/smash/get/diva2:380720/FULLTEXT01.pdf

/smash/get/diva2:380720/FULLTEXT01.pdf
TVE 10 021
Examensarbete 30 hp
December 2010
Experimental study of Cu2ZnSnS4
thin films for solar cells
Hendrik Flammersberger
Institutionen för teknikvetenskaper
Department of Engineering Sciences
Abstract
Experimental study of Cu2ZnSnS4 thin films for solar
cells
Hendrik Flammersberger
Teknisk- naturvetenskaplig fakultet
UTH-enheten
Besöksadress:
Ångströmlaboratoriet
Lägerhyddsvägen 1
Hus 4, Plan 0
Postadress:
Box 536
751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Cu2ZnSnS4 (CZTS) is a semiconductor with a direct band gap of about 1,5 eV and an
absorption coefficient of 10^4 cm^-1, and is for this reason a potential thin film solar
cell material. Demonstrated efficiencies of up to 6,8% as well as use of cheap and
abundant elements make CZTS a promising alternative to current solar cells.
The aim of this study was to fabricate and characterize CZTS films and to evaluate
their performance in complete solar cells. For the fabrication of CZTS we applied a
two-step process consisting of co-sputtering of the metal or metal-sulphur
precursors, and subsequent sulphurization by heating at 520°C in sulphur atmosphere
using sealed quartz ampoules.
The work included a systematic comparison of the influence of composition on quality
and efficiency of CZTS solar cells. For this purpose films with various metallic ratios
were produced. The results show that the composition has a major impact on the
efficiency of the solar cells in these experiments. Especially zinc-rich, copper-poor and
tin-rich films proved to be suitable for good cells. The worst results were received for
zinc-poor films. An increase in efficiency with zinc content has been reported
previously and was confirmed in this study. This can be explained by segregation of
different secondary phases for off-stochiometric compositions. According to the
phase diagram, zinc-poor films segregate mainly copper sulfide and copper tin sulfide
compounds which are conductive and therefore detrimental for the solar cell. Zinc
sulfide, that is supposed to be present in the other regions of the phase diagram
examined in this study, could be comparatively harmless as this secondary phase is
only isolating and by this ’just’ reduces the active area. This is less disadvantageous
than the shunting that can be caused by copper sulfides. Contrary to the efficiency
results, metal composition had no major impact on the morphology.
A comparison of the composition before and after the sulphurization revealed that
metal precursors showed higher tin losses than sulphur containing precursors. A
possible explanations for this was given.
Another central point of this work was the examination of the influence of sulphur in
the precursor. Less need of additional sulphur in the film might lead to better material
quality. This is based on the assumption that the film is subjected to less diffusion of
the elements and so to less dramatic changes within the film, which might result in
fewer voids and defects. However, our experiments could find only a weak trend that
sulphur in the precursor increases the performance of the solar cells; concerning
morphology it was observed that more compact films with smaller grains develop
from metal-sulphur-precursors.
The best efficiency measured within this work was 3,2%.
Handledare: Tomas Kubart
Ämnesgranskare: Charlotte Platzer-Björkman
Examinator: Nora Maszzi
UPTEC FRIST10 021
Contents
1
Introduction
2
Theory
2.1 Solar cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 Solar radiation . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 Intrinsic, p– and n–type semiconductor . . . . . . . . . .
2.1.3 Fermi energy, valence and conduction band . . . . . . . .
2.1.4 Formation of the space charge region at the p–n–junction
2.1.5 Currents in a diode . . . . . . . . . . . . . . . . . . . . . .
2.1.6 IV characteristics of a diode . . . . . . . . . . . . . . . . .
2.1.7 The illuminated diode . . . . . . . . . . . . . . . . . . . .
2.1.8 Equivalent circuit of a solar cell . . . . . . . . . . . . . . .
2.1.9 Losses in solar cells . . . . . . . . . . . . . . . . . . . . . .
2.2 Thin film solar cells . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Device structure and fabrication techniques . . . . . . . .
2.2.2 Possible materials . . . . . . . . . . . . . . . . . . . . . .
2.3 CZTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 The ternary phase diagram (TPD) . . . . . . . . . . . . .
2.3.3 Secondary phases . . . . . . . . . . . . . . . . . . . . . . .
2.3.4 Reaction path for formation of CZTS . . . . . . . . . . .
2.3.5 Previous studies . . . . . . . . . . . . . . . . . . . . . . .
2.3.6 Aim of this study . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
4
4
4
6
7
8
9
11
13
14
17
17
19
21
21
22
23
25
26
29
Experimental
3.1 Fabrication techniques . . . . . . . . . . . . . . . . . . . . .
3.1.1 Precursor deposition by sputtering . . . . . . . . . .
3.1.2 Sulphurization . . . . . . . . . . . . . . . . . . . . .
3.1.3 Processing of the solar cell . . . . . . . . . . . . . . .
3.2 Analysis techniques . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Scanning Electron Microscopy (SEM) . . . . . . . .
3.2.2 Energy Dispersive X-ray Spectroscopy (EDS/EDX) .
3.2.3 X-ray Photoelectron Spectroscopy (XPS) . . . . . .
3.2.4 X-Ray Diffraction (XRD) . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
31
31
33
34
35
35
35
39
39
3
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
i
Contents
ii
3.2.5
3.2.6
4
5
Quantum efficiency (QE) measurements . . . . . . . . . . . . . . .
Current–voltage (IV) measurements . . . . . . . . . . . . . . . . .
Results and discussion
4.1 Sputtering of precursors . . . .
4.1.1 Composition . . . . . .
4.1.2 Structure . . . . . . . .
4.1.3 Special settings . . . .
4.1.4 Conclusions . . . . . . .
4.2 Properties of sulphurized films
4.2.1 Data set . . . . . . . . .
4.2.2 Composition . . . . . .
4.2.3 Morphology . . . . . . .
4.2.4 Conclusions . . . . . . .
4.3 Solar cells . . . . . . . . . . . .
4.3.1 Efficiency . . . . . . . .
4.3.2 QE measurements . . .
4.3.3 IV measurements . . . .
4.3.4 Conclusions . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Conclusions and suggestions for future work
Bibliography
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
40
42
43
43
43
45
46
48
49
49
49
55
73
75
75
83
89
93
95
99
List of Abbreviations
104
List of Figures
107
List of Tables
109
A Appendix
111
A.1 Trends in photovoltaics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
A.2 XPS measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
1 Introduction
In recent years, climate change and a sustainable development of energy resources were put
into the limelight to a greater extend. Among other things, the United Nations Framework
Conventions on Climate Change1 made the broad public aware that the finiteness of
primary fossil fuels like coal and oil on the one hand, and the climate change as a result
of the CO2 –emission by the use of burning those fuels on the other hand [1], lead to an
indispensable change from fossil fuels to renewable energies. As there is only a certain
amount of fossil fuels, there is already now an increasing trend of prices (Fig. 1.1).
Furthermore, already now the maximum of the oil production could have been reached
(so called peak oil). That means, sooner or later one has to search for alternatives. At the
same time, the world’s energy consumption increases massively , especially in countries
like China. It is also a question of equity that countries for example from the Third World
Figure 1.1: Increase of prices for the non-renewable resources coal (black line), oil (blue),
natural gas (red) and uranium (yellow). Units: US-Dollar per barrel of oil equivalent
(equates ca. 6 GJ). From [2].
1
Weltklimagipfel
1
2
1 Introduction
achieve the standard of western civilization, which requires much more energy as well.
Consequently, other energy sources, that both accomplish the increasing energy consumption and present a CO2 –neutral technology, are essential. One famous CO2 –free
technology is the nuclear power plant. Due to several aspects, this is not a real long term
alternative. Of course, uranium – as coal, oil and every other fuel – is of limited availability.
Furthermore, there is still the unsolved problem of final storage for the radioactive waste.
So far, no country in the world has storage for the ultimate waste disposal of high-level
radioactive waste [3]. Needless to mention the omnipresent danger of nuclear accidents.
So what is needed is real renewable energy, i.e. technologies based on energy that is
inexhaustible. Thereof are available on earth geothermal energy, energy from interaction of
earth and moon (tidal forces), and solar energy. The latter can be subdivided in hydropower,
wind power, biomass/biofuel and photovoltaic/solar thermal power plants/solar heating
systems (eventually all these forms of energy are created by the sun, i.e. solar energy).
The perhaps most promising renewable energy is solar energy, as it can potentially
cover the world’s energy consumption [4]. However, photovoltaics today have not yet
reached so called grid parity, which means electricity from solar cells is more expensive
than energy from conventional sources like coal or gas. Consequently, there is further
research essential to increase the efficiency of solar cells and to make them cheaper. One
approach to this is the thin film solar cell. Thin film solar cells have a thickness of only few
micrometers (regarding the absorber), which means that much less material is used (saves
energy and money). Further possible savings result from the easier production process,
which means it is very much automatable. For example is the absorber material directly
applied to the substrate (by sputtering, evaporation or the like), so that a complicated and
material-consuming process like the sawing of silicon wafers from ingots can be omitted.
So far, three thin film materials have become industrially produced solar cells: Amorphous silicon (a-Si), Cadmium telluride (CdTe) and Copper-Indium-Gallium-Selenide/Sulfide
(CIGS), whereof CIGS reached the highest efficiencies and can compete with polycrystalline
silicon [5]. Admittedly also CIGS will have to face some difficulties. One problem is that
Indium is a rare element and could run low within the next 10-20 years, while the price is
already now increasing rapidly [6].
That means that further research has to be done, and one approach is the material
CZTS, which is the topic of this diploma thesis. CZTS is an abbreviation for Cu2 ZnSnS4 ,
i.e. it is a compound semiconductor made of copper, zink, tin and sulphur, which are in
each case for the time being sufficiently abundant elements, none of them harmful to the
environment in the used amounts. Although it is a comparatively new material, there are
already promising results that indicate that CZTS could be used as a solar cell absorber
material. The world record today is 6,8% achieved by IBM [7].
The aim of this study is to fabricate Cu2 ZnSnS4 by sputtering the metal precursor
and subsequent annealing in sulphur atmosphere. The influence of parameters like metal
composition or the presence of sulphur in the precursor before annealing will be studied for
example regarding grain size, surface smoothness and homogeneity. At the end, applicative
material will be used to fabricate complete solar cells.
2 Theory
Within the renewable energies, solar energy is the most promising. As can be seen from
Fig. 2.1, the direct sunlight presents the by far biggest source of renewable energy. The
technically usable potential of renewable energies is by the factor six higher than what is
needed today, whereof solar energy amounts to 65%.
One of the most promising techniques are solar cells, which combine several advantages.
They can be used more or less in any dimension, from the small one in a calculator up to
solar power plants in the GW range. This also makes solar cells an autonomous source
of energy. It is possible to cover the energy demand of small villages or street lamps at
bus stops, which is especially interesting if it is not possible or too expensive to connect
them to the grid. Moreover, it is a technology that is quiet, has no emissions and that
has no moving parts, which makes it a technology with a quite long lifetime. Producer
give guaranties of 20–25 years, but in principle much longer lifetimes are possible [8]. For
this reasons, solar cells are part of a growing industry [9] (see Appendix, Fig. A.2) with a
decreasing development of costs [10].
Figure 2.1: Worldwide energy demand (grey), existing (big cubes) and usable (small
cubes) renewable energies. The latter includes structural and ecological restrictions, as well
as limited efficiencies of the available techniques. Solar energy on its own would be enough
to cover world’s energy consumption. From [4].
3
4
2 Theory
2.1 Solar cells
Solar cells directly convert radiation into electricity. All solar cells are based on semiconductors. The radiation produces electron-hole-pairs in the semiconductor, which are
segregated by a voltage. Those charge carriers can then perform work when the solar cell
is connected to a load. In the following it will be explained how the needed voltage is
generated, and on which principles a solar cell is based.
2.1.1 Solar radiation
The sun presents from the physical point of view a black body, which means it absorbs
all kinds of radiation completely and itself emits a characteristic, temperature-dependent
black body radiation. For the temperature on the surface of the sun, the spectrum looks
like the orange line in Fig. 2.2. The maximum radiation intensity is around 500 nm.
However, the sun is not an ideal black body, and so the spectrum that reaches the earth’s
atmosphere looks like the black curve. The radiation power there is 1353 W/m2 [11], and
the radiation on earth but outside of the atmosphere is denoted with AM0. AM stands
for Air Mass and indicates how far the radiation has to travel trough the atmosphere, so
in this case 0. The radiation on earth’s surface with perpendicular incidence of light is
denoted with AM1. Crucial for a solar cell is of course the actual incoming light, and the
standard spectrum for measurements on solar cells is AM1,5, which means an angle of
incidence of ca. 48°. Standard test conditions (STC) used for certification of solar cells
and modules is the AM1,5 spectrum with a power of 1000 W/m2 and cell temperature of
25°C. The AM1,5 spectrum is the lowermost, blue line in Fig. 2.2. Absorption by water,
oxygen, ozone and other molecules in the atmosphere causes the peak pattern. In that
way also the wavelength for the maximum photon number shifts slightly. Of course it
is necessary to make optimal use of the sunlight, so that a solar cell is optimized (for
example by anti-reflection coating) for its application (space, satellite, roof tops, etc.).
2.1.2 Intrinsic, p– and n–type semiconductor
In all semiconductor solar cells, the voltage occurs in the contact area of a p- and an n-type
semiconductor. This shall be shown here using the example of silicon, as it is particularly
good understandable here how a p- and an n-type semiconductor are formed.
Silicon is a tetravalent element. In a 2D-projection of a silicon crystal, silicon atoms
form a lattice like shown in Fig. 2.3 a). Every atom has four bindings. If a silicon atom
is substituted by a pentavalent atom (group V element) like phosphor or arsenic, there
is one electron left that cannot form a covalent bond (Fig. 2.3 b) ). This electron is only
2.1 Solar cells
5
Figure 2.2: Spectral irradiance of the sun for different wavelengths. The orange line is the
spectrum of a perfect black body at the temperature of the sun (5800 K), the black one is
for extraterrestrial radiation from the sun (AM0), and the lowermost curve (blue) shows
the sun spectrum that reaches earth under an angle of ca. 48° (AM1,5). Data from NREL
[12].
very weakly bonded1 , so that it is possible to remove it from its atom with little energy,
i.e. room temperature can be sufficient. If a fraction of the silicon atoms in a silicon crystal
is substituted by pentavalent atoms, this is referred to as n-type doping. n stands for
negative, as there are negative charges (electrons) that can be removed easily from their
atoms. That is also why the doping atoms on the n-side are called donors. Note that the
n-type doped material itself is neutral, not negatively charged!
If a silicon atom is on the other hand substituted by an element with three valence
electrons, there is something called hole, because there is one electron missing to form four
bonds (Fig. 2.3 c) ). Such a trivalent atom in silicon is called acceptor, because it would
be able to accept one more electron. If electrons move over to this vacancy, it can be
treated like a positive charge because it looks like the hole moves in the opposite direction.
Materials that exhibit quasi free positive charges (holes) are called p-type doped, because
1
4
me ·e
1
Formula for energy level of an H-atom: E = 2(4πε
2 · n2 , with me : electron mass, e: elemental charge,
0 h̄)
n: shell. For n=1, the ionization energy is 13,6 eV. In this particular case, the electron mass has to be
substituted by the effective mass m* = 0,3 me , because it is not a free electron but influenced by the
lattice, and the dielectric coefficient of the vacuum ε0 has to be substituted by the one of Si. From this
it follows that the ionization energy is ≈ 30 meV. From [13].
2 Theory
6
of the positive hole. As for n-type semiconductors, p-type semiconductors do not exhibit
excess positive charges.
If no doping is carried out the semiconductor is called intrinsic (Fig. 2.3 a) ).
2.1.3 Fermi energy, valence and conduction band
In theory, a crystal is an infinitely continued sequence of unit cells. The infiniteness
leads to some unique phenomena; one of those is the formation of bands.1 As there are
numberless amounts of energy levels in an atom, there is also a large number of bands
in a crystal. Anyway, only the bands around the Fermi level are interesting.2 All states
up to the Fermi level are occupied with electrons. The highest band where electrons are
present at zero temperature is called valence band, the next band above (empty at zero
temperature) is called conduction band [13]. If the valence band is completely occupied,
electrons cannot take up any small amount of energy, which means that no current can
Figure 2.3: a) Intrinsic, b) n-doped and c) p-doped silicon lattice. Donors (n-type) have
one an electron that does not have a partner to form a bond and which can easily be removed from the atom. Acceptors (p-type) form one bond less compared to the intrinsic
case as they have too few electrons. The remaining vacancy is called hole.
1
The deeper reason for this is the following: In an atom, all electrons sit on so called energy levels, which
means they cannot have just any energy when bounded to an atom. Each level can only be filled with
one electron or rather with two electrons of different spin, according to Pauli Exclusion Principle that
says that no two equal electrons may have the same energy (more precisely: same quantum numbers).
However, in a bond of equal atoms (like it is the case in silicon), electrons on the same energy level
come very near to each other, so that the energy levels have to split up slightly. In a perfect, infinite
crystal, though, there are so many electrons, that all the slightly splitted energy levels form quasi
continuous bands. The different bands, belonging to different energy shells, can be divided by so called
band gaps Eg , i.e. energetic regions where electrons are not allowed to be.
2 The electrons in a crystal (like in an atom) fill up the energy levels/free states, of course beginning with
the lowest. As only two Fermions (which electrons are) may occupy the same level (Pauli Exclusion
Principle), higher energy levels have to be filled up as well; only two electrons can sit on the lowest level.
In that way, little by little from the bottom up all energy levels are filled up until the last electron
has found a place. The energy that the last electron needs to occupy a place is called Fermi energy or
Fermi level.
2.1 Solar cells
7
flow. If the conduction band is ’considerably’ (often means Eg more than 3 eV) apart from
the valence band, this refers to an insulator. If valence and conduction band overlap or
the valence band is occupied only partly, this is called a conductor. A semiconductor is a
case in between, when a band gap between valence and conduction band is existing (at
zero temperature an insulator), but it is so small that some electrons are lifted up to the
conduction band at room temperature so that it is occupied with some few electrons [13].
2.1.4 Formation of the space charge region at the p–n–junction
In an intrinsic silicon crystal, only very few electrons are in the conduction band at room
temperature. To increase the charge carrier density the material is doped. In n-doped
material, electrons are the majority carriers, which means that at room temperature
predominantly electrons carry the current. The same is the case in p-doped material
for holes. Fig. 2.4 shows the band structure of a semiconductor. The red line is the
Fermi level, the dashed line the doping level. For the n-side, the energy level for the
doping atoms is called ED , for the p-side EA . The blue balls symbolize electrons in the
conduction band, the purple ones holes in the valence band. To the left in a) there is
an n-type semiconductor. As most electrons come from the doping level just under the
conduction band, electrons represent the majority carrier (only few holes in the valence
band). The Fermi level is between doping level and the conduction band as there is the
highest occupied electron level at zero temperature now. The same applied to the case of
p-type semiconductor in the middle (b)), with holes and valence band instead, though. To
the right (c)) the intrinsic band scheme is shown, with the Fermi level in the middle of
valence and conduction band. It isexactly in the middle in case of zero temperature and
equal density of states.
If the n-doped and the p-doped semiconductor are interconnected – called p-n-junction
–, the following happens: the Fermi level has to be the same everywhere in the crystal as we
consider thermodynamic equilibrium. The relative position of the bands with respect to
the Fermi level does of course not change. Consequently, a band bending occurs (Fig. 2.5).
The deeper physical reason for this is that electrons from the n-type semiconductor move
Figure 2.4: a) n-doped, b) p-doped and c) intrinsic semiconductor. The blue balls symbolize electrons in the conduction band, the purple ones holes in the valence band. The dashed
lines are the doping levels, i.e. ED for the n-doped and EA for the p-doped semiconductor.
The red line is EF and the energy difference between valence and conduction band is Eg .
2 Theory
8
to lower energetic levels in the p-type material. By this, an electron excess in the p-type
semiconductor is generated (at the same time an electron depletion in the n-type material
near the contact, Fig. 2.6), which leads to a voltage that raises the bands of the p-type
side. This voltage is the crucial point: It falls across the region where the ionized n-doping
atoms lost their weakly bonded electrons to the p-region where the electrons ended up.
This region is called the space charge region (SCR) or depletion region, as it is depleted of
free carriers.
Looking at Fig. 2.5, one can directly see what happens with electrons that are generated
in the p-type material and reach (by diffusion) the space charge region: they fall down the
hill, or physically expressed drift in the built-in field of the space charge region. In this
way the separation of charges takes place.
This electrical component is called diode.
2.1.5 Currents in a diode
In equilibrium there are two currents in the described system. One flows from the p- to
the n-side. It is caused by electrons that are generated in the p-side and reach by diffusion
the space charge region, where they feel the voltage and drift to the n-side. For this reason
the current is called drift current or reverse current jR .
On the n-side of the semiconductor, the concentration of free electrons is considerably
larger than on the p-side. This difference in concentration causes a diffusion current from
the n-side to the p-side and is consequently called diffusion current or forward current jF .
As we consider equilibrium, the two currents are equal.
Figure 2.5: Band-bending at the p-n-junction. To the left is the n-side, to the right the pside. Electrons (blue) on the p-side – which are minority carriers there – flow to the n-side
if they reach the space charge region and feel the electric field. The same applies for the
holes (purple) on the n-side.
2.1 Solar cells
9
Figure 2.6: p-n-junction and space charge region. The blue cubes are positively charged
donors, the blue balls electrons, the purple cubes negatively charged acceptors and the purple balls represent holes. The space charge region consists of ionized donors and acceptors,
which originate from the electrons that moved from the n-side to the energetic lower levels
of the acceptors at the n-side. Between the ionized atoms an electric field occurs (symbolized by arrows).
2.1.6 IV characteristics of a diode
If a voltage V is applied to a diode there is no equilibrium anymore and so a net current
can flow. There are two situations possible, referring to the two possible polarities.
Situation A: The negative pole at the p-side (reverse direction)
The p-side in the space charge region already has an excess of electrons and is negatively
charged (we are talking about net charges, not about the amount of electrons in the
conduction band, see chapter 2.1.4). This leads to an increase of the conduction band
on the p-side compared to the n-side (see Fig. 2.7 a) ), namely about the value e · V . By
this, the energy barrier for the electrons on the n-side becomes even bigger, so that the
diffusion current vanishes at some point. The drift current, however, is not influenced by
the voltage. Electrons that reach the space charge region are attracted as before. These
considerations lead to the formula
j (Vex ) = − (jR (C) + jR (V ))
(2.1)
(with j(Vex ): current in the diode for the voltage Vex ; jR (C): drift current in the
conduction band; jR (V ): drift current in the valence band).
This polarity is called reverse direction.
This applied voltage leads furthermore to an increase of the space charge region,
as in this situation more electrons from the n-side are pulled to the (negative)
p-side. Considering Fig. 2.6, this means more positively charged donors and
2 Theory
10
negatively charged acceptors and therefore a larger space charge region. For
good material this effect makes no difference for the performance of a solar cell,
but for bad material with low diffusion lengths of the electrons this increases
the number of collected electrons and in doing so the current, as more electrons
can reach the larger SCR. This phenomenon is referred to as voltage dependent
carrier collection and causes a deviation from the standard diode curve since
an increasing (negative) voltage results in an increasing (negative) current.
Situation B: The negative pole at the n–side (forward/conducting direction)
If the negative pole is connected to the n-side, the band bending is reduced (Fig. 2.7 b) ).
More free electrons from the conduction band on the n-side can now diffuse to the p-side
because the potential holding them back is lower. That means the diffusion/forward
current jF increases, and that exponentially:
e·Vex
(2.2)
j = −jR · e kB ·T
(with Vex : applied voltage, kB : Boltzmann-constant, T : temperature).
At the same time, the drift current is unchanged as long as some of the voltage drop
between p- and n-side is left, but it becomes with increasing voltage negligible as the
forward current increases exponentially. This is called forward or conducting direction of
the diode.
The resulting formula, taking into account that there are both electron and hole currents,
is
j (Vex ) = − (jR (C) + jR (V )) · e
e·Vex
kB ·T
−1
1
(2.3)
Figure 2.7: Band bending as a result of an applied voltage V. a) shows the negative pole
on the p-side. The potential wall increases and the diffusion current decreases. b) Negative
pole on the n-side, the barrier decreases and the diffusion current increases exponentially.
The drift current is in both cases not influenced (as long as some potential gradient from
the p- to the n-side is left).
2.1 Solar cells
11
(jR (C): reverse current from the conduction band, jR (V ): reverse current from the
valence band).
The resulting curve can be seen in Fig. 2.8.
A more precise formula includes the current from the space charge region:
e·Vex
e·Vex
j = − (jR (C) + jR (V )) · e kB ·T − 1 − (jR (C) + jR (V )) · e 2·kB ·T − 1
(2.4)
2.1.7 The illuminated diode
Another non equilibrium situation is when the device is illuminated. The incoming light
produces electron-hole-pairs. Those that are minorities (i.e. electrons in the p-type and
holes in the n-type material, respectively) drift to the other side and are available as
external current, if they can reach the space charge region.
Figure 2.8: Behavior of a diode under external voltage (IV characteristics). If the negative pole is applied to the p-side only a very small drift current flows. This is the reverse
direction of the diode. The opposite polarity leads to an exponential growth of the current.
1
The "-1" stands for the forward current jF . As it is equal to jR in equilibrium, this formula can be
e·Vex
written like this instead of j(Vex ) = −(jR (C) + jR (V )) · e kB ·T + (jF (C) + jF (V )), so that the formula
looks much clearer.
2 Theory
12
That has three consequences for the final solar cell: First, the space charge
region should be as wide as possible, as all charge carriers that are already in
the space charge region can feel the potential and add to the current; second,
the space charge region should be near the surface or rather near to where the
light comes in, because much light is already absorbed near the surface; third,
the diffusion length of charge carriers in the bulk has to be as large as possible
(which requires good material quality, i.e. large grains, few defects...), so that
many minority carriers can reach the space charge region.
