Sie sind auf Seite 1von 204

Influence of the electric polarization on

carrier transport and recombination


dynamics in ZnO-based heterostructures

Von der Fakultt fr Physik und Geowissenschaften


der Universitt Leipzig
genehmigte
DISSERTATION
zur Erlangung des akademischen Grades
Doctor rerum naturalium
Dr. rer. nat.

vorgelegt
von Diplom-Physiker Matthias Brandt
geboren am 10. Mai 1981 in Hohenmlsen

Gutachter:

Prof. Dr. Marius Grundmann, Universitt Leipzig


Prof. Dr. David C. Look, Wright State University

Datum der Einreichung: 2. Dezember 2009


Datum der Beschlussfassung: 16. August 2010
ii
Bibliographische Beschreibung
Brandt, Matthias
Influence of the electric polarization on carrier transport and recombination
dynamics in ZnO-based heterostructures
Universitt Leipzig, Dissertation
201 S. , 269 Lit. , 78 Abb., 3 Anlagen

Referat:
Die vorliegende Arbeit befasst sich mit dem Einfluss der elektrischen Polari-
sation auf Eigenschaften freier Trger in ZnO basierten Halbleiterheterostruk-
turen. Dabei werden insbesondere Transporteigenschaften freier Trger sowie
deren Rekombinationsdynamik untersucht. Die Arbeit behandelt vier inhalt-
liche Schwerpunkte. Der erste Schwerpunkt liegt auf den physikalischen Eigen-
schaften der verwendeten Materialen, hier wird der Zusammenhang der Ban-
dlcke und der Gitterkonstanten von Mgx Zn1x O Dnnfilmen und deren Mag-
nesiumgehalt beschrieben. Weiterhin wird die Morphologie solcher Filme disku-
tiert. Auf unterschiedliche Substrate und Abscheidebedingungen wird dabei
detailliert eingegangen. Der zweite Schwerpunkt behandelt die Eigenschaften
undotierter und phosphordotierter ZnO und Mgx Zn1x O Dnnfilme. Die struk-
turellen, Transport- und Lumineszenzeigenschaften werden hier verglichen und
Rckschlsse auf die Zchtungsbedingungen gezogen. Im dritten Schwerpunkt
werden Quanteneffekte an ZnO/Mgx Zn1x O Grenzflaechen behandelt. Hier-
bei wird insbesondere auf den Einfluss der elektrischen Polarisation eingegan-
gen. Die Prsenz eines zweidimensionalen Elektronengases wird nachgewiesen,
und die notwendigen Bedingungen zur Entstehung des sogenannten qunatum
confined Stark-effects werden dargelegt. Insbesondere wird hier auf zch-
tungsrelevante Parameter eingegangen. Den vierten Schwerpunkt stellen Kop-
plungsphnomene in ZnO/BaTiO3 Heterostrukturen dar. Dabei werden zuerst
die experimentell beobachten Eigenschaften verschiedener Heterostrukturen
die auf unterschiedlichen Substraten gezchtet wurden aufgezeigt. Hier ste-
hen strukturelle und Transporteigenschaften im Vordergrund. Ein Modell
zur Beschreibung der Ausbildung von Raumladungszonen in derartigen Het-
erostrukturen wird eingefhrt und zur Beschreibung der experimentellen Ergeb-
nisse angewandt. Die Nutzbarkeit der ferroelektrischen Eigenschaften des Ma-
terials BaTiO3 in Kombination mit halbleitendem ZnO wurden untersucht. Hi-
erzu wurden ferroelektrische Feldeffekttransistoren unter Verwendung beider
Materialien hergestellt. Die prinzipielle Eignung der Bauelemente als nicht-
flchtige Speicherelemente wurde nachgewiesen.


... S. (Seitenzahl insgesamt)
... Lit. (Anzahl der im Literaturverzeichnis ausgewiesenen Literaturangaben

iii
iv
Summary of the scientific results of the dissertation
Influence of the electrical polarization on carrier transport and recombination
dynamics in ZnObased heterostructures
submitted by
Matthias Brandt
prepared at
Fakultt fr Physik und Geowissenschaften der Universitt Leipzig
Institut fr Experimentelle Physik II, Abteilung Halbleiterphysik
November 2009

1 Introduction
ZnO is a transparent oxide semiconductor with a bandgap energy of Eg 3.37 eV at room temper-
ature. The wide gap and the strong exciton binding energy of about 60 meV make it an interesting
material for optoelectronic applications in the blue and near UV range of the electromagnetic spec-
trum. These interesting material properties triggered a wave of scientific investigation in the past
decade, although reports of the materials properties date back as late as the 1920s [1] . A first report
on transport measurements in ZnO single crystals and thin films emerged in 1935 [2] . Intrinsic n
type conductivity was confirmed in these primary studies. It was found, that a highly doped layer
can exist on the surface of the single crystals, and that its presence depended on the surface polarity
of the single crystal [3] .
It has been argued, that the internal polarization is responsible for the formation of accumulation
and depletion layers at the Zn and Ofaces, respectively [3] . The properties of such accumulation
layers were investigated in the present study in ZnO and Mgx Zn1x O thin films.
As the wurtzite crystal system lacks inversion symmetry, and the ZnO crystal is partially ionic, a
spontaneous polarization must arise in parallel to the [0001] direction. In contrast to GaN and its
ternary alloys, where detailed experimental data on the spontaneous and piezoelectric polarization
and their dependence on the composition of the ternary materials are available, only a few experi-
mental reports are given for ZnO and its related compounds [48] .
In a heterostructure the electric polarization in adjacent materials can be different due to the dif-
ferent layer composition, and in the case of pseudomorph layers due to the strain of the layers. A
sheet charge at the heterointerface will result from that difference. Accumulation of free carriers is
possible at the heterointerfaces and can lead to the formation of two dimensional carrier gases. In
this study ntype ZnO was investigated. Therefore only the formation of a twodimensional elec-
tron gas (2DEG) is expected in equilibrium. The transport properties inside this 2DEG differ from
that of a three dimensional electron gas in the bulk, for example an increased mobility is expected.
This can be employed for the development of highly efficient devices, such as high electron mobility
transistors. The properties of a 2DEG forming in the ZnO/Mgx Zn1x O system were studied in
this work.
In contrast, the discontinuities in the electric polarization can also induce effects which are detrimen-
tal to device performance. One of the effects investigated in detail in the recent past is the quantum
confined Starkeffect (QCSE). The difference in polarization between the barrier Mgx Zn1x O and
the well material of a quantum well (QW) structure ZnO leads to different sheet charges on both
sides of the well, resulting in an electric field within. The square well potential is changed into a
trapezoid potential. The transition energy between electrons and holes is reduced, but more im-
portantly, the maxima of the electron and hole occupation probability become spatially separated.
This results in a decrease of the overlap of the electron and holes wavefunctions, a decrease in the
oscillator strength and the radiative recombination efficiency. Lately extensive efforts are under-
taken to avoid this effect and the corresponding reduction of the radiative recombination efficiency
in light emitting devices. Therefore, growth studies of GaN related devices in non polar orienta-

v
tion are frequent in the recent past. Despite these considerations from the device design point of
view, the observation of the QCSE is interesting, as basic properties, such as the dependence of the
polarization on the alloy composition and the motion of carriers in the plane of the QW become
accessible to experimental observation. It was shown [9] that the QCSE can be diminished by a
controlled damage of the crystal, resulting in a smear out of the heterointerfaces. In the converse
argument the presence of the effect is an indicator for atomically abrupt interfaces. The conditions
necessary for the occurence of the QCSE were therefore investigated, both in experiment and by
calculations.
All these considerations apply to a static electric polarization. It is however possible to change
the electric polarization externally, for example by mechanical stress and electric and magnetic
fields. In the present study, the material ZnO was coupled with the ferroelectric BaTiO3 in order
to reveal the effects arising from the presence of a fixed and a switchable polarization component
in a heterostructure. The combination of these materials is of special interest, as both materials
represent model materials for their individual material classes, and a combination does not face
chemical incompatibilities, as both materials are oxides. The influence of the ZnO polarization
on the ferroelectric hysteresis and the possibility of an application of such heterostructures as non
volatile memory elements were studied.

2 Epitaxy of ZnO and Mgx Zn1x O on ZnO substrates


In any material system investigated, the homoepitaxial growth of the respective material has sig-
nificant advantages over the growth on heterogenous materials. Among them are the perfect lattice
match, leading to a significant reduction of the density of dislocations upon growth as well as the
equal lattice expansion coefficient. This avoids the risk of the formation of cracks and dislocations
upon cooling from elevated temperatures, at which the growth process takes place. Further, the
substrate material equals that of the film, reducing indiffusion of impurities from the substrate.
The state of the art ZnO substrate material available was investigated and compared with respect to
the applicability in the present study. Important issues are the substrate quality, the conductivity of
the substrate, which should be as low as possible in order to observe the transport processes in the
thin films only, and the substrate price. Only hydrothermally grown material proved applicable in
the light of transport investigations. A thermal pretreatment of the substrates necessary to produce
surfaces suitable for epitaxy is discussed [10] .
The mechanisms involved in the homoepitaxial growth of ZnO were investigated. Within the tem-
perature range available layer by layer growth was observed rather than step flow growth [11] . A
significant reduction of the density of dislocations was seen in comparison to ZnO grown on sapphire
substrates [12] . The density of donors in undoped homoepitaxial ZnO thin films was investigated
and a net doping concentration of 6 1014 cm3 was found [10] . The very pure material is ideally
suited for doping experiments. Rutherfordbackscattering channeling studies confirmed that the
crystalline quality of the thin films is close to that of single crystals [13]
The properties of nominally undoped ZnO and Mgx Zn1x O thin films grown on ZnO are strongly
influenced by the oxygen partial pressure during growth. The IIVI ratio in the films is controlled by
the partial pressure of oxygen radicals in the growth chamber. For films grown at low oxygen partial
pressures an increased caxis lattice constant is observed for both the ZnO and Mgx Zn1x O thin
films, most likely due to an increasing density of charged intrinsic defects. This thesis is supported
by an increase in the zinc interstitial related luminescence and the free electron concentration upon
decrease of the oxygen partial pressure [14] . Very similar results were seen in phosphorous doped
ZnO and Mgx Zn1x O thin films grown in ZnO. This indicates that intrinsic defects (and therein
mainly zinc on interstitial site) are responsible for the apparent ntype conductivity and the free
electron concentration in these samples.
Transport measurements on phosphorous doped homoepitaxial ZnO thin films showed high electron
mobilities of up to 800 cm2 /Vs around 70 K and 170 cm2 /Vs at room temperature. Exact modeling

vi
(a) (1 0 4) MgZnO on ZnO Log10(Intensity (counts))

2.5 (b) 70
7.705
60
7.7 2
50

ED (meV)
q (nm )
1

7.695 1.5 40

7.69 30
1
20
7.685
0.5 10
7.68 0
3.542 3.543 3.544 3.545 3.546 3.547 0.0 0.2 0.4 0.6 0.8 1.0 1.2
1
q|| (nm ) ND1/3 (106 -1
cm )

Figure 1: (a) RSM of the (104) reflection of a Mgx Zn1x O thin film pseudomorphically grown on ZnO.
(b) Donor activation energy versus donor concentration in ZnO:P of series I and Mgx Zn1x O :P thin films

of the transport properties using a numerical solution of the Boltzmanntransportequation indi-


cated, that the scattering processes involved equal that of ZnO single crystals [13] . Vicinal growth
was observed for one of the ZnO:P films with a phosphorous content of 0.01 wt.% P2 O5 in the target
grown at 0.1 mbar oxygen partial pressure [15] .
The growth of Mgx Zn1x O thin films on ZnO substrates was investigated. It was shown, that the
temperature required for step flow growth was lower than for pure ZnO. This was explained by the
presence of compressive strain in the films, in accordance with the argumentation given in Refs. [16] .
Perfect vicinal surfaces were prepared, depending on the miscut of the substrate [17] . Such layers
represent an ideal starting point for the fabrication of heterostructures with atomically flat inter-
faces and are therefore of great value both to fundamental studies and for device applications.
The lattice constants a and c of Mgx Zn1x O systematically change with the Mg content x in
relaxed Mgx Zn1x O layers, where a increases with x and c decreases with x. It was shown that
Mgx Zn1x O can be grown pseudomorphically on ZnO [18] (see Fig. 1 (a)). The c axis lattice constant
changes systematically with x. The linear relationship cps = c0 + kc,ps x with kc,ps = 0.064 was
found from data obtained in the present study and in accordance with the experimental data given
in Ref. [19] , and are explained by classical theory of elasticity.
Also for the Mgx Zn1x O thin films grown on ZnO, the influence of the growth conditions was
studied. The composition of the group II constituents depends on the background pressure in the
chamber. The concentration of the lighter Mg atoms compared to Zn atoms is reduced by almost
a factor of three within the range of oxygen partial pressures available in the present growth setup.
Interestingly the concentration of P in doped thin films does change accordingly. It is concluded,
that the atomic mass (which differs between Mg and P by only 20%) of the individual element
influences the kinetic scattering processes with the background gas and therefore the composition
of the plasma at the substrate position.
The similarities of the incorporation of Mg and P indicate that P is like Mg mainly introduced into
the ZnO host lattice on Zn site, in accordance with theoretical calculations [20] . This contradicts
the assumption that phosphorous would form a PO acceptor. Instead, the defect PZn is a donor,
which could in principle emit three electrons. Nevertheless, the formation of phosphorous related
acceptors is predicted [20] , by the formation of a defect complex involving two zinc vacancies.
A transport study revealed that the mobility in the phosphorous doped Mgx Zn1x O thin films
grown on ZnO is comparable to that of the homoepitaxial ZnO:P films. Also the donor concen-
tration, and the dependence on the oxygen partial pressure were found very similar. The critical
donor concentration and activation energy in the dilute limit of the dominant donor in phospho-

vii
rous doped ZnO and Mgx Zn1x O layers grown on ZnO was for the first time determined to be of
NC = 8.3 1018 cm3 and ED0 = 56 meV, respectively [21] (see Fig. 1 (b)). The thermal activation
of phosphorous related acceptor was studied in ZnO and Mgx Zn1x O thin films grown on ZnO. A
transition to ptype conduction was never observed. However an increase of the resistivity by more
than 6 orders of magnitude was observed in the Mgx Zn1x O layers deposited at the highest oxygen
partial pressures. This provides evidence of the activation of the phosphorous related acceptors in
Mgx Zn1x O . The contribution of a degenerate layer to the sample conduction was often found
in homoepitaxial samples and was removed by annealing in oxygen, indicating that the origin of
the degenerate layer are intrinsic defects, likely to be formed at the surface. The separation of
the conduction of the degenerate layer and the bulk of the film was partially possible employing a
correction scheme as proposed in Ref. [22] . A determination of the properties of the individual layers
however is not possible without further assumptions, as the incorporation of the defects, responsible
for the formation of the degenerate layer is not homogeneous. A more sophisticated approach was
employed in this study, which allowed an unambiguous evaluation of the transport properties of the
individual layers. The separation of the conductivity of individual conductive paths is possible from
magnetic field dependent measurements by the maximum entropy mobility spectrum approach. The
individual properties of bulk conduction of a Mgx Zn1x O thin film grown on ZnO and the conduc-
tion of the degenerate layer were revealed over a wide range of temperatures. Indications of two
dimensional transport were found. Such a detailed study of the parallel conduction processes in
ZnO and related heterostructures has not been reported so far.

3 Quantum effects in ZnO/Mgx Zn1x O heterostructures


2D growth in the BurtonCabreraFrank mode is observed for the ZnO/Mgx Zn1x O single het-
erostructures and ZnO/Mgx Zn1x O quantum wells grown on asapphire. The effective surface
roughness however in these samples is of the order of the thickness of the quantumwells present in
the structures. However, polarization induced electron accumulation layers were observed both for
single heterostructures and quantum wells capacitancevoltage spectroscopy and temperature de-
pendent Hall effect measurements [23] . Indications for the presence of a 2DEG in ZnO/Mgx Zn1x O
QWs grown on ZnO were also found by temperature dependent Hall effect measurements [17] .
Quantum confinement effects were confirmed previously by photoluminescence (PL) measurements [24] .
A narrowing of the PL peak width is observed in quantumwells grown on ZnO, especially for low
well widths, due to the higher influence of the interface roughness on thinner quantum wells [17] .
Careful control of the laser fluence applied to the PLD target resulted in a significant improvement
of the quality of the heterointerface in ZnO/Mgx Zn1x O heterostructures. The use of a slanted
QW allowed the observation of quantum confinement effects in QW of different thicknesses in one
sample with a well defined barrier composition. A pronounced QCSE was observed in PLD grown
samples for the first time, indicating that atomically abrupt interfaces exist in the samples. Com-
parison with data obtained on MBE grown samples given in the literature [25] revealed, that the
samples discussed here equal the MBE grown samples both in interface quality and in density of
nonradiative recombination centers (see Fig. 2).

4 Effects of the polarization in ZnO/BTO heterostructures


ZnO/BTO heterojunctions were successful synthesized on both Si and STO substrates. A sub-
strate preparation routine following Ref. [26] was applied to commercially available STO substrates
and produced perfect vicinal substrates with step edges of unit cell height. The structural and
electrical properties of the structures grown on the lattice matched STO substrates are by far su-
perior to these of the samples grown on Si [27] . Vicinal BTO thin films with step edges of unit cell
step height were grown on STO up to a critical thickness of roughly 45 nm in concordance with

viii
10-3

Source - Drain Current Isd (A)


(a) (b)

QW Emission Energy [eV]


3.6 10-4 10-8

Decay time [s]


-5
10
3.4 10-9
10-6
3.2 10-7 10-10
Isd (A) (switched on)
-8
10
3.0 10 -11 Isd (A) (switched off)
10-9 Vsd = 2V
2.8 10-10 10-12
0 2 4 6 8 10 1.5 2.0 2.5 3.0 3.5 4.0
QW thickness [nm] Gate Voltage Vg (V)

Figure 2: (a) Dependence of the luminescence energy (left scale) and decay time (right scale) at the
maximum of the QW luminescence. Closed symbols give data obtained in this study, while open symbols are
taken from Ref. [25] . Solid lines give the expected dependence from the variational calculations. (b) Transfer
characteristics of a ZnO/BTO MISFET after applying a for and backwards pulse, respectively.

Ref. [28] . A leakage current density as low as jl < 109 A/cm2 was found from IV measurements on
ZnO/BTO heterostructures grown on STO. From the curves obtained the polarization of the ZnO
layer PZnO = 0.2 0.05 C and the coercive field of the BTO EC = 12.5 1 kV/cm were
estimated.
A detailed model was developed to understand the coupling of the ZnO and BTO polarization and
the formation of depletion layers in the ZnO introduced by the sheet charge related to the change
of the polarization at a ZnO/BTO heterointerface [29] . Employing the aforementioned model the
SawyerTower loops of single BTO layers, ZnO/BTO single heterojunctions and ZnO/BTO double
heterojunctions could be accurately described, and the material parameters of the individual layer
were extracted. Qualitative description of the CV loops of ZnO/BTO double heterostructures was
also possible from this model [30] .
Ferroelectric thin film field effect transistors were demonstrated on the basis of ZnO/BTO het-
erostructures. The transistors were normallyon and showed output currents of up to 10 A,
which can be controlled over more than 6 orders of magnitude by the gate voltage. A reproducible
nonvolatile memory effect was observed due to the ferroelectric polarization in the BTO [31] (see
Fig. 2 (b)).

References
[1] N. E. Eberly, C. V. Gross, and W. S. Crowell. J. Am. Chem. Soc. 42, 1433 (1920).
[2] O. Fritsch. Ann. Phys. 414, 375 (1935).
[3] G. Heiland and P. Kunstmann. Surf. Sci. 13, 72 (1969).
[4] J. Jerphagnon and H. W. Newkirk. Appl. Phys. Lett. 18, 245 (1971).
[5] A. Tsukazaki, A. Ohtomo, T. Kita, Y. Ohno, H. Ohno, and M. Kawasaki. Science 315, 1388
(2007).
[6] M. Yano, K. Hashimoto, K. Fujimoto, K. Koike, S. Sasa, M. Inoue, Y. Uetsuji, T. Ohnishi, and
K. Inaba. J. Cryst. Growth 301, 353 (2007).
[7] H. Tampo, H. Shibata, K. Maejima, A. Yamada, K. Matsubara, P. Fons, S. Kashiwaya, S. Niki,
Y. Chiba, T. Wakamatsu et al.. Appl. Phys. Lett. 93, 202104 (2008).
[8] T. Bretagnon, P. Lefebvre, T. Guillet, T. Taliercio, B. Gil, and C. Morhain. Appl. Phys. Lett.
90, 201912 (2007).
[9] J. A. Davis and C. Jagadish. Laser Photonics Rev. 3, 85 (2009).
[10] H. von Wenckstern, H. Schmidt, C. Hanisch, M. Brandt, C. Czekalla, G. Benndorf, G. Biehne,
A. Rahm, H. Hochmuth, M. Lorenz et al.. Phys. Status Solidi RRL 1, 129 (2007).
[11] H. von Wenckstern, M. Brandt, H. Schmidt, G. Benndorf, J. Zippel, H. Hochmuth, M. Lorenz,

ix
and M. Grundmann. vol. 6895, 689505. SPIE (2008).
[12] M. Lorenz, G. Wagner, A. Rahm, H. Schmidt, H. Hochmuth, H. Schmid, W. Mader, M. Brandt,
H. von Wenckstern, and M. Grundmann. Phys. Status Solidi C 5, 3280 (2008).
[13] M. Brandt, H. von Wenckstern, H. Schmidt, A. Rahm, G. Biehne, G. Benndorf, H. Hochmuth,
M. Lorenz, C. Meinecke, T. Butz et al.. J. Appl. Phys. 104, 013708 (2008).
[14] A. Lajn, H. von Wenckstern, G. Benndorf, C. Dietrich, M. Brandt, G. Biehne, H. Hochmuth,
M. Lorenz, and M. Grundmann. J. Electron. Mater. in press (2009).
[15] H. von Wenckstern, M. Brandt, H. Schmidt, C. Hanisch, G. Benndorf, H. Hochmuth, M. Lorenz,
and M. Grundmann. J. Korean Phys. Soc. 53, 3064 (2008).
[16] H. Brune, K. Bromann, H. Rder, K. Kern, J. Jacobsen, P. Stoltze, K. Jacobsen, and
J. No/rskov. Phys. Rev. B 52, R14380 (1995).
[17] M. Brandt, H. von Wenckstern, G. Benndorf, H. Hochmuth, M. Lorenz, and M. Grundmann.
Thin Solid Films 518, 1048 (2009).
[18] M. Lorenz, M. Brandt, G. Wagner, H. Hochmuth, G. Zimmermann, H. von Wenckstern, and
M. Grundmann. Zinc Oxide Materials and Devices IV 7217, 72170N (2009).
[19] Y. Nishimoto, K. Nakahara, D. Takamizu, A. Sasaki, K. Tamura, S. Akasaka, H. Yuji, T. Fujii,
T. Tanabe, H. Takasu et al.. Appl. Phys. Express 1, 091202 (2008).
[20] W.-J. Lee, J. Kang, and K. J. Chang. Phys. Rev. B 73, 024117 (2006).
[21] M. Brandt, H. von Wenckstern, C. Meinecke, T. Butz, H. Hochmuth, M. Lorenz, and M. Grund-
mann. J. Vac. Sci. Technol. B 27, 1604 (2009).
[22] D. C. Look and R. J. Molnar. Appl. Phys. Lett. 70, 3377 (1997).
[23] H. von Wenckstern, J. Zippel, A. Lajn, M. Brandt, G. Biehne, H. Hochmuth, M. Lorenz, and
M. Grundmann. AIP conference proceedings (in press).
[24] S. Heitsch, G. Zimmermann, A. Mller, J. Lenzner, H. Hochmuth, G. Benndorf, M. Lorenz,
and M. Grundmann. Proc. MRS 957, K07 (2006).
[25] T. Guillet, T. Bretagnon, T. Taliercio, P. Lefebvre, B. Gil, C. Morhain, and X. Tang. Super-
lattices Microstruct. 41, 352 (2007).
[26] G. Koster, G. Rijnders, D. H. A. Blank, and H. Rogalla. Physica C 339, 215 (2000).
[27] M. Lorenz, M. Brandt, J. Schubert, H. Hochmuth, H. von Wenckstern, M. Schubert, and
M. Grundmann. Zinc Oxide Materials and Devices II 6474, 64741S (2007).
[28] A. Petraru, N. A. Pertsev, H. Kohlstedt, U. Poppe, R. Waser, A. Solbach, and U. Klemradt.
J. Appl. Phys. 101, 114106 (2007).
[29] V. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann, N. Ashkenov, and M. Schubert.
J. Electron. Mater. 37, 1029 (2008).
[30] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, N. Ashkenov, M. Grundmann, and M. Schu-
bert. Appl. Phys. Lett. 95, 082902 (2009).
[31] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, M. Grundmann, and J. Schubert. J. Vac.
Sci. Technol. B 27, 1789 (2009).

x
Contents

1 Introduction 1

2 Experimental Techniques 5
2.1 Pulsed laser deposition . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Atomic force microscopy (AFM) . . . . . . . . . . . . . . . . . . 6
2.3 X-ray diffraction (XRD) . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Hall-effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Luminescence measurements . . . . . . . . . . . . . . . . . . . . 9
2.6 Rectifying contancts . . . . . . . . . . . . . . . . . . . . . . . . 9
2.7 Currentvoltage measurements (IV) . . . . . . . . . . . . . . . 10
2.8 CapacitanceVoltage spectroscopy (CV) . . . . . . . . . . . . . 10
2.9 SawyerTower measurements (ST) . . . . . . . . . . . . . . . . . 11
2.10 Other techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Physical properties of materials employed 13


3.1 Binary ZnO and ternary Mgx Zn1x O . . . . . . . . . . . . . . . 13
3.1.1 Change of the bangap energy . . . . . . . . . . . . . . . 14
3.1.2 Influence of the Mg content x on the lattice constants . . 15
3.2 Growth of ZnO and Mgx Zn1x O thin films . . . . . . . . . . . . 23
3.2.1 Growth of ZnO on sapphire substrates . . . . . . . . . . 23
3.2.2 Growth of Mgx Zn1x O on sapphire substrates . . . . . . 31
3.2.3 ZnO substrate material . . . . . . . . . . . . . . . . . . . 33
3.2.4 ZnO substrate preparation . . . . . . . . . . . . . . . . . 36
3.2.5 Growth of ZnO on ZnO substrates . . . . . . . . . . . . 38
3.2.6 Growth of Mgx Zn1x O on ZnO substrates . . . . . . . . 45
3.3 The perovskites BaTiO3 and SrRuO3 . . . . . . . . . . . . . . . 49
3.4 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . 53

4 Properties of ZnO and Mgx Zn1x O single layers 55


4.1 Properties of homoepitaxial ZnO thin films . . . . . . . . . . . . 56
4.1.1 Properties of nominally undoped homoepitaxial ZnO thin
films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.1.2 Properties of phosphorous doped homoepitaxial ZnO thin
films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Evaluation of the transport properties by mobility spectrum
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

xi
Contents

4.3 Properties of phosphorous doped Mgx Zn1x O thin films grown


on ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.4 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . 90

5 Quantum effects on carriers in ZnO/Mgx Zn1x O heterostruc-


tures 93
5.1 Principle considerations . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Carrier confinement in ZnO/Mgx Zn1x O single heterostructures 96
5.3 Electron accumulation in ZnO/Mgx Zn1x O quantum wells . . . 98
5.3.1 Single quantum wells grown on a-plane sapphire . . . . . 98
5.3.2 Single quantum wells grown on ZnO substrates . . . . . 99
5.4 The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O sys-
tem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.4.1 Literature data on the QCSE in ZnO/Mgx Zn1x O Quan-
tum wells . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.4.2 Luminescence phenomena related to presence of the QCSE106
5.4.3 Observation of the QCSE in ZnO/Mgx Zn1x O quantum
wells grown by PLD . . . . . . . . . . . . . . . . . . . . 108
5.5 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . 118

6 Polarization related effects in ZnO/BaTiO3 heterostructures 121


6.1 Effects observed in ZnO/BTO heterostructures . . . . . . . . . . 122
6.1.1 ZnO/BTO samples grown on Si . . . . . . . . . . . . . . 122
6.1.2 ZnO/BTO samples grown on lattice matched substrates 124
6.2 Theory of coupling phenomena in ZnO/BTO heterostructures . 131
6.2.1 Derivation of a model describing the coupling phenomena 131
6.2.2 Application of the model to experimental data . . . . . . 137
6.3 ZnO/BTO based ferroelectric thin film transistors . . . . . . . . 143
6.4 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . 147

7 Summary and outlook 149


7.1 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . 149
7.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Acknowledgement 187

Curriculum Vitae 191


Address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Personal Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Language Skills . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192

xii
Chapter 1

Introduction

ZnO is a transparent oxide semiconductor with a bandgap energy of Eg =


3.37 eV at room temperature [1] . The wide gap and the strong exciton binding
energy of 60 meV [1] make it an interesting material for optoelectronic appli-
cations in the blue and near UV range of the electromagnetic spectrum. Its
outstanding material properties triggered a wave of scientific investigations in
the past decade. However, numerous reports of the materials properties date
back as late as the 1920s [2] . First experiments revealing the conductivity
of polycrystalline ZnO were already reported in that time [3,4] . A first report
on transport measurements in ZnO single crystals and thin films emerged in
1935 [5] . An intrinsic n-type conductivity was discovered in these first studies.
Detailed information on the thermal activation of the conductivity was given
in the 1950s [6,7] , and first tentative assignments of the defects involved to their
chemical nature were given. Further the mechanisms limiting the mobility were
investigated and theoretically modeled. Fundamental material parameters like
the effective electron mass and Debye temperature were estimated in this pro-
cess [8] . At the same time first indications were found, that a highly doped
layer exists near the surface of the single crystals, and that its presence can be
controlled by the atmosphere the crystal is exposed to [9] . More detailed inves-
tigations showed, that the formation of these accumulation layers depends on
the polarity of the surface where the treatment is applied to [10] .
As the wurtzite crystal system lacks inversion symmetry, and the ZnO crystal
is partially ionic, a spontaneous polarization must arise parallel to the [0001]
direction [11] . It was argued, that this internal polarization is responsible for
the formation of accumulation and depletion layers at the Zn- and O-faces,
respectively [10] and that quantization effects will play a role in the case of ac-
cumulation [12] . The properties of such accumulation layers will be investigated
in the present study in ZnO and Mgx Zn1x O thin films.
Theoreticians have calculated the charge per atom, and the lattice parame-
ters of the material, resulting in values for the spontaneous polarization of
PZnO = 0.032 C/m2 [13] , PZnO = 0.05 C/m2 [11] and PZnO = 0.057 C/m2 [14] .
The spread of the values obtained in this way is larger in other studies, which
partially rely on experimental values for the lattice constants [15,16] . First ex-
perimental studies employing second harmonic generation yielded a value of

1
1. Introduction

PZnO = 0.07 0.015 C/m2 [17] .


The influence of the ZnO polarization on the electronic structure and the for-
mation of defects was studied by photoluminescence spectroscopy, current
voltage, capacitancevoltage, and Hall-effect investigations [1823] . The relative
intensity of luminescence features differs reproducibly, which was explained by
phenomenological models involving the different positions of the Fermi-level [20] .
Different free carrier concentrations are apparent [19,2123] , chemical differences
are reported [18] and a remarkable difference in the work function was observed
between the O- and the Zn-face [19] .
Recently more elaborate calculations of surface reconstruction mechanisms
were carried out, which showed, that both polar ZnO and nonpolar surfaces
show different surface reconstructions [24] which contribute to the compensa-
tion of the internal polarization creating an externally neutral crystal. The
polarity of ZnO single crystals can be easily determined by etching the polar
surface [25] . Nondestructive methods involve more complexity, especially if ap-
plied to thin films. Rather complex techniques like anomalous dispersion X-ray
diffraction [2628] , scanning nonlinear dielectric microscopy [29] , angular resolved
X-ray photoelectron diffraction [30,31] and coaxial impact collision ion scattering
spectroscopy [3234] were successfully employed in order to determine the polar
orientation of ZnO thin films. However, non of these investigations was able
to reveal the magnitude of the spontaneous polarization.
A different approach can, however, be adapted from ongoing research into
other wurtzite type material systems. Detailed investigations were carried out
in the GaN/AlGaN system [39,40] . Therein, the dependence of both spontaneous
polarization and piezoelectric polarization on the Al concentration was deter-
mined from the carrier density in two-dimensional electron gases forming at het-
ereointerfaces. Similar investigations were carried out in the ZnO/Mgx Zn1x O
system [3537] . Further experimental values have been obtained by numerical
evaluation of the photoluminescence shift in ZnO/Mgx Zn1x O quantum well
structures due to the internal electric fields arising from the different electric
polarization of the well and the barrier material [38] . The values of the effective
polarization, the sum of piezoelectric and spontaneous polarization, extracted
from these different considerations differ by up to 30% among themselves and
from theoretical values as given in [13] . The dependence of the spontaneous po-
larization derived in these studies on the Mg content x is given in Fig. 1.1. The
decomposition of the effective polarization into the spontaneous and piezoelec-
tric polarization introduces further uncertainties, as the lattice constants and
piezoelectric coefficients have to be known, which are both subject to exper-
imental error. Therefore it seems beneficial to think of alternative routes for
the determination of the spontaneous polarization in ZnO and its compounds.
The difference in the electric polarization in ZnO/Mgx Zn1x O heterostruc-
tures gives rise to unique effects. Carrier accumulation is possible at the het-
erointerfaces and can lead to the formation of two-dimensional carrier gases
(2DEG) [41] . The transport properties inside this 2DEG differ from that of a
three-dimensional electron gas (3DEG) in the bulk, which can be exploited
for the development of highly efficient devices, such as high electron mobil-

2
0.04

0.03

0.02
P (C/m2 )

0.01

0.01

0.02
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Mg content x

Figure 1.1: Change in the spontaneous polarization Psp (dotted lines), piezoelec-
tric polarization Ppe (dashed lines) and resultant polarization (solid lines) at a
ZnO/Mgx Zn1x O heterointerface as a function of the x. Red lines after Ref. [35] ,
blue lines after Ref. [36] , green line after Ref. [37] and black line after Ref. [38] .

ity transistors. The mechanisms involved in the formation of a 2DEG in the


ZnO/Mgx Zn1x O system will be studied in this work.
In contrast, the discontinuities in the electric polarization can also induce ef-
fects which are detrimental to device performance. One of the effects investi-
gated in detail in the recent past is the quantum confined Stark-effect (QCSE).
The change in polarization between the barrier and the well material of a quan-
tum well (QW) structure in a wurtzite material such as ZnO or GaN and their
compounds leads to an electric field inside the quantum well. The square well
potential is changed into a triangular potential. The transition energy between
electron and hole states is reduced, but more importantly, the maxima of the
electron and hole occupation probability become spatially separated. This re-
sults in a decrease of the overlap of the electron and hole wave-functions, a
decrease in the oscillator strength and the radiative recombination efficiency.
Lately extensive efforts are undertaken to avoid this effect in light emitting
devices. Therefore, growth studies of GaN-related materials in non polar ori-
entation are frequent in the recent past [42,43] . First attempts of the realization
of non-polar material are also reported from the ZnO system [44,45] . Despite
these considerations from the device design point of view, the observation of
the QCSE is interesting, as basic properties, such as the dependence of the
polarization on the alloy composition and the motion of carriers in the QW

3
1. Introduction

become accessible to experimental observation. The observable phenomena


related to the QCSE and the conditions necessary to its observation will be
investigated.
All the above mentioned considerations apply to a static electric polarization.
It is however possible to change the electric polarization externally, for exam-
ple by mechanical stress and by electric and magnetic fields. In the present
study, the material ZnO will be coupled with the ferroelectric BaTiO3 in order
to reveal the physical mechanisms arising from the presence of a fixed and a
switchable polarization component in a heterostructure. The combination of
these materials is of special interest, as both materials represent model materi-
als for their individual material classes, and a combination does not face chem-
ical incompatibilities, as both materials are oxides. The influence of the ZnO
polarization on the ferroelectric hysteresis will be studied, and the possibility
of an application of such heterostructures as non volatile memory elements will
be investigated.
This thesis is organized as follows: In Chap. 2 experimental techniques and
equipment involved will be briefly discussed. Chapter 3 gives an overview over
the physical parameters of the material systems involved and growth conditions
and morphology of the samples investigated. The structural, electric and lumi-
nescence properties of undoped and phosphorous doped ZnO and Mgx Zn1x O
thin films grown on ZnO are discussed in Chap. 4. Chapter 5 is devoted to
quantum effects in ZnO/Mgx Zn1x O films, the focus here is on the formation
of 2DEGS and the presence or absence of the QCSE. The unique coupling
effects discovered in ZnO/BTO heterostructures are disscussed in Chap. 6, a
quantitative model is introduced and nonvolatile memory device demonstrated.
Chapter 7 summarizes the thesis and gives an outlook on further work planned
or recommended by the author.

4
Chapter 2

Experimental Techniques

Within this chapter experimental techniques involved in the present study will
be briefly described. Herein an illustration of these techniques will be empha-
sized which have been personally employed by the author. Other techniques
will only be mentioned with reference to literature giving the full experimental
details and the persons performing the experiments.

2.1 Pulsed laser deposition


Pulsed laser deposition (PLD) was used for the fabrication of all samples re-
garded in the following. A detailed description of the method, its development
and the material system deposited in the semiconductor physics group of the
Universitt Leipzig can be found in Ref. [46] . In the PLD processes a target
of the respective material is hit by a focused laser beam. The wavelength of
the laser beam is 248 nm, which is emitted from a LPX305 KrF excimer laser
bought from Coherent Lambda Physik. The laser is pulsed, with a pulse length
of 5 ns, and a repetition frequency of 130 Hz. The energy per pulse can be
set between 500 and 1,000 mJ. Around 20% of this intensity is lost by reflec-
tion and absortion at the optical elements within the light path (mirrors, the
lense, and the entrance window of the chamber). The laser fluence, meaning
the power density of the inbound laser beam on the target can be selected by
moving the lense off focus.
The target consists of a sintered ceramic pelled. The poweder is homogenized
before in an agate mill, and pressed into the pelled.
The substrate is mounted onto a circular rotating substrate holder, which is
radiatively heated from the back side. The heater itself consists of Kantal wire,
which is periodially folded and anclosed by alumina tubes. behind the heater
several metal reflection screens are mounted, to ensure a highly efficient heat
transport to the substrate holder.
The laser beam excites the electron system of the target, which is partially
relaxed to the phonon system of the target material. Multiple processes take
place in parallel which are summarized in Ref. [47] . It is important to note that
the energy introduced by the laser beam both results in evaporation of the
target and thermal heating, which will eventually result in a liquid phase on

5
2. Experimental Techniques

the target surface. The evaporated components will be further heated by the
laser pulse, and partially ionized. The plasma in the vicinity of the target sur-
face can also contribute to the sputtering of the target and will speed up the
creation of atomic and ionic species. In contrast, the molten target material is
typically not emitted from the target. It is however possible, that due to the
rapid expansion of the heated material parts of the liquid material will form
a drop moving into the direction of the target. Such a large portion of tar-
get material will not redistribute on the substrate surface as the single atoms,
which results in the formation of micrometer sized hillocks, named droplets.
Detailed investigations were carried out on the kinetic of the individual con-
stituent of the plasma [4850] . Nevertheless, the exact interactions are not yet
completely understood. This study does not aim to contribute to the un-
derstanding of the processes in the laser plasma. Nevertheless, as the plasma
kinetic influences the composition of the film it will be regarded in the presenty
study. Among the parameters varied during the growth of the thin films, the
substrate temperature, the chamber background pressure and the laser fluence
will be varied.

2.2 Atomic force microscopy (AFM)


In the present study three different AFM setups were used. The reason is the
temporal availability of the individual setups. Initial studies were carried out
on a Veeco Metrology Dimension 3100 with a nanoscope IVa controller in the
semiconductor physics group. Additionally a Veeco Metrology Dimension 3000
with a nanoscope III controller in the group of Prof. Dr. Friedrich Kremer at
the Universitt Leipzig was used. Further a Park system XE-150 microscope
as employed in the study. All measurements were performed using standard
silicon non contact cantilevers, and the images were obtained in the non-contact
mode. In the case of the Veeco Metrology instruments, a special non-contact
mode named tapping was used. The measurements were performed by the
author, Helena Hilmer, Stefan Schche, and Gregor Zimmermann.

2.3 X-ray diffraction (XRD)


For all XRD investigations presented in this study a Philips Analytical Ma-
terials Research XPert Diffractometer was used. The instrument possesses
two different goniometers, one for wide angle measurements and one for high
resolution (HR-XRD). Cu K radiation is used in both cases, however for the
HR-XRD measurements a Bartels monochromator is introduced into the light
path. Herein a pointshaped Cu K1 beam is selected by four Ge (220) reflec-
tion. The detector was either kept open (for rocking curve measurements) or
a BonseHart analyzer crystal with acceptance angle of 12 arcsec (triple axis
configuration, for reciprocal space maps) was used. The measurements were
performed by Dr. habil. Michael Lorenz, Christof Dietrich, Stefan Schche,
and Jan Zippel.

6
2.4. Hall-effect

2.4 Hall-effect
For the magnetotransport measurements conducted in this study two different
setup were employed. The setup available at the semiconductor physics group
features a Keithley 220 current source, with a current range between 1 nA and
1 A. All measurements were performed in the van-der-Pauw geometry. Differ-
ent contact configurations are scanned by a Keithley 7065 Hall-measurement
card. The signal is detected with a Keithley 2000 multimeter. A temperature
range between 20 K and 800 K is accessible. A magnetic field of up to 0.43 T
can be applied.
For the magnetic field dependent measurements a setup at the Forschungszen-
trum Dresden Rossendorf is used. The measurement device is a Lakeshore
9700A cryostat, herein magnetic field of up to 9 T can be applied. The mea-
surements were performed by Danilo Buerger.
For the evaluation of the measured data a exact solution of the Boltzmann-
transport-equation (BTE) is used, as outlined in Refs. [51,52] . Herein the re-
lationship between the Hall- and drift mobility one is calculated via the per-
turbations of the electron distribution function caused by a external electrical
field F and magnetic field B respectively. In this case the BTE can be written
as:     
e ~ky f ~kx f df
F +B B = , (2.1)
~ me kx me ky dt coll
where e is the elementary charge, ~ is Plancks constant, me the electron effective
mass, f the electron distribution function, kx and ky are the components of
the electron wavevector in the plane perpendicular to the magnetic field and
df
dt coll
denotes the change in f due to collisions of the carriers. This equation
can be solved by the Ansatz of two coupled perturbation terms x and y :

e~F f0 eB
f = f0 (kx x + c ky y ) where c = , (2.2)
me E me

which are linked by


Lc x = 1 + c2 y (2.3)
and
Lc y = x . (2.4)
Herein E demarks the carrier energy and Lc the so called collision-operator

Lc (E) = So (E) + el (E) Si ((E)) , (2.5)

where So and Si denote the inelastic rates of scattering in and out respectively
and el the combined rates of all elastic scattering processes, which can be
calculated assuming a relaxation time i for each scattering process by
X X1
el = i = . (2.6)
i i
i

7
2. Experimental Techniques

The scattering processes included are: scattering at ionized impurities, scat-


tering at the acoustic deformation potential, piezoelectric scattering and polar-
optical scattering. All relaxation rates were calculated following the assump-
tion of isotropic and parabolic bands and the absence of screening apart form
ionized impurity scattering, which is given in the BrooksHerring approxima-
tion [53,54] . The scattering rate of ionized impurity scattering is
 2 2 2 2 2
NI e ln(1 + BH ) BH (1 + BH )
ionizedimpurity = 3/2 , (2.7)
E 0 16 2me
where  1/2
2m 6kT
BH = LD . (2.8)
~ me
LD is the Debye length, = 1/(kB T ) with kB the Boltzmann constant and
T the absolute temperature, the relative dielectric constant of ZnO, 0 the
dielectric constant of free space and NI is related to the density of ionized
impurities and can be expressed as NI = 2NA + n by assuming singly ionized
donors and acceptors. For piezoelectric scattering [53]
1/2
me e 2 P 2 1
piezoelectric = 3/2 2 (2.9)
2 ~ 0 E 1/2
e2 /c
is used, where P = 0 +e 2 /c is the electromechanical coupling coefficient con-
pz l
pz l
taining the isotropic piezoelectric coefficient epz and the longitudinal elastic
constant cl . The acoustic deformation potential scattering is modeled by [53]
3/2 2
2me XC 1/2
acousticdeformationpotential = E , (2.10)
~4 cl
where XC stands for the hydrostatic deformation potential of the conduction
band. For a non degenerate semiconductor the scattering in and out rates of
polar optical phonon scattering can be written as:
~o /T
So = Npo (+
o + e o ) = Npo (+
o + e o ) (2.11)

and

Si = Npo (
i + e i ) = Npo (
~o + +
i + e
/T + +
i ) , (2.12)

where Npo designates the phonon occupation number given by the BoseEinstein

distribution, and o and i are the scattering in and out rates respectively,
independent of the phonon occupation number. The indices + and give ref-
erence to the fact that the terms are evaluated at the Energies E + kB and
E kB respectively, where is the Debye temperature and o the optical
phonon frequency. For these rates the following equations hold:

e2 kB me
   
1 1 a + 1
o = ln , (2.13)
4 2~2 0 E a 1

8
2.5. Luminescence measurements

!
e2 kB me
 
1 1 2E + kB a + 1
1 , (2.14)
i = ln
2 E(E + kB ) a 1
p
4 2~2 0 E
where r
~o
a = 1 . (2.15)
E
The perturbation terms were evaluated in a seven step iteration process for
any temperature recorded in the experiment. From the perturbation terms the
components xx and xy of the conductivity tensor can be computed by nu-
merical integration over energy space, and the Hall-mobility can be calculated
as well as the Hall-factor, which in turn allows to calculate the drift mobility.
f
R
2 ne2 dE x E 3/2 E
xx = R , (2.16)
3 me dE Ef (E, T )
f
R
2 ne2 dE y E 3/2 E
xy = c R . (2.17)
3 me dE Ef (E, T )
An iterative technique has to be applied for a complete modeling of a data
set, as the free electron concentration enters the term for ionized impurity
scattering. This is altered by the Hall-factor in the experiment and must be
corrected after a calculation the drift mobility. This corrected value of the free
electron concentration enters a succeeding calculation of the mobility, and so
on. A self consistent solution can be obtained within two or three iterations.

2.5 Luminescence measurements


Within this study photoluminescence (PL) and cathodoluminescence (CL)
measurements were perfomed. The PL measurements were conducted at 2 K
and room temperature, while the CL measurements take place at roughly 10 K.
In the PL measurements the samples were excited, depending on the Mg con-
tent in the barrier, by the 325 nm line of a cw HeCd laser (KIMMON IK3552R-
G) or a frequency multiplied tuneable 200 fs pulse titanium doped sapphire laser
(Coherent Mira 900F, pumped by an Nd:YAG-Laser Verdi V 10). The lumi-
nescence was dispersed by either a 320 mm or a 1,000 mm monocromator using
different gratings. For time resolved measurements a Hamamatsu R3809U-52
microchannel photomultipliertube was used. The experimental conditions are
in great detail described in Refs. [55,56] . The PL measurements were performed
by Gabriele Benndorf, Christof Dietrich, Susanne Heitsch, Johannes Kupper,
Martin Lange, Alexander Mller, and Marko Stlzel. The CL was measured
by Jrg Lenzner and Jan Zippel.

2.6 Rectifying contancts


In order to perform capacitance spectroscopic measurements rectifying contacts
have to be prepared. Great expertise exist in the semiconductor physics group

9
2. Experimental Techniques

regarding this topic. Two types of contacts are used, thermally evaporated
Pd Shottky-contacts and reactively sputtered metal contacts. A comprehen-
sive review on the properties of Pd contacts is given in Ref. [47,57] . Details on
sputtered contacts can be found in Refs. [58,59] .

2.7 Currentvoltage measurements (IV)


The current transport trough a space charge region is limited a potential barrier
accompanied by the bending of the bands in the semiconductor. The current
voltage dependence can be modeled as
   
e(V IRs V IRs
I(V ) = I0 exp 1 + , (2.18)
kB T Rp

where I and V are the current and voltage respectively, Rs and Rp denote the
sample series and parallel resistance is the ideality factor and I0 describes the
saturation current. A detailed derivation of the model equation can be found
in Ref. [60] . In this work the current flowing through a diode is modeled using a
least squares fit. Therein the characteristic is calculated numerically following
Eqn. 2.18.

2.8 CapacitanceVoltage spectroscopy (CV)


The capacitance of a Schottky diode can be written as
s
0 eNt
C(V ) = A , (2.19)
2(b V )

where C and V are capacitance and voltage, respectively, r and 0 denote


the relative permittivity of the sample and the permittivity of free space, A is
the sample cross sectional area and b is the barrier height. The net dopant
concentration Nt can be calculated from the inversion of the CV dependence.
Further the density of accumulated carriers in potential minima such as QWs
can be determined, as outlined in Ref. [61] . Additionally a time dependence
of the emission of carriers from their parent defects exist, which depends on
their activation energy. Thermal admission spectroscopy (TAS) measures the
frequency and temperature dependence of the conductance and capacitance of
a sample. The activation energy of the defects in the probed sample volume
can be derived from the resonance frequency found in the conductance versus
frequency and capacitance versus frequency dependence and the temperature.
A detailed derivation can be found in Ref. [47,62] . The CV and TAS mea-
surements discussed in the present thesis were performed by Dr. Holger von
Wenckstern and Alexander Lajn.

10
2.9. SawyerTower measurements (ST)

V
Vin

Cst Vo

Figure 2.1: Schematics of the Sawyer Tower measurement circuit. (Courtesy of


Venkata M. Voora)

2.9 SawyerTower measurements (ST)


In order to determine the electric polarization of a ferroelectric, the Sawyer
Tower measurement method is frequently employed [63] . Herein a sinusoidal
voltage is applied to a series of a known standard capacitor and the ferroelec-
tric sample. The total voltage and the voltage across the sample are monitored,
classically on an two plate oscilloscope. The electric field versus voltage de-
pendence can be derived from the measurement signal. In the simplest case
the electric field in the sample is proportional to the voltage across the sam-
ple, and the polarization is proportional to the input voltage. However, this
requires that the electric field is constant within the sample. The analysis
fails, if this simplifying assumption is not fulfilled. A detailed model for the
response of complex semiconductorferroelectric heterostructures will be dis-
cussed in the present thesis. The measurements presented in the present thesis
were performed by the author, Nurdin Ashkenov and Venkata M. Voora.

