You are on page 1of 108

Diss. ETH No.

15877

Low-cost

Cu(ln,Ga)Se2

absorbers from

nanosized precursor materials


for thin-film solar cells

A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH


for the

degree

of

DOCTOR OF SCIENCES

presented by
MARC-ROLAND KAELIN

Dipl. Ing.
born
citizen of

Mikrotechnik EPFL

on

July 16th,

1975

Ostermundigen BE, Switzerland

Accepted

on

the recommendation of

Prof. Dr. S.E. Pratsinis, examiner Prof. Dr. A.N. Tiwari, co-examiner
Dr. T.

Meyer,

co-examiner

Zurich, 2005

1I.

Seite Leer / i Blank leaf

Ill

Acknowledgements
First of all I would like to

express my

gratitude

to Prof.

Ayodhya

N. Tiwari and Prof.

Sotiris E. Pratsinis for their

motivation, advice and supervision of my PhD thesis.

I would also like to thank PD Dr. Hans

Zogg for giving

me

the

opportunity to work
and Andreas

in

his
I

laboratory.
a

Further I

acknowledge

the

help from

Dr.

Toby Meyer

Meyer.

profited

lot from the close collaboration.

This work would not have been

possible

without the contribution and

help

of my

colleagues,

former and

present

members of the Thin Film

Physics

group:

Dominik Rudmann, David Brmaud, Franz the thin-film solar cell makers,
Karim

Haug,

Alessandro Romeo, Derk Btzner,

Alchalabi, Dmitry Zimin, Klaus Kellermann, Martin Arnold, the infrared-sensor

team,
Fedor Kurdesau, Antonio F. da academic

Cunha,

Maxim

Ganchev, Gennady Khrypunov,

our

guests during my thesis work,


in his semesterwork in
our

Christoph Langhammer, having investigated electrospray


group,
Thomas

Kmpfer,

Michael

Leopold,

our

technical staff,
secretaries.

GabyStrahm,

Paulette Pfamatter, the

group's

Many

thanks go to the members of the

particle technology laboratory.

The fruitful

discussions with Olliver Wilhelm, Wendelin Stark, Lutz Mdler und Karsten
to many
new

Wegner

led

ideas.

am

very

grateful

for the XPS measurements and their

interpretation by

Dr. Roland

Hauert from the Swiss Federal Laboratories for Materials

Testing

and Research.

I would also like to thank the


Mr.

workshop

of the

physical department, design

Mr.

Brhwiler,

Nussbergerand
of

co-workers for their advice in construction


mechanical

and excellence in

manufacturing

complicated

parts.

IV

Further,
Marc

appreciated
and his wife

very much the company of my

good friends, especially sharing


a

Jantime

Lehky

Jaqueline

with which I have been

flat for

some

and

finally,

I would like to thank my

family for their support during the whole time.

The financial

support by

the Commission for

Technology

and Innovation

(KTI)

is

kindly acknowledged.

This thesis resulted from the work done in the frame of the

TopNano2i Programme (KTI Project 5491).

Table of Contents

Acknowledgements
Table of Contents

III V
IX XI

Zusammenfassung
Summary

11ntroduction
1.1

Development
1.1.1

of

photovoltaics

2
2

Historical

Background
silicon solar cells

i.i.2

Crystalline

(1st generation cells)

1.1.3 Thin-film solar cells 1.1.4 Future


i.2

(2nd generation cells)

3 4 5 7

prospects (3rd generation cells)


of the CIGS

Properties

Compound

1.3 The CIGS solar cell 1.31 Formation of the p-n


1.3.2

junction

7
8 9
10

Layer

structure of

CIGS solar cell

1.3.3 Band

diagram

and CIGS

layer grading

1.3.4 Electrical

properties

1.3.5 Sola reel I

efficiency deposition techniques

11

1.4 Conventional CIGS


1.5 Low cost CIGS
1

12

deposition techniques

13

51

Spray pyrolysis

14 15
^

1.5.2

Electrodeposition
coating
non-vacuum

1.5.3 Paste 1.6


1.7

Completely Objectives

processed

CIGS solar cells

16 17

of this work

Precursor material evaluation


2.1

19

Precursor Characterization
2.1.1 2.1.2

19
2i 22

Metals

Metal-oxides

2.1.3 Metal-selenides

24

VI

2.2

Chemical conversion in selenium vapor

25
25

2.2.1 2.2.2
2

Metals Metal-oxides

28
30

23 Metal-selenides

2.3 Conclusion

31

3 Precursor

deposition
methods

33

3.1

Deposition

33 33
34

3.1.1 Doctor-blade 3.1.2 Screen 3.1.3

printing

Electrospray

34

3.2 Evaluation 3.2.1 Mixed-metal precursor 3.2.2 Metal-oxide precursor 3.3 Conclusion

36 36
39 41

4Selenization
4.1 Chemical conversion
in vacuum

43 reactor

44 44
47

4.1.1 Conversion reaction 4.1.2 4.1.3

Impurities

Crystallization
in non-vacuum

48
reactors

4.2 Chemical conversion 4.2.1 Reactor 4.2.2 4.2.3 4

49 50
51

design

Impurity phases

Crystallization
non-vacuum

53
reactor

2.4Open

55 59

4.3 Conclusions

5 Solar cell devices from

nanoparticle

precursor

61

5.1 Device

processing deposition
of

61

5.1.1 Chemical bath

Cd(S,OH) layer

61

5.1.2 ZnO/ZnO:Al TCO


5.1 3 Cell

deposition

62
62

scribing

VII

5.2 Characterization

62 62

5.2.1 5.2.2

Contacting

Current-Voltage

characteristic

63
65

5.3 Conclusion

Inorganic

salt precursors and solar cells

66

6.1 Doctor-blade

(Paste coating)

66

6.1.1 Paste formulation 6.1.2 Paste

67
68

deposition

6.1.3 Preheating 6.1.4


6.1.5
Selenization of Precursor
CIGS

68 7 75

growth

mechanisms

6.1.6 Solar Cell Device characterization


6.2

76
79
82

Electrospray
Conclusions

6.3

7 Research recommendations

83

References

85

Abbreviations

93

Curriculum vitae

95

Refereed articles

96

VIII

Seite Leer /
Blank leaf

IX

Zusammenfassung
Im

Rahmen

dieser

Arbeit

wurden

kostengnstige

Solarzellen

basierend

auf

polykristallinem Culn^Ga^e^ (CIGS

oder CIS fr

x=o) entwickelt. Dabei

wurde fr die

Abscheidung

der CIGS Schicht auf die konventionellen, teuren Hochvakuum-Techniken

verzichtet und stattdessen einfache Druck- und

Sprayverfahren

bei
in

Atmosphrendruck
Form eines NanoIm

eingesetzt.
Pulvers
zweiten

In einem ersten Schritt wird ein Vorlufermaterial

mit

der

gewnschten
wird die

metallischen
in

Zusammensetzung aufgetragen.
kontrollierter

Schritt

Vorluferschicht

Atmosphre

zu

CIGS Gase

konvertiert. Dabei werden fr die chemische benutzt und die

Umwandlung
dass

keine

gefhrlichen

Temperaturen

so

tief

gehalten,

kostengnstige

Substrate wie bis


zu

Fensterglas

und Metallfolien

eingesetzt

werden knnen.

Wirkungsgrade

6.7%

konnten mit den entwickelten Techniken


Das erste

aufgezeigt werden.
in das Gebiet der Photovoltaik. In einem

Kapitel gibt
berblick

eine

Einfhrung
die

historischen

wird

Entwicklung
Die

der

ersten
von

Solarzelle

bis

zu

den

Dnnschichtsolarzellen effizientesten

aufgezeigt.

Eigenschaften
liefert,

CIGS, das bis anhin die

Dnnschichtzellen

werden

besprochen.

Obwohl

Dnnschichtzellen im

Vergleich

zu

Zellen

aus

kristallinem Silizium
eine weitere

bereits grosses

Kostenreduktionspotenzial
Produktionskosten

haben,
damit

wird
die

noch

Senkung

der

ntig sein,

Photovoltaik

gegenber

konventionellen

Energiequellen konkurrenzfhig

wird. Ein

berblick ber den aktuellen Stand der


und die

kostengnstigsten CIGS-Abscheidungstechniken
Arbeit werden
In

Zielsetzung

der

vorliegenden

gezeigt. Entwicklung
von

Kapitel

zwei bis vier wird die

CIGS Schichten mit

pulverfrmigem

Vorlufermaterial beschrieben:
Im zweiten

Kapitel

werden verschiedene Nano-Pulver


in

(reine Metalle, Legierungen,

Metallselenide und

Metalloxide)

Bezug

auf

Partikelgrsse, Form, Agglomeration,


Eignung
bilden
zum

Zusammensetzung
Sintern

und Kristallinitt charakterisiert. Zustzlich wird ihre

und

zur

Konvertierung

in

Selendampf

untersucht:

Zwar

alle

Vorlufermaterialien die CIGS-Phase aus,

zeigen aber auch Verunreinigungen durch (nur


Cu und In; Ga wurde wegen dem

ln203

und

Cu2.xSe.

Rein metallisches Nano-Pulver

tiefen

Schmelzpunkt
sich

nicht
den

verwendet)
getesteten

bildet

zwar

bis

zu

um-grosse
um

Agglomerate
dichte

eignet

jedoch

von

Materialien

am

besten,

grosskrnige,

CIS-Schichten herzustellen.

Im dritten eine

Kapitel

werden

typische Billigdruck-Verfahren
ihrer

wie

Rakeln, Siebdruck und


dnner

Spraymethode bezglich
Herstellung
der

Eignung

zur

Abscheidung

Schichten

untersucht. Die
werden

Druckpasten

und die verwendeten Druckmethoden

beschrieben.

Spraybeschichtungen

zeigen

die

beste

Homogenitt

der

Schichtdicke.
In

Kapitel

vier wird der Einfluss verschiedener

Selenisierungsbedingungen
und

auf die
von

chemische

Umwandlung

des

Vorlufermaterials

Eliminierung

Verunreinigungen untersucht.

Mit drei verschiedenen

Reaktortypen (Vakuumkammer,

geschlossene Ouarzkammer,

offenes

Ouarzrohr)

wird der

Hintergrunddruck (Stickstoff)
um

und Selen-Partialdruck ber mehrere

Grssenordnungen verndert, gezeigt,

die

optimalen ntig

Umwandlungsbedingungen
ist,
um

zu

finden. Es wird

dass ein hoher Selendruck

Oxid-Phasen umzuwandeln. fnf werden die

In

Kapitel

Fertigungsschritte

von

der CIGS-Schicht bis

zur

fertigen

Solarzelle beschrieben. Die Solarzellen werden anhand der charakterisiert. Ein maximaler Vorlufermaterial diskutiert. erreicht.

Strom-Spannungs-Kennlinie
mit dem

Wirkungsgrad
Grnde fr

von

1.5% wurde

pulverfrmigen
werden

Die

die

niedrigen Wirkungsgrade

Kapitel
basierend

sechs auf

befasst

sich

mit

einem

neuartigen CIGS-Abscheidungsverfahren
Die

Metallsalz-Vorlufermaterial.

Druckpaste,

bestehend
in

aus

den

gelsten
mittels

Salzen und einem Rakelmethode


CIGS

organischen

Bindematerial
in

(Ethylcellulose)
der
ca.

Alkohol, wird
und
im

aufgetragen,

anschliessend

Luft

getrocknet

Selendampf zu
auf einer

umgewandelt.

Dabei entsteht eine

600
zu

nm

dicke CIGS Schicht


mit dieser

amorphen

Kohlenstoffschicht.

Wirkungsgrade

bis

6.7 % wurden
es

Methode erreicht. Noch hhere CIGS-Schichtdicke


zu

Wirkungsgrade
die

werden erwartet, falls


zu

gelingt,
und

die
ein

erhhen,

Kohlenstoffschicht
der CIGS-Schicht
zu

eliminieren

optimiertes Zusammensetzungsprofil

erhalten.

XI

Summary
Within the scope of this
or

work, low-cost solar cells

based
were

on

polycrystalline
The CIGS with

Culn1.)CGaxSe2 (called CIGS,

CIS for

x=o)

absorber

layers

developed.
was

layer, being

the most

complex

and

important part step


a

of the solar cell,


a

deposited

simple

non-vacuum

equipment.

In the first

thin film of

nano-powder

precursor
The

material is

deposited,

which contains the metals in the desired

stoichiometry.
in
a

second step converts the precursor material into the CIGS

compound

controlled

atmosphere.
and reaction lime
were

Conversion conditions

were

restricted to the

use

of non-hazardous gases
as

temperatures compatible

with low-cost substrate materials such

soda-

glass

or

metal foils. Efficiencies up to

6.7% with techniques scalable

to

large-area

obtained in this thesis.

The first

chapter gives

an

introduction into the field of

photovoltaics.

In

brief

historical summary the cells is discussed. The

development

from the first solar cell to current thin-film solar


on

chapter then

focuses

the

properties of CIGS, which yielded

the

highest

efficiencies in thin-film solar cells to date. Even if thin-film solar modules,


to wafer based

compared

crystalline

silicon modules, have the

potential

for lower with

production costs,

further cost reduction is necessary for


sources.

photovoltaics to compete

conventional energy
is reviewed and

The current status of low-cost CIGS


are

deposition techniques

targets for the present work

elaborated.

Chapters
based
on

two to four describe the research and

development

work of CIGS

layers

particulate

precursor materials:

Chapter
and

two describes the precursor materials

(pure metals, alloys,

metal-selenides
and

metal-oxides)

with

respect

to

particle

size,

shape, agglomeration, composition

crystalline phase

state. After

reacting

the precursor materials in selenium vapor, the CIGS


are

formation and

sintering properties

evaluated. All precursor materials form and

the CIGS

phase

but also material

impurity phases, mainly ln203


and
Ga
~i

Cu2.xSe form depending


precursors

on

the

precursor

selenization
was

process.

Metal

(nanopowder
even

containing

Cu and In

only;
of

not used due to its low


are

melting point),

though
CIS

forming agglomerates compound Chapter


formation three

urn,

most suited for dense and

large-grained

during

selenization.

compares

important

non-vacuum

deposition methods, namely

doctor-blade, screen-printing and electro-spray and investigates their suitability for

XII

CIGS precursor thin-film the used

deposition.
Films

Precursor

paste preparation, deposition


from spray

as

well

as

set-up

is described.

resulting

deposition
if

show

superior
fixed

homogeneity
substrate.

in thickness but lack of

compositional homogeneity

sprayed

on a

Chapter four shows the


conversion and

influence of different selenization processes elimination. box and

on

the chemical

impurity phase

Testing
open

three different reactor

types

(vacuum chamber, closed quartz


pressure

an

quartz tube) the background


varied

(nitrogen)
to find

and selenium

partial

pressure

were

by

several orders of

magnitude

optimal

conversion conditions.

High

selenium pressure is found to

be necessary for oxide

impurity elimination. processing from


the CIGS

Chapter
cell

five describes the cell

layer

to the finished solar

by depositing the

n-CdS

part

of the

heterojunction

and the ZnO/ZnO:AI

transparent
the

conducting

electrical contact. Characterization of the solar cells


curve

by recording

current-voltage

show maximum efficiencies of

-1.5% for the particulate precursor

materials. Reasons for the low

efficiency are discussed.


novel
CIGS

Chapter

six

describes

deposition

process

based

on

metal

salt

precursors. The precursor material

paste, prepared by
in

dissolution of the salts and the

organic

binder

(ethyl cellulose)

alcohol,

is

deposited by

doctor-blade
600
on nm

method,

preheated

in air and selenized in selenium vapor. An


on

approximately

thick CIGS

layer is formed

top

of

an

amorphous

carbon

layer.
the

Solar cells based

this

approach
largeand

have reached efficiencies up to


area

6.7%. Considering

simplicity, non-toxicity

and

scalability

this is

good efficiency
improvements

with further

potential

for

improvement
the CIGS

cost-reduction. Further

can

be reached

by increasing
and

layer

thickness, while reducing the amorphous carbon


elemental

layer

optimization

of the

grading profile

of the CIGS

layer.

11ntroduction
Photovoltaic

systems,

also

called

solar

cells, present

an

interesting

and

clean

alternative to fossil and nuclear fuels for energy which is

generation. They work

with

sunlight,
into

freely

available for the next

10

billion years, and convert it

directly

electrical energy. Solar cell with sales

industry

continues to grow at to reach US

an

astonishing by
2010.

pace

over

35% annually

easily expected
on

8 billion

With further

cost-cutting
a

innovations

the horizon and the effects of economies of scale, there may be

bright

future for solar cells. reached with

Large

scale

production

with drastic cost reductions could be

simple,
costs

fast and affordable

non-vacuum
as

deposition equipment.
or

Direct

manufacturing

given by typical techniques


are

printing

spraying

are

estimated
costs

to be in the 50-100

$/m2 range, and


based

well below current industrial

production

(200 $/m2)
The

with

vacuum

technology.
reasonable solar cell efficiencies is

development required.

of such
For

techniques yielding

therefore

targeted

energy costs of 0.8

$/WP (Watt peak),

technique

capable

of

producing

6% efficient modules at 50 $/m2 does

already

meet the aim, and

high efficiency always desired)

modules
as seen

(tj>12%)

are

not

mandatory (although higher

efficiencies

are

in Table

i.i.

Table

1.1

Module cost in

$/WP calculated

for direct

manufacturing

costs between 50 and

200

$/m2 and

produced

module efficiencies in the range of 6 to 15%

(Zweibel, 1999)
Manufactu

'ing

costs

Module

Efficiency

$200/m2
3-3

$i5o/m2
2-5 1.9 1-5 1.25
1

$ioo/m2
17

$ 50/171*
0.8

6% 8%
10%

2-5
2

i-35
1

0.63
0.5
0.42

12%

1-7 i-3

0.83 0.67

15%

0.33

i.i

Development
Historical

of

photovoltaics

i.i.i

Background
the basis of

The

photoelectric effect, being


Alexander Edmund

photovoltaics (PV),

was
an

first observed in

1839 by

Becquerel,

while

experimenting with

electrolytic
a

cell

made up of two metal electrodes

(Becquerel, 1839).
was

His observation that

voltage
there
a

developed
were

when

light

fell

on

the electrodes
nor

considered very

interesting, yet
In the

neither

practical applications
effect
was

any theoretical
a

explanations.

1870's

similar

photovoltaic

first studied in

solid, namely selenium (Smith, 1873),

and allowed to prove that without heat


or

solid material

can

change light directly


These

into

electricity

moving parts (Adams

and

Day, 1877).
cell

findings
cell
a

led to the
was

development

of the first

photoelectric (or solar)


a

by

Fritts in

1883. The

made

of selenium wafers coated with

nearly transparent layer of gold, yielding revolutionary explanation


for the

conversion

efficiency
was

of -1%.

Only

in 1905,

photoelectric

effect

given by

Einstein
as

(Einstein, 1905)

for which he received the Nobel Prize in 1921.


were

Other materials, selenium and

thin films of copper oxide,


a

found to show the

same
one

effects

as

in 1914 the existence of in


PV

barrier to current flow in

direction metal-

(rectifying mechanism)
semiconductor barriers

cells

was

reported,
1938 by

where
W.

the

theory

for

was

developed

later in

Schottky

and F.N. Mott.

i.i.2

Crystalline
1930's

silicon solar cells

(ist generation cells)


understanding

In the
were

selenium and the copper oxide cell still had efficiencies less than 1% and

mainly employed

for

light-meters

in

photography. Despite
did
not

better

of the

physics involved, photovoltaic systems


on

improve significantly until


into the silicon that Pearson

research started to focus allowed the formation of


at Bell
a

silicon.

By introducing impurities
Calvin

p-n

junction, Daryl Chapin,

Fuller, and Gerald

Laboratories
et

produced

silicon solar cells with efficiencies up to 6% in 1954

(Chapin
grew,

al., 1954). This

was a

major breakthrough

and demand for the

technology
During

especially from

the US

military,

which used solar cells to power satellites.

the energy crisis in the

early 1970's public

sectors became interested in PV energy in

generation
over

for terrestrial

applications. Improvements

technology

of

manufacturing

three decades led to efficiencies up to


et

24.7% with single crystal silicon wafers

(Green

al., 2004). Researchers began

to

use

polycrystalline

silicon in order to

provide

lower cost alternative, but solar cells still remained too silicon is
a

expensive

for

widespread
with

adoption. Crystalline
extreme

relatively

poor

light absorbing

material

(Fig. i.i)

sensibility

on

impurities.

Therefore silicon cells need

typically

200-300UITI

thick absorbers of
time

solar-grade
PV

silicon wafers.

They
to

have

projected

energy

payback

(the

time

system

has to

operate

recover

the energy needed for its

production)
and

of 2-3 years and

an

assumed life

expectance of 20-30 years. Single crystal


current industrial PV market.

polycrystalline

silicon cells share about

90% of the

1.1.3 Thin-film solar cells


Efforts
were

(2nd generation cells)


high absorption
coefficients
in the form of thin films
can

initiated to find alternative materials with

(Fig. i.i)
based

that could be

processed

(00 urn).
on

In contrast to wafer

technology,

thin-film absorber materials

be grown

different substrate

materials

including

flexible foils and allow monolithic module


on a common

production (series
processing.

connection of many solar cells

substrate)
as

and roll-to-roll
as one

Projected

energy

payback time for thin-film

cells is

short

year.