The current that can be reached under illumination when the device is short-circuited
is referred to as short circuit current Isc . The theoretical maximum can be calculated
as follows: assuming that all photons reaching the solar cell per second create each one
electron-hole pair, this number times e (elemental charge: 1,6 · 10−19 C) is the current that
is possible without any losses. This calculation depends on the band gap, as all photons
with energy below the band gap produce no electron-hole-pairs [14]. For our band gap of
mA
roughly 1,5 eV (see chapter 4.3.2.2) we assume around 22-24 cm
2.
If the circuit is not closed a voltage is established under illumination. This maximum
voltage that is possible to achieve under illumination is referred to as open circuit voltage
Voc . The maximum Voc that is theoretically possible is difficult to estimate. Note that not
the whole band gap can be gained as Voc but only the built in voltage of the pn-junction
Vbi :
Vbi = q · (Eg − ∆ED − ∆EA ),
(2.5)
i.e. the band gap Eg minus the distance of the doping levels from the conduction (ED )
and valence band (EA ), respectively. This voltage is further reduced by recombination
losses in the bulk of the semiconductor and at interfaces.
This means for the diode curve that it is shifted down by the open circuit current and
intercepts the x-axis at Voc . That can be seen in Fig. 2.9.
Hence, the equation for the illuminated diode is now
j (Vex ) = − (jR (C) + jR (V )) · e
e·Vex
kB ·T
− 1 − jR (solar)
(2.6)
(with jR (solar): current density under illumination without voltage; i.e. jR (solar) · A =
−ISC , A: device area).
Nevertheless, the current that can be used is never Isc , as a circuit needs both current
and voltage. Thus, a point on the curve is searched that has the highest power (current
times voltage) that is possible. This operating point is called maximum power point, and
the associated voltage and current are Vmp and Imp , respectively. The ratio between the
theoretical power Isc · Voc and the maximal possible power is called fill factor FF :
FF =
Imp · Vmp
.
Isc · Voc
(2.7)
2.1 Solar cells
13
Figure 2.9: IV characteristics of an illuminated diode. The curve is shifted down by the
light-induced current Isc and intercepts the voltage axis at Voc . The maximum power is
received for Imp and Vmp . The rectangle at this point is called fill factor FF and is aimed
to be as large as possible, corresponding to the maximum power that is possible to get.
That means a fill factor as high as possible is aimed for. Graphically, the fill factor is the
area of a rectangle within the IV-curve, determined by the maximum power point (which
is represented by Vmp and Imp ), see Fig. 2.9. The efficiency of a solar cell, i.e. how much
of the incoming light can be converted into electrical energy, is
η=
Pout
F F · Isc · Voc
Imp · Vmp
=
=
Pin
Plight
Plight
(2.8)
(with Pout : gained electric power, Pin = Plight : incoming power in form of light).
Isc , Voc , Imp , Vmp , F F and η are the important parameter that characterize a solar cell. FF
and η are redundant, though, but allow often a quick comparison of solar cells.
2.1.8 Equivalent circuit of a solar cell
Real solar cells cannot be described by equation 2.6. The reason for that is that a real cell
has resistances: the resistance of the bulk of the semiconductor material, of the contact
between metal and semiconductor and of the metal contacts themselves. These resistances
are in summary called series resistance RS and shall be as small as possible. Furthermore,
the cell can be shunted, which means that short-circuits exist across the p-n-junction in
the form of defects (crystal defects, impurities and precipitates), which is characterized by
the shunt resistance RSH . This one is supposed to be as small as possible.
2 Theory
14
That is, while an ideal solar cell could be modeled by a diode in parallel with a current
source, for a real solar cell a series and a shunt resistance have to be added [14]. The
equivalent circuit for a solar cell looks then like in Fig. 2.10.
A further approximation to the real solar cell is the two-diode modell, but this is not
discussed here.
2.1.9 Losses in solar cells
As already brought up in the last chapter, there are several losses that reduce the possible
output of a solar cell or its efficiency, respectively. They are described below.
1. Optical losses
Processes that inhibit that photons actually can produce electron-hole-pairs in the
cell are summarized as optical losses. As charge carriers that are never produced
cannot contribute to a current, these losses reduce the current Isc .
• Reflection: As long as the solar cell is not completely black, always some light
is reflected. The glass that is customarily on modules reflects light as well.
Reflection losses can be reduced by antireflexion coating and non-reflecting
glass.
• Shading: Silicon cells usually have a metal grid on the top to contact the cell.
This shades the active area of the cell between 5 and 15% [14]. CIGS thin film
cells do not have such a grid; they lose light by a TCO (transparent conductive
oxide), though, which is not completely transparent but absorbs some light.
Further area losses result from frames and interconnect zones in a module.
• Transmission: Especially photons with long wavelengths can be transmitted
through the absorber. All photons that have an energy lower than the band gap
Figure 2.10: Equivalent circuit of a solar cell. It includes next to the diode and a current
source I a series resistance RS and a parallel resistance RSH . According to [14].
2.1 Solar cells
15
are always lost, they cannot produce electron-hole-pairs and transmit the cell.
But also light that in principle could generate charge carriers can be transmitted
if the cell is too thin. The absorption coefficient is a factor that tells how strong
the absorption per depth is. It is for example for CZTS 104 cm−1 , i.e. a one
centimeter thick film of CZTS reduces the intensity by the factor 104 .
Anyway, every solar cell is wished to be as thin as possible to save material
and with it costs. One solution for this is a reflecting back-contact. For silicon,
aluminium is widely-used [15], for CIGS zirconium nitride (ZrN) is possible [16].
Nevertheless, thicker cells or back contacts can never completely inhibit losses,
as charge carriers can only travel a certain distance before they recombine, that
means they have to be able to reach the space charge region. That leads to the
next loss.
2. Recombination
Imperfections in the cell lead to recombination of the charge carriers, i.e. electron
and hole recombine and send out a photon. Those charge carriers are then lost for
the current, which means that this effect also reduces Isc . But it also has a big
influence on the voltage Voc which is the higher the lower the recombination is.
Recombination can occur at so called traps, which means impurity atoms like iron.
They simplify the recombination process as it is more likely for an electron to loose
its energy in small portions than completely at once. Especially traps in the middle
of the band gap (which are e.g. for silicon: copper, iron and gold [15]) are very
bad for a solar cell. Further recombination centers are crystal defects like grain
boundaries.
Recombination in the depletion region reduces also the fill factor.
3. Thermalization
Photons that exhibit energy higher than the band gap can nevertheless only produce
one electron-hole-pair. The excess energy is lost by thermalization, which means
that the created electron in the conduction band emits photons that only heat up
the device. Impact ionization, that is that one of the emitted photons creates a
second or even more electron-hole-pairs, is negligible.
Thermalization and Transmission (see point 1. Optical losses ) present an optimization problem. On the one hand, the band gap should be as high as possible, to reduce
thermalization. On the other hand, a low band gap would be desirable to collect as
many photons as possible witout loosing them by transmission. The optimum band
gap is somewhere around 1,4-1,5 eV, which would allow a theoretical efficiency of
around 30% [14]. The only possibility to avoid this problem is a multijunction cell,
i.e. several cells with different band gaps stacked.
2 Theory
16
4. Electrical losses
As mentioned in the last chapter, series resistance RS (which results from the internal
resistance of the used materials) and shunt resistance RSH (which results from short
circuits across the p-n-junction) decrease the efficiency of a cell. At a series resistance
the voltage drops. The parallel resistance results in a reduction of the FF since the
diode curve is influenced by an ohmic resistance in parallel. For very low RSH there
is also a loss in Voc (see Fig. 2.11).
As can be seen from Fig. 2.11, the fill factor is reduced by both resistances.
Figure 2.11: Impact of a) RSH and b) RS on the diode curve. a) If a diode is shunted,
the current can partly flow back within the diode, i.e. this resistance lowers I: ∆ I = V/RSH .
If the diode is (almost) completely shunted it is a ohmic resistance. This means also a
drop of the Voc . b) At a series resistance the voltage drops: ∆ V = I· RS . The voltage drop
for currents smaller than Isc leads to more linear IV-curves. In cases of very large series
resistances, Isc can be reduced.
In both cases the loss of the rectangular form of the diode curve results in a loss of the FF.
Figure according to [17].
2.2 Thin film solar cells
17
2.2 Thin film solar cells
All the advantages of solar cells should not hide the fact that there are still some issues
that have to be worked on. Like wind power, photovoltaics base on an inconstant energy
supply, which poses the problem of energy storage. Furthermore, so far solar electricity is
more expensive than conventional produced electricity. There is until today no so called
grid parity [15]. To solve the latter problem, solar cells have to become cheaper to shorten
the economical payback period and make electricity cheaper. One approach is to use less
material, which leads to thin film solar cells. Compared to crystalline silicon solar cells,
much less material is expended. While crystalline silicon needs 200 cm3 (200 µm · 1m · 1m)
material for 1 m2 solar cell, only 1 cm3 is needed for thin film material (the production of
the pure feedstock requires the main part of energy consumption during the production
process of a solar cell). Furthermore, silicon has losses of more than 50% of the material
when it is sawn from the ingots [15].1
Another advantage of thin film solar cells is the monolithic integration. That means
that the serial connection of the cells in a module (which is always necessary because of
the low voltage in one cell) is directly done during the fabrication of the cells and does
not need an individual production step, saving money and time.
One more advantage is that it is possible to adjust the band gap in some materials
like CIGS (CuInGaS(e)2 ) by varying the composition. By this, the solar spectrum can
be utilized much better and a higher efficiency can be reached, because the theoretical
possible efficiencies depend strongly on the band gap, as can be seen from Fig. 2.12.
As thin film solar cells today can compete in efficiency at least with polycrystalline
silicon (20,4% efficiency for multicrystalline Si and 19,4% for CIGS, [5]), the proportion of
thin films of the whole photovoltaic module production is increasing [9].
2.2.1 Device structure and fabrication techniques
In spite of the differences of the various semiconductor materials, many thin film solar
cells have a similar device structure, which shall be shown here with CIGS as an example.
Fig. 2.13 shows a cross section of a basic CIGS solar cell as described below.
As thin film solar cells are so thin and to protect the back side, they have to be deposited
on a substrate. One advantage of the thin films is that they can be deposited on flexible
materials like metal foils or polimides (plastic), which allows completely new applications.
Anyway, still the most common substrate is glass.
On the substrate, some kind of back contact is needed. The demands for a back contact
are good conductivity, a good work function and stability against corrosion, oxidation etc..
1
Mass yield ingot to column: 70%, mass yield column to wafer: 60%. ⇒ ca. 58% loss. Indeed do thin
film techniques have losses in the order of 50% as well, as not only the substrates are coated but also
the surrounding area, but this is as well 50% of 2 cm3 compared to 50% of 400 cm3 . In both cases, the
material can quite easy be recycled.
18
2 Theory
Figure 2.12: Theoretical efficiencies of solar cells for different band gaps. One maximum
is around 1,4 eV. Band gaps of different solar cell materials are marked. From [18].
For CIGS, molybdenum has proven suitable. The thickness is about half a micrometer
and the molybdenum is commonly sputtered on the glass.
On top of the back contact follows the most important part of a solar cell, the absorber,
where the main part of the electron-hole-pair production takes place. It is eponymous for
a solar cell. Both CdTe and CIGS are p-doped, the latter by intrinsic defects and not
by extrinsic doping like in Si. Various techniques are possible to deposit the material:
(co-)evaporation, (reactive) sputtering, CVD (chemical vapor deposition) and several more
[18]. For CIGS, co-evaporation of the four components (Cu, In, Ga, S/Se) is perhaps
the most prevalent method. Other methods like sputtering require after the deposition a
second step called sulphurization/selenization, to form the final material from the metal
precursor. The final absorber has a thickness of about 2–4 µm.
On the absorber often follows some kind of buffer layer. It can have several functions,
for example improving the lattice matching between the absorber and the n-doped-layer
on the top. For CIGS, the buffer layers are cadmium sulfide (CdS; n-type buffer layer) and
intrinsic zinc oxide (i-ZnO); the reason for improved performance by adding these layers is
nevertheless not yet completely understood [19]. CdS is normally deposited by chemical
bath deposition (CBD) and has a thickness of ca. 50 nm; ZnO with an approximately
double thickness of 100 nm can again be sputtered.
The buffer layer(s) are then capped by the n-layer. This is done to form the p-n-junction,
but at the same time it is used in CIGS, CdTe and a-Si cells to carry away the charge
carriers, while they are not allowed to absorb to much of the incoming light. This layer
is mostly a TCO (transparent conducting oxide), a so called window layer. The name
results from the transmissibility for visible light. In case of CIGS this is done with heavily
Al-doped ZnO (ZnO:Al), band gap ≈ 3,3 eV [18]); the heavy doping provides the needed
good conductivity. The ca. 300–500 nm thick layer can be deposited by sputtering.
2.2 Thin film solar cells
19
For further improvement, an anti-reflection coating is possible, to increase the amount
of incoming light. Moreover, a reflecting back contact (as described in chapter 2.1.9) could
be added to have less loss by transmission. Fig. 2.13 a) shows a cross section of a basic
CIGS solar cell as described above, in b) a cross section of a CZTS film can be seen.
Figure 2.13: a) Cross section of a typical CIGS solar cell with the different layers
(schematic). b) SEM cross section of a CZTS film.
2.2.2 Possible materials
Material for thin film solar cells has to fulfill some important conditions to be usable.
An essential precondition is of course a large absorption coefficient, as all (suitable) light
should be absorbed in only a few micrometers. Furthermore, the band gap should be
in the range of roughly 1–1,6 eV (see Fig. 2.12) to provide the theoretical opportunity
to reach sufficient efficiencies. Anyway, quite a lot materials fulfill these conditions (see
Fig. 2.14). On the basis of silicon, one can deduce at first the III–V– (like GaAs) and II–VI–
semiconductors (like CdTe) [18]. Further compound semiconductor can be formed by
substituting in the latter one half of the group-II element with a group-I and one half with
a group-III element. A common example for such an I–III–VI-compound semiconductor is
CIS (CuInS2 ) or – replacing partly the Indium by Gallium to modify the band gap – CIGS
(CuInGaS2 /CuInGaSe2 ). Various more substitutions are possible, for example replacing
half of the group-III element with a group-II element and half with a group-IV element.
For CIGS, substituting In/Ga with Zn and Sn, this leads to CZTS (Cu2 ZnSnS4 ).
Nevertheless, not all thinkable compounds give viable solar cell materials. A lot more
conditions have to be fulfilled, like availability, producibility in industrial scale, costs
and environmental safety (e.g. toxicity). Thus only few materials actually made the step
from an interesting semiconductor to a solar cell ready for the broad market: amorphous
silicon (a-Si), cadmium telluride (CdTe) and copper indium (gallium) sulfide or selenide,
respectively, (CIGS).
20
2 Theory
Figure 2.14: Various possible compound semiconductors, obtained by gradual substitution
of elements by elements of groups from higher and lower group numbers. According to [18].
CIGS is at the moment the industrial used thin film material with the highest efficiency
and can compete with multicrystalline silicon [5]. The production technique is a bit more
difficult than for CdTe, but still cheaper and less material consuming than for silicon. But
although only little material is used, availability of needed material could become one of
the big problems for CIGS thin film solar cells. Especially indium will run low within the
next years. The consumption is already now higher than the production, which cannot
just be increased because it is only a by-product of zinc mining. According to estimates
there are reserves for 6 years and resources for approximately 15 years [6]. In this context
the price for indium increased massively, for example about 463% from 2001 to 2004 [6].
Of course the present situation will be intensified in future by the fact that other industries
use those resources as well; just as an example, indium is very much used in flat screens
(ITO, Indium Tin Oxides).
The named problems of the current thin film solar cell materials indicate that further
research has to be done, and one approach is the material CZTS. CZTS is a compound
semiconductor made of copper, zink, tin and sulphur, which are in each case for the time
being sufficiently abundant elements, none of them harmful to the environment in the
used amounts. Although it is a comparatively new material, there are already promising
results that indicate that CZTS could be used as a solar cell absorber material. The next
chapter deals with the theoretical foundations of CZTS.
2.3 CZTS
21
2.3 CZTS
2.3.1 Properties
Cu2 ZnSnS4 (CZTS) is a p-type semiconductor with a direct band gap of approximately
1,5 eV. It is suitable for thin film solar cells due to its high absorption coefficient of more
than 104 cm−1 1 [20].
As mentioned before, CZTS is derived from the CIGS structure by the isoelectronic
substitution of two In (or Ga, respectively) atoms by one Zn and one Sn atom. As a
consequence, CZTS has some similar properties as CIGS. One main advantage of this is
that the standard device structure of the solar cells, shown in Fig. 2.13, can be adopted.
Of course it is not sure that the combination of CZTS with CdS and ZnO yields the best
results that are possible for this absorber material, but it allows starting directly without
spending too much time in searching for a working device structure. Instead one can
concentrate on the properties of CZTS and leave subtleties of the solar cell structure for
future work.
The crystal structure of CZTS is shown in Fig. 2.15. It is referred to as kesterite (space
group I4) and can be derived from the sphalerite2 structure by duplicating the unit cell.
The kesterite structure was found to be the most stable phase of CZTS [21]. The lattice
constants for CZTS are a = 0,54 nm and c = 1,09 nm [22]; from that one can calculate with
g
the atomic masses of Cu, Zn, Sn and S [23] the density of CZTS, which is ≈ 4,6 cm
3.
Figure 2.15: Kesterite structure in which CZTS crystallizes. It is derived from the sphalerite structure by duplicating the unit cell. From [24].
1
2
That means, for example, that a CZTS film with a thickness of 1 µm absorbs 99% of the incoming light.
Zinkblende
22
2 Theory
The doping of this material occurs by internal defects. Cu-atoms sitting on the places
of Zn atoms (Cu on Zn antisite) causes p-conductivity [25]. That means that one would
not necessarily aim for stoichiometric CZTS. Small deviations from stoichiometry lead
also to the formation of secondary phases, though. Which secondary phases may develop
can be seen from a phase diagram. It shows the phases that can be present in equilibrium
for certain temperatures and material contents.
2.3.2 The ternary phase diagram (TPD)
Such a phase diagram is of course possible for CZTS as well, but since it consists of four
kinds of atoms this would need a three dimensional diagram. However, one can assume
that always the right amount of sulphur is in the film as the sulphur is introduced by the
reactions with the metals and therefore depends on how much of those are present. This
assumption will be supported by our measurements where all of our samples contained
≈ 50% sulphur.
This reduces the degrees of freedom of the system to three and the phase diagram can
be simplified to a ternary phase diagram (TPD). In this study the TPD developed by
Scragg [26] on the basis of comprehensive measurements done by Olekseyuk et al. [27] is
used. It should be noted that this phase diagram is valid in equilibrium at 400°C. Both is
strictly speaking not the case for the experiments performed in this work. Anyhow, as
the sulphurization process used in this work comprised very slow ramping (< 0,15 °C/s)
and a long dwell time (2h at 520°C) we assume to have a quasi-equilibrium. Furthermore,
other experiments at comparable conditions (e.g. [28]) obtained secondary phases that
are predicted by this phase diagram. Therefore we will use it on a number of occasions,
mostly just to depict the compositions of the samples (i.e. not taking into account any
precondition except of having the right amount of sulphur), but partly also to support or
deduce assumptions concerning composition and secondary phases.
The ternary phase diagram is shown in Fig. 2.16. As can be seen from the scale this is
only a part of the whole diagram. In order to provide a better overview and as no samples
with a metallic ratio outside this section were produced, it will always be shown a zoomed
in version.
There are ten fields drawn in the phase diagram. Each field means the presence of CZTS
plus the one or two secondary phases that are noted. The eleventh region quite in the
middle (marked with an asterisk) means that only CZTS is supposed to be existing. All
secondary phases contain sulphur. As a sufficient amount of sulphur is provided during
the sulphurization process it is assumed that no metallic phases form but only sulfides.
However, not all secondary phases that have been found in the diverse studies have been
found by Olekseyuk, probably due to different conditions. One important secondary phase
that will play a role in this work as well is SnS2 . It was for example found by Schurr et
al. [29] and could be found for films with Sn excess.
Talking about regions in the phase diagram it is very helpful to divide it into regions
that are labeled in an unambiguous way and that already indicates which secondary phases
2.3 CZTS
23
Figure 2.16: Ternary phase diagram of CZTS. A fraction of 50% sulphur is assumed. In
the different regions indicated in the phase diagram, secondary phases appearing next to
CZTS are given. In the middle (marked with an asterisk) only pure CZTS occurs. Blue
arrows indicate lines of constant Zn, Sn or Cu ratio, respectively, in this case chosen for the
ratios that mean stoichiometry. According to [26].
can be expected. This shall be done according to the notation of Scragg (Fig. 2.17). In
the Zn-rich region, for example, ZnS is the expected (main) secondary phase formed by
the excess Zn. The Zn-poor region covers several fields with various possible secondary
phases. It should be noted that this notation is very clear, but different to the notations
used in most publications, where for example only Cu-poor and Cu-rich are distinguished
(e.g. [29]), not taking into account the ratio between the remaining metals.1
2.3.3 Secondary phases
As secondary phases can of course – depending on their fraction – have a big impact on
the characteristics of the cell, the most relevant for this thesis shall be specified in the
following.
1
Example: Schurr [29] describes a film with the ratios Cu/(Zn+Sn)=0,9 and Cu/Sn=2 as ’Cu-poor’.
Indeed this sample contains an excess of Zn, while the Cu/Sn ratio is stoichiometric. This is why this
sample would according to Scragg be referred to as ’Zn-rich’, and this is how it will be done in this
work as well.
24
2 Theory
Figure 2.17: Ternary phase diagram with different regions of composition. The labeling is
done according to Scragg [26].
2.3.3.1 Cu(2) S
Copper sulfides can be expected in the Cu-rich as well as in the Sn- and Zn-poor region.
These secondary phases are metals, or semiconductors that are heavily doped by intrinsic
defects so that they act as metals ([30], [31]). The major hazard of Cu(2) S is that it shunts
the cell, meaning that front and back contact are connected within the cell so that the
current cannot be used for an external load. However, the copper sulfide has not to have
grains reaching through the whole cell to reduce the performance significantly. Conducting
phases within a solar cell can present a serious problem as they enhance recombination.
2.3.3.2 SnS2
Tin sulfide (SnS2 ) is a n-type semiconductor with a band gap of 2,2 eV [32]. This secondary
phase could work as an insulator, but if existing in larger amounts it is also possible that
it forms a second diode with opposite polarity to CZTS, which would act as a barrier to
carrier collection and reduce the fill factor.
SnS2 is not noted in the TPD but could be found in films with Sn-excess, i.e. especially
for Sn-rich and Cu-poor (and partly for Zn-poor) samples in the TPD.
As Weber found out [28] do tin sulfides due to their high vapour pressure evaporate
from CZTS films during sulphurization if they are not prevented from that (for example
by being covered by other phases).
2.3 CZTS
25
2.3.3.3 ZnS
Zinc sulfide is a secondary phase in the Sn- and Cu-poor as well as in the Zn-rich region.
Due to the high band gap this material could even be called insulator (3,54 eV), and
this means that the presence of ZnS can both reduce the active area (i.e. the area where
electron-hole-pairs are produced) and inhibit the current conduction in the absorber.
It crystallizes in the sphalerite and the wurtzite structure and presents in both cases a
semiconductor with a wide band gap of 3,54 or 3,68 eV, respectively [23]. As mentioned
before is the crystal structure very similar to the CZTS one. CZTS, as the compound CTS
(Cu2 SnS3 ), is a superlattice to sphalerite. This results in a serious problem concerning XRD
measurements, which are one analyzing method in this thesis. Actually, ZnS, CTS and
CZTS are not possible to distinguish by XRD, they are therefore in literature commonly
summarizing referred to as Σ–signal. Indeed do CZTS and CTS have an additional peak
compared to ZnS, but that means that ZnS as well as CTS can never be excluded to be
present only from XRD measurements. Only a look on the phase diagram might give
a hint which secondary phase is more likely; ZnS and CTS do not appear in the same
region of the phase diagram together but ZnS for Zn-rich/Cu-poor and CTS for Zn-poor
compositions.
2.3.3.4 CTS
Cu2 SnS3 (CTS) is a secondary phase that should according to the TPD appear for Zn-poor
phases. CTS shows metallic properties [33] which makes it like copper sulfides a very
detrimental secondary phase. As mentioned before can CTS in our experiments never be
proven or excluded, as XRD is our only method to identify secondary phases. It is only
possible to assume that CTS can be avoided by producing films that are further away
from being Zn-poor.
The phases Cu2 ZnSn3 S8 and Cu4 SnS4 are not discussed here. The former was not
reported besides the studies of Olekseyuk, the latter is supposed to be found only in
regions of the phase diagram where we did not produce any samples.
Summing up one can state that Cu(2) S and CTS would probably be the most detrimental
phases, while ZnS and SnS2 might be less harmful. However, CTS and ZnS will in our
study never be proven or excluded for sure.
2.3.4 Reaction path for formation of CZTS
Even though several experiments have been made on the reaction paths for CZTS (e.g.
[29], [26], [28]), for example by in-situ XRD, there is not one universally valid reaction
path known. First, all studies were done with different precursors and sulphurization
conditions. Second, as mentioned phases like CZTS, CTS and ZnS are hard to distinguish,
even if several analyzing methods are combined. Third, there are procedural uncertainties,
like the assumption that the situation can be "frozen" by rapid cooling and that the phases
present at a certain temperature can be analysed in the cooled state [26].
26
2 Theory
What they have in common is, however, that the reaction path starts with the elements
(or the compounds like ZnS) in the precursor, via formation of binary compounds (Cu6 Sn5 ,
SnS2 , ZnS, Cu(2) S, Cu5 Zn8 , etc.) to more complex compounds like CTS and CZTS. Due
to the very similar crystal structure it is plausible that the final CZTS is formed of ZnS
and CTS ([28], [29]). The reaction path requires of course the interdiffusion of all elements,
which depending on the precursor type is not always given from the beginning.
