is much larger than O, and should not substitute for it readily. Some
theoreticians have recently suggested that As does not enter the lattice
simply as AsO, but as a much more complex acceptor-type defect, AsZn-
45
2VZn11. Although this defect is consistent with our data, still it must be
confirmed by more direct means.
The most studied acceptor dopant up to now, by far, is N, and this
element has been shown by electron paramagnetic resonance data to go
predominantly onto the O site and have acceptor nature12,13. It has an
acceptor transition at ~ 90 meV, for [N] ~ 1019 cm-3, and this energy can be
estimated to be about 130 150 meV at infinite dilution14. Thus, it is perhaps
not surprising that p-type ZnO can be produced with N doping;
however, it is still somewhat of a mystery as to why As, P, and Sb also work.
In any case, the recent success in creating p-type ZnO bodes well for the
future of the ZnO LED industry, although much more research will be
necessary before the situation is well understood.
5.
SUMMARY
Electrical measurements, mainly temperature-dependent Hall-effect

measurements, have been critical in the elucidation of donors and acceptors
in ZnO. The main background donors in state-of-the-art VP-grown ZnO
have been shown to be H and Al, and the acceptor, VZn. Other possible
donors that must be investigated further are the defects VO and ZnI.
Although VZn seems to be the main electrically active acceptor, still N is
evidently present at a much higher concentration. If this is true, then N must
be passivated, and the likely passivator is H. Indeed, annealing experiments
lead to a much higher acceptor concentration, presumable due to the
breaking of the H-N bond. Deliberate doping of N can produce [N] ~ 1019
cm-3 or higher, and p-type ZnO has been produced in this way. Surprisingly,
however, even P, As, and Sb have been successfully used for p-doping.
Research into these matters is proceeding at a rapid pace, and the next few
years should produce a much clearer understanding.
REFERENCES
1. D. C. Look, Mater. Sci. and Eng. B 80, 383 (2001).
2. D .C. Look and B. Claflin, phys. stat. sol. (b) 241, 624 (2004).
3. D. C. Look, Electrical Characterization of GaAs Materials and Devices (Wiley, New
York, 1989).
4. B. R. Nag, Electron Transport in Compound Semiconductors (Springer, Berlin, 1980).
5. F. Tuomisto, K. Saarinen, and D. C. Look, Phys. Rev. Lett. 91, 205502 (2003).
6. D. C. Look, D. C. Reynolds, C. W. Litton, R. L. Jones, D. B. Eason, and G. Cantwell,
Appl. Phys. Lett. 81, 1830 (2002).
46
7. N. Y. Garces, L. Wang, N. C. Giles, L. E. Halliburton, G. Cantwell, and D. B. Eason, J.

Appl. Phys. 94, 519 (2003).
8. N. H. Nickel and K. Fleischer, Phys. Rev. Lett. 90, 197402 (2003).
9. H. Shibata, M. Watanabe, M. Sakai, K. Oka, P. Fons, K. Iwata, A. Yamada, K. Matsubara,
K. Sakurai, H. Tampo, K. Nakahara, and S. Niki, phys. stat. sol. (c) 1, 872 (2004).
10. K. Minegishi, Y. Koiwai, and Y. Kikuchi, K. Yano, M. Kasuga, and A. Shimizu, Jpn. J.
Appl. Phys. 36, L1453 (1997).
11. S. Limpijumnong, S. B. Zhang, S. H. Wei, and C. H. Park, Phys. Rev. Lett. 92, 155504
(2004).
12. W. E. Carlos, E. R. Glaser, and D. C. Look, Physica B 308-310, 976 - 979 (2001).
13. N. Y. Garces, N. C. Giles, L. E. Halliburton, G. Cantwell, D. B. Eason, D. C. Reynolds,
and D. C. Look, Appl. Phys. Lett. 80, 1334 - 1336 (2002).
14. D. C. Look, B. Claflin, Ya. I. Alivov, and S. J. Park, phys. stat. sol. (a) (in press).
Chapter 5
ELECTRICAL PROPERTIES OF ZnO THIN
FILMS AND SINGLE CRYSTALS
M. Grundmann1, H. von Wenckstern1, R. Pickenhain1, S. Weinhold1, B.
Chengnui1, and O. Breitenstein2
1
IInstitut fr Experimentelle Physik II, Fakultt fr Physik und Geowissenschaften, Universitt

Leipzig, Linnstr. 5, D-04103 Leipzig, Germany; 2Max-Planck-Institut fr Mikrostrukturphysik, Experimental Department II, Weinberg 2, D- 06120 Halle / Saale
Abstract:
We have investigated the electrical properties of ZnO single crystals and thin
films. Low resistive ohmic contacts were realized by annealing thermally
evaporated Ti-Au contacts at 400C in vacuum. Hall effect measurements
provided information about dominant scattering mechanisms in the single
crystals and the epitaxial thin films. Dominant donor levels were found in an
energy range from 37 meV to 350 meV below the conduction band for the
different samples. Schottky contacts of good quality ((j(-5V) = 4 x 10-5 A/cm2, n
= 1.3) were produced on thin films after etching with HNO3. Deep level
transient spectroscopy yielded the activation energy of deep traps for single
crystals of 300 meV and 540 meV and for the thin films of 300 meV,
respectively.
Key words:
Hall effect, schottky contacts, DLTS, thermography, PLD
1.
INTRODUCTION
The wide band gap semiconductor ZnO has attracted much attention in
recent years. This is stimulated by the development of transparent
electronics, the growing need of emitters and detectors working in the ultraviolet range of the electromagnetic spectrum, and the progress in the
realization of ferromagnetic semiconductors with Curie temperatures close
to or above room temperature. ZnO is nowadays applied as transparent front
contacts of solar cells competing with the commonly used material indium
tin oxide (ITO). The low cost of ZnO as well as the radiation hardness1 make
47
48
M. Grundmann et al.
this material favorable compared to ITO. Material properties as the high

exciton binding energy of 61 meV2, the comparatively high optical gain3,
unproblematic wet-chemical etching4, or the widely tunable band gap by
alloying with CdO or MgO5 give ZnO an extraordinary role compared to
other wide band gap semiconductors. The possibility of low-temperature
fabrication (300 800 K) makes it favorable compared to GaN. The
availability of ZnO substrates and the growing quality of ZnO thin films
have put forward systematic studies. We will focus in this article on the
electrical properties of commercially available ZnO single crystals and
heteroepitaxially grown ZnO thin films produced by pulsed laser deposition
(PLD).
2.
SAMPLES
We investigated three different kinds of ZnO single crystals. One kind

was produced by seeded chemical vapor transport (SCVT) by Eagle Picher
Technologies. Two single crystals analyzed here were grown by
hydrothermal vapor phase epitaxy (HVPE) one of them purchased from
Crystec (labeled from now on HVPE I) the other was obtained from MaTeck
(referred to as HVPE II). The thickness of these single crystals is 500 m. In
addition we probed samples grown by PLD on sapphire substrates. For that
we used the 248 nm line of a KrF excimer laser to ablate polycrystalline ZnO
targets produced by standard ball milling, pressing, and sintering. After the
growth of a low temperature nucleation layer a about 1 m thick ZnO layer
was deposited at about 750C at an oxygen partial pressure of 0.016 mbar.
Details about the growth and the properties of the ZnO thin films are
published in.5-7
3.
OHMIC CONTACTS ON THIN FILMS
The contact resistivities Uc of different ohmic contacts were determined

with the transmission line model. We investigated Ti, Al, and Ti-Au ohmic
contacts produced by thermal evaporation of the contact metal. Ti contacts
result in poor ohmic contacts with Uc = 3.2 :cm2. Annealing these contacts
at 300C doubles Uc due to the formation of titanium oxide at the interface.
The contact structure becomes insulating after annealing at temperatures of
about 600C. As-deposited Al contacts have with Uc = 0.14 :cm2 a lower
contact resistivity than Ti-contacts. Annealing these contacts at 700C
decreases Uc to 0.028 :cm2 due to the in diffusion of Al forming a highly
doped surface layer that electrons may pass by tunneling. The lowest contact
Electrical properties of ZnO thin films and single crystals
49
resistivity was obtained for Ti-Au contacts. As-deposited contacts have Uc =

0.01 :cm2 which decreases after annealing in vacuum at about 400C to Uc =
2.4 x 10-3 :cm2. If the contacts are annealed in a N2 ambient Uc does not
change compared to the as-deposited contacts. Annealing at higher
temperatures increases Uc as depicted in Fig. 1.
0.008
as-deposited
400C 3min vacuum
400C 3min N2
600C 10min N2
0.006
0.004
Current (A)
0.002
0.000
-0.002
-0.004
-0.006
-0.008
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
Voltage (V)
Figure 1. Current-voltage characteristic of as-deposited and annealed Ti-Au contacts
4.
LOCK-IN THERMOGRAPHY ON THIN FILMS
In lock-in thermography8 the surface temperature of a sample is imaged

by an infrared (IR) camera and Joule heat is periodically created by injecting
a pulsed current flow. The output signal of the IR camera is digitally
processed according to the lock-in principle to convert the periodic surface
temperature modulation into an image of the local temperature modulation
amplitude. This amplitude image can be interpreted to visualize the local
dissipated power density. If in a homogeneous conducting thin film with
sheet resistivity of Rs a local current density of j(x,y) flows, the local
dissipated power density is p(x,y) = Rs (j(x,y))2. If the current flows between
two contacts in opposite corners of a square Hall sample, the current density
decreases in the vicinity of the contacts proportional to the distance to the
contact. Hence, we expect an increase of the lock-in thermography signal
towards the current injecting contacts as shown in Fig. 2a. This pattern is the
same if the injecting contacts are exchanged, or if the sample is rotated by
50
M. Grundmann et al.
90. Any deviation of the C2v-symmetry is an indication of a spatial

inhomogeneity of the electric properties of the conducting film.
We found, that the current paths are usually not homogeneous for
samples grown using a fixed substrate holder. Often the current flows along
an edge indicating regions with lower resistance (Fig. 2b). Such kinds of
samples are not appropriate for Hall measurements in the van der Pauw
geometry. A substantial improvement of the reliability of Hall measurements
of PLD thin films in the van der Pauw geometry is obtained if the outer 2
mm of the sample are mechanically removed. The current paths show the
desired behavior after the removal (Fig. 2c) and Hall measurements on
samples prepared in that manner like the PLD thin films investigated in this
study are reliable. If the substrates are mounted to a rotating substrate holder
as-grown samples can be used directly (Fig. 1a).
Figure 2. Lock-in thermography images of PLD sample grown with a) rotating substrate
holder, b) fixed substrate holder, and c) fixed substrate holder but the outer 2 mm of the
sample where removed. In (a, c) the circular contacts are in the top and bottom corners. In part
(b) the contacts are stripe-like at the top and the bottom edge.
5.
HALL MEASUREMENTS ON SINGLE

CRYSTALS AND THIN FILMS
We investigated the transport properties of ZnO single crystals and PLD

thin films by Hall measurements using the van der Pauw method.9
5.1
Room Temperature Measurements on thin films
The Hall mobility H of the thin films is highest within a narrow region of
free electron concentrations n ranging from about 2 to 5 x 1016 cm-3 as
depicted in Fig. 3. In general, the Hall mobility of the thin films is lower
51
Figure 3. Dependence of the Hall mobility on the free electron concentration.
than expected by considering intrinsic scattering mechanisms and ionized

impurity scattering only. The reason is grain boundary scattering, which
occurs at potential barriers at the grain walls. The potential barriers arise
from the occupation of acceptor states at the grain walls leading to depletion
zones around these walls. Free carriers have to overcome these barriers,
which reduces the mobility. The shape of Fig. 3 can be explained by
considering the dependence of grain boundary scattering on the grain size
and on the density of donors (which determines the depletion width, if the
density of traps at the grain walls is assumed to be in average similar for all
samples). As the carrier concentration increases the grain size decreases
leading to a larger number of scattering events and with that to a reduction of
the mobility. For low carrier concentrations the depletion width around the
grain walls increases and the amount to which a grain is depleted increases.
The peak in Fig. 3 corresponds to a regime where these two processes are
comparatively small and grain boundary scattering is least important for the
investigated thin films. The maximum Hall mobility of 155 cm2/Vs defines
the present state of the art for heteroepitaxial grown thin films on c-plane
sapphire.
5.2
Temperature dependent measurements on single

crystals and thin films
To model the temperature dependence of the Hall data we first fitted the
Hall mobility accounting the intrinsic scattering mechanisms piezoelectric
scattering and deformation potential scattering on longitudinal acoustic
phonons and polar optical scattering (for them we used material constants, no
fitting parameters are needed for the intrinsic scattering mechanisms) as
52
M. Grundmann et al.
103
H (cm/Vs)
SCVT
HVPE II
HVPE I
thin film
102
10
20
40
60 80 100
200
400
T (K)
Figure 4. Hall mobility of single crystals and a thin film.
well as the extrinsic scattering at ionized impurities containing the fitting

parameter NA the concentration of compensating acceptors. For the thin films
we also accounted grain boundary scattering containing two more fitting
parameters, namely the barrier potential EB and the grain size LG.
Matthiesens rule was applied for the fits and the obtained density of
compensating acceptors was used as input for the fit of n(T).
Figure 4 shows the Hall mobility of the single crystal grown by SCVT, of
the two hydrothermally grown single crystals (HVPE I and HVPE II), and a
PLD thin film. The single crystal grown by SCVT has the highest H at room
temperature. This is due to a low concentration of compensating acceptors.
The samples HVPE I and HVPE II are compensated for temperatures lower
than 240 K. This originates in higher values of NA that explains also the
lower values of H at room temperature. The Hall mobility of the thin films
has a different temperature dependence compared to the single crystal grown
by SCVT. The reason is the additional scattering at the grain boundaries this
sample contains. From the fit we find that grain boundary scattering is the
dominant scattering mechanisms in thin films up to room temperature.
A one donor charge balance equation considering compensating acceptors
was used to model n(T). Figure 5 depicts the experimental data and the
corresponding fits. The shallowest dominating donor level with a density of
ND = 1.3 x 1017 cm-3 is found in the single crystal grown by SCVT. The
thermal activation energy of about 37 meV is due to hydrogen donors at
interstitial site. Effective mass donors as Al and Ga with lower concentration
53
1017
1016
SCVT
thin film
1012
1014
n (cm 3)
n (cm 3)
1015
1013
HVPE I
1011
1012
1011
1010
HVPE II
1010
3.00
3.25
HVPE II HVPE I
2
3.50
3.75
1000/T (1/K)
10
12
4.00
14
1000/T (1/K)
Figure 5. Free electron concentration (markers) and corresponding fits (lines) of single
crystals and a thin film. The inset shows n(T) of HVPE I and HVPE II.
are also present in this sample as photoluminescence (PL) measurements

suggest. The dominant donor in the HVPE grown samples is much deeper
than the one in the SCVT grown sample. The activation energies are 299
meV or 352 meV for the samples HVPE I and HVPE II, respectively. An
electron trap with an activation energy of 300 meV was also found in the
SCVT grown single crystal and in PLD thin films by deep level transient
spectroscopy as discussed in 6.2. The existence of this level in samples
grown by three different methods suggests that it is an intrinsic defect
probably the oxygen vacancy. The thin film has a donor level 65 meV below
the conduction band which corresponds to the expected thermal activation.
Table 1. Parameters obtained from fits of the temperature dependent Hall measurements for
single crystals and a PLD thin film.
SCVT
HVPE I
HVPE II
PLD thin film
ND
13 r 2
19 r 0.6
12 r 0.6
6r1
(1016 cm-3)
ED (meV)
37 r 2
299 r 5
352 r 6
65 r 2
NA
0.3 r 0.07
18 r 0.04
8 r 0.04
0.15 r 0.06
(1016 cm-3)
J
0.021
0.945
0.69
0.024
LG (nm)
40 r 5
13 r 2
EB (meV)
54
M. Grundmann et al.
energy of effective mass donors. The most probable cause of this level is Al
diffusing during growth from the sapphire substrate in to the thin film. The
lowest compensation ratio J of about 0.02 have the single crystal grown by
SCVT and the PLD thin film. This is a very low ratio indicating the high
quality of present state of the art grown ZnO. The parameters of the Hall
analysis are summarized in table 1.
6.
SCHOTTKY CONTACTS ON SINGLE CRYSTALS

AND THIN FILMS
Schottky contacts on ZnO were realized by the thermal evaporation of

Ag, Au, Ni, or Pd, respectively. We used different surface preparation
techniques prior to the deposition of the contact metal. For the single crystals
a front-back contact configuration was used while a front-front configuration
has to be used for thin films grown on insulating sapphire substrates. The
homogeneity of the Schottky contacts depends on the surface preparation as
revealed by electron beam induced current (EBIC) measurements (Fig. 6).
Figure 6. Electron beam induced current scans on a) a sample treated with N2O plasma and b)
a sample treated with HCl prior to the metal deposition.
The dependence of the Schottky barrier height on the metal work function
is weak10 and the ideality factors of the diodes are far from ideal indicating
the presence of surface states, recombination in the space charge region, or
diffusion limited currents. Etching the surface with HNO3 leads to Schottky
contacts of good quality on thin films as we found recently. The current
density at 5V is about 4 x 10-5 A/cm and the ideality factor determined for
voltages between 0.05 V and 0.22 V is 1.3. The current density vs. voltage
for this diode is depicted in Fig. 7.
55
10-3
jd (A/cm)
10-4
10-5
10-6
10-7
10-8
-5
-4
-3
-2
-1
Ud (V)
Figure 7. Current density in dependence on the voltage for a Schottky diode realized on a ZnO
PLD thin film with Pd after etching with HNO3.
6.1
Photocurrent Measurements on Single Crystals
We have measured the photocurrent of a Pd Schottky diode on a single

crystal grown by SCVT in a temperature range from 4 K to 300 K. The
spectra are shown in Fig. 8. The band edge shifts to lower energies as the
temperature increases. For T < 175 K the spectra exhibit two peaks in the
near band edge region. Their intensity ratio changes with temperature. The
peak lower in energy is dominant for T > 50 K (this is the temperature where
both peaks have about the same intensity) and for T < 50 K the higher
energetic peak is most intense. To explain this behavior further investigations
are on the way.
We have also realized interdigital metal-semiconductor-metal (MSM)
contacts. We found a strong increase of the current through the structure by
illuminating it with UV-light. If the stability under reverse bias of regular
Schottky diodes (see Fig. 7) is achieved for interdigital structures we are
confident that MSM structures are well suited for the realization of solarblind UV-detectors.
6.2
Deep Level Transient Spectroscopy on single crystals

and thin films
Deep level transient spectroscopy (DLTS) measurements were carried out

between 4 K and 300 K. We investigated single crystals grown by SCVT
56
M. Grundmann et al.
80
photocurrent (arb. u.)
70
300 K
60
175 K
50
40
50 K
30
20
4K
10
3.2
3.3
3.4
3.5
3.6
energy (eV)
Figure 8. Photocurrent of a Pd Schottky contact on a single crystal grown by SCVT.
and PLD thin films. The series resistance of the samples HVPE I and II is too
large for measuring DLTS with a capacitance bridge working at 1 MHz
which we have at our disposal. We applied rate windows between 64 and
10725 Hz and found two traps in the single crystal and one trap in the thin
film. Arrhenius plots of the temperature shift of the peak maximums with
changing rate window are shown in Fig. 9. The slope in this plot corresponds
0
-1
SCVT
thin film
ln(en/T2)
-2
-3
-4
-5
-6
-7
-8
3.0
3.5
4.0
4.5
5.0
5.5
6.0
1000/T (1/K)
Figure 9. Arrhenius plot for deep levels found in the single crystal grown by SCVT and the
thin film
57
to the activation energy Et. A defect with Et = 300 meV is found in both
samples and was also found in HVPE I and II by the TDH experiments. The
existence of the defect in all samples suggests that it is an intrinsic defect
likely the oxygen vacancy. A deeper trap with an activation energy of 540
meV is contained in the SCVT sample.
7.
SUMMARY
In summary we have reviewed the progress in the characterization of

electrical properties of ZnO. Improved ohmic and Schottky contacts have
been fabricated, the latter suitable for depletion layer spectroscopy. The
shallow and deep donor levels have been identified for ZnO from various
sources. Further control of deep donors seems necessary for achieving pconductivity.
REFERENCES
1. D. C. Look, Mater. Sci. Eng., B 80, 383 (2001).
2. R. Schmidt, B. Rheinlnder, M. Schubert, D. Spemann, T. Butz, J. Lenzner,E. M.
Kaidashev, M. Lorenz, A. Rahm, H. C. Semmelhack, and M. Grundmann, Appl. Phys.
Lett. 82, 2260 (2003).
3. E. M. Wong and P. C. Searson, Appl. Phys. Lett. 74, 2393 (1999).
4. M. J. Vellekoop, C. C. G. Visser, P .M . Sarro, A. Venema, Sensors and Actuators, A21A23, 1027 (1990).
5. M. Lorenz, E. M. Kaidashev, H. v. Wenckstern, V. Riede, C. Bundesmann, D. Spemann,
G. Benndorf, H. Hochmuth, A. Rahm, H.-C. Semmelhack und M. Grundmann, Solid-State
Electronics 47, 2205 (2003).
6. C. Bundesmann, M. Schubert, D. Spemann, T. Butz, M. Lorenz, E. M. Kaidashev, N.
Ashkenov, H. Neumann, and M. Grundmann, Appl. Phys. Lett. 81, 2376 (2002).
7. E. M. Kaidashev, M. Lorenz, H. v. Wenckstern, A. Rahm, H.-C. Semmelhack, K.-H.
Han,G. Benndorf, C. Bundesmann, H. Hochmuth und M. Grundmann, Appl. Phys. Lett. 82,
3901 (2003).
8. O. Breitenstein, M. Langenkamp: Lock-in Thermography - Basics and Use for Functional
Diagnostics of Electronic Components, Springer-Verlag, Berlin, 2003.
9. L. J. van der Pauw, "A Method of Measuring Specific Resistivity and Hall Effect of Discs
of Arbitrary Shape", Philips Research Reports 13, No. 1, pp. 1-9 (1958).
10. H. von Wenckstern, E. M. Kaidashev, M. Lorenz, H. Hochmuth, G. Biehne, J. Lenzner, V.
Gottschalch, R. Pickenhain, and M. Grundmann., Appl. Phys. Lett. 84, 79 (2004).
Chapter 6
STRUCTURE, MORPHOLOGY, AND
PHOTOLUMINESCENCE OF ZnO FILMS
V. A. Karpina1, V. D. Khranovskyy1, V. I. Lazorenko1, G. V. Lashkarev1, I.
V. Blonsky2 , and V. A. Baturin3
1
IInstitute for Problems of Material Science, National Academy of Sciences of Ukraine,3,

Krzhyzhanovsky str., Kyiv, 03142, Ukraine, 2Institute of Physics, National Academy of
Sciences of Ukraine, 46, Prospect Nauky, Kyiv, 03650, Ukraine, 3IInstitute of Applied Physics,
National Academy of Sciences of Ukraine, 40034, 58 Petropavlowska str., Sumy, Ukraine
Abstract:
Thin films of ZnO were deposited by reactive thermal evaporation, r.f.

magnetron sputtering and PEMOCVD. Crystal structure, surface morphology
as well as ultraviolet photoluminescence at optical excitation were
investigated. It is determined that ZnO films especially deposited by reactive
thermal evaporation exhibit strong ultraviolet luminescence. We suggest that
observed luminescence is obliged to recombination of excitons bound on two
different defects.
Key words:
PEMOCVD, magnetron sputtering, AFM, photoluminescence
1.
INTRODUCTION
ZnO is a wide band gap semiconductor, which is used for various

applications. Based on textured ZnO films one can build highly effective
piezo field emitters. On the other hand ZnO is a very effective electronexcited phosphor. ZnO films easily withstand electron fluence more than 1
W/cm2.1 ZnO films doped with Al, Ga, or In have a low resistivity of about
10-4 :cm and a high transparency of about 90%. This is sufficient for
applications as a front contact in solar cells, liquid crystal displays etc.
Dielectric ZnO films have a high electromechanical coupling factor that
allow using ZnO in various surface acoustic wave (SAW) devices such as
delay lines, delay-line filters, resonators, transducers and SAW convolvers.
59
60
V. A. Karpina et al.
Besides that, ZnO has some interesting properties, which make it very
perspective for functional devices nowadays. ZnO is recognized as a
promising optoelectronic material. Due to a large exciton binding energy the
effective excitonic generation in ZnO could be achieved at and above room
temperature. Indeed, optically pumped stimulated emission and lasing have
been demonstrated at room temperature.4-6 There are papers devoted to the
deposition of p-type ZnO.7,8 Hence, we can expect that effective lightemitting devices based on ZnO will be available in the near future.
As semiconductor ZnO can be used for detection of UV-A (300-400 nm)
radiation. By doping with Mg the range can be tuned to the UV-B and UV-C
(200-280 nm) regions. Such a wide interval of sensing spectra makes
ZnxMg1-xO UV detectors to be used in many applications such as solar UV
radiation monitoring, ultra-high temperature flame detection, and airborne
missile warning systems. There are reports on the design of photoconductive
as well as Schottky types of ZnO UV detectors.2,3
Single crystal ZnO films have a strong optical anisotropy and can be used
for modulation of UV radiation. ZnO modulators with a contrast of 70:1 and
an operation speed of 100 ps have been designed already.9
ZnO appears to be a ferroelectric material upon doping with non-central
ions such as Li+, Mg2+.10,11 The Curie temperature can be increased to 340 K
when ZnO is simultaneously doped with Li and Mg. Recently, ferroelectric
thin films have been used for random access memory and other integrated
ferroelectric devices. So, ZnO can be one of the candidates for such systems.
ZnO is recognized as the best candidate for diluted magnetic
semiconductor in the recently emerged new field of electronics spintronics.
Spintronics (spin + electronics) focuses on the creation of functional devices
based on the control of ferromagnetic ordering by an electric field or light.
According to several theoretical predictions ZnO doped with transition
metals is expected to be ferromagnetic with a high Curie temperature.12,13
Therefore, a transparent ferromagnetic could be created. This shows that
ZnO is a promising material for the development of a variety of functional
devices.
2.
EXPERIMENTAL PROCEDURE
ZnO films were obtained by evaporation of zinc metal from a special cell
in oxygen atmosphere. The deposition was carried out on mica and glass
substrates. The deposited films were annealed at 400 C in air and at higher
temperatures. For RF magnetron sputtering we used pure Zn as a target and a
gas mixture of Ar+O2 for the working gas. These samples were deposited on
glass and sapphire substrates. Zinc acetylacetonate Zn(AA)2 was used as a
Structure, morphology, and photoluminescence of ZnO films
61
precursor for PEMOCVD depositions. The base system was equipped by

additional device with a resistive evaporator, to which small portions of
Zn(AA)2 powder were introduced periodically. The argon carrier gas
transports the vapor into the chamber that contains an oxygen-argon plasma.
The Zn(AA)2 vapor is decomposed under plasma discharge and reacts with
oxygen. Polycrystalline ZnO thin films were deposited at substrate
temperatures in a range of 150 C - 300 C. The crystal structure was
examined by XRD analysis and the surface morphology by AFM and SEM.
UV photoluminescence was measured at high excitation with a nitrogen
laser (337 nm, 10 ns, 100 Hz). The measurements were carried out at room
temperature and at 77K.
3.
RESULTS AND DISCUSSION
3.1
Structure and morphology of ZnO films
Fig. 1 shows XRD spectra of ZnO films deposited by reactive thermal

evaporation (RTE), magnetron sputtering (MS) and PEMOCVD. As can be
seen from Fig. 1(a) ZnO films deposited by RTE have no preferred crystallite
orientation. It may be assumed that the ZnO films are not textured due to a
high growth rate of 0.3-0.5 m/min. Att such deposition rates ZnO films have
excess of Zn and their structure is not perfect. Thus, these films require
subsequent annealing. On the contrary, ZnO films deposited by MS (Fig
1(b)) can be textured under certain conditions with the c-axis perpendicular
to the substrate. It is known that increasing substrate temperature leads to
high quality textured ZnO films. In our experiments we obtain highly
oriented ZnO films deposited by MS at 300 C and 20% O2 for films
deposited on glass substrate and at 200 C, 20% O2 for films deposited on
sapphire substrates. Thin films, prepared by PEMOCVD showed some
texture (Fig 1(c)), and the optimal growth conditions the following: TS = 200
- 250 C, Wp = 300 W, QAr = 40 ml/min, P = 0.3 Torr.
Fig. 2 represents the surface morphology of ZnO films deposited by RTE
and MS. It can be seen that films deposited by RTE have a large grain size of
0.7-1 m. The samples deposited by MS have a smoother surface. AFM
images of these films are shown on Figs. 3 and 4. ZnO films deposited by
MS have a maximum grain size of about 100 nm and an average roughness
of 2 nm. As the substrate temperature increases within 100-300 C the
average roughness increases from 0.7 to 3 nm. Simultaneously, the average
grain size increases from 40 to 60 nm. AFM analysis demonstrates that the
grain size of ZnO films deposited by PEMOCVD in the same temperature
62
Figure 1. XRD spectra of ZnO/glass films deposited by RTE (a), MS (b), PEMOCVD (c).
ZnO films deposited by RTE and PEMOCVD were annealed at 500 C for two hour in
ambient air.
range amounts to 25 75 nm for Ts=200 C, 30 60 nm for Ts=250 C, and

25 - 70 nm for Ts=300 C. The average roughness was minimal at Ts=250 C
and changes slightly with increasing temperature. From it we conclude that
that Ts=250 C is the optimal temperature to obtain high quality ZnO films
using PEMOCVD.
a)
b)
Figure 2. SEM images of ZnO deposited by RTE (a) and MS (b) and annealed at 500 C.
63
Fig. 1
Figure 3. AFM images of ZnO films deposited by MS at Ts=150 C (a) and Ts=300 C (b).
64
Fig. 2
Figure 4. AFM images of ZnO deposited by PEMOCVD at Ts=200 C (a), Ts=250 C (b), and
Ts=300 C (c).
3.2
Optical properties and UV photoluminescence
All deposited ZnO films show an optical transmittance of 65-85% in the

range of 400 800 nm. High transparency and an abrupt edge of absorption
(at 380 nm) indicate a high optical quality of ZnO films. Fig. 5 shows RT PL
spectra of ZnO films deposited by MS and RTE. It should be noted that we
have observed UV PL in all deposited ZnO films only after thermal
annealing at 500 C or at higher temperatures. It is surprising that ZnO films
deposited by RTE exhibit a stronger UV PL than ZnO films deposited by
MS. The excitation density at which we clearly observed UV PL was not
higher than 1.2 MW/cm2 for films deposited by RTE and 9 MW/cm2 for
films deposited by MS. PL spectra taken at 77 K under the same excitation
source are shown on Fig. 6(a). Fig. 6(b) shows the PL spectrum at low
65
excitation using a cw He-Cd laser (325nm, 6 mW). Super linear integrated

PL intensity was observed (Fig. 6(c)).
1,0
9.5 MW/cm
m
PL intensity,a.u.
0,8
0,6
0,4
0,2
0,0
1.2 MW/cm
m
-0,2
350
400
450
500
550
600
Wavelength,nm
a)
4 .4 M W /c m
3 ,5
PL intensity, a.u.
3 ,0
2 ,5
2 ,0
1 ,5
1 ,0
1 .2 M W /c m
0 ,5
0 ,0
-0 ,5
340
360
380
400
420
W a ve le n g th , n m
b)
Figure 5. RT PL spectra of ZnO films deposited by MS on sapphire (a) and RTE on mica (b).
ZnO films were annealed at 500 C.
66
77 K
2.5
10
77 K
8
Io
1.5
PL intensity, a.u.
PL Intensity, a.u.
2.0
0.5Io
0.1Io
1.0
0.5
0.0
6
4
2
0
360
380
400
420
350 360 370 380 390 400 410 420
Wavelength (nm)
Wavelength (nm)
(b)
(a)
Integrated PL intensity
35
30
25
20
15
10
5
0
10
2
excitation density (M W /cm )

(c)
Figure 6. PL spectra of ZnO films deposited on mica by RTE and annealed at 500 C (a), at
high excitation by an N2 laser (b), and at low excitation with a He-Cd laser (c). PL integrated
intensity at N2 excitation.
At low intensity there are two PL lines located at 369 and 374 nm (Fig.
6a). As the excitation intensity increases the intensity of the peaks also
increases. These two bands widen and overlap. Besides, the long-wave part
of the spectra strongly widens. At low excitation using a He-Cd laser we
67
observed three lines (Fig. 6b). Two lines are located at the same positions
and the third line is positioned at 382.5 nm. The energy difference between
the second and third line is 73 meV, what corresponds to LO phonon replica.
Whereas we observed a broadening of the line at increasing of pumping
intensity it could be supposed that stimulated emission with simultaneously
generation of acoustic phonons occurs. Besides that we also suggest that
twofold line is obliged to the recombination of excitons bound on two
different types of defects.
Finally, we have measured the time-dependent behavior of the PL
spectra at 77 K. At a delay time of 0-2 ns we observed strong UV PL (Fig.
7). At steady state conditions a deep-level yellow-red emission connected
with defect luminescence is observed. So it could be concluded that exciton
recombination is a faster process than recombination via deep levels.
normalized PL intensity
1.2
1.0
1 ns
steady state
0.8
0.6
0.4
0.2
0.0
300
400
500
600
700
800
Wavelength (nm)
Figure 7. Normalized time-dependent PL spectra of ZnO films deposited by PEMOCVD.
4.
SUMMARY
ZnO films were deposited by RTE, MS, and PEMOCVD. In the case of
RTE the ZnO films have a large-grained structure with no preferred
crystallite orientation. On the contrary, ZnO films deposited by PEMOCVD
and MS have fine-grained structure with an average grain-size of 40-60 nm.
Under certain growth conditions ZnO films show a preferred orientation with
the c-axis normal to the substrate. However, ZnO films prepared by
68
RTE have larger RT UV photoluminescence. The UV PL peaks at 383 nm at

RT. At 77K this peak splits into two located at 369 and 374 nm. Their
intensity increases super linearly with increasing excitation. We suggest that
the observed UV PL is due to recombination of excitons bound to two
different centers. A long-wave tail is observed at high-level excitation. Such
behavior testifies to the lasing effect accompanied by the generation of
acoustic phonons. The transition time of the excitonic radiative process is in
the range of sub nanoseconds.
ACKNOWLEDGEMENTS
We are thankful to A. Yu. Pyryatinsky, V. N. Kadan, A. N. Komarov, V.
M. Vereschyak, and P. M. Lytvyn for their interest and input to this paper.
We are indepted to the Institute of Single Crystals of the National Academy
of Sciences of the Ukraine for providing us with sapphire substrates. This
work is partially supported by the program "Nanosystems, nanomaterials,
nanotechnologies" (project H27-04).
REFERENCES
1. P. P. Gorbick, A. A. Dadykin, and I. V. Dubrovin, Reports of The National Academy of
Science of Ukraine, 1, 84 (2001).
2. W. Yang, R. D. Vispute, S. Choopun, R. P. Sharma, T. Venkates, and H. Shen,
Appl.Phys.Lett. 78, 2787 (2001).
3. S. Liang, H. Sheng, Y. Liu, Z. Huo, Y. Lu, and H. Shen, J. Cryst. Growth, 225, 110
(2001).
4. Z. K. Tang, G. K. L. Wong, P. Yu, M. Kawasaki, A. Ohtomo, H. Koinuma, and Y.
Segawa, Appl. Phys. Lett. 72, 3270 (1998).
5. H. Cao, Y. G. Zhao, H. C. Ong, S. T. Ho, J. Y. Dai, J. Y. Wu, and R. P. H. Chang, Appl.
Phys. Lett. 73, 3656 (1998).
6. Y. Chen, H. Ko, S. Hong, and T. Yao, Appl. Phys. Lett. 78, 1469 (2001).
7. Y. R. Ryu, S. Zhu, D. C. Look, J. M. Wrobel, H. M. Jeong, and H. W. White et al., J.
Cryst. Growth, 216, 330 (2000).
8. M. Joseph, H. Tabata, and T. Kawai, Jpn.J.Appl.Phys. 38, L1205 (1999).
9. M. Wraback, H. Shen, S. Liang, C. R. Gorla, and Y. Lu, Appl.Phys.Lett. 74, 507 (1999).
10. A. Onodera, N. Tamaki, Y. Kawamura, T. Sawada, and H. Yamashita, Jpn. J. Appl. Phys.
35, 5160 (1996).
11. M. Joseph, H. Tabata, and T. Kawai, Appl. Phys. Lett. 74, 2534 (1999).
12. T. Dietl, H. Ohno, F. Matsukura, J. Gibert, and D. Ferrand, Science 287, 1019 (2000).
13. K. Sato and H. Katayama-Yoshida, Jpn. J. Appl. Phys. 39, L555 (2000).
Chapter 7
OPTICS AND SPECTROSCOPY OF POINT
DEFECTS IN ZnO
Vladimir Nikitenko
Moscow State Transport University, Obraztsova str. 15, 103055 Moscow, Russia
Abstract:
This work investigates spectra of luminescence and photo-electron spin

resonance (EPR) of zinc oxide single crystals, powders and films. We used
single crystals of zinc oxide produced by the hydrothermal method and by the
method of gas-transport reactions. The samples were subjected to irradiation of
electrons (E = 4 MeV, f = 7.51018 cm-2 and f = 1.21019 cm-2) and protons (E
= 7.7 MeV, f = 1015 cm-2). Using both thermoluminescence and EPR electronhole processes are investigated (T = 4 - 450 K), which cause luminescence and
decharging of self and impurity point defects. The mechanism of violet, green,
yellow-orange and red (and accompanying it infra-red: = 11000 - 12000 )
luminescence is discussed.
Key words:
Point defects, photoluminescence, photo-EPR, single crystals, hydrothermal

method, exciton, donor, acceptor, thermoluminescence
1.
INTRODUCTION
Among the well-known semiconducting materials employed in various

technical solutions, a unique position is held by zinc oxide, which is applied
in electronic and space engineering, chemical industry, and medicine.1-3
Despite such different applications, there is no general agreement
regarding the problem of the involvementt of the intrinsic and extrinsic point
defects in the optical, electrophysical, and radiative properties of this
material. The mechanism underlying the photoexcitation and luminescence
of zinc oxide is still a matter of debate. The simultaneous implementation of
spectral luminescence and photo-EPR (electron paramagnetic resonance)
methods has led to considerable improvement of this situation. This report is
devoted to review our data obtained by these methods.
69
70
2.
Vladimir Nikitenko
EXPERIMENT
In the present work, we investigated powders, films and single crystals of