Figure

1.1

Absorption

coefficients for

common

solar absorber materials

as a

function of incident

photon energy (Lewerenz

and

Jungblut, 1995).
thin-film solar cell materials
a

Requirements for optimal

are:

Semiconductor with

bandgapof 1.4 eV, (Direct


band

however 1.1-1.5 eV

are

also suitable

High absorption
Consisting

coefficient

gap)

of abundant and non-toxic materials

Suitable for Should

large

area

thin-film

deposition

with

simple

methods

yield long term

stable and

high efficiency cells

A material

fulfilling

all these criteria has not been discovered

yet. However, high

efficiency

thin-film cells have been silicon

developed

with different material classes

(Figure

1.2): amorphous
indium

(a-Si:H),
and

cadmium-telluride

(CdTe),

gallium-arsenide (GaAs),

phosphide (InP)

copper-indium-gallium-selenide-sulfide (Cu(ln,Ga)(Se2,S2)).
silicon

Hydrogenated amorphous
were

(a-Si:H) and polycrystalline CuJ/CdS


severe

solar cells

first

developed. Cu2S/CdS
and the

cells showed
was

stability problems

due to

high

Cu

diffusivity
was more

development

stopped by

the

early i98o's. Amorphous

silicon

successful and devices became

commercially available, although

these cells

have low conversion efficiencies


To increase the

(-6% stabilized).
cells, other materials
were

efficiency

of thin-film solar

investigated

and research efforts in

polycrystalline

CdTe and

Cu(ln,Ga)(Se2,S2),
(Wu
et

called CIGS, based

solar cells led to 16.5% and 19.3%

efficiency

cells

al.,

2002; Ramanathan et

al.,

2004).

1.1.4 Future
Even

prospects (3rd generation cells)


second

though high efficiency


market,
new

generation

cells

are on

the

spot

to enter the

solar cell

concepts of

future PV devices

are

already theoretically being


limited to
a

elaborated. As conventional
maximum conversion

single

band gap solar cells


new

are

theoretical

efficiency of -33%,
limit

concepts

to reach conversion efficiencies

beyond
Such

the

Shockley-Queisser
include

(Shockley

and Queisser,

1961)

are

being pursued.

concepts
level

tandem

systems,

fluorescent

mirrors,

impact ionization,

impurity
stacked

systems,

intermediate bands and

hot electron collection.


are

Except stage

for and

multi-junction

solar cells other

concepts

at

an

early

theoretical

suitable materials and structures still have to be found.

1.2

Properties
a

of the CIGS

compound
which
and
can

CIGS is

l-lll-VI

compound material,

be derived

by splitting
and

the bivalent

component
VI

of ll-VI materials into

a mono-

trivalent

component

doubling

the

component (Fig. 1.2).

The

specifically high absorption


well

coefficient

(Fig. i.i)

makes the
an

tetragonal chalcopyrite compounds


absorption
2 urn are

suited for thin-film solar cells. With

coefficient of the order of io5cm_1, CIGS

compound layer

thicknesses of

1-

sufficient for

complete absorption

of

light.

IV Si.Ge

Cd,Zn

S,Se,Te

In, Ca, AI As,P

^
I VL
In, Ca, AI S, Se,Te
a

Cu.Ag.Au

Figure 1.2

Some

important semiconducting

material classes. CIGS is

l-lll-V, compound derived

from ll-VI class materials.

The

phase composition

of the

ternary

CIS

compound
and
/

is

mostly

described

by

the

pseudo-binary phase diagram


region
are

of the

binary Cu2Se

ln2Se3 phase (Fig. 1.3).

In the

of the stoichiometric

composition (50% Cu2Se

50%

ln2Se3) four different phases


the

observed: The

a-phase (chalcopyrite, CulnSe2, Fig. 1.4a),


the

-phase (ordered
CulnSeJ
and

defect

compound, Culn3Se5),

8-phase (sphalerite,

cation disordered

Cu2_xSe.
room-

The existence range of the

a-phase

is

limited to

24-24.5%

Cu

content at

temperature,
content in

not

even

including

the stoichiometric

composition

of

25%
At

Cu. The Cu

high efficiency
of

solar cells is

typically

in the range of

22-24%.
At
room

deposition

temperatures
the the

~500C this composition yields single a-phase.


shows
a

temperature

equilibrium phase diagram partial replacement


can

phase separation

for these

compositions. By
of sodium the

of In

by

Ga and the addition of small

quantities

single phase region

be extended

(Herberholz

et

al., 1999).

1200

Liquid

600

500
0 10 20 30 40 X 50 60 70 80 90 100

Cu3Se

[mol. %]

In2Se3

Figure 1.3

Cu2Se and ln2Se3 pseudo-binary phase diagram (Stanbery, 2002). The chalcopyrite o>

phase

shows

only little tolerance to stoichiometric distortions


copper rich

at

thermodynamic
Cu2Se phase, slightly

equilibrium. Slightly
indium rich

compositions form
an

in addition the

compositions form

in addition

ordered defect

compound (-phase).

I-III-VI2 compounds containing Cu, In, Ga,

Se and S

give great flexibility

in band gap

tailoring.

The obtainable band gaps range from


a

1.04eV for CulnSe2

to 2.43 eV for

CuGaS2 while the lattice constant


as

changes

from

5.78 for CulnSe2

to 5.35

forCuGaS2

shown in

Fig. 1.4b.

'

,,..

'

'

2.422-

-CuGaSj

2.0-

%
S
D

1-816-

CuGaSe2

1-41.2-

CulnSj
v

1.0-

^'^P
CulnSe2
1 1

5 3

5.4

5.5

56
a

S.7

5.8

b) a)
Unit cell of

Lattice constant

()
a

chalcopyrite

structure

(a-phase) with

lattice constants

and c, named
are

after the mineral

Chalcopyrite (CuFeS2).

Cation and Anion atom

arrangements

the

same as

in the Zincblende mineral diamond

(ZnS) and the basic

arrangement corresponds to the


CIGS

crystal

structure,

b) Band gap and lattice constants range for

compounds.

1.3 The CIGS solar cell


1.3.1 Formation of the p-n
In addition to
an

junction
which creates free electrons and

efficient
a

light absorbing material,

holes when illuminated, distinct electric with


an

charge separation

mechanism is

required
is

to form two

poles.

In semiconductor

materials, charge separation


up in the

implemented
of the p and
carriers in the

electric field

(called

built-in

field) building

junction

area

n-semiconductor
near n-

(Fig. 1.5).

The electric field is due to the

depletion

of

charge

the junction and the

resulting

space

charge region (SCR), oppositely charged


are

and

p-semiconductor.

Most absorber materials used for solar cells

p-type.

This is hole
to

due to the fact that the electron

mobility

in semiconductors is

higher

than

mobility

and efficient

transport

of

light generated minority


CIGS absorber

carriers is

important

avoid recombination.

Polycrystalline

layers

are

grown

slightly

copper

deficient and

are

p-type

conductors with carrier concentrations of the order of io1B to

1017 cm"3. In order to extend the SCR


are

deep

into the absorber

layer, higher doping


the

levels

required

for the

n-type

semiconductor.

Additionally,

n-type

semiconductor

must be

nearly transparent to

allow the

light to

reach the absorber

layer.

Figure

1.5

Schematic

drawing

of the p-n

junction of

solar cell

(in substrate
to
a

or

back-wall

configuration) showing light-generated


field in the space extend

carrier

separation due

built-in electrical
allows the SCR to
remains
near

charge region (SCR).

An

asymmetrical doping profile


A

deeply

into the

p-type absorber layer.

quasi-neutral region (QNR)

the back contact.

The

efficiency
of

of CIGS solar cells

can

be

improved by applying
p-absorber

thin

n-type

buffer

layers

Cd(S,OH)

and i-ZnO between the

and the n+-front contact TCO

(transparent

conducting

oxide).

The

mechanisms

responsible

for

the

cell

enhancements

are

not

fully
as

understood

and

they probably

serve

more

than

one

purpose.

They

may act

diffusion reduce

barriers between the TCO and absorber

layer,
of

provide

interface

passivation,
at the

optical reflectivity (although absorbing


or

part

the solar

spectrum

same

time)
a

establish

an

inverted

region

in the absorber

and thus lead to the formation of

homojunction.

A detailed

explanation

of the

physics

of solar cells

can

be found elsewhere

(Nelson, 2003)

1.3.2

Layer

structure of

CIGS solar cell


a

The conventional CIGS solar cell consists of substrate material first

stack of 5
A

layers deposited

on

(substrate configuration, Fig. 1.6).


Thereafter the CIGS

molybdenum

back contact

is

deposited by dc-sputtering.

layer (p-type,
a

1-2 urn

thickness)

is

grown
-50

typically by

vacuum

co-evaporation
chemical bath

and covered with

Cd(S,OH) layer (n-type,


is

nm

thickness) using

deposition

process. The front contact


nm

realized

by sputtering

thin

intrinsic ZnO

layer (-50

thickness)
urn

followed

by

thicker ZnO:AI
increase

transparent conducting
a

oxide
is

layer (TCO, deposited


an

0.3-1

thickness).

In order to

current collection
can

Ni/AI

grid

on

the TCO. Further

efficiency
the

enhancement cells/modules

be obtained

by depositing

antireflection

layer. Eventually,

are

protected against

environmental

impact by encapsulation.

ZnO:AI
ZnO

(300nm)

CdS

(50nm) (50nm)

CIGS

Molybdenum
Substrate Figure
1.6
SEM

image showing the


vacuum

cross-section of

conventional CIGS cell. The CIGS

layer was

deposited by

co-evaporation

of the elements.

1.3.3 Band
The band

diagram

and CIGS

layer grading
Figure
1.7a. The front contact buffer

diagram

of the CIGS solar cell is shown in material. It is followed

(ZnO:AI)

is

heavily doped n-type


charge region.

by the

layers

which
on

are

within the space

The band gap of the CIGS

compound depends

its

Ga/(ln+Ga)

ratio which lifts the level of the conduction band minima with

higher

Ga

contents while the valence band remains

nearly

unaffected

(Wei, 1995).
on

Therefore the

spike height produced


ratio of the CIGS

between

CdS(-OH)

and the CIGS

depends
(1.4)

the

Ga/(ln+Ga)
CIGS

layer.

Some of the

deposition

processes

lead to

graded
a

layers

where Ga accumulation at the back contact is observed. This leads to

higher

band gap towards the CIGS/Mo interface, interface where the recombination is

keeping photoelectrons
to be

away from this

supposed

high.

A detailed

investigation

of

grading effects

is

given by Lundberg (Lundberg

et

al., 2003).

n
ZnO Al
mm.

n
j
nm

p
CIGS

ZnO CdS

300

[MM (llll 50 50

1.2 eV

EF
-2.5 eV

layer
a

material

nr

type
108
-

NA,ND
-

nti-reflection

MgF2
ZnO:AI
i-ZnO

137
1.9 19

3.3 eV

front contact buffer buffer absorber back contact

3.6
3-3 2-5
1 2

n+
n

1020

1016
1015

Cd(S,OH)
CIGS

24
2.9
-

P
-

1016
-

Mo

a).
Figure 1.7 a)
Band

b) diagram
of the p-n

junction of

CIGS solar cell with

normal Ga

grading

increasing the band-gap b)


Material

at the back contact. in the conventional CIGS solar cell. The

properties of the layers used (nr), band-gap

refractive index levels

(Eg),

conduction

type and acceptor and donor doping

(NA, ND)

are

given for the optically and electrically important layers.

10

1.3-4 Electrical
When there is characteristics

properties
illumination diode. The

no a

(dark)

the solar

cell, being

p-n

junction,
in

has the

same

as

rectifying

behavior is shown
an

Fig i.8,

where the

characteristic dark current

density (JDark)

is indicated. For

ideal diode the dark current

density

is

given by
(
qv
e
"

JaV)=h
where J0 is
a

-1

(i.i)
current

constant

(reverse
in K.

saturation

density), KB

is

Boltzmann's

constant and T the

temperature

Dark

-O

( "mr'm)

dark

charge

b) a)
Solar cell current

density-voltage

characteristic under dark and illuminated

(light)
circuit

conditions. The solar cell parameters Jsc

(short-circuit current density), Voc (open


can

voltage)

and the maximum power

point {Vm,Jm)

be determined.

b) Equivalent

electrical circuit of ideal solar cell

Illumination of the cell generates


characteristic
can

photocurrent.
the
sum

The overall current

density-voltage

be

approximated

as

of the short-circuit

photocurrent (Jsc)

and the dark current

(photocurrent

is

positive by convention):

qV
e
*

J(V) Jx-Jl
=

(1-2)

Real solar cells may suffer from power


as

dissipation

in

parasitic

series resistances such

electrical contacts

or

in

parallel

resistances created

by

shorts at

pinholes

or

at the

sides of the devices. The influence

can

be modeled

by two

resistances in the

equivalent

11

circuit

(Fig. 1.9a)
The

and leads to
current

deteriorated

current-voltage
as:

characteristic

as

seen

in

Fig. 1.9b.

according

density

is written

q(V-R~*,,-J)
!nsiJLJ-

J{V)=Jsc-Jo

A-K.-T

-1

V-R RShunt

(1.3)

^
a)
Figure
1.9

V
vshunt

^charge

a) Equivalent circuit of solar cell

with

parasitic parallel (shunt) and


of
a

series resistances,
a

b)

Current

density-voltage characteristic

real solar cell

showing

reduced
resistances.

maximum power

point (VmJ Jj due to the influence of series and parallel

1.3.5 Solar cell


The solar cell

efficiency
defined

efficiency,

by

the ratio of the electrical

output

power to the
a

input power
Standard

of the solar radiation, is the most


measurements
use an

important characteristic of
factor 1.5

solar cell. which

efficiency
to

air

mass

(AM1.5) spectrum,
with
an

corresponds

standard

terrestrial

conditions

(ASTM 1999)
power

integrated
solar cell is

irradiance power

density

of ioooW/m2

(Ps).

The

output

density

of

given by

P
out

-J

V
m

(14)

This value

corresponds
(FF)

to the grey

rectangular

area

in the J-V characteristic

(Fig.
It

i.8a).

The fill factor

is often used to characterize the


areas

quality of a

solar cell 1.8a


:

junction.

expresses the ratio of the

of the two

rectangles

drawn in

Figure

FF

J
=

-V

(1-5)

12

Therewith, output power and efficiency

can

be

expressed

as:

'out

^j J

.v
SC

.ff

vOC

rr

(1.6)

P '

77:

out

J J

-V

SC

YOC

FF rr

Ps

(1-7)

1.4 Conventional CIGS

deposition techniques
the best control of
CIGS

Sophisticated
composition

vacuum

co-evaporation techniques (Fig. 1.9) give


the of

and

compositional grading through

films.
a

Maximum

cell

efficiencies up to 19.3 % have been obtained process


in

by the

use

three-stage co-evaporation

(Ramanathan

et

al., 2004). The three-stage process has not been implemented


module

industrial

large-area

production. However, production


1.10 as

other
as

co-evaporation process
Solar,

variants have been used for module

large

1.2x0.6 m2 at Wrih

Germany (Powalla

et

al., 2003, Figure

b).

r-^Jlongauige
Substrate Holder
+

Heating

r i

Flux

Flux

me; measurement

ld

Substrats

1G~7mbar

Turbomolecular
pump

MH
Se Ga Cu

"NaF

Figure

1.10

High a)

vacuum

CIGS

deposition systems have high complexity.


on an

Schematic

drawing of laboratory-scale equipment based


for each element.
at Wrth

MBE

system with

point effusion cells

b) Industrial in-line deposition system

Solar, Germany.

Shell Solar and Showa Shell

preferred

selenization processes

(also

called 1wo

stage

processes)
of
a

for industrial module


a

production.

These processes consist of the

deposition

precursor material and

subsequent

thermal

annealing

in

controlled reactive

13

atmosphere. Typically,

the

sputter deposition

of

metal-alloy

precursor is followed
or

by
a

selenization and sulfurization in reactive

H25e

and

H2 ambient
thermal

by evaporation

of

selenium

layer

on

the metal precursor followed


band gap

by rapid

annealing (Palm etal.,


is
more

2003).

A controlled
use

grading (grading

of Ga/ln
to obtain
on

ratio)
a

difficult to
at the

achieve, but the

of

post-sulfurization step

higher band-gap

CIGS surface led to module efficiencies above 14 %

30x30 cm2 substrates with


et

good

process

yield, showing

the mature status of this solar modules initial and


can

technology (Kushiya
produced by

al., 2003).

Although,

low-cost CIGS
a

be

such vacuum-based

deposition techniques,
vacuum

high

investment is

required
will

for the

sophisticated
on

deposition equipment
volume and
vacuum
-

module

costs

depend strongly mandatory.


1.11.

the

production

of course-also

high

efficiencies
is

are

An overview

of the different

methods for CIGS

growth

given

in

Figure

Method
M;llii!P!iP'"j

Deposition
Various thermal

evaporation stages with defined metal flux ratios and increasing


while

Co-evaporation
*..,

substrate

temperatures (400C-600C).

maintaining

an

excess

selenium

background pressure Evaporation


or

m-

Selenization

Sputtering of precursor layers


temperatures (<200C):
without selenium

Selenization reaction at
substrate

a&

at low substrate

high

(two-stag process)

Cu, In, Ga stacked elemental layers, alloys


or

temperature (450-600C)

compounds with

in Se/S,

or

H2Se/S

gas ambient

alloyed

or

stacked metal

layers

polycrystalline

CIGS

layer

Substrate

Figure 1.11

Vacuum methods for CIGS

growth.

1.5 Low cost CIGS


CIGS

deposition techniques
challenging step High
for low cost cell

layer deposition

is the most

processing,

since

it is the most 100% are

complex layer

of the cell.

material utilization efficiencies close to

important
In

in order to

keep

material costs and

processing

waste

as

low

as

possible.

general,

these low-cost

techniques

follow the idea of the

sequential
at low

techniques:
temperature

A precursor material and

having

defined metal ratios is


reaction
or

deposited
at

followed
aim is to

by

chemical

annealing step

higher
and

temperature. The

develop

fast processes with

technologically simple

14

reliable low-cost the


non-vacuum

equipment. Undesired
environment have to

contamination of the precursor materials

by

be avoided

by choosing
evolved:
In

suitable precursor
one

materials and
materials

additives. Two different

approaches

the

precursor

decompose during deposition

on

the substrate and form the CIGS

compound

directly (spray pyrolysis, electrodeposition approach,


the precursor material is

of

compound layers)
as a

whereas in the second that is

deposited

thin

layer

subjected

to

chemical treatment and selenization thereafter metal

(paste coating, electrodeposition


is
an

of the be

layers).

Even if CIGS formation

during deposition
includes

appealing approach,
and may

as-produced amorphous quality


is

material
or

is

of

poor

quality,

impurity phases

microcrystalline

due to the low


with
a

deposition temperatures (<4ooC).


processing step,

The

generally improved

second

where the material is

annealed at

highertemperature.

1.5.1

Spray pyrolysis
pyrolysis technique chalcopyrites,
is
one

The spray

of the
are

best-investigated
few
et

non-vacuum

deposition
The

processes for

but there

only

reports

on

cell

performance.
a

efficiency
area

record dates back to

1989 (Duchemin

al., 1989) and is of 5 % for

small The

cell. The spray

technique

is very well suited for uniform

large-area coatings.
precursors,

process consists of the


metal chlorides and
a

decomposition

and reaction of

premixed

generally
or

chalcogen compound (typically N,N-dimethyl

selenourea

thiourea),
Bates,

on a

heated substrate

(300-400^) (Pamplin
technique
are

and

Feigelson,

1979; Brown and oxide and and

1990).

Drawbacks of this

impurity phases (binary


and

chalcogenide phases),
the

traces from the reaction size obtained.

by-products (chlorine annealing

carbon)

relatively

small

grain

However,

treatments in controlled

atmosphere
al.,

can

reduce the

impurity concentration

and enhance

crystallinity (Krunks

et

2002; Marsillac et

al., 2003).
on

Alternative precursor solutions based been

organic single
of

source

precursors have also precursors is to

investigated.

The idea behind the

development

single

source

lower the CIGS formation


since the

temperature
on a

and to avoid the formation of

impurity phases,
Banger
et

synthesis

is controlled

molecular level and

(Hirpo
form
are

et

al.,

1993;

al.,
at

2002).

These

organic
below

precursors

decompose

the
not

CIGS

compound

temperatures
and
no

4ooC. However,

such precursors
so

commercially

available

cell efficiencies have been

reported

far.

15

1.5.2

Electrodeposition
has been

Electrodeposition resulting
(Guimard
in small
et

successfully

used to
a

deposit quaternary

CIGS

films,

area

efficiencies above io% after

subsequent
a

thermal

annealing

al., 2003). By applying additional


an

In and Ga and
was

high-temperature

annealing

treatment in vacuum,

efficiency

of 15.4 %

obtained

(Bhattacharya

et

al., 2000). The stability of the chemical solution, large-area uniformity and high

deposition

rates

are

still

challenge. Electrodeposition
a

of

an

elemental-layer

stack

subsequently subjected

to

selenization treatment led to


et

7-10% efficient devices

(Kapur et al., 1987; Kampmann

al., 2003).

1.5.3 Paste

coating
are screen

Typical paste coating techniques


curtain

printing,
can

doctor-blade

coating

and

coating.

Paste

coating
is

is

fast process and

be

applied

in continuous roll-to-

roll

deposition.

The

paste

prepared
and
a

with the precursor materials binder used


can
as a

(e.g. nanoparticles)
(US
Pat No.

in the desired

stoichiometry

liquid

transfer media

5,985,691,

Basol B. M., Nov

1999).

Paste

rheology

be tailored with material load and

additives, which affect the final film thickness


atomization

and

homogeneity.

In contrast to
are

liquid

(spraying),
allows
a

where valuable very efficient


et

portions

of the precursor material

wasted,

paste coating
densities

use

of the material and exhibits

higher packing

(Eberspacher

al., 2003). Material loss

minimization is very desirable for

precursors
the

containing expensive
layer
is

elements like indium and and converted


to

gallium.
CIGS

After
in

deposition,
controlled

precursor

sintered

the

layer

atmosphere.
After
some
a

preliminary
powdered

work
CIS

on

high temperature sintering (78oC)


et

of CulnSe2
a

(CIS)
to

pellets

from

ingot (Gombia

al., 1983), it
of

was

quite

challenge
et

produce homogenous
Vervaet et

thin

layers by sintering

ground
was

material

(Arita

al., 1988;

al., 1989). The motivation for the approach

the successful

development
of CIS, the
size

of

screen

printed

and sintered CdTe/CdS solar cells. However, in the used to formulate


screen

case

precursor

powders

printable pastes

had too

large particle
even

(5-50 urn)

and did not lead to

compact single-phase thin layers.