Another interesting aspect that could be shown is the diffusion of Cu to the surface
[34]. As Cu has a high diffusion rate, especially compared to sulphur [26], it does more or
less completely diffuse to the surface to form copper sulfides. This phenomenon means
also that copper sulfides are often found at the surface and can be removed by etching
[26]. The major problem is, however, that this process forms voids that are left by the
Cu atoms, even if the copper sulfides in a later stage react with the remaining phases to
CZTS.
To tackle this problem, one approach is the integration of sulphur already in the precursor.
As the reason for the Cu diffusion to the surface is the affinity of Cu to react with sulphur,
this issue can be alleviated if the Cu can at least partly react with sulphur in the bulk
directly from the beginning. The consequence would be less diffusion of the elements and
hence less dramatically changes within the film. Another beneficial effect associated would
be less expansion of the film due to less diffusion of sulphur into the film. Considering that
the film expands by a factor of more than two ([35]) and a lot of stresses and cracks can
occur during this, the sulphur in the precursor from the beginning could lead to a more
homogeneous growth of CZTS and in doing so to bigger grains and less voids. Katagiri
[36] could in this way improve grain size, uniformity and adhesion of the film substantially.
The best CZTS solar cells published until today included S in the precursor ([7], [37]).
The above described means that phases in and composition of the final film depend
on the precursor and the sulphurization process. According to the degree of diffusion of
the elements, the content of sulphur and several more aspects, the formation of CZTS
might be slower or faster and in doing so allow more or less loss of elements/secondary
phases by evaporation as well as influence the segregation of secondary phases (e.g. as
conglomerates).
A hypothesis about a possible reaction path for these experiments is derived in the
analysis part of this thesis.
2.3.5 Previous studies
In recent years the number of publications on CZTS rose. This is accompanied with an
increase of the best efficiency that could be found for this material. The current record is
6,8% by IBM [7]. The years before, Katagiris group dominated the progress for CZTS
and hold the world record with 6,7% [37]. Fig. 2.18 shows the development of efficiencies
for CZTS. It should be noted that the graph contains all fabrication methods, and for
this reason the efficiencies spread a lot. Besides that a clear trend to better efficiencies is
obvious.
2.3 CZTS
27
Figure 2.18: Development of the efficiencies published for CZTS. The red point shows the
highest efficiency measured in this study. Note that the graph shows only publications were
an efficiency was given, the total number of published CZTS related papers was higher.
There are various methods for fabrication of CZTS. Some produce directly CZTS (so
called single-stage processes, like thermal evaporation), but most processes use two steps:
First, the fabrication of what is referred to as precursor is done, a film that contains only
the metals or metals plus some amount of sulphur. The deposition of the precursor can be
performed by any physical or chemical method (see Table 2.1 for a list of several methods
including the maximum efficiency reached with this method so far). The elements can
be deposited as stacked layers (i.e. for example one layer of Cu, one of Zn and one of Sn)
or for all elements at the same time. The latter is indicated with a "co-", for example
"co-sputtering" where all elements that are supposed to be in the precursor are sputtered
at the same time and are hence alloyed.
The second step is the sulphurization, a process where the sample is annealed to
temperatures mostly in the region of 500°C in an atmosphere containing sulphur. This
can be elemental sulphur or for example H2 S.
The so far highest efficiencies were reached by a two-step-process with co-deposition of
the precursor: co-evaporation (6,8%) and co-sputtering (6,7%). This might be due to the
formation of stable alloy phases instead of losses by evaporation of elements/compounds
(see chapter 2.3.4). For this reason such a process (co-sputtering) was used for this work.
Next to the deposition method, manifold more aspects were studied (examples without
reference will be discussed in more detail below): The order of the layers in stacked
precursors [38], the influence of sulphur in the precursor, the composition of the precursor,
the sulphurization process (temperature, ramping, dwell time, pre-annealing,...), in-situ
measurements to reveal the reactions during sulphurizations [29], etching to remove
secondary phases from the CZTS film, and some more. The most important results shall
be briefly mentioned.
2 Theory
28
Electrochemical deposition
Sol-gel
Evaporation
Thermally co-evaporated
Co-Sputtering
3,4% [39]
1,6% [40]
5,5% [41]
6,8% [7]
6,7% [37]
Table 2.1: Some common deposition methods for CZTS and their reached maximum
efficiency.
• Katagiri’s group studied among other things the influence of the temperature [42].
The optimum temperature turned out to be 520°C.
• The composition of the films was subject to several studies. However, most concentrated only on the influence of Cu and gave as a result that films with Cu deficiency
gave the best results concerning efficiency and optoelectronic properties ([42], [43]).
• Grain size appears to be related to the Cu-content. Tanaka et al. [44] found increasing
grain size with increasing Cu/(Zn+Sn) ratio. This is related to the results that were
found for CIGS where the beneficial influence of Cu on grain size is well known (see
for example [45]).
• Copper sulfides that appear as secondary phases for films with Cu-excess could be
removed by KCN etching. However, this process can instead lead to undesirable
voids ([35], [34]).
• Loss of elements: Both loss of Zn and Sn (by evaporation, elemental or as secondary
phases) could be observed as a consequence of the sulphurization process ([42], [46]).
• The presence of sulphur already in the precursor showed beneficial effects on the
CZTS films. An increase in grain size, better adhesion to the substrate and smoother
films could be shown [42].
• Films with Zn excess and Cu deficiency (Cu/(Zn+Sn) ≈ 0,85, Zn/Sn ≈ 1,1–1,2)
gave so far the best results (see for example [37]).
It is worth to note that efficiencies up to almost 10% could be reached for CZTS
where a certain amount of selenium was added [47]. However, as this approach is
not part of this study it will not be discussed here any further.
2.3 CZTS
29
2.3.6 Aim of this study
The results of previous studies were of course considered in this work. We used the
process that showed so far the most promising results, namely co-sputtering1 , and used
sulphurization temperatures of 520°C. No precursors from the Cu-rich region of the phase
diagram were used to make films for solar cells.
This diploma work was supposed to explore the potential of CZTS as a new absorber
material for thin film solar cells and to establish CZTS in this group that so far concentrated
on CIGS devices. This included both the determination of sputter conditions that allow a
control of the precursor composition, as well as to find and arrange a way to sulphurize
these precursors. For the further processing of the solar cell the standard process for CIGS
solar cells was applied.
With regards to the content the main intentions of this work were to analyse the following
two aspects:
• Influence of sulphur in the precursor: one series with precursors that only contained
metals (referred to as metal- or metallic precursors) and one series with precursors
that had an additional content of sulphur (referred to as ZnS precursors) were
performed and compared.
• Influence of the composition: we studied films from 6 of the 7 regions of the phase
diagram (including stoichiometric). This shall give deeper insights which composition
is best and complement previous studies that mostly concentrated on the influence
of only one element (e.g. Cu-poor vs. Cu-rich films, [29]).
For analysis of the films scanning electron microscopy (used for: analysis of morphology),
energy dispersive X-ray spectroscopy (determination of composition), X-ray diffraction
(identification of crystalline phases) and X-ray photoelectron spectroscopy (compositional
gradients) were used. Further analysis of the solar cells included IV (current + voltage)
and quantum efficiency measurements (determination of possible losses).
1
The slightly better result of 6,8% for co-evaporation was published after these experiments started.
3 Experimental
3.1 Fabrication techniques
Fabrication of CZTS thin films in this study was performed by a two-stage process,
sputtering a metal precursor or a metal-sulphur precursor, respectively, followed by a
sulphurization process.
3.1.1 Precursor deposition by sputtering
3.1.1.1 Sputter process
To deposit the precursors on the substrate, the magnetron sputtering technique is applied.
It is based on momentum transfer from bombarding ions of working gas to the target
atoms. By this, the atoms are knocked out from the target and are transported to the
surfaces. When they get in contact with a surface, the atoms adhere, for example on the
substrate. As a result, the substrates get coated by a thin film of the target material.
DC and AC sputtering
A sputter system is set up as follows. The target, i.e. the source of the used material, acts
as the cathode and is connected to the several kilovolts high direct current (DC ) voltage
[48]. The anode is the substrate holder, or even the whole chamber like in the system used
here, and typically connected to ground. In the evacuated chamber, a so called working
gas is introduced. Often – and thus also in our case – Argon is used for this, as it is a noble
gas (i.e. not reacting with the sputtered atoms) with low ionization energy. Furthermore it
is low-cost and has a suitable mass for the momentum transfer to the target. Then a glow
discharge is initiated and maintained in the chamber when a critical voltage is reached.
The argon atoms get ionized by the high voltage and the electrons move to the anode (and
ionize on their way further argon atoms), while the positively charged ions are accelerated
towards the target and there knock out atoms. The aggregate state where electrons and
atoms exist separated from each other is called plasma.
Sputtering in this way – with DC voltage – works only for conducting materials. In
the case of insulating materials, the surface charges up because of the charge brought by
argon ions. To avoid this problem, alternating current (AC ) voltage is used. In our case,
ZnS and ZnO required this approach.
31
32
3 Experimental
Magnetron sputtering
In magnetron sputtering, a magnetic field is applied in order to confine the electrons in the
target vicinity. The electrons then suffer more collisions with argon atoms thus increasing
plasma density. That makes possible to reduce the working gas pressure, which, in turn,
lets more energetic particles reach the substrate and also results in higher deposition rates
[48]. Another positive effect of trapping the electrons near the target is reduced heating of
the substrate.
Co-sputtering
It is even possible to sputter several targets at the same time. This could be desirable
to dope a material by co-sputtering an appropriate small amount of the doping material,
or to achieve a homogenous mixture of several compounds. If a sufficient energy supply
is given – for example by heating the substrate –, even crystal growth already during
sputtering can be enhanced.
We used co-sputtering in order to reduce diffusion of the sputtered metals during
sulphurization.
3.1.1.2 Process used in this thesis
The metals copper, tin and zink were sputtered as precursor material. For this, a von
Ardenne CS 6005 was used. As this system has only room for two targets, a Cu/Sn alloy
(99,99% purity) was used for the one target, the other one was a Zn-target (99,994-99,995%
purity) or a ZnS-target (99,99% purity), respectively. The targets were produced by Kurt
J. Lesker Inc..
The sputtering was performed under a base pressure of between 3,8 · 10−6 and 1,5 · 10−7
mTorr. The working pressure was 2,7-6,7 mTorr with an Argon flow of 20 sccm (standard
cubic centimeters per minute) in the first series, and fixed to 3,3 mTorr in the second series
(Ar flow: 40 sccm).
5 samples were deposited on a heated substrate (300°C); all other depositions took place
at room temperature.
The sputtering was performed as co-sputtering, i.e. all metals were sputtered at the
same time, except of some films where multi-layers were deposited. The thicknesses of the
films were varied between 100 nm and 1 µm. The deposition rate was changed between
20 and 80 nm/min. First, films only from one target (Cu/Sn or Zn, respectively) were
sputtered to find out the rate for different settings. It turned out that the film thickness is
proportional to sputter time and power. After this, co-sputtering was performed, and it
emerged that the film thickness was approximate to the sum that one would expect from
the individual thicknesses of each target at the certain power.
Concerning the composition, we aimed for stoichiometry, but also wanted to have
variations in some samples to investigate the influence of non-stoichiometry such as Cudepletion. Anyway, it was not possible to set an exact composition, all samples ended up
with less Sn in the film than in the target.
3.1 Fabrication techniques
33
In this diploma thesis, two sputter series were performed. The first one contained 47
different films and was done with a Cu/Sn-target (ratio: 60/40) and a Zn-target. These
samples are referred to as metallic precursors. The second series included 44 films. The
same Cu/Sn-target and a ZnS target were used. Those samples from the second sputter
series are referred to as ZnS precursors. In this second series, also the substrate heating
was tested for 5 samples.
The thickness of the precursors was measured using a Dektak V 200-Si profilometer.
Storage
In the first series, the sputtered samples just lay in plastic boxes with no further precaution,
that means they were exposed to air, (sun-)light etc..
In the second series, two pieces of each sample (i.e. 2 pieces of SLG, Mo-SLG and Si)
already cut in the right size for the sulphurization process (ca. 0,5x1,25 cm), were stored
in a vacuum storage at (5x10−8 mbar). The remaining parts of the samples were stored in
plastic boxes in a nitrogen-flooded cabinet in the clean room, but not in vacuum.
3.1.2 Sulphurization
Quartz tubes
Due to the absence of an adequate oven that could be flown with H2 S or sulphur vapor, a
quite special method had to be accomplished. Small pieces were cut off from the samples
and sealed with sulphur powder into small quartz tubes, which then were heated in a
normal oven.
As the glass had to be sealed by hand, only small quartz tubes could be used, otherwise
they became too warm to hold them with hand. For this reason, only very tiny pieces of
the samples could be sulphurized, i.e. a size of roughly 0,5-0,7 times 1-1,5 cm.
The original quartz tubes had a length of 1 m (outer diameter: 1 cm, inner diameter:
0,8 cm). They were first cut in the middle. The remaining 0,5 m pieces were warmed in
the middle with a Bunsen burner and then carefully extended by pulling on both sides.
By this, the quartz tube was melt thoroughly and two quartz tubes with melted end were
made.
The quartz tubes with one closed end were then cleaned by immersing into a tank with
heated water (60°C) and detergent. After it they were shower rinsed with de-ionized water.
Preparations
The sulphur used for this sulphurization process was available in form of powder (Scharlau,
synthesis grade, 99% purity). Ca. 2–4 mg were weighted and put into the quartz tube.
The amount of sulphur ending up at the bottom of the quartz tube was unknown and
quite various, as on the one hand the weighting machine had an uncertainness of 0,1 mg
(2,5–10%), and on the other hand always some powder adhered at the inner surface of the
funnel used to pour the sulphur into the quartz tube (which was the considerably bigger
source of error compared to the measurement inaccuracy). In the end, 1–4 mg of sulphur
3 Experimental
34
could be assumed to be in the quartz tube. Then, the tiny sample piece was put into the
quartz tube. Samples on SLG and on Mo-SLG came into the same quartz tube, back to
back, the ones on silicon into one quartz tube by itself.
To provide a high vapor pressure and to have as little oxygen, carbon etc. in the quartz
tube under the reactive time in the oven as possible, a vacuum in the quartz tube was
requested. To preserve the vacuum in the quartz tube, it should be melt thoroughly in the
middle while it is still connected to the vacuum pump. So, the open end of a quartz tube
was connected with the flexible tube of a vacuum pump (varian SD-91 ) and evacuated for
around 20 minutes down to a pressure in the order of some mbar. Then – still connected
to the pump – the quartz tube was melted thoroughly in the middle.
Sulphurization process
The prepared quartz tubes could then be put into the furnace. There was room for
maximum 6 quartz tubes. The furnace was a simple upright cement cylinder with a hole
inside (Eurotherm 904 ). It was possible to run programs on this oven, i.e. ramping (raising
the temperature in a wished velocity, here typically 4,2–8,4 °C/min) and dwell time at a
certain temperature (or even on several stages) could be programmed. Setting used here
were 1h ramping and dwell for 2h at 520°C. All samples cooled naturally, this means that
the oven was switched off and the samples were taken out at room temperature.
To test if the vacuum is still present, a spark tester (TMF Spark Tester MK.3, Buckleys
(Uvral) Ltd.) was used. It induces plasma in the closed quartz tube if there is a vacuum
(see Fig. 3.1). Finally, the quartz tube was scored with a circular saw, to make it easier to
break it up later.
3.1.3 Processing of the solar cell
The processing of the CZTS films to solar cells included substrate preparation, sputtering of
the molybdenum back contact, deposition of CdS by CBD and sputtering of a ZnO/ZnO:Al
TCO layer. This process was performed according to the "baseline" process of the Ångström
solar center [49].
Figure 3.1: Spark tester induces plasma in the evacuated quartz tube.
3.2 Analysis techniques
35
3.2 Analysis techniques
3.2.1 Scanning Electron Microscopy (SEM)
SEM measurements are an important tool for the analysis of the morphology of films.
The possibility to immediately see structures gives a direct and intuitive approach, which
makes it to a popular analyzing method.
The functionality of an electron microscope is similar to an optical microscope, but
instead of light, electrons are used. The advantage is the much smaller wavelength and
the associated higher resolution. Typical electron energies are 10–30 keV, which refers
to significantly smaller wavelengths than for light. For example, electrons with 20 keV
like used in these experiments corresponds to a wavelength of 0,009 nm (light: ≈ 400–
h
h
700 nm), according to de Broglies formula λ = mv
= √2qmV
, where h is the Planck
−34
−31
constant (6,626·10
Js), m the mass of the electron (9,1·10
kg), q the elemental charge
(1,602·10−19 C), v the velocity and V the acceleration voltage of the electron [50].
In a SEM, electrons are emitted from a filament and accelerated to an anode. Like in
an optical microscope the beam is focused, but here with magnetic or electrostatic lenses
instead of glass. The electron beam is than scanned over the sample. Secondary and/or
backscattered electrons that are produced in this process are collected by a detector and a
picture is created [50].
In this study a LEO 440 SEM was used for EDS measurements (see next chapter) and
a LEO 1550 to make pictures of cross sections of the samples.
3.2.2 Energy Dispersive X-ray Spectroscopy (EDS/EDX)
EDS is a method to analyze the composition of a sample. It bases on the principle that
electrons hit the sample where they excite bound electrons to leave the atomic shells.
This process creates vacancies in the atomic shell, which are then filled up by electrons
from higher shells. Hence, X-ray photons with characteristic element-specific energies are
produced and can be detected to determine the content.
The characteristic photons are labeled with a Latin and a Greece letter. The Latin
letter corresponds to the shell where the electron has been knocked out (from the K-, L-,
M-shell etc.), the Greece letter indicates from which shell it has been substituted (α for
the next higher shell, β for the next but one shell etc.). For example, if an electron from
the K-shell is substituted by an electron from the L-shell, Kα X-rays are emitted.
When electrons impinge on a solid, they can interact in several ways: They can
inelastically interact with the matter, i.e. get reflected or deflected (which widens the
electron beam); they can interact with the atomic nucleus which produces bremsstrahlung
(continuous radiation); and the electrons can ionize the atoms which causes characteristic
radiation. The interaction volume of the electron beam with the material is a bulb as
shown in Fig. 3.2. As can be seen the penetration depth is in around one micrometer. That
means that the result of the EDS measurement might be dependent on the thickness of the
36
3 Experimental
film, especially if there are depth gradients. For example, a secondary phase segregating
as a sublayer below a film thicker than 1 µm would not or only partly contribute to the
result. This should be kept in mind.
To determine the composition of the analyzed sample is not that easy. The first problem
is that not only and not the whole characteristic radiation from the elements reaches the
detector. On the one hand, the bremsstrahlung caused by the deceleration of the electrons
in the material adds to the background of the spectrum. The software of the EDS program
has to filter out arithmetically this noise. On the other hand, the deeper the electrons
penetrate into the material, the more photons produced deep in the sample do not get
out of it. This attenuation is not the same for all elements but depends of course on
the energy of the photons, i.e. suppressed more for light elements. Consequently, higher
electron beam energies are not necessarily better, as they increase the excitation but also
the depth of penetration, which again increases the absorption.
Another fact that hinders the analysis is the overlap of lines of different elements due
to noise in the detector and other electronic components, and as the channels that count
certain energies have a certain width. For this reason, for each element different spectral
lines are taken into account and compared, as they have a certain ratio. To remove the
other named effects (background, absorption), complicated calculations have to be carried
out by the software, or reliable reference samples with known composition are needed.
Figure 3.2: Interaction of electrons with matter. The form of interaction is different for
different depths of penetration. For EDS, characteristic X-rays are used. Re depends on
the density of the examined material and on the electron energy [50]. In our case Re can be
calculated to be between 1 and 2 µm. From [50].
3.2 Analysis techniques
37
Anyway, the detection limit is typically not higher than 0,1% [48].
In the software used in this diploma thesis, elements or their peaks, respectively, are
identified automatically and the atomic and mass percentage is immediately computed.
Elements that are known to be not part of the sample or that are irrelevant for the
considerations (like in our case the silicon from the substrate) can be removed manually,
and the composition is then calculated for the remaining elements.
It should be noted that the height of the peaks is not linearly connected to the concentration, and not even the elements in one measurement can be estimated relatively
to each other just by the peak height. The excitation is different for varying incoming
electron energies, and different for the particular element. Furthermore the named effects
background (which is not the same for all energies) and absorption play an important role.
Often – and also in our case – the incoming and exciting electron beam comes from
an SEM where the EDS is attached to. The photons are detected by a silicon detector
doped with lithium. The incoming photons produce electron-hole-pairs, and the number is
linear to the photon energy. The pulses are in the following amplified and then sorted into
channels according to their energy. The resulting spectrum is then among other things
dependent on the intensity of the incoming electron beam, the atomic concentration, the
ionization cross section, the X-ray absorption coefficient in the material and the detector
efficiency, which leads to a complicated formula to analyze the data. [48]
The equipment used for the EDS measurements was a LEO 440.
3.2.2.1 EDS reliability measurements
EDS was one of the most important analyzing methods in this work. For an estimation of
the error range of the EDS results diverse measurements were performed.
• Same sample - same area: Deviation for exactly the same conditions.
• Same sample - different area: Exploration of lateral gradients.
• Same sample - different resolution: Also influence of different areas.
• Same sample - different date: Stability of the EDS measurement.
• Different substrates: Si, SLG and Mo.
The first case, where exactly the same area was measured, gave as one would expect
the lowest deviations: in average the results for the elements differed by 2% (standard
deviation divided by the mean value; average over all samples and elements).
For different regions on the same sample, where lateral gradients within the film could
come into play, the average deviation accounted for 4%. This slightly increased variation
compared to measurements on the same area might be explained by small lateral gradients
as well as non-uniform segregation of secondary phases within the film.
3 Experimental
38
For the same reason, i.e. to see if it plays a role which area of the substrate is measured,
compositions for different resolutions on the same sample were measured. This was done for
100x, 250x and 1000x magnification (250x was the default value used in our experiments).
No significant deviations could be observed, in average 2% difference was seen.
The most critical cases are when samples are measured on different days (i.e. it cannot
be guaranteed that exactly the same conditions prevail). In average, the mean values
of the different days differ up to ca. 10%. Even taking into account that most probably
different areas were measured (as mentioned this would mean ca. 4% variation) this is
quite a lot. This might be due to variations in the sensitivity of the detector and the exact
properties of the electron beam.
In Table 3.1 the results for the measurements on different substrates (Si, SLG, Mo-SLG)
are given. The last row shows the variation that was measured for the same sample but
on different days.
The compositional results between the (assumed) same films on Si and Mo as well as
Si and SLG were compared to each other and the average deviation between the values
was calculated. The difference is mostly below 6%, in one case up to ca. 11% deviation
between same films on different substrates.
This result means that the composition of equally processed films on different substrates
can be assumed to be equal, as the EDS results do not show more deviation between same
films on different substrates compared to the same film on the same substrate. Nevertheless
it is not impossible that some variation also results from different forming of CZTS on the
different substrates, which might affect both the loss of certain elements and the structure
or density of the film, respectively. The latter influences also the thickness, which can
itself affect the result if there is a depth gradient.
Si ↔ Mo
Si ↔ SLG
Si ↔ Si
Cu/Zn
1,7%
5,7%
8,2%
Cu/Sn
11,2%
4,2%
9,5%
Cu/(Zn+Sn)
5,9%
5,9%
10,3%
Table 3.1: Deviations in the EDS measurements concerning different substrates. The
data for SLG and Mo-SLG are referred to measurements on Si on different dates (last
row). Compared to this, the error range for glass and glass coated with molybdenum is
comparable (even smaller). Hence, the same composition on the different substrates can be
assumed.
3.2 Analysis techniques
39
In default of an appropriate calibration sample, the EDS measurements had to be
verified in a different way.1 This was done by comparison of the Cu/Zn ratio with XRF2
measurements. The average deviation was about 9%, which is near the variation of EDS
itself for different days.
The conclusion is that EDS is possible to be used for composition measurements.
However, it should be kept in mind that variations up to 10% can occur.
3.2.3 X-ray Photoelectron Spectroscopy (XPS)
XPS is a method closely related to EDS and XRF. Incoming X-rays knock out electrons
from the atomic shells of the sample, which is referred to as photoelectric effect. These
electrons are detected by a spectrometer and allow conclusions about the composition of
the irradiated substance. Furthermore, also information about the bonds of the elements
can be obtained as the electron levels are influenced by bindings. This allows to draw
interesting conclusions for example if oxides or sulfides are formed [50].
As electrons are already absorbed by very thin layers of material, XPS is very surface
sensitive (much more than for example EDS or XRF). In this work this method was mainly
used to get information about the depth profile of the composition of the samples. For
this the sample had to be sputtered, that means the material was gradually removed while
measuring the composition.
For this thesis, a Quantum 2000 Scanning ESCA Microprobe from Physical Electronics
with monochromatic Al Kα X-ray radiation (1486,7 eV) was used. All XPS sputterings
considered here were performed at 0,5 keV.
3.2.4 X-Ray Diffraction (XRD)
The technique XRD was used in this study to identify secondary phases in our films. It
bases on the diffraction of waves by a lattice. The wavelength of the incoming light has to
be in the order of the lattice constant for this effect to take place. As crystals have lattice
constants in the order of nanometers or Ångströms, X-rays with a similar wavelength are
used.
Considering a crystal, the incoming light waves are reflected by the crystal planes. For
a certain angle θ that depends on wavelength λ and distance of the crystal planes d, the
reflected waves interfere constructively and give a measurable intensity. This is described
λ
) [50].
by Braggs law: Θ = sin−1 ( 2d
1
There was actually one more confirmation that cannot be called a "calibration" but was used before
XRF measurements to see how reasonable EDS is. The Cu/Sn = 60/40 sputter target was measured
with EDS and the result was ≈ 59/40 (1% other elements).
