Zinc Oxide-Silicon Heterojunction Solar Cells Sputtering...
-
Upload
truongngoc -
Category
Documents
-
view
221 -
download
0
Transcript of Zinc Oxide-Silicon Heterojunction Solar Cells Sputtering...
Zinc Oxide- Silicon Heterojunction Solar Cells by Sputtering
by
Jeanne-Louise Shih
A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements for the degree of Master of Engineering
Department of Electrical & Computer Engineering McGill University Montreal, Quebec
Canada
© November, 2007
1+1 Library and Archives Canada
Bibliothèque et Archives Canada
Published Heritage Bran ch
Direction du Patrimoine de l'édition
395 Wellington Street Ottawa ON K1A ON4 Canada
395, rue Wellington Ottawa ON K1A ON4 Canada
NOTICE: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats.
The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
ln compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis.
While these forms may be included in the document page count, their removal does not represent any loss of content from the thesis.
• •• Canada
AVIS:
Your file Votre référence ISBN: 978-0-494-51474-0 Our file Notre référence ISBN: 978-0-494-51474-0
L'auteur a accordé une licence non exclusive permettant à la Bibliothèque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par télécommunication ou par l'Internet, prêter, distribuer et vendre des thèses partout dans le monde, à des fins commerciales ou autres, sur support microforme, papier, électronique et/ou autres formats.
L'auteur conserve la propriété du droit d'auteur et des droits moraux qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Conformément à la loi canadienne sur la protection de la vie privée, quelques formulaires secondaires ont été enlevés de cette thèse.
Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant.
Abstract
Heterojunctions of n-ZnO/p-Si solar cells were fabricated by RF sputtering
ZnO:Al onto boron-doped (1 00) silicon (Si) substrates. Zinc Oxide (ZnO) films were also
deposited onto soda lime glass for electrical measurements. Sheet resistance
measurements were performed with a four-point-probe on the glass samples. Values for
samples evacuated for 14 hours prior to deposition increased from 7.9 to 10.17 and 11.5
0/o for 40 W, 120 and 160 Win RF power respectively. In contrast, those evacuated for
2 hours started with a higher value of22.5 0/o, and decreased down to 7.6 and 5.8 0/o.
Vacuum annealing was performed for both the glass and the Si samples. Current-voltage
measurements were performed on the ZnO/Si junctions in the dark and under illumination.
Parameters such as open-circuit voltage, V0 e; short-circuit current, Ise; fill factor, FF; and
efficiency, 1J were determined. A maximum efficiency of 0.25% among all samples was
produced, with an Ise of 2.16 mA, Voe of 0.31V and aFF of 0.37. This was a sample
fabricated at an RF power of 80 W. Efficiency was found to decline with vacuum
annealing. Furthermore, interfacial state density calculated based on capacitance-voltage
measurements showed an increase in the value with vacuum annealing. The results found
suggest that the interface states may be due to an interdiffusion of atoms, possibly those
of Zn into the Si surface. The Electron Bearn lnduced Current (EBIC) method was used to
determine diffusion length to be at a value ~40-80 ~-tm and therefore a minority carrier
lifetime calculated of 3 ~-tsec. It was also used to determine the surface recombination
velocity (SRV) of the fractured surface of the Si bulk from the fabricated solar cells. An
SRV of ~500 cm/sec was determined from the fractured Si surface, at a point located at
30 and 20 ~-tm away from the junction interface.
Résumé
Des cellules solaires composées de hétérojonctions n-ZnO/p-Si ont été fabriquées
par pulvérisation irradiation à fréquence radio (RF). Les échantillons ont été fabriqué en
pulvérisant du ZnO:Al sur des substrats de silicium dopés de bore. Du ZnO a aussi été
pulvérisé sur des substrats de verre pour faire des études. Les résistances par carré ont été
mesurées sur ces derniers avec une sonde à quatre points. Les résultats mesurés pour les
échantillons qui ont été évacuées dans la chambre à vide pendant 14 heures avant la
pulvérisation ont augmenté d'une valeur de 7.9 à 10.17 et ensuite 11.5 0 /o pour une
puissance RF de 40W, 120W et 160W respectivement. Tandis que, ceux qui ont été
évacués pendant seulement 2 heures ont commencé avec une valeur de 22.5 0 /o, et ont
diminuées à 7.6 et 5.8 0/o. Les substrats en verre et en silicium ont subi des recuits dans
le vide. Les caractéristiques courant-tension ont été réalisées dans le noir et dans la
lumiére. Les paramètres de tension circuit-ouvert (Voe), courant court-circuit (Ise), facteur
remplissage (FF) et efficacité (YJ) ont été établis. Une efficacité maximum de 0.25 %
parmi tous les échantillons a été réussie avec un Ise de 2.16 mA, Voe de 0.31 V et un FF de
0.37. Cet échantillon a été fabriqué avec une puissance RF de 80 W. L'efficacité a baissé
avec les recuits dans le vide. La densité état d'interface a été calculées par les
caractéristiques capacité-tension et montre une augmentation avec recuit dans le vide. Les
résultats suggèrent que les états d'interface pourraient être dû à une interdiffusion
d'atomes, et ceux du zinc vers la surface du slicium. La longueur de diffusion a été
mesuré par la méthode de courant induit par faisceau électronique (EBIC). Cette valeur se
trouve à être ~ 40-80 JLm. La durée de vie des porteurs minoritaires calculée a une valeur
de 3 f.Lm. La vitesse de recombinaison superficielle sur la surface fracturée du silicium sur
11
les cellules fabriquées ont été mesurées deux fois par la méthode EBIC à des point situés
à des distances de 30 et de 20 pm. Les valeurs calculées sont semblables et reviennent à
~500 cm/sec.
lll
Acknowledgements
1 would like to thank Dr. C. K. Jen, Dr. 1. Shih for their continued help and
support throughout the course of my study. 1 would also like to thank the graduate
students in the Electronic Deviees and Materials lab at McGill University for their
encouragement and friendship. I'd finally like to acknowledge the support of my family
for whom this would not have been possible.
lV
Table of Contents
Abstract ................................................................................................................................ i
Résumé ................................................................................................................................ ii
Acknowledgements ............................................................................................................ iv
Table of Contents ................................................................................................................ v
Chapter 1 INTRODUCTION .............................................................................................. 1
1.1 ZnO-Si based photovoltaics ................................................................................ 1 1.2 Theory ofheterojunction solar cells ................................................................... 2
1.2.2 Built-in electric field .......................................................................................... 3 1.2.3 Op ti cal absorption .............................................................................................. 3 1.2.4 Electron ho le generation .................................................................................... 4 1.2.5 Charge separation and collection ....................................................................... 5 1.2.6 Open circuit voltage, short circuit current, fill factor and conversion efficiency
····································································································································· 5 1.2. 7 Recombination ................................................................................................... 8
1.3 Effects of series and shunt resistances ................................................................ 9 1.4 Effects of interface states .................................................................................. 10 1.5 Objective and Outlines ofthe thesis ................................................................. 11
Chapter 2 ZnO FILM DEPOSITION AND GROWTH ................................................... 12
2.1 The princip le of RF sputtering ................................................................................. 12 2.2 The effects of pressure and RF power on the deposition rate and film quality ...... 15
2.2.1 Crystallinity and grain size .............................................................................. 15 2.2.2 Crystallinity vs. op ti cal properties ................................................................... 16 2.2.3 RF power effects .............................................................................................. 17 2.2.4 Argon Pressure effects ..................................................................................... 17
2.3 Sample preparation ................................................................................................. 18 2.4 Principles of film growth ........................................................................................ 21 2.5 Experimental steps ofthe film deposition .............................................................. 23 2.5.1 Experimental conditions ...................................................................................... 24 2.6 Stoichiometric content of ZnO films deposited by RF magnetron sputtering ........ 25 2.7 Experimental results ................................................................................................ 26
2.7.1 Resistivity ........................................................................................................ 26 2.7.2 Effects ofannealing ......................................................................................... 27 2. 7.3 Resistivity measurements for annealed samples .............................................. 28
2.8 Conclusions ............................................................................................................. 32
Chapter 3 1-V CHARACTERIZATION ........................................................................... 33
3.1 Heterojunction barrier formation ............................................................................ 33
v
3.2 Current transport mechanisms ................................................................................ 3 7 3.3 Previous work on ZnO-Si heterojunctions .............................................................. 42 3.4 Effects of Substrate Annealing ............................................................................... 44
3.4.1 Experimental results ......................................................................................... 44 3.5 Conclusions ............................................................................................................. 53
Chapter 4 C-V CHARACTERIZATION .......................................................................... 54
4.1 Capacitance of a heterojunction .............................................................................. 54 4.2 Princip les of interface state density calculations .................................................... 63 4.3 Experimental results ................................................................................................ 64 4.4 Conclusions ............................................................................................................. 82
Chapter 5 ELECTRON BEAM INDUCED CURRENT MEASUREMENTS ................ 83
5.1 Principles ofthe Electron Bearn Induced Current method ..................................... 83 5.2 Principles of diffusion length and surface recombination velocity measurements. 84 5.3 Experimental results ................................................................................................ 89
5.3.1 Characterization ofZnO-Si heterojunctions .................................................... 89 5.3.2 Diffusion length measurements ....................................................................... 90 5.3.3 Surface recombination velocity measurements ................................................ 97 5.5 Conclusions ....................................................................................................... 100
Chapter 6 CONCLUSIONS ............................................................................................ 101
References ....................................................................................................................... 105
VI
Chapter 1
INTRODUCTION
This chapter gives an introduction to the ZnO/Si heterojunction structure and
properties for heterojunction solar cells. Topics such as carrier generation, charge
separation and collection are discussed. As well, the method of characterizing the
performance of a solar cell with parameters such as open circuit voltage, short circuit
current, fill factor and efficiency are presented.
1.1 ZnO-Si based photovoltaics
Zinc Oxide (ZnO), a group II-VI semiconductor is widely applied and has useful
properties to act as a window layer in a heterojunction solar cell. It is considered among
other transparent and electrically conducting films (TCO) as having high conductivity,
low optical absorption and as being resistive to high energy radiation [1.1]. ZnO has a
direct wide band gap whose value is ~3.3 eV at 300 K and a large exciton binding energy
[ 1.1]. Its crystalline structure varies between the zinc-blende, hexagonal wurtzite and the
rocksalt. The thermodynamically stable phase in the ambient is the wurzite whose unit
cell is hexagonal with two lattice lengths 'a' and 'c' and a 'c/a' ratio of 8/3 112• Epitaxial
growth of ZnO films is performed by methods such as magnetron sputtering, chemical
vapor deposition (CVD), molecular bearn epitaxy (MBE) [1.1], spray pyrolysis [1.2], sol
gel, as well as thermal oxidation [1.3]. Magnetron sputtering has been of interest due to
its low cost, simplicity and low operating temperatures [ 1.1]. It is also able to grow highly
oriented films i.e. films of high quality [1.3]. The majority of work related to ZnO films
1
used as solar cells involves the study of n-type ZnO deposited on a p-type material other
than ZnO. The growth of reproducible p-type ZnO films is difficult [1.4] and materials
such as Si, GaN, AlGaN, CdTe, GaAs, etc are used to form the heterojunction with the n
type ZnO instead. The topic of this work is focused on n-type ZnO:Al deposited by the
RF magnetron sputtering technique onto p-type Si.
When producing n-type ZnO, doping it with aluminum (Al) has excellent benefits.
The Znü's optical bandgap widens in proportion to the Al doping concentration.
Furthermore, the material's conductivity, charge carrier density and mobility are
improved [1.5]. It has been reported that maximum conductivities and the highest
mobilities are achieved at Al-doping concentrations of 2-3 at. %. Furthermore, these
parameters are optimized for maximum crystalline sizes [ 1.5].
1.2 Theory of heterojunction solar cells
In a semiconductor heterojunction, one semiconductor is grown onto of a different
semiconductor to form a junction. Each has its own lattice structure with different lattice
constants and are chosen based on their individual and common properties. Conduction
and valence bands are therefore misaligned and when joined form one of the three types
of alignments: straddled, staggered, or broken. Heterojunction energy bands are
characterized by discontinuities at the conduction and valence band edges which
determine carrier transport mechanisms in the deviee. This topic will be further discussed
in chapter 3.
2
1.2.2 Built-in electric field
No matter whether the semiconductor materials being joined are the same or
different, a built-in electric field will result when an n-type and p-type are brought
together to make a junction. The large amount of electrons in the n-type region will have
a tendency to diffuse into the p-type region, giving rise to a diffusion current. The same
occurs for holes in the p-type region. As electrons leave the n-type region, they willleave
behind a fixed, positive, ionized charge in its place. Similarly, holes will leave behind in
the p-type region, fixed, negative, ionized charges. At a certain point, equilibrium is
reached and the donor ions in the n-type region close to the interface will no longer be
compensated by electrons, and acceptor ions in the p-type region also close to the
interface will not be compensated by holes. A charged region, called the depletion region
results in an electrostatic potential barrier being created, and hence an electric field. This
built-in electric field will thereafter help newly generated carriers to drift across the
junction and into the extemal contacts of the solar cell.
1.2.3 Optical absorption
Carriers are generated in the solar cell by photons. Photons have different levels of
energy and wavelengths. When light shines onto a solar cell, sorne of the energy will
reflect off and sorne will be absorbed. In the case of a ZnO/Si heterojunction solar cell,
photons will tend to travel through the highly doped ZnO at the front surface and be
absorbed in the depletion region within the Si. This is convenient since most of the
depletion region, hence the built-in electric field, is in fact contained at the interface, but
within the Si. This is in fact the reason why the highly conductive ZnO is chosen as a
3
"window layer" to be deposited onto Si when fabricating a solar cell. Si has a high
absorption coefficient while ZnO has a low one in the visible region. The absorption
coefficient is therefore a parameter which describes to what extent incident light can be
converted into electricity. Since the goal is to minimize the photon absorption in the
window layer (ZnO) and to maximize it in the Si, it is best to achieve a thin deposition of
ZnO (~0.1 pm). The absorption coefficient of crystalline Si varies between 2xl06 and 1
cm-1 in the spectral range between 300 and 1150 nm [1.6]. The absorption coefficient of
ZnO is in the range of 1-4 cm-1• The absorption coefficient is determined by two factors:
the excitation of an electron from the valence band to the conduction band and the
transitions within a band. The former which is known as the fundamental absorption
occurs when a photon with energy slightly lower or equal to the energy band gap of the
semiconductor transfers energy to an electron in the valence band. The latter type, which
is known as the free carrier absorption, occurs when a photon with much larger energy is
involved. In this case, one electron-hale pair is generated and the excess energy will be
dispersed as thermal energy, becoming a loss. The absorption coefficient o(J...) is the sum
of the fundamental absorption coefficient and the free carrier absorption coefficient.
1.2.4 Electron hole generation
Absorption of light generates electron-hale pmrs via the mechanisms just
described. For a planar slab, a photon absorbed will generate g(x)* ~x electron-hale pairs
in a layer at a depth x- x+ ~x. The generation function g(x) is described by Equation
1.1.
4
g(x) = a(À)exp[-a(À)·x] (1.1)
The generation rate is denoted as Gand is related to g(x) by G = g(x)/A where Ais the
illuminated area of the sample.
1.2.5 Charge separation and collection
Electron-hole pairs generated at a rate G have a finite lifetime 7. In order for the
generated carriers to be carried off into the extemal circuit, an electric field must be
present in order for the free carriers to be separated and drawn out through the front and
back contacts of the solar cell. This is the purpose and reason for the solar cell's built-in
potential and is wh y the conversion of sunlight into an electrical current is possible with a
structure like the p-n junction. The electric field will attract the newly generated electrons
from the p-type region (Si) towards the n-type region (ZnO) and out the front contact and
the newly generated holes from the n-type into the p-type and out the back contact,
creating an extemal current flow. The heterojunction solar cell's top window layer, ZnO,
is a thin, highly doped material, whereas the lower bulk Si is thick and less doped. High
doping in the thin layer of ZnO achieves a high conductivity for separated carriers to be
drawn out through the front contact.
1.2.6 Open circuit voltage, short circuit current, fill factor and conversion efficiency
The equivalent circuit for an ideal solar cell is represented by a current source,
connected in parallel with a rectifying diode. The 1-V characteristic described is shown in
5
Equation 1.2, where Imum is the photogenerated current, k is the Boltzmann constant, T is
the temperature, q is the electronic charge and 10 is the dark saturation current.
(1.2)
Under thermal equilibrium conditions, the diffusion current and the drift current due to
both electrons and holes cancel each other out so that when the solar cell is not
illuminated, no current is induced. Taking a closer look at the diffusion mechanism, only
the electrons in the n region which have enough energy to overcome the built-in barrier
will diffuse across and recombine with a hole in the p region. Under equilibrium, the
recombination process is characterized by an exponential relationship with the built-in
potential. The recombination current is shown in Equation 1.3 where Vbi is the built-in
potential.
- (-q~iJ-lrecomb - lo exp kT - lgen_dark (1.3)
The recombination current, Irecomb, is being compensated by the generation of electrons in
the p-type region. These thermally generated electrons under equilibrium conditions make
up the dark generation current, Igen_dark· It is independent of the barrier height and is
determined by the amount of minority carriers on the p-side. Once the solar cell is
illuminated, the electron concentration in the p-region will increase, increasing the
magnitude of the initial dark generation current. In order for charge neutrality to be
maintained, the recombination current will increase to compensate for the generation
6
current. The electrostatic potential will be lowered by an amount of Yoe, known as the
open-circuit voltage. The generated current expression is shown in Equation 1.4 and is
made up of the dark thermal generation current, lgen_dark and the generated current under
illuminated conditions, lgen_illum·
I I - I (- q(Vbi- Voc)) gen _ dark + gen _ mum - o exp kT (1.4)
The ideal equation relating the photo-generated current with Yoe is shown in Equation 1.5.
(1.5)
The short circuit current Ose) is the measure of the induced current at zero voltage
as a result of electron-hole generation from light. Both the Yoe and Ise are measures of the
solar cell's performance. The standard expressions goveming open-circuit voltage
behavior of n + -p junctions, relating it to the short circuit current are presented in the
following equations,
(1.6)
(1.7)
where, A is the cell area, ni is the intrinsic carrier concentration, DN is the electron
diffusion constant in the bulk, LN is the electron diffusion length in the bulk, p is the
7
-- equilibrium hole concentration at the front of the bulk and q is the electron charge. The
fill factor (FF) is also a very good indicator to the quality in performance as it is the ratio
of the maximum power incurred by the sol ar cell at a voltage V max and current Imax and an
ideal maximum power characterized by the V oc and Ise· A good FF is dependent on a high
mobility of free charge carriers, and low traps. Efficiency is another measure of
performance and is calculated by dividing the maximum power generated by the solar cell
(V max *Imax) over the incident power of the source oflight.