2.10 Other techniques


In this work results obtained through other techniques are evaluated and dis-
cussed, which will not be explained in great detail. Dr. Rdiger Schmidt-
Grund, Helena Hilmer, Philipp Khne, Stefan Schche, Chris Sturm and David
Schumacher performed the spectroscopic ellipsometry measurements on the
samples studied in this work. Christoph Meinecke performed the RBS and
PIXE measurements discussed in this work. Dr. Gerald Wagner, Prof. Dr.
Werner Mader, Dr. Herbert Schmid (both Universitt Bonn) obtained the
CBED and TEM images given in this work.

11
2. Experimental Techniques

12
Chapter 3

Physical properties of materials


employed

3.1 Binary ZnO and ternary MgxZn1xO


ZnO crystallizes in the wurtzite structure (Fig. 3.1). Among its remarkable
properties are the direct bandgap at 3.37 eV [64] at room temperature and the
high exciton binding energy of 60 meV [1] . The high bandgap energy results
in transparency of ZnO in the visibe range, making it suitable as a material for
visible-blind photodetectors [65] . The high exciton binding energy in principle
allows the fabrication of highly efficient excitonic devices operating at room
temperature, making ZnO an interesting candidate for UV light emitters. In-
deed LEDs [66] and lasing from ZnO structures [67] was reported. The internal
quantum efficiency of ZnO-based devices can be improved by the formation of
heterostructures with its ternary alloys [68,69] . Both the group II and group
VI constituent can be replaced. Oxygen was substituted successfully by sul-
phur [70] . The focus of this work will solely be on substituting Zn, as the PLD
system employed in all the growth studies is, like most PLD systems avail-
able, excellently applicable for the deposition of oxides, but lacks capability
for the deposition of sulfides, selenides or telurides. In the literature the ma-
terials MgZnO [71] , CdZnO [72] and BeZnO [68] were proposed, even quartanery
BeMgZnO [73] and MgCdZnO [74] were considered. Both MgZnO and BeZnO
show a higher electronic bandgap than pure ZnO, the bandgap of CdZnO is
reduced in comparison to ZnO. The synthesis of CdZnO is problematic because
of the low solubility of Cd in ZnO [74] . Finally, it is important to mention that
Cd falls under the substances controlled by the EUs RoHS directive, making
it an unfavorable material in any application. In contrast to rocksalt MgO,
BeO has the hexagonal wurtzite structure, making it an interesting option for
bandgap engineering. However it is classified as highly toxic in the EU, ex-
cluding its use within the present study. We will therefore rely on the use of
ZnO/Mgx Zn1x O heterostructures in the following. The incorporation of Mg
into the ZnO lattice changes its properties. Over a wide range of Mg contents
a linear change of the properties can be assumed (Vegards rule) with respect to
the Mg content. As MgO and ZnO cyrstallize in different crystal structures, an

13
3. Physical properties of materials employed

Figure 3.1: The two polar orientations of the wurtzite material ZnO: (a) O-face
polarity and (b) Zn-face polarity.

abrupt phase transition is expected at a certain Mg content [75] . At this point


the material properties change instantly. Therefore, the material properties are
not directly a weighted average of these of MgO and ZnO. For a more exact
description a parabolic bowing behaviour of the individual material constant
can be modeled in the continuous part of the individual dependence on the Mg
content x.

3.1.1 Change of the bangap energy


First of all, with increasing Mg content x the bandgap of Mgx Zn1x O increases.
Therefore both the optical absorption and emission spectra are shifted to-
wards higher energies. Numerous experimental data was reported, employing
techniques as PL, spectroscopic ellipsometry and optical absorption measure-
ments [7678] . Whereas the absorption measurements yield the optical bandgap
by an extrapolation or exact modeling of the data in the proximity of a steep
increase in the absorption, PL measurements can give information about the
energetical position of the excitonic recombination. Exemplarily, data on the
luminescence peak position obtained by PL are given in Fig. 3.2. Whereas
the PL technique is more accurate in terms of energetic position, it does not
probe the bandgap directly. The values measured have to be corrected by the
exciton binding energy, which is in principle a function of x, investigated in
detail in [79] . It appears that for low Mg concentration an excitonic spectrum is
observed which resembles that of pure ZnO. With increasing Mg content, the
excitons become localized in minima of spatial potential fluctuations [80] . These

14
3.1. Binary ZnO and ternary Mgx Zn1x O

fluctuations originate from the inhomogenous distribution of Mg atoms in the


lattice which can be purely stochastic or induced by a clustering, as outlined in
Ref. [81] . However, the binding or localization energy of the exciton is a slowly
varying function of x, allowing the interpolation of the PL peak position versus
the Mg content x by a linear or quadratic approximation function.
In order to correlate the PL peak position with the actual Mg content, the
chemical composition has to be determined. The Mg content x was probed
by RBS, as published in [55,82] . However, the error determining the Mg con-
tent in such a measurement is comparably large. From the experimental data
an empirical relationship between the x content and the PL peak position
was obtained and reported in previous studies [55,79,81] . In these three studies
Mgx Zn1x O was investigated, however, the focus was different. While [81] fo-
cuses on Mgx Zn1x O thin films and especially QW structures incorporating
these as a barrier material, the range of x values investigated is limited to
0.05 x 0.25. This range was extended in [55] up to values of x = 0.35,
with a lower bound of x = 0.08. In Ref. [79] samples with comparatively low
Mg content were investigated, including x values between 0.005 and 0.1. The
relations obtained in these studies are plotted in Fig. 3.2. In the present thesis
the empirical relationship given in [55] is used, as the Mgx Zn1x O materials em-
ployed here typically show a Mg concentration of x 0.1. The determination
of the PL peak energy is, especially at room temperature, much easier than the
chemical analysis using RBS. The inversion of the derived empirical relation-
ship allows a fast determination of the Mg content. As the measurement itself
is very accurate, the main source of error remains that of the RBS measure-
ment used for derivation of the empirical standard curve. The PL investigation
therefore represents a valid substitute for the more complicated RBS measure-
ment, once the empirical standard curve is known.

3.1.2 Influence of the Mg content x on the lattice con-


stants
Lattice constants of relaxed Mgx Zn1x O films
An increase in the Mg concentration in the ternary alloy Mgx Zn1x O leads to
a change in the lattice constants. Various authors have evaluated this rela-
tionship for at least one decade. An enormous amount of data was gathered,
however the values reported vary. Several reasons account for this. Most impor-
tantly the changes in the lattice constants are small compared to other material
system as GaN/AlGaN and GaAs/AlGaAs. Much data was obtained by XRD
investigations of Mgx Zn1x O thin films. Having a comparatively small sample
volume, the intensity of the X-ray diffraction peaks is low for such samples.
Crystal imperfections are unavoidable in heteroepitaxial ZnO and Mgx Zn1x O
thin films and broaden the diffraction peaks. The broadening of the peak hin-
ders investigations employing high-resolution X-ray scattering techniques as
the wider angular distribution reduces the peak intensity per unit angle. As
the growth is carried out at high temperatures on not lattice matched sub-

15
3. Physical properties of materials employed

4.4
PL peak energy [eV]

4.2

4.0

3.8

3.6

3.4

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35


Mg content
Figure 3.2: Peak position of photoluminescence of Mgx Zn1x O thin films as taken
from Ref. [55] (red dots) and Ref. [79] (blue triangles). The red line indicates a quadratic
least squares fit taken from Ref. [55] .

strates which even have different thermal expansion coefficients, the films can
be under mechanical stress at room temperature, even if they are relaxed at
the growth temperature. Often, the thickness of the thin films is not sufficient
for a complete relaxation of the film upon cooling.
An excerpt of experimental data reported in literature is given in Tab. 3.1. The
absolute change in the lattice constants per unit of Mg content ka , kc is given.
The contraction of the c-axis constant with increasing Mg content x is appar-
ent from all experiments. Due to some spread in the experimental values, the
four closest results are averaged (Refs. [8386] ), giving a value of kc = 0.1235 ,
which will be used in the following analysis. The situation is more complicated
for the a-axis constant. ZnO and its ternary compounds show a preferential
growth along the c-axis [8790] . Therefore, the a-axis lattice constant can only be
determinated from asymmetric reflexes, which are broadened compared to the
symmetric reflexes, or sometimes are not resolved at all. Two possible values of
the a-axis expansion will be regarded, ka = 0.036 as reported in Refs. [9193]
and ka = 0.05 as reported in Ref. [36] . The value reported in Refs. [84,94] is
omitted, as a shrinking of the a-axis constant is reported, which might be
caused by the different substrate material. From investigations carried out in
the semiconductor physics group in Leipzig an expansion in the a-axis constant
is observed [79,86] . Also, from both works it is evident, that the c-axis lattice
constant is not only a function of the Mg content x, but also of the oxygen
partial pressure during growth. Such an influence cannot be clearly resolved

16
3.1. Binary ZnO and ternary Mgx Zn1x O

Table 3.1: Change of lattice constants with increasing magnesium content for vari-
ous growth methods: molecular beam epitaxy (MBE), radical source molecular beam
epitaxy (rs-MBE), metal-organic vapor phase epitaxy (MOVPE), laser molecular
beam epitaxy (LMBE) and combinatorial laser molecular beam epitaxy (C-LMBE).

Reference Growth Substrate Thickness kc [] ka []


[95]
rs-MBE a-sapphire 700 nm -0.17
[83]
MOVPE c-sapphire 300-700 nm -0.126
[9193]
PLD c-sapphire 300 nm -0.11 0.036
[36]
MBE a-sapphire 500 nm 0.05
[84,94]
MBE Si (111) 400 nm -0.138 -0.05
[85]
C-LMBE c-sapphire -0.12
[86]
PLD a-sapphire 1000 nm -0.11 0.058
single p(O2 ) -0.16 0.058
[96]
LMBE ZnO 200300 nm -0.077
[97]
MBE ZnO 800 nm -0.063
this work PLD ZnO 300 nm -0.064

for the a-axis lattice constant. However, it was demonstrated in Ref. [79] that
both lattice constants of pure ZnO thin films change with the oxygen partial
pressure. With lower oxygen partial pressure during growth the c-axis constant
expands while the a-axis constant shrinks. The actual shift of the lattice con-
stant introduced by the oxygen partial pressure is of the order of the shift due
to the Mg incorporation, which hinders the determination of the coefficients kc
and ka .

Lattice constants of pseudomorphic Mgx Zn1x O films


In contrast to relaxed Mgx Zn1x O films, where a sizeable amount of publica-
tions exist, only few reports are available on Mgx Zn1x O thin films grown on
ZnO substrates [96,97] . As these lattice-matched substrates have both a very
similiar lattice constant and a nearly identical thermal expansion coefficient
pseudomorphic growth of the Mgx Zn1x O layers can be achieved. In that
case, up to a critical thickness the a-axis constant coincides with that of the
substrate, but the c-axis constant differs. As the a-axis lattice constant of
Mgx Zn1x O is larger than that of ZnO, the Mgx Zn1x O will be under com-
pressive strain, which results in an increase of the c-axis. However, the c-axis
constant of relaxed Mgx Zn1x O is smaller than that of ZnO, therefore the c-
axis lattice constant of Mgx Zn1x O is not trivially predictable. In a wurtzite
crystal the strain in an epilayer pseudomorphically grown along the c-axis can
be calculated following Ref. [98] :
C13
0001
= 2 k . (3.1)
C33

17
3. Physical properties of materials employed

MgxZn1-xO cps
a0
ZnO

c0

Figure 3.3: Principle of pseudomorphic growth, unit cells are exemplarily shown to
the right. A Mgx Zn1x O film grown on top of a unstrained pure ZnO substrate adopts
the a-axis lattice constant of the underlying substrate a0 . The film is compressively
strained leading to expansion of the unlocked c-axis lattice constant to the value cps .

Where = (cstrained crelaxed )/crelaxed and k = (astrained arelaxed )/arelaxed and


C13 and C33 are elastic constants, taken to be C13 = 106.1GPa and C33 =
209.5GPa [99] .
Assuming a linear change of the lattice constants with Mg content x

cx = c0 + kc x and ax = a0 + ka x (3.2)

where a0 and c0 are the lattice constants of an unstrained pure ZnO substrate.
This assumption does not necessarily hold true for the hydrothermal substrates
employed in this study due to their comparatively high Li content, resulting
in a slight change of the substrate lattice constant. Therefore only the change
of the film lattice constant with respect to the substrate is monitored. A
Mgx Zn1x O film grown on such a substrate adopts the a-axis constant of the
substrate a0 (see Fig. 3.3),
a0 ax ka x
k = = . (3.3)
ax a0 + ka x
As outlined in 3.1.2 the a-axis constant expands with increasing Mg content
x, therefore the pseudomorphic Mgx Zn1x O film is under compressive strain.
From (3.3) the c lattice constants of a pseudomorphic Mgx Zn1x O layer on
ZnO can be calculated and a quantity kc,ps , the change of the c-axis constant
per unit of Mg content x can be defined:
C13 ka C13 ka
kc,ps = kc + (c0 + kc x)2 kc + 2c0 , (3.4)
C33 a0 + ka x C33 a0
where the approximation in 3.4 is for small x, making kc,ps a constant. It has

18
3.1. Binary ZnO and ternary Mgx Zn1x O

(a) (0 0 2) on O130 Log10(Intensity (counts))


4
3.87
3.5

3.86 3
q (nm )
1

2.5
3.85
2

3.84 1.5

1
3.83
0.5

0.014 0.012 0.01 0.008


1
q (nm )
||

(b) (1 0 4) on O130 Log10(Intensity (counts))


7.7

2.5
7.69
2
q (nm )
1

7.68 1.5

1
7.67
0.5

7.66
3.57 3.575 3.58 3.585
1
q|| (nm )

Figure 3.4: Reciprocal space map (RSM) of the (a) (002) and (b) (104) reflection
of a Mgx Zn1x O thin film containing a ZnO QW grown on a ZnO substrate (O130)
obtained from CrysTec. The color scale represents the log10 of the scattered X-ray
intensity (counts). (Data measured by Dr. Michael Lorenz)

19
3. Physical properties of materials employed

(a) (0 0 2) on TD2 Log (Intensity (counts))


10

3.87
3.5

3.86 3

2.5
q (nm )
1

3.85
2
3.84 1.5

3.83 1

0.5
3.82
2 3 4 5 6
1 3
q|| (nm ) x 10

(b) (1 0 4) on TD2 Log10(Intensity (counts))

7.705 2.5

7.7 2
q (nm )
1

7.695
1.5
7.69
1
7.685
0.5
7.68

3.542 3.543 3.544 3.545 3.546 3.547


1
q (nm )
||

Figure 3.5: RSM of the (a) (002) and (b) (104) reflection of a Mgx Zn1x O thin
film containing a ZnO QW grown on a ZnO (00010 substrate (TD2) obtained from
Tokio Denpa. The color scale represents the log10 of the scattered X-ray intensity
(counts). (Data measured by Dr. Michael Lorenz)

20
3.1. Binary ZnO and ternary Mgx Zn1x O

to be mentioned that even for x 7 1, kc x and ka x are small compared to c0


and a0 , and therefore the error introduced by the approximation are below 2%
for the values of kc , ka , c0 and a0 obtained in the ZnO/Mgx Zn1x O system.
The coefficient kc,ps is determined from the values of the lattice constants given
in Refs. [96,97] and tabulated in Tab. 3.1. The experimental values of the lattice
constants are plotted in Fig. 3.6.
Mgx Zn1x O layers containing ZnO single quantum wells were deposited on the
O-face of hydrothermally grown ZnO wafers obtained from CrysTec GmbH
(CT) and Tokio Denpa (TD), respectively. The properties of the QWs will
be discussed in Chap. 5.4. As the QWs are embedded into the Mgx Zn1x O
layers, it is reasonable to assume that their lattice constant equals that of the
substrate, while the Mgx Zn1x O is under biaxial strain. High resolution X-ray
diffraction experiments (HR-XRD) were carried out and confirm pseudomor-
phic growth. Reciprocal space maps of the (002) and (104) reflections of a
Mgx Zn1x O thin film containing a QW grown on a substrate obtained from
CrysTec (here O130) and Tokio Denpa (here TD2) are shown in Fig. 3.4 and
Fig. 3.5, respectively. The total film thickness of the sample grown on the
CT substrate is around 100 nm, resulting in comparatively weak film peaks,
especially in the asymmetric reflex. Nevertheless, the a-axis constant can be
resolved to equal that of the substrate, proving pseudomorphic growth. The
c-axis constant of both substrate and film was obtained from the (002) reflex.
Despite the total thickness of the film grown on the TD substrate being around
1200 nm and a Mg content of x = 0.175, the film is not relaxed. The critical
thickness of a film of this Mg content was calculated to be 128 nm following
the Matthew and Blakeslee model [100] .
The results obtained from these experiments are compared with values ob-
tained from Refs. [96,97] in Fig. 3.6. A linear fit to the data obtained in the
present study is applied in order to determine the coefficient kc,ps from the
experimental data yielding a value of kc,ps = 0.064 . This is in remarkable
agreement with the value obtained in a fit to the data given in Ref. [97] (see
Tab.3.1). Further the expected change in c-axis lattice constant versus Mg
content x for a relaxed layer and a pseudomorphic film are shown assuming
kc = 0.1235 and ka = 0.036 , again resulting in an excellent agreement
with the experimental data in this work and that reported in Ref. [97] . A value
of kc,ps = 0.065 is calculated following (3.4) for an ka = 0.036 . Using
the value of ka = 0.05 would result in kc,ps = 0.042 , which deviates
from the experimental value by about 30 percent. One might of course argue,
that this can be compensated by a larger change in the c-axis lattice con-
stant with Mg content. A kc = 0.145 would be necessary to give the same
kc,ps = 0.065 for ka = 0.05 . In principle any pair of values solving (3.4)
with kc,ps 0.064 can be regarded. Among these are the values obtained
in Ref. [86] for individual p(O2 ), giving kc,ps = 0.066 , in very good agree-
ment with the lattice constant of the pseudomorphic films. In the following
the values for kc and ka will be limited to 0.16 kc 0.1235 and
0.036 ka 0.058 . This however gives an error bar of 20% of the mean
value for both material constants, and shows the neccessity of more detailed

21
3. Physical properties of materials employed

0.01
A]
Change in c-axis constant [

0.02

0.03

0.04
relaxed
pseudomorphic
Nishimoto et al. Ref (96)
0.05
Matsui et al. Ref (97)
own data
0.06
0 0.1 0.2 0.3 0.4
Mg content x

Figure 3.6: Change in the c-axis lattice constant of Mgx Zn1x O thin films vs. Mg
content x. The solid blue line shows the contraction of the c-axis in a relaxed layer,
while the red line shows the contraction in a pseudomorphic film. and show
values for pseudomorphic layers as published in Ref. [97] and Ref. [96] respectively, +
denotes data points obtained in this study. The solid black line shows the change in
c-axis constant assuming a kc = 0.064 , as obtained from a linear fit to the own
experimental data.

22
3.2. Growth of ZnO and Mgx Zn1x O thin films

investigations on the Mgx Zn1x O lattice constant in relaxed films. Also the
growth of single crystals of homogenous composition and variable Mg content
would be desirable, increasing the sample volume and reducing the peak width
significantly.

3.2 Growth of ZnO and MgxZn1xO thin films


In any material system investigated, the epitaxial growth of the respective ma-
terial has significant advantages over the growth on heterogenous materials.
Among them are the perfect lattice match, leading to a significant reduction of
the density of dislocations upon growth as well as the equal lattice expansion
coefficient. This avoids the risk of the formation of cracks and dislocations upon
cooling from elevated temperatures where the growth process takes place. Fur-
ther, the substrate material equals that of the film, reducing contamination by
diffusion from the substrate. Nevertheless, even under optimal conditions, the
substrate material available is always contaminated by some kind of impurities.
The density and nature of impurities mostly depends on the growth process of
the substrate itself. While for many semiconductors (especially Si) purification
methods like zone melting exist, the application of these methods to ZnO has
not yet been published. In order to understand the advantages over the growth
of ZnO on sapphire a summary of morphologic and structural properties of ZnO
thin films grown on sapphire will be given in Chap. 3.2.1. The properties of
available substrate materials will be reviewed in Chap. 3.2.5. The choice of
substrate material used in the present work will be motivated from comparing
these properties. As differences exist in the growth of ZnO and Mgx Zn1x O,
the growth of Mgx Zn1x O on sapphire is documented in Chap. 3.2.2, while
details on Mgx Zn1x O films grown on ZnO are given in Chap. 3.2.6.

3.2.1 Growth of ZnO on sapphire substrates


Like for GaN sapphire is the most frequently used substrate in ZnO epitaxy.
This is understandable, as both GaN and ZnO are transparent in the visible
range, and the fabrication of transparent devices requires an equally trans-
parent substrate. Sapphire delivers this property. Further the substrate is
insulating, leaving the conductivity of the samples in principle uninfluenced.
It has to be remarked, however, that sapphire contains Al, which is a donor in
ZnO. The influence of interdiffusion between the ZnO film and the substrate is
discussed in Ref. [101] . The crystal structure and the lattice constants of GaN
and ZnO are very similar, (for GaN: c = 0.5185 nm, a = 0.3189 nm). Also the
thermal expansion coefficients of both materials are very close (6.51 106 K1
for ZnO and 5.59 106 K1 for GaN [102] ) compared to that of sapphire with
7.0106 K1 [103] and silicon 2.59106 K1 . Considering these similarities,
GaN has also been proposed as a buffer layer for ZnO growth on sapphire [102] .
It is further easy to understand why sapphire is a favourable choice compared
to the cheaper Si substrate. Within the present study the influence of GaN
buffer layers will not be regarded, as the PLD system employed is solely de-

23
3. Physical properties of materials employed

veloped for deposition of oxide materials. Only ZnO and Mgx Zn1x O layers
grown on sapphire and ZnO substrates will be regarded. The morphology of
the ZnO thin films grown on sapphire will be compared with the morphological
results of GaN films on the same substrate reported in the literature, due to
the similarities outlined above.
Growth experiments were carried out on thermally pretreated a-plane sap-
phire substrates. These are annealed in air at 1200 C. After annealing the
surface is perfectly flat, and atomic steps of 4.8 nm height are visible. Then
the samples are inserted into the growth chamber and heated to the growth
temperature under the oxygen pressure present during growth. A laser power
density of around 1.6 mJ/cm2 is applied to ablate the target. The substrates are
rotated during growth for the fabrication of a film of homogeneous thickness.
A number of 10,000 laser pulses was used for all of the samples discussed in
the following. A final thickness ranging from 140 nm to 210 nm was measured.
Thickness variations can however be observed, due to the following reasons.
First kinetic processes in the PLD plasma are controlled by the chamber pres-
sure and influence the plasma density at the substrate location. Second the
adsortion, reorganization and desorption processes taking place on the film sur-
face are influenced by both the partial pressure of the film constituents and the
temperature of the film surface. Finally, the oxygen partial pressure controls
the II/VI ratio during the deposition process. Within the experiments carried
out, the oxygen partial pressure and the substrate temperature were varied.
Figure 3.7 summarizes results obtained by AFM on films grown at 650 C and
700 C. Four samples are shown, which were grown at (a) 700 C and p(O2 ) =
0.1 mbar; (b) 700 C and p(O2 ) = 0.002 mbar; (c) 650 C and p(O2 ) = 0.1 and
(d) 650 C and p(O2 ) = 0.002. All of the films deposited clearly reveal mono-
layer steps at the surface. However the density of surface defects is different.
Especially for the films grown at 700 C large hexagonal pits are observed on the
surface. The density of these pits is counted in the 2 m2 m AFM images,
and is compared in Fig. 3.7 (e). It can be seen that the pit density is always
lower for samples grown at 650 C, and that for this temperature the density of
pits increases with increasing oxygen partial pressure. Also, if pits are present
they are always smaller in size compared to pits found in the samples grown
at 700 C. Figure 3.7 (f) shows again the density of the pits along with the
thickness of the thin films as determined by spectroscopic ellipsometry mea-
surements. The film thickness increases by 15% within the range of oxygen
partial pressures.
The presence of pits is well known from heteroepitaxial growth of GaN, and
ascribed to multiple origins. In general, all ascriptions rely on dislocations pen-
etrating the film, and emerging at the pit position. Such defects have influence
on the formation of heterostructures in the films and can lead to significant
interface degradation. Pit-like defects originating from threading dislocations
have been claimed to be responsible for the so called V-defect in InGaN/GaN
multiple quantum wells leading to a significant broadening of the luminescence
observed [104] . Therefore surface defects have been investigated in the GaN sys-
tem in great detail.

24
3.2. Growth of ZnO and Mgx Zn1x O thin films

4.0 210
1.0
3.5
200
3.0 0.8
Film thickness (nm)

Pit density (m-2)


Pit density (m-2)

2.5 190
flach 700 0.6
2.0
flach 650 180
1.5
0.4
1.0 170
0.5 0.2
160
0.0
0.0
-0.5 150
0.001 0.010 0.100 0.001 0.010 0.100
p[O2] (mbar) p[O2] (mbar)

Figure 3.7: 2 m2 m AFM images of ZnO thin films grown on a-plane sapphire.
The substrate temperature is 700 C for (a), (b) and 650 C for (c), (d). The oxygen
partial pressure was 0.1 mbar for (a), (c) and 0.002 mbar for (b), (d). (e) Pit
density vs. oxygen partial pressure as obtained from the AFM images. (f ) Pit
density and film thickness as determined by spectroscopic ellipsometry vs. oxygen
partial pressure for samples grown at 650 C.

25
3. Physical properties of materials employed

Figure 3.8: Example for double spirals and islands present on one ZnO thin film
grown in sapphire. Black circles indicate the double spiral mounds containing a
threading dislocation with a screw component parallel to [0001], white encircled
mounds are islands without a dislocation. (Courtesy of Gregor Zimmermann)

Heying et al. [105] regarded the surface tension introduced by the dislocation
line, following Burton, Cabrera and Frank [106] and Frank [107] . GaN films were
grown by MBE on MOCVD grown GaN templates. Whereas the MOCVD tem-
plates show straight steps and two types of depressions, hexagonally shaped
spiral hillock without depressions are observed in the MBE grown films. The
depressions are assigned to two types of dislocations. Small depressions (di-
ameter 36 nm, depths 0.6 nm) are assigned to pure edge type dislocations
(Burgers vector b = 1/3[1120]) while larger depressions (diameter 62 nm,
depths 1.8 nm) are assigned to mixed type dislocations (b = 1/3[1123]).
This is supported by two steps being terminated by the larger depressions. It
was argued that the presence of depressions is less likely in an non-equilibrium
growth process like MBE (and even more PLD, which is of importance for the
present work), and in such a process otherwise straight steps will wind around
threading dislocations, leading to the formation of hexagonal hillocks. How-
ever, the critical radius is a function of the III/V ratio and depressions were
also observed for a III/V ratio slightly larger then 1. The same authors pointed
out in another paper (Ref. [108] ) that the size of depressions can increase with in-
creasing film thickness, however no conclusions were drawn about the origins of

26
3.2. Growth of ZnO and Mgx Zn1x O thin films

this process. Scanning tunneling microscopy investigations revealed [109] , that


the depressions terminating edge type dislocations with Burgers vector b =
1/3[1120] have a triangular shape. The base of the depression is an isosceles
triangle of 7 a-axis constant edge length with one side reduced to 6 a-axis
edge length, due to the dislocation. The base of the depression is surrounded
by elevated plateaus of an equally triangular shape with its edge length being
integer powers of two of the base edge length. It is interesting to note, that
among these values 32 nm and 64 nm can be found, being close to the
size of the small and larger depressions found in Ref. [105] . Liu et al. [110] have
shown that among the dislocations with a screw component the distribution
of Burgers vectors b = j[0001] is not uniform. Each value of j is represented
by a different number of monolayer steps emerging from the dislocation at the
surface. Values of j equaling 1/2 (one step emerging) and 1 (two steps emerg-
ing) are found frequently, but even up to j = 5 (10 steps emerging from the
dislocation) are present in the samples in that study. According to Frank [107]
screw dislocation in equilibrium can have an open core. The size of this void
is given by
b2
r= 2 , (3.5)
8
based on the surface tension introduced by a dislocation in a fluid-like mate-
rial [107] . Here is the rigidity modulus and is the surface energy. Thereas,
the size of the open core is proportional to the square of the magnitude of the
Burgers vector. However Northrup et al. [111] argued, that giving the material
constants of GaN, the size of this open core dislocations should not exceed
3 for a Burgers vector with a screw component of [0001] while experiments
revealed sizes of around 60 nm [112,113] . The authors therefore theoretically in-
vestigated a mechanism leading to a reduction of the GaN(1010) surface energy
by hydrogen termination of dangling bonds stabilizing large core diameters. It
is remarked, that the pits observed in our experiments can reach diameters of
up to 100 nm. Cui and Li [114] discussed the possibility of merging of disloca-
tions with a screw component, forming double spirals or closed loop spirals
and thereby reducing the overall dislocation density with increasing film thick-
ness. In their paper they further argue, that a double spiral should grow faster
in height than a single spiral. Therefore, given a sufficiently large film thick-
ness, only double spirals would be visible on the sample. Figure 3.8 shows a
2 m2 m AFM image of a ZnO sample grown on a-plane sappire of roughly
1 m thickness. The sample is grown at 700 C and p(O2 ) = 0.002 mbar. Dou-
ble spirals, comparatively large pits and mounds without a spiral-like shape
are visible on one sample. The growth speed of mounds without a threading
dislocation almost equals that of spiral mounds, even after a film thickness of
1 m no distinct height differences are seen for both types of mounds.
This is somewhat unexpected, as other studies have shown that spiral mounds
should, given sufficiently large adatom concentration, grow always faster than
mounds without spirals [115] . It has to be mentioned however, that the authors
investigated Pt thin film growth, a comparatively simple material system. In
ZnO additional complications in a theoretical prediction occur from the inter-

27
3. Physical properties of materials employed

play of its two constituents and the polar nature of the material. Further, the
non equilibrium growth conditions introduced by the PLD process provides a
certain kinetic energy for the adatoms, which gives them the ability to over-
come energetic barriers forming at the step edges, at least to some extend,
and therefore increases the speed of island growth. In order to further un-
derstand the formation of pits, and their comparatively large size seen in this
study, a growth evolution study was carried out. Therefore films were grown
at 700 C and p(O2 ) = 0.016 mbar, a condition resulting in a high density of
large pits. A number of 100, 300, 1,000, 3,000 and 10,000 pulses was used for
deposition. AFM measurements of the surface morphology were carried out
and are shown in Fig. 3.9. For 100 pulses (Fig. 3.9 (a)) the steps of the sap-
phire substrate are still observable, and only a few point shaped hillocks are
visible on the film. The density of these hillocks increases after the deposition
of 300 pulses (Fig. 3.9 (b)), but the steps on the sapphire substrates are still
observable. The hillocks are separated, a coherent film growth has not yet
started. After depositing 1,000 pulses (Fig. 3.9 (c)) beginning coalescence of
neighboring hillocks is observed. The surface contains a significant fraction
of holes which most likely originate at the sapphire substrate. Note that the
height scale of Fig. 3.9 (c) is 6.4 nm, the height determined by spectroscopic
ellipsometry instead is 10.5 nm (see Fig. 3.9 (f)). It is known that AFM mea-
surements underestimate the depth of narrow features due to the tip curvature.
It is therefore well possible that the pits go down to the substrate. After 3,000
pulses (Fig. 3.9 (d)) coalescence of the growth seeds has progressed, and uni-
form terraces of width of several tens to hundreds of nm are visible. Still a
comparatively high density of pits is observed. For the film thickness attained
it is impossible to judge whether the pits penetrate the film down to the sub-
strate surface or whether they are at least partially filled with material. For a
number of 10,000 pulses deposited (Fig. 3.9 (e)) terraces are clearly observed
on the film, and separated by steps of c/2 and c. The width of the terraces
is of the order of tens to hundreds of nm, and depends largely on the miscut
of the substrate chosen. The thickness of the films shown in Fig. 3.9 (a)(e)
was measured by spectroscopic ellipsometry and is depicted versus the number
of laser pulses in Fig. 3.9 (f). Along with the experimental data a thickness
extrapolation is given assuming a strictly linear dependence of film thickness
on the number of laser pulses. It can be seen, that the film thickness remains
smaller than the extrapolated value in the initial stage of the film growth, but
then advances to the expected value with a steeper slope than in the extrapo-
lation. Assuming that the growth after 10,000 pulses represents a steady state
condition or being at least very close to this, the following conclusions can be
drawn:

The growth of ZnO thin films starts in a nucleation process. The growth
rate during nucleation is reduced compared to the steady state growth,
resulting in an offset of the experimental thickness in the initial stages of
the growth towards lower film thickness. This nucleation is observed in
Fig. 3.9 (a) and (b).

28
3.2. Growth of ZnO and Mgx Zn1x O thin films

100.0
Film thickness (nm)

10.0

1.0

0.1
100 1,000 10,000
Number of PLD pulses

Figure 3.9: 2 m2 m AFM images of ZnO thin films after (a) 100, (b) 300, (c)
1,000, (d) 3,000 and (e) 10,000 pulses. The thickness of the films, as determined by
spectroscopic ellipsometry, is depicted in (f ). The solid line shows the extrapolated
thickness assuming a steady state growth after 10,000 pulses.

29
3. Physical properties of materials employed

Once the growth seeds are formed they start growing in a 3-dimensional
growth mode. Seeds gather material from the surrounding area, and pref-
erentially grow in the c direction. This is evident by the steeper increase
of the film thickness in comparison to the steady state growth. The pref-
erential growth of ZnO along this direction is known from investigations
of single crystals, thin films and nanostructures [8790] . It is possible to
grow nanowires with high aspect ratios [116] under adequate growth condi-
tions. During this process further nucleation might take place, depending
on the diffusion length of the adatoms on the substrate. This might fur-
ther be a reason why many experimenters show an improvement in ZnO
growth if a low temperature buffer layer is introduced during the growth
process. Regarding the growth evaluation presented in this study a low
temperature will lead to low diffusion length of ZnO adatoms on the sub-
strate, increasing the density of growth seeds. Further, the adatoms, as
gathered from the substrate will arrive at a side facet of the growth seed.
The surface energy is minimal on the c-plane, but has a local minimum
on the m-plane [117] . Upon a lower temperature the probability decreases
that an adatom arrives at the lowest point in energy space, but rather
sticks to a local minimum. The preference of c-axis orientated growth
of the seeds will therefore be not as pronounced as at higher tempera-
tures. This will lead to an increased in-plane growth and speeds up the
following growth step.

During 3-dimensional growth the seeds also expand in the in-plane di-
rection, leading to the inevitable coalescence of the seeds. This process
is fast in the beginning, but slows down as more and more of the surface
is covered by ZnO. If further adatoms are adsorbed onto the ZnO, they
will reorganize to the c-planes facing upwards, as these are the surfaces
of minimal energy. However on this surfaces step edges represent even
lower energy minima, due to the higher number of bonds they provide to
the adatoms. In this step the growth mode changes from 3-dimensional
to 2-dimensional. The actual nature of the 2-dimensional growth now
depends on the density of dislocations, the availability of the ZnO con-
stituents and the temperature. As for most applications a flat surface
is advantegeous, step-flow growth mode is desired. This growth mode
replaces layer by layer growth if the temperature is sufficient that the dif-
fusion length of adatoms exceeds the average terrace width [118] . At this
point a low growth temperature would be disadvantegeous, giving a limit
to the usability of the low temperature buffer. It has to be remarked,
that control of the 2-dimensional growth mode is not only possible by
the temperature but also by the miscut of the substrate, which reduces
the average step width. It has indeed been reported in the GaN system
that growth on vicinal substrates improves the surface morphology [119] .

As outlined, the transition between 3-dimensional and 2-dimensional


growth is not abrupt. Parts of the sample will be growing 3-dimensionally,
thus in out-of-plane and in-plane direction simultaneously, while other

30
3.2. Growth of ZnO and Mgx Zn1x O thin films

parts show step mediated 2-dimensional growth. Therefore pits can be


present on the sample up to comparatively large thicknesses and essen-
tially represent voids which go down towards the substrate. It can of
cause not be excluded that the pits are partially filled with material,
but this seems rather unlikely, regarding the tendency of the adatoms to
reach positions of minimal energy. Further the steeper increase in thick-
ness with increasing number of laser pulses in the 3-dimensional growth
regime compared to the steady state condition suggests that a part of
the film has to be unfilled, as the steady state condition represents the
growth rate at complete layer filling. The reduction of the lateral pit size
is most likely induced by a form of lateral overgrowth slowly reducing the
diameter. It can be stated, that most of the pits observed in this study
are not of the same origin as the pits observed in the GaN system. It
can of course not be concluded, that upon closure of a pit a dislocation
remains due to different orientation of the growth seeds surrounding the
pit, however the dislocation is not the main driving force for generating
a hollow structure within the film.

3.2.2 Growth of Mgx Zn1x O on sapphire substrates


As the aim of the present study is the investigation of the effects of the electric
polarization on carrier transport and recombination properties, for example in
ZnO/Mgx Zn1x O heterostructures, the growth of Mgx Zn1x O thin films is an
important issue. The same argumentation for the choice of sapphire as sub-
strate holds as it does for pure ZnO. The lattice constants vary only slightly
(see Chap. 3.1.2), which should in principle result in very similar growth con-
ditions. This assumption was extensively tested in Ref. [81] , and will therefore
not be discussed in great detail here. However a short summary will be given,
along with some extended explanations on the growth mechanism exceeding
these in Ref. [81] . Fig. 3.10 shows AFM images obtained on four representative
Mgx Zn1x O thin films. A ZnO QW of variable thickness is included in all of
the samples. A comparison between layers with and without a QW however
does not show a qualitative difference in surface morphology. Two of them
(depicted in Fig. 3.10 (a) and (b)) are directly deposited onto sapphire, while
for the other two (depicted in Fig. 3.10 (c) and (d)) a roughly 100 nm thick
ZnO buffer layer was introduced. The Mg content x is given in the caption.
The thickness of the films is intentionally unhomogeneous, as described in
Chap. 5.4. The thickness of the layers however, exceeds 100 nm even for the
thinnest part of the film. Therefore it is reasonable to assume that a steady
state growth is reached, as described in Chap. 3.2.1. For the sample with the
lowest Mg content (x = 0.17) directly grown on sapphire (Fig. 3.10 (a)) steps
of c/2 height are visible. The morphology however is clearly distinguished from
that of pure ZnO thin films. Even as the film is completely closed rather steep
hillocks are observed on the sample. The growth mode is clearly 2-dimensional,
however a transition to island growth is observed. The width of steps is of the
order of tens of nm, an exciton localized in a quantum well with a radius of

31
3. Physical properties of materials employed

Figure 3.10: 2 m2 m AFM images of Mgx Zn1x O thin films grown on sapphire
substrates. The growth temperature was about 650 C, the oxygen partial pressure
0.004 mbar. The Mg content x was determined by RT PL to be (a) 0.17 (b) 0.28 (c)
0.3 and (d) 0.32 respectively. The samples shown in (a) and (b) are grown directly
onto sapphire, while for (c) and (d) a 100 nm ZnO buffer layer was used.

roughly 2 nm will therefore not face a variation of the thickness more than a
single step. It is however not clear, if the thickness of a QW is homogeneous
over the whole sample, as parts of the hillocks might grow faster than for ex-
ample valleys between them, resulting in an inhomogeneous distribution of the
ZnO between the Mgx Zn1x O layers. This might contribute to a broadening
in the QW luminescence of the sample. Figure 3.10 (b) shows a sample with
an Mg content x = 0.3. The surface of the sample is very rough and no atomic
steps can be observed. The growth is clearly 3-dimensional, and no features
corresponding to hexagonal structures can be observed. It can be speculated
that the film is highly disordered due to the Mg concentration being signifi-
cantly above the solution limit in equilibrium (2%) [120] . XRD measurements
however revealed, that the sample crystallized in the wurtzite structure and
is solely c-axis oriented. Figure 3.10 (c) shows a thin film of equal Mg con-
tent x deposited onto a approximately 100 nm thick ZnO buffer layer. The

32
3.2. Growth of ZnO and Mgx Zn1x O thin films

morphology resamples that of ZnO thin films grown on sapphire. Steps of c/2
height are visible. Two lines are marked in the AFM image, labeled as A and
B edge, respectively. The labeling is adopted from Ref. [121] , where both kinds
of step edges have been observed in the GaN system. The origin of this phe-
nomenon is due to the fact that despite having a hexagonal symmetry in the
arrangement of atoms, the tetrahedral coordination of ZnO or GaN results in
an angular separation of the in-plane bonds by 120 . The {1010} family of
directions therefore splits into two subfamilies, one of them being parallel to a
bond (perpendicular to a B edge, from now on called B direction), the other
one being centered between two bonds (perpendicular to an A edge, from now
on called A direction). If a step propagates into the A direction two dangling
bonds are available for adatoms, enhancing the growth speed. In contrast a
step propagating into the B direction offers only one dangling bond to adatoms
leading to a lower growth speed. As an A edge propagates, it tends to facet
into B egdes, leaving a zigzag shape behind, as visible in Fig. 3.10 (c). As the
wurtzite lattice has a binary stacking sequence, the arrangement of bonds is
rotated by 180 in every successive layer. Necessarily A and B edges have to
alternate. The B edge neighboring the A edge is marked likewise, a strait edge
without remarkable facets. Double spirals are partially observed, however due
to the vicinal substrate propagating steps dominate the surface. As no island
formation takes place, and no hillocks are observed apart from that of spiral
winding around dislocations with a screw component, it is concluded that the
sample grows in the step flow mode. This is an improvement compared with
the growth of pure ZnO and Mgx Zn1x O directly on sapphire, which both tend
to form islands in the layer by layer growth mode. As both temperature and
miscut of the substrate are roughly equal, the origin of this behavior has to be
clarified. A possible explanation is the reduction of the surface diffusion barrier
by compressive strain, as has been experimentally [122] and theoretically [123,124]
shown in the literature. As the a-axis constant of Mgx Zn1x O is larger than
that of ZnO the introduction of a ZnO buffer layer might give rise to the step
flow growth of the Mgx Zn1x O layer via the strain formed within. Figure 3.10
(d) shows an AFM image of a Mgx Zn1x O thin film with x = 0.32 grown on
a ZnO buffer layer. The film shows a number of pits, and additionally double
spirals are visible. Also a closed loop spiral can be seen, indicative for partial
annealing of dislocations with a screw component in this film.

3.2.3 ZnO substrate material


In order to choose ZnO substrates suitable for epitaxy, the properties of avail-
able substrate material will be reviewed in the following. Three major growth
techniques are used nowadays for the production of ZnO single crystals: growth
from the vapor phase, from the melt and from aqueous solution. All of these
processes are carried out at elevated temperatures and pressures, each of them
having its advantages and disadvantages. While the cleanest material can be
grown from the vapor phase [125,126] , the throughput of this process is very lim-
ited, resulting in high prices of this substrate material. Recently, the availabil-

33
3. Physical properties of materials employed

Table 3.2: Suppliers and growth methods of ZnO single crystals investigated. Some
of these crystals were investigated by Holger von Wenckstern, which is denoted by
the abbreviation HvW. SCVT stands for seeded chemical vapor transport, while IKZ
denotes the Institut fr Kristallzchtung in Berlin. Prices are given if applicable.

Label Supplier Method Ref. Price Data


VP1 Eagle-Picher SCVT [125]
n.a. HvW
VP2 Zn-Technologies SCVT [126]
n.a. HvW
MG1 Cermet Inc. Skull melting [127,128]
300$
MG2 IKZ, Berlin Bridgeman [129131]
n.a
MG3 IKZ, Berlin Bridgeman [129131]
n.a
HT1 Mateck GmbH Hydrothermal [132]
n.a HvW
HT2 CrysTec Hydrothermal [133]
50 e
HT3 Tokyo Denpa Hydrothermal [134]
300 e

ity of the material also became very limited, removing that kind of substrates
virtually from the free market. The purity of melt grown substrates is not
as high as that of vapor phase material [128] , and lattice imperfections are a
big issue in this material [130,131] . For the crystals grown from aqueous solution
it is important to note that solving agents, so called mineralizers have to be
added to the material [88,135,136] . For ZnO these mineralizers are typical group
I-hydroxides, especially KOH and LiOH are very popular [88,135,136] . While the
crystalline quality of these materials can be very high [88] , a contamination of
the substrate by the mineralizer is unavoidable. This has a significant influence
on the properties of the crystals. The free carrier concentration and mobility
of selected single crystals is displayed in Fig. 3.11 respectively. The labels in
the graphs are given in Tab. 3.2 along with information on the supplier, growth
technique and pricing, if available. The crystals labeled MG2 and MG3 were
kindly provided by the Institut fr Kristallzchtung in Berlin. Some of the
crystals have been measured by Holger von Wenckstern, which is also denoted
in Tab. 3.2. Reviewing the transport data, the mobility is the highest in the
vapor-phase grown samples and the commercially available melt-grown sam-
ple. The free carrier concentration exceeds that of the hydrothermally grown
samples, but is below that of the samples MG2 and MG3. The latter can be
easily explained by the purity of the powder employed in the growth process,
being 99.99% only [131] , and the incorporation of intrinsic donors due to oxygen
deficiency during growth. Annealing experiments could therefore somewhat
reduce the free carrier concentration and increase the room temperature mo-
bility [129] . It has further been shown, that the crystals were suitable for epitaxy
of pseudomorphic Mgx Zn1x O thin films of high structural quality [137] . Nev-
ertheless, these cyrstals are not suitable to serve as a substrate in the present
study as the observation of 2-dimensional transport in the samples would be
concealed by the high conductivity of the substrate. Similar arguments hold
for the substrate MG1 and the substrates grown from the vapor phase. In all

34
3.2. Growth of ZnO and Mgx Zn1x O thin films

1017

1015
n (cm -3)

1013

VP1 HT1
VP2 HT2
1011 MG1 HT3
MG2
MG3

109
0 5 10 15 20 25 30 35 40 45 50
1000/T (1/K)

1.000

VP1 HT1
VP2 HT2
H (cm 2/Vs)

MG1 HT3
MG2
MG3

100
30 50 70 100 200 300
T (K)

Figure 3.11: Free electron concentration and mobility in various ZnO substrates
versus temperature. Labels are explained in Tab. 3.2. Some of the crystals were
measured by Holger von Wenckstern, as denoted in Tab. 3.2.

35
3. Physical properties of materials employed

of these crystals the conductivity is greater than 103 1 m1 , and therefore


not suitable for the utilization in this study. The situation is different for the
crystals grown from aqueous solution. The main impurities in these crystals
are the group I elements Li and K [88,138] , as far as data is available. If these
elements are incorporated on Zn site they act as acceptors and reduce the free
electron concentration in the samples significantly [139] . It is however notewor-
thy, that these elements might also be introduced on interstitial site [140,141] .
This fact has been mentioned as one possible explanation for the difficulties in
obtaining p-type ZnO by the doping with group I elements, despite the high
concentrations incorporated.
Regarding the single crystals investigated in this study two different regimes
are found. The crystals HT1 and HT2 show a very low free carrier concentra-
tion (1013 cm3 and below) and are only accessible to Hall-effect investigation at
temperature above 180 K. Below this value the conductivity of the substrates is
virtually zero. Further, by an annealing step, which is required in the substrate
preparation steps prior to epitaxy (see Chap. 3.2.4), the resistivity increases
even further. These substrates are highly suitable for electrical investigations
of the films thereon, as they do not contribute to the overall conductivity. How-
ever, care has to be taken regarding sample HT3. As the composition varies
between individual samples (a detailed investigation is given in Ref. [88] , Fig. 13)
the conductivity might be higher. Therefore prior to the growth the conductiv-
ity was measured and only substrates with very low conductivity are employed
in the present study, so influence of the substrate conductivity can be neglected
in all following considerations. For the experiments two types of substrates have
been used: substrates provided by CrysTec GmbH, Germany, and substrates
provided by Tokio Denpa Inc., Japan. The quality of the substrates obtained
from both suppliers will be discussed in the following section.