But

chemically

synthesized
mill

colloidal CIS did not lead to

pore-free

CIS

layers (Schulz

et

al., 1998). BallCIS

grinding

of soft and reactive metals resulted in

mechanically alloyed

powder,

which

requires higher sintering temperatures than pure metal powder.

A solution to

i6

this

problem

was

presented by using copper-indium-alloyed powder produced by


The

gas

atomization

(Basol, 2000).
to

alloy

is

more

brittle than pure In and therefore allows

ball

milling

produce

finer

powder.

CIS Devices

produced by

selenization in diluted

H2Se gas yielded efficiencies


in the

in the range of lo-n

%,

although

some

porosity remained

film, which

was

attributed to the

randomly shaped

precursor

particles. Layers

of

metal oxide

nanoparticles uniformly

allowed
round

an

apparently higher packing density

due to smaller

particle

size and

shape. Furthermore, gallium


to
a

could be introduced into


and devices with

the precursor

material, which led

higher open-circuit voltage


et

maximum efficiencies of 11.7 % US Pat No.

(Eberspacher
Oct

al., 2002) and 13.6 % (Kapur et al., 2003,

6,127,202 Kapur V.K.,


a

2000).
Ga

Still

higher
is

efficiencies above 15 % may be achieved

reached

if

better
near

control

of the

grading

(e.g. optimum

Ga

concentration

the CdS-CIGS interface for

higher Voc).

The chemical conversion of


reactions. Such
a

the stable metal-oxides to CIGS process includes


a

requires well optimized


in diluted

two

stage

reduction

step
an

hydrogen prior to

selenization with diluted

H2Se (Kapur
diluted to
a

et

al, 2003). It is

ongoing

discussion if the

highly toxic H2Se

gas,

though

small concentration in in

N2, orthe less hazardous

selenium vapour should be

used for selenization concluded that


at

industrial

production.
more

An

experimental comparison study

H2Se selenization gives


selenization
restrictions

uniform of

layers

with

better control,

especially

moderate
are some

temperatures
on

400-500^ (Alberts, 2002).


in many

However, there
and

the

use

of

H2Se

European

countries

costly security

measures are

required

in industrial

production. Figure

1.12

gives

an

overview of different non-vacuum CIGS

deposition

processes.

1.6

Completely
all the
can

non-vacuum

processed
of the cell

CIGS solar cells


back contact,

Although
contact)

component layers

(metal
more

buffer, front
is

be grown with

non-vacuum

methods,

research and

development

necessary to find suitable low-cost


non-vacuum

deposition

methods and conditions. Some issues


were

on

TCO

layers

from sintered

nanoparticles

presented, high

but the

sintering
in

temperature required
efficient solar cells

for low sheet resistances is still too


et

for the

application

(Ederth

al., 2003; Eberspacher

et

al., 2002).

17

Precursor (low

Deposition
temperature)
Materials

Selenization/Annealing/Recrystallization
(high
substrate

substrate

temperature 450-600X)
Selenization
',: ^''.'."i'.'^w1: ^""wwi"

Methods

Pretreatment

Electrodeposition

Cu, In, Ga stacked


elemental

Recrystallization
or

layers,

Selenization
in inert
or

alloys

or com

pounds with Se

Se/S, HjSe/S gas ambient

5#riy pyrolysis
:lIlf'tlfP^!.
>'

Cu, In, Ga, Se/S

Recrystallizatictn

compounds
Cu. In, Ga, 0

Tiftj

MM^vmim'Binder elimination
Selenization reaction Se/S, H2Se/S gas ambient

Paste coating
1
""'

"""""*'

compounds
stacked

H2

reduction to metal alloy

Small

grained, alloyed

or

layer

with defined metal ratios

polycrystalline CIGS layer

Substrate

Figure 1.12

Overview of

non-vacuum

CIGS

deposition

methods

1.7
The

Objectives
objective
method

of this work

of this thesis is to

develop

solar cells with

low-cost CIGS absorber

deposition

using simple agents.

non-vacuum

equipment

and

avoiding
a

the

use

of

hazardous chemical precursor

The most
on a

promising concept

is to start with

nanosized
a

powder that
The

is coated

substrate and

subsequently
allows

converted into
a

CIGS

layer (Fig. 1.13).

separation

into two

independent steps

better control of the

layer properties:

Deposition

of

precursor film
structure to

with

good

control

of the

resulting

film

composition (and

some

extent)

Conversion of the precursor reaction

layer

to the l-lll-VI

compound layer by

suitable

pathway to

control the CIGS

phase

formation and

grain growth

i8

i)

Precursor

2)

Chemical
Conversion

Deposition

Figure

1.13

Two-step process used

in this thesis: precursor

deposition

and chemical conversion

The

experimental

work for the thesis concentrated

on

the

following aspects:

1) Finding

suitable precursor material which fulfills the

following requirements:
impurity free
CIGS

Suitable chemical

properties for the

conversion into

Nano-sized material to allow the

deposition

of

homogenous

precursor films

with

thickness below 5

urn.

Economical

production

in

large quantities

2) Evaluating

different
a

non-vacuum

deposition techniques
of the substrate.

for

specific

precursor

material, allowing

reliable and fast

coating

3)

Construction

and

evaluation

of different

reactor

designs

for the

chemical

conversion of the precursor to

high quality CIGS layer.

4) Development

of solar cells with

non-vacuum

processed

CIGS absorber

layers.

19

Precursor material evaluation


In
non-vacuum

deposition methods,
can

the constituents of CIGS


as

(Copper, Indium,
it is done in

Gallium and
vacuum

Selenium)

not be reacted in their pure form

high

co-evaporation deposition.
the

As CIGS has to be formed at

temperatures
to

above

3ooC,

atmospheric
of

environment
into the

would

instantly
a

lead

oxidation
is

and

incorporation
which first

impurities
a

layer.

Therefore

two-stage process

preferred,
(at
room

deposits

film

containing

the constituents in the correct ratio

temperature) and, second,


high temperature.
Because of the

converts that precursor

layer

into the desired

compound

at

required (<<
i

final

layer

thickness of

only

1-2 urn,

precursor material with


a

small

particle

size
nm

urn)

has to be

employed.

Particles with

diameter smaller than

about 50-100

show different

properties compared
for many
new

to the bulk material. The

high
and

surface to volume ratio is

responsible

aspects, e.g.

lower

melting

sintering temperature, magnetic


and

increased

catalytic activity,
as

modified

thermo-physical, electric,

and mechanical

properties

plasticity

and hardness. Lower the used two

melting points

high reactivity

will facilitate
are

layer sintering by
as

step process (1.7).

Therefore

nanoparticles

well suited

precursor materials.

2.1

Precursor characterization
can

Nanoparticles
flame aerosol

be

produced by

different

physical

or

chemical methods like e.g.


ballor

or

plasma
etc.

processes, of the

co-precipitation, sol-gel, high energy


methods
are

cryogenic milling,

Some

specialized

to

several

kinds

of

nanoparticles.

While

milling

methods work well with brittle materials


as

they

can

not be
co-

used for ductile

materials such
can

indium. On the other hand, the


a

vacuum

precipitation

method

be

applied

for

large

range of materials.

20

Four classes of materials

were

chosen for

investigation:

Table

2.1

Investigated

precursor materials.

Precursor Material

Supplier

Production Method

Metal
Cu and In

nanoparticles

Aid rich Co.


EPFL

gas

phase

condensation

CulnGa

alloy

laser

spark ablation

Metal-Oxide
Cu/ln/Ga Oxide

nanoparticles

NEI

(Nanoparticle

gas

phase

condensation/

Enterprises Inc.)
Selenide precursor
Mixed metal selenide Salt precursor
SOLARONIX

oxidation

wet chemical

nanoparticles
Alfa Aesar
and

synthesis

Inorganic salts (chlorides

nitrates)

The criteria for evaluation of the

resulting layer were:

uniformity

of

deposition

(smooth

and

crack-free

morphology,

homogenous thickness)

purity of

CIS/CIGS

phase

formation

(no binary

selenide

and

oxide

impurities)

CIS/CIGS

grain

size after selenization

( large grains

of-i urn, dense

layer)

Porosity (elimination

of pores

by sintering

of the precursor

layer)

The

deposition pastes

were

all

prepared

as

suspensions

of the

nanoparticles

in

alcohol
The
are

(pentanol).

The

pastes

were

deposited

with the doctor-blade

technique ( 3.1.1).
thesis

deposition
described

and conversion of the salt precursor


in

layers developed during this

separately

Chapter

6.

2.1.1

Metals
metal precursors,
were

The

copper
as

(purity >99-9%)

and

indium with
a

(purity >99.g%)
particle
size of

nanoparticles,
-loonm.

obtained

nanosized activated

powder
have
a

As indicated

by the

vendor

(Aldrich Co.), they

high

defect structure and processes. The

internal energy which should facilitate chemical reaction and

sintering

powders

are

shipped

in inert hexanes.

Figure

2.1

SEM

images

of precursor

powder, (a)

Cu and

(b)

In

nanoparticles

Fig.

2.1

shows SEM

micrographs

of the Cu and In form

powder.
of

The

particle

size is in

range of ioo-20onm, but the


of

particles

agglomerates

i-2um.

The XRD

patterns
that the

mixed-powder deposition (2.2.1)


in ambient at
room

show Cu, In and leads to

in703 peaks indicating partial

deposition

atmosphere already

oxidation of the indium

nanoparticles

temperature.

Gallium, having

melting point
a

of

297C,

is not available

as

metallic nanosized

precursor material. If and Cu. Therefore metals in


an
a

pure metallic precursor is needed, it has to be

alloyed

with In of the
et

Cu-ln-Ga

ingot, prepared by melting appropriate quantities


was

evacuated

quartz tube,

pulverized by

laser

spark

ablation

(Houriet

al., 1999).

Figure

2.2

shows the Cu-ln-Ga precursor material. The

particle

size is in the

range of ioo-30onm.

Figure

2.2

SEM

image of ingot

Cu-ln-Ga

nanoparticle precursor produced by laser spark ablation of

Cu-

ln-Ga

2.1.2

Metal-oxides
of oxide precursors is motivated
metal-oxide

The

use

by

the

high flexibility
of several

and

good

control in

(mixed-)
processes.

nanoparticles
oxide

production
from

available

production
Inc.
were

The
in
a

metal vapor

precursors

Nanoparticle Enterprises
are

produced
desired

phase

condensation method. The metals

vaporized

in the

stoichiometry.

This vapor is

rapidly

cooled in the presence of oxygen,

causing

nucleation and condensation of blade


SEM

copper-indium-gallium-oxide particles.
mixed and ball milled in

Before doctor2.3 shows


a

deposition,

the

powder

was

pentanol. Fig.

picture

of this

powder.

Figure

2.3

SEM

image

of

Cu,ln,Ga mixed-oxide nanopowder

23

The

powder

from NEI

was

ordered in three different chemical


was

composition

ratios

Cu/(ln+Ga) (table 2.2)

where the In/Ga ratio

fixed to 4:1.

Table

2.2

Investigated

oxide precursor materials.

Powder; Cu/ln/Ga ratio


NEI
1:

Group
3 1.05 0.9

lll/l ratio

O.33/O.8/O.2 O.95/O.8/O.2

group III rich

NEI

2:

slightly group
group I rich

III rich

NEI 3: 1.1/0.8/0.2

X-ray

diffraction

(XRD)

revealed the main constituent


can

phases

of these

powders (Fig.

2.4).

The very

pronounced ln203 peaks


there

be detected in the XRD

pattern of the oxides

precursor.

Additionally,

are some

minor

peaks

attributable to other oxide


20

phases,

which have their most intense reflections in the 30-479;

32-34

region (Cu2ln20;,
of
an a

JCPDS file

CulnGa207, 38-840; CulnGa04, 38-839).


even

The

occurrence

very

pronounced

ln203 peak
precursor
and

in the

case

of the copper rich precursor may be


a

indication that the

powder mainly

is

mixture of

amorphous

or

microcrystalline

copper oxide

polycrystalline

indium oxide.

IrijOj

35
2 Theta

40

60

(degree)

Figure

2.4
on

XRD

patterns of mixed

oxide precursor

layers

with different Cu/ln/Ga ratios. Cards:

(Joint

Committee 30-479; Mo,

Powder Diffraction Standards, ASTM,

Philadelphia, PA, 1967,

ln203, 6-416; CUjln205,

42-1120)

24

2.1.3 Metal-selenides
The formation of CIGS is often

reported

to be

reaction of

binary

selenides

(Cu2Se

-ElnSe;
suited

Orbey
as

et

al., 1998). Therefore, these compounds have the potential

to be well
reaction

precursor materials. A

ternary selenide precursor

was

synthesized by

(2.1)

at ambient

temperature.
H20

2Cu(SOj + 2lnCI3 + 4SeC(NH2)2


The obtained

->

CulnSe2i

+ 2

H2(SOj+C H2N2 + 3Cl2


composition although
The

(2.1)

powder
were

did

not

yield

stoichiometric

the

reaction chemicals

mixed in stoichiometric

quantities.
are

powder

is indium rich

and deficient in selenium. Less indium rich precursors


Cu/ln ratio

obtained

by changing
The

the

of

the

reaction

chemicals

(Table

2.3a

and
are

2.3b).

composition

measurements
as

by energy-dispersive X-ray analysis (EDX)


porous films
was
were

however not very accurate


into the solution
a

the

deposited

only

1-2 urn

thin.

By adding GaCI3
of the

Cu-ln-Ga-Se

powder

obtained. The XRD


to

pattern

produced powder

does not

show distinct indicates


an

peaks corresponding
state.

crystalline phases (Figure 2.13)

and therefore

amorphous

Table 2.3a
Cu/ln ratio

Composition of CIS precursors obtained after


Cu

reaction 2.1.

chemicals
1 1.2

(at %)
32-5 30.5 33-5

In

(at %)
42

Cu/ln

Se(at%)
256 37

0.77 0.94
1.1

32.5 30.5

14

36

Table

2.3b.

Composition of CIGS precursors


Cu

as

measured with EDX.

Cu/(ln+Ga)
chemicals
1.2

(at %)
29.5

In

(at %)
29.5

Cu/(ln+Ga)
0.88

Ga

(at %)
4

Se

(at %)
37

The obtained

powder

was

highly aggregated
an

and therefore

ball-milling

was

used to
a

disperse

the brittle

powder

in

alcoholic

suspension paste. Figure 2.5(a) shows

Cu-

In-Se precursor

powder

obtained
are

by the

chemical reaction

(2.1).

After 1.5 h of ball

milling
some

the

big

1-3

urn

agglomerates
are

reduced to submicron

particles, although

still

bigger particles

present (figure 2.5(b)).

25

Figure

2.5

SEM

images

of CIS precursor

powder

obtained
can

by

wet chemical

synthesis (a). The


aggregates by

powder
ball

consists of loose

aggregated particles (i-3pm) which

be broken into smaller

milling (b).

2.2

Chemical conversion in selenium vapor suitability


of the precursor material, the doctor-bladed precursor
in selenium vapor. The used

To evaluate the

layers

were

annealed at of
a

high temperature (500-6ooC)


zone

reactor consisted

two-temperature

semi-closed
one

quartz

box in
to

vacuum

chamber

(Chapter 3).

The

sample

is heated in

temperature

zone

450-550X
the second

while the selenium

partial-pressure

and selenium-feed is controlled

by

temperature
To
were

zone.

Prior to reaction the chamber is

pumped

down toio-5 mbar.

investigate

the

degree

of conversion of the precursor,

X-ray

diffraction

patterns

recorded of the precursor

layer

before and after selenization. The


a

composition
metal ratio

was

measured

by

EDX before and after selenization to determine

possible

distortion.

2.2.1

Metals
doctor-blade of

The

deposited

mixed

metal

precursor
was

layer
a

was

selenized
of

at

temperature
which

50oCfor
to
a

25 minutes. The selenium

hold at

temperature

?50C

corresponds
of the

selenium

partial

pressure of 102 mbar.

Figure

2.6 shows the

morphology

layer

before and after selenization.

26

a)^l^^
Figure
2.6 SEM

b)

images showing the morphology of a metal precursor- (a) and selenized layer (b).

The selenized

layer

shows

grains

up to

i-2um

in diameter.

X-ray

diffraction confirms

the conversion from Cu and In to the CIS

compound (figure 2.7).

35
2 Theta

40

[degree]

Figure

2.7

XRD

pattern of precursor and

selenized

layer. (Joint

Committee

on

Powder Diffraction
42-

Standards, ASTM, Philadelphia, PA, 1967, Cards: CulnSej, 40-1487; ln203, 6-416; Cu2XSe, 6-680; Mo,

1120)

The precursor

layer

is not

only composed

of pure In and Cu but also of


some

ln203.

The that

selenized

layer

contains in addition to

CulnSe2 still

ln?03 crystals indicating

the reaction is not

complete.

As

second

impurity phase, Cu?.xSe

is formed in small to

quantity. The average composition during selenization changed from indium rich

slightly

copper rich

as seen

in Table 2,4. The

compositions

in different

places

varied and
of the

therefore the

significance
is

of these measurements is limited. Better


a

mixing

precursor

layer

required

for

homogenous composition.

27

Table 2.4

Composition

before and after selenization


Cu

(in

atomic

percentage).
Cu/ln ratio Se

(at %)
48
275

In

(at%)
52

(at %)

Precursor

0.92
1.02

Selenized

45-5

The

layers

from the

alloyed

precursor

produced by

laser

spark

ablation did not show

pronounced grain growth (fig. 2.8). Moreover, they

cracked

during

selenization.

Figure

2.8

SEM

images showing the morphology


at two different

of selenized laserablated

metal-alloy

precursor

layer

magnifications.

The XRD

patterns

revealed

an

incomplete

conversion to CIGS. Where the precursor

layer only

shows

some

Cu and In

peaks,
and

the selenized

layer

contains the CIGS

phase

as

main constituent

along

with

ln203

Cu2.xSe impurity phases (Figure 2.9).

(112)

selenized
precurosr

(204)

CIGS

(220)

phase Impurity phase


(116)
(312)

|><Brty<|l^'*V*l*tfW^W>frA^WWfrl^MH
a DC x

Cu, Se
Cu
Cu

P
1

I
'

C^

"

'

ir-T,

,,

,-

-111i
3540 3540

~t-

15

20

25

30

45

50

55

60

2 Theta

(degree)
and selenized

Figure

2.9

XRD

patterns

of laserablated

alloyed metal precursor

layer.

28

2.2.2

Metal-oxides
was

Selenization

performed

at

500C
to

for
a

duration of 45 minutes. The selenium selenium pressure of 102 mbar The

temperature
selenized

was

250C, corresponding

partial
and

layer

shows very limited gram

growth
in

sintering (Figure
parts (Figure

210

a) During

the

selenization reaction the

layer

is

cracked

different

210

b),

where the

crack width scales with the


of the alcohol

layer

thickness Care had to be taken that the


to

evaporation

proceeded slowly,

prevent micro-crack formation


in

in

the precursor could not be

layer However, completely

the formation of micro-cracks Different

the precursor
were

layer

excluded

deposition

methods

tested for this precursor

(Chapter 3)

Figure

2.10

SEM

image

of

morphology

of selenized metal oxide precursor

layer

at two different

magnifications.

(112)

selenized

precursor
(204)

(220)
3 TO

CIGS

phase Impurity phase


(116) (312)

<0

a on
x

f-f&M

r-i

,|n

15

20

25

30

35
2 Theta

40

45

50

55

60

(degree)

Figure

2 11

XRD

patterns of metal-oxide precursor (NEI 3) layer before and after selenization.

29

The selenization

reaction

was

investigated by

XRD before and after selenization

(Figure 2.n).
oxide

This precursor showed the best conversion results. Whereas the


in the precursor is

ternary
is still

phase (Cu2ln205) present


in the selenized

converted, the ln?03 phase

present

layer.
for
a

Another selenized

prerequisite
2.5,

precursor material is

controllable metal ratio in the

layer (Table

2.6).

Table 2.5

Metal ratios of

oxide-precursor layers
Elemental

before selenization.
measured

Powder

type
Cu

:omposition

by

EDX

inat% and

and Cu/ln/Ga ratio

(elemental fractions)
In 66 47.5 34 Ga

NEM NEI 2 NEI 3

(0.33/0.8/0.2) (0.95/0.8/0.2)
(1.1/0.8/0.2)

23

(o-33)
(0.95) (1.1)

(o-95)
(0.98)

(0.16) (0.13)

46
53

6.5
13

(0.71)

(0.27)

Table 2.6

Metal ratios of selenized

layers

obtained from oxide precursors mentioned in table 2.5.

Elemental
Powder

type
Cu

composition measured by

EDX

in at% and In

and Cu/ln/Ga ratio

(elemental fractions)
Ga Se

NE11 NEI
2

(0.33/0.8/0.2) (0.95/0.8/0.2)
(1.1/0.8/0.2)

18.9
34.5
32.3

(0.33)
(0.95) (1.1)

58.1
32.1 29.9

(1) (0.88) (1)

8.7
5.4
4.9

(0.15) (0.15) (0.16)

14.3 28 33

(0.25) (0.77) (1.12)

NEI 3

Taking

into account is

measurement

error

of

5%,

it

can

be concluded that the metal

composition
copper rich
content.
In

generally

retained

during
a

selenization of oxide precursors.

Only

for the

composition there may be


all three
cases

slight tendency

for

an

increase in the indium

the selenium content is below the stoichiometric CIGS

composition, indicating
via
a

that the reaction is not

completed,

since the reaction

proceeds

replacement

of the oxygen
2.11 was

by

selenium.

The data of

Fig.

obtained for the

NEI3 powder which has

copper rich

composition.

This

composition showed

the most intense CIGS

peaks

of the three oxide This fact

precursors and had the

highest
more

amount of

incorporated

selenium

(table 2.6).

suggests

that copper is

easily

converted than indium and

gallium.

3o

2.2.3 Metal-selenides
The selenide precursor

layers

were

selenized at
to
a

500T

for 15 minutes. The selenium

temperature

was

250C,

which

corresponds

partial

selenium pressure of io-2 mbar.