1.2.7 Recombination
There are several ways to categorize recombination mechanisms. One of which
distinguishes between bulk and surface recombination. The other between band-to-band
recombination and recombination that is aided by defect levels within the band gap. Other
methods involve differentiating between radiative and non-radiative recombination.
Surface recombination is related to the defects at the surface or interface. It affects
in particular the dark saturation current and the efficiency of the solar cell. Two methods
of optimizing performance may be undertaken. These are to passivate the surface or
improve the window layer. It is important to note that solar cells may not be completely
passivated with oxide since the current would not be extractable. Interestingly, it has been
reported that during RF sputtering of ZnO onto Si, it has been previously reported that a
thin native oxide layer, with a thickness of a few Angstroms is unintentionally grown.
The effect of this layer during ZnO growth in the sputtering chamber has been previously
discussed by other authors [1.7, 1.8, 1.9, 1.10]. The discussion in this work therefore
8
- extends to understanding the growth process of ZnO during sputtering and how the
stoichiometric content between Zn and 0 is affected at different deposition conditions.
Retuming to the topic of recombination at the surface or interface of a
semiconductor, the investigation of the process is important as it affects the minority
carrier lifetime of the deviee. The lifetime is the time a minority carrier is able to exist in
a sea of majority carriers before it recombines. The diffusion length is the length at which
it can travel within that time. The relation between these is shown in Equation 1.8, where
D is the diffusion coefficient, L is the diffusion length and 7 is the minority carrier
lifetime.
L =-fi).; (1.8)
1.3 Effects of series and shunt resistances
The effect of series resistance is to lower the fill factor value from its ideal. Series
resistance is composed of resistance along the mean current path of the collected carriers.
These may be bulk resistance, front and back contact resistance, grating resistance, etc.
Series resistance may also be due to leakage from junction defects. The fill factor
relationship with series resistance is shown in Equation 1.9, where FF0 is the ideal fill
factor whose expression is shown in Equation 1.10. FF0 is the fill factor ofthe solar cell
with the ideal I-V characteristic described in Equation 1.2 [1.3].
FF = FF0 (l- rs) (1.9)
9
Voc -In(Voc +O. 72)
voc +l (1.10)
Fill factor may also reduce due to low shunt resistance. It has been observed in the
laboratory with the HP parameter analyzer that low shunt resistance decreases the open
circuit voltage whereas high series resistance decreases the short circuit current.
Furthermore, these resistance values may be easily observed from the I-V characteristic.
Specifically, the series resistance is the inverse of the slope of the I-V curve in the 1 st
quadrant whereas the shunt resistance is that in the 3rd quadrant.
1.4 Effects of interface states
Interfacial states or surface states are caused by defects at the interface of a
junction or at the surface of the semiconductor. Under ideal conditions, where interfacial
states do not exist, the junction will ohey the current superposition principle after series
resistance corrections. The presence of interfacial states will results in non adherence and
changes the electrical characteristics of the heterojunction such as a reduction in the open
circuit voltage. The effect of interfacial states may be mode led by band diagrams and may
be observed by performing capacitance measurements. This will be discussed in chapter 4.
Since epitaxial ZnO has a crystalline structure slightly different than that of Si, interfacial
states are likely to exist after ZnO sputtering. ZnO's wurzite structure is characterized by
unit size lengths a= 0.324 nm and c = 0.519 nm whereas Si's has a length a= 0.543 nm
and it is easy to guess the presence of effects from lattice mismatch.
10
1.5 Objective and Outlines of the thesis
The object of this work is to perform fabrication and characterization of n-ZnO/p
Si solar cells that enables the understanding of film growth and its properties which
contribute to energy conversion. The method of ZnO deposition is by RF sputtering since
it is considered a relatively simple procedure that may be performed at low temperatures
and at lower costs. The objective is to present data that may support understanding of the
growth of ZnO via RF sputtering, the effect that sputtered growth has on the junction's
stoichiometric content and how the latter may affect solar cell performance. Further work
related to determine the diffusion length and the surface recombination velocity from a
method called Electron Bearn lnduced Current (EBIC) will also be discussed.
This document is organized as follows. Each chapter starts with an overview of
studies and work done by other authors on a specifie topic. It continues with a description
of measurements performed on fabricated samples of ZnO/Si solar cells and provides
discussions and interpretations. To start, Chapter 2 relates to the method of fabrication of
the ZnO/Si solar cells through RF sputtering. It describes the method of deposition and
the growth action of ZnO. Chapter 3 discusses current-voltage measurements performed
on samples that subsequently go through annealing steps under vacuum conditions. The
topic of stoichiometry is presented. Chapter 4 presents capacitance-voltage and
capacitance-frequency measurements performed a:fter each annealing step and a
discussion on interfacial states. Chapter 5 relates to diffusion length and surface
recombination velocity measurements taken via the EBIC method on samples that have
already gone through all annealing steps.
11
Chapter 2
ZoO FILM DEPOSITION AND GROWTH
This chapter describes the method and principle of ZnO film deposition by RF
magnetron sputtering performed on Si and soda lime glass substrates. The experimental
steps taken for fabricating the n-ZnO/p-Si solar cell are also described here. Previous
work done related to the study of crystallinity and film quality with respect to the change
in sputtering conditions are mentioned in this chapter. Resistivity measurements
performed under different conditions on the ZnO/glass samples are presented. The
purpose is to introduce material that is useful in interpreting results related to the growth
and oxidation of RF magnetron sputtered ZnO:Al films.
2.1 The principle of RF sputtering
Sputtering designates the action of atoms being ejected out from a target as a
result of bombardment from source particles. The purpose is to deposit a layer of the
target material onto the substrate. This is performed under vacuum conditions. The source
of particles is a supply of argon atoms whose flow rate into the vacuum chamber is
controlled by the mass flow controller as shown in Fig. 2.1. Argon atoms pumped into the
chamber will receive energy from the electric field that is being supplied by the RF power
source connected to the target. The target, which is composed of sintered ZnO doped with
Alz03 is held by a water-cooled support. lt acts as the cathode whereas the substrate,
which is suspended above the target and clipped to a grounded Al support, acts as the
anode. An electric field is induced between these two by the RF power source. As Ar
12
-- atoms are pumped into the chamber, collisions result and produce Ar+ ions and electrons
creating a soft purple glow within the chamber. The positive argon ions will be attracted
to the cathode target. The electric field will accelerate these particles onto the target
whereas the electrons will be attracted towards the anode. As the positive argon ions
bombard the surface of the ZnO:Al target, the atoms of the sintered target are knocked
loose. Electrons are sputtered out and are prompted by the electric field to deposit onto
the substrate. It is in this manner that a layer of ZnO:Al is grown onto the substrate.
13
Aluminum support
Vacuum chamber
Target
Mass flow
controller for Argon c=[> .___..JI9'---t
Cooling water c=[> L...----.--------1
RF Power
Diffusion pump Backing pump
Fig. 2.1 A schematic diagram of the RF magnetron sputtering system.
14
2.2 The effects of pressure and RF power on the deposition rate and film quality.
Sputtering conditions may be controlled during deposition to affect deposition rate
and film quality. For example, it is known that RF power regulates the sputtering yield
rate since it controls the rate of bombardment of Ar+ ions onto the ZnO target. [2.1]. In
fact, the higher the power, the higher is the yield rate. In this work, films are deposited at
different RF powers and therefore for higher power levels, the amount of time for which
deposition takes place is shortened. For example, films which were deposited at 40 W
were sputtered for 16 hours, whereas films which were deposited at 80 W were sputtered
for 6 hours.
2.2.1 Crystallinity and grain size
Crystallinity as weil as grain size increase with substrate temperature during
deposition [2.2]; however, at a given substrate temperature, the appropriate growth rate
for which an optimum crystalline morphology can be achieved may be controlled by the
RF power [2.1]. Setting the optimum combination of RF power and deposition
temperature has been shown to yield the highest level of c-axis oriented columnar
structures [2.3]. For example in the work presented by Kim et al. [2.3], among the ZnO
films deposited at 550 oc the optimum power required was 80 W to produce a highly c-
axis oriented columnar structure. They have used sintered ZnO (99.999%) targets with 2
inch diameters [2.4], and a mixed plasma of Ar and 0 2 for the sputtering. It was observed
that with an increase or decrease in power, crystallinity was se en to have degraded. If the
RF power was lower than the optimum value, the sputtered atoms did not have enough
15
energy to migrate to the substrate surface. If the RF power was higher than the optimum
value, the decrease in crystallinity was said to be due to the possibility that the power was
too high and therefore gave insufficient time for atoms to find a stable site for stable grain
growth. It has been concluded that growth quality of ZnO via RF magnetron sputtering is
affected by two variables. The first is the energy supplied by the RF power source as it
regulates bombardment and sputtering. The second is the amount of thermal energy
supplied to the system during the deposition process. An optimum lev el of thermal energy
during deposition at a given RF power being supplied encourages stable deposition and
larger grain size which greatly improves optical properties in the deviee. In the current
work, substrate temperature was not regulated; however the points mentioned above are
important in understanding the changes in film characteristics for changes observed after
post-deposition annealing. Post-deposition annealing is performed under vacuum
conditions for fabricated samples in this work. Annealing is known to achieve lattice re
ordering and increased crystalline quality [2.5].
2.2.2 Crystallinity vs. optical properties
It is interesting to note that optimum quality in terms of crystallinity is not an
indicator of optimum optical performance. In Kim et al. 's work [2.3], although films
deposited at 80 W showed the highest level of crystallinity they demonstrated lower
photoluminescence properties (PL spectra) than those that were deposited at 120 W.
Crystallinity had degraded at 120 W and polycrystalline tendencies were said to have
resulted. These are said to have had an influence on the formation of defects such as
dislocations, vacancy and interstitial defects. The diffusion of defects during the growth
16
process increased. The result was that the density of defects was said to have been
reduced inside the columns themselves as a result of higher power, which improved the
optical photoluminescence property. If deposition temperature was increased to 600 °C,
with power maintained at 120 W, both crystallinity and optical property were
simultaneously improved. The suggestion was made that this would be due to enough
thermal energy being supplied, subsequent! y helping to lead atoms to move to stable sites.
As well, impurities would move to grain boundaries and defect density within the
columnar structures themselves would decrease, resulting in improved PL properties.
2.2.3 RF power effects
At an optimum temperature of 300 oc reported by Das et al. and high pressures,
for an increase in RF power, stronger sputtering action results in better stoichiometry
which transpires into a higher resistivity value [2.6]. Lower power creates oxygen
deficiency and non-stoichiometric films. In this case, oxygen vacancies or interstitial zinc
atoms will act as donor electrons thereby decreasing resistivity. Higher conductivity is
therefore a result of non-stoichiometry and of doping [2. 7].
2.2.4 Argon Pressure effects
Higher pressure serves to decrease ZnO film quality for the purposes of a solar
cell. According to Y oo et al. [2.2], lower gas pressure pressure results in higher
conductivity. In addition, according to Jeong et al. [2.8], combining low pressure (<2
mTorr) with a high substrate temperatures (> 573 K) results in a denser and more
17
compact morphological structure with effective light-trapping capabilities. On the other
hand, when the pressure is increased, ZnO film becomes less crystalline, evidenced by its
increase in resistivity [2.9, 2.2]. An explanation given by Das et al. [2.6] is that precursors
tend to collide severa! times with other gas molecules before they arrive at the substrate.
Other effects that are reported were that the deposition rate of the film was found to
decrease from 105 to 60 A/min with an increase in argon pressure from 0.04 to 1.33 Pa
[2.2]. Grain size as weil has been noted to decrease with increase in argon pressure [2.9].
Overall, a lower pressure is said to produce a more ordered film. As a result, in this work,
pressure has been maintained at a low 5 mTorr for each deposition performed.
2.3 Sample preparation
The heterojunction solar cell sample was prepared on a singled-side-polished p
type (1 00), 1.5 0 -cm Si wafer. The wafers were eut with a diamond scriber into ~4 x 2
cm2 sample sizes. The samples went through the following cleaning procedure. They
were immersed in acetone for 2 minutes, in de-ionized (DI) water for 2 minutes, in
buffered hydrfluoric acid (BHF) for 10 seconds and then in DI water for 10 seconds.
Samples were then spun dry at 3000 rpm for 20 seconds and then loaded into the
sputtering chamber for ZnO deposition. After ZnO deposition was performed on the front
substrate surface, the samples were prepared for the evaporation of Al contact grids. A
thin flexible mask containing the contact grid pattern was mounted on an Al support as
shown in Fig. 2.2. The Al support was then clipped onto the ZnO/Si samples and loaded
into the vacuum chamber. 6 pieces of 1 mm diameter Al wires with a length of about 2
cm were placed on the tungsten heating filament. Evacuation was performed for a
18
minimum of 2 hours. Once evaporation was performed, the samples were removed and 6
drops of photoresist AZ-1827 was deposited onto the ZnO and contact surface. The
photoresist was spun at 3000 rpm for 20 seconds and soft-baked at 90 oc for 10 min. The
photoresist was then exposed through the transparency-printed mask as shown in Fig. 2.3
to an Ultra-Violet source for 6 minutes at 150 W. The samples were then immersed in
concentrated developer for a few seconds and in DI water for 1 minute. Etching was then
performed to remove ZnO from the peripheries and was done with a solution of 10% HCl
diluted in 90% DI water for 2-3 minutes. The samples were then rinsed in DI water for 1
minute and light1y swabbed with a q-tip while immersed in DI water. The samp1es were
then rinsed in DI water' for 10 minutes. Photoresist was then removed by immersing the
samples in acetone. The samples were again rinsed in DI water for 2 minutes and spun
dry at 3000 rpm for 20 seconds. The resulting sample is shown in Fig. 2.4.
19
Fig. 2.2 A photograph of the flexible contact grid mask taped to the Al support.
Fig. 2.3 A photograph of the window mask printed on a transparency
20
Fig. 2.4 A photograph of a finished sample with three n-ZnO/p-Si solar cells and their respective Al contact grids
2.4 Principles of film growth
It has been found by Song et al. [2.1 0] through transmission electron microscopy
(TEM) that during RF magnetron sputtering deposition, an unintentional thin layer of
oxide is grown between the Si and the ZnO layer. In Song et al.'s work, the ZnO:Al/n-Si
heterojunction was formed by RF magnetron sputtering with an RF power of 150 W, Ar
pressure of 0.5 Pa and a substrate temperature of 250 °C. TEM high-reso1ution bright-
field images showed that the thickness of the amorphous Si oxide layer at the interface
was about 12 A. The native oxide growth occurring during RF magnetron sputtering of
ZnO has been observed by other authors and is often referenced [2.11, 2.12, 2.13].
21
It would be a good guess that an oxide layer grown between two non-compatible
materials such as Si (a= 0.543 nrn) and epitaxial ZnO (a= 0.324 nrn and c = 0.519 nrn)
would serve to decrease the high defect density resulting from the lattice mismatch
between these two materials. However, it has been noted in Song et al.'s work [2.10] that
despite it, a rather high density of active defects within the Si depletion layer exists which
is evidenced by a large n, ideality factor. The high ideality factor is the basis for which
the forward bias current transport mechanism has been reported as being dominated by
trap-assisted multistep tunnelling. Current transport mechanisms of ZnO-Si
heterojunctions are discussed in chapter 3.
It has been docurnented that film growth of ZnO tends to an orientation of
crystallites perpendicular to the substrate surface during sputtering. Along this c-axis, the
axis or direction for which there exists the highest rotational syrnmetry, a conductive
channel is predestined [2.14]. It has been shown through lattice images that strictly c-axis
oriented ZnO:Al crystals may be grown on a Si wafer despite the amorphous native
surface oxide at the interface [2.11]. The conclusion is that orientation is not affected by
the crystallographic structure of the substrate and is rather a result of a self-ordering effect
caused by the minimization of the crystal surface free energy and by the interaction
between the deposit material and the substrate surface. This so-called self-ordering effect
is known as self-texturing. It is important to note that during the sputtering process,
conditions must remain stable to achieve a highly ordered and therefore highly conductive
film since it is still possible to change the crystal structure during growth and produce a
highly resistive polycrystalline film by changing the sputter conditions [2.14].
22
Growth of crystalline films depends on the energy conditions of the reactive
components prior to bombardment. It also depends on the surface and the substrate. Three
growth phases have been noted for crystalline films: nucleation, horizontal growth of the
nuclei to a closed layer and a vertical growth of the closed layer. High resistivity
polycrystalline films may be described as largely dominated by growing mechanisms
such as nucleation and horizontal growth of the nuclei to a closed layer such is the case
for Moeller et al.'s reports [2.14] of films with resistivities ranging between 8x1010 and
2x 1011 0 -cm. It may th en be inferred that low resistivity, crystalline films are dominated
by the vertical growth.
2.5 Experimental steps of the film deposition
RF magnetron sputtering of ZnO was performed under vacuum conditions. The
vacuum chamber is connected to a diffusion and backing pump for evacuation, and a
sputtering gun for argon pumping. The sintered target is made of 99.9999 % pure ZnO
and doped with 2 wt % of Ah03 and is held in a metal holder. The target' s diameter
measures 7 cm. The metal holder is attached to a water-cooled support. Prior to vacuum
chamber loading, the Si substrates were cleaned as described in section 2.3. The cleaned
Si substrates and four-finger-probe soda lime glass substrates were then clipped side-by
side onto the Al substrate holder and placed face down at a 45 degree angle from the edge
of the target as shown in Fig. 2.1. The approximate distance between the substrate and the
target is 4-5 cm. A glass slide with four-fingered Au probes were clipped onto the Al
substrate holder along side the Si substrates during deposition. The vacuum chamber was
first evacuated ovemight. Following this, the evacuated chamber is then filled with pure
23
argon. The RF power is then tumed on at 40W for 20 minutes and argon pressure is
maintained at 20 mtorr. This step is done to rem ove contamination from the surface of the
target and ensure that the system is stable prior to actual film deposition. Actual
deposition is performed thereafter under the conditions listed in Table 2.1.
2.5.1 Experimental conditions
Compiled in Table 2.1 is a list of samples that were characterized in this study.
The first set of ZnO films were deposited on four-fingered soda lime glass. They are
identified by an alpha numeric number which describes the evacuation time prior to
deposition and the RF power used during deposition. The second set of ZnO films were
deposited on p-type (1 00) Si wafers. These are identified by a letter showing the length of
time during which dry oxidization was performed prior to ZnO deposition. Round 3
samples were exposed to air for ~3 weeks prior to ZnO deposition.
Table 2.1 A summary of samples fabricated with RF power, deposition time and pressure listed.