3.2.4 ZnO substrate preparation


As outlined in the previous section, substrates grown by the hydrothermal
method are the most promising candidates for the investigation of conductive
ZnO thin films, as these substrates do not produce an additional conductive
channel in parallel to the thin film in question. This is ideal for the determi-
nation of the carrier concentration and mobility in the thin films, for example
in doping experiments. Despite these advantages, the number of reports re-
garding successful homoepitaxy are very limited (see Fig. 9.1 in Ref. [47] ). One
particular reason for that is the state the substrates are delivered in. It was
shown, that the surface of the crystals is covered by a damaged layer, which
extends into the bulk material up to 1 m [142,143] . It has further been argued,
that this damage is introduced by the polishing process the vendors apply to
the substrates. Indeed, significant scratches have been observed by AFM in-
vestigations on the substrate surfaces [47] .
In an attempt to improve the substrate quality chemical cleaning, etching and
thermal annealing experiments have been carried out [47,142,144,145] . It is shown,
that the oxygen-face of the substrates recovers upon annealing above 1,000 C

36
3.2. Growth of ZnO and Mgx Zn1x O thin films

TD5 Oface (0 0 2) Log10(Intensity (counts)) TD5 Oface (1 0 1) Log10(Intensity (counts)) TD5 Oface (1 0 4) Log10(Intensity (counts))
3.855 4.05 3 7.695
2.5
3.5
3.85 2.5
3 4.045 7.69
2
3.845 2.5
q (nm )

q (nm1)

q (nm1)
2
1

4.04 7.685
1.5
3.84 2
1.5
4.035 7.68

1.5 1
3.835 1
1
4.03 7.675 0.5
3.83 0.5
0.5
3.825 4.025 7.67
0.02 0.018 0.016 0.014 8 6 4 2 0 2 3.535 3.54 3.545 3.55 3.555
1 1 3 1
q|| (nm ) q|| (nm ) x 10 q|| (nm )

(a) (b) (c)


O64 Oface (0 0 2) Log10(Intensity (counts)) O64 Oface (1 0 1) Log10(Intensity (counts)) O64 Oface (1 0 4) Log10(Intensity (counts))
3.855 4.05 7.695
3.5
2.5
3.85 2
3 4.045 7.69

3.845 2.5 2
q (nm )

q (nm1)

q (nm1)
1

4.04 7.685 1.5


3.84 2 1.5
4.035 7.68

1.5 1
3.835 1
1
4.03 7.675 0.5
3.83 0.5
0.5
3.825 4.025 7.67
2 0 2 4 6 4 2 0 2 4 6 3.535 3.54 3.545 3.55 3.555
1 3 1 3 1
q|| (nm ) x 10 q|| (nm ) x 10 q|| (nm )

(d) (e) (f)


O65 Oface (0 0 2) Log10(Intensity (counts)) O65 Oface (1 0 1) Log10(Intensity (counts)) O65 Oface (1 0 4) Log (Intensity (counts))
10
3.855 4.05 7.695
2.5
3 2
3.85
4.045 7.69
2.5 2
3.845
q (nm )

q (nm1)

q (nm1)

1.5
1

2 4.04 7.685
1.5
3.84
1.5 4.035 1 7.68

3.835 1
1
4.03 0.5 7.675 0.5
3.83
0.5

3.825 4.025 7.67


8 6 4 2 10 8 6 4 2 0 2 3.535 3.54 3.545 3.55 3.555
q|| (nm1) q|| (nm1) q|| (nm1)
3 3
x 10 x 10

(g) (h) (i)

Figure 3.12: Comparison of reciprocal space maps of the (from left to right) (002),
(101) and (104) reflexes obtained on 3 different substrates, namely (from top to
bottom) a TD substrate and two selected CT substrates. The color scale represents
the log10 of the scattered X-ray intensity (counts). (Data obtained by Dr. Michael
Lorenz)

37
3. Physical properties of materials employed

in an oxygen atmosphere [144148] . Similar improvement was found for other sur-
faces [145,147,149] . The substrates employed in this study are typically of oxygen-
face polarity.
HR-XRD measurements carried out on the substrates reveal that the struc-
tural quality of the substrates is varying. A comparison of reciprocal space
maps obtained from the samples in the as delivered state is given in Fig. 3.12.
Substrates obtained from TD show very narrow peaks. No splitting of the
symmetric and the asymmetric reflexes is observed for these substrates, as
shown in Fig. 3.12 (a)(c). In contrast the structural quality of the CT sub-
strates varies strongly. Substrates with structural properties similar to the TD
substrates have been measured, however most of the material shows inferior
quality in the as delivered state. Two examples are shown in Fig. 3.12 (d)(f)
and Fig. 3.12 (g)(i). As reported in the literature [144] , the structural proper-
ties of annealed substrates improve significantly.
Further the surface morphology of the substrates improved, as also reported
in Refs. [144148] . In order to illustrate this, AFM height images from substrates
used in this study are shown in Fig. 3.13. The as-received state is shown in
Fig. 3.13 (a). The substrate shows a very rough surface. After annealing for
2 hours in 700 mbar oxygen at a temperature of 1,000 C the surface of the
substrates shows steps of c/2 height, as depicted in Fig. 3.13 (b)(d). The
morphology of the substrates can be different, due to the miscut of the sub-
strates. A substrate with a very small miscut is shown in Fig. 3.13 (b). The
few steps available due to the non zero miscut, however, tend to meander and
form small islands and holes of c/2 height on the substrate. In contrast, the
substrate shown in Fig. 3.13 (c) has a miscut of approximately 0.3 , resulting in
a vicinal surface with parallel step edges. It is interesting to note, the signs of
scratches from the polishing process have remained on the substrate, meaning
that nearly no ZnO material has evaporated from the surface at the temper-
atures used in the annealing. Some of the substrate show surface defects (as
depicted in Fig. 3.13 (d)), which might be connected to crystal defects extend-
ing into the bulk of the substrate. The different morphological and structural
properties of the substrates influence the thin films grown on them, which will
be extensively discussed in the following sections.

3.2.5 Growth of ZnO on ZnO substrates


Following the approach employed in the study of heteroepitaxial growth on sap-
phire, measurements of the morphology of homoepitaxially grown ZnO films
were carried out. Therefore undoped ZnO layers were deposited onto the O-
face of CT crystals. Figure 3.14 shows homoepitaxial ZnO thin films grown
at different temperatures and oxygen partial pressures. Resulting AFM height
images of size 2 m1 m are summarized in Fig. 3.14. For low growth temper-
ature (470 and 550 C, as shown in Fig. 3.14 (f) and Fig. 3.14 (e) respectively)
a 3-dimensional growth is observed. For Tg = 470 C nanocrystallites with a
size of 30 nm form. Increasing the growth temperature to 550 C leads to a
coarsening of the crystallites due to the increased diffusion length of adatoms

38
3.2. Growth of ZnO and Mgx Zn1x O thin films

Figure 3.13: 2 m2 m AFM height images of ZnO substrates purchased from


CrysTec GmbH. (a) without thermal annealing, (b)(d) after thermal annealing.
Different substrate miscut can be seen in (b) and (c). Extended crystal defects were
observed on some of the substrates as depicted in (d).

39
3. Physical properties of materials employed

p(O2) = 0.1 mbar

Figure 3.14: 2 m1 m AFM height images of homoepitaxial ZnO thin films. All
films are grown on the O-face of CT crystals. The oxygen partial pressure was (a)
0.1 and (b)(f ) 0.016 mbar. The substrate temperature was set to (a) and (c)
650 C; (b) 700 C; (d) 610 C; (e) 550 C and (f ) 470 C. (Measurements performed
by Heidemarie Schmidt (now Forschungszentrum Rossendorf))

on the surface. The average size of the crystallites is about 400 nm. Begin-
ning at a temperature of 610 C (visualized in Fig. 3.14 (d)) a transition into a
2-dimensional growth mode occurs. While the lateral dimension of surface fea-
tures reduces slightly 250 nm compared to the film grown at 550 C the height
of the features is dramatically reduced. Hexagonal symmetry is observed, and
even c/2 steps can be seen. The surface is still somewhat roughened by island
nucleation of the next layers to be formed, making the overall image seeming
rough. This changes as the temperature is raised to 650 C (Fig. 3.14 (c)).
Hexagonal hillocks with a average diameter of 400 nm are observed, defined by
steps of c/2 height. The average width of terraces is 30 nm. Increasing the tem-
perature further to 700 C (Fig. 3.14 (b)) results in both an increase of hillock
diameter and terrace width, indicative for the increase in the surface diffusion
length. Also the strictly hexagonal symmetry of surface features breaks down,
as neighboring islands coalesce. Further increase of the temperature would
increase the island size even more, and lead to a reproduction of the stepped
terraces of the substrates by the thin films. Interestingly the island size also

40
3.2. Growth of ZnO and Mgx Zn1x O thin films

increases upon a change in oxygen partial pressure from p(O2 ) = 0.016 mbar to
p(O2 ) = 0.1 mbar (compare Fig. 3.14 (a) and (c)). An average hillock diameter
of 700 nm is reached, along with a terrace width of 80 nm. It has to be pointed
out, that changing the oxygen partial pressure influences the film growth in
more then one way. The motion of particles in the plasma plume is influ-
enced by the pressure inside the growth chamber, as is the stoichiometry of the
film. As oxygen atoms are much lighter then Zn atoms (16 amu/65.4 amu) the
scattering probability for oxygen increases with respect to zinc atoms as the
oxygen partial pressure increases. Therefore during the comparatively short
travel time of the plasma, the surface of the growing film might be subject to
excess zinc. As all of the films are grown in the oxygen polarity (as confirmed
by convergent beam electron diffraction (CBED), compare Fig. 3.17), only a
single oxygen bond points upwards from completely filled layer. Under slightly
oxygen deficient conditions, arriving zinc atoms will therefore be only weakly
bonded and have an increased surface diffusion length. Following the 5ns laser
pulse and the traveling time of the plasma a large window is available for a
complete oxidation of the forming layer. Such a behavior would lead to an im-
provement of the surface morphology. Despite being a possible scenario, this
mechanism is rather speculative, as the processes involved are not yet com-
pletely understood. It is worthwhile to investigate the influence of the oxygen
partial pressure on the thin films grown, therefore, for example in the doping
experiments reported in Chap. 4 series of samples are grown under different
oxygen partial pressures, leaving all other parameters constant.
The structural properties of homoepitaxial ZnO thin films have been studied
by HR-XRD, TEM and HR-TEM. Reciprocal space maps of a selected ho-
moepitaxial ZnO thin film grown on the O-face of a CT substrate are shown
in Fig. 3.15. No splitting between the film and the substrate peak is observed
in the symmetrical (002) reflex (Fig. 3.15 (a)). The film peak is as least as
narrow as the peak of the substrate (compare Fig. 3.12). A modulation in the
out of plane reciprocal lattice vector q is observed due to Pendellsung oscil-
lations. This points towards a discontinuity at the interface between film and
substrate, and the film thickness can be determined from these oscillations.
In this particular case the thickness was determined to be 980 30 nm, being
in perfect agreement with the aspired thickness of 1 m, which was controlled
by the number of laser pulses during the deposition process. An empirical
growth rate for ZnO grown on sapphire determined like in Fig. 3.9 (f) was
used. In the asymmetric (104) reflection a finite broadening is observed (as
seen in Fig. 3.15 (b)), which could be interpreted as a mismatch between the
a-axis lattice constants of film and substrate, or which can originate from the
mosaicity of the substrate. A broadening of the (104) is observed for some of
the CT substrates (refer to Fig. 3.12).
TEM images were obtained under different imaging conditions. Fig. 3.16
shows a homoepitaxial thin film under (a) ~g = (0002) and (b) ~g = (1100)
imaging condition. The interface between substrate and film is seen in both
images by a dark horizontal line. In the image obtained under ~g = (0002) imag-
ing conditions dislocations are visible (dark lines in Fig. 3.16 (a)) above the

41
3. Physical properties of materials employed

(a) ZnO (0 0 2) Log10(Intensity (counts))

3.855 3.5

3.85 3
q (nm )

3.845 2.5
1

3.84 2
3.835 1.5
3.83 1
3.825 0.5
3.82
4 2 0 2 4
1 3
q|| (nm ) x 10

(b) ZnO (1 0 4) Log10(Intensity (counts))


7.69 3

2.5
7.685
q (nm )

2
1

1.5
7.68
1

0.5
7.675

3.545 3.55 3.555


1
q|| (nm )

Figure 3.15: Reciprocal space maps of the (a) (002) and (b) (104) reflex of a
homoepitaxial ZnO thin film determined by HR-XRD. The color scale represents
the log10 of the scattered X-ray intensity (counts). (Data obtained by Dr. Michael
Lorenz)

42
3.2. Growth of ZnO and Mgx Zn1x O thin films

Figure 3.16: Cross section TEM bright field images of a homoepitaxial ZnO thin
film under (a) (0002) and (b) (-1100) imaging condition. Beam direction parallel
to [1120]. (Images obtained by Dr. Gerald Wagner (ZnO on ZnO) and Dr. Herbert
Schmid (Universitt Bonn) (ZnO on sapphire))

Figure 3.17: CBED diffraction patterns as obtained (a) from measurements on


an O-polar homoepitaxial ZnO thin film (b) from simulations for an O-polar ZnO
crystal. The beam direction is [0110]. (Courtesy of Dr. Gerald Wagner)

horizontal line. The contrast vanishes under ~g = (1100) imaging conditions,


indicating that the burgers vectors of the dislocations have to be of c type. In
the inset a bright field TEM image of a ZnO thin film grown on sapphire is
shown under (nachgucken) imaging conditions. The density of dislocations is
significantly reduced in the homoepitaxial thin film due to the perfect lattice
match with the substrate.
The nature of the interface layer however remains unresolved. In order to ex-
clude that it corresponds to an inversion boundary, within the TEM study the
polarity of the thin films was determined by convergent beam electron diffrac-
tion (CBED). The fine structure patterns of the (0002), (0000) and (0002)
diffraction are shown in Fig. 3.17 as obtained from (a) measurement and (b)
simulation for oxygen polar ZnO. The exact congruence of the patterns con-
firms the O-face polarity of the homoepitaxial ZnO thin film.
HR-TEM images of the interface between the film and the substrate were

43
3. Physical properties of materials employed

Figure 3.18: Cross section HR-TEM image of the interface between a homoepitaxial
thin film and the substrate. Beam direction is parallel [1120]. (Courtesy of Dr.
Gerald Wagner)

recorded and are shown in Fig. 3.18. It can be seen that the interface layer has
an extension of one to two unit cells. It is further observed that the interface
is not related to a stacking fault or any other kind of extended lattice imper-
fection. The nature of the interface layer remains unclear, it might however
be speculated, that, as it is not due to extended defects, an increased density
of points defects might play a role. These could be of vacancy type, substi-
tutional atoms which gather at the crystal surface [150] or hydrogen which was
involved in surface reconstruction [151,152] of the substrate and is introduced on
interstitial site. Also intrinsic interstitial defects might be a possible origin.
Regardless of the actual nature of the defects, the density be must consider-
able, given the contrast in the TEM images. It is important to note, that any
of the defects discussed will form electronic states in the bandgap and there-
fore change the local density of free carriers. Depending on the nature of the
defects introduced, an accumulation of donor or acceptor defects is possible.
While acceptor defects would lower the Fermi-level and lead to a depletion of
free electrons, donor defects would rise the Fermi-level and increase the free
electron density in the vicinity of the interface. Exceeding a critical density,
which is experimentally determined in Chap. 4.3 this would introduce a degen-
erate layer. It was reported that especially the density of group III elements is
increased at the surface of O-face ZnO single crystals [150,153] . A contribution
from a degenerate layer was indeed reported for such crystals [153,154] . Also the
influence of hydrogen on the conductivity of a surface layer on single crystals is
discussed in the literature [155157] . The contribution of these layers to the over-
all conduction of the single crystal is not negligible [153,154] . The interface layer
may therefore have an equally important influence on the transport properties
measured in the homoepitaxial thin films.
In order to understand the mechanisms involved in the growth of the homoepi-
taxial thin films a growth evolution study was carried out, as already conducted
for the growth of ZnO thin films on sapphire in Chap. 3.2.1. Figure 3.19 shows

44
3.2. Growth of ZnO and Mgx Zn1x O thin films

d)

Figure 3.19: 2 m0.5 m AFM height images of homoepitaxial ZnO thin films of
different thickness. All films were deposited at a temperature of 650 C and a oxygen
partial pressure of 0.016 mbar. A substrate prior to growth is shown in (a), films with
a thickness of (b) 8 nm, (c) 25 nm and (d) 1,000 nm. (Data acquired by Heidemarie
Schmidt (now Forschungszentrum Rossendorf))

the surface morphology for homoepitaxial ZnO thin films of different thickness.
The substrate surfaces show vicinal steps, from which the miscut can be deter-
mined. For the substrate in Fig. 3.19 (a) a miscut of 0.1 was determined from
the AFM height image. Figure 3.19 (b) shows the morphology of a homoepi-
taxial ZnO thin film of approximately 8 nm thickness. It can be seen, that the
film is not completely closed, indicating an inhomogeneous nucleation of the
ZnO film on the substrate. Adatoms gather around these islands and become
incorporated at upsteps. However the lateral size of the randomly distributed
islands is more than a factor of 10 higher than the thickness of the film grown,
indicating a 2-dimensional growth mode. Also the existence of circular shaped
depressions indicates a coalescence of the growth seeds, which is dominated by
diffusion of adatoms to the nearest upsteps. As the film thickness increases, the
layers close completely , as can be seen for the film with a thickness of 25 nm
(Fig. 3.19 (c)). Meandering steps and islands of the next overlayer are visible,
as typical for layer by layer growth. The miscut of the substrate is resembled
by the sequence of the meandering steps. However, a stable step flow growth
cannot be achieved for the combination of growth temperature and substrate
miscut, as the diffusion length of adatoms is below the average step distance of
the vicinal substrate (in this case roughly 125 nm). Therefore upon increase of
the layer thickness hexagonal hillocks form, the height and size of the hillocks
increases with increasing film thickness (Fig. 3.19).

3.2.6 Growth of Mgx Zn1x O on ZnO substrates


For the realization of single heterostructures and quantum wells, Mgx Zn1x O
thin films were grown onto ZnO substrates. It was anticipated that sample
structure and morphology would improve in comparison to Mgx Zn1x O thin
films grown on sapphire as was observed for the ZnO thin films. Therefore mor-
phological investigations were carried out on Mgx Zn1x O thin films grown on
the O-face of CT substrates. All films discussed in this section contain a ZnO
quantum well, again following the assumption that the very thin quantum well

45
3. Physical properties of materials employed

Figure 3.20: AFM height images of Mgx Zn1x O thin films grown on ZnO. (a)
2 m2 m height image of a sample grown at p(O2 ) = 0.002 mbar and Tg = 680 C;
(b) 10 m10 m height image of a sample grown at p(O2 ) = 0.002 mbar and Tg =
660 C; (c) 2 m2 m height image of a sample grown at p(O2 ) = 0.004 mbar and
Tg = 630 C (d) temperature progress during the deposition of the films shown in
(a)(c)

layer does not influence the final morphology. Figure 3.20 shows AFM height
images of three samples grown under different deposition conditions. HR-XRD
measurements confirm pseudomorphic growth for all the thin films discussed
in the section. Structural data is given in Chap. 3.1.2. The sample shown in
Fig. 3.20 (a) was deposited at p(O2 ) = 0.002 mbar and an approximate growth
temperature of 680 C. It has to be mentioned that the substrate heater in the
deposition chamber is operated at constant power. The temperature is moni-
tored, however depending on the oxygen partial pressure, the heating time prior
to deposition and the temperature of the outer walls of the deposition chamber
the actual substrate temperature is not trivialy predictable (see Fig. 3.20 (d)).
The sample grown at 680 C shows a high density of pits on the surface. Fur-
ther a mostly vicinal arrangement of the steps edges is visible. This results are
somewhat unexpected, as for similar growth conditions hexagonal hillocks are
formed for homoepitaxial ZnO thin films. A similar morphology is reported in
literature [158] . In that study ZnO thin films are grown homoepitaxially onto
the Zn- and the O-face of ZnO single crystals. At temperatures above 850 C
hexagonal pits are formed on the O-polar films. The Zn-face shows a stepped

46
3.2. Growth of ZnO and Mgx Zn1x O thin films

surface without pit formation. The differences are explained in the framework
of Ref. [159] by the number of dangling bonds emerging from the surface of the
film to bind arriving adatoms. Reducing the growth temperature leads to a
layer by layer growth mode with hexagonal hillocks similar to that reported in
Chap. 3.2.5. A more complete explanation is given by Chen et al. [160] and Ko
et al. [161] who observed in MBE grown O-polar ZnO thin films that pits only
form under zinc rich conditions. Again considering the mobility of adatoms
they concluded that excess zinc adatoms are able to overcome the Ehrlich
Schwoebel barrier [118] and migrate to an higher level terrace. This leaves a
depression behind. Given the stoichiometric target applied in PLD, and the
higher vapor pressure of oxygen, an increased growth temperature would also
increase the Zn/O ratio, explaining the observations in Ref. [158] . Further, as
the vapor pressure of Mg is lower than that of Zn [92] , an increased Mg content
x would result in an even increased excess of the group II constituent. This
could lower the temperature required for the process to take place, as indeed
observed in our experiments. Fig. 3.20 (b) shows an 10 m10 m AFM height
image of a sample deposited at p(O2 ) = 0.002 mbar and an approximate growth
temperature of 660 C. A vicinal surface is visible, and the formation of pits
was completely suppressed. A limited meandering and the formation of islands
of the next overlayer is visible indicating that the sample is at the transition
between step flow growth and layer by layer growth. The diffusion length of
adatoms on the surface must roughly equal the terrace width. Fig. 3.20 (c)
shows a 2 m2 m AFM height image of a sample deposited at p(O2 ) =
0.004 mbar and an average growth temperature of 630 C. The sample shows a
perfectly vicinal surface with equally spaced and ideal parallel terrace edges.
It is noteworthy, that the miscut of the substrate is with 0.32 much higher
than for the sample shown in Fig. 3.20 (b), being 0.05 . At least one pit is
visible on the film surface, however in this case the pit most likely originates
from a threading dislocation. As no steps emerge from the pits, the dislocation
can only be of pure edge type. Nevertheless, the perfection of the surface is
unsurpassed by any other film grown in this study.
In order to understand the influence of the substrate miscut and Mg content
x in more detail, a series of Mgx Zn1x O thin films with varying Mg content x
was deposited on two batches of substrates, one with a high miscut ( 0.3 )
and one with a low miscut (< 0.1 ). 2 m2 m AFM images obtained on
these samples are shown in Fig. 3.21. All samples were grown under an oxygen
partial pressure of 0.002 mbar and at 650 C in order to exclude any influence of
the temperature and partial pressure on the sample morphology. The samples
shown in Fig. 3.21 (a) and (b) are both deposited on substrates with a mis-
cut around 0.3 . Both samples show vicinal steps, very similar to the sample
shown in Fig. 3.20 (c). The sample shown in Fig. 3.21 (b) with x = 0.33 shows
a relative high density of pits, which are not present in the sample shown in
Fig. 3.21 (a) with x = 0.23. In comparison, the samples shown in Fig. 3.21 (c)
and (d) were grown on substrates with smaller miscuts of 0.05 and 0.1 , re-
spectively. The sample depicted in Fig. 3.21 (c) is grown at conditions which
favour the beginning of island formation, very much like the sample shown in

47
3. Physical properties of materials employed

Figure 3.21: 2 m2 m AFM height images of Mgx Zn1x O samples with various
Mg contents grown on substrates of different miscuts. (a) x = 0.23, substrate miscut
0.31 (b) x = 0.33, substrate miscut 0.3 (c) x = 0.18, substrate miscut 0.05 (d)
x = 0.32, substrate miscut 0.1 .

Fig. 3.20 (b). Both samples are grown under conditions close to the transition
between step flow growth and layer by layer growth. In contrast, Fig. 3.21 (d)
displays a sample which is clearly grown in the layer by layer growth mode,
despite the larger miscut of the substrate. Additionally small pits are visible,
as in the sample given in Fig. 3.21 (b). Interestingly both Mgx Zn1x O thin
films have a high Mg content x above 0.3. Therefore the pits can be ascribed
to the piercing points of misfit dislocations introduced by the high strain be-
tween film and substrate. It is therefore also understandable, why the sample
shown in Fig. 3.21 (d) is grown in the layer by layer growth mode. Upon the
introduction of misfit dislocations part of the compressive strain is relaxed.
Therefore the reduction of the surface diffusion barrier decreases, and the con-
ditions favouring the step flow growth mode are not present. In contrast, the
sample shown in Fig. 3.21 (b) retains the step flow growth mode, despite the
higher density of pits and therefore dislocations due to the higher miscut of
the substrate. In summary the following conclusions can be drawn from the
growth experiments. Three main parameters influence the surface morphology

48
3.3. The perovskites BaTiO3 and SrRuO3

of Mgx Zn1x O thin films grown pseudomorphically on ZnO:

The growth temperature A higher growth temperature increases the sur-


face diffusion length of the adatoms. If the temperature exceeds a critical
value, however, the formation of pits becomes likely. The actual mech-
anism involved remains unclear, as the c-plane represents the absolute
minimum in surface energy. It has been lined out, that a change in com-
position towards Zn rich condition give rise to a motion of adatoms to
higher terraces, leaving the depressions behind [160,161] .

The miscut of the substrate Upon increase of the miscut the distance be-
tween vicinal steps decreases, increasing the probability for an adatom to
arrive at an upstep during its diffusive motion on the underlying terrace.

The Mg content An increase in the Mg content increases the compressive


strain imposed on the film. As already argued for Mgx Zn1x O films
grown on sapphire the surface diffusion is enhanced via reduction of the
surface diffusion barrier by the compressive strain [122124] . Therefore an
effective adatom diffusion length can be defined for different Mg contents.
The strain of the Mgx Zn1x O layer can, however, be partially relaxed by
the introduction of misfit dislocations. The probability for this process
increases with increasing Mg content x. Also, the density of misfit dislo-
cations increases with increasing Mgx Zn1x O film thickness, making the
effective adatom diffusion length a function of sample thickness. Addi-
tionally, as the vapor pressures of Mg and Zn differ, an increased Mg
content x leads to an excess of the group II constituent, which can fur-
ther give rise to the formation of pits by a mechanism involving adatom
migration across upsteps [160,161] .

3.3 The perovskites BaTiO3 and SrRuO3


In contrast to ZnO and Mgx Zn1x O, where the electric polarization is fixed to
the lattice, ferroelectric materials allow a control of the electric polarization by
an external electric field. Within the class of oxide ferroelectrics numberous
materials crystallize in the perovskite structure. Among these BaTiO3 (BTO)
is one of the most popular materials, and is often chosen as a model material.
The perovskite structure is displayed in Fig. 3.22. Like for ZnO, initial exper-
iments regarding the growth of BTO thin films were carried out. Two basic
systems are regarded, BTO grown on Si substrates and on lattice matched
SrTiO3 (STO) substrates. While BTO substrates exist, the quality of these
substrates is questionable (Ref. [162] ). Also, the substrates are with more the
600 e very pricey, while STO substrates are available from 30 e on. Fur-
ther, due to the different lattice constants a strain induced enhancement of
the ferroelectric polarization of very thin BTO layers was predicted [163165] and
experimentally observed in films deposited on STO [166] .
Fig. 3.23 shows 2 m2 m images of the surface of as delivered STO sub-
strates from two suppliers (Fig. 3.23 (a) Crystal GmbH [167] and Fig. 3.23 (b)

49
3. Physical properties of materials employed

Figure 3.22: Perovskite unit cell (ABO3 ) with blue A, cyan B and red O atoms. In
the case of BaTiO3 the smaller Ti atom in the center of the unit cell is able to move
to off center position in the low symmetry phases.

Crystec GmbH [133] ). Both substrates show stepped surfaces, however multiple
different height steps are observed on the substrate. Additionally no sharp
edges are seen in the AFM images. It was shown in the literature, that a
substrate preparation technique is necessary to improve the quality of the sub-
strates [168170] . Approaches employing thermal annealing [171173] , etching in
buffered HF [168,169] and the combination of both approaches [170] are reported.
A comprehensive review of the literature on substrate treatments and a sum-
mary of the experiments carried out in the framework of the present study can
be found in Ref. [162] . For the preparation of the substrates used herein a com-
bination of etching and annealing is employed, and the substrate is soaked in
a water bath for 10 min prior to the etching step, as proposed in Refs. [174,175] .
The annealing step was either carried out in a rapid thermal annealing chamber
under an oxygen atmosphere of 700 mbar or in a conventional tube furnace in
air. The annealing temperature was set to 1,000 C in both cases. The effects
on the surfaces were investigated by AFM, 2 m2 m images are shown in
Fig. 3.24. In both cases vicinal substrates were prepared which show only steps
of unit cell height. It is therefore concluded, that the substrates are only termi-
nated by TiO2 layers, which gives ideal conditions for the epitaxy of perovskite
thin films [169] . Additionally the morphology of (110) and (111) STO substrates

50
3.3. The perovskites BaTiO3 and SrRuO3

7,4 nm 2,6 nm

0,0 nm 0,0 nm

Figure 3.23: 2 m2 m AFM images of untreated (100) STO substrates as re-


ceived from (a) Crystal GmbH and (b) Crystec GmbH. (Courtesy of Stefan Schche)

8,6 nm 7,2 nm

0,0 nm 0,0 nm

Figure 3.24: 2 m2 m AFM images of STO substrates annealed in (a) a rapid


thermal annealing chamber in 700 mbar O2 (b) a tube furnace in air. Both samples
are annealed at 1,000 C. (Courtesy of Stefan Schche)

is investigated in Ref. [162] . Both types of orientation will not be regarded in


this study, as the ferroelectric polarization is parallel to the [001] direction in
BTO layers at room temperature. The effects of the ferroelectric polarization
at a heterointerfaces will therefore be maximal if the surface normal of the
interface plane is parallel to this direction.
Growth experiments summarized in Ref. [162] showed, that the optimal oxy-
gen partial pressure for the growth of BTO thin films on STO substrates was
2 103 mbar, and the optimal temperature was 730 C. It has to be remarked,
that the temperature is close to the attainable maximum in the PLD cham-
ber employed, a higher structural quality could be possible, if the temperature
was raised to even higher levels. Structural investigations indicated pseudo-
morphic growth for thin layers [176] . Within this study an evaluation of the
asymmetric reflexes was not possible due to low reflection intensity, and there-
fore the growth mode had to be deduced from the sample morphology. Fig. 3.25
shows 2 m1 m images of BTO thin films of different thicknesses grown on
STO substrates under optimized growth conditions (2 103 mbar oxygen and
730 C). A transition from very well developed vicinal surfaces with unit cell
height to surfaces without clear developed step edges takes place at a layer
thickness of around 4550 nm. It is concluded that the deterioration of the

51
3. Physical properties of materials employed

1,46 nm 1,02 nm

(a) 0,00 nm (b) 0,13 nm

2,48 nm 1,57 nm

(c) 0,00 nm (d) 0,70 nm

1,09 nm 3,21 nm

(e) 0,00 nm (f) 0,17 nm

Figure 3.25: 2 m1 m AFM images of BTO thin films of (a) 15.9 nm, (b)
26.2 nm, (c) 43.4 nm, (d) 42.2 nm, (e) 53.9 nm, and (f ) 56.1 nm thickness grown on
STO substrates. (Courtesy of Stefan Schche)

surface is induced by the formation of misfit dislocation above the critical


thickness, and that the critical thickness is roughly 45 nm, in concordance with
Ref. [176] . As for the samples investigated in the following the BTO thickness
is typically above 500 nm it is concluded, that the film is relaxed, and misfit
dislocations exist inside the BTO films.
Fig. 3.26 shows SRO thin films of roughly 30 nm thickness. A stepped sur-
face with step edges of unit cell height can be seen for the highest deposition
pressure of 0.1 mbarmbar, however holes in the films are visible. It is assumed,
that these steps are parallel to the step edges of the substrate, and originate
from incomplete substrate preparation [162] . More detailed work on the subject
is required.

2,51 nm 9,6 nm

0,00 nm 7,3 nm

15,9 nm 10,5 nm

13,8 nm 8,4 nm

Figure 3.26: SRO thin films of roughly 30 nm thickness grown on STO substrates
at 730 C and (a) 3 104 mbar (b) 2 103 mbar (c) 0.016 mbar and (d) 0.1 mbar.
(Courtesy of Stefan Schche)

52
3.4. Summary and conclusion

3.4 Summary and conclusion


The results of the present chapter can be summarized as follows:
The lattice constants a and c of Mgx Zn1x O change linearly with the Mg
content x in relaxed Mgx Zn1x O layers, where a increases with x and c
decreases with x.

It was shown that Mgx Zn1x O can be grown pseudomorphically on ZnO.


The c-axis lattice constant changes systematically with x. The linear
relationship cps = c0 + kc,ps x with kc,ps = 0.064 was found from data
obtained in the present study and in accordance with the literature [96,97] .

The mechanisms involved in the growth of ZnO on sapphire were investi-


gated, a transition from initial 3D growth to 2D growth was revealed. A
steady state is reached after roughly 100 nm film thickness, which allows
determination of the growth rate.

The observation given in Ref. [81] that the morphology of Mgx Zn1x O
layers improves upon the introduction of a ZnO buffer layer was con-
firmed, and understood within the argumentation given in Refs [122124] .
The Mgx Zn1x O layer grown on the ZnO buffer is compressively strained,
which increases the surface diffusion length and therefore promotes step
flow growth.

The state of the art ZnO substrate material available was reviewed, and
the individual advantages discussed. Only hydrothermally grown mate-
rial proved applicable in the light of transport investigations. A thermal
pretreatment of the substrates necessary to produce surfaces suitable for
epitaxy was discussed.

The mechanisms involved in the homoepitaxial growth of ZnO were inves-


tigated. Within the temperature range available layer by layer growth
was observed rather than step flow growth. A significant reduction of
the defect density was observed in comparison to ZnO grown on sapphire
substrates.

The mechanisms involved in the growth of Mgx Zn1x O on ZnO substrates


were investigated. It was shown, that the temperature allowing step flow
growth was lower than for pure ZnO. This was again explained by the
presence of compressive strain in the films, in accordance with the argu-
mentation given in Refs. [122124] . Perfect vicinal surfaces were prepared,
depending on the miscut of the substrate.

A substrate preparation routine was applied to commercially available


STO substrates and produced perfect vicinal substrates with step edges
of unit cell step height.

Vicinal BTO thin films with step edges of unit cell step height were grown
up to a critical thickness of roughly 45 nm in concordance with Ref. [176] .

53
3. Physical properties of materials employed

Vicinal SRO thin films with step edges of unit cell step height were grown,
however in the films investigated holes parallel to the steps of the sub-
strate are visible. An improvement of the growth of this material will be
necessary in the future.

54
Chapter 4

Properties of ZnO and


MgxZn1xO single layers

For the fabrication of light emitting devices several conditions have to be satis-
fied. Primarily the design of the active region is of importance. High densities
of electrons and holes have to be available to recombine in this region. For the
sake of radiative recombination efficiency the density of defects in the active
region has to be as low as possible. Further a low series resistance of the injec-
tion regions is required for high injection efficiencies. Hence, high conductivity
of the layers adjacent to the active region is necessary. The photons gener-
ated in the active region have to leave the crystal, which demands additional
optimization of the device. Reabsorption has to be low, and the extraction ef-
ficiency has to be high. Many of the considerations regarding this issue are of
geometrical nature, but refinement in the material properties is also required.
Again, low defect densities are required, as they act as absorption centers.
One structure is most suitable to fulfill all the requirements listed above, a
pn heterodiode. The most simple structure applicable would be a double-
heterostructure with a higher bandgap in the injection regions than in the
active region. Much more sophisticated structures can be designed, containing
(multi-)quantum wells, graded injection regions, dielectric mirrors and lateral
carrier confinement barriers. From these complicating device improvements
only quantum wells will be regarded in the following. It is further important
to know, that reliable reports on ZnO based heterostructure light emitting
devices are sparse. pn homodiodes are reported in the literature [66,177179] ,
but show low recombination efficiency. Despite the inherent drawbacks in the
structure itself, the low conductivity of the hole injection layer reduces the
recombination efficiency. Another promising approach is the developement of
more efficient heterodiodes [180] . It is therefore necessary to transfer the abil-
ity of p-type doping from ZnO to ternary systems, in the case of this study
Mgx Zn1x O. In the literature successful p-type doping of Mgx Zn1x O with
phosphorous was reported [181184] . Like for pure ZnO numerous other dopants
(all group I and all group V elements) are conceivable alternatives. Among
them nitrogen and phosphorous are the most promising candidates [141,177,185] .
Sucessful fabrication of p-type material employing arsenic and antimony are

55
4. Properties of ZnO and Mgx Zn1x O single layers

also reported [186,187] , but will not be regarded in the present study. Given the
successful reports of p-doping of PLD grown Mgx Zn1x O [188,189] with phospho-
rous and its much easier handling (it can be introduced into the target from
solid powder) this route is followed in the present study.

4.1 Properties of homoepitaxial ZnO thin films

4.1.1 Properties of nominally undoped homoepitaxial ZnO


thin films

Structural properties

The structural and morphological properties of typical homoepitaxial thin films


were extensively discussed in Chap. 3.2.5. In the following section the elec-
tronic properties of nominally undoped and phosphorous doped thin films will
be reported. Therefore four samples will be compared in the following. All
samples were grown at 700 C but at four different oxygen partial pressures in
the range between 3 104 and 0.1 mbar. The films were approximately 1m
in thickness. The results given in this section are partially reported in [190] . The
sample morphology will not be discussed in detail here, as it is very similar
to the morphology observed in Fig. 3.14 (b) and only shows a slight variation
with the change in oxygen partial pressure.
Figure 4.1 depicts the X-ray diffraction intensity measured on the films in the
vicinity of the (002) reflection. All films show a common peak of high intensity,
which is related to the substrate. The peak is shifted to zero angle for com-
parison of the strain in the individual layers. No peak splitting is visible for
the two samples grown at the highest oxygen partial pressure. The presence of
the thin film is evident from the Pendellsung oscillations best visible for the
film grown at 0.016 mbar. It can therefore be concluded, that the two films
grown at the highest oxygen partial pressures possess the same c-axis lattice
constant as the substrate. No splitting can be seen for the in-plane reciprocal
lattice vector from asymmetric reflexes, indicating pseudomorphic growth of
all four thin films. Interestingly, the film peak is shifted towards lower angles
for the two films deposited at the lower oxygen partial pressures. As the film is
grown pseudomorphically, an increase in the c-axis lattice constant corresponds
to a compressive strain. The relative change in the c-axis lattice constant is
4.3 103 and 2.3 103 for sample grown at 2 103 mbar and 3 104 mbar,
respectively. The change in the lattice constant is most likely related to the
increased incorporation of intrinsic defects. Several theoretical studies have
shown, that the unit cell volume changes upon the incorporation of charged
intrinsic defects [191194] .

56
4.1. Properties of homoepitaxial ZnO thin films

Figure 4.1: 2 X-ray diffraction patterns of the (002) reflection of homoepitax-


ial ZnO thin films grown at various oxygen partial pressures. The oxygen partial
pressures during growth are marked in the graph. (Data acquired by Dr. Michael
Lorenz, graphed by Alexander Lajn)

Recombination spectra

Recombination spectra were detected by photoluminescence spectroscopy. The


samples were subject to 325 nm excitation, as described in Chap. 2.5. The
luminescence spectra of the near band edge region are graphed in Fig. 4.2. The
recomination lines are labeled following the nomenclature given in Ref. [195] . In
order to identify the features visible, and correlate them to their chemical
origin, the spectrum obtained from the sample grown at 0.016 mbar will first
be regarded, as the features are best visible in this sample. Identification of the
respective lines is done by comparing the peak positions with the compilation
given in Ref. [139] . The free exciton recombination is visible (labeled FXA ), along
with the recombination features I0 , I3a and I6a . While the I6a line corresponds to
an exciton bound to a neutral aluminum donor (Al0 X) [139] , the I0 line originates
from the recombination of an exciton bound to an ionized Al donor (Al+ X) [196] .
The I3a feature is related to the recombination of an exciton bound to a zinc
atom on interstitial site (Zni X) [197] . The presence of this defect indicates a zinc
rich stoichiometry, and a dependence on the oxygen partial pressure during
growth is anticipated. For the aluminum donors, however, an extrinisic source
has to exist. It is reported, that metallic impurities gather at the oxygen
face of ZnO single crystals [150,153] . Also the ZnO target itself contains trace
amounts of aluminium, which are most likely introduced during the sintering
process. Therein the pressed ZnO powder is heated up to 1150 C under an

57
4. Properties of ZnO and Mgx Zn1x O single layers

Figure 4.2: Near band edge recombination spectra of homoepitaxially ZnO thin
films grown at various oxygen partial pressures. The labeling follows a nomenclature
given in Ref. [195] . The chemical origin is given in the graph and explained in the
text. (Spectra measured by Gabriele Benndorf, graphed by Alexander Lajn)

alumina cover. Despite the very high (2072 C) melting point of alumina, a
finite mass transport might take place due to sublimation of the alumina.
Finally a contamination of the powder used in the target preparation process
can not be excluded. The chemical purity of the powder is analyzed by the
supplier, and the aluminum concentration is below 2 ppm. It has however been
found, that the concentration of shallow donors in homoepitaxially grown ZnO
thin films is as low as 6 1014 cm3 [146] , which is significantly below this level.
Whatever the origin of the aluminum donors is, it is reasonable to assume, that
the concentration does not change drastically with the oxygen partial pressure.
A small spread might however exist, as the different atomic mass will lead to
different scattering kinetics in the plasma plume compared to Zn and O. Also
a stochastic scattering of the aluminum concentration is thinkable regarding
the substrate material. The substrates employed in this study are however cut
from nearby sectors of the same crystal, thus it seems reasonable to assume
an equal distribution. Comparing the relative intensity of the I6a and I3a line,
the sample grown at 0.016 mbar shows the lowest contribution from I3a which
corresponds to nearly stoichiometric conditions. This is also supported by the
sharper lineshape and the observation of the free exciton recombination line.
In contrast the films grown at lower oxygen partial pressure show a relative
increase of the I3a line, indicating Zn rich conditions. Further a blueshift of
the spectra is observed for the samples grown at the lowest oxygen partial
pressures. This can be understood by a shift of the fundamental band edge
due to the compressive strain. In order to quantify the shift, the difference in
the energetic position of the I6a line is taken with respect to the sample grown
at 0.016 mbar. The blue-shifts are 2.2 meV and 1.3 meV for the sample grown

58
4.1. Properties of homoepitaxial ZnO thin films

at an oxygen partial pressure of 2 103 mbar and 3 104 mbar, respectively.


In order to clarify if the blue-shift is strain induced, the experimental shift
is compared with theoretical values. Therefore the change in the energetic
position of the free exciton emission E(XA ) is calculated following Ref. [198] :

E(XA ) = EA0 + (D1 D2 /)c + b+ (D3 D4 /)c


(4.1)
b+ b [(D3 D4/)c ]2 (EC0 EA0 )1

where EA0 = 3.3759 eV and EC0 = 3.4199 eV are the energies of the free A and
C exciton transitions in unstrained bulk ZnO [139] . The deformation potentials
used are D1 = 3.9 eV, D2 = 4.13 eV, D3 = 1.15 eV and D4 = 1.22 eV [199] ;
the Poisson ratio = C13 /C33 is calculated using the elastic constants given
in Ref. [99] . The b are calculated following

2b = 1 [1 2 ][(1 2 )2 + 822 ]1/2 (4.2)

where 1 and 2 are the crystal-field and spinorbit-coupling splitting parame-


ters 1 = 38.3 meV and 2 = 2.1 meV [199] . Taking the strain values determined
from the XRD analysis, blue-shifts of 3.1 meV and 1.6 meV were calculated for
the samples grown at a p(O2 ) of 2 103 mbar and 3 104 mbar, respec-
tively. Both values a slightly above the experimentally observed blue-shift,
which might relate to a partial relaxation of the films by misfit dislocations.
Again PL measurements only probe a small fraction of the sample volume,
approximately 100 nm from the surface, while XRD measures the whole thin
film and part of the substrate. As the reduction of the strain due to misfit
dislocations increases with the film thickness the results would be explained
from this viewpoint.

Transport properties and defect states


In order to further probe the electronic properties electric measurements were
carried out. Hall-effect investigations are undertaken using the setup described
in Chap.2.4. Ohmic gold contacts were sputtered onto the corners of the sam-
ples in argon. The temperature dependence of the free carrier concentration
and the Hall-mobility are depicted in Fig. 4.3. Only three of the samples
could be measured by means of temperature dependent Hall-effect. The sam-
ple grown at 0.016 mbar was too insulating to allow a meaningful determination
of the free carrier concentration and Hall-mobility. Regarding the free carrier
concentration, the samples can be phenomenologically divided into two sub-
sets: one being dominated by rather shallow donors (group A), in this case the
samples grown at lowest oxygen partial pressures, the other one with very deep
dominant donors (group B), in this case the sample deposited at 0.1 mbar. For
rigorousness groups A and B will be defined as the samples grown at pressures
at and below 2 103 mbar and at and above 0.016 mbar respectively. The
samples of group A can be measured over a wide range of temperatures and
show rather high Hall-mobilities. The determination of the donor activation
energy and donor and acceptor concentrations as well as an evaluation of the

59
4. Properties of ZnO and Mgx Zn1x O single layers

1017 0.1 mbar


210-3 mbar
310-4 mbar
1016
n (cm -3)

1015

1014

1013
5 10 15 20 25
1000/T (1/K)

1.000
0.1 mbar
800 210-3 mbar
310-4 mbar
H (cm 2/Vs)

600

400

200

0
50 100 150 200 250 300
T (K)

Figure 4.3: Temperature dependence of (a) the free electron concentration and
(b) the Hall-mobility in nominally undoped homoepitaxial ZnO thin films deposited
under various oxygen partial pressures. The solid lines represent fits to the data, as
described in the text.

60
4.1. Properties of homoepitaxial ZnO thin films

scattering processes limiting the Hall-mobility is possible in these samples. In


contrast the samples with deep dominant donors can only be measured in a
rather small range of high temperatures, revealing only a limited amount of
data. While the determination of the donor activation energy is still possible,
the scattering processes limiting the Hall-mobility can not be assessed in detail.
Especially the influence of ionized impurity scattering which limits the mobility
at low temperatures is not revealed. As described in Chap. 2.4 the determina-
tion of the acceptor concentration is only possible from the mobility if this low
temperature region is measurable, and the mobility in this region is not affected
by additionally complicating conductivity contributions. The determination of
the donor and acceptor concentration is therefore limited to the evaluation of
the temperature dependence of the free carrier concentration, which can only
yield both values in certain cases (refer to Chap. 2.4). It is possible however to
understand the mechanisms involved. As the PL measurements have shown,
shallow donors like aluminum and interstitial zinc are present in all four sam-
ples, however in different concentrations. The fact that the shallow donors
can be observed in the Hall-effect measurements in the samples of group A
shows that they dominate the Fermi-level. In contrast, in the sample grown at
0.1 mbar only a deep donor is visible, the Fermi-level must therefore lie below
the levels of the shallow donors at all temperatures. This can be explained
by a significant increase in the compensation, the concentration of acceptors
has to exceed the concentration of shallow donors. As discussed before, the
concentration of aluminum donors is expected to be roughly constant in all
four films. A change in the ratio between shallow donors and acceptors can
therefore only be explained by a change in the sample stoichiometry. The sam-
ple grown at 0.016 mbar therefore shows a behaviour typical for the samples of
group B. The conductivity in this sample is even lower than that in the sample
grown at 0.1 mbar, most likely due to the higher compensation ratio.
For the samples of group A the acceptor concentration was determined from the
low temperature part of the Hall-mobility. The Hall-mobility in all samples was
modeled employing the exact numerical solution of the Boltzmann-transport
equation including inelastic scattering effects and the self consistently deter-
mined Hall-factor, as described in Chap. 2.4. All intrinsic scattering processes
are left unaltered. The solid line shown in Fig. 4.3 (b) gives the best fit to
the experimental data, with the acceptor concentration NA being the only free
parameter. In an iterative sequence the drift mobility is calculated, and the
experimental Hall-mobility is converted into the drift mobility by the Hall-
factor obtained. This is repeated until a self consistent solution is obtained.
Finally the Hall-mobility is calculated from the drift mobility and Hall-factor
obtained. For the sample grown at 2103 mbar an almost perfect coincidence
between the experimental and theoretical data can be obtained, indicating that
no additional scattering processes contribute to the mobility, and no parallel
conduction takes place in layers with different material properties. The scat-
tering mechanisms involved equal that present in a single crystal, indicating
the superior quality of the film compared to heteroepitaxially grown ZnO thin
films. The mobility in the film exceeds 855 cm2 /Vs. For the sample grown at

61
4. Properties of ZnO and Mgx Zn1x O single layers

Table 4.1: Donor activation energy, donor and acceptor concentration as obtained
from Hall-effect investigations (index H), net dopant concentration Nt = ND NA
measured by capacitance voltage spectroscopy (index C) and donor activation energy
and capture cross section (in cm2 ) as determined by Arrhenius evaluation of thermal
admittance spectroscopy (index T) for homoepitaxial ZnO thin films grown at various
oxygen partial pressure (given in mbar). Only the most shallow donor is tabulated as
would be seen by Hall-effect measurements. All concentrations are given in 1017 cm3 ,
activation energies in eV.

p(O2 ) ED,H ND,H NA,H NN,C ED,T T


0.1 350 50 70 30 < 0.003
0.016 0.01
2 103 29 1 3.4 0.2 0.24 0.01 0.07 37 1 2 1016
3 104 34 6 0.9 0.2 0.25 0.05 0.6 42 3 1 1016

3 104 mbar an equally good fit is not possible, which gives rise to the as-
sumption that additional scattering at crystal imperfections takes place which
is not accounted for in the model. This is supported by the broadening of the
XRD diffraction curve for this sample (compare Fig. 4.1), which might give a
hint towards the presence of misfit dislocations. These would represent one-
dimensional scattering centers which impede the transport. Nevertheless the
acceptor concentration could be estimated from the low temperature flank of
the mobility. All values are summarizes in Tab. 4.1. The temperature depen-
dence of the free carrier concentration was modeled using a simple one donor
charge balance equation with compensation. The data for the sample grown
at 2 103 mbar was corrected by the Hall-factor as obtained from the mobil-
ity fit, prior to the analysis. Solid lines in Fig. 4.3 (a) represent the modeled
results. The donor activation energies and concentrations are summarized in
Tab. 4.1.
In order to apply capacitance spectroscopic methods rectifying PtOx contacts
are formed on the four samples. Pt was reactively sputtered in and Ar/O2
mixture, as described in Chap. 2.6. The contacts were capped by a metal-
lic Pt layer sputtered in pure Ar afterwards to form an equipotential surface.
CapacitanceVoltage and thermal admittance spectroscopy measurements were
performed on the contacts. From the CV measurements the net dopant con-
centration was derived and is tabulated in Tab. 4.1. For the samples of group A
the net dopant concentration can be very well explained by the shallow donor
and the acceptor concentration as determined by Hall-effect, it equals there
difference within the experimental error. The situation is significantly differ-
ent for the sample grown at 0.1 mbar, where the values obtained differ by four
orders of magnitude. Several facts have to be considered explaining this dras-
tic difference. First of all the determination of both values is subject to large
experimental error due to its low value in the CV case and the low conduc-
tivity in the case of the Hall-effect measurement. Additionally the low number

62
4.1. Properties of homoepitaxial ZnO thin films

Figure 4.4: Temperature dependence of capacitance of PtOx Schottky diodes fab-


ricated from homoepitaxial ZnO thin films. (Courtesy of Alexander Lajn)

of points obtained in the temperature dependent Hall-effect measurement in-


creases the uncertainty of the results obtained from the numerical evaluation.
It is also important to note, that the simple assumption of a one donor charge
balance equation is not valid in the case discussed here, as at least two types of
shallow donors are present additional to the deep donor observed. While this
does not introduce a large error for the shallow donors in the samples of group
A which influence the Fermi-level, the error is considerable in the samples of
group B, where numerous donors are already ionized before the ionization of
deep donors. Finally, the acceptor concentration could not be determined from
the Hall-effect data. It is however anticipated that the compensation is very
high, and values above 95% were reported [47,200] . The net dopant concentration
will therefore be significantly reduced in comparisson with the donor concen-
tration.
Figure 4.4 depicts the temperature dependence of the capacitance of the Schot-
tky diodes fabricated from the samples. Again a clear distinction between sam-
ples of group A and samples of group B is visible. While for the samples of
group A two steps in the capacitance can be seen, only a single step is visible
for the samples of group B. The first step, visible at temperatures around 50 K
corresponds to a resonance between the excitation frequency and the emission
rate of the shallow donors, while the second step corresponds to the resonance
between the excitation frequency and the emission rate of the deep donors. A
emission from the shallow donors is not observed in the samples of group B,
as these donors are ionized at all temperatures, and no recharging takes place.
The activation energy of the shallow donors and two deep donors were deter-

63
4. Properties of ZnO and Mgx Zn1x O single layers

mined for the samples of group A by a conventional Arrhenius evaluation (not


depicted here) and are tabulated in Tab. 4.1. Additionally the activation energy
of the deep donor in both samples could be evaluated and is 41030 meV and
24410 meV for the sample grown at 2 103 mbar and 3 104 mbar, respec-
tively. The latter is similar to the previously reported defect L2 [201] . A donor
with an activation energy of 41030 meV has not been reported to the best of
my knowledge, however considering the comparatively large error bar, it might
be related to the level E3 with an activation energy of roughly 370 meV [202] .
Regarding the shallow donors a difference between the activation energy ob-
tained by TAS and Hall-effect is observed. The activation energy determined
by Hall-effect is always smaller than that from the TAS analysis. This might
originate from an inhomogenous distribution of the donors within the sample.
As TAS probes a rather small sample volume at the edge of the space charge
region introduced by the Schottky contact, the Hall-effect measures essentially
the whole sample. In Hall-effect, however, one measures a weighted average
over the conductivities and magnetoelectric properties of the sample, with the
largest contribution from layers with the highest conductivity. It is reported,
that metallic impurities gather at the surface of ZnO single crystals, there-
fore inducing a layer of higher conduction, up to degeneracy [150,153] . In such
a layer (assuming it not to be degenerate, as this would give clear signatures
in the Hall-effect measurement), the donor activation energy might be reduced
due to beginning overlap of the donor wave functions. This scenario receives
additional support by the results of the PL measurements, which resolve the
presence of both aluminum and zinc interstitial related donors, while only a
single level is observed in TAS, most likely zinc interstitial related. Following
Haynes rule, an activation energy of 38.1 meV is anticipated from the PL mea-
surements [196] , being in very good agreement with the values determined by
TAS.