Figure

2.12

shows the selenized

layer morphology. During

the selenization

step

the

layer

cracked and

grain

size remained in the sub-micron range. Therefore

only

very limited

grain growth

occurred in the selenized

layer compared

to the precursor

(figure 2.5).

Figure

2.12

SEM

image showing the morphology of grains.

selenized CIS precursor

showing large

cracks

and small

The

composition

of the precursor and the selenized

layer

is

given

in table 2.7. The

copper to indium ratio is

slightly

increased

during

selenization but the

change

is not

significant, given

the

precision

of the measurements. The low selenium content in the

precursor material

was

increased but is still below the values for stoichiometric CIS be concluded that the CIS formation reaction is not

(-50%). Therefore,

it

can

complete.

Table 2.7

Composition

of metal selenide precursor- and selenized Cu


In

layer (in at%).


Se

Cu/ln ratio

Precursor

32.5 26
30.5

077

25.5 43.5

Selenized

085

Fig.
also

2.13 shows the and

phases

formed upon selenization. Beside the desired CIS


are

phase

ln703

Cu2.xSe impurity phases


vacuum

present.
can

Since the introduction of oxygen


must

during
be

selenization in the
in
an

chamber

be

excluded, the oxygen

already

present

amorphous

or

microcrystalline phase

in the precursor material.

31

I
(112)

precursor selenized precursor (204)


(220)

i/lo

CIS

phase Impurity phase


(116)
(312)

I
Cu2.S<

I
o

I
ln203

I Cu,
Se

in203

ln203

c
40
45
50 55

15

20

25

30

35

60

2 Theta

(degree)
and after selenization.

Figure 2.13

XRD

pattern of selenide precursor layer before

2.3 Conclusion
The results of this

chapter

show that all the tested precursor materials

(pure metals,
a

alloyed metals,
reaction

metal

selenides, metal oxides) form the CIS/CIGS phase by


can

simple phase

step

in selenium vapor. The metal ratios it is

change during the

CIS/CIGS

formation, but
precursor

possible

to control

the

composition by choosing appropriate


of dense

concentrations.

However,

sintering

homogenous films, grain


not achieved. The main
are

growth,

and

complete

elimination of

impurity phases

were

impurity phases forming during the


and the copper selenide

reaction in selenium vapor

the indium oxide


For
a

phases,

which

always

occurred

together.

complete

selenization reaction the indium oxide impurity

phase

needs to be converted to indium

selenides which would react with the copper selenide to form the CIS/CIGS

phase.

Cracking

of the

layers during the heating layers

is another

problem encountered. Except for


the selenization
a

the pure metal precursor, all the

cracked

during

step.

One

reason

for this may be

seen

in the fact that the


is

layer undergoes

threefold volume

expansion
Metal-

during

CIGS

formation, which
a

filling

out the pores in the precursor

layer.

oxides and selenides have crack formation.

much lower volume

expansion and may lead

to island and

Cracking

is

absolutely
because

detrimental for the

subsequent deposition

of

front contact and buffer cracks. Table 2.8 shows


a

layer,

they

would short-circuit the cell

through

the

summary of the results.

32

Table 2.8

Layer properties resulting from

selenization

of

various

precursor

materials

in

selenium vapou r

(+ satisfactory; 0 medium; poor).


-

Nanoparticle
Precursor

CIS

phase

No

impurity

Crack-free

Low

formation
+

phases
0

Large grain
+

size

porosity

layers
+

Mixed metal

Alloyed

metal

Metal-oxide
Metal-selenide

The mixed metal precursors

obviously provide

the best

properties

of the selenized introduction

layer and
of

will therefore have the main focus in the


into this precursor material remains
an

following chapters. The


issue.

gallium

Therefore, the other precursor

materials have not been

completely

abandoned at this

point. Cracking

of the

layers

may be linked with the


an

deposition

methods and may therefore be eliminated


Grain

by using
are

appropriate

method

(Chapter 3).

growth

and

impurity phase elimination

very much

dependent on

the conditions of the conversion reaction

(Chapter 4).

33

3 Precursor

deposition
chapter
is to

The aim of the work in this


vacuum

investigate

and evaluate different

non-

deposition

methods for

nanoparticle

precursor materials.

3.1

Deposition methods
are

The tested methods


Svec et
Even

doctor-blade

(also
et

known

as

cup

coating

or

tape casting,

al., 2002),

screen

printing (Jabbour
are

al., 2001) and electrospray (Siefert, 1984).

though
on

these

techniques

readily
given

used to

deposit thick

films

(>

10

urn), only

few

reports

paste

formulations

are

in the literature. In

general,

it is

agreed

that the

paste viscosity
for
screen

should be low for spray


3.1 shows
a

liquids, higher

for doctor-blade and

even

higher

printing. Figure

schematic illustration of the three methods.

a) Figure
3.1
Common

b)
non-vacuum

c) deposition
methods: doctor-blade

(a), screen-printing (b),

spraying (c).

3.1.1 Doctor-blade
Doctor-blade is
is
a

one

of the

simplest

methods for film

deposition. Everything required

blade and

mechanism that fixes the distance between blade and substrate. The
be realized with
a

fixed distance

can

spacer that is

applied
urn

on

both sides of the


were

substrate.

In

this work

stripes

of adhesive

tape (-55 particles

thick)

used. The

deposition suspensions containing

the precursor
Suitable

were

prepared

with alcohol

(1,2-propanediol
doctor-blade

or

i,2-propylene glycol).
was

viscosity

of the precursor

paste for

deposition

obtained with 5 ml alcohol per gram of precursor

powder.

34

3-1.2 Screen
Screen

printing
is

printing
et

widely

used for ceramic

layer deposition
et

in

dye

sensitized solar

cells

(Spth
a

al., 2003) and fuel-cells (Weber


material

al., 1996). This technique consists in

printing

high-viscosity

through

mesh that is in contact with the substrate


as

at the moment of

paste deposition (fig. 3.1b). The mesh properties such

mesh size

and mesh thickness will define the


Screen

deposited layer

thickness for

given paste rheology.


as

printing

has been

widely

used in thick film

applications

such

printed

circuit

board

(PCB)

metallization and color


a

printing

where

usually layers

of several

100 urn are

deposited. By

careful control of the

paste rheology, especially thixotropy and viscosity,


with submicron thickness. A vast

this method is

applied

to

print homogenous layers


been
an

chemistry required
a

of additives

has

developed

to tailor the

paste properties
can

for the

results. Most often


in air

organic

binder material is used that


at

be removed

by

heating step
well

prior

to

sintering

higher temperature.
but
are

These
to

processing steps
introduce
a

are

suited for oxide


into

layer deposition,

suspected

oxide

impurities

oxygen-free compounds
impurities
can

like CIGS. However, there may be

workaround

if the introduced lower the risk of

be removed

by

the

subsequent

selenization
were

step.
used

To
as

organic
where
a

contamination

pure alcoholic

suspensions

deposition pastes, deposition


was

ratio of iml alcohol per gram of

powder
a

was

chosen. The
screen.

done with

manual

screen

printer set-up using

250 mesh

3.1.3
The

Electrospray
has

electrospray

recently

been

developed

for chemical spray

pyrolysis.

The main

advantage compared
While

to traditional sprays is the very


a

high droplet deposition efficiency.


out of the

pneumatic sprays carry


an

large

fraction of the

droplets

deposition

area,

electrospray generates
the and

electric field between nozzle and substrate which attracts

charged droplets driving


force
allows

to the

deposition target.
also

The electric field acts


the

as

an

atomizing

which
a

speeds

up

droplet transport (Siefert, 1984).


phase
reaction
near

Electrospray

precise control
a

of the gas

the

heated
even

substrate, good film adhesion and


with
are a

very

high efficiency

of material usage

(>

80%

carrier

gas).

In the

present

work

no

gas is used for atomization but

some

vapors

produced during
has

the

evaporation

of the

suspension liquid.
in

Electrospray

successfully
as

been

applied

chemical

spray

pyrolysis
et

to

grow

various oxide films such

lithium manganese oxide

(Vanzomeren

al., 1994), indium

35

tin

oxide, titanium oxide (Choy, 2001) and non-oxide films like ZnS, CdS, and CdSe (Su

and

Choy, 2000).
chemical
was

While

spray

pyrolysis

uses

diluted

precursor

materials,

in

this

work

electrospray
Precursor

investigated
were

for the transfer of solid

particles suspended

in alcohols.

liquids

prepared

with suspensions of ethanol and 1,2

propanediol

with

approximately 35ml

of alcohol per gram of

powder.

Heating
Substrate
i i i I

5kV

Figure

3.2

Schematic

drawing

of the used

electrospray setup.

Figure

3.2 shows the

electrospray setup consisting


et

of

an

electrified

plate-nozzle
of the

configuration (Vanzomeren
suspension
reservoir. is controlled
on

al., 1994)

in

normal

air

ambient. The feed

by the hydrostatic
the

pressure between nozzle


one or

tip
are

and

suspension

Depending
The

applied potential
mode

several

jets

emitted at the

nozzle

tip.

Taylor
by

cone

(Taylor, 1964),
a

which is used for the

deposition
with
mm

process, is defined

the emission of

single,

stable

jet, breaking

into

droplets
2.1

nearly monodisperse
outer

size distribution. A Hamilton Luer

Tip probe (Pt. Style 3,


mm

diameter, 1.6

mm

inner

diameter)

was

used

as

nozzle, located 50

below the
ratio of

substrate. An evaluation of the material

deposition efficiency (MDE, weight sprayed particles)


a was

deposited particles
spraying
cm2 in
1

on a

defined surface to the total


on

achieved

by

ml of the

1,2-propanediol suspension

heated stainless steel foil of 5x5

area.

The increase in

weight

of the foil

was

compared
i

to the

weight
same

of the

residues after

heating
MDE

and

evaporation

of the alcohol of
an

ml of the of

spray
As
a

suspension. The

measurements

yielded

efficiency
a

nearly

100%.

comparison, conventional pneumatic spray yields typically


ultrasonic spray

MDE of

25% while for

-85% MDE has been reported (Eberspacher

et

al.,

2001).

36

3-2 Evaluation
3.2.1 Mixed-metal precursor

Typical

SEM
are

pictures

of

as-deposited pastes
Fig.
3.3. In the
case

for

doctor-blade,

screen

printing

and

electrospray

shown in

of doctor-bladed

layers (Fig. 3.3a), deposition

the

formation of clusters of

agglomerated particles prevents


During evaporation
of the

uniform

of thin

layers

of less than the

urn.

suspension liquid,

the risk of

dislocating

deposited particles, liquid


to

which may result in clusters and cracked


ratios. A better

layers,

rises for increased

particle

uniformity
present

of the film is obtained


a

with the

screen

printing
a

method

(Fig. 3.3b);

such films

clustered structure the


screen.

as

well, but they have

smoother surface due to the mechanical


was

forming by

In

electrosprayed layers (Fig. 3.3c) agglomeration


arising
from the small

limited to submicron

clusters

droplet
or

size of the

spray. It is desirable to evaporate the


on

suspension liquid before


film remains

just after the droplets impinge


are

the substrate,

so

that the

dry

and the

particles

unable to

move

after

deposition.

Figure

3.3

SEM

images showing

the

morphology
screen

of mixed metal

nanoparticulate

precursor films

deposited by doctor-blade (a),

printing (b), and electrospray (c).

The substrate the


is

temperatures needed
of the

to

evaporate the liquid during spray


More

were

below

boiling temperatures
to

suspension liquids.

precisely, sprayed

room

temperature

enough

evaporate
the

ethanol

(bp

78C)

so

that the

film remains

dry.
a

Charging prevents
good

particles

from

agglomeration composition
to the
to
0.2

after atomization and results in

adhesion to the substrate. The

of

layers deposited by spraying


of the

the
A

ethanol

suspension does

not

correspond
i

composition

suspension.

decrease of the Cu/ln ratio, distortion increased with effects of

typically from

is observed

(Table 3.1).

The Cu/ln ratio

increasing deposition

times. This may be

explained by settling
distortion,
a

agglomerates

in tube and nozzle. In order to avoid the

alcohol

37

with

higher viscosity (1,2-propanediol)

was

tested

(Table 3.2).

At

25T

the

Cu/ln

distortion could be reduced but not eliminated

(Table 3.1).

Table 3.1

Cu/ln ratio of

sprayed

films.

Suspension liquid
Ethanol

Suspended powder
Cu/ln ratio
i

Sprayed
0.2

film

Selenized film
Cu/ln ratio

Cu/ln ratio

025
0.8

1,2-Propanediol

0.7

Table 3.2

Boiling temperature and viscosity

of alcohols used for the spray

suspension liquid

(Lide, 2000).
Alcohol

Boiling
temperature [C] 78 184

viscosity [mPa s]
o

viscosity [mPa s]
25 C
1.1

viscosity [mPa s]
50 C 0.7 113

Ethanol

1.8

1,2-Propanediol

248

404

Electrospray

of ethanol

suspensions

is

delicate,

as

little vibrations and

small

changes

in the electrostatic

configuration (charges
produce
the

near

spray)

lead to instabilities in
was

the spray jet. The

voltage
the

used to

Taylor

cone

spray mode

around 5 kV.

Using i,2-propanediol, Figure

jet

is

more

stable and does not of

get disturbed that easily.


films for ethanol and 1,2at

3.4 compares the

morphology
For

electrosprayed

propanediol suspensions.

ethanol

suspensions sprayed
distributed

room-temperature
If

(Figure 3.4a)
used
as
as

the film is

composed

of

uniformly

particles.

1,2-propanediol
are

is

suspension liquid, bigger agglomerates Figure 3.4b

of several micrometers

deposited, by

shown in

and 3.4c. The thickness of the

sprayed
one

films
hour

was

controlled

the spray time. While ethanol

suspensions required nearly


10

spraying

time for

precursor films with

~2um

thickness,

minutes

were

sufficient with
of the

1,2-propanediol

suspensions.

This

indicates that the

higher viscosity
The

suspension prevents
of 1,2

particles

and

agglomerates

from

sinking.

higher boiling point


90C-iooC
to

propanediol dry

(i84C) required heating

of the substrate to

keep

the substrate

during spraying.

38

Figure

3-4

SEM

images

showing

morphology
using
an

and

cross-section

of

electrosprayed

metal
1-

nanoparticle layers. Layer (a)


2-propanediol suspensions.

was

deposited

ethanol

suspension. Layer (b) and (c)

result from

The measured

composition
as a

varied

across

the

sample Figure

35 shows the copper


center

and indium fractions small offset the

function of distance from the


shows
a

sample

Except

for

composition
is

symmetrical
from the

distribution around the towards the

sample sample
is

center. The copper content center

decreasing

sample edge

This

problem
in

can

be circumvented

by

using moving substrates, which

common

practice

industrial spray processing

39

-10

-5

0
on

10

Position

sample (mm from center)


in function of the distance

Figure
from the

3.5

Compositional analysis of electrosprayed precursor layer


center.

sample

3.2.2 Metal-oxide precursor


Selenization of doctor-blade

deposited oxide

precursors led to

severly

cracked

layers
an

(Figure 2.10). Figures 3.6 a)


electrosprayed

and

b)

compare the
from

morphology

of

screen-printed
The

and

oxide precursor

layer
are

1,2-propanediol suspension.

deposition

results for metal-oxide precursors


is more

similar to the metal precursors: The small round

sprayed layer
due to
un-

homogeneous

and shows

only

deepenings, probably
printed layer
However,

evaporated droplets impinging

on

the

surface.

The

shows

bigger

agglomerated

particles

and

indicates

micro-cracks.

the

subsequent

selenization of the
The width and

electrosprayed layer length


of the

also led to crack formation.

cracks

diminish when the


and to

deposition

method

is

changed
the

from doctor-blade to

screen

printing

electrospray.

This may be linked to

high liquid

content in the doctor-blade

deposited layer

precursor before

evaporation of
of the

the alcohol. Crack formation in the precursor


alcohol.

results from

evaporation

During

the conversion of the precursor the cracks widen. The The small cracks appear
a

electrosprayed

layer

remains

dry during deposition.


due to

during

the conversion at coefficient of the


et

high temperature probably


substrate and the

difference in the

expansion
literature

particulate
of

precursor
a

layer.

In the

(Eberspacher

al.,

2000),

volume

expansion

factor of 2-3 times is

reported

for the conversion of

oxides to

selenides, which should be sufficient for pore elimination and avoidance of

crack formed

by tensile

stress.

However, the conversion reaction is not

complete (2.2.2)

and the cracks form


reaction.

probably during heating

up and before the chemical conversion

40

Figure 3.6

Morphology
a

of

screen

printed (a)

and

electrosprayed

precursor

layer (b).

The

sprayed
of

layer presents
both

smoother surface

compared

to the

screen

printed layer.

Selenization led to

cracking

layers,

screen

printed (c)

and

electrosprayed (d) layers.


a

Mechanical

compression of the
but

screen-

printed

precursor

layer prior

to selenization led to

very smooth

layer,

cracking

could not be

prevented (e).

In order to

investigate layer
was

the influence of

higher packing density


with
a

on

crack

formation,
After

the precursor precursor

mechanically compressed sample


was

manual hand press

deposition

the

covered with

3mm thick

glass plate

and

put
was

under the press. After selenization, the


not

layer

shows

smooth surface, but


a

cracking

prevented (Figure 3.6e) Figure


a

3.7 compares the microstructure of

compressed

with

normal metal-oxide precursor

layer.

41

Figure

3.7

Morphology

of selenized

electrosprayed

oxide precursor

layers

without

(a)

and with

compression step prior to

selenization

(b)

In the
on

compressed layer large copper


a

rich

crystals (as

measured

by EDX)

are

formed

top

of

smooth surface. This may be

explained by
though

fast diffusion and

dewetting
selenized

properties
in the

of the copper selenide


run

phase.

Even

the two

samples

were

same

(~450C), layer

the

compressed layer
be

contained

only

20% of selenium

whereas the normal films

had

24%. This may


to

an

indication that denser precursor

require longer selenization times due

more

restricted selenium in-diffusion.

3.3 Conclusion
Three
were

non-vacuum

deposition techniques commonly

used for thick film

deposition deposition

tested for thin-film


were

deposition

of nanosized precursor materials. The


in pure alcohols to

pastes

prepared
fillers

as

suspensions

prevent pollution by organic


the

additives

as

or

dispersants. Doctor-blade, being


poorest layers compared
on

simplest deposition

technique, yielded printing gives


more

the

with the other

techniques.

Screen and

control

layer homogeneity by using higher viscosity pastes


the
screen.

by

the mechanical

forming by
a

Electrospray deposition yields homogeneous


time. To

layers

at the cost of

much

longer deposition

improve

the spray

jet stability

and the control of For


even

composition higher viscosity alcohols have


on

to be used.

better results

layer homogeneity,
are

the

deposition pastes
due to

have to be

improved.

The bad results of doctor-blade

mainly
Ball

agglomerated particles
can

which should be better

dispersed

the

liquid.

milling (Figure 2.5) being optimum

be used to

improve the pastes

but the results

are

still far from

for

homogenous

42

layer deposition.
be avoided

The

cracking

of oxide and selenide


methods.

layers during

selenization could not

by any of the tested deposition

43

4 Selenization
The conversion of the precursor material to the CIGS selenization reaction. CIGS,

phase
a

is achieved

by

the

being
and

refractory

material with

melting point
as

of

989C,
is

generally

exhibits

sintering

grain growth

at the same time

the CIGS

phase

formed. Low-cost therefore

glass
and

substrates do not support temperatures

higher
to

than

6ooC,

reaction

deposition temperatures
common

are

limited

this

maximum

temperature.
thermal

The most

selenization

agent

is selenium vapor

produced by

evaporation.

Selenium is known to

produce

various

allotropie

modifications

(cyclo-hexaselenium

Se6, cyclo-heptaselenium Se7, cyclo-octaselenium Se8, ...) with different crystalline and

amorphous
at

structures

showing red,

grey, and black colors. It melts at 22oC and boils

685C. Selenium vapor

consists of
on

Sen molecules (n

2-10)

with

molecular size
shift the

distribution that

depends

the vapor

temperature. Higher temperatures

distribution to low atomic the

molecules, which have higher enthalpies of formation than

higher

atomic molecules is

(Steudel

and Strauss,

1987).

The existence of different Another

vapour

compositions

supposed

to influence

reproducibility.

commonly

used

reactant is

hydrogen

selenide gas

(H2Se).
are

Selenium and its

compounds

toxic for the human

body. Consumption
such
as

of

more

than lug Se per g of food leads to

poisoning phenomena

inflammation of the

respiratory organs
if less then
0.2

and loss of hair. On the other

hand, Se deficiency symptoms appear


an

ug Se per g of food is consumed because Se is

essential mineral with


to

powerful
immune

antioxidant effects in the

body.

Se also

seems

indispensable
Cd and

appropriate
their

response. It

can

chelate various metals

(e.g.
of

Hg) allowing
room

detoxication

(Dubois

and Belleville,

1988).

Elemental Se is solid at
use

temperature requires

and therefore

unproblematic

to handle. But the

hydrogen
and

selenide gas

sophisticated safety precautions (scrubber).

and

expensive storing

disposal equipment

44

4-1 Chemical conversion in Figure


4.1 shows
a

vacuum

reactor

schematic

drawing
a

of the reactor used for the selenization

treatment. The reactor is set up in

stainless steel

high

vacuum

chamber and consists

of

heated

quartz plate

and

heated cap which

can

be lifted. Two

temperature

zones

allow

separate heating

of the selenium
a

pellets

and substrate. Prior to selenization, the

vacuum

chamber is evacuated to

pressure below 10'5 mbar with the lifted cap. Then which results in
on

the cap is lowered onto the


selenium vapor. However, the

plate,

better confinement of the

fitting

of the cap

the

plate
on

is not

tight

and the
vacuum

selenium
chamber.

can

diffuse out of the reactor and then condenses

the walls of the

high

vacuum

chamber

high vacuum pump roughing pump

Figure 4.1

Schematic

drawing

of vacuum reactor for selenization.