Sample no. RFpower(W) Sputtering Pressure (mtorr) Remarks Time (h)
ZnO:Al/soda lime glass
A-80 80 6 5 More uniform than B B-80 80 6 5
A-120 120 4 5 More uniform than B B-120 120 4 5 A-160 160 4 5 More opaque than B B-160 160 4 5
ZnO:Al/p-Si
Round 1-a 40 16 5
24
Round 1- b 40 16 5 Round2-a 120 2 5 Round2- b 120 2 5 Round 3- a 80 6 5 Round3- b 80 6 5
*Ail samples were deposited w1th an Ar pressure of 5 mTorr. **Prior to deposition as described above, Ar pressure was set to 16 mTorr and RF power to 40 W for 20 minutes to decontaminate the target and stabilize the system.
A = evacuated for 14 hrs B = evacuated for 2 hrs a = not oxidized b = oxidized for 15 min
2.6 Stoichiometric content of ZnO films deposited by RF magnetron sputtering
A reported useful method of detecting film quality in terms of stoichiometric
content in Znü is through photoluminescence spectra. If visible emission in PL spectra is
suppressed without suppression in the ultraviolet emission spectra, that is an indication of
a reduction of oxygen vacancies and zinc-interstitial-related defects [2.8]. This is the case
reported by Jeong et al. [2.8] for a Znü film deposited by RF magnetron sputtering using
a ZnO target (99.999%), at a Si substrate temperature of 400 oc and an increase in
lm et al. [2.15] showed that UV luminescence intensity changes are more strongly
dependent on stoichiometry rather than on the crystal's microstructural quality. A
dependency nevertheless exists with respect to the latter. Conclusions resulting from this
work show that with increasing deposition temperature, the grain of c-axis oriented
texture increases in size. Furthermore, it was observed that UV emission was only
observed for ZnO deposited at 550 oc which means that samples deposited at this
25
temperature were found to have the best stoichiometric quality. What is meant by good
stoichiometric quality is that the film con tains as close as possible, a 1: 1 Zn to 0 content
in the film. Interestingly, even though stoichiometric film quality was best for Znü
deposited at 550 oc, the best photocurrents that were observed through 1-V measurements
were for those deposited at 480 °C. This is an interesting result since it shows that
photocurrent depends on something other than or in addition to stoichiometry.
2. 7 Experimental results
2.7.1 Resistivity
Resistivity depends on the crystalline volume fraction in the film, as weil as on the
carrier concentration and mobility [2.9]. For Znü:Al films, two types of donors are
considered. The first originates from Zn interstitials or oxygen vacancies. The second
exists due to substitutional Al atoms. Mobility depends upon impurity and grain boundary
scattering. Steady increase in concentration with doping was observed by Y oo et al. [2.9]
and caused a decrease in mobility. Similarly, for a grain size decrease, grain boundary
density increases, increasing scattering effects thereby lowering mobility. A lower
mobility or a lower carrier concentration results in a lower resistivity film. It has been
noted by Jeong et al. [2.7] that resistivity is also dependent on target-substrate distance as
weil as Al(OH)3 content in the target. When the distance is below 50 mm, resisitivity
remains constant. The lowest resistivity obtained in their study (5.0x10-1 n -cm) was for a
distance of 45 mm. With increasing Al content from 0 to 4 wt%, resistivity decreases
from 5.0 x 10-1 ü -cm for pure Znü to 9.8 x 10-2 ü -cm. According to Jeong et al. [2.13],
Znü deposited on glass substrates showed increase in resistivity with decrease in layer
26
thickness at values below 100-300 mn. The reason for this is possibly due to the different
growth and morphology in early stages ofthe film growth [2.13].
2.7.2 Effects of annealing
Annealing has been performed in previous studies [2.16, 2.17]. For reports of
ZnO:Allayers deposited onto soda lime glass by DC planar magnetron sputtering [2.16],
annealing was performed under vacuum conditions. Resistivity was seen to decrease after
annealing at 250 oc and was attributed to the increase in carrier concentration. Increase in
carrier concentration is a result of the outgassing of oxygen from the film and into the
vacuum ambient. Outgasing of oxygen severs bonds, creating oxygen deficiencies in the
film and frees electrons. The Hall mobility that was measured after the 250 oc annealing
step decreased, as is expected when an increase in carrier concentration is observed. The
resistivity decrease was therefore confirmed to be attributed to the increase in carrier
concentration and dominated by grain boundary scattering. Between annealing
temperatures of 250 and 450 oc, resistivity was reported to continue to decrease, but in
this case, Hall mobility increased. The possible explanation for this is that creation of new
vacancies would have started to cease and instead, the lattice was being changed by the
annealing temperature in such a way as to improve crystallinity, thereby increasing
mobility and hence lowering resistivity. In essence, the measure of resistivity with
annealing has shown to be effective in detecting two parameters of the film: the
stoichiometric content of Zn vs. 0 in the film and the crystallinity via the decrease and
increase in Hall mobility.
27
Interesting is the case of ZnO:Al deposited by an evaporation method where
annealing was performed at temperatures within the range of 27 to 427 oc in air and
vacuum. Resistivity increased by two orders of magnitude for those annealed m air
whereas it decreased by five orders of magnitude for th ose annealed in vacuum.
Also interesting is resisitivity measurements performed for samples deposited at
different deposition temperatures [2.18]. For low temperature, low resistivity values were
obtained with a higher carrier concentration. For 500 oc, resistivity increases and the
compound reaches optimum stoichiometery and the carrier concentration starts to
decrease. At higher temperatures, resistivity starts to decrease and ZnO becomes oxygen
deficient. Two depositions were performed at a substrate temperature of 500 °C. The first
was at an Ar/02 oxygen supply ratio of 4:1 and the second at 6:1. Zn/0 stoichiometry was
observed to be similar between the two with a slight difference in resistivity. lt may be
said that either oxidation of zinc is easier at a higher temperature since oxygen is more
reactive or it is possible that at a higher temperature, the native oxide is thicker and
becomes a source of 0 to the ZnO being sputtered and therefore a source of SiOx defects.
2. 7.3 Resistivity measurements for annealed sam pies
Sheet resistance measurements performed on ZnO sputtered onto soda lime glass
for this work are shown in Fig. 2.6 and 2.7. These glass samples have been sputtered at
the same time as their Si counterpart and have been annealed under vacuum conditions, as
well at the same time. Fig. 2.5 shows the sheet resistance measurement setup performed
after each annealing step, where step 0 represents measurements taken for as-deposited
ZnO on the glass. Measurements were taken with a four-point-probe and the HP 4145A
28
parameter analyzer as shown in Fig. 2.5. Equation 2.1 shows the relationship between V
and I taken from the parameter analyzer, where Rs is the sheet resistance and p is the
resistivity of the Znü film. S is the spacing between the contacts and Le is the length of
the contact fingers. S has been measured with a microscope to be 0.2 mm and Le is 5 mm.
Rs = Le . V = Le . p (Q. 1 sq) s 1 s
Glass with ZnO layer
(2.1)
1
Fig. 2.5 Four-point-probe measurement method for the resistivity of Znü sputtered onto soda lime glass.
29
In Fig. 2.6, the trends for 40 W and 120 W are similar; however, for the 40 W
sample, sheet resistance slightly decreases after the 200 oc annealing step and increases
steadily thereafter. As noted above, a decrease is indicative of oxygen outgassing. The
sample fabricated at 120 W increases in sheet resistance with each annealing step. The
trend for the ZnO deposited at 80 W on the other hand has a much larger decrease in sheet
resistance until the annealing step at 300 oc.
In Fig. 2. 7, sheet resistance was also measured for samples that were deposited at
different RF powers and were evacuated in the vacuum chamber prior to ZnO deposition
either for 2 hours or 14 hours. The sheet resistance trend for samples which were
evacuated for 2 hours shows a decrease with RF power, whereas those which were
evacuated for 14 hours show an increasing trend. Since a decrease in sheet resistance has
been noted to be associated with an outgassing of oxygen during vacuum annealing, it
may be qualitative! y said that the sample which was evacuated in the vacuum chamber for
only 2 hours had higher oxygen content. In other words, an RF power increase results in
an increase in the sheet resistance of samples which have lower oxygen content. The
opposite is true for samples which appear to have higher oxygen content.
Furthermore, the difference in resistivity measurements might be an indication
that the native oxide, which has been mentioned in other reports to grow during RF
sputtering, is affected by the amount of time that evacuation is performed. This has not
been verified in this work.
30
Sheet resistance for ZnO on soda lime glass
70
60
_.,_ ZnO deposited at 40 W _.,._ ZnO deposited at 80 W __,._ ZnO deposited at 120 W
0 +-------~--------~--------~-------,
0 2
Annealing step
3 4
Fig. 2.6 Sheet resistance for ZnO deposited at 40, 80 and 120 Won soda lime glass. Each annealing step lasts 30 minutes and is performed under vacuum conditions. Step 0 is the as-deposited sample. Step 1 annealing was performed at 200 °C, step 2 at 300 oc, step 3 at 350 oc and step 4 at 400 oc.
Sheet Resistance vs RF power 24~------------------------------~
22
20
(1) 18 u_ c: (1) 16 ~ ... 1i) ~ 14 ·- :::1 t/) C" 12 (1) t/)
o::: "E 10 -J:: (1) 0 8 .ê-(1) 6
4
Ill
2
0~~--~--~~--~--~~--~--~~ 70 80 90 100 110 120 130 140 150 160 170
RF Power(W)
.....-sputtering time: 6h- Evacuated for 2h -111- Sputtering ti me: 6h- Evacuated for 14h --.- Sputtering ti me: 4h- Evacuated for 2h
"' Sputtered ti me: 4h- Evacuated for 2h -•- Sputtering ti me: 4h- Evacuated for 14h -1!1- Sputtering ti me: 4h- Evacuated for 14h
Fig. 2.7 Sheet resistance for soda lime glass evacuated in the vacuum charnber for 2 or 14 hours before ZnO deposition at 40, 80 and 120 W.
31
2.8 Conclusions
It has been shown that resistivity has increased continuously for ZnO sputtered
onto soda lime glass at high RF power (120 W) with vacuum annealing. Film deposited at
80 W showed a large initial decrease in resistivity after 200 and 300 oc annealing and
then a continued increase thereafter. The film deposited at 40 W showed a very slight
initial decrease after the 200 oc annealing step and a graduai increase in resistivity,
similar in trend to the film deposited at 120 W.
It has also been shown that resistivity measurements of ZnO films sputtered onto
soda lime glass samples are characterized by an increase in sheet resistance with an
increase in RF power for samples which were evacuated for 14 hours prior to deposition.
The opposite is true for samples which were evacuated for only 2 hours prior to
deposition. The decrease in sheet resistance for samples only evacuated for 2 hours prior
to deposition is probably due to outgassing of O. The difference in resistivity trends is a
possible indication that the amount of time spent on evacuation prior to ZnO deposition
affects the growth of native oxide during reported by others for the sputtering process.
Specifically, it is possible that traces of water and oxygen reside on the surface and a
longer evacuation period is required to remove these. It would be interesting to examine
native oxide thickness via Transmission Electron Microscopy (TEM) and to examine the
change in quality with annealing via Hall mobility and resistivity measurements with
different evacuation conditions prior to sputtering.
32
Chapter 3
1-V CHARACTERIZATION
This chapter describes the current-voltage (1-V) characterization results of n
Znü/p-Si solar cells with the ZnO layers prepared by the RF magnetron sputtering
method. It describes previous work done by others and also presents results achieved in
this work. The purpose is to understand transport mechanisms in the deviee. As well, it is
to understand the changes in the 1-V curves and therefore the photoelectrical performance
due to high temperature annealing under vacuum conditions.
3.1 Heterojunction barrier formation
The energy band dia gram of an ideal ZnO/Si heterojunction is shown in Fig. 3 .1.
Features of the band alignment are determined based on the Anderson energy-band model.
It is worth noting that the Anderson model neglects parameters such as lattice mismatch
and interface traps; however it is useful in sketching the heterojunction band alignment
showing ideal band edge differences. Known also as the electron affinity model, the
Anderson model has been used to model ZnO/Si and ZnO/SiC heterostructures [3.1, 3.2]
and is based on the energy equation x1 - ~Ec- Xz =O. This represents the fact that the
energy of an electron moving from the vacuum level to the conduction band of
semiconductor "l" to the conduction band of semiconductor "2" and back to the vacuum
level must be zero. The conduction band discontinuity is therefore the difference between
the two electron affinities, ~Ec = Xzno - Xsi where Xzno and Xsi are the electron affinities of
ZnO and Si. Similarly, the valence band discontinuity is ~Ev= ~Ec + ~Eg, where ~Eg is
33
the energy band gap difference between ZnO and Si. Two sources have quoted values for
L1Ec and L1Ev to be 0.4 eV and 2.55 eV respectively [3.3, 3.4] following the above
equations for conduction and valence band discontinuities, where the ZnO work function
is assumed to be 4.45 eV [3.1, 3.3], though other sources have stated other values [3.2]. In
fact, work function and electron affinity for ZnO have shown to vary in the literature.
Sundaram et al. have determined values of work function of sputter deposited ZnO films
via the Shottky barrier mode! for ZnO/p-Si and ZnO/n-Si junctions, including the
assumption of the presence of a very thin Si02 layer at the interface. The values they
found were to be between 4.45 and 4.5 eV. For the purposes ofFig. 3.1, a value of 4.45 is
used.
With respect to ZnO films, work function is assumed to be nearly equal to the
electron affinity [3.3]. This is due to the fact that doped ZnO films are highly conductive.
The assumption is that the Fermi leve! coïncides with the conduction band edge in the
ZnO, hence work function is considered almost equal to the electron affinity [3.3]. The
electron affinity therefore also has a value ofranging between 4.45 and 4.5 eV, according
to Sundaram et al.'s evaluations. In the case of Si, the electron affinity is weil known to
be 4.05 eV [3.3]. L1Ec, according to the above equation is therefore equal to ~0.4 eV and is
schematically pointed out in Fig. 3 .1. L1Ev is of a much larger value of 2.55 eV which is a
great indication of wh ether current density is dominated by electron or ho les during non
illuminated conditions. This point will be discussed shortly.
There are two shortcomings with the Anderson mode! [3.4]. Specifically related to
the modeling of ZnO/Si junction, the first is of the lack of interface trap and lattice
mismatch considerations to determine band edge differences. This has been addressed by
34
Ruan and Ching [3.5]. According to them, the valence band discontinuity is changed due
to the formation of the junction interface, which they have modeled as a newly formed
effective dipole at the interface. Band discontinuities ~Ec and ~Ev are affected by this
dipole effect. Interestingly however, for lattice mismatched junctions, the dipole effect
induced by electron transfer is said to be destroyed or severely altered and reduced when
there is a large amount of localized interfacial defects, such is assumed to be the case for
the ZnO/Si heterojunction. Furthermore, deviations in electron affinities due to dipole
layers has also been previously noted to be small (1 %) by Shay et al [3.4]. The second
shortcoming with the Anderson model is that it neglects electron correlation effects.
Correlation effects arise when an electron is moved into the vacuum level. The
surrounding electrons will rearrange themselves to reduce the total energy of the electron
system.
Considering the fact that there is very high lattice mismatch between ZnO and Si
and therefore that a high defect density is assumed to exist, the dipole theory is neglected
here. Whether correlation effects are small for ZnO/Si is not determined in this work;
however, according to [3.4], the magnitude of correlation effects is generally small.
Anderson modeling is therefore validated for determining band edge differences in the
case ofhighly mismatched ZnO/Si heterojunction.
In terms of the heterojunction band lineup shown in Fig. 3.1, ZnO/Si is
characterized as a type II structure [3.4, 3.6]. This means that the conduction and valence
bands of ZnO and Si are staggered with respect to each other as is shown in Fig. 3.1. The
staggered structure clearly demonstrates the location and magnitude of the conduction
band and valence band discontinuities ~Ec and ~Ev.
35
As discussed previously, the conduction band discontinuity is much smaller
compared to that of the valence band discontinuity. The result of this is that majority
carriers (electrons) in the ZnO will not be as greatly impeded from flowing across the
junction as holes from the Si are during forward biasing. Under non-illuminated
conditions, the forward bias current density is therefore largely dominated by electrons.
Once the junction is illuminated, specifically by low energy photons, photo-generated
electron-hale pairs will be created in the depleted p-Si region. The depletion region is
extremely narrow for the ZnO region since ZnO is highly doped and highly conductive.
As discussed, the Fermi leve! is assumed to coïncide with the conduction band edge.
Generated electrons will start to flow down the potential barrier whereas generated holes
will have a harder time doing so towards the ZnO region as is shown in Fig. 3.1. It is clear
that under illuminated forward biasing, two competing carrier flows are present, one
resulting from the forward bias, the other from the photogeneration. In fact, the
illuminated I-V curves, as may be seen in Fig. 3.2 a, show two oppositely directed current
densities. Positive current appearing on the I-V curve is largely the result of the injection
of electrons coming from the n-ZnO, overcoming the potential barrier and diffusing into
the p-Si. Negative current on the I-V curve is the result of the electron-ho le generation
occurring in the depleted region of Si, due to low energy photon illumination as is shown
in Fig. 3.1.
Since it has been documented that an unintentional thin oxide layer results from
RF magnetron sputtering deposition of ZnO [3.13.7, 3.8, 3.9], it has been reported that
photocurrent and carrier flow across the junction is affected depending on the thickness
and the quality of the junction. [3.9, 3.10]
36
Ec
Ef Ev
Vacuum Level
\ ~Visible light
r Photo- electron-hole pair generation fê\ - D~
Forward bias electron Dow
1 0-->f---1-.--,_ AEc
'."-® Eg= 1.12 eV
1
l
~\ p-Si
.o. Ev Eg= 3.27 eV
n-ZnO \
XZno = 4.45-4.5 eV
Fig. 3.1 A schematic of the n-ZnO/p-Si band dia gram including the depiction of photogeneration and forward bias electron flow.
3.2 Current transport mechanisms
The dominant current transport mechanism across the n-ZnO/p-Si heterojunction
at intermediate forward bias (~0.25- 0.3 V) has been determined to be via trap assisted
37
multistep tunneling [3.7, 3.11]. This has been proven by the observation of large ideality
factors (~n=3) that are independent oftemperature [3.7, 3.11].
The ideality factor, shown in equation 3.1 is often used to characterize the
relationship between the 1-V curve and transport mechanisms.