4.1.2 Properties of phosphorous doped homoepitaxial ZnO


thin films
As outlined before phosphorous in one of the most promising candidates for
achieving reliable p-type conductivity. Several publications report on p-type
doping with phosphorous [203207] , and even ZnO homojunction diodes were re-
ported [205,208,209] . The investigated nominally undoped homoepitaxial ZnO thin
films show very high purity, significantly reducing the influence of background
doping. These samples are therefore perfectly suited to perform a doping
study. Phosphorous doped thin films were deposited from targets contain-
ing 0.01 wt.%, 0.1 wt.% and 1 wt.% P2 O5 added to the ZnO powder. From now
on these samples will be labeled series I, series II and series III. For each target
samples were grown at four different oxygen partial pressures, namely 3 104 ,
2 103 , 0.016 and 0.1 mbar, the same pressures as for the nominally undoped
samples shown in the previous section. For all samples a thickness of 1m
was deposited. Some of the results presented in this section were published
previously [210213] .

64
4.1. Properties of homoepitaxial ZnO thin films

Morphology
Figure 4.5 shows AFM height images obtained from samples of series I grown
at different oxygen partial pressures. A clear 2-dimensional growth is observed
for all samples, however the actual growth mode differs. The samples grown at
oxygen partial pressures below 0.1 mbar show the formation of islands of 200
400 nm diameter. Island growth is present in this case. The height and shape
of hillocks differs somewhat from the clearly hexagonal shapes observed in the
case of undoped homoepitaxial ZnO films. For the sample grown 0.1 mbar a
much flatter surface is observed and vicinal steps are visible. Meandering of
this steps and the formation of nucleation islands of the next overlayer are
visible. Nevertheless, as with a film thickness of 1m no transition to island
growth is observed, it is concluded, that the growth mode must be at the
transition between layer by layer growth and step flow growth.

Structural and chemical properties


XRD and HR-XRD measurements were carried out to reveal the structural
properties of phosphorous doped homoepitaxial ZnO thin films. Fig. 4.6 de-
picts reciprocal space maps of the (002) and (104) reflection of the sample
grown at 0.1 mbar from series I. No peak splitting is observed in any of the
reciprocal space maps. A small asymmetric broadening of the reflections is
visible, which is ascribed to minor mosaicity and the finite crystallite size of
the substrate.
Results of 2 X-ray diffraction measurements are graphed in Fig. 4.7. In
Fig. 4.7 (a) the diffraction patterns from series I are compared. For all sam-
ples the substrate peak is visible. Only for the film deposited at 0.016 mbar a
splitting between the film and the substrate peak is visible. The lattice con-
stant of all other films in this series equals that of the substrate. Pendellsung
oscillations are visible, proving that the film peak indeed overlaps with that
of the substrate and allowing the determination of the film thickness, which
equals 1 m within the experimental error. For the film grown at 0.016 mbar
an expansion of the c-axis lattice constant of 27015 ppm can be determined.
This gives indication that the incorporation of phosphorous expands the ZnO
lattice. As the film is grown pseudomorphic an expansion of the c-axis lattice
constant corresponds to an compressive in-plane strain.
This is supported by a comparison of the lattice constants of the films of the
individual series as depicted in Fig. 4.7 (b). It can be seen that the c-axis
lattice constants expands further with increasing phosphorous content in the
target. It is also visible that the lattice constant differs within series III, that
one with the highest phosphorous content. The largest splitting is observed
for the smallest oxygen partial pressure. This behavior infers a dependency
of the phosphorous content on the oxygen partial pressure. Also, the growth
mode changes from pseudomorphic to relaxed as the phosphorous content in-
creases [212] . Both the a-axis and c-axis lattice constant are expanded with
respect to the lattice constant of the substrate, giving clear evidence of the
expansion of the lattice with increasing phosphorous content.

65
4. Properties of ZnO and Mgx Zn1x O single layers

2 nm
(a)

0
0 m 2
2 nm
(b)

0
0 m 2
2 nm
(c)

0
0 m 2

Figure 4.5: 2 m1 m AFM height images for samples from series I. The samples
were deposited at (a) 0.1 mbar (b) 0.016 mbar and (c) 2103 mbar. (Data measured
by Dr. Heidemarie Schmidt (now Forschungszentrum Rossendorf))

66
4.1. Properties of homoepitaxial ZnO thin films

(a) ZnO:P (0 0 2) Log10(Intensity (counts))


3.85

3.5
3.845 3
q (nm )
1

2.5
3.84 2

1.5
3.835 1

0.5
3.83
7 6 5 4 3
1 3
q|| (nm ) x 10

(b) ZnO:P (1 0 4) Log10(Intensity (counts))

7.684 3

7.682 2.5
q (nm )
1

7.68 2

1.5
7.678
1
7.676
0.5
7.674

3.55 3.552 3.554 3.556 3.558 3.56


1
q|| (nm )

Figure 4.6: HR-XRD reciprocal space map of the (a) (002) and (b) (104) reflection
of a homoepitaxial ZnO:P sample from series I deposited at 0.1 mbar. The color scale
represents the log10 of the scattered X-ray intensity (counts). (Data obtained by Dr.
Michael Lorenz)

67
4. Properties of ZnO and Mgx Zn1x O single layers

p(O2) [ mbar ]
10,000
0.1 mbar
Intensity [cps]

0.016 mbar
210-3 mbar
1,000 310-4 mbar

100

34.30 34.35 34.40 34.45 34.50

2 []

E1328 0.01?P
Intensity [cps]

4 E1320 0.1?P
10 E1324 1? P
E1325 1? P
3
10

2
10

1
10

34.0 34.1 34.2 34.3 34.4 34.5


2Theta []

Figure 4.7: (a) 2 X-ray diffraction patterns of the (002) reflection of homoepi-
taxial ZnO:P thin films of series I grown at various oxygen partial pressures. The
oxygen partial pressures during growth are marked in the graph. (b) 2 X-ray
diffraction patterns of the (002) reflection of homoepitaxial ZnO:P thin films from
different series with different splitting in the film and substrate c-axis lattice constant.
(Data obtained and (b) graphed by Dr. Michael Lorenz.)

68
4.1. Properties of homoepitaxial ZnO thin films

Table 4.2: Summary of all impurities detected by the PIXE experiments in the
samples of series I, and their apparent concentrations

p(O2 )(mbar) 0.1 0.016 2 103 3 104


impurity concentration in weight ppm
Al 1390 530 1450 500
Fe 3070 130 2980 180 2200 210 2480 130
Si 3490 490 2530 460 3140 600 3020 340
P 320 110 830 280 780 290 840 140
S 780 170 1000 300 350 80
Cl 170 60 310 70 2080 130
K 1500 80
Ca 90 40 490 40
Cr 260 60 130 60 130 70
Mn 240 90

RBS channeling experiments were carried out at the LIPSION ion beamline
as described in Ref. [214] . For comparison the channeling spectrum of a ther-
mally pretreated CT substrate was recorded. A channeling axis perpendicular
to the sample surface was observed only for the substrate and the sample of
series I deposited at 0.1 mbar oxygen partial pressure. This gives evidence for
the single crystalline nature of this ZnO:P thin film. It further indicates a high
density of lattice imperfections for the films grown at lower oxygen partial pres-
sures, incorporated most likely due to the phosphorous doping. The spectra
of the substrate and the sample grown at 0.1 mbar oxygen partial pressure are
depicted in Fig. 4.8 (a). A min as low as 0.034 can be seen for the film grown
at 0.1 mbar oxygen partial pressure which is comparable to the value obtained
for the substrate (0.033), indicating very good crystal quality (cf. for pure
single crystalline Si min = 0.03).
Within this structural study the chemical composition of the samples of series
I was also evaluated. The phosphorous content of the ZnO:P thin films of
series I was determined by PIXE experiments and is graphed versus the oxy-
gen partial pressure in Fig. 4.8 (b). The phosphorous content in the samples
differs significantly from the nominal concentration in the target (nominally
0.01 wt% P2 O5 , thus 43 g/g P in ZnO). PIXE investigations of the target
revealed, that the concentration in the ablated region is 4511 g/g P in ZnO
in line with the expected concentration. The concentration of phosphorous
in the films is significantly higher, and further depends on the oxygen partial
pressure. The actual concentrations are only slightly above the detection limit
of the PIXE analysis and therefore subject to a relative large error, which is
depicted in Fig. 4.8 (b) as well. The apparent concentration in the film ex-
ceeds that of the target by a factor of 5 to 20, in contradiction to a previous
analysis [215] , which resulted in a constant phosphorous transfer ratio between
the target and the film of 0.5. It has to be mentioned that the phosphorous

69
4. Properties of ZnO and Mgx Zn1x O single layers

Sample thickness [m]


12 8.4 4.8 1.2 0

1.0
Channeling yield

0.8 p[O2] = 0.1 mbar

sample surface
0.6
substrate

0.4

0.2

0.0
200 300 400 500 600

Channel number

1,200
Phosphorous content [g/g]

1,000

800

600

400

200

0
0.0001 0.0010 0.0100 0.1000
p(O2) [mbar]

Figure 4.8: (a) Channeling yield as extracted from RBS measurements on the
homoepitaxial ZnO:P film from series I deposited at 0.1 mbar compared to a typical
CT substrate after thermal pretreatment. (b) Phosphorous content of of the samples
of series I as a function of the oxygen partial pressure applied during growth. (Data
obtained by Christoph Meinecke)

70
4.1. Properties of homoepitaxial ZnO thin films

concentrations in the previous analysis was at least by a factor of 20 larger


than that in the present investigation.
Further within this study the concentration of other impurities was deduced
from the PIXE data, and several unintentionally introduced elements were
found in the samples of series I. Among them the most prominent detected
were Fe and Si. Further Al, S, Cl, K, Ca, Cr and Mn were detected in small
concentrations, or not for all samples. Traces of Ti, V, Ni, Co and Cu were de-
tected, but their concentration is below the detection limit. The concentrations
of all the impurities are given in detail in Tab. 4.2. The results give further
insight into the origin of the impurities. The high fluctuation of Al between the
samples is unexpected. From the possible mechanisms of aluminum incorpora-
tion discussed in Chap. 4.1.1 the substrate therefore becomes the most likely
source. Other origins are not excluded, but their contribution must be below
the detection limit of the PIXE analysis. Rather high concentrations of both
iron and silicon are observed. There incorporation can easily be understood.
The silicon enters the target during the target preparation, where the powder
is homogenized in an agate mill. Iron most likely originates from the substrate
holder in the PLD chamber, which like most of the parts of the chamber con-
sists of steel and is subject to the same temperatures as the substrate. This is
also the most probable origin of Cr and Mn, which are used in steel refinement.
For the other impurities, it is remarkable to note that all of these elements take
share in biological processes, especially Cl, Ca and K and might therefore be
introduced by the sample handling.

Recombination spectra
Photoluminescence spectra of the near band edge region of series I are depicted
in Fig. 4.9 (a). Again, like in in the undoped series (see Chap. 4.1.1) the spec-
tra are dominated by the lines I6a and I3a . Their ratio is very similar to that
in the nominally undoped films. This indicates, that the film stoichiometry is
not strongly altered by the P content in the films. It further shows, that the
high aluminum concentrations measured by the PIXE analysis cannot be ho-
mogeneously distributed throughout the affected films, which would most likely
change the height and shape of the I6a peak in the samples with high aluminum
concentrations, namely the samples deposited at 0.1 mbar and 3 104 mbar.
Further the spectrum of the sample deposited at 0.016 mbar is blueshifted by
1.9 meV. In principle the blue-shift can again be explained by the compressive
strain in the sample (see Chap. 4.1.2). However the magnitude of the strain
is to small to explain the blue-shift completely, it can only account for around
20%. It is also interesting to note that the features I0 and the free exciton
recombination peak can not be observed for the sample grown at 0.016 mbar,
which were present in the undoped sample grown at the same pressure (refer
to Chap. 4.1.1). This is reasoned by the overall reduction of the luminescence
intensity, which is reduced by over three orders of magnitude in the phospho-
rous doped samples of series I (not shown here) It has to be concluded that
additional recombination centers are present in the phosphorous doped films.
Therefore luminescence spectra of all three series including the visible re-

71
4. Properties of ZnO and Mgx Zn1x O single layers

I6 I3a

0.1 mbar

PL intensity [arb.u.]
0.016 mbar

210-3 mbar

310-4 mbar

3.34 3.35 3.36 3.37 3.38

Energy [eV]

0.01 wt.% P2O5 0.1 wt.% P2O5 1 wt.% P2O5


10-1 mbar
10 K 10-1 mbar 10 K 10 K
10-1 mbar

1.6 10-2 mbar


1.6 10-2 mbar
CL intensity (arb. u.)

CL intensity (arb. u.)

CL intensity (arb. u.)

2 10-3 mbar
1.6 10-2 mbar
2 10-3 mbar
2 10-3 mbar
3 10-4 mbar
3 10-4 mbar

3 10-4 mbar

a) b) c)
1.5 2.0 2.5 3.0 3.5 1.5 2.0 2.5 3.0 3.5 1.5 2.0 2.5 3.0 3.5
energy (eV) energy (eV) energy (eV)

Figure 4.9: (a) Near band edge region of low temperature PL spectra of the samples
of series I. (Data obtained by Gabriele Benndorf) (b) Near band edge and deep
emission CL spectra of all three series. (Data obtained by Jan Zippel)

gion were obtained by means of cathoduluminescence and are depicted in


Fig. 4.9 (b). Four main luminescence features are visible: the near band edge
luminescence in the UV region (nbe), as discussed previously, a blueviolet
luminescence band (bv) centered between 2.8 and 3.0 eV, a green lumines-
cence band (g) centered around 2.5 eV and a redorange luminescence band
(ro) observed between 1.8 and 2.0 eV. All of these luminescence features were
observed previously. The bv luminescence was ascribed to a donor acceptor
transition probably involving the zinc vacancy [206,216] . The green luminescence
is commonly ascribed to two possible origins, copper on zinc site CuZn and the
oxygen vacancy VO [217219] . The ro luminescence was proven to be acceptor
related, and is tentatively assigned to oxygen on interstitial site Oi [220] . The

72
4.1. Properties of homoepitaxial ZnO thin films

high intensity of the deep luminescence gives an explanation for the quenching
of the nbe luminescence, as a number of alternative recombination channels is
present. The ro luminescence is observed in all of the samples investigated
and almost independent of the oxygen partial pressure applied during growth.
Again, like in the undoped samples, the samples of the three series discussed
can be collected in two groups, group B grown at oxygen partial pressures of
0.016 mbar and above and group A grown at oxygen partial pressures of 2103
and below. While in the samples of group A the bv luminescence is dominant,
the samples of group B are dominated by the nbe luminescence, apart from
the samples of series III, which all show a significant contribution to the lu-
minescence in the nbe region. The appearance of the bv luminescence in the
samples grown at the smallest oxygen partial pressures is somewhat disturb-
ing, as if it is related to the zinc vacancy its formation should be favored under
oxygen rich conditions. Nevertheless, regarding the broad luminescence it is
most likely that this feature is related to a donor acceptor pair recombination.
The observation of this defect therefore must be related with the formation of
a more shallow acceptor which is not present in the samples of group B. This
acceptor might be related to phosphorous, however the zinc vacancy cannot
be ruled out completely, as this defect has the highest probability among ac-
ceptor defects to form upon a high Fermi-level especially under oxygen rich
conditions [191194,221,222] . Even under zinc rich conditions it is very likely to be
formed if the Fermi-level is high, but results of ab-inito studies vary regarding
this issue. Some studies identify the zinc vacany as the most probably formed
defect under the presence of a high Fermi-level [191,193,194] . Such conditions are
present in the samples of group A, as proven by the transport measurements
and space charge region spectroscopy given in the following section. It is also
evident, that phosphorous doping must play a role in the formation of the ac-
ceptor, as it is not present in the nominally undoped films. It however remains
unclear, if phosphorous itself takes part in the acceptor on a microscopical
level, or if the lattice distortion or the shift in the Fermi-level introduced by
the phosphorous incorporation promotes the formation of the acceptor. More
detailed investigations will be necessary to understand the behavior observed.

Transport properties and defect states


To further evaluate the influence of the incorporation of phosphorous on the
electronic defect structure Hall-effect and TAS measurements were carried out
on the samples. In order to validate the approach of Hall-effect measurements
the conductivity of typical homoepitaxial ZnO:P thin films is compared to the
conductivity of the substrate, as depicted in Fig. 4.10. It can be clearly seen,
that the conductivity even of the film with the lowest conductivity exceeds that
of the substrate by more than 3 orders of magnitude at room temperature. As
the substrates are affected by a high degree of compensation the activation of
deeper donors takes place in the substrate than in the films. This leads to a
faster decrease of the substrate conductivity towards low temperatures com-
pared with the thin films. The conductivity of the substrates further decreases
by several orders of magnitude in the annealing process preceding the growth.

73
4. Properties of ZnO and Mgx Zn1x O single layers

103

102
[-1cm-1]
101 Thin film (high conductivity)
Thin film (low conductivity)
100 Substrate (as received)

10-1

10-2

10-3

10-4
2 4 6 8 10 12 14 16
1000/T [1/K]

Figure 4.10: Conductivity of two typical homoepitaxial ZnO:P thin films compared
to the conductivity of a CT substrate. Two samples from series I were chosen, the
one with the highest and the one with the lowest conductivity.

Table 4.3: Donor and acceptor concentration and donor activation energy in ZnO:P
thin films.

p(O2 )(mbar) ED (meV) ND (1016 cm3 ) NA (1016 cm3 ) ED (TAS) (meV)


0.1 >8 200 50 unknown unknown
0.016 > 50 >5 >1 255
2 103 36 4 35 3 2.6 0.1 19 5
3 104 34 5 43 4 1.5 0.1 20 5

It has to be pointed out however, that the composition of the substrates varies
(as outlined in Chap. 3.2.3), between the batches and even within one indi-
vidual batch. This of cause influences the transport properties. Therefore the
substrates are measured at room temperature before growth of the thin films
and compared to the data of the thin films.
Hall-effect measurements revealed the mobility and free carrier concentration
in the samples of series I, which are given in Fig. 4.11. A very good qualita-
tive agreement exists between the transport properties of these films and the
nominally undoped homoepitaxial thin films discussed in Chap. 4.1.1 (compare
Fig. 4.3) for both the mobility (Fig. 4.11 (a)) and the free carrier concentra-
tion (Fig. 4.11 (b)). Maximum mobilities of more than 800 cm2 /Vs are reached
in the film grown at 2 103 mbar. Again a modeling of the mobility was
done employing an exact self consistent solution of the Boltzmann transport
equation using the acceptor concentration in the films as only free parameter.
However an equally good agreement of the calculated and measured data as

74
4.1. Properties of homoepitaxial ZnO thin films

1.000 1.000

H [cm2/Vs]
100
800
10
H [cm2/Vs]

600
1
30 50 70 100 200 300

400 T [K]
p(O2) [mbar]

310-4
200 210-3
0.016
0.1
0
0 100 200 300 400
T [K]

1018

1017 p(O2) [mbar]


0.1 0.1 corrected
nH (cm-3)

210-3 fit 0.1


310-4 fit 310-4
0.016 fit 210-3
1016

1015

1014
0 5 10 15 20 25
1000/T (1/K)

Figure 4.11: Hall-mobility (a) and free carrier concentration (b) as a function of
temperature in the thin films of series I.

75
4. Properties of ZnO and Mgx Zn1x O single layers

for the undoped film grown at 2 103 mbar was not achieved. The reason-
ing is comparatively simple, as the partial contribution of a highly conductive
layer is observed for these samples. Its influence can be easily seen from the
temperature dependence of the free electron concentration (see Fig. 4.11 (b)),
where at low temperatures a higher free carrier concentration is observed from
the experimental data than it would be theoretically expected. This behav-
ior is most pronounced for the sample grown at 0.1 mbar, where a correction
following Refs. [223,224] was employed to separate the characteristic transport
parameters of the non degenerate layer. It is however visible, that a com-
plete correction of the influence of the degenerate layer was not possible, which
can be understood by the fact, that the dopant concentration does not drop
sharply at a imaginary interface, as assumed in the model [223,224] . The dopant
concentration will therefore be high enough in the bulk of the film, leading
to an overlap of the donor wave functions and to a reduction of the appar-
ent dopant concentration. For the films of series I grown at oxygen partial
pressures below 0.1 mbar the correction scheme was not applied, as the deter-
mination of the properties of a degenerate layer was not unabigiously possible.
Also, the influence of the correction would only be of minor importance for the
determination of the donor ionization energy, as it contributed to the free car-
rier concentration measured only at low temperatures. A significant influence
on the low temperature mobility however exists, which can not be corrected
within the model and with the experimental data available. An overestimation
of the acceptor concentration is therefore possible, which in turn will lead to
an uncertainty in the donor concentration identified.
The donor activation energies and cocentrations were determined from the
free carrier concentration using a single donor charge balance equation and are
given in Tab. 4.3. Also the donor activation energies as determined by TAS are
tabulated. Remarkable differences are visible between these values indicating
that again both methods do not probe the same sample regions. This again
supports the assumption that the distribution of donors in the films is not ho-
mogeneous.
Hall-effect measurements of the samples of series II and III revealed that these
samples were degenerately doped in the 1019 cm3 region. Temperature de-
pended measurements were carried out on these samples and again revealed
that the high dopant concentrations observed were not homogeneously dis-
tributed. A distinct temperature dependence of both the mobility (Fig. 4.12 (a))
and the free electron concentration (Fig. 4.12 (b)) is observed. Such a behavior
is not expected if the whole layer is degenerately doped. Only a small layer
in the sample can show this high carrier concentration, and its influence can
be separated using the correction scheme discussed before [223,224] . Fig. 4.12 (a)
shows the mobility and Fig. 4.12 (b) free carrier concentrations of the samples
of series II group A as measured and after the application of the correction
scheme. It is visible that the transport data of the bulk of the film can be
largely reconstructed. The abilities of the correction scheme are particularly
impressive regarding the mobility, which increases by one order of magnitude.
Also, the free carrier concentration reduces by roughly one order of magnitude,

76
4.1. Properties of homoepitaxial ZnO thin films

1.000
210-3
210-3 corrected
210-3 annealed
310-4
310-4 corrected
310-4 annealed
H (cm 2/Vs)

100

10

30 50 70 100 200 300


T (K)

1020

1019

1018 210-3
n (cm -3)

210-3 corrected
210-3 annealed
17 310-4
10
310-4 corrected
310-4 annealed
16
10

1015

1014
0 10 20 30 40 50
1000/T (1/K)

Figure 4.12: Hall-mobility (a) and free carrier concentration (b) as a function of
temperature in the thin films of series II, group A.

77
4. Properties of ZnO and Mgx Zn1x O single layers

101

100

10-1
annealed (cm)

10-2

10-3

10-4

10-5
10-5 10-4 10-3 10-2 10-1 100 101
as-grown (cm)

Figure 4.13: Resistivity of homoepitaxial phosphorous doped ZnO thin film as


grown and after annealing. Meaning of the symbols: Series I (red ), series II (blue) and
series III (green). Oxygen partial pressures during growth 0.1 mbar (), 0.016 mbar
(N), 2 103 (H) and 2 103 ().

still remaining in the high 1018 cm3 to 1019 cm3 region, which should still be
very close to the degenerate limit. Despite the remarkable success of the cor-
rection scheme, a full reconstruction of the bulk properties remains beyond its
abilities. This is explained by the comparatively simple model used. Addi-
tional complicating elements like a temperature dependence of the mobility in
the degenerate layer and a inhomogeneous distribution of the dopants in the
degenerately doped region were considered [153,154,225] , however these approaches
introduce further unknown parameters. The determination of these parame-
ters is not trivial, if possible at all, from the Hall-effect data available in this
study. Certain parameters, like the distribution of dopants in the film can not
be determined unambiguously. It is therefore questionable if the data obtained
by such an approach would give a better representation of the physical reality.
Such a reasoning is also followed in the paper by D. C. Look introducing some
of these improvements in the correction scheme [154] and it is concluded that a
Hall-effect analysis alone can not yield the dopant profile. In the present work
a different approach will be followed to improve the informative values of the
Hall-effect measurements, refer to Chap. 4.2.
As the aim of this study is the exploration of the properties of homoepitax-
ial phosphorous doped ZnO thin films, and the assertion of the applicability
of phosphorous as a p-type dopant in this layers a short review of the liter-
ature will be given at this point. Several authors have reported the transi-
tion to p-type material with phosphorous doping, both in heteroepitaxial and
homoepitaxial samples [205207] . However, all of these samples showed n-type

78
4.2. Evaluation of the transport properties by mobility spectrum analysis

conductivity in the as grown state. A thermal annealing step at temperature


between 800 C and 900 C was necessary in all cases to activate the phospho-
rous related acceptors and transform the samples into p-type material. Such
annealing experiments were likewise carried out within this study. The mate-
rial was exposed to a 700 mbar oxygen atmosphere at 850 C for 5 minutes. The
transport data measured after the annealing step are given for the samples of
series II group A in Fig. 4.12. The degenerate layer was completely removed.
However a transition to p-type conductivity is not observed. Also the concen-
tration of acceptors can not be higher than the concentration of shallow donors,
as can be inferred from their influence on the free carrier concentration.
The influence of the annealing process on the transport properties was studied
and is summarized in Fig. 4.13. The resistivity of the films prior to and after
the annealing are depicted. No change of the carrier type is observed in any
of the samples. The change is most distinct for the samples having a high
initial resistance. In this case the resistance decreases by at least three order
of magnitude, which must originate from the annealing of acceptor defects.
For some samples a slight increase of the resistance is observed, however in
this case the effect consists mostly of the annealing of the defects necessary to
form the degenerate layer, as discussed for the samples of series II. If at all a
tendency towards increased acceptor concentration is observed this occurs for
the samples of group B, grown at high oxygen partial pressures.

4.2 Evaluation of the transport properties by


mobility spectrum analysis
As outlined above, the determination of the mobility and free carrier concen-
tration from Hall-effect measurements in ZnO samples can be complicated by
the presence of several conducting paths with different mobilities. A separation
of the individual contributions to the overall transport properties can be done
from their different dependence on temperature [153,154,223225] . This approach
however requires detailed theoretical knowledge of the temperature dependence
of the individual contributions. The applicability of this approach is typically
limited to two-layer system, and even in this case several assumptions have to
be made on the nature of the transport processes involved. Therefore other
possibilities for the separation of multiple transport channels have to be consid-
ered. The most promising approach employed in order to extend the informa-
tional value of Hall-effect measurements in multilayer system is the application
of magnetic field dependent measurements. All of these approaches rely on the
fact that the components of the conductivity are weighted sums of the differ-
ent conduction paths in the sample. Rather simple approaches include three
distinct conduction paths with different mobilities and carrier concentrations
(which are typically electrons and holes in different bands) [226,227] .
However, following a seminal work of Beck and Anderson a mobility spec-
trum can be defined, which can in principle account for any finite number of
carriers of different mobilities contributing to the total conductivity [228] . They

79
4. Properties of ZnO and Mgx Zn1x O single layers

3000
0.5

1
2500

1.5

2000 2
[cm2 /Vs]

2.5
1500

1000 3.5

4
500
4.5

5
30 35 40 45 50 60 70 80 105 120 145160 180 220 270 320
Temperature [K]

Figure 4.14: Mobility spectra of a Mgx Zn1x O layer grown on the (0001) face of a
ZnO single crystal as obtained from the f-MEMSA algorithm. The color scale repre-
sents the log10 of the occupation probability at the respective mobility channel. Red
dots indicate the mobility calculated from a conventional Hall-effect analysis. The
solid red line gives the mobility as a function of temperature only limited by lat-
tice scattering. (Data obtained by Danilo Brger (Forschungszentrum Rossendorf),
spectra deconvolved by Robert Heinhold)

showed, following the argumentation of Ref. [229] that if a relaxation time can
be defined, and if that relaxation time is constant around a cyclotron orbit,
the components of the conductivity tensor can be written as
Z
s()d
xx (B) = 2
(4.3)
1 + (B)

and
Bs()d
Z
xy (B) = . (4.4)
1 + (B)2
The problem of the determination of the mobility spectrum now consists of
the inversion of these integrals. Several approaches have been proposed in
order to solve the problem, which will briefly be discussed. Apart from the
original approach techniques such as quantitative mobility spectrum analysis
(QMSA) [230] , improved quantitative mobility spectrum analysis (i-QMSA) [231] ,
maximum entropy mobility spectrum analysis (MEMSA) [232] and full maxi-
mum entropy mobility spectrum analysis (f-MEMSA) [233] have been developed.
A brief summary of the advantages and disadvantages of the specific methods

80
4.2. Evaluation of the transport properties by mobility spectrum analysis

is summarized in Ref. [234] by the authors of Ref. [232] . The main advantage of
the MEMSA algorithms over the competing methods are that no additional
corrections and manipulations of the measured data are necessary and no ar-
tificial ghost peaks are seen in the spectra. The f-MEMSA method introduces
two additional advantages, namely a matrix based inversion algorithm, which
both increases the speed of convergence and reduces the sensitivity to errors,
and the evaluation of the full set of available measurement data. The MEMSA
algorithm instead only includes 50% of the available data as constrictions to
the inversion problem, which is justified by the fact that the components of the
conductivity tensor are ideally odd or ideally even functions of the magnetic
field in noise free experiment. An additional improvement to the MEMSA al-
gorithm is given in Ref. [235] , where a more accurate treatment of the noise level
involved is given. Several approaches have been implemented and tested on
both synthetical and experimental data by R. Heinhold, indicating, that the
most stable algorithm is indeed the f-MEMSA algorithm. It has to be noted
however, that the f-MEMSA algorithm uses a data truncation algorithm in
the matrix inversion problem. Herein Eigenvalues under a certain magnitude
will be removed from the set of solutions, reducing the dimensionality of the
problem. While this both increases the speed of convergence and reduces the
influence of small disturbances, the actual solution now depends on the limit
chosen. This again introduces some ambiguity to the solution.
As outlined in the preceding section multiple parallel conduction paths can be
found in ZnO thin films and single crystals. Further complications are possible
in the presence of a polarization induced 2DEG which will be discussed in sec-
tion 5.3. In principle at least three different transport channels can be found
in a ZnO/Mgx Zn1x O single heterostructure, namely bulk conduction, surface
conduction and interface conduction, where the mobility will differ between
the individual processes. The application of the MEMSA algorithm does in
principle allow the separation of these individual conduction paths.
In the present study a ZnO/Mgx Zn1x O single heterostructure was grown on
the Zn face of a hydrothermal ZnO single crystal. A 1 m thick ZnO buffer
layer was introduced below the 1.3 m thick Mgx Zn1x O layer, in order to re-
duce the density of impurities at the heterointerface, and to avoid structural
defects influencing the interface conduction. Bulk conduction can therefore
occur in the substrate, the ZnO buffer layer and the Mgx Zn1x O layer. As the
substrates show a very low conductivity after annealing (refer to Chap. 4.1.2),
no significant contribution to the overall conductivity should be expected. The
ZnO buffer layer is grown at an oxygen partial pressure of 0.016 mbar, which
again results in a low conductivity of this layer (refer to Chap. 4.1.1). The
Mgx Zn1x O layer however is grown at 0.004 mbar, and therefore a significant
contribution to the conductivity is assumed in this layer. However, due to the
[0001] polarity of the sample, the formation of a 2DEG is assumed to take place
at the heterointerface (see Fig. 5.1) partially draining free electrons from the
Mgx Zn1x O layer. Interface conduction can both occur at the heterointerface
and the interface between the substrate and the buffer layer. At the heteroint-
erface the mobility of the 2DEG should exceed that of the bulk electrons. The

81
4. Properties of ZnO and Mgx Zn1x O single layers

interface transport channel between the buffer layer and the substrate does
only exist if a high density of impurities is found near the interface, resulting
in a reduced mobility. An equal argumentation holds for the surface conduction
channel. Figure 4.14 shows a compilation of the individual mobility spectra
calculated from field dependent measurements at different temperatures. Mag-
netic fields between -8 and 8 T were applied with widths between 0.2 and 0.5
T. The mobility derived from standard Hall-effect analysis is given as well as
the maximal mobility possible in bulk ZnO in the limit of dilute doping, as cal-
culated from an iterative solution of the Boltzmann transport equation. Only
positive mobilities are displayed, which in this case means that only electron
transport is present. The determination of the mobility spectra was carried
out allowing for the possibility of hole transport, which however was not found
in the spectra. Some contribution can be seen below zero mobility, which is
introduced by a broadening of the peak at low electron mobilities. The cor-
responding (here negative) values in the mobility are therefore omitted from
the plot. The f-MEMSA evaluation allows the separation of two individual
transport channels with different mobilities at almost all temperatures. The
channel with low mobility corresponds to the degenerate layer either found at
the surface or the buffer layer substrate interface. The channel with higher
mobility most likely originates from bulk conduction, as the contribution from
this transport mechanism is only seen above 35 K. It is therefore concluded,
that the carriers contributing to the transport process freeze out at low temper-
atures, which is expected from bulk conduction, but does not take place in the
case of a 2DEG. One might now speculate that a 2DEG is present, and this as-
sumption will receive some support by the data given in Fig 5.7. Between 50 K
and 120 K a third peak is visible in the spectra at higher mobilities. Though
no systematic behavior is observed, most likely due to the high conductivity
of the other two conduction processes, the values obtained from the spectra
lie in a range which is physical reasonable. No contribution of the transport
mechanism is however seen below 50 K, which is again consistent with the data
given in Fig 5.7. Such a behavior is surprising as metallic conduction would be
expected for the 2DEG. One possible explanation is, that the energetic states
related to the degenerate layer are found lower than the quantized states at
the heterointerface, and some mechanism exists which allows the relaxation
of free electrons from the 2DEG into the degenerate layer. An influence of
surface states on the properties of a 2DEG has indeed been seen in the related
GaN/AlGaN system [236] . However the nature and position of the degenerate
layer is still not fully known in ZnO, and more detailed investigations are re-
quired to support this assumption.

4.3 Properties of phosphorous doped MgxZn1xO


thin films grown on ZnO
In order to investigate the influence of phosphorous doping on the properties
of Mgx Zn1x O thin films two series of samples were grown from Mgx Zn1x O:P

82
4.3. Properties of phosphorous doped Mgx Zn1x O thin films grown on ZnO

(a) 4.5
concentration (at.%)
4
magnesium

3.5

2.5

1.5

1
10-4 10-3 10-2 10-1

p(O2) (mbar)

(b) 0.8
concentration (at.%)

0.75
phosphorous

0.7
0.65
0.6
0.55
0.5
0.45
0.4
0.35
10-4 10-3 10-2 10-1

p(O2) (mbar)
Figure 4.15: Magnesium (a) and phosphorous (b) content of the Mgx Zn1x O thin
films deposited on ZnO substrates as determined by PIXE versus the oxygen partial
pressure applied during growth. (Data obtained by Christoph Meinecke, graphed by
Dr. Holger von Wenckstern)

83
4. Properties of ZnO and Mgx Zn1x O single layers

targets with a nominal P2 O5 -content of 1 wt and a nominal MgO-content of


2 wt.%. The Mg- and P- content of the thin films was determined by PIXE
and is depicted in Fig. 4.15 versus the oxygen partial pressure applied during
growth. With increasing oxygen partial pressure, the concentration of both
Mg and P decreases. This suggests, that the incorporation of both elements
follows the same mechanism. It is known, that Mg replaces Zn. In concordance
with a defect model for Sb and As doped samples [237] it was calculated, that
phosphorous is incorporated mainly onto Zn site as well [238] . This hypothesis
is supported by the present results. Further experimental indications are given
in Ref. [206,216] . It is further interesting, that the incorporation of dopants on Zn
site is found to be higher for lower oxygen partial pressure. A possible expla-
nation for this behavior is the fact, that at the growth temperatures employed
the vapor pressure of Zn is more than one order of magnitude higher than that
of for example Mg [92] . The film composition is not only determined by the
target composition and the background gas, but further controlled by dynamic
processes on the sample surface. Zinc evaporates partially from the surface
leading to an increase in the dopant concentration in the film. This behavior is
experimentally observed for Mg [215] . It is also in accordance with recent theo-
retical calculations [159] . Further, the background pressure influences the spatial
distribution of the plasma composition, as the atomic momentum depends on
the actual atomic weight of the element, and therefore light elements (in this
case the oxygen) are more affected by scattering processes in the plasma than
heavier elements (as Mg, P and Zn) at the same kinetic energy [48] .
The HR-XRD 2 scans of the (002) reflex for the MgZnO:P samples are
depicted in Fig. 4.16. On top of the reflexes Pendellsung oscillations are visi-
ble, indicating a very smooth interface between the substrate and the film. All
samples, grown at an oxygen partial pressure > 0.1 mbar, show two peaks, one
corresponding to the substrate and the other to the film. The c-axis of the
films is expanded compared to that of the substrate. From relaxed films grown
on sapphire [85,91] it is known, that with increasing Mg concentration the a-axis
expands, while the c-axis shrinks. In the pseudormorphic growth mode the
a-axis is fixed due to the in-plane lattice match of the film and the underlying
ZnO substrate, resulting in an relative expansion of the c-axis. Therefore in
Mgx Zn1x O samples grown pseudomorphically the c-axis constant is reduced
compared to the substrate but larger than that of a relaxed Mgx Zn1x O thin
film. For a detailed discussion refer to Chap. 3.1.2. It is concluded, that the
incorporation of phosphorous expands the ZnO lattice significantly, as already
observed for the homoepitaxial ZnO thin films doped with phosphorous, ex-
ceeding the reduction of the c-axis due to the Mg. For the sample grown at
0.1 mbar only a single peak is observed in the 2 scans, suggesting a per-
fect, both in-plane and out-of-plane lattice match between the film and the
substrate.

Reciprocal space maps of the symmetric (002) (Fig. 4.17 (a)) and the asym-
metric skew (101) (Fig. 4.17 (b)) reflex of the MgZnO:P sample with the highest
Mg and P concentration are shown exemplarily. As can be seen, the peak posi-
tions only differ in the out-of-plane direction, but have the same in-plane value.

84
4.3. Properties of phosphorous doped Mgx Zn1x O thin films grown on ZnO

0.1 mbar
0.016 mbar

Intensity (arb. units)


2 10-3 mbar
3 10-4 mbar

34.1 34.2 34.3 34.4 34.5 34.6

2 ()

Figure 4.16: HR-XRD triple axis 2 scans of the symmetric (002) reflex of
Mgx Zn1x O:P films grown at the indicated oxygen partial pressures. For the sam-
ples grown at oxygen partial pressures below 0.1 mbar thickness fringes are visible,
indicating the very smooth interface between the substrate and the film. The peaks
at 2 around 34.42 correspond to the ZnO substrate reflections. (Data obtained by
Dr. Michael Lorenz, graphed by Dr. Holger von Wenckstern)

This indicates a pseudomorpic growth for all samples grown at oxygen partial
pressures below 0.1 mbar. The a- and c-axis values for all films investigated
are given in Tab. 4.4.
The temperature dependence of the Hall-mobility is depicted in Fig. 4.18 (a),
that of the free carrier concentration is shown in Fig. 4.18 (b). A minor influ-
ence of a degenerate layer is detected for some of the samples, and eliminated
using the correction scheme outlined in Ref. [223] . High electron mobilities are
observed in the MgZnO:P samples grown at oxygen partial pressures below
0.1 mbar, comparable to that in the previously discussed homoepitaxial ZnO:P
samples underlining the very high structural quality of the samples. While for
these homoepitaxial ZnO:P samples the maximal mobility was observed in the
sample grown at 3 104 mbar, the maximal mobility in the homoepitaxial
MgZnO:P was observed for the samples grown at 2 103 mbar for a nominal
MgO content of 1 wt.% and 2 wt.% likewise.
From the temperature dependence of the free carrier concentration the donor
activation energy ED and concentration ND was determined using a single
donor model. As the donor concentration is in the 1017 cm3 range, the activa-
tion energy is reduced by the beginning overlap of the donor states. The donor
activation energy then follows
 1/3 !
ND
ED = ED0 1 , (4.5)
NC

as given e.g. in Ref. [98] . Here NC denotes the critical donor concentration, ED0
is the donor activation energy in the dilute limit. The donor activation energy
versus donor concentration is depicted in Fig. 4.19 (a). A least squares routine

85
4. Properties of ZnO and Mgx Zn1x O single layers

(a) MgZnO:P (0 0 2) Log10(Intensity (counts))

3.5
3.85
3

3.84 2.5
q (nm )
1

2
3.83
1.5

3.82 1

0.5
3.81
10 9 8 7 6 5
1 3
q (nm ) x 10
||

(b) MgZnO:P (1 0 1) Log10(Intensity (counts))

2.5
4.042

4.04 2
q (nm )
1

4.038
1.5
4.036
1
4.034

4.032 0.5

4.03
6 4 2 0 2
1 3
q|| (nm ) x 10

Figure 4.17: HR-XRD reciprocal space maps of (a) the symmetric (002) and (b)
the asymmetric skew (101) reflex of a Mgx Zn1x O:P film grown on ZnO at 2
103 mbar. The color scale represents the log10 of the scattered X-ray intensity
(counts). The horizontal broadening of the reciprocal lattice points of both substrate
(top reflection) and thin film (bottom reflection) is due to the mosaizity of the used
ZnO substrate. The homoepitaxial film follows the substrate mosaizity, resulting in
an equal broadening. (Data obtained by Dr. Michael Lorenz)

86
4.3. Properties of phosphorous doped Mgx Zn1x O thin films grown on ZnO

1.000
1% MgO 2% MgO
1.6 10-2
800
2 10-3 2 10-3
3 10-4 3 10-4
(cm 2/Vs)

600

400

200

0
0 50 100 150 200 250 300 350
T (K)

1% MgO 2% MgO
1018 1.6 10-2
2 10-3 2 10-3
3 10-4 3 10-4
n (cm -3)

1017

1016

1015
0 5 10 15 20 25
1000/T (1/K)

Figure 4.18: Electron mobility (a) and free electron concentration (b) versus tem-
perature in selected Mgx Zn1x O:P thin films grown on ZnO, prepared from targets
with a nominal MgO-content of 1 wt.% and 2 wt.%, respectively, as labeled. (Graphed
by Dr. Holger von Wenckstern)

87
4. Properties of ZnO and Mgx Zn1x O single layers

Table 4.4: Oxygen partial pressure p(O2 )during growth, Mg content x and P
content cp both in atomic percent, lattice parameters, and room temperature electron
concentration n and Hall-mobility H of MgZnO:P thin films grown on ZnO. The
lattice constants of a substrate are given for reference.

p(O2 ) (mbar) x (%) cp (%) c (nm) a (nm) n (cm3 ) H (cm2 /Vs)


0.1 1.09 0.43 0.52070 0.32498 1.6 1017 0.6
0.016 2.25 0.47 0.52208 0.32500 1.8 1018 4.7
2 103 3.81 0.67 0.52305 0.32498 4.9 1017 140
3 104 4.71 0.84 0.52285 0.32499 1.3 1018 84
substrate 0.52069 0.32499

was applied to model the experimental data using Eqn. 4.5. From the fit, a
critical concentration of NC = 8.3 1018 cm3 and a donor activation energy of
ED0 = 56 meV is derived, which is close to the activation energy of the effective
mass donor [239] .
In order to investigate effects of thermal activation of the phosphorus dopant,
the Mgx Zn1x O:P layers were annealed in a 700 mbar oxygen atmosphere at
850 C for 5 minutes, the same conditions as for the homoepitaxial ZnO:P
samples. The resistivity of the samples prior to and after the annealing are
compared in Fig. 4.19 (b). The change in resistivity strongly depends on the
oxygen partial pressure during growth. The samples grown at the highest oxy-
gen partial pressures p(O2 )= 0.016 mbar and p(O2 )= 0.1 mbar show the highest
increase in resistivity by 6 and 4 orders of magnitude, respectively. This result
is in concordance with that obtained on the homoepitaxial ZnO:P thin films
were high resistivities are only achieved in the samples grown at high oxygen
partial pressures. The change in the resistivities upon annealing, however, is
much stronger in the Mgx Zn1x O thin films. The samples of group A, having a
higher phosphorous content show only slight changes in the resistivity. Again
these results coincide with that obtained on the homoepitaxial ZnO:P thin
films, where only a slight change is observed for these samples. It is concluded,
that despite having the highest phosphorous concentrations, the conditions are
not present in these films to form a high density of phosphorous related ac-
ceptors. One reason therefore might be the preferential incorporation of P on
Zn site. The mechanism involving the formation of PZn -2VZn might not work
completely in these samples due to a high density of interstitial Zinc, which
is observed for the nominally undoped homoepitaxial ZnO thin films. For the
samples grown at higher oxygen partial pressures a transition to p-type con-
ductivity might well be possible from the high density of acceptors present in
the sample, however the density of compensating donors is still to high for the
transition to occur. One possible reason for that is the high density of iron and
silicon impurities in the sample as determined by PIXE in the homoepitaxial
ZnO:P samples of series I (see Tab. 4.2). While these impurities do not have a
significant influence on the transport properties as the defect levels introduced

88
4.3. Properties of phosphorous doped Mgx Zn1x O thin films grown on ZnO

70

60

50
ED (meV)

40

30

20

10

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
ND1/3 (106 cm -1)

107

106
annealed / as-grown

105

104

103

102

101

100

10-1
10-4 10-3 10-2 10-1

p(O2) (mbar)

Figure 4.19: (a) Donor activation energy versus donor concentration in ZnO:P of
series I and Mgx Zn1x O:P thin films. The red line is a least squares fit to Eqn. 4.5.
(b) Ratio of the resistivities of Mgx Zn1x O:P thin films grown on ZnO after and
before annealing versus oxygen partial pressure during growth. Full and open symbols
indicate films with a nominal MgO content of 1 wt.% and 2 wt.%, respectively.

89
4. Properties of ZnO and Mgx Zn1x O single layers

by them are deep, their high density will still compensate acceptors and pre-
vent the creation of free holes. It is therefore necessary to reduce the density
of these impurities. This however is not trivial in the PLD process as these
impurities were already present in the target, being most likely introduced by
the homogenisation and sintering process. The target preparation process is
required to add the dopant, and in the case of Mgx Zn1x O also to produce the
host material. For ZnO the possibility exist to use single crystalline targets of
much higher purity, and to add the dopants via an alternative route. Nitrogen
for example was introduced from a plasma, which allows a very high purity of
the doping agent and finally led to p-type conduction and the fabrication of
LEDs [66] . This route however is not accessible for Mgx Zn1x O or other ternary
alloys which in turn are necessary for the fabrication of highly efficient devices.
Therefore either complicating steps involving multiple single crystalline tar-
gets or different growth methods like MBE have to be considered. All these
steps would result in a significant complication of the growth process. In order
to maintain the advantages of the PLD process such as simplicity, scalabil-
ity and comparatively cheap design, one of the most critical challenges is the
production of highly pure target material.

4.4 Summary and conclusion


The results of the present chapter can be summarized as follows:
The properties of undoped and phosphorous doped ZnO and Mgx Zn1x O
thin films grown on ZnO are strongly influenced by the oxygen partial
pressure during PLD growth. The oxygen partial pressure influences the
growing films in more than one way.
1. The IIVI ratio in the films is controlled by the oxygen partial pres-
sure. For films grown at low oxygen partial pressures an increased
c-axis lattice constant is observed for both the ZnO and Mgx Zn1x O
thin films, most likely due to an increasing density of charged de-
fects. The thesis of an increase of the density of intrinsic defects
related to oxygen deficiency in the films is supported by an increase
in the zinc interstitial related luminescence and the free electron
concentration upon decrease of the oxygen partial pressure. It is
concluded, that oxygen in the background gas atmosphere is neces-
sary in order to grow stoichiometric ZnO and Mgx Zn1x O film.
2. The composition of the group II constituents depends on the back-
ground pressure in the chamber. The concentration of the lighter
Mg atoms compared to Zn atoms is reduced by almost a factor of
three within the range of oxygen partial pressures available in the
present growth setup. Interestingly the concentration of P does
change accordingly. It is concluded, that the atomic mass (which
differs between Mg and P by only 20%) of the individual element
influences the kinetic scattering processes with the background gas
and therefore the plasma composition at the substrate position.

90
4.4. Summary and conclusion

It is therefore concluded that the versatility of the growth setup can be


improved if the influences of the background gas on the sample compo-
sition could be separated. Two modifications of the PLD setup would
probably allow at least a partial realization of this goal.

1. The use of a non uniform gas atmosphere.


The application of a gas mixture, for example oxygen and an inert
gas, inside the growth chamber could probably allow a control of
the composition of the group II constituent without changing the
IIIV ratio, therefore maintaining stoichiometric growth. However,
a minimum oxygen partial pressure is necessary in order to maintain
stoichiometry. This limits the total pressure attainable to values
above the oxygen partial pressure.
2. The introduction of a plasma source.
During laser plasma formation only a small fraction of the available
oxygen molecules is cracked. It is reasonable to assume in the light
of literature work [240] , that molecular oxygen is only responsible for
a vanishing fraction of the total oxygen incorporation. The range
of available gas mixtures in the growth chamber can be further ex-
panded, if the availability of ionic oxygen species can be controlled
without changing the base chamber pressure. Such a approach could
allow a chamber pressure virtually below the oxygen partial pres-
sure, as the availability of activated oxygen increases. Additionally,
it is found in the literature [241] that the incorporation of nitrogen
is promoted by the activation in a plasma environment, eventually
leading to p-type conductivity [66,242] .

The similarities of the incorporation of Mg and P indicate that P is like


Mg mainly introduced into the ZnO host lattice on Zn site, in accordance
with theoretical calculations [238] .