4.1.1 Conversion reaction


For the formation

of CIGS

variety

of reaction

pathways
were

must

be considered
are

depending
stacked

on

the precursor material. The precursors that


of metal
reactors

best

investigated,
are

layers

(evaporated
under

or

sputtered)

and selenium which

annealed in
XRD

high

vacuum

excess

selenium. Time

resolved

in-situ

during

annealing

of

substrate/Cu/ln/Se precursor identified the

following (crystalline) phase

transformations

(Brummer

et

al., 2003):

45

Table 4.1 stacked


vacuum

Phases and reaction

paths investigated by

in-situ

high

energy with

powder

diffraction for
Cu/ln/Se
=

deposited

precursors

(substrate/Cu/ln/Se configuration

composition

1/1/2.5) (Brummer et al., 2003).

Annealing
Temperature
RT

Ph

ases

and Phase transformations

Cu, Culn:,

Cujn^

Sea morphous ^

150-200C 125-175T 208-225C

Culn2+Cu

Culn9
Sec (crystalline)
Sei (liquid)

Se3 (amorphous)
Sec

->
-

225C
35C 375C

Cujng + Se, ln4Se3


2
+

^
+

ln4Se3 + CuSe2
lnSe
+

CuSe

Cu2.Seafter275.c

CuSe2

CuSe
+

-
^

Cu2.xSe

lnSe

Cu2.xSe

Se

2CulnSe2

In contrast to

vacuum

evaporated

stacked precursors,

nanoparticulate

precursors

have

porous structure and may react contained in the metal

differently. Additionally,
precursors

the observed

ln203

impurity phase
the

nanoparticle

(2.2.1)

may influence

typical

reaction

path

shown in table 4.1.

To

investigate

the conversion reaction with

nanoparticulate

metal precursors,

series of selenization treatments was carried out at various

temperatures (steps
temperatures
reaction time

of

iooC between

too

and

6ooC). During
source was

the

annealing

at different

the
was

temperature
30 minutes.

of the selenium

kept

constant at

3ooC. The

Figure

4.2 shows the XRD

patterns
this
ex

of the selenized and

quenched samples.
phase
back
at

In contrast to in situ

phase monitoring

situ method may suffer from

transformation
room

(including

solidification of

liquids) during quenching

and

storing

temperature.
contrast
no

In

to

the

observations

made
are

on

vacuum

deposited

stacked
and

metal
are

precursors,

intermetallic

phases

observed. The

phases Culn2

Cujn9

known to form
in the

already

at

room

temperature but only

elemental Cu and In
is

are

observed

nanoparticle

precursor. The fact that the

layer
a

composed

of small elemental

particles

with little contact surface may result in


on

very slow

intermixing.

The

ln203

impurity phase, probably present


inhibit the intermetallic

the surface of the In

nanoparticles,

may further

phase formation.
evaporated

Selenization at iooC does not lead to any detectable reaction. The

selenium is

deposited

on

the precursor and

crystallizes

in its

thermodynamically stable

hexagonal

structure.

46

CulnSe,

3.
"5
c

Q 300*C

200"C

ioo-c

Precursor
i

|
28

; ii
30

|
32

rr'i'i rr\
34

rn

\
38

rn |

rryrn |
42
44

i"|"i
46

|
48

p? r|
SO

f
54

26

36

40

52

2 Theta

(degree)

Figure 4.2

XRD

patterns of selenized

Cu-ln

metal precursors at different

temperatures. (Joint CulnSe2, 40-1487;

Committee on Powder Diffraction

Standards, ASTM, Philadelphia,

PA, 1967, Cards:

InjOj, 6-416; Cu2.xSe, 6-680; CuSe219-400; Mo, 42-1120)

After

passing the melting point

of indium

(i56C)

and

approaching the melting point


CuSe2 phase
is

of selenium

(220C)

first reaction between selenium and copper to the


vacuum

detected. In contrast to the

evaporated

stacked precursors
some

(Table 4.1),

no

indium selenide

phases

could be identified and there is still

metallic indium

present that may


At

have been stabilized

by the ln203 phase.


CIS
a

300C

the indium and the

CuSe2 phase have disappeared and the


is still

phase newly

started to form. However, the formed and the

ln203 impurity phase


selenization

present together with


the CIS

Cu2.xSe phase.

With

higher

temperatures

peaks get sharper

Cu2.xSe peaks diminish indicating higher interdiffusion

rates and increased CIS

crystallinity (Figure 4.4).

47

Table 4.2

Phases

formed

during

selenization

of

metal

nanoparticle precursors

at

various

temperatures.
Selenization Phases

Temperature (eC)
25
100
200

Cu, In,
Cu, In,

ln203
Se

ln203,

In, ln203, CuSe2

300-600

CulnSe^, ln203, Cu2.xSe

The conversion of the oxide precursor material


reactions where first the metal-oxides to CIS. However,
no are

most

probably

consists of two

converted to their selenides and then react


of the conversion reaction is available.

data of in situ

monitoring

4.1.2
The

Impurities
has shown that it is not

previous chapter

possible

to eliminate the

impurity

phases

in the metal precursor with the used selenization reactor and

temperatures up

to 6ooC. As

presented

in Table 4.1, the

Cu2_xSe phase
But
as

reacts with indium selenide and


no

selenium to form the CIS

compound.
the

there is

indium selenide cannot be

(indium

is

trapped

in

the

oxide

phase)

copper
to
a

selenide

phase

converted.

Temperatures higher than 6ooC lead


are

strong bending
In

of the

glass

substrate and

therefore of

no

interest for

production.

Figure

4.3 the
was

phase

fractions of CIS and

ln203

in the selenized

layer

are

shown. This data

obtained

by fitting

the XRD

patterns

with

crystal
on

diffraction data of the identified Powder Diffraction

phases (Powder

Diffraction Files,

Joint Committee

Standards, ASTM, Philadelphia, PA, 1967, ln203:

Card

6-416;

CulnSe2: Card

40-1487).

The oxide

phase

contributes

nearly

constant

amount of-i8wt%.

The conversion reaction in the other precursor materials has not been
in detail. Similar observations
are

investigated

made for the oxide precursor material


can

(Figure 2.11).
is not in

Whereas

some

of the oxide

phases

be converted to CIS, the

ln203 phase

reacted and remains in the small concentration

layer. Also,

the selenide precursor forms the

ln203 phase

during

selenization

(Figure 2.13).

48

;?

ioo 80 60 40 20 0 300 400 500 600

w
c

8
Rt Q.

CulnSe2 ln2o3

selenization

temperature (C)
in

Figure 4.3
selenization tested

Phase fractions of

ln203

and CIS

the selenized

layer

in

function
2.0
on

of different

temperatures calculated by fitting the

XRD

patterns with the Topas

software. For the the conversion of

temperature range, the selenization temperature has only little influence

ln203toCIS.

Enthalpy

and free energy calculations show that the conversion of

ln:03

to selenide

is difficult to achieve with the

given

reaction.

ln203 + 9

Se->

ln2Se3 + 3 Se02

AH, AGr

490

kJ7mol;n7kcai/moi
; i2ikcal/mol

(4.1)

5O5 kJ/mol

The

use

of

higher

selenium

partial

pressures allows

partial

conversion of the oxide

phases (2.4).
A
more

efficient reaction for the conversion of the oxides is

reported by Kapur

who

used

hydrogen
to CIGS

gas to reduce the oxides to metal and H2Se gas to convert the metal

alloy

(Kapur

et

al., 2003). Even though this reaction is


is

not

energetically
conditions

favorable

either, the

conversion

achieved

by using non-equilibrium

(excess hydrogen

and successive removal of the volatile conversion

byproduct H20).

4.1.3

Crystallization
phase
starts to form at

The CIS

300C (Fig. 4.2). During

the thermal treatment,

gradual

conversion and coalescence of small


is

grains

lead to

larger grain
CIS

size. This

phenomenon
patterns
for

reflected with selenization

sharper

and

higher-intensity
The average

peaks

in the XRD

higher

temperatures.

crystallite

size has been

49

extracted with the Rietveld method software

(Cheary

and

Coelho, 1998) using the Topas

2.0

(Bruker AXS, 2000)


data

and the

Inorganic Crystal

Structure Database for CIS and

ln203 crystal
size
as a

(CIS:

ICSD

6397, ln203: ICSD14387). Figure


temperature.

4.4 shows the average

grain

function of the selenization


nm

The average CIS

crystal

size grows

from 40 to 170

with

an

increase in selenization
no

temperature

from 300 to 6ooC.

The

ln203 crystals

cannot grow much since

further oxygen is

provided during

selenization.

c
<D

160-

U>

120-

JS
(A

fr
O o>O) re
i_

80.

> (0

401
'

'

'

300

400

500

600

selenization

temperature (C) phases


in

Figure 4.4

Average crystal

size of the XRD

ln203

and CIS

dependence
software.

of the selenization

temperature calculated by fitting the


lead to

patterns with the Topas

2.0

Higher temperatures
constant

gradually larger

CIS

crystals,

while the

ln203 crystal

size remains

nearly

during

selenization.

4.2 Chemical conversion in High


solar
vacuum

non-vacuum

reactors
times. Low-cost Since
non-

equipment

is

expensive

and

requires long pumping


for
non-vacuum

cell

production

methods

therefore

aim

processes.
a

atmospheric
vacuum

oxygen would lead to oxide formation

during annealing,
argon
or

suitable

process

requires

an

inert

atmosphere (typically

the

considerably
the

cheaper nitrogen)
reactor may be

which is then mixed with selenium vapor. For for

safety precautions

designed

sub-atmospheric
or

pressure
A further

operation

which reduces the of


a non-vacuum

risk of out-diffusion of reactants


reactor is the
vacuum use

explosion.
some

advantage

of reactant pressures

orders of

magnitude higher
of
a

than used in

reactors. This allows

shifting

the chemical

equilibrium

reaction to the

desired

product.

4.2.1 Reactor

design
a

Figure
precursor
reactor
can

4.5 shows

schematic illustration of the


in
a

employed

non-vacuum
a

reactor. The

sample

is

placed

quartz

vessel. It is difficult to construct


to 6ooC with
uses

closed hot-wall
and which
Two inlets

capable

of

working

at

temperatures up
This reactor The

tight sealing

be

opened

to load

samples.

quartz-quartz sealing. part


have

allow

purging

the vessel with

nitrogen.

top

and bottom

independent

heaters for

independent

control of substrate temperature and selenium pressure. A

second container around the

quartz

vessel is

pumped
the

to the

same

base pressure and

condenses any selenium vapor


water cooled. A

diffusing through
a

sealing.

The outer container is

membrane

pump with

base

pressure of 1-5 mbar is used for

evacuating the

reactor.

roughing pump

N2-*^

Jt

pump

%
C

re

quartz container

Figure 4.5

Schematic illustration of

non-vacuum

reactor, allowing the

use

of

higher

selenium

pressures than the

vacuum

reactor.

The elimination of the

ln203 phase

in the precursor has turned out to be very difficult

with selenium vapor. In order to

improve

the

efficiency

of this reaction, selenium has to the selenium

be available in

large quantities,
gives
an

which is achieved

by increasing

partial

pressure. Table 4.3

overview of the process

parameters

of the two reactor

systems.

51

Table 4.3 Comparison of reactor parameters Vacuum reactor

Non-vacuum reactor
1

Base pressure

io"5mbar
10"1 mbar

mbar mbar

Selenium pressure Selenium Substrate


source

100

temperature

~300C 500-550^

~500C 500-550C

temperature

4.2.2

Impurity phases
layers
similar to the
ones

Precursor
non-vacuum

used in the

vacuum

reactor

were

selenized in the

reactor.

Figure 4.6

compares the selenized and

layers

obtained from the two

reactor

types. Impurity phases, namely the Cu2.xSe


with the
vacuum

ln203

are

only detected

in

layers

processed

reactor. Selenium vapor at

higher

pressure is therefore

advantageous
allow

for the conversion of the oxide

phase. However,

the XRD results do not

excluding

the presence of small concentrations of oxides in the

layers

because of

the detection limit of XRD.

(112)

(204)/(220)

CIS

phase

(116)/{312)

w c

(101)

(103)

(211)

Mo

(301)
"

Q x

kT

A.

non-vacuum reactor

,.

,.

A... I

..*

i
1 1 "T 1

1
c uz.xSe
1 1
11t11-r

,-,,

A
1

vacuum

reactor

Cuz-xSe
25

lnz03
1
1

ln203
1
1 1 1

11|1111)111r

111111

15

20

30

35 2 Theta

40

45

50

55

60

(degree)
in the vacuum

Figure 4.6
and the
vacuum

Comparison of XRD patterns of metal nanoparticle precursors selenized


reactors, where the selenium pressure is 3 orders of

non-vacuum

magnitude higher

in the

non-

reactor

(Table 4.3). The high selenium pressure allows the conversion of the ln203 and Cu2XSe
CIS

impurity phases to the

compound.

52

Whereas the metal precursor contained

only

small

quantity

of the oxide

phase

on

the

particle surface,

the oxide precursor contains

only

pure oxides and will therefore be

more

difficult to convert. In which


was

addition, the oxide precursor also contained the gallium


to

component

not

possible

incorporate

in the pure elemental metal


a

paste.

Ga203will

be

even more

difficult to convert,
than that of

having

free energy of formation of AH/But since the

1090, which is

even

higher

ln203 ( AH/=-926).

gallium
of

content is rather low it is not detected

by XRD. Figure

4.7 compares the XRD

patterns

the oxide precursor selenized in different reactors. The the oxide precursor much better than the
vacuum

non-vacuum reactor

converts
a

reactor. But there is still

broad,

low-intensity peak

of the

ln203 phase around

20=

30.50.

CIGS

phase

(112)1

Mo

(204)/(220)
(116)/(312)

(0 c
at

(101)
X

(301)

>^.
non-vacuum reactor

X
ln23
I 1" T
'

^^v^^^
vacuum reactor
T -r
t11r-

Cuz.xSe
-111r

lnz03
T 1

Cu2.xSe2.*Se
1
1 1 1 I

"

15

20

25

35
2 Theta

40

50

55

60

(degree)

Figure 4.7
vacuum

Comparison
non-vacuum

of XRD

patterns of

NEI2 oxide

nanoparticle

precursors selenized in the

and the

reactors, the selenium pressure being 3 orders of magnitude higher in the


conversion of the

non-vacuum

reactor

(Table 4.3). The high pressure of selenium allows the

ln203

and

Cu2.xSe to the CIS compound..

Figure 4.8

shows the XRD

patterns

of the selenized CIS


more

powder

from Solaronix. The


vacuum

conversion in the

non-vacuum

reactor is

efficient than in the

reactor. in the

This is

seen

in the

complete layer.

conversion of the oxide

phase
two

which is still

present

vacuum

selenized

In addition to the CIS

peaks

large
top

and broad

peaks

of the back

MoSe2 phase appeared. This phase


contact,

must have formed on

of the

molybdenum
was

resulting

from reaction of

molybdenum

with selenium, and

not observed

53

in other precursor materials. This

can

be

explained by
of metal

the different

sintering

behavior leads
to

of the

precursors:

While

selenization
a

and

oxide

precursors

densification of the

layer by

2-3 fold volume


a

expansion (Eberspacher, 2000),


no

the

selenide precursor does not

undergo

chemical reaction and

densification takes
access

place (Figure 4.10).


the

Therefore the porous

layer gives

the selenium vapor direct much faster in the

to

molybdenum

back contact. This reaction is

again

non-vacuum

process.

CIS

phase

(101)

non-vacuum reactor

^fmttfV*** ^^tff*fe*^r^l*h' ' '0*4*&b*iiirig&*Ait&i,

vacuum reactor rT^

15

20

35 2 Theta

40

60

(degree)
in the
non-

Figure 4.8
vacuum

Comparison
non-vacuum

of XRD

patterns of the selenide nanoparticle precursors selenized

and

reactor, the selenium pressure being 3 orders of magnitude higher in the


is able to convert most of the

vacuum

reactor

(Table 4.3). The high pressure of selenium


present
in

ln203
In

and

Cu2.

xSe, which

is still

large

amounts in
are

vacuum

selenized

layers,

to the CIS

compound.

addition,

peaks indicating the MoSe2 phase


contact and the CIS to the back contact.

identified. This

phase forms

at the interface between the Mo back

layer. The remaining porosity

in this precursor

gives the selenium vapor direct

access

4.2.3
The

Crystallization
of the selenized metal

morphology

nanoparticle layer

indicates

compact

and of
a

large-grained
5
urn

structure

(Figure 4.9a). The in-depth


(55o"C
for 5

sinter reaction and

grain growth

thick

sprayed

precursor

layer was analyzed. Figure 4.9b


minutes).
The

shows the cross-section of


a

the

layer

after selenization
and
a

layer
layer.

consists of
As the

large-grained,
surface is first

compact top layer

porous

small-grained

bottom

layer

54

exposed

to the selenium vapor, the chemical conversion starts

on

the surface of the volume

precursor. The reaction of the metal to the CIS

phase

results in

expansion
vapor
the

which

leads to

compact, pore

free

layers. However,
layer.

the densification

stops

transport

of selenium to the bottom

Therefore after

sintering

of the

top layer

further conversion reaction is slowed down. thin

Nevertheless, the formation of a

1.5 to 2pm

compact

CIS

layer

is obtained in

short time and the reaction may therefore be

suitable for

rapid

thermal

processing

of thin precursor

layers.

Figure

4.9

SEM

micrograph

of metal precursor

layer

selenized in

non-vacuum

reactor

showing

the

morphology (a)
layer.

and cross-section

(b), showing

densified top

layer

and

porous, small

grained

bottom

Although
the

the chemical

conversion

of the oxide precursor could be

improved using

non-vacuum

reactor, the grain size of the

layer

remained small. Also

cracking

of the

layer could

not be

prevented.

Figure 4.10
precursor.

SEM

micrograph

of

a vacuum

(a)

and

a non-vacuum

(b)

selenized selenide

nanoparticle
it is still

Although

the obtained

grain

size in the

non-vacuum

selenized

layer

is much

larger

highly porous.

55

Sintering
difficult

and densification of the selenide precursors has turned out to be very The
use

(Figure 2.12).
a

of

a non-vacuum

process with

high

selenium pressure

resulted in

markedly changed morphology. Figure


each method.

4.10 compares the


does

resulting layers
change
the

obtained

by

Whereas

vacuum

selenization

not
a

morphology
the

of the precursor

layer,

the the

non-vacuum

reaction leads to

coarsening

of

layer with larger grains although

layer

remains porous.

4.2.4
The

Open

non-vacuum

reactor

design

of

closed hot-wall reactor


was

imposed

some

difficult construction issues.

The main

problem
parts

the

sealing

between the two caps of the


to fit and to

quartz

vessel

(Figure
high

4.1).

The two

were

carefully adjusted
system
was

give

nearly tight system.


a

Therefore, after

evacuation the

closed and allowed to maintain


were

selenium pressure. The

presented

results in the last sections


runs were

obtained within the

first

runs

after

set-up. Subsequent

not able to

reproduce

the
can

same

results

and the

ln203 phase appeared again

in the

layers.

This

phenomenon

be

explained

by two

reasons:

The

quartz sealing

of the caps started to leak

so

that the selenium pressure in


some

the vessel

dropped.
on

Direct evidence

was

given by the fact that

selenium

vapor condensed

the water-cooled outer walls of the reactor.

The second the

reason

may be the accumulation of oxide in the reactor. When


are

impurity phases

converted to the CIS

phase

the

byproducts

may

condensate in the reactor

(e.g.

in the form of

Se02).

The oxygen stored in the

reactor may then lead to concurrent reactions which favor the formation of

oxide

phases.

These drawbacks

were

eliminated

by using

an

open reactor

design

as

shown in

figure 4.11.

56

r-,

It

ij:iiniiimnin

Substrate N,
If

Cooling ""X traP

t
-

* i

Pump

350C

T~550C
non-vacuum

Figure 4,11

Scheme of the

employed

reactor with open ends.

Nitrogen

gas flows

over

the heated selenium

source

and carries the selenium vapor

to the substrate

region.

The outlet

region

of the reactor is cold and the selenium vapor

condenses there. In order to

keep
at
as

the selenium

consumption low, the selenium

evaporation temperature is kept

350-400^.

This

design

does not allow

as

high

partial selenium vapor pressures


reaction is less efficient

the closed reactor, and therefore the conversion

(Figure 4.12).
reactor does not

The open

non-vacuum

produce

the

Cu2.xSe phase but could

not

convert all the

ln203 crystals.
so

In contrast to the vacuum reactor, some of the oxides

could be converted

that the copper selenide reacted to CIS. The Cu2.xSe

phase

is

known to be

highly conductive

and detrimental for solar cells. It is therefore very

important

to eliminate it in order to avoid shunts in the CIS


a

layer.

The influence of the

ln203 phase, being


resistivity and
The open

very stable and less conductive

compound
et

leads to increased

reduced

mobility

and is also detrimental

(Haba

al., 1990).

non-vacuum

reactor is easy to use and

produces reproducible results, but

is not efficient for

converting the ln203 compound. nanoparticle


films
were

In order to characterize this reaction,

indium and copper


minutes at

deposited

and selenized

separately

for

20

550C (Figure 4.13). The

selenized copper

layer

shows

clearly

defined

large
In

crystals

with diameters up to 10pm. However, many pores form between the

grains.

contrast to the copper

sample, the grains

of the selenized indium

layer do

neither show

sharp edges

nor a

smooth surface.

57

(112)

(204) (220)

(hkl)

CIS

phase

Impurity phases

(116) (312)

ce X 550C 5 min 100 mbar closed


non-vacuum

Mo

(101)
reactor

jfl

(211)

(301)

550C 20 min
1 mbar

open

non-vacuum

reactor

600C 30 min 0.1 mbar


vacuum

reactor

Cu2.xSe
20

ln203
1!!!!!! !