(3.1)
An ideality factor of n = 1 is an indication of the presence of mechanisms such as thermal
generation, minority carrier injection and recombination. Current and voltage relationship
for common diode rectification behavior is generally modeled by J a [exp(qV/nkT)]
where k is the Boltzmann' s constant and T is the temperature. Ideality factors for
ZnO/Si junctions have been observed to be larger than 1 and do not vary with temperature
therefore mechanisms mentioned above no longer apply. In fact, the new relationship
between current and voltage is J = J0[exp(Ao·V)] where A0 is a constant independent of
temperature. Whether simple tunneling is a possible dominant mechanism was ruled out
by Song et al. in the case of the n-type Si substrate sample i.e. n--Si/Si02/n+ZnO:Al [3.7]
since the Si is lightly doped (~5x1015 cm-3) which means that the barrier created at the
depleted Si region is considered too thick for simple tunneling to occur. In the case of a p
Si/n-ZnO heterojunction presented in the work by Song et al. [3.11], multi-step tunneling
is inferred to be the dominant mechanism, the evidence being those mentioned above: an
ideality factor larger than 1 and a forward-bias current-voltage relationship that is
independent of temperature. Simple tunneling is not mentioned by Song et al. for this p
type Si substrate case.
38
At lower forward bias voltages ( <1 00 rn V), Song et al. have reported that current
is due to a shunt resistance of about 5.1 kO cm2 and that it is due to series resistance
effects at large forward voltages (>400 rn V). Dhananjay et al. [3.15] and Song et al. have
reported similar results in ideality factor values. Dhananjay et al. have however grown Zn
films on cleaned Si and thermally oxidized them to create crystalline ZnO films.
Dhananjay et al. have described the low forward voltages to be dominated by
recombination within the space charge region. They have deposited films from a Zn metal
target (50 mm diameter) through DC sputtering inside a vacuum chamber. DC power was
fixed at 35 W and Argon gas was used as the sputter gas. Deposition was carried out at
200 oc at 100 rn Torr. Subsequent oxidation was performed in a tube furnace under 0 2
flow at temperatures ranging from 300 to 500 oc. They have not reported the presence of
an oxide interlayer.
Both Dhananjay et al. and Song et al. show very similar trends in ideality factor.
That is they report an ideality factor (n) of around 2.8 for a temperature of ~350 oc and a
value of n ~ 1.4 for ~ 250 °C. In fact, the description made by Dhananjay et al. that the
change in ideality factor with temperature as being affected by recombination within the
space charge region is equivalent to Song et al.'s description that a shunting effect is
present. In summary, under dark forward bias, sorne electrons that flow from the ZnO
region in Fig. 3.1, up the potential barrier into the Si depletion region will fall into
interfacial traps and recombine. These electrons will not be able to contribute to the dark
forward current and therefore, the recombination of these trapped electrons is equivalent
to current being drawn by a shunt resistor. This is the effect described for low forward
voltages.
39
.~. Song et al. describe three regions in the ideality factor graph. The low forward
bias region characterized by shunting, the mid-forward bias region characterized by
constant n with change in temperature and the high forward bias region where the ideality
factor increases. Song et al. describe the high-forward bias region as being due to the
series resistance effect coming from, bulk resistances in the Si and the ZnO, as weil as the
resistance associated with tunneling through the Si depletion layer and the interfacial
oxide. [3.7]
Interestingly, although Dhananjay et al. and Song et al. have reported similar
ideality factor trends, Dhananjay et al. have not reported the presence of an oxide
inter layer.
Another report of the presence of a thin interlayer of oxide is provided by Sieber
et al. [3.12] whereby ZnO:Al films were prepared by reactive DC magnetron co
sputtering from Zn and Al targets in an Ar/Oz gas mixture on Si substrates. Si substrates
were cleaned with HF dip prior to deposition. Unlike Dhananjay et al. who have used Ar
as the sputtering gas, Sieber et al. have used Ar/Oz gas mixture. The result is the evidence
of an amorphous Si oxide layer. They state that Si surfaces that are oxygen free such as
those that have H-terminated surfaces also oxidize during the supply of oxygen or oxygen
containing components. Since Dhananjay et al. have deposited Zn films by sputtering
without Oz gas, have performed oxidation after the Zn deposition and have not reported
an oxide interlayer; it is interesting to note that ideality factor trends are still similar to
those by Song et al. for whom an oxide interlayer was reported.
40
The above has characterized the different current transport mechanisms reported
by other authors for the forward bias case. Reverse bias current was shown by Song et al.
to be direct1y proportiona1 to (Vbi-V)112 [3.11] where Vbi is the built-in voltage and Vis
the app1ied voltage, and by extension is directly proportional to the depletion width of the
junction [3.11]. Reverse bias current is therefore dominated by thermal carrier generation.
The location of electron-hole pair generation depends on the photon energy i.e.
whether high or low energy photons are used to illuminate the ZnO/Si junction. In work
developed by Jeong et al. [3.1 ], it was found that low-energy visible photons transmit
through the ZnO layers and are absorbed mainly in the depletion region of the p-Si where
electron-hole pairs are generated. On the other hand, for high energy U.V. light, photons
were absorbed in the depleted n-ZnO region.
It is worth noting that when photon energy increases, the penetration depth of
photons will decrease rapidly. Therefore when the depletion thickness gets larger
compared to the light penetration depth, the photogenerated current level will saturate
which is exactly what is observed on the 1-V curve when a large reverse bias voltage is
applied.
Fig. 3.1 is based on this model. Assuming the presence of a thin layer of oxide at
the interface, in the case where visible photons were involved and generation occurred at
the depleted Si region, photoelectrons were said to encounter the thin oxide barrier on
their way to the front contact whereas holes did not. The total photocurrent density is
therefore dependent on and dominated by hole current density. According to their
equation, it is also dependent on the square root of the potential drop across the depleted
p-Si region. In the case of high energy photons, generation occurs in the depleted ZnO
41
region and photogenerated holes are the ones that encounter the oxide barrier and travel
through the p-Si to reach the back contact. The photocurrent in the high energy case is in
contrast, dependent on the electron density and according to their equation, the voltage
across the n-Znü depleted region. The conclusion is that under the assumption of an
oxide barrier layer at the interface of the ZnO/Si junction, carrier generation occurs in the
depleted region in Si causing current density to be dominated by hole flow when low
energy photons are illuminating the junction whereas high energy photons cause carrier
generation in the depleted region of ZnO resulting in current density being dominated by
electron carrier flow.
3.3 Previous work on ZnO-Si heterojunctions
According to Kim et al. [3.13, 3.14], crystallographic morphology is not the most
influential factor in photoelectrical performance. Poor photoelectric effects have been
observed in [3.3.10] for ZnO films deposited on Si at 550 oc by the sputtering method
that had shown the best stoichiometric quality. Interpretations include the possibility of
defectiveness in the oxide layer since the reaction between Si and oxygen gases is more
probable at higher temperatures. A defective Siüx layer may be the source of large
leakage currents. Lee et al. [3.9] propose that the layer may be a mixture of SiOx-ZnO
that could be acting as light-absorbing media. The conclusion is that for the purposes of a
photodiode, the quality of the diode junction is as important as that of the ZnO film itself.
Furthermore, Kim et al. [3.14] report that since crystallinity is shown to be the highest for
films deposited at 600 °C, and that photocurrent degrades, it is concluded that the overall
photoelectric performance of diodes is not directly proportional to the crystallinity of the
42
ZnO films. Although crystallinity is considered a factor in terms of achieving good
photoelectric performance, it is obviously not sufficient in terms of achieving an optimum
performance. Stoichiometry is the important factor and has been determined to be
improved by optimizing process conditions such as substrate temperature and/or Ar/02
ratio. Increasing the deposition temperature makes the heterojunction cell oxygen
deficient [3 .14]. lt has been shown by Kim et al. that films deposited at 600 oc have the
lowest resistivity values which have been determined to be a direct result of the oxygen
deficiency. According to 1-V measurements shown by Kim et al., low resistivity and
therefore oxygen deficiency may be observed through an increase in dark current leakage.
In fact, Lee et al. [3 .1 0] show that dark leakage current gradually increases with
deposition temperature. The phenomenon is first guessed to be attributed to the formation
of a defective interlayer between n-ZnO and p-Si. Further analysis with X-ray
photoelectron spectroscopy (XPS) shows that the interlayer is Si02 which was clearly
unintentionally grown since the Si substrates were cleaned in buffered HF (HF:H20 = 1:
7) for 1 minute, prior to ZnO deposition. Probing of the junctions by X-ray Photoelectron
Spectroscopy (XPS) shows that the thickness of the insulating layer grows with
deposition temperature [3.10]. The overall conclusions derived from the literature indicate
that an increase in deposition temperature will i) increase the interfacial oxide thickness
and ii) lower resistivity, an indication that the film is oxygen deficient. The latter may
also be observed through an increase in dark current leakage.
43
3.4 Effects of Substrate Annealing
When deposition temperature is increased for an appropriate level of RF power, as
discussed in chapter 2, it supplies thermal energy to the growth process, helps to
encourage stable deposition and larger grain size. Post-deposition annealing provides
thermal energy and in this work, has shown to improve photoelectric performance for
temperatures below and up to 350 oc for Round 2 samples which were deposited at 120
W. A subsequent annealing at 400 °C shows a decrease in photoelectric performance.
This is an interesting result since Chapter 4 will present an increasing trend of interfacial
state density with annealing up to 350 oc for these samples.
3.4.1 Experimental results
In this section, 1-V results obtained in this work from the ZnO-Si heterojunctions
prepared by RF sputtering will be described. It is noted that annealing in this work was
performed under vacuum conditions. The glass and Si heterojunctions were
simultaneously loaded into a glass tube connected to the vacuum system and evacuated
ovemight prior to annealing. Annealing was performed at 200, 300, 350, 400 oc for 30
minute intervals each. Three rounds were performed for ZnO films deposited at different
RF powers, namely, 40 W, 120 W, and 80 Win this order. The Si substrates which were
used to deposit ZnO at 80 W remained in air for an approximate period of three weeks
prior to deposition. Furthermore these fabricated samples were annealed at the 350 oc
round twice, for 30 minutes each.
The 1-V characteristics were measured with a HP 4145A Semiconductor
Parameter Analyzer after each annealing step. The results are shown in Fig. 3.2. The light
44
source used to illuminate the sample during measurements has a power density
approximated to be about 100 mW/cm2• Fig. 3.2 a shows the dark leakage current for the
as-deposited sample to be close to zero. Annealing steps performed at 200, 300 and 350
oc increase the leakage current whereas the subsequent annealing step performed at 400
°C decreases it. Leakage current in the dark reverse biasing situation is generally a result
of thermal generation. Thermally generated electrons are attracted to the n-side whereas
thermally generated holes are attracted to the p-side. Leakage current in the forward
biasing situation results from the trapping of carriers in interfacial states as is described in
page thirty nine of this thesis. Trapped carriers which recombine do not have the
opportunity to contribute to dark forward current. The increase in dark reverse-bias
leakage with annealing is possibly an indication of an increase in trapping effects
resulting from Zn having diffused into the oxide layer. Fig. 3.2 b shows a continued
increase in dark leakage current up until the 400 °C annealing step for the Round 2,
oxidized sample. The interpretation is however qualitative in nature and requires further
experimental analysis.
Table 3.1 lists the measured results of the short circuit current (Ise), open circuit
voltage (V oc), fill factor (FF), efficiency ('YI). Calculations for efficiency were made with
equation 3.2 and for fill factor, equation 3.3 where V max and Imax are the voltage and
current at the maximum power point.
Maximum power density 17 = Power of incident light
45
(3.2)
(3.3)
-$ -0.5 -s:: e ._ ::s 0
Round 2 ZnO-Si not oxidized
2.00E-03
1.00E-03
-4.00E-03
Voltage (V)
Fig. 3.2 (a)
-+- dark at 400 degree anneal
--400 degree anneal
0.5 350 degree anneal
300 degree anneal
_,_ 200 degree anneal
--As-deposited -- dark as-deposited
-- dark at 350 degree anneal
Round 2 ZnO-Si oxidized
2.00E-03
1.00E-03
-2.00E-03
-3.00E-03
-4.00E-03
Voltage (V)
Fig. 3.2 (b)
46
--+- dark at 400 degree anne al -- 400 degree anneal '* 350 degree anneal
300 degree anneal _,._ 200 degree anneal --As-deposited _,._ dark as-deposited
< --c a.t ::::: -0.5 ::::J 0
-<C --c ~ -0.6 ::::s 0
Round 3 ZnO-Si not oxidized
3.00E-03
2.00E-03
1.00E-03
-3.00E-03
Voltage (V)
Fig. 3.2 (c)
--+- dark at 400 degree anneal --- 400 degree anneal
350 degree annealed twice 300 degree anneal
0.5 --200 degree anneal
--As-deposited -- dark as-deposited
Round 3 ZnO-Si oxidized
3.00E-03
2.00E-03
1.00E-03
-3.00E-03
Voltage (V)
Fig. 3.2 (d)
--+- dark at 400 degree anneal --400 degree anneal ~·CM· 350 degree annealed twice
300 degree anneal _,._ 200 degree anneal
--As-deposited " "dark as-deposited"
Fig. 3.2 Plots ofl-V characteristics offabricated samples for which ZnO was deposited at 120 W (a, b) and 80 W (c, d). Illuminated and dark currents are shown.
47
For samples that were deposited at 120 W (Round 2), in the case for the non-
oxidized sample, Ise, FF and 'YI improve with every annealing step as shown in Table 3.1.
On the other hand, for the sample that was oxidized for 15 min prior to ZnO deposition,
Ise and 'YI are highest as-deposited whereas the FF is highest after the 350 °C annealing
step and the V oc is highest after the 200 °C annealing step. Round 3 samples which had
not been oxidized showed a decreasing trend with annealing for ali four parameters.
Th ose which had been oxidized had maximum values for Ise, V oc and 'YI in as-deposited
films; however the FF was best after the 200 oc annealing step. Round 3 samples were
exposed to air for ~ 3 weeks prior to ZnO deposition. Efficiency values for Round 3
samples were of one order of magnitude higher than for Round 2 samples. The highest
values in ali four parameters were achieved for the Round 3 samples which had not been
oxidized prior to deposition.
Table 3.1 A summary oflsc, V oc, FF and 'YI for fabricated samples after each annealing where the maximum values are shaded.
Sample Annealing step: Ise (mA) V oc FF 'YI(%) no.
Temperature COC) Time (V)
Round2 As-deposited 0 0.62 0.16 0.018 not 200 30 1.50 0.16 0.23 0.054
oxidized 300 30 1.49 0.16 0.29 0.069 350 30 0.15 400 30 0.14
Round2 As-deposited 0 0.18 oxidized 200 30
300 30 350 30 400 30
Round 3 As-deposited 0 not 200 30 1.63 0.27 0.34 0.15
oxidized 300 30 1.81 0.22 0.26 0.10 350 30 0.03 0.09 0.31 0.0009
48
,- 400 30 0.01 0.00 0.000008 Round 3 As-deposited 0 0.27 oxidized 200 30 1.22 0.25 • 0.10
300 30 0.68 0.24 0.22 0.04 350 30 0.004 0.07 0.28 0.00 400 30 0.003 0.05 0.29 0.00
The ideality factors shown in Fig. 3.3, for both oxidized and non-oxidized samples
start from a value ofn ~ 1-2 at a bias ofless than 0.1 (shown as q/k:T*V = 5 where kT/q =
0.026) and reaching as high as n ~ 19 for voltages of~ 0.7 V. The most significant
changes in the ideality factors with vacuum annealing is observed at the higher forward
bias region, apparent in Fig. 3.3 b and 3.3 d. Low voltages however display only slight
changes in ideality factor values with annealing, especially so for the non-oxidized
sample. Mid forward bias voltages show a region between ~0.25 - 0.3 V that has lower
ideality factor values than lower or higher voltages which is similar to the results
observed by Song et al. (3.7]. This region has current density that is modeled by the
equation J = J0[ exp(Ao·V)] where Ao is a constant independent of temperature. It can be
initially inferred that prior to any annealing step, current transport mechanism is
dominated by trap assisted multistep tunneling in this forward bias voltage range.
Subsequent annealing will change perhaps the oxide thickness and therefore change
current transport mechanism and the idea that current transport mechanism is dominated
by multi-step tunneling may no longer be fully useful after annealing.
An increase in ideality factor values is nevertheless apparent from Fig. 3.3 b and
3.3 d with annealing. For non oxidized samples, ideality factors increase to maximum
values after the 350 oc annealing step and come back down close to the original values
49
with 400 oc annealing. The oxidized samples however show continued increase in
ideality factor values until the 350 oc annealing where the next 400 oc annealing step
does not show any significant change.
At high forward bias voltages large increases in the ideality factors are observed
with annealing, except after the 400 oc annealing step. The increase may be an indication
of an increase in either one of the following variables: bulk resistances in the Si and the
Znü, as weil as the resistance associated with tunneling through the Si depletion layer
and the interfacial oxide.
Annealing also seems to increase the reverse bias current, indicative of an increase
in leakage current at the junction. As previously mentioned, leakage current has been
reported to be due to a deficiency in oxygen as is the case when deposition temperature is
increased [3.14]. In this work, annealing is performed under vacuum conditions, which
leads to oxygen out-gassing and therefore a deficiency in oxygen. Leakage may be due to
carrier trapping due to increased interfacial state density. This topic will be discussed in
Chapter4.
50
10.0
5.0
::::::-z -5.0 ..J
·10.0
-15.0
·20.0
16.0
14.0
t: 12.0 ,_ 0 10.0 -u CIS
8.0 .... >. ~
6.0 ëij C1l E 4.0
2.0
0.0
0.0 0.1
Round 2 - Oxidized ZnO-Si
1/n factor
q/KT *V
Fig. 3.3 (a)
30.0
" Non ideal: Alter 200 degree anneal
Non ideal: Alter 300 degree anneal
~ Non ideal: Alter 350 degree anneal
---Non ideal: Alter 400 degree anneal
• n=2
Round 2- Oxidized ZnO-Si ldeality factor values
0.2 0.3 0.4 0.5 0.6
Voltage (V)
Fig. 3.3 (b)
51
0.7
· As-deposited
" After 200 degree anneal
After 300 degree anneal
• After 350 degree anneal
_,._ After 400 degree anneal
10.0
-z ....
16.0
14.0
1: 12.0
.... 0 10.0 -u «1
8.0 ~
>-~
'ii 6.0 (1,1
:E 4.0
2.0
0.0
0.0 0.1
Round 2 - Not oxidized ZnO-Si
1/n factor
q/KT*V
Fig. 3.3 (c)
• n=6 ·• n=19
Round 2- Not Oxidized ZnO-Si ldeality factor values
0.2 0.3 0.4 0.5
Voltage (V)
Fig. 3.3 (d)
0.6 0.7
---- As-deposited
-- After 200 degree anneal ' After 300 degree anneal
" After 350 degree anneal _.,_ After 400 degree anneal
Fig. 3.3 Plots ofln(I) vs. 0.026*V for Round 2 oxidized (a) and non-oxidized (c) samples. Plots of n vs. V are shown in (b) and ( d) respective! y.