ZnO produced by the hydrothermal method at the Institute of
Crystallography of the Russian Academy of Sciences and by the method of
gas-transport reactions at the Institute for the Physics of Semiconductor of
the Siberian Branch of the Russian Academy of Science.3
The use of a helium flow cryostat in the our spectrometer made it
possible to carry out EPR measurements within the temperature range 4-300
K with a rapid change in the sample temperature. In most cases of low
temperatures, an identical initial state of the crystal before illumination was
achieved by annealing the sample in a cryostat at room temperature for 5-10
min. The concentration N(T) of paramagnetic centers in thermally stimulated
processes of charge transfer in the crystal under investigation was
determined by the EPR signal intensity in the pulsed heating mode.3
3.
INVESTIGATION OF POINT DEFECTS IN ZnO
Typical luminescence spectra of as-grown crystals show ultra-violet

emission due to free excitons (including their longitudinal optical phononsLO replicas) as well as typical bands of green and yellow-orange
luminescence (the temperature is 77-80 K). When the temperature is 4.2 K,
the luminescence of bound excitons (including their LO replicas)
dominates.1-3
An exciton bound to a shallow neutral donor of interstitial zinc (Fig 1 a)
and of interstitial lithium (Fig. 1b) is presented, for example, in our spectra.
In some instances the radiative recombination of an exciton bound to a
neutral defect may not lead to the ground state of the respective defect but to
an excited state of the carrier at this occupied center (2 electron transition).
In a hydrogenic model we can calculate an ionization energy of the neutral
donor state of interstitial zinc to 0.05 eV and of interstitial lithium to 0.033
eV.
In zinc oxide, various impurities of both substitution and interstitial type
act as shallow donor centers with an ionization energy Ed = 0.03 - 0.05 eV
(for example, interstitial hydrogen, the substitutional impurities of the 3-rd
group of the periodic table of elements and so on).
The lithium dopant in ZnO may occupy interstitial sites (Lii) or may
substitute for the Zn site (LiZn) acting as a shallow donor or as an accepter,
respectively. The optical depth, measured from the edge of the valence band,
amounts to about 1 eV and the thermal depths to 0.2 - 0.4 eV.3
Optics and spectroscopy of point defects in ZnO
71
400
1
J (arb. units)
JD(Zni )
(a)
300
2
JD(Zni)ohQ=E(JD)-Ed(1-1/n
200
Ed=0.05 eV
100
0
1
J (arb. units)
100
JD(Lii )
(b)
2
JD(Lii)ohQ=E(JD)-Ed(1-1/n
1
JD(Zni)
50
3700
3750
Ed=0.033 eV
3800
3850
O ()
Figure 1. Photoluminescence spectra of a hydrothermally grown ZnO single crystal (a) before
and (b) after Li-doping by heat treatment in Li2CO3 at 700 qC; T=4.2 K; laser excitation at
O=2950 .
The formation of defects of the first type is favored by reducing

conditions for the doping of crystals while acceptors are created in oxidizing
regimes. In order to investigate the lithium acceptor microstructure in ZnO,
the photo EPR method was used, which turned out to be very efficient in
this respect.3,4
It was found that the EPR of lithium ions is observed only on optical
excitation (mainly by UV light) and at low temperature (< 200 K). The
resonance possesses axial symmetry with respect to the hexagonal C axis of
a crystal. In a magnetic field directed normally to the C axis, the signal of the
axial center is a quartet of closely spaced lines with g factor gA=2.0260,
which corresponds to the hyperfine interaction of the unpaired electron spin
72
Vladimir Nikitenko
moment of the oxygen ion O| with the magnetic moment of the nucleus of
isotope 7Li (j = 3/2).
In a non-excited crystal, the substituting lithium ion is surrounded by four
oxygen ions O|| (with two electrons). The center becomes paramagnetic LiZn
(in a more complete notation [Li+Zn O|] when a hole is trapped (or an
electron is released) by one of the nearest neighbor oxygen ions (or by one of
the four oxygen ligands).
As an example needed for further discussions we give Fig. 2. Optical
transitions (3) and (4) during excitation of the sample form paramagnetic
states of lithium at T = 30 K. The capture of free electrons by the traps keeps
N, I (arb. units)
100
80
60
40
20
0
1(LiZn-O')
2(TL)
0
20 40 60 80 100 120 140 160 180 200
T (K)
Figure 2. Dependence of the concentration of lithium paramagnetic centers (LiZn+-O|),

shortened to LiZn, at T = 30 K on the temperature of annealing the ZnO-Li single crystal (1),
the intensity of thermoluminescence (2) at exe = 380 nm, and the schematic diagram of optical
transitions during excitation of the luminophor (3 and 4) and radiation (5).
73
a paramagnetic state of lithium. Vacating of the electron traps during

heating (first at T1 = 50 K, then at T2 = 145 K) leads to the appearance of
electrons in the conduction band. The subsequent capture of these electrons
by the lithium acceptor is accompanied not only by a decrease in the
concentration of the paramagnetic lithium states (curve 1) but also by
yellow-orange thermoluminescence (curve 2 and transition 5). As a result,
clearly observed is the mechanism of the yellow-orange luminescence of
ZnO and the lithium nature of the corresponding luminescence centers
(Li+ZnnO|), which have been subject of lasting discussions. An analysis of
thermoluminescence curves by various methods3 showed that the detected
electron traps have depths of E1 = 0.03 - 0.05 eV and E2 = 0.18 eV.
Heat treatment of as-grown crystals in Li2CO3 at 400 to 700qC for 2 to 7
days causes a partial substitution of lithium for Zn sites causing considerable
growth of resistivity to U~108-1010 :cm because of the Li+I = Li|Zn and
noticeable lowering of the Fermi level. The luminescence spectra of such
samples are characterized by enhanced yellow-orange emission noticeable
quenching of emission due to excitons bound to neutral shallow donors.
8000
7192
6989
6000
7209
7240
7307
3000
7121
7135
4000
7054
5000
6989
J (arb. units)
7000
2000
1000
0
7000
7200
7400
7600
7800
O ()
igure 3. Red photoluminescence spectrum of grown hydrothermally ZnO(Fe) single crystal
annealed in Li2CO3 at 973 K for seven days; T=4.2 K; excitation with a DKSSh-3000 lamp
(O=3500 ).
74
Vladimir Nikitenko
Li2CO3, their luminescence spectra show a composite violet band, which

was attributed to donor-acceptor pairs is which the acceptors are not isolated
Li|Zn defects (as in the case of the yellow-orange luminescence) but
associates including shallow donors, such as (Zn+i Li|Zn), (In+Zn Li|Zn) so on.2,5
Characteristic red luminescence is observed for ZnO crystals doped with
iron and then heat-treated in Li2CO3 (iron doping can be accomplished
during hydrothermal growth, yielding black ZnO crystals) (see Fig. 3).
The strong EPR signal, which is observed in these crystals, corresponds
to the g-factor g__ = 2.006. It is associated with transition M = -1/2 o +1/2 of
the fine structure of the EPR spectrum of the Fe3+ ions. The EPR of the
Fe3+Zn (3d5) ions in zinc position is not observed in the dark in low-ohmic asgrown ZnO single crystals. This is attributed to the fact that as the Fermi
level raises, three valent iron ions go over into the two-valent diamagnetic
state of Fe2+Zn with the configuration 3d6. Therefore, to observe the EPR
signal of the Fe3+ ions, low-ohmic single crystals are often exposed to a
preliminary thermal treatment, which lowers the Fermi level. Or,
alternatively, the specimens can be irradiated with elementary particles in
250
N, J (arb. units)
1(LiZn-O')
200
3+
3(FeZn)
2(LiZn-O')
150
5(TL)
4(TL)
100
6(TL)
x10
50
0
50
100
150
200
T (K)
Figure 4. Dependence of the spin density N of the paramagnetic centers and of the
thermoluminescence intensity (TL): (LiZn+-O|) at =30 K (1) and at =70 K (2), FeZn3+ at
=70 K (3) on the annealing temperature of the hydrothermal ZnO single crystal and TL
curves the temperature of excitation =5 K (4) and 80 K (5) TL at O=5900 ; at =70 K
and O=7200 (6). The original single crystal (1,4,5); ZnO:Fe single crystal annealed in
Li2CO3 at T = 973 K for seven days (2,3,6). Oexc. = 3800 .
75
regimes causing an increase in their specific resistance. A systematic

relation between the intensity of the red luminescence and the intensity of
EPR signal of the Fe3+Zn ions was noticed.
In other ZnO single crystals, we observed the appearance of an isotopic
EPR signal with the factor g = 2.0134, in which hyper fine structure was
discovered (6 lines of identical intensity). Such an EPR signal was ascribed
to the 2S1/2 state of Pb3+Zn ions.
At low temperatures, the EPR signal of Fe3+ ions and Pb3+ ions in zinc
oxide is photosensitive. We observed this signal in relatively low-ohmic
single crystals of ZnO. A pronounced photo excitation band shows up.3 It
can be related to the photo ionization of the deep donor Fe2+ ions. The
analysis of the form of the photo ionization spectrum allows to calculate the
optical ionization energy, which turned out to be equal to 1.4 eV for Fe2+Zn
ions and 1.65 eV for Pb2+Zn ions.
Fig. 4 shows the dependence of spin density of the paramagnetic centers
and thermoluminescence intensity on the annealing temperature of the used
single crystals. We should note the possibility of observing the Fe4+Zn state in
crystals, which have an artificially lowered Fermi level. The transition
Fe4+Zn + e o Fe3+Zn manifests itself in an increase of the concentration of
paramagnetic Fe3+Zn states in the process of the release of electrons from
electron traps (at 110 to 160 K; Fig. 4, curve 3). This leads to a simultaneous
decrease in the concentration lithium paramagnetic states (curve 2) and
appearance of the yellow-orange (curve 5) and red termoluminescence (curve
6).
As is seen in Fig. 4, the concentration of Li induced paramagnetic states
drops in the range 160-200 K (the yellow-orange luminescence is absent).
Taking into account the hole character of the process, this indicates that the
electrons are transferred from the valence band to lithium paramagnetic
states (LiZn+kToLi|Zn+e+). Therefore, the lithium paramagnetic centers can
be identified with the hole traps responsible for red thermoluminecseence
and, in some cases, for green thermoluminescence with a peak at T=180 K.
A strong peak of red TL appears at 160-200 K in our crystals (curve 6). The
concentration of paramagnetic Fe3+Zn ions increases at this moment (curve
3). This result can be accounted for by the reaction (Fe2+Zn+e+oFe3+Zn),
which is highly probably for the Fe center in ZnO.3 It should be noticed that
the thermal degradation of Li paramagnetic states in the range 160-200 K is
also confirmed by the absence of the corresponding EPR signal for T>200
K.
Numerous measurements show that the traps responsible for the
thermoluminescence peak at = 180 K are located 0.35-0.45 eV above the
edge of the valence band. According to a model of yellow-orange
76
Vladimir Nikitenko
luminescence, the optical depth of the lithium level is 1-1.3 eV above the
valence band.1-4 Thus, thermal and optical measurement yield different
defect ionization and neutralization energies for ZnO which is a natural
consequence of the high bond ionicity and may result from a charge-transferinduced polarization of the crystal lattice.
The spectrum of the red luminescence has a complex structure. We were
able to separate three components of red luminescence that each have their
own kinetics and that differ by the character of excitation. The radiation band
at 7300-7800 is predominantly excited by light with O1 = 3700-4800 and
has a three exponential time dependence of afterglow with time constants
of 3 msec, 15 msec, and 75 msec. The relatively strong emission line at 7175
has maxima in the excitation spectrum at 3700 and 5200 . It is
characterized by an afterglow time constant of W = 33 sec. Several narrow
lines in the 6900-7200 region are well excited with O=4000-4800 and
are characterized by W1 = 3 and W2 = 22 msec.
The red luminescence extends into the infra red region (Fig. 5). The
results obtained point to the requirement of further refinements of the
emission transition scheme in the case of the red luminescence of ZnO.
The model, according to which the red luminescence of ZnO represents
luminescence of ions Fe3+Zn (3d5 configuration) is the most encompassing
Omaxx ()
I (arb. units)
11722
10000
11863
11564
12393
12000
14000
16000
O ()
Figure 5. The infra-red photoluminescence spectrum of a ZnO-Fe,Li single crystal at T=4.2
K; excitation at 4880 (argon laser).
77
on our view now. The red luminescence of ZnO is the analogue of the
well-known orange-red luminescence of manganese ions Mn2+ (for example,
in ZnS), which have the same configuration 3d5 or the red luminescence of
ions Fe3+Zn in ZnS. This radiation corresponds to the optical transitions
4
T1(4G) o 6A1(6S).
Irradiation of the ZnO Li samples by protons leads to the appearance of
a photosensitive EPR signal (f signal) which corresponds to the center with
an axial symmetry and g-factor g|| = 1.9948 and gA = 1.9963. This signal is
very important for understanding the processes of defect formation, since it
leads to gives rise to two alternative interpretations for the microscopic
structure of the intrinsic defects that determine the deviation from the
stoichiometry in ZnO.6
According to the one viewpoint, the f-signal should be assigned to the
oxygen vacancies V+O (F+ center) that are formed by capture of one
electron by anionic vacancies with double positive charge. In this case the
detected hyperfine structure of the signal is explained by the interaction of
the unpaired spin of the electron with the nuclei of the 67Zn isotope (content
is 4.1%, spin I=5/2). The condition for photoexcitation of the f-signal of
EPR is such that the oxygen vacancy VO (F center) represents a deep
donor with a photoionization energy of 2.3 eV. In this case the low
resistivity in ZnO is determined by shallow donors of the interstitial zinc
Zni with an ionization energy Ed = 0.05 eV, which by forming the impurity
band or by recoiling the electrons to the empty band, are responsible for the
specific EPR signal with the factor g=1.96.
From another viewpoint the recorded f-signal of EPR should be
attributed to the interstitial oxygen in the paramagnetic state O|i or Oi, and
VO should be considered as the shallow donor specifying the EPR signal
with the factor g | 1.96 (there is an anisotropy g|| = 1.9557 and gA = 1.9573
at low temperature) and high electrical conductivity of ZnO.
The investigations of the processes of the charge exchange of the centers
by the EPR method in the mode of pulsed heating made it possible to
determine the temperature dependence of the relative concentration of the
paramagnetic centers that specify the f-signal of EPR and of the reference
LiZn-centers in the ZnO-Li single crystals irradiated with protons.
Concurrently, we studied thermoluminescence (Fig. 6). The spectra of the
photoluminescence of crystals and those of the excitation of the F+ and LiZn
paramagnetic centers are shown in Ref. 6.
An important feature of the results obtained is the recapture of the
electrons released from traps at T1 | 30-60 K and T2 | 110-150 K by the
paramagnetic centers that form the f-signal of EPR (Fig. 6, curves 1). Only
when these centers finish their work does the transformation of the lithium
paramagnetic centers into the diamagnetic state begin (T2 | 150-180 K,
78
Vladimir Nikitenko
curve 2); this is accompanied by yellow-orange thermoluminescence (Fig. 6,

curve 3). This fact indicates that the paramagnetic centers responsible for the
f-signal with g|| = 1.9948 and gA = 1.9963 have a much greater effective cross
section for the capture of mobile electrons than the neutral paramagnetic
centers of the form (LiZnnO|), which is characteristic just for the F+-centers.
The Oi states must behave in the same way as the lithium paramagnetic
must be substantially less effective in the capture of

centers, and the O|I states
the electrons than the centers of the (LiZnnO|) type.
The concept of the F+-center in interpreting the f-signal of EPR is
confirmed by a theoretical model of the F+-centers in ZnO and by the data for
other A2B6 compounds.3 The absence of the EPR-signal with the factor g |
1.96 in the crystals irradiated with proton (the resistance of a sample
increases by 2-3 orders of magnitude) supports the argument in favor of the
description of the signal by free carriers in the conduction band (or electrons
in the band of shallow donors) and also confirms the conclusion about the
vacancy nature of the f-signal of EPR. The valence band electrons fill the
lithium paramagnetic centers in the 180-200 K temperature range. The holes
formed in the valence band are captured by the F-centers following the
reaction Fx + e+ > F+, which leads to an increase in the intensity of the fsignal of EPR (Fig. 6, curve 1). A similar filling of the F+ states by the
valence band electrons in the temperature region of about 250 K leads to the
complete disappearance of the f-signal (Fig. 6).
Attention is drawn to the radiation-less character of the transition of
electrons from the conduction band to the F+-centers (a decrease in the
concentration of the paramagnetic states F+ does not cause
thermoluminescence, Fig. 6). This contradicts rather widespread models
attributing the green luminescence of ZnO just to that sort of process.
Quenching of the complex band of the green-orange luminescence and
redistribution of its maximum to the long-wavelength side are observed in
ZnO single crystals irradiated with protons. This is related with the presence
of a green-radiation absorption band in the spectrum of photoexcitation of
F+-centers.3 This appears to be also an explanation of the absence of the
green thermoluminescence which is typical for ZnO. Irradiation of initial
ZnO-Li single crystals with electrons is also accompanied by the appearance
of the f-signal of EPR. The character of the spectrum of the visible
luminescence of these crystals corresponds to that in Fig. 6. An interesting
feature of the ZnO-Li single crystals irradiated with electrons is the existence
of the EPR signal from FA-centers,6,7 which is an additional argument in
favor of the selected model of the f-signal of EPR.
The centers of FA type have deeply been studied in alkali-earth oxides
and have recently been detected in ZnS.7 Unlike F+, this center contains an
impurity ion in the immediate neighborhood with the anionic vacancy (for
79
2(LiZn-O')
100
N, I (arb. units)
80
+
1(F )
60
4(red TL)
40
20
0
3(y-oTL)
50
100
150
200
250
T (K)
Figure 6. Dependence of the concentration of paramagnetic centers N at T = 30 K (full
symbols) and of the intensity of thermoluminescence (curves 3,4) on the temperature of
annealing of the ZnO single crystal irradiated with protons (E = 7.7 MeV. Tdose 1015
protons/cm2): 1) F+-centers; 2) paramagnetic states LiZn; 3 and 4) the intensity of
thermoluminescence at = 6000 and 7200 (the excitation hv = 2.6 eV).
example, [FMg]+ and [FLi] in CaO or [FCu] in ZnS). The presence of the
impurity in the FA -center is usually confirmed by the hyperfine interaction of
the spin moment of the unpaired electron with the magnetic moment of the
impurity nucleus, which is fixed in the ZnO-Li crystals irradiated with
electrons.6,7
The values of the g-factor (g|| = 1.9951 and gA = 1.9956), hyperfine
interaction constants, and the electron wave function for FA-centers in ZnO
correlate well with similar parameters for the FA- and F+-centers in CaO and
ZnS. Here, LiZn seems to be the impurity in the FA-center in the crystals we
have investigated. This assumption is based on the fact that the investigated
hydrothermal single crystals were doped with lithium with a concentration of
NLi | 1017 cm-3 during the process of growing. The unpaired electron of the
FA-center interacts with the impurity having nuclear spin J = 3/2, which is
also well satisfied for lithium, which in the natural state contains 92.58% of
the isotope 7Li with J = 3/2.
In the temperature range 530-660 K in the investigated ZnO crystals,
thermally stimulated ionic processes were detected, at the first stage of
which (T = 530-630 K) "pure" anionic vacancies disappear and at the second
80
Vladimir Nikitenko
of which (T = 610-660 K) the FLi-centers also are healed. A similar behavior

was observed for the F- and FCu-centers in ZnS.7
The green luminescence is typical for ZnO-Cu.1-3 An interesting feature
of the isovalent impurity of copper in ZnO is that the photoexcitation of a
valence electron into the impurity d-shell is accompanied by the formation of
a hole which on the hydrogen like large-radius orbit is bound with the
impurity atom onto which the electron has gone over. Thus, as a result of the
Figure 7. Energy scheme of basic point defects in zinc oxide. The figure shows the energy
level position relative to the edge of the conduction and valence band; the optical and thermal
depths of the corresponding level are given in parentheses and square brackets, respectively
(eV); T = 77 K. Luminescence: violet (V), yellow-range (YO) and green (g). The unbroken
line is the energy level occupied with one electron; the dotted line is the energy level of the
empty state.
optical transition on the impurity, the exciton - like state of two particles is
formed, 3d10(Cu+)h , the so-called acceptor exciton, whose formation and
breakup leads to the appearance of the characteristic structure of edge
absorption (the , and lines) and of green multi phonon luminescence.8-10
It should be noted that in recent years the concept of acceptor and donor
excitons bound with the impurities of transition metals in semiconductors has
found theoretical and experimental justification for a variety of compounds
A2 B 6 .
The given model is confirmed by the application of green emission
intensity in low voltage cathodoluminescent phosphors, single crystals and
ZnO films doped with Cu.11
4.
81
CONCLUSION
In conclusion, the simultaneous application of the photo-EPR and

luminescence methods is extremely useful for studying the role of point
defects in thermo- and photo stimulated electron-hole processes in crystals.
As a result of the investigations, the energy position and the microstructure
of basic local centers that influence the optical and electro physical
properties of zinc oxide have been determined. Fig 7 summarizes our results
and interpretations by showing the energy scheme of the defects and the
assignments of luminescence transitions. It is shown that when constructing
the energy scheme of optical transitions with participation of point defects in
ZnO it is necessary to take into account the nonequivalence of the position of
occupied and "awaiting" energy levels of the center and also the difference
between the optical and thermal energies of the ionization of defects.
ACKNOWLEDGMENT
The author is grateful to . E. Tarkpea (investigations of EPR), A.
Freiberg, I. P. Kuz'mina, S. V. Mukhin and I. V. Pykanov for their help in
this work.
REFERENCES
1. W. Hirschwald, P. Banasewicz, L. Ernst, Curr. Top. Mater. Science. 7, 143 (1981).
2. J. P. Kuz`mina, V.A. Nikitenko, Zinc Oxide. Production and Optical Properties [in
Russian], Moscow (1984).
3. V. A. Nikitenko, J. Appl. Spectroscopy , 57, 783 (1992).
4. D. Zwingel, J. Luminescence. 5, 385 (1972).
5. V. A. Nikitenko, S. V. Mukhin, J. P. Kuz`mina, V. G. Galstyan, S. G. Stoyukhin,
Inorganic Materials, 31, 1243 (1995).
6. V. A. Nikitenko, K. E. Tarkpea, I. V. Pykanov, S. G. Stoyukhin, J. Appl. Spectroscopy, 68,
502 (2001).
7. K. Tarkpea, A. Ots, V. A. Nikitenko, J. Phys. Chem. Sol. 55, 1353 (1994).
8. R. Dingle, Phys. Rev. Lett. 23, 579 (1969).
9. D. J. Robbins, S. G. Bishop, J. A. Savage, P. Porteous, J. Phys. C. 14, 2847 (1981).
10. D. J. Robbins, D. C. Herbert, P. J. Dean, J. Phys. C. 14, 2859 (1981).
11. Ya.J. Alivov, M. V. Chukichev, V. A. Nikitenko, Phys. and Tech. Semiconductors, 38, 34
(2004).
Chapter 8
WHISPERING GALLERY MODES IN
HEXAGONAL ZINC OXIDE MICRO- AND
NANOCRYSTALS
Thomas Nobis, Evgeni M. Kaidashev, Andreas Rahm, Michael Lorenz, and
Marius Grundmann
Universitt Leipzig, Fakultt fr Physik und Geowissenschaften, Institut fr Experimentelle
Physik II, Linnstrae 5, D-04103 Leipzig, Germany
Abstract:
The resonator properties of zinc oxide (ZnO) micro- and nanocrystals grown
by a novel high pressure pulsed laser deposition process have been
investigated at room temperature by cathodoluminescence (CL), spatially
resolved CL-imaging and polarization resolved micro-photoluminescence (PL) within the visible spectral range. The spectra exhibit a series of
comparatively sharp and almost equidistant resonance lines. Using a simple
plane wave interference model and taking into account the spectral
characteristic n(Z) of the refractive index of ZnO, we can unambiguously
attribute those lines to whispering gallery modes (WGMs) of a two
dimensional hexagonal resonator. The predicted resonator diameters agree
well with the measured crystal sizes. Tapered, high aspect ratio ZnO
nanoneedles furthermore allow systematic investigations of the WGMs as a
function of cavity diameter D down to zero. Hence, the transition from a
multi-mode to a single mode cavity is directly observed. -PL experiments
demonstrate that the WGMs are mainly TM polarized.
Key words:
Whispering
polarization
1.
gallery
modes,
nanostructure,
microcrystal,
nanocavity,
INTRODUCTION
If as mechanical, acoustical, electrical, or optical ones, resonators have

always been part and parcel of unique kinds of technical applications. As an
integral part of semiconductor lasers, e.g., optical cavities are essential
83
84
Thomas Nobis et al.
home-electronic devices and therefore even entered peoples daily life. In

the last decades, as a result of miniaturization within the field of electronics
and photonics, very small optical resonators strongly gained importance. In
this context, an interesting type of resonator is the so called dielectric
whispering gallery resonator, e.g. shaped as a microdisc1-4, microcylinder5 or
microsphere.6,7 Inside such a resonator, the light wave can be considered to
circulate around due to multiple total internal reflection (TIR) at the
resonators boundary. This effect typically leads to high Q-factors and low
laser threshold power. In recent years, particularly in the course of analysing
and employing wurtzite structured materials, research focussed on
hexagonally shaped resonators, and whispering gallery mode (WGM) lasing
has been successfully demonstrated also in hexagonal cavities.8,9
However, the size of all afore mentioned cavities is typically much larger
than the emitted light wavelength O. Characteristic cavity diameters usually
exceed 5 m. In order to create nano-sized and laterally single mode devices,
the prospering branch of nanophotonics requires resonators whose
lateral size is at least comparable to or even smaller than O.
In this paper we present the systematic investigation of WGMs in
hexagonal resonators within the micro- and
d nanometer regime for small
mode numbers N down to N = 1. Single zinc oxide (ZnO) microcrystals and
high aspect ratio ZnO nanoneedles have been analysed, together covering
the range from D | 3 m down to D = 0. The needles tapered cross section
furthermore easily allows direct studies of resonator properties in
dependence of the cavity diameter. In our measurements, the continuous
transition from a multi mode to a single mode cavity is observed.
Surprisingly, a simple plane wave interference model without free
parameters agrees well with our experimental data for all N.
N
2.
THEORY OF HEXAGONAL WHISPERING

GALLERY MODES
2.1
Dielectric resonators
From a general point of view a dielectric resonator is a transparent body

of certain shape consisting of a substance with a high refractive index n,
usually surrounded by air. Such a structure has the ability to confine light
inside, e.g. utilized in any kinds of optical fibers or waveguides. To explain
this phenomenon in terms of classical electrodynamics, and any similar
problem respectively, it is necessary to solve the tree-dimensional wave
Whispering gallery modes in hexagonal zinc oxide
85
equation for the given geometry. Steady state solutions of this equation are
then usually called eigenmodes of the considered resonator.
In the case of a hollow metallic resonator as an example for a so-called
closed
d cavity, a discrete set of eigenmodes exists related to a discrete
spectrum of eigenmode energies.10 This is due to the boundary conditions for
the electric and magnetic field that lead to a complete suppression of wave
propagation outside the resonator. Hence, no energy is radiated from the
hollow resonator; the whole radiation field is restricted inside the metallic
cavity.
In contrary, a dielectric resonator is an open cavity. Because of the
continuity conditions for the electric and magnetic field at the cavitys
boundary, the radiation field is not exclusively restricted to the resonator
itself, but actually extended to infinity. This results in a permanent radiation
loss of energy and hence, every imaginable field distribution decays
exponentially with a lifetimeW, therefore not representing a steady state
solution.11 The actual solutions of this problem form a continuum of socalled scattering states consisting of an incoming plane wave and an
outgoing scattered wave. Therefore, the eigenmodes of a dielectric cavity
appear as maxima within the continuous spectrum of the total scattering
cross section in dependence of the energy of the incoming wave. Those
maxima, i.e. those incoming wave energies that exhibit the strongest
interaction with the given resonator, are usually called resonance energies or
resonances of the dielectric cavity.11 Due to the above mentioned finite
lifetime, all resonances exhibit a finite line width. The larger the losses of a
given resonance the shorter is the life time and the larger is the resonances
line width.
2.2
Two-dimensional hexagonal resonator
An important class of dielectric resonators are two-dimensional cavities,

i.e. their in-plane dimensions are much larger compared to their height. In
this case, wave propagation parallel to the longitudinal axis c is negligible.
The light can be assumed to circulate around only within the resonators
cross section. The respective mode patterns are so-called whispering gallery
modes (WGMs), obtained by solving the wave equation in its twodimensional form. Hence, two decoupled types of polarizations occur,11
namely TE polarized modes ((E A c), and TM polarized modes ((E
E || c), as
explained in Figs. 1b,c.
86
Thomas Nobis et al.
Figure 1. (a) The light confined within a hexagonal cross section can be assumed to circulate
inside the cavity as indicated by the white arrows. The geometry of the cavity can be
described by the radius of the incircle Ri, by the radius of the circumscribing circle R, or by its
related diameter D = 2R using the geometric relation R = 2Ri/3. Note that the circumference
of the inscribed white hexagon has a length of 6Ri. (b), (c) Two possible types of polarization
of two-dimensional whispering gallery modes, exemplary given for a hexagonal resonator.
For TE modes the electric field is transverse to the longitudinal axis of the resonator, for TM
modes the magnetic field is.
Eigenmodes of such two dimensional whispering gallery resonators have

been calculated in the past; the solutions for a circular cavity can be
expressed in an analytical way.12 Other kinds of geometries have been
subject to present literature, as e.g. deformed circles,13,14 rectangles15 or
hexagons,16 that necessarily have to be calculated numerically.16,17 A scheme
of the geometry of the two-dimensional hexagonal cavity is given in Figure
1a.
The results of theoretical modeling show,16 that the eigenmodes of such a
hexagonal cavity can be numbered by an integer mode number N.
N
Additionally, the modes can be distinguished concerning their symmetry, e.g.

leading to pairs of degenerated modes N and N , related to their
respective chirality. In Ref 16, hexagonal WGMs have been calculated for a
fixed n = 1.466 and in the mode number range of about N = 20 to 70.
2.2.1
Plane wave model
For large mode numbers N > 70 a simple plane wave model (PWM) has
been deduced.16 Its main idea is, that the light wave circulates around
exploiting the process of multiple total internal reflection (TIR, see Fig. 1a)
and finally interferes with itself when having completed one full circulation
within the resonator. To enforce constructive interference the total phase
87
shift of the wave along its path has to be an integer multiple of 2S, i.e. only
entire wave trains are allowed to perform multiple circulations generating a
standing wave. Taking into account the polarization-dependent negative
phase shift that occurs during the process of TIR,10 we obtain the following
equation:
h c
n E
a cta
arctan
t
(1)
The factor E regards polarization, for TM-polarization, E = ETM = n-1 has

to be used; TE-polarization leads to E = ETE = n. Due to the spectral
( ), eq. (1) is an implicit equation to
dependence of the refractive index n = n(E
determine the discrete resonance energies E = EN (Ri) in terms of the
geometric parameter Ri, Plancks constant h and vacuum speed of light c.
The first factor of the right side of eq. (1) corresponds to the wavelength in
matter. The integer N t 1 characterizes the interference order of the
resonance, which is in this case identical with the respective WGM
number.16 The following term containing E refers to the additional phase
shift mentioned above. Furthermore, since ZnO is uniaxial, n = n||(E) and n =
nA(E) have to be applied for TM- and TE-polarisation, respectively. We note
that eq. (1) for TM-polarisation and n = const. is identical with eq. (17) of
Ref. 16. Neglecting the spectral dependency of n, eq. (1) leads to
EN v
1
Ri
(2)
and thus, WGMs generally shift to higher energies with decreasing cavity
diameter. Note that, although TIR suggests no way for the light to leave the
resonator, emission nevertheless occurs at the corners of the hexagon.16 After
all, radiation losses are always allowed due to the reasons given in part 2.1.
Therefore it is possible to optically detect WGMs. Since the geometrical
model leading to eq. (1) is a combination of basic ray and plane wave optics,
it is expected to be valid only for R >> O, i.e. for N >> 1.16 Anyhow, we show
in this work, that eq. (1) describes the experimentally found WGMs
even down to mode number N = 1.
88
Thomas Nobis et al.
3.
EXPERIMENTAL DETECTION OF HEXAGONAL

WHISPERING GALLERY MODES
3.1
Sample properties
3.1.1
Growth
As a model system for the investigation of hexagonal cavities we use zinc

oxide micro- and nanocrystals naturally exhibiting a hexagonal cross section
due to their wurtzite structure. Those crystals have been grown selfassembled on a-plane sapphire substrates by high pressure pulsed laser
deposition (PLD) utilizing a gold assisted growth mechanism similar to
Morales and Lieber.18 For this purpose the substrates have been sputtered
partially or fully by a few nanometer thin gold film. When heating up to the
growth temperature of about 870C to 950C the gold is melting and
subdividing into small clusters that provide the seed for columnar
nanocrystal-growth. Within the PLD chamber a KrF excimer laser is used to
evaporate the ZnO target, and argon acts as transport gas, set under a
background pressure ranging from about 1 mbar to 100 mbar. More detailed
information about this technique will be published elsewhere.19 Depending
on the respective growth conditions the PLD process provides a variety of
differently shaped ZnO micro- and nanocrystals shown in
Figure 2.
3.1.2
Luminescence
ZnO is a direct semiconductor with a band gap of about 3.4 eV at room

temperature. Its room temperature luminescence typically consists of two
bands, the first (UV) around 3.25 eV, the second (visible, VIS) around
2.35 eV. The UV band is associated with free-exciton emission. The latter is
also called green band in the literature; its exact nature, involving deep
levels, is still under debate. Since the luminescence of our PLD grown
micro- and nanocrystals is dominated by VIS emission, and furthermore
since VIS emission provides a bright and broad-band light source, it can
easily be utilized in cathodoluminescence (CL) experiments to excite the
micro- or nanocavities. Within the VIS range ZnO is optically transparent
and exhibits a refractive index of n | 2,20 i.e. ZnO micro- and nanocrystals
represent an interesting system to investigate resonator properties in a wide
spectral range.
89
Figure 2. Different shapes of PLD grown ZnO micro- and nanocrystals. (a) Hexagonal
microcrystals with diameters of about 2 m. (b) ZnO nanowires with diameters around 100
nm. (c) Needle like nanostructures. (d) Microcrystals exhibiting a dodecagonal cross section.
3.2
Whispering Gallery Modes in hexagonal ZnO

microcrystals
3.2.1
Detection of Whispering Gallery Modes
Figure 3 shows a typical VIS emission CL spectrum from a ZnO

microcrystal. An electron microscopy image of the crystal is given in the
inset. In comparison to the broad and unstructured VIS emission of thin
films or bulk material, VIS emission of the microcrystal is accompanied by a
series of comparatively sharp peaks.
Using the PWM, those peaks can unambiguously be attributed to WGMs
of a hexagonal cavity. Therefore it is necessary to determine the correct
mode number N of every single detected peak. Provided that this mode
number is known, eq. (1) enables to calculate the theoretical diameter of the
crystal D = 4Ri/3 out of every single detected peak energy. Fortunately,
since the peaks have to be numbered in ascending order, one only has to find
the correct mode number to start. The best fitting and therefore the final peak
numbering is found, if every single peak predicts the same radius as all the
90
Thomas Nobis et al.
other resonance peaks, or at least if the variations in the predicted radius

become smallest. This procedure is explained more detailed in Fig. 4.
Figure 3. Room temperature VIS emission CL spectra of a ZnO microcrystal (curve 1) and of
thin film material (curve 2) scaled to the same intensity and vertically shifted for clarity.
Arrows mark theoretical energetic positions of WGMs due to eq. (1) for TM polarization and
D = 2.86 m. The inset shows a SEM image of the investigated microcrystal. The scalebar has
a length of 2 m. The experimental cavity diameter is Dexp = (2.90 r 0.06) m.
3.2.2
Theory vs. Experiment
For the investigated crystal, the correct mode numbering is shown also in
Fig. 3, the respective theoretical diameter yields to Dtheory = 2.86 m with a
small minimum-maximum spreading of less than 10 nm. Although theory
predicts both TM and TE polarization, all calculations have been performed
for TM type, since polarization-dependent micro-photoluminescence
experiments showed, that WGMs are preferentially TM polarized (see
Figure 7 below). The required data for n||(E) were obtained from
ellipsometry measurements on PLD grown ZnO thin film samples.20
91
Figure 4. Diagram Ri(E) to determine the correct mode numbering for the spectrum of
Figure 3. Equation (1) enables to calculate the radius of the cavity-incircle Ri=R
= i(E,N)
N out of a
given resonance number N and resonance energy E. The resulting set of curves is given here
for N = 1041 as a set of black lines approximately exhibiting a hyperbolic shape (as
expected from eq. 2). The index of the curves, i.e. the respective resonance number, is shown
on the left side of the scheme marked by black arrows. The experimentally detected resonance
energies are given as white circles. Since the resonances have to be numbered in ascending
order, every white line represents a possible peak numbering. As can be seen from the
diagram there exists only one peak numbering that results in an energy-independent incirlce
radius, denoted by the horizontal line. Hence, the starting number for the peak numbering of
the spectrum in Figure 3 yields to N0=21 with Ri = 1.24 m, this leads to a respective cavity
diameter D=2.86 m.
The experimentally determined cavity diameter obtained from electron

microscopy amounts to Dexp = (2.90 r 0.06) m. Hence, the deviations
between theory and experiment are in the range of only 2%! This means, that
the simple PWM fits very well, even if N is in the range of only 20 to 30. To
emphasize this fact, the theoretical values EN calculated for fixed diameter D
= Dtheory are given as black arrows in Figure 3; they appear very close to the
measured peaks. We note, that if n z n(E) = const. is used, the agreement
between theory and experiment becomes worse.
92
3.3
Thomas Nobis et al.
Whispering Gallery Modes in ZnO nanostructures
Although the simple PWM obviously is valid for WGMs with mode
numbers in the range of about N = 2030, i.e. for optical cavities in the
micrometer regime, it does not necessarily need to be valid for
nanostructures. Furthermore, prism-shaped microcrystals only enable
analysis of WGMs for fixed cavity diameter given by the cross section of the
particular microcolumn. Those restrictions can be vanquished by focusing on
nanostructures with a needle-like shape.
3.3.1
ZnO nanoneedles
An example of an array of PLD-grown ZnO nanoneedles is shown in

Figure 5a. As can be seen in Figs. 5b-d, the width of the hexagonal cross
section of a selected needle is continuously tapered when approaching the
needles top. Since the aspect ratio of this structure is very high, every plane
perpendicular to the needles longitudinal axis creates a hexagonal cavity
and actually, those structures allow for systematic investigations of WGMs
in the nanometer regime from D = 800 nm down to D = 0!
Figure 5. SEM image of the investigated nanoneedle. All images have been obtained at 45
perspective. (a) SEM image of an array of ZnO nanoneedles containing the reported one
marked by a black rectangle. The scale bar has a length of 10 m. (b) Larger scale SEM
image of the investigated nanoneedle. The needles diameter is continuously tapered
approaching zero at the top. (c) High resolution SEM image indicating the hexagonal cross
section of the needle. The scale bar has a length of 300 nm. (d) Experimentally determined
shape of the needle. The obtained characteristic of the diameter D vs. scan position x can be
fitted using a potential law D v x0.523 r 0.007, leading to a square root like shape of the needle.

3.3.2
93
Detection
For this purpose, we performed spatially resolved CL experiments,21,22

i.e., point wise CL spectra have been recorded across the whole needle using
a scan grid of 16 u 150 points. Typical spectra and their interpretation are
given in Fig. 6. As in the case of microcrystals, the broad VIS band of
luminescence again is modulated due to a preferred constructive interference
of light whose photon energy again fulfills the resonance condition of eq.
(1). For decreasing diameter, approaching the top of the needle, these
maxima continuously shift to higher energies. This effect supplies the
unambiguous proof that the measured spectral modulation actually originates
from WGMs. When the dominating WGM with number N is blue shifted so
far that it leaves the VIS range, the next lower resonance N
N-1 occurs at lower
energies indicated by a discontinuous red shift of the spectral maximum in
Fig. 6c and 6d. This process continues until N = 1 is reached referring to the
last resonance that can be observed.
Figure 6. Spatially resolved CL investigation of a single tapered ZnO nanoneedle. (a) CL

spectra shifted vertically for clarity collected att 8 equidistant locations marked on the needles
longitudinal axis by white dots in Fig. 6b. The VIS band of ZnO is modulated such that
maxima can clearly be distinguished from each other and attributed to WGMs labelled
according to eq. (1). Since the radius of the needle decreases along the longitudinal axis, the
spectral maxima continuously shift to higher energies as indicated by the red dashed lines.
Blue arrows and error bars mark selected TM resonance energies obtained from eq. (1) and
their error. (b) Experimentally determined shape of the needle. (c) Map of the energy of the
spectral maximum within the visible spectral range. (d) Line scan along the white line shown
in Fig. 6c. The red dashed line gives the maximum position of the unstructured VIS band in
bulk material.
94
3.3.3
Thomas Nobis et al.