A degenerate contribution to the sample conductivity was found in ho-


moepitaxial samples with a phosphorous concentration with 0.1% and
above, which could partially be removed by annealing in oxygen, indicat-
ing that at least in this case the origin of the degenerate layer is a layer
of intrinsic defects, likely to be formed at the surface.

It was shown that the separation of the conduction of the degenerate


layer and the bulk of the film was partially possible in the evaluation of
the Hall-effect measurements employing a correction scheme as proposed
in Refs. [223,224] . A complete separation however was not possible without
further assumptions, as the incorporation of the defects responsible for
the formation of the degenerate layer is not homogeneous.

It was shown that the separation of the conduction of individual con-


ductive paths is possible without any assumption about the distribution
of the individual carriers from magnetic field dependent measurements
by the maximum entropy mobility spectrum approach. The individual

91
4. Properties of ZnO and Mgx Zn1x O single layers

properties of bulk conduction and the conduction of the degenerate layer


were revealed over a wide range of temperatures. Such a detailed study
of the parallel conduction processes in ZnO has not been reported so far.

The critical donor concentration and activation energy in the dilute limit
of the dominant donor in phosphorous doped ZnO and Mgx Zn1x O layers
grown on ZnO was determined to be of NC = 8.3 1018 cm3 and ED0 =
56 meV, respectively.

Transport measurements on phosphorous doped homoepitaxial ZnO thin


films showed high electron mobilities of up to 800 cm2 /Vs around 70 K and
170 cm2 /Vs at room temperature. Exact modeling of the transport prop-
erties using a numerical solution of the Boltzmann-transport-equation
indicated, that the scattering processes involved equal that of ZnO single
crystals [210] .

The thermal activation of phosphorous related acceptors was studied in


ZnO and Mgx Zn1x O thin films grown on ZnO. A transition to p-type
conduction was never observed. However an increase of the resistivity
by more than 6 orders of magnitude was observed in the Mgx Zn1x O:P
layers deposited at the highest oxygen partial pressures. It is concluded
that high oxygen partial pressures are necessary if p-type conduction is
to be achieved, as the compensation of acceptors is reduced by the low
density of intrinsic donors.

From the chemical analysis of the sample composition it is concluded that


concentrated efforts are necessary to improve the purity of the target
material, as the purity of the films can not exceed the purity of the
starting material.

92
Chapter 5

Quantum effects on carriers in


ZnO/MgxZn1xO heterostructures

5.1 Principle considerations


Free carriers in semiconductor heterostructures will be influences by the band
gap discontinuities between the constituents. If the spatial dimensions involved
are of the order of the carrier wavelength, quantum effects will become impor-
tant in the vicinity of the heterointerface. Quantized energy levels can be found
in potential minima, and tunneling processes can occur at potential maxima.
A detailed compilation of the numberous effects observed can be found in
Ref. [243] . Within this work only effects of quantization in potential minima in
heterostructures will be regarded. Two classes of minima will be investigated
and differentiated by the heterostructures where they occur in. First a single
ZnO/Mgx Zn1x O heterojunction represents a one sided potential barrier due
to the band offset at the heterointerface. Secondly a double heterostructure in-
troduces two energetic barriers. The ZnO/Mgx Zn1x O material system allows
the fabrication of type-I heterostructures [244] , therefore for both electrons and
holes the potential energy is lower in the ZnO than in the Mgx Zn1x O part of
the junction. If the layer sequence is Mgx Zn1x O/ZnO/Mgx Zn1x O and the
thickness of the ZnO layer is sufficiently small so that quantization occurs, such
a structure can be regarded as a quantum well with finite barriers. Electron
and holes will be confined in such a structure, leading to an increase of the
carrier density and recombination efficiency. Such a behavior was observed in
semiconductor junctions [245] . As both the ZnO and Mgx Zn1x O constituent
show a spontaneous polarization parallel to the c-axis which can be superim-
posed by a piezoelectric polarization likewise parallel to the crystallographic
c-axis additional complicating aspects have to be regarded. High quality het-
erostructures, as investigated in the present study grow pseudomorphic, and
therefore at least one of the layers is strained. The polarization necessarily
contains a piezoelectric component, which is discontinuous at the heterointer-
face. Additionally the spontaneous polarization in Mgx Zn1x O is a function
of the Mg content x, and an additional discontinuity will be introduced by
the change in x at the heterointerface. If the layer sequence is organized in

93
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

(0001)
(a) ZnO MgZnO

0 50 100 150 200


z (nm)
1020
(b) MgZnO ZnO

1019

n (cm-3)
1018

1017

1016

0 50 100 150 200


z (nm)

1020
(c) MgZnO ZnO MgZnO

1019
n (cm-3)

1018

1017

1016
195 200 205 210
z (nm)

Figure 5.1: Band diagram of ZnO/Mgx Zn1x O heterostructures. Stack-


ing sequence (a) ZnO/Mgx Zn1x O, (b) Mgx Zn1x O/ZnO, and (c)
Mgx Zn1x O/ZnO/Mgx Zn1x O. Solid black lines show the respective conduc-
tion and valence bands, vertical black lines indicate heterointerfaces. Dotted black
lines represent quantized energy levels of confined carriers. Solid green and blueish
lines symbolize hole and electron wavefunctions, respectively. The density of
electrons in the vicinity of the junctions is shown by a red solid line. In the QW
structure only the lowest energy levels are indicated.

94
5.1. Principle considerations

a fashion that the heterointerfaces are perpendicular to the crystallographic


c-axis, an electric field forms due to the discontinuity of the polarization. Such
a stacking sequence is very common due to the preferential growth of ZnO and
its compounds along the c-axis [8790] . All samples investigated in the present
study have their growth direction z parallel to the crystallographic c-axis. The
piezoelectric polarization counteracts the change in spontaneous polarization in
c-axis oriented growth, and might in principle cancel its influence completely.
Such a behavior is suggested in some studies for small Mg contents [35] , see
the discussion in Chap. 1. However for x > 0.15 a finite change in the polar-
ization at the heterointerface is evident from most studies [3538] . The electric
field introduced will add a linear change in the potential parallel to the growth
direction. Therefore a triangular potential is formed in the vicinity of the het-
erointerface for single heterojunctions. Depending of the layer sequence this
represents of a minimum in the E(z) relation for either electrons or holes,
which leads to the accumulation of the respective carriers, and the formation
of quantized states. The change in polarization will also produce an electric
field across a quantum well, which leads to the occurrence of the quantum
confined stark effect. In all cases the accumulation of carriers leads to the
formation of a space charge incorporating an additional potential, which in
turn influences the potential leading to the confinement of the carriers. Self
consistent solutions of the Poisson- and Schroedinger equation are necessary
to determine the actual distribution of carriers in position and energy space.
The calculations were carried out using the poisson solver Beta8c developed by
Gregory Snider [246,247] . The results of such calculations are shown exemplarily
in Fig. 5.1. For the Mgx Zn1x O layers an Mg content x of 0.2 was used. All
layers were assumed to be n-type with a donor concentration of 1 1017 cm3 .
Figure 5.1 (a) and (b) illustrate the possible situations in single heterostruc-
tures at 300 K. A 2DEG will only form in a single heterostructure if the O-face
of the Mgx Zn1x O is adjacent to the Zn-face of the ZnO, due to a positive
sheet charge at the interface. Such a situation is displayed in Fig. 5.1 (b). A
significant increase in the electron concentration is visible in the vicinity of
the heterointerface. Four quantized states are found in the triangular poten-
tial, as visualized in the inset of Fig. 5.1 (b). Nevertheless, depending on the
doping level, free electrons can be found in the bulk of the thin films, which
will additionally contribute to conductivity of the stacked sample. Hall-effect
measurements will record a weighted average over the whole film, complicating
the unambiguous determination of the properties of the 2DEG. It is therefore
crucial to control the dopant concentration in the adjacent layers, in order to
determine the properties of the 2DEG correctly. The problem is reduced at
low temperatures, where the bulk carriers are frozen out, and transport is only
possible via the 2DEG. In contrast a heterostructure in which the Zn-face of the
Mgx Zn1x O is adjacent to the O-face of the ZnO will lead to a negative sheet
charge at the interface and the repulsion of free electrons. Such a situation is
graphed in Fig. 5.1 (a). The structure would in principle allow the accumula-
tion of free holes at the heterointerface, which would require the Fermi-level to
be close to the valence band. The situation displayed in Fig. 5.1 (a) fulfills this

95
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

40 1.61017
(a) (b)
39
1.41017
38
37 1.21017
36

Nd - Na (cm -3)
1.01017
35
C (pF)

34 0.81017
33
0.61017
32 ZnMgO
31 ZnO 0.41017
ZnO:Al
30
a-Al 2O3 0.21017
29
28 0.01017
-2 -1 0 1 220 230 240 250
U (V) w (nm)

Figure 5.2: (a): Capacitance versus voltage dependence of a Schottky diode formed
on a ZnO/Mgx Zn1x O heterostructure grown on a-plane sapphire (b): Spatial distri-
bution of the apparent dopant concentration in the heterostructure as calculated from
Fig. 5.2 (a), the peak indicated by the solid rectangle corresponds to the 2DEG. The
sheet charge concentration was calculated to be n2d = 9.0 1010 cm2 , integrating
over the range inside the solid box. (Courtesy of Dr. Holger von Wenckstern)

condition, it can however be argued, that such a situation is not very likely in
real ZnO/Mgx Zn1x O samples. A high density of deep donors is frequently ob-
served [101] which will hinder the Fermi-level from sinking to the valence band.
The observation of a free hole gas is therefore very unlikely. Deep donors and
acceptors are purposely excluded from the calculations, as their density and
energy levels are not exactly know. Any inclusion would be rather specula-
tive. Fig. 5.1 (c) shows the situation of a Mgx Zn1x O/ZnO/Mgx Zn1x O QW
structure at 2K. The well width is 4 nm. The influence of the electric field is
clearly visible. Quantized states for both electrons and holes are found, and
the lowest level and respective wave function are displayed in Fig. 5.1 (c). It
is seen that the overlap of the wave functions is reduced by the electric field.
Further, in an equilibrium situation only electrons are confined in the quantum
well, leading to carrier densities similar to that in the single heterostructures.
A low temperature reduces the influence of bulk conduction significantly, al-
lowing in principle the determination of the transport properties of the 2DEG
in Hall-effect measurements.

5.2 Carrier confinement in ZnO/MgxZn1xO sin-


gle heterostructures
The formation of a 2DEG was studied in heteroepitaxial ZnO/Mgx Zn1x O sam-
ples by CV measurements (Fig. 5.2) and temperature dependent Hall-effect
investigations (Fig. 5.3). For the CV measurements the heterostructures were

96
5.2. Carrier confinement in ZnO/Mgx Zn1x O single heterostructures

1013
(a) ZnO/MgZnO single
heterostructure
n 2D (cm -2)

1012
0 10 20 30 40
1000/T (1/K)

350
(b) ZnO/MgZnO single
300 heterostructure
250 ZnO thin film
H (cm/Vs)

200

150

100

50

0
0 100 200 300
T (K)

Figure 5.3: (a) Free electron concentration versus inverse temperature and (b) mo-
bility versus temperature in a ZnO/Mgx Zn1x O heterostructure on a-plane sapphire
as determined by temperature dependend Hall-effect, its sheet charge concentration
was calculated to be n2d = 31012 cm2 . The mobility of a bulk ZnO thin film grown
on sapphire is shown for comparison. (Courtesy of Dr. Holger von Wenckstern)

97
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

deposited on a highly aluminum doped ZnO layer, which represents a back-


contact to the structure. Thereby the series resistance of the films is reduced,
allowing the application of higher measurement frequencies before cutoff oc-
curs [57,101] . Additionally the field distribution is more homogeneous in the
film, which leads to a more accurate determination of the spatial distribution
of carriers. This is of special importance for the determination of the sheet
charge density of the 2DEG. From the CV characteristic (Fig. 5.2 (a)) the
net dopant concentration was calculated (Fig. 5.2 (b)). The accumulation of
electrons at the hetero-interface is visible (indicated by the solid rectangle in
(Fig. 5.2 (b)). The sheet charge concentration was calculated using [61] to be
n2d = 9.0 1010 cm2 . The sheet charge density could be increased by doping
of the Mgx Zn1x O barrier, as reported in [248] .
For Hall-effect investigations the heterostructures have been directly grown
onto a-plane sapphire. A sheet carrier concentration of n2d = 3 1012 cm2
was observed, which is almost independent of the temperature (Fig. 5.3 (a)).
The mobility of the 2DEG exceeds that of a 3DEG in a ZnO thin film shown for
comparison, especially at low temperatures (Fig. 5.3 (b)). As the ZnO buffer
layer and the Mgx Zn1x O barrier material are not fully depleted in the struc-
ture and the Hall-effect integrates over all parallel layers in the sample, a slight
influence of the 3DEG on the mobility can be seen, as indicated by the circle
in (Fig. 5.3 (b)). Further, a degenerate layer forming at the interface between
the film and the substrate might generally contribute to the transport [101] . In-
dicative for the existence of such a layer is the steep drop in mobility in the
ZnO reference sample at low temperatures (marked by the solid rectangle),
and the fact, that the contribution of a 2DEG is only observed in a very small
number of heterostructure samples grown on sapphire. This suggests, that if
a degenerate layer is present, its conductivity exceeds that of the 2DEG at all
temperatures. The presence of such a layer would at least result in an under-
estimation of the 2DEG mobility [249] . Although the low temperature mobility
is higher and the influence of ionized impurity scattering on the transport in
the 2DEG is reduced, the room temperature value of the mobility is below
that of a single crystal. Possible reasons are scattering at the heterointerface,
potential fluctuations in the barrier and charged dislocation lines as well as
grain boundaries intersecting the quantum well.

5.3 Electron accumulation in ZnO/MgxZn1xO


quantum wells
5.3.1 Single quantum wells grown on a-plane sapphire
An AFM image of a typical heteroepitaxial MgZnO/ZnO QW structure is
shown in Fig. 5.4 (a). The sample shows two-dimensional growth in the
BurtonCabreraFrank mode [106] , with a visible surface roughness of several
c/2 steps, which is of the order of the QW thickness (rRSM = 0.394 nm,
rA = 0.495 nm). The crystal growth is enhanced by threading dislocations

98
5.3. Electron accumulation in ZnO/Mgx Zn1x O quantum wells

with a screw component present in the center of each of the crystallites. It is


anticipated, that the interfaces of the ZnO/Mgx Zn1x O QWs grown on sap-
phire, as well as the heterointerface in the ZnO/Mgx Zn1x O structures grown
on sapphire shows similar roughnesses. This results in interface scattering of
electrons reducing the mobility, and in inhomogeneous QW thickness, broad-
ening the QW related luminescence. Fig. 5.5 shows the temperature dependent
transport data as obtained by Hall-effect investigations. A monotoneous in-
crease of the mobility is observed towards decreasing temperatures, indicative
for the contribution of a 2DEG. However, in contrast to the data obtained
from the single heterostructure (see Fig. 5.3) a monotonic decrease of the free
carrier concentration is observed. This can be ascribed to the contribution of a
3DEG to the overall conductivity, as already seen for the single heterostructure
sample grown on sapphire. In the case of the QW structure, the influence of
the 3DEG is even more pronounced, and a temperature independent free car-
rier concentration is not seen, even not at low temperatures. However a slight
decrease of the slope in log n versus 1, 000/T is seen, which hints to the reduc-
tion of the three-dimensional transport in comparison to the 2DEG. Further
the temperature dependence of the mobility shows a H T 3 behavior. As
for the 2DEG the main scattering mechanism should be scattering at phonons,
at the smallest exponent anticipated should be 3/2 for scattering at acoustic
phonons via the deformation potential, this behavior is unexpected. As how-
ever the mobility in the three-dimensional transport process is lower than that
in the 2DEG, the systematic increase of the contribution of two-dimensional
transport can lead to an increase of the mobility, which is steeper than that
observed in a isolated 2DEG. Again the overall mobility remains smaller than
that observed in other ZnO/Mgx Zn1x O heterostructures [37,250252] . The pres-
ence of bulk transport and a degenerate layer are possible explanations, along
with interface imperfections. In principle a high density of ionized impurities
near the heterointerface is a possible explanation as well, however the mono-
tonic increase of the mobility towards low temperatures indicates that this
limitation is of minor importance.

5.3.2 Single quantum wells grown on ZnO substrates


Following the results obtained in the growth studies (see Chap. 3.2.6), the
surface roughness can be significantly reduced by growth on ZnO substrate.
An AFM image of a ZnO/Mgx Zn1x O QW grown under optimized conditions
(650 C, oxygen partial pressure of 4 103 mbar, substrate miscut of 0.36 )
is shown in Fig. 5.4 (b). Step flow growth mode is achieved, with perfect
parallel alignment of subsequent layer steps [253] . The average terrace width of
w = 43 nm corresponds to that expected for the present miscut w0 = 41 nm.
The sample shown represents the minimal surface roughness possible for the
given substrate miscut (rRSM = 0.121 nm, rA = 0.152 nm). Reciprocal space
maps of the symmetric (00.2) (Fig. 5.6 (a)) and the asymmetric skew (10.1)
(Fig. 5.6 (b)) reflex of a homoepitaxial MgZnO/ZnO sample are shown. The
peak positions only differ in the out-of-plane direction, but have the same in-

99
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

(a)

(b)

Figure 5.4: (a) 4 m4 m AFM image of a ZnO/Mgx Zn1x O QW grown on


sapphire. (b) 2 m2 m AFM image of a ZnO/Mgx Zn1x O QW grown on ZnO.
(Top image obtained by Gregor Zimmermann)

100
5.3. Electron accumulation in ZnO/Mgx Zn1x O quantum wells

T (K)
20 40 60 80
1018
250

100 200

muH (cm2/Vs)
mu H~T-3
1017 150

muH (cm2/Vs)
n (cm-3)

10
30 40 50 60 70 80 90 100
T (K)

50
1016 ND =5.6x1017cm -3
ED =35 meV
NA =1.0x1016cm -3 0

10 20 30 40 50
1000/T (1/K)

Figure 5.5: Temperature dependence of (Red ): the free carrier concentration and
(Blue): the mobility in a ZnO/Mgx Zn1x O quantum well structure grown on sap-
phire as determined by Hall-effect investigation. The upper right inset shows the
temperature dependence of the mobility in double logarithmical plot. The donor
concentration and activation energy is given along with the acceptor concentration
in lower left corner.

plane value. This indicates a perfect match of the a-axis lattice constant, thus
growth is pseudomorphic. The c-axis of the film is compressively strained with
respect to the substrate by 2.36 103 .
The temperature dependence of the mobility of selected samples grown on
ZnO is depicted in Fig.5.7 along with the temperature dependence of the mobil-
ity of a single crystal grown by the skull-melting method (Cermet) [127,128] . The
mobility data of a homoepitaxial ZnO film are presented, as well as that of a
ZnO/Mgx Zn1x O QW structure. The mobility measured on the QW structure
exceeds that of the homoepitaxial ZnO thin film at all temperatures, which
itself shows significantly higher mobility than any ZnO thin film grown on
sapphire investigated within the present study. However, the mobility of the
single crystal still exceeds that of both the homoepitaxial ZnO thin film and
the ZnO/Mgx Zn1x O QW structure. Further a pronounced increase of the mo-
bility at low temperatures, as expected for a 2DEG, is not observed for the QW
structure. It is also observed, that the mobility of the homoepitaxial ZnO thin
film and especially that of the ZnO/Mgx Zn1x O QW structure drop sharply
for low temperatures (indicated by the solid box in Fig. 5.7). In literature
this behavior is connected to the presence of a degenerate layer in the sam-
ple, dominating carrier transport at low temperatures [101,225] . Again the TDH
experiment is sensitive only to the integral properties of all parallel layers in

101
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

(a) (00.2)
q (nm -1 )

4.05
(b) (10.1) 3

4.045 2.5
q (nm -1 )

2
4.04
1.5

1
4.035

0.5

4.03
-4 -3 -2 -1 0
-1 -3
q (nm ) x 10

Figure 5.6: HR-XRD reciprocal space maps of (a) the symmetric (002) and (b)
the asymmetric skew (101) reflex of a ZnO/Mgx Zn1x O QW structure grown at
4 103 mbar on ZnO (0001). The color scale represents the log10 of the scattered
X-ray intensity (counts). The topmost peaks in each of the plots are from the thin
film and the stronger bottom peaks from the ZnO substrate. (Data obtained by Dr.
Michael Lorenz)

102
5.3. Electron accumulation in ZnO/Mgx Zn1x O quantum wells

103

H (cm2/Vs)

ZnO single crystal


homoepitaxial MgZnO/ZnO QW
homoepitaxial ZnO thin film

102
30 50 70 100 200 300
T (K)

Figure 5.7: Electron mobility versus temperature in a homoepitaxial ZnO thin film,
a ZnO/Mgx Zn1x O QW structure grown on ZnO and a ZnO single crystal.

the sample. The dominance of the degenerate layer masks the presence of the
3DEG in the homoepitaxial ZnO thin film as well as the presence of the 2DEG
in the ZnO/Mgx Zn1x O QW structure and that of the most likely existing
3DEG in the Mgx Zn1x O barrier material. However, the presence of the 2DEG
in the ZnO/Mgx Zn1x O QW structure can still be deduced from its mobility.
At about 45 K a peak in the mobility versus temperature behavior is observed
(marked by the circle in Fig. 5.7). This peak is neither present in the mobility
versus temperature dependence of the homoepitaxial ZnO thin film nor in that
of the single crystal, and should not appear for a 3DEG at all. The peak is
attributed to the presence of a 2DEG in the ZnO/Mgx Zn1x O QW structure.
However, the sheet carrier density of the 2DEG must decrease towards lower
temperatures, indicating that the electronic states of the degenerate layer are
energetically preferred. Detailed investigations of this phenomena have to be
carried out, as well as investigations on samples without the degenerate layer.
Nevertheless the deducible mobility of the 2DEG in the ZnO/Mgx Zn1x O QW
grown on ZnO by far exceeds that in the heterostructures grown on sapphire
in this study. This is even more remarkable, as the mobility is expected to
further increase towards low temperatures and is most likely underestimated
in the present structure due to the overlay with the 3DEG and the degenerate
layer.
The FWHM of the QW related PL peaks of ZnO/Mgx Zn1x O QW structures
grown on sapphire and ZnO is depicted in Fig. 5.8 a. The peak width is always
slightly smaller for the ZnO/Mgx Zn1x O QW structures grown on ZnO, and
the difference is more pronounced for smaller thicknesses. This difference is
attributed to the more homogeneous growth of the QWs on ZnO compared to
the QWs grown on sapphire.

103
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

50

40
FWHM (meV)

30

20

10 QWs on a-sapphire
QWs on c-ZnO
0
0 2 4 6
L w (nm)
Figure 5.8: Peak width (FWHM) of the QW related PL peak versus QW thickness
LW for ZnO/Mgx Zn1x O QW structures grown on sapphire and ZnO substrates.
(The data for the QWs grown on sapphire are taken from Ref. [81] )

5.4 The quantum confined Stark-Effect in the


ZnO/MgxZn1xO system
In the previous sections the principle physics involved in the acculumation
and quantum confinement of free carriers in the vicinity of a ZnO/Mgx Zn1x O
heterojunctions were discussed. Morphological, structural and transport prop-
erties of single heterostructures and structures containing a ZnO QW were
discussed. This section is devoted to a detailed study of the luminescence prop-
erties of ZnO/Mgx Zn1x O QW well structures. All structures discussed in the
following are grown along the c-axis direction, and will therefore be subject
to an electric field forming inside the heterostructures due to the discontinuity
in the electric polarization at the heterointerfaces. All samples discussed in
the following are of very high crystalline quality. AFM investigations of the
surfaces revealed step-flow-growth exclusively, though the interface roughness
is different. For the samples grown on ZnO solely pseudomorphic growth is
observed, and depending on the miscut, a perfect vicinal surface is observed as
shown in Fig 3.21. It therefore seems reasonable to believe, that the interfaces
defining the quantum well are of equal quality. Nevertheless, deterioration of
the interfaces is possible. Among them, interdiffusion of the constituents has
the most tremendous influence. While the surface of the samples remains per-
fect, the interfaces buried inside the junctions might degrade during growth,
making the evaluation of the interface quality impossible from surface obser-
vation, even if the growth is stopped after the layer in question. The rate of
interdiffusion depends on the kinetic energy of the particles that are about to
diffuse. Apart from the inevitable thermal energy given to all atoms inside the

104
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

growing crystal at the growth temperature another source of kinetic energy


becomes important in the case of PLD growth. The ablation of the target
results in the formation of a high energy plasma with a plasma temperature of
several 103 K [49,50] . This gives energies in the range of up to several hundreds
of eV to the impinging adatoms [49] . This can be advantageous, as the surface
diffusion is enhanced, which will lead to an improvement of the surface perfec-
tion. This does in principle allow lower growth temperature than in processes
which involve incorporation of adatoms from a gas phase, such as CVD and
MBE. In contrast, the high energies of impinging species does have significant
drawbacks, as it can result in an ion implantation, and will damage the bulk
of the film, at least the first few atomic layers [49,254] . While such a process
is not disturbing in the growth of bulk films, it will dramatically reduce the
attainable interface quality. The appearance of the quantum confined Stark-
effect can be a measure for the interface quality, as it has been shown, that
the influence of polarization discontinuities on the wave-function separation of
electrons and holes can be reduced by smoothing the interfaces [255257] . The
authors showed that ion implantation of ZnO/Mgx Zn1x O samples showing a
pronounced QCSE resulted in a blueshift of the emission. They concluded, that
softening of the interfaces changes the triangular well potential. The electron
and hole wave-function therefore have an increases overlap, and their energy
levels are closer.

5.4.1 Literature data on the QCSE in ZnO/Mgx Zn1x O


Quantum wells
In preceding studies [81,258] , luminescence properties of ZnO/Mgx Zn1x O quan-
tum wells were investigated. Luminescence spectra were recorded as a function
of the well widths in different samples. The well width was extrapolated by
the number of laser pulses required to grow a certain ZnO layer by PLD, and
accurately measured by X-ray reflectometry. The quality of the interfaces was
optimized employing AFM studies of the surfaces, and step flow growth was
obtained, at least in the case of growth on a single crystal ZnO substrate.
However, despite having thicknesses of up to 6.5 nm only a slight influence of
the QCSE was seen. The lowest value of an QW related emission was observed
at 3.36 eV, which is roughly the position of the donor bound transition line I6
typically observed in bulk ZnO. This line is present in the emission of all the
samples discussed in these studies, as all samples are deposited onto a ZnO
buffer layer introduced in order to improve the sample morphology. It is there-
fore not possible to judge, if a luminescence lower in energy than that of bulk
ZnO occurs.
In contrast, a strong redshift of the QW luminescence below the bulk ZnO lumi-
nescence emission energy due to the QCSE was observed in ZnO/Mgx Zn1x O
samples grown by MOCVD [259] , MBE [38,255,257,260263] and Laser-MBE [264267]
(which is technically equal to PLD, with the difference, that the distance be-
tween the target and the substrate is much larger than the mean free path

105
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

10

Inverse Oscillator strength (arb. u.)


3.8 10

Transition energy (eV)


12
3.6 10

14
3.4 10

16
3.2 10

18
3 10
0 2 4 6 8 10 0 2 4 6 8 10
QW width (nm) QW width (nm)

9
x 10
0.05 8
Exciton binding energy (eV)

Exciton Bohr radius (m)


7
0.04
6

0.03 5

4
0.02
3

0.01 2
0 2 4 6 8 10 0 2 4 6 8 10
QW width (nm) QW width (nm)

Figure 5.9: Properties of an exciton in a c-oriented ZnO/Mgx Zn1x O QW with


x = 0.18 as function of the QW width obtained from variational calculations. (a)
transition energy (b) inverse oscillator strength (c) exciton binding energy and (d)
Exciton Bohr radius.

of the particles in the plasma). Interestingly the redshift associated with the
QCSE is less pronounced in the samples grown by LMBE, compared to MBE
and MOCVD grown samples for the same well width. Further the QCSE was
observed in MBE grown ZnO/CdZnO QWs [268,269] .

5.4.2 Luminescence phenomena related to presence of the


QCSE
In order to understand the luminescence phenomena to be expected in QWs
subject to the QCSE numerical calculations have been performed. The wave-
function of the uncorrelated electrons and holes have been calculated using
the poisson solver Beta8c developed by Gregory Snider [246,247] . The exciton
binding energy and wave-function have been calculated in the envelope func-
tion approximation by a variational approach [270273] . Herein the exciton is
treated as 3-dimensional, which is required due to the strong electric field in
the quantum well. While for small well widths a 2-dimensional approach would
in principle be valid, this is not the case in broad quantum wells, which is how-
ever the case of interest here. The two-body exciton wave function is written

106
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

as

= e (ze )h (zh )r (r , |ze zh |) , (5.1)

where e and h symbolize the uncorrelated electron and hole wave functions,
respectively, and

r2 + 2 |ze zh |2
 
r = exp , (5.2)

where is a dimensionless parameter describing the dimensionality of the ex-


citon, and is the exciton Bohr radius. Only the lowest electron and heavy hole
eigenstates of the uncorrelated system are regarded in the following. The total
energy of the exciton is minimized in the calculation, by variation of the pa-
rameters and . The oscillator strength of the exciton is calculated with this
parameters following Ref. [273] . The properties of a single exciton determined
from these calculations are displayed in Fig. 5.9 as a function of the QW width.
Figure 5.9 (a) shows the transition energy of the exciton, which is the energy of
the photon observed in an luminescence experiment. Upon increase of the well
width Lz the transition energy decreases, due to the lowering of the electron
energy level (and the rise of the hole energy level) in the presence of the polar-
ization induced electric field. This effect was observed in all publications cited
in the previous section. In fact, the existence of a luminescence peak clearly
lower in energy than the I6 transition, which is very often observed in bulk
ZnO, and its dependence on the thickness was chosen as a criterion to list the
specific publication. Likewise the overlap of electron and hole wave functions
decreases due to the increasing spatial separation, and the inverse oscillator
strength increases (see Fig. 5.9 (b)). This quantity is proportional to the lu-
minescence decay time, and an increase of the luminescence decay time with
increasing well width is frequently observed in the literature [255,257,260263,267,268] .
Further the separation of the electron and hole leads to an increase of the ex-
citon Bohr radius (see Fig. 5.9 (d)) which results in a decrease of the exciton
binding energy via the Coulomb interaction (see Fig. 5.9 (c)). In experiments
this manifests in an increase of the exciton phonon coupling, and therefore an
increase in the HuangRhys factor S [264267] . All the considerations outlined
here are drawn for a single exciton. In a real experiment however, the optical
excitation will generate an electron hole plasma, which thermally relaxes into
the respective minimum, accompanying exciton formation. During this process
however additional radiative recombinations can take place, resulting in pho-
tons with higher energy than the characteristic luminescence. Additionally the
high density of free carriers leads to a partial screening of the polarization in-
duced interface charges in the system, and therefore a reduction of the electric
field within the QW. The QCSE will therefore diminish, and a blueshift of the
luminescence is expected with increasing excitation strength. Such a behavior
has indeed been observed in the ZnO/Mgx Zn1x O system [260] .

107
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

Figure 5.10: Schematics of ZnO/Mgx Zn1x O QW samples investigated in the


present study. (a) sample grown on sapphire substrate and (b) on ZnO (0001).
As indicated the surface polarity of the samples differs in both cases. An additional
ZnO buffer layer can be introduced for the sample grown on sapphire, in order to
improve the morphology of the Mgx Zn1x O film.

5.4.3 Observation of the QCSE in ZnO/Mgx Zn1x O quan-


tum wells grown by PLD
The samples grown for this study follow an idea outlined by Zhang et al. [274] .
In spite of having multiple samples of a single thickness each, the sample shows
a thickness gradient, result in different QW widths on one sample. This has
the advantage of an exactly equal composition of the Mgx Zn1x O barrier layer
for all QW widths. This is not guaranteed if multiple samples of individual
thickness are used, as the Mg content x of the layers deposited increases slightly
during the lifetime of a ceramic target. The reason for this behavior is most
likely a selective evaporation of Zn from the target, which has a much higher
vapor pressure than Mg. The resulting films therefore have to be carefully
compared regarding their composition. This complication is completely ex-
cluded by the present approach. The gradient of the thickness is small enough,
that within the excitation spot of the PL experiment of 100 m diameter a
thickness variation of less than 0.1 nm occurs, even for the thickest QWs inves-
tigated. This results in a maximal change of a single unit cell height of the QW
width, which is inevitable anyway, as the growth of an exact number of atomic
layers can not be realized over the whole sample by conventional PLD. It has
to be outlined, that introducing an in-situ RHEED analysis could in principle
allow for such high accuracy, the realization itself however is demanding. As
the growth setup employ does not include such equipment, no additional error
sources are introduced by the shape of the sample in the PL experiment. An
additional complication however arises, which is the determination of the QW
width Lz at the individual point investigated in the experiment. This is much
easier in a sample of uniform thickness as techniques like X-ray reflectome-
try or spectroscopic ellipsometry can be employed, which would be obscured
by the thickness variation inside the detection spot. It has nevertheless been
shown in a previous study [81] and also in this work, that the number of laser
pulses relates linearly to the thickness of the film above a critical thickness.

108
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

CL-Intensity (arb. units)

1000

100

10

1
3.0 3.1 3.2 3.3 3.4

Photon energy (eV)

1000
CL-Intensity (arb. units)

100

10

1
3.20 3.25 3.30 3.35 3.40 3.45 3.50 3.55 3.60

Photon energy (eV)

Figure 5.11: Low temperature CL spectra obtained from two otherwise equal
Zno/Mgx Zn1x O QWs deposited at two different laser fluences: (a) lense in po-
sition 2 (low fluence) (b) lense in position 1 (high fluence). The spectra are color
coded, red corresponds to a high, blue to a low QW thickness. (Data obtained by
Jan Zippel)

109
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

10000
CL-Intensity (arb. units)

1000

100

10

3.0 3.1 3.2 3.3 3.4

Photon energy (eV)

Figure 5.12: Excitation dependence of the CL spectra of a sample showing a pro-


nounced QCSE. The excitation currents are 800 nA, 1600 nA, and 2,0008,000 nA in
steps of 1,000 nA. (Data obtained by Jan Zippel)

The error introduced by this approximate thickness determination is below


the experimental error of the thickness determination by X-ray reflectometry
or spectroscopic ellipsometry, as the coefficients can be determined from mea-
surements of comparatively thick films by the same techniques, reducing the
relative error by more than two orders of magnitude. Additionally, as discussed
in Chap. 3.2.1, the error introduced by a 3-dimensional nucleation process in
the beginning of the thin film growth is negligible allowing the exact determi-
nation of the growth rate from the absolute thickness of a film thicker than
100 nm. This leads to the conclusion that an equally exact determination of
the QW thickness is possible from the total thickness of the layer stack in the
individual spot investigated.
The thickness of the whole stack was assessed for selected samples by scanning
electron microscopy on a cross section prepared by in-situ ion milling in an
focused ion beam microscope. Cross sections were prepared on both sides of
the sample, and the thickness inside was assumed to show a linear gradient.
As the QW thickness is not determined directly, for all evaluations, were this
thickness is not required, the actual measured quantity, which is the position
on the sample is given in the following.
In the preceding sections it was shown, that the QCSE can in general be
observed in ZnO/Mgx Zn1x O QW structures. It became clear, that growth
methods involving low energy atomic species, such as MBE and MOCVD are
suitable for the growth of the sharp heterointerfaces required to see the QCSE.
In contrast no convincing report has yet emerged showing the QCSE in PLD

110
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

grown samples. The occurrence of the QCSE in samples grown by the PLD
related technique LMBE is much less pronounced than in the case of MBE and
MOCVD. It has been speculated that the high energy of impinging adatoms
leads to a damage of the topmost atomic layers of the sample, and therefore
enhances the interdiffusion of ZnO and Mgx Zn1x O. This thesis will be proven
unambiguously by the experimental data given in the following.
The mean to control the kinetic energy of the impinging particles is the flu-
ence of the laser pulses used to ablate the target. In the setup employed in
the present study, the fluence is controlled by the position of a focusing lense
in the light path. The CL spectra shown in Fig. 5.11 have been obtained
on two selected samples grown with different lense positions, noted as posi-
tion 1 (corresponding fluence 2.4 J/cm2 ) and position 2 (corresponding fluence
1.75 J/cm2 ). The CL spectra were obtained at similar positions on the sample,
covering the whole thickness gradient. A remarkable difference in the spectra
is observed. For the sample grown at the lower laser fluence a pronounced
redshift of the luminescence maximum below the I6 transition is observed (see
Fig. 5.11 (a)), which systematically depends on the position investigated on
the sample. In contrast, the luminescence spectra of the sample grown at the
high laser fluence have their maximum either at the spectral position of the
I6 line, or at even higher energy (see Fig. 5.11 (b)). The observation of the I6
luminescence originates in the ZnO buffer layer. A systematic shift is indeed
observed, however the luminescence energy of the maximum is always at least
80 meV higher than in the sample grown at the lower laser fluence. This indi-
cates, that the QCSE is significantly reduced in this sample, which is attributed
to an interdiffusion of the ZnO and Mgx Zn1x O layers in the vicinity of the
interface. However, in principle another origin is thinkable. A high density
of residual donors in the sample grown at higher laser fluence would result in
an increase of the carrier density accumulating at the interface, and therefore
in partial compensation of the polarization induced sheet charge at the het-
erointerfaces responsible for the electric field inside the QW. The free electron
concentration was determined by Hall-effect measurements, and was roughly
1016 cm3 in both samples. Although the density of free carriers in the vicinity
of the QW can not be directly inferred from this value, it indicates that no
significant differences exits between the samples. Numerical calculations show
that the QCSE is as well expected in samples with a donor concentration of
1018 cm3 . Therefore this alternative origin of the diminishing of the QCSE in
samples grown at a high laser fluence is ruled out.
In order to evaluate the magnitude of the change of the field induced carrier
separation due to high carrier densities in the QW, CL experiments have been
carried out with different excitation currents. The current has been varied
between 800 and 8,000 nA at an acceleration voltage of 5 kV. The resulting
spectra are depicted in Fig 5.12. A systematic blueshift with increasing exci-
tation current is indeed visible, as has already been observed in Ref. [260] . The
overall magnitude of the spectral shift is 60 meV in the range of currents
employed in the present study. The luminescence intensity increases by more
then a factor of 20, which is roughly equal to the increase in the excitation

111
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

power. The overlinear increase can be explained by the increase in the oscilla-
tor strength of the excitonic transition due to an increased overlap of electron
and hole wave functions.
Photoluminescence investigations have been carried out on samples of differ-
ent Mg contents x. Both, samples grown on sapphire and on O-face ZnO were
investigated. The spectra obtained on two selected samples are displayed in
Fig. 5.13. The spectra depicted in Fig. 5.13 (a) are obtained from a sample
directly deposited onto an a-plane sapphire substrate, without a buffer layer.
The spectra given in Fig. 5.13 (b) are measured on a sample grown on a ZnO
substrate. All spectra were normalized to the intensity of their QW lumines-
cence. The measurement position is color coded where blue indicates low and
red high QW thickness, respectively. A clear redshift of the QW luminescence
with increasing QW thickness is observed for both samples. For the sample
directly deposited onto sapphire an additional feature is observed for the high
QW thicknesses at 3.379 eV. The spectral position is not a function of the QW
thickness, which indicates that the origin of the luminescence feature is not
the QW. A spectral feature labeled XI was observed in a previous work in the
range of 3.393.4 eV in QW samples [81] . It has been speculated in this work,
that this luminescence feature originates from the interface between the ZnO
buffer and the Mgx Zn1x O barrier layer in the samples discussed therein. This
assertion can not be supported here, as no ZnO buffer layer is present in the
sample in question. It however remains unclear, if the origin of the peak re-
ported here is the same as in Ref. [81] .
For the sample grown on ZnO additional to the QW luminescence contributions
from the luminescence of the underlying substrate is visible. The spectral posi-
tion of typical ZnO luminescence lines is indicated in Fig. 5.13 (b), namely the
lines I6 , I3a , I0 and FX are given in the plot, corresponding to the luminescence
from excitons bound to the neutral aluminum donor (Al0 X), the zinc intersti-
tial donor (Zni X), the ionized aluminum donor (Al+ X) and the free A-exciton
respectively. Interestingly, if the spectral position of the luminescence of the
quantum well coincides with that of the ZnO luminescene, an absorption line
is visible in the QW luminescence at the position of the free exciton, indication
that

A sizeable fraction of the photons measured in the PL experiment are


not directly emitted from the sample, but undergo a reflection, either at
the back side of the substrate or at at the filmsubstrate interface.

An excitation in the substrate takes place via the formation of free exci-
tons by the incident light from the QW luminescence.

The free excitons created in the substrate by the QW luminescence do not


recombine instantly, but are likely to be bind to defects, and recombine
there with a lower emission energy.

The luminescence from the quantum well shows pronounced phonon replica,
which indicate a strong exciton phonon coupling in the quantum well. It can
be observed that

112
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

Normalized Intensity (arb. units)

3,379 eV

3.10 3.15 3.20 3.25 3.30 3.35 3.40

Photon energy (eV)


Normalized Intensity (arb. units)

3.10 3.15 3.20 3.25 3.30 3.35 3.40

Photon energy (eV)

Figure 5.13: Normalized PL spectra of two samples with different Mg content x in


the barrier (a) x = 0.23 and (b) x = 0.28. The sample shown in Fig. 5.13 (b) was
grown on the O-face of a ZnO substrate. The measurement position is color coded:
blue denotes low QW thickness, while red corresponds to high QW thickness. (Data
obtained by Martin Lange)

113
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

The exciton phonon coupling strength increases with increasing well


width, as the intensity of the first phonon replica increases systemati-
cally with increasing well width.

The exciton phonon coupling strength of the QW well luminescence ex-


ceeds that of the donor bound excitons in bulk ZnO, which is seen for
the thinnest QWs. There a significant peak is seen at the I6 position,
exceeding the intensity of the QW luminescence by more than a factor
of 5. In contrast for the first phonon replica the peak shape has changed
and only a weak shoulder is observed corresponding to the replica of the
I6 line.

The spectral position of the maximum of the QW related emission was evalu-
ated and is given in Fig. 5.14 (a) for different sample positions. The location
on the sample was carefully controlled by an adjustable sample holder, and the
10x10 mm sample was scanned over the whole width. The spectral position
of the ZnO luminescence is indicated. Though the thickness of the quantum
well is not directly known, it can be inferred from the position on the sub-
strate. The samples are all grown under the same conditions and employing
the same pulse numbers. Electron micrographs for one of the samples allow
to calculate a thickness variation of the QW from 2.94.3 nm. A systematic
redshift is observed for all samples with increasing thickness. An increase in
the shift is observed with increasing Mg content x. However a significant dif-
ference in spectral the position of the QW luminescence is seen between the
samples grown an sapphire and these grown on ZnO. The QW emission energy
measured in the samples grown on sapphire is always lower than that in the
samples grown, typically by 20 meV. Several possible reasons for the change
exist. First of all, in the samples grown on ZnO the Mgx Zn1x O layer is com-
pressively strained, while the ZnO layer is relaxed. This situation is reversed
if the Mgx Zn1x O layer is directly deposited onto sapphire. The Mgx Zn1x O
layer is relaxed, while the ZnO is under tensile strain, which induced a re-
duction of the band gap of the ZnO. The shift in the bandgap does however
only account for x16 meV, which is below the experimental observed values.
Additional processes have to play a role here. It has to be remarked, that the
different strain does not influence the change in polarization trough the piezo-
electric polarization, which is proportional only to the change of the a-axis
lattice constant in the strained layer. This change is exactly equal if either
the Mgx Zn1x O barrier layer or the ZnO QW layer is strained. Also a depen-
dence of the QW thickness on the substrate chosen is unlikely, but can not be
completely excluded. High resolution TEM investigations of the QW region
should reveal the thickness of the QW in detail, and allow determination of
the change in polarization in both cases. HR-XRD measurements and Raman
investigations are required in order to unambiguously determine the strain of
the barrier layer, and allow a calculation of the strain in the quantum well.
Figure 5.14 (b) gives the dependence of the HuangRhys factor of the first
phonon replica of the QW luminescence on the spectral position of the QW
luminescence maximum. A systematic increase is observed with decreasing

114
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

3.40

QW Luminescence Energy [eV]


3.38 FX
3.36 I6
3.34

3.32
x=0.23
3.30
x=0.22
3.28 x=0.28
x=0.27
3.26

0 2 4 6 8 10

Position on sample [mm]

I6 FX
0.18
0.16
Huang-Rhys Faktor S0

0.14
0.12
0.10
0.08
x=0.23
0.06
x=0.22
0.04 x=0.28
0.02 x=0.27
0.00
3.26 3.28 3.30 3.32 3.34 3.36 3.38

QW Luminescence Energy [eV]

Figure 5.14: (a) Position of the QW related luminescence maximum for two sam-
ples grown on sapphire (open symbols) and two grown on ZnO (full symbols) with
different Mg content x measured at different positions on the sample. (b) Huang
Rhys factor S0 calculated from the main transition and the first phonon replica as a
function of the spectral position of the main peak.  corresponds to samples grown
on ZnO substrates, while indicates the samples grown on sapphire. The position
of important ZnO related emission lines is indicated. (Measurements performed by
Martin Lange)

115
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

100
PL-Intensity [arb. units]

Decay-time [ns]
10

0.1

3.15 3.20 3.25 3.30 3.35 3.40 3.45

Energy [eV]

3.7 10-3
3.6
10-4
QW Emission Energy [eV]

3.5
10-5
Decay time [s]

3.4
3.3 10-6

3.2 10-7
3.1
10-8
3.0
10-9
2.9
2.8 10-10
0 2 4 6 8 10

QW thickness [nm]

Figure 5.15: (a) Pl maximum position (left scale) and decay time (right scale) as a
function of the luminescence energy. (b) Dependence of the luminescence energy (left
scale) and decay time (right scale) at the maximum of the QW luminescence on the
QW widths. Full symbols give data obtained in this study, while open symbols are
taken from Ref. [262] . Solid lines give the expected dependence from the variational
calculations. (Spectra obtained and time constants determined by Marko Stlzel,
QW thickness calculated from measurements by Jrg Lenzner)

116
5.4. The quantum confined Stark-Effect in the ZnO/Mgx Zn1x O system

transition energy, in accordance with previous publications [264267] .


In addition an increase in the luminescence decay time is observed from time
dependent luminescence experiments. The dependence of the decay time on
the emission energy is given in Fig. 5.15 (a). The spectrum chosen contains
essentially three parts, which are, from left to right: the phonon replica of
the QW luminescence (3.13.27 eV), the QW luminescence (3.273.35 eV) and
a luminescence of unknown origin (3.353.45). While the luminescence decay
time is as low as 300 ps for the luminescence of unknown origin, a system-
atic increase in the decay time is observed, for lower luminescence energies.
Interestingly a maximum is reached at a transition energy where a minimum
in the luminescence is observed. The QW luminescence dominates at higher
energy, the phonon replica at lower energy. It is therefore concluded, that the
QW luminescence itself does not consist of a single luminescence process with
a single time constant, but represents a dynamic relaxations of the free carri-
ers created by the excitation. After the excitation the free carriers screen the
polarization induced sheet charges partially, which leads to a higher transition
energy than in the dilute limit. However, carriers decay radiative and non
radiative and additionally diffuse in the plane of the quantum well to areas
which are not subject to the excitation. The overall density of carriers in the
active region of the QW therefore decreases with time, and the luminescence
red-shifts, as the screening of the polarization induced sheet charges by the free
carriers is reduced. The luminescence shifts to lower energies while increasing
the time interval after excitation, as the system approaches the dilute limit.
In principle, the carrier dynamics can be understood from the present data
already, but sophisticated models would be required to include the in-plane
motion of the free carriers and excitons, exciton formation and radiative and
non radiative recombination rates. Such an evaluation will not be performed in
the present thesis. A detailed investigation of the luminescence energy versus
time behavior is highly helpful to understand the recombination dynamics com-
pletely. Such data can be obtained employing a streak camera, which allows
the observation of the time dependence of several energetic channels simulta-
neously. Additionally, the spatial dependence of the luminescence should be
investigated, possibly by using a micro-luminescence setup as described in [275]
or by observation of spatially masked samples, as described in Ref. [276] .
Figure 5.15 (b) shows the luminescence energy of the maximum of the QW
emission and the related decay time as a function of the QW thickness as
calculated from the cross section micrographs and the position on the sam-
ple where the spectra were obtained. The Mg content x in the barriers was
x = 0.18, the spectra at the individual positions are displayed in Fig 5.13 (a).
Data from Ref. [262] is included for reference. The solid lines show the expected
transition energy and a quantity proportional to the inverse oscillator strength
as given by the variational calculations. This quantity was normalizes to the
low thickness decay time obtained in Ref. [262] , as for the lowest QW thickness
the influence of non radiative recombination is the smallest, assuming a thick-
ness independent non radiative recombination rate. The coincidence of the
values is remarkable, underlining, that the samples present in this study are

117
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

equal in both interface quality and density of non radiative defects to the MBE
grown samples studied in Ref. [262] . It also shows, that a determination of the
polarization as function of the Mg content is possible from the present samples,
if the recombination properties of the excitons in the dilute limit are known.
An estimation is already possible at present yielding a value of 0.24 C/cm2 in
close agreement with previous studies. Further investigation will be required
to improve the accuracy of this value, and could in principle help to accurately
quantify further unknown material parameters as the conduction band offset
in Zno/Mgx Zn1x O heterostructures and the effective hole masses.

5.5 Summary and conclusion


The results of this chapter can be summarized as follows:

2D growth in the BurtonCabreraFrank mode has been observed for the


ZnO/Mgx Zn1x O single heterostructures and ZnO/Mgx Zn1x O quantum
wells grown on a-sapphire. The effective surface roughness in these sam-
ples is of the order of the thickness of the quantum wells present in the
structures.

In contrast, a significant improvement of the structural and morpholog-


ical properties of the ZnO/Mgx Zn1x O quantum wells grown on ZnO
is observed. Step flow growth leading to perfect vicinal surfaces could
be achieved for these samples as confirmed by atomic force microscopy.
The Mgx Zn1x O layers were confirmed to be grown pseudomorphic by
high-resolution X-ray diffraction.

Electron accumulation layers have been observed for ZnO/Mgx Zn1x O


single heterostructures grown on sapphire by CapacitanceVoltage spec-
troscopy and temperature dependent Hall-effect measurements.

Indications for the presence of a 2DEG in ZnO/Mgx Zn1x O quantum


wells grown on ZnO were found by temperature dependent Hall-effect
measurements. For a clear observation of the transport properties of the
2DEG further improvements is still necessary. The contributions of the
degenerate layer have to be removed and the conductivity of the bulk
material has to be reduced.