52
-' '
3.5 Conclusions
Efficiency has been observed to increase for the Round 2 non oxidized sample
with annealing up until the 350 oc annealing step, but decreases with the subsequent 400
oc step. On the other hand, samples that were exposed to air prior to ZnO deposition
showed no efficiency improvement with annealing even if they had been intentionally
oxidized prior to air exposure, perhaps an indication that the samples contained too much
oxygen components. Ideality factors have shown significant increase in the higher
forward bias voltage region with annealing. Only small increases are seen in the low
forward bias voltage region. Ideality factor value curves consistently show a region where
values are lower than both low and high forward voltage. Values increase in this region
with annealing. Annealing has shown to increase the leakage current observable on the I-
V curves of Round 2 samples, up to 350 oc and further decreases with subsequent
annealing steps. Comparing the efficiency trend, it appears that vacuum annealing
increases the leakage current and the efficiency up to the 350 oc annealing step. These
may be associated either with increased oxygen deficiency associated with vacuum
annealing out-gassing. An increase in leakage is also associated with an increase in
interfacial state density. This topic is discussed in the next chapter.
53
Chapter 4
C-V CHARACTERIZATION
This chapter describes the principle behind capacitance-voltage measurements on
heterojunction solar cells of ZnO film deposited on Si by RF magnetron sputtering. The
experimental steps taken for measuring the n-ZnO/p-Si solar cell is described. Previous
work done related to the study of 1/C2- V relationships with deep levels and interfacial
states are reviewed in this chapter. Capacitance measurements were made after each
annealing step. The results are discussed. The purpose is to present material that is useful
in interpreting results related to understanding junction parameters and the capacitance
effects and changes with annealing.
4.1 Capacitance of a heterojunction
Capacitance measurements are useful in determining the built in voltage (Vbi) and carrier
density following equation 4.1 where Eo is the permittivity of free space, Ks is the
dielectric constant, A is the cross-sectional area, NA is the acceptor concentration.
Equation 4.1 is the depletion capacitance equation for a metal-semiconductor, which is a
good approximation and applicable for a highly doped n-ZnO layer sputtered onto a p-Si.
2(Vbi- VA) qNAKs&oA2
54
(4.1)
In the ideal system, plotting 1/C2 vs. V A, the applied voltage, yields a straight line
intercepting at 1/C2 = 0 with the voltage axis. Following equation 4.1, the slope of this
line is proportional to the carrier density and therefore a straight line is indicative of a
uniform doping distribution. Capacitance measurements, which are ideally frequency
independent, indicate the capacitance of the space charge layer. It is however often the
case that these plots are frequency dependent. Two types of plots are possible: curves that
are parallel to each other and therefore having different built-in voltages and curves that
have different slopes, converging to the same built-in voltage. The latter behavior is
attributed to the presence of surface irregularities and/or surface states in the
semiconductor [ 4.1].
C-V measurements are also often used to study deep impurity levels [4.2, 4.3].
Studies have been performed on Schottky barriers by Roberts and Crowell [ 4.3, 4.4]. The
analysis is similarly applicable to the p-n junction. Deep level effects are presented in Fig.
4.1. Deep levels are impurity levels that lie deep in the band gap, as opposed to shallow
levels lying close to the valence band for a p-type semiconductor or close to the
conduction band edge for an n-type. In Fig. 4.1, deep levels are mode led to be in the p-Si
side due to the n-ZnO being heavily doped and highly conductive. As a result of heavy
doping, the depletion region on the n-side is very small and therefore band bending is
minimal when bias voltage changes are applied. The effects of deep levels are however
observed in depletion regions for which band bending is obvious. The absence of deep
levels is modeled on the n-ZnO side in Fig. 4.1. Under high forward biases, when banding
ben ding is less, deep levels in the depletion region of the p-Si lie that lie completely over
the Fermi level are filled and do not contribute to charge storage. When the forward bias
55
is reduced, deep levels closest to the interface will bend downwards, below the Fermi
level, thereby allowing the carriers trapped in these deep levels to jump into the valence
band, creating an ionized charge in its place. The charge from these ionized sites
con tri butes to the overall capacitance measured. It is therefore at the crossing of the deep
level with respect to the Fermi level that a peak in terms of the change in charge density
with the change in barrier bending will be observed
56
0 Filled deep levels + Free holes - lonized levels
Ec ----------------------~
Deep level2 ;-;: oooooooooooooo~
Deep levell ooooooooooooooOê;"c;O\ EfSi--------------------~~:::....>.n
OO<-Xo
p-Si
1
1
~
dP dljfs
<--x-----------;
EfZnO
n-ZnO
Fig. 4.1: Band diagram of p-Si/Si02/n-ZnO with deep level modeling and dp/dl/; schematically plotted as a function of x where x2, x1 represent mid points between peak or demarcations for the different capacitive contributions from different energy levels.
57
Measurements for deep level analysis for p-n junctions are however usually clone
under reverse bias. The reason being that in forward bias, minority carrier injection
produces diffusion capacitance which makes deep level effects hard to detect. Under
reverse bias, band bending increases, depletion width widens and an increasing amount of
deep level sites move under the Fermi level in the depleted p-Si. The number of ionized
acceptors increases as the depletion width widens and the concentration level of free
holes increases in the valence band.
According to Equation 4.1, a non linear relationship between 1/C2 and V are
expected when deep levels are present and exposed and when their sites are ionized.
Concentration levels increase non-linearly as a result of bias changes. Nonlinearity in
1/C2 vs. V curves is therefore a direct result ofthe effect of deep level charge [ 4.3].
Further indication of the presence of deep level effects is observed in a C vs. V
relationship that is dependent on frequency. When an a.c. signal is added to the d.c. signal,
change in the depletion width is observed. During the positive half-cycle, the depletion
width will decrease while the opposite occurs for the negative half-cycle. When the
depletion width increases, deep level sites lying below the Fermi level in the p-Si, will be
ionized, creating a new charge and an increase in the concentration level in free carriers.
When the depletion width shrinks, the ionized site is neutralized by a free carrier from the
valence band. This change in charge is a direct relationship to the capacitance measured
and is inversely proportional to the deletion width as is shown in equation 4.2 where V ac
is the a.c. voltage, Q is the charge in Coulombs, W is the depletion width, A is the cross
sectional area, Eo is the permittivity of free space, and Er is relative permeability. In fact,
since equation 4.2 shows well known relationships that do not take into consideration the
58
effect of deep levels, the depletion width derived from the total capacitance measured
which includes the effect from deep level impurities is not a true value of width. In fact,
when low frequency is used, the depletion width profile is shifted towards the interface;
however, at high frequency, it shifts away from it. At low frequency, deep level sites have
enough time to react and therefore more of them become ionized. More charges are
present and therefore according to equation 4.2, the apparent width Wapp will decrease
towards the interface.
(4.2)
Low-frequency curves help to give properties of deep impurities since the deep
level sites are most likely to be able to follow the a.c. signal and contribute to the
capacitance measured. When the frequency is increased and the deep level sites are no
longer able to follow due to their long time constant for charging and discharging even
though they are below the Fermi level, the contribution from deep level sites is not
detectable on 1/C2 vs. V curves. A value for V bi and dopant concentration is best
measured at high frequency as the time constants of the dopants are substantially shorter
than that of the deep levels. It is clear that if C vs. V measurements change with
frequency, deep level effects are known to be present [4.3].
The change in apparent width with frequency is a result of the total capacitance
being composed of two elements, not properly modeled by the capacitance equation
containing Wapp· The total capacitance is composed of the capacitance due to shallow
dopant impurities and the capacitance due to deep level impurities. This second
59
capacitance term is a summation of the capacitance equations shown as equation 4.3. A
change in bias voltage will result in a non linear change in 1/C2 vs. V when deep level
effects are present. The following is a more detailed explanation of the relationship
between capacitance and a change in bias voltage as described by Roberts and Crowell
[4.4]. Let p(x) be the total charge density at any point x and ql/;8 be the energydifference
between the conduction band and the Fermi level at the interface. When the d.c. bias is
changed, the quantity dp/dl/;s which is shown in Fig. 4.1 will peak at the edge of the
depletion region and at points within the depletion region where deep impurity levels pass
under the Fermi level. The integral of dp!dl/ts over the depletion region is the capacitance
per unit area of the barrier. According to Roberts and Crowell's model, for each value of
d.c. bias, the capacitances may be written as:
(4.3)
Sin ce the midpoints between the dp/dl/;s peaks Xi (i = 0, 1, 2, etc) appearing in Fig. 4.1,
change with applied bias, capacitance values that take into account deep level effects
change with applied bias.
The effect of interface states may also be observed through C vs. V curves.
Surface state density according to Cowley and Sze's study on metal-semiconductor
systems [ 4.5] is a property only of the semiconductor surface. Interface charge appears at
the immediate vicinity of the heterojunction interface. This is an effect that differentiates
the heterojunction from the homojunction. Surface states are produced when foreign
60
atoms are absorbed at the surface and the termination of the periodic structure of the
semiconductor' s crystal is disturbed. A first assumption may be made to assume that the
sputtered ZnO is more likely to be defective than the industrial grade Si wafer, interface
states may be modeled in Fig. 4.2 to be at the interface of the Si02 layer and the ZnO.
Altemately, deep levels discussed previously and modeled in Fig. 4.1 may be present in
the Si side. Evidence of continued change in C-V curves with annealing, specifically a
decrease in V bi derived from the deep regions of the junction, is an indication that it is
more likely that interface states reside on the Si side as is modeled in Fig. 4.3. The results
for this will be presented shortly. Intuitively, this makes sense since ZnO is much more
heavily doped than Si. In this case, during the annealing process, it is possible to expect
that sorne of the zinc atoms may diffuse across from the ZnO si de.
61
Deep level2----------
Deep levell ---------~ EfSi-------------->.;:----'-::l
p-Si
• Filled surlace state + Empty surlace state
:------------ Ec
EfZnO
-----~--------Ev
n-ZnO
Fig. 4.2 Band diagram of p-Si/Si02/n-Zn0 with interface state modeling on the n-Znü si de.
62
Deep level2--------~..__
• Filled smface state - Empty smface state
Deeplevell-----------~ Ec
EfSi-----------~:----4!!'! f------------ Ef ZnO
p-Si ----------Ev
n-ZnO
Fig. 4.3 Band diagram of p-Si/Si02/n-Zn0 with interface state modeling on the p-Si side.
4.2 Princip les of interface state density calcula ti ons
The density of interface states (Nrs) at the Fermi level in the depletion region may be
calculated via Equation 4.4 using capacitance vs. frequency curves.
N - CIS IS- q
where C18 = CLF- CHF
63
(4.4)
where CLF is the low frequency capacitance, CHF is the high frequency capacitance.
Density of interface states may also be calculated via Equations 4.5 and 4.6 [4.1.50].
Vbi_app = Vbi -V a
qŒ2 V=---=-----
a 2(cnND +&pNA)
(4.5)
(4.6)
Where Vbi_app is the apparent built-in voltage, V bi is the actual diffusion voltage, a is the
number of charged states per cm2, q is the electron charge in Coulombs, Nn is the donor
density, NA is the acceptor density, En is the permittivity of the n-type semiconductor and
Ep is the permittivity of the p-type semiconductor. The value of Vbi may be calculated
using the Equation 4.7 [4.6].
(4.7)
where ~Ec is the conduction band discontinuity, Eg_si is the band gap of Si, Ûv_Si is the
energy between the valence band of Si and the Fermi level, and finally Ûc_zno is the
energy between the conduction band of ZnO and the Fermi level.
4.3 Experimental results
As mentioned in chapter 2, Round 1 samples are characterized by ZnO
depositions performed at 40 W for 6 hours, Round 2 samples at 120 W for 2 hours,
Round 3 samples at 80 W for 4 hours. Dry oxidation was performed at 450 oc for 15
64
minutes on sorne samples prior to Znü sputtering. The following is a list of the V bi and
dopant concentration measured via the 1/C2 vs. V curve with each annealing step on the
heterojunctions fabricated in the present work. Annealing step conditions are shown.
Table 4.1 A summary ofVbi derived from C~V results from high reverse bias regions.
Sample no. Annealing step: vbi(V)
Temperature Time (OC) (min)
*Round 1 As-deposited 0 1.40 not oxidized 200 30 0.10
300 30 0.20 350 30 NIA 400 30 NIA
*Round 1 As-deposited 0 1.40 oxidized 200 30 0.10
300 30 0.15 350 30 NIA 400 30 NIA
**Round 2 As-deposited 0 0 not oxidized 200 30 -0.05
300 30 -0.15 350 30 -0.20 400 30 ~ij~lO
**Round 2 As-deposited 0 0.20 oxidized 200 30 0.10
300 30 -0.05 350 30 -0.15 400 30 -0.20
Round 3 As-deposited 0 1.11 oxidized 200 30 0.41
300 30 0.05 350 30 -0.21 400 30 -0.22
Round 3 not As-deposited 0 1.10 oxidized 200 30 0.40
300 30 0.02 350 30 -0.15 400 30 -0.25
*: values are not accurate due to instrument-use error **: Slopes used are those resulting at high reverse biases
65
The 1/C2 vs. V intercept is shown to decrease with increase in annealing
temperature. This is the case for Round 2 and 3 samples except for the non-oxidized in
Round 2 after an annealing step at 400 °C. Round 1 samples show an increase in Vbi after
the second annealing step, i.e. after an annealing step at 300 °C. The results of Vbi
tabulated in Table 4.1 are shown in Fig. 4.4.
-> -:.c >
1.6
1.4
1.2
0.8
0.6
0.4
0.2
0
-0.2
-0.4
Vbi vs annealing step
Annealing step (30 minutes each)
--Round 2 non-oxidized w Round 2 oxidized
-Round 3 non-oxididized ---Round 3 oxidized -Round 1 non-oxidized
-Round 1 oxidized
Fig. 4.4 Change in V bi after each annealing step performed under the conditions noted in Table 4.1
The behavior of the apparent V bi shown in Fig. 4.4 is an indication of the presence
of surface states which is schematically modeled in Fig. 4.2. Surface states decrease
barrier height c/>b whose expression is shown in Equation 4.8, where Vbi is the built-in
voltage and Vp is the potential difference between the Fermi level (EF) and the valence
energy lev el (Ev).
(4.8)
66
V P may be deduced from impurity concentrations via Equation 4.9 where Nv is the
density of states for the valence band, NA is the impurity concentration and k is the
Boltzmann constant.
(4.9)
A high Vbi is therefore indicative of a higher barrier, one which may be degraded by
interface states. Decrease in barrier height leads to a decrease in open circuit voltage (V oc)
[4.7].
Prior to annealing (annealing step = 0 in Fig. 4.4), Round 1 samples are
characterized by a high and positive Vbi, Round 3 samples are characterized by a slightly
lower Vbi and Round 2 samples are characterized by the lowest Vbi whose values are
negative. This is an indication that surface states are present. Round 2 samples which
were subjected to higher RF power during ZnO sputtering have higher surface states due
to the lowest V bi measured. This is probably due to the bombardment of atoms onto the
surface. Annealing appears to increase the amount of surface states since there is a
decrease in Vbi values with annealing; however, for the Round 2 non-oxidized sample,
after the 350 oc annealing step, the Vbi starts to increase slightly. This is in agreement
with the dark leakage current trend observed in Fig. 3.2 a) whereby the shunt resistance
depicted by the dark leakage decreases up until the 350 oc annealing step and increases
after the 400 °C annealing step. The V bi trend for the oxidized Round 2 sample is also in
agreement with the dark leakage trend.
67
Fig. 4.5 a) and b) show the as-deposited results of C vs. V for the sample that was
oxidized for 15 minutes and the other that was not oxidized prior to ZnO deposition in
Round 2 where ZnO films were deposited at 120 W. The capacitance measurements were
performed at frequencies of 200, 1K, 10K, and 100KHz with the 4247A HP multi
frequency LCR meter. The curves show that for Round 2 samples, for which ZnO films
were deposited at 120 W in RF power, capacitance generally decreases with increase in
frequency. Round 2 samples are therefore characterized by deep level effects. Fig. 4.6 a)
and b) show as-deposited results of C vs. V for the samples fabricated in Round 1 for
which ZnO films were deposited at 40 W of RF power. Interestingly, the curves do not
vary as much with frequency, indicative of reduced effects from deep levels. 1t is possible
that when RF power is increased, bombardment of atoms onto the substrate surface
creates more deep level defects. Fig. 4.8 a) and b) show results of 1/C2 vs. V for as
deposited round 2 samples. Fig. 4.7 a) and b) show results of as-deposited round 1
samples. Round 2 samples show high frequency curves for which 3 different slopes are
detectible from high to low reverse bias. Round 1 samples are characterized by one slope
in the high reverse bias voltage range and another smaller slope at low reverse bias.
68
Round 2: Non oxidized ZnO-Si CvsV
-1 .0.9 .0.8 .0.7 .0.6 .0.5 .0.4 .0.3 .0.2 .0.1 0
Voltage (V)
Fig. 4.5 (a)
Round 2: oxidized ZnO-Si CvsV
90 ~--------------------------------~
80 -LL 70 s:::: -60 IV (,J 50 s::::
~ 40
~ 30
~ 20 0 10~~_.~~-+--.--
0 ~~--~--~--~--~--~~--~--~~ -1 .0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Voltage (V)
Fig. 4.5 (b)
-+-100KHz -~t--10 KHz
--.k-1 KHz
-+-200Hz
-+-100KHz -~t-10 KHz -+-1KHz -+-200Hz
Fig. 4.5 Reverse bias measurements of capacitance vs. voltage for as-deposited Round 2 samples, which were (a) non-oxidized, and (b) oxidized prior to ZnO deposition. ZnO deposition was performed at 120 W for 2 hours.
69
i:L 8.s
t: 8 -
-1
Round 1: Non oxidized ZnO-Si CvsV
.0.8 .0.6 .0.4 .0.2 0
Voltage (V)
Fig. 4.6 (a)
Round 1: oxidized ZnO-Si CvsV
-+-100KHz
--10KHZ
-.:--1KHz
-+-100Hz
12 ~------------------------------~
10 -LL 1::
- 8 -+-100KHz
-Il-10KHz
-.:--1KHz
-+-100Hz
0 ~----~----~------~----~----~ -1 .().8 .0.6 .0.4 .0.2 0
Voltage (V)
Fig. 4.6 (b)
Fig. 4.6 Reverse bias measurements of capacitance vs. voltage for as-deposited Round 1 samples, which were (a) non-oxidized, and (b) oxidized prior to ZnO deposition. ZnO deposition was performed at 40 W for 6 hours.