Polarization
To obtain essential information about the state of polarization of the

detected WGMs we performed local polarization-dependent microphotoluminescence experiments. These show that the WGMs are
preferentially TM polarized, since detecting TE polarization causes an
almost complete suppression of WGM modulations (see Fig. 7). The
remaining spectrally unstructured emission occurs due to the unpolarized
character of ZnO photoluminescence originating from the needles center.
However, this result is consistent with former investigations of lasing in
hexagonal cavities within the micrometer regime that are reported to only
emit TM modes.8 Hence, in comparison to TE modes, TM modes obviously
exhibit lower losses. This can be generally understood since, leaving the
regime of TIR, the reflection coefficient for TM polarized waves due to
Fresnels equations is always larger than that for TE polarized ones.
Figure 7. Micro-photoluminescence spectra of the nanoneedle of Fig. 5 for D | 790 nm at

two different polarizer azimuth orientations shifted vertically for clarity. The modulation of
the VIS band almost disappears when detecting the TE mode. Thus, the resonances are
mainly TM polarized. The inset shows the ratio P = (ITM ITE) / ((IITM + ITE), that visualizes the
TM-WGMs as distinguishable peaks. The polarization effect has been found to be
independent of the polarization of excitation.
3.3.4
Theory vs. Experiment
To compare the measured energies of the WGMs in nanostructures with

theoretically predicted ones for both types of polarization, a spectral line
95
scan along the needles longitudinal axis is given in Fig. 8. In grey scales,
the resonances appear as a set of bright lines, that actually exhibit a
curvature as expected from eq. (2). The theoretical values are shown with
dots and cross symbols in Fig. 8. They are directly calculated without free
parameter using the experimentally determined shape of the needle, i.e. Ri(x
( ),
see Fig. 5, and the above mentioned thin-film data for n||(E) and
nA(E).20 The dots (TM) match the measured resonance energies with a very
good agreement. Additionally, the well fitting blue arrows within Fig. 6a
confirm, that the simple PWM gives a good description of the basic physics
of WGMs even in nano-sized crystal geometries within the limits of the
resonance line widths.
Two dimensional plot of spectra recorded along a line scan on the needles
longitudinal axis. The left vertical axis shows the line scan position x, the right one refers to
the respective needle diameter D. The grey scales refer to the spectral CL-intensity. The
spectral maxima, i.e. the measured WGM energies, appear as bright belts going from the
bottom left corner to the right upper one. With
h decreasing diameter all resonances shift
systematically to higher energies. The white dots give theoretical TM-resonance energy
positions obtained from eq. (1), white crosses give the same for TE-polarisation. Without
adjusting free parameters there is a very good agreement between experiment and theory.
96
3.3.5
Thomas Nobis et al.

Line broadening effects
As shown in Figure 6, compared to microcavities, nano-sized crystals

show broadened WGMs due to the increase of losses with decreasing cavity
diameter. Nevertheless, CLI of the broad VIS band still enables the
visualization of WGMs in the ZnO nanoneedle even at diameters
D < O/n # 270 nm. For mode numbers N t 4 single TM-WGMs can be
detected unambiguously as distinguishable maxima modulating the broad
VIS band. For smaller mode numbers N d 3 WGMs can only be visualized
as a continuous blue shift of the intrinsic VIS maximum, and the
discontinuities in Fig. 6c and 6d become smooth. This is due to the extreme
line broadening of both types of WGMs with decreasing cavity diameter.
Theory16 predicts line broadening by loss processes due to boundary waves
that are scattered out of the resonator and due to light paths whose angle of
incidence is slightly deviating from 60. A relation FWHM v E-1 R-2 is
obtained, that explains the dramatic broadening of the measured resonances.
For TM polarization and N = 6 (spectrum 1 of Fig. 6a) a FWHM
M of about
350 meV is predicted. Considering that the underlying theory again is
prepared for N >> 1, the obtained spectrum verifies this value within the
limits of resolution. For resonance N = 3 the theoretical FWHM
M is about 900
meV, which exceeds even the intrinsic line width of the VIS band. This
explains the difficulty to obtain distinct WGM peaks for N d 3. Nevertheless,
VIS emission is still affected when a WGM crosses the VIS range, as Fig. 8
clearly shows intensity modulations at the respective line scan positions. We
note that losses due to surface roughness of the cavity faces could also lead
to line broadening. However, as the high resolution SEM image of Fig. 5c
shows well shaped resonator faces, we do not adopt this loss process to be
the dominating one.
3.3.6
Discussion
As mentioned above, there is an energy difference between predicted

TM- and TE-modes. This difference is mostly due to E in eq. (1), since the
ZnO birefringence (n|| - nA)/(n|| + nA) | 1.2% is small.20 However, TE-mode
with mode number N is predicted (coincidentally, due to the particular value
of n) to appear very close to the spectral position of TM-mode N
N+1 (Fig. 8),
and hence, this could be one reason for the missing TE mode series. Possible
TE-WGM maxima should always lie beyond higher ordered TM maxima,
and as the former ones suffer larger losses, they lead to much broader
resonance peaks hardly detectable in the VIS band.
Certainly there are, however, small deviations between the detected peak
energies and the predicted TM-WGMs according to eq. (1). (See Figs. 6a
97
and 8.) These deviations probably reveal the limits of the simple PWM. To
discuss this fact more in detail we note, that the application of numerical
methods to predict resonance energies at the end always depends on a
precise determination of the considered geometry, e.g. Ri, and a good
knowledge of the refractive index n. SE microscopy measurements of the
cavity diameter D is performed with an error of at least 1%. The
ellipsometrically determined values for n|| for bulk material yield an error in
n of about 2%. The actual refractive index of the nanocrystals may possibly
differ additionally. These effects lead to an error in E of about 3%, i.e. 50 80 meV in the considered spectral range, which is indicated as error bars in
Fig. 6a. Compared to the line widths of the WGMs, eq. (1) for TM
polarization gives a very good description of the spectral position of the
resonant modes.
4.
CONCLUSION
In conclusion we investigated the resonant optical behavior of ZnO

micro- and nanocrystals, giving detailed insight into optical whispering
gallery modes within the micro and
d nanometer regime. In particular, for the
first time we analyzed hexagonal WGMs in nanocavities for mode numbers
in the range from N = 1 to N = 6. The energy shift and broadening of the
modes for cavity diameter decreasing to zero is well described without free
parameter by eq. (1) from a plane wave model.
ACKNOWLEDGMENTS
This work was supported by the Deutsche Forschungsgemeinschaft
within FOR 522 (Project Gr 1011/12-1).
REFERENCES
1. S.L. McCall, A.F.J. Levi, R.E. Slusher, S.J.Pearton, and R.A. Logan, Appl. Phys. Lett. 60,
289 (1992).
2. S. Chang, N.B. Rex, R.K. Chang, G. Chong, and L.J. Guido, Appl. Phys. Lett. 75, 166
(1999).
3. K.J. Luo, J.Y. Xu, H. Cao, Y. Ma, S.H. Chang, S.T. Ho, and G.S. Solomon, Appl. Phys.
Lett. 77, 2304 (2000).
4. X. Liu, W. Fang, Y. Huang, X.H. Wu, S.T. Ho, H. Cao, and R.P.H. Chang, Appl. Phys.
Lett. 84, 2488 (2004).
98
Thomas Nobis et al.
5. A.F.J. Levi, R.E. Slusher, S.L. McCall, S.J. Pearton, and W.S. Hobson, Appl. Phys. Lett.
62, 2021 (1993).
6. S.M. Spillane, T.J. Kippenberg, and K.J. Vahala, Nature 415, 621 (2002).
7. M. V. Artemyev, U. Woggon, R. Wannemacher, Appl. Phys. Lett. 78, 1032 (2001).
8. I. Braun, G. Ihlein, F. Laeri, J.U. Nckel, G. Schulz-Ekloff, F. Schth, U. Vietze, .
Weiss, and D. Whrle, Appl. Phys. B 70, 335 (2000).
9. U. Vietze, O. Krau, F. Laeri, G. Ihlein, F. Schth, B. Limburg, and M. Abraham, Phys.
Rev. Lett. 81, 4628 (1998).
10. J. D. Jackson, Classical electrodynamics (Wiley, New York, 1999).
11. J.U. Nckel, 2-d Microcavities: Theory and Experiments, in Cavity-Enhanced
Spectroscopies, edited by R.D. van Zee and J.P.Looney, (Academic Press, San Diego,
2002) pp. 185-226.
12. N.S. Kapany and J.J. Burke, Optical Waveguides, (Academic Press, New York, 1972)
13. S.-B. Lee, J.H. Lee, J.S. Chang, H.J. Moon, S.W. Kim, and K. An, Phys. Rev. Lett 88,
033903 (2002).
14. C. Gmachl, F. Capasso, E.E. Narimanov, J.U. Nckel, A.D. Stone, J. Faist, D.L. Sivco, and
A.Y. Cho, Science 280, 1556 (1998).
15. M.Lohmeyer, Opt. Quant. Electron. 34, 541 (2002).
16. J. Wiersig, Phys. Rev. A 67, 023807 (2003).
17. J. Wiersig, J. Opt. A. 5, 53 (2003).
18.A. M. Morales and C. M. Lieber, Science 279, 208 (1998).
19. M. Lorenz et al., unpublished.
20. R. Schmidt, B. Rheinlnder, M. Schubert, D. Spemann, T. Butz, J. Lenzner, E.M.
Kaidashev, M. Lorenz, A. Rahm, H.C. Semmelhack, and M. Grundmann, Appl. Phys. Lett.
82, 2260 (2003).
21. J. Christen, M. Grundmann, and D. Bimberg, J. Vac. Sci. Technol. B 4, 2358 (1991).
22. Th. Nobis, E.M. Kaidashev, A. Rahm, M. Lorenz, J. Lenzner, and Marius Grundmann,
Nano Lett. 4, 797 (2004).
Chapter 9
PROPERTIES OF DISLOCATIONS IN
EPITAXIAL ZnO LAYERS ANALYZED BY
TRANSMISSION ELECTRON MICROSCOPY
E. Mller1, D. Livinov1, D. Gerthsen1, C. Kirchner2, A. Waag3, N. Oleynik4,
A. Dadgar4, and A. Krost4
1
Laboratorium fr Elektronenmikroskopie, Universitt Karlsruhe (TH), 76128 Karlsruhe,

Germany; 2Abteilung Halbleiterphysik, Universitt Ulm, Albert-Einstein Allee 45, 89081 Ulm,
Germany; 3Institut fr Halbleitertechnik, Universitt Braunschweig, Hans-Sommer-Strae 66,
38106 Braunschweig, Germany; 4Otto von Guericke Universitt Magdeburg, FNW-IEP,
I
Postbox 4120, 39016 Magdeburg, Germany
Abstract:
The dislocation configuration in epitaxial ZnO layers grown by metal organic

vapor phase epitaxy (MOVPE) was analyzed by transmission electron
microscopy. Misfit dislocations and the majority of threading dislocations are
characterized by Burgers vectors of the type 1/3 <11-20>. First results on the
electrical activity of dislocations are presented which are obtained by electron
holography in a transmission electron microscope. The evaluation of the phase
change of the electron wave in the vicinity of a dislocation yields a negative
line charge of approximately 3 e/nm.
Key words:
Metal organic vapor phase epitaxy (MOVPE), dislocations, electrical activity

of dislocations, transmission electron microscopy, electron holography
1.
INTRODUCTION
ZnO was rediscovered some years ago as a promising material for

optoelectronic devices because ZnO could be grown epitaxially at that time
by metal organic vapor phase epitaxy (MOVPE) and molecular beam
epitaxy (MBE). Although significant progress has been achieved in
improving the structural quality of epitaxial ZnO layers, the defect density
has to be further reduced before the fabrication of optoelectronic devices
becomes feasible. Apart from point defects, which determine the electrical
properties of the material, high concentrations of extended defects like
99
100
E. Mller et al.
dislocations and stacking faults are present which are known to be

detrimental for light-emitting devices. It is therefore of considerable interest
to study the configuration of extended defects in ZnO epilayers and correlate
the microstructure with the growth procedure. For this purpose, transmission
electron microscopy (TEM) has proven to be a suitable technique because it
allows the analysis of extended defects down to the atomic scale. Despite the
strongly increasing activity in epitaxial ZnO growth, only few TEM studies
have been devoted to a detailed study of the defect structure in ZnO. Narayan
et al. (1998) have analyzed defects in ZnO deposited by pulsed layer
deposition on (0001)Al2O3. The defect structure in MBE-grown ZnO layers
was studied by Hong et al. (2000) and Vigu et al. (2001). For ZnO
deposited directly on Al2O3(0001) substrates, a 30o rotation of the in-plane
orientation was observed independent of the growth technique, which
reduces the lattice-parameter mismatch from 32 % to 15 %.
The possible influence of dislocations on the overall electrical properties
of epilayers was recognized recently when electron holography in a
transmission electron microscope was used to reveal the electrical activity of
edge dislocations in epitaxial GaN layers (Cherns and Jiao, 2001). It is
known already for a long time that the electrical activity of dislocations can
be described by electronic states in a partially filled band acting as donors or
acceptors depending on the position of the Fermi energy with respect to the
occupation limit of the dislocation band (see review of Alexander and
Teichler, 1991). If present at a high density, dislocations in particular
charged dislocations - act as scattering centers which strongly reduces the
charge carrier mobility (Gerthsen, 1986). In addition, the charge carrier
concentration can be modified by several orders of magnitude. In a review by
Ossipyan et al. (1986), the electronic properties of dislocations in several IIVI semiconductors are summarized and a negative line charge of 1.5 e/nm
was assigned to 60o Zn(glide) dislocations on the basal plane of n-ZnO.
The present study is focused on dislocations in ZnO epilayers grown by
MOVPE. The distribution of edge, mixed and screw dislocations was
analyzed for layers grown under different MOVPE conditions using TEM.
Electron holography was applied to study the electrical activity of threading
dislocations.
2.
EXPERIMENTAL TECHNIQUES
The TEM investigations were performed in a 200 keV Philips CM 200

FEG/ST microscope which is equipped with a field emission gun.
Dislocation Burgers vectors b were analyzed on the basis of the b g = 0
extinction criterion using different imaging vectors g. The weak-beam
Properties of dislocations in epitaxial ZnO layers
101
technique was applied to improve the resolution compared to conventional

bright-field and dark-field images (see e.g. Williams and Carter, 1996).
Cross-section samples along the <11-20>- and <1-100>-zone axes were
prepared applying the technique described by Strecker et al. (1993) using 3.5
keV Xe+ ions for the final ion milling to minimize radiation damage with an
ion current of 1 mA and an incidence angle of 14 degrees.
For electron holography, the microscope is equipped with a Mllenstedt
biprism, which is installed in the selected-area aperture holder. The
electrostatic potential of the biprism was set at approximately 180 V leading
to an interference fringe distance of 0.2 nm. Holograms were obtained by
orienting a straight dislocation line parallel to the biprism whereas the
reference wave is transmitted through the undisturbed crystal on the other
side of the biprism. The reconstruction procedure for the amplitude and
phase of the electron wave is outlined in detail by Lehmann and Lichte
(2002).
All investigated samples were grown by MOVPE on Al2O3(0001)
substrates. Details of the growth procedure can be found in context with the
analyzed samples in chapter 3.
3.
DISLOCATIONS IN EPITAXIAL ZnO LAYERS
3.1
Structural and morphological properties
Dislocations are characterized by a Burgers vector and line direction,

which determine the dislocation type. Perfect dislocations in hexagonal
crystals have Burgers vectors ba=1/3<11-20>, bc=[0001] and ba+c=1/3<1123> (Hull, 1975). Misfit dislocations are generated at the interface between
the substrate and epitaxial layers, if they are grown on a substrate with a
different lattice parameter. The density of misfit dislocations depends on the
lattice-parameter mismatch and - during the early stages of the growth on
the layer thickness. The threading dislocation density is correlated with the
density of misfit dislocations but can be minimized by optimizing the
epitaxial growth.
Figure 1 shows cross-section TEM images of a MOVPE-grown ZnO
layer on Al2O3(0001) with a 40 nm GaN buffer layer deposited prior to the
ZnO growth. Isopropanol and Diethylzinc were used as precursors for the
ZnO deposition with a VI/II-flux ratio of 36 at total reactor pressure of 400
mbar and hydrogen as a carrier gas. The substrate temperature was 380 oC.
102
E. Mller et al.
Figure 1. Cross-section weak-beam images of the same area of a MOVPE-grown ZnO layer
on Al2O3(0001) with a 40 nm GaN buffer layer taken with (a) g = {0002} under g,3g
conditions and (b) g = {11-20} under g,3g conditions.
Figure 2. Bright-field image of the interface between the GaN/Al2O3(0001) interface and the
ZnO layer taken with g = {11-20}. The black line marks a GaN island.
Threading dislocations extend typically along the [0001] direction in

epilayers with a hexagonal crystal structure, which has been also observed
for MBE-grown ZnO epilayers (Hong et al., 2000). One possible origin of
this preferred line direction was attributed to the fact that the dislocations
form small angle grain boundaries between columnar grains (Vigu et al.,
2001) slightly twisted around the [0001] direction which is visible in Fig.1(b)
by contrast variations in the ZnO layer. Figure 1(a) was taken with g =
{0002} where edge dislocations with ba = 1/3<11-20> do not show any
contrast. The contrast of screw dislocation is extinct in Fig. 1(b) taken with g
= {1-100}. It is obvious from Fig. 1, that the majority of dislocations exhibit
edge-type character while the density of dislocations with bc and ba+c is
relatively low.
Imaging the interface region (Fig.2) shows, that the GaN buffer layer
consists of relatively large islands (a GaN island is indicated by the black
line in Fig.2), which could contribute to the mosaicity of the ZnO film.
103
The threading dislocation configuration depends significantly on the

buffer layer morphology and growth procedure, which can be inferred from
Fig. 3. This ZnO film was grown on a thick GaN buffer layer using a
procedure outlined in detail by Dadgar et al. (2004). A high-temperature
annealing treatment was applied after the buffer layer growth and again after
the deposition of the ZnO. Due to the thick GaN buffer layer, the mismatch
between the Al2O3(0001) and ZnO is reduced from 15 % to 1.8 % which is
accompanied by a reduction of the threading dislocation density. It is
remarkable that the threading dislocations in the GaN buffer layer do not
continue in the ZnO film but annihilate at the ZnO/GaN interface. The
dislocation lines in the ZnO deviate frequently from the [0001] direction
indicating a low degree of mosaicity, which is supported by the
homogeneous contrast of the ZnO layer. The extinction of the dislocation
contrast in Fig.2(a,b) shows that the majority of the dislocations is
characterized by Burgers vectors of the type ba=1/3<11-20> as in the layer
shown in Fig.1. However, different dislocation types are the result of the
random dislocation line orientations.
Figure 3. Cross-section weak-beam images of the same area of a MOPVE-grown ZnO layer
on Al2O3(0001) with a thick GaN buffer layer taken with (a) g = {0002} and (b) g = (11-20)
taken under g,3g conditions.
The large fraction of threading dislocations with ba can be understood by

analyzing the misfit dislocations at the interface between the buffer layer and
the ZnO film. Figure 4(a) displays a <11-20> zone-axis high-resolution
TEM (HRTEM) image of the interface of the sample depicted in Fig.3. The
{1-100} Fourier-filtered image Fig.3(b) shows that the misfit dislocations
104
E. Mller et al.
are characterized by one {1-100} plane inserted from below as expected due
to the larger a lattice parameter of the ZnO.
Figure 4. (a) HRTEM image of the interface between the GaN buffer layer and ZnO along the
<11-20>-zone axis and (b) Fourier-filtered image showing only {1-100} lattice fringes. The
arrows indicate {1-100} fringes inserted from the substrate side.
Figure 5 shows schematically the misfit dislocation orientation and

possible Burgers vectors (grey vectors). The electron-beam direction and
dislocation lines are aligned parallel to the [11-20] direction. 60o or 120o
misfit dislocations are characterized by ba = 1/3 [-12-10] and 1/3 [-2110].
The edge component be = [1-100] is compatible with an inserted (1-100)
plane as observed in Fig.4(b). It can assumed that three set of misfit
dislocations along the three <11-20> directions are present in the (0001)
interface plane.
The generation of misfit dislocations with ba can be understood by
considering the efficiency of the different possible Burgers vectors to relax
the lattice-parameter mismatch. Dislocations with ba are most efficient
because they are characterized by the longest Burgers vector edge
component be in the (0001) interface plane among all possible dislocation
Burgers vectors. In addition, the strain energy of dislocations is a b2 which
favors the formation of dislocations with the shortest overall Burgers vector
ba. Since dislocations cannot end inside a perfect crystal and the Burgers
vector is conserved along the dislocation line, the threading segments of
misfit dislocations are also characterized by ba as observed is Fig.1 and
Fig.3. The above analysis suggests that improving the structural quality of
the interface and using a thick GaN buffer layer should minimize threading
dislocations with Burgers vectors of the type 1/3 <11-20>.
105
Figure 5. Schematic representation of the orientation of a misfit dislocation (bold black line)
and possible Burgers vectors (grey vectors) in the (0001) interface plane
3.2
Electrical activity of dislocations
Due to the strong disturbance of the translational symmetry of the crystal,

dislocations with an edge component are expected to generate deep
electronic states in the energy gap of a semiconductor. If the Fermi level
does not coincide with the occupation limit of the band of electronic states
the dislocation will be charged. To maintain the overall neutrality of the
material, a charged dislocation is required to be shielded by a space charge
region of opposite sign. The resulting potential V at a distance r from the
dislocation line was already calculated by the Read (1954),
V( )
R2
r2
l 2 2 1
ln
4SHH 0 r R
Q
(1)
where Q=q e is the charge per dislocation length and H = 8.4 the dielectric
constant of the material (Viswanatha et al., 2004). The Read radius R =
q/
describes the radius of space charge region where Nsc is
density of compensating charges available for screening.
Electron holography is a TEM technique that allows the retrieval of the
phase of the electron wave. This provides the possibility to study the
interaction of the incident electrons with the electrostatic potential of a
charged dislocation with high spatial resolution. Fig.6 indicates the
106
E. Mller et al.
orientation of the dislocation with respect to the electron-beam direction,

which has been used in our study.
Figure 6. Geometry used for electron holography investigations: the electron beam is oriented
along the z direction and the dislocation along y perpendicular to the electron-beam direction.
After passing a sample with a thickness t, the phase of the transmitted

beam 'M is shifted with respect to the vacuum. If a charged dislocation is
present, the mean inner potential V0 of the material is locally modified by
the potential V(r) ( r
Eq.(2).
'M
x 2 z 2 ) which induces a phase shift given by
CE (V0 V (r )) dz
0
(2)
The constant CE = 7.29.106 is determined by the electron energy which was

measured precisely for our electron microscope by convergent beam electron
diffraction (Kruse et. al., 2003). Equation (2) only applies if kinematical
diffraction conditions are chosen. To eliminate the influence of dynamical
electron diffraction, the sample was tilted out of the zone axis by minimizing
the excitation of the Bragg reflections in the diffraction pattern.
In previous work (Cherns and Jiao, 2001) dislocations with [0001]-line
directions were aligned parallel to the electron beam, which is adequate to
maximize the contribution of the line charge to the phase shift of the electron
wave. However, the formation of pits at the intersection line of the
dislocation with the surface due to ion etching of the sample is difficult to
avoid and to detect. By analyzing embedded dislocations, thickness
modifications, which also shift the phase with respect to the surrounding
material according to Eq.(2), can be eliminated definitely. In addition,
dynamical contributions to the phase shift are more difficult to exclude if
107
dislocations and electron beam are parallel. The dislocation strain field can
locally tilt the lattice such that dynamical diffraction conditions apply in the
vicinity the dislocation, i.e. Bragg reflections are strongly excited, even if
kinematical conditions are chosen for the undisturbed crystal. The
contribution of dynamical phase shifts can be minimized in our approach as
will be shown later in this section.
On the basis of Eq.(2), a comparison of the phase shift measured by
electron holography and the calculated 'I yields the line charge q. Figure 7
shows the evaluated amplitude and phase of an electron hologram for the
transmitted beam. The dislocation line direction is indicated by the dashed
line in the amplitude image. Due to the noise, the position of the dislocation
is visible only by a weak dark contrast in the phase image.
Figure 7. Reconstructed amplitude and phase of the electron wave in an area containing a
dislocation indicated by the dashed line in the reconstructed amplitude.
To reduce the noise, the phase profile perpendicular to the dislocation

line along the x direction was obtained by averaging the phase along the y
direction inside the dark frame indicated in the phase image. The resulting
phase profile is plotted in Fig.8. The line represents the best fit of the
expected phase shift from Eq. (1) and (2) using the line-charge density as a
fit paramter. The best fit was obtained for a line charge q = 3 e/nm-1. The
calculation was carried out iteratively by optimising both, the line charge q
and R, because the Read radius R also depends on the line charge. Since the
density of positive screening charges NSC is not exactly known we have
varied Nsc in the range from 1017cm-3 to 1018cm-3, which yields Read radii
between 98 nm and 30 nm. The error due to the variation of Nsc corresponds
to 'q = r 0.024 e/nm-1. The retardation of the electron wave at the
dislocation indicates that the dislocation is negatively charged. As only the
electrical field induced by the charged dislocation contributes to the
108
E. Mller et al.
measured phase change with respect to the undisturbed crystal, the thickness
integration interval in Eq.(2) was chosen to be twice the Read radius.
Therefore, an accurate value for the sample thickness is not required.
Figure 8. Measured phase change perpendicular to the dislocation line and fit according to Eq.
(1) and Eq.(2).
The value for the line charge is reliable only if the effects of dynamical
diffraction can be neglected. To assess the influence of dynamical diffraction
on the phase shift, the worst case was considered by assuming that the
dislocation locally tilts the crystal from kinematical into {0002} or {11-20}
two-beam conditions. The resulting dynamical phase shift was calculated
using the Bloch wave formalism of the EMS program package (Stadelmann,
1987). Figure 9 shows the phase difference for the transmitted beam in the
undisturbed crystal as a function of the thickness comparing the phase under
kinematical conditions and for {0002} or {11-20} two-beam conditions. For
the tilt angle the experimental values (1.5 perpendicular and 5 along the
dislocation line with respect to the <1-100> zone axis) were taken into the
account. It follows from Fig. 9, that dynamical diffraction is negligible if the
distorted region in Bragg condition is smaller than 7 nm. With the
experimental geometry with a dislocation orientation perpendicular to the
electron-beam direction the assumption of a laterally limited distortion field
is more likely to be satisfied compared to a parallel orientation.
109
Figure 9. Phase difference for the transmitted beam in the undisturbed crystal comparing the
phase under kinematical conditions and for (lower curve) {0002} or (upper curve) {11-20}
two-beam conditions as a function of the thickness.
Regarding the accuracy of the determined line charge, it has to be taken

into account that the fit of the phase profile strongly depends on the
dielectric constant of ZnO with significantly differing values published in
the literature (Landolt Brnstein, 1987). Another source of error is the
sample thickness. If the sample thickness is smaller than the diameter of the
Read cylinder (2 R) the electrical field of the line charge leaks into the
vacuum region. Thus the effective dielectric constant decreases, leading to a
smaller value of q. For an accurate treatment of this situation the field
distribution outside the sample has to be calculated requiring an additional
parameter for the fit of the measured phase curve.
The determined value for q is large, but it agrees with the trend towards
high line charges in ZnO according to Ossipyan et al. (1986). The
discrepancy between our measured value and the value published for 60o
(Zn)gilde dislocation can be explained by the different core structure of
threading dislocations which do not lie on the (0001) plane. In our case, we
did not analyse the dislocation Burgers vector because the dislocation could
not be identified anymore after taking electron holograms. It has to be noted
that the electrical activity of dislocations is not necessarily induced by
intrinsic electronic dislocation states but can be also due to a high
110
E. Mller et al.
concentration of point defect accumulated in the vicinity of the dislocation

core.
The high negative dislocation line charge in n-ZnO indicates that
dislocations can provide acceptor states with a high concentration if the
dislocation density is high. This will lower the free electron concentration.
Due to the amphoteric nature of the dislocation states they act as donors, if
the Fermi energy is located below the occupation limit of the dislocation
states. Pinning of the Fermi energy at the occupation limit of the band is
expected if the concentration of shallow donors is reduced and the shallow
acceptor density is increased. Therefore, dislocation states could contribute
to the difficulty of obtaining p-type ZnO with low resistivity.
4.
SUMMARY
The morphology and Burgers vectors of threading and misfit dislocations

in MOVPE-grown ZnO heterostructures grown on Al2O3(0001) substrates
were analyzed by transmission electron microscopy. Most threading
dislocation are characterized by 1/3 <11-20> Burgers vectors which can be
considered as threading segments of misfit dislocations with the same
Burgers vectors. The line directions of threading dislocations extend
typically along the [0001] growth direction yielding a high fraction of edge
dislocations. High-temperature annealing treatments during the ZnO growth
lead to different line directions and consequently different threading
dislocations types. The threading dislocation density can be reduced by the
growth of a thick GaN buffer layer and the optimization of the interface
quality between the GaN buffer and the ZnO film. Analyzing the phase of
the transmitted beam by electron holography, indications for a high negative
line charge at threading dislocations are found. This observation emphasizes
that dislocations can have a strong influence on the mobility and
concentration of charge carriers in ZnO if they are present at a high density.
REFERENCES
Alexander, H., and Teichler, H., Dislocations, 1991, in: Materials Science and Technology,
Vol. 4, chap.6, W. Schrter, ed., North-Holland Pub. Company, pp. 249.
Cherns, D., and Jiao, C.G., 2001, Phys. Rev. Lett. 87, 205504.
Dadgar, A., Oleynik, N., Forster, D., Deiter, S., Witek, H., Blsing, J., Bertram, F., Krtschil,
A., Diez, A., Christen, J., and Krost, A., 2004, J. Cryst. Growth 267, 140 .
Gerthsen, D., 1986, phys. stat. sol. (a) 97, 527.
Hong, Soon-Ku, Ko, Hang-Ju, Chen, Y., Yao, T, 2000, J. Cryst. Growth 209, 537.
Hull, D., Introduction to Dislocations, 1975, 2ndd edition, Pergamon Press, pp.122.
111
Kruse, P., Rosenauer, A., and Gerthsen, D., 2003, Ultramicroscopy 96, 11.
Landolt Brnstein, 1987, Springer-Verlag, Heidelberg, 165
Lehmann, M., and Lichte, H., 2002, Microscopy and Microanalysis 8, 447.
Narayan, J., Dovidenko, K., Sharma, A.K., and Oktyabrsky, 1998, S., J. Appl. Phys. 84, 2597.
Osipyan, Yu. A., Petrenko, V.F., Zaretski, A.V., and Whitworth R.W., 1986, Adv. Phys. 35
115.
Read, W.T., 1954, Phil. Mag. 45, 775.
Stadelmann, P.A., 1987, Ultramicroscopy 21, 131.
Strecker, A., Salzgeber, U., and Mayer, J., 1993, Prakt. Metallographie 30, 482.
Vigu, F., Venngus, P., Vzian, S., Lagt, M., and Faurie, J.-P., 2001, Appl. Phys. Lett. 79,
194.
Viswanatha, R., Sapra, S., Satpati, B., Satyam, P.V., Dev, B. N., and Sarma, D.D., 2004, J.
Mater. Chem. 14, 661.
Williams, D.B., and Carter, C.B., Transmission Electron Microscopy, 1996, Plenum Press,
New York.
PART III: ROLE OF HYDROGEN
Chapter 10
MUON SPIN ROTATION MEASUREMENTS ON
ZINC OXIDE
E. A. Davis
Department of Materials Science and Metallurgy, University of Cambridge, Pembroke Street,
Cambridge CB2 3QZ, United Kingdom
Abstract:
The theoretical suggestion that hydrogen might form shallow donor states in
zinc oxide and hence account for the n-type conductivity normally found in
undoped samples has been confirmed by experimental studies using
mounium, a light pseudo-isotope of hydrogen. Characteristic frequencies in
muon spin rotation experiments yield a hyperfine constant that is ~104 times
smaller than that of vacuum-state muonium, indicating an extended orbital and
a shallow centre. Temperature-dependence studies yield an ionization energy
of about 30 meV. Band-offset diagrams and measurements on other
semiconductors suggest that hydrogen forms shallow donor centres when the
electron affinity of the host material is more than about 4 eV; otherwise the
hydrogen level lies deep in the energy gap.
Key words:
Hydrogen, muons, muonium, shallow donors
1.
INTRODUCTION
The simulation of hydrogen by muons has proved to be extremely

valuable in the identification of potential sites for hydrogen in
semiconductors and insulators. Although the muon has a mass one-ninth that
of the proton, its interaction with the host lattice, both electronically and
chemically, is virtually identical to that of a proton. During its 2.2
microsecond lifetime (experiments are frequently undertaken over a
timescale of up to ten lifetimes), the muon can diffuse, interact with, and
adopt positions in the lattice that protons themselves would occupy. If the
temperature is sufficiently low, muons can capture electrons to form
muonium atoms. The reduced mass of muonium is within 0.5% of that of
115
116
E. A. Davis
hydrogen and so its Bohr radius and ionization energy are essentially the
same as those of hydrogen.
Experiments using muons are generally easier and more direct than
experiments1 using hydrogen itself. One reason for this is the high sensitivity
of the technique, by which one can essentially see individual muons.
Another and more fundamental aspect is that hydrogen frequently forms a
U system, making the neutral state inaccessible under thermal
negative-U
equilibrium conditions. Such conditions do not apply in the muon
experiments, permitting observation of both the ionized (charged) and
unionized (neutral) states.
The theoretical prediction2 that hydrogen forms a shallow donor state in
ZnO has been confirmed by muon implantation studies3,4. Below 40 K a
distinctive beating of the muon precession signal provides the required
signature. Fourier transformation reveals a central line at the muon Larmor
frequency accompanied by two symmetrically disposed satellites, the
separation of which yields a hyperfine constant about 10,000 times smaller
than that for vacuum-state muonium. Confirmation that this signal represents
the extended orbital of a shallow centre has been obtained by studies as a
function of temperature, which reveal the disappearance of the satellites with
an ionization energy of approximately 30 meV. More recent measurements5
in longitudinal magnetic fields provide further evidence for these findings.
2.
EXPERIMENTAL DETAILS AND RESULTS
The experiments were undertaken at the ISIS pulsed muon facility at the
Rutherford Appleton Laboratory situated near Oxford in England. Positive
muons of energy 4 MeV and 100% spin polarized are brought to rest within a
few tenths of a millimetre in powder or single-crystal samples mounted
within a cryostat.
The technique of muon spin rotation involves applying a magnetic field
perpendicular to the direction of the incoming beam of muons (transverse to
their spin) and monitoring the resulting precession signal via the emission of
positrons that are emitted preferentially in the direction of the muon spin at
the moment of its radioactive decay. For the bare muons this is simply the
Larmor frequency but for muonium several frequencies are observed. In the
case of a small hyperfine constant, one can easily reach the so-called
Paschen-Back regime in moderate fields and then a triplet of lines is seen in
a Fourier transform of the raw data.
Asymmetry (%)
Muon spin rotation measurements in zinc oxide
117
20
0
-20
0
Probability Density
0.06
0.05
10
Time (s)
15
20
ZnO powder
5K
0.04
0.03
0.02
0.01
0.00
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
Frequency (MHz)
Figure 1. Muonium spin precession signal for a ZnO powder sample at 5 K. The upper plot is
the raw time-domain spectrum (corrected for the muon decay) while the lower plot is the
corresponding frequency spectrum. The central line corresponds to the Larmor frequency of
the bare muon (ionized muonium) and the two symmetrically disposed satellites are associated
with muonium. The dotted curve is a theoretical fit using a powder-pattern lineshape.
The triplet of frequencies gives rise to a beating in the time-domain

spectrum as seen for a powder sample of ZnO in Figure 1.The separation of
the two satellite lines provides a direct measure of the hyperfine coupling
constant between the muon and the electron. This is 500 r 20 kHz, which is
0.011% of the free-muonium value of 4463 MHz, indicating immediately a
small electron spin density at the site of the muon and an extended
wavefunction associated with a shallow donor state.
Studies of the temperature dependences of the lines reveal that, as the
temperature is raised, the central line increases in amplitude at the expense of
the satellites. This is to be expected for ionization of the donor state. The
corresponding Arrhenius plots, shown in Figure 2, produce activation
energies for the central (diamagnetic) line of ~ 26 meV and of the satellite
(paramagnetic) lines of ~ 33 meV. The results imply that muonium in ZnO
ionizes above ~ 40 K with an activation energy of about 30 r 3 meV.
118
E. A. Davis
100 60
40
20
10
Asymmetry (%)
20
26(7) meV
2.5(9) meV
15
Diamagnetic fraction
10
10
20
30
40
50
60
70
10
100
00
800 60
100
60
12
2
12.5
ZnO powder
TF 200G
25
Asymmetry (%)
T (K)
80
-1
1000 / T (K )
40
40
T (K)
20
20
12.5
ZnO powder
TF 200G
2.4(8) meV
33(12) meV
4
2
0
10
Paramagnetic fraction
20
30
40
50
60
70
80
-1
1000 / T (K )
Figure 2. Arrhenius plots of the temperature dependences of the amplitudes of the lines shown
in Figure 1. (Courtesy of J M Gil.)
The smaller activation energies seen at lower temperatures are probably

associated with an exchange of electrons between muonium and other donors
(hydrogen?) already present in the samples.
One might question whether the energy of 30 meV represents the depth of
the donor level below the conduction band edge? Under conditions of
thermal equilibrium, carrier activation energies yield the donor depth if
compensating acceptors are present but one-half the donor depth for the case
of no compensation. In the case of implanted muons, thermal equilibrium is
not reached and the correct interpretation is a matter for debate. Doubling our
activation energies would imply a donorr depth for muonium or hydrogen of
~ 60 r 6 meV, which is closer to the theoretical hydrogenic value see
next section.
Investigations by others have yielded a variety of possible values for the
depth of hydrogen donors in ZnO. Hall effect studies6 have been analysed in
terms of the ionization of two donors having energies of 31 meV and
61meV, with the authors preferring to associate the lowerr of these two values
with hydrogen donors. Hofmann et al.1, as part of their ENDOR
investigations, also made Hall effect measurements and found two activation
energies with values 35 and 66 meV. Early photoluminescence data7 gave 52
meV as the depth of a hydrogen donor. More recent photoluminescence
studies8 suggests a lower value of 40 meV.
3.
FURTHER ANALYSIS
Assuming for the sake of simplicity an isotropic centre, the effective Bohr
radius a can be obtained directly from the hyperfine constant A ,
since this scales as the third power of the radius:

a
13
a0 A0 A
119
(1)
where A0 is the free-muonium hyperfine constant and a 0 is the Bohr

radius. Taking A0 = 4463 MHz and a 0 = 0.053 nm, we find a = 1.1 nm.
This can be compared with the value estimated from a hydrogenic model:
a
a0H me m
(2)
where H is the relative permittivity and m the effective mass for electrons
in ZnO. Using the values H = 8 and m me = 0.24, gives a = 1.7 nm, in
fair agreement with the value deduced from the data.
In a similar vein we can estimate the ionization energy from the
hyperfine constant:
I
13
I 0 A A0 / H
(3)
where I 0 is the Rydberg = 13.6 eV. This yields I = 51 meV, which can be
compared with the hydrogenic value, given by:
I
I 0 m me / H 2
(4)
of 50 meV. The agreement is good, bearing in mind the assumptions in the

hydrogenic model no central-cell corrections for example.
4.
SINGLE-CRYSTAL STUDIES
The experiments described above were made on powder samples of