2 XRF (X-Ray Fluorescence) analysis is related to EDS. Electrons are by incoming radiation (X-rays)
removed from the atoms and substituted by electrons from higher shells. During this process they emit
fluorescent X-rays that are characteristic for the element. The difference to EDS is the excitation by
polychromatic X-rays instead of electrons [51]. In these experiments the Spectro X-lab 2000 was used.
40
3 Experimental
That makes it possible to determine for known angle and wavelength the lattice spacing.
The other way round, unknown phases can be identified by the characteristic reflection
pattern that depends on the lattice parameters. This is used here.
The advantage of this technique is that it is nondestructive and needs no elaborate
sample preparations [48]. For thin films, however, it is a problem that the volume of
material that contributes to the signal is very small. This would require very long counting
times. A way to avoid this is the technique of grazing incidence. A very small angle is
applied which increases the effective volume by some factor in the order of 10 [48].
In our experiments, a Siemens D 5000 was used. The wavelength of the X-rays was
1,54 Å (gained from a Cu-target), the angle of gracing incidence was 1°.
The XRD peaks are often of different intensity, and in principle the measured signal has
only to be compared with the pattern of the related phase in a data base. One problem is,
though, that peaks often overlap. For example, since the peaks of Cu2 ZnSnS4 (CZTS),
Cu2 SnS3 (CTS) and ZnS – which all have the sphalerite structure (see chapter 2.3.3) –
overlap, it is complicated to identify CZTS clearly. Indeed, the first two compounds have
some small extra peaks which ZnS does not show, but ZnS on the other hand has no
unique peak. That means that ZnS can never be proven or excluded, as long as there is
crystalline CZTS (and/or CTS). In literature this signal is referred to as Σ-signal.
Another common problem is that often only a part of the peaks of the phase one would
assume is present. That means that if only a few expected peaks appear, this phase
cannot be proven but neither be excluded. This was a major problem in our XRD analysis
and can be due to preferred orientation of the crystals. Typically, the reference spectra
are for powder diffraction and in case of preferred orientation, some peaks may not be
present. Furthermore, the absence of an XRD signal does not have to mean absence of
the secondary phase, since it could be present in amorphous phase.
3.2.5 Quantum efficiency (QE) measurements
One technique for analysis of solar cells is quantum efficiency (QE) measurements. It
allows conclusions about the losses in the cell, effectiveness of charge separation and the
absorption, or charge generation. Quantum efficiency means how many of the incoming
light quanta are converted into electron-hole-pairs and collected in the outer circuit. Here,
always the external quantum efficiency will be meant, that means the incident and not
only the absorbed photons are counted. The incoming light is monochromatic and varied
stepwise (for our cells from 360 to 900 nm in 10 nm steps). If all incoming photons excite
each one electron-hole-pair that is collected, this would mean a quantum efficiency of 100%
or 1, respectively. Theoretically a QE higher than 1 is possible if photons with an energy
higher than the double band gap produce two electrons. Practically this effect is irrelevant,
though.
In our experiments the signal of the monochromatic light is split by a chopper. That
3.2 Analysis techniques
41
allows to refer the measurement signal to the incoming monochromatic light.1 The light
can be superposed by biaslight to analyse the influence of light intensity on the cell. The
result is calibrated with a reference sample of known quantum efficiency.
The QE curve allows drawing conclusions about the absorption and recombination losses
in the solar cell. Furthermore the band gap can be estimated. For this the x-scale is
converted from wavelength to energy (E = h · f = h · λc =p1240 eV λ·nm ) and the QE-values
are squared. Reason for this is the relation QE (E (λ)) ≈ E − Eg . That means that the
squared QE curve in theory intercepts the energy axis at Eg . Practically one has to take
into account that the material contains secondary phases with a lower band gap than the
considered semiconductor. Those as well as defects could lead to some absorption of the
photons even below the band gap. Therefore not the energy at QE 2 = 0 is taken, but the
slope is extrapolated to 0, like shown in Fig. 3.3. The described procedure is a common
way in literature to determine QE (see for example [52]). However, this includes of course
some freedom where exactly to draw the extrapolation line. For this reason we consider
this more as an estimation for the band gap than an exact definition.
Next to the information that can be extracted from the shape of the QE curve, by
integrating quantum efficiency with photon flux the short circuit current can be calculated.
Figure 3.3: Squared QE curve to extract the band gap. The slope of the curve on the low
energy side is extrapolated lineary. The interception with the energy axis gives the energy
of the band gap. Here the band gap is roughly between 1,6 and 1,7 eV.
1
As the measurement is not performed in complete darkness, otherwise the ambient light would contribute
to the counting.
42
3 Experimental
3.2.6 Current–voltage (IV) measurements
The functionality of a diode concerning (short circuit) current and (open circuit) voltage
are described in detail in the theory part of this thesis (chapter 2.1.7). To measure the
diode characteristic the solar cell is contacted at the surface and the back contact. Then
the solar cell is illuminated and the current is measured while at the same time varying
the voltage that is applied, which gives the IV-curve.
Additionally, the measurement can be performed in darkness as well. This can give
further insights about the cell. For example, an increasing current for negative voltages
could be due to a shunt through the space charge region or due to voltage dependent
carrier collection (for more details see chapter 2.1.6). As the latter effect occurs only under
radiation, i.e. when electrons are produced and can be collected by the SCR, it can be
excluded if the dark curve does no show an increasing current for increasing (negative)
voltage.
In this work the light source is a halogen lamp, calibrated to provide an intensity giving
the same short circuit current as that obtained from QE measurements on CZTS devices.
The temperature is maintained at 25°C by a water-cooled peltier element.
4 Results and discussion
4.1 Sputtering of precursors
This chapter will cover the composition of the precursors, structural insights, as well as
considerations on deviations from the standard sputtering process.
4.1.1 Composition
4.1.1.1 Growth of films at different powers and for different periods
In our experiments, altogether 91 different depositions were performed, except of some
test runs all of them on silicon, glass and glass coated with molybdenum in each run. The
depositions were split into two series: 47 depositions with a Cu/Sn and a Zn target, and
44 depositions with a Cu/Sn and a ZnS target in order to get S-containing precursors.
The first aim was to find out the deposition rate in order to determine the needed
deposition time for desired thicknesses. For this, depositions from single target of only
Cu/Sn and Zn (ZnS), respectively, were done, and the film thicknesses were measured.
For the Cu/Sn-target the sputter process was very stable. Two series at 200 W and
400 W were performed, each with several different sputter times. The results showed a
linear increase in thickness with deposition time, so that a deposition rate for each power
could be calculated. These were 29,6 nm/min for 200 W, and 58,5 nm/min for 400 W. Thus,
the sputtering process for the Cu/Sn target also showed a linear dependence in sputtering
power, which made it possible to vary the sputter power arbitrarily and interpolate the
expected thickness.
For the Zn target, the results were not that well-defined. Three series (at 50, 70 and
100 W) were made, that also showed linearity in deposition time. Plotted for different
powers, though, the growth per minute is not linear, especially not if the linear slope
should intercept 0 (see Fig. 4.1). Therefore no interpolation was done and for this target
and only the deposition rates 50, 70 and 100 W were used.
On basis of the gained results, it could be calculated which thickness one would obtain
for certain power settings. It turned out that in average the calculated and measured total
film thickness were differing by 6%. This is a very good value, considering that there is
also an error of measurement from the profilometer, which was found to be ≈ 3%. For
this reason we can expect that the sputter rates for the single (Cu/Sn and Zn) films are
preserved in co-sputtering.
In the second series, the Zn target was replaced by a ZnS target. This target was always
43
44
4 Results and discussion
Figure 4.1: Zn deposition rate for different powers. The slope cannot be proven to be
linear. For this reason no extrapolated but only measured values of Zn growth are used for
composition calculations.
sputtered at 400 W. At this power, the target showed a constant and stable behaviour
with a deposition rate of 24,4 nm/min.
In spite of the linear behaviors of the Cu/Sn and the ZnS target, the expected and the
measured thicknesses of co-sputtered films were on average differing by 13%. As all films
were thicker than expected, this allows the conclusion that either the co-sputtered film is
more porose or less dense, respectively, or that anomaly in the co-sputtering causes that
more material is sputtered compared to the single material layers. As the ZnS target was
operated in RF sputtering mode, it might have been influenced by the presence of the DC
sputtering of the Cu/Sn target during co-sputtering.
4.1.1.2 XPS studies on precursors
On some of the precursors, XPS measurements were performed to get information about
composition (details including the graphs can be found in Appendix A.2). XPS is very
surface sensitive. That offers the possibility to determine gradients in the film, if the
material is uniformly removed by sputtering, which was the case here.
We compared 4 precursors with XPS. Each one sample pair was sulphurized in the
same run, but on different substrates: Si and Mo, respectively. One sample pair was pure
metal precursors, the other one ZnS precursors. In both cases the composition was very
uniform within the film. Furthermore, no difference between Si and Mo substrate could
be observed, both showed homogenous element distribution within the film. The only
gradients that could be observed for all 4 films were related to the surface, which was
found to be Cu-poor and Sn- and Zn-rich.
That means that no difference in composition and gradients could be found for the
different substrates Mo/Si and for ZnS/metallic precursors from XPS.
4.1 Sputtering of precursors
45
4.1.2 Structure
To get insights about our precursors beyond composition, XRD analysis was done to reveal
information about structure and existing phases.
Firstly, we take a look at the metallic precursors as well as single Zn and Cu/Sn films. In
Fig. 4.2, the black pattern belongs to a pure Zn film, sputtered from the Zn target. The red
line shows the data of a Cu/Sn film as sputtered from our Cu/Sn 60/40 target. Last, the
green line is a typical precursor, composition according to EDS of Cu/Zn/Sn = 46/25/30.
The vertical lines in the diagram – black, blue, green and red – denote XRD peaks for
different crystal structures, in which the length is related to the intensity of the peak from
randomly oriented powder samples. The reference data is taken from the JCPDS (Joint
Committee on Powder Diffraction Standards) data base.
Figure 4.2: XRD pattern of a Zn (black line), Cu/Sn (red) and Cu/Sn/Zn (precursor)
film (green). The vertical lines indicate where peaks for zinc (black), tin (blue), molybdenum (green) and bronze (red) would be expected. The Zn film shows hexagonal zinc
structure, the pattern for bronze is slightly shifted as compared to our curve. The metal
precursor contains crystalline tin (tetragonal structure) and probably bronze according to
XRD.
The Zn line (black) matches clearly with the expected signals for hexagonal zinc structure
(black as well). Obviously, sputtered pure Zn forms this structure for pure films.
The red pattern, belonging to the Cu/Sn film, is a bit more difficult to clarify. Peaks
referring to Cu6 Sn5 (bronze) would match perfectly (red), anyway, they are all shifted
compared to the Cu/Sn peaks for around 0,5°. The reason for this could be that the
stoichiometry is slightly different, for example Cu7 Sn5 or similar instead of Cu6 Sn5 . This
46
4 Results and discussion
could lead to a slightly different signal for the same structure, if the lattice is stretched.
Additionally to Cu6 Sn5 or related, more Cu-rich compounds would be expected, as a
Cu/Sn ratio of around 1,7 was measured by EDS. Indeed, pure Cu cannot be excluded by
this pattern, as it’s main peak is at around 43°, a peak that exists here. Further weaker
peaks, like for 50,5°, cannot clearly be identified here, though. The existence of amorphous
Cu or Cu-compounds is another possibility, but that cannot be detected with this method.
The precursor pattern (green line) shows – in contrast to the Cu/Sn film – a clear match
to elemental Sn (tetragonal structure). All blue vertical lines showing the expected Sn
peaks find a related peak in the precursor pattern. Further peaks can be explained by
bronze (red lines), which means that the existence of this phase is very likely here.
Several more peaks, in contrast, are not so easily to assign to a certain phase, while at
the same time several phases cannot be excluded to occur. Cu (43° and weaker at 50,5°)
and CuZn (43°) are quite possible and likely. Cu5 Zn8 (major peak at 43°) is often reported
for metal precursors ([38], [26]), but no clear evidence for this can be found here.
In Fig. 4.3, ZnS precursors are shown. Here we concentrate on the black line, the others
will be discussed in the next chapter as they were sputtered under special conditions.
It can be seen that the XRD pattern shows a much more amorphous structure. Except
of three distinct peaks that can be related to the substrate (Mo, red vertical lines), the
peaks are very broad, indicating an amorphous or nanocrystalline structure. For this
reason, a clear assignment to certain phases is not possible. Both broad ’peaks’ could
belong to bronze (pink lines) and Sn. The broad peak between 40° and 46° could include
elemental Zn and Cu as well as CuZn (all have the major peak at 43°).
4.1.3 Special settings
During the second sputter series with the ZnS target, some settings that vary from the
otherwise adjusted settings were carried out. To the best of our knowledge, none of these
special settings caused any major change in the precursor, as we will support in this
chapter, so that we treat them in the following as ordinary samples, if they were used to
make solar cells.
4.1.3.1 Slowly sputtered
5 precursors were sputtered with only half of the usual power. These precursors are
referred to as ’slowly sputtered’. Compared to the average composition of ’normal sputtered’ precursors, no abnormal variations could be observed. Looking at the XRD data
(Fig. 4.3, red line), the pattern looks very similar to the normal sputtered reference (black
line). From this the conclusion is, that slowly sputtered samples can be treated without
as normal samples.
4.1 Sputtering of precursors
47
Figure 4.3: XRD pattern of ZnS precursors, sputtered under normal conditions (black
line), slowly sputtered (red), with higher gas pressure (blue) and with heated substrate
(green). Except for the sample with heated substrate, all films have an amorphous structure
and are very similar. The crystalline XRD pattern for the heated substrate sample matches
to bronze (pink vertical lines); ZnS (or CTS/CZTS) is very probable (brown). Several
peaks cannot be clearly assigned.
4.1.3.2 Higher gas pressure
For 13 out of 44 samples in the second series, the pressure in the sputter chamber was
different from the default settings. That was not a wished effect. However, no unusual
variations in composition were found, and like for the ’slowly sputtered’ case, no difference
for the XRD pattern (blue line in Fig. 4.3) compared to the reference sample can be found.
The weaker XRD signal results from an exchange of the X-ray tube in the used XRD
machine for samples with the stronger signals (red and black lines).
4.1.3.3 Heated substrates
As for the previous special cases, the composition of samples with heated substrates (4
precursors) had no unusual deviations from the average compositions for the used power
settings. The XRD data, however, shows a significant different pattern than the three
other precursors presented in Fig. 4.3. In contrast to those samples, the pattern for the
heated substrate precursor shows clear crystallinity. Bronze (Cu6 Sn5 ) is clear identified
(pink vertical lines). ZnS (brown) looks very probable, but at the same time CZTS/CTS
cannot be excluded as both patterns coincide very much in XRD (see chapters 2.3.3, 3.2.4).
48
4 Results and discussion
Several peaks can again not or not clearly be assigned to a certain phase, as well as several
elements or compounds cannot be unambiguously identified. Those are in this case Cu,
Cu5 Zn8 and Zn (Peak at 43°), as well as Cu-Sn compounds and copper sulfides.
4.1.3.4 Summary
As precursors sputtered slowly or with higher pressure do not change in structure and
composition, they will simply be treated as ’normal’ samples. For heated substrates, the
situation is a bit more complicated, as the structure appears to be much more crystalline
than for other ZnS-precursors. But as only one single cell was made of heated substrate
precursors, and no conclusions are drawn only from this specific cell, while it on the other
hand does no show any major abnormality in composition, morphology or efficiency, we
won’t consider this special setting any further.
4.1.4 Conclusions
Sputter experiments showed that the Cu/Sn target behaved very stable while for the Zn
target non-linear growth of the films was observed. This could be due to some interaction
of DC (Cu/Sn target) and RF (ZnS target) sputtering that could have occurred during
co-sputtering.
XPS studies on some samples indicated that precursors on Mo and Si can be treated as
equal, as they showed same composition and behaviour regarding gradients. For metallic
precursors no gradients within the film but a Cu-poor and Zn-rich surface were observed;
ZnS precursors had a slight Cu gradient within the film and a very Sn-rich surface.
Further analysis on the precursors was done with XRD. A general problem here was
that not all peaks could be assigned to a certain phase, which means that only few phases
could be proven to be present.
Pure Zn films matched clearly to the signal of crystalline Zn. Pure Cu/Sn films showed
the peaks for Bronze (Cu6 Sn5 ) but shifted about 0,5°. This might be due to a different
composition. Further Cu-containing phases could neither be proven nor be excluded.
For the metallic precursors (Cu+Zn+Sn) crystalline Sn was found, Cu and CuZn are
very possible.
ZnS precursors (Cu+Zn+Sn+S) turned out to be amorphous, except for one that had a
heated substrate. A lot of phases (Cu, Zn, Sn, CuZn, Cu6 Sn5 and more) might contribute
to the broad peaks that were seen here. The sample with heated substrate indicated the
existence of Cu6 Sn5 and ZnS (or CZTS, CTS); Cu, Cu5 Zn8 ,Zn as well as Cu-Sn compounds
and copper sulfides are possible.
Another conclusion from XRD combined with compositional considerations was that
precursors with different sputtering processes (slow sputtering, different pressure, heated
substrate) can be treated as equal.
4.2 Properties of sulphurized films
49
4.2 Properties of sulphurized films
This chapter shows the results of the analysis done on CZTS films obtained by sulphurisation
of precursors. The measurements included SEM, EDS, XPS and XRD, and were used
to identify compositional evolution (e.g. element losses under sulphurization), structure,
appearance (e.g. grain size) and existence of secondary phases. The main goal was to
analyze the quality of the synthesized material.
4.2.1 Data set
Table 4.1 shows the most important data for all sulphurized samples that were part of this
study. They are grouped by the region of the phase diagram where the final CZTS films
belong to (column ’category’). The first column contains the sample number that will
be referred to in this thesis. The second column gives type of the precursor which either
contained sulphur or was a metallic one (for the labeling metallic and ZnS see chapter
3.1.1.2). Next, the composition of the CZTS film is shown. It should be noted that only
the three metals and their fraction relative to the total metal amount is given; the amount
of sulphur was always 50% within the measurement accuracy. The next two columns show
thickness of the precursor and CZTS, in which the CZTS thickness shows the range of
thicknesses that could be estimated from SEM cross section images. For samples 24 and 25
(last two rows) no SEM images were made, and as the CZTS thickness was extracted from
these no information about the CZTS thickness is available here. The further columns
show the solar cell parameters of the solar cell with highest efficiency of each sample.
These 25 samples were used to make solar cells and are discussed in this thesis, and the
table shall provide all available information.
4.2.2 Composition
4.2.2.1 Compositional changes during sulphurization and CZTS formation
CZTS from ZnS precursors
An appropriate tool to get a quick overview of composition related properties is the phase
diagram. Fig. 4.4 a) illustrates the compositional changes during sulphurization of ZnS
precursors. Inset as visual help are the directions of Sn- and Zn-loss. Arrows point in the
direction from precursor to the sulphurized sample composition.
A clear trend is the movement of most samples roughly in a direction corresponding to
Sn-loss, even if they often show minor components of other directions like Zn-loss.
Some samples, however, point in completely different directions; they are marked in grey.
Those directions make, seen from a physical point of view, no sense, as they would indicate
compositional changes like Cu-loss (no physical mechanism for this shown for CZTS) or
Zn-enrichment. Those changes in composition are probably caused by measurement errors,
most likely by the accuracy of EDS.
4 Results and discussion
50
#
Type
11
21
22
23
15
16
7
8
6
17
9
10
20
3
2
1
4
5
14
19
18
13
12
24
25
ZnS
Metal
Metal
Metal
Metal
Metal
ZnS
ZnS
ZnS
Metal
ZnS
ZnS
Metal
ZnS
ZnS
ZnS
ZnS
ZnS
ZnS
Metal
Metal
ZnS
ZnS
ZnS
ZnS
Cu
[%]
47,2
46,2
46,5
45,7
48,2
46,6
46,5
47,1
45,8
47,2
47,7
47,8
49,1
51,4
49,2
48,9
50,7
50,2
49,1
48,8
48,3
49,5
48,3
52,8
53,7
Zn
[%]
27,4
30,3
29,9
30,8
27,6
25,9
25,5
25,6
26,3
26,5
25,2
25,3
24,6
21,2
22,8
21,7
22,0
23,5
24,9
25,4
26,1
24,9
27,0
34,7
34,6
Sn
[%]
25,3
23,5
23,6
23,5
24,2
27,6
28,0
27,3
27,9
26,3
27,1
26,9
26,3
27,4
28,0
29,4
27,3
26,3
26,0
25,8
25,5
25,5
24,6
12,4
11,7
Prec
[µm]
0,40
0,49
0,49
0,49
0,60
0,31
0,49
0,49
0,40
0,46
0,49
0,49
0,46
0,80
0,39
0,47
0,47
0,39
0,47
0,46
0,31
0,47
≈ 0,20
0,49
0,49
CZTS
[µm]
0,6-0,8
1,5-2,2
1-1,3
1,0-2,0
1-1,6
0,7-1,3
1-1,3
0,9-1
0,7
1,1-1,4
1-1,1
1,2
1-1,5
0,85-1,15
0,7-1,2
2-2,5
1,7-2,1
0,8-1
1
1,3-1,6
0,55-1
0,5-1
0,4
-
η
[%]
2,7
2,6
2,3
2,7
0,5
0,5
1,8
1,9
1,9
2,1
2,5
3,2
0,8
0,0
0,0
0,0
0,0
0,1
0,0
0,1
0,2
0,4
1,1
0,0
0,0
Voc
[V]
0,61
0,66
0,64
0,69
0,53
0,22
0,57
0,71
0,49
0,55
0,62
0,72
0,34
0,08
0,06
0,01
0,08
0,09
0,04
0,06
0,11
0,28
0,44
0,02
0,02
Isc
mA
[ cm
2]
10,3
12,6
10,5
9,0
3,6
7,7
10,4
9,6
8
8,2
9,3
10,9
7,5
2,5
1
0,8
1,1
1,8
1,1
3,3
6
3,6
6,1
0,6
0,5
FF
[%]
42,6
35,1
35,0
43,2
28,2
26,8
29,7
28,6
47,9
47,3
43,6
38,9
31,3
25,5
26,5
25,0
26,5
30,7
24,9
24,9
27,1
38,1
39,1
23,2
25,8
TPD
Zn-rich
Zn-rich
Zn-rich
Zn-rich
Zn-rich
Cu-poor
Cu-poor
Cu-poor
Cu-poor
Cu-poor
Cu-poor
Sn-rich
Sn-rich
Zn-poor
Zn-poor
Zn-poor
Zn-poor
Zn-poor
stoic
stoic
stoic
stoic
stoic
Sn-poor
Sn-poor
Table 4.1: Data set of all samples that were fabricated to solar cells. The data set contains sample number, precursor type, composition regarding the metals, thickness of precursor and final CZTS film, the solar cell parameters efficiency, open circuit voltages, short
circuit current and fill factor, as well as the region of the phase diagram where the sample
is assigned to.
4.2 Properties of sulphurized films
51
Figure 4.4: Ternary phase diagram showing the compositional changes for a) ZnS and b)
metallic films. Arrows point in the direction from the precursor to the sulphurized sample.
CZTS from metallic precursors
For metallic precursors, the general impression is the same as for the ZnS samples
(Fig. 4.4 b)). Most samples show Sn-loss, but some other directions exist as well. However,
here the trend appears to be much more pronounced, i.e. a stronger Sn-loss occurs.
It is also worth noting that no remarkable Zn-losses can be observed which was reported
in many publications (e.g. [46]), even for those samples that end up in the Zn-rich region
and could have reached stoichiometry by Zn-loss.
Mechanism of Sn–loss and possible explanation for differences between metal and ZnS
It is very likely that the Sn-loss occurs by evaporation of SnS2 that is formed early in the
sulphurization process at 180°C for Cu-poor films like used here [29]. The material in this
way looses the excess tin. That means that almost all samples achieved a Cu/Sn ratio
closer to stoichiometric (which would be 2). At the same time, the Sn loss for our samples
was never so large that films would end up in the Sn-poor region of the phase diagram.
The tin loss observed here seems to be a beneficial process for the CZTS formation in
cases like ours, where the initial Cu/Sn ratio is below 2. This observation is contrary to
frequently reported tin loss beyond a beneficial dimension (e.g. [28]). We contribute this
to the sealed quartz ampoules used in this process, where an atmosphere of tin sulfide is
established which prevents further tin loss.
As mentioned before, the trend is much more pronounced for metallic precursors. The
examined precursors changed from Cu/Sn = 1,58 to 1,89 (near stoichiometric) for metallic,
and from 1,68 to 1,79 for ZnS precursors. The question is what the cause for this difference
is. It seems very likely that the reason could be related to the fact that ZnS precursors
contain S already before sulphurization.
4 Results and discussion
52
For metallic precursors, the following reaction path1 has been proposed2 ([29], [28], [26]).
In the beginning, the S has to diffuse into the film of the precursor to get in contact with
the elements to react. The first sulfides that are formed are the binary sulfides, i.e. copper
sulfides (Cu(2−x) S, 2≤x≥1), zinc sulfides (ZnS) and tin sulfides (SnS2 ). This happens
below temperatures of 300–350°C ([28], [26]).
In the next stage, these sulfides react with sulphur to more complex structures like
Cu2 SnS3 (that itself reacts with ZnS to CZTS) and Cu2 ZnSnS4 . This step starts at around
350°C ([28]3 ) to 400°C ([26]).
The reactions are summarized below.
Cu+Zn+Sn+S
Cu(2−x) S+SnS2
Cu(2−x) S+ZnS+SnS2
Cu2 SnS3 +ZnS
→
→
→
→
Cu(2−x) S+ZnS+SnS2
Cu2 SnS3
Cu2 ZnSnS4
Cu2 ZnSnS4
(<300–350°C)
(>350–400°C)
(>350–400°C)
(>350–400°C)
For the process in our samples, it seems very likely that very early formed SnS2 evaporates
partly. Possibly existing further excess of Sn (most likely in form of SnS2 as there is excess
sulphur) that is formed deeper in the film and does not reach the surface but is enclosed
by the CZTS (and other, secondary phases) will then remain as conglomerates or maybe
as a SnS2 layer.