70
N < 0 -or-
N < 0 -or-
Round 1: Non oxidized ZnO-Si 1JCA2 VS V 0.03 r-------=-:..~--==--.:....=;.___:;__ ___ ____,
0.025
0.02 --+--100KHz
--10KHz
O.Q15 ---1KHz
0.01
0.005
0.03
0.025
0.02
0.015
0.01
0.005
-1 .0.8 .0.6 .0.4 .0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Voltage (V)
Fig. 4.7 (a)
Round 1: oxidized ZnO-Si 1JCA2 VS V
o~~~-~~~-~~~-~~~;.___:~
-1 .0.8 .0.6 .0.4 .0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Voltage (V)
Fig. 4.7 (b)
--+--100Hz --slope 1
-+-100KHz
--10KHz --.k--1 KHz
~100Hz
--Siope 1
Fig. 4.7 Reverse bias measurements of 1/C2 vs. V for as-deposited Round 1 samples, (a) non-oxidized, and (b) oxidized prior to ZnO deposition.
71
Round 2: Non oxidized ZnO-Si 1/CA2 VS V
0.007
0.006
0.005
N 0.004 < 0 - 0.003 'l""
0.002
0.001
0
-1
0.008
0.007
0.006
N 0.005
< 0.004 0 -~ 0.003
0.002
0.001
0
.().8 .().6 .().4 .().2 0 0.2
Voltage (V) Fig. 4.8 (a)
0.4 0.6 0.8
Round 2: oxidized ZnO-Si 1JCA2 VS V
1
~100KHz
-~t-10 KHz _.,.,_1 KHz
~200Hz
--slope 1 ,,~,Siope 2
-+-100KHz -10KHz ...,..1KHz
-+-200Hz -slope 1
Slope 2
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Voltage (V)
Fig. 4.8 (b)
Fig. 4.8 Reverse bias measurements of 1/C2 vs. V for as-deposited Round 2 samples, (a) non-oxidized, and (b) oxidized prior to ZnO deposition.
72
Round 2 samples for which ZnO films were deposited at an RF power of 120 W
for 2 hours show deviations in 1/C2 vs. v curves from straight lines and is therefore an
indication of a non-uniform doping profile due to deep level effects mentioned previously.
Other possibilities include stray capacitance or high series resistance. Since samples
fabricated in Round 2 and Round 3 show the same initial trend of non linearity, deviations
are most likely due to deep level effects. Linearity appears to be improved with annealing,
meaning the slope in high reverse bias becomes similar to that in low reverse bias. This is
an indication that the states at the near vicinity of the heterojunction interface are either 1)
being remedied by annealing or 2) are overshadowed by impurities in the deeper junction
region. Interestingly, Round 1 samples for which ZnO films were deposited at an RF
power of 40 W for 6 hours show C vs. V relationships that have no dependency on
frequency. The author would like to note that the voltage scale is not complete! y accurate
due to equipment use error. However, the trend of frequency independence is apparent
and an indication of low deep level effects in the samples characterized by low power
ZnO deposition. More specifically, deep levels most probably lie more than 5 kT away
from the Fermi level in neutral bulk [ 4.8] which would therefore mean that during reverse
biasing, the deep levels unlikely reach a level at which they can cross the Fermi level and
have deep level sites ionized. Contribution from the deep levels to the capacitance
becomes less likely and therefore C vs. V curves do not change with frequency. A section
of non-linearity is still however observed for low reverse voltages in the Round 2 curves
for which ZnO films were deposited at 120 W after all annealing steps. Deviation is larger
for non-oxidized samples and become more apparent at high annealing steps. Fig. 4.11 a)
73
,,..-...\ and b) show the results for Round 2 samples annealed after 200 oc degree annealing. Fig.
4.12 a) and b) show their results after 400 oc degree annealing. These non linear tails are
also observed in the Round 3 curves for which ZnO films were deposited at 80 W, but
appear only after 350 and 400 oc annealing. Similar trends of non linear faU offs are
reported in Tavakolian and Sites work with CdS/CulnSe2 [4.9]. The difference is that
their fall offs show an increase in capacitance whereas Round 3 curves show a decrease in
capacitance after each annealing step. Tavakolian and Sites have attributed their non
linear faU offs to interfacial states. Furthermore, beyond a certain threshold in the number
of interfacial states, effects on open-circuit voltage are noticeable such has been presented
by Tavakolian and Sites for their CdS/CulnSe2 ceUs. In this work, it has been shown in
chapter 3 that the highest V oc values for non-oxidized samples are found in as-deposited
samples. They decrease in value with each annealing step. Samples which were oxidized
for 15 minutes prior to ZnO deposition however show a slight increase in V oc after the
200 oc annealing step and then a decrease with subsequent annealing step.
Round 2 samples for which an RF power of 120 W was used show significant
non-linearity at low reverse biasing voltages for low and high frequency curves as shown
in Fig. 4.8, high indication of the presence of deep levels. Round 3 samples for which an
RF power of 80 W was used show non linearity at low reverse biasing voltages only for
low frequency curves as shown in Fig. 4.9 indicating that the film may be ofbetter quality
in terms of deep levelslinterfacial states compared to Round 2 samples. This is in
agreement with conversion efficiency values presented in chapter 3 whereby Round 3
samples were characterized with better Ise, V oc, FF and 'Y/. Round 1 samples for which an
RF power of 40 W was used show only a very slight non-linear portion at low reverse
74
biasing voltages for both high and low frequency curves which follow closely to one
another as shown in Fig. 4. 7. This is an indication that the lower the RF power, the less
interfacial state effects are present. Non-linear tails from Round 1 samples prior to
annealing are most likely not due to deep level effects since the C vs. V curves do not
vary with frequency.
0.012
0.01
~0.008
0 ;::o.oo6
0.004
0.002
Round 3: Non oxidized ZnO-Si 1/CA2 VS V
-1 .0.8 .0.6 .0.4 .0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Voltage (V)
Fig. 4.9 (a)
75
-+-100KHz
--10KHz
-.-1KHz
-+-200Hz
--stope 1
0.012
0.01
N < 0.008
~0.006 oc-
0.004
0.002
Round 3: Oxidized ZnO-Si 1fCA2 VS V
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Voltage (V)
Fig. 4.9 (b)
-+-100KHz
--10KHz .....,.1KHz
-+-200Hz -slope 1
Fig. 4.9 Reverse bias measurements of 1/C2 vs. V for as-deposited Round 3 samples, (a) non-oxidized, and (b) oxidized prior to ZnO deposition.
The number of interface states may be determined via C-V as well as C-F measurements
as discussed previously. Via the C-V method, Equations 4.5 and 4.6 are used to determine
the amount of interface charges present. To determine the actual diffusion voltage,
Equation 4.7 is used and repeated here.
(4.7)
The calculated value for Ôv_sds 0.15 eV assuming NA~lx10 17 cm·3 with a mobility of331
cm2N-sec for the boron-doped (100) Si wafers. For Ôc_zno, an approximate value of
~0.023 eV is calculated, assuming that N0 ~ 1019 cm·3 and ni~ 1.18xto·8 cm·3 which was
calculated with the effective densities of states for Si, since these values were assumed to
be very close to those of ZnO. The calculated V bi is 1.335 V. Using Equation 4.5 and 4.6
and assuming Er for Si is 11.7, Er for ZnO is 8.72, NA= lx1017 cm·3 and N0 =lx1019 cm·3,
76
the calculated parameters used and the calculated interface state densities are shown in
Table 4.2. It is worth noting that equations 4.5, 4.6, 4.7 used for determining the
interfacial state densities follow the assumption that the shallow region doping density is
the same as the deep region density. The deep region is observed at high reverse biasing
whereas the shallow region is observed at low reverse biasing and low forward biasing.
The shallow region may is characterized by low reverse biasing and forward biasing since
the depletion edges remain close to the interface. The deep region is characterized by a
high reverse bias since the depletion edges are deeper into the bulk. Since, in the cases
where there were more than one slopes in the 1/C2 vs. V curves, the slope in high reverse
bias was used to determine the apparent Vbi values, interfacial state densities (a)
calculated and tabulated in Table 4.2 are truly for those in the deep region density.
Table 4.2 A summary of calculated values for the interface state density at -0.18 V bias via C-V measurements.
Sampleno. Annealing step: vbi_app Va a (states-
Temperature Time (V) (V) cm-2
)
(OC) (min) with C-V
*Round 1 As-deposited 0 1.40 0.065 -not oxidized 200 30 0.10 -1.235 1.10x1013
300 30 0.20 -1.135 1.05 x1013
350 30 NIA NIA NIA 400 30 NIA NIA NIA
*Round 1 As-deposited 0 1.40 0.065 -oxidized 200 30 0.10 -1.235 1.10 x1013
300 30 0.15 -1.185 1.08 x1013
350 30 NIA NIA NIA 400 30 NIA NIA NIA
**Round2 As-deposited 0 0 -1.335 1.14 xl013
not oxidized 200 30 -0.05 -1.385 1.16 x1013
300 30 -0.15 -1.485 1.21 x1013
350 30 -0.20 -1.535 1.23 x1013
400 30 -0.10 -1.435 1.18 x1013
**Round 2 As-deposited 0 0.20 -1.135 1.05 x1013
oxidized 200 30 0.10 -1.235 1.09 x1013
300 30 -0.05 -1.385 1.16x1013
77
350 30 -0.15 -1.485 1.21 x1013
400 30 -0.20 -1.535 1.23 x1013
Round 3 As-deposited 0 1.11 -0.225 4.69 x1012
oxidized 200 30 0.41 -0.925 9.51 xl012
300 30 0.05 -1.285 1.12 xl013
350 30 -0.21 -1.545 1.23 x1013
400 30 -0.22 -1.555 1.23 x1013
Round 3 not As-deposited 0 1.10 -0.235 4.79 x1012
oxidized 200 30 0.40 -0.935 9.56 x10 12
300 30 0.02 -1.315 1.13 x1013
350 30 -0.15 -1.485 1.21 x1013
400 30 -0.25 -1.585 1.25 x1013
*: values are not accurate due to instrument-use error **: Slopes used are those resulting at high reverse biases
Interface state densities summarized in Table 4.2 are plotted in Fig. 4.1 O. It can be
observed for Round 2 samples, that the sample which was oxidized for 15 minutes prior
to Znü deposition has a lower deep region interface state density than the sample which
was not oxidized. Similarly, Round 3 samples which were exposed to air for a long period
of time are shown to have much lower interface state densities. The values for interface
state density are similar for oxidized and non-oxidized Round 3 samples with annealing
throughout the annealing processes however after 400 oc annealing, Round 2 samples
which were not oxidized show a decrease in interface state density whereas Round 2
oxidized samples show an increase. The decrease in interface state density for the non-
oxidized sample agrees with the decrease in dark leakage current and by the increase in
vbi at this 400 oc point. It appears that oxidation, whether intentionally performed at the
interface prior to Znü deposition or through exposure to air helps to decrease the
interface state density. Exposure to air however appears to decrease the value
significantly compared to an intentional oxide growth at the interface. The difference is
almost of one magnitude. Interface state density values appear to increase with annealing,
which is a peculiar trend as Si02 thickness is expected to increase. Furthermore, as
78
i---· mentioned earlier, the calculated values are those for the deep regions. Assuming that an
interfacial oxide exists, the increase in interfacial state density with annealing occurs at
the Si/Si02 boundary. The location ofinterfacial state may again be visualized by Fig. 4.3.
According to the increase in linearity for 1/C2 vs. V curves with annealing, interfacial
states at the vicinity of the interface appear to be either removed or overshadowed by the
deep region states. Uniformity in state density from the deeper region to the shallow
region is achieved with annealing though the overall magnitude in density increases.
-~N ·- < tn E s:: u cv-"Ctn cv cv--CQ: CQ:-
- tn tn ..... cv 0 u ... ~~ SE s:: ::::s - s:: -
Interface state density vs annealing step
1.4E+13
1.2E+13
1E+13
8E+12
6E+12
4E+12
2E+12
0
0 1 2 3
Annealing step (30 minutes each)
4
--Round 1 non oxidized _,__Round 1 oxidized
--Round 2 non oxidized
Round 2 oxidized
-- Round 3 non oxidized ---Round 3 oxidized
Fig. 4.10 Change in interface state density with annealing step.
79
N < 0 --
N < 0 --
Round 2: Non oxidized ZnO-Si After annealing at 200 degrees for
30 min
0.0016
0.0014
0.0012
0.001
0.0008
0.0006
0.0004
0.0002
1/CA2 VS V
Voltage (V)
Fig. 4.11 (a)
-+-100KHz
--10KHz ...._1KHz
-+-200Hz ---slope 1
Round 2: Oxidized ZnO-Si After annealing at 200 degrees for 30
min 1JCA2 VS V
0.003
0.0025
0.002 -+-200Hz --10KHz -.-1KHz
0.0015 -+-200Hz ----slope 1
0.001
0.0005
0 -1 .0.9 .0.8 .0.7 .0.6 .0.5 .0.4 .0.3 .0.2 .0.1 0 0.1 0.2
Voltage (V)
Fig. 4.11 (b)
Fig. 4.11 Reverse bias measurements of 1/C2 vs. V after 200 oc annealing for 30 minutes, for Round 2 samples (a) non-oxidized, and (b) oxidized prior to ZnO deposition.
80
Round 2: Non oxidized ZnO-Si After annealing at 400 degrees for
30 min 1/CJ\2 vs V
0.0016 ..--------------------,
0.0014
0.0012
N 0.001 < ~ 0.0008
"l"" 0.0006
N < 0 -"l""
0.0004
0.0002
0.002
0.0018
0.0016
0.0014
0.0012
0.001
0.0008
0.0006
0.0004
0 L-~~~~~~~~~~~--~
-1 -0.9 .Q.8 -0.7 .Q.6 .Q.5 .Q.4 .Q.3 -0.2 .Q.1 0 0.1 0.2
Voltage (V)
Fig. 4.12 (a)
Round 2: Oxidized ZnO-Si After annealing at 400 degrees
for 30 min 1/CA2 VS V
0.0002
0 -1 -0.9 -0.8 .Q.7 .Q.6 .Q.5 -0.4 -0.3 .Q.2 -0.1 0 0.1 0.2
Voltage (V)
Fig. 4.12 (b)
-+-100KHz
--10KHz _._1KHz
-+-200Hz --slope 1
-+-100KHz
-11-10KHz
--1KHz
-+-200Hz
--series5
Fig. 4.12 Reverse bias measurements of 1/C2 vs. V after 400 °C annealing for 30 minutes, for Round 2 samples (a) non-oxidized, and (b) oxidized prior to Znü deposition.
81
4.4 Conclusions
Samples for which ZnO films were sputtered at higher RF power show a higher
amount of interface states. This is probably due to the bombardment of a toms onto the Si
surface as a result of higher power during the sputtering process. It has also been shown
that annealing increases the amount of surface states in the deep to shallow region of the
junction since V bi decreases with annealing; however, for Round 2 non-oxidized samples,
after the 350 oc annealing step, Vbi starts to increase again. On the other hand, samples
which had been intentionally oxidized prior to ZnO deposition show a continued decrease
in V bi with a subsequent annealing. It appears therefore that oxidized samples show a
continued increase in interfacial states with annealing.
Samples for which ZnO films were sputtered at higher RF power also show higher
amount of deep level effects compared to samples for which low RF power was used.
Annealing appears to make uniform the interfacial state density from the deep to shallow
junction regions. A small non-linear tail is however observed in 1/C2 vs. V curves in low
reverse biasing and is not removed with annealing. This has been cited to be due to
interfacial states from other sources. It appears that oxidation, whether intentionally
performed at the interface prior to ZnO deposition or through exposure to air helps to
decrease the interface state density. Long exposure to air appears to lower the initial
interfacial state density measured much more significantly. Overall, the interface state
density increases with annealing. This result suggests that the interface states may be due
to interdiffusion of atoms, likely those of Zn into the Si surface, but possibly also due to
0 diffusion or less likely, Al.
82
Chapter 5
ELECTRON BEAM INDUCED CURRENT MEASUREMENTS
This chapter describes the principle behind the performance of Electron Bearn
Induced Current (EBIC) measurements on heterojunction solar cells. The experimental
steps tak:en for measuring the diffusion length of the n-ZnO/p-Si solar cell fabricated in
this work is described.
5.1 Princip les of the Electron Bearn Induced Current method
The performance of a solar cell depends on the diffusion length of minority
carriers (L) and surface recombination (S) which can are affected by fabrication and
manufacturing steps. The Electron Bearn Induced Current (EBIC) method, performed
with the scanning electron microscope (SEM), has often been used to study and
characterize semiconductor materials and deviees. The method has been successful in
determining carrier lifetime, bulk minority carrier diffusion length, surface recombination
velocities and defect energy levels. The technique involves the collection of carriers that
are generated by an electron bearn and then separated by the electric field in the depletion
region of the p-n junction. Other electron bearn carriers that do not enter the depletion
region are recombined. Once the carriers are collected, they are tumed into a video signal
on the SEM and reveal the location of the junction, as well as recombination centers such
as dislocations and doping inhomogeneities.
Excess carriers are removed via the recombination mechanism, at a rate
that is proportional to the carrier density. Carriers are also removed from a region by
83
mechanisms such as diffusion and drift and they are replenished by generating
mechanisms for steady state equilibrium to be maintained. Equation 5.2 summarizes these
events for an n-type semiconductor region, where G is the generation rate of electron-hole
pmrs.
5.2 Principles of diffusion length and surface recombination velocity measurements
The scanning electron microscope is a system of two cathode ray tubes (CRT),
one of which emits a bearn of electrons onto the sample. Bombardment effects are
captured by a detector. The other CRT, operating in synch with the first, detects the
voltage signal created in proportion to the effects captured by the detector. A point by
point image is then produced of the sample. When the sample is a p-n junction, the
junction's electric field, serves as the detector. When the electron bearn bombards the
sarnple, it generates electron ho le pairs, sorne of which will recombine with one another
most likely with the help oftraps. Other electron-hole pairs which are created in depletion
region, will encounter the junction's electric field, be separated and collected as current in
an extemal circuit [5.1]. Similarly, electron-hole pairs which are created in the n-p-type
bulk region diffuse across the junction and be separated as well into the extemal circuit.
The spatial dependence of EBIC profiles allow for the characterization of the junction,
detection of its diffusion length and determination of the location of recombina ti on sites.
