You are on page 1of 160

Diss. ETH No.

16401

Controlled Synthesis of Mixed Oxide


Nanoparticles by Flame Spray Pyrolysis

A dissertation submitted to the

SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH


for the degree of
DOCTOR OF TECHNICAL SCIENCES
presented by

RAINER JOSSEN
Dipl. Chem. Ing. ETH Zurich

born on August 19th , 1974


citizen of Naters VS, Switzerland

Accepted on the recommendation of


Prof. Dr. Sotiris E. Pratsinis, examiner
Prof. Dr. Greg Beaucage, co-examiner
Zurich, 2006

Rose, oh reiner Widerspruch, Lust,


Niemandes, Schlaf zu sein unter so vielen Lidern.
Rainer Maria Rilke

Acknowledgement

This work was carried out at the Particle Technology Laboratory at the ETH Zurich. I
would like to acknowledge the financial support by the TH-Gesuch No. 34/02-3, WilliStuder-Fonds, and the Millennium Chemical Research Center, USA. Furthermore, I would
like to express my thanks to the following people, who contributed significantly to the
success of this work:
Prof. Sotiris E. Pratsinis, for giving me the opportunity to carry out this Ph.D
thesis in his team, for his continuous encouragement and his valuable advice as well as
for creating a dynamic and stimulating atmosphere; Prof. Gregory W. Beaucage for
co-advising this work, for the introduction in small angle X-ray scattering and in light
scattering during his stay in Zurich and also the time I could spend with him.
Furthermore, I appreciate the work of help assistants: Sarah Lanfranchi, Tobias
Weber, Frederick Marxer, Tim Patey and Juan Carlos Andresen as well as the fruitful
discussion and collaboration with Dr. Hendrik K. Kammler, Dr. Lutz Madler, Dr. Roger
M
uller, Martin C. Heine and Alexandra Teleki, Dr. Frank Krumeich for the TEM-images
and the whole PTL-team. I would like to acknowledge the work and support from the IPE
machine shop especially to Rene Pl
uss. Finally, to Ileana Eugster for helpful discussion
beside technology and science.

CONTENTS

Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Zusammenfassung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI
1 Criteria for flame spray synthesis of hollow, shell-like or inhomogeneous
oxides

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2.1

Powder synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2.2

Powder characterization . . . . . . . . . . . . . . . . . . . . . . . .

Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3.1

SiO2 Particle Size and Morphology . . . . . . . . . . . . . . . . . .

1.3.2

Bi2 O3 Particle Size and Morphology . . . . . . . . . . . . . . . . . . 11

1.3.3

Morphology mapping of FSP-made oxides . . . . . . . . . . . . . . 17

1.3

1.4

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Morphology and composition of spray-flame-made yttria-stabilized zirconia nanoparticles

25
III

IV
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.2

Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.3

2.4

2.2.1

Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.2.2

Precursor solution selection and preparation . . . . . . . . . . . . . 28

2.2.3

Powder characterization . . . . . . . . . . . . . . . . . . . . . . . . 30

Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31


2.3.1

Effect of precursor on particle morphology . . . . . . . . . . . . . . 31

2.3.2

YSZ crystal and primary particle sizes . . . . . . . . . . . . . . . . 38

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3 Thermal stability of flame-made zirconia-based mixed oxides

49

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.2

Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.3

3.4

3.2.1

Precursor preparation and thermal stability characterization . . . . 51

3.2.2

Apparatus and processing . . . . . . . . . . . . . . . . . . . . . . . 51

3.2.3

Powder characterization . . . . . . . . . . . . . . . . . . . . . . . . 52

Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53


3.3.1

Pure ZrO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.3.2

Yttria-doped zirconia . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.3.3

Ceria-doped zirconia . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.3.4

Silica-doped zirconia . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4 Thermal Stability and Catalytic Activity of Flame-made Silica-VanadiaTungsten oxide-Titania

67

V
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.2

Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.3

4.4

4.2.1

Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.2.2

Material Characterization . . . . . . . . . . . . . . . . . . . . . . . 70

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71


4.3.1

Powder characterization . . . . . . . . . . . . . . . . . . . . . . . . 71

4.3.2

Catalytic performance . . . . . . . . . . . . . . . . . . . . . . . . . 82

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5 Non-intrusive droplet and particle dynamics during spray flame synthesis of nano ZrO2

89

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

5.2

Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2.1

Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.2.2

Spray flame characterization . . . . . . . . . . . . . . . . . . . . . . 94

5.3

Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.4

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.5

5.4.1

Spray flame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.4.2

Droplet evaporation by SAXS . . . . . . . . . . . . . . . . . . . . . 103

5.4.3

Primary particle growth by SAXS . . . . . . . . . . . . . . . . . . . 104

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6 Research Recommendations

117

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

VI
A Reproducibility of nanoparticle production and color investigation on
titania particle made by pilot scale FSP

123

A.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123


A.1.1 Reproducibility of pilot scale FSP . . . . . . . . . . . . . . . . . . . 124
A.1.2 Investigation on the yellowish color of titania made by pilot scale
FSP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
A.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
A.2.1 Reproducibility of pilot scale FSP . . . . . . . . . . . . . . . . . . . 125
A.2.2 Investigation on the yellowish color of titania made by pilot scale
FSP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
B Flame spray scale-up investigation