Cuz.xSe
,
,

rh2o,
,r ~i:ii i )r

15

25

30

35

40

45

50

55

60

2 Thta

(degree)

Figure

4.12

XRD data for metal

nanoparticle
partial

precursor films selenized under different conditions. selenium pressure and reactor

Substrate temperature, selenization time, the left-hand side of the

type

are

indicated

on

diagram.

28fc.V.';.':X2-.:0'0#

a)HHHHHHHHHMHHHiHMHM

b)
of
a

Figure

4.13

SEM

pictures showing the morphology

selenized copper

(a)

and selenized indium

(b) nanoparticle layer.

58

Compositional depth profiling by sputtering


to

and XPS surface

analysis

was

performed

investigate

the selenization reaction in the two


a

layers (Figure 4.14).


the measured

The selenized

copper
no

layer

has

constant selenium content

through

depth

range and

oxygen is detected in the film, except for surface contamination. Therefore the
can

selenization reaction of copper The selenized indium

be considered

as

complete through

the whole film.


2oonm.

layer

shows

complete

reaction

only

for the first

Below

20onm, the oxygen content is

increasing

while the selenium content decreases and

they

stabilize after 500

nm.

The results indicate that the conversion reaction of indium

nanoparticles

is much slower than the conversion of copper

particles.

,...y y-yy.~

60

s?
-60
o

50-|
c

5 40-

g
c

40

o
o

g
o

30

20

*-*-*-*-

i
Q

20

I
0.0 0.4 0.8 1.2

10-

'

0.0

0.4

0.8

1.2

a)
Figure 4.14
XPS

Sputter Depth (nm)

b)

Sputter Depth (um)

depth profile

measurements for selenized copper

(a)

and indium

(b) nanoparticle

layers.

59

4-3 Conclusions
The selenization treatment is used to convert various precursor
absorber

layers

into CIS/CIGS

layers. Different

conditions for the conversion

reaction

were

tested and

evaluated

by

means

of CIGS

phase purity, grain size,


precursor
was

cracks and
in

density.

The conversion
a

of the metallic

nanoparticle

investigated

more

detail and showed

distinct behavior from

vacuum

evaporated,

stacked metal precursors. No


was

alloying

of the
to the

nanoparticle

metals
area

prior to selenide formation

observed, which may be due

limited contact

between the metal surfaces of the reduces the

particles,
alloy

since the oxidized

surface of the In

particles

reactivity

for metal

formation. Vacuum

selenization with selenium vapor pressures up to 10'1 mbar the

was

not efficient to convert

indium-oxide, hence the


reactor
was

reaction to the

ternary

CIS

phase

is not

complete.

A closed

non-vacuum

designed

to allow
were

higher

selenium

partial

pressures.

High

partial

selenium pressure reactions

very efficient in
even

removing impurity phases

during

selenization of the metal precursors and

converted most of the oxide selenized under different

precursors to the CIGS conditions lead


to

compound.

The

resulting layers
a

the conclusion that

high

selenium

partial

pressure

during

annealing

is more efficient for


or

complete

conversion than
vacuum

reaction time,
i

substrate

temperature reactors).

base pressure

(<io'5

mbar for

reactors,

mbar for

non-vacuum

As the conversion reaction and

layer

densification starts from

top

of the

layer,

further selenium inclusion and oxygen removal from precursor material


contact is much
more

near

the back

difficult.

Layers

with

thickness of

1-2 urn were

easily converted.
nano

However, the

deposition
include

of thin

(<i urn)

and
1

homogeneous layers

with the used

particles (which

agglomerates

up to

urn)

is

an

unsolved

problem.
CIGS

Metal-oxide
even

and metal-selenide precursors did not

yield compact, large grained

layers,

if

converted

in

higher-pressure

selenium vapor.

Although coarsening

resulted in the

selenide precursor, the the

porosity remained and allowed the selenium

to

directly
on

access

molybdenum

back contact, where it formed the

MoSe2 phase

top of

the

molybdenum

back contact. Pure CIS

phase

was

only

obtained with the metal precursor

selenized in the closed reactor.

Despite
obtain

the

promising

results with the closed


in successive
runs

non-vacuum

reactor

reproducibility to

single phase layers

was

not

possible.

Reaction

byproducts

remain in the reactor and

compete

with the selenization reaction unless the reactor is


overcome

completely

cleaned

periodically.

To

this

problem,

an

open

non-vacuum

6o

reactor

was

designed

and tested. Due to lower selenium


non-vacuum

partial

pressures, it is less
in
a

efficient for oxide conversion, but allows


manner.

processing

reproducible

The difficulties of
of the oxide

homogeneous

precursor

layer deposition
a new

and

complete

conversion

impurities

motivated the search for

precursor material, which allows of the metals

depositing
and

more

homogeneous
of

and thinner

layers, improved intermixing


layer.

optimal incorporation

gallium

in the selenized

6i

5 Solar cell devices from


This

nanoparticle

precursor
Only
solar

chapter
on

describes the

processing

and characterization of solar cells.

cells based
were

CIS obtained from selenized metallic Cu and In


as

nanoparticle

precursors

characterized in detail, and the

the other precursor materials did not lead to crack-free


were

CI(G)S layers

resulting

cells

shunted.

5.1 Device
The
CIGS

processing
of

processing

complete

CIGS solar cells

(Figure 1.4)

includes three

steps

after

deposition:

CdS buffer

deposition

in

chemical bath

TCO front contact

deposition (ZnO/ZnO:AI layers deposited by rf-sputtering)


grid (Ni/Al)
is

optionally
collection

metallic

applied

on

top

for

improved

current

Cell

area

definition

by scribing

5.1.1 Chemical bath


The thin

deposition

of Cd(S,OH)

layer
in conventional CIGS

n-doped Cd(S,OH) layer deposition


the

forms the p-n

heterojunction

solar cells. The

of the buffer

by

the

non-vacuum
so

chemical bath

deposition
of this

process has

yielded

highest-efficient
respect
to

CIGS solar cells

far. The
the

advantages

deposition

method with

vacuum

evaporation

are

etching
even

of the CIGS

layer

surface
50

prior

to

deposition

and the of

perfect

surface coverage

for very thin


and

layers (<
oxides)
CdS. The

nm).

The
a

incorporation

impurities from the bath (namely hydroxides

results in

material with different

properties compared

to

vacuum

deposited

recipe

used in this work consists in

mixing

the precursor chemicals cadmium-

acetate and thiourea in alkaline ammonium

hydroxide

solution before
to

immersing

the

substrate and
reaction

heating

of the bath from


as:

room

temperature

65C.

The chemical

taking place

is written

Cd(CH3COO)2 + SC(NH2)2

CdS^

+ 2

CH3COOH+ CH2N2

(5.1)

62

The

typical deposition

time is 20 minutes

during

which the bath is heated to


on

65C.

The thickness is controlled


where the

by
can

simultaneous CdS
be

deposition
by

Mo coated

glass,

layer

thickness
-50
nm

easily

estimated
an

its color. A dark blue color

corresponding to

Cd(S,OH)

is found to be

appropriate thickness (1.3).

5.1.2 ZnO/ZnO:Al TCO

deposition
cells,
an

As for conventional CIGS solar

rf-sputtered

ZnO

bilayer

front contact

was

applied. First,
buffer

first 50

nm

thin, high resistivity


a

intrinsic ZnO

layer

is

deposited
is

on

the

layer.

Then the conductive ZnO:AI with

thickness of 0.3 to
as a

1 urn

sputtered.

The thickness of the ZnO:AI and

layer

must be chosen is

trade-off between

transparency
a

conductivity.
Qn

If

metallic

grid

deposited

on

top

of the TCO

layer,

sheet
et

resistance of 70

is sufficient for
can

high

fill factors and efficiencies


maximum

(Ramanathan

al.,

2003)

and the thickness

be

optimized for

transparency.

5.1.3 Cell
After the

scribing
of the CIGS solar cell, the cell
area

complete processing efficiency

has to be defined

by

scribing clearly

for

measurements. The

layer

is therefore

partitioned

into smaller

defined cells
x 1

by scratching through
defined for

all the

layers

on

the

molybdenum. Typically,

cells of 0.5

cm2

were

laboratory

scale characterization.

5.2 Characterization

5.2.1
If
a

Contacting
grid
has been

metallic
as

deposited

on

top

of the TCO, it is connected

by the probe
on

needles

the front contact. Otherwise the front contact is connected back contact is connected at the

directly

the

TCO. The

molybdenum
are

edge

of the

sample.

Contact

resistances

monitored

by 4-terminal

Kelvin connections.

63

5.2.2
The

Current-Voltage

characteristic

current-voltage

measurement determines the electrical

output performance

of

solar cells
maximum

(1.3.4): open-circuit voltage (Voc),


power output of the device,

short-circuit current

(lsc),

fill factor

(FF),

the

voltage

at maximum power

(Vm),

current at

maximum power

(lmax)

and the conversion channels for

efficiency (r|). 4-terminal

Kelvin connections

enable the
minimizes

use

of

separate

voltage

and current measurements, which


losses

measurement

errors

by eliminating voltage-drop

resulting

from

resistances due to cables and contacts.

The most efficient cell obtained with based cell


vacuum

nanoparticle

precursors
was

was a

metal precursornon-

deposited by electrospraying.
a

The precursor

selenized in the open

reactor at

maximum

temperature

of 6ooC.

Figure

5.1 shows the current-

voltage

characteristics of this solar cell.

C30
E

FF:

33%

V0C:219mV

/
/

20
c

Jsc

17.3 mA/cm

Efficiency:

1.24%

i
1 1 1 1

D.2

-0.1

0.1

&

0.3

0.

Voltage (V)
-10

^^__^-~~'3o
Figure 5.1
from metallic
Current
-

Voltage

characteristics of

non-vacuum

processed

CIS solar cell

prepared

nanoparticles.

With

FFof 33%, Vqc of 219mV, lsc of 17.3 mA/cm2 all the values

are

about half of the

values of
measures

good, vacuum-processed
the

CIS cells. The low value of the fill


a

factor, which
a

"squareness"

of the l-V curve, indicates

large

series resistance and

low

shunt resistance.
In order to

investigate
CIS

the

layer composition
was

at different

depth levels,

XPS

depth

profiling

of

layer

was

done. Material

removed from the surface

by argon-ion

64

sputtering

in-between

consecutively acquired
Figure
5.2 shows the

XPS

spectra.

The

sputter

rate

is

approximately 100

nm/min.

resulting depth profiles.

50 r

Sputter

rate:

100 nm/min

OL
0 5 10 15 20

Sputter time (min)


Figure
5.2
XPS

depth profiles

of

selenized metal

nanoparticle precursor layer.

No clear interface between the CIS and the substrate

(soda-lime glass)

can

be

seen.

Although

the

molybdenum

concentration

was

not measured
or

(perturbation by Se3s
allows

peak)

the

binding energies

of the oxygen

(either ln203
and
12

to

SiOx)

identifying the
a

start of the substrate material between

10

min

sputter

time. With
nm

thickness of

about 300

nm

for the Mo

layer,

the average CIS

layer

is 800

thick.

Remaining

porosity

and the

inhomogeneous

thickness of the film

are

responsible

for the blurred


in the first

concentration

profile. However,

the

profile

shows

clear copper

deficiency

500

nm

layer

thickness. This fact may result

partly

from the spray process which

yielded slight composition


the

deviations due to

settling

effects in the spray


the

liquid

so

that

sprayed

film
are

was

copper deficient in the


vacuum

top part of

layer (3.2).

But similar

observations

made for

deposited
(Kurdesau

Cu-ln et

alloy layers

where the surface turns

copper deficient after selenization

al., 2004).

65

5-3 Conclusion
A

variety of
on

CIS

and

CIGS

layers

from

different

precursor

powders

were

deposited

molybdenum

coated

glass

substrates and

processed

to solar cells. Oxide

and selenide precursors

yielded only

resistive devices without

photovoltaic

behavior.

Measurable
and

photovoltaic properties

were

observed in devices obtained from These CIS solar cells

sprayed
a

screen-printed

copper and indium of

nanoparticles.

yielded

maximum

efficiency

1.4%. The eel I parameters lsc, Vocand


vacuum

FF of the best cell have

only

half of the values of There


are

good

processed

CIS cells.

different

reasons

for the low

efficiency

values obtained. XPS

depth
have
a

profiles rough

and cross-section SEM

images

have shown that the selenized


some

layers layers

surface and

high porosity
is

which lead in

places

to very thin

where and

electrical

shunting
are

possible.

The

mixing

of the precursor

particles
with

is not

perfect

agglomerates
deviations. But copper

formed which

lead to

precursor
a

layers

local

composition
of

single-phase already

CIS is limited to

narrow
excess

composition range

24-24.5%

(1.2)

and

small deviations with

copper lead to the formation of

the low-resistive

Cu2.xSe phase which shunts the cell.

66

Inorganic

salt precursors and solar cells


a

Metal salt precursor materials have been used for

long

time for

non-vacuum

CIGS

layer deposition
containing salts

in
are

the

spray

pyrolysis

process.
a

Copper-,

indium-

and

galliumin
a

dissolved
on a

together

with

soluble

chalcogenide compound
react

liquid

and

sprayed

heated cell

substrate where efficiencies


were

they

and form the CIGS

compound.

Obtained

solar

rather

poor,

probably

due

to

nanocrystallinity
et

of the CIGS and inclusions of

byproducts during

the reaction

(Krunks

al., 2002).
Within the work of this

thesis, inorganic metal salts dissolved

in alcohol solvent

are

used to

deposit

mixed salt precursor

layer

instead of

depositing

the CIGS

compound
the solvent

directly.
and

After precursor materials.

layer deposition,
The

the

sample

is dried to
is

evaporate

binder

dry

precursor

material

then

converted

to

CIGS

by

selenization in selenium vapor. The


use

of

liquid

precursor materials avoids most of the

problems
particles

encountered in

nanoparticle
mixing.
reaction

precursor

pastes

such

as

agglomeration

of

and

non-perfect
lowers the

Atomic

scale-mixing generally

increases the reaction

rate and

temperature. However,

the anion

part of the salt,

the

solvent, and possibly any


to be released and

binder materials introduce further elements into the

paste that have


were

extracted

during

chemical conversion.

Inorganic

salts

chosen to

prevent

carbon

contamination.

6.1 Doctor-blade
The precursor
an

(Paste coating)
deposition
of
an

paste

should allow the

amorphous layer consisting

of

atomic scale-mixture of the precursor salts. Nitrate and chloride precursors


were

(Table

6.1)

chosen because of their low cost and


in water and
a

good availability.
as

The precursors also

show

good solubility

variety using
and

of alcohols such
nitrates include

methanol, ethanol,
process,

propanol
exclusion

and others. The of carbon

advantages

of

simplicity of the

contamination,
salts.

their

low

decomposition temperatures

compared

to other

inorganic

67
Table 6.1

Inorganic salts
Precursor

used for precursor

paste preparation

in this work

Molecular formula

Purity (%)
99-999 99-99 99-99 Techn.

mp (PC)

(C)
-

Cupric chloride, anhydrous

CuCl,

630/300 dec.
115

Copper

nitrate

hydrate anhydrous hydrate


hydrate

Cu(N03)2xH20
InClj

170 dec.

Indium chloride, Indium nitrate

583
100

500 subi.
-

ln(N03)3xH20 Ga(N03)3xH20

grade

dec. dec.

Gallium nitrate

99-999

no

6.1.1 Paste formulation


Dissolved precursor materials
can

easily

be

applied

to the substrate surface

by spin by

coating, dip coating,

or

spraying. However,
et

two main

problems

have to be solved

adjusting

the

paste chemistry (Schwartz

al., 2004):
the mixed metal salts lead to local

When the solvent is allowed to


cationic

evaporate,

segregation

because of their different solubilities and

decomposition

temperatures/rates.

Dewetting
extent of

of the substrate may

occur

due to the absence of

gelation

and the

hydrogen bonding.
binder material, which
is

Therefore,

an

organic

decomposes
as a

upon heat-treatment, is

added to the

paste. Ethylcellulose
was

widely

used

binder material for

coating

applications
the

and

adopted

in this

approach.

Because the

ethylcellulose
of
a

influences after
in

solubility

of the metal salts and leads to the


was

precipitation

white

powder

some

days (which

not further
are

investigated),
before

the precursor

paste

was

prepared

two

components,

which

mixed

just

deposition:
of Cu nitrate, In
is

Precursor

pastes

were

prepared by dissolving appropriate quantities


in methanol. The precursor
1

chloride and Ga nitrate

hydrate
a

composition

adjusted

to

the desired metal ratio with

concentration of

mili-mole

(mmol)
is

copper atoms per

gram methanol. In

parallel,

high viscosity
with
a

cellulose

paste

prepared: Ethylcellulose
pastes
are

(30-50 mPas)
in
a

is dissolved in ratio of 1/2 to

1-pentanol yield
a

weight
paste

ratio of i/io. The

mixed

weight

precursor

with suitable

rheology

for doctor-

blade

coating. pastes
were

Alternative precursor dissolved in methanol

prepared containing

Cu chloride

or

In

nitrate
were

or

1,2-propanediol. Resulting layers


paste.

and cell

properties

inferior to those obtained with the above described

68

6.1.2 Paste

deposition
glass
substrate is coated with
For doctor blade
a

A1 mm-thick soda-lime

400

nm

thick

molybdenum
of scotch
area.

layer by

conventional

dc-sputtering.
1 cm

deposition,

two

stripes

tape (-50 pm thickness,

wide)

are

applied
is

on

each side of the distributed


on

deposition

In

one-pass movement the precursor

paste

evenly

the substrate surface. the metal

By weighing

the

deposited

precursor

layer
has
a

and

calculating

parts
10

in the

layer,
After

it

can

be estimated that the

layer

copper content of about

mmol/m2.

complete yield
a

conversion without metal losses CIGS

during preheating
600
nm.

and selenization,

this will

layer

thickness of

roughly

Thicker CIGS films would

require higher metal concentrations,


thicker films, the metal concentration

but due to adhesion


was

problems

encountered for

kept

low.

6.1.3 Preheating
The

as-deposited sample

is

placed
to

on a

hotplate

and heated to

250-350^

for

few

minutes in ambient

atmosphere
on

evaporate

the volatile alcoholic solvent and to burn

the cellulose.
in different

Depending

the precursor salts used, the heated

layer changes

its color

steps.

For the

paste described

in

6.1.1 the

color

changes
a

from white to

brown and then to black. When the

layer

turns from black to

metallic blue, the

sample

is removed from the

hotplate

and allowed to cool down.

Longer heating

and and

higher temperatures
changes

lead to fast oxidation of the uncovered

molybdenum layer
further to
a

the color of the precursor

layer

to silver metallic and

rough

reddish surface.

Homogenous

film thickness and

heating

rates

are

important to
colors.

avoid the formation of

regions

where the

preheated layer

shows different

Depending phases. Figure


conditions
no

on

the

drying conditions,

the heated film contains different

crystalline

6.1 shows the XRD

patterns

of the heated films. Common to all


or

heating
can

crystalline phases containing gallium


even

indium

are

detected. It

therefore be concluded that

though

the precursor

layer was

heated to

350C

in air

for several minutes the group III metals did not form

crystalline heating

oxides which would

complicate

the chemical conversion


a

(Chapter 4).
ethyl

After

the

sample

to

50C

to

evaporate the methanol


formed.

dry

film of

cellulose

containing
are

the metal salts is

During evaporation

of the

solvent, crystals of CuCI

formed. This is due to its

69

CuCI

(111)
o

Mo

cl

'c a

CuCI

(220)
50C
0*4*4hMl%MMW*<vMl*l

Wi^WVrfJW

(air)

3 (S

3
IT*
mi

CuCI

(311)
mu

f*nnn>iiji

iimiit-nii

**

rt

n'

350C

MMHIWWm*'

Wy WHMWI

*hKiimtuHnAv^mflMimMMWHiW **Am
*^W*ti^^.,jjwy'vimfmwWW^^aff*1^!!!

Mo I

(N2)
420C
__iW
i

Uli T'

T"- "il

'"**-

^Tifl

_,

(N2)
t
r rrr

|
30

|
40

|-i

|
50

|'
60

20

2 Thta

(degree)
at different conditions:

Figure 6.1
alcohol at

XRD

patterns of precursors layer after drying


to

evaporation
in

of the

5o"C, heating

35o"C (Joint

in

air, heating
on

to

350C

in

nitrogen, heating

to

420C

nitrogen for

two to three minutes each.

Committee

Powder Diffraction Standards, ASTM,

Philadelphia, PA,

1967, Cards: CuCI, 06-0344; CuO, 05-0661; Cu, 85-1326; Mo, 42-1120)

low

solubility

and

early precipitation
for

in alcohol. If heated in air to in this

350C,

which is the

standard of the show If the

procedure

sample preparation
a

work, fumes

are

produced (burning
the CuCI
are

ethylcellulose)
a

until

metallic blue color appears. At this

stage

crystals

strong texturing
is done in

in the

(in)

direction and traces of the CuO the XRD

phase

detected.

heating

nitrogen atmosphere,

pattern

does not show


in

peaks

from the CuO


to
even more

phase

but elemental Cu

crystals form. Heating to 420C

nitrogen leads

pronounced peaks
20

from copper, while the CuCI

phase

is not detected

anymore. The low

region,

with its very broad

peak,

indicates

some

amorphous

phases, containing the

rest of the precursor material.

The observations of these

experiments
It

do

provide

detailed reaction be

path

of the

precursor

material

during preheating.

could

nevertheless

shown

that the

introduced salts may

partly change their anionic part

and oxidation state

(Table 6.2):

Table 6.2
Precursor

Modification of precursor salts

during preheating

in air

or

nitrogen

Modified Precursor

Preheating

Condition

ln(lll,CI3

lnl?)X

Air, 25C, Solvent evaporation


Air, 25C, Solvent evaporatiion

Cu(l)CI

Cu("CI, Cu(ll)0
3

air,3SOC
N 350C N 420C

Cu"CI, Cu<>
Cu(o)

6.1.4

Selenization of precursor

Chemical conversion of the


non-vacuum

preheated

precursor

samples
is

was

performed
up to
is

in the open

reactor

(4.2.4).