99.999% purity from Alfa Aesar. Single-crystal studies have also been
undertaken5. Figure 3 shows spectra of an Eagle-Picher sample taken at two
different orientations of the crystallographic c-axis with respect to the muon
beam; the inset gives the orientation dependence of the separation of the
main satellite lines.
For an anisotropic centre this splitting is given by
'Q
D
E
2
(5)
120
E. A. Davis
where A is the isotropic part of the hyperfine tensor and D is a dipolar

term. These data yield a very similar value of A (namely 490 r 10 kHz) to
the powder sample studies but confirm an anisotropy of the centre with D =
260 r 20 kHz. From the angular dependence one can deduce that the
symmetry axis of the centre is parallel to the c-axis.
Figure 3. Frequency spectrum derived from muonium spin rotation measurements on a singlecrystal sample of ZnO as a function of orientation (from reference 5).
Of the possible sites that hydrogen (muonium) can occupy in the wurtzite
lattice (see Figure 4) the BCparr site has been shown theoretically to have the
lowest energy, i.e. to be the most stable site9. Experimentally we cannot
distinguish between this site and the site antibonding to oxygen along the caxis.
BCpar
BCperp
ABApar
ABAperp
ABCpar
ABCperp
Figure 4. Six principal sites for hydrogen in the wurtzite structure. BC = bond centre and AB
= antibonding. Superscripts denote sites adjacent to the cation(C) and anion (A). Perp and
par refer to directions perpendicular (not exactly) and parallel to the c-axis (from ref. 11).
121
Shimomura et al.10 have also made muonium studies on single crystals

of ZnO and claim to have identified two distinct shallow centres, one
associated with each of the above sites.
5.
SHALLOW VERSUS DEEP
The reason why muonium (and by implication hydrogen) acts as a

shallow centre in some materials and as a deep centre in others is intriguing.
It appears that the electron affinity of the host is a crucial parameter, as
revealed by the plot shown in Figure 5.
1.2
SiO2
Normalized Hyperfine constant
1.0
AlN
MgO
ZnSe
ZnS
0.8
GaAs
GaP
SiC
0.6
Ge
Si
0.4
0.2
CdTe/S/Se
ZnO
InN
0.0
Electron Affinity (eV)
Figure 5. Correlation of the normalized muonium hyperfine constant with electron affinity.
(Courtesy of S F J Cox reference 11.)
Here the muonium hyperfine constant, relative to the vacuum-state value,

is plotted versus electron affinity for a variety of materials, from insulators
122
E. A. Davis
such as SiO2 and diamond on the left (for which A A0 is close to unity) to
semiconductors in the middle of the plot. The five semiconductors with an
effectively zero (on this scale) hyperfine constant are those in which
muonium forms shallow centres. It is evident that the dilation of the
wavefunction from atomic-like to extended occurs rather suddenly on this
plot, corresponding to an electron affinity of about 3.7 eV.
The concept of the electron affinity of the host being the all important
factor influencing the deep to shallow transition is implicit in the band-offset
diagrams proposed by Van de Walle12. An example is shown in Figure 6 for
a few materials. In this diagram Ec represents the energy of the bottom of the
conduction band and Ev the top of the valence band, both plotted on an
absolute energy scale, i.e. with respectt to the vacuum level. The dashed line
marked +/- gives the energy at which the formation energies of the positively
and negatively charged states of hydrogen are equal. This represents the
position at which the Fermi level would be pinned in a negative U system.
Figure 6. Band-offset diagram for several semiconductors (from reference 13).
If this level lies in the conduction band (as in ZnO and InN) hydrogen
forms a shallow donor level. If the level lies in the band gap (as for the other
three materials) then hydrogen forms a deep centre. The predictive nature of
this model is currently being tested for other semiconductors, in particular
oxides other than ZnO.
Peacock and Robertson14 have questioned whether the hydrogen level
really does lie at a constant depth below the vacuum level. A subtle point
relevant to this question is that the donor state actually lies at the level +/0,
i.e. the energy through which the Fermi level would pass when the centre
ionizes, changing from the neutral to the positively charged state. This level
differs by U
U/2 from the +/- level and so, even if the +/- level is invariant, the
123
+/0 level is not expected to be so. Under thermal equilibrium conditions the
Fermi level is pinned at +/- (for a ve U system). Such conditions do not
apply in the muonium experiments and so we are able to explore the higher
+/0 level directly, without the need to illuminate the sample to reveal the
neutral state, as required for, say, ENDOR experiments using hydrogen
itselff1.
ACKNOWLEDGEMENTS
The author wishes to acknowledge fruitful and enjoyable collaboration
with several colleagues during the experimental work reported here. These
are S F J Cox (Rutherford Appleton Laboratory and University College
London), P J C King, J S Lord and S P Cottrell (Rutherford Appleton
Laboratory), J M Gil, H Alberto, R Vilo, J Piroto Duarte and N Ayres de
Campos (Coimbra University, Portugal), and R Lichti (Texas Tech
University, USA).
REFERENCES
1. D. M. Hofmann, A. Hofstaetter, F. Leiter, H. Zhou, F. Henecker, B. K. Meyer, S. B.
Orlinskii, J. Schmidt, and P. G. Baranov, Phys Rev Lett. 88, 045504 (2002).
2. C. G. Van de Walle, Phys. Rev. Lett. 85, 1012 (2000).
3. S. F. J. Cox, E. A. Davis, S. P. Cottrell, P. J. C. King, J. S. Lord, J. M. Gil, H. V. Alberto,
R. C. Vilo, J. Piroto Duarte, N. Ayres de Campos, A. Weidinger, R. L. Lichti, and S. F.
C. Irvine, Phys. Rev. Lett. 86, 2601 (2001).
4. J. M. Gil, H. V. Alberto, R. C. Vilo, J. Piroto Duarte, P. J. Mendes, L. P. Ferreira, N.
Ayres de Campos, A. Weidinger, J. Krause, E. A. Davis, S. P. Cottrell, and S. F. J. Cox,
Phys. Rev. B 64, 075205 (2001).
5. H. V. Alberto, R. C. Vilo, J. Piroto Duarte, N. Ayres de Campos, R. L. Lichti, E. A. Davis,
S. P. Cottrell, and S. F. J. Cox, Hyperfine Interact. 136/137, 471 (2001).
6. D. C. Look, D. C. Reynolds, J. R. Sizelove, R. L. Jones, C. W. Litton, G. Cantwell, and W.
C. Harsch, Solid State Commun. 105, 399 (1998).
7. D. C. Reynolds and T. C. Collins, Phys. Rev. 185, 1099 (1969).
8. D. C. Look, C. Coskun, B. Clafin, and G. C. Farlow, Physica B 340-342, 32 (2003)
9. E V. Lavrov, J. Weber, F. Brrnert, C. G. Van de Walle, and R. Helbig, , Phys. Rev. B 66,
165205 (2002).
10. K. Shimomura, K. Nishiyama, and R. Kadono, Phys Rev Lett. 89, 255505 (2002).
11. S. F. J. Cox, J. Phys.: Condens. Matter 15, R1727 (2003).
12. C. G. Van de Walle and J. Neugebauer, Nature 423, 626 (2003).
13. E. A. Davis, S. F. J. Cox, R. L. Lichti, and C. G. Van de Walle, Appl. Phys. Lett. 82, 592
(2003).
14. P. W. Peacock and J. Robertson, Appl. Phys. Lett. 83, 2025 (2003).
Chapter 11
HYDROGEN DONORS IN ZINC OXIDE
M. D. McCluskey and S. J. Jokela
Department of Physics, Washington State University, Pullman, WA 99164, U.S.A.
Abstract:
Zinc oxide (ZnO) has emerged as a leading material for micro- and
optoelectronic applications.
Although the fabrication of ZnO, from
nanocrystals to bulk single crystals, is well established, a major roadblock for
fabricating optoelectronic devices is the lack of reliable p-type doping. The
presence of compensating donors inhibits the growth of p-type ZnO. In this
paper, studies pertaining to the microscopic structure and doping properties of
hydrogen in ZnO are described. Results from infrared (IR) spectroscopy are
consistent with a model where the hydrogen attaches to a host oxygen atom, in
an anti-bonding orientation, which is not aligned along the c axis. These
hydrogen complexes are unstable, however, perhaps due to the formation of
H2 molecules.
Key words:
IR spectroscopy, infrared, hydrogen, vibrational modes, H donors, O-H

complexes
1.
INTRODUCTION
Zinc oxide (ZnO) has a wide range of industrial applications. An

environmentally friendly, wide-bandgap semiconductor, ZnO may prove to
be a valuable material for blue lasers and light-emitting diodes.1 The lack of
reliable p-type doping, however, is major roadblock to the realization of
such devices. Compensating donors play a detrimental role in preventing ptype conductivity. Early work by Mollwo2 and Thomas and Lander3 showed
that the diffusion of hydrogen into ZnO could increase the electrical
conductivity. The more recent theoretical work by Van de Walle,4 which
showed that hydrogen should be a shallow donor in ZnO, re-ignited interest
in this subject. Subsequent experiments involving muonium in ZnO5 and
electron-nuclear double resonance (ENDOR) spectroscopy6 verified that
hydrogen is a shallow donor in ZnO. In this paper, we review results from
125
126
infrared (IR) spectroscopy of hydrogen diffused into ZnO.7-9 Preliminary

Hall-effect and high-pressure results are also presented.
2.
EXPERIMENTAL TECHNIQUES
Single crystal, c-cut ZnO samples were purchased from Cermet, Inc.
Samples were sealed in quartz ampoules along with hydrogen or deuterium
gas. In addition, ZnO powder was placed in the ampoule, in order to prevent
the decomposition of the ZnO crystal due to hydrogen reduction. Hydrogen
diffusion occurred at a temperature of 700-800C for 10-100 hr, followed by
a rapid quench to room temperature by immersion of the ampoule in water.
The sample was then retrieved by breaking the ampoule.
IR spectra were taken at room temperature (300 K) and liquid-helium
temperatures (5-15 K), using a Bomem DA8 Fourier transform infrared
(FTIR) spectrometer and an InSb detector. For the low-temperature
measurements, a Janis continuous-flow liquid-helium cryostat with wedged,
IR-transparent windows was utilized. Hall-effect measurements, in the Van
der Pauw geometry, were performed at room temperature using a system
from MMR Technologies. Wires were attached to the ZnO using silver paint,
which provided adequate Ohmic contacts for the electron concentrations
(1017 cm-3) in these samples.
3.
RESULTS
3.1
IR spectroscopy and Hall-effect measurements
At liquid-helium temperatures, hydrogenated ZnO samples exhibit an

O-H stretch mode at a frequency of 3326 cm-1 at liquid-helium temperatures
(Fig 1). The IR peak corresponding to this mode broadens and shifts to
higher frequency when the sample is warmed to room temperature.
Deuterated ZnO samples, at liquid-helium temperatures, have an O-D stretch
mode at 2470 cm-1.7 The ratio of the O-H and O-D frequencies are in good
agreement with a simple diatomic model. More importantly, the frequency
ratio is nearly identical to that of O-H and O-D complexes in GaP.10 Our
observed O-H mode is different from those observed by Lavrov et al.11 and
Nickel and Fleischer.12
At a temperature of 300 K, the O-H complexes are unstable. Over the
course of a few months, the intensity of the IR peak decreases substantially.
In order to quantify this decay process, we prepared two identical
Hydrogen donors in zinc oxide
127
Absorbance
0.15
0.10
T = 300 K
0.05
T=8K
0.00
3300
3320
3340
3360
-1
Wave numbers (cm )

Figure 1. IR spectra of O-H complexes in ZnO at room- and liquid-helium temperatures.
hydrogenated samples. For one sample, the electron concentration was

determined as a function of time, using room-temperature Hall-effect
measurements. For the other sample, room-temperature IR spectroscopy was
used to measure the intensity of the O-H peak as a function of time. The
results of these measurements are plotted in Fig. 2. The circles and squares
refer to the Hall-effect and IR measurements, respectively.
In Fig. 2, there are two features worth noting. First, the decay of the O-H
peak and the decay of the free-electron concentration are correlated,
providing evidence that the O-H complexes are shallow donors. Second, the
data were fit using a biexponential decay model,
N
1
1/ N 0
At
(1)
where N is the number of O-H complexes, N0 and A are adjustable

parameters, and t is time. This model is consistent with the idea that
hydrogen donors may form H2 molecules.13 Since H2 molecules are
0.25
Absorbance
-3
Electron Density (10 cm )
4.4
4.0
17
0.20
3.6
3.2
0.15
2.8
0.10
2.4
0.05
Absorbance (arb. units)
128
2.0
0
200
400
600
800
1000
0.00
Time (hr)
Figure 2. Electron density (circles) and IR absorbance of the O-H peak (squares) as a function
of time, in hydrogenated ZnO.
essentially IR inactive, and electrically neutral, the formation of H2

molecules results in a decrease in the O-H IR mode and
d free-electron
concentration.
3.2
Polarized IR spectroscopy
The IR measurements described in the previous section were performed

in a normal incidence geometry on c-cut ZnO samples. Since the
polarization of the light is perpendicular to the c axis, it follows that the O-H
complexes that we observed are nott aligned along the c axis. This
observation contradicts some theoretical reports, which claim that the lowest
energy orientation of the hydrogen is parallel to the c axis.4 The calculated
energy differences are small, however, on the order of 0.1 eV.
To check this result, we performed polarized IR spectroscopy on an a-cut
sample. For light that was polarized along the c axis, the O-H absorption is
reduced as compared to light perpendicular to the c axis. As discussed in
Ref. 8, our results are consistent with O-H dipoles that are aligned at an
angle of ~112 to the c axis.
129
Figure 3. Polarized IR spectra of O-H complexes in ZnO, at room temperature.
3.3
High-pressure measurements
Regarding the structure of the O-H complexes, one question remains: is

the configuration bond-centered or anti-bonding? Since both structures have
identical symmetry, it is not obvious how to distinguish between them. One
possible method involves measuring the O-H stretch-mode frequency as a
function of hydrostatic pressure. By comparing the results with predictions
of ab initio calculations, it should be possible to determine the structure with
reasonable certainty.
[0001]
Zn
O
H
Figure 4. Ball-and-stick diagrams of the antibonding (left) and bond-centered (right) models
for hydrogen donors in ZnO.
130
High-pressure measurements were performed at room temperature, using

a moissanite-anvil cell. As shown in Fig. 5, there is only a very slight shift
with pressure, approximately -1 cm-1/GPa. Qualitatively, this small shift
would seem to favor the antibonding orientation, in which the hydrogen is
not crowded by neighboring atoms. However, it will be necessary to compare
our results with first-principles calculations. In addition, IR measurements
over a larger range of pressures will be performed.
Absorbance (arb. units)
1.1 GPa
0 GPa
3300
3320
3340
3360
-1
Wave numbers (cm )
Figure 5. IR spectra of hydrogenated ZnO, at room temperature, at hydrostatic pressures of 0

and 1.1 GPa.
4.
UNSOLVED MYSTERIES
In addition to the bond-centered versus antibonding question, there are

two results that do not have good explanations. First, as shown in Fig. 6,
hydrogenated samples from Cermet show a much stronger peak than
identically treated samples from Eagle-Pitcher. This may be due to a higher
concentration of oxygen vacancies in Eagle-Picher material, which could
trap hydrogen atoms. The second mystery is the larger intensity of the O-H
peak as compared to the O-D peak (Fig. 7). The larger intensity may be
related to hydrogens faster diffusion and larger vibrational amplitude, but
the size of the difference (x10) is puzzling.
131
Figure 6. IR spectra of hydrogenated ZnO, at liquid-helium temperatures, for samples

purchased from Cermet and Eagle-Picher.
Figure 7. IR spectra of O-H (left) and O-D (right) complexes in ZnO.
132
5.
CONCLUSIONS
Hydrogen donors can be introduced by annealing ZnO in a hydrogen

ambient, resulting in O-H complexes. These donor complexes are unstable,
however, and decay away in a few months at room temperature. It is likely
that most of the O-H complexes form H2 molecules in the sample. Results of
IR spectroscopy are consistent with an O-H complex that is in an
antibonding configuration, oriented at some angle to the c axis.
ACKNOWLDGMENTS
The authors would like to thank S. Limpijumnong, M. Stavola, C.G. Van
de Walle, and S.B. Zhang for helpful discussions. This work was supported
by NSF Grant No. DMR-0203832.
REFERENCES
1.
2.
3.
4.
5.
D. C. Look, Mater. Sci. Eng. B 80, 383 (2001).

E. Mollwo, Z. Phys. 138, 478 (1954).
D. G. Thomas and J.J. Lander, J. Chem. Phys. 25, 1136 (1956).
C. G. Van de Walle, Phys. Rev. Lett. 85, 1012 (2000).
S. F. J. Cox, E. A. Davis, S. P. Cottrell, P. J. C. King, J. S. Lord, J. M. Gil, H. V. Alberto,
R. C. Vilo, J. Piroto Duarte, N. Ayres de Campos, A. Weidinger, R. L. Lichti, and S. J. C.
Irvine, Phys. Rev. Lett. 86, 2601 (2001).
6. D. M. Hoffman, A. Hofstaetter, F. Leiter, H. Zhou, F. Henecker, B. K. Meyer, S. B.
Orlinskii, J. Schmidt, and P. G. Baranov, Phys. Rev. Lett. 88, 045504 (2002).
7. M. D. McCluskey, S. J. Jokela, K. K. Zhuravlev, P. J. Simpson, and K. G. Lynn, Appl.
Phys. Lett. 81, 3807 (2002).
8. S. J. Jokela, M. D. McCluskey, and K. G. Lynn, Physica B 340-342, 221 (2003).
9. M. D. McCluskey and S. J. Jokela, in MRS Proc. Vol. 813, edited by N. H. Nickel, M. D.
McCluskey, and S. B. Zhang (Materials Research Society, PA, 2004).
10. W. Ulrici, M. Czupalla, and M. Siefert, Phys. Status Solidi B 210, 551 (1998).
11. E. V. Lavrov, J. Weber, F. Brrnert, C. G. Van de Walle, R. Helbig, Phys. Rev. B 66,
165205 (2002).
12. N. H. Nickel and K. Fleischer, Phys. Rev. Lett. 90, 197402 (2003).
13. G. Alvin Shi, M. Saboktakin, M. Stavola, and S. J. Pearton (unpublished).
Chapter 12
HYDROGEN-RELATED DEFECTS IN ZnO
STUDIED BY IR ABSORPTION SPECTROSCOPY
E.V. Lavrov,1,2 F. Brrnert,1 and J. Weber1
1
Technical University Dresden, 01062 Dresden, Germany; 2Institute of Radioengineering and
Electronics of RAS, 101999 Moscow, Russia
Abstract:
Most of the ZnO growth techniques result in n-type conductivity of the crystal,
which hinders the progress of ZnO applications for electronic devices.
Hydrogen incorporated during the process of crystal growth is now considered
as a likely source of the n-type conductivity of ZnO. Infrared absorption
spectroscopy provides detailed insights into the physical properties of the light
impurities in the semiconductor matrix and is, therefore, an excellent tool to
explore the structure of the hydrogen-related defects embedded in ZnO. We
report a number of hydrogen-related defects observed in hydrothermally
grown ZnO and ZnO grown from the vapor phase studied by Fourier
transform infrared absorption spectroscopy. Three IR absorption lines at
3611.3, 3349.6, and 3312.2 cm1 at 10 K are observed after hydrogenation of
the vapor phase grown ZnO. The line at 3611.3 cm1 is tentatively assigned to
a bond-centered hydrogen, whereas the other two are identified as a Zn
vacancy decorated with two hydrogen atoms. The nature of a hydrogen center
responsible for the 3577.3 cm1 line observed at 10K in as-grown
hydrothermal ZnO is discussed.
Key words:
Hydrogen, hydrogen-related defects, IR absorption, Zn vacancy
1.
INTRODUCTION
Due to recent progress in crystal growth1 and the unique piezoelectric,

optical, and electrical properties, ZnO is considered at the moment as a
serious alternative to GaN for use in optoelectronic devices.2 ZnO nearly
always exhibits n-type conductivity. The nature of this conductivity has been
discussed for years and is normally attributed to native defects, such as the
Zn-on-O antisite (ZnO), the Zn interstitial (Zni), and the O vacancy (VO).3
133
134
E. V. Lavrov et al.
However, a new cause for the n-type conductivity of ZnO has recently been
proposed: First-principles investigations based on density functional theory
suggest that hydrogen in ZnO occurs exclusively in the positive charge state,
i.e., it always acts as a donor.4 Recent muon spin rotation (PSR)5 and EPR6
studies confirmed the presence of hydrogen-related shallow donors.
However, the molecular structure of these defects remains unknown.
Local vibrational mode (LVM) spectroscopy is a tool suited to study this
problem. The knowledge about LVMs supplies a detailed information about
the physical properties of light impurities embedded in ZnO. The
frequencies of the LVMs reveal directly the chemical binding of hydrogen
with its neighbors due to the dependence on the atomic structure of the
hydrogen-related defects.
To our knowledge, the first infrared absorption study of hydrogen-related
LVMs in ZnO was made in the end of the 70th by Grtner and E. Mollwo.7,8
They investigated copper doped ZnO annealed in a hydrogen atmosphere
and identified a set of LVMs originating from different copper-hydrogen
defects.
Interest for hydrogen behavior in ZnO stimulated further studies of
LVMs hydrogen-related defects. McCluskey et al. reported an O-H stretch
mode at 3326.3 cm1 observed at 8 K in ZnO samples that were annealed at
700 oC in hydrogen atmosphere.9 The 3326.3 cm1 line was assigned to a
hydrogen atom in an antibonding configuration oriented at an angle of 100o
to the c axis.
Recently, Nickel and Fleischer reported a Raman scattering study of asgrown state-of-the-art ZnO. A number of Raman lines were detected, which
were ascribed to LVMs of N-H and C-Hn complexes with n=1,2,3.10
Here, we summarize our results of an infrared absorption study of
hydrogen-related centers in ZnO grown by different techniques.11,12 A
number of O-H LVMs has been detected. The lines at 3611.3, 3349.6, and
3312.2 cm1 are observed after hydrogen plasma treatment in nominally
undoped ZnO grown from the vapor phase. The 3611.3 cm1 line is
tentatively identified as a LVM of bond centered (BC) hydrogen, whereas
the other two modes are assigned to LVMs of a zinc vacancy decorated with
two hydrogen atoms. The origin of an IR absorption line at 3577.3 cm1
detected at 10 K in as-grown hydrothermal ZnO is discussed.
2.
EXPERIMENTAL DETAILS
Two types of ZnO crystals have been employed in the current study.
Nominally undoped n-type ZnO crystals grown from the vapor phase (VP) at
Hydrogen-related defects in ZnO studied by IR absorption
135
the Institute for Applied Physics, University Erlangen (Germany) with

resistivity of 10-100 : cm had a shape of hexagonal prisms with a diameter
of ~2 mm and a length of ~20 mm.13,14 Hydrothermally grown (HT) n-type
(1120) ZnO substrates supplied by CrysTec GmbH had sizes of 5x5x1 mm3
with the c axis parallel to one of the edges of the substrate. The carrier
concentration at RT was 1017 cm3.
For hydrogenation the samples were exposed for 1-19 h to a remote
hydrogen and/or deuterium dc plasma in a parallel plate system, with a plate
voltage of 1000 V. The samples were mounted on a heater block held at a
temperature of 150-380 oC and placed 10 cm downstream from the plasma
with a bias voltage of about 300 V, which fixed the bias current to ~40 PA.
The gas pressure was held at ~1 mbar.
In order to study the annealing behavior of the hydrogen-related defects,
infrared absorbance spectra were recorded at 10 K after each step in a series
of isochronal heat treatments (annealings) at temperatures in the range 4001200 oC. The annealings were performed in a furnace purged with argon gas
and the duration of each treatment was 30 min.
Infrared absorbance spectra were recorded with a BOMEM DA3.01
Fourier transform spectrometer equipped with a globar light source, a KBr
beam splitter, and a liquid-nitrogen-cooled mercury cadmium telluride
(MCT) detector. The samples were free standing in the sample compartment
of an exchange-gas cryostat, where the temperature was measured with a
platinum resistor. The measurements were performed in the temperature
range 9-120 K. Polarized light was produced by placing a wire-grid polarizer
with KRS-5 substrate in front of the cryostat.
3.
VAPOR PHASE GROWN ZnO
3.1
Results
Figure 1 shows typical infrared absorption spectra measured at 9 K for a

VP ZnO sample after exposure to a hydrogen and/or deuterium plasma at
350 oC for two hours.11 The spectra have been measured with the c axis of
the crystal perpendicular to the beam. Three IR absorption lines at 3611.3,
3349.6, and 3312.2 cm1 are seen in the hydrogen treated sample (top
spectra).11 When hydrogen is substituted by deuterium (mid spectra), the
lines shift downwards in frequency to 2668.0, 2486.3, and 2460.7 cm1,
respectively. The frequency ratio of the hydrogen- and deuterium-related
lines is 1.35, which is close to the value expected for a harmonic oscillator
comprising a hydrogen atom bound to an oxygen atom, ((Dr/H
Hr)1/2=1.37,
136
E. V. Lavrov et al.
where Hr and Dr are the reduced masses of the 16O-H and 16O-D species,
respectively. The LVM frequencies strongly suggest that the new lines are
stretch LVMs of the O-H species. When the sample is treated with a mixed
hydrogen and deuterium plasma, four additional lines at 3346.6, 3315.2,
2484.6, and 2463.0 cm1 are observed (bottom spectra in Fig. 1). No extra
lines are seen near the 3611.3 and 2668.0 cm1 lines. This implies, that two
different defects are responsible for the 3611.3, 3349.6, and 3312.2 cm1
lines. The LVM at 3611.3 cm1 belongs to a defect containing one hydrogen
atom, whereas the LVMs at 3349.6 and 3312.2 cm1 originate from a defect
containing two nonequivalentt hydrogen atoms. We label these defects H-I
and H-II.
Figure 2 shows the results of isochronal annealing experiments on ZnO
samples treated with a hydrogen plasma at 350 oC. It follows from the figure
that the H-I defect anneals out at 500 oC, whereas H-II is more stable and
disappears from the spectra at 600 oC. Parallel to the annealing of H-I and HII, another line at 3191.6 cm1 grows in the spectra. It represents a stretch
LVM of the well-known Cu-H defect.7,8 It reaches its maximum intensity at
600 oC, when both H-I and H-II are already gone from the spectra.
Figure 1. Absorption spectra measured for vapor phase grown ZnO samples at 10 K after: H-,
D-, and H+D-plasma treatments at 350 oC. Unpolarized light.
137
Figure 2. The intensity (integrated absorbance) of the 3191.6, 3312.2, 3349.6, and 3611.3
cm1 lines measured at 10 K for a vapor phase grown ZnO sample vs. the annealing
temperature. The sample was exposed to a hydrogen plasma at 350 C.
Bond angles of the O-H species comprising the H-I and H-II defects with
respect to the c axis of the crystal may be obtained from the polarized IR
absorption spectra. For the stretch modes, the direction of the dipole moment
should coincide with the bond direction, provided each bond vibrates
independently.
The H-I defect contains only one O-H bond, therefore the assumption that
its direction coincides with the induced dipole moment is fulfilled
automatically. The H-II defect contains two O-H species and according to
the isotope substitution data (see bottom spectrum in Fig. 1), the coupling
between the two LVMs of the defect is rather weak, i.e. substitution of one
hydrogen atom comprising the defect with deuterium does not change the
LVM frequency of the second O-H bond considerably. Thus, the O-H bonds
of H-II vibrate independently, and from the polarized absorption spectra we
get directly the bond angles.
The results of polarization dependent absorption measurements are
presented in Fig.3. It shows the normalized intensities of the LVMs
138
E. V. Lavrov et al.
belonging to the H-I and H-II defects as a function of the polarizer angle (-)
with respect to the c axis of the sample. As one can see from the figure, the
LVM of H-I and the LVM of H-II at 3312.2 cm1 are polarized along the c
axis, while the LVM of H-II at 3349.6 cm1 is polarized perpendicular to c.
Figure 3. Normalized LVMs intensities of the H-I (a) and H-II (b) defects measured at 10 K
as a function of the angle - between polarizer and the c axis of the sample.
The best fit to the experimental data shown in Fig. 3 by the solid lines is
achieved when the O-H bond of the H-I defect is parallel to the c axis with
the accuracy of 10o. The same result is obtained for the O-H bond
responsible for the 3312.2 cm1 line of H-II, whereas the O-H bond giving
rise to the 3349.6 cm1 line is nearly perpendicular to the c axis, i.e.:
- =100o.
3.2
Molecular models
3.2.1
H-I complex
The experimental observations indicate that this defect has one hydrogen
atom primarily bound to an oxygen atom with the O-H bond aligned close to
the c axis. H-I is observed in all our nominally undoped VP ZnO samples.
We propose that interstitial hydrogen at the BC|| site is the most likely
candidate for this center. Here, the symbol || stands for the configuration
139
with the O-H unit parallel to the c axis; the other bond directions are denoted
by the symbol A. Indeed, the first-principles calculations indicate that in
this configuration hydrogen is strongly bound to an oxygen atom, with the
neighboring Zn atom relaxing outward by almost 40 % of the bond length.4
The results of Ref. 4 showed that the BC||, BCA, AB|| and ABA configurations
all have very similar formation energies, although calculations with
increased convergence and additional relaxation indicate that BC|| may be
lower than the others by about 0.1 eV.11 The experimental results
conclusively indicate a configuration oriented parallel to c, but experiment
alone cannot distinguish between BC|| and AB||. Both the calculated
formation energies4 and the calculated vibrational frequencies11 strongly
favor BC||, indeed, the calculated LVM for BC|| is about 300 cm1 lower than
the LVM at BC||. All these findings point to interstitial hydrogen at BC||
(H+BC) as the most likely candidate for the H-I defect.
Nonetheless, we cannot exclude that this interstitial hydrogen might be
bound to an impurity. Indeed, the fact thatt the H-I signal is stable up to about
350 oC in the annealing experiments (see Fig. 2) may point to a
configuration that is more strongly bound than isolated interstitial hydrogen,
which has a diffusion barrier of about 0.9 eV.15 Interstitial hydrogen bound
to an isoelectronic or acceptor impurity could account for the H-I line,
provided the hydrogen would still be in the BC|| configuration.
It is known that the conductivity of ZnO crystals exposed to hydrogen
increases.16 With our samples we have found that the resistivity is reduced
from 50 to 5 :cm after hydrogen plasma treatment. This finding is definitely
consistent with hydrogen being introduced as an interstitial and acting as a
shallow donor. However, the resistivity data alone cannot help to distinguish
between hydrogen being present as an isolated donor (H+BC) or passivating
an acceptor impurity or native defect, since both scenarios increase the ntype conductivity of ZnO.
3.2.2
H-II complex
The defect has two non-equivalent hydrogen atoms primarily bound to

oxygen atoms. It is observed in all our nominally undoped ZnO samples.
One of the two O-H bonds comprising H-II is close to the c axis, whereas the
other one is nearly perpendicular to c. Two interstitial hydrogen atoms, both
acting as shallow donors, are unlikely to spontaneously come together,
pointing strongly towards involvement of another defect. We believe that
VZn is the most likely candidate. Indeed, the Zn-H bond is much less stable
as compared to the O-H bond, and the Zn-H stretch LVM frequencies should
be in the range 1500-1700 cm1, which is too far away from what we see in
our spectra.17 Thus, among the native defects only VZn, Oi or OZn decorated
140
E. V. Lavrov et al.
with two hydrogen atoms may explain our data. Interstitial oxygen and
oxygen antisites are ruled out by the weak coupling between the two LVMs
of H-II. Therefore, we propose that a zinc vacancy decorated with two
hydrogen atoms, VZnH2, is the most probable model for H-II.11
This assignment is also supported by the results of cathodoluminescence
studies of ZnO after hydrogen plasma treatment. Sekiguchi et al. have shown
that such a treatment made at 400 oC fully suppresses the, so-called, green
luminescence band centered around 2.2 eV.18 We have observed the same
behavior by means of photoluminescence for our ZnO samples. On the other
hand, based on first-principles pseudopotential calculations, Kohan et al.
suggested that VZn is involved in the green luminescence in ZnO.19 Thus,
passivation of VZn with two hydrogen atoms leads to a reduction of the
green luminescence, and appearance of the H-II LVMs.
As mentioned above, the resistivity of our ZnO samples is reduced from
50 to 5 : cm after hydrogen plasma treatment, which is also in favor of the
VZnH2 model. Indeed, passivation of VZn acting as an acceptor in ZnO should
increase the n-type conductivity of the samples.
4.
HYDROTHERMALLY GROWN ZnO
We found that hydrogen influences on the n-type conductivity of the VP

ZnO and suggested molecular structures responsible for this observation. It
follows from these findings that hydrogen does not seem to be the main
source for the n-type conductivity of ZnO because all hydrogen-related
centers found in VP ZnO appear only after intentional hydrogenation.
However, this situation changes in the case of hydrothermally grown
material.
Figure 4 shows typical IR absorption spectrum of an as-grown HT ZnO
sample.12 The lines at 4216.2, 4240.0, and 4246.6 cm1 were previously
identified as internal transitions of substitutional Ni2+ split by the local
crystalline field.20,21 The new lines at 3335.6, 3482.9, 3516.3, and 3577.3
cm1 are located in the region characteristic of O-H stretch modes.17
An additional deuterium plasma treatment at 350 oC results in one additional
IR absorption line at 2644.4 cm1. The frequency ratio of the 3577.3 cm1
line and the line at 2644.4 cm1 is 1.35, which is close to the value expected
for a harmonic oscillator consisting of a hydrogen atom bound to an oxygen
atom (1.37). Based on this, we identify the 3577.3 cm1 line as a LVM of an
O-H bond. No extra lines appear in the spectra after H+D plasma treatment.
This strongly indicates that the defect responsible for the 3577.3 cm1 line
includes a single hydrogen atom.
141
Thus, in contrast to ZnO grown from the vapor phase, as-grown

hydrothermal ZnO contains hydrogen. Note that we could not find any
deuterium replica of the 3335.6, 3482.9, and 3516.3 cm1 lines. Hence, the
nature of these remains unknown.
Figure 4. Absorption spectra measured at 10 K for an as-grown hydrothermal ZnO sample.
All the lines presented in Fig. 4 display distinct polarization properties,

which is demonstrated in Fig. 5. The figure shows the IR absorption spectra
measured for an as-grown HT ZnO sample with the polarization vector of
the absorbed light parallel (top spectrum) and perpendicular (bottom
spectrum) to the c axis of the crystal. The polarization properties of the Nirelated lines at 4216.2, 4240.0, and 4246.6 cm1 are identical to those
reported in Ref. 21. It follows also from the figure that the 3577.3 cm1 line
is predominantly polarized with the c axis of the crystal, which implies that
the O-H bond of the defect is aligned close to c. We label this defect as H-I*.
The H-I* defect is extremely stable: The 3577.3 cm1 line does not change its
intensity even after annealing at 1100 oC for 30 min. Only thermal treatment
at 1200 oC for several hours eliminates the signal from the spectra.
Interestingly, subsequent hydrogen plasma treatment at 350 oC does not
result in the formation of this defect. No changes in the Ni2+ signals occur
142
E. V. Lavrov et al.
after thermal and hydrogen plasma treatment. The last finding strongly
indicates that the H-I* defect does not include a substitutional Ni atom, as
tentatively suggested in Ref. 12.
We could not establish a correlation/anticorrelation between the 3577.3
cm1 line and any other spectroscopic signal of HT ZnO. We take this as an
indication that the center giving rise to this line is related to the presence of
native defects. The LVM frequency of the center points to VZn, Oi, or OZn as
possible candidates. Of these three, we rule out VZn because, as we have seen
Fig. 3
Figure 5. Polarized absorption spectra measured at 10 K for a virgin hydrothermally grown
ZnO sample: top E || c, bottom E A c.
in the previous section, zinc vacancies tend to bind two hydrogen atoms
resulting in VZnH2 defects, which are not stable at annealing temperatures
above 600 oC.
Indirect indication that Oi or OZn might be responsible for H-I* is the
annealing temperature of this center. It is known that ZnO crystals, when
heated in air at temperatures above 1000 oC become nonstoichiometric in
composition through formation of excess zinc atoms resulting from the more
rapid loss of oxygen atoms.22 It means that of all native defects Oi and OZn
143
should be destroyed first at elevated temperatures. This may explain why HT