The described reactions refer to metallic precursors. For S containing precursors the
reaction path could be different. Sulphur is now – even though not in sufficient amount –
available throughout the whole film already from the beginning. Binary sulfides can be
formed everywhere much earlier, or may even exist from the beginning. XRD analysis
of the precursors (chapter 4.1.2) did not show any crystalline phases for ZnS precursors.
However, the results for the heated substrates could give an indication. XRD measurements
of precursors heated up to around 300°C suggest that ZnS (or even CTS or CZTS) exists
in those samples. That means that either ZnS could be present in all ZnS precursors
(in an amorphous, or crystalline but disordered form), or at least should be formed at
1
2
For further details see chapter 2.3.4
Schurr et al. [29] and Weber [28] performed XRD analysis during the annealing process, Scragg [26]
interrupted the sulphurization process at certain temperatures and cooled rapidly, assuming that the
existing phases at the interruption temperature "freeze".
3 As mentioned before, it is not possible to distinguish between CTS, CZTS and ZnS from XRD. Anyway,
Weber was able to find strong evidences for the present phases by additional XRF measurements. His
precursors were layered: Mo/SnS/CuS/ZnS. The XRF signal of Cu and Sn is attenuated by the upper
layers, i.e. occurance of a mixture of layers can be found by an increasing signal of Cu or Sn, respectively.
An increase of the XRD Σ–signal, without any change of the XRF signal of Zn indicates that Zn did
not diffuse and Cu2 SnS3 is formed. The amount of ZnS cannot be increased as all Zn is already present
in the form of ZnS from the beginning. This reaction takes place between 250 and 300°C. At 350°C, the
XRF Cu signal increases, while the Zn signal decreases. This indicates interdiffusion of both elements.
CZTS formation then likely starts from Cu2 SnS3 and ZnS.
4.2 Properties of sulphurized films
53
temperatures below 300°C. The copper and tin sulfides should be formed very early in the
sulphurization process as well. So the question is, could these sulfides in the ZnS precursors
form CTS and CZTS much earlier than in the metallic precursors, or is the formation
limited not by sulphur but requires achieving a certain temperature? According to Weber
[28], copper and tin sulfides form Cu2 SnS3 already at 250°C. This means that temperature
can obviously not have been the limiting factor in experiments with metallic precursors,
but rather the fact that first binary sulfides had to be formed before the sulphurization
could proceed.
Therefore, following mechanism for our ZnS samples may be proposed: copper, tin and
zinc sulfides are formed very early or are even present already throughout the whole ZnS
precursor. That makes it possible to form Cu2 SnS3 and Cu2 ZnSnS4 at a very early stage
of the sulphurization process, so that the Sn is bound and cannot evaporate from the film.
Only SnS2 compounds near the surface would have the possibility to evaporate and lead
to a slight Sn-loss. The latter assumption is supported by Weber [28], who found out that
the Sn-S loss rate increases for the different phases in the order
Cu2 ZnSnS4 → Cu4 SnS4 → Cu2 SnS3 → SnS.
This means that excess tin cannot evaporate from the film in the same way as in the case of
metallic precursors, but would either be bound in non volatile structures or encapsulated
by the early formed CTS and CZTS phases. We therefore suggest that the excess tin
would evaporate as SnS or SnS2 for metallic precursors, while it would rather be enclosed
as conglomerates or SnS2 layer for ZnS precursors.
It is difficult to find a definite proof for this mechanism that we conclude from the
different Sn-losses for metallic and ZnS precursors, but it is at least supported by the fact
that all CZTS films showing clear SnS2 signals in XRD (3 samples) were made of ZnS
precursors.
4.2.2.2 Reproducibility
It is interesting to see how reproducible our sulphurizations are in order to evaluate
to which extent we can control the sulphurization process. This is also important to
verify whether we can draw conclusions from the observed compositional changes. During
our experimental series, three precursors were sulphurized twice under exactly the same
conditions to answer this question. We took pieces of the same precursor and subjected
them to the same sulphurization on different days. The results for metal composition and
efficiency are listed in Table 4.2.
It can be seen that our experiments are very reproducible concerning composition and
also in efficiency. For this reason we assume that samples from the same precursors are
comparable, which we will use for example in chapter 4.3.2 for the QE measurements.
4 Results and discussion
54
Sample #
21 (metallic)
23 (metallic)
9 (ZnS)
10 (ZnS)
7 (ZnS)
8 (ZnS)
Cu
0,46
0,46
0,48
0,48
0,47
0,47
Zn
0,30
0,31
0,25
0,25
0,25
0,26
Sn
0,23
0,24
0,27
0,27
0,28
0,27
Efficiency
2,6%
2,7%
2,5%
3,2%
1,8%
1,9%
Table 4.2: The sample pairs #21 and #23, #9 and #10 as well as #7 and #8 were in
each case made of the same precursor. The results for composition and efficiency show the
reproducibility of our experiments.
4.2.2.3 Role of the substrate
All precursors were deposited on three substrates: silicon, glass, and glass coated with
molybdenum. The precursors on the three substrates are assumed to be identical as they
were sputtered in the same deposition process. The substrates lay on a rotating substrate
holder, so that there should be no differences between the several substrates. Reason for
this is the advantage of different substrates for different analyzing methods. For example
suits the conducting silicon very well for EDS, while molybdenum-coated glass (Mo-SLG)
is needed for making a solar cell.
In order to test possible differences of films deposited on Mo, Si and SLG, some EDS
comparison measurements were done (chapter 3.2.2.1). Measured compositions of the
(assumed) same films on Si and Mo as well as Si and SLG were compared to each other
and the average deviation between the values was calculated. The difference is mostly
below 6%, but in one case up to ca. 11% between same films on different substrates. This
is within the error range we found for EDS (3.2.2.1). That means, that the composition of
equally processed films on different substrates can be assumed to be equal, as the EDS
results do not show more deviation between same films on different substrates compared
the same film on the same substrate.
This is in good agreement with the results from chapter 4.1.1.2, where we could conclude
that there was no difference in composition or compositional gradients between precursors
on Mo and Si.
4.2.2.4 Summary
For our set of samples we observed Sn-loss during sulphurization. The Sn-loss is more
pronounced for the metallic precursors. A possible explanation including a theory how
CZTS is formed in our films was presented. As a result, it is assumed that a different
formation pathway of CZTS in ZnS films (compared to metallic films) leads to enclosure of
SnS2 or other tin-rich secondary phases. For metallic precursors it is assumed that excess
tin evaporates during sulphurization.
4.2 Properties of sulphurized films
55
Furthermore we analysed the role of the substrate comparing the composition of films on
glass, glass coated with molybdenum and on silicon. The conclusion was that the variation
between composition on the several substrates lies within the error range of EDS.
One experiment was performed to find how reproducible our experiments are. This
was done by sulphurizing pieces of the same precursors in different runs. The result
showed a very high reproducibility regarding the composition and quite good concerning
efficiency. For this reason we believe that the sulphurization process is stable with regard
to composition, i.e. compositional changes like the observed Sn-loss should not depend on
the day of sulphurization but have more fundamental reasons.
4.2.3 Morphology
In addition to the composition, the morphology of the sulphurized films is another important
aspect that can help to understand the processes occurring during sulphurization as well
as to reveal information about the film quality.
In this chapter we attempted to identify secondary phases by XRD and SEM combined
with EDS, to examine the ’quality’ of the CZTS films (grain size, voids, compactness),
and to evaluate the influence of ZnS in the precursor. Measurements of the thickness of
the CZTS layers and of a presumed molybdenum sulfide layer from SEM cross sections
were also performed.
As a further aspect, the influence of the substrate for the morphology of the CZTS was
studied.
4.2.3.1 Surface features
A SEM top view makes possible to identify secondary phases that precipitate at the surface.
As mentioned before, analysis of final CZTS films on Mo was difficult as the samples were
in most cases further processed directly after sulphurization to minimize oxidation. Films
with CdS and ZnO as top layers, however, cannot be easily analyzed, as the top layers
hide or at least attenuate the desired details.
For the surface features, we could distinguish between two appearances: round, more
spherical features (Fig. 4.5 a) ), and features that looked like small plates or flakes
(Fig. 4.5 b) ). Those two were the most common structures seen on the surface, and
both were in the range of micrometers.
The round features were extremely Cu-rich. Compared to the average Cu content in
the whole film, the amount of copper increased roughly 25 times (e.g. Cu/(Zn+Sn):
0,51→13,52). The other metals appear only in small amounts.
Unfortunately it is very difficult to distinguish sulphur and molybdenum with EDS
as the peaks partly overlap. This is one of the reasons why EDS measurements were
usually performed on silicon. Therefore we could not accurately determine the contribution
from sulphur from molybdenum. However, it is very likely that this surface feature is
Cu(2) S. First, EDS indicates an increase of the sulphur and a decrease of the molybdenum
56
4 Results and discussion
Figure 4.5: Surface features on CZTS films (substrate: Mo). EDS measurements were
performed in the areas marked with a cross. a) shows round features that were very Curich . Hence, it is assumed that these surface features are copper sulfides. In b) the flakes
can be seen. As EDS measurements showed Sn-richness, these features are possibly tin
sulfides.
signal, which would be expected when looking at a sulphur rich surface feature (i.e. the
molybdenum signal should be more attenuated). Second, the combined Mo/S signal is
still very strong compared to the average value of the surface, while Zn and Sn signal
almost disappear. In the end, Cu2 S is frequently reported (e.g. [35]) to precipitate at
the surface (see chapter 2.3.4), especially if the precursors is Cu-rich like in this case
(Cu/(Zn+Sn) = 1,12 where 1 would be stoichiometric). For this reason we assume that we
had formation of copper sulfides as secondary phases.
The flakes like structures shown in Fig. 4.5 b) are Sn-rich. The point marked in the
image showed ca. 50% more Sn than the surrounding area, while the other metals were
reduced by ca. 50%. The combined S/Mo signal increased (less Mo, more S). What we
assume from this result is that we have SnS(2) flakes. The signal from Cu, Zn and Mo
could result from the film below, considering that the flakes seem to be much thinner
than the penetration depth of the electron beam. As the sample was slightly Sn-rich
(Cu/Sn = 1,87 where 2 would be stoichiometric), this result seems to be plausible, and tin
sulfides are apparently another possible secondary phase for our CZTS films.
4.2.3.2 Cross sections
Cross section studies of sulphurized films are an important and useful tool which provides an indication about material quality. They can give information about grain size,
voids, compactness, homogeneity and uniformity of the film, which can potentially allow
drawing conclusions about the material formation process, secondary phases and possibly
disadvantageous structures or features.
4.2 Properties of sulphurized films
57
Cross sections of all solar cells were made. However, it is often not easy to draw
unambiguous conclusions. 25 solar cells are part of this study, and more than 200 SEM
images were taken and analysed. Still, often only 3–4 images for each sample remained
where the interesting aspect (such as grain size or surface features) could be seen clearly
and accurately. Furthermore, the solar cells were cut only once, that means the images
show only the structure along a single line. Therefore, it can of course not be excluded
that in some other part of the cell different structures occur. The images should at least
give an impression about the general appearance of the film, even if we have to exercise
caution in drawing conclusions.
Grain size
It is often very difficult to determine the grain size from SEM images. Main problem is that
the cross section does not show an even fracture plane, but a very square-edged side. It is
often impossible to determine where grains begin and where they end, or whether several
grains "overlap" in the image but appear as one; a general topographical problem when
a three-dimensional structure is pictured in a two-dimensional image. Another problem
is that it sometimes was not possible to get a sufficient contrast for the images (see for
example Fig. 4.8 e) ) due to charging effects, which made it nearly impossible to estimate
grain size.
However, for most samples an impression could be received if the grains appear ’larger’,
by which we mean here clearly bigger than 0,5 µm, or smaller (less than 0,5 µm). In these
categories, no clear trend for grain sizes related to composition could be found. As can be
seen in Fig. 4.6, there exist large-grained samples in the Zn-rich, Cu-poor, stoichiometric
and Sn-rich region of the phase diagram, i.e. in four of five regions of the phase diagram
where we have made solar cells1 . Only Zn-poor cells with reasonable grain sizes could not
be found.
Comparison of samples with similar composition could give more specific insights in
addition to the general conclusions. That could support – or disprove – the assumption
that composition is not the driving force for grain size.
In Fig. 4.7 a) one can see a sample with Cu/Sn ratio of 1,69 and Cu/(Zn+Sn) ratio
of 0,87, that is Cu-poor. It shows huge grains in the micrometer range. Next to it, in
Fig. 4.7 b), a sample with almost the same composition from the same region (Cu/Sn = 1,66,
Cu/(Zn+Sn) = 0,87) can be seen, and the grains are significantly smaller.
Another example with samples of similar composition but different in grain size can be
seen in Fig. 4.27 b) and c) on page 87.
From this we conclude that composition is not a parameter that controls grain size in
our experiments.
1
Actually, two solar cells from a fifth region of the TPD exists (Sn-poor), but no SEM-images were made
for these samples.
58
4 Results and discussion
Figure 4.6: Solar cell cross sections from the a) Zn-rich, b) Cu-poor, c) stoichiometric
and d) Sn-rich region of the TPD. It can be seen that grain sizes larger than 0,5 µm can be
found in each of these regions. Inset are the sample numbers.
4.2 Properties of sulphurized films
59
Figure 4.7: Grain size comparison for two Cu-poor samples near in composition. Even
though the two samples lie in the same region of the TPD, the grain size is very different.
Inset are the sample numbers.
Voids and compactness
Considerably easier to determine is if a film appears compact or shows a lot of voids. As for
grain size, the composition seems to have no impact on the formation of voids or density
of films. The Figures 4.8 a)–h) give an impression: a) and b) are Zn-rich samples, c) and
d) Cu-poor, e) and f) stoichiometric, and g) and h) Zn-poor (for Sn-rich it was difficult to
see a clear trend). The left image always shows a dense, compact film, almost without
voids, while the right image from the same region of the phase diagram is more porous and
shows (often huge) voids. Note that we consider three-dimensional holes; cracks between
different layers like in c) are examined in the next chapter and do not count as voids.
Voids at the back contact are frequently seen for CZTS films (for example in [7]). As
described in chapter 2.3.4 it is very likely that this effect results from Cu diffusion to the
front, leaving voids within the film.
a) and b) as well as e) and f) are at the same time examples of samples that are very
close in composition and still very different concerning compactness, as can be seen from
Table 4.3.
4 Results and discussion
60
Figure 4.8: Void comparison for different regions of the TPD. a) and b) are Zn-rich samples, c) and d) Cu-poor, e) and f) stoichiometric, and g) and h) Zn-poor. Each left image
shows dense films with no or only few voids, the right image films with large and/or several
voids. Inset are the sample numbers.
a (#23)
b (#21)
e (#14)
f (#19)
Cu/Sn
1,94
1,97
1,89
1,89
Cu/(Zn+Sn)
0,84
0,86
0,96
0,95
Table 4.3: Sample set with similar composition (concerns Fig. 4.8, for more information
see the text). Samples #23 and #21 are Zn-rich, #14 and #19 stoichiometric.
4.2 Properties of sulphurized films
61
Layering
In several cases we observed that the CZTS film was clearly divided into two or even
three sublayers. This phenomenon we refer to as layering, as examplified can be seen in
Fig. 4.10 (page 62).
Like for grain size and voids, we were interested in if the occurrence of layers in some
way is connected to composition. For this we investigated all samples with respect to
layering. In doing so we encountered the problem that many samples were not clearly
defined, as layering was not very pronounced. Fig. 4.8h) for example might show layering
or just a random accumulation of voids near each other. Another problem was when only
some images of the same sample showed layering while others did not exhibit any such
features. This class of samples was categorized as ill-defined.
Definite layering could be observed in all regions of the phase diagram that were covered
in our experiments. Anyway, we had a closer look at samples where we could definitely
exclude the existence of layering, and plotted samples without layering as one group (black
in Fig. 4.9), and samples with or with unclear layering as another group (marked in blue).
There is a trend that ’no layering’ mainly occurs in or near the stoichiometric region (there
is only one sample far off in the Zn-poor region). That means to avoid layering it could be
helpful to end up not too far away from the stoichiometric region of the phase diagram,
even if this cannot guarantee layer-free structures.
Figure 4.9: Composition (plotted in the TPD) of samples with no (black) and with
clear/ill-defined layering (blue). Samples that definitely showed no layering are grouped
in and near the stoichiometric region.
62
4 Results and discussion
Of course it is of interest why layering occurs and especially if it just means a crack
within the CZTS layer, or if the borderline actually separates two different phases. For
this reason we examined one sample more closely.
Fig. 4.10 shows sample #10. The image illustrates what we refer to as layering: a crack
divides along the whole image a top from a sublayer. Obviously the adhesion between
these two layers is poor, as the top layer flaked off near the breaking edge (that is why
the top layer is – contrary to the bottom layers Mo and glass – out of contrast). This
adhesion problem was observed in several images and Fig. 4.11 shows it even more clearly:
to the left is the bare substrate, next to it only the bottom layer is covering the substrate.
The top layer including CdS and ZnO flaked off and starts in some distance away from
the fracture edge.
Zooming in (Fig. 4.12) one can see that the sublayer seen from the top has a needle
shaped structure in plane of the sublayer, which is different from the CZTS structure. That
indicates that this layer is not CZTS but another, secondary, phase. EDS measurements
reveal that while the top layer exhibits the expected Cu/Sn ratio of 1,9, the bottom layer
is extremely Sn-rich: Cu/Sn = 0,28. Measurements of the whole film including the bottom
layer gave 1,78; 67% of the metals in the bottom layer are tin. A determination of sulphur
content was not possible due to overlap between sulphur and molybdenum. Although we
cannot prove it, we strongly suggest that this layer contains large amounts of tin sulfide.
There is no Cu-Zn-Sn-S phase next to CZTS known, especially not with those metal ratios,
and no Cu-Zn-Sn phase either. On the other hand, the presence of pure Sn after 2 hours of
sulphurization is very unlikely. Therefore it is likely that this layer consists mainly of SnS2 .
This assumption is supported by the fact that XRD analysis for this sample shows the
existence of a crystalline SnS2 phase (see chapter 4.2.3.6, Fig. 4.17), and that precursor as
well as final sulphurized layer are Sn-rich (Cu/Sn = 1,56 and 1,78, respectively, while 2
would be stoichiometric).
Figure 4.10: Cross section of sample #10. The film is clearly divided into two layers. The
upper layer contain the (presumed) CZTS film as well as the CdS and ZnO layer.
4.2 Properties of sulphurized films
63
Figure 4.11: Top view on sample #10. The different layers can be seen as they partly
separated from each other during the sample breaking. This shows also the bad adhesion
between the sub and the top layer: near the breaking edge the top layer has flaked off,
which caused the contrast problems that can be observed in Fig. 4.10.
This observation supports our considerations in chapter 4.2.2.1, where we suggested
that ZnS precursors tend to keep their excess tin in form of tin sulfide as conglomerates or
layers, while metallic precursors loose tin sulfide by evaporation early in the sulphurization
process before it can be enclosed by CZTS.
Figure 4.12: Zooming into the breaking edge. The sublayer has a needle shaped structure.
EDS in combination with XRD measurements leads to the assumption that this layer is
SnS2 .
64
4 Results and discussion
Two more ZnS samples (#8&9) show layering in the way shown in Fig. 4.10 with bad
adhesion between the two layers (#9 is shown in Fig. 4.13 a, bad adhesion arises from the
flaked off top layer near the breaking edge), and for both samples SnS2 was found by XRD
as well. There is also a sample made of a metal precursor where the cross section could
suggest two layers with bad adhesion, but XRD could not prove the existence of SnS2 .
Note that the considerations about this kind of layer concern only those mentioned
three ZnS samples. There are other layer structures (see Fig. 4.13b)) that look different
and where we cannot clarify if the layers consist of different phases. We assume that these
different layers consist of CZTS divided by cracks, as the structures of the layers are very
similar to each other. This kind of layering has also been observed in publications [7].
Summary
As grain size, voids and layering are assumed to influence the performance of a device we
would like to be able to control these properties. However, a correlation with composition
– the main parameter we changed in our experimental series – could not be observed. This
was shown likewise for a general overview as for a comparison between each two samples
very near in composition. Only Zn-poor samples were all small grained.
Layering, in contrast, might be avoidable by choosing (near-) stoichiometric samples.
It is not always clear if the two to three layers are CZTS or consist of different phases,
but combined EDS and XRD measurements indicate that at least for a part of the ZnS
precursors the bottom layer is mainly tin sulfide.
The question in which way efficiency is influenced by those morphological parameters is
discussed in chapter 4.3.1.2.
Formation of molybdenum sulfide
In all of our cross sections we observed that the molybdenum film is discoloured at the
top after sulphurization. We assume that this discolouring results from a sulphurization
of the molybdenum, i.e. we suppose that molybdenum sulfide (more precisely: MoS2 ) is
formed. This assumption is supported by XPS analysis (see Appendix, Fig. A.5), where an
Figure 4.13: Different layer structures observed in this study. a) shows layering where
the sublayer has a different structure and where the adhesion between sub- and top layer
is poor. b) shows layering that supposes same material, and the top layer is not flaked off
which suggests better adehesion than in a).
4.2 Properties of sulphurized films
65
increase of the Mo signal can be seen, while the S signal is still unchanged and decreases
significantly deeper in the film and slower than the metals Cu, Zn and Sn. That means
that Mo and S coexist in a certain area and supports our assumption.
4.2.3.3 The influence of sulphur in the precursor
The substantial difference between our two experimental series was that we used a ZnS
target instead of the Zn target in the second series. The aim was to incorporate sulphur
already into the precursor, expecting that the film is subjected to less diffusion of the
elements and so to less dramatical changes within the film (see chapter 2.3.4).
In our case the fraction of S in the precursors varied between 14 and 20%. The
sulphurization process remained the same to assure that the results are comparable to the
experiments with the metallic precursors.
General impression
It is not straightforward to get an objective overview about the possible difference between
metallic and ZnS CZTS-films. However, the fact that it is difficult to see a general
difference between these two types of films means that the use of ZnS actually did not
make any major improvement in our experiments. Especially considering grain size it
seems more that the impact could even be the opposite, i.e. that grain size for ZnS films
rather decreased. Most ZnS samples show grains not bigger than half a micrometer,
bigger grains are nearly exclusively observed for metallic films. Concerning voids and
compactness, there actually seems be a slight trend. With only one exception, all metallic
films show voids, often even very large ones, while ZnS films tend to produce less and
smaller voids. The latter is probably related to less diffusion of Cu to the surface (see
chapter 2.3.4). Furthermore, the ZnS samples are smoother, i.e. show less variation in
thickness (see chapter 4.2.3.4). Fig. 4.14 gives an impression of the described effect. a)
shows a metallic film, b) a film made of a ZnS precursor.
Hence, the general conclusion is: ZnS precursor show a trend to more compact films
with smaller grains as well as less and smaller voids.
Figure 4.14: Illustration of the general impression that metallic precursors a) tend to
have larger voids than ZnS precursors b).
4 Results and discussion
66
Comparison of similar compositions: Experimental matrix
When comparing all samples for a general discussion one should keep in mind that
composition is the main parameter and that it is not possible to distinguish between the
impact of the S in the precursor and the composition of the films. To support the results
from the previous chapter it would therefore be illustrative to compare samples with as
similar precursor compositions as possible. For this reason we made an experimental
matrix with metallic and ZnS precursors from the same region of the phase diagram.
Same composition means in this case the ratios between the metals; as only one kind
of precursors contains sulphur, the overall composition is of course not the same. More
information about the chosen samples is provided in Table 4.4. Two pieces of the same ZnS
precursor were sulphurized two times under the same conditions, which is why we have
two samples to compare with here. As can be seen, all samples keep similar composition
after sulphurization, which makes the comparison sensible.
Fig. 4.15 shows the cross sections of the three sulphurized films, a) made of a metallic
precursor, b) and c) the ZnS ones. An increase in grain size cannot be stated, rather the
opposite: while the grains in sample #17 are noticeable larger than 0,5 µm (often up to
≈ 0,9x0,7 µm), the grains for the ZnS samples #9 and #10 are difficult to estimate but
already limited by the obvious layering to sizes less than 0,5 µm. The impression is that
the grains are even smaller than that. On the other hand, comparing #17 and #10 could
support the general impression of the previous chapter that ZnS films are more compact
and contain less and also smaller voids. The result is not very clear for sample #9, the
crack within the film seems to be the origin of voids that reach into the CZTS film. Sample
#9 illustrates also very well both the difficulty to distinguish clearly between for example
voids and cracks, and also that the use of ZnS could not prevent voids in every case.
Next to this example from the Cu-poor region, ZnS and metallic samples with similar
composition from the Zn-rich and stoichiometric region were compared. They are not
discussed here1 but showed the same impression as described for the samples above: no
increase in grain size, but films that are more compact and show fewer large voids.
Precursor # and type
#17 (metallic)
#9* (ZnS)
#10* (ZnS)
Precursor
Cu/Sn Cu/(Zn+Sn)
1,53
0,83
1,56
0,86
1,56
0,86
Sulphurized
Cu/Sn (sulph) Cu/(Zn+Sn)
1,79
0,89
1,76
0,91
1,78
0,92
Table 4.4: Sample set with similar composition that is examined to analyse the impact of
sulphur in the precursor on film morphology. They were chosen to minimize the influence
of the composition. Samples marked with an asterisk (*) were made of the same (ZnS-)
precursor, sample #17 is made of a metal precursor.