Quantum confinement effects have been confirmed previously by photo-


luminescence PL measurements [258] . A narrowing of the PL peak width
is observed in quantum wells grown on ZnO, especially for low well width,
due to the higher influence of the interface roughness on thinner quantum
wells.

Careful control of the laser fluence applied to the PLD target resulted
in a significant improvement of the quality of the heterointerface in
ZnO/Mgx Zn1x O heterostructures. A pronounced QCSE was observed
in PLD grown samples for the first time. Comparison with data obtained

118
5.5. Summary and conclusion

on MBE grown samples given in the literature revealed, that the samples
discussed here equal the MBE grown samples both in interface quality
and in density of non-radiative recombination centers.

The use of a slanted QW allowed the observation of quantum confine-


ment effects in QW of different thicknesses in one sample with a well
defined barrier composition. However, uncertainties in the thickness of
the determination are present. The use of an in-situ RHEED observation
would help to determine and control the thickness of individual layers.

119
5. Quantum effects on carriers in ZnO/Mgx Zn1x O heterostructures

120
Chapter 6

Polarization related effects in


ZnO/BaTiO3 heterostructures

The preceding chapter discussed the influence of a change in the electric polar-
ization at a semiconductor heterointerface on the locaization of free carriers.
The interface was controlled to be atomically abrupt, which allowed the ob-
servation of quantization effects. While the change in polarization could be
controlled via the Mg content x, and a determination of this change as a func-
tion of x was possible, it remained a constant value in one single structure.
The effects on mobile carriers induced by the polarization where the same in
equilibrium, regardless of the history of the sample. Nevertheless, the occupa-
tion of the individual accumulation layers can be modified by optical excitation
(see Chap. 5.4), and the dynamics of the system approaching equilibrium was
monitored. In principle the equilibrium configuration can be modified by the
application of an electric field. This modification however is only effective as
long the electric field is applied.
This behavior is contrasted by the material class of ferroelectrics. As out-
lined in Chap. 3.3 BaTiO3 (BTO) will serve as a model ferroelectric within the
present study. Upon the formation of heterostructures between ZnO and BTO
interesting physical phenomena can be anticipated. First the polarization of
the ZnO will, if aligned perpendicular to the interface between the layers, in-
fluence the polarization in the BTO. This polarization can be controlled by the
external electric field, and will, at least to some extent remain in the state it
was polarized to, even in the absence of the external field. Like in the case of
the ZnO/Mgx Zn1x O heterostructures the density of free carriers in the ZnO
will be influenced by the change in polarization across the interface. Depend-
ing on the orientation of this difference either carrier accumulation or depletion
is expected in the ZnO layer, again in accordance with the ZnO/Mgx Zn1x O
heterojunctions. This however will have a subtle influence on the value of the
external electric field forming in the individual parts of the heterostructure, as
a function of external voltage. This in turn will influence the polarization in
the BTO layer. The behavior of the structure can therefore not be predicted
from the applied voltage to the heterostructure alone, but is also a function of
the history of applied voltages.

121
6. Polarization related effects in ZnO/BaTiO3 heterostructures

Within the present thesis a detailed understanding of the behavior of ZnO/BTO


heterostructures has been developed. The work was carried out in close col-
laboration with Venkata M. Voora and Prof. Dr. habil. Mathias Schubert
from the University of Nebraska Lincoln. A description of the model employed
to understand the properties of ZnO/BTO heterostuctures will be given in
Chap. 6.2. Before that, the effects observed experimentally will be summa-
rized in Chap.6.1. An application in the form of ZnO/BTO based ferroelectric
thin film transistors will be presented in Chap.6.3.

6.1 Effects observed in ZnO/BTO heterostruc-


tures
6.1.1 ZnO/BTO samples grown on Si
For the first investigations into ZnO/BTO heterostructures such junctions were
grown by PLD on Si substrates. In order to form ohmic contacts to both sides
of the heterojunction, Pt was used as a bottom and top electrode. The BTO
and subsequently the ZnO were than deposited onto the bottom Pt contact,
which was previously DC-sputtered on the Si substrate. ZnO grows highly tex-
tured polycrystalline on the polycrystalline BTO, with a preferred c-axis (002)
orientation of the ZnO and multiple different orientations of the BTO [277] . The
phase transition from the tetragonal ferroelectric phase to the cubic paraelectric
phase was observed in such structures by Raman and CV measurements [277] .
Polarization measurements revealed a significant change in the ferroelectric
hysteresis loop in the presence of a ZnO layer adjacent to the BTO [278] . While
the ST loop of a single BTO layer is highly symmetric, a strong asymmetry is
observed in the ZnO/BTO heterojunctions. The actual shape depends strongly
on the measurement parameters, as cycling frequency and voltage amplitude.
This behavior was already reported in Ref. [278] , and is exemplified in Fig 6.1.
In general the area surrounded by the loop increases with increasing voltage
amplitude, indicative for the fact, that the polarization of the ferroelectric in-
creases. This further shows, that the polarization of the ferroelectric in the
investigated junction is not in saturation yet.
Additionally the current flowing trough the heterostructure shows a hysteresic
behavior as depicted in Fig. 6.2. The overall shape of the IV dependence re-
sembles that of a semiconductor diode. For such a structure the current voltage
characteristic can be modeled following Eqn. 2.18. The determination of the
barrier height and ideality factor is obscured by the presence of the hysteresis,
however the parallel and series resistance can be estimated with high accurancy.
Especially the parallel resistance can be easily calculated from the current in
backwards direction. For the modeling the average of both parts of the hyster-
sis loops was used. Therefore numerous heterojunctions were measured on a 3
inch wafer of Si. Maps of the parallel and series resistance are given in Fig. 6.3.
The spread in the parallel resistance is almost 3 orders of magnitude, and the
parallel resistance can be as low < 104 for individual contacts. The series

122
6.1. Effects observed in ZnO/BTO heterostructures

100.00 Hz 1000.00 Hz
4

3 1

2 0.5

1 0

Vo
Vo

-0.5
-1
-1
-2
-1.5
-3
-2
2V
-4
3V
-5 0 5 -5 0 5
Vi Vi 4V
5V
10000.00 Hz 100000.00 Hz 6V
1 7V
1
8V
0.5
0.5
0

0
Vo
Vo

-0.5

-0.5
-1

-1.5 -1

-2 -1.5
-5 0 5 -5 0 5
Vi Vi

Figure 6.1: SawyerTower-circuit response of a ZnO/BTO heterostructure grown


on Si with Pt contacts measured at various cycling frequencies and voltage amplitudes
as indicated. The data was obtained from selected contacts on the wafer shown in
Fig. 6.3.

resistance has a much smaller spread, but remains of the order of 100 . This
value is much to small, if the BTO is to be considered as an insulator. The poly-
crystalline nature of the BTO introduces a high density of grain boundaries,
which most likely serve as parallel conducting channels trough the BTO layer.
Also the density of shunts trough the ZnO must be very high, resulting in the
low parallel resistance. The structure is able to serve as a model to understand
physics of ZnO/BTO heterojunctions, an application in an electrically driven
device is unlikely in the present form, as parasitic power consumption would
be huge.

123
6. Polarization related effects in ZnO/BaTiO3 heterostructures

current density (A/cm2)


10-2

10-3

10-4

10-5

10-6
-2 0 2

Voltage (V)

Figure 6.2: Current voltage characteristic of a ZnO/BTO heterojunction grown on


Si with Pt contacts.

6.1.2 ZnO/BTO samples grown on lattice matched sub-


strates
In order to improve the properties of the BTO layers, thin BTO films were
grown on isostructural substrates, in this case STO substrates [279] . In order
to form contacts to the back side of the BTO films, either a conductive SRO
layer was deposited prior to the deposition of the BTto the samples grown on
silicon. First of all, the structural properties differ largely. Figure 6.4 shows
the 2 scans of a sample grown on silicon and one deposited onto STO. The
same angular range is chosen for better comparison. The peaks related to the
ZnO and the BTO layers are encircled in blue and green respectively. While in
both cases only one (002) ZnO reflex is found in the whole angular range due
to the tendency of ZnO to grow c-axis oriented [8790] , the situation is different
for the BTO. Three different orientations of BTO crystallites are observed for
the material grown on Si, while only the (001) orientation is present in the
material grown on the STO substrates. Further, the peak width of both (002)
ZnO and (001) BTO is reduced on the STO substrate. The peak intensity of
the ZnO (002) reflection is increased by more the two orders of magnitude for
the ZnO/BTO heterostructure grown on the STO substrate, indicating that a
uniform, defined BTO (001) underlayer results in a much better orientation of
the ZnO layer on top of it.
While no in-plane epitaxial relationship between ZnO and BTO was found by
XRD in ZnO/BTO heterostructures grown on Si, such a relationship is seen
for the films grown on STO substrates. Figure 6.5 gives -scans of the STO
substrate, and the BTO and ZnO layers. From the alignment of the four (103)
BTO peak with the (310) STO peak it can be inferred, that the BTO grows
with the [100] and [010] directions of the STO and the BTO aligning. In con-
trast, the (104) ZnO reflection shows a 12 fold in-plane symmetry, where none
of the peaks coincide with that of the (103) BTO and (310) STO peak. Four
of the 12 peaks are however shifted by 45 with respect to the (103) BTO and

124
6.1. Effects observed in ZnO/BTO heterostructures

> 5103 W

> 1104 W

> 2104 W

> 5104 W

> 1105 W

> 2105 W

> 5105 W

> 1106 W

> 2106 W

> 5106 W

> 40 W

> 50 W

> 60 W

> 70 W

> 80 W

> 90 W

> 100 W

> 110 W

> 120 W

> 130 W

Figure 6.3: Maps of the (a) parallel and (b) series resistance of ZnO/BTO het-
erojunctions of a 3 inch silicon wafer. Individual contacts are color coded. Colors
represent certain values of the parallel and series resistance, whereas gray contacts
show an ohmic behavior.

125
6. Polarization related effects in ZnO/BaTiO3 heterostructures

105

Pt (111)
Intensity counts (arb.u.)

104

BTO(110)
103

BTO(111)

BTO(002)
BTO(001)

ZnO(002)
Pt/Si(011)
102

101

100
20 25 30 35 40 45 50

2? ()

STO(200)
BTO(002)
BTO(001)
STO(100)

107
Intensity counts (arb.u.)

ZnO(002)

106
Au (111)

105

104

103
K
K
K?

102

101

100
20 25 30 35 40 45 50

2? ()
Figure 6.4: Wide angle 2 scans of ZnO/BTO heterostructures grown on (a)
Si/Pt and (b) STO/SRO. The Peaks corresponding to ZnO and BTO are encircled in
blue and green, respectively. On Si/Pt, the BTO shows a polysrystalline orientation,
whereas on STO/SRO the BTO is (001) oriented. (Data obtained by Aziz Abdullah
and Dr. Michael Lorenz)

126
6.1. Effects observed in ZnO/BTO heterostructures

STO:Nb (310) ZnO (104)


BTO (103)
103
Intensity (cps)

102

101
-180 -90 0 90 180
()
Figure 6.5: XRD -scans of a ZnO/BTO heterostructure grown on a STO:Nb
substrate. The scans were performed using the asymmetric (104), (103), and (310)
reflexes of ZnO, BTO, and STO, respectively on a wide-angle goniometer. (Data
obtained by Dr. Michael Lorenz)

(310) STO peaks. This indicates an in-plane epitaxial relationship, where the
[100] ZnO direction is parallel to the [110] BTO direction. As there exist four
equivalent [110] direction in the BTO lattice, which are rotated by 90 each,
and ZnO does not contain the 90 degree rotation as a symmetry element, two
different types of ZnO crystallites must be present in the sample. As ZnO
contains the 60 rotation as a symmetry element, both types of crystallites are
rotated by 30 , giving a twelfefold reflex pattern. The presence of such 30
rotational domains is reported for ZnO grown on c-plane sapphire [280] . It has
also been reported that these 30 rotational domains correspond to an inver-
sion in the ZnO polarity [34] , in that case rendering a sample with both Zn- and
O-polar crystallites.
In addition to these changes in the crystal structure, the electrical properties
of the ZnO/BTO heterostructures grown on STO improved compared to the
samples grown on silicon. The leakage current reduced significantly, down to
values in the nA range for BTO layers of roughly 1m thickness. The leak-
age current is a function of the BTO thickness, and increases with decreasing
thickness. Fig. 6.6 shows the current voltage characteristic of a ZnO/BTO het-
erostructure consisting of a 800 nm thick BTO and a 500 nm thick ZnO layer.
A significant hysteresis is again observed, however the shape differs signifi-
cantly from that in the heterojunctions grown on Si. This can be understood
if the origin of the hysteresis is considered. In the ZnO/BTO heterostructures
a depletion layer is formed, if a positive voltage is applied to the ZnO side

127
6. Polarization related effects in ZnO/BaTiO3 heterostructures

2dBTOEC
6 t i (ms)
0.2
4 2
Current (nA)

20
2 200
2000
0

-2
dZnOPZnO/
-4
0 5 10 15 20
Voltage (V)
Figure 6.6: Current flowing into a ZnO/BTO heterostructure grown on a STO:Nb
as function of applied voltage and integration time. The ZnO has a thickness of
500 nm, the BTO 800 nm.

of the junction, which is large enough to remove possible accumulation in the


vicinity of the heterointerface caused by the difference in the electric polariza-
tion of ZnO and BTO. Such a depletion region will lead to the formation of a
potential barrier, which impedes the transport of carriers. If the structure has
a low series resistance, the transport is largely controlled by the presence of
this potential barrier. The current flowing is controlled by the presence of the
depletion layer, which in turn is also a function of the change in polarization
at the interface between ZnO and BTO. This difference is a function of the
polarization in the BTO, which shows a typical ferroelectric hysteresis. This
case is present in the ZnO/BTO heterostructures grown on Si.
In contrast, if the resistance of the heterostructure is very large, the current
is mostly limited by the series resistance. The formation of the depletion layer
will therefore only have a small influence on the transport properties, and no
significant hysteresis will be observable related to the different extension of the
space charge region and the difference in the polarization at the interface. The
currents measured are not only currents flowing trough the structure, but also
currents transporting the charge necessary to change the ferroelectric polariza-
tion. As the current trough the structure is now small enough, the polarization
current determines the IV characteristic. Such a behavior is also observed for
highly insulating BTO single layers [281] . The current will be a function of time,
as the charge necessary to switch the polarization of the ferroelectric domains
is finite. This behavior is observed in Fig. 6.6, where two contribution of the

128
6.1. Effects observed in ZnO/BTO heterostructures

current can be distinguished. The hysterestic part of the current decreases with
increasing integration time. It is remarkable, that this current can be negative,
even if the applied voltage is positive. This can be understood by the influence
of the ZnO polarization, which can not be changed by the applied voltage.
In a ferroelectric, the current transporting the charge necessary to change the
polarization will show a maximum at the voltage position of the steepest incli-
nation in the ferroelectric hysteresis loops [281] . At this point the largest change
in polarization per unit voltage occurs. Assuming a constant voltage cycling
rate (which can be controlled in the experiment), the rate of unit voltage per
unit time is constant. Therefore, at the steepest point in the ferroelectric hys-
teresis the largest current will flow into the structure. In an ideal undisturbed
ferroelectric, the hysteresis loop will be symmetric with respect to zero. In the
presence of a ZnO layer, however, the BTO polarization will have a preferred
orientation which is related to the magnitude of the polarization in the ZnO.
The hysteresis loop will therefore be shifted along the voltage axis. The value
of the ZnO polarization can be deduced from the magnitude of this shift, which
in this case is PZnO = 0.2 0.05 C. Additionally, in an ideal ferroelectric, the
steepest change in the polarization versus voltage occurs at the point where
the polarization is zero. If the presentation of P (E) is considered, the electric
field at this point is the coercive field. The maxima observed in the current
voltage dependence can therefore be used to determine the coercive field, in
this case EC = 12.5 1 kV/cm. This value is comparable to typical values
obtained from BTO thin films [282] . The ZnO polarization however, is about on
order of magnitude smaller than in the literature [16,17,283,284] .
Figure 6.7 shows the IV curves obtained on two contacts of the same sample
with a ZnO thickness of 50 nm and a BTO thickness of 800 nm. The leakage
currents are again below 1 nA, and peaks are observed in the IV curves. The
separation of the peaks is much less pronounced in this case, and the peaks
are significantly broadened. Also the mean value of the peaks lies at smaller
voltage amplitudes, which is understood regarding the smaller film thickness.
It has therefore to be pointed out, that the relative width of the peaks is larger
compared with that measured in the sample with the 500 nm thick ZnO film.
The experimental errors are therefore increased. It is nevertheless obvious,
that differences exist in both curves. While in Fig. 6.7 (a) the center of gravity
of curves is at a positive voltage, it is at a negative voltage in Fig. 6.7 (b).
This indicates, that two different ZnO polarization orientations influence the
transport properties of both contacts, which are on the same sample. This can
be understood assuming the same situation is present in ZnO growth on BTO
as it is on c-sapphire [34] . It has however to be remarked, that the origin of the
two ZnO polarities might be slightly different. In the case of ZnO on c-plane
sapphire, the reason for the existence of both polarities is that the first layer
of ZnO might either be a O layer or a Zn layer, depending on the bonding
partner in the topmost layer of the sapphire. The binding to aluminum atoms
is energetically preferred considering a single bond, however the formation of
crystallites of the Zn face polarity results in a high (31.8%) lattice mismatch.
The bonding to oxygen atoms in the topmost layer is less favorable for a single

129
6. Polarization related effects in ZnO/BaTiO3 heterostructures

2
a)
1.5

1
Current (nA)

0.5

-0.5

-1

-1.5
-5.0 -2.5 0.0 2.5 5.0
Voltage (V)

b) 0.4

0.2
Current (nA)

-0.2

-0.4

-0.6

-6 -4 -2 0 2
Voltage (V)
Figure 6.7: Currentvoltage characteristic of two selected contacts of a ZnO/BTO
heterstructure grown on STO:Nb with a reduced ZnO thickness of 50 nm.

130
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

bond, however represents the global energy minimum. It is a question of the


growth temperature if this global minimum is reached. Both growth orienta-
tions can coexiste in one sample [34,280] . The situation is different for the growth
of ZnO on BTO. The existence of 30 rotational domains in ZnO is inherent
to the symmetry of the BTO lattice. Nevertheless the topmost layer contains
at least two species, Ba and O. If the BTO unit cell is not completely filled,
Ti is also present in the topmost layer. Bonding sites therefore exist for both
Zn and O, and the coexistence of both polarities is thinkable. In order to form
high quality ZnO films on BTO the polarity of the ZnO has to be controlled.
The presence of both 30 rotational and inversion domains will hinder in-plane
transport in the ZnO, which is a key issue for the formation of devices such as
field effect transistors.

6.2 Theory of coupling phenomena in ZnO/BTO


heterostructures
From the experimental observations considerations about the nature of the
coupling of the polarization of ZnO and BTO, and the formation of deple-
tion layers within the heterojunctions arose. Those considerations do not only
include the single ZnO/BTO heterostructures, as discussed in the previous sec-
tion, but also double ZnO/BTO/ZnO heterostructures. No extension of the
physical assumptions included in the model is required to predict the prop-
erties of such a structure, once single heterostructures are understood. The
work was carried out in close cooperation with Venkata M. Voora and Prof.
Dr. Mathias Schubert (University of Nebraska, Lincoln). Some of the results
presented in this section were published previously [283,285,286] .

6.2.1 Derivation of a model describing the coupling phe-


nomena
The properties of the heterostructures are derived from the distribution of
charges within the individual layers. For simplification several assumptions
are drawn in the beginning.
1. The BTO is treated as to be ideally insulating. As outlined in the pre-
ceding section, this is not self-evident. It can however be motivated, that
upon adequate experimental conditions BTO layers of very high resis-
tivity can be grown. For the best insulating layers the resistivity of the
BTO layers was more then 12 orders of magnitude above that of the ZnO.
Even for inferior samples, the assumption is still valid in the form that the
conductivity of the BTO is negligible compared to that of the ZnO. This
assumption however introduces a limitation to the model, that is, that
the current passing through the heterojunctions will not be predictable.
The model will therefore be limited to a capacitive response of the het-
erostructures. Nevertheless, a detailed understanding of the distribution

131
6. Polarization related effects in ZnO/BaTiO3 heterostructures

of space charge regions will be gained from the model, which will prove
valueable, at least for a qualitative understanding of the current flow in
the structures where the space charge regions introduce the main limita-
tion to the current flow. As a consequence, however, the current flowing
through the structure must be zero.
2. Any charge transport within the junction will be instantaneous. No ex-
plicit time dependence exists for the ferroelectric response to an applied
field, and for the ionization of the donors in ZnO upon formation of a de-
pletion layer. Tough neither of these two facts is true in a real sample, the
assumption introduces significant simplifications in the calculation of the
properties as any explicit time dependence is removed from the quan-
tities involved. Additionally, the experiment can be easily performed
sufficiently slow that explicit time dependence of as well the ferroelectric
response as the donor ionization does not play a role, and the assumption
holds true.
3. The ZnO has a finite net donor concentration Nt , which can be removed
2
from sublayers of thickness w by an external voltage of size eN2t wz , where
e is the unit charge, and z the permitivity of ZnO. This will lead to the
formation of a space charge region. Additionally a voltage is required
to drive carriers through the ZnO due to its finite conductance. This
however will be neglected as a consequence of the first two assumption,
due to the absence of steady state currents, and currents resulting from
the redistribution of charge within the structure. Further, carriers can
be accumulated within the ZnO layer, which also requires an external
voltage. This will also be neglected, and the ZnO will be treated as ideally
conductive. Even this is not the case in reality, the error introduced by
this assumption is small.
Following these assumption two possible responses of the heterojunctions to
external excitations exist. First the ferroelectric component can be polarized
by the external field. The response is described by a phenomenological electric
field dependent polarization model [287]
 
Ef EC
+
Psat (Ef ) = Ps tanh . (6.1)
2
Here the plus sign in the superscript indicates the saturation polarization Psat
+

on the lower branch of the saturated polarization hysteresis loop at electric


field Ef after a given electric field increment. EC denotes the coercive field.
The quantity is obtained from
  1
1 + Pr /Ps
= EC log , (6.2)
1 Pr /Ps
where Pr denotes the remanent polarization and Ps is the maximum saturation

polarization. Psat after a given electric field decrement is obtained from Eq. 6.1
as follows

Psat +
(Ef ) = Psat (Ef ) . (6.3)

132
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

For an unsaturated polarization loop Pd does not necessarily follow the sat-
uration hysteresis loop. In order to calculate an unsaturated hysteresis loop
an approach suggested by Miller et al. is employed [288] . In this model, the
saturated and unsaturated polarizations at a given field Ef can be related by a
factor , which ensures smooth and monotonic approaching of the unsaturated
polarization to the saturation polarization loop. This factor relates the first
partial derivatives of both loops
Pd Psat
= , (6.4)
Ef Ef
and  
Pd Psat
= 1 tanh . (6.5)
Ps Pd
The factors = +1 and = -1 are for incrementing and decrementing fields
Ef , respectively, and Pd is the ferroelectric polarization on the unsaturated
hysteresis loop at Ef .
Second, upon the appearance of a difference in the polarization at a heteroin-
terface, a depletion layer can form in the ZnO layer adjacent to the interface.
As discussed in the third assumption, for removing the mobile charge from
2
sublayers of thickness w by the potential of size eN2t wz is required. The ar-
rangement of space charge regions is exemplarily depicted in Fig. 6.8 for a
double heterostructure. All further considerations will be developed for a dou-
ble heterostructure, as the properties of a single heterostructure can be easily
determined from setting the thickness of one of the ZnO layers to zero. In
the following discussions the polarization will be counted positive, if it is ori-
ented as given in Fig. 6.8, pointing from the ground contact of the structure
towards the contact with the applied voltage. Further both the thickness and
the magnitude of the polarization for the ZnO layers will be set to a single
value. For the thickness any influence is canceled out, as the voltages applied
are small, so that none of the layers will be fully depleted. As the non depleted
ZnO is treated as ideal conductive, the part of the ZnO layers included in the
calculations is only the depletion layer. Nevertheless care was taken during
the deposition of the layers to ensure equal thicknesses, as the assumption of
ideal conductive layers is not fulfilled, in real samples. For the polarization,
fixing the amplitude to a certain value is explained by the fact, that it is a con-
stant for ZnO, however the growth orientation of both layers can be different,
therefore the sign will be left a variable. For further discussions, the ZnO layer
adjacent to the ground contact will be denoted as layer 2, that adjacent to the
excitation bias contact will be denoted as layer 1. The ferroelectric layer will
be described as layer F.
In order to calculate the polarization of the BTO the electric field distribution
inside the sample has to be related to the applied voltage. The voltage drop
across the sample is the sum of the voltage drops across the sublayers. Again,
a voltage will only drop over the depleted parts of the ZnO layers.
Z w1 Z w2
V = df Ef + E1 dx + E2 dx , (6.6)
0 0

133
6. Polarization related effects in ZnO/BaTiO3 heterostructures

ZnO BTO ZnO


dz df dz

w2 w1
V

ez ef ez
Psz Pd Psz

Figure 6.8: Schematic of the distribution of space charge regions within the hetero-
junctions. Here dz and df are the thicknesses of the ZnO and BTO layers, w1 and w2
are the extension of the space charge regions in layer 1 and 2, respectively. PZnO and
Pd denote the electric polarization in the ZnO and BTO layers, respectively, while
z and f represent their relative permittivity. (Courtesy of Venkata M. Voora)

where Ef is the electric field in layer F, E1 and E2 are the fields and PZnO1 and
PZnO2 the spontaneous ZnO polarizations in layer 1 and layer 2, respectively.
The depletion layers will be assumed to be abrupt (Schottky-approximation),
and the concentration of donors will be treated independently, as the donor
concentration Nt1 in layer 1 might differ from Nt2 in layer 2 due to slightly
different growth conditions. The electric field will therefore be a linear function
of x, and the voltage drop will become a quadratic function of the individual
depletion width, as outlined before. The field in the part of the ZnO layer
away from the junction will be zero, which results in different dependencies of
E1 and E2 on x,
eNt1 w1
E1 (x = 0) = E1 (x = w1 ) = 0 (6.7)
z
eNt2 w1
E1 (x = 0) = 0 E1 (x = w2 ) = . (6.8)
z
With Eq. 6.7, and replacing Ef = 1
f (Df Pd ) one obtains

df e
V = (Df Pd ) + (Nt1 w12 Nt2 w22 ) , (6.9)
f 2z
where Pd , Df , f are the ferroelectric polarization, displacement, and dielectric
constant of layer F, respectively. Df is equal to the surface charge density on
the left side of layer F, b . Dielectric displacement continuity across interface
1F requires
eNt w1 + PZnO1 = Ef f + Pd , (6.10)
and across F2
Ef f + Pd = eNt w2 + PZnO2 . (6.11)

134
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

Table 6.1: Operation schemes according to depletion layer formation, variables w1 ,


w2 , and corresponding junction conditions (forward-like F, reverse-like R ) between
layer 1 and layer F (1F) and layer F and layer 2 (F2), respectively.

Configuration w1 w2 Junction 1F Junction F2


(A) 0 0 F F
(B) 0 < w1 < dz1 0 R F
(C) 0 0 < w1 < dz1 F R
(D) 0 < w1 < dz1 0 < w1 < dz1 R R
(E) dz1 0 R F
(F) 0 dz1 F R
(G) 0 < w1 < dz1 dz1 R R
(H) dz1 0 < w1 < dz1 R R
(I) dz1 dz1 R R

The values w1 and w2 can be taken to distinguish different operation schemes


the heterostructure is in, their labeling is summarized in Tab. 6.1. Operation
schemes (E)(I) are included for completeness, within the present study the
driving voltage will be limited to values where a full depletion of the ZnO lay-
ers is not reached. Therefore only the schemes (A)(D) will be of importance.
In order to probe the properties of the samples SawyerTower measurements
where performed on them. Therein a sinusodial voltage is applied to the sam-
ple. In order to compare the experimentally obtained with the calculated loops,
the time derivative of the charge transfered to the heterostructure, a current
ji is required as a function of the time derivative of the excitation voltage.
One of the basic assumptions of the present model is, that none of the internal
charge transfer processes has a explicit time dependence. Nonetheless all in-
ternal quantities implicitly depend on time through the excitation voltage. In
order to evaluate the charge transferred to the structure a new quantity will be
defined, the charge on the virtual metal plate within layer 1 m . The introduc-
tion of this virtual metal plate is required by assumption 3, wherein the ZnO is
treated as ideal conductive (metallic), as long as no depletion region is present.
The position of the virtual metal plate is not constant, but determined by the
extension of the space charge region in layer 1.
As long as no space charge region is present, the metal plate is situated at the
interface between layer 1 and layer F. The charge density on the metal plate
will be positive, and equal b = Df . In this case, the current flowing into the
structure directly equals the time derivative of the dielectric displacement at
the interface 1F
df dDf
ji = = . (6.12)
dt dt
As b crosses zero and becomes negative, m will not follow this response.
The reason for that is that free electrons inside the ZnO are repelled by the
negative sheet charge b and withdrawn from layer 1. The virtual metal plate
will wander towards the electrode. The negative sheet charge b that would be
required on a plate at the heterointerface 1F is completely compensated by the
positive space charge within the depletion layer. Nevertheless, a finite positive

135
6. Polarization related effects in ZnO/BaTiO3 heterostructures

current will be flowing into the structure, in order to remove the electrons
from the depletion layer. This current will equal the change of the charge
stored in the depletion layer, as every change in the sheet charge density at the
heterointerface 1F is expressed by a change in the size of the depletion region

d(eNt1 w1 ) df dDf
ji = = = . (6.13)
dt dt dt
For the theoretical case of full depletion a finite negative charge density will
be found on the metal plate, which will we reduced in comparison to b by the
constant charge stored within the ZnO layer eNt1 dz1 . The current flowing into
the structure will again be
df dDf
ji = = . (6.14)
dt dt
This current can now be calculated via the dielectric displacement from the
electric field and polarization within the BTO layer
dDf d
ji = = (f E + Pd ) . (6.15)
dt dt
Herein the time dependence of the Polarization is related to the time depen-
dence of the electric field by the chain rule
dPd dPd dE
= , (6.16)
dt dE dt
giving  
dPd dV
ji = f + , (6.17)
dE dt
where gamma relates the external voltage and the electric field within the BTO
layer   1
w1 + w2 dPd
= df + f + , (6.18)
z dE
and therefore represents the inverse of the effective thickness of the sample,
which can exceed the actual thickness of the BTO layer, whenever a depletion
layer is present. The contribution of the depletion layer
 width to the effective
sample thickness is amplified by the term z f + dE which is typically much
1 dPd


larger than 1, due to the different permitivities of ZnO and BTO. Within this
framework of equations the extension of the space charge regions inside the
ZnO layers can be calculated at all times. Following the approach of Miller
et al. [287] , the response measured in the SawyerTower circuit is given by the
Kirschhoff rules
d(Cst Vo ) Vo Vin Vo
ji Af = , (6.19)
dt Rn Rs
where Vin is the sinusodial voltage applied to the circuit, Vo is the voltage
across the normal capacitor, Cst is the capacitance of the standard capacitor,
and Rs and Rn are parallel resistances of the sample and standard capacitor

136
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

Table 6.2: Best-match model calculation parameters.

Parameter Best-match model values


BTO BTOZnO ZnOBTOZnO
dz (m) 0.5 0.5
df (m) 1.5 1.45 1.4
Nt (cm3 ) 5.51016 21017
Vbi (V) 0 -0.780.2 0
f 202 250 80
ZnO 8 6
EC (kV/cm) 11.5 11 11
Ps (C/cm2 ) 7.54 14.1 6
Pr (C/cm2 ) 1.39 5.6 0.11
Rs (k) 1000 13 1000
PZnO (C/cm2 ) -4.1 -0.1

respectively. In order to predict the output measured in the experiment, the


dependence of Vo on Vin has to be calculated. An iterative technique is applied
following Ref. [287] . Herein the differential relation yielding Pd is integrated, and
the change in Pd in step l is calculated using the slope obtained in the previous
iteration step l 1  
dPd
Pd = Efl . (6.20)
dE l1
After each differential change in the applied voltage the actual depletion widths
w1 and w2 are calculated from the input voltage and the polarization obtained
in the previous step. Only positive values for w1 and w2 ate taken into account.
The field within the ferroelectric is then calculated following Eqn. (6.18). Fi-
nally the polarization of the ferroelectric is calculated after Eqn. (6.20).

6.2.2 Application of the model to experimental data


The predictions of the model have been tested on experimental data obtained
from several samples. The model was applied to single BTO layers, ZnO/BTO
single heterostructures and ZnO/BTO/ZnO double heterostructures. Repre-
sentative ST output loops are given for all three types of samples in Fig. 6.9.
The parameters varied in order to match the experimental loops are summa-
rized in Tab. 6.2. The frequency of the sinusodial excitation was fixed to
1.5 kHz, the capacitance and resistance of the normal capacitor were 1 C and
1 M, respectively. The thicknesses of the individual layers were measured
by spectroscopic ellipsometry, and are not varied during the calculations. The
quantity Vbi is introduced in the case of the ZnO/BTO heterostructure in order
to account for the finite offset of the conduction band between ZnO and BTO.
For all three structures a very good agreement between the calculated and the
experimentally measured data was obtained. In the case of the single BTO
layer (see Fig. 6.9 (a)), the voltage across the standard capacitor is propor-

137
6. Polarization related effects in ZnO/BaTiO3 heterostructures

(a)

(b)

(c)

Figure 6.9: Experimentally measured (dots) and model calculated (solid lines)
Sawyer-Tower loops for (a) a single BTO layer, (b) a ZnO/BTO single heterostruc-
ture and (c) a ZnO/BTO/ZnO double heterostructure. The left panels give the
voltage across the standard capacitor as a function of the sweep voltage, the right
panels show the polarization of the ferroelectric as a function of the input voltage.
(Courtesy of Venkata M. Voora, data partially obtained by Nurdin Ashkenov)

138
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

tional to the polarization in the ferroelectric layer. The values obtained for
the coercive field and the remanent and saturation polarization agree well with
the values estimated from the IV measurements in Chap. 6.1.2 and literature
values for polycrystalline BTO thin films [289,290] . They further agree exactly
with what would be calculated from a conventional analysis of the ST loops.
The situation changes significantly in the case of the ZnO/BTO single het-
erostructures (see Fig. 6.9 (b)). Again a very good agreement between the
measured and calculated loop is achieved, however in this case the polariza-
tion inside the ferroelectric and the voltage across the standard capacitor are
not equal. A conventional analysis is not able to determine the parameters of
the individual layers in this case. The difference in the shape of the ST loop
and the real voltage dependence of the ferroelectric polarization is of cause
introduced by the presence of a space charge region within the ZnO. The cal-
culations allow to determine the the voltage points at which the depletion layer
starts to form, these points are marked by triangles in Fig. 6.9 (b). The de-
pletion layer is present on the right side of the loops, and a significant fraction
of the applied voltage now drops across this layer. The effective thickness of
the sample in this case is larger than the thickness of the BTO layer, and the
field within will be reduced Ef < V /df . The change in polarization per unit
voltage will therefore also be reduced, as can be seen from the right panel of
(see Fig. 6.9 (b)). Only a subloop of the possible ferroelectric hysteresis loop
is therefore reached in this structure.
The situation becomes more complicated in the case of the ZnO/BTO/ZnO
double heterostructure, as seen in Fig. 6.9 (c). Here the depletion layer forms
for both positive and negative voltages, and the pure ferroelectric response of
the layer system is, if present at all, only observed for a small range of input
voltages. The question for which part the hysteresis curve of a single BTO layer
is still reached, is not trivially answered. Here the orientation of the ZnO po-
larization is important, and four different arrangements can be distinguished,
giving the two possible orientations in both ZnO layers. The four possible ar-
rangements are graphed in Fig. 6.10. For two of them the ZnO polarization
vectors are aligned in parallel (arrangements P FP and P+ FP+ ). In this case
there is always exactly one depletion layer present either in layer 1 or in layer
2. The effective sample thickness then always exceeds the BTO layer thick-
ness. The attainable polarization of the BTO is therefore significantly reduced
in such samples with respect to single BTO layers (compare the scales in the
right panels of Fig. 6.9 (a) and (c)). Depending on the actual direction of the
ZnO polarization vectors, the points where the transition from the depletion
of layer 1 to the depletion of layer 2 occurs changes with respect to the input
voltage.
In the other two possible arrangements the ZnO polarization vectors are aligned
antiparallel (arrangements P+ FP and P FP+ ). In order to understand the
differences between both arrangements, it is helpful to consider the state of
the heterostructure in the absence of a ferroelectric polarization. The sheet
charge density at the heterointerface will then only be influenced by the ZnO
polarization. As the individual polarization vectors are antiparallel, the sheet

139
6. Polarization related effects in ZnO/BaTiO3 heterostructures

0.5
w w 140
-0.1
140
w w
ZnO BTO ZnO 1
2

0.4 -P -P
2 1
sz sz
-0.2
[V]

[V]
0.3

[nm]

[nm]
-0.3
o

70 70

o
V

V
w

w
0.2 ZnO BTO ZnO
-0.4
P P
sz sz

0.1
-0.5
0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
V [V] V [V] V [V] V [V]
in in in in

(a) Arrangement P FP (b) Arrangement P+ FP+

0.03
0.04
ZnO BTO ZnO

w
140
w w
-P P
w 2 1
2 1
140
sz sz [nm]
[V]

[V]

[nm]
0.00 0.00
70
70
o

o
w
V

w
ZnO BTO ZnO

P -P
sz sz

-0.04 -0.03 0
0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
V [V] V [V] V [V] V [V]
in in in in

(c) Arrangement P+ FP (d) Arrangement P FP+

Figure 6.10: Model calculated SawyerTower loops (left panels) and depletion layer
widths (right panel ) in ZnO/BTO double heterostuctures for different arrangements
of the ZnO polarizations in both layers, as explained in the text. (Courtesy of Venkata
M. Voora)

charges on both sides will be equal, positive in the P+ FP arrangement, and


negative in the P FP+ arrangement. It is clear, that regardless of the polar-
ization of the BTO the sum of the depletion layer widths will be always larger
in the P FP+ arrangement than in the P+ FP arrangement. The effective
thickness is therefore always larger in the P FP+ case, and the polarization
has to remain smaller (this manifests also in smaller voltages across the stan-
dard capacitor, as seen in comparison between the left panels of Fig. 6.10(c)
and Fig. 6.10(d). The accuracy of the parameters describing the ferroelectric
properties derived from the model is therefore not as good as for the cases of a
single BTO layer, or the ZnO/BTO heterostructure, especially the saturation
polarization, as the structures are typically far away from saturation.
From the model, further prediction of the capacitance of the structures is
possible. The capacitance will be a function of the external voltage, through
the depletion layer width within the ZnO layers. The capacitance of the BTO
layer will be treated as a constant, which is not true for single BTO layers.
As the structure represents a series of capacitances, the overall capacitance is
determined by the smallest individual capacitance, which is that of the ZnO.
The changes in the BTO capacitance will therefore be negligible compared to
the changes introduced by the occurrence of the depletion layer. The ferroelec-

140
6.2. Theory of coupling phenomena in ZnO/BTO heterostructures

C [arbitrary units]
(a) (b)

C [arbitrary units]
SIS SFS

(c) (d)

C [arbitrary units]
C [arbitrary units]

- - + +
P FP P FP

-5 0 5 -5 0 5
V [V] V [V]
in
in

Figure 6.11: Principle of CV response of (a) a semiconductorinsulator


semiconductor (SIS) heterostucture (b) a semiconductorferroelectricsemiconductor
(SFS) heterostructure and (c) and (d) of ZnO/BTO double heterostructures with
two different parallel arrangements of the ZnO polarization. (Courtesy of Venkata
M. Voora)

tric nature of the BTO will however be regarded, as it implicitly influences the
depletion layer widths. From the previous considerations the depletion layer
width is know as a function of the external voltage.
Figure 6.11 depicts a model calculated CV hysteresis curve for a semiconductor
insulatorsemiconductor heterostructure without ferroelectric hysteresis. For
semiconductor layers with equal donor concentrations and interface properties,
the CV curve is symmetric with respect to zero voltage (Fig. 6.11 (a)), and
decreases monotonically for increasing voltages due to formation of identical
space charge regions in either of the semiconductor layers depending on the
sign of the applied voltage. The space charge region infers a series capacitance
which decreases with increasing voltage, leading to an overall reduction of the
total capacitance. Figure 6.11 (b) shows the CV response if the insulator
between the two semiconductor layers is a ferroelectric. In that case the CV
response is shifted along the voltage axis, depending on the actual polariza-
tion in the ferroelectric constituent. The ferroelectric polarization produces
an additional interface charge, which offsets the space charge region formation
depending on the sign of the charge. At certain external voltages, the electric
field within the ferroelectric layer is sufficient to switch the polarization, iden-

141
6. Polarization related effects in ZnO/BaTiO3 heterostructures

(a) (b)

10

BTO
ZnO

ZnO
V
-7 in P
10
sz

P
log|I| [A]

sz

C [nf]
-9
7
10

-11
10 4
-5 0 5 -5 0 5
V [V] V [V]
in in

Figure 6.12: Experimental (a) IV and (b) CV characteristic of a ZnO/BTO/ZnO


double heterostructure. (Courtesy of Venkata M. Voora)

tified by the maxima in the CV response. A typical symmetric butterfly-type


hysteresis is observed centered around zero voltage. Figure 6.11 (c) depicts the
scenarios when the two semiconductor layers carry spontaneous polarizations
with their associated polarization vectors parallel to each other (arrangements
P FP and P+ FP+ ). As in the arrangements with parallel orientation of the
ZnO polarization the contribution of the ZnO polarizations will generate sheet
charges of different sign on both heterointerfaces, and the overall CV response
will be shifted with respect to the excitation voltage. The direction and mag-
nitude of the shift depends on the arrangement and the size of PZnO .
For antiparallel arrangement, P+ FP and P FP+ , identical sheet charges are
located at both interface. While these charges still affect the formation of a
space charge region, the changes are symmetric with respect to the applied
external voltage, and the CV response is qualitatively equivalent to the case
depicted in Fig. 6.11 (b).
The theoretical predicted curves were indeed seen in experiments. Figure 6.12 (b)
gives the CV response of a ZnO/BTO/ZnO heterostructure. The correspond-
ing IV curve is given in Fig. 6.12 (a). From the shift of the capacitance peaks
with respect to the voltage axis and the relative peak height in both sweep
directions it is concluded, that the arrangement of the ZnO polarizations in
the heterostructure is of P FP type, thereas, both ZnO layers are grown in
the Zn polar orientation.

142
6.3. ZnO/BTO based ferroelectric thin film transistors

Source Drain

Au Au
ZnO Channel PLD 2

BTO Gate oxide PLD 1

STO:Nb Substrate

Gate

Figure 6.13: Principle layout of ZnO/BTO MISFET transistors.

6.3 ZnO/BTO based ferroelectric thin film tran-


sistors
Additionally to theoretical understanding of the properties of ZnO/BTO het-
erojunctions, experimental investigations to exploit their unique properties for
device application were performed within the present study. The ferroelectric
polarization can be used to control the conductivity in the ZnO layer, enabling
the fabrication of ferroelectric field-effect-transistors [291] . The first ferroelectric
field effect transistors have been reported by Wu et al. [292] , and since than sev-
eral reports emerged on such devices involving different materials materials.
In the ZnO system metalinsulatorsemiconductor field effect transistors (MIS-
FET) have been demonstrated using different gate oxides [293,294] . Further,
transparent metalsemiconductor field effect transistors (MESFET) have been
demonstrated for the ZnO system [295] . Recently, it has been shown [296] that
the fabrication of ZnO/Bax Sr1x TiO3 (BST) thin film transistors is possible.
However, the used BST was polycrystalline, and was not ferroelectric at room
temperature, due to its composition. However a dielectric constant of around
500 was obtained, which, in principle, allows a rather thick gate insulator.
The device discussed in this study was realized by the deposition of ZnO
thin films of 50 nm thickness by pulsed laser deposition (PLD) at a substrate
temperature of 670 C and an oxygen partial pressure of 0.002 mbar on ex-situ
prepared BTO templates. These templates were likewise prepared by PLD
by Dr. Jrgen Schubert from Forschungszentrum Jlich and consist of a BTO
layer of a thickness of 800 nm, deposited on a lattice matched Nb doped SrTiO3
(100) (STO:Nb) substrate. The ZnO target was ablated using a KrF excimer
laser with a pulse energy of 600 mJ and a repetition rate of 10 Hz. At first,
a 300 pulse nucleation layer was deposited at a reduced ablation rate of 1 Hz.
MISFET structures were patterned by photolithography. The ZnO was etched
by phosphoric acid in order to form mesas acting as the ZnO channels. The

143
6. Polarization related effects in ZnO/BaTiO3 heterostructures

Source - Drain Current Isd (A)


70
Ug =5V
60

50
Ug =4V
40

30 Ug =3V
20
Ug =2V
10
Ug =1V
0 Ug =0V
0 2 4 6 8 10
Source - Drain Voltage Vsd (V)
Figure 6.14: Output characteristics of a ZnO/BTO MISFET at different gate volt-
ages. The dashed line is a guide to the eye, underlining the quadratic behavior of
the position of the pinch-off-points.

channel length (L) is 30 m and the channel width W = 430 m. The con-
ductive STO:Nb substrate was used as a back-gate electrode. The back gate
geometry was chosen due to different etchability of ZnO and BTO. While ZnO
is easily wet-etched, dry etching methods are typically employed in order to
pattern BTO. Ohmic source and drain contacts, of 500 m 400 m size were
deposited by DC-sputtering of Au onto the ZnO, without overlapping onto the
underlying BTO layer. The schematic layout of the device is given in Fig. 6.13.
Metalinsulatorsemiconductor (MIS) diodes have been fabricated as a refer-
ence, by DC-sputtering of circular Au contacts on the ZnO with a diameter of
2 mm.
One necessary condition for the fabrication of high quality MISFETs is a
low leakage current of the gate insulator material. As discussed in Chap. 6.1.2
the use of lattice matched substrates reduces the leakage current significantly.
CurrentVoltage measurements on reference samples with circular contacts of
2 mm diameter revealed leakage currents as low as 6 nA at 20 V gate voltage,
giving a leakage current density of 200 nA/cm2 . The leakage current density
was significantly smaller for lower gate voltages (jl < 109 A/cm2 at 5 V).
Field-effect measurements on the samples show a linear region and a typi-
cal saturation behavior of the sourcedrain-current upon an increase of the
sourcedrain-voltage in the output characteristic (Fig. 6.14). The pinch-off
points follow the theoretically expected quadratic behavior [98] . The transfer
characteristic is shown in Fig. 6.15. The MISFETs are normally-on with a

144
6.3. ZnO/BTO based ferroelectric thin film transistors

Channel mobility ch (cm 2/Vs)


Source - Drain Current Isd (A)
0.30
10-5

10-6 0.25

10-7 0.20

10-8 0.15

10-9 0.10

10-10 0.05

10-11 0.00
-4 -2 0 2 4
Gate Voltage Vg (V)
Figure 6.15: Transfer characteristics and calculated channel mobility of a
ZnO/BTO MISFET at a fixed sourcedrain-voltage of VSD = 2V.

threshold voltage of VT = 2.4V, and the sourcedrain-current can be varied


by at least 6 orders of magnitude within a change in VG of 7V, reaching values
of up to 10 A. This results in a sub-threshold slope of 400 mV per decade.
The field effect mobility was calculated from the output and the transfer char-
acteristic, following

ID L 1
ch = , (6.21)
VD VD 0 W Ci (VG VT )

and is depicted in Fig. 6.15. The insulator capacitance was determined by


quasi-static-CV measurements (not depicted here), to be Ci = 7nF, yielding
a dielectric constant of the BTO of 197. The field effect mobility reaches a
maximum of 0.27, which is significantly below the mobility in bulk ZnO at
room temperature, and even significantly below the mobility in high quality
heteroepitaxial ZnO thin films [297] , however, the value is close to that reported
in literature for bottom gate ZnO MISFETs [298,299] . Three important scatter-
ing effects might contribute: First, the density of defects at the interface can
be large, accounting for a significant interface scattering [300] , and these defects
might create a high density of dislocation lines in the thin film, especially due
to the small thickness. Further, charged grain boundaries will also contribute
to carrier scattering, and a high density of these is likewise probable, due to
the low film thickness. This assumption is supported by the existence of the
30 rotational domains (see for detailed explanation Chap 6.1.2). It has to
be remarked, that the scattering at the 30 rotational domain barriers and at
inversion domain barriers, if they are present in the material, will have a more
dramatic effect on the mobility than scattering at low angle grain boundaries,

145
6. Polarization related effects in ZnO/BaTiO3 heterostructures

Source - Drain Current Isd (A)

a)
10-8

10-9

10-10
Isd (A) (switched on)
10-11 Isd (A) (switched off)
Vsd = 2V
10-12
1.5 2.0 2.5 3.0 3.5 4.0
Gate Voltage Vg (V)

1.000
b)
Isd(on) / Isd(off)

100

10

1
1.5 2.0 2.5 3.0 3.5 4.0
Gate Voltage Vg (V)
Figure 6.16: (a) Transfer characteristics of a ZnO/BTO MISFET after applying
a for- and backwards pulse, respectively. (b) Ratio of the current in the permanent
on and the permanent off state.

146
6.4. Summary and conclusion

as typically treated in literature [301] . In sum, a high density of charged crystal


imperfections is the most likely reason for the rather low mobility. In order
to improve the performance of the field-effect transistor the mobility has to
be improved significantly. Inversion of the transistor structure might solve the
problem, especially if the ZnO is grown homoepitaxially. One has to be aware,
however, of the fact that this might reduce the quality of the gate oxide leading
to increased leakage currents.
The transistor structures have been poled by exposure to increased gate volt-
ages (VG 20V) and the transfer characteristic was recorded between VG = 0 V
and VG = 5 V. After that, a sufficiently large negative erase voltage (e.g.
VG = 7V) was applied and the transfer characteristic was recorded again.
The process was repeated multiple times in order to check for reproducibility.
The results of these experiments have been depicted in Fig. 6.16 (a). The
transistor can be switched into two polarization states, and the transfer char-
acteristics in both states differ largely. The ratio of the sourcedrain-current
between these states is depicted in Fig. 6.16 (b). A positive gate voltage po-
larizes the BTO with the positive side towards the ZnO, creating an electron
depletion therein. A negative gate voltage causes a polarization in the BTO,
with the negative side facing the ZnO, leading to the formation of an accumu-
lation layer. The ratio of the currents in the both operation schemes can reach
values of up to 1,000, depending on the gate voltage. However, the source
drain-current is larger following the negative voltage pulse, which cannot be
explained assuming a simple ferroelectric polarization model. An equal behav-
ior has already been reported for ferroelectric field effect transistors and was
assigned to interface defects between the channel and the ferroelectric [292] . The
mechanism involved in the formation of these interface defects will have to be
studied in detail, before a commercial application of the device becomes an op-
tion. Nevertheless, as the process is reversible and reproducible, the structure
forms a transparent nonvolatile memory element, allowing a resistive readout
process. Such a method is always advantageous compared to capacitive read-
out, as the information stored on the ferroelectric is not destroyed, and will
not have to be rewritten.