ZnO annealed at 1200 oC for several hours and subsequently treated with a
hydrogen plasma does not reveal the 3577.3 cm1 line in the IR absorption
spectra.
Nonetheless, participation of some other impurity is not excluded.
Further theoretical studies in combination with other experimental
techniques such as EPR, electron-nuclear double resonance (ENDOR), etc.,
are needed to establish microscopic structure of the H-I* defect.
5.
CONCLUSIONS
Vibrational spectroscopy has been employed to study hydrogen-related

defects in ZnO samples grown by different techniques. Hydrogenation of
vapor phase grown ZnO results in three IR absorption lines at 3611.3,
3349.6, and 3312.2 cm1. The 3611.3 cm1 line corresponds to a defect
labeled H-I that contains a single hydrogen atom. We tentatively assign H-I
to an interstitial H at a BC|| site. The lines at 3349.6 and 3312.2 cm1 are
associated with a defect labeled H-II that contains two hydrogen atoms. The
H-II defect is identified as a hydrogenated Zn vacancy (VZnH2).
Absorption spectra of as-grown hydrothermal ZnO samples reveal a
number of lines located in the region characteristic of O-H stretch LVMs.
Based on the isotope substitution experiments as well as on the polarization
studies, it is shown that at least one of these lines, at 3577.3 cm1, originates
from a defect containing one O-H bond primarily aligned with the c axis of
the crystal and labeled as H-I*. The presence of a native defect is tentatively
suggested.
ACKNOWLEDGMENTS
The authors greatly acknowledge C. G. Van de Walle, and B. Bech
Nielsen for helpful discussions, R. Helbig for supplying the ZnO samples,
and the Russian foundation for Basic Research (Grant No. 02-02-16030).
REFERENCES
1. D.C. Look, D.C. Reynolds, J.R. Sizelove, R.L. Jones, C.W. Litton, G. Cantwell, and W.C.
Harsch, Solid State Comm., 105, 399 (1998).
2. M. Joseph, H. Tabata, and T. Kawai, Jpn. J. Appl. Phys., 38, L1205 (1999).
144
E. V. Lavrov et al.
3. D. C. Look, J. W. Hemsky, nad J. R. Sizelove, Phys. Rev. Lett., 82, 2552 (1999) and
references therein.
4. C. G. Van de Walle, Phys. Rev. Lett., 85, 1012 (2000).
5. S.F.J. Cox, E.A. Davis, S.P. Cottrell, P.J.C. King, J.S. Lord, J.M. Gil, H.V. Alberto, R.C.
Vilo, J. Piroto Duarte, N. Ayres de Campos, A. Weidinger, R.L. Lichti, and S.J.C. Irvine,
Phys. Rev. Lett., 86, 2601 (2001).
6. D.M. Hofmann, A. Hofstaetter, F. Leitner, H. Zhou, F. Henecker, B.K. Meyer, S.B.
Orlinskii, J. Schmidt, and P.G. Baranov, Phys. Rev. Lett., 88, 45504 (2002).
7. F. G. Grtner and E. Mollwo, phys. status solidi B, 89, 381 (1978).
8. F. G. Grtner and E. Mollwo, phys. status solidi B, 90, 33 (1978).
9. M. D. McCluskey, S. J. Jokela, K. K. Zhuravlev, P. J. Simpson, and K. G. Lynn, J. Appl.
Phys., 81, 3807 (2002).
10. N. H. Nickel and K. Fleischer, Phys. Rev. Lett., 90, 197402 (2003).
11. E. V. Lavrov, J. Weber, F. Brrnert, C. G. Van de Walle, and R. Helbig, Phys. Rev. B, 66,
165205 (2002).
12. E. V. Lavrov, Physica B, 340-342, 195 (2003).
13. G. Mller and R. Helbig, J. Phys. Chem. Solids, 32, 1971 (1971).
14. R. Helbig, J. Cryst. Growth, 15, 25 (1972).
15. D. G. Thomas and J. J. Lander, J. Chem. Phys., 25, 1136 (1956).
16. E. Mollwo, Z. Phys., 138, 478 (1954).
17.CRC Handbook of Chemistry and Physics, 82ndd edition, David R. Lide (ed.), CRC Press
LLC, 2001-2002, p. 9-77.
18. T. Sekiguchi, N. Ohashi, and Y. Terada, Jpn. J. Appl. Phys., 36, L289 (1997).
19. A. F. Kohan, G. Ceder, D. Morgan, and C. G. Van de Walle, Phys. Rev. B, 61, 15019
(2000).
20. U. Kaufmann, P. Koidl, and O. F. Schirmer, J. Phys. C, 6, 310 (1973).
21. H. J. Schulz and M. Thiede, Phys. Rev. B, 35, 18 (1987).
22. H. E. Brown, Zinc Oxide: Properties and Applications, New York, Int. Lead Zinc
Research Organization, 1976.
Chapter 13
INFLUENCE OF THE HYDROGEN
CONCENTRATION ON H BONDING IN ZINC
OXIDE
N. H. Nickel
Hahn-Meitner-Institut Berlin, Kekulstr. 5, 12489 Berlin, Germany
Abstract:
Despite of the recent developments the properties of hydrogen in ZnO are not
fully understood. While in some ZnO single crystals the amount of H detected
by gas effusion experiments is similar in concentration to the number of free
electrons, a large number of samples exhibits much higher H concentrations.
Raman backscattering experiments reveal that a significant amount of H is
accommodated at sites that do not give rise to enhanced conductivity; a large
amount of H is bound to carbon and nitrogen impurities forming CHX and
NH complexes. Depending on the method of the sample preparation H
concentrations as high as 3.0u1021 cm-3 have been observed. To gain further
insight into the properties of H in ZnO data obtained from H effusion
measurements of single crystal and sputter deposited ZnO have been analyzed
to deduce the hydrogen binding energy as a function of the H chemical
potential. In samples with a low H content six peaks were observed in the H
density-of-states distribution. With increasing H concentration the peaks
broaden. In addition, an increase of the H concentration results in a
pronounced increase of the average H binding energy. This observation
indicates that the properties of H expand well beyond the formation of donors
and the neutralization of impurities and deep defects.
Key words:
Hydrogen binding energy, hydrogen local vibrational modes, hydrogen

concentration, hydrogen density-of-states distribution
1.
INTRODUCTION
Recently, there has been a significant increase of research in the field of

zinc oxide. This is mainly driven by its extraordinary electrical and optical
properties that enable a variety of applications ranging from piezoelectrical
145
146
N. H. Nickel
devices, transparent conducting layers for solar cells, transparent thin-film

transistors to ultra violet light emitting diodes and lasers. Almost all of these
applications require the control of the conductivity from n-type to p-type.
While n-type doping can be controlled easily the reproducible fabrication of
p-type ZnO is a major obstacle that has to be overcome to allow for the
successful development of LEDs and UV lasers. The major drawback is the
fact that zinc oxide generally exhibits n-type conductivity that has been
attributed to the presence of native defects, traditionally. However, recent
theoretical and experimental investigations have identified a new source that
can easily account for the observed n-type conductivity. Van de Walle1
suggested that hydrogen atoms can act as shallow donors. This prediction has
been confirmed experimentally by muon spin rotation2 and electron nuclear
double resonance measurements.3 In addition, Hall-effect measurements
revealed the presence of a second shallow donor.3
Thus, to achieve p-type doping of ZnO a number of aspects have to be
considered. First of all the formation of hydrogen shallow donors has to be
prevented or their concentration has to be minimized. Recently it has been
shown that a simple post anneal of state-of-the-art ZnO at 850 C causes the
dissociation of hydrogen shallow donors which results in a decrease of the
free carrier concentration from 1u1017 cm-3 to 6u1016 cm-3.4
A suitable acceptor for ZnO is nitrogen. In fact, substitutional nitrogen
incorporated at an oxygen site has been observed using electronparamagnetic-resonance (EPR) measurements.5 Although nitrogen can be
incorporated as an acceptor the presence of hydrogen could give rise to the
formation of N-H complexes that are electrically neutral. Hence it is of great
importance to investigate the properties of hydrogen in ZnO.
In this paper, we present information on hydrogen bonding and hydrogen
concentration in state-of-the-art single crystal and sputter deposited ZnO thin
films. Raman backscattering measurements reveal 6 H-related local
vibrational modes in the wave number range between 2800 and 3150 cm-1.
Hydrogen effusion measurements clearly show that the observed local
vibrational modes are due to the presence of H in ZnO. From the effusion
data the H density-of-states (DOS) distribution is derived, which provides
detailed insight into H binding energies. In single crystal ZnO reveals six
peaks in the H DOS. On the other hand, in sputtered ZnO thin films the total
H concentration increases by up to 5 orders of magnitude. This is
accompanied by a pronounced increase of the average H binding energy.
Influence of the H concentration on H bonding in ZnO
2.
147
EXPERIMENTAL
Two sets of samples were used in this investigation. The first set consists
of commercially available ZnO single crystals with a (0001) orientation and
an O face. These crystals were fabricated by Eagle-Picher. The second set
of samples were Al-doped ZnO thin-films. These specimens were grown on
quartz substrates by dc magnetron sputtering at a power density of 22.6
W/cm2 and a substrate temperature of 150 C. The ZnO films had a thickness
ranging from 0.11 m to 3.83 m. From a structural point of view the
sputtered ZnO samples were composed of columnar grains with an average
grain diameter of about 0.16 m. The length of the grains is only limited by
the thickness of the film. The samples were characterized with Raman
backscattering experiments in a conventional macro-Raman setup using the
488 nm laser line of an Ar+ ion laser. Information on hydrogen content and H
bonding was obtained from H effusion measurements. For this purpose the
samples were placed in an ultrahigh vacuum system and heated to a
temperature of up to 950 C with a constant heating rate of 20 K/min. During
the temperature ramp the flux of molecular hydrogen was measured with a
quadrupole mass spectrometer. Prior to each measurement background
spectra were taken and subsequently subtracted from the sample data. To
obtain absolute values for the H2 flux the effusion spectra were calibrated by
measuring the know neon flux through a capillary.
3.
HYDROGEN LOCAL VIBRATIONAL MODES
Raman backscattering measurements were performed in the wave number

range between 200 and 4300 cm-1. Besides the well known phonon modes
that are located between 200 and 800 cm-1 additional lines were observed at
wave numbers ranging from 2800 to 3150 cm-1 (see spectrum (a) in Fig. 1).
At higher wave numbers Raman lines were not observed.
The local vibrational modes (LVMs) shown in Fig. 1 are due to the
presence of hydrogen. This is supported by Raman backscattering
experiments performed after hydrogen effusion experiment. Spectrum (b) in
Fig. 1 was obtained after the sample was heated up to 950 C in an effusion
experiment. The measured molecular hydrogen flux as a function of
temperature is plotted in Fig. 2. The background signal taken in a separate
measurement was subsequently subtracted from the data. The spectrum
shows a maximum H2 flux at a temperature of 473 C. From the H2 flux the
total hydrogen concentration can be determined by integration and taking
into account the heating rate and sample thickness. The H concentration
148
N. H. Nickel
Raman intensity (arb. units)
amounts to 5.21016 cm-3. This value is remarkably close to the

concentration of shallow donors that was attributed to the presence of
hydrogen donor states in state-of-the-arte undoped ZnO 6. After the diffusion
experiment the sample is depleted of hydrogen and a second scan merely
yields the background data taken without a sample. Raman backscattering
measurements performed on H depleted specimens did not show any local
vibration modes. This clearly indicates that the LVMs observed between
2800 and 3150 cm-1 are due to the presence of H atoms. However, when H
acts as a donor in ZnO it has to be accommodated in a bond center site or
anti bonding site forming an O-H bond and the local vibration modes should
occur at 3680 cm-1 and 3550 cm-1, respectively.1 These values for the LVMs
are close to those of free O-H molecules and were recently observed in single
crystal ZnO.7,8 This raises the question regarding the origin of the observed
hydrogen LVMs.
Local vibrational modes between 2800 and 3150 cm-1 (see Fig. 1) have
been observed in a number of semiconductors such as amorphous silicon
carbide,9 GaAs,10 and GaN.11 In these materials the vibrational modes are
assigned to symmetric and antisymmetric stretching modes of CHX
complexes with x = 1, 2, 3. Therefore, the vibrational modes located at 2854
cm-1, 2890 cm-1, 2918 cm-1, 2948 cm-1, and 2988 cm-1 should be due to the
488 nm, 190 mW
(a)
(b)
after H effusion
2800
2900
3000
3100
3200
-1
Z (cm )
Figure 1. Raman backscattering spectra of single crystal ZnO before (a) and after hydrogen
effusion (b). The measurements were performed using the 488 nm line of an Ar+ laser and a
laser power of 190 mW. Background data were subtracted.
149
13
10
-2 -1
H2 flux (cm s )
12
10
11
10
crystalline ZnO
10
10
10
200
400
600
800
1000
T (C)
Figure 2. Molecular hydrogen flux of single crystal ZnO as a function of temperature. The
sample was heated with a heating rate of 20 K/min.
stretching modes of C-H, symmetric stretching modes of C-H3, symmetric

stretching modes of C-H2, antisymmetric stretching modes of C-H3, and
antisymmetric stretching modes of C-H2, respectively. The measured LVMs
(see Fig. 1) are in excellent agreement with CHX stretching modes except for
the local vibrational mode located at 2988 cm-1 that is shifted to higher wave
numbers by about 65 cm-1.
The H local vibrational mode at 3096 cm-1 has been assigned to N-H
centers. This assignment is consistent with a recent theoretical investigation
by Van de Walle12 who calculated a vibrational frequency of approximately
3100 cm-1 for the N-H center. It is interesting to note that this frequency
deviates by about 10 % from the value observed for the N-H vibration in
ammonia molecules.
The results of this section clearly demonstrate that hydrogen does not
only form shallow donors in ZnO. The properties of H and the underlying
chemistry are manifold. Apparently a high concentration of H atoms can be
accommodated in single crystal ZnO without the need to form a significant
number of shallow donors. The presence of impurity-hydrogen complexes is
one important aspect that has to be investigated further.
150
4.
N. H. Nickel
HYDROGEN DENSITY OF STATES

DISTRIBUTION
From hydrogen effusion spectra as shown in Fig. 2 a variety of

information such as the H concentration, diffusion coefficient, and hydrogen
binding energies can be obtained. This section will focus on the H binding
energies. However, to avoid errors in the analysis it is important to know if H
transport from the bulk to the surface is governed by molecular or atomic
hydrogen diffusion. This question can be solved either by changing the
heating rate of the effusion experiment or by measuring samples of various
thicknesses. In both cases peaks observed in the effusion spectra should shift
to different temperatures; with decreasing heating rate and increasing sample
thickness the maxima of the H2 flux should move to higher temperatures.
Fig. 3 shows H effusion spectra obtained on sputter deposited ZnO thin films
with thicknesses varying from 0.11 to 3.83 m.
The effusion spectra of the sputter deposited ZnO films reveal two
pronounced maxima around 420 C and 600 C. A detailed analysis of the
spectra yielded the thickness dependence of the maxima (see Fig. 4). Both
14
10
d (m)
0.11
13
2.06
-2 -1
H2 flux (cm s )
10
3.83
12
10
11
10
sputtered ZnO:Al
10
10
200
400
600
800
T (C)
Figure 3. Molecular hydrogen flux vs temperature for sputter deposited ZnO thin films. The
experiments were performed with a heating rate of 20 K/min. For clarity only three effusion
spectra are plotted.
151
Table 1. Hydrogen concentration in single crystal and sputter deposited ZnO. The asterisk
marks single crystal ZnO.
H concentration (cm-3)
3.31021
7.91020
1.01020
1.01020
5.21016
Thickness (m)
0.11
0.5
2.06
3.83
500 (*)
effusion peaks shift to higher temperatures as the sample thickness increases.

This behavior indicates that the effusion process is governed by atomic
hydrogen diffusion from the bulk to the surface.15
When comparing the data of Fig. 3 with the H2 flux of the single crystal
ZnO (Fig. 2) it is obvious that the total H concentration in sputtered ZnO is
significantly enhanced. A summary of the H concentration of the
investigated specimens is given in Tab. 1. The highest H content is observed
in the thinnest ZnO film. A more detailed investigation revealed that the H
concentration does not vary with thickness. This effect is observed only
because the ZnO thin-films were deposited in a single run without breaking
the vacuum and starting with the thinnest sample. The large amount of H
incorporated in the thin films originates from the residual water vapor in the
vacuum system that dissociates upon electron impact during sputtering.13, 14
Since the effusion experiment is governed by the diffusion of monatomic
H, the hydrogen binding energy can be deduced from effusion spectra
applying the following formalism. The total H content, CH, is related to the
hydrogen density-of-states distribution, NH(E
( ) by
f
CH
N
H
dE
(1)
f
where E is the H binding energy, T the temperature, f is the occupation

function, and H is the hydrogen chemical potential. Since wf
wf/wH has its
maximum value at E | H the H DOS distribution can be estimated by16
wCH
.
wP H
(2)
152
N. H. Nickel
700
high T peak
T (C)
600
500
400
low T peak
sample thickness (m)

Figure 4. Temperature of the effuson peaks as a function of the sample thickness of sputtered
ZnO thin films.
The hydrogen chemical potential can be estimated using the following

equation
E PH
F
kT ln
F0
(3)
where k is Boltzmanns constant, T the temperature, F the molecular

hydrogen flux, and F0 a prefactor. The prefactor can be estimated using the
equation dF
F0 = 2QaN
Nsurface, where a is the mean free path, Q is an attempt
frequency, and Nsurface is the concentration of H surface states. With
reasonable values of a | 3u10-8 cm, Q |1013 Hz, and Nsurface | 1015 cm-3, a
value of dF
F0 | 6u1020 s-1cm-1 is obtained. This prefactor represents an
average value. However, since F0 is in the argument of logarithm when
computing the H chemical potential even an order of magnitude error results
in an error of only 0.1 eV in the absolute energy scale.
Fig. 5 shows the H density-of-states distribution derived from effusion
spectra. The lowest H DOS is obtained for single crystal ZnO (Fig. 5a).
Deconvolution of the NH reveals that the distribution is composed of six
peaks; the peak positions are marked in the figure by arrows
-3
0.59
1.35
(b)
0.92
1
0
0.0
-1
-3
single crystal
y
ZnO
1.55
0.5
1.0
E H (eV)
10
1.5
2.0
(c)
153
0.9
1.1
1
1.3
5
0
4
-1
0
2
1.03
1.18
1.4
20
(a)
21
20
NH (10 eV cm ) NH (10 eV cm )
-1
-3
16
-1
-3
NH (10 eV cm ) NH (10 eV cm )
0
0.96
(d)
1.44
2
0
0.0
0.5
1.0
1.5
2.0
E H (eV)
Figure 5. Hydrogen density-of-states distribution in single crystal ZnO (a) and Al doped ZnO
thin films (b) (d). Zero energy corresponds to the H transport site.
For sputter deposited ZnO the H content and thus NH increases up to 81021
cm-3eV-1 (see Fig. 5d). The data plotted in Fig. 5(b) show an important
result. The H effusion spectra were obtained on 2 different samples with
different thickness and the analysis yielded identical density-of-states
distributions indicating that the method to derive the H DOS is reproducible
and reliable.
Further analyzing the data in Fig. 5 reveals a striking observation. While
in single crystal ZnO about 45 % of all H atoms are accommodated with
binding energies of less than 1.0 eV with respect to the H transport site an
increase of the H concentration is accompanied with an increase of NH in the
energy range 1.0 eV < E H < 1.8 eV. With increasing H content the
amount of H atoms accommodated with binding energies of less than 1.0 eV
decreases to about 25 %. This suggests that H creates larger clusters
presumable in a similar way as it has been observed in other
semiconductors.17
5.
SUMMARY
In summary, Raman backscattering measurements showed the presence

of C-HX and N-H local vibrational modes in single crystal ZnO. Heating the
specimens to temperatures of up to 950 C caused hydrogen out diffusion.
After dehydrogenation the local vibrational modes disappeared indicating
that they are related to the presence of H. From H effusion measurements the
154
N. H. Nickel
total H concentration of the samples was determined. Single crystal ZnO

exhibits the lowest H concentration while sputter deposited ZnO thin-films
contain H concentrations of up to 31021cm-3. The H content depends on the
precondition of the deposition chamber; e.g.: the amount of residual water
vapor. From the effusion spectra the H density-of-states distribution was
derived. Single crystal ZnO reveals a H DOS that deconvolutes into six
peaks. With increasing H concentration an pronounced increase of the H
binding energies is observed. In sputtered ZnO about 75 % of the H atoms
are incorporated with binding energies larger than 1.0 eV while in single
crystal ZnO only about 55 % of the accommodated H atoms reveal binding
energies larger than 1.0 eV.
ACKNOWLEDGEMENT
The author is grateful to C. Klimm for the preparation of the ZnO thin
films and to K. Brendel for stimulating discussions and his constant interest
in the topic.
REFERENCES:
1. C. G. Van de Walle, Phys. Rev. Lett. 85, 1012 (2000).
2. S. F. Cox, E. A. Davis, S. P. Cottrell, P. J. C. King, J. S. Lord, J. M. Gil, H. V. Alberto, R.
C. Vilo, J. Pironto Duarte, N. Ares de Campos, A. Weidinger, R. L. Lichti, and S. J. C.
Irving, Phys. Rev. Lett. 86, 2601 (2001).
3. D. M. Hofmann, A. Hofstaetter, F. Leiter, H. Zhou, F. Henecker, B. K. Meyer, S. B.
Orlinskii, J. Schmidt, and P. G. Baranov, Phys. Rev. Lett. 88, 045504 (2002).
4. B. K. Meyer, H. Alves. D. M. Hofmann, W. Kriegseis, D. Forster, F. Bertram, J. Christen,
A. Hoffmann, M. Straburg, M. Dworzak, U. Haboeck, and A. V. Rodina, phys. stat. sol
(b) 241, 231 (2004), and references therein.
5. N. Y. Garces, N. C. Giles, L. E. Halliburton, G. Cantwell, D. B. Eason, D. C. Reynolds,
and D. C. Look, Appl. Phys. Lett. 80, 1334 (2002).
6. D. C. Look, J. W. Hemsky, and J. R. Sizelove, Phys. Rev. Lett. 82, 2552 (1999).
7. M. D. McCluskey, S. J. Jokela, K. K. Zhuravlev, P. J. Simpson, and K. G. Lynn, Appl.
Phys. Lett. 81, 3807 (2002).
8. E. V. Lavrov, J. Weber, F. Brrnert, C. G. Van de Walle, R. Helbig, Phys. Rev. B 66,
165205 (2002).
9. S.-Y. Lin, J. Appl. Phys. 80, 1399 (1996), and references therein.
10. D. M. Joseph, R. Balagopal, R. F. Hicks, L. P. Sadwick, and K. L. Wang, Appl. Phys. Lett.
53, 2203 (1988).
11. M. O. Manasreh, J. M. Baranowski, K. Pakula, H. X. Jiang, and J. Lin, Appl. Phys. Lett.
75, 659 (1999).
12. C. G. Van de Walle, presentation at the MRS Spring Meeting in San Francisco 2004.
13. J. W. Bozzelli and R. B. Barat, Plasma Chem. Plasma Proc. 8, 293 (1983).
155
14. S. Roychowdhury, U. K. Roychowdhury, and M.Venugopalan, Plasma Chem. Plasma

Proc. 2, 157 (1982).
15. W. Beyer, in Hydrogen in Semiconductors II;
I Vol. 61, edited by N. H. Nickel (Academic
Press, San Diego, 1999), p. 165.
16. W. B. Jackson, A. J. Franz, H.-C. Jin, J. R. Abelson, and J. L. Gland, J. Non-Cryst. Sol.
227-230, 143 (1998).
17.Hydrogen in Semiconductors II,
I edited by N. H. Nickel (Academic Press, San Diego,
1999), Vol. 61. p. 165.
PART IV: FUNDAMENTAL PROPERTIES
Chapter 14
VALENCE BAND ORDERING AND MAGNETOOPTICAL PROPERTIES OF FREE AND BOUND
EXCITONS IN ZnO
A.V. Rodina1, M. Strassburg2, M. Dworzak2, U. Haboeck2, A. Hoffmann2,
H. R. Alves3, A. Zeuner3, D. M. Hofmann3, and B. K. Meyer3
1
A.F. Ioffe Physico-Technical Institute, St.-Petersburg, Russia; 2Institute of Solid State

Physics, Technical University of Berlin, Berlin, Germany; 3I. Physics Institute, Justus-LiebigUniversity Giessen, Giessen, Germany
Abstract:
The Zeeman splitting of free and bound excitons is analysed theoretically for
both * and *symmetries of the upper valence band. The Zeeman splitting of
the ** holes in magnetic field parallel to the hexagonal c axes of the
crystal is found to be smaller (larger) than the Zeeman splitting of the
electrons. The magnetic-field dependence of the bound exciton optical
transition energies measured in the Faraday geometry with right and left
circular polarized light can be well described with * as well as with *holes.
However, the fitting of the experimental dependence of the Zeeman splitting
on the angle between magnetic field and c axis proves the * symmetry of the
holes involved in all low temperature transitions and thus of the upper valence
band in bulk ZnO. The determined parameters are in excellent agreement with
known experimental data on the fine structure of the free exciton ground state
in ZnO. The hole g-factors determined for excitons bound to an ionized and
neutral impurity centers are close to each other and to the g-factor calculated
for the *7 hole in the free exciton ground state. This points to a donor character
of the neutral impurity centers. This conclusion is confirmed by temperature
dependent magneto-transmission and magneto-luminescence measurements.
Key words:
Photoluminescence, magneto-luminescence, Zeeman effect, g-factor, excitons,

valence band symmetry
1.
INTRODUCTION
Bulk ZnO is a direct band gap wurtzite semiconductor with the valence
band maximum split by the crystal field and spin-orbit interaction. The
159
160
A. V. Rodina et al.
symmetry of the upper valence subband (A-subband) has been the subject of
controversy (9 or 7 character) for more than 40 years.1-15 Based on the
polarization properties of the free exciton transitions, the inverted ordering of
the valence subbands as shown in Figure 1(a) was first suggested by
Thomas1 and supported later in many articles.3-7,11-18,21 However, different
argumentation led some researches2,8-10,19,20 to the conclusion that the
ordering of valence subbands in ZnO is the same as in other wurtzite
semiconductors like GaN, CdSe, CdS (see Figure 1(b)).
(a)
(b)
E
*7
*7
Egap=3.4376 eV
T=4.2 K
*7
A
A
*9
AB = 4.9 meV
B
*7
43 7 meV
V
EBC = 43.7
B
*7
C
Figure 1. Energy band structure and the symmetry of the free exciton ground state in wurtzitetype semiconductors for (a) ZnO and (b) GaN, CdSe, CdS.
Magneto-optical studies are a powerful tool to distinguish the symmetry

of the free and bound exciton states.5-8,14-21 Unfortunately, most of the
magneto-optical investigations in ZnO were incomplete and allowed both
interpretations concerning the symmetry of the holes involved. For example,
recent magneto-optical studies8 of the free A exciton fine structure were
interpreted with the assumption of the 9 symmetry for the valence band
maximum, while the more elaborated theoretical analysis11 enabled the
explanation of the same date applying the 7 symmetry for the A valence
band. To resolve this controversy, detailed theoretical and experimental
studies of the magneto-optical properties of bound excitons in bulk ZnO
were undertaken.14 In this paper we show that the detailed comparison
between the theoretical predictions and the experimentally observed Zeeman
behavior of free excitons and bound exciton complexes in bulk ZnO
confirms the *7 symmetry of the upper valence subband.
Valence band ordering and magneto-optical properties
2.
161
THEORY
The inverted ordering of the valence subbands (Fig. 1(a)) can be

understood in terms of an effective negative spin-orbit splitting. The firstprinciples band structure calculations of Ref. 11 confirmed the inverted
ordering in bulk ZnO. It was found that the negative spin-orbit coupling
arises from the contribution of the lower lying Zn 3d bands. The optimal
value for the Zn 3d bands position was derived which gives good agreement
with the basic valence band splittings.11 A self consistent set of the effective
mass parameters obtained from these band structure calculations was used in
the analysis of the exciton binding energies and effective g values.11
In Figure 2 we sketch all possible transitions from the free exciton and
from the exciton bound to the ionized impurity center corresponding to the
calculated values11 of the hole effective g-factors. We consider the holes
from the *7 (case (a)) and *9 (case (b)) valence bands. We take in account
only the short-range exchange interaction in Fig. 2 and do not consider the
long range exchange interaction or polariton effects. The zero field exchange
splitting between the *1 and *2 exciton states is neglected in Fig. 2 (a) as it is
known to be small in ZnO.11 The zero field spin-exchange splitting of the
*1/*2 and *5 exciton states (case (a)) or *6 and *5 exciton states (case (b)) is
about 0.95 meV23 and governs the Zeeman behavior in the magnetic field
perpendicular to the c axis. In the magnetic field parallel to the c axis the
linear Zeeman splitting of the *5 exciton state is expected. The *6 state and
the *1/*2 states mixed by the magnetic field also show the linear splitting.
The splitting is given by
P B Bg
B exc
P B B(
//
h
),
(1)
where PB is the (positive) value of the Bohr magneton, ge=1.95 is the

electron (isotropic) effective g-factor and the calculated values of the hole
effective g factor are gh/// (7 )= - 1.40.15 for the 7 hole in the exciton
ground state and gh//(
( 9 ) = 3.00.15 for the 9 hole. The resulting values of
the exciton effective g-factors gexc for all states are shown in the Figure 2.
For the transitions allowed with E A c the gexc values are positive in both (a)
and (b) cases. Therefore in the B // c // k configuration the upper-energy state
is expected to be active for the right-circular polarized light V+. In the k A c
configuration, the transitions with E // c are allowed in the case (a); in the
case (b) they are forbidden but might be observed due to high-order
perturbations. The calculated gexc values for the 7 hole (case (a)) in B // c
describe well the experimental data reported in Ref. 8 for the free exciton
ground state in both E A c and E // c polarization. We discuss the
162
A. V. Rodina et al.
(a) excitons with *7 holes: *7 x *7 =*
* 5+ * 1+ * 2
B
B = 0
B ||c
*5
geexc | gh/// ge || 3.35
* 1, * 2
geexc ge gh/// | 0.55
E||c
V V-
E||c
*1
(b) excitons with *9 holes: *7 x *7 =*

* 5+ * 6
B
B = 0
B ||c
*5
geexc | gh/// ge || 4.95
*6
geexc gh/// ge |1.05
E||c
V V-
E||c
*1
Figure 2. Level schemes and selection rules for the ionized bound exciton transitions.
experimentally Zeeman splitting of the ionized donor bound exciton

complex in the next section.
In Figure 3 we sketch all possible transitions for excitons bound to neutral
impurities considering the holes from the *7 (cases (a), (c)) and *9
(cases (b), (d)) valence bands. In the case of a donor bound exciton complex
(Fig. 3 (a) and (b)), the spins of two electrons are anti-parallel and the
Zeeman splitting of the excited (
) state is determined by the
anisotropic hole effective g factor ((ggh), while the splitting of the ground D0
state is given by ge. In the case of an acceptor bound exciton complex (Fig. 3
(c) and (d)), two holes have the same symmetry and anti-parallel spins.
Hence, the splitting of the excited ((A0;X)
X state is determined by ge, while the
ground A0 state splits according to gh. For the holes in the exciton states we
use in Fig. 3 (a) and (b) the g values calculated in Ref. 11. For the holes in
the acceptor ground states we use in Fig. 3 (c) and (d) the values calculated
using the theory developed in Ref. 24 with effective mass Luttinger
parameters from Ref. 11.
Our calculations show that the Zeeman splitting of the * * holes in
B /c determined by |ggh//| is smaller (larger) than the Zeeman splitting of the
B//
electrons determined by ge. For the arbitrary angle 4 between the c axis and
the direction of the magnetic field the absolute value of the hole anisotropic
g factor is given by22

gh(
//
h
|2 c o s 2
163
(2)
|2 s in 2 4
Therefore, for the * holes the Zeeman splitting in any magnetic field
remains always smaller than the Zeeman splitting of the electrons. In
contrary, for the * holes some critical angle 4crr should exist so that
gh(4crr)=ge, and gh(4)>ge for 0q d 4 < 4cr and
d gh(4 )< ge for 90q t 4 >4cr.
Examination of the selection rules for optical transitions in the B // c // k
configuration shows that the upper-energy transition is expected to be active
for the right-circular polarized light V+ for all cases presented in Fig. 3 cases.
(a) (D
D0,X((*7))
D0
g |0
A
h
| g h|||| | 1.4

(c) (A
A0,X)
, )
h
B
+
+1/2
B ||c
-1/2
1/22
A0(*9)
+1/22
g e | 1.95
| gh|||| | 0.75
ghA |0
+1/22
d)) ((A
A0,,X)
X))
((d)
-1/2
1/2
-3/2
3/22
V-
ge | 1.95
-1/2
1/22
g e | 1.95
| gh|||| | 3.0
V
+1/22
A0((*
*7)
D0 +3/22
ge | 1.95
g hA
+1/2
V
e
-1/2
1/22
D0,X((*9))
(b) (D
-1/2
1/22
+3/22
ghA 0
-1/2
+1/2
||
h
| g || 3.3
B ||c
-3/2
3/22
Figure 3. Level schemes and selection rules for neutral bound exciton transitions.
3.
EXSPERIMENTAL SETUP AND SAMPLE

CHARACTERIZATION
We used nominally undoped (as grown from Eagle-Picher, see for details
Ref. 15) and nitrogen doped bulk ZnO crystals. A nitrogen doped ZnO:N
sample was prepared from the as grown one by ion implantation (2 MeV,1013
cm2) followed by a thermal annealing process at 900 C for 15 min (see also
Ref. 15).
164
A. V. Rodina et al.
The magneto-PL measurements were performed at liquid helium

temperature in a split-coil magnetocryostat allowing the variation of both
temperature (2300 K) and magnetic field (05 T). For some measurements,
a second split-coil magnetocryostat was applied providing magnetic fields up
to 15 T. Photoluminescence was excited by the 325 nm line of a HeCd laser.
A 450 W XBO lamp was used as excitation source for the transmission
investigations. The spectral resolution of the detection system was better than
0.15 meV for the 5 T cryostat and >0.3 meV for the 15 T cryostat.
The magneto-PL and magneto-PT measurements were performed in
Faraday configuration (magnetic field B parallel to the c axis of the crystal
and parallel to the k vector of the detected light) and in Voigt configuration
( A k). Additionally, the angles between B and c axes were varied from 0
(B
to 90q at fixed magnetic field of 5 T. The circular polarization of the light
(V and V ) in the Faraday configuration was analyzed using l/4 plate and
linear polarizer. Additionally, PL spectra with B A c, k A c and E // c were
recorded for the ZnO:N. Temperature-dependent measurements reveals the
thermalization behavior of the Zeeman split components in the emission and
transmission spectra.
The zero-field PL spectra of the as grown (nominally undoped) ZnO and
nitrogen doped ZnO:N are shown in Figs. 4(a) and 4(b), respectively. Only
the lines that will be discussed in this paper in details are marked (for other
lines see Ref. 14,15). The strongest line in the as grown ZnO (see Fig. 4 (a))
is the I4 (368.59 nm or 3.3628 eV). We attribute the I4 transition as an
exciton bound to a neutral shallow H donor that can be easily removed by an
annealing process.15,25 Indeed, the I4 PL line is absentt in ion implanted ZnO:N
(see Fig. 4(b)). This is because of the post annealing processes at 900 C
included in the doping procedure. The PL spectrum of ZnO:N is dominated
by two lines: a new line I2 (368.07 nm or 3.3676 eV) and the I9 transition.
The latter considerably gains in intensity compared to the undoped ZnO. The
wavelengths of the lines are given here in air with an accuracy of 0.02 nm
and have been transformed into the respective wavelength values in vacuum
before calculating the transition energies.
4.
ZEEMAN BEHAVIOR OF THE BOUND EXCITON

COMPLEXES: EXPERIMENTAL DATA AND
DISCUSSION
Figure 5 presents the magneto-PL spectra in the energy region of the I2

transition for the ZnO:N for Faraday configuration (a), Voigt configuration
(c) and for the different angles between the magnetic field direction and the

I9
I4
PL Intensity (log. arb. units)
(b) ZnO:N
I2
k//c
(a) ZnO
as grown
3,356
3,360
165
k//c
3,364
,
Energy (eV)
3,368
Figure 4. Photoluminescence spectra of undoped as grown ZnO (a) and nitrogen doped
ZnO:N (b) at 4.2 K.
c axis at B = 5 T (b). The most important feature of the spectra in Figures 2

(b) and (c) is the appereance of a new transition line at an energy about 1
meV below I2. This I3 transition (368 18 nm or 3 3666 eV in zero field as
determined by the extrapolating procedure) is forbidden in B // k // c,
E A c geometry but has been also observed in parallel magnetic field with
the light polarized along the c axis ((E
E // c, k A c configuration). Such
behavior leads to the conclusion that the I2 and I3 transitions originate from
excitons bound to an ionized impurity. The stability of such complex in
semiconductors strongly depends on the ratio of electron and hole effective
masses.26 In the case of ZnO, the single band isotropic mass approximation
predicts the existence of the excitons bound to ionized donors.23 Next, the
presence of acceptors in the ZnO:N may induce the ionization of the shallow
donors26 as in the case of Li and Na doped ZnO.19 Therefore, we attribute the
I2 and I3 lines in the doped ZnO:N to transitions of excitons bound to
respective ionized donors (( D +, X ((7)) complex). The value of the zero-field
spin-exchange splitting is 0.98 meV, in a good agreement with Refs. 19,23.
166
A. V. Rodina et al.
(a) B ||| c
V-
PL Intensity
y [norm.
[
u.]
u]
(c) B A c
( B=5T
(b)
I3/I2
90
5T
80
4T
60
3T
40
2T
4T
3T
2T
20
0T
I2
3,3675
3,3680
Energy [eV]
0T
0
3,366
3,367
3,368
Energy [eV]
3,366
3,367
3,368
Energy [eV]
Figure 5. Photoluminescence spectra of ZnO:N in the energy region of the I2/I3 lines at 4:2 K
for different magnetic fields in Faraday configuration (case (a): B // k // c, E A c) and Voigt
configuration (case (c): B A c, k // c, E A c), and for different angles between the c axis and
the direction of the magnetic field B at 5 T (case (b)).
The magnetic field and angular dependencies of the emission lines I2 and
I3 allow us to determine the symmetry and the g values of the valence band
hole involved in the corresponding bound excitons.14,15 The magnetic field B
/// c linearly splits the I2 transition (allowed in E A c configuration) as well as
the I3 transition (allowed in E // c configuration). The upper-energy
component of the I2 line is active for the right-circular polarized light (see
dashed line in Fig. 5 (a)) as predicted in Fig. 2. The linear splitting of the I2
line in B // c can be described with the positive exciton effective g- factor
gexc = 0.71 that is close to the calculated values in Fig.2 (a) and (b). However,
the experimental value gexc = 2.66 determined for the I3 line is in better
agreement with the expected value for the * holes (case (a) in Fig. 2) and
differs significantly from the respective value in case (b). Furthermore, from
the fitting of the angular dependence of the transition energies14,15 we
obtained gh// = -1.24 in the excellent agreement with the calculated value for
the * hole in the exciton ground state. This demonstrates that the hole in the
lowest exciton state and thus the upper A valence subband in ZnO also have
the * symmetry.
The Zeeman behavior of the emission lines from I4 to I9 is different from
that of the I2/II3 lines. Their linear splitting in B A c indicates transitions
originating from excitons bound to neutral impurities. We discuss the
emission lines I9 and I4 in detail. These lines representt the dominant

(a) B || c
(c) B A c
(b) B = 5 T
V+
PL Intensity
y [norm.
[
u.]
V-
167
5T
90
5T
80
4T
60
3T
4T
3T
40
2T
2T
20
0T
I9
3,3565
0T
3,3570
I9
3,3565
Energy [eV]
3,3570
Energy [eV]
3,3565
3,3570
Energy [eV]
Figure 6. Photoluminescence spectra of ZnO:N in the energy region of the I9 line at 4:2 K.
Cases (a), (b) and (c) are the same as in Fig. 5.
recombination in ZnO:N and as grown ZnO bulk crystals, respectively, and

are spectrally resolved from the other observed emission lines. Figure 6
presents the magneto-PL spectra in the energy region of the I9 transition for
the ZnO:N, the Zeeman behavior of this line and the I4 in the undoped ZnO
is very similar. The transitions allowed for E A c split with magnetic fields
parallel and perpendicular to the c axis. The upper Zeeman split energy
component in B // c // k configuration is active in V+ polarization (see Fig. 6
(a)) as predicted for all cases in Fig. 3 (a-d). Four transition components can
be distinguished for an arbitrary angle 4 between the direction of the
magnetic field and the c axis.
The B-field dependencies up to 5 T of the I4 transition energies in B A c
and B // c are shown in Figure 7 together with the angular dependence at B =
5 T. These dependencies can be described as
E ( B, )
E0 1/ 2
( ))) ,
(3)
where E0 is the zero field transition energy and gh(T) is given by Eq. (2). For
the *9 holes the ghA =0 by the symmetry and the Eq. (3) leads to the
linear dependence off E(B,4) on cos((4) (see dashed lines in Fig. 7 (b)). For
the *7 holes in ZnO the value of ghA | 0 and the nonlinear behavior of
E(B,4) on cos((4) can not be detected as well (see solid lines in Fig. 7 (b)).
However, the symmetry of the holes can be unambiguously established by
the analysis of the angular dependence in Fig. 7 (b). Indeed, no critical
168
A. V. Rodina et al.
angle Tcr corresponding to gh(4crr)=ge and no respective crossing of the

transition energies expected for the *9 holes ( see dashed lines in Fig. 7 (b))
was observed. The excellent agreement between experimental data and
theoretical modeling is possible if and only if we assume that the hole g
factor is smaller than electron's g factor: |g
gh///| < ge. Together with the
selection rules observed in B // c // k configuration this corresponds to the
holes of the *7 symmetry. The values of the electron and hole effective g
factors used for the fitting (solid curves) are ge = 1.97, gh// = -1.21 and ghA =
0.1 for the I4 in Fig. 7 and ge = 1.86, gh// = -1.27 and ghA = 0.06 for the I9
line14.
The values of gh// derived for the I4 and I9 lines are very close to the gh// =
-1.24 obtained for the hole involved in the exciton bound to ionized donor
and to the g factor of the hole in 1S free exciton state.5,7,11 On the other hand,
the expected gh// values of the holes involved in the acceptor bound exciton
transitions differ significantly from the g values of the holes involved into
excitons bound to ionized or neutral donors. This is similar to the situation
found in CdS.22 Therefore we conclude that both I4 and I9 transitions should
be assigned to the ((D0;X
XA(*7)) complex rather than to the ((A0(*7);X
XA(*7)).
3,3640
(a) k||B || c
(b) B = 5 T
(c) B A c, k || c
Energy (eV)
3,3635
I4
3,3630
3,3625
V + ZnO
V - ZnO
3,3620
B (Tesla)
EA c ZnO
1,0
0,5
Cos(T)
0,0
B (Tesla)
Figure 7. Zeeman splitting of the bound exciton line I4. Case (a): B // k // c, E A c. Case (c):
B A c, k // c, E A c. Case (b): the angular dependence for B = 5 T. Symbols are experimental
data measured with circular polarized light (V + and V -) and with unpolarized light (E A c) as
denoted in the Figure, and lines are fits. For the fits the hole is assumed to be of *7
symmetry with | gh|||< ge (solid lines) and of *9 symmetry with | gh|||> ge (dashed lines).
169
This conclusion is supported by the analysis of the thermalization

properties of the I4 and I9 lines14 in the temperature dependent emission and
transmission spectra and by the observation of the two electron satellite
transitions for these lines.15,21,25
5.
CONCLUSION
The inverted *7(A

( ), *B(B
( ), *7(C)
C ordering of the valence subbands in
bulk ZnO was confirmed by the detailed analysis of the Zeeman splitting of
the free and bound excitons. The polarization properties and the angular
dependence of the transition energies from excitons bound to ionized and
neutral impurity centers indicated the *7 character of the upper A valence
band. The obtained *7 effective g values are in good with theoretical
calculations. We observed no low temperature PL transitions involving the
*9 hole states from the B valence subband.
ACKNOWLEDGEMENTS
The work of A. V. Rodina was carried out during the stay at the Institute
for Solid State Physics, TU Berlin, and supported by the Deutsche
Forschungsgemeinschaft (DFG).
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
D. G. Thomas, J. Phys. Chem. 15, 86 (1960).