1
Fig. 4.8 e) (ZnS) and f) (metallic) show a cross section, though
4.2 Properties of sulphurized films
67
Figure 4.15: Metallic (#17) and two ZnS (#9 and #10) CZTS films. No increase in grain
size for the samples that contained S in the precursor can be observed. Instead, a trend to
smaller grains and more compact films is visible, especially comparing #17 and #10.
Summary
Both the general overview and the experimental matrix with similar-composition samples
indicate that replacing Zn with ZnS in the precursor does not increase grain size but rather
leads to even smaller grains. However, the use of S in the precursor appears to lead to
more dense films with less or smaller grains, even if that is a weak trend. Further work
could enlighten this question by an extended series with metallic and ZnS samples that
are very close to each other in composition to exclude compositional effects.
4.2.3.4 Thickness development
Both the metallic and the ZnS precursors contained less sulphur than needed for CZTS.
For this reason, the film is subjected to strong expansion: CZTS has a lower density than
68
4 Results and discussion
the metals, and up to 50% of the final number of atoms has to diffuse into the film and
build structures under sulphurization (see chapter 2.3.4).
To get an idea how strong the changes within a film are when CZTS is formed under
sulphurization, it can be helpful to evaluate the expansion from the precursor to the final
CZTS film and look at surface features as well as roughness of the film.
The thickness data for all samples can be found in Table 4.1. The expansion for metallic
precursors varies from just less than 2 up to even 4 times. At the same time, the CZTS
films are in most cases very uneven so that expansion factor could vary a lot between
different areas of the same sample. In the most extreme cases, the expansion was between
2 and 4 times, depending on the considered area. On average, the maximum and the
minimum thickness measured for one sample varied around 50%. On the other hand only
two of the nine samples showed remarkable surface features on the viewed images.
The impression for the ZnS precursors is, in contrast, that the expansion of the films
was less dramatic than for the metallic ones. First, the increase in thickness was except of
one case in maximum 2,5 or less. In most cases the expansion was around two, which was
the lower limit for the metallic precursors. This is not surprising as the ZnS precursors
already contained up to 20% sulphur, so less incorporation of additional material was
needed. Second, the ZnS films were in general smoother, that means they had less hills
and valleys and also showed less thickness variation between different images. In average,
the thickness for a sample varied around 25%, i.e. only half of the value found for the
metallic films. However, it should be noted that almost all ZnS samples showed enormous
surface features at some parts and piling up of material. This structure resembles the
surface features seen in Fig. 4.5 b), and therefore it might be tin sulfide. However, deeper
analysis would be necessary and could be part of future work.
In conclusion, the use of ZnS in the precursor obviously leads to smoother films and
at the same time decreased expansion of our films. Considering that expansion always
can cause stresses and cracks within the film and that a film that is more even and shows
less distortion probably is desirable, this is a beneficial effect of S incorporation in the
precursor. At this point, the origin of the obviously increased surface features for ZnS
films is unclear. Further experiments could focus on finding out their nature and the
possibility to remove them, for example by etching. This might increase the performance
of the solar cells. Moreover, stresses within the films could be measured to reveal if ZnS
films show less of it, which might be assumed as the expansion is lower.
4.2.3.5 Role of the substrate
In chapter 3.2.2.1 we showed that – within the error range – the composition of the films
deposited on silicon, glass and glass coated with molybdenum are the same. That is
expected as for each sample all the three substrates were sputtered in the same run.
However, even if the samples were then sulphurized in the same run, the morphology does
not have to be the same. We would rather expect that the substrate plays an important
role for the growth. First, the orientation of the substrate crystals might influence the
4.2 Properties of sulphurized films
69
grain growth; second, elements from the substrate could diffuse into the CZTS, which can
potentially have major influence on the film quality. For example it is known for CIGS
that sodium has a beneficial effect on the material such that films grown on sodium free
glass or other substrates need supplementary Na to achieve same efficiency [53].
Therefore, it was not surprising that the cross sections for the different substrates
revealed completely different images. Typical images for the three substrates coming from
the same precursor deposition are shown in Fig. 4.16.
Fig. 4.16 a) shows a CZTS film on molybdenum substrate. Note that it shows a final
solar cell, i.e. there is a top layer of ZnO (CdS is too thin to be seen), even though it
varies in thickness. The film shows relatively large grains that are clearly defined. The
surface is quite even and no larger surface features can be seen. There are few voids and
the films seem to adhere to each other.
b) shows the corresponding film on silicon. No further layers like ZnO were deposited
here. The adhesion to the substrate appears to be very poor. Some voids can be seen;
some other pictures show a more porous structure. The grain size is very small, in most
cases only one to two hundred nanometers, but it is very difficult to estimate it as the
grains are not as well defined as for the Mo-substrate. Some surface features could be
observed.
c) shows CZTS grown on glass substrate. In this image, the substrate cannot be seen as
the film peeled off from it. Other pictures showed better adhesion, but never as good as for
the Mo. The grain size seems to be a mixture of what was seen on Mo and Si: Some larger,
well-defined grains, next to extremely small grained without contours. Most conspicuous
are, however, the gigantic surface features and distortions that dominate the image. Other
images support the existence of unidentified surface features in the micrometer range and
show huge voids where the film still adhered to the glass.
Summary
It should be noted that these conclusions were not drawn only on the basis of the three
presented pictures. They served only to illustrate a general trend that was observed on
several samples.
Film growth and quality appear to be completely different for each of our substrates.
Fortunately, the by far best results are achieved for molybdenum coated glass, which is
quite satisfying as these are the substrates we used for solar cells. Films grown on silicon
showed much smaller grain size and were partly lifted up from the substrate as a result
of poor adhesion. As these films only were used to measure composition with EDS, this
is not a basic problem, as composition was shown not to vary significantly between the
substrates. The glass substrates, however, which were used for XRD measurements (see
next chapter), showed a clear difference in appearance. We cannot exclude occurrence of
different secondary phases as compared to Mo. Especially for the surface features we could
not identify if they consist of secondary phases that on Mo substrates where enclosed into
the film, or if growth on glass enhanced the appearance of further secondary phases. It is
70
4 Results and discussion
Figure 4.16: Comparison of CZTS films on different substrates. a) On Mo: The film looks
quite even and has only few voids. The adhesion between the different layers is good. b)
On Si: The adhesion between CZTS and substrate is very poor. The grains are considerably smaller than on the Mo substrate. c) On SLG: The films is completely lifted off. The
film is very uneven, large surface features can be observed. Grain size is a mixture of very
small and some larger grains.
4.2 Properties of sulphurized films
71
possible that these structures are related to the surface features seen in Fig. 4.5 b), which
would mean that they are probably tin sulfides. However, further work would be required
to identify these secondary phases; for this we have examined too few samples.
4.2.3.6 XRD analysis
The analysis of samples by XRD can provide important insights in the structure of films.
Especially the identification of crystalline secondary phases is a powerful tool that is not
possible with other analyzing methods used. However, the drawback of this technique is
that XRD results are difficult to interpret in cases of peak overlap and multiple-phase
systems. As discussed in more detail in chapter 3.2.4 crystalline phases show several peaks
in X-ray diffraction. Different phases can show overlapping peaks, as in this thesis CZTS,
CTS and ZnS. They cannot be distuingished in our XRD analyses.
Not related to XRD, but an issue in our case, is that all measurements were performed
on glass, only few samples got an additional measurement on Mo or Si. Even though they
did not show any fundamental difference to those on glass, no hypothesis shall be educed
only from XRD measurements.
The first result is that all of our sulphurized samples showed the Σ–signal. As mentioned,
this is not a definite proof of CZTS, but as the composition is always close to CZTS
stoichiometry and we could in most cases produce working cells with a band gap of roughly
1,6–1,7 eV (see chapter 4.3.2.2), i.e. the presence of CZTS is very likely. Cu2 SnS3 as
single phase would give a different band gap (≈ 0,9 eV) and has a completely different
composition. For this reason we assume to have mainly CZTS. CTS and ZnS might occur
as secondary phases (from the phase diagram in Fig. 2.16 we would expect ZnS for Cu-poor
and Zn-rich samples and CTS for the Zn-poor ones) and increase recombination or reduce
the active area, respectively.
The second observation is that all samples show additional peaks next to the Σ–signal.
The spectrum reaches from one additional peak to more than ten. We conclude this means
that we never obtained phase pure CZTS, not even in the cases where the samples lay in
the stoichiometric region of the phase diagram and the absence of secondary phases might
be assumed.
Concerning identification of the secondary phases, our third conclusion from XRD
measurements is that it appears that most peaks cannot clearly be assigned to a certain
phase. There are only four cases where we believe that we can claim to have found a
certain phase.
Three samples show signals that can clearly be assigned to SnS2 (hexagonal structure)
as a secondary phase (Fig. 4.17). All samples lie in the Sn-rich (or Cu-poor, respectively,
which is still on the opposite side of Sn-rich) region of the phase diagram, i.e. compositional
measurements support this result. Deeper analysis shown in chapter 4.2.3.2 could at least
for one of these three samples show that the secondary phase SnS2 is present as a layer
under the CZTS. For the other two samples there is a strong evidence from images of the
cross section.
72
4 Results and discussion
In the XRD pattern of one sample, CuS (hexagonal structure) could be found (Fig. 4.18).
This sample was in the Sn-poor region of the phase diagram where copper sulfides would
according to the TPD (Fig. 2.16) would be assumed.
Figure 4.17: Samples #8 (red line), #9 (black) and #10 (blue). The red vertical lines
indicate the kesterite structure. These samples matched to the SnS2 signal (Berndtite-2T
SnS2 , hexagonal structure) which is inset with blue vertical lines.
Figure 4.18: Sample #24. The red vertical line indicate the kesterite structure. This
sample matched to the CuS signal (hexagonal structure) which is inset with blue vertical
lines.
4.2 Properties of sulphurized films
73
Summary
XRD analysis show – in combination with other analyzing methods like EDS – that CZTS
is most likely the main phase in our films. All samples show also additional signals that
cannot be assigned to the Σ–signal, though. In most cases we could not clearly identify
secondary phases. In three cases, however, SnS2 could be found; one sample shows copper
sulfide. All these secondary phases are supported by what can be expected from the
samples’ position in the ternary phase diagram.
4.2.4 Conclusions
All sulphurized films were subjected to SEM and XRD analysis. The main focus was on
the examination of the cross section pictures from which we drew conclusions concerning
morphology (grain size, voids etc.), especially relating to composition and the use of ZnS
in the precursor.
Taking all samples into account, composition had no influence on grain size, voids and
compactness of our samples, with the only exception that Zn-poor material showed no
films with big grains at all. The only impact of composition on morphology could be seen
for layering, where (near-)stoichiometric films tended to show less layering.
Concerning secondary phases we can state from XRD that no sample could be proven
to be phase pure CZTS, because additional peaks were always seen, even if they could not
clearly be assigned to a certain phase. It has to be noted that XRD was done on glass,
though, and SEM pictures showed a different film structure for glass samples. Therefore,
this conclusion does not have to be valid for the solar cells. On the other hand, even the
absence of additional peaks in the XRD pattern does not prove the absence of secondary
phases as they can be amorphous.
For the layering, however, XRD results match to the observation that several ZnS
samples show a sublayer of SnS2 . There were three samples that lay in the Sn-rich region,
showing a sublayer that looked different from the (supposed CZTS) top layer and at the
same time had been proven to contain SnS2 by XRD. For one sample the existence of SnS2
in the sublayer could even be confirmed by EDS. These results support the assumption
from chapter 4.2.2.1 that ZnS samples rather enclose excess tin as secondary phase while
metallic samples loose excess tin mainly by evaporation of tin sulfide.
One sample far away from the Cu-poor region showed copper sulfide as secondary phase
in XRD. This and the results for SnS2 indicate that XRD obviously needed larger amounts
of secondary phases to reveal their existence.
Further secondary phases could be found on the surface of films. Compositional
measurements with EDS suggest here that these surface features are copper- and tinsulfides as well.
One important aspect in our experimental series was to examine the influence of sulphur
in the precursor by using a ZnS instead of a Zn target for sputtering. The conclusions
we could draw were that those ZnS samples did not produce larger but rather smaller
grains. Instead, more compact films with less voids have been observed, although this
74
4 Results and discussion
is more a trend and not valid for each sample. Another consequence was that the films
were subjected to less expansion during the sulphurization process due to less integration
of further sulphur. This led to smoother films with less thickness variation as well (but
possibly more surface features). As discussed in more detail in chapter 2.3.4, those
beneficial effects of incorporated sulphur in the precursor are due to less Cu diffusion to
the surface, which would leave voids within the film.
MoS2 could be found as an additional layer on the Mo.
In conclusion we can say that it was possible to control film quality only to a very limited
extend. The use of ZnS in the precursor could reduce voids and increase compactness;
the choice of films that lie in the stoichiometric region of the phase diagram might
reduce layering and the formation of secondary phases. As secondary phases copper- and
tin-sulfides could be identified for our samples.
However, these results concern only the appearance of the material. The next chapter
will enlighten the impact of these parameters on efficiency, which is after all the most
important parameter for a cell.
4.3 Solar cells
75
4.3 Solar cells
The final analyses that was done concerns the solar cell performance. The most important
aspect for us to evaluate the quality of the cell is the efficiency, determined from IV
measurements. Next to it, also QE measurements were performed.
4.3.1 Efficiency
To get an overview about the efficiencies of our fabricated solar cells, the phase diagram is a
very helpful tool. Especially trends in terms of composition can much easier be discovered
than from the raw data.
Fig. 4.19 shows the phase diagram with all our samples that were part of this study.
They are classified in different efficiency groups and for it marked with colours. The black
points symbolize cells where no efficiency could be measured, for example because they
were completely shunted. All cells with better efficiency up to < 1,2% are marked in brown,
with 1,2% and more up to < 2% in dark green, and the best cells with more than 2%
efficiency are shown in blue.
It is conspicuous that Zn-poor samples obviously gave really bad results, there was
only one cell that was working at all (0,1% efficiency). A bit more surprising is that cells
within the stoichiometric region of the phase diagram are not really good either. Except
of one sample they have all efficiencies of 0,5% and less. The exception has 1,1% and is
the sample nearest to the Zn-rich region. This observation goes with the fact that the
best solar cell efficiencies can be found in the Zn-rich region, in average slightly better
than those from the Cu-poor region. For this reason we conclude that we have a clear
trend that – to a certain extend – more Zn in the film increases the efficiency.
To illustrate this we plot the measured efficiencies against the Zn content, more precisely
against the Zn/(Cu+Sn) ratio. Included are Zn-poor, stoichiometric and Zn-rich samples.
The plot can be seen in Fig. 4.20 and supports the first impression from the phase diagram:
An increase of the Zn-content leads to an increase in efficiency.
However, it has to be noted that this observation refers to samples that have (almost)
the same Cu/Sn ratio. Spoken in terms of the phase diagram one would say they lie on a
pseudo binary tie line1 , which is plotted red in Fig. 4.21. Looking for example parallel to
that line above the stoichiometric samples, there are samples in the Sn-rich and Cu-poor
region that have efficiencies comparable to those from the Zn-rich region, even though
they have a lower Zn/(Cu+Sn) ratio.
1
The pseudo binary tie line is a line that assumes the presence of only two components (here: Zn and a
Cu-Sn phase), where the amount x of the one component and the amount 1–x of the other component
are existent. That is for example that x = 1 would mean presence of only one of the two components.
The Cu-Sn phase or the Cu/Sn ratio, respectively, has in this case to be constant; otherwise one would
end up above or below the pseudo binary tie line.
76
4 Results and discussion
Figure 4.19: TPD with all samples of this study ordered by efficiency. While samples
from the Zn- and Sn-poor as well as from the stoichiometric region are almost without
exception poor in efficiency, the best cells come from the Zn- and Sn-rich as well as from
the Cu-poor region.
Figure 4.20: Influence of Zn content on efficiency. A clear increase of efficiency with increasing Zn content is observed.
4.3 Solar cells
77
Unfortunately we don’t have enough samples to make a plot like in Fig. 4.20 for a
parallel pseudo binary tie line like the light blue dashed one in Fig. 4.21. More data points
for example in the region that is marked with an ’X’ would be needed for this. Future work
could examine this very interesting aspect that has so far not been shown in literature.
Considering physical reasons for the observed trend it is not that surprising anymore.
A deficiency of Zn means – provided a sufficient amount of sulphur like assumed in
our experiments – that Cu-, Sn- and Cu-Sn- sulfides are formed. Especially compounds
containing Cu, like CuS or Cu2 SnS3 turn out to be metallic ([31], [33]), which is detrimental
for a solar cell. Conglomerates of those Cu-compounds that extend over the space charge
region would let the electrons flow back directly and they would recombine with the holes
before they could do work at a load. This is referred to as shunting.
Looking at the Zn-rich region of the phase diagram, ZnS is supposed to appear as
secondary phase. ZnS is a less harmful secondary phase (see chapter 2.3.3) as it is a
semiconductor with a band gap of 3,54 eV. That means that in most cases it would just
act isolating, and by this ’only’ reduce the active area, but at least would not shunt the
sample.
This consideration shows: As we assume that we in all of our films have secondary
phases (see chapter 4.2.3), it would be preferable to end up with the least harmful ones.
ZnS might be such a secondary phase. For this reason samples should preferably not come
from the Zn-poor region, to be on the safe side perhaps even from the Zn-rich side. This
approach is strongly supported by our results.
Figure 4.21: Pseudo binary tie line (PBTL, line of constant Cu/Sn ratio, red). Increasing Zn content (left direction) increases efficency of the cells. The blue dotted line is a
PBTL for a different Cu/Sn ratio. A region where samples would have been needed for an
extended analysis of the influence of the Zn content on efficiency is marked with an ’X’ .
4 Results and discussion
78
For samples that are not on the red pseudo binary tie line in Fig. 4.21, one can at least
see that samples that lie in regions of the phase diagram where no Cu- or Cu-Sn-sulfides
would be expected show much better efficiencies than those in the Zn-poor region. For a
real trend more data points with a higher Zn/(Cu+Sn) ratio would be needed, though.
Note: Of course we suspect that this trend will be beneficial only to a certain Zn/(Cu+Sn)
ratio and then make the efficiency drop again.
The two samples lying in the Sn-poor region of the phase diagram are negligible as they
are so far away from stoichiometry which results in 0% solar cells. Most probably copper
sulfides shunt the cell; this secondary phase is both expected from the TPD (Fig. 2.16)
and has been found by XRD (4.2.3.6).
More interesting is that – next to the Zn-rich region – also in the Sn-rich and the
Cu-poor region near the Sn-rich region very good solar cells were produced. The secondary
phase found here with XRD (4.2.3.6, 4.2.3.2) was SnS2 , which is a semiconductor with a
band gap of 2,2 eV (2.3.3). This secondary phase is less detrimental for the solar cell than
metallic phases like copper sulfides.
4.3.1.1 Influence of sulphur in the precursor
Like for the morphological considerations, the influence of sulphur already in the precursor
shall be examined in respect of efficiency as well. As we concluded from the studies of the
cross sections (4.2.3.3) sulphur in the precursor leads in our case assumedly to smaller
grains, but more compact films with less voids. However, next to the question if one would
assume rather compact films or films with large grains to be better for a good solar cell, at
this stage we want to consider the correlation ZnS precursor ↔ efficiency from a general
point of view, not taking into account morphological parameters. This is done in a further
chapter (4.3.1.2).
General impression
As we have seen from the last chapter (4.3.1) the composition has a major influence on
the efficiency. Zn-poor and stoichiometric samples give rather poor cells, while Zn- rich
and Cu-poor cells show in general relatively high efficiencies1 . Therefore it is reasonable
to take this into account comparing metallic and ZnS precursors.
In Table 4.5 the average efficiencies for metallic and ZnS precursors ordered by phase
diagram region is listed. No solar cells made of metallic precursors lay in the Zn-poor
region, so here is no comparison possible. For the other regions the average efficiency for
ZnS samples is (slightly) better than for the metallic ones, but the number of samples
(mostly 2–5 samples per each precursor and compositional region) requires being careful in
drawing major conclusions from this. The comparison of samples very close in composition
1
There are two samples that would actually be referred to as "Sn-rich". As only two samples would
not allow to see a trend here and as they lie on the border to the Cu-poor or stoichiometric region,
respectively, they are just added to these categories in this case.
4.3 Solar cells
79
might support or disprove this trend. This will be done in the next subchapter. For the
general overview, however, we state that in average a slightly higher efficiency in favour of
the ZnS cells could be observed.
Comparison of similar compositions: Experimental matrix
The same samples that were used to find morphological differences between metallic and
ZnS samples very near in composition in the chapter 4.2.3.3 will now serve to consider
correlations between efficiency and sulphur in the precursor. Table 4.6 shows the examined
solar cells. #17 is made of a metallic precursor, #9 and #10 were made of the same
ZnS precursors, but in different runs (same sulphurization conditions, though). The
composition can be considered as very close.
The ZnS samples show better efficiencies. However, one has to take into account that
the two ZnS samples #9 and #10 differ more from each other than metal sample #17 and
ZnS sample #9. This makes it again very difficult to draw an unambiguous conclusion,
but it at least does not contradict the weak trend from the general consideration that ZnS
samples are slightly better, but rather supports it.
Summary
The influence of sulphur in the precursors was analysed regarding efficiency. The general
overview, where the solar cells were grouped by compositional regions of the phase diagram,
as well as a comparison of samples near in composition lets us assume that there is a slight
improvement by using ZnS precursors. However, experiments with a larger set of samples
than it was possible to do in this work would be needed to prove or disprove this trend.
Metal
ZnS
Cu-poor
1,3% (2,1%*)
2,3%
Zn-poor
–
0,02%
Stoichiometric
0,4%
0,5%
Zn-rich
2,0%
2,7%
Table 4.5: Average efficiency of metallic and ZnS precursors from the different regions of
the phase diagram. One can see a slight trend that ZnS samples produce better efficiencies.
No Zn-poor metallic films were sulphurized.
*All Cu-poor samples have an efficiency between 1,8% and 3,2%, with only one exception.
As one of the 2 metal samples has only 0,5%, this lowers the metal average efficiency from
2,1% to 1,3%. It is likely that this sample is only an outlier.
#17 (metal)
#9* (ZnS)
#10* (ZnS)
Cu/Sn (sulph)
1,79
1,76
1,78
Cu/(Zn+Sn) (sulph)
0,89
0,91
0,92
efficiency
2,1%
2,5%
3,2%
category (TPD)
Cu-poor
Cu-poor
Cu-poor
Table 4.6: Sample set with similar composition that is examined to analyse the influence
of sulphur in the precursor on efficiency. ZnS samples show better efficiencies.
80
4 Results and discussion
4.3.1.2 Correlation between morphology and efficiency
It is of interest to study if a correlation between the appearance of a film and its performance
can be seen. From this one could deduce which parameters – like grain size or compactness
of the film – should be subjected to more effort to enhance them in future experiments.
However, we cannot see any correlation between material quality and efficiency. First
we consider grain size. One would suspect that larger grains and the hence resulting fewer
grain boundaries lead to a larger mean free path of the electrons and reduce recombination.
Less recombination should result in higher current and higher efficiency. Looking at
Fig. 4.22 one can see two samples with comparable, large grain sizes. The larger grains are
roughly between 0,7 to 1 µm. However, the efficiency is significantly different: the sample
in Fig. 4.22 a) has 0,5%, while the sample in b) shows 2,3% efficiency.
For small grains the result is the same, even if we do not show such an explicit comparison
here. Fig. 4.24 can serve as an example since those samples have both small grains but
very different efficiencies.
We conclude that grain size in our experiments has obviously no influence on the
efficiency.
Figure 4.22: Illustration of the correlation between morphology and efficiency. Both samples have large grains, but while the sample in a) (#16) has 0,5% efficiency the cell in b)
shows 2,3% (cell #22).
4.3 Solar cells
81
Figure 4.23: Cross section of sample #17. Huge holes can be seen, nevertheless 2,1%
could be reached with this cell.
Next, we consider voids and compactness. Fig. 4.23 shows a sample with considerable
voids and a very uneven film with 2,1% efficiency. This sample might be compared to
Fig. 4.22 b), where a film can be seen with less voids and a smoother surface, but roughly
the same efficiency (2,3%).
For small grained films the picture is the same: Compactness or number of voids seem
to have no impact on the efficiency. Fig. 4.24 shows two samples, both small grained and
very compact. The upper one has an efficiency of 0%, while the lower one shows in spite
of the sublayer 3,2%.
Considering the sublayer it was found (chapter 4.2.3.2) that this is most probably a
segregated SnS2 phase. As mentioned in the theory part (chapter 2.3.3) this phase is an
n-type semiconductor which would be detrimental for the p-type solar cell. However, the
solar cell with the best efficiency of 3,2% was a cell with a SnS2 sublayer.
We conclude that voids, compactness and even clear layering did not have an influence
on the performance of our solar cells either.
Summarizing one can say that morphology and efficiency were independent from each
other in our experiments. It was not possible to judge from the visual appearance if a
solar cell might show good or bad performance.
The reason for this is quite unclear. One would suggest that films with large grains or
compact CZTS would lead to better conditions for the electrons (diffusion length) and
in doing so increase the efficiency. That we do not see this correlation might relate to
several aspects. First, the level of our efficiencies and other data gained from QE and IV
measurements (Isc , Voc , FF) is rather low (see the following chapters 4.3.2, 4.3.3). That
means that a lot of parameters probably influence and decrease the performance of our
cells. Shunting or isolating phases might be present, the CZTS might be lifted from the
substrate at some places, or the top layers (CdS + ZnO) might not be perfect everywhere
(pictures with few or not ZnO at all could be seen). Second, we can not be sure that
all these parameters that have influence on the efficiency (secondary phases, ZnO, voids,
layering, bad adhesion etc.) could be seen on the images for every sample. We divided
every sample in different cells and measured efficiency separately. Not always did the
several cells on one sample show the same result, but often the efficiencies varied a lot.
82
4 Results and discussion
Figure 4.24: Two samples with comparable compactness. a) #14, efficiency 0%. b) #10
with 3,2% efficiency.