The minority carrier diffusion equation in a p-type semiconductor is shown below
1 flnP G--\IJ --N q r
84
dflnP --=--= 0
dt (5.1)
It describes several mechanisms which are present and required for charge neutrality to
occur. Excess carriers such as those that might be provided by the electron bearn are
removed via the recombination mechanism, at a rate that is proportional to the carrier
density i.e. /':,.np/7. Carriers are also removed from a region by mechanisms such as
diffusion and drift and they are replenished by generating mechanisms, at a rate G for
steady state equilibrium to be maintained. G is related to the generation function g(x)
presented in Chapter 1 by G = g(x)/ A, where A is the illuminated area of the sample. The
Equation 5.1 's solution is of the form presented in Equation 5.2 where K is a constant and
Lp=(DN*7N) 112 is the minority carrier diffusion length. DN is the electron diffusion current
and 7N is the lifetime of an electron.
!J..n = Ke-jr.p p
(5.2)
The induced current from the electron bearn used to measure the diffusion length and the
surface recombination velocity in the EBIC method is proportional to the minority carrier
density, /':,.np.
The generated carrier profile resulting from the electron beam's penetration into
the semiconductor is shown in Fig. 5.1. The contours represent areas with the same
carrier generation rate. What is called "the mushroom" effect by Leamy [5.1] is due to the
scattering of the kilovolt electrons originating from the focused bearn during penetration.
Angular deviations in their original trajectories result in the contour profile shown in Fig.
5.1. Furthermore, it has been previously reported [5.1] that the carrier generation rate
85
decreases with increase in bearn energy while the depth of penetration increases. Reis the
maximum electron range or depth attained for a specified bearn energy.
Electron bearn
w = depletion region
Fig. 5.1 A cross-sectional view of the electron bearn penetration
The diffusion length L is a material parameter which represents the average
distance minority carriers can diffuse among majority carries before they recombine. It is
dependent upon intrinsic properties such as v, the drift velocity and therefore D, the
diffusion coefficient, but it is also dependent on trap density [5.1]. Trap density increases
due to fabrication processes and decreases the diffusion length. The EBIC method has
been extensively used to determine its diffusion length in deviees [5.1, 5.2, 5.3].
It is possible to determine the value for the diffusion length via the EBIC method
by fracturing the sample to expose the junction and then scanning the electron bearn
across it perpendicularly. It is also possible to determine the value by scanning the bearn
86
across the front surface of the heterojunction which is a non-destructive method. It is the
former case that will be dealt with in this work. It is therefore useful to know that the
EBIC peaks at the junction and decays exponentially away from it when the bearn is
scanned perpendicularly across the junction [ 5.1].
Equation 5.1 shows the relationship between lee the induced current, L the
diffusion length and x the distance between the bearn and the junction. The relationship is
valid under the following conditions: that the carrier generation source is a point source
and that the sample width is infinitely large with respect to the diffusion length [5.4, 5.5].
The first assumption is met in practice when a low energy bearn is used whereas the
second is met when the distance between the bearn and the interface is two diffusion
lengths. Equation 5.3 also holds for cases where the surface recombination velocity is
negligible.
(5.3)
where Imax is the current at the depletion edge. Plotting the natural log of the normalized
collected current Ieellmax vs. position x gives an approximate value of the reciprocal of the
diffusion length L. In non ideal cases where the above mentioned assumptions are not met,
the collected current data must be fitted with analytic expressions reported for general
conditions or in the case of Ong et al.'s work, implementing a modified binary search
algorithm to extract the values of diffusion length and surface recombination velocity.
It is worth noting that EBIC images which change with time [5.6] is an indication
of charging at the surface. This effect is expected for Si since the samples used for EBIC
87
measurements had gone through the annealing steps mentioned in previous chapters. An
oxide layer is likely to be present and act as the charger. This has been inferred to be the
case in a previous work done in this lab by Garanzotis [5.7]. It was shown for an Al/Si
Schottky barrier whose EBIC images changed after a few seconds of repeated
bombardment that when the bearn was tumed off for a few seconds, the original EBIC
profile was restored [5.7].
In addition to determining the diffusion length with the EBIC method, it is
possible to determine the surface recombination velocity. Surface recombination velocity
(cm/sec) is the product of trap density per unit are a, capture cross section per trap and the
thermal velocity [ 5.1]. These are part of the set of assumptions made on physical
boundaries and the recombination rate R that are specified to solve continuity equations
in practical situations. Recombination rate at the surface or boundary is proportional to
the surface recombination velocity.
It is commonly reported that actual lifetime for a semiconductor is difficult to
determine due to its relationship with surface recombination effects and those of bulk
recombination [5.8, 5.9]. Nevertheless, the determination of surface recombination
velocity is straightforward with EBIC performed across the fractured surface of the bulk
Si. The expression for the minority carrier flux that is absorbed at a surface recombination
velocity S with the appropriate boundary condition for diffusion to the surface is shown in
Equation 5.4, where z is the surface normal, De is the diffusion coefficient of the minority
carrier in the p-type Si and ~n is the excess electron density.
Sl1n = D 811n e 8zz=O
(5.4)
88
Since the induced current lee is proportional to 8.n, assuming a point source for
which z* ~ Re/3 where Reis the electron range [5.1], the expression shown as Equation
5.5 may be rewritten as Equation 5.6 when z* approaches zero.
alec !cc-l az
limln/cc Z---)-0
Blnlcc az
s
(5.5)
(5.6)
This result shows that with a change in bearn energy, it is possible to deduce from
the trend in induced current lee the value for the surface recombination density.
5.3 Experimental results
5.3.1 Characterization of ZnO-Si heterojunctions
In preparation of EBIC line-scanning across the heterojunction, the sample was
fractured by scribing over the front side and applying pressure with the sample lying on a
glass plate. Fracturing is performed in order to expose the junction. The fracture appeared
relatively clean. Prior to line scanning the deviee, images were grabbed of the junction.
Fig. 5.2 shows the image of one of the ZnO/Si heterojunction samples fabricated, taken
from the secondary electron detcctor of the SEM. The sample used has gone through the 4
annealing steps mentioned in the previous chapters. The image shows the thickness of the
sputtered ZnO film to be about 0.1 pm. The thickness of the ZnO sputtered layer is
89
clearly shown since on the le ft si de of the image, a section of ZnO that has pee led off the
Si is observable. No interfacial, native oxide layer is perceivable on the SEM image,
which is an expected result since documented thicknesses of native oxide growth from RF
magnetron sputtering are in the range of a few nanometers. This is beyond the SEM
resolution range.
Fig. 5.2 Image taken from the SEM secondary electron detector of the ZnO/Si heterojunction fabricated in this work.
5.3.2 Diffusion length measurements
The electron bearn is scanned across the fractured heterojunction and travels
parallel to the barrier as shown in Fig. 5.3. The induced current is preamplified by GW
Electronics preamplifier model 1 03B, and th en measured by GW Electronics Specimen
90
Current Amplifier model 1 03B. The plotting of the induced current was performed
manually by tabulating the induced current value displayed on the LED panel meter and
then plotting vs. time. A snapshot of the section being scanned which includes the
heterojunction interface is displayed on the CRT and saved after measurements are taken.
The images of two different samples that were fabricated under the same conditions and
previously annealed together are shown in Fig. 5.4 (a) and 5.5 (a). The junctions are
magnified at x500 (lcm=lum) and scanned with a 29 KeV bearn energy. The
magnification and actuallength (22.5 cm) ofthe displayed section on the CRT are used to
convert the total scanned time into a position value. A plot of induced current vs. position
is therefore achieved and shown in Fig. 5.4 (b) and Fig. 5.5 (b). Diffusion lengths are
determined according to the slope ofln(lccllmax) asper Equation 5.3. These are determined
from Fig. 5.4 (c) and 5.5 (c). The results are consistent between the two samples.
electron-bearn
w = depletion width Electrode
p-Si
n-ZnO
~-~CRT
LED panel meter
Fig. 5.3 A schematic diagram of the EBIC method performed on the fabricated ZnO/Si heterojunction with the SEM.
91
-c ê< ~uO -"Cl 111-·- < 't-0 =~ C.X E <(
Fig. 5.4 (a)
lnduced current vs position Round 2 - non oxidized sample 1 -post 15 annealing - n-ZnO/p-Si 14 13 12 11 10 9 8
~ l--1cm=10uml 5 4 3 2 1 o+-~-+~~.-.--.-*~~~~
-1 0 20 40 60 80 100 120 140 160 180 200
Position (um) Fig. 5.4 (b)
92
10
->< ca E 'ù ,g -z ...1
0
LN (lcc/lmax) vs position Round 2 - non oxidized sample 1 - post
annealing - n-ZnO/p-Si
-+-1cm = 10 um -•- Slope 1
20 40 60 80 100 120 140 160 Position (um)
Fig. 5.4 (c)
Fig. 5.4 A CRT snapshot of the first scanned junction (a) for which the plot of the induced current taken from the LED panel meter vs. position is shown in (b ), where zero is the leftmost point of the scanned section appearing on the CRT snapshot. The ln(Icc/Imax) relationship is displayed in (c) where x= 0 is the location ofthe EBIC peak.
Fig. 5.5 (a)
93
-x co E 0 0
::::.. z ...J
10
24 22 20 18
lnduced current vs position Round 2 - non oxidized sample 2 - post
annealing - n-ZnO/p-Si E-beam energy of 29 KeV
l--1cm=10uml
20 40 60 80 100 120 140 160 180 200 Position (um)
Fig. 5.5 (b)
LN (lcc/lmax) vs position Round 2- non oxidized sample 2- post
annealing - n-ZnO/p-Si E-beam energy of 29 KeV
-+-1cm= 10um _,._ Slope 1
0 10 20 30 40 50 60 70 80 90 100 110
Position (um)
Fig. 5.5 (c)
Fig. 5.5 A CRT snapshot of the second scanned junction (a) for which the plot of the induced current taken from the LED panel meter vs. position is shown in (b ), where x =0 is the leftmost point of the scanned section appearing on the CRT snapshot. The ln(Icc/Imax) relationship is displayed in (c) where x'= 0 is the location of the EBIC peak.
94
According to the results of the two scanned samples, the range in values for the diffusion
length is between 40 and 80 /lill. In order to further verify results, a higher magnification
was used to scan across the junction of the second sample to verify whether the slope of
the semi-logarithmic plot was the sarne as that at a lower magnification. Fig. 5.6 shows
that this is the case. Since the diffusion length LN= (DN*7N)112, the approximate minority
carrier lifetime is 3.07 11sec.
10
->< ca E 0 u -z .J
0
LN (lcc/lmax) vs position Round 2- non oxidized sample 2- post
annealing - n-ZnO/p-Si
2 3 4 5 6 7 8 9 10 Position (um)
Fig. 5.6 A plot of ln(Icc/Imax) vs. position for the second scanned sample at a magnification ofxlOOO, where x'= 0 is the location ofthe EBIC peak.
lt has been mentioned previously that Equation 5.3 is most valid for low bearn
energy. The previous samples have been scanned at a relatively high energy of 29 KeV.
Fig. 5.7 shows the result for a lower energy of20 KeV. The semi-logarithmic plot shown
in Fig. 5.7 (b) shows a more linear slope than those previously displayed in Fig. 5.5 and
5.6 showing that there is less scattering occurring for a lower bearn energy. The diffusion
length is still within the sarne range described from the previous scans i.e. 40-80 /lill.
95
-[: ~<( ::::J-uo ...... "0' Q)-·- < ~0 - ...... a. x E <(
8
6
4
2
0
-2
lnduced current vs position Round 2 - non oxidized sample 2 - post
annealing - n-ZnO/p-Si E-beam energy of 20 KeV
!--1cm = 10 umJ
20 40 60 80 1 00 120 140 160 180 200
Position (um)
Fig. 5.7 (a)
LN (lcc/lmax) vs position Round 2- non oxidized sample 2- post
annealing - n-ZnO/p-Si E-beam energy of 20 KeV
10
î ·~. u 1~~~--~~--~~--~~~~ u 1 0 20 30 40 50 60 70 80 90 100
2 ...J
0.1
Position (um)
Fig. 5.7 (b)
Fig. 5.7 A plot of the induced current for a bearn energy of20 KeY, taken from the LED panel meter vs. position (a), where x = 0 is the leftmost point of the scanned section appearing on the CRT snapshot. The ln(Iccllmax) relationship is displayed in (b) where x'= 0 is the location of the EBIC peak.
96
5.3.3 Surface recombination velocity measurements
The surface recombination velocity for the p-type Si has been determined
according to Equation 5.6. Scanning was performed at different bearn energies. Since the
value for the diffusion length has been determined to be ~40 f.J-m, the induced current was
measured at a point located at 30 f.J,m from the interface. The magnification was kept at
x500 (lem = 10 f.J-ID) and re-focusing was performed prior to every scan. The surface
recombination velocity which was determined from the value of ln(Icc) at 0 Ke V in bearn
energy gathered from Fig. 5.8 (c) shows that the measured velocity for the bulk Si which
have been annealed previously is ~520 cm/sec. Since the resistivity of the Si wafers used
were ~0.3-0.5 0-cm, the dopant concentration was estimated to be ~lx1017 cm-3 [5.10],
therefore an electron mobility of ~801 cm2N-sec [5.10] and hence a diffusion coefficient
of ~20.826 cm2 1 sec. Values measured by spectral conductance for wafers with p~0.3 n
cm were in the range of 105 cm/sec [5.10]. Monocrystalline p-type Si wafers doped to
2x1015 cm-3 are reported to have surface recombination velocity of 5020 cm/sec [5.11].
Palais et al. have also passivated the wafers with an iodine solution which subsequently
lowered the value to 80 cm/sec. The effect of passivating oxides has also been reported to
lower surface recombination velocities in Si [5.12] in sorne cases, with additional
techniques [5.13]. For diffusion lengths above 100 f.J,m, Baek et al. have calculated values
of 500-50 000 cm/sec for Si wafers doped with boron, arsenic, phosphorous or antimony.
The value of ~520 cm/sec found in this work when the induced current was taken at a
point located at 30 f.J-m from the interface has also been repeated for the induced current
taken at 20 f.J-m away from the interface. The results for the latter are shown in Fig. 5.9.
97
Whether this value is an accurate estimation of the surface recombination velocity
appears plausible considering the above, however it has not been full y verified.
4.5
4
...... 3.5 c ~- 3 .... 0 r5 s. 2.5
~ ~ 2 u-::J >< 1.5 "C c
0.5
Fig. 5.8 (a)
lnduced current vs bearn energy
-+- Measured data, 30 um away from the interface
0+-.-.-~--~--------~~~~--~~
0 2 4 6 8 1 0 12 14 16 18 20 22 24 26 28 30
Bearn energy (KeV)
Fig. 5.8 (b)
98
-5
- -10 0 .E -15 -z ...J -20
LN(Icc) vs beam energy
2 4 6 8 10 12 14 16 18 20 22 24 26 28
-+- Measured data, 30 um away from the interface
--Siope 1
-25 ---' -------~v~~~
-30
Bearn energy (KeV) Fig. 5.8 (c)
Fig. 5.8 A CRT snapshot of the scannedjunction (a) for which the plot of the induced current tak:en from the LED panel meter vs. electron bearn energy tak:en at a point 30 f1m away from the interface is shown in (b). The semi-logarithmic plot ofln(Icc) vs. bearn energy is shown in (c).
... c
14
12
cv 10 .__ ... 0 :::l- 8 U..,!...
"C < 6 cvo u-:::l >C 4
"C c
2
lnduced current vs bearn energy
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Bearn energy (KeV)
Fig. 5.9 (a)
99
-+- Measured data, 20 um away from the interface
LN(Icc) vs bearn energy
-5
........ -10 0 .f! -15 -z ...J -20
-25
-30
2 4 6 8 1 0 12 14 16 18 20 22 24 26 28
Bearn energy (KeV)
Fig. 5.9 (b)
--+-- Measured data, 20 um away from the interface
·--Siope 1
Fig. 5.9 A plot of the induced current taken from the LED panel meter is shown vs. electron bearn energy taken at a point located 20 flm away from the interface (a). The semi-logarithmic plot ofln(lcc) vs. bearn energy is shown in (b ).
5.5 Conclusions
The EBIC method has been used to measure the diffusion length and the surface
recombination velocity. The diffusion length measurements were conducted at high bearn
energy (29 KeV) and at a lower one of20 KeV. A diffusion length in the range of 40-80
flm was found. The minority carrier lifetime is therefore calculated to be around 3 flSec.
The surface recombination velocity has been measured in two instances, at 30 and 20 flm
away from the interface. The value is in the range of 500 cm/sec and appears to be low
compared to other reports for Si wafers; however it may be a reasonable range
considering that the sarnples that were scanned were previously annealed at 200, 300, 350
and 400 oc for 30 minutes each step.
100
Chapter 6
CONCLUSIONS
The main objective of this work was to fabricate n-ZnO/p-Si solar cells with the
RF sputtering method and characterize them by performing resistivity, current-voltage,
capacitance-voltage, capacitance-frequency measurements with vacuum annealing. The
diffusion length and surface recombination velocity were also measured with the Electron
Bearn Induced Current (EBIC) method.
Sheet resistance measurements were performed with a four-point-probe on the
glass samples. The values for samples evacuated in the vacuum chamber for 14 hours
prior to deposition increased from 7.9 to 10.17 and 11.5 Q/o for 40 W, 120 and 160 Win
RF power respectively. In contrast, those evacuated for 2 hours respect started with a
higher value of22.5 0/o, and decreased down to 7.6 and 5.8 Q/o for 40 W, 120 and 160
W in RF power respectively.
Since resistivity measurements have been linked to stoichiometric content of Zn/0,
it is likely that the amount of time that evacuation is performed has influence on
stoichiometry and film quality. Water and oxygen molecules diffuse into the surface of
the substrate when exposed to air and may need time to evacuate. A decrease in resistivity
is associated with oxygen outgassing during vacuum annealing and would indicate that
the film evacuated for 2 hours had a higher 0 content. The difference in resistivity trend
is a possible indication that the amount of time spent on evacuation prior to ZnO
deposition affects the growth of native oxide. It would be interesting to examine native
oxide thickness via Transmission Electron Microscopy (TEM) and to examine the change
101
in quality with annealing via Hall mobility and resistivity measurements with different
evacuation conditions prior to sputtering.
Samples which were oxidized for 15 minutes prior to ZnO sputtering showed no
efficiency improvement with annealing. For samples deposited at higher RF power and
which were not oxidized had improved up until an annealing temperature of 350 oc
though interfacial state densities and leakage current had increased up to this point.
Samples which had been exposed to air for a few weeks showed no efficiency
improvement with vacuum annealing, but had the highest magnitude in Ise, V0c, FF and TJ.
This was perhaps an indication that the samples surfaces oxide has formed and outgassing
in vacuum was unable to improve stoichiometry and that a passivation layer had grown at
the surface prior to ZnO deposition. These air-exposed samples had the lowest interfacial
state density ..