Y. S. Park, C. W. Litton, T. C. Collins, and D. C. Reynolds, Phys. Rev. 143, 512 (1966).
B. Segall, Phys. Rev. 163, 769 (1967).
W. Y. Liang and A. D. Yoffe, Phys. Rev. Lett. 20, 59 (1968).
K. Hmmer, phys. stat. sol. 56, 249 (1973).
M. Rosenzweig, Diploma Thesis, TU Berlin, 1975.
G. Blattner, G. Kurtze, G. Schmieder, and C. Klingshirn, Phys. Rev. B 25, 7413 (1982).
D. C. Reynolds, D. C. Look, B. Jogai, C. W. Litton, G. Cantwell, and W. C. Harsch, Phys.
Rev. B 60, 2340 (1999).
9. B. Gil, Phys. Rev. B 64, 201310(R) (2001)
10. B. Gil, A. Lusson, V. Sallet, R. Triboulet, and P.Bigenwald, Jpn. J. Appl. Phys. Lett. 40,
L1089 (2001).
11. W. R. L. Lambrecht, A. V. Rodina, S. Limpijumnong, B. Segall, and B. K. Meyer, Phys.
Rev. B 65, 075207 (2002).
12. V. Val. Sobolev and V. V. Sobolev, Lith. J. of Physics 42, 189 (2002).
170
A. V. Rodina et al.
13. K. Hazu, T. Sota, K. Suzuki, S. Adachi, SF. Chichibu, G. Gantwell, D. B. Eason, D. S.

Reynolds, and C. W. Litton, Phys. Rev. B 68, 033205 (2003).
14. A. V. Rodina, M. Strassburg, M. Dworzak, U. Haboeck, A. Hoffman, A. Zeuner, H. R.
Alves, D. M. Hoffmann, and B. K. Meyer, Phys. Rev. B 69, 125206 (2004).
15. B. K. Meyer, H. Alves, D. M. Hofman, W. Kriegseis, T. Riemann, J. Christen, A.
Hoffmann, M. Strassburg, M. Dworzak, U. Haboeck, and A. V. Rodina, phys. stat. sol. (b)
2411, 231 (2004).
16. J. Gutowski, N. Presser, and I. Broser, Phys. Rev. B 38, 9746 (1988).
17. G. Blattner, C. Klingshirn, R. Helbig, R. Meinl, phys. stat. sol. (b) 107, 105 (1981).
18.P. Loose, M. Rosenzweig, and M. Wohlecke, phys. stat. sol. (b) 75, 137 (1976).
19.D. C. Reynolds, C. W. Litton, and T. C. Collins, Phys. Rev. 140, A1726 (1965).
20.D. C. Reynolds and T. C. Collins, Phys. Rev. 185, 1099 (1969).
21.K. Thonke, N. Kerwien, A. Wysmolek, M. Potemski, A. Waag, and R. Sauer, Mat. Res.
Soc. Symp. Proc. Vol. 719, F2.4.1 (2002); K. Thonke, N. Kerwien, A. Wysmolek, M.
Potemski, A. Waag, and R. Sauer, in Proc. 26th Int'l Conf. on the Physics of
Semiconductors, edited by R. Long and J. H. Davies, Institute of Physics Publishing,
Bristol (UK) and Philadelphia (USA), Institute of Physics Conference Series Number 171,
P22, (2003).
22. D. G. Thomas and J. J. Hopfield, Phys. Rev. 128, 2135 (1962).
23. T. Skettrup, M. Suffczynski, and W. Gorzkowski, Phys. Rev. B 4, 512 (1971).
24. A. V. Malyshev, I. A. Merkulov, and A. V. Rodina, Fiz. Tverd. Tela 40, 1002 (1998)
[Physics of the Solid State 40, 917 (1998)].
25.H. Alves, D. Pfisterer, A. Zeuner, T. Riemann, J. Christen, D. M. Hofmann, and B. K.
Meyer, Optical Materials 23, 33 (2003).
26.M. A. Lampert, Phys. Rev. Lett. 1, 450 (1958).
27. A. Zeuner, H. R. Alves, D. M. Hofmann, B. K. Meyer, A. Hoffmann, U. Haboeck, M.
Strassburg, and M. Dworzak, phys. stat. sol. (b) 234, R7 (2002).
Chapter 15
FUNDAMENTAL OPTICAL SPECTRA AND
ELECTRONIC STRUCTURE OF ZnO CRYSTALS
V. Val. Sobolev and V. V. Sobolev
Udmurt State University, 426034 Izhevsk, Russia
Abstract:
The analysis of ZnO exciton spectra is given. In the band and exciton theory,
necessity of taking into account hybridization of O 2p and Zn 3d states is
confirmed. Full sets of ZnO exciton spectra are obtained for 1.6, 4.2, and 90 K
for the first time. The main parameters of the three exciton series are
determined. Five variants of full sets of the fundamental polarized optical
functions of ZnO crystal were determined in the energy range 026 eV for
polarizations E A c and E || c. Three variants of dielectric permittivity and
volume electron energy loss polarized spectra of ZnO crystal were
decomposed into the transverse and longitudinal transition band components.
The main features and parameters of the five variants of the optical functions
spectra were established. A theoretical analysis of the obtained components
was performed, based on the known theoretical band schemes.
Key words:
Exciton, interband transition, optical spectra, reflectance, permitivity,

polarization, band structure
1.
INTRODUCTION
The polarization of the exciton spectra of ZnO does not follow the
classical model. All the three exciton series are strongly polarized and can be
observed only with E A c (A, B) and E || c (C). This has led to contradictory
position schemes of the two highest valence bands (*9, *7) and values of 'so,
'cr. There are also certain discrepancies in the results of different papers on
the nature of the lines of complex discrete reflectivity and absorption spectra
of ZnO at 4.2 K, as well as on the bands and excitons parameters.113
Polarized reflectivity spectra of ZnO have long been measured in the
wide energy region of fundamental absorption 026 eV, and some other
optical functions have been calculated from them.4,7,14,15 However, full sets
171
172
of optical functions and their decompositions into the elemental components

have not yet been determined, neither have the results of the three mentioned
papers been compared with each other.
In the present report, three subjects are to be reviewed: (i) The triplet
structure of excitons and the upper valence band; (ii) the full set of the
fundamental optical functions in the 030 eV energy range for polarizations
E || c, E A c, and their theoretical analysis; and (iii) the main parameters of
the elementary transverse and longitudinal transition components in the 030
eV energy range for polarizations E || c, E A c, and their theoretical analysis.
2.
CALCULATION TECHNIQUES
A full set of optical functions consists of reflectivity R and absorption

coefficients, P, the imaginary H2 and real H1 parts of the dielectric function H,
the absorption and refraction indices k and
d n, the product of the integral joint
density of states (DOS) function and the transition probability, equal within
constant factor to H2E2; the effective number of valence electrons neff(E)
participating in the transitions to given energy level E; the effective dielectric
coefficient Heff, and the characteristic electron energy functions for volume
( ImH-1) and surface (-Im(1+H)-1) losses.
The fine structure of optical transitions is determined by several main
parameters: the energy values of the maxima Ei and half-widths Hi of the
component bands, oscillator strengths f , and the areas Si of the bands which
equal transition intensity, to within a universal constant factor, and the
amplitudes Ii of the components maxima. The techniques used in the paper
were discussed16 and applied.17,18
3.
ANALYSIS OF THE STRUCTURE OF EXCITON

REFLECTIVITY, ABSORPTION AND EMISSION
SPECTRA
The exciton transitions of three series with n = 1, 2, 3 were found in the

reflectivity, absorption, and radiation spectra at a temperature T d 5 K, and
their binding energies R were determined.14 The ZnO samples were obtained
using different techniques, and their surfaces were treated in different ways
(cleavages, etched surfaces). The spectra were measured with
photographic and photoelectric techniques. Despite these substantially
different surface treatment and measurement techniques, a comparatively
good agreement was reached between the results of many papers concerning
Fundamental optical spectra and electronic structure of ZnO
173
the positions of the exciton lines. This agrees well with the simple classical
exciton model and contradicts the polariton model.
The problem has long been discussed of a connection between the
differences in the exciton spectra and the triplet band structure of ZnO and
other hexagonal crystals.111 In all these papers, the simplest model is
accepted on the assumption of invariability of the symmetry of the three
bands: in some papers, inversion of two upper bands is assumed,1,12 i.e.: the
highest band is *7 instead of *9, while other papers propose to retain the
conventional interleaving of bands *9, *7, and *7.2,3,10 This problem is
theoretically considered.9
It was shown in the quasicubic approximation of the group theory that the
relative energies of three upper valence bands with energies E1, E2, E3
are determined by the values of 'so and 'cr:
E2
A2
2 B 3, E3
A2 2 B 3
(1)
where 2 A ' so ' cr ; B ' so ' cr ; E1 = 0; 'so, 'crr << Eg. The values of 'so
for ZnO obtained with Eq. (1) vary in their absolute values and even in their
sign: 0.009,1 0.015,12 0.01610 eV. The calculations of 'so and 'crr are
known that use the classical model of band interleaving (*9, *7, *7) and
inversion (*7, *9, *7). The values of 'crr proved to be almost identical in all
the papers, regardless of the band model: 'crr = 0.041,1 0.035512 eV for *7,
*9, *7; 0.041,3 0.04310 eV for *9, *7, *7. However, the values of the second
parameter ('so) turned out to be mostly negative: 'so 0.009,1 0.01512 eV for
*7, *9, *7; 0.005,3 0.01610 eV for *9, *7, *7. Large divergence of the
calculation results for 'so seem to be due to too small a distance between two
( , B) exciton series and larger relative errors of ((BA
(A
)/A
/ .
Two experimental results have been known, which show a possibility of
considerable change of the parameters of all the three valence bands of ZnO.
First, polarization of all the three exciton series of ZnO conceptually differs
from their polarization in other A2B6 crystals that is in good agreement with
the conventional band model. This fact undoubtedly shows that in ZnO, not
only the symmetry of the second band changes, but the symmetry of the
third band also. Second, according to the photoemission spectra (PES), the
valence band of zinc d-states (Zn 3d) is located below the top of the valence
band of ZnO by ~ 8 eV, while below its bottom only by ~ 3 eV.19 Therefore,
Zn 3d states can intermix with O 2p states and change the symmetry of the
upper valence bands.
According to the theoretical calculations,20 O 2p states form ~ 67 74%
of the highest valence band, while Zn 3d states form the remaining almost
174
quarter of it. The analysis of the experimental X-ray photoemission spectra

(XPS) showed a lower contribution of d-states about 9%.21
Among the theoretical papers, the most interesting results are found by
Wei, et al.23 and Lambrecht et al.11 Wei, et al.23 show that interaction of
oxygen p-levels with zinc d-levels, i.e.: p d repulsion and hybridization,
reduces the spin-orbital splitting and alters the sign of the crystal-field
splitting; p d coupling mixes d-character into the wave function at the
valence-band maxima:
' so
' so ( p O) a
so
(d Zn)
(2)
According to the estimations, 'so(d Zn) | 0.37 eV, a = 0.07. Oxides BeO
and MgO, lacking d-bands, have 'so(p O) | 0.034 eV. Hence, 'so(ZnO) |
(0.0340.026) eV < 0.01 eV.
Lambrecht and his colleagues have theoretically shown11 that the
energies 'so and 'crr depend very strongly and almost linearly on the location
of the d-levels of zinc ((E
Ed). According to the authors, the most optimal
variant is that with Ed | 6.25 eV together with the band inversion model.
However, the experimental results give Ed | 8.5 r 0.4 eV.19 Unfortunately, it
is in the energy region E > 6.7 eV that the calculations of this paper are not
reliable. Thus, both the theoretical and experimental results indicate that zinc
d-states make a substantial contribution to the formation of the highest
valence band of ZnO. It is the hybridization of O 2p and Zn 3d states that
strongly changes the symmetry of the three valence bands and, consequently,
the polarization of three exciton series A, B, C, as well as the splitting
energies of the three bands.
4.
CALCULATIONS OF OPTICAL FUNCTIONS

AND EXCITON PARAMETERS
From the experimental polarized exciton reflectivity spectra of papers2,4,5

and the reflectivity spectra in wide energy region up to 25 eV,14 three full
sets of optical fundamental functions of excitons were calculated at 1.6 K4
(L), 4.2 K2 (P), and 90 K5 (K). For the sake of briefness, some of them are
given in Figs. 1.
The analysis of the H2 and ImH-1 spectra allowed to determine the energy
values of transverse and longitudinal exciton transitions, and to evaluate
their band areas and oscillator strengths (Tab. 1); the values of ff, given here,
have been enlarged 105.
175
Table 1. Energy values E (eV) of H2 and Im H1 maxima, bands areas S, and oscillator
strengths f u 105 of ZnO excitons, calculated from the R(E) spectra of papers Lagois, et al.4
(L), Park, et al.2 (P), and Klucker, et al.5 (K)
N
B1
C1
A1
H2
ImH1
H2
ImH1
H2
ImH1
L,
E
3.3766
3.3792
3.3830
3.3948
3.421 3.439
0.16
0.0005
0.20
0.008
0.32
0.007
1.6 K S
f 105 (107)
fu
7.5
2.4
9.4
38
15
33
P,
E
3.3746
3.3772
3.3796
3.3904
3.419 3.433
0.20
0.0006
0.16
0.006
0.50
0.006
4.2 K S
f 105 (107)
fu
9.4
2.8
7.5
2.8
23.5
28
K,
E
3.349
3.38
3.349
3.38
3.387 3.424
90 K
S
0.50
0.008
0.50
0.008
0.50
0.008
f 105 (107)
fu
24
38
24
38
24
38
N
A2
B2
C2
ImH1
H2
ImH1
H2
ImH1
H2
L,
E
3.43
3.43
3.44
3.43
3.47
3.47
1.6 K S
0.02
0.0005
0.03
0.0005 0.05
0.0007
f 105 (107)
fu
0.94
2.35
14
2.4
2.35
3
P,
E
3.42
3.42
3.43
3.43
3.46
3.47
4.2 K S
0.04
0.04
0.02
0.001
0.1
0.003
f 105 (107)
fu
1.9
1.9
0.94
4.7
4.7
1.4
K,
E
3.41
3.41
3.41
3.43
3.453
3.49
S
0.10
0.10
(0.10)
0.008
0.05
0.001
90 K
f 105 (107)
fu
4.4
4.4
4.7
38
2.4
4.7
5.
RESULTS OF CALCULATIONS OF OPTICAL

FUNCTIONS IN THE 0 - 26 eV ENERGY RANGE
Using the R(E

( ) spectra of the three papers,5,14,15 we calculated three
variants of sets of polarized spectra of all the optical functions ((A, B, C from
the R of papers15,14,5). Some of them are given in Figs. 1. The polarized
reflectivity spectra of ZnO in the energy range 018 eV were calculated
theoretically in three papers.23,24,25 The results of Wei, et al.23 differ very
much from the experimental spectra.5,14,15 Because of this, our calculations of
polarized spectra of all the remaining optical functions are based on the
theoretical R(E
( )24 and H2(E)25 spectra. Some of the calculated spectra are
given in Fig. 1 (curves 4 and 5).
The most long-wavelength maximum of the experimental reflectivity
curves at E A c and E || c and that of the other optical functions spectra,
calculated from them (the experimental-calculated ones), is caused by free
excitons. Therefore, there should be no such maximum in the theoretical
spectra obtained from the bands.
176
0.25
0.02
0.15
0.00
2.5
3
1 4
0.01
-1
P (10 cm )
0.20
2
5
3.0
(a)
3.5
4.0
0.10
0.05
0.00
10
15
E (eV)
20
0.25
(b)
2
0.02
-1
P (10 cm )
0.20
0.15
0.00
2.5
25
5
3.0
0.10
3.5
4.0
0.05
0.00
10
15
E (eV)
20
25
Figure 1. The P (a, b) and E2H2 (c, d) spectra of ZnO crystal for EAc (a, c) and E||c (b, d) in
the range of 025 eV (in the inset in the range of 2.54.0 eV), calculated from the
experimental R spectra (1, 2, 3) and from the theoretical H2 and H1 spectra (4, 5)
1000
2
30
H (E )
800
20
600
(c)
3
4
10
0
2.5
5
3.0
177
3.5
4.0
400
200
0
40
800
20
0
2.5
600
15
E (eV)
4
5
3.0
3.5
4.0
200
25
(d)
400
20
H2 (E )
1000
10
10
15
E (eV)
20
25
Figure 1. Continued: The P (a, b) and E2H2 (c, d) spectra of ZnO crystal for EAc (a, c) and
E||c (b, d) in the range of 025 eV (in the inset in the range of 2.54.0 eV), calculated from
the experimental R spectra (1, 2, 3) and from the theoretical H2 and H1 spectra (4, 5)
178
The maxima of the theoretical R(E) spectra of papers ref. 24 and 25 differ
one from the other by ~ 0.21 eV, while the differences between these and
the maxima from papers5,14,15 come up to 1.5 eV, with the intensity
distribution of the maxima remaining mostly similar. A noticeable general
similarity was also discovered as a result of direct comparison of the
remaining experimental and calculated, and theoretical spectra of the optical
functions (Fig. 1). However, the detailed comparison of the theoretical and
experimental spectra of the optical functions showed, naturally, substantial
differences in the intensities and locations of the maxima.
In papers 3336, the photoemission spectra of ZnO crystal were
experimentally studied. According to the presented results Zn 3d electron
energy levels of ZnO ((E
Ed) are lower than the maximum of the valence bands
for ~ 8.5 eV,19 7.5 eV,34 and 7.5 eV,21 while the total width of the highest
valence band ('E
'Ev) is equal to ~ 5.0 eV,34 8.0 eV,19 and 6.0 eV.21 From the
theoretical band calculations, 'E
'Ev is equal to ~ 6 eV24 and 4.2 eV,25 while Ed
25
| 4.2 eV; zinc d-levels being ignored in paper ref. 21. This means that both
important band parameters of ZnO ('E
'Ev and Ed) are undervalued in paper
ref. 24 by ~ 2 eV and 4 eV for 'E
'Ev and Ed, respectively. This must have led,
in the first place, to underestimation of the energy values of the theoretical
transitions from the lower sub bands of the highest valence band and
alteration of the transitions intensities.
6.
THE H2 AND Im H -1 SPECTRA IN THE 026 eV

ENERGY RANGE
Spectra of H2 and ImH-1 of the three variants of optical functions sets ((A,
B, C
C) were deconvoluted by us into components. The main parameters of the
obtained components are given in Tables 1, 2 ((A), 3, 4 ((B), and 5 (C),
C as well
as in Fig. 1.
In Table 2, the calculated values of energy maxima Ei and band areas Si
of transverse transitions were compared for the three variants of calculations
of optical function sets. These variants differ comparatively little in energy
Ei. The principal inconsistencies are observed in the values of Si. This is
caused by the differences in the experimental R(E) spectra, i.e. by using of
ZnO samples of different quality and by different registration techniques for
polarized reflectivity spectra. The determined components of H2 and ImH-1
spectra are caused by direct interband transitions or metastable excitons,
except for the most long-wavelength of them, which are associated with free
excitons. The theoretical band calculations of ZnO20,2325 strongly differ in
the bands dispersion and positions. This makes it difficult to propose a
179
model of specific and unambiguous connection between the determined

components and transitions between pairs of bands. It is commonly accepted
that transitions can be expected between those pairs of bands that are parallel
in the largest portion of the Brillouin zone.16 Using this simplified model, we
estimated the energies of such transitions in the vicinity of points *, A, M,
K, and L for the bands of papers20,2325 (Table 2,3). These results are in
accordance with our calculations of the components energies (Tables 2,3).
Table 2. Energy (eV) of the maxima Ei, band areas Si, oscillator strengths f of the transverse
(H2) and longitudinal (H1) transition components of ZnO crystal for E A c (A) and E __ c (__)
and theoretical interband transitions En
Ei, K
N
Si
f
Ei
A
||
A
||
A
||
H1
H2
H1
H2
H2
H2
1
3.28
3.57
0.36
1.59
3.4
0.26
2.56
3
3.58
0.62
1.01
3.7
3.7
0.84
1.9
4
4.34
4.42
1.11
0.9
3.8, *
5
4.7
4.9
1.11
1.01
5.0, *
6
6.2
6.1
2.67
1.21
7.5, '
7
6.7
6.7
2.36
1.2
8
8.4
8
0.57
0.21
9.5, ', 6
9
8.8
9.3
9
9.3
4.22
1.39
1.05
0.2
10, U
10 11.1
11.4
11.2
11.4
0.67
1.66
0.12
0.35
11.0, S
11 12.5
12.7
12.9
12.9
6.85
2.2
0.84
0.32
12.6, U
12 13.7
13.7
13.7
13.8
1.78
6.83
0.17
0.77
14.0, S
13 14.7
14.7
15
15
3.8
3.51
0.31
0.3
14.6, U
14 16
16
15.5
1.87
1.69
0.14
0.13
16.7, '
15'
16.4
16.3
0.07
15 17.2
17.3
17.5
17.5
0.77
0.45
0.06
0.03
17.8, U
16 18.7
18.4
18.7
18.3
0.48
0.89
0.03
0.06
16
19.1
19.1
17'
20.1
19.7
20.3, U
17 20.7
21.5
20.9
20.9
1.47
1.89
0.11
0.28
18 22.9
23.4
23.4
23.7
0.34
1.08
0.03
0.08
22.0, U
19 25.1
25.1
2.46
0.19
20
26.8
26.8
2.31
0.17
7.
CONCLUSIONS
Thus, an analysis of the known polarized exciton reflectivity, absorption,

and radiation spectra of ZnO crystal has been performed for the first time.
High stability of the positions of the lines of three exciton series has been
180
found despite the substantial differences in the monocrystal's growing

processes and the spectra recording techniques. This argues against the
necessity of application of the polariton model in these papers.
The conventional or inverse model of triplet structure of the highest
valence band applied to explanation of the exciton spectra of ZnO has been
shown to be much too simplified. It is necessary to take into account
Table 3. Energy Ei (eV) of the maxima and band area Si of the components of H2 spectrum
decomposition at E A c (A) and E __ c (__) in three variants of optical functions sets ((A, B, C
C) of
ZnO crystal
Ei
Si
N
A
||
A
||
A
B
C
A
B
A
B
C
A
B
1
3.28
3.10
3.10
3.10
0.36
0.44
0.15
0.35
2
3.18
3.19
3.20
0.08
0.22
0.09
2
3.46
3.48
3.40
3.35
0.68
0.05
0.26
0.95
3
3.58
0.62
3.69
3.70
3.70
0.91
0.84
0.05
4
4.34
4.10
4.25
4.15
1.11
0.75
0.14
2.16
5
4.60
4.70
4.80
0.63
0.13
6
6.20
5.38
5.13
5.25
2.67
0.39
0.96
1.19
7
7.20
7.3
6.70
7.40
1.10
0.22
2.36
1.91
8
8.2
8.40
0.06
0.57
9
8.80
8.8
9.00
9.1
4.22
0.53
1.39
0.65
10 11.1
11.4
11.2
11.1
0.67
0.21
1.66
1.08
11 12.5
12.8
12.3
12.9
12.0
6.85
2.74
0.35
2.20
0.67
12 13.7
13.7
12.7
13.7
13.7
1.78
0.38
1.68
6.83
6.77
13 14.7
15.1
14.0
15.0
14.8
3.80
7.58
1.20
3.57
0.38
14 16.0
15.5
15.5
16.0
1.87
4.32
1.69
15 17.2
17.3
17.6
17.5
17.9
0.77
3.75
3.88
0.45
1.46
16 18.7
19.5
18.8
18.7
0.48
1.42
0.16
0.89
17 20.7
21.9
20.4
20.9
21.9
1.47
2.60
2.26
1.89
5.05
18 22.9
23.3
23.4
23.3
0.34
0.24
1.08
19 25.1
24.7
24.3
24,7
2.46
1.09
2.50
0.05
20
26.8
2.31
hybridization of O 2p- and Zn 3d-states and change of symmetry of all

the three valence bands, as well as the relative values of their splitting
energy.
For the first time, the full sets of exciton optical fundamental functions of
ZnO have been calculated at 1.6, 4.2, and 90 K. The most correct energy
values of transverse and longitudinal exciton transitions of the three series,
together with their areas and oscillator strengths, have been determined, as
well as their characteristic features.
The spectra of full sets of optical functions of ZnO crystals in the range
024 eV were calculated for polarizations E A c and E || c from three
181
experimental reflectivity spectra5,14,15 and from two theoretical R(E)24 and

H2(E
( )25 spectra. Very complex structures and strong polarization of most
transition band maxima were found for five sets of optical functions. Good
general agreement was established between the experimental and the
theoretical optical function spectra.
The differences between the experimental R(E
( ) spectra of ZnO for
polarizations E A c and E || c in the range 324 eV (together with the spectra
of the other optical functions, calculated from them)68 and the theoretical
R(E
( )21 and H2(E)26 spectra (together with the spectra of the other optical
functions, calculated from them) are caused mainly by simplifications of the
theoretical band calculation models, as well as by not taking into account the
competitiveness of the metastable excitons model.
Thus, for the first time, three variants of H2 and ImH-1 polarized spectra
of ZnO crystal in the energy region 026 eV have been decomposed into
elementary components. The main parameters of the transverse and
longitudinal transition components were found. The main features of the
transitions parameters were determined. Having analyzed the theoretical
bands provided by seven calculations of the electronic structure of ZnO,
performed using different techniques, we estimated the energy values of
possible interband transitions and their localization in various points of the
Brillouin zone in the range of 330 eV. These simplified estimations proved
to be in good agreement with the calculated energy values of the H2(E
( )
spectra decomposition components.
The obtained results allow to study the electronic structure and optical
properties of ZnO crystal on the basis of conceptually deeper and more
detailed background, than it was previously available on integral reflectivity
spectra, and to base theoretical calculations of the properties of zinc oxide
upon new foundations.
ACKNOWLEDGEMENTS
The paper was supported by the Fundamental Natural Sciences Center,
State University of Saint-Petersburg, Russia.
REFERENCES
1.
2.
3.
4.
D. G. Thomas, J. Phys. Chem. Sol. 15 (1), 8696 (1960).

Y. S. Park, C. W. Litton, T. C. Collins, and D. C. Reynolds, Phys. Rev. 143, 512 (1966).
W. Y. Liang and A. D. Yoffe, Phys. Rev. Lett. 20, 59 (1968).
J. Lagois and K. Hummer, phys. stat. sol. (b) 72, 393(1975).
182
5. R. Klucker, H. Nelkowski, Y. S. Park, M. Skibowski, and T. S. Wagner, phys. stat. sol. (b)
45, 265 (1971).
6. V. V. Sobolev, V. I. Donetskich, and E. F. Zagainov, Fizika i Tekhnika Poluprovodnikov
12, 1089 (1978).
7. V. Sobolev, (Stiintza, Kishinev, 1980).
8. P. Kuzmina and V. A. Nikitenko, Poluchenie i opticheskie svoystva, (Nauka, Moscow,
1984).
9. J. Hopfield, J. Phys. Chem. Sol. 15, 97 (1960).
10. Reynolds, D. C. Look, B. Jogai, and C. W. Litton, Phys. Rev. B 60, 2340 (1999).
11. L. Lambrecht, A. V. Rodina, and B. Segal, Phys. Rev. B 65, 75207 (2002).
12. H. Yoshikawa and S. Adachi, Jpn. J. Appl. Phys. 36, 6237 (1997).
13. G. E. Jellison and L. A. Boatner, Phys. Rev. B 58, 3586 (1998).
14. J. L. Freeouf, Phys. Rev. 7, 3810 (1973).
15. R. L. Hengehold, R. J. Almassy, and F. L. Pedrotti, Phys. Rev. B 1, 4784 (1970).
16. V. V. Sobolev and V. V. Nemoshkalenko, Elektronnaya struktura poluprovodnikov
(Naukova dumka, Kiev, 1988).
17. V. V. Sobolev, A. I. Kalugin, V. Val. Sobolev, and V. I. Kormilets, Wide Bandgap
Materials 8, 87 (2001).
18. V. Val. Sobolev and V. V. Sobolev, Phys. Low Dim. Struct. 5/6, 65 (2003).
19. C. J. Vesely, R. L. Hengehold, and D. W. Langer, Phys. Rev. B 5, 2296 (1972).
20. D. Vogel, P. Kruger, and J. Pollmann, Phys. Rev. B 54, 5495 (1996).
21. S. A. Leontyev, S. V. Koshcheyev, V. G. Devyatov, A. E. Cherkashin, and E. P.
Mikheyeva, Zh. Strukturnoy Khimii 38, 867 (1997).
22. S. Bloom, and J. Ortenburger, phys. stat. sol. (b) 58, 561 (1973).
23. S. B. Zhang, S.-H. Wei, and A. Zunger, Phys. Rev. B 37, 8958 (1995).
24. J. R. Chelikowsky, Solid State Commun. 22, 351 (1977).
25. J. Xu and W. J. Ching, Phys. Rev. B 48, 4335 (1993).
Chapter 16
PHOTO-INDUCED LOCALIZED LATTICE
VIBRATIONS IN ZnO DOPED WITH 3D
TRANSITION METAL IMPURITIES
Alexey Kislov
Ural State Technical University, Mira Str. 19, 620002 Ekaterinburg, Russia
Abstract:
The purpose of this work is twofold: to simulate vibrational spectra of ZnO crystals
with nickel impurities in different charge states and to analyze vibrational
structures of electroabsorption spectra of ZnO:Ni based on calculations of
localized vibrations induced by the impurities.
Key words:
Impurity exciton, electroabsorption spectra, molecular static method,

vibrational density of states, recursion method
1.
INTRODUCTION
The progress in development of optoelectronics demands the further

investigation of new materials and new effects being a basis for the
production of various optoelectronic systems. A special interest is focuses on
II-VI semiconductor compounds containing 3d transition metals as impurities.
Due to the unique combination of physical properties ZnO crystals are
potential candidates for numerous industrial applications.
In recent years some researchers were attracted by the photorefractive
effect in such systems. It is based on the photoionization of 3d impurities by
light. Note that at the photoionization of these impurities a non-stationary
charge with respect to the lattice is created. Owing to the change of an
impurity charge there can be a lattice deformation around 3d impurities.
Besides, in the distorted part of the lattice photo-induced localized modes
connected to the charged impurity can appear. Such modes are difficult to
observe with traditional methods such as infrared absorption and Raman
light scattering because the concentration of the charged centers is small.
183
184
Alexey Kislov
Under illumination the 3d impurity can pass in excited states of a

hydrogen-like type as well. When the impurity loses the electron from the
3d-shell it is positively charged with respect to the lattice and the electron is
confined by the long-range Coulomb field of the 3d impurity to a hydrogenlike orbit (see Fig. 1a and Fig. 1c). As a result a donor exciton is formed:1 dn
+ Zd o [dn-1e]. The acceptor exciton is formed after the transition of the
electron from the valence band to the impurity and the Coulomb field of the
negatively charged impurity keeps the hole in the hydrogen-like orbit (see
Fig. 1b and Fig. 1c). Schematically, the process of formation of the acceptor
exciton is written as follows: dn + a o [dn+1h].
Figure 1. Diagrams of energy levels of a 3d impurity [(a) and (b)]: transitions involving an
ionization of the impurity (D, A) and transitions to hydrogen-like states (d, a) are shown. Fig.
1(c) is an illustration of the transitions (d, a) in a space near a Ni impurity.
Photo-induced localized lattice vibrations in ZnO
185
The formation of the impurity exciton leads to an appearance of photoinduced lattice vibrations. Sokolov1 has offered an experimental
electroabsorption (EA) method, which has a very high sensitivity for
detecting such vibrations. However, the analysis of experimental data
supposes a theoretical study of the vibrational spectrum of the defective crystal.
The purpose of our work was to simulate vibrational spectra of ZnO crystals
containing nickel impurities in different charge states with respect to the
lattice and also to analyze vibrational structures of EA spectra of ZnO:Ni based
on calculations of localized vibrations induced by the impurities.
Figure 2. Illustrating the way the defective crystal is modeled.
2.
METHODS OF CALCULATION
The lattice relaxation near the impurities was calculated on the basis of
the molecular static method2. The crystal with the impurity was dividing into
regions: an inner region 1 and an outer region 2+3 (see Fig. 2). The energy of
the static lattice of this system can be written as2
&
Ests E
E(( ,[ )
( )
&
( , )
&
([ )
(1)
186
Alexey Kislov
&
where& E1( ) is the energy of the region 1, E3 ([ ) is the energy of region 3, and
&
E2 ( , ) is the interaction energy of regions 1 and &3. x is the coordinate
describing the lattice configuration of region 1 while [ is the displacement in
&
region 3 and is formally distinguished from x .
The ion displacements in region 1 can be determined by direct
&
minimization of the defect energy with respect to x , i.e.: by solving the
equation
i,
i j
wwE
E st
wx
wx
x(
i, j, ,
i j
w 2 E st
w xi w x j E
&
x(
u stjE
(2)
The defect formation energy is the difference in energy between the

relaxed defective crystal and the ideal crystal: 'Ed = Edstt Eidst.
The theoretical study of the lattice dynamics of the defective crystals
conventionally makes use of the local vibrational density of states (LVDS)3
giD(Z)3. It is related to the diagonal element of the Greens function GuiD(Z)
by the expression3
2Z
g iD ( )
Im
uiD
( ) .
(3)
The latter in turn is determined by the dynamical matrix D of the crystal

GuiD ( )
iD
| (
|[
0))
] 1 | uiD
3N
s 1
| isD |2
.
Z i00 Z s2
2
(4)
Using the recursive relationship,4
u n 1
(D
)u n
bn 1u n 1
(5)
we transform the dynamical matrix into tri-diagonal form

a1
1
D 0
...
0
b1
0 0 ... 0
b2 0 ... 0
2
1 ... ... ... ... .
... ... ... ... ...

0 ... ... 1 a3 N
(6)
187
In this case the diagonal element of the Green's function are described by the
easily calculated continued fraction
Gu1 ( )
Z 2 a1
b1
(7)
b2
Z 2 a2
...
bn1
2
Z an bnt ( )
where t() is a truncation function.

The well-known shell model5 was used for modeling the local atomic
structure and the lattice dynamics of the defective crystal. In this model, the
lattice energy is written as
E
lat
1
2 i, j
1
(r )
k P2
ij 2 i i
i
j i
(8)
1
[
2 i, j
1
( )]
k P2
ij
2i i i
( )
ij
j i
The structure and dynamics of the lattice were simulated by using

empirical potentials of pair wise interactions. The interaction between ion cores
is assumed to be long-range purely Coulombic. The interaction between electron
shells has two components: a long-range purely Coulombic interaction and a shortrange interaction described by the Buckingham potential.
V Coul (rr )
ij
V Buck ( ij )
X X
XY
YX
YY
i j
i j
i j
i j

c
s
c
c
s
s
r
|r
r | |r
r | |r
rs |
ij
j
i
j
i
j
i
ij
exp( |
s
j
s
i
|/
ij
)
Ciij
|
s
j
s 6
i
(9)
(10)
188
3.
Alexey Kislov
The parameters of the ion-ion interaction potentials in ZnO crystals are

listed in Table 1. The values of the parameters were obtained by theoretical fits
to the measured elastic and dielectric constants, the lattice constants, the
lattice energy (see Table 2), and the phonon dispersion curves.
Table 1. The parameters Aij, ij and Cij of the Buckingham potential.
Uij ()
i j ions
Aij (eV)
Zn+2Zn+2
Zn+2O-2
O-2O-2
890.0
1265.0
88760.0
0.300
0.300
0.149
Table 2. Potential charges and spring constants.