6.4 Summary and conclusion


The results of this chapter can be summarized as follows:

ZnO/BTO heterojunctions have been successful synthesized on both Si


and STO substrates. The structural and electrical properties of the struc-
tures grown on the lattice matched STO substrates are by far superior
to that of the samples grown on Si.

From the IV curves measured on ZnO/BTO heterostructures grown on


STO the polarization of the ZnO layer PZnO = 0.2 0.05 C and the
coercive field of the BTO EC = 12.5 1 kV/cm were estimated.

A detailed model was developed in close cooperation with Venkata M.

147
6. Polarization related effects in ZnO/BaTiO3 heterostructures

Voora and Prof. Dr. Mathias Schubert (University of Nebraska Lincoln)


to understand the coupling of the ZnO and BTO polarization and the
formation of depletion layers in the ZnO introduced by the sheet charge
related to the offset of the polarization at a ZnO/BTO heterointerface.

Employing the aforementioned model, the SawyerTower loops of sin-


gle BTO layers, ZnO/BTO single heterojunctions and ZnO/BTO double
heterojunctions could be accurately described, and the material parame-
ters of the individual layer were extracted. Qualitative description of the
CV loops of ZnO/BTO double was also possible from the model.

Ferroelectric thin film transistors have been demonstrated on the basis of


ZnO/BTO heterostructures. The transistors were normally-on and show
output currents of up to 10 A, which could be controlled over more than
6 orders of magnitude by the gate voltage. A reproducible nonvolatile
memory effect was observed due to the ferroelectric polarization in the
BTO.

148
Chapter 7

Summary and outlook

7.1 Summary and conclusion


The results of the present study can be summarized as follows:
The lattice constants a and c of Mgx Zn1x O systematically change with
the Mg content x in relaxed Mgx Zn1x O layers, where a increases with
x and c decreases with x.
It was shown that Mgx Zn1x O can be grown pseudomorphically on ZnO.
The c-axis lattice constant changes systematically with x. The linear
relationship cps = c0 + kc,ps x with kc,ps = 0.064 was found from data
obtained in the present study and in accordance with the literature [96,97] .
The mechanisms involved in the growth of ZnO on sapphire were investi-
gated, a transition from initial 3D growth to 2D growth was revealed. A
steady state is reached after roughly 100 nm film thickness, which allows
determination of the growth rate.
The observation given in Ref. [81] that the morphology of Mgx Zn1x O
layers improves upon the introduction of a ZnO buffer layer was con-
firmed, and understood within the argumentation given in Refs [122124] .
The Mgx Zn1x O layer grown on the ZnO buffer is compressively strained,
which increases the surface diffusion length and therefore promotes step
flow growth.
The state of the art ZnO substrate material available was reviewed, and
the individual advantages discussed. Only hydrothermally grown mate-
rial proved applicable in the light of transport investigations. A thermal
pretreatment of the substrates necessary to produce surfaces suitable for
epitaxy was discussed [146] .
The mechanisms involved in the homoepitaxial growth of ZnO were in-
vestigated. Within the temperature range available layer by layer growth
was observed rather then step flow growth. A significant reduction of the
defect density was observed in comparison to ZnO grown on sapphire
substrates.

149
7. Summary and outlook

The mechanisms involved in the growth of Mgx Zn1x O on ZnO substrates


were investigated. It was shown, that the temperature allowing step flow
growth was lower than for pure ZnO. This was again explained by the
presence of compressive strain in the films, in accordance with the argu-
mentation given in Refs. [122124] . Perfect vicinal surfaces were prepared,
depending on the miscut of the substrate.

A substrate preparation routine following Ref. [174,175] was applied to com-


mercially available STO substrates and produced perfect vicinal sub-
strates with step edges of unit cell step height.

Vicinal BTO thin films with step edges of unit cell step height were grown
up to a critical thickness of roughly 45 nm in concordance with Ref. [176] .

Vicinal SRO thin films with step edges of unit cell step height were grown,
however in the films investigated holes parallel to the steps of the sub-
strate are visible. An improvement of the growth of this material will be
necessary in the future.

The properties of undoped and phosphorous doped ZnO and Mgx Zn1x O
thin films grown on ZnO are strongly influenced by the oxygen partial
pressure during growth. The oxygen partial pressure influences the grow-
ing films in more than one way.

1. The IIVI ratio in the films is controlled by the oxygen partial pres-
sure. For films grown at low oxygen partial pressures an increased
c-axis lattice constant is observed for both the ZnO and Mgx Zn1x O
thin films, most likely due to an increasing density of charged de-
fects. The thesis of an increase of the density of intrinsic defects
related to oxygen deficiency in the films is supported by an increase
in the zinc interstitial related luminescence and the free electron
concentration upon decrease of the oxygen partial pressure. It is
concluded, that oxygen in the background gas atmosphere is neces-
sary in order to grow stoichiometric ZnO and Mgx Zn1x O film.
2. The composition of the group II constituents depends on the back-
ground pressure in the chamber. The concentration of the lighter
Mg atoms compared to Zn atoms is reduced by almost a factor of
three within the range of oxygen partial pressures available in the
present growth setup. Interestingly the concentration of P does
change accordingly. It is concluded, that the atomic mass (which
differs between Mg and P by only 20%) of the individual element
influences the kinetic scattering processes with the background gas
and therefore the composition of the plasma at the substrate posi-
tion.

The similarities of the incorporation of Mg and P indicate that P is like


Mg mainly introduced into the ZnO host lattice on Zn site, in accordance
with theoretical calculations [238] .

150
7.1. Summary and conclusion

A degenerate contribution to the sample conduction was found in ho-


moepitaxial samples with a phosphorous concentration with 0.1% and
above, which could partially be removed by annealing in oxygen, indicat-
ing that at least in this case the origin of the degenerate layer is a layer
of intrinsic defects, likely to be formed at the surface.

It was shown that the separation of the conduction of the degenerate


layer and the bulk of the film was partially possible employing a correc-
tion scheme as proposed in Refs. [223,224] . A complete separation however
was not possible without further assumptions, as the incorporation of
the defects responsible for the formation of the degenerate layer is not
homogeneous.

It was shown that the separation of the conduction of individual con-


ductive paths is possible without any assumption about the distribution
of the individual carriers from magnetic field dependent measurements
by the maximum entropy mobility spectrum approach. The individual
properties of bulk conduction and the conduction of the degenerate layer
were revealed over a wide range of temperatures. Such a detailed study
of the parallel conduction processes in ZnO has not been reported so far.

The critical donor concentration and activation energy in the dilute limit
of the dominant donor in phosphorous doped ZnO and Mgx Zn1x O layers
grown on ZnO was determined to be of NC = 8.3 1018 cm3 and ED0 =
56 meV, respectively.

The thermal activation of phosphorous related acceptor was studied in


ZnO and Mgx Zn1x O thin films grown on ZnO. A transition to p-type
conduction was never observed. However an increase of the resistivity by
more than 6 orders of magnitude was observed in the Mgx Zn1x O layers
deposited at the highest oxygen partial pressures. It is concluded that
high oxygen partial pressures are necessary if p-type conduction is to be
achieved, as the compensation of acceptors is reduced by the low density
of intrinsic donors.

2D growth in the BurtonCabreraFrank mode has been observed for the


ZnO/Mgx Zn1x O single heterostructures and ZnO/Mgx Zn1x O quantum
wells grown on a-sapphire. The effective surface roughness in these sam-
ples is of the order of the thickness of the qunatum wells present in the
structures.

A significant improvement of the structural and morphological properties


of the ZnO/Mgx Zn1x O quantum wells grown on ZnO is observed. Step
flow growth leading to perfect vicinal surfaces could be achieved for these
samples as confirmed by atomic force microscopy. The Mgx Zn1x O lay-
ers were confirmed to be grown pseudomorphic by high-resolution X-ray
diffraction.

151
7. Summary and outlook

Electron accumulation layers have been observed for ZnO/Mgx Zn1x O


single heterostructures grown on sapphire by CapacitanceVoltage spec-
troscopy and temperature dependent Hall-effect measurements.
Indications for the presence of a 2DEG in ZnO/Mgx Zn1x O quantum
wells grown on ZnO were found by temperature dependent Hall-effect
measurements. For a clear observation of the transport properties of the
2DEG further improvements is still necessary. The contributions of the
degenerate layer have to be removed and the conductivity of the bulk
material has to be reduced.
Quantum confinement effects have been confirmed previously by photo-
luminescence PL measurements [258] . A narrowing of the PL peak width
is observed in qunatum wells grown on ZnO, especially for low well width,
due to the higher influence of the interface roughness on thinner quantum
wells.
Careful control of the laser fluence applied to the PLD target resulted
in a significant improvement of the quality of the heterointerface in
ZnO/Mgx Zn1x O heterostructures. A pronounced QCSE was observed
in PLD grown samples for the first time. Comparison with data obtained
on MBE grown samples given in the literature revealed, that the samples
discussed here equal the MBE grown samples both in interface quality
and in density of non-radiative recombination centers.
The use of a slanted QW allowed the observation of quantum confine-
ment effects in QW of different thicknesses in one sample with a well
defined barrier composition. However, uncertainties in the thickness of
the determination are present. The use of an in-situ RHEED observation
would help to determine and control the thickness of individual layers.
ZnO/BTO heterojunctions have been successful synthesized on both Si
and STO substrates. The structural and electrical properties of the struc-
tures grown on the lattice matched STO substrates are by far superior
to that of the samples grown on Si.
From the IV curves measured on ZnO/BTO heterostructures grown on
STO the polarization of the ZnO layer PZnO = 0.2 0.05 C and the
coercive field of the BTO EC = 12.5 1 kV/cm were estimated.
A detailed model was developed to understand the coupling of the ZnO
and BTO polarization and the formation of depletion layers in the ZnO
introduced by the sheet charge related to the offset of the polarization at
a ZnO/BTO heterointerface.
Employing the beforementioned model the SawyerTower loops of sin-
gle BTO layers, ZnO/BTO single heterojunctions and ZnO/BTO double
heterojunctions could be accurately described, and the material parame-
ters of the individual layer were extracted. Qualitative description of the
CV loops of ZnO/BTO double was also possible from the model.

152
7.2. Outlook

Ferroelectric thin film transistors have been demonstrated on the basis of


ZnO/BTO heterostructures. The transistors were normally-on and show
output currents of up to 10 A, which could be controlled over more than
6 orders of magnitude by the gate voltage. A reproducible nonvolatile
memory effect was observed due to the ferroelectric polarization in the
BTO.

7.2 Outlook
From the present work further questions arise which should be addressed in
the future.
The versatility of the setup could be improved if the influences of the
background gas on the sample composition could be separated. Two
constructive changes of the setup would allow at least a partial realization
of this goal.
1. The use of a non uniform gas atmosphere.
The application of a gas mixture, for example oxygen and an inert
gas, inside the growth chamber will allow a control of the compo-
sition of the group II constituent without changing the IIIV ratio,
therefore maintaining stoichiometric growth. However, a minimum
oxygen partial pressure is necessary in order to maintain stoichiom-
etry. This limits the total pressure attainable to values above the
oxygen partial pressure.
2. The introduction of a plasma source.
During laser plasma formation only a small fraction of the available
oxygen molecules is cracked. It is reasonable to assume in the light
of literature work [240] , that molecular oxygen is only responsible for
a vanishing fraction of the total oxygen incorporation. The range
of available gas mixtures in the growth chamber can be further ex-
panded, if the availability of ionic oxygen species can be controlled
without changing the base chamber pressure. Such an approach
could allow a chamber pressure virtually below the oxygen partial
pressure, as the availability of activated oxygen increases. Addi-
tionally, it is found in the literature [241] that the incorporation of
nitrogen is promoted by the activation in a plasma environment,
eventually leading to p-type conductivity [66,242] .
3. The use of an in-situ RHEED analysis A LMBE chamber has
been set up recently by members the semiconductor physics group.
This device will become a very valuable tool for the analysis of
growth phenomena and the very precise evaluation of layer thick-
nesses during the growth process.
4. From the chemical analysis of the sample composition it is con-
cluded that concentrated efforts are necessary to improve the purity
of the target material, as the purity of the films can not exceed the

153
7. Summary and outlook

purity of the starting material. New zirconia and tungsten carbide


mills have become available only recently. A study on the impurity
introduction in the homogenization process is motivated.

A degenerate layer is often seen in ZnO and Mgx Zn1x O samples in this
study. While the properties of such layer were investigated in transport
measurements in great detail, the microscopical origin remains unclear.
Further investigations are necessary, regarding the localization of the
degenerate layer. Chemical investigation of the substrate surfaces could
help to understand if the substrate introduces a high density of point
defects at the film substrate interface as speculated. A plasma treatment
of the substrate could help to remove impurities creating these defects.

Removal of the degenerate layer would allow more detailed measurements


on the 2DEG found in several samples. In that case low temperature
studies in high magnetic fields would be interesting, in order to observe
quantum transport phenomena.

In principle, the carrier dynamics can be understood from the present


data on the QCSE. Sophisticated models are required to include the in-
plane motion of the free carriers and excitons, exciton formation and
radiative and non radiative recombination rates. A detailed investiga-
tion of the luminescence energy versus time behavior is highly helpful to
understand the recombination dynamics completely. Such data can be
obtained employing a streak camera, which allows the observation of the
time dependence of several energetic channels simultaneously. Addition-
ally, the spatial dependence of the luminescence should be investigated,
possibly by using a micro-luminescence setup as described in [275] or by
observation of spatially masked samples, as described in Ref. [276] .

More detailed studies on the growth of SRO on STO will be necessary


in order to produce perfect vicinal surfaces as a starting point for BTO
growth.

The use of STO/BTO superlattices is a promising approach in order to


increase the effective critical thickness of BTO on STO and to enhance
certain material properties, as the dielectric constant [302] and the ferro-
electric polarization.

The mechanisms involved in the growth of ZnO on BTO should be studied


in more detail. Especially the polarity of the ZnO layers is of interest here,
and CBED measurements are perfomed at present. Also the formation
of interface defects is of high interest in order to improve the properties
of nonvolatile memory elements based on ferroelectric ZnO/BTO FETs.

154
Bibliography

[1] L. Brstein, Numerical Data and Functional Relationships in Science and


Technology, chap. III/17b, 49pp, (Springer Verlag) (1990).
[2] N. E. Eberly, C. V. Gross, and W. S. Crowell, J. Am. Chem. Soc. 42,
1433 (1920).
[3] F. Skaupy, Z. Phys. A 1, 259 (1920).
[4] W. Jander and W. Stamm, Z. Anorg. Allg. Chem. 199, 165 (1931).
[5] O. Fritsch, Ann. Phys. 414, 375 (1935).
[6] S. E. Harrison, Phys. Rev. 93, 52 (1954).
[7] A. R. Hutson, Phys. Rev. 108, 222 (1957).
[8] A. Hutson, J. Phys. Chem. Solids 8, 467 (1959).
[9] G. Heiland, Z. Phys. A 138, 459 (1954).
[10] G. Heiland and P. Kunstmann, Surf. Sci. 13, 72 (1969).
[11] A. Dal Corso, M. Posternak, R. Resta, and A. Baldereschi, Phys. Rev. B
50, 10715 (1994).
[12] D. Eger, A. Many, and Y. Goldstein, Surf. Sci. 58, 18 (1976).
[13] A. Malashevich and D. Vanderbilt, Phys. Rev. B 75, 045106 (2007).
[14] Y. Noel, C. M. Zicovich-Wilson, B. Civalleri, P. DArco, and R. Dovesi,
Phys. Rev. B 65, 014111 (2001).
[15] P. Gopal and N. Spaldin, J. Electron. Mater. 35, 538 (2006).
[16] Y.-I. Kim, K. Page, and R. Seshadri, Appl. Phys. Lett. 90, 101904 (2007).
[17] J. Jerphagnon and H. W. Newkirk, Appl. Phys. Lett. 18, 245 (1971).
[18] S. Lautenschlaeger, J. Sann, N. Volbers, B. K. Meyer, A. Hoffmann,
U. Haboeck, and M. R. Wagner, Phys. Rev. B 77, 144108 (2008).
[19] M. W. Allen, S. M. Durbin, and J. B. Metson, Appl. Phys. Lett. 91,
053512 (2007).
[20] M. W. Allen, P. Miller, R. J. Reeves, and S. M. Durbin, Appl. Phys. Lett.
90, 062104 (2007).
[21] Y. Dong, Z.-Q. Fang, D. C. Look, G. Cantwell, J. Zhang, J. J. Song, and
L. J. Brillson, Appl. Phys. Lett. 93, 072111 (2008).
[22] R. Y. Gunji, M. Nakano, A. Tsukazaki, A. Ohtomo, T. Fukumura, and

155
Bibliography

M. Kawasaki, Appl. Phys. Lett. 93, 012104 (2008).


[23] D. Kohl, M. Henzler, and G. Heiland, Surf. Sci. 41, 403 (1974).
[24] O. Dulub, U. Diebold, and G. Kresse, Phys. Rev. Lett. 90, 016102 (2003).
[25] G. Heiland, P. Kunstmann, and H. Pfister, Z. Phys. A 176, 485 (1963).
[26] H. Tampo, P. Fons, A. Yamada, K.-K. Kim, H. Shibata, K. Matsubara,
S. Niki, H. Yoshikawa, and H. Kanie, Appl. Phys. Lett. 87, 141904 (2005).
[27] R. L. Barns, E. T. Keve, and S. C. Abrahams, J. Appl. Crystallogr. 3,
27 (1970).
[28] A. N. Mariano and R. E. Hanneman, J. Appl. Phys. 34, 384 (1963).
[29] S. Kazuta, Y. Cho, H. Odagawa, and M. Kadota, Jpn. J. Appl. Phys.
39, 3121 (2000).
[30] M. Sambi, G. Granozzi, G. A. Rizzi, M. Casarin, and E. Tondello, Surf.
Sci. 319, 149 (1994).
[31] M. Galeotti, A. Atrei, U. Bardi, G. Rovida, M. Torrini, E. Zanazzi,
A. Santucci, and A. Klimov, Chem. Phys. Lett. 222, 349 (1994).
[32] T. Ohnishi, A. Ohtomo, I. Ohkuboa, M. Kawasaki, M. Yoshimoto, and
H. Koinuma, Mater. Sci. Eng., B 56, 256 (1998).
[33] S.-K. Hong, T. Hanada, H.-J. Ko, Y. Chen, T. Yao, D. Imai, K. Araki,
and M. Shinohara, Appl. Phys. Lett. 77, 3571 (2000).
[34] I. Ohkubo, A. Ohtomo, T. Ohnishi, Y. Mastumoto, H. Koinuma, and
M. Kawasaki, Surf. Sci. 443, L1043 (1999).
[35] A. Tsukazaki, A. Ohtomo, T. Kita, Y. Ohno, H. Ohno, and M. Kawasaki,
Science 315, 1388 (2007).
[36] M. Yano, K. Hashimoto, K. Fujimoto, K. Koike, S. Sasa, M. Inoue,
Y. Uetsuji, T. Ohnishi, and K. Inaba, J. Cryst. Growth 301, 353 (2007).
[37] H. Tampo, H. Shibata, K. Maejima, A. Yamada, K. Matsubara, P. Fons,
S. Kashiwaya, S. Niki, Y. Chiba, T. Wakamatsu, and H. Kanie, Appl.
Phys. Lett. 93, 202104 (2008).
[38] T. Bretagnon, P. Lefebvre, T. Guillet, T. Taliercio, B. Gil, and
C. Morhain, Appl. Phys. Lett. 90, 201912 (2007).
[39] O. Ambacher, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu, M. Mur-
phy, W. J. Schaff, L. F. Eastman, R. Dimitrov, L. Wittmer, M. Stutz-
mann, W. Rieger, and J. Hilsenbeck, J. Appl. Phys. 85, 3222 (1999).
[40] O. Ambacher, B. Foutz, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu,
M. Murphy, A. J. Sierakowski, W. J. Schaff, L. F. Eastman, R. Dimitrov,
A. Mitchell, and M. Stutzmann, J. Appl. Phys. 87, 334 (2000).
[41] H. L. Strmer, R. Dingle, A. C. Gossard, W. Wiegmann, and M. D.
Sturge, Solid State Commun. 29, 705 (1979).
[42] J. J. M. Law, E. T. Yu, B. A. Haskell, P. T. Fini, S. Nakamura, J. S.

156
Bibliography

Speck, and S. P. DenBaars, J. Appl. Phys. 103, 014305 (2008).


[43] B. Bastek, F. Bertram, J. Christen, T. Wernicke, M. Weyers, and
M. Kneissl, Appl. Phys. Lett. 92, 212111 (2008).
[44] J. Ziga Prez, V. Muoz Sanjos, E. Palacios-Lidn, and J. Colchero,
Phys. Rev. Lett. 95, 226105 (2005).
[45] J. Ziga Prez, V. Muoz Sanjos, E. Palacios-Lidn, and J. Colchero,
Appl. Phys. Lett. 88, 261912 (2006).
[46] M. Lorenz, Gepulste Laser-Plasmaabscheidung (PLD) von oxidischen
Dnnfilm- und Nanostrukturen, Habilitationsschrift, Universitt Leipzig
(2008).
[47] H. von Wenckstern, Doping, contacting, defect levels and transport prop-
erties of ZnO, Dissertationsschrift, Universitt Leipzig (2008).
[48] R. K. Singh and J. Narayan, Phys. Rev. B 41, 8843 (1990).
[49] P. R. Willmott and J. R. Huber, Rev. Mod. Phys. 72, 315 (2000).
[50] R. W. Dreyfus, J. Appl. Phys. 69, 1721 (1991).
[51] B. R. Nag and G. M. Dutta, Phys. Status Solidi B 71, 401 (1975).
[52] B. R. Nag, J. Phys. C 7, 3541 (1974).
[53] K. Seeger, Semiconductor Physics, An Introduction, Springer Verlag, 9th
edn. (2004).
[54] H. Brooks and Herring, Phys. Rev. 83, 879 (1951).
[55] A. Mller, Optische Untersuchungen an Mgx Zn1x O-Schichten im UV-
Bereich, Diplomarbeit, Universitt Leipzig, (2007).
[56] M. Stlzel, Zeitaufgelste Photolumineszenz an ZnO Strukturen, Diplo-
marbeit (2009).
[57] H. von Wenckstern, G. Biehne, R. A. Rahman, H. Hochmuth, M. Lorenz,
and M. Grundmann, Appl. Phys. Lett. 88, 092102 (2006).
[58] A. Lajn, H. v. Wenckstern, Z. Zhang, C. Czekalla, G. Biehne, J. Lenzner,
H. Hochmuth, M. Lorenz, M. Grundmann, S. Wickert, C. Vogt, and
R. Denecke, J. Vac. Sci. Technol. B 27, 1769 (2009).
[59] H. von Wenckstern, S. Mller, G. Biehne, H. Hochmuth, M. Lorenz, and
M. Grundmann, J. Electron. Mater. 39, 559 (2010).
[60] S. M. Sze, Physics Of Semiconductor Devices, Wiley - Interscience, 1st
edn. (1969).
[61] H. Kroemer, W.-Y. Chien, J. J. S. Harris, and D. D. Edwall, Appl. Phys.
Lett. 36, 295 (1980).
[62] P. Blood and J. W. Orton, The electrical characterization of semiconduc-
tors: majority carriers and electron states., Academic Press, Inc. (1992).
[63] C. B. Sawyer and C. H. Tower, Phys. Rev. 35, 269 (1930).
[64]

157
Bibliography

[65] M. Nakano, T. Makino, A. Tsukazaki, K. Ueno, A. Ohtomo, T. Fuku-


mura, H. Yuji, S. Akasaka, K. Tamura, K. Nakahara, T. Tanabe,
A. Kamisawa, and M. Kawasaki, Appl. Phys. Lett. 93, 123309 (2008).
[66] A. Tsukazaki, A. Ohtomo, T. Onuma, M. Ohtani, T. Makino, M. Sumiya,
K. Ohtani, S. F. Chichibu, S. Fuke, Y. Segawa, H. Ohno, H. Koinuma,
and M. Kawasaki, Nat. Mater. 4, 42 (2005).
[67] D. M. Bagnall, Y. F. Chen, Z. Zhu, T. Yao, S. Koyama, M. Y. Shen, and
T. Goto, Appl. Phys. Lett. 70, 2230 (1997).
[68] Y. R. Ryu, T. S. Lee, J. A. Lubguban, A. B. Corman, H. W. White,
J. H. Leem, M. S. Han, Y. S. Park, C. J. Youn, and W. J. Kim, Appl.
Phys. Lett. 88, 052103 (2006).
[69] J. Dong, A. Osinsky, B. Hertog, A. Dabiran, P. Chow, Y. Heo, D. Norton,
and S. Pearton, J. Electron. Mater. 34, 416 (2005).
[70] B. K. Meyer, A. Polity, B. Farangis, Y. He, D. Hasselkamp, T. Krmer,
and C. Wang, Appl. Phys. Lett. 85, 4929 (2004).
[71] R. Schmidt, C. Bundesmann, N. Ashkenov, B. Rheinlnder, M. Schubert,
M. Lorenz, E. M. Kaidashev, D. Spemann, T. Butz, J. Lenzner, and
M. Grundmann, Institute of Physics Conference Series 171, 11 (2002).
[72] M. Lorenz, E. M. Kaidashev, H. von Wenckstern, V. Riede, C. Bun-
desmann, D. Spemann, G. Benndorf, H. Hochmuth, A. Rahm, H. C.
Semmelhack, and M. Grundmann, Solid-State Electron. 47, 2205 (2003).
[73] C. Yang, X. M. Li, Y. F. Gu, W. D. Yu, X. D. Gao, and Y. W. Zhang,
Appl. Phys. Lett. 93, 112114 (2008).
[74] T. Makino, Y. Segawa, M. Kawasaki, A. Ohtomo, R. Shiroki, K. Tamura,
T. Yasuda, and H. Koinuma, Appl. Phys. Lett. 78, 1237 (2001).
[75] C. Bundesmann, M. Schubert, A. Rahm, D. Spemann, H. Hochmuth,
M. Lorenz, and M. Grundmann, Appl. Phys. Lett. 85, 905 (2004).
[76] M. Lorenz, E. M. Kaidashev, H. von Wenckstern, V. Riede, C. Bun-
desmann, D. Spemann, G. Benndorf, H. Hochmuth, A. Rahm, H. C.
Semmelhack, and M. Grundmann, Solid-State Electron. 47, 2205 (2003).
[77] R. Schmidt-Grund, M. Schubert, B. Rheinlnder, D. Fritsch, H. Schmidt,
E. M. Kaidashev, M. Lorenz, C. M. Herzinger, and M. Grundmann, Thin
Solid Films 455-456, 500 (2004).
[78] S. Heitsch, G. Benndorf, G. Zimmermann, C. Schulz, D. Spemann,
H. Hochmuth, H. Schmidt, T. Nobis, M. Lorenz, and M. Grundmann,
Appl. Phys. A 88, 99 (2007).
[79] C. P. Dietrich, Strukturelle und Optische Eigenschaften von Magnesium-
zinkoxid-Dnnfilmen, Diplomarbeit, Universitt Leipzig (2009).
[80] S. Heitsch, G. Zimmermann, J. Lenzner, H. Hochmuth, G. Benndorf,
M. Lorenz, and M. Grundmann, AIP Conference Proceedings 893, 409
(2007).

158
Bibliography

[81] S. Heitsch, Optimierung der PLD-Abscheidung und Lumineszenzunter-


suchungen an Mgx Zn1x O-Dnnfilmen und -Quantengrabenstrukturen,
Dissertationsschrift, Universitt Leipzig (2008).
[82] S. Heitsch, G. Zimmermann, D. Fritsch, C. Sturm, R. Schmidt-Grund,
C. Schulz, H. Hochmuth, D. Spemann, G. Benndorf, B. Rheinlnder,
T. Nobis, M. Lorenz, and M. Grundmann, J. Appl. Phys. 101, 083521
(2007).
[83] W. I. Park, G.-C. Yi, and H. M. Jang, Appl. Phys. Lett. 79, 2022 (2001).
[84] K. Koike, K. Hama, I. Nakashima, G. Takada, K. Ogata, S. Sasa, M. In-
oue, and M. Yano, J. Cryst. Growth 278, 288 (2005).
[85] Y. Matsumoto, M. Murakami, Z. Jin, A. Ohtomo, M. Lippmaa,
M. Kawasaki, and H. Koinuma, Jpn. J. Appl. Phys. 38, L603 (1999).
[86] A. Rahm, High Resolution X-ray Diffraction of ZnO Based Thin Films,
Diplomarbeit, Universitt Leipzig (2003).
[87] N. Ashkenov, B. N. Mbenkum, C. Bundesmann, V. Riede, M. Lorenz,
D. Spemann, E. M. Kaidashev, A. Kasic, M. Schubert, M. Grundmann,
G. Wagner, H. Neumann, V. Darakchieva, H. Arwin, and B. Monemar,
J. Appl. Phys. 93, 126 (2003).
[88] D. Ehrentraut, H. Sato, Y. Kagamitani, H. Sato, A. Yoshikawa, and
T. Fukuda, Prog. Cryst. Growth Charact. Mater. 52, 280 (2006).
[89] T. Sekiguchi, S. Miyashita, K. Obara, T. Shishido, and N. Sakagami, J.
Cryst. Growth 214-215, 72 (2000).
[90] G. Dhanaraj, M. Dudley, D. Bliss, M. Callahan, and M. Harris, J. Cryst.
Growth 297, 74 (2006).
[91] A. Ohtomo, M. Kawasaki, I. Ohkubo, H. Koinuma, T. Yasuda, and
Y. Segawa, Appl. Phys. Lett. 75, 980 (1999).
[92] A. Ohtomo, M. Kawasaki, T. Koida, K. Masubuchi, H. Koinuma,
Y. Sakurai, Y. Yoshida, T. Yasuda, and Y. Segawa, Appl. Phys. Lett.
72, 2466 (1998).
[93] A. Ohtomo and A. Tsukazaki, Semicond. Sci. Technol 20, S1 (2005).
[94] M. Yano, K. Ogata, F. Yan, K. Koike, S. Sasa, and M. Inoue, MRS Proc.
744 (2003).
[95] S. Sadofev, S. Blumstengel, J. Cui, J. Puls, S. Rogaschewski, P. Schfer,
Y. G. Sadofyev, and F. Henneberger, Appl. Phys. Lett. 87, 091903 (2005).
[96] H. Matsui, H. Tabata, N. Hasuike, and H. Harima, J. Appl. Phys. 99,
024902 (2006).
[97] Y. Nishimoto, K. Nakahara, D. Takamizu, A. Sasaki, K. Tamura,
S. Akasaka, H. Yuji, T. Fujii, T. Tanabe, H. Takasu, A. Tsukazaki,
A. Ohtomo, T. Onuma, S. F. Chichibu, and M. Kawasaki, Appl. Phys.
Express 1, 091202 (2008).

159
Bibliography

[98] M. Grundmann, The Physics of Semiconductors An Introduction In-


cluding Devices and Nanophysics, Springer, Berlin, Heidelberg, New York
(2006).
[99] I. Kobiakov, Solid State Commun. 35, 305 (1980).
[100] J. Matthews and A. Blakeslee, J. Cryst. Growth 27, 118 (1974).
[101] H. von Wenckstern, M. Brandt, H. Schmidt, G. Biehne, R. Pickenhain,
H. Hochmuth, M. Lorenz, and M. Grundmann, Appl. Phys. A 88, 135
(2007).
[102] B. M. Ataev, W. V. Lundin, V. V. Mamedov, A. M. Bagamadova, and
E. E. Zavarin, J. Phys. Condens. Matter 13, L211 (2001).
[103] KYOCERA, Single Crystal Sapphire, KYOCERA Corporation, 6 Takeda
Tobadono-cho, Fushimi-ku, Kyoto 612-8501, Japan (2007).
[104] X. H. Wu, C. R. Elsass, A. Abare, M. Mack, S. Keller, P. M. Petroff,
S. P. DenBaars, J. S. Speck, and S. J. Rosner, Appl. Phys. Lett. 72, 692
(1998).
[105] B. Heying, E. J. Tarsa, C. R. Elsass, P. Fini, S. P. DenBaars, and J. S.
Speck, J. Appl. Phys. 85, 6470 (1999).
[106] W. K. Burton, N. Cabrera, and F. C. Frank, Philos. Trans. R. Soc.
London, Ser. A 243, 299 (1951).
[107] F. C. Frank, Acta Crystallogr. 4, 497 (1951).
[108] E. J. Tarsa, B. Heying, X. H. Wu, P. Fini, S. P. DenBaars, and J. S.
Speck, J. Appl. Phys. 82, 5472 (1997).
[109] S. Vzian, J. Massies, F. Semond, N. Grandjean, and P. Venngus, Phys.
Rev. B 61, 7618 (2000).
[110] B. Liu, S. R. Leone, T. Kitajima, T. H. Zhang, and C. Borca, J. Appl.
Phys. 97, 023509 (2005).
[111] J. E. Northrup, R. Di Felice, and J. Neugebauer, Phys. Rev. B 56, R4325
(1997).
[112] W. Qian, G. S. Rohrer, M. Skowronski, K. Doverspike, L. B. Rowland,
and D. K. Gaskill, Appl. Phys. Lett. 67, 2284 (1995).
[113] D. Cherns, W. T. Young, J. W. Steeds, F. A. Ponce, and S. Nakamura,
J. Cryst. Growth 178, 201 (1997).
[114] Y. Cui and L. Li, Phys. Rev. B 66, 155330 (2002).
[115] A. Redinger, O. Ricken, P. Kuhn, A. Rtz, A. Voigt, J. Krug, and
T. Michely, Phys. Rev. Lett. 100, 035506 (2008).
[116] M. Lorenz, J. Lenzner, E. Kaidashev, H. Hochmuth, and M. Grundmann,
Ann. Phys. 13, 39 (2004).
[117] U. Diebold, L. V. Koplitz, and O. Dulub, Appl. Surf. Sci. 237, 336
(2004).

160
Bibliography

[118] W. Hong, H. N. Lee, M. Yoon, H. M. Christen, D. H. Lowndes, Z. Suo,


and Z. Zhang, Phys. Rev. Lett. 95, 095501 (2005).
[119] Y. Cui and L. Li, Phys. Status Solidi A 188, 583 (2001).
[120] C. Jin and R. Narayan, J. Electron. Mater. 35, 869 (2006).
[121] M. H. Xie, S. M. Seutter, W. K. Zhu, L. X. Zheng, H. Wu, and S. Y.
Tong, Phys. Rev. Lett. 82, 2749 (1999).
[122] H. Brune, K. Bromann, H. Rder, K. Kern, J. Jacobsen, P. Stoltze,
K. Jacobsen, and J. No/rskov, Phys. Rev. B 52, R14380 (1995).
[123] C. Ratsch and A. Zangwill, Appl. Phys. Lett. 63, 2348 (1993).
[124] M. Schroeder and D. E. Wolf, Surf. Sci. 375, 129 (1997).
[125] Eagle-Picher Technologies L.L.C., 200 B.J. Tunnell Blvd., Miami, OK
74354, USA.
[126] Zn Technology Inc, 910 Columbia, St. Brea, CA , U.S.A.
[127] Cermet Inc., 1019 Collier Rd, Atlanta, GA, U.S.A.
[128] J. Nause and B. Nemeth, Semicond. Sci. Technol 20, S45 (2005).
[129] D. Schulz, S. Ganschow, D. Klimm, and K. Struve, J. Cryst. Growth
310, 1832 (2008).
[130] D. Schulz, S. Ganschow, D. Klimm, M. Neubert, M. Roberg, M. Schmid-
bauer, and R. Fornari, J. Cryst. Growth 296, 27 (2006).
[131] D. Klimm, S. Ganschow, D. Schulz, and R. Fornari, J. Cryst. Growth
310, 3009 (2008).
[132] MaTecK GmbH, Im Langenbroich 20, D-52428 Juelich, Germany.
[133] Crystec GmbH, Kpenicker Str. 325, D-12555 Berlin, Germany.
[134] TOKYO DENPA CO. LTD., 5-6-11 Chuo, Ohta-ku, Tokyo 143-0024,
Japan.
[135] L. N. Demianets, D. V. Kostomarov, I. P. Kuzmina, and S. V. Pushko,
Crystallogr. Rep. 47, S86 (2002).
[136] L. N. Demyanets, N. L. Sizova, and L. E. Li, Crystallogr. Rep. 50, 646
(2005).
[137] S. Sadofev, P. Schfer, Y.-H. Fan, S. Blumstengel, F. Henneberger,
D. Schulz, and D. Klimm, Appl. Phys. Lett. 91, 201923 (2007).
[138] K. Maeda, M. Sato, I. Niikura, and T. Fukuda, Semicond. Sci. Technol
20, S49 (2005).
[139] 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. Status Solidi B 241, 231 (2004).
[140] Y. Yan, M. M. Al-Jassim, and S.-H. Wei, Appl. Phys. Lett. 89, 181912
(2006).

161
Bibliography

[141] C. H. Park, S. B. Zhang, and S.-H. Wei, Phys. Rev. B 66, 073202 (2002).
[142] M. Cho, C. Harada, H. Suzuki, T. Minegishi, T. Yao, H. Ko, K. Maeda,
and I. Nikura, Superlattices Microstruct. 38, 349 (2005).
[143] H. Wenisch, V. Kirchner, S. K. Hong, Y. F. Chen, H. J. Ko, and T. Yao,
J. Cryst. Growth 227-228, 944 (2001).
[144] H.-J. Ko, M.-S. Han, Y.-S. Park, Y.-S. Yu, B.-I. Kim, S. S. Kim, and
J.-H. Kim, J. Cryst. Growth 269, 493 (2004).
[145] H. Matsui, H. Saeki, T. Kawai, A. Sasaki, M. Yoshimoto, M. Tsubaki,
and H. Tabata, J. Vac. Sci. Technol., B 22, 2454 (2004).
[146] H. von Wenckstern, H. Schmidt, C. Hanisch, M. Brandt, C. Czekalla,
G. Benndorf, G. Biehne, A. Rahm, H. Hochmuth, M. Lorenz, and
M. Grundmann, Phys. Status Solidi RRL 1, 129 (2007).
[147] S. Graubner, C. Neumann, N. Volbers, B. K. Meyer, J. Blsing, and
A. Krost, Appl. Phys. Lett. 90, 042103 (2007).
[148] S. Heinze, A. Krtschil, J. Blsing, T. Hempel, P. Veit, A. Dadgar,
J. Christen, and A. Krost, J. Cryst. Growth 308, 170 (2007).
[149] H. Matsui and H. Tabata, J. Appl. Phys. 99, 124307 (2006).
[150] N. Volbers, Sekundrionenmassenspektrometrie an Zinkoxid, Disserta-
tionsschrift, Justus-Liebig-Universitt Gieen (2007).
[151] M. Valtiner, M. Todorova, G. Grundmeier, and J. Neugebauer, Phys.
Rev. Lett. 103, 065502 (2009).
[152] C. Wll, Prog. Surf. Sci. 82, 55 (2007).
[153] D. C. Look, B. Claflin, and H. E. Smith, Appl. Phys. Lett. 92, 122108
(2008).
[154] D. C. Look, J. Appl. Phys. 104, 063718 (2008).
[155] Y. Wang, B. Meyer, X. Yin, M. Kunat, D. Langenberg, F. Traeger,
A. Birkner, and C. Wll, Phys. Rev. Lett. 95, 266104 (2005).
[156] G. Heiland and P. Kunstmann, Surf. Sci. 13, 72 (1969).
[157] G. Heiland, J. Phys. Chem. Solids 6, 155 (1958).
[158] H. Xu, K. Ohtani, M. Yamao, and H. Ohno, Appl. Phys. Lett. 89, 071918
(2006).
[159] M. Kubo, Y. Oumi, H. Takaba, A. Chatterjee, A. Miyamoto,
M. Kawasaki, M. Yoshimoto, and H. Koinuma, Phys. Rev. B 61, 16187
(2000).
[160] Y. Chen, H.-J. Ko, S.-K. Hong, T. Yao, and Y. Segawa, Appl. Phys. Lett.
80, 1358 (2002).
[161] H.-J. Ko, T. Yao, Y. Chen, and S.-K. Hong, J. Appl. Phys. 92, 4354
(2002).
[162] S. Schche, Kristallstruktur und dielektrische Funktion von

162
Bibliography

Bariumtitanat-Einkristallen und -Dnnfilmen, Diplomarbeit, Universitt


Leipzig (2009).
[163] N. A. Pertsev, A. G. Zembilgotov, and A. K. Tagantsev, Ferroelectrics
223, 79 (1999).
[164] A. G. Zembilgotov, N. A. Pertsev, H. Kohlstedt, and R. Waser, J. Appl.
Phys. 91, 2247 (2002).
[165] K. J. Choi, M. Biegalski, Y. L. Li, A. Sharan, J. Schubert, R. Uecker,
P. Reiche, Y. B. Chen, X. Q. Pan, V. Gopalan, L.-Q. Chen, D. G. Schlom,
and C. B. Eom, Science 306, 1005 (2004).
[166] N. Yanase, K. Abe, N. Fukushima, and T. Kawakubo, Jpn. J. Appl.
Phys. 38, 5305 (1999).
[167] Crystal GmbH, Ostendstrae 25, Haus 4, D-12459 Berlin.
[168] M. Kawasaki, K. Takahashi, T. Maeda, R. Tsuchiya, M. Shinohara,
O. Ishiyama, T. Yonezawa, M. Yoshimoto, and H. Koinuma, Science
266, 1540 (1994).
[169] M. Kawasaki, A. Ohtomo, T. Arakane, K. Takahashi, M. Yoshimoto, and
H. Koinuma, Appl. Surf. Sci. 107, 102 (1996).
[170] G. Koster, B. L. Kropman, G. J. H. M. Rijnders, D. H. A. Blank, and
H. Rogalla, Mater. Sci. Eng., B 56, 209 (1998).
[171] Y. Liang and D. A. Bonnell, Surf. Sci. 285, L510 (1993).
[172] Y. Liang and D. A. Bonnell, Surf. Sci. 310, 128 (1994).
[173] K. Szot and W. Speier, Phys. Rev. B 60, 5909 (1999).
[174] G. Koster, B. L. Kropman, G. J. H. M. Rijnders, D. H. A. Blank, and
H. Rogalla, Appl. Phys. Lett. 73, 2920 (1998).
[175] G. Koster, G. Rijnders, D. H. A. Blank, and H. Rogalla, Physica C 339,
215 (2000).
[176] A. Petraru, N. A. Pertsev, H. Kohlstedt, U. Poppe, R. Waser, A. Solbach,
and U. Klemradt, J. Appl. Phys. 101, 114106 (2007).
[177] T. Aoki, Y. Hatanaka, and D. C. Look, Appl. Phys. Lett. 76, 3257 (2000).
[178] W. Z. Xu, Z. Z. Ye, Y. J. Zeng, L. P. Zhu, B. H. Zhao, L. Jiang, J. G.
Lu, H. P. He, and S. B. Zhang, Appl. Phys. Lett. 88, 173506 (2006).
[179] A. Tsukazaki, M. Kubota, A. Ohtomo, T. Onuma, K. Ohtani, H. Ohno,
S. F. Chichibu, and M. Kawasaki, Jpn. J. Appl. Phys. 44, L643 (2005).
[180] Y. Ryu, T.-S. Lee, J. A. Lubguban, H. W. White, B.-J. Kim, Y.-S. Park,
and C.-J. Youn, Appl. Phys. Lett. 88, 241108 (2006).
[181] K.-K. Kim, H.-S. Kim, D.-K. Hwang, J.-H. Lim, and S.-J. Park, Appl.
Phys. Lett. 83, 63 (2003).
[182] Y. W. Heo, S. J. Park, K. Ip, S. J. Pearton, and D. P. Norton, Appl.
Phys. Lett. 83, 1128 (2003).

163
Bibliography

[183] Y. J. Li, Y. W. Heo, Y. Kwon, K. Ip, S. J. Pearton, and D. P. Norton,


Appl. Phys. Lett. 87, 072101 (2005).
[184] Y. W. Heo, Y. W. Kwon, Y. Li, S. J. Pearton, and D. P. Norton, Appl.
Phys. Lett. 84, 3474 (2004).
[185] H. von Wenckstern, R. Pickenhain, H. Schmidt, M. Brandt, G. Biehne,
M. Lorenz, M. Grundmann, and G. Brauer, Appl. Phys. Lett. 89, 092122
(2006).
[186] Y. R. Ryu, T. S. Lee, and H. W. White, Appl. Phys. Lett. 83, 87 (2003).
[187] W. Guo, A. Allenic, Y. B. Chen, X. Q. Pan, Y. Che, Z. D. Hu, and
B. Liu, Appl. Phys. Lett. 90, 242108 (2007).
[188] K. Ip, Y. W. Heo, D. P. Norton, S. J. Pearton, J. R. LaRoche, and
F. Ren, Appl. Phys. Lett. 85, 1169 (2004).
[189] G. Shukla, J. Phys. D 42, 075105 (6pp) (2009).
[190] A. Lajn, H. von Wenckstern, G. Benndorf, C. Dietrich, M. Brandt,
G. Biehne, H. Hochmuth, M. Lorenz, and M. Grundmann, J. Electron.
Mater. 39 (2009).
[191] A. F. Kohan, G. Ceder, D. Morgan, and C. G. Van de Walle, Phys. Rev.
B 61, 15019 (2000).
[192] F. Oba, S. R. Nishitani, S. Isotani, H. Adachi, and I. Tanaka, J. Appl.
Phys. 90, 824 (2001).
[193] A. Janotti and C. G. V. de Walle, J. Cryst. Growth 287, 58 (2006).
[194] P. Erhart, K. Albe, and A. Klein, Phys. Rev. B 73, 205203 (2006).
[195] D. C. Reynolds, C. W. Litton, and T. C. Collins, Phys. Rev. 140, A1726
(1965).
[196] B. K. Meyer, J. Sann, S. Lautenschlger, M. R. Wagner, and A. Hoff-
mann, Phys. Rev. B 76, 184120 (2007).
[197] J. Sann, J. Stehr, A. Hofstaetter, D. M. Hofmann, A. Neumann,
M. Lerch, U. Haboeck, A. Hoffmann, and C. Thomsen, Phys. Rev. B
76, 195203 (2007).
[198] M. R. Wagner, T. P. Bartel, R. Kirste, A. Hoffmann, J. Sann, S. Laut-
enschlger, B. K. Meyer, and C. Kisielowski, Phys. Rev. B 79, 035307
(2009).
[199] J. Wrzesinski and D. Frhlich, Phys. Rev. B 56, 13087 (1997).
[200] M. Brandt, Elektrische Eigenschaften von Zinkoxid bei erhhten Temper-
aturen, Diplomarbeit, Universitt Leipzig (2006).
[201] A. Rohatgi, S. K. Pang, T. K. Gupta, and W. D. Straub, J. Appl. Phys.
63, 5375 (1988).
[202] F. D. Auret, W. Meyer, P. Janse van Rensburg, M. Hayes, J. Nel, H. von
Wenckstern, H. Schmidt, G. Biehne, H. Hochmuth, M. Lorenz, and
M. Grundmann, Physica B 401-402, 378 (2007).