127

B.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127


C Systematical study of silica contaminated YSZ

131

C.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131


C.1.1 Precursor preparation

. . . . . . . . . . . . . . . . . . . . . . . . . 131

C.1.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131


C.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
D Thermophoretice sampling in spray flames

135

D.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135


D.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Appendix: References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Curriculum Vitae

140

Refereed Publications

141

Presentations

141

Zusammenfassung

Funktionnelle Oxide konnen mit Flammen-Spr


uh-Pyrolyse (FSP) Technik hergestellt werden, die als Katalysatoren, stabile Quantenpunkte oder als Elektrolyt in Brennstoffzellen
angewended werden konnen. Jede dieser Anwendung benotig aber ein bestimmte durchschnittliche Partikelgrosse, Partikelverteilung, Agglomerate mit bestimmter fraktalen Dimension, Morphology und zuletzt bei Mehrkomponentensystem die richtige chemische
Zusammensetzung.
In FSP werden ein oder mehrere metallische Ausgangsprodukte in einem Losungsmittel gelost, mittels einer D
use zu Tropfen verspr
uht und anschliessend gez
undet. Im
ersten Kapitel wird der Einfluss der Ausgangsprodukte und der Losungsmittelkomposition
erforscht. Hierzu wurde Silizumoxid und Bismuthoxid hergestellt und analysiert. Zur Herstellung von Siliziumoxid wurden Siliziumalkoxide mit verschieden Siedepunkte in Xylol
gelost oder Tetraethylorthosilikat wurde in verschiedenen Alkane gelost auch mit verschieden Siedpunkte. F
ur ein nicht verdampfbares Systhem wurde Bismuth Nitrat Pentahydrat als Ausgangschemikalie verwendet und in verschiedenen Alkohlen mit verschiedenen Siedepunkten gelost. Mittels Transmission-Elektronenmikroskopieanalyise (TEM)
und Rontgenbeugung (XRD) wurde die Morphologie der Partikel bestimmt. Es hat sich
gezeigt, dass inhomogene und/oder hohle Partikel nur bei tiefer Verbrennungswarmedichte
und wenn der Siedepunkt des Losungsmittel tiefer als der Schmelz- oder Zerfallpunkt der
Ausgansprodukte ist, entstehen. Dies stimmt auch mit Daten aus der Literatur u
berein.

VII

VIII
Im zweiten Kapitel wurden die Untersuchungen auf Yttriumoxid stabilisiertes Zirkoniumoxid (YSZ), einem Zweistoffsystem, erweitert. Es zeigte sich, dass homogenes YSZ
mit organometallischen Ausgangsprodukten oder mit Yttrium Nitrat Hexahydrat und
Zirkoniumkarbonat, behandelt mit 2- Ethylhexansaure, hergestellt werden kann. Wenn
organometallische Chemikalien mit Ausganstoffen, die Hydratwasser beinhalten gemischt werden, werden Partikel mit inhomogener Morphologie und chemischer Komposition
produziert. Alkoxide sind sehr wasserempfindlich und reagieren rasch zu Hydroxiden,
welche in der Metalllosung ausfallen konne. Mit XRD konnten zwei durchschnitts Kristalldurchmesser gemessen werden, deren Verhaltnis die Hohe der Inhomogenitat angibt.
Im dritten und vierten Kapitel wird das Gefundene von Kapitel eins und zwei
angewendet. Zirkonoxid basierende Partikel wurden auf ihre Thermostabilitat getestet
bez
uglich Oberflachenverlust und Kristallinitat. Siliziumoxide und Vanadiumoxid dotietes Wolframoxide/Titanoxide wird ebenfalls auf seine Thermostabilitat gepr
uft. Spezielle
Wert wurde auf den Effekt von Siliziumoxid auf die Thermo-, Kristallstabilitat und die
katalytisches Verhalten im DeNOx Pozess gelegt.
Zur Zeit werden nur die Endprodukte analytisch untersucht. Die Kombination
von ortlichen Messmethoden, die die Partikelbildung nicht beeinflussen, konnen Aufschluss geben, wie die Evolution der Partikelmorphologie im FSP vor sich geht. Fourie
transformierte Infrarot Spektroskopie (FTIR) wird eingesetzt um Gastemperaturen zu
messen, Phasen Doppler Anometrie (PDA) f
ur Tropfen- und Gasgeschwindikeit, Thermophoretische Probenahme (TS) um das Partikelwachsum zu zeigen, und KleinwinkelRontgenstreuung (SAXS) und Lichtstreuung (LS) f
ur detailliertes Partikelwachstum und
Agglomerationsentwicklung zu studieren.
Partikle- und Tropfendynamik wird in Kaptiel f
unf mit ortlich aufgelosten und nicht
storenden (in-situ) Techniken studiert. In einer 15 g/h produzierenden Zirkoniumoxid
Spr
uhflamme wurde mit PDA Gasgeschwindigkeiten und mit FTIR Gastemperaturen
gemessen. Mit SAXS kann man gleichzeitig Primarpartikelradius, fractal Dimenssione
der Agglomerate, geometrische Standartabweichung, Partikelvolumenfraktion, Partikel-

IX
anzahlkonzentration, Agglomeratenradius und die Anzahl Primarpartikel pro