Yi (~e~)
Ion
Zn+2
3.749
O-2
-2.930
Table 3. Calculated and observed crystal properties for Zn.
Properties
Calculated
Lattice constants ()
a
3.25
c
5.21
Lattice energy (eV)
40.8
Elastic constants (GPa)
C11
237.4
C12
127.1
C13
115.8
C33
258.4
60.2
C44
C66 = (C11 C12)/2
55.2
Dielectric constants
H0A
5.60
H0__
6.10
Hf A
3.82
Hf__
3.91
Phonon spectrum gap (THz)
8.8 11.3
Qmax (THz)
16.6
Cij (eV/ )
(Ref. 6)
0.0
0.0
0.0
ki (eV/ 2)
116.4
51.8
Observed
3.25 (Ref. 6)
5.21 (Ref. 6)
42.2 (Ref. 7)
209.7 (Ref. 8)
121.1 (Ref. 9)
105.1 (Ref. 9)
210.9 (Ref. 8)
42.47 (Ref. 8)
44.29 (Ref. 8)
8.33 (Ref. 6)
8.84 (Ref. 6)
3.94 (Ref. 8)
4.00 (Ref. 8)
8.0 11.5 (Ref. 8)
17.7 (Ref. 10)
As an example, Fig. 3 plots the phonon dispersion curves for three highly
symmetric directions in the Brillouin zone of the perfect ZnO crystal.
Comparison of the theoretical and experimental frequencies8 shows good
agreement for the acoustic branches. The densities of phonon states of the
perfect ZnO crystal calculated by integrating over the Brillouin zone are
displayed in Fig. 4. Comparison of the results of our calculation and a calcu-
189
Figure 3. Phonon dispersion curves for ZnO crystals. The points denote experimental values.8
lation based on the valence force model11 shows clearly that the approximation
of pairwise central interactions is satisfactory. The difference in the position of the
phonon forbidden region and of some maxima is due to limitations associated
with the pairwise interaction approximation used in the calculations.
The formation of the nickel excitons results in the lattice distortion near
them and the induced lattice vibrations. Conditions of their occurrence are
defined by the charged impurity because the removed hydrogen-like type
carrier practically does not influence the deformation of the lattice near the
charged impurity. Thus, our analysis of the vibrational background of the
zero phonon line of the EA spectrum of the nickel exciton is based on results of
a simulation of the lattice dynamics of the ZnO crystal with Ni+1 or Ni+3 ions.
An important question is how to correctly describe the interaction between
the defect and the remaining atoms of the crystal. The parameters of the shortrange interaction potentials between the Ni impurities in various charge
states and its surrounding ions being taken as the corresponding parameters
190
Alexey Kislov
Figure 4. Total density of phonon states in ZnO crystals. Dashed line - result of the
calculation from Ref. 11; solid line - our calculation.
for Zn ions. Of course, we took into account the change of mass and charge
associated with the substitution of Ni+1 or Ni+3 for Zn.
Table 3 lists the displacements of nearest neighbors around the charged
nickel. The calculations show that the nearest neighbors of the Ni+1 impurity are
quite removed from it. For Ni+3 they shift toward the impurity.
The effect of charged nickel ions on the vibration spectrum of ZnO crystals
was studied by calculating the symmetrized LVDS in perfect and defective
crystals using a cluster of 1000 ions (region 1). For the calculation of the
Coulombic component of the diagonal elements of the force constant
matrixes of the ion-ion interactions by the Ewald method about 4000 ions
were considered for region 2.
We have calculated the symmetrized LVDS projected onto a region
containing two coordination spheres around a zinc ion or its nickel
replacement. Nickel ions in ZnO crystals are in the positions with the
symmetry point group C3v. The 36-dimensional displacement space splits
into the irreducible representations of the C3vv point group: 3(36u36)= 91 +
32 + 12.
191
Table 4. The defect formation energy and the displacements around the defect.
Defect
Ni+1
'Ed (eV)
18.4
Ni+3
-30.3
Ion (coord. Sphere)

O-2 (1)
Zn+2 (2)
O-2 (1)
Zn+2 (2)
Displacement ()
0.251
-0.017
-0.240
0.027
In Fig. 5 LVDS projected onto displacements with A1-symmetry of the center

ion and the O-2(z) ion for ZnO:Ni crystals are presented. Specific features of the
symmetrized LVDS in a defective crystal, which differ from those of the
symmetrized LVDS in a perfect crystal, corresponded to both resonant
vibrations and gap vibrations induced by defects. Local vibration frequencies
were determined by observing the change of the sign of the real part of the
Greens function. For example, the resonant vibration located in the acoustic
band at the frequency 3.0 THz (see Fig.5a) is associated with motion of the Ni+1
impurity.
Figure 5. LVDS projected onto displacements with A1-symmetry of the center ion (a) and the
O-2(z) ion (b).
Special attention is given to the localized vibrations with the symmetries

A1 and E. According to a group-theoretic analysis, they are active in the EA
spectra of ZnO:Ni crystals. The frequencies of such localized vibrations
induced by nickel impurities in ZnO crystals are listed in Table 4.
192
Alexey Kislov
For ZnO:Ni the EA spectrum depicted in Fig. 6 (a) was measured at liquidhelium temperature T=4.2 K;12 therefore, its vibrational part is shifted into the
Stokes region of the spectrum and is associated with phonon creation.
Table 5. The frequencies of localized vibrations in ZnO with nickel impurities.
Defect
Type of mode
Type of symmetry
Frequency (THz)
Resonance
A1
3.0, 13.8
Gap
A1
10.0
Ni+1
Local
A1
16.8
Resonance
E
3.0, 13.8
Local
E
16.8
Ni+3
Resonance
A1
Gap
Local
Resonance
A1
A1
E
Gap
Local
E
E
4.6, 6.8, 7.8, 14.0,

14.8
10.8
17.4, 19.6
4.6, 6.8, 7.8, 14.0,
14.8
9.8, 10.8
18.0
In the first-order phonon sidebands of the EA spectrum there are four

features (arrows in Fig. 6(a)) that correspond to the maxima of the absorption
spectrum. In our case, these features in the EA spectrum are zero-crossing points
or points associated with the largest slope in the regions of fall-off of the
absorption coefficient. These features correlate very well with the four strong
peaks in the symmetrized LVDS projected onto displacements with A1- and Esymmetries of the ions of two coordination spheres around a Ni+1 impurity (see
Fig. 6(b)). These peaks correspond to the localized vibrations induced by the
Ni+1 impurity.
Our calculation on defect vibrations enables the assignment of the EA
spectrum features. For example, one is found at a frequency of 10.0 THz from
the zero phonon line and falls in the region between the acoustic and optical
bands of the phonon spectrum of the ZnO crystal. Perhaps it is due to an A1motion of the Ni+1 ion, the O-2 ions of the first sphere and the Zn+2 ions of the
second sphere. According to our calculation, the EA spectrum for ZnO:Ni
corresponds to the nickel acceptor exciton [d9h].
4.
CONCLUSIONS
The effect of charged nickel ions on the vibrational spectra of ZnO

crystals was investigated by the recursion method using the shell model. The
d vibrations of different symmetries were calculated
frequencies of localized
with allowance for the atomic displacements in the vicinity of the charged
impurities. These data on defect vibrations were applied to the interpretation
193
Figure 6. Spectral dependence of the EA second harmonic D2 for ZnO:Ni (a). Symmetrized
LVDS projected onto displacements with A1- and E-symmetries of the ions of two
coordination spheres around a Ni+1 impurity (b).
of the vibrational background of the zero phonon line of the EA spectrum of

ZnO:Ni. The analysis of the EA spectrum based on model calculations of
localized modes near a Nii+1 charged impurity has advanced our understanding of
the vibrational structure of the EA spectrum. It was shown that the EA
spectrum for ZnO:Ni corresponds to the nickel acceptor exciton.
194
Alexey Kislov
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
V. I. Sokolov, Semiconductors 28, 329 (1994).

M. P. Puls, C. H. Woo, and M. J. Norgett, Phil. Mag. 36, 1457 (1977).
P. E. Meek, Phil. Mag. 33, 897 (1976).
S. K. Bose, Phil. Mag. B 49, 631 (1984).
B. G. Dick, A. W. Overhause, Phys. Rev. 112, 90 (1958).
P. P. Kuzmina and V. A. Nikitenko, (Nauka, oscow, 1984).
G. V. Lewis and C. R. A. Catlow, J. Phys. C: Solid State Phys. 18, 1149 (1985).
K. Thoma, B. Dorner, G. Duesing, and W. Wegener, Solid State Commun. 15, 1111
(1974).
9. T. B. Bateman, J. Appl. Phys. 33, 3309 (1962).
10. J. M. alleja and M. Cardona, Phys. Rev. B 16, 3753 (1977).
11. S. V. Melnichuk, V. I. Sokolov, . P. Surkova, and V. . Chernov, Phys. Sol. State. 33,
1833 (1991).
12. A. N. Kislov, V. G. Mazurenko, V. I. Sokolov, and A. N. Varaksin, Phys. Sol. State. 41,
897 (1999).
PART V: DEVICE APPLICATIONS
Chapter 17
ZnO WINDOW LAYERS FOR SOLAR CELLS
Walther Fuhs
Hahn-Meitner-Institut Berlin, Kekulstrae 5, D-12489 Berlin, Germany
Abstract:
ZnO layers are very frequently used as window layers in photovoltaic solar cell
structures. In this application high electrical conductivity should be
combined with a low value of the optical absorption constant in the visible
range. Thus an appropriate measure of the performance is the ratio r = / and
the optimization consists in realizing high by high carrier concentration and
mobility with a minimum optical loss due to free carrier absorption at long
wavelength. Such layers can be prepared by numerous techniques most
frequently by magnetron sputtering either from compound targets or reactively
from metallic targets. Optimized doped ZnO-layers have resistivities of 1.52u10-4 cm with Hall mobilities of up to 60 cm2/Vs, refractive indices of
approximately 2.0 and an average transmittance of 85 % in the visible range. It
appears that for the value of the physical limit as defined by ionized
impurity scattering has been reached for homogeneously doped thick layers.
This paper will discuss various aspects of the application of ZnO window
layers in photovoltaic solar cells.
Key words:
Polycrystalline films, heterointerface window layers, electrical transport,

doping, light trapping
1.
INTRODUCTION
Transparent, electrically conducting films (TCO) have been prepared

from a large variety of materials such as semiconducting oxides of Sn, In,
Zn, and Cd.1 The uses of such materials are manifold. Particularly important
are low emissivity glass for building applications, energy-saving windows,
front electrodes in flat-panel displays, thermal solar collectors, solar cells or
smart windows. In such applications high electrical conductivity should be
combined with a low value of the optical absorption constant in the visible
range. Thus an appropriate measure of the performance is the ratio r = /
and the optimization consists in realizing high by high carrier
197
198
Walther Fuhs
concentration and mobility with a minimum optical loss due to free carrier
absorption at long wavelength. However, for a particular application this
figure of merit is not the only parameter that has to be considered. Further
aspects may become important such as physical, chemical and thermal
durability, stability, homogeneity, chemical etchability, workfunction,
method of preparation, toxicity, availability of the materials or simply cost
factors. So far tin oxide, SnO2, prepared by chemical vapor deposition
(CVD) is most commonly used in low-energy windows and indium tin oxide
(ITO) made by spray pyrolysis or sputter deposition is the preferred material
for flat panel displays.1 ZnO films are used in particular for transparent
contacts in photovoltaic solar cells (window layers). However, due to
potentially lower cost, material availability and easy chemical etchability
ZnO may become attractive also for display applications in the future.
This article deals with various aspects of ZnO window layers in
photovoltaic cells. The copper indium diselenide CuInSe2 (CIS) thin-film
solar cell uses a ZnO window with a thickness of about one m. Here ZnO
forms a heterojunction with a CdS buffer layer.2 In the superstrate
configuration of the amorphous silicon a-Si:H pin solar cell light enters
through the glass substrate which is covered by a thick ZnO layer. This layer
is chemically structured to enhance light scattering to enable the efficient use
of sun light.3 A third example is a solar cell on basis of a silicon wafer which
uses a heterojunction formed between a highly doped thin a-Si:H emitter
layer an the crystalline silicon wafer. In this case the task of the window
layer is twofold: enhancement of the lateral conductivity and reduction of the
reflection losses. The latter limits the thickness to 80 nm only.4 These
examples demonstrate the various aspects in the application: high lateral
conductivity, formation of heterojunctions (ZnO/CdS, ZnO/a-Si:H) and the
generation of textures for enhanced light trapping by chemical etching.
2.
PREPARATION AND STRUCTURE
The research on polycrystalline ZnO layers so far has been performed on

an empirical basis mainly. ZnO films on glass substrates have been prepared
by a variety of techniques such as chemical vapor deposition (CVD),
magnetron sputtering (MS), spray pyrolysis, metal organic chemical vapor
deposition (MOCVD), or pulsed laser ablation.1 Of particular importance has
been magnetron sputtering due to a number of advantages: low process
temperatures, high deposition rate, reasonable controllability of the process,
potentially low cost and scalability to large areas. Highly conducting n-type
films can be made either in argon atmosphere from compound targets
(ZnO/Al2O3) or by sputtering reactively in argon/oxygen atmosphere from
ZnO window layers for solar cells
199
metallic targets (Zn/Al). The dependence of the physical properties of the

films on the most important deposition parameters substrate temperature,
pressure and oxygen flow rate has been studied intensively.5,6 The growth
can qualitatively be described in a modified structure zone model as
developed by Thornton.7 This model predicts that with increasing substrate
temperature and decreasing pressure there is a transition from a porous
structure (tapered crystallites with voids) to a densely packed film (columnar
structure). A qualitative explanation of this trend may be that the reduction of
pressure results in an increase of the mean free path to values close to the
substrate/target separation and thus to a higher energy of the species
impinging on the growing surface.
22
10
9
21
6
20
19
10
18
10
II
10
P H (cm / Vs)
-3
n (cm )
10
10
15
20
25
30
III
f(O2) (sccm)
Figure 1. Electron concentration n and electron mobility measured by Hall effect of ZnO
films prepared by reactive magnetron sputtering as a function of the oxygen flow rate f(O2).
SEM micrographs taken on samples in the three regions of low, medium and high f(O2).8,9
In a reactive sputtering process the oxygen flow rate f(O2) is the most
relevant parameter. Fig.1 displays a typical example of the influence of f(O2)
on physical properties and structure. Hall effect measurements show that the
free carrier concentration n decreases continuously with f(O2) whereas the
electron mobility attains a maximum at medium values of f(O2). This
variation of the n and clearly reflects the change from metallic behavior at
low f(O2) (region I) to oxide formation (region III) at high f(O2) which is
related with an increase of the optical transmission T. These changes are
accompanied by structural variations in the ZnO layers. The SEM
200
Walther Fuhs
micrographs taken on samples in the three regions show granular

morphology with high surface roughness in region I and a flaky morphology
in region III. Around f(O2) = 15 sccm (range II) where the mobility attains a
maximum value the films are compact with columnar structure which has
preferential orientation of the c-axis perpendicular to the glass substrate.9
Independent of the deposition method used ZnO films contain a high
amount of hydrogen. Fig. 2 displays photoelectron spectra (XPS-O1sspectra) that have been obtained on layers prepared by magnetron sputtering
in the UHV.10 The structure at EB = 533 eV in the O1s-spectrum of the asdeposited layer has been assigned to Zn(OH)2. Annealing at 400 C in O2
atmosphere results in hydrogen effusion and a reduction of the hydrogen
content to values below of the XPS detection limit. Fig. 2b shows the
hydrogen profiles of the same samples determined with the nuclear reaction
technique (15N-profiling). The hydrogen concentration in the bulk of the
layers is as high as 1021cm-3 and decreases upon annealing by about two
orders of magnitude. The presence of hydrogen in such high amounts in ZnO
window layers has been ignored so far. At present it is not clear whether the
hydrogen is homogeneously distributed in the bulk of the films or whether
the hydrogen terminates grain boundaries. It is also an open question whether
hydrogen is directly or indirectly responsible for n-type conduction of
nominally undoped films. In single crystalline ZnO hydrogen has been
shown to form a flat donor state.11 It appears also reasonable that hydrogen in
such high concentrations affects the defect structure considerably and forms
complexes with intrinsic defects and/or impurities.
3.
TRANSPORT PROPERTIES
The transport properties of polycrystalline ZnO films have mainly been

studied on an empirical basis focused by the requirement of high
conductivity and high transparency for conducting window layers. ZnO films
in general exhibit n-type conduction with values for the resistivity strongly
dependent on the deposition conditions and/or doping level.6,12 The influence
of the preparation conditions is rather confusing. The scatter in the data from
various laboratories often prevents general conclusions. An example is the
influence of the substrate temperature. If one compares data from various
laboratories no trend can be observed. However, the resistivity clearly
decreases with increasing substrate temperature within one preparational set
from a single laboratory. Among the preparation methods magnetron
sputtering appears to be superior. Besides of practical and cost
201
Intensity (a.u.)
(a)
2
ZnO
Zn(OH) x
1
as grown
536
534
532
530
528
Binding energy (eV)
-3
C H (cm )
10
22
10
21
10
20
10
19
(b)
50
100
150
200
250
depth (nm )
Figure 2. (a) XPS-Photoelectron spectra of ZnO deposited by magnetron sputtering taken with
an excitation energy of Eex = 1486 eV (O1s-spectra): (1) as deposited, (2) after annealing at
400 C; (b) Depth profile of the hydrogen concentration CH determined by nuclear reaction
technique (15N). Same ZnO samples as in (a).10
arguments it has been emphasized that the physical advantage lies in the
incorporation of particle energy from the plasma into the growing film. Two
directions have been pursued to obtain the low resistivity values at room
temperature:
202
Walther Fuhs
1. Creation and control of intrinsic donor levels by defects such as oxygen

vacancies or Zn atoms on interstial sites. Undoped ZnO films with
resistivities of less than 5u10-4 :cm have been prepared by magnetron
sputtering without adding impurities but were found to be unstable at T >
150 C.13
2. Impurity doping with group III elements (Al, Ga, In, B) or with group IV
elements (Si,Ge, Ti; Zr, or Hf) incorporated substitutionally on Zn-sites.
Alternatively also group VII elements have been used such as F
incorporated on an O-site. The lowest values achieved today range at 1.5
- 2u10-4 :cm rather independent of the method of preparation.
Fig. 3 summarizes literature data on the resistivity U and Hall mobility P
from various laboratories with different techniques for preparation and
doping.6 U and P are plotted as a function of the carrier concentration n. The
figure includes results for undoped and doped ZnO single crystals for
comparison. As one expects, U decreases monotonously with n. The lowest
values reached are 1.5-2u10-4 :cm at about 1.5u1021cm-3. Most interestingly
these values are reached almost independently of the kind of dopant used. It
is obvious that there is an intrinsic limitation to the dopability which is given
by the solubility of the dopants which for all elements used is in the order of
a few at%.
The scatter in the data for P is more pronounced than for U and probably
due to the influence of structural defects such as grain boundaries on the
scattering mechanism. At n < 1019 cm-3 scattering at structural defects
appears to be dominant. The results obtained without impurity doping where
P increases with n fall in this range. It is interesting to notice that at n > 1019
cm-3 the mobility values of single crystals and polycrystalline ZnO films are
of very similar magnitude and both decrease with increasing carrier
concentration. This supports the view that the mobility in highly doped
material is not limited by the crystalline structure. Scattering at ionized
impurities has been proposed as the dominant scattering process in this
doping range. The dashed curve shows the result of the theoretical estimate
for impurity scattering basing on parabolic bands.14 Clearly, the mobility
calculated from this approach is considerably too high. Various corrections
to this theory have been proposed such as additional contribution of grain
boundary scattering, non-parabolic bands, multiply charged scattering
centers, or cluster formation at higher doping levels. In most applications
thick ZnO layers (typically 1 m) are used for window layers. However,
there are also situations where thin ZnO layers are required in order to meet
two demands: high lateral conductivity and low reflectivity. The latter
requires very thin ZnO layers (about 80 nm). It has been reported quite
generally that the physical properties of ZnO depend on the layer thickness.
203
U :
:cm
cm
101
10-1
10-3
(a)
(cm/Vs)
10-5
100
10
1015
(b)
1017
1019
n (cm-3)
1021
Figure 3. Resistivity U (a) and Hall mobility (b) as a function of the carrier
concentration n.6 Data from various laboratories: ZnO single crystals, (, ) magnetron
sputtering impurity doped, # magnetron sputtering without impurity doping. The dashed line
is the result of a theoretical estimate for impurity scattering.14
Using glass substrates it has been observed that the resistivity increases
appreciably with decreasing layer thickness at values below of 100-300 nm
mainly due to a decrease of the electron mobility. A possible reason for such
behavior is the different growth and morphology in the early stages of film
growth.15
204
Walther Fuhs
4.
ZnO/SEMICONDUCTOR INTERFACE
In many applications not only the bulk properties of the ZnO layers are
important but also the morphology and electronic structure of the
ZnO/semiconductor interface which may have influence on the band profiles
of the solar cells, losses by contact resistance and instability.
t D = 500 s
Intensity (a.u.)
90s
SiO x
30s
Si(111):H
N an
104
102
100
98
Binding energy (eV)

Figure 4. Photoelectron spectra (Si 2p core levels) of magnetron sputtered ZnO layers after
different deposition times tD (different thickness).10
For PV-applications the stability of the cell efficiency is a very critical

parameter. Therefore CIS solar cells have to pass a test, a so called dampheat-test, where the modules are held in an atmosphere with a humidity of
85 % at a temperature of 85 C. For non-encapsulated Cu(In,Ga)Se2 cells
under such conditions in particular the open circuit voltage and the fill factor
degrade as a function of time causing severe performance loss.16 The
physical background of this effect is quite unclear. The observed small
increase of the sheet resistance of the ZnO window apparently is not
sufficient to explain this effect. Some investigations suggest that the
degradation is strongly related with the use of CdS as bufferlayer which may
react with ZnO and thus cause an increase of the contact resistance of the
window.16
205
The ZnO/Si heterointerface has been of particular interest in

photovoltaics. The preparation of interfaces with high electronic quality is
complicated by the problem that oxidation of the surface of the Si substrate
may be induced by the deposition process. The photoelectron spectra (Fig. 4)
show that indeed oxide formation occurs in the early stages of film growth
when the films are deposited by magnetron sputtering.10 The finger print of
SiOx are the bands at a binding energy of about 103 eV. For the Hterminated Si(111) surface no structure is found in this spectral range.
Clearly the suboxide band develops at small deposition times.
Concomitantly the Si-2p bands decrease due to the decreasing sensitivity to
the surface of the Si wafer (1-2 nm). This is also the reason why no structure
from neither the Si-wafer nor the oxide surface layer can be observed when
after longer deposition time (tD = 500 s) a thicker ZnO layer has grown.
Besides of the oxidation of the Si substrate reduction of Zn is observed.
Surface sensitive photoelectron spectroscopy clearly reveals a neutral Znphase as well as Zn2SiO4 phases at the interface.
nc-Si
a-SiOx
c-Si
Figure 5. Cross sectional TEM micrographs of the ZnO/c-Si interface.17
The complex structure of the ZnO/Si interface has also directly been
observed in cross-sectional TEM micrographs. An example is given in
Fig. 5. This micrograph clearly reveals the formation of a thin amorphous
film on the Si-surface (suboxide phase), the formation of nano-crystalline
phases followed by columnar ZnO growth.17
The above results showed that the ZnO/Si interface prepared by
magnetron sputtering is of high complexity and difficult to control. This
explains why all attempts to develop photovoltaic devices with reasonable
performance on basis of sputtered ZnO/Si heterojunctions have failed. This
may be a particular problem of the deposition method used. Recently it was
206
Walther Fuhs
shown that the formation of silicon suboxides at the interface does not occur
when the ZnO film is deposited by metal organic chemical vapor deposition
(MOCVD) from diethylzinc/H2O vapor on a heated substrate.10 The different
result obtained from these two methods may be explained by the fact that in
case of MOCVD only thermal energy is involved while magnetron
sputtering is characterized by additional energy input from the plasma into
the growing film. The quality of the interface may therefore strongly depend
on the energy of the species, which are responsible for film growth, and on
the ion-to-neutral ratio in the plasma.
5.
LIGHT TRAPPING
Silicon thin-film solar cells on basis of hydrogenated amorphous silicon,

a-Si:H, and microcrystalline silicon, c-Si:H, are one of the most promising
technology options for cost effective photovoltaic modules. A common
device structure for such cells uses a superstrate configuration (glass/ZnO/pi-n/Al) where the light penetrates in through the glass. Effective light
trapping is essential to achieve high cell efficiencies with absorber layers the
thickness of which is in the range of one m only. This can be achieved by
an optimized surface texture of the window layer which results in light
scattering and a back reflector for the transmitted part of the light which
often consists of a ZnO/Al double layer.3 In the superstrate configuration
used, ZnO is required as window layer forming the front contact and has to
combine high conductivity (low series resistance) with high transparency
and excellent surface texture (milky appearance). The required surface
texture can be realized with sputtered ZnO:Al films by a combination of a
proper choice of the process parameters of the sputter process and postdeposition wet chemical etching in diluted hydrochloric acid HCl.7 With
increasing substrate temperature and decreasing pressure there is a transition
from a porous structure (tapered crystallites with voids) to a densely packed
film (columnar structure). In the same direction the etching rate decreases.
Films which, are prepared at low substrate temperature and high deposition
pressure, have a porous rough surface structure and are homogeneously
etched without surface texturing. For optimized surface texture resulting in
optimized light scattering over a wide wavelength range the starting
structure has to be densely packed with a smooth surface. Fig. 6 shows an
example for an optimized surface texture and the resulting optical properties.
The layer is highly transparent and the reflection amounts to less than 10 %
in the entire visible spectral range.3 Fig. 7 compares the external quantum
efficiencies of c-Si:H solar cells prepared as glass/ZnO/pin-structures on
207
smooth (as deposited) and textured ZnO (short HCl-dip, optimized) window
layers. The cells had been prepared in the same run with an absorber layer
(a)
1.0
(b)
T
T, R
0.8
0.6
textured ZnO film:

Rsheet = 7.2 :
0.4
0.2
0.0
R
400
600
800
1000
wavelength (nm)
Figure 6. Optimized surface texture of ZnO-layer (a) and optical properties: transmission T
and reflectivity R (b).3
208
Walther Fuhs
1.0
1.0
0.9 1 m -Si
0.9
optimised
0.8
0.8
short dip
smooth
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
300 400 500 600 700 800 900 1000 1100
wavelength (nm)
absorbance
external quantum efficiency
thickness of 1.1 m. Texturization enhances the external quantum efficiency

in the entire spectral range. On the short wavelength side this is due to a
reduction of the reflection of the cells. The enhancement at long wavelengths
is mainly due to an enhancement of light trapping. From a comparison of the
Figure 7. External quantum efficiency of c-Si:H solar cells, glass/ZnO(textured)/pin/ZnOAl,

co-deposited on smooth, shortly etched and optimized ZnO window layer. For comparison the
absorbance of a 1 m thick c-Si:H layer is shown.3
absorbance of a 1 m thick c-Si:H layer with the quantum efficiency it has

been concluded that the effective light path through the i-layer must be
larger than 10 m.3 Although at present the a-Si:H based solar modules quite
commonly use tin oxide (SnO2:F) coated glass substrates, doped ZnO
prepared by sputtering and post deposition etching presents a promising
alternative for the future.
6.
CONCLUSION
Polycrystalline ZnO for window layers in solar cells are most frequently
prepared by magnetron sputtering. Optimized layers have resistivities of
1.5-2u10-4 cm with Hall mobilities of up to 60 cm2/Vs, refractive indices
of approximately 2.0 and an average transmittance of 85 % in the visible
range. The best values of the mobility in optimized films have reached the
209
intrinsic limit set by ionized impurity scattering. Optimized films grow with
columnar structure with the c-axis oriented perpendicular to the substrate.
Depending on the type of solar cell various aspects have to be considered:
high lateral conductivity, formation of heterojunctions (ZnO/CdS, ZnO/aSi:H) and the generation of textures for enhanced light trapping by chemical
etching.
ACKNOWLEDGMENT
The author gratefully to C. Pettenkofer for allowing the use of so far
unpublished results (Figs. 2 and 4) and to Ina Sieber and Frank Fenske for
fruitful discussion.
REFERENCES:
1. D. S. Ginley and C. Bright (guest editors), MRS Bulletin, 15 (2000).
2. J. Hedstrm, H. Ohlsen, M. Bodegard, A. Kylner, L. Stolt, D. Hariskos, M. Ruckh, and H.
W. Schock, Proc. 23rdd IEEE PVSC
C (1993).
3. B. Rech, J. Mller, T. Repmann, O. Kluth, T. Roschek, J. Hpkes, H. Stiebig, and W.
Appenzeller, Mat. Res. Soc. Symp. Proc. 762, A 3.1. (2003).
4. M. Schmidt, A. Schpke, O. Milch, T. Lussky, and W. Fuhs, Mat. Res. Soc. Symp. Proc.
762, A19.11.1. (2003).
5. K. Ellmer, J. Phys. D: Appl. Phys. 33, R1 (2000).
6. K. Ellmer, J. Phys. D: Appl. Phys. 34, 3097 (2001).
7. O. Kluth, G. Schpe, J. Hpkes, C. Agashe, J. Mller, and B. Rech, Thin Solid Films 442,
80 (2003).
8. S. Brehme, F. Fenske, W. Fuhs, E. Neubauer, M. Poschenrieder, B. Selle, and I. Sieber,
Thin Solid Films 342, 167 (1999).
9. F. Fenske, W. Fuhs, E. Neubauer, A. Schpke, B. Selle, and I. Sieber, Thin Solid Films
343, 130 (1999).
10. U. Meyer, Dissertation Universitt Cottbus (2004), U. Meyer and C. Pettenkofer,
unpublished.
11. S. F. J. Cox, E. A. Davis, S. P. Cottrell, P. J. C. King, J. S. Lord, J. M. Gil, H. V. Alberto,
R. C. Vilao, J. Pironto Duarte, N. Ayres de Campos, A. Weidinger, R. L. Lichti, and S. J.
C. Irvine, Phys. Rev. Lett. 86, 2601 (2001).
12. T. Minami, MRS Bulletin, 38 (2000).
13. T. Minami, H. Nanto, and S. Takata, Appl.Phys.Lett. 41, 958 (1982).
14. J. R. Bellingham, W. A. Phillips, and C. J. Adkins, J. Mater. Sci. Lett. 11, 263 (1992).
15. M. Birkholz, B. Selle, F. Fenske, and W. Fuhs, Phys. Rev. B 68, 205414 (2003).
16. J. Wennerberg, J. Kessler, and L. Stolt, Solar Energy Materials and Solar Cells 75, 47
(2003).
17. I. Sieber, N. Wanderka, I. Urban, I. Drfel, E. Schierhorn, F. Fenske, and W. Fuhs, Thin
Films 330, 108 (1998).
Chapter 18
ZnO/AlGaN ULTRAVIOLET LIGHT EMITTING
DIODES
E. V. Kalinina1, A. E. Cherenkov1, G. A. Onushkin1, Ya. I. Alivov2, D. C.
Look3, B. M. Ataev4, A. K. Omaev4, and C. M. Chukichev5
1
A. F. Ioffe Physico-Technical Institute, RAS, 194021 St. Peterburg, Russia, 2Institute of

Microel. Technology, RAS, Chernogolovka, Moscow dist., 142432 Russia, 3Semiconductor
Research Center, Wright State University, Dayton, Ohio 45435, U.S.A., 4Institute of Physics,
Daghestan Scientific Centre of RAS, Makhachkala, 367003 Russia, 5Department of Physics,
Moscow State University, 119899 Moscow, Russia
Abstract:
We report on the optical and electrical properties of n-ZnO/p

/ -AlGaN
heterojunctions. Ga doped n-type ZnO layers were grown using chemical
vapor deposition on Mg doped p-type AlGaN epitaxial layers. AlGaN epitaxial
layers with 12 at.% Al were grown on 6H-SiC by hydride vapor phase epitaxy.
Rectifying diode-like behavior with a threshold voltage of 3.2 V was achieved.
Intense ultraviolet electroluminescence peaking at a wavelength of 390 nm was
observed at 300 and 500 K as a result of hole-injection from the n-ZnO layer
into the p- AlGaN layer of the heterostructure.
Key words:
Electroluminescence, epitaxial layers, heterojunction, hole injection, ultraviolet

emission
1.
INTRODUCTION
Zinc oxide, with a direct bandgap of a3.3 eV is a promising

semiconductor for the fabrication of ultraviolet light emitting diodes and
lasers.1-3 ZnO has some properties that are perhaps more favorable than those
of other wide bandgap semiconductors such as GaN and SiC. These include
intense exciton luminescence with a large exciton binding energy of a60
meV that provides effective exciton luminescence at high temperatures,
higher quantum efficiency of the photo- and cathodoluminescence (PL, CL)
as well as the relative simplicity of the growth of ZnO epitaxial layers.
However, despite much progress in ZnO technology in recent years, high
211
212
E. V. Kalinina et al.
quality p-type ZnO and therefore, the manufacture of ZnO homojunctions,

continues to be problematic. The alternative means for attaining functional
device structures with ZnO is the fabrication of ZnO based heterojunctions
with other p-type semiconductors.4-7
The AlGaN alloys are suited to this application since they can be doped
p-type, have a larger bandgap, and are relatively well lattice-matched to
ZnO.8 In this work, we present the results on optical and electrical properties
of n-ZnO/p
/ -AlGaN heterojunctions.
2.
EXPERIMENTAL PROCEDURE
Commercial n-type 6H-SiC wafers with a concentration of

uncompensated donors Nd - Na = 3u1018 cm -3 were used as substrates. These
substrates were covered with a thin epitaxial n-GaN buffer layer grown by
hybrid vapor phase epitaxy (HVPE). This buffer layer had a thickness of 0.2
Pm and a donor concentration of Nd-Na = (2-3)u1017cm-3. On top, Mg doped
p-type AlGaN epitaxial layers with thickness of 0.8 Pm and an Al content of
12 at.% were also grown by HVPE with an acceptor concentration of Na-Nd
= (5-8)u1017 cm-3. Then, Ga doped n-ZnO epitaxial layers were deposited
using chemical vapor deposition (CVD) with a thickness of 0.8 Pm and a
donor concentration of Nd - Na = 7u1017 cm-3. The growth process was
stimulated by a discharge plasma, that allowed to reduce the substrate
temperature by up to 400 C and thus improve the structural quality of the
ZnO layers.9
The heterojunction devices were fabricated by masking the surface, then
using a 10 HNO3 aqueous solution to etch down through 2u2 mm2
windows to remove ZnO and leave regions of the AlGaN layer exposed.
Ohmic contacts to the n-ZnO and p-Al0.12Ga0.88N were made by thermal
vacuum evaporation of Al and Ni, respectively.8
Al compositions in the AlGaN epilayers and the structural quality of
epitaxial layers were estimated using x-ray diffraction. The optical properties
of the ZnO and AlGaN films were explored with cathodoluminescence (CL)
spectroscopy at 80 and 300 K using accelerating voltage of 20 kV and an
electron-beam current of 0.1-1 PA. Electroluminescense (EL) measurements
were performed under forward DC-biased conditions at 300 and 500 K. The
concentration of uncompensated acceptors and donors in epitaxial layers
were determined from capacitance-voltage (C-V) characteristics of Cr
Schottky barriers and Hall measurements, respectively. Forward and reverse
current-voltage characteristics (I-V) were measured under DC conditions.
ZnO / AlGaN ultraviolet light emitting diodes
3.
213
According to x-ray studies, the Al content in the p-AlGaN epitaxial layers

was estimated to about 12 at.% and the lattice mismatch to the ZnO layer
amounted to 2.2 %. At the same time, for both epitaxial layers x-ray
measurements revealed a good crystallinity. This is corroborated by CL
measurements.10
The CL spectrum of the ZnO film consisted of intense, near-band-edge
ultraviolet emission with a wavelength maximum at 387 nm and a full width
at half maximum of 21 nm. This emission is of excitonic nature and is a
result of the radiative annihilation of free and bound excitons. A broad
defect-related green band with much lower intensity near 510 nm, typical for
ZnO, was also observed (Fig. 1). The CL spectrum of the pAl0.12Ga0.88N(Mg) film consisted of a very weak near-band-edge emission
with Omax at ~ 356 nm, and of a more intense broad band with a maximum at
~410 nm.
The current-voltage characteristics of the n-ZnO/p
/ -AlGaN heterostructure
were typical for rectified diodes: the direct current amounted to 60 mA at a
voltage of 4 V and a current of about 10-7 A was observed under reverse
voltage of about 30 V (Fig. 2). The cut-off voltage was equal to 3.2 V that is
very close to the band-gap energy of ZnO suggesting that ZnO is the active
region of the device. The other reason for this supposition we can see from
the Anderson model of the energy band-diagram forr n-ZnO/p
/ -AlGaN
8
heterostructures. In this diagram, the electron affinity and
25
20
10
Intensity, a.u.
15
x7
ZnO
AlGaN
0
350
400
450
500
550
Wavelength, nm
Figure 1. Cathodoluminescence spectra of the ZnO and Al0.12Ga0.88N layers measured at room
temperature. Imax (ZnO) = 387 nm, FWHM (ZnO) = 21 nm.
214
60
Current, mA
50
40
30
20
10
0
-1
-10
10
Bias Voltage, V
-20
Figure 2. Current - voltage (I - V) characteristics of the n-ZnO / p-Al0.12Ga0.88N structures

measured at room temperature: Uc = 4 V, Ic = 60 mA, Urev = 30 V, Irev d 10-7 A.
10
EL intensity, a.u.
12
300 K
500 K
0
355
365
375
385
395
405
415
Wavelength, nm
425
435
445
Figure 3. Electroluminescence spectra of the n-ZnO/p-Al0.12Ga0.88N heterostructure of a lightemitting diode at temperatures of 300 and 500 K (Ic = 20 mA).
ZnO / AlGaN ultraviolet light emitting diodes
215
Under forward bias of 20 mA the device produced intense ultraviolet

electroluminescence with a peak emission near 389 nm (3.19 eV) and a full
bandgap energy values for ZnO were taken as 4.35 and 3.3 eV, respectively.
Taking into account, that the electron affinity and bandgap energy values for
AlGaN are assumed to be linearly dependent on the Al content, for our
composition with 12 at.% Al these values equal to 3.68 and 3.72 eV,
respectively. As a result, the energetic barrier for electrons is 0.67 eV, while
the energetic barrier for holes is 0.25 eV. Thus, the energetic barrier for
holes is about three times less than the barrier for electrons. Hence, the
predominant device current is hole injection from the p-AlGaN into the ntype ZnO active region. width at half maximum of 26 nm (Fig. 3).
Significantly, no other emission bands were observed in the EL spectrum.
Also, significant EL emission at temperatures up to 500 K was observed due
to a very large exciton binding energy of ZnO of about 60 meV.
4.
CONCLUSION
Bright ultraviolet light emitting diodes based on n-ZnO/p

/ -Al012Ga0.88N
heterojunctions were formed for the first time. EL emission was obtained
from these heterostructures with a maximum at 389 nm. CL spectra,
obtained from the ZnO layer, consisted of an intense emission with a
maximum at 387 nm. Hence, we suggest that the dominant EL emission was
observed from the ZnO layer. The dominant device current was a hole
injection from p-Al012Ga0.88N into the n-type ZnO active region of the
heterojunction. The lattice mismatch between ZnO and p-Al012Ga0.88N was
estimated to be 2.2 %.
REFERENCES
1. D. C. Look, Mater. Sci. Eng. B 80, 383 (2001).
2. D. M. Bagnall, Y. F. Chen, Z. Zhu, T. Yao, M. Y. Shen, and T. Gata, H, Appl. Phys. Lett.
73, 1038 (1998).
3. V. I. Zinenko and Yu. A. Agafonov, Solid State Electronics, 48, 2343 (2004).
4. A. E. Tsurkan, N. D. Fedotova, L. V. Kicherman, and P. G. Pasko, Semiconductors 6,
1183 (1975).
5. I. T. Drapak, Semiconductors 2, 624 (1968).
6. Ya. I. Alivov, J. E. Van Nostrand, D. C. Look, M. V. Chukichev, and B. M. Ataev, Appl.
Phys. Letters 83, 1 (2003).
7. B. M. Ataev, Ya. I. Alivov, E. Kalinina, V. V. Mamedov, G. A. Onushkin, S. Sh.
Makhmudov, and A. K. Omaev, ICPS2004, July 26-30, 2004, Flagstaff, Arizona, USA.
216
8. Ya. I. Alivov, E. V. Kalinina, A. E. Cherenkov, D. C. Look, B. M. Ataev, A. K. Omaev,

M. V. Chukichev, and D. M. Bagnall, Appl. Phys. Lett. 83, 1 (2003).
9. B. M. Ataev, Ya. I. Alivov, V. V. Mamedov, S. Sh. Makhmudov, and B. A. Magomedov,
Semiconductors 38, 699 (2004).
10. D. M. Bagnall, Ya. I. Alivov, E. V. Kalinina, D. C. Look, B. M. Ataev, M. V. Chukichev,
A. E. Cherenkov, A. K. Omaev, Mat. Res. Soc. Symp. Proc. Vol. 798, Y3.9.1 (2004).
Chapter 19
ZnO TRANSPARENT THIN-FILM TRANSISTOR
DEVICE PHYSICS
J. F. Wager
School of EECS, Oregon State University, Corvallis, Oregon, 97331, U.S.A.
Abstract:
ZnO transparent thin-film transistors (TTFTs) are a recent and important

development in the emerging field of transparent electronics. The objective of
this contribution is to elucidate TTFT device physics issues related to channel
mobility and output conductance.
Key words:
Transparent thin-film transistor, device physics, effective mobility, field-effect

mobility, saturation mobility, average mobility, incremental mobility, output
conductance
1.
INTRODUCTION
ZnO transparent thin-film transistors (TTFTs) are a very recent

development.1-8 The focus of this article is elucidation of channel mobility
and output conductance. Although these device physics issues are important
for the development of TTFTs, they are also of relevance for other types of
inorganic or organic TFTs.
2.
DRAIN CURRENT-DRAIN VOLTAGE

MODELING
The simplest approach for specifying the drain current-drain voltage (IDSVDS) electrical characteristics of a TTFT involves use of the square-law
model,9 as expressed by
I DS
V2
Z
Ci VGS VT VDSS DDS
L
2
217