164
Bibliography

[203] B. Q. Cao, M. Lorenz, M. Brandt, H. von Wenckstern, J. Lenzner,


G. Biehne, and M. Grundmann, Phys. Status Solidi RRL 2, 37 (2008).
[204] B. Q. Cao, M. Lorenz, H. von Wenckstern, C. Czekalla, M. Brandt,
J. Lenzner, G. Benndorf, G. Biehne, and M. Grundmann, Proc. SPIE
6895, 68950V (2008).
[205] A. Allenic, W. Guo, Y. Chen, M. Katz, G. Zhao, Y. Che, Z. Hu, B. Liu,
S. Zhang, and X. Pan, Adv. Mater. 19, 3333 (2007).
[206] A. L. Allenic, Structural, electrical and optical properties of p-type ZnO
epitaxial films, Ph.D. thesis, The University of Michigan (2008).
[207] H. S. Kim, S. J. Pearton, D. P. Norton, and F. Ren, J. Appl. Phys. 102,
104904 (2007).
[208] M. Lorenz, B. Cao, G. Zimmermann, G. Biehne, C. Czekalla, H. Fren-
zel, M. Brandt, H. von Wenckstern, and M. Grundmann, J. Vac. Sci.
Technol. B 27, 1693 (2009).
[209] P.-J. Li, Z.-M. Liao, X.-Z. Zhang, X.-J. Zhang, H.-C. Zhu, J.-Y. Gao,
K. Laurent, Y. Leprince-Wang, N. Wang, and D.-P. Yu, Nano Lett. 9,
2513 (2009).
[210] M. Brandt, H. von Wenckstern, H. Schmidt, A. Rahm, G. Biehne,
G. Benndorf, H. Hochmuth, M. Lorenz, C. Meinecke, T. Butz, and
M. Grundmann, J. Appl. Phys. 104, 013708 (2008).
[211] H. von Wenckstern, M. Brandt, H. Schmidt, G. Benndorf, J. Zippel,
H. Hochmuth, M. Lorenz, and M. Grundmann, Proc. SPIE 6895, 689505
(2008).
[212] M. Lorenz, G. Wagner, A. Rahm, H. Schmidt, H. Hochmuth, H. Schmid,
W. Mader, M. Brandt, H. von Wenckstern, and M. Grundmann, Phys.
Status Solidi C 5, 3280 (2008).
[213] H. von Wenckstern, M. Brandt, H. Schmidt, C. Hanisch, G. Benndorf,
H. Hochmuth, M. Lorenz, and M. Grundmann, J. Korean Phys. Soc. 53,
3064 (2008).
[214] T. Reinert, Ionenstrahluntersuchungen am Gelenkknorpel, Dissertation-
sschrift, Universitt Leipzig (2001).
[215] D. Spemann, Anwendung hochenergetischer Ionenstrahlen in den Mate-
rialwissenschaften, Dissertationsschrift, Universitt Leipzig (2007).
[216] A. Allenic, X. Q. Pan, Y. Che, Z. D. Hu, and B. Liu, Appl. Phys. Lett.
92, 022107 (2008).
[217] R. Dingle, Phys. Rev. Lett. 23, 579 (1969).
[218] F. A. Krger and H. J. Vink, J. Chem. Phys. 22, 250 (1954).
[219] K. Vanheusden, W. L. Warren, C. H. Seager, D. R. Tallant, J. A. Voigt,
and B. E. Gnade, J. Appl. Phys. 79, 7983 (1996).
[220] H. von Wenckstern, G. Benndorf, S. Heitsch, J. Sann, M. Brandt,

165
Bibliography

H. Schmidt, J. Lenzner, M. Lorenz, A. Y. Kuznetsov, B. K. Meyer, and


M. Grundmann, Appl. Phys. A 88, 125 (2007).
[221] S. Lany, J. Osorio-Guilln, and A. Zunger, Phys. Rev. B 75, 241203
(2007).
[222] P. Pagliuso, E. Miranda, M. Continentino, and E. B. Saitovitch, Physica
B 404, xiv (2009).
[223] D. C. Look and R. J. Molnar, Appl. Phys. Lett. 70, 3377 (1997).
[224] D. C. Look, H. L. Mosbacker, Y. M. Strzhemechny, and L. J. Brillson,
Superlattices Microstruct. 38, 406 (2005).
[225] D. C. Look, Surf. Sci. 601, 5315 (2007).
[226] R. K. Willardson, T. C. Harman, and A. C. Beer, Phys. Rev. 96, 1512
(1954).
[227] T. Harman, J. Honig, and P. Trent, J. Phys. Chem. Solids 28, 1995
(1967).
[228] W. A. Beck and J. R. Anderson, J. Appl. Phys. 62, 541 (1987).
[229] J. W. McClure, Phys. Rev. 101, 1642 (1956).
[230] J. R. Meyer, C. A. Hoffman, J. Antoszewski, and L. Faraone, J. Appl.
Phys. 81, 709 (1997).
[231] I. Vurgaftman, J. R. Meyer, C. A. Hoffman, D. Redfern, J. Antoszewski,
L. Faraone, and J. R. Lindemuth, J. Appl. Phys. 84, 4966 (1998).
[232] S. Kiatgamolchai, M. Myronov, O. A. Mironov, V. G. Kantser, E. H. C.
Parker, and T. E. Whall, Phys. Rev. E 66, 036705 (2002).
[233] J. Rothman, J. Meilhan, G. Perrais, J.-P. Belle, and O. Gravrand, J.
Electron. Mater. 35, 1174 (2006).
[234] O. A. Mironov, M. Myronov, S. Kiatgamolchai, and V. G. Kantser, Phys.
Rev. E 69, 038702 (2004).
[235] D. Chrastina, J. P. Hague, and D. R. Leadley, J. Appl. Phys. 94, 6583
(2003).
[236] J. P. Ibbetson, P. T. Fini, K. D. Ness, S. P. DenBaars, J. S. Speck, and
U. K. Mishra, Appl. Phys. Lett. 77, 250 (2000).
[237] S. Limpijumnong, S. B. Zhang, S.-H. Wei, and C. H. Park, Phys. Rev.
Lett. 92, 155504 (2004).
[238] W.-J. Lee, J. Kang, and K. J. Chang, Phys. Rev. B 73, 024117 (2006).
[239] M. Brandt, H. von Wenckstern, C. Meinecke, T. Butz, H. Hochmuth,
M. Lorenz, and M. Grundmann, J. Vac. Sci. Technol. B 27, 1604 (2009).
[240] K. Sakurai, M. Kanehiro, K. Nakahara, T. Tanabe, S. Fujita, and S. Fu-
jita, J. Cryst. Growth 209, 522 (2000).
[241] H. Matsui, H. Saeki, T. Kawai, H. Tabata, and B. Mizobuchi, J. Appl.
Phys. 95, 5882 (2004).

166
Bibliography

[242] D. C. Look, D. C. Reynolds, C. W. Litton, R. L. Jones, D. B. Eason,


and G. Cantwell, Appl. Phys. Lett. 81, 1830 (2002).
[243] B. V. Claude Weisbuch, Quantum Semiconductors Structures, Academic
Press, Inc. (1991).
[244] A. Janotti and C. G. V. de Walle, Phys. Rev. B 75, 121201 (2007).
[245] E. O. Gbel, H. Jung, J. Kuhl, and K. Ploog, Phys. Rev. Lett. 51, 1588
(1983).
[246] G. Snider (2009), http://www.nd.edu/~gsnider/.
[247] I.-H. Tan, G. L. Snider, L. D. Chang, and E. L. Hu, J. Appl. Phys. 68,
4071 (1990).
[248] H. von Wenckstern, J. Zippel, A. Lajn, M. Brandt, G. Biehne,
H. Hochmuth, M. Lorenz, and M. Grundmann, AIP conference proceed-
ings 1199, 469 (2009).
[249] K. Koike, K. Hama, I. Nakashima, G. Takada, M. Ozaki, K. Ogata,
S. Sasa, M. Inoue, and M. Yano, Jpn. J. Appl. Phys. 43, L1372 (2004).
[250] A. Tsukazaki, A. Ohtomo, and M. Kawasaki, Appl. Phys. Lett. 88, 152106
(2006).
[251] A. Tsukazaki, A. Ohtomo, M. Nakano, and M. Kawasaki, Appl. Phys.
Lett. 92, 052105 (2008).
[252] A. Tsukazaki, H. Yuji, S. Akasaka, K. Tamura, K. Nakahara, T. Tanabe,
H. Takasu, A. Ohtomo, and M. Kawasaki, Appl. Phys. Express 1, 055004
(2008).
[253] M. Brandt, H. von Wenckstern, G. Benndorf, H. Hochmuth, M. Lorenz,
and M. Grundmann, Thin Solid Films 518, 1048 (2009).
[254] A. Kalabukhov, R. Gunnarsson, J. Borjesson, E. Olsson, D. Winkler, and
T. Claeson, J. Phys. Conf. Ser. 100, 082039 (4pp) (2008).
[255] J. A. Davis, L. V. Dao, X. Wen, C. Ticknor, P. Hannaford, V. A. Cole-
man, H. H. Tan, C. Jagadish, K. Koike, S. Sasa, M. Inoue, and M. Yano,
Nanotech. 19, 055205 (4pp) (2008).
[256] V. A. Coleman, M. Buda, H. H. Tan, C. Jagadish, M. R. Phillips,
K. Koike, S. Sasa, M. Inoue, and M. Yano, Semicond. Sci. Technol 21,
L25 (2006).
[257] J. A. Davis and C. Jagadish, Laser Photonics Rev. 3, 85 (2009).
[258] S. Heitsch, G. Zimmermann, A. Mller, J. Lenzner, H. Hochmuth,
G. Benndorf, M. Lorenz, and M. Grundmann, Proc. MRS 957, K07
(2006).
[259] B. P. Zhang, B. L. Liu, J. Z. Yu, Q. M. Wang, C. Y. Liu, Y. C. Liu, and
Y. Segawa, Appl. Phys. Lett. 90, 132113 (2007).
[260] C. Morhain, T. Bretagnon, P. Lefebvre, X. Tang, P. Valvin, T. Guillet,
B. Gil, T. Taliercio, M. Teisseire-Doninelli, B. Vinter, and C. Deparis,

167
Bibliography

Phys. Rev. B 72, 241305 (2005).


[261] T. Bretagnon, P. Lefebvre, P. Valvin, B. Gil, C. Morhain, and X. Tang,
J. Cryst. Growth 287, 12 (2006).
[262] T. Guillet, T. Bretagnon, T. Taliercio, P. Lefebvre, B. Gil, C. Morhain,
and X. Tang, Superlattices Microstruct. 41, 352 (2007).
[263] B. Gil, P. Lefebvre, T. Bretagnon, T. Guillet, J. A. Sans, T. Taliercio,
and C. Morhain, Phys. Rev. B 74, 153302 (2006).
[264] T. Makino, K. Tamura, C. H. Chia, Y. Segawa, M. Kawasaki, A. Ohtomo,
and H. Koinuma, Phys. Rev. B 66, 233305 (2002).
[265] T. Makino, K. Tamura, C. H. Chia, Y. Segawa, M. Kawasaki, A. Ohtomo,
and H. Koinuma, Appl. Phys. Lett. 81, 2355 (2002).
[266] T. Makino, Y. Segawa, M. Kawasaki, and H. Koinuma, Semicond. Sci.
Technol 20, S78 (2005).
[267] T. Makino, A. Ohtomo, C. H. Chia, Y. Segawa, H. Koinuma, and
M. Kawasaki, Physica E 21, 671 (2004).
[268] S. Kalusniak, S. Sadofev, J. Puls, and F. Henneberger, Laser Photonics
Rev. 3, 233 (2009).
[269] S. Kalusniak, S. Sadofev, J. Puls, H. J. Wnsche, and F. Henneberger,
Phys. Rev. B 77, 113312 (2008).
[270] C. P. Hilton, W. E. Hagston, and J. E. Nicholls, J. Phys. A 25, 2395
(1992).
[271] P. Hilton, J. Goodwin, P. Harrison, and W. E. Hagston, J. Phys. A 25,
5365 (1992).
[272] P. Harrison, T. Piorek, W. E. Hagston, and T. Stirner, Superlattices
Microstruct. 20, 45 (1996).
[273] P. Harrison, Quantum Wells, Wires and Dots, John Wiley & Sons, LTD.
(2000).
[274] B. P. Zhang, N. T. Binh, K. Wakatsuki, C. Y. Liu, Y. Segawa, and
N. Usami, Appl. Phys. Lett. 86, 032105 (2005).
[275] T. Nobis, Ortsaufgelste und zeitabhngige Photostromuntersuchungen
an Zinkoxid-Strukturen, Diplomarbeit, Universitt Leipzig (2003).
[276] U. Jahn, J. Miguel-Snchez, T. Flissikowski, H. T. Grahn, R. Hey, and
A. Guzmn, Phys. Status Solidi (c) 3, 627 (2006).
[277] B. N. Mbenkum, N. Ashkenov, M. Schubert, M. Lorenz, H. Hochmuth,
D. Michel, M. Grundmann, and G. Wagner, Appl. Phys. Lett. 86, 091904
(2005).
[278] N. Ashkenov, M. Schubert, E. Twerdowski, H. v. Wenckstern,
B. Mbenkum, H. Hochmuth, M. Lorenz, W. Grill, and M. Grundmann,
Thin Solid Films 486, 153 (2005).
[279] M. Lorenz, M. Brandt, J. Schubert, H. Hochmuth, H. von Wenckstern,

168
Bibliography

M. Schubert, and M. Grundmann, Proc. SPIE 6474, 64741S (2007).


[280] R. D. Vispute, V. Talyansky, Z. Trajanovic, S. Choopun, M. Downes,
R. P. Sharma, T. Venkatesan, M. C. Woods, R. T. Lareau, K. A. Jones,
and A. A. Iliadis, Appl. Phys. Lett. 70, 2735 (1997).
[281] L. Pintilie, I. Vrejoiu, D. Hesse, G. LeRhun, and M. Alexe, Phys. Rev. B
75, 104103 (2007).
[282] I. Vrejoiu, Y. Zhu, G. L. Rhun, M. A. Schubert, D. Hesse, and M. Alexe,
Appl. Phys. Lett. 90, 072909 (2007).
[283] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann,
N. Ashkenov, and M. Schubert, J. Electron. Mater. 37, 1029 (2008).
[284] F. Bernardini, V. Fiorentini, and D. Vanderbilt, Phys. Rev. B 56, R10024
(1997).
[285] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, N. Ashkenov,
M. Grundmann, and M. Schubert, Appl. Phys. Lett. 95, 082902 (2009).
[286] V. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann,
N. Ashkenov, and M. Schubert, J. Electron. Mater. 37, 1029 (2008).
[287] S. L. Miller, R. D. Nasby, J. R. Schwank, M. S. Rodgers, and P. V.
Dressendorfer, J. Appl. Phys. 68, 6463 (1990).
[288] S. L. Miller, J. R. Schwank, R. D. Nasby, and M. S. Rodgers, J. Appl.
Phys. 70, 2849 (1991).
[289] A. Li, C. Ge, P. Lu, D. Wu, S. Xiong, and N. Ming, Appl. Phys. Lett.
70, 1616 (1997).
[290] W. J. Lin, T. Y. Tseng, H. B. Lu, S. L. Tu, S. J. Yang, and I. N. Lin, J.
Appl. Phys. 77, 6466 (1995).
[291] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, M. Grundmann, and
J. Schubert, J. Vac. Sci. Technol. B 27, 1789 (2009).
[292] S.-Y. Wu, IEEE Trans. Electron. Devices 21, 499 (1974).
[293] J. Nishii, A. Ohtomo, K. Ohtani, H. Ohno, and M. Kawasaki, Jpn. J.
Appl. Phys. 44, L1193 (2005).
[294] G. Adamopoulos, A. Bashir, P. H. Wbkenberg, D. D. C. Bradley, and
T. D. Anthopoulos, Appl. Phys. Lett. 95, 133507 (2009).
[295] H. Frenzel, A. Lajn, M. Brandt, H. von Wenckstern, G. Biehne,
H. Hochmuth, M. Lorenz, and M. Grundmann, Appl. Phys. Lett. 92,
192108 (2008).
[296] J. Siddiqui, E. Cagin, D. Chen, and J. D. Phillips, Appl. Phys. Lett. 88,
212903 (2006).
[297] E. M. Kaidashev, M. Lorenz, H. von Wenckstern, A. Rahm, H.-C. Sem-
melhack, K.-H. Han, G. Benndorf, C. Bundesmann, H. Hochmuth, and
M. Grundmann, Appl. Phys. Lett. 82, 3901 (2003).
[298] R. L. Hoffman, B. J. Norris, and J. F. Wager, Appl. Phys. Lett. 82, 733

169
Bibliography

(2003).
[299] S. Masuda, K. Kitamura, Y. Okumura, S. Miyatake, H. Tabata, and
T. Kawai, J. Appl. Phys. 93, 1624 (2003).
[300] B. V. Zeghbroeck, Principles of Semiconductor Devices, e-book:

http://ece.colorado.edu/bart/book/book/index.html (2004).
[301] J. W. Orton, B. J. Goldsmith, M. J. Powell, and J. A. Chapman, Appl.
Phys. Lett. 37, 557 (1980).
[302] T. Harigai, S.-M. Nam, H. Kakemoto, S. Wada, K. Saito, and T. Tsurumi,
Thin Solid Films 509, 13 (2006).

170
List of own publications

[1] M. Brandt, M. Lange, M. Stlzel, A. Mller, G. Benndorf, J. Zippel,


J. Lenzner, M. Lorenz, and M. Grundmann, Appl. Phys. Lett. 97, 052101
(2010).
[2] A. Lajn, H. von Wenckstern, G. Benndorf, C. Dietrich, M. Brandt,
G. Biehne, H. Hochmuth, M. Lorenz, and M. Grundmann, J. Electron.
Mater. 39, 595 (2010).
[3] M. Brandt, H. von Wenckstern, G. Benndorf, M. Lange, C. P. Dietrich,
C. Kranert, C. Sturm, R. Schmidt-Grund, H. Hochmuth, M. Lorenz,
M. Grundmann, M. R. Wagner, M. Alic, C. Nenstiel, and A. Hoffmann,
Phys. Rev. B 81, 073306 (2010).
[4] M. Brandt, H. von Wenckstern, M. Stlzel, H. Hochmuth, M. Lorenz,
and M. Grundmann, Semicond. Sci. Tech. in press (2010).
[5] M. Lorenz, M. Brandt, M. Lange, G. Benndorf, H. von Wenckstern,
D. Klimm, and M. Grundmann, Thin Solid Films 518, 4623 (2010).
[6] B. Cao, M. Lorenz, G. Zimmermann, C. Czekalla, M. Brandt, H. von
Wenckstern, and M. Grundmann, Nanowires, 117132 (2010), ISBN 978-
953-7619-79-4.
[7] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, N. Ashkenov,
M. Grundmann, and M. Schubert, Appl. Phys. Lett. 95, 082902 (2009).
[8] H. von Wenckstern, H. Schmidt, M. Brandt, A. Lajn, R. Pickenhain,
M. Lorenz, M. Grundmann, D. M. Hofmann, A. Polity, B. K. Meyer,
H. Saal, M. Binnewies, A. Brger, K.-D. Becker, V. A. Tikhomirov, and
K. Jug, Prog. Sol. Stat. Chem. 37, 153 (2009).
[9] M. Brandt, H. von Wenckstern, G. Benndorf, H. Hochmuth, M. Lorenz,
and M. Grundmann, Thin Solid Films 518, 1048 (2009).
[10] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann,
N. Ashkenov, and M. Schubert, Appl. Phys. Lett. 94, 142904 (2009).
[11] M. Lorenz, M. Brandt, G. Wagner, H. Hochmuth, G. Zimmermann,
H. von Wenckstern, and M. Grundmann, Proc. SPIE 7217, 72170N
(2009).
[12] M. Brandt, H. von Wenckstern, C. Meinecke, T. Butz, H. Hochmuth,
M. Lorenz, and M. Grundmann, J. Vac. Sci. Technol. B 27, 1604 (2009).
[13] M. Lorenz, B. Cao, G. Zimmermann, G. Biehne, C. Czekalla, H. Fren-

171
List of own publications

zel, M. Brandt, H. von Wenckstern, and M. Grundmann, J. Vac. Sci.


Technol. B 27, 1693 (2009).
[14] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, M. Grundmann, and
J. Schubert, J. Vac. Sci. Technol. B 27, 1789 (2009).
[15] H. Frenzel, H. von Wenckstern, A. Lajn, M. Brandt, G. Biehne,
H. Hochmuth, M. Lorenz, and M. Grundmann, AIP conference proceed-
ings 1199, 469 (2009).
[16] H. von Wenckstern, J. Zippel, A. Lajn, M. Brandt, G. Biehne,
H. Hochmuth, M. Lorenz, and M. Grundmann, AIP conference proceed-
ings 1199, 99 (2009).
[17] H. von Wenckstern, M. Brandt, H. Schmidt, G. Benndorf, J. Zippel,
H. Hochmuth, M. Lorenz, and M. Grundmann, Proc. SPIE 6895, 689505
(2008).
[18] B. Q. Cao, M. Lorenz, H. von Wenckstern, C. Czekalla, M. Brandt,
J. Lenzner, G. Benndorf, G. Biehne, and M. Grundmann, Proc. SPIE
6895, 68950V (2008).
[19] V. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann,
N. Ashkenov, and M. Schubert, J. Electron Mater. 37, 1029 (2008).
[20] G. Brauer, W. Anwand, W. Skorupa, J. Kuriplach, O. Melikhova,
J. Cizek, I. Prochazka, H. von Wenckstern, M. Brandt, M. Lorenz,
and M. Grundmann, Phys. Status Solidi C 4, 3642 (2007).
[21] M. Brandt, H. von Wenckstern, H. Schmidt, A. Rahm, G. Biehne,
G. Benndorf, H. Hochmuth, M. Lorenz, C. Meinecke, T. Butz, and
M. Grundmann, J. Appl. Phys. 104, 013708 (2008).
[22] M. Lorenz, G. Wagner, A. Rahm, H. Schmidt, H. Hochmuth, H. Schmid,
W. Mader, M. Brandt, H. von Wenckstern, and M. Grundmann, Phys.
Status Solidi C 5, 3280 (2008).
[23] H. Frenzel, A. Lajn, M. Brandt, H. von Wenckstern, G. Biehne,
H. Hochmuth, M. Lorenz, and M. Grundmann, Appl. Phys. Lett. 92,
192108 (2008).
[24] H. von Wenckstern, M. Brandt, H. Schmidt, C. Hanisch, G. Benndorf,
H. Hochmuth, M. Lorenz, and M. Grundmann, J. Korean Phys. Soc. 53,
3064 (2008).
[25] B. Q. Cao, M. Lorenz, M. Brandt, H. von Wenckstern, J. Lenzner,
G. Biehne, and M. Grundmann, Phys. Status Solidi RRL 2, 37 (2008).
[26] V. M. Voora, T. Hofmann, M. Brandt, M. Lorenz, M. Grundmann, and
M. Schubert, Phys. Status Solidi C 5, 1328 (2008).
[27] H. von Wenckstern, R. Pickenhain, H. Schmidt, M. Brandt, G. Biehne,
M. Lorenz, M. Grundmann, and G. Brauer, Superlattices Microstruct.
42, 14 (2007).
[28] H. von Wenckstern, H. Schmidt, C. Hanisch, M. Brandt, C. Czekalla,

172
List of own publications

G. Benndorf, G. Biehne, A. Rahm, H. Hochmuth, M. Lorenz, and


M. Grundmann, Phys. Status Solidi RRL 1, 129 (2007).
[29] M. Lorenz, M. Brandt, J. Schubert, H. Hochmuth, H. von Wenckstern,
M. Schubert, and M. Grundmann, Proc. SPIE 6474, 64741S (2007).
[30] H. von Wenckstern, M. Brandt, G. Zimmermann, J. Lenzner,
H. Hochmuth, M. Lorenz, and M. Grundmann, Proc. MRS 957, 23
(2007).
[31] H. von Wenckstern, R. Pickenhain, H. Schmidt, M. Brandt, G. Biehne,
M. Lorenz, M. Grundmann, and G. Brauer, Appl. Phys. Lett. 89, 092122
(2006).
[32] H. von Wenckstern, G. Benndorf, S. Heitsch, J. Sann, M. Brandt,
H. Schmidt, J. Lenzner, M. Lorenz, A. Y. Kuznetsov, B. K. Meyer, and
M. Grundmann, Appl. Phys. A 88, 125 (2007).
[33] H. von Wenckstern, M. Brandt, H. Schmidt, G. Biehne, R. Pickenhain,
H. Hochmuth, M. Lorenz, and M. Grundmann, Appl. Phys. A 88, 135
(2007).
[34] I. Konovalov, L. Hartmann, M. Brandt, and L. Makhova, Phys. Status
Solidi C 3, 2891 (2006).
[35] M. Lorenz, H. Hochmuth, J. Lenzner, M. Brandt, H. von Wenckstern,
G. Benndorf, and M. Grundmann, Wissenschaftlich-Technische Berichte
des FZ Rossendorf FZR-433, 57 (2005).

173
List of own publications

174
List of own conference
contributions

[1] M. Brandt, H. von Wenckstern, M. Stlzel, M. Lange, C. P. Dietrich,


A. Mller, G. Benndorf, J. Zippel, J. Lenzner, M. Lorenz et al.. Obser-
vation of a strong polarization induced quantum-confined Stark effect in
Mgx Zn1x O/ZnO quantum wells. 52st Electronic Materials Conference,
Notre Dame, IN, 2010 (2010).
[2] M. Brandt, H. von Wenckstern, M. Lorenz, H. Frenzel, G. Benndorf,
M. Lange, M. Stlzel, C. Dietrich, J. Zippel, and M. Grundmann. Com-
plex oxide heterostructures grown by pulsed laser deposition. 2010 SPIE
Photonics West, San Francisco, 2010 (2010).
[3] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, and M. Grundmann.
Ferroelectric thin film field-effect transistors based on ZnO/BaTiO3 het-
erostructures. 51st Electronic Materials Conference, Penn State Univer-
sity, 2009 (2009).
[4] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, and M. Grundmann.
Transparent non-volatile memory devices based on ZnO/BaTiO3 ferro-
electric FETs. 2009 MRS Spring Meeting, San Francisco, 2009 (2009).
[5] M. Brandt, H. von Wenckstern, G. Benndorf, J. Zippel, J. Lenzner,
H. Hochmuth, M. Lorenz, and M. Grundmann. Formation of a 2DEG
in ZnO/MgZnO single heterostructures and quantum wells. 2009 MRS
Spring Meeting, San Francisco, 2009 (2007).
[6] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, and M. Grundmann.
Ferroelectric thin film transistors based on ZnO/BaTiO3 heterojunc-
tions. German Physical Society Spring Meeting, Dresden, March 2009
(2009).
[7] M. Brandt, H. von Wenckstern, H. Hochmuth, M. Lorenz, J. Zippel,
J. Lenzner, G. Benndorf, and M. Grundmann. Accumulation of carriers
in ZnO/MgZnO heterostructures and quantum wells. 2nd International
Symposium on Transparent Conducting Oxides, Hersonissos, Crete, 2008
(2008).
[8] M. Brandt, H. von Wenckstern, C. Dietrich, G. Benndorf, J. Zippel, A. Ab-
dullah, C. Meinecke, T. Butz, H. Hochmuth, M. Lorenz et al.. Dopant
activation in homoepitaxial ZnO:P and ZnMgO:P thin films. The 5th

175
List of own conference contributions

International Wokshop on ZnO and Related Materials, Ann Arbor, Michi-


gan, 2008 (2008).
[9] M. Brandt, H. Frenzel, H. Hochmuth, M. Lorenz, J. Schubert, and
M. Grundmann. Ferroelectric thin film field-effect transistors based on
ZnO/BaTiO3 heterostructures. The 5th International Wokshop on ZnO
and Related Materials, Ann Arbor, Michigan, 2008 (2008).
[10] M. Brandt, H. Hochmuth, M. Lorenz, M. Grundmann, N. Ashkenov,
V. Voora, and M. Schubert. Ferroelectric properties of BaTiO3 ZnO
heterojunctions. German Physical Society Spring Meeting, Berlin, March
2008 (2008).
[11] M. Brandt, H. von Wenckstern, H. Hochmuth, M. Lorenz, G. Biehne,
G. Benndorf, C. Meinecke, T. Butz, H. Schmidt, A. Rahm et al.. Ho-
moepitaxial growth of ZnO thin films by pulsed laser deposition (PLD).
German Physical Society Spring Meeting, Berlin, March 2008 (2008).
[12] M. Brandt, H. Hochmuth, M. Lorenz, M. Grundmann, M. Schubert,
T. Hofmann, V. M. Voora, N. Ashkenov, and J. Schubert. Electrical
and electro-optical investigations of the polarization coupling in epitax-
ial ferroelectric BTO/ZnO heterostructures. 2007 MRS Fall Meetingg,
Boston, Fall 2007 (2007).
[13] M. Brandt, H. von Wenckstern, C. Hanisch, H. Hochmuth, M. Lorenz,
H. Schmidt, and M. Grundmann. Homoepitaxial growth of ZnO thin
films by pulsed laser deposition (PLD). German Physical Society Spring
Meeting, Regensburg, March 2007 (2007).
[14] M. Brandt, H. von Wenckstern, H. Hochmuth, M. Lorenz, M. Grund-
mann, J. Schubert, V. Voora, and M. Schubert. Ferroelectric properties
of BaTiO3 - ZnO heterojunctions. German Physical Society Spring Meet-
ing, Regensburg, March 2007 (2007).
[15] M. Brandt, H. Schmidt, H. von Wenckstern, C. Henkel, H. Hochmuth,
M. Lorenz, and M. Grundmann. Charge carrier profiling of ZnO. The
4th International Wokshop on ZnO and Related Materials, 03.-06.10.2006,
Giessen, Germany P 221 (2006).
[16] M. Brandt, H. von Wenckstern, G. Benndorf, J. Lenzner, H. Schmidt,
M. Lorenz, M. Grundmann, G. Braunstein, and G. Brauer. Optical
and microelectrical characterization of ZnO single crystals implanted with
group V elements. German Physical Society Spring Meeting, Dresden,
March 2006 HL14 4.
[17] M. Brandt, H. von Wenckstern, G. Zimmermann, S. Heitsch, G. Benndorf,
H. Hochmuth, M. Lorenz, and M. Grundmann. Hall effect measurements
on ZnO thin films. German Physical Society Spring Meeting, Berlin, 2005
HL 17.4 (2005).

176
List of Symbols

Variable Meaning First Occurrence


a, c Lattice constants in a hexagonal crystal . . . . . . . . . . . . . . . . . 16
a0 , c 0 Lattice constants of an unstrained pure ZnO substrate . . 18
b Burgers vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
B Magnetic induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
inverse thermal energy = kB1T . . . . . . . . . . . . . . . . . . . . . . . . . . 8
BH Brooks-Herring correction factor . . . . . . . . . . . . . . . . . . . . . . . . . 8
C Capacitance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
C13 , C33 Elastic constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Ci Insulator Capacitance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
cl Longitudinal elasticity constant . . . . . . . . . . . . . . . . . . . . . . . . . . 8
cp Phosphorus content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
cps c-axis lattice constant under pseudomorphic strain . . . . . . 18
Cst Capacitance of standard capacitor . . . . . . . . . . . . . . . . . . . . . 136
cx c-axis lattice constant as function Mg content x . . . . . . . . . 18
Dn Deformation potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Df Dielectric displacement in ferroelectric . . . . . . . . . . . . . . . . . 134
df Thickness of ferroelectric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
dz Thickness of individual ZnO layers . . . . . . . . . . . . . . . . . . . . . 134
1 , 2 Crystal-field and Spin-Orbit splitting parameters . . . . . . . . 59
E Electron energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
e Elementary charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
EC Coercive field strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
ED Donor activation energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
ED0 Donor activation energy in the dilute limit . . . . . . . . . . . . . . 85
Ef Electric field in the ferroelectric . . . . . . . . . . . . . . . . . . . . . . . . 132
epz Isotropic piezoelectric coefficient . . . . . . . . . . . . . . . . . . . . . . . . . .8
Relative permittivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
0 Permittivity of free space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
High frequency permittivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Strain perpendicular to the substrate . . . . . . . . . . . . . . . . . . . .17
k Strain parallel to the substrate . . . . . . . . . . . . . . . . . . . . . . . . . . 17
f Relative permittivity of ferroelectric layer . . . . . . . . . . . . . . 134
z Relative permittivity of ZnO layer . . . . . . . . . . . . . . . . . . . . . 134
F Electric field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
f Distribution function of electrons . . . . . . . . . . . . . . . . . . . . . . . . . 7
b Barrier height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
e , h Electron and Hole wave function . . . . . . . . . . . . . . . . . . . . . . . 107
x , y Perturbation terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
r Exciton radial wave function . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Surface energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
~ Plancks constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Ideality factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
I Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
ji Current density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
ka , kc Change of lattice constant with magnesium content . . . . . 16

177
List of symbols

kx , ky Components of the electron wave vector . . . . . . . . . . . . . . . . . . 7


kB Boltzmann constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
kc,ps Pseudomorphic c-axis lattice constant . . . . . . . . . . . . . . . . . . . 18
L Channel length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
LD Debye length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Lc Collision operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Exciton radius . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
i Differential rate of scattering in . . . . . . . . . . . . . . . . . . . . . . . . . . 8
o Differential rate of scattering out . . . . . . . . . . . . . . . . . . . . . . . . . 8
Lz Quantum well width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
me Effective electron mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Rigidity modulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
H Hall-mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .60
NA Acceptor concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62
NC Critical donor concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
ND Donor concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
NI Concentration of ionized impurities . . . . . . . . . . . . . . . . . . . . . . . 8
Npo Phonon occupation number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Nt Net dopant concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
n Free electron concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
el Elastic scattering rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
o Phonon frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
c Cyclotron frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
P Electromechanical coupling coefficient . . . . . . . . . . . . . . . . . . . . 8
Pd Polarization of the ferroelectric . . . . . . . . . . . . . . . . . . . . . . . . 132
Pr Remanent polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Ps Spontaneous polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Psat Saturation polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
PZnO Spontaneous polarization of the individual ZnO layer . . .134
p(O2 ) Oxygen partial pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
r Radius of a depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
RS Series resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
RP Parallel resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Sn HuangRhys factor of n-th phonon replica . . . . . . . . . . . . . 107
Si Rate of scattering into a specific state . . . . . . . . . . . . . . . . . . . . 7
So Rate of scattering out of a specific state . . . . . . . . . . . . . . . . . . 7
s Mobility distribution function . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
b , f , m Charge at individual interfaces of a heterojuction . . . . . . 134
ij Components of the conductivity tensor . . . . . . . . . . . . . . . . . . . 9
T Effective capture cross section . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
T Absolute temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
t Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
Debye temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Scattering time of elastic scattering . . . . . . . . . . . . . . . . . . . . . . 7
V Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
VD Drain voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

178
List of symbols

VG Gate voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145


VT Threshold voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
W Channel width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
w Depletion layer width in the individual ZnO layer . . . . . . 132
x Magnesium content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
XC Acoustic deformation potential . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Exciton dimensionality parameter . . . . . . . . . . . . . . . . . . . . . . 107

179
List of symbols

180
List of Figures

1.1 Change in the spontaneous polarization vs Mg content x. . . . . 3

2.1 Schematics of the Sawyer Tower measurement circuit. (Courtesy


of Venkata M. Voora) . . . . . . . . . . . . . . . . . . . . . . . . 11

3.1 Possible orientantions of the wurtzite material ZnO. . . . . . . . 14


3.2 Peak position of photoluminescence spectra vs Mg content x. . . 16
3.3 Principle of pseudomorphic growth. . . . . . . . . . . . . . . . . 18
3.4 Reciprocal space maps of (002) and (104) reflection of a Mgx Zn1x O
thin film grown on a CT substrate. . . . . . . . . . . . . . . . . 19
3.5 Reciprocal space maps of (002) and (104) reflection of a Mgx Zn1x O
thin film grown on a CT substrate. . . . . . . . . . . . . . . . . 20
3.6 Change in the c-axis lattice constant vs. Mg content x in pseu-
domorphic Mgx Zn1x O thin films. . . . . . . . . . . . . . . . . . 22
3.7 2 m2 m AFM images of ZnO thin films grown on a-plane
sapphire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.8 Example for double spirals and islands present on one ZnO sample. 26
3.9 2 m2 m AFM images and film thickness of ZnO thin films
after 100, 300, 1,000, 3,000 and 10,000 pulses. . . . . . . . . . . 29
3.10 2 m2 m AFM images of Mgx Zn1x O thin films grown on
sapphire substrates. . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.11 Free electron concentration and mobility in various ZnO sub-
strates versus temperature. . . . . . . . . . . . . . . . . . . . . . 35
3.12 Comparison of reciprocal space maps of the (002), (101) and
(104) reflexes obtained on a TD substrate and two selected CT
substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.13 2 m2 m AFM height images of ZnO substrates purchased
from CrysTec GmbH before and after thermal annealing. . . . . 39
3.14 2 m1 m AFM height images of homoepitaxial ZnO thin films
grown at various temperatures. . . . . . . . . . . . . . . . . . . 40
3.15 Reciprocal space maps of the (002) and (104) reflex of a ho-
moepitaxial ZnO thin film. . . . . . . . . . . . . . . . . . . . . . 42
3.16 Cross section TEM bright field images of a homoepitaxial ZnO
thin film. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.17 CBED diffraction patterns as obtained on an O-polar homoepi-
taxial ZnO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

181
List of Figures

3.18 Cross section HR-TEM image of the interface between a ho-


moepitaxial thin film and the substrate. . . . . . . . . . . . . . 44
3.19 2 m0.5 m AFM height images of homoepitaxial ZnO thin
films of different thickness. . . . . . . . . . . . . . . . . . . . . . 45
3.20 AFM height images of Mgx Zn1x O thin films grown on ZnO. . . 46
3.21 2 m2 m AFM height images of Mgx Zn1x O samples with
various Mg contents grown on substrates of different miscuts. . . 48
3.22 Perovskite unit cell (ABO3 ). . . . . . . . . . . . . . . . . . . . . 50
3.23 2 m2 m AFM images of untreated (100) STO substrates as
received from (a) Crystal GmbH and (b) Crystec GmbH. (Cour-
tesy of Stefan Schche) . . . . . . . . . . . . . . . . . . . . . . . 51
3.24 2 m2 m AFM images of annealed STO substrates. . . . . . . 51
3.25 2 m1 m AFM images of BTO thin films. . . . . . . . . . . . 52
3.26 SRO thin films of roughly 30 nm thickness grown on STO sub-
strates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.1 2 X-ray diffraction patterns of the (002) reflection of ho-


moepitaxial ZnO thin films. . . . . . . . . . . . . . . . . . . . . 57
4.2 Near band edge PL spectra of homoepitaxially ZnO thin films. . 58
4.3 Temperature dependence of the free electron concentration and
the Hall-mobility in nominally undoped homoepitaxial ZnO thin
films. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.4 Temperature dependence of capacitance of PtOx Schottky diodes
fabricated from homoepitaxial ZnO thin films. . . . . . . . . . . 63
4.5 2 m1 m AFM height images for homoepitaxial ZnO:P samples. 66
4.6 HR-XRD reciprocal space map of the (002) and (104) reflection
of a homoepitaxial ZnO:P sample. . . . . . . . . . . . . . . . . . 67
4.7 2 X-ray diffraction patterns of the (002) reflection of ho-
moepitaxial ZnO:P thin films. . . . . . . . . . . . . . . . . . . . 68
4.8 Channeling yield and phosphorous content of homoepitaxial
ZnO:P thin films as extracted from RBS and PIXE measurements. 70
4.9 Low temperature recombination spectra of homoepitaxial ZnO:P
samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.10 Conductivity of two typical homoepitaxial ZnO:P thin films
compared to the conductivity of a CT substrate. . . . . . . . . . 74
4.11 Hall-mobility and free carrier concentration as a function of tem-
perature in homoepitaxial ZnO:P thin films. . . . . . . . . . . . 75
4.12 Hall-mobility and free carrier concentration as a function of tem-
perature in homoepitaxial ZnO:P thin films before and after an-
nealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.13 Resistivity of homoepitaxial phosphorous doped ZnO thin film
as grown and after annealing. . . . . . . . . . . . . . . . . . . . 78
4.14 Mobility spectra of a Mgx Zn1x O layer grown on the (0001) face
of a ZnO single crystal as obtained from the f-MEMSA algorithm. 80
4.15 Magnesium and phosphorous content of the Mgx Zn1x O thin
films deposited on ZnO substrates as determined by PIXE. . . . 83

182
List of Figures

4.16 HR-XRD triple axis 2 scans of the symmetric (002) reflex


of Mgx Zn1x O:P films. . . . . . . . . . . . . . . . . . . . . . . . 85
4.17 HR-XRD reciprocal space maps of the symmetric (002) and the
asymmetric skew (101) reflex of a Mgx Zn1x O:P film grown on
ZnO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.18 Electron mobility and free electron concentration versus tem-
perature in selected Mgx Zn1x O:P thin films grown on ZnO. . . 87
4.19 (a) Donor activation energy versus donor concentration in ZnO:P
and Mgx Zn1x O:P films grown on ZnO. (b) Ratio of the resistiv-
ities of Mgx Zn1x O:P thin films grown on ZnO after and before
annealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.1 Band diagram of ZnO/Mgx Zn1x O heterostructures investigated. 94


5.2 Capacitance versus voltage dependence and apparent dopant
concentration of a Schottky diode on a ZnO/Mgx Zn1x O het-
erostructure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.3 Free electron concentration and mobility versus temperature in
a ZnO/Mgx Zn1x O heterostructure. . . . . . . . . . . . . . . . . 97
5.4 AFM images of a ZnO/Mgx Zn1x O QW grown on sapphire and
ZnO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5 Temperature dependence of the free carrier concentration and
the mobility in a ZnO/Mgx Zn1x O quantum well structure. . . . 101
5.6 HR-XRD reciprocal space maps of the symmetric (002) and the
asymmetric skew (101) reflex of a ZnO/Mgx Zn1x O QW. . . . . 102
5.7 Electron mobility versus temperature in a homoepitaxial ZnO
thin film, a ZnO/Mgx Zn1x O QW structure grown on ZnO and
a ZnO single crystal. . . . . . . . . . . . . . . . . . . . . . . . . 103
5.8 Peak width (FWHM) of the QW related PL peak versus QW
thickness LW for ZnO/Mgx Zn1x O QW structures grown on sap-
phire and ZnO substrates. . . . . . . . . . . . . . . . . . . . . . 104
5.9 Properties of an exciton in a c-oriented ZnO/Mgx Zn1x O QW
with x = 0.18 as function of the QW width obtained from vari-
ational calculations. . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.10 Schematics of ZnO/Mgx Zn1x O QW samples investigated in the
present study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.11 Low temperature CL spectra obtained from two otherwise equal
Zno/Mgx Zn1x O QWs deposited at two different laser fluences:
(a) lense in position 2 (low fluence) (b) lense in position 1 (high
fluence). The spectra are color coded, red corresponds to a high,
blue to a low QW thickness. (Data obtained by Jan Zippel) . . 109
5.12 Excitation dependence of the CL spectra of a sample showing a
pronounced QCSE. . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.13 Normalized PL spectra of two samples with different Mg content
x in the barrier. . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

183
List of Figures

5.14 (a) Spectral maximum of QW related luminescence in various


QW samples (b) HuangRhys factor as a function of the spectral
position of the main peak. . . . . . . . . . . . . . . . . . . . . . 115
5.15 PL maximum position and decay time as a function of the QW
thickness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

6.1 SawyerTower-circuit response of a ZnO/BTO heterostructure


grown on Si with Pt contacts measured at various cycling fre-
quencies and voltage amplitudes as indicated. The data was
obtained from selected contacts on the wafer shown in Fig. 6.3. . 123
6.2 Current voltage characteristic of a ZnO/BTO heterojunction
grown on Si with Pt contacts. . . . . . . . . . . . . . . . . . . . 124
6.3 Maps of the parallel and series resistance of ZnO/BTO hetero-
junctions of a 3 inch silicon wafer. . . . . . . . . . . . . . . . . . 125
6.4 XRD Wide angle 2 scans of ZnO/BTO heterostructures
grown on Si/Pt and STO/SRO. . . . . . . . . . . . . . . . . . . 126
6.5 XRD -scans of a ZnO/BTO heterostructure grown on an STO
substrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.6 Current flowing into a ZnO/BTO heterostructure grown on a
STO:Nb as function of applied voltage and integration time.
The ZnO has a thickness of 500 nm, the BTO 800 nm. . . . . . . 128
6.7 Currentvoltage characteristic of two selected contacts with a
reduced ZnO thickness of 50 nm. . . . . . . . . . . . . . . . . . . 130
6.8 Schematic of the distribution of space charge regions within the
heterojunctions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.9 Experimentally measured and model calculated Sawyer-Tower
loops. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
6.10 Model calculated SawyerTower loops and depletion layer widths
in ZnO/BTO double heterostuctures. . . . . . . . . . . . . . . . 140
6.11 Principal of CV response of a SIS, a SFS and of ZnO/BTO
double heterostructures with two different parallel arrangements
of the ZnO polarization. . . . . . . . . . . . . . . . . . . . . . . 141
6.12 Experimental (a) IV and (b) CV characteristic of a ZnO/BTO/ZnO
double heterostructure. . . . . . . . . . . . . . . . . . . . . . . . 142
6.13 Principle layout of ZnO/BTO MISFET transistors. . . . . . . . 143
6.14 Output characteristics of a ZnO/BTO MISFET at different gate
voltages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.15 Transfer characteristics and calculated channel mobility of a
ZnO/BTO MISFET at a fixed sourcedrain-voltage of VSD = 2V.145
6.16 (a) Transfer characteristics of a ZnO/BTO MISFET after ap-
plying a for- and backwards pulse, respectively. (b) Ratio of the
current in the permanent on and the permanent off state. . . 146

184
List of Tables

3.1 Change of lattice constants of Mgx Zn1x O thin films with in-
creasing magnesium content for various growth methods on dif-
ferent substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Suppliers and growth methods of ZnO single crystals investigated. 34

4.1 Donor activation energy, donor and acceptor concentration and


capture cross section homoepitaxial ZnO thin films. . . . . . . . 62
4.2 Impurity concentration as detected by PIXE in homoepitaxial
ZnO:P samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3 Donor and acceptor concentration and donor activation energy
in ZnO:P thin films. . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Mg and P content, lattice parameters and room temperature
electron concentration and Hall-mobility in MgZnO:P thin films
grown on ZnO. . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

6.1 Operation schemes of a ZnO/BTO/ZnO heterostructure. . . . . 135


6.2 Best-match model calculation parameters. . . . . . . . . . . . . 137

185
List of Tables

186
Acknowledgement

First of all, I acknowledge the guidance by my supervisor Prof. Dr. Marius


Grundmann and the countless opportunities he gave me to ask questions and
to improve my knowledge. Without the discussions, seminars, invited talks
and conferences he organized or enabled me to visit the thesis certainly would
not be what it is today.

I am indebted to PD Dr. habil Michael Lorenz, who is the supervisor of


the A2 project of the Sonderforschungsbereich 762, for numerous discussions
and extensive support especially in regard to PLD related questions. Im very
grateful for the scientific freedom he admitted to me and for providing the
means of realization of these ideas wherever possible.

I want to thank Dr. Holger von Wenckstern for all the support in the past
years, starting from supervising my diploma thesis. I always benefited from
his respectable knowledge on the matter of electrical characterization, and was
highly influenced by his encouraged and encouraging support. Thank you for
the always open door, and the friendship that developed.

I also want to thank Gabriele Benndorf for her help regarding optical and lumi-
nescence related questions, her accurate and critical comments, and countless
nightly discussions on scientific problems. I deeply appreciate the personal
relation to her, her warm hearted attitude and the help she provided on any
personal matter. Danke Gabi!

I am greatly indebted to all the people who helped in carrying out the work
discussed in this thesis, namely

Holger Hochmuth for the help regarding the growth of the films studied,
RHEED measurements as well as technical support in multiple ways. I
appreciate his patience on the topic of mysteriously disappearing tools.

Helena Hilmer and Martin Lange for mutual help in sample growth and
company during nights and weekends.

Hannes Mnch for the help regarding sample growth and metalization.

Dr. Jrgen Schubert (Forschungszentrum Jlich) for providing BTO tem-


plates.

187
. Acknowledgement

Gisela Biehne and Monika Hahn for the help in sample processing. Beson-
deren Dank an Gisela fr die Hilfe auerhalb aller Arbeitszeiten.

Jrg Lenzner for several microscopy and CL measurements and extensive


help regarding technical questions.

Gabriele Ramm for the preparation of the target material used in the
study and technical help.

PD Dr. habil Michael Lorenz, Christof Dietrich, Stefan Schche and Jan
Zippel for XRD measurements.

Dr. Heidemarie Schmidt (now Forschungszentrum Rossendorf), Helena


Hilmer, Stefan Schche and Gregor Zimmermann for AFM investigations.

Gabriele Benndorf, Christian Czekalla, Chistof P. Dietrich, Martin Lange,


Alexander Mller, Marko Stlzel, Jan Zippel and Johannes Kupper for
the various luminescence spectroscopy measurements.

Roswitha Riedel for various technical help.

Dr. Holger von Wenckstern and Robert Heinhold for various Hall-effect
measurements.

Dr. Heidemarie Schmidt and Danilo Brger (both Forschungszentrum


Rossendorf) for the magnetotransport measurements.

Robert Heinhold for the MEMSA evaluation.

Dr. Holger von Wenckstern, Martin Ellguth, Alexander Lajn, Matthias


Schmidt, Kerstin Brachwitz, Florian Schmidt and Stefan Mller for nu-
merous measurements on rectifying contacts, including IV, CV, TAS
and DLTS measurements.

Heiko Frenzel and Peter Schwinkendorf for the help regarding the char-
acterization of field effect transistors.

Dr. Rdiger Schmidt-Grund, Helena Hilmer, Philipp Khne, Stefan


Schche, Chris Sturm and David Schumacher for numerous ellipsome-
try measurements.

Christoph Meinecke for the RBS and PIXE measurements discussed in


this work.

Dr. Gerald Wagner, Prof. Dr. Werner Mader, Dr. Herbert Schmid (both
Universitt Bonn) and Dr. Jess Dr. Ziga-Prez (Centre de Recherche
sur lHtro-Epitaxie et ses Applications) for the TEM measurements.

Nurdin Ashkenov and Venkata M. Voora for ST, IV and CV measure-


ments on ZnO/BTO heterostructures.

188
Venkata M. Voora and Mathias Schubert (University of Nebraska, Lin-
coln) for the cooperation in developing the model of ZnO/BTO coupling
and for numerous mutual visits.

PD Dr. habil. Rainer Pickenhain for helpful discussions.

Anja Heck and Birgit Wendisch for help with all administrative issues.

Prof. Dr. Kremer for providing his AFM instrument.

Dr. Atsushi Tsukazaki for the inspiring opportunity to visit his lab.

My room members Matthias Schmidt, Venkata M. Voora, Thomas Lder,


Zhipeng Zhang, Michael Lorenz, Robert Heinhold, Heiko Frenzel and
Martin Lange for various invaluable help and exciting discussions.

My librarians Annika Brandt (Universitt Halle) and Stefanie Benkwitz


(Universitt Jena) for the help in acquiring various scientific articles and
their kind motivation.

Gregor Zimmermann for the help with LATEXrelated questions.

PD. Dr. habil. Michael Lorenz, Dr. Holger von Wenckstern, Dr. Gerald
Wagner, Gabriele Benndorf and Andreas Kraus for help in proofreading
the present work.

I acknowledge the financial support in the framework of the Forschergruppe


404 and the Sonderforschungsbereich 762.

I am very grateful for the opportunity to work in the semiconductor physics


group at the Universitt Leipzig and will always remember the outstanding
teamwork which is both scientifically and personally inspiring. Thank you all!

Last but not least I want to acknowledge the ongoing support, help and moti-
vation I received through my friends and family. Vielen Dank!

189
. Acknowledgement

190
Curriculum Vitae

Address
Matthias Brandt
Mierendorffstr. 52
04318 Leipzig
Germany
Tel: +49 341 97 32 642
Fax: +49 341 97 32 668
Email: m_brandt@uni-leipzig.de

Personal Details
Surname Brandt

Given Name Matthias, Andreas

Date of Birth 10th of May, 1981

Place of Birth Hohenmlsen

Nationality German

Education
early 2010 Award of the doctorate

end 2009 Completion of PhD thesis, prospective title:


"Influence of the electrical polarization on carrier trans-
port and recombination dynamics in ZnO-based het-
erostructures"

since 07/2006 Multiple monthly visits in Lincoln, Nebraska, USA,


in the group of Prof. M. Schubert

since 07/2006 Postgraduate student, Universitt Leipzig,


Fakultt fr Physik und Geowissenschaften,
Semiconductor physics group of Prof. M. Grundmann

191
. Curriculum Vitae

30.06.2006 Award of the Diplom, title of the thesis:


"Elektrische Eigenschaften von Zinkoxid bei erhhten
Temperaturen"
("Electrical properties of zinkoxide at elevated temper-
atures")

20032004 Semester abroad, University of Leeds, UK

20002006 Physics study program, Universitt Leipzig

19992000 Civilian Service in a public hospital,


"Kreiskrankenhaus", Altenburg

19971999 Secondary school:


"Friedrichgymnasium", Altenburg, Award of the Abitur

19911997 Secondary school:


"Veit Ludwig von Seckendorff Gymnasium", Meusel-
witz

19871991 Primary school:


"Polytechnische Oberschule Rosa Luxemburg", Meusel-
witz

Language Skills

German native
English near native
French basic knowledge

192