The

sample temperature
whereas the

ramped
source

56oC

and then
at
a

held

constant for

10

minutes

selenium
mass

maintained selenium

temperature

of

~350C during

selenization. The
run.

of the

evaporated

was

measured after each selenization

Typical

amounts of 20-30img

were

evaporated

and used to convert the 5x5 cm2 substrate


To test

sample.
precursors

the

conversion

of the
or

metal

separately,
were

precursor
in

pastes
to

containing only

the copper

only

the indium precursor

prepared
a

analogy

the CIGS pastes


conversion to the

(6.1.1).
Cu2.xSe

EDX measurements
and

(Table 6.3) suggest

nearly complete

ln2Se3 phases, respectively.

Table 6.3 Selenized


Cu Se

Composition of separately selenized copper nitrate and indium chloride precursor (EDX).
precursor

Cu(N03)2

(at %)

Selenized
In Se

InCI,

precusor

(at %)
39.7

62.3
37.7

60.3
o

Cl

71

Figure

6.2

Morphology Cu2.xSe (a)

and

InSe, (b) thin-films obtained by

non-vacuum

deposition

and

selenization of salt precursor materials.

Figure layers

6.2 shows the

morphology

of

separately

selenized Cu and In precursors. The


to selenized metal

show different

morphology
liquid

when

compared

nanoparticle
of thin films

layers (4.14).

Due to the

state of the precursor materials

deposition

is facilitated and the selenized

layers

are

smooth, while the copper selenide does not


isolated

coverall of the substrate and forms


dense and smooth surface with the In selenide

large

grains,

the indium selenide formed

clearly

visible

grains. However, only poor

adhesion of

layer

on

the Mo coated

glass

was

obtained and most of the converted

layer delaminated during

selenization.

Figure 6.3

Morphology
a

of

non-vacuum

deposited

and

selenized

CIGS

precursor

layer.

The

polycrystalline layer has

smooth surface

showing large grains.

The

morphology

of

selenized CIGS The

layer containing
has
a

the precursors

as

described in

6.1.1
as

is shown in

Figure 6.3.

compact layer top

smooth surface with

large grains
formed.

desired for efficient solar cells. On

of the surface few small

grains

are

72

Energy (keV)
Figure 6.4
EDX
a

Energy (keV)

spectra

of CIGS precursor and selenized

layer

on

glass

substrate

(therefore

showing

no

Mo- but

Si-signal).

In

Figure 6.4

the EDX
are

spectra

of the precursor

(after heating

to

350C)

and the

selenized

layer

shown. The conversion


a

introduced

selenium and

reduced the
in

chlorine and oxygen to selenized

level below the detection limit of EDX

(1-2

at

%)

completely

layers (Table 6.3).

Table 6.3 Element

Typical composition
Precursor

of precursor and selenized

layer

measured

by

EDX.

Precursor

layer

Selenized

layer

Changes

selenized

layer/

paste
Cu
In Ga
i

% relative to Cu 19-5 23.0


1

% relative to Cu
22.0
i

precursor

paste

reference
+

0.85
0.4 2-55
-

1.18 0.41

21.4

0.97 0.25
0

14%

7-9
49-5
-

5-5
0

-60% -100%
+

CI
Se

2.53
-

51.1

2.32

100%

The

Cu/(ln+Ga)

ratio of 0.8 remained constant


a

during selenization, although


paste

the

In/Ga ratios

changed nearly by
explained by
a

factor

from the precursor

to the selenized

layer.

This may be

the

inhomogeneous distribution

of In and Ga in the

layer (Figure 6.6)

and

difference in vapor pressure of intermediate

binary phases.
a

The

high

In content measured in the precursor

layer

may be

explained

with

matrix effect

73

resulting

from the formation of copper chloride


material.

crystals

in

an

amorphous

In and Ga

containing

CIS precursor CIGS precursor

(101)
I
' ' '

'

TI

1II

1-

'

'

'

'

'

'

'

'

15

20

25

30

35
2 Theta

40

45

50

55

60

(degree)

Figure 6.5

XRD of converted CIS and CIGS precursors. The addition of the Ga precursor results in

non-symmetrical

and broad CIGS

peaks. (Cu

Ka

radiation)

The conversion reaction and

incorporation

of Ga into CIS
a

was

investigated by 1/1.1)

XRD

analysis. Fig. 6.5


(Cu/ln/Ga

compares the XRD patterns of

CIS

(Cu/ln

ratio of

and CIGS

ratio of 1/0.8/0.6)

layers.
asymmetrical peaks
with their maxima shifted to

Addition of Ga leads to broad and

slightly higher peak


shift
is

20

values than

indicating
expected

change
for the

in the lattice

parameters (Figure 1.4). The


and
the

less

measured

layer composition

asymmetric, broad peaks suggest


For further

non-uniform electron

composition of the precursorfilm.


was

investigation, Auger

spectroscopy (AES) depth profiling

performed
Figure
6.6

to determine the elemental distribution in the

layer (Fig. 6.6).

As

seen

in

(a) except

for the

top part,

the

layer

is not formed of pure CIGS but contains

large

amounts of carbon. The carbon

signal

increases towards the back contact. Hence,


not efficient in

paste drying and evaporation of the organic binder was organic


metals material

removing all

the the

(Figure

6.6a and
a

6.7). Figure

6.6b

gives the depth profiles for


across

only.

The data suggests

Ga/ln
a

graded composition
low Ga content
on

the

layer thickness.
a

The CIGS

composition changes

from

the surface to

high

Ga

74

content at the back contact.

Such

grading

can

lead to

back surface field which is

favorable for thin absorber

layers (<i urn)


a

since

it reduces fill factor deterioration

(Lundberg
limiting

et

al., 2003). But

low Ga content

near

the p-n junction of the device is

the

Voc parameter

in solar cells

(6.1.6).

U-

/
y y

70-

(%}

/
< Ui fr

actions
u.

0
X

/
v

/'
\

\ Se^.'c
<=_Cu

*
J

1
0

30-

^^~~_

100-

'

'

'

T"

<

m"l

3
Time

a)
Figure 6.6
sputter
time
a

Sputter

(min)

b)
of
a

Sputter Time (min)


selenized precursor

Auger
carbon

electron

spectroscopy depth profile

layer. After

short

signal

is detected with

increasing intensity signals

towards the back

(a). Depth profile

showing the

metal concentrations

only,

while the C and Se

are

ignored (b).

Figure 6.7 polycrystalline

Cross section of selenized precursor CIGS

layer showing

double

layer

structure with the

layer formed

on

top of an amorphous carbon layer.

The cross-section SEM precursor

micrograph picture (Figure 6.7)


shows
a

is taken from

selenized is

layer.

The

figure

double

layer structure,

where the

bottom-layer

75

responsible analysis
on

for the carbon


a

signal

in the AES

depth profile (Figure 6.6).

EDX

composition
which

focused

point

in the lower

layer

showed C, O, Se and Ga
a

signals, layer

corresponds

well to the results of the AES


nm

depth profile. Finally,


compared
for
to 1.5-2

CIGS

thickness

of about 6oo

in average is very thin

urn

which is considered

optimal

for total

light absorption

and

required

high efficiency solar cells.

6.1.5

CIGS

growth

mechanisms
of the precursor and
an

Doctor-blade the desired

deposition

paste containing
binder

dissolved metal salts in leads to


a

stoichiometry
precursor

organic

material

thin

and

homogenous
are

layer

upon

evaporation

of the alcohol solvent.


an

Crystals

of CuCI

formed in the

layer whereas

In and Ga remain in in ambient

amorphous

state in the

organic organic

matrix. When

heated to

300-350C

atmosphere,

some

of the

material

evaporates although large

amounts remain in the precursor

layer.

In addition

to the CuCI

crystals,

small

Cu20 crystals form due

to reaction with ambient oxygen. In

and Ga remain embedded in the

organic

matrix.

During selenization,
CIGS

the selenium The

vapor reacts with the metal


conversion

on

top of the layer and forms the


of

compound.

taking place

at

temperature

500C

allows most of the metal atoms to


react with in-diffused selenium

diffuse to the surface of the and


are

organic layer,

where

they

added to the formed CIGS film. Because of the difference in the reaction

kinetics of CIS and CGS in the first

(Marudachalam

et

al., 1997), preferably the

CIS

compound

forms
in the

phase.

This leads to

depletion

of indium and accumulation of


a

gallium

remaining precursor layer


back.

and the formation of CIGS with of the

higher

Ga content in the

Complete decomposition
the

remaining organic layer

which resulted from

burning
with its
CIGS

ethylcellulose

could not be achieved with this method. The residual

layer

high

carbon content is

trapped

between the

molybdenum

back contact and the

layer.
an

Having

amorphous organic (and

therefore

possibly an electrically insulating) layer

between the CIGS absorber and the back contact may be detrimental for the solar cell

device. However, device characteristics have shown

encouraging results (6.1.6).


was

Spray deposition

of the precursor salts in cellulose-free solutions


a

tested in order

to avoid the formation of

carbon

layer (6.2).

76
6.1.6 Solar cell device characterization
Selenized precursor
estimate the

layers

were

finished to

complete
a

solar cells
on a

according

to

5.1.

To

homogeneity of the
21

selenized
area

layer,

solar cell

5x5 cm2 substrate


and which

was

scribed into

small cells with

an

of 0.6 cm2 each

(Figure 6.8)

were

individually
conditions.

characterized

by

l-V measurements under simulated

AM1.5 illumination

Figure 6.9

shows the solar cell


values
are

parameters

of the cells. Efficiencies vary


over

between 2.4 and

4.9%. While lsc

nearly

constant

the whole

sample (and

may

only

reflect the

inhomogeneity
the FF and

of illumination from the centered show

single-lightparameters

source sun

simulator)

mainly the Voc

larger variations.

The

are

more

consistent in the direction of the blade movement


to column

during deposition (from


paste composition
blade

column
are

3)

than

perpendicularly
may be found in

to it. As differences in
a

not

probable,

the

reason

non-uniform

(manual)

deposition

resulting

in different

layer thicknesses.

Figure 6.8

5x5 cm2 substrate with


rows

21

cells of 0.6 cm*

area

arranged

in three columns

(11

mm

spacing)

and 7

(5.5

mm

spacing).

77

a)

b)

c) Figure 6.9
over

d)
The cell
area.

parameters Voc (a), Jsc (b),

FF

(c) and efficiency (d) show moderate homogeneity

the substrate

Degradation

of the parameters is observed towards the substrate

edges.

Table 6.4

Average values with standard deviation of the cell parameters.


Jsc (mA/cm3)

Voc(mV)
340
34

FF(%)
43.6
3-5

Efficiency (%)
3.52 072

average value

23.5
i.i

standard deviation

Maximum efficiencies above 6%

were

obtained

on

smaller

area

cells.

Figure

6.10

shows the IV

curve

and cell
is

parameters

of the best cell with

an area

of 7.3 mm2. The

higher efficiency of 6.7%


is lowfora CIGS

mainly

due to the

high

FF above 60%. The

Voc with 404

mV

compound containing

6 at %Ga.

78

730
FF
:

61 %

/
/

.20
c

Voc: Jsc:

404 mV 27.2

mA/cm2
:

/
/

I10
[).2
-0.1

Efficiency

6.7 %

0.1

0.2

0.3

0/4

0.

Voltage (V)
-10

-20

-30

Figure
vacuum

6.io

Current-Voltage
CIGS

characteristics of the best efficient solar cell obtained from


on

non-

deposited

layers

based

inorganic

salts.

External

quantum efficiency (EOE)


device to different
EOE

measurements were

performed

to

identify

the
A
a

sensitivity of the

wavelengths
a

of the illumination

spectrum.
with

comparison
conventional

of the

diagram

of

non-vacuum

processed

solar cell

co-evaporated

solar cell

(Figure 6.n)

reveals

significant

losses in the

o z

400

600

800

1000

1200

Wavelength (nm)

Figure processed

6.11

Comparison of the normalized external quantum efficiencies for


on

non-vacuum

CIGS solar cell based


non-vacuum

inorganic salt precursors and

conventionally vacuum-processed
losses in the middle and
cell is at
i20onm

CIGS solar cell. The

solar cell shows

pronounced absorption
non-vacuum

long
a

wavelength region.
band gap energy of

The cut-off

wavelength

of the

processed

indicating

only -1

eV.

79

medium to

long wavelength region (500-noo nm), completely


absorb the AM 1.5 solar
1200 nm

which confirms that the CIGS


The band gap for
a

layer

is too thin to

spectrum.

cut-off
with
a

wavelength

of

approximately

corresponds

well to the CIS

compound
by

band gap of 1.04 eV. The addition of Ga should increase the band gap

0.2-0.3 eV DUt

the observations indicate


AES

inhomogeneous

Ga

distribution, which
The

is

supported by

the

depth profile
a

measurements

(Figure 6.6b).
a

top part

of the

layer consists

of CIGS

with

low concentration of Ga and

band gap neartothe CIS

compound.

6.2

Electrospray
the spray method,

Using

deposition

of low

viscosity liquids
a

as

metal salts dissolved

in methanol is

possible. Therefore,

the addition of

cellulose binder material is not

necessary. The aim is to reduce the carbon content in the

paste formulation

in order to

prevent
the CIGS

the formation of

carbon

layer

between the

molybdenum

back contact and

layer. layers
would allow
a

Carbon free CIGS and

better contact for CIGS solar cells in substrate

superstrate configuration.
on

In contrast to substrate solar

cells, where the


is formed
a

CIGS is

deposited

the

molybdenum

back contact and the p-n cells form the


an

junction

near

the and

surface of the CIGS the thereon

layer, superstrate

junction

between

TCO

layer

deposited

CIGS. The formation of

amorphous
properties

carbon

layer

in the

junction region would completely deteriorate

the cell

and carbon free CIGS

layers

are a

prerequisite. pastes
were

Precursor

prepared accordingly

to

6.1.1

but

only containing

the

dissolved salt precursor in methanol and

1,2-propanediol (Ga

was

not added at this

stage).
was

In order to
on a

prevent incorporation

of moisture into the precursor

layer, spraying
After

done

heated

glass

substrate

(iooC)

in

inert

nitrogen atmosphere.

deposition
at

the

sample was quickly transferred

to the selenization reactor and selenized

550C. Figure
6.12 shows the XPS

depth profile

of

selenized precursor The precursor

layer deposited

on a

molybdenum

coated substrate

by electrospray.

paste

did not contain the

gallium signal
more

precursor. Because the Cis


was

signal gets perturbed by the by


the
area

selenium

LMM4 Auger
which is

the carbon content


difficult to

evaluated

of the C KLL

Auger peak,

quantify.

The

resulting

carbon concentration may therefore vary up to

factor of three. However, the C KLL

peak

is very low and the indicated carbon

8o

concentrations

are

covered

by

noise

(0-2 at%). Hence,

the carbon contamination

was

reduced below the detection limit of XPS.

70

60

50Mo
.Ml*.
>

40-

V.
V Se*
***

0.5

1.5

Layer Depth (pm)

Figure

6.12

XPS

depth profile: Electrospray deposition of

salt precursor and selenization allows

carbon free

compound formation.

The reduction of the carbon content should also be reflected carbon blade of the

by

elimination of the
in the doctoron

layer

between the CIGS and the substrate which The

was

present

deposited samples (Fig. 6.7). layer.

morphology

is dominated

by large grains
a

top

Surface and cross-section SEM

images (Fig. 6.13)


on

show

1.5

urn

thick smalla

grained
surface

bottom

layer

with

large crystals towering layer

the surface, which induce formed.


Hence the

high

roughness.

No

distinct carbon

was

electrospray

deposition

method is suitable to eliminate the carbon

layer

which is otherwise formed

in the doctor-blade

deposited layers.
of the selenized surface.

Compositional analysis
the elements
across

layers

showed non-uniform distribution of


to

the

sample

Contrary

electrosprayed
copper rich

metal

nanoparticle sample
of

layers (3.2.1)
center

the

tendency
poor

for salt precursors

was a

region

in the

and

copper

composition
may be

in

the

edges. However,

this

problem

compositional non-uniformity

overcome

by a moving

substrate.

8i

Figure 6.13
precursor

SEM

images: Morphology

and

cross-section

of

selenized

electrospray-deposited

layer.

Solar cells

produced

from
0.1

electrosprayed
or

and selenized salt precursors showed either

very poor efficiencies

(<

%)

just

an

ohmic behavior

produced by

shunts in the cell.

Considering

the CIS structure shown in

Figure 6.13

it is assumed that the

large

Cu-rich

crystals

are

responsible

for

shunting

the cells. The formation of such Cu-rich

crystals growth

results from of

non-homogenous crystals during

precursor precursor will

layers

which may be

explained by
substrate

the

large

CuCI

deposition. Higher
the precursor film

temperatures

during electrospray deposition


the

keep

completely dry preventing


Further

sprayed layer

from

recrystallization

to

larger

salt

particles.

investigation

of

the precursor spray process is necessary to find

optimized deposition

conditions.

82

6.3 Conclusions
In

spite

of the notable cell efficiencies, many

aspects of the

conversion reaction from

the precursors to the CIGS

compound

remain unclear. For instance, nitrate precursors

usually
(Kaelin

result in metal-oxides upon


et

heating (Beck

and Cocivera,

1996).

previous study

al., 2004a) revealed difficulties in converting the stable compound ln203, and
into their selenide
are

probably Ga203,
no

phase using selenium

vapors. As

seen

in

Figure 6.1,
organic

oxide

phases

detected upon

preheating the

precursor to

250C
a

in air. The

matrix may

prevent

the formation of metal oxides and thus have

beneficial effect

by

reducing the

conversion time.

The addition of cellulose material in the precursor

paste

allows

adjustments
CuCI

in

paste
upon

rheology

for

optimal coating
of the

results and avoids formation of

larger

crystals

evaporation
However,
obtained
an

solvent, which otherwise deteriorate local film stoichiometry.


carbon

amorphous

layer

forms between the Mo and the CIGS that


it is
i

layer.

The

cell

efficiencies

demonstrate

the pm

carbon

layer

is

not

completely

detrimental for cell

performance, although

thick, porous, and amorphous and

although by

it increases the series resistance of the cell. Thicker precursor

layers deposited

doctor-blade resulted in poor adhesion of the CIGS/carbon structure to the Mo


are

surface; hence further experiments


structures
on

required

to

improve the

adhesion. Carbon-free
in

were

prepared by electrospraying
heated to iooC followed
and had

cellulose-free

pastes

nitrogen
CIGS

ambient
were

substrates

by

selenization. Such

layers

extremely rough
Ga
can

irreproducible compositions.
into the selenized
near

be

incorporated

layer

but its distribution

has to be

improved (higher

Ga concentration

the CdS buffer

layer)

to achieve
or an

higher Voc
additional

values. Better results may be obtained with

longer

selenization times

annealing step (Marudachalam


CuCI2 and
were

et

al., 1997).
as

ln(N03)3

have also been tested

precursors in

paste formulations

and

successfully

converted to the CIGS

compound, but

the obtained cell efficiencies

were

lower than for the described paste formulation. However, further

investigation

is

needed to

optimize

the

paste chemistry and

conversion processes.

83

7 Research recommendations
A novel CIGS absorber formation process
was

developed, which,

due to its

simplicity

and the exclusion of hazardous reactants, allows the


solar cells. Cost-effectiveness is
a

manufacturing

of low-cost CIGS
costs and the remains the

trade-off between direct

manufacturing

efficiency

of the

produced

solar cells.

Optimization

of the cell

efficiency

prior development target.

Comparing nanoparticles
conversion

the

results

obtained

with

metal/metal-oxide/metal-selenide

to those with

inorganic

precursor salts, it results that, under the


a

given

conditions, the salt precursor based cells have


solar cells. The

higher potential

to

yield

high-efficiency
research and The

following topics

should

be considered for further

development activities:
reaction

chemical conversion

of the

precursor

layer containing chlorides,


to know which volatile

nitrates and cellulose is not

fully

clear

yet.

It is

important

products
chloride

form in the selenium

atmosphere.
(in
to

An

investigation

of

remaining

oxide and
or

impurity traces

and their location would

the CIGS

crystals,

at

grain boundaries,

in the carbon

bottom-layer)

help

optimize

selenization conditions

(time, layers

temperature
but lacked of blade

ramps, Se vapor

pressure). Electrosprayed layers yielded

carbon-free

compositional
allows

and structural control. The cellulose binder used in doctorcontrol of the

pastes
an

good

composition

and the

layer
a

structure

but

introduces which

amorphous

carbon

layer

into the cell structure.

Finding

binder material

undergoes

controlled

decomposition during

the selenization treatment would

solve this

problem.
process is

The CIGS formation

lacking

Ga/ln

composition
are

control for band gap

engineering
near

in the CIGS

layer.

Selenization processes

known to accumulate

gallium

the

molybdenum
region.

back contact

yielding mainly

CIS

compound

in the CdS-CIGS

interface

An additional sulfurization treatment

introducing

sulfur into the

top
cell
an

layer

would increase the band gap in the junction thanks to


a

region,

which leads to

higher during

efficiency

higher Voc. Therefore,


step
and

the diffusion behavior of sulfur structures should be

additional sulfurization

resulting band-gap
of

investigated.
cells,

In order to follow the

concept

completely
has to be

non-vacuum

processed
a

CIGS solar

the

sputter-deposited
TCO

TCO

layer

replaced by
as

low-cost,

non-vacuum

processed

layer.

Similar precursor

paste chemistry

used for CIGS

synthesis

84
should be

investigated

for TCO

deposition

processes,

using

oxidation treatments and

instead of selenization. The TCO therefore the aim would be to

layer

is one of the last

layers applied (1.3.2)


to
a

bring

the reaction

temperatures
and

minimum

(below

250C,

to avoid elemental diffusion between

layers

junction degradation) while


TCO

maintaining good transparency and

electrical

conductivity of the

layer.

85

References

Adams, W. G. and Day,

R. E.

(1877) The

action of

light

on

selenium. Proc

Roy

Soc A 25,113.

Alberts,

V.