On the other hand we looked at only one cross section, and in doing so perhaps saw only
bad film quality while the efficiency measurement was performed at another, better region
of the sample. Unfortunately it was technically not possible to measure efficiency and
cross section at exactly the same place. For this reason we have to trust that we got an
overview about the general morphology of each sample, even if we made only one cross
section. It is very possible that influences that cannot be seen with an electron microscope
(like secondary phases) decrease the performance of a cell a lot, even if the appearance
shows nice grains or compact films.
4.3.1.3 Result from the efficiency measurement
We examined the influence of morphology, sulphur in the precursor and the composition of
the sulphurized film in terms of efficiency. It could be shown that only composition has a
clear impact on the performance of a cell. Here we found that the efficiency increased going
from the Zn-poor to the Zn-rich region on a pseudo binary tie line with almost constant
Cu/Sn ratio. The best cell efficiencies were gained for the Zn-rich and Cu-poor region. For
a general beneficial effect of sulphur in the precursors only a weak trend was noticeable.
For parameters like grain size, voids and layering no correlation to the efficiency could be
found at all.
4.3 Solar cells
83
4.3.2 QE measurements
Quantum efficiency measurements are a useful tool to get informations about the losses in
a cell. For more details about QE measurements see chapter 3.2.5.
Fig. 4.25 shows the QE data for all samples with reasonable results, i.e. where the typical
pattern is observable. ’Typical pattern’ means for example the kink in the graph starting
below ≈ 520 nm. This is the "CdS-kink", resulting from the CdS band gap at 2,42 eV [53]
where the CdS layer starts absorbing photons and is not completely transparent any longer.
Electrons produced in the CdS are assumed to be lost. The ZnO absorbs most at > 900 nm
(not relevant here), where free carrier absorption comes into play, and below 400 nm at
the band gap for ZnO [53]. For longer wavelengths the band gap of CZTS is limiting.
As shown in chapter 4.3.2.2 this starts around 750–800 nm for our cells. Furthermore,
recombination at the back contact concerns mainly electrons produced by long wavelengths.
Next to these losses, reflection of the layers (CZTS/CdS/ZnO) lowers the whole level.
The latter loss should only reduce quantum efficiency by some percentage, the other
losses are supposed to influence the QE only above or below certain wavelengths (those
named before). However, what we see is that the overall level of our quantum efficiency
is very low and reaches in the best case a bit more than 60%. In most cases half of the
incoming energy or even more is not converted into usable current. That means that
fundamental losses reduce the whole level. These losses could be
1. Shunting: Metallic secondary phases or pinholes bypass the space charge region and
the separated electron-hole-pairs recombine directly.
2. Reduced active area: The actual area that contributes to the current might be
reduced by several influences:
• Isolating secondary phases reduce the area as they do not produce electronhole-pairs and hinder the electrons to come to the front contact.
• The same is valid for voids.
• Parts of the film can due to bad adhesion flake off. Even if not the hole film
falls off, if it happens after deposition of CdS and ZnO those layers fall off as
well (seen in some pictures) and the current cannot be transported away.
3. Layering: Parts of our samples showed that the absorber was divided into layers.
Besides the fact that the sublayer turned out to be a secondary phase (SnS2 ) at least
in some cases, a crack within the absorber will definitely impair electron migration.
4. Interface recombination: Recombination at the interfaces, i.e. between CZTS and
CdS, CdS and ZnO as well as surface recombination, concern all electrons and
therefore lower the whole current level.
5. Recombination at defects: Secondary phases, holes, unwanted impurities etc..
6. Voids: Voids, like for example in Fig. 4.14, can reduce the active area of the cell.
84
4 Results and discussion
The differences regarding the level between the different curves in Fig. 4.25 can in the
main be explained by the different efficiencies. The red curves have an efficiency better
than 2%, the blue ones between 1,2% and 2%, and the lowest of 0,4% (black). As shown in
chapter 4.3.1 this has for our samples mainly to do with the composition, which probably
effects secondary phases, even if we cannot point out here if certain phases are correlated
to certain quantum efficiencies or losses of QE.
Figure 4.25: QE curves of all samples with reasonable curves, sorted by efficiency.
There is one distinctive difference between the curves, though, that does not concern the
overall level but only the long wave lengths. Looking at sample #21 and #22 (Fig. 4.26)
one can see that these curves show some kind of plateau, which means that they in contrast
to the other curves do not slope after reaching the maximum value. Even if two samples
are the minority one could assume that they represent the optimum case. The slope after
the maximum QE value means an increasing loss for higher photon wavelengths. This
effect was also observed in literature (see for example [54]). A first idea would be that the
films are too thin and low energy photons just pass the solar cell without being absorbed.
However, these two samples are not particularly thick compared to the other cells. Too
thin films could thus only mean a general lowering of the QE level for higher wavelengths
for all samples. Indeed there are some more reasons that could cause this effect which we
want to discuss in the following.
4.3 Solar cells
85
4.3.2.1 Possible reasons for slopes in the QE curves
In order not to complicate the discussion we want to consider samples very near in
composition as we already know that composition influences both efficiency and the
presence of secondary phases strongly. Next to the two samples with a plateau, there is
one more sample made of the same precursor, and the three samples are very near in
composition after sulphurization as well (see Table 4.1). For the sake of clarity the three
graphs are plotted again without all the others in Fig. 4.26, which makes the difference
even clearer. The samples #21 and #22 show a plateau that #23 does not have. Three
possible explanations will be discussed below.
1. A small space charge region
As discussed in the theory part (chapter 2.1.7) is a large space charge region (SCR)
beneficial for a solar cell as more electron-hole-pairs are excited within it and directly
feel the electrical field. All other electrons have to reach the SCR by diffusion;
therefore a smaller SCR results in fewer electrons reaching it. The recombination
probability increases with increasing distance to the SRC, i.e. for electrons produced
deeper in the absorber. Those electrons are generated by long wave length photons,
which is the correlation of interest here.
As described in the theory part of this work (chapter 2.1.4) the extension of the
SCR depends on the doping: Less doping of the absorber leads to a larger SCR. For
our material the doping is intrinsic, that means no material is added for doping but
this happens by intrinsic defects. The doping is thus influenced by the composition,
which for our three samples is very similar. Even for a look on all samples and their
composition no trend to a correlation between composition and absorption of long
wave photons can be seen. This means that we cannot proof this theory.
2. A bad back contact
Like all interfaces the back contact is a region of increased recombination. Electrons
produced deep in the absorber, i.e. nearer to the back contact, have of course a higher
probability to reach this interface by diffusion. If we would find some observable
difference for the back contacts (we have only the tool "cross section") this might be
an explanation for the different QE pattern of the three cells.
Considering the film, especially sample #21 and #23 are quite similar; both show
some kind of layering near the back contact (Fig. 4.27). This makes obviously not
the difference. However, looking at the other side of the back contact, the Mo or
rather the MoS2 (see chapter 4.2.3.2), a difference between the samples with plateau
for the QE curve and the sample that has a slope can be seen. The MoS2 is much
thinner for the latter, which might indicate some influence of this layer on the back
contact. For CIGS it is shown that a very thin MoSe2 layer at the back contact
is a so called back surface field. This can be seen as an electronic mirror where
electrons are reflected (by a potential wall) and so the back contact recombination is
4 Results and discussion
86
Figure 4.26: 3 samples with similar composition (made of the same precursor). The
curves of sample #21 and #23 show a plateau, while the curve for #22 slopes after reaching the maximum QE value.
reduced [55]. In the same way the MoS2 might have some beneficial influence for
our back contacts. This is admittedly a very unconfident theory. No such an effect
was reported for CZTS so far and the band structure for CZTS (here needed for the
back contact) is not known either. Furthermore, samples #21 and #22 are the only
ones with a plateau in QE but not the only ones with thick MoS2 . Therefore we
conclude that an influence of MoS2 on the back contact and the recombination at
the same place is not impossible, but we cannot prove it.
3. A low diffusion length
To contribute to a current, the electrons produced deep in the absorber need to reach
the SCR by diffusion before they recombine. The diffusion length of the electrons
depends on the material quality, i.e. grain size, compactness, as well as on voids and
other crystal defects that enhance recombination of the electrons. To see if such
aspects might be responsible for this QE results we take a look at Fig. 4.27. a) and b)
show the samples that have a plateau in the QE curve (#21 and #22). Both samples
have at the same time comparatively large grains in the µm range. Fig. 4.27 c) shows
sample #23 which has similar composition but a slope in QE measurements for
higher wavelengths, and the grains are significantly smaller (<0,5µm).
4.3 Solar cells
87
Figure 4.27: Cross sections of the three samples #21 a), #22 b) and #23 c) that were
analysed regarding the QE curve. While the samples with the plateau in the QE curve have
large grains (#21 and #22) sample #23 has considerably smaller grains, which might be
the reason for the worse quantum efficiency for higher wavelengths.
4 Results and discussion
88
This means that the grain size could be a possible explanation for the differences of
the QE curves for these three samples. As grain boundaries are (like other crystal
defects) a region of potentially increased recombination, this looks like a plausible
explanation. However, samples with large grains and a slope in QE exist as well,
even if they are due to their different composition not comparable with these three
samples. Hence, we conclude that the presence of large grains could be a necessary
but not sufficient condition for the collection of electrons produced by long wave
photons.1 It is noticeable that at the same time efficiency was not influenced by
this aspect: The samples with plateau in QE and large grains have 2,6% and 2,3%
efficiency, respectively, while the sample with the slope shows 2,7%.
4.3.2.2 The bandgap
A common way to estimate the band gap Eg is to extract it from the squared quantum
efficiency, like described in chapter 3.2.5. For it the x-scale is converted from wavelength to
energy and the QE-values are squared. The slope at the long wave length side (corresponds
to the low energy side) of the curve is extrapolated to 0, like shown in Fig. 4.28. As this
includes of course some freedom where exactly to draw the extrapolation line, we show
the extrapolations of the examined QE2 -curve for sample #22. We consider this more as
estimation for the band gap than an exact definition since the edge in general is not sharp.
Fig. 4.28 shows also three more curves, even if only the detail of the extrapolation. The
samples are #21–23 which were examined in the last chapter (4.3.2.1) and additionally
a sample with a much lower QE level (sample #15, dark blue in Fig. 4.28). The band
gaps for our CZTS lie between 1,5 and 1,65 eV (corresponding to ≈ 750–800 nm). This is
slightly higher than what is reported in publications (1,45–1,5 eV, [20], [56]).
4.3.2.3 Summary
The QE measurements revealed that fundamental losses reduce the quantum efficiencies
of our cells. Possible reasons are conducting and isolating secondary phases, voids, bad
adhesion between the layers as well as layering. We assume that all these aspects contribute
in varying proportions to the low QE level, as they all have been proven by our analyses.
For the fact that most curves slope after reaching the maximum QE value we suggest after
comparing near-compositional samples that grain size could be one responsible parameter.
Larger grains might increase the probability for the collection of more electrons that were
produced by long wavelength photons. An influence of space charge region and back
contact could not be proven.
1
Morphological aspects that could be responsible for a lower diffusion length (like voids, layering,
compactness etc.) but did not turn out to be different for the "plateau-samples" and the "slope-sample"
on the cross sections are not discussed here. Grain size was the only clear difference.
4.3 Solar cells
89
Figure 4.28: Possibility to extract the band gap from the QE curve. The band gap is the
interception of the interpolated slope of the squared QE-curve.
From squared QE-curves we extracted a rough estimation for the band gap. It is for
our samples between 1,5 and 1,65 eV.
4.3.3 IV measurements
For further analysis we performed IV measurements. During IV measurements the cell is
illuminated and the current is measured while at the same time the voltage is varied. The
result is (for good cells) a diode curve like discussed in chapter 2.1.7.
For this discussion we want to concentrate on the clearest cases to point out possible
influences on the IV behaviour of the cells. The other cells lie somewhere in between and
are thus subjects to several influences.
Fig. 4.29 a) shows the three solar cells with the best parameters we got. Sample #10 had
the highest efficiency (3,2%) and open current voltage (0,72 V), #21 showed the highest
mA
short circuit current density (Isc = 12,6 cm
2 ), and sample #17 had the best fill factor of all
our cells (FF = 47,9%).
The blue line in Fig. 4.29 a) represents sample #10. It has the highest η and Voc of all
mA
our solar cells. Even the current is with Isc = 10,9 cm
2 in the range of the best values we
got. The voltage is not the maximum one would expect from theoretical foundations but
90
4 Results and discussion
exceeds the so far best published results by ca. 10%.1 The efficiency is clearly lower than
the so far world record of 6,8% (IBM, [7]). The high slope at forward bias suggests that the
cell has a high series resistance. First, the curve is not exponential for high voltages like for
an ideal diode but a linear slope (I = RUS ). Second, the curve is not almost "rectangular
shaped" like an ideal diode curve, but the current is very voltage-dependent. That means,
as soon as current is flowing, a part of the voltage drops due to the series resistance.
The same is valid even to a greater extend for sample #21 (red curve). It has the
highest Isc found for our cells. It drops even more even for quite low voltages. Furthermore,
increasing the voltage in the reverse direction of the diode results in increased current. A
possible reason for this shunting, which means we would also have a parallel resistance
RSH . A high parallel conductance could even be responsible for the lower Voc compared
to the maximum value we found (for more details see chapter 2.1.9). Furthermore, voltage
dependent carrier collection (see chapter 2.1.6) can cause this phenomenon.
The maximum current for our solar cells is significantly lower than one could expect
mA
mA
from theory (see chapter 2.1.7) and what other groups received (22–24 cm
2 and 15–20 cm2 ,
respectively; [37], [7]). This problem is related to what was said about the QE measurements
(4.3.2): General losses caused by secondary phases, interface recombination and such effects
mentioned before reduce the overall electron collection and in doing so reduce the current
compared to what would be possible in a perfect cell.
The most significant loss we got was in the fill factor (FF). As defined in the theory part
I ·Vmp
(equation 2.7) it is FF = Imp
. That means even compared to our open circuit voltage
sc ·Voc
Voc and the comparably low Isc we gain very low power. Best solar cells made of CZTS
reached above 60% ([37], [7]), our maximum value was ≈ 48%. This is seen for sample
#17 (green line in Fig. 4.29 a) ). It shows the most "rectangular" shape of the curve; the
linear behaviour of the curve for high voltages indicates again series resistances.
However, several samples were far away from reaching that; while the average fill factor
for all of our cells is 33,3%, some samples have a FF of only 25%. This is the fill factor
of a linear slope (conductor), and three examples of this kind can be seen in Fig. 4.29 b).
The reason is most probably that the cells are completely shunted. A voltage dependent
1
The maximum voltage that is theoretically possible is difficult to estimate. Not the whole band gap can
be gained as Voc but only the band gap minus the distance of the doping levels from the conduction
and valence band, respectively (see chapter 2.1.7). Futher reduction is due to radiative losses at
room temperature. For example for CIGS, a material close related to CZTS, there is an empirical
E
formula Voc = qq − 500 mV ([57]). This would suggest a value of around 1 V for a band gap of 1,5 eV.
Publications on CZTS gave up to around 0,65 V in open current voltage (640 mV (IBM,[47]), 662 mV
(Katagiri, [37])). Voc is reduced by recombination throughout the cell and future improvements of
CZTS devices are expected to give at least some improvements in Voc . Wide band gap solar cells often
show a larger loss in Voc relative to their theoretical maxima than for smaller band gap materials. The
reason for this is not completely clear. For CZTS it is assumed that very short life times and interface
recombination might be responsible. For example is always CdS used a a buffer layer like for CIGS;
however, it is unknown if other buffer layers might suit better to CZTS. So far it is unclear to what
extent fundamental material properties of CZTS will limit Voc in the future.
4.3 Solar cells
91
carrier collection can most likely be excluded as the dark curve of the samples looks the
same, which should then not be the case (but look like a normal diode).
Figure 4.29: a) Samples with the best IV data measured in our experiments. #10 has the
best efficiency and the best Voc , #17 the best fill factor and #21 the best value for Isc .
b) Samples #3, #14 and #19 with bad fill factor and efficiency. They were most likely
shunted. For comparison sample #10 that showed the best efficiency in our experiments is
added in grey.
Figure 4.30: FF plotted against Zn content. For ZnS samples generally higher FFs are
achieved than for metallic samples with the same Zn content. The circled samples are
discussed in the text. They have excess Zn, as the vertical dotted line that shows stoichiometry indicates.
92
4 Results and discussion
Two of the curves in Fig. 4.29 b) are by the way of stoichiometric samples which again
indicates that we could have secondary phases for films with stoichiometric composition
as well. However, shunting can also be caused by pinholes.
Considering the fill factor an interesting observation could be made concerning the
difference between metallic and ZnS films. Looking at Fig. 4.30 where the fill factor is
plotted against the Zn/(Cu+Sn) ratio (i.e. the Zn content, compare to Fig. 4.20), one sees
not only the trend of improvement with increasing Zn content. Furthermore, for the fill
factor obviously the ZnS had a major impact. There is a clear trend that samples for a
certain Zn/(Cu+Sn) ratio became better for the films with sulphur in the precursor.
A closer look at the cross sections of the 3 metallic and ZnS samples, respectively, which
are similar in Zn-content but very different in the FF (see marked samples in Fig. 4.30),
reveals that the morphology is actually quite different. While the ZnS samples show only
some voids within the CZTS, do the metallic ones have quite poor morphology. They
have large voids or even cracks, in particular between CZTS and Mo. Pinholes or cracks
through the whole CZTS film could shunt the cell and by this decrease FF. Furthermore
the metallic films may have a different segregation of secondary phases, and isolating
secondary phases like ZnS might increase the series resistance for the solar cells and in this
way reduce FF. As the samples have Zn excess, ZnS would be expected from the TPD.
This observation suggests that the film growth for similar composition obviously is
enhanced by the use of ZnS in the precursor instead of Zn (metallic precursor). Possibly
due to less diffusion in the ZnS samples, fewer voids are formed by migrating elements.
That supports the observation in chapter 4.3.1.1 that the integration of S in the precursor
leads to smoother films with less voids.
However, such a general trend that ZnS samples are better could be seen neither for Isc
nor for Voc .
Summary
mA
The IV-measurements showed the maximum values η = 3,2%, FF = 48%, Isc = 12,6 cm
2 and
Voc = 0,72 V. While the open circuit voltage is higher than reported in other publications
are efficiency, fill factor and short circuit current lower than for published record cells.
The low current indicates general losses for electron collection; the fill factor is reduced by
parallel and series resistances in our cells. The efficiency is affected by these losses (i.e.
lower than it would be possible for this material). The results show that we probably have
secondary phases that cause loss of active area (affects Isc ), increase the series resistance
(FF) and can shunt the cell (FF). Pinholes, voids and other defects could be further reasons
for the described effects. For this experimental series it was not possible to correlate the
observed phenomena to certain defects, in can only be stated that – like for the efficiency
part – the best IV results where found for Cu-poor and Zn-rich samples; stoichiometric
and Zn-poor samples showed a bad IV performance. For the fill factor we observed a
beneficial effect of sulphur in the precursor and assume that this is due to a smoother
sulphurization process with less diffusion resulting in less voids and more compact films.
4.3 Solar cells
93
4.3.4 Conclusions
The best solar cell fabricated in our experimental series had 3,2% efficiency. Further
maximum values for solar cell parameters are a fill factor of roughly 48%, and open circuit
mA
voltage of 0,72 V and a short circuit current of 12,6 cm
2 . For the band gap we estimate
that our CZTS lies between 1,5 and 1,65 eV.
However, the potential for this material is much higher than what we could measure
for our samples. QE and IV measurements revealed that we have significant losses.
Low QE level and Isc indicate general losses in charge carrier generation and collection.
Secondary phases, voids, layering and bad adhesion can by increasing series resistance and
recombination as well as by shunting be responsible for this and have been found to be
present in our samples in a previous chapter. Clear correlations between certain losses
and defects were difficult to find. However, the main influence on efficiency turned out
to be the composition of the absorber (best results for Cu-poor and Zn-rich films), and
large grains might be one factor that is beneficial for the collection of electrons produced
by long wave photons. The former is related to the fact that samples from the Cu-poor
or Zn-rich region would according to the phase diagram mainly have ZnS as secondary
phase, which is the least bad secondary phase for CZTS solar cells, while samples in the
Zn-poor region (where most of our bad films were) can show copper sulfides or CTS which
are conductive and can shunt the sample. A trend that could be seen in our experiments
(but should be examined for more samples in different regions of the phase diagram) is the
beneficial effect of more Zn; on a pseudo binary tie line from the Zn-poor to the Zn-rich
region increased efficiency with the Zn content.
Another observation is that ZnS obviously has a clear beneficial effect concerning the fill
factor. While the FF increases with increasing Zn-content (just like the efficiency), almost
all ZnS samples show a higher fill factor than metallic ones for the same Zn/(Cu+Sn)
ratio. We refer this effect to a better film quality seen for these samples, that might result
from a smoother sulphurization process, which could be enhanced by less diffusion as a
consequence of S already in the precursor.
For a general beneficial influence of ZnS in the precursor next to the fill factor, however,
only a weak trend could be observed.
5 Conclusions and suggestions for future work
The purpose of the presented work was to fabricate and analyse the thin film material
Cu2 ZnSnS4 (CZTS) and to produce solar cells based on this material. In the following the
most important results are summarized.
We used a two-step process to produce CZTS. The first step comprised the co-sputtering
of metals (metallic precursors) and metals + sulphur (ZnS precursors), respectively. Analysis of the precursors showed that they were all Cu-poor and Sn-rich compared to stoichiometry (Cu/Sn in average 1,7 where 2 would be stoichiometric). The composition
was uniform within the precursors, only the surfaces showed some gradients. For metallic
precursors, crystalline phases were found, even though it was not possible to assign all
XRD peaks to certain phases. The ZnS precursors were amorphous as seen by XRD.
As all precursors were fabricated on different substrates (glass, glass coated with
molybdenum and silicon) we studied the influence of the substrate as well. No difference
in composition or compositional gradients could be found here.
The second process step was sulphurization of the precursors. Excess of sulphur was
provided, however all samples showed almost exactly 50% of sulphur content, which is
stoichiometric. That made it possible to examine the samples using the ternary phase
diagram of Cu, Zn and Sn. The main observation here was that all samples showed some
Sn-loss. This was more distinct for the metal precursors, and the conclusion was that
metal films tend to loose Sn by evaporation of SnS2 , while this secondary phase is partly
enclosed for ZnS films. An explanation for different forming of CZTS for the two different
precursors types was suggested.
Good reproducibility of the sulphurization process considering composition and efficiency
was shown.
Further analysis concerned the influence of the composition on the films. A correlation
between composition and morphology was not found, except that the formation of several
layers might be avoided by stoichiometric composition. Moreover, all films showed additional peaks next to the CZTS (Σ) pattern, indicating that all films contained secondary
phases to some extent. Most could not be identified, but for some samples from the Sn-rich
region a Sn-rich sublayer - most likely SnS2 - was found. As those samples were made of
ZnS precursors this supports the theory about the enclosure of tin sulfides in those films.
EDS measurement showed copper- and tin-sulphide conglomerates on the surface.
Even if an influence of composition on morphology could not be proven, the influence
on efficiency was strong. Here we found that more Zn is to a certain extent beneficial for
the performance of the solar cell. The best results were found for Cu-poor and Sn- as well
as Zn-rich films. We attribute this to the fact that Zn-poor samples (which showed the
95
96
5 Conclusions and suggestions for future work
worst performance) probably lead to formation of the most detrimental secondary phases.
Cu- and Cu-Sn-sulfides are conducting and can cause shunting, while ZnS (expected for
Zn-rich films) is insulating and thus mainly reduces the active area of the solar cell.
Another important point we studied was the influence of S in the precursor. A weak
trend could be observed that those precursors developed to more compact films with
smaller grains and slightly higher efficiencies. Concerning the fill factor this trend was a
bit clearer.
The best solar cell parameters that were measured for our cells had 3,2% efficiency, a
fill factor of roughly 48%, an open circuit voltage of 0,72 V and a short circuit current of
mA
12,6 cm
2 . We estimate that the band gap of our CZTS lies between 1,5 and 1,65 eV. Those
values are comparable to previously published data. Nevertheless, we assume that the
potential of our CZTS is even higher, as QE and IV measurements suggest high losses
for the generation and collection of the charge carriers. Increasing grain size and most
of all reducing secondary phases, pinholes and voids might decrease series and parallel
resistance as well as other losses and lead to higher efficiencies. However, it is not clear if
that is possible with this simple sulphurization process.
The importance of composition on the performance of the solar cells shall be emphasized
as the main conclusions from this work. This study systematically compared CZTS films
from almost all different regions of the phase diagram (Zn-rich, Cu-poor, Sn-rich, Zn-poor,
Sn-poor and stoichiometric). The result is that Zn-deficiency produces poor solar cells
and the composition should rather be Zn-rich, Cu-poor or Sn-rich.
Recommendations for future work
The sulphurization process used in this work was very simple and for example concerning
sulphur content and pressure not controllable. Due to the simple furnace used we could
sulphurize only very tiny samples which complicated the analyses in several cases. Therefore
we would recommend using a furnace that allows control of more parameters and the use
of larger samples, which might directly solve some of our problems and would also allow
studying variations in the sulphurization process, which is probably a very important
parameter.
Besides this practical aspect, further experiments concerning results that were somewhat
unclear in this study could be performed. The increase of efficiency with the Zn-content
could only be shown for one fixedCu/Sn ratio and it would be interesting to extend this
study for more compositions. Perhaps even the missing region of the phase diagram could
be examined concerning the correlation between composition and efficiency, even if we
strongly assume that Cu-rich samples will perform badly due to the secondary phase Cu2 S,
which has to be expected according to the phase diagram.
Finally, the reported beneficial influence of sulphur in the precursor was weak in our
experiments. Comparison of more metallic and ZnS precursors near or even equal in
composition could enlighten this question.