Annealing has shown to increase the leakage current observable on the I-V curves
for the Round 2 ZnO/Si samples, for which ZnO was deposited at 120 W, up to an
annealing temperature of 350 °C. The subsequent annealing step performed at 400 oc
decreases leakage. Dark leakage current trends appearing on the I-V curves are a clear
indication of the increases and decrease of interfacial states. Leakage has also been
previously associated with increased oxygen deficiency. It is likely that interdiffusion of
Zn is occurring across the interface with annealing. Interfacial state density increase has
been calculated based on the assumption that the shallow junction region is characterized
by the same density as the deep junction region. The increase indicates that in the vicinity
of the Si/Si02 interface, at the deep regions of the junction, is getting worse in terms of
deep levels/interfacial state density with annealing. Furthermore, annealing serves to
102
make uniform the interfacial state density from the deep to shallow junction regions since
1/C2 vs. V curves become linear. What is interesting is that although evidence shows an
increase in interfacial states in the deep regions of the junction with vacuum annealing,
the measured conversion efficiency, fill factor and short circuit current of the solar cell
increases with annealing up until the 350 oc step.
Round 2 oxidized samples on the other hand show a continued increase in interfacial
states with annealing. 11 and Ise have followed this trend and show a continued decrease
with annealing; however FF has increased up until the 350 oc annealing step, despite the
increasing trend in interfacial state.
It appears that the samples which had been exposed to air for a long period of time
prior to ZnO deposition have the highest values in Ise, V oc, FF and 11 among all samples
fabricated, but also, their values are highest prior to annealing, whether an oxide layer had
been intentionally grown or not. Trends in interfacial state density are very similar
between the two types of samples, that is, deep region interfacial state density increase
with annealing.
Samples for which ZnO films were sputtered at higher RF power show a higher
amount of interface states, likely due to the bombardment of atoms onto the surface
during sputtering. A small non-linear tail, probably due to interfacial states, is however
observed in 1/C2 vs. V curves in low reverse biasing and is not removed with annealing,
perhaps an indication that the very shallow interfacial states are not affected by annealing.
It appears that oxidation, whether intentionally performed at the interface prior to ZnO
deposition or through exposure to air helps to decrease the deep region interface state
density. Long exposure to air appears to lower the initial interfacial state density much
103
more significantly than intentionally growing oxide at the interface at 450 oc for 15
minutes. For these air-exposed samples, the calculated interface state density has been
shown to increase with vacuum annealing, while conversion efficiency decreases with it.
The trend is different from that seen for Round 2 non-oxidized samples whose calculated
conversion efficiency increases with interfacial state density up to an annealing of 350 °C.
The EBIC method has been used to measure the diffusion length and the surface
recombination velocity of the Si bulk. A diffusion length in the range of 40-80 pm was
measured, with a minority carrier lifetime calculated to be around 3 JlSec. It has been
shown that the surface recombination velocity may be measured either at 30 and 20 Jlm
away from the interface. The SRV on the fractured surface of the bulk Si used to fabricate
the solar cells and determined by the EBIC method was calculated to be around 500
cm/sec. It would be worthwhile to measure the surface recombination velocity at the
interface of the ZnO/Si heterojunction and compare the change with annealing to
parameters such as V oc, Ise, FF and 'Y/.
104
References
[1.1] N. Izyumskaya, V. Avrutin, U. Ozgur, Y. I. Alivov, and H. Morkoc, Preparation
and Properties ofZnO and Deviees, phys. Stat. Sol. (b) 244 No. 5, 1439-1450 (2007)
[1.2] A. A. Ibrahim, A. Ashour, ZnO/Si cell fabricated by Spray Pyrolysis, J. Mater. Sei,:
Mater Electron (2006) 17:835-839
[1.3] T. Markvart and L. Castaner, Solar Cells; Materials, Manufacture and Operation,
Elsevier, First edition, 2005.
[1.4] Y. I. Alivov, U. Ozgur, S. Dogan, D. Johnstone, V, Avrutin, N. Onojima, C. Liu, J.
Xie, Photoresponse ofn-ZnO/p-SiC Heterojunction Diodes Grown by Plasma-Assisted
Molecular-Beam Epitaxy, Applied Physics Letters, 86, 241108, 2005
[1.5] I. Sieber, N. Wanderka, I. Urban, I. Dorfel, E. Schierhom, F. Fenske, and W. Fuhs,
Electron Microscope Characterization ofReactively Sputtered ZnO Films with Different
Al-Doping Leve1s, Thin Solid Films 330 (1998) 108-113
[1.6] K. Bucher, J. Bruns, and H. G. Wagemann, Absorption Coefficient of Si: An
Assessment ofMeasurements and the Simulation ofTemperature Variation, J. Appl. Phys.
Vol. 75, No. 2, 15 January, 1994
[1.7] J.Y. Lee, Y.S. Choi, J.H. Kim, M.O. Park, S. lm, optimizingn-ZnO/p-Si
heterojunctions for photodiode applications, Thin Solid Films, 403-404 (2002) 553-557
[1.8] I. S. Jeong, J.H. Kim and S. lm, Ultraviolet-enhanced photodiode employing n
ZnO/p-Si Structure, Applied Physics Letters, 83, 2946, 2003
[1.9] D. Song, D.-H. Neuhaus and A. G. Aberle, Interfacial Structure and Current
Transport Properties of Sputter-Deposited ZnO:Al/c-Si Hetereojunction Solar Cells, 3rd
World Conference on Photovoltaic Energy Conversion, May 11-18, 2003, Osaka, Japan
[ 1.1 0] D. Song, B. Guo, A. G. Aberle, Heterojunction Properties of ZnO :Al/p-Si
Prepared by RF Magnetron Sputtering, IEEE 2002
[2.1] U. Ozgur, Y a. I. Alivov, C. Liu, A. Teke, M. A. Reshchikov, S. Dogan, V. Avrutin,
S.-J. Cho and H. Morkoc, A Comprehensive Review ofZnO Materia1s and Deviees,
Journal of Applied Physics, 98, 041301, 2005
105
[2.2] J. Yoo, J. Lee, S. Kim, K. yoon, I. J. Park, S. K. Dhungel, B. Karunagaran, D.
Mangalaraj, and J. Yi, The Properties of Surface Textured ZnO:Al Films for Thin Film
Solar Cells, Phys. Stat. Sol. 2, 1228-1232 (2005)
[2.3] K. K. Kim, J. H. Song, H. J. Jung, W. K. Choi, S. J. Park, J. H. Song, The Grain
Size Effects on the Photoluminescence of ZnO/a-Al203 Grown by Radio-Frequency
Magnetron Sputtering, Journal of Applied Physics, 87, 3573
[2.4] K. K. Kim, J.H. Song, H. J. Jung, W. K. Choi, S. J. Park, J.H. Song, J. Y. Lee,
Photoluminescence and heteroepitaxy of ZnO on sapphire substrate (0001) grown by RF
Magnetron Sputtering.
[2.5] Y. I. Alivov, X. Bo, S. Akarca-Biyikli, Q. Fan, J. Xie, N. Biyikli, K. Zhu, D.
Johnstone, and H. Morkoc, Effect of Anneaoing on Electrical Properties ofRadio
Frequency-Sputtered ZnO Films, J. Elec. Mat. Vol. 35, No. 4, 2006
[2.6] R. Das and s. Ray, Zinc Oxide- a Transparent, conducting, IR-Reflector Prepared
by RF-Magnetron Sputtering, Journal ofPhysics D: Applied Physics, 36 (2003) 152
[2.7] S. H. Jeong, J. W. Lee, S. B. Lee, J.H. Boo, Deposition of Aluminum-Doped Zinc
Oxide films by RF Magnetron Sputtering and Study of their Structural, Electrical and
Optical Properties.
[2.8] S. H. Jeong, B. S. Kim, B. T. Lee, Photoluminescenece Dependence of ZnO Films
Grown on Si(1 00) by Radio-Frequency Magnetron Sputtering on the Growth Ambient,
Applied Physics Letters 82, 2625, 2003
[2.9] J. S. Yoo, J. C. Lee, S. K. Kim, K. H. Yoon, I. J. Park, d. Y. Kim, and J. Yi, Surface
Textured ZnO:Al Films by RF Magnetron Sputtering Deposition for Thin Film Solar
Cellls
[2.10] D. Song, D.-H. Neuhaus and A. G. Aberle, Interfacial Structure and Current
Transport Properties of Sputter-Deposited ZnO:Al/c-Si Hetereojunction Solar Cells, 3rd
World Conference on Photovoltaic Energy Conversion, May 11-18, 2003, Osaka, Japan
[2.11] X. Jiang, C.L. Jia, B. Szyszka, Manufacture of Specifie Structure of Aluminum
Doped Zinc Oxide Films by Patterning the Substrate Surface, Appl. Phys. Lett. Vol. 80,
No. 17, 3090, 29 Apri12002.
[2.12] J.Y. Lee, Y.S. Choi, J.H. Kim, M.O. Park, S. lm, optimizing n-ZnO/p-Si
heterojunctions for photodiode applications, Thin Solid Films, 403-404 (2002) 553-557
106
[2.13] I. S. Jeong, J.H. Kim and S. lm, Ultraviolet-enhanced photodiode employing n
ZnO/p-Si Structure, Applied Physics Letters, 83, 2946, 2003
[2.14] F. Moeller, T. Vandahl, D.C. Malocha, N. Schwesinger, W. Buff, Properties of
thick ZnO layers on oxidized Si, Ultrasonics Symposium, 1994, 403-406.
[2.15] S. lm, B.J. Jin, S. Yi, Ultraviolet Emission and Microstructural Evolution in
Pulsed-Laser Deposited ZnO films, Journal of Applied Physics, 87, 4558, May 2000
[2.16] O. breitenstein, P. altermatt, K. Ramspeck, and A. Schenk, The Origin ofldeality
Factors N>2 of Shunts and Surfaces in the Dark I-V curves of Si Solar Cells, Proceedings
21st European Photovoltaic Solar Energy Conference, pp. 625-628, 2006
[2.17] M. S. Aida, e. Tomasella,j. Cellier, M. Jacquet, N. Bouhssira, S. Abed, Z. Mosbah,
Annealing and Oxidation Mechanism of Evaporated Zinc Thin Films from Zinc Oxide
Powder, Thin Solid Films, 515, (2006) 1494-1499.
[2.18] J.D. Ye, S.L. Gu, S.M. Zhu, W. Liu, S.M. Liu, R. Zhang, Y. Shi, and Y.D. Zheng,
Electroluminescent and transport mechanisms of n-ZnO/p-Si heterojunctions, Applied
Physics Letters 88, 182112 (2006)
[3.1] I. S. Jeong, J.H. Kim and S. lm, Ultraviolet-enhanced photodiode employing n
ZnO/p-Si Structure, Applied Physics Letters, 83, 2946, 2003
[3.2] Y. I. Alivov, U. Ozgur, S. Dogan, D. Johnstone, V, Avrutin, N. Onojima, C. Liu, J.
Xie, Photoresponse of n-ZnO/p-SiC Heterojunction Diodes Grown by Plasma-Assisted
Molecular-Beam Epitaxy, Applied Physics Letters, 86, 241108, 2005
[3 .3] K. B. Sundaram and A. Khan, Work Function Determination of Zinc Oxide Films, J.
Vac. Sei. Technol. A 15(2) 1997
[3.4] E. F. Schubert., Band Diagrams ofHeterostructures, Rensselaer Polytechnic
Institute, New York, U.S.A.
[3.5] Y. C. Ruan and W. Y. Ching, An Effective Dipole Theory for Band Lineups in
Semicondutctor Heterojunctions, J. Appl. Phys. 62 (7), 1987
[3.6] M. Grundmann, The Physics ofSemiconductors, Springer Berlin Heidelberg
Publishing, pp. 277-301, 2006
107
[3.7] D. Song, D.-H. Neuhaus and A. G. Aberle, Interfacial Structure and Current
Transport Properties of Sputter-Deposited Znü:Al/c-Si Hetereojunction Solar Cells, 3rd
World Conference on Photovoltaic Energy Conversion, May 11-18,2003, Osaka, Japan
[3.8] X. Jiang, C.L. Jia, B. Szyszka, Manufacture of Specifie Structure of Aluminum
Doped Zinc Oxide Films by Patteming the Substrate Surface, Appl. Phys. Lett. Vol. 80,
No. 17, 29 April2002.
[3.9] J.Y. Lee, Y.S. Choi, J.H. Kim, M.O. Park, S. lm, optimizing n-Znü/p-Si
heterojunctions for photodiode applications, Thin Solid Films, 403-404 (2002) 553-557
[3.10] J.Y. Lee, Y.S. Choi, W.H. Choi, H.W. Yeom, Y.K. Yoon, J.H. Kim, and S. lm,
Characterization of films and interfaces in n-Znü/p-Si photodiodes, Thin Solid Films
420-421,2002,112-116
[3.11] D. Song, B. Guo, A. G. Aberle, Heterojunction Properties ofZnO :Al/p-Si
Prepared by RF Magnetron Sputtering, IEEE 2002
[3.12] 1. Sieber, N. Wanderka, 1. Urban, 1. Dorfel, E. Schierhom, F. Fenske, and W. Fuhs,
Electron Microscope Characterization ofReactively Sputtered Znü Films with Different
Al-Doping Levels, Thin Solid Films 330 (1998) 108-113
[3.13] K.H. Kim, K. C. Park, and D.Y. Ma, Structural, electrical and optical properties of
aluminum doped zinc oxide films prepared by radio frequency magnetron sputtering,
Journal of Applied Physics, 81 (12), 15 June 1997.
[3.14] H. Y. Kim, J.H. Kim, M.O. Park, S. lm, Photoelectric, Stoichiometric and
structural properties ofn-Znü film on p-Si, Thin Solid Films 398-399 (2001) 93-98.
[3.15] Dhananjay, J. Nagaraju, S. B. Krupanidhi, Investigations on Zinc Oxide Thin
Films Grown on Si (100) by Termal Oxidiation, Materials Science and Engineering B 137
(2007) 126-130.
[4.1] A. J. Nozik, Photoelectrochemistry: Applications to Solar Energy Conversion, Ann.
Rev. Phys. Chem. 1978,29: 189-222
[ 4.2] S. M. Sze, Physics of Semiconductor Deviees, Second Edition, John Wiley & Sons
Taipei, Taiwan, 1981
[ 4.3] G. 1. Roberts and C. R. Crowell, Capacitive Effects of Au and Cu lmpurity Levels in
Pt-N Type Si Schottky Barriers, Solid State Electronics, Vol. 16, pp. 29-38, 1973
108
[4.4] G.I. Roberts and C. R. Crowell, Capacitance Energy Level Spectroscopy ofDeep
Lying Semiconductor Impurities Using Schottky Barriers, J. Appl. Phys., 15 March 1970
[4.5] A. M. Cowley and S. M. Sze, Surface States and Barrier Height ofMetal
Semiconductor Systems, J. Appl. Phys. Vol. 36, 3212, October 1965
[4.6] J.P. Donnelly and A.G. Milnes, The Capacitance ofp-n Heterojunctions Including
the Effects of Interface States, IEEE Transactions on Electron Deviees, Vol. ED-14, No. 2,
February 1967
[4.7] S. Darwish, Dark and Photovoltaic Properties ofp-CoPc/n-GaAs Heterojunction
Cells, Egypt. J. Sol., Vol. 26, No. 1, 2003
[ 4.8] M. Beguwala and C. R. Crowell, Characterization of Multiple Deep Level Systems
in Semiconductor Junctions by Admittance Measurements, Solid State Electronics, 1974,
Vol. 17, pp. 203-214
[ 4.9] H. Tavakolian and J. R. Sites, Effect oflnterfacial States on Open-Circuit Voltage,
IEEE 1998, 1608
[5.1] H. J. Leamy, Charge Collection Scanning Electron Microscopy, J. Appl. Phys. 53,
R51, 1982
[5.2] Z. Z. Bandic, P. M. Bridger, E. C. piquette, and T. C. McGill, Minority Carrier
Diffusion Length and Lifetime in GaN, App. Phys. Lett. 72, no. 5, pp. 3166, 15 June 1998.
[5.3] V. K. S. Ong, J. C. H. Phang, and D. S. H. Chan, A direct and Accurate Method for
the Extraction of Diffusion Length and Surface Recombination V elocity From an EBIC
Line Scan, Solid State Electronics Vol. 37, No. 1, pp. 1-7, 1994
[5.4] V. K. S. Ong, Determination ofDiffusion Length from Within a Confined Region
with the Use ofEBIC, IEEE Trans. Elect. Dev. Vol. 48, No. 2, February 2001.
[5.5] L. Chemyak, A. Osinsky, H. Temkin, J. W. Yang, Q. Chen and A. AsifKhan,
Electron Bearn Induced Current Measurements ofMinority Carrier Diffusion Length in
Gallium Nitride, Appl. Phys. Lett., Vol. 69, no. 17, pp.2531-2533, October 1996
[5.6] T. Kato, T. Matsukawa, H. Koyama, and K. Fujikawa, Scanning Electron
Microscopy of charging Effect on Si, J. Appl. Phys. Vol. 46, 2288, May 1975
[ 5. 7] T. Garanzotis, Electrical Characterization of CulnSe2 Crystals, M.Eng Thesis,
Department ofElectrical Engineering, McGill University, May 1998.
109
[5.8] Y. 1. Ogita, Bulk Lifetime and Surface Recombination Velocity Measurement
Method in Semiconductor Wafers, J. Appl. Phys. Vol. 79, No. 9, 1 May 1996, pp.6954-
6960
[5.9] D. Baek, S. Rouvimov, B. Kim, T.-C. Jo, D. K. Schroder, Surface Recombination
Velocity of Si Wafers by Photoluminescence, Appl. Phys. Lett. Vol. 86, 2005,112110-1
[5.10] R.F. Pierret, Semiconductor Fundamentals Volum 1, Modular Series on Solid
State Deviees, Second Edition Addison Wesley Publishing Company, 1988.
[5.11] O. Palais and A. Arcari, Contactless Measurement ofBulk Lifetime and Surface
Recombination Velocity in Si Wafers, J. Appl. Phys. Vol. 93, No. 8, 15 April, 2003
[5.12] A. Poggi andE. Susi, Effect ofPassivating Oxides on the Surface Recombination
Velocity in Si, Phys. Physica Status Solidi (a) Volume 113, Issue 1, pp. K61-K65.
[5.13] E. Yablonovitch, D. L. Allara, C. C. Chang, T. Gmitter, and T. B. Bright,
Unusually Low Surface-Recombination Velocity on Si and Germanium Surfaces,
Physical Review Letters, Vol. 57, Number 2, 14 July 1986
110