(2002) Comparison

of material

properties

of CulnSe2 films
va

produced by Phys
7

reaction of metallic

alloys to

H2Se/Arand elemental Se

pou r.Jpn J Appl

41,

518-523.
Arita, T., Suyama, N., Kita, Y., Kitamura, S., Hibino, T., Takada, H., Omura, K., Ueno, N. and

Murozono,
method.

M.

(1988) CulnSe2

films

prepared by screen-printing
Las

and

sintering

Twentieth-IEEE-Photovoltaic-Specialists-Conference,

Vegas, NV, USA,

1650
ASTM

(1999)

American Solar for

Society for Testing Spectral


a

and
at

Materials, G173-03 Standard Tables for


Air Mass 1.5:
Direct

Reference

Irradiance

Normal

and

Hemispherical
Banger,
K.

37

Degree Tilted
J. A.,

Surface. ISO

9845-1,1992

K.,

Hollingsworth,

Harris,

J.

D., Cowen, J., Buhro, W. E. and Hepp, A. F.

(2002) Ternary single-source

precursors for

polycrystalline

thin-film solar cells.

Appl Organomet Chem 16, 617-627.


Basol, B. M. (2000) Low
cost

techniques for the preparation

of

Cu(ln,Ga)(Se,S)2

absorber

layers.

Thin Solid Films 361, 514-519.

Beck, M. E. and Cocivera,

M.

(1996)

Thin-film copper indium diselenide

prepared by

selenization of copper indium oxide formed

by

spray

pyrolysis.

Thin Solid Films 272,

71-82.

Becquerel,

E.

(1839)

Mmoires

sur

les effets

lectriques produits

sous

l'influence des

rayons.

Comptes Rendues
R. N.,

9,

561-567
Batchelor, W., Contreras,
M.

Bhattacharya,
R.

Hiltner,

J. F.,

A., Noufi, R. N. and Sites, J.


from

(2000) 15.4% Culn^Ga^Se.-based photovoltaic


361,396-399.

cells

solution-based

precursor films. Thin Solid Films

86

Brown, B.

J. and

Bates, C. W. (1990) Similarities in the Chemical Mechanisms of CulnSe2

and CdS Thin-Film Formation 305.

by Chemical Spray Pyrolysis.

Thin Solid Films 188, 301-

Brummer, A., Honkimaki, V., Berwian, P., Probst, V., Palm, J. and Hock, R.
Formation of

(2003)
in

CulnSe2 by the annealing of stacked elemental layers-analysis by


Thin Solid Films 437, 297-307.

situ

high-energy powder diffraction.


M., Fuller,
C. S. and

Chapin,
for

D.

Pearson, G. L

(1954)

A New Silicon P-N Junction Photocell

Converting Solar
R. W. and

Radiation into Electrical PowerJ Appl

Phys

25,

676-677.

Cheary,

Coelho, A. A. (1998) Axial divergence in

conventional

X-ray powder

diffractometer. I. Theoretical foundations. J Appl

Ciysta/logr^, 851-861.

Choy,

K. L

(2001) Processing-structure-property

of

nanocrystalline

materials

produced
S16,139-

using
145.

novel and cost-effective ESAVD-based methods. Mat Sei

Eng

C-Bio

Dubois,

F. and

Belleville, F. (1988) Selenium: physiologic role and value in human


Pathol Biol

pathology.
Duchemin, S.,

36,1017-1025.
A.

Bougnot, J., El-Ghzizal,

and

Belghit,

K.

(1989)

Studies

on

the

improvement of sprayed CdS-CulnSe2 solar cells. Proceedings of the gth European


Photovoltaic Solar

Energy Conference, Freiburg,

West

Germany,

Kluwer Academic

Publishers, Dordrecht, Netherlands, 1217.

Eberspacher,

C, Pauls, K. and Serra, J.

(2000)

Nonvacuum

techniques

for

fabricating
15-22

thin-film CIGS.

Proceedings of 28th

IEEE Photovoltaic

Specialists Conference.

Sept.

2000

Anchorage,

AK, USA, IEEE, Piscataway, NJ, USA.

Eberspacher,
of

C, Pauls, K. and Serra, J.


materials. ll-VI

(2001) Improvements

in

non-vacuum

deposition

particulate

Compound

Semiconductor Photovoltaic Materials

Symposium.

Mater. Res. Soc,

Warrendale, PA, USA, 570.

87

Eberspacher, C, Pauls, Proceedings of

K. and

Serra, J.

(2002)

Non-vacuum

processing

of CIGS solar cells.

the

29th
1792.

IEEE Photovoltaic

Specialists Conference, LA, USA, IEEE,

Piscataway, NJ, USA, Eberspacher, C, Pauls,

K. L and Serra, J. P.

(2003)

Non-vacuum thin-film CIGS modules. Mater. Res.

Proceedings of the Compound Semiconductor Photovoltaics Symposium,


Soc, Warrendale, PA, USA, 483.

Ederth, J., Johnsson, P., Niklasson,


van

G. A.,

Hoel, A., Hultaker, A., Heszler, P., Granqvist,

C. G.,

Doom, A. R., Jongerius, M. J. and


of thin films
no.

Burgard,

D.

(2003)

Electrical and

optical

properties
RevB

consisting

of

tin-doped

indium oxide

nanoparticles. Phys

68,

art.

155410.

Einstein,

A.

(1905)

ber

einen

die

Erzeugung

und

Verwandlung

des

Lichtes

betreffenden heuristischen

Gesichtspunkt. Annalen
C.

der

Physika, 132-148. (1983)


and

Gombia,

E.,

Leccabue,

F.,

Pelosi,

and

Seuret,

D.

Vapor

Growth,

Thermodynamical Study

and Characterization of

CulnTe2

CuGaTe, Single-

Crystals. Journal of Crystal Growth 65,391-396.


Green, M. A.,
tables

Emery, K., King,

D.

L, Igari, S. and Warta, W.

(2004)

Solar cell

efficiency

(Version 24). Prog

Photovoltaics 12,

365-372.

Guimard, D., Bodereau, N., Kurdi, J., Guillemoles, J. F., Lincot, D., Grand, P. P., BenFarrah,
M., Taunier, S., Kerrec, O. and Mogensen, P.

(2003)
the

Efficient CIGS solar cells

prepared by electrodeposition. Proceedings of


Photovoltaic

3rd

World

Conference

on

Energy Conversion,

V0I.1 Osaka japan, Arisumi

Printing Inc, Japan.

515-

518. Haba, B., Bates,


W., Jr., Brown, B. J., Ndlela, Z. and Smestad, G. (1990) Characterization
defects
in the

C.

of structural

presence of

ln203

in

CulnSe2 prepared by spray

pyrolysis. Solar Cells 28,177-183.


Herberholz, R., Rau, U., Schock,
Benz, K. W. and Cahen, D.
formation in
H. W.,

Haalboom, T., Godecke, T., Ernst, F., Beilharz, C,


Phase

(1999)
Eur

segregation,

Cu

migration

and

junction

Cu(ln, Ga)Se2.

Phys J-ApplPhys 6,131-139.

88

Hirpo, W., Dhingra, S., Sutorik, Copper


Indium

A. C. and
-

Kanatzidis, M. G. (1993) Synthesis of Mixed


Precursors for the Photovoltaic

Chalcogenolates

Single-Source

Materials

CulnQ2 (Q=S, Se). J Am Chem

Soc 115,1597-1599.

Houriet, R., Vacassy, R. and Hofmann, H.


new

(1999) Synthesis

of

powders

and films

using

laser ablation

technique.
R. and

Nanostruct Mater n,

1155-1163.

Jabbour,

G. E.,

Radspinner,

Peyghambarian,

N.

(2001)

Screen

printing

for the
in

fabrication of

organic light-emitting

devices. IEEE Journal

of

Selected

Topics

Quantum Electronics 7,769-773.

Kaelin, M., Meyer, T., Kurdesau, F., Rudmann, D., Zogg, H. and Tiwari, A. N. (2003) Low
cost

Cu(ln,Ga)Se2

absorber
World

layers

from
on

selenization Photovoltaic

of

precursor

materials. V0I.1

Proceedings of the 3rd


Osaka, Japan,
Arisumi

Conference

Energy Conversion,

Printing Inc, Japan.

404-405.

Kaelin, M., Rudmann, D., Kurdesau, F., Meyer, T., Zogg, H. and Tiwari, A. N. (2003) CIS and
CIGS

layers from

selenized

nanoparticle

precursors. Thin Solid Films 431-432, 58-62.

Kaelin, M., Zogg, H., Tiwari, A., Wilhelm, O., Pratsinis,

S. E.,

Meyer,

T. and

Meyer,

A.

(2004a) Electrosprayed
457.

and selenized Cu/ln metal

particle

films. Thin Solid Films

391-396.

Kaelin, M., Rudmann, D., Kurdesau, F., Zogg, H., Meyer,


cost CIGS solar cells

T and Tiwari, A. N.

(2004b)

Low-

by paste coating

and selenization. Thin Solid Films in press.

Kampmann, A., Rechid, J., Raitzig, A., Wulff, S., Mihhailova, M., Thyen,

R. and

Kalberlah,

K.

(2003) Electrodeposition Compound


Semiconductor

of

CIGS

on

metal

substrates.
22

Proceedings of
25

the
San

Photovoltaics

Symposium,

April

2003

Francisco, CA, USA, Mater. Res. Soc, Warrendale, PA, USA, 483.

Kapur,

V. K.,

Basol,

B. M. and

Tseng,

E. S.

(1987)

Low-Cost Methods for the Production of

Semiconductor-Films for

CulnSe2/CdS

Solar-Cells. Solar Cells 21,

65-72.

89

Kapur,

V. K.,

Bansal, A., Le,


on

P. and

Asensio,

O. I.

(2003)

Non-vacuum

processing

of

Culn,.

xGaxSe2 solar cells

rigid

and flexible substrates

using nanoparticle precursor

inks. Thin Solid Films 431, 53-57.

Krunks, M., Kijatkina, O., Rebane, H., Oja, I., Mikli,

V. and

Mere, A. (2002) Composition of

CulnS2 thin films prepared by spray pyrolysis. Thin Solid Films 403,71-75.
Kurdesau, F., Kaelin, M., Zalesski, V. B., Kovalewsky, V. I., Gremenok, V. F., Zaretskaya, E. P.
and Tiwari, A. N.

(2004)

In situ

resistivity

measurements

during

selenization

process. Thin Solid Films 451-452, 245-249.

Kushiya,

K., Ohshita, M., Hara, I., Tanaka, Y.,


issues
on

Sang, B., Nagoya, Y., Tachiyuki,


of 30
cm x

M. and

Yamase, 0. (2003) Yield

the fabrication

30

cm-sized

Cu(ln,Ga)Se2-based
Lewerenz, J. and

thin-film modules. Sol

Energ Mat Sol C 75,171-178.

Jungblut,

H.

(1995).

Photovoltaik 31

Grundlagen

und

Anwendungen,

Springer Verlag Berlin, Heidelberg,

Lide, D. R. (2000). CRC Handbook of Chemistry and Physics


chemical and

ready-rfre nee

book of

physical

data. Boca Raton, CRC Press.

Lundberg, 0., Bodegard, M., Malmstrom, Cu(ln,Ga)Se2


thickness and Ga

J.

and
on

Stolt, L (2003) Influence of the


solar cell

grading

performance. Progress

in

Photovoltaics: Research and Applications

11,77-88.
Nasrallah, T. and Belgacem, S. (2003)

Marsillac, S., Zouaghi, M. C, Bernede, J. C,


Evolution of the

Ben

properties of spray-deposited CulnS, thin films with postSol

annealing treatment.

Energ
R. W.,

Mat Sol

C76,125-134.
J.

Marudachalam, M., Birkmire,

Hichri, H., Schultz,

M., Swartzlander, A. and Alin

Jassim, M. M. (1997) Phases, morphology, and diffusion

Cul^Ga^Se., thin

films. J

Appl Phys 82, 2896-2905.


Nelson, J. (2003). The Physics of Solar Cells. Imperial College Press, London

90

Orbey, N., Norsworthy, G., Birkmire,


the formation of

R. W. and
in
a

Russell,

T. W. F.

(1998)

Reaction
reactor.

analysis
Progress

of
in

CulnSe2 films

physical

vapor

deposition

Photovoltaics: Research and Applications

6,79-86.

Palm, J., Probst, V., Brummer, A., Stetter, W., Tolle, R., Niesen, T. P., Visbeck, S., Hernandez,
0., Wendl, M., Vogt, H., Calwer, H., Freienstein, B. and Karg, F.

(2003)

CIS module

pilot processing applying


area

concurrent

rapid

selenization and sulfurization of

large

thin film precursors. Thin Solid Films 431,514-522.

Pamplin,

B. and

Feigelson,

R. S.

(1979) Spray Pyrolysis


Thin Solid Films

of

CulnSe2 and Related Ternary

Semiconducting Compounds.

60,141-146.
B.

Powalla, M., Hariskos, D., Lotter, E., Oertel, M., Springer, J., Stellbogen, D., Dimmler,
and

Schaffler,

R.

(2003) Large-area

CIGS modules: processes and

properties. Thin

Solid Films 431, 523-533.

Ramanathan, K., Contreras, M. A., Perkins,

C.

L, Asher, S., Hasoon, F. S., Keane, J.,

Young,
of
n,

D., Romero, M., Metzger, W., Noufi, R., Ward, J. and Duda, A.

(2003) Properties

19.2% efficiency ZnO/CdS/CulnGaSe2 thin-film solar cells. Prog Photovoltaics


225-230.

Ramanathan, K., Teeter, G., Keane, J. C. and Noufi, R. (2004) Properties of High Efficiency
CulnGaSe2

(CIGS)

Thin Film Solar Cells. Thin Solid Films in press.

Schulz,

D.

L, Curtis, C. J., Flitton, R. A., Wiesner, H., Keane, J., Matson, R. J., Jones, K. M
P.

Parilla,
spray

A., Noufi, R. and

Ginley,

D. S.

(1998)

Cu-ln-Ga-Se solar cell

nanoparticle

colloids

as

deposition

precursors for

Cu(ln,Ga)Se2

materials. Journal

of

Electronic Materials 27,433-437.

Schwartz, R. W., Schneller, T. and Waser,


electronic oxide films. Cr Chim

R.

(2004)

Chemical solution

deposition

of

7,433-461.

Shockley,

W. and

Queisser, H. J. (1961) Detailed Balance Limit of Efficiency of P-N

Junction Solar Cells. J

Appl Phys

32, 510.

91

Siefert,

W.

(1984)

Corona spray

pyrolysis:

new

coating technique 120,267-274.


the

with

an

extremely

enhanced

deposition efficiency.
on

Thin Solid Films

Smith, W. (1873) Effect of Light


Nature 20, 303.

Selenium

during

Passage

of

an

Electric current.

Spath, M., Sommeling,

P. M.,

van-Roosmalen, J. A. M., Smit, H. J. P., van-der-Burg, N. P. G.,

Mahieu, D. R., Bakker, N. J. and Kroon, J. M. (2003) Reproducible manufacturing of

dye-sensitized

solar cells

on a

semi-automated baseline.

Progress

in Photovoltaics:

Research and Applications 11, 207-220.

Stanbery,

B. J.

(2002) Copper

indium selenides and related materials for

photovoltaic

devices. Critical Reviews in Solid State and Materials Sciences 27,73-117.

Steudel,

R.

and

Strauss

E.

M.

(1987)

Selenium

Homocycles

and

Sulphur-Selenium
I.

Heterocycles.

The

Chemistry of Inorganic

Homo- and

Heterocycles,

Haiduc, D.B.

Sowerby (eds.),
Su, B. and

Vol. 2, Academic Press, London,

769-792 jet deposition


of CdS, CdSe

Choy,

K. L

(2000)

Electrostatic assisted aerosol

and ZnS thin films. Thin Solid Films

361,102-106.

Svec, M., Roosen, A., Schmidt, M., Munstedt, H., Betz, T. and Koppe, F. (2002)
and local flow behaviour of ceramic slurries in the

Rheology
Mat

tape casting

head.

Key Eng

206-2, 219-222.

Taylor,

G.I.

(1964) Disintegration

of Water

Drops

in Electric Field. Proc R Soc Lon Ser-A

280,383-397Vanzomeren, A. A., Kelder, E. M., Marijnissen, J. C. M. and Schoonman, J.


Production of Thin-Films of

(1994)

The

LiMn204 by ElectrosprayingJ Aerosol Sei


Land

25,1229-1235.

Vervaet, A.,
films.

Burgelman, M., Clemminck,

Capon,

J.

(1989)

Screen

printing

of

CulnSe2
25-29

Proceedings of the 9th European


West

Photovoltaic Solar

Energy Conference,

Sept. 1989 Freiburg,


Netherlands, 480-483.

Germany,

Kluwer Academic

Publishers, Dordrecht,

92

Weber, A., Manner, R., Waser, R. and


microstructure and electrical

IversTiffee, E. (1996) Interaction


of
screen

between

properties

printed

cathodes in SOFC

single

cells. Denki

Kagaku 64, 582-589.

Wu, X., Keane, J. C, Dhere, R. G., C, D., Albin, D. S., Duda, A., Gessert, T. A., Asher, S., Levi,
D. H. and

Sheldon,

P.

(2002) 16.5%

efficient CdS/CdTe

polycrystalline thin-film
Energy Conference, II,

solar 995-

cells.
1000.

Proceedings of the ifh European

Photovoltaic Solar

Zweibel, K. (1999) Issues in thin film PV manufacturing

cost reduction. Sol

Energ Mat Sol

C59,i-i8.

93

Abbreviations

AES

Auger electron spectroscopy


Air
mass

AMi.5

factor 1.5

bp
CBD CIGS CIS EDX

Boiling point temperature


Chemical bath

deposition
Ga, Se
;
;

Compound formed by Cu, In, Compound formed by Cu,

usually Culn^Ga^ej

In, Se

usually CulnSe2

Energy-dispersive X-ray analysis Band-gap


External

Eg
EOE
l-V

energy

quantum efficiency

Current-voltage
Current

J-V

density-voltage
density

Jsc
MBE

Short circuit current Molecular beam

epitaxy

mp
PV SCR

melting point temperature


Photovoltaic

space

charge region
electron

SEM SLG
TCO
UHV

Scanning

microscopy

Soda-lime

glass
oxide

transparent conducting
Ultra

high

vacuum

Voc WP

Open
Watt
1000

circuit

voltage
Unit for

peak,

indicating

the

output power

of solar cells

at

W/m2 solar irradiation

XRD
XPS

X-ray diffraction X-ray photoelectron spectroscopy

94

/ Seite Leer Blank leaf

95

Curriculum vitae

Ma re-Roi and Kaelin

Date of birth:

i6

July 1975

Place of birth:

Bern, Switzerland
Swiss

Nationality:

May 2000

present

Research and

Teaching Associate,
Solid
State

Thin Film ETH

Physics Group,
(Swiss
Federal

Laboratory

of

Physics,

Institute of Technology)

Zrich, Switzerland

Oct.!995

Apr.

2000

Studies in Microtechnical
Institute of

Engineering,

EPF

(Swiss

Federal

Technology) Lausanne,

Switzerland

Diploma

Thesis in semiconductor device

processing

and

characterization at ETH Zrich Jul.

1998

Oct.

1998

Student

Exchange for Technical Experience, Celaya, development at Campbell's de


Army

Mexico
Mexico

Database front-end Jul.


Oct.

1995-Oct. 1995

Military

Service in the Swiss

1989-J un.

1995

Mathematisch
Bern

Naturwissenschaftliches

Gymnasium

in

Apr. 1981 -Jul. 1989

Primary

School in Moosseedorf, Switzerland

96

Refereed articles

Kaelin M., Rudmann D., Kurdesau F., solar cells

Zogg H., Meyer T.,

Tiwari A.N.

(2004)

Low-cost CIGS

by paste coating

and

selenization, Thin Solid Films,

in press.

Rudmann D., Brmaud D., da Cunha, A.F.,


A.N.

Bilger, G.,

Strohm A., Kaelin M., for


CIGS

Zogg H., Tiwari,


at

(2004)

Sodium

incorporation strategies

growth

different

temperatures,

Thin Solid Films, in press.

Kaelin, M., Rudmann, D., Tiwari, A.N. (2004) cells, Solar Energy 77(6): 749-756.

Low cost

processing of

CIGS thin film solar

Kaelin, M., Zogg H., Tiwari, A., Wilhelm, O., Pratsinis,

S.

E., Meyer, T., Meyer, A.

(2004)

Electrosprayed
396.

and selenized Cu/ln metal

particle films,

Thin Solid Films 457: 391-

Kurdesau, F., Kaelin, M., Zalesski, V. B., Kovalewsky,


Tiwari, A. N.

V.

I., Gremenok, V. F.,

Zaretskaya,

E. P.,

(2004)

In situ

resistivity

measurements

during

selenization process,

Thin Solid Films 451-52: 245-249.

Rudmann, D., da Cunha, A. F., Kaelin, M., Kurdesau, F., Zogg, H., Tiwari,

A.

N.,

Bilger,

G.

(2004) Efficiency
Na

enhancement of

Cu(ln,Ga)Se2
84(7):

solar cells due to

post-deposition

incorporation, Applied Physics

Letters

1129-31.

Romeo, A.,

Terheggen, A., Abou-Ras,

D. Batzner, D. L,

Haug,

F.

J., Kaelin, M., Rudmann, D.,


and CdTe solar cells.

Tiwari, A. N.

(2004) Development 12(2-3):

of thin-film

Cu(ln,Ga)Se2

Progress

in Photovoltaics

93-111.

Kaelin, M., Rudmann, D., Kurdesau, F., Meyer, T., Zogg, H., Tiwari,
CIGS

A. N.

(2003)

CIS and

layers from

selenized

nanoparticle

precursors, Thin Solid Films 431-432:

58-62.

Rudmann, D., Bilger, G., Kaelin, M., Haug, F. J., Zogg, H., Tiwari, A. N. (2003) Effects of NaF

coevaporation
431-432: 37-40.

on

structural

properties

of

Cu(ln,Ga)Se2

thin films, Thin Solid Films