(1)
218
J. F. Wager
in the pre-pinch-off regime, defined by VDS < VDSAT, and by

I DS
Z
Ci
2L
(2)
in the post-pinch-off regime, defined by VDS > VDSAT. In these equations, Z =

gate width, L = gate length, Ci = gate capacitance per unit area, P = channel
mobility, and VT = threshold voltage. Also, VDSAT is the drain voltage at
which the channel pinches off, which is specified as VDSAT = VGS VT,
where VGS is the gate voltage referenced to the source. Figure 1 illustrates a
typical IDS-VDS curve for a specific VGS.
IDS
pre-pinch-off
Z
Ci
L
IDS
VDS
VDS
2
x
post-pinch-off
Z
Ci
2L
IDS
2
VDS
VDSAT = VGS-VT
Figure 1. Drain current-drain voltage (IDS-VDS) curve at a given gate voltage (VGS),
illustrating the pre-pinch-off and post-pinch-off regimes of operation, which are distinguished
by VDSAT = VGS - VT.
It is clear from these square-law equations that mobility is an important

physical property of the channel material since its magnitude directly affects
the current-drive capability of the device. Additionally, mobility is directly
related to the device frequency response, as given by the unity-gain
frequency,
ft
PVDS
2S 2
GS
(3)
pre pinch off
2 L
post pinch off
ZnO transparent thin-film transistor device physics
219
Given the importance of mobility in establishing the current gain and

frequency-response characteristics of a TTFT, it is worth considering
channel mobility assessment.
3.
CHANNEL MOBLITY ASSESSMENT
There are three common mobility assessment methods,9 giving rise to (i)
effective mobility, Pefff, (ii) field-effect mobility, PFE, and (iii) saturation
mobility, Psat. Each type of mobility is now individually considered.
3.1
Effective Mobility
Effective mobility, Pefff, is obtained by considering the linear portion of

the pre-pinch-off regime, defined by VDS << VGS VT, in which case,
I DS
Z
Ci
L
VGS
VT VDDS ,
(4)
evaluating the output conductance for this situation,

gd
wII DS
D
wVDS V
GS
G
Z
Ci VGS VT
L
wP
VDDS ,
VDS
D
(5)
and solving for the mobility arising from this procedure, while neglecting the
term involving the change in mobility with respect to drain voltage yields,
Peff
gd
Z
Ci
L
(6)
VDDS o 0
Neglect of the term involving the change in mobility with respect to drain
voltage is appropriate for long-channel TTFTs for which velocity saturation
is not a relevant consideration.
Of the three common mobilities used for TTFT assessment, the effective
mobility is probably the best channel mobility figure-of merit since (i) it is
evaluated in the linear regime of operation where the channel is most
uniform and the concept of mobility as a proportionality constant between
the electric field and carrier velocity is most appropriate, and (ii) the term
220
J. F. Wager
neglected in deriving Pefff is indeed negligible for the long-channel devices

currently of interest in the development of TTFTs. However, there are
several disadvantages of the effective mobility, (i) VT is required, (ii) Pefff is
singular when VGS = VT, and (iii) Pefff estimation is sensitive to the magnitude
of the source-drain resistance.
3.2
Field-effect Mobility
In a similar manner as to how Pefff is assessed, the field-effect mobility,

PFE, is also evaluated in the linear regime. However, PFE evaluation involves
use of the transconductance rather than the output conductance,
gm
wII DS
wVGS V
DS
D
(
Ci VDDS 1
L
GS
eff
) wP
.
wVGS
(7)
If the term involving the change in mobility with respect to a change in the
gate voltage is neglected, the defining equation for the field-effect emerges,
P FE
gm
Z
CiVDDS
L
VDDS
.
o
(8)
The most serious liability of PFE is the neglect in the derivation of PFE of
the term involving the change in mobility with respect to a change in the
gate voltage. This approximation is clearly not accurate at large gate
voltages where the accumulation channel thickness decreases and hence
electrons drifting in the channel experience more intense scattering at the
interface, resulting in a reduction of the channel mobility. Another PFE
demerit is that its estimation is sensitive to the magnitude of the source-drain
resistance, as is Peff. On the other hand, the use of PFE offers several
advantages, (i) it is evaluated in the linear regime of operation where the
channel is most uniform and the concept of mobility as a proportionality
constant between the electric field and carrier velocity is most appropriate,
(ii) VT is not required, and (iii) PFE is not singular when VGS = VT. The fieldeffect mobility is probably the most popular method used for channel
mobility estimation.
3.3
221
Saturation Mobility
Unlike Pefff and PFE, the saturation mobility, Psat, is evaluated from an
assessment of IDS in the post-pinch-off or saturation regime by taking the
square root of the post pinch-off expression for IDSAT, differentiating with
respect to VGS, and then solving the resulting equation for mobility to obtain,
P sat
w I DDSAT
wVGGS
Z
Ci
2L
,
VDS
GS
(9)
where the condition VDS > (VGS VT) ensures that this procedure is
accomplished in the post-pinch-off or saturation regime of the IDS-VDS curve.
Note that the numerator term is just the square of the slope of the linear
portion of a (IDSAT)1/2 VGS plot.
Of the three most common channel mobilities, Psat is probably used the
least. The most serious liability of Psatt is that it is evaluated in the postpinch-off regime of operation where the channel thickness and electric field
profile are highly non-uniform, the gradual channel approximation is not
valid, and the concept of mobility as a proportionality constant between the
electric field and carrier velocity is of questionable viability. However, Psat
offers several advantages, (i) VT is not required, and (ii) PFE is not singular
when VGS = VT, and (iii) its estimation is relatively insensitive to the
magnitude of the source-drain resistance, unlike Pefff and PFE.
3.4
A New Approach to Mobility Assessment
Recently Hoffman8 has proposed a new approach for channel mobility

assessment that is novel in its interpretation of mobility and yet is closely
related to conventional mobility assessment methodologies. This approach
involve defining two new mobilities, an average mobility, Pavg,
P avg (
GS
gd (
Z
Ci (
L
GS
GS
G
,
ON
VDDS o 0
(10)
222
J. F. Wager
and an incremental mobility, Pinc,
Pinc (
GS
wg
g d ( GGS )
wVGGS
Z
Ci
L
V
DS
D
(11)
o0
Note that both of these mobilities are written as functions of VGS,

explicitly indicating that these mobilities are expected to vary with gate
voltage. A comparison of the equations defining Pefff and Pavg indicates that
these mobilities are identical, except for the use of VT for Pefff, and VON for
Pavg. VON is defined as the voltage at which IDS begins to noticeably increase
when assessed using a Log(IDS)-VGS curve.8 Use of VON circumvents
problems associated with the VGS = VT singularity inherent in Peff. In a
similar fashion, a comparison of the equations defining PFE and Pinc indicates
that these mobilities are quite similar, except that the transconductance,
which is used in the defining relation for Pefff, is replaced by the derivative of
the channel conductance with respect to the gate voltage, i.e., the differential
channel conductance, for the case of Pinc.
The compelling aspect of this new approach to channel mobility
assessment is the physical picture that emerges. Hoffman argues that Pavg
corresponds to the average mobility of alll of the carriers moving in the
channel, whereas Pinc corresponds to the mobility of carriers added
incrementally to the channel with an incremental increase in the gate voltage.
Thus, a Pinc(VGS) provides a physical picture of the channel mobility trend in
terms of injected carrier trapping and mobility degradation, whereas
Pavg(VGS) provides information of relevance to projecting how a device will
perform when used in a circuit application. Thus, both of these mobilities are
useful, informative, and complementary.
4.
OUTPUT RESISTANCE
Figure 2 illustrates a two-port network large-signal equivalent model

appropriate for TTFT modeling. The gate-source is the input port and the
drain-source is the output port.
IGS
CGD
+
VGS C
GS
S
223
IDS(total)
IDS
Rout
+
VDS
-
Figure 2. A large-signal TTFT equivalent circuit model.
The important feature of Fig. 1 is that the output current, IDS(total), can
arise from either or both of the two output branches shown, and identified by
IDS and Rout. In the context of the present discussion, IDS represents the
square-law theory current-voltage characteristics, which can be written in
functional form as IDS(VDS,VGS), where VDS and VGS act as independent
variables. Thus, this IDS branch of the equivalent circuit represents transistor
behavior. In contrast, Routt is a simple resistor that accounts for the
conductivity of the bulk channel layer. If the channel layer is conductive,
Routt is small, most of the output current flows through Rout, and the TTFT
operates as a resistor. In contrast, if the channel is highly resistive, Routt is
large, and most of the output current flows through IDS so that the TTFT
functions as a transistor is supposed to function. Intermediate cases in which
the output current flows through both branches give rise to voltagecontrolled resistor operation. It is our experience that early-stage TTFT
development more often results in resistor or voltage-controlled resistor
behavior than transistor behavior, at least until the fabrication process is in
control.
ACKNOWLEGMENTS
This work was funded by the U.S. National Science Foundation under
Grant No. DMR-0071727 and by the Army Research Office under Contract
No. MURI E-18-667-G3.
224
J. F. Wager
REFERENCES
1. S. Masuda, K. Kitamura, Y. Okumura, S. Miyatake, and T. Kawai, J. Appl. Phys. 93, 1624
(2003).
2. R. L. Hoffman, B. J. Norris, and J. F. Wager, Appl. Phys. Lett. 82, 733 (2003).
t 82, 1117 (2003).
3. P. F. Carcia, R. S. McLean, M. H. Reilly, and G. Nunes, Appl. Phys. Lett.
4. J. Nishii, F. M. Hossain, S. Takagi, T. Aita, K. Saikusa, Y. Ohmaki, I. Ohkubo, S.
Kishimoto, A. Ohtomo, T. Fukumura, F. Matsukura, Y. Ohno, H. Koinuma, H. Ohno, and
M. Kawasaki, Jpn. J. Appl. Phys. 42, L347 (2003).
5. J. F. Wager, Science 300, 1245 (2003).
6. B. J. Norris, J. Anderson, J. F. Wager and D. A. Keszler, J. Phys. D 36, L105 (2003).
7. Y. Kwon, Y. Li, Y. W. Heo, M. Jones, P. H. Holloway, D. P. Norton, and S. Li, Proc.
Electrochem. Soc. 20003, 68 (2003).
8. R. L. Hoffman, J. Appl. Phys 95, 5813 (2003).
9. D. K. Schroder, Semiconductor Material and Device Characterization, 2nd ed., (Wiley,
New York, 1998).
Chapter 20
ZINC OXIDE THIN-FILM TRANSISTORS
E. Fortunato, P. Barquinha, A. Pimentel, A. Gonalves, A. Marques, L.
Pereira, and R. Martins
Materials Science Department, Faculty of Sciences and Technology of New University oof
Lisbon and CEMOP-UNINOVA, Campus da Caparica, 2829-516 Caparica, Portugal
Abstract:
ZnO thin film transistors (ZnO-TFT) have been fabricated by rf magnetron

sputtering at room temperature with a bottom-gate configuration. The ZnOTFT operates in the enhancement mode with a threshold voltage of 21 V, a
field effect mobility of 20 cm2/Vs, a gate voltage swing of 1.24 V/decade and
an on/off ratio of 2105. The ZnO-TFT present an average optical
transmission (including the glass substrate) of 80 % in the visible part of the
spectrum. The combination of transparency, high channel mobility and room
temperature processing makes the ZnO-TFT a very promising low cost
optoelectronic device for the next generation of invisible and flexible
electronics. Moreover, the processing technology used to fabricate this device
is relatively simple and it is compatible with inexpensive plastic/flexible
substrate technology.
Key words:
Thin-film transistors, rf magnetron sputtering, transparent electronics,

optoelectronic devices
1.
THIN FILM TRANSISTORS: (R)EVOLUTION
Even though the field-effect transistor did not come into widespread use
until the 1960s, its invention predated both the junction and point-contact
transistors by many years. As it is normal with many innovations, its
practical realization was delayed until adequate materials and technologies
were available for its fabrication. We can even say that the Thin Film
Transistor (TFT) was the first solid-state amplifier ever patented. The basic
principle of the field-effect transistor (what we call now JFET) was proposed
by the first time by Julius Edgar Lilienfeld as early as 1925 and patented in
1930 Ref. (1) (see Fig. 1a) where an adaptation of the cross section of the
225
226
E. Fortunato et al.
device is shown). The field-effect transistor he describes was probably the

first successful solid-state amplifier invented. The advantages of this device
over the vacuum tubes, the only alternative high frequency amplifier
available at that time, are also described in this patent and in later ones.
Lilienfelds later patent from 1933 (Ref. 2) describe a new field-effect
transistor (now well-known as MISFET) were he specifies the thickness of
the dielectric layer, which insulates the metal control electrode from the
copper sulfide channel, of about 10-5 cm (1000 ). This is, in fact, the typical
thickness used in the insulated-gate transistors, which were developed many
years later. Claim 1 of this patent reads:
An electrical device comprising a conducting solid material having one
surface an insulating layer of a thickness of the order of magnitude of 10-4
mm, and a superposed stratum of solid material integral therewith, of
substantially greater conductivity and insulating from the conducting
solid material by the said insulating layer.
Nevertheless, the first functional working TFT was demonstrated by
Weimer in 1962 (Ref. 2). He used thin films of polycrystalline cadmium
sulfide, similar to those ones developed for photodetectors. The simplified
structure is shown in Fig. 1(b). Other TFT semiconductor materials like
CdSe, Te, InSb and Ge were investigated, but in the mid-1960s the
emergence of the metal oxide semiconductor field effect transistor
(MOSFET) based on the crystalline silicon technology and the possibility to
perform integrated circuits, led to a decline in TFT development activity by
the end of the 1960s.
What dramatically changed the prospects for TFTs in the 1970s was the
realization that with crystalline silicon, low cost was inseparable from
miniaturization, whereas some applications required large arrays of low cost
electronics, like for example displays. By this time many researchers and
engineers had been engaged in improving the characteristics of liquid crystal
displays (LCDs), recently discovered. The first active matrix LCD was
successfully demonstrated by Brody3 in 1973, where a CdSe TFT was used
as a switching element for each pixel of a 120120 matrix.
In spite of the many successful demonstrations of CdSe TFT LCDs, the
industry did not enter in this market, until the report on the feasibility of
doping amorphous silicon (a-Si:H) by the glow discharge technique.4,5 After
that and in 1979, LeComber, Spear and Ghaith6 described a TFT using aSi:H as the active semiconductor material, whose structure is indicated in
Fig. 1(c).
After LeComber reported the first a-Si:H TFT, many laboratories started
the development of AMLCDs formed on glass substrate. Although this result
attracted much attention, the major disadvantage of a-Si:H TFT is its low
Zinc oxide thin-film transistors
227
electron mobility that limits the ultimate speed of devices. However, an

adequate device speed for switching applications in LCDs has been
achieved.
a)
Cu2S
Al
gate
drain
source
control gate
b)
insulator
CdS
source
drain
insulating substrate
a-Si:H
c)
source
drain
silicon nitride
gate
Fig. 1. Schematic cross sectional views of TFTs: a) Adapted from Lilienfeld,1 b) after
Weimer et al.,2 and c) after LeComber et al.6
Since the mid-1980s the silicon-based thin-film transistors become the

most important devices for active matrix liquid crystal displays (AMLCDs),
and have successful dominated the large area LCD product market.7
The year 1990 marked the debut of a new class of TFT, based upon
organic semiconductor active layer material,8 with electron mobilities similar
to that of a-Si:H. These new TFTs are very promising candidates for
integration onto flexible plastic substrates for a future generation of rugged,
lightweight displays than can be rolled up like a map.
More recently a new generation of oxide semiconductors are being
studied and applied as the active material to TFT, in special zinc oxide
(ZnO). This will be described in the next sections.
228
2.
E. Fortunato et al.
ZINC OXIDE TFTs: MOTIVATION
Transparent electronics are nowadays an emerging technology for the

next generation of optoelectronic devices. Oxide semiconductors are very
interesting materials because they combine simultaneously high/low
conductivity with high visual transparency and have been widely used in a
variety of applications (e.g. antistatic coatings, touch display panels, solar
cells, flat panel displays, heaters, defrosters, optical coatings, among others)
for more than a half-century. Transparent oxide semiconductor based
transistors have recently been proposed using as active channel non-doped
ZnO.9-15 One of the main advantages exhibited by these transistors lies on
the magnitude of the electron channel mobility leading to higher drive
currents and faster device operating speeds. The mobility reported in the
literature is ranging from 0.2 to 7 cm2/Vs with an on/off current ratio from
105 to 107 and a threshold voltage (VTH) between -1 and 15 V. To date, ZnO
channel layers have been deposited using substrate heating or submitted to
post thermal annealing in order to increase, mainly the crystallinity of the
ZnO layer and so the films mobility. The main advantage of using ZnO
deals with the fact that it is possible to growth at/near room temperature high
quality polycrystalline zinc oxide, which is a particular advantage for
electronic drivers, where the response speed is of major importance. Besides
that, since ZnO is a wide band gap material (3.2 eV), it is transparent in the
visible region of the spectra and therefore, also less light sensitive.
Table 1. State of the art concerning ZnO TFTs.
Ion Beam
Technique
600-800 C (ref. 11)
Processing
temperature
Mobility (cm2/Vs)
0.3-2.5
on/off ratio
106
VTH (V)
10-20
TFT operation mode enhancement
Isat (PA)
70
Sol Gel
700 C (ref. 15)
Pulsed Laser
450 C (ref. 10)
0.2
107
not mentioned
enhancement
14
0.031-0.97
106
-1-2.5
enhancement
80
Thin film transistors based on a-Si:H technology present some limitations

such as light sensitivity and light degradation accompanied with a low
mobility (< 2 cm2/Vs). On the other hand, the polysilicon TFT in spite
exhibiting a high mobility (50 cm2/Vs d d 500 cm2/Vs) their opacity limits
the aperture ratio for active matrix arrays, highly important when for
instance, organic light emitting diodes (OLEDS) have to be addressed.
Apart from that, if flexible substrates based on polymers are intended to be
used, the processing temperature is also a quite limiting factor. One possible
229
way to overcome such problems is the utilization of efficient and reliable

oxide based thin film transistors.
Table 1 presents a comparison of fully transparent ZnO-TFTs (data as of
March 2004).
3.
EXPERIMENTAL DETAILS
3.1
ZnO optimization
The ZnO films (doped and undoped) were deposited onto soda lime glass
substrates by rf (13.56 MHz) magnetron sputtering using a ceramic oxide
ZnO target from Super Conductor Materials, Inc. with a purity of 99.99%
and 2 diameter. The sputtering was carried out at room temperature and the
argon deposition pressure was 0.15 Pa. The distance between the substrate
and the target was 10 cm and the rf power was varied between 50 and 175
W. The deposition rate was varied between 15 and 30 nm/min. The film
thickness was measured using a surface profilometer (Dektak 3 from Sloan
Tech.). The electrical resistivity (U) was measured as a function of
temperature in the range of 300 to 500 K using thermally evaporated
aluminum electrodes in a coplanar configuration. X-ray diffraction
measurements were performed at RT in air, using the Cu-KD line of a Rigaku
DMAX III-C diffractometer. The surface morphologies were analyzed using
a Field Effect Scanning Electron Microscope (FE-SEM, S-1400 Hitachi).
The optical transmittance measurements were performed with a Shimadzu
UV/VIS 3100 PC double beam spectrophotometer in the wavelength from
200 nm to 2500 nm.
3.2
ZnO-TFT fabrication
Fig. 2 shows a schematic illustration of the ZnO-TFT produced, using

only oxide materials. In order to evaluate the performances of the TFTs,
different channel width and channel lengths were used, varying the W/L ratio
from 1 to 65. The devices were electrically characterized in air, at room
temperature, and in the dark using a microprobe station connected to a
semiconductor parameter analyzer (model HP 4145B). The reproducibility of
the device characteristics was confirmed by measuring more than 200 TFTs,
each one several times. The gate dielectric of the ZnO-TFT is based on a
engineered insulator consisting of a superlattice of Al2O3 and TiO2 (ATO)16
deposited on a glass coated with an Indium Tin Oxide (ITO). The
230
E. Fortunato et al.
Fig. 2. Schematic illustration of the ZnO based TFT structure. The ZnO channel layer and the
ATO gate insulator are 100 nm and 220 nm in thickness, respectively, while the ITO gate and
the GZO source and drain are 200 nm and 150 nm in thickness, respectively. The channel
length and gate width ratio varied between 1 and 65.
dielectric had a thickness of 200 nm, exhibiting a transparency of 85%, a

resistivity of 2.310-4 :cm, a carrier concentration of 7.71020 cm-3, and a
mobility of 36 cm2/Vs. The average capacitance of the ATO layer measured
nearby the TFTs produced was 55 nF/cm2, having a total thickness of 220
nm. From this measurement the relative dielectric constant, was determined
as being around 14.
The channel is based on undoped ZnO, 100 nm thick, whose electrical
properties were properly controlled during the deposition pressure aiming to
obtain a material where the conduction is dominated by electrons. In this
case the background electron concentration must be reduced in order to
increase the carriers mobility. This attempted was achieved by optimizing
the deposition conditions for undoped ZnO films produced by rf magnetron
sputtering. To do so we choose to vary the rf power instead of the oxygen
partial pressure (easier for controlling the stoichiometry), because the use of
oxygen during the deposition is not compatible with the lift-off technique
used (the photoresist is removed by a plasma that contains oxygen ions) and
also to minimize the defects at the channel-insulator interface. Moreover, the
processing technology used is relatively simple and it is compatible with
inexpensive plastic/flexible substrate technology.
The drain and source are based on highly conductive gallium doped zinc
oxide (GZO)17, 150 nm thick and patterned by the lift-off technique. These
depositions were also carried out at room temperature.
4.
231
RESULTS
Fig. 3 shows the dependence of the electrical resistivity and the average
optical transmittance at the visible spectra (between 400 and 700 nm), as a
function of rf power density (P). The highest resistivity (| 108 :cm) was
obtained for P = 5 W/cm2. For P around 5 W/cm2 the films became close to
stoichiometry with low structural defects and consequently higher resistive.
As we decrease or increase the rf power density from 5 W/cm2, a deviation
from stoichiometry is obtained accompanied by a decrease on the electrical
resistivity, due to a lower carrier concentration or/and electron mobility.
100
10
10
10
10
10
60
40
10
-1
10
-3
20
Transmittance (%)
Resistivity (:cm)
80
0
10
8
2
rf power density (W/cm )

Fig. 3. Electrical resistivity (measurements were performed at room temperature in vacuum
using thermally evaporated aluminum contacts on stabilized samples) and average optical
transmittance, versus rf power density.
This is also confirmed by a decrease on the optical transmittance,

especially for rf power densities lower than 5 W/cm2. For the films with
lower resistivity (< 102 :cm), we measured Hall mobilities of about 2
cm2/Vs and a carrier concentration of 31016 cm-3 (n-type). For these
undoped films an excess of interstitial Zn ions or/and oxygen vacancies can
contribute with free electrons to the electrical conduction. Concerning the
ZnO films used for the TFTs, due to the high value of the electrical
resistivity (~108 :cm) it was not possible to measure the Hall.
The X-ray diffraction patterns of these films are shown in Fig. 4(a). As
the rf power density increases, the diffraction peak corresponding to the
232
E. Fortunato et al.
0.35
Normalized intensity (a.u.)
8
8.9
0
0.30
7
7.7
0
0.25
6.4
0
0.20
0
0.15
0.10
0
5.
5.1
3.8
0.05
0
2.
2.6
31
32
33
34
35
36
37
(degrees)
(b)
Intensity (a.u.)
500
(002)
600
400
4
300
3
200
2
100
20
30
40
50
60
70
2T (degrees)
Fig. 4. (a) X-ray diffractograms of ZnO films corrected to the thickness of each sample
(ranging from 190 nm to 360 nm), deposited on soda lime substrates, using the Cu-KD line.
The large width of the peaks reveals a limited crystalline domain. For the film produced at 5
W/cm2, using the full-width at half-maximum of the peak at an angle of 34.1 , we calculated
an average crystallite size of the order of 10 r 1 nm using the Scherrer equation. The low
crystallite size is also associated with the fact that the films were processed at room
temperature without any type of post thermal annealing; (b) X-ray diffractogram of a ZnO
film produced with an rf power density of 5 W/cm2 for a wider 2T range. The film thickness is
130 nm.
233
Fig. 5. SEM micrograph with an apparent viewing angle of 40 of the same ZnO film
presented in Fig. 4(b).
(002) orientation become predominant up to a rf power density of 5 W/cm2,

where the resistivity is maximum. Besides that the data also show that this
film is mainly constituted of dense and small crystallites (nanocrystalline
with sizes up to 10 nm). For higher rf power densities the peak disappears,
indicating that the crystallinity of the ZnO starts to degrade. This is
attributed to a decrease of the oxidation of the films. The amorphous phase
starts to become predominant.
Fig. 4(b) shows the X-ray diffraction pattern for the ZnO film produced
at P = 5 W/cm2 for a wider 2T range. For all the films produced only the
ZnO (002) peak at 2T | 34q is observed, revealing that the films are
nanocrystalline with a hexagonal structure and a preferred orientation with
the c-axis perpendicular to the substrate. This is also consistent with the
results obtained by scanning electron microscopy (see Fig. 5).
Fig. 5(a) shows the source-to-drain (IDS) current as a function of the gate
voltage (VGS). It is observed that the ZnO-TFT has an n-channel, since
electrons are generated by the positive VGS. A high IDS (230 PA) is obtained
for VGS = 40 V and VDS = 20 V. Besides that a hard saturation current at the
post pinch-off voltage is observed that is in accordance with the standard
theory of field-effect transistors. The saturation mobility (Psat) and the
threshold voltage (VTH) were calculated by fitting a straight line to the plot of
the square root of IDS vs VGS, according to the expression (saturation region)
I DS
DS
Ci sattW

22L
L
ffor VDS
DS
VGS
G - VT
TH
(1)
234
E. Fortunato et al.
where Ci is the capacitance per unit area of the gate insulator. The obtained
Psat is around 20 cm2/Vs and the VTH is 21 V, showing that the ZnO-TFT
operates in the enhancement mode. Enhancement mode is preferable to
depletion mode since it is not necessary to apply a gate voltage to switch off
the transistor, because the circuit design is simpler and the power dissipation
is lower. The high value of Psatt agrees with the effective mobility (Pefff)18
determined for VDS d 2 V (in order to guarantee a uniform distribution of
carriers in the channel along the source to drain region) that has a value of
about 19 cm2/Vs. Here we also have to notice that the average field-effect
mobility (PFE) obtained from the transconductance18 data in the same VDS
region and for 30 V d VGS d 40 V is around 21 cm2/Vs. These data confirm
the high quality of the undoped ZnO films (improved crystallinity fraction of
the nanocrystals and low oxygen vacancies and/or Zn interstitials working as
donors) as well as the good channel-insulator interface obtained. The
magnitude obtained for VTH is directly proportional to the gate insulator
thickness. Reducing the gate insulator thickness could easily reduce these
values. The off-current is low, on the order of 1.310-9 A, and the on/off
ratio is 2105. That is, the TFT is able to sustain currents as high as 2.310-4
A. The gate voltage swing, S defined as the voltage required to increase the
drain current by a factor of 10 was 1.24 V/decade for the ZnO-TFT under
analysis. S is given by the maximum slope in the transfer curve (Fig. 6a).
Fig. 6(b) shows the IDS current curves as a function of the drain current
(VDS) for different VGS. The saturation was about 230 PA under a gate bias of
40 V. Besides the high value obtained for the saturation current, the device
exhibits hard saturation, evidenced by the flatness of slope of each IDS
curve, for large VDS, without the need of post heating treatments. This
indicates that the entire thickness of the ZnO channel layer is depleted and so
the device exhibits a high resistance (defined as the inverse of the slope of
the I-V curve for a given gate voltage, in the off saturation regime), which is
larger than 20 M:, as required for most circuit applications. The on
resistance (RON = 1/gd) defined as
RON
gd
1
L
Peff CiW ( GS -
TH
(2)
is of about 45 K: (VGS = 30 V). The relation between the ROFF and RON is
within what is expected for fast switching device behavior. Besides that, the
device response time, defined as being proportional to RONCi amounts to
values of about 25 ps/Pm2, which is highly promising for device
miniaturization.
-4
10
-5
10
-6
10
-7
10
-8
10
-9
(a)
0.015
1/2
10
(A )
-3
0.010
1/2
10
235
IDS
IDS (A)
0.005
10
20
30
0.000
40
VGS (V)
-4
2.5x10
VGS (V)
40
(b)
-4
2.0x10
IDS (A)
-4
1.5x10
35
-4
1.0x10
-5
5.0x10
30
25
20
10
20
30
40
VDS (V)
Figure 6. Typical ZnO-TFT characteristics with the channel layer deposited at room
temperature by rf magnetron sputtering, for a TFT with a width-to-length ratio of 1.4. (a)
Transfer characteristics for VDS = 20 V. The on/off ratio is 2105. The ZnO-TFT operates in
the enhancement mode with a threshold voltage of 21 V and a saturation mobility of 20
cm2/Vs. (b) Output characteristics for a ZnO-TFT. The saturation was about 230 PA under a
gate bias of 40 V. The ZnO-TFT exhibits hard saturation, evidenced by the flatness of slope
of each IDS curve, for large VDS. The dashed line represents the saturation drain current that
follows an exponential dependence on the voltage.
The exposure to ambient light of these ZnO-TFTs has no effect on the

current-voltage characteristics, which is an advantage in electronic drivers
for displays.
236
E. Fortunato et al.
Fig. 7 shows the optical transmittance spectrum of the entire ZnO-TFT in

the wavelength range between 200 nm and 2500 nm (including the glass
substrate with 1.1 mm thickness). The average optical transmission in the
visible part of the spectrum is 80% while at 550 nm (maximum sensitivity
for the human eye) it is 85%, which indicates that transmission losses due to
the ZnO-TFTs in comparison with the uncoated glass substrate are negligible
100
Transmittance (%)
glass substrate
ITO film
80
60
40
ITO/ATO film
20
ZnO-TFT
0
500
1000
1500
2000
2500
Wavelength (nm)
Figure 7. Optical transmission spectra for the entire ZnO-TFT structure including the glass
substrate, were it is possible to observe that the ZnO-TFT is fully transparent to visible light.
For comparison the optical transmission spectra for the glass substrate (1.1 mm thickness), the
ITO, and the ITO/ATO films are also shown. The baseline and reference were measured in air
to quantify the total amount of transmitted light. The lower part shows a photograph of a glass
substrate (2.5 2.5 cm2) with ZnO based TFTs.
237
(7%). The absorbance at the near infrared region is associated with the
increase of the carrier concentration of the ITO and GZO films that leads to a
lower value associated to the plasma frequency.19
5.
CONCLUSIONS
In conclusion we have demonstrated the possibility to produce fully

transparent high quality ZnO based TFTs at room temperature by rf
magnetron sputtering on glass substrates. ZnO is one of a few oxides that can
be grown as a polycrystalline or even as a high compact and dense
nanocrystalline material at relatively low or even at room temperature, on a
variety of substrates, such as amorphous glasses and plastics or metal foils.
This particularity together with the possibility to realize efficient and reliable
TFTs exhibiting high mobilitys, opens new doors for the so called invisible
electronic circuits that are highly important for the next generation of
invisible and flexible electronics.
ACKNOWLEDGEMENTS
We thank A. Pakkala from Planar Systems, Inc., Espoo, Finland, for
supplying the ITO/ATO substrates. The authors acknowledge A. Lopes for
the SEM analysis.
REFERENCES
1.
2.
2
3.
J. E. Lilienfeld, U.S. Patentt 1,745,175 (1930).

J. E. Lilienfeld, U.S. Patentt 1,900,018 (1933).
P. K. Weimer, Proceedings of the IRE, 1462 (1962).
T. P. Brody, J. A. Asars, and G. D. Dixon, IEEE Trans. Electron Devices, ED-20, 995
(1973).
4. W. E. Spear and P. G. LeComber, Solid State Commun. 17, 1193 (1975).
5. W. E. Spear and P. G. LeComber, Philosophical Magazine 33, 935 (1976).
6. P. G. LeComber, W. E. Spear, and A. Gaith, Elec. Lett., 15 179 (1979).
7. Y. Kuo in Thin Film Transistors Materials and Processes Volume 1: Amorphous Silicon
Thin Film Transistors, Kluwer Academic Publishers, 2004.
8. F. Garnier, Adv. Mat. 2, 592 (1990).
9. Y. Ohya, T. Niwa, T. Ban, and Y. Takahashi, Jpn. J. Appl. Phys., 40, 297 (2001).
10. S. Masuda, K. Kitamura, Y. Okumura, and S. Miyatake, J. Appl. Phys. 93, 1624 (2003).
11. R. L. Hoffman, B. J. Norris, and J. F. Wager, Appl. Phys. Lett. 82, 733 (2003).
12. P. F. Carcia, R. S. McLean, M. H. Reilly, and G. Nunes, Jr., Appl. Phys. Lett. 82, 1117
(2003).
238
E. Fortunato et al.
13. P. F. Carcia, R. S. McLean, M. H. Reilly, I. Malajovich, K. G. Sharp, S. Agrawal, and G.

Nunes, Jr., Mater. Res. Soc. Symp. Proc. 769, H7.2.1 (2003).
14. J. Nishii, F. M. Hossain, S. Takagi, T. Aita, K. Saikusa, Y. Ohmaki, I. Ohkubo, S.
Kishimoto, A. Ohtomo, T. Fukumura, F. Matsukura, Y. Ohno, H. Koinuma, H. Ohno, and
M. Kawasaki, Jpn. J. Appl. Phys., 42, L347 (2003).
15. B. J. Norris, J. Anderson, J. F. Wager, D. A. Keszler, J. Phys. D: Appl. Phys. 36, L105
(2003).
16. ITO/ATO-coated glass substrates from Planar Systems, Inc. Espoo, Finland.
17. V. Assuno, E. Fortunato, A. Marques, A. Gonalves, I. Ferreira, H. guas and R.
Martins, Thin Solid Films 442, 102. (2003).
18. D. K. Schroder, in Semiconductor Material and Device Characterization, 2nd ed. Wiley,
New York, 1998.
19. T. J. Coutts and D. L. Li Young. MRS Bull. 25, 58 (2000).
Index
Acceptor 37, 69
Atomic force microscopy
Field effect mobility 225

g-factor 159
Band structure 171

Hall effect 37, 47
single crystal 47
thin films 47
Heterointerface 197
Heterojunction 211
Heterostructures 25
Hexagonal resonator 83
Hole injection 211
Hydrogen 115, 125
Cathodoluminescence 83
Cluster 15
Columnar structure 15
Output conductance 217
dc-magnetron sputtering 15
Defects 133
Device physics 217
Dislocations 99
electrical activity 99
DLTS 47
Donor 37, 69, 115
hydrogen 115
Doping 197
Hydrogen 115, 125, 133

binding energy 145
complexes 145
concentration 145
density of states 145
donor 125
local vibrational modes 145
Hyperfine constant 115
Electrical properties 3
Electrical transport 197
Electroabsorption spectra 183
Electroluminescence 211
Electron affinity 115
Electron holography 99
Epitaxial layers 211
Electron paramagnetic resonance 69
photo-EPR 69
Exciton 69, 159, 171, 183
Interband transition 171

IR absorption 133
IR spectroscopy 125
Kinetic analysis 25
Light trapping 197
239
240
Magneto-luminescence 159
Magnetron sputtering 59
Metal organic vapor phase epitaxy 99
Microcrystals 83
Mobility 217
average 217
effective 217
field-effect 217
incremental 217
saturation 217
Molecular static method 183
MOVPE See Metal organic vapor phase
epitaxy
Muon spin rotation 115
Muonium 115
Muons 115
Nanocavity 83
Nanostructure 83
n-type 25, 37
O-H complexes 125
Optical spectra 171
Optoelectronic devices 225
Oxygen radicals 25
PEMOCVD
plasma enhanced MOCVD 59
Permitivity 171
Photoluminescence 59, 69, 83, 159
thermoluminescence 69
Puldes laser deposition 47
Point defects 69
Polarization 83, 171
Polycrystalline films 197
p-type 25, 37
Index
Raman spectroscopy 145
Recursion method 183
Reflectance 171
rf Magnetron sputtering 225
Schottky contacts 47
Structural properties 3
Thermal evaporation 59
Thermography 47
Thin-film transistors 217, 225
Transmission electron microscopy 99
Transparent electronics 225
Transparent thin-film transistor 217, 225
Ultraviolet emission 211
Valence band symmetry 159
Vapor-liquid-solid mechanism 15
Vibrational density of states 183
Vibrational modes 125
WGM See whispering gallery modes
Whisker 15
Whispering gallery modes 83
Window layers 197
Zeeman effect 159
Zn vacancy 133
ZnO synthesis 3
hydrothermal growth 3
melt growth 3
solution growth 3
vapor growth 3

(Norbert H. Nickel, Evgenii Terukov) Zinc Oxide

Uploaded by

Copyright:

Available Formats

(Norbert H. Nickel, Evgenii Terukov) Zinc Oxide

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

(Norbert H. Nickel, Evgenii Terukov) Zinc Oxide

Uploaded by

Copyright:

Available Formats

Zinc Oxide -- A Material for

Micro- and Optoelectronic Applications

NATO Science Series

Life and Behavioural Sciences

Series II: Mathematics, Physics and Chemistry Vol. 194

Zinc Oxide -- A Material for

Published in cooperation with NATO Public Diplomacy Division

Proceedings of the NATO Advanced Research Workshop on

Printed on acid-free paper

All Rights Reserved

PART I: ZnO Bulk and Layer Growth

The Scope of Zinc Oxide Bulk Growth

Growth Mechanism of ZnO Layers

PART II: Electrical, Optical, and Structural Properties

Electrical Properties of ZnO

Zinc oxide - a material for micro- and optoelectronic applications

Electrical Properties of ZnO Thinn Films and Single Crystals

Structure, Morphology, and Photoluminescence of ZnO Films

Optics and Spectroscopy of Point Defects in ZnO

Whispering Gallery Modes in Hexagonal Zinc Oxide Micro- and

PART III: Role of Hydrogen

Muon Spin Rotation Measurements on Zinc Oxide

Hydrogen Donors in Zinc Oxide

Hydrogen-Related Defects in Zno Studied by IR Absorption

Influence of the Hydrogen Concentration on H Bonding in Zinc Oxide 145

Zinc oxide - a material for micro- and optoelectronic applications

Fundamental Optical Spectra and Electronic Structure of Zno Crystals 171

PART V: Device Applications

ZnO Window Layers for Solar Cells

ZnO/AlGaN Ultraviolet Light Emitting Diodes

ZnO Transparent Thin-Film Transistor Device Physics

Zinc Oxide Thin-Film Transistors

Zinc oxide - a material for micro- and optoelectronic applications

Zinc oxide - a material for micro- and optoelectronic applications

Zinc oxide - a material for micro- and optoelectronic applications

Zinc oxide - a material for micro- and optoelectronic applications

Zinc oxide - a material for micro- and optoelectronic applications

Recently, a significant effort has been devoted to the investigation of

PART I: ZnO BULK AND LAYER GROWTH

CNRS, Laboratoire de Physique des Solides et de Cristallogense, 1 Place Aristide Briand,

ZnO can be considered as an 'old' semiconductor which has been

ZINC OXIDE SYNTHESIS

It is worth noting that ZnO occupies already an enviable place in the

The scope of zinc oxide growth

ZINC OXIDE MELT-GROWTH

In spite of its very high melting-point (~ 1900C) and very high

ZINC OXIDE VAPOR GROWTH

Because of the high melting-point of ZnO, there have been many

amount of water to maintain the proper stoichiometry:19 The crystals are

CSVT allows short experiments and easy determination of the growth

The scope of zinc oxide growth

Growth rate ln(A/min)

Figure 2. Instantaneous growth rate as a function of time.

In addition to this chemically assisted sublimation suggested with such

The scope of zinc oxide growth

in the ZnO decomposition, as proposed in the CSVT approach. Some kind of

ZINC OXIDE SOLUTION GROWTH

Figure 4. ZnO crystal grown using NaCl as a solvent.

ZINC OXIDE HYDROTHERMAL GROWTH

where m* is the effective mass and Wn! is defined by