97
Bibliography
[1]
Solomon, S. ; Qin, D. ; Manning, M. ; Chen, Z. ; Marquis, M. ; Averyt, K.B. ;
M.Tignor ; Miller, H.L.: Zusammenfassung für politische Entscheidungsträger. In:
Klimaänderung 2007: Wissenschaftliche Grundlagen.
http://www.ipcc.ch/pdf/reports-nonUN-translations/deutch/IPCC2007-WG1.pdf.
Version: 2007
[2]
Bundesanstalt für Geowissenschaften und Rohstoffe: Reserven, Ressourcen und
Verfügbarkeit von Energierohstoffen. http://www.bgr.bund.de/cln_109/nn_322848/
DE/Themen/Energie/Downloads/Energiestudie-Kurzf-2009,templateId=raw,
property=publicationFile.pdf/Energiestudie-Kurzf-2009.pdf. Version: 2009
[3]
IAEA: The Long Term Storage of Radioactive Waste: Safety and Sustainability.
http://www-pub.iaea.org/MTCD/publications/PDF/LTS-RW_web.pdf.
Version: 2003
[4]
Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit: Erneuerbare
Energien. http://www.dlr.de/tt/Portaldata/41/Resources/dokumente/institut/
system/publications/broschuere_ee_innov_zukunft.pdf. Version: 2004
[5]
Green, M. A. ; Emery, K. ; Hishikawa, Y. ; Warta, W.: Solar cell efficiency tables
(version 35). In: Progress in photovoltaics: Research and applications 18 (2010), S.
144–150
[6]
Umweltbundesamt: Seltene Metalle.
http://www.umweltdaten.de/publikationen/fpdf-l/3182.pdf. Version: 2007
[7]
Wang, K. ; Gunawan, O. ; Todorov, T. ; Shin, B. ; Chey, S. J. ; Bojarczuk, N.A. ;
Mitzi, D. ; Guha, S.: Thermally evaporated Cu2 ZnSnS4 solar cells. In: Applied
Physics Letters 97 (2010)
[8]
Neuberger, F.: Module power evaluation report.
http://www.schottsolar.com/global/solar-yield/long-term-study/. Version: 2009
[9]
ISI: Rohstoffe für Zukunftstechnologien. http://www.isi.fraunhofer.de/isi-de/n/
projekte/rohstoffbedarf.php?WSESSIONID=edc588012d961c68488601f69504ffa0.
Version: 2009
[10] Poortmans, J. (Hrsg.) ; Arkhipov, V. (Hrsg.): Thin Film Solar Cells. Wiley, 2006
99
100
Bibliography
[11] Würfel, P. ; Würfel, U.: Physics of Solar Cells. Bd. 2. Wiley-VCH, 2009
[12] http://rredc.nrel.gov/solar/spectra/
[13] Ibach, H. ; Lüth, H.: Festkörperphysik. Bd. 6. Springer, 2002
[14] Green, Martin A.: Solar cells. The University of New South Wales, 1998
[15] Hahn, G. ; Willeke, G.: Halbleitertechnologie und Physik der Solarzelle. Printout.
http://www.uni-konstanz.de/pv/index_l.htm. Version: 2008-2009
[16] Schleußner, S.: Cu(In,Ga)Se2 Thin Film Solar Cells with ZrN as a Back Contact,
Universität Stuttgart, Diplomarbeit, 2003
[17] Zimmermann, U.: Solar Cell technology. lecture script PhD course, 2007
[18] Lewerenz, H.-J. ; Jungblut, H.: Photovoltaik - Grundlagen und Anwendungen.
Springer-Verlag, 1995
[19] Bhattacharya, R. N. ; Ramanathan, K.: Cu(In,Ga)Se2 thin film solar cells with
buffer layer alternative to CdS. In: Solar Energy 77 (2004), S. 679–683
[20] Ito, K. ; Nakazawa, T.: Electrical and Optical Properties of Stannite-Type
Quaternary Semiconductor Thin Films. In: Jpn. J. Appl. Phys. 27 (1988), S.
2094–2097
[21] Maeda, T. ; Nakamura, S. ; Wada, T.: Electronic Structure and Phase Stability of
In-free Photovoltaic Semiconductors, Cu2 ZnSnSe4 and Cu2 ZnSnS4 by First-principles
Calculation. In: Mater. Res. Soc. Symp. Proc. 1165 (2009)
[22] Hall, S.R. ; Szymanski, J.T. ; Stewart, J.M.: Kesterite, Cu2 (Zn,Fe)SnS4 , and
stannite, Cu2 (Fe,Zn)SnS4 , structurally similar but distinct minerals. In: Canadian
Mineralogist 16 (1978), S. 131–137
[23] Lide, D. R. (Hrsg.): Handbook of Chemistry and Physics. 79th. CRC press,
1998-1999
[24] Bernardini, G. P. ; Borrini, D. ; Caneschi, A. ; Benedetto, F. D. ; Gatteschi, D. ;
Ristori, S. ; Romanelli, M.: EPR and SQUID magnetometry study of Cu2 FeSnS4
(stannite) and Cu2 ZnSnS4 (kesterite). In: Physics and Chemistry of Minerals 27
(2000), S. 453–461
[25] Chen, S. ; Yang, J. ; Gong, X. G. ; Walsh, A. ; Wei, S.: Intrinsic point defects and
complexes in the quaternary kesterite semiconductor Cu2 ZnSnS4 . In: Phys. Rev. B
81 (2010), Nr. 24, S. 245204
[26] Scragg, J. J.: Studies of Cu2 ZnSnS4 films prepared by sulphurisation of
electrodeposited precursors, University of Bath, Diss., 2010
Bibliography
101
[27] Olekseyuk, I.D. ; Dudchak, I.V. ; Piskach, L.V.: Phase equilibria in the
Cu2 S-ZnS-SnS2 system. In: Journal of Alloys and Compounds 368(1-2) (2004), S.
135–143
[28] Weber, A.: Wachstum von Dünnschichten des Materialsystems Cu-Zn-Sn-S,
Friedrich-Alexander-Universität Erlangen-Nürnberg, Diss., 2009
[29] Schurr, R. ; Hölzing, A. ; Jost, S. ; Hock, R. ; Voß, T. ; Schulze, J. ; Kirbs, A. ;
Ennaoui, A. ; Lux-Steiner, M. ; Weber, A. ; Kötschau, I. ; Schock, H.-W.: The
crystallisation of Cu2 ZnSnS4 thin film solar cell absorbers from co-electroplated
Cu–Zn–Sn precursors. In: Thin Solid Films 517 (2009), S. 2465–2468
[30] Wagner, R. ; Wiemhöfer, H. D.: Hall effect and conductivity in thin films of low
temperature chalcocite Cu2 S at 20°C as a function of stoichiometry. In: Journal of
Physics and Chemistry of Solids 44 (1983), S. 801–805
[31] Nozaki, H. ; Shibata, K. ; Ohhashi, N.: Metallic hole conduction in CuS. In: Journal
of Solid State Chemistry 91 (1991), S. 306–311
[32] Shibata, T. ; Muranushi, Y. ; Miura, T. ; Kishi, T.: Photoconductive properties of
single-crystal 2H-SnS2 . In: Journal of Physics and Chemistry of Solids 51 (1990), S.
1297–1300
[33] Wu, C. ; Hu, Z. ; Wang, C. ; Sheng, H. ; Yang, J. ; Xie, Y.: Hexagonal Cu2 SnS3 with
metallic character: Another category of conducting sulfides. In: Applied Physics
Letters 91 (2007), S. 143104
[34] Yoo, H. ; Kim, J.: Comparative study of Cu2 ZnSnS4 film growth. In: Solar Energy
Materials and Solar Cells 95 (2010), S. 239–244
[35] Fernandes, P. A. ; Salomé, P. M. P. ; Cunha, A. F.: Precursors’ order effect on the
properties of sulfurized Cu2 ZnSnS4 thin films. In: Semiconductor Science and
Technology 24 (2009), Nr. 10, S. 105013
[36] Katagiri, H. ; Ishigaki, N. ; Ishida, T.: Characterization of Cu2 ZnSnS4 thin films
prepared by vapor phase sulfurization. In: 2nd world conference on photovoltaic solar
energy conversion Bd. 1, 1998, S. 640–643
[37] Katagiri, H. ; Jimbo, K. ; Yamada, S. ; Kamimura, T. ; Maw, W. S. ; Fukano, T. ;
Ito, T. ; Motohiro, T.: Enhanced Conversion Efficiencies of Cu2 ZnSnS4 -Based Thin
Film Solar Cells by Using Preferential Etching Technique. In: Applied Physics
Express 1 (2008), Nr. 4, 041201
[38] Araki, H. ; Mikaduki, A. ; Kubo, Y. ; Sato, T. ; Jimbo, K. ; Maw, W. S. ; Katagiri,
H.i ; Yamazaki, M. ; Oishi, K. ; Takeuchi, A.: Preparation of Cu2 ZnSnS4 thin films
by sulfurization of stacked metallic layers. In: Thin Solid Films 517 (2008), Nr. 4, S.
1457–1460
102
Bibliography
[39] Ennaoui, A. ; Lux-Steiner, M. ; Weber, A. ; Abou-Ras, D. ; Kötschau, I. ; Schock,
H.-W. ; Schurr, R. ; Hölzing, A. ; Jost, S. ; Hock, R. ; Voß, T. ; Schulze, J. ; Kirbs,
A.: Cu2 ZnSnS4 thin film solar cells from electroplated precursors: Novel low-cost
perspective. In: Thin Solid Films 517 (2009), Nr. 7, S. 2511 – 2514. – Thin Film
Chalogenide Photovoltaic Materials (EMRS, Symposium L). – ISSN 0040–6090
[40] Moritake, N. ; Fukui, Y. ; Oonuki, M. ; Tanaka, K. ; Uchiki, H.: Preparation of
Cu2 ZnSnS4 thin film solar cells under non-vacuum condition. In: Phys. Status Solidi
(c) 6 (2009), Nr. 5, S. 1233–1236. – ISSN 1610–1642
[41] Katagiri, H. ; Jimbo, K. ; Moriya, K. ; Tsuchida, K.: Solar cell without
environmental pollution by using CZTS thin film, 2003, S. 2874–2879
[42] Kobayashi, T. ; Jimbo, K. ; Tsuchida, K. ; Shinoda, S. ; Oyanagi, T. ; Katagiri, H.:
Investigation of Cu2 ZnSnS4 -Based Thin Film Solar Cells Using Abundant Materials.
In: Japanese Journal of Applied Physics 44 (2005), Nr. 1B, 783-787
[43] Scragg, J. J. ; Dale, P. J. ; Peter, L. M.: Towards sustainable materials for solar
energy conversion: Preparation and photoelectrochemical characterization of
Cu2 ZnSnS4 . In: Electrochemistry Communications 10 (2008), Nr. 4, S. 639–642
[44] Tanakaa, T. ; Yoshida, A. ; Saiki, D. ; Saito, K. ; Guo, Q. ; Nishio, M. ; Yamaguchi,
T.: Influence of composition ratio on properties of Cu2 ZnSnS4 thin films fabricated
by co-evaporation. In: Thin Solid Films 518 (2010), Nr. 21, S. S29–S33
[45] Tuttle, J. ; Contreras, M. ; Albin, D. ; Noufi, R.: Physical, chemical and structural
modifications to thin-film CuInSe2 -based photovoltaic devices. In: 22nd IEEE
Photovoltaic Specialists Conference (PVSC), 1991, S. 1062–1067
[46] Fernandes, P.A. ; Salomé, P.M.P. ; ; Cunha, A.F. da: Growth and Raman scattering
characterization of Cu2 ZnSnS4 thin films. In: Thin Solid Films 517 (2009), Nr. 7, S.
2519–2523
[47] Todorov, T. K. ; Reuter, K. B. ; Mitzi, D. B.: High-Efficiency Solar Cell with
Earth-Abundant Liquid-Processed Absorber. In: Advanced Materials 22 (2010), Nr.
20, S. E156–E159
[48] Ohring, M.: The Materials Science of Thin Films. Academic Press, 1992
[49] Kessler, J. ; Bodegård, M. ; Hedström, J. ; Stolt, L.: Baseline Cu(In,Ga)Se2 device
production: Control and statistical significance. In: Solar Energy Materials and
Solar Cells 67 (2001), S. 61–76
[50] Schroeder, D. K.: Semiconductor Material and Device Characterization. 2nd edition.
Wiley-Interscience, 1998
Bibliography
103
[51] Klenk, M. ; Schenker, O. ; Probst, U. ; Bucher, E.: X-ray fluorescence measurements
of thin film chalcopyrite solar cells. In: Solar Energy Materials and Solar Cells 58
(1999), S. 299–319
[52] Lundberg, O. ; Bodegård, M. ; Malmström, J. ; Stolt, L.: Influence of the
Cu(In,Ga)Se2 thickness and Ga grading on solar cell performance. In: Progress in
photovoltaics: Research and applications 11 (2003), S. 77–88
[53] Shafarman, W. N. ; Stolt, L.: Cu(InGa)Se2 Solar Cells. In: Luque, A. (Hrsg.) ;
Hegedus, S. (Hrsg.): Handbook of Photovoltaic Science and Engineering. Wiley, 2003,
Kapitel 13, S. 567–616
[54] Schubert, B. ; Marsen, B. ; Cinque, S. ; Unold, T. ; Klenk, R. ; Schorr, S. ; Schock,
H.-W.: Cu2 ZnSnS4 thin film solar cells by fast coevaporation. In: Prog. Photovolt:
Res. Appl. (2010). – ISSN 1099–159X
[55] Archer, M. D. (Hrsg.) ; Hill, R. (Hrsg.): Clean electricity from photovoltaics.
Imperial College Press, 2001
[56] Katagiri, H.: Cu2 ZnSnS4 thin film solar cells. In: Thin Solid Films 480-481 (2005),
426 - 432. – EMRS 2004. – ISSN 0040–6090
[57] Schock, H.-W.: High voltage solar cells by band gap and defect engineering in
Cu(In,Ga)(S,Se)2 chalcopyrite semiconductors. In: Proceedings of the 16th European
Photovoltaic Solar Energy Conference, 2000, S. 304–308
List of Abbreviations
CBD
CIGS
CIS
CTS
CZTS
EDS
KCN
Mo-SLG
QE
SEM
SLG
TPD
XPS
XRD
XRF
Chemical Bath Deposition
Cu(In,Ga)(S,Se)2
CuInS2
Cu2 SnS3
Cu2 ZnSnS4
Energy Dispersive x-ray Spectroscopy
Potassium Cyanide
Soda-Lime Glass coated with Molybdenum
Quantum Efficiency
Scanning Electron Microscopy
Soda-Lime Glass
Ternary Phase Diagram
X-ray Photoelectron Spectroscopy
X-Ray Diffraction
X-Ray Fluorescence
105
List of Figures
1.1
Rise in prices for resources . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
2.11
2.12
2.13
2.14
2.15
2.16
2.17
2.18
Available renewable energy . . . . . . . . . . . .
Black body AM0 AM1,5 . . . . . . . . . . . . . .
Silicon lattice . . . . . . . . . . . . . . . . . . . .
Energy bands . . . . . . . . . . . . . . . . . . . .
Band-bending . . . . . . . . . . . . . . . . . . . .
Space charge region . . . . . . . . . . . . . . . .
Band-bending applying voltage . . . . . . . . . .
Diode characteristic . . . . . . . . . . . . . . . .
Illuminated diode characteristic . . . . . . . . . .
Equivalent circuit of a solar cell . . . . . . . . . .
Impact of resistances on the diode characteristic
Possible efficiencies for certain bandgaps . . . . .
CIGS and CZTS cell cross section . . . . . . . .
Semiconductor compounds . . . . . . . . . . . . .
CZTS crystal structure . . . . . . . . . . . . . . .
Ternary phase diagram (secondary phases) . . . .
Ternary phase diagram regions . . . . . . . . . .
CZTS efficiency development . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
5
6
7
8
9
10
11
13
14
16
18
19
20
21
23
24
27
3.1
3.2
3.3
Spark tester . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electron-matter interaction . . . . . . . . . . . . . . . . . . . . . . . . . .
QE squared . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
36
41
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
Zn deposition rate for different powers . . .
XRD of Zn, Cu/Sn and Cu/Sn/Zn films . .
XRD of ZnS precursors . . . . . . . . . . .
Sn loss . . . . . . . . . . . . . . . . . . . . .
Surface features . . . . . . . . . . . . . . . .
Grain sizes for different TPD regions . . . .
Grain sizes for two Cu-poor samples . . . .
Comparison voids for different TPD regions
Layering: TPD . . . . . . . . . . . . . . . .
Layering: Cross section . . . . . . . . . . .
44
45
47
51
56
58
59
60
61
62
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
107
List of Figures
108
4.11
4.12
4.13
4.14
4.15
4.16
4.17
4.18
4.19
4.20
4.21
4.22
4.23
4.24
4.25
4.26
4.27
4.28
4.29
4.30
Layering: Top view . . . . . . . . . . . . .
Layering: Zoom in . . . . . . . . . . . . .
Layering: Comparison . . . . . . . . . . .
Impact of ZnS: General . . . . . . . . . .
Impact of ZnS: Matrix (cross sections) . .
Substrate comparison . . . . . . . . . . .
XRD SnS2-signal . . . . . . . . . . . . . .
XRD CuS-signal . . . . . . . . . . . . . .
TPD ordered by composition . . . . . . .
Influence of Zn content on efficiency . . .
Pseudo binary tie line . . . . . . . . . . .
Correlation grain size and efficiency . . .
Correlation voids and efficiency . . . . . .
Correlation compactness and efficiency . .
QE data of most samples . . . . . . . . .
QE comparison 3 samples . . . . . . . . .
QE comparison 3 samples: cross sections .
Band gap determination . . . . . . . . . .
IV measurements: good and bad samples
FF plotted against Zn content . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
63
63
64
65
67
70
72
72
76
76
77
80
81
82
84
86
87
89
91
91
A.1
A.2
A.3
A.4
A.5
Energy consumption . . . . . . . .
Thin film solar cells fraction . . . .
XPS metal precursor on Si and Mo
XPS ZnS precursor on Si and Mo .
XPS concerning MoS2 . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
111
111
113
114
115
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
List of Tables
2.1
Deposition methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
3.1
Substrate composition comparison . . . . . . . . . . . . . . . . . . . . . .
38
4.1
4.2
4.3
4.4
4.5
4.6
All sample data . . . . . . . . .
Reproducibility . . . . . . . . .
Voids table . . . . . . . . . . .
Sample matrix . . . . . . . . .
Average efficiency ZnS vs metal
Sample matrix efficiency . . . .
50
54
60
66
79
79
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A.1 Sample matrix XPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
109
A Appendix
A.1 Trends in photovoltaics
Figure A.1: Predicted increase of worldwide energy consumption. (Shell energy scenario
to 2050, http://www-static.shell.com/static/aboutshell/downloads/)
Figure A.2: Market share of thin film solar cells until 2008 (IEA study 2009 "Trends
in Photovoltaic Applications", from http://www.iea-pvps.org/products/download/rep1_
18.pdf).
111
A Appendix
112
A.2 XPS measurements
Precursors
On 4 precursors XPS was performed to reveal possible differences in composition and
gradients between metal and ZnS precursors as well as between films on Si and Mo
substrates, see Table A.1.
The compositional values from XPS differ strongly from those obtained by EDS. For
example, for the metallic precursor seen in Fig. A.3a), the XPS result shows about 60% Cu,
while EDS measurements give about 40%. Indeed is the EDS not calibrated either, but
confirmed by several measurements and calculations (see chapter 3.2.2.1). For this reason,
we do not use XPS measurements for the determination of absolute atomic percentages,
since XPS in general is less exact for absolute composition measurements. However, the
relative changes in composition can be used assuming all samples of the same material
and for different sputtering depths give the same error.
In the following we compare 4 samples: A is a metal precursor on Si, B the same on Mo,
C is a sulphur containing precursor on Si and finally D is the same on Mo. By this we can
get an overview of the homogeneity of the precursors, and also study possible differences
for different substrates.
In Fig. A.3a) we see sample A, a metal precursor on Si. Near the surface1 a strong
decrease of the Cu signal (almost down to 0) is accompanied by a similar strong increase
of the Sn-signal. The Zn signal shows a strong increase at the very surface, but reaches a
plateau very soon. These gradients indicate that the very surface is very Cu-poor and
Zn-rich. A bit deeper, but still near the surface, the film appears Cu-poor and Sn-rich.
Despite these gradients near the surface, the overall impression of the precursor is that
the composition is very uniform within the film.
Not visible in this graph is that the film was almost oxygen free, only the surface showed
a slight increase in signal intensity and indicates the presence of Zn- and Sn-oxides.
The next graph, Fig. A.3 b), shows a precursor from the same run as a but on molybdenum. Qualitatively – and to a certain extend also quantitatively – the picture shows
Substrate \Precursor type
Silicon
Molybdenum
Metallic
A
B
ZnS
C
D
Table A.1: Sample matrix used for XPS to study differences between metallic and ZnS
precursors as well as precursors on the substrates silicon and molybdenum.
1
Unfortunately, it was not possible to determine the sputter depth. Sometimes it was sputtered through
down to the substrate (which could be seen by an increase of the signal of the substrate material), but
otherwise expressions like "near the surface" or "in the bulk" have to be understood relatively to the
distance (indicated by the sputter time) that has been sputtered in the particular sample.
A.2 XPS measurements
113
Figure A.3: a) Metal precursor on silicon substrate (sample A). The film shows uniform
composition, only the surface has gradients (Cu-poor and Zn-rich). b) Metal precursor on
molybdenum substrate (sample B). The film shows very similar composition to B (uniform
composition besides at the surface).
the same result. Again the surface is Cu-poor and Zn- or Sn-rich, respectively; just the
gradients start probably a bit closer to the surface. Except of that, this comparison
indicates that precursors on Mo and on Si can be treated as equal, considering gradients
and composition.
Sample c can be seen in Fig. A.4 a). Here, Si is used as a substrate, but the precursor
is from the second sputter series, i.e. it contains sulphur. Again, apart from the surface,
the elements Zn, Sn and here also S, show good homogeneity. Only Cu seems to reveal
a slight decrease of some atomic percent towards the surface. At the surface, the Sn
content increases rapidly. However, the Cu-, Zn- and S-concentration decreases, but in a
minor degree than the Sn-increase. Hence, this sample has a Sn-rich surface and a quite
homogenous bulk.
The last picture in this consideration (Fig. A.4 b)) shows the data of sample D, produced
together with sample C but on Mo instead of Si. Note the different scale compared
to Fig. A.4 a) due to a spike in the data caused by a measurement error. The overall
impression of this sample is the same as for C: Zn, Sn and S lie on the same level at
around 20%, while the Cu-concentration lies around 45% and decreases slightly down to
roughly 40%. Again, the Sn content increases strongly very near the surface, connected
with a decrease of the other elements. This comparison indicates (like for the other two
samples), that precursors on Si and Mo can be treated as equal, both considering gradient
and composition.
Not yet discussed, as it cannot be seen from these pictures where only the sputtered
114
A Appendix
Figure A.4: a) ZnS precursor on silicon substrate (sample C). Except of the surface the
composition is constant within the film. For Cu content there might be a slight decrease
towards the surface, though. b)ZnS precursor on molybdenum substrate (sample D). The
film shows very similar results to C in a).
elements are shown, is the content of oxygen. Two ZnS samples were studied and showed
the same oxygen level. It turns out to be higher in the S-containing precursor through
the whole film. This indicates that the ZnS-containing precursors are less dense than the
metal one. That is not surprising, as Zn has an almost twice as high density than ZnS
(Zn: 7,1 g/cm3 , ZnS: 4,1 g/cm3 ; [23]). The higher porosity allows the oxygen to diffuse
into the film even after finalizing the deposition.
Another aspect that can be observed is that the metallic precursors show Sn- and
Zn-oxides at the surface, which can be seen from the shift of the elemental peaks. The
peak position in XPS can be shifted by changes in chemical bonding environment of the
atom. In this way compounds like in this case oxides can be determined.
The presence of oxides at the surface is not surprising as the samples have not been
stored in vacuum, but should not have any strong impact on the sulphurization.
A.2 XPS measurements
115
Molybdenum sulfide
Figure A.5: XPS data showing the increase of the Mo-signal (light blue) attended by a
decrease of the metal signals. The sulphur signal remains meanwhile almost the same and
decreases only sligthly. That indicates the presence of molybdenum sulfide.
Acknowledgments
First of all I want to thank Prof. Marika Edoff, Dr. Mikhail Fonin and Prof. Giso Hahn,
who offered me the possibility to do this thesis in Uppsala, for what I’m deeply grateful.
Furthermore I thank my supervisors Lotten Platzer–Björkman and Tomas Kubart for
many fruitful discussions and supporting my work a lot. I thank Lotten for doing XPS,
XRD and QE measurements, and Tomas for help with sputtering, EDS and searching for
blueberries.
I thank Jonathan Scragg for helping me with SEM and for several useful discussions about
CZTS and the ternary phase diagram.
Moreover I want to thank the whole group of the Ångström Solar Center for a nice year
in a pleasant working atmosphere and a lot of ’15 o’clock fikas’. In particular I want to
bring out the help of Uwe Zimmermann for always answering all questions covering the
range from physics to planting, and Sebastian Schleußner who helped me a lot with any
concern in the stressful final period of my writing.
Manuel Brendle is acknowledged gratefully for troubleshooting Latex and proof reading
my thesis.
Großer Dank gebührt vor allem meinen Eltern. Sie haben mich das ganze Studium über
unterstützt und bei jeder Entscheidung hinter mir gestanden, und mir damit meinen
Traum von Schweden erst möglich gemacht.
Und abschließend, aber bestimmt nicht als Letzter, möchte ich von ganzem Herzen Hanna
danken, die sich nicht nur zwei Tage und Nächte um die Ohren geschlagen hat, um Fehler
in der Diplomarbeit aufzudecken, sondern mir vor allem ganz viel Kraft und Halt in
schwierigen Momenten gegeben hat!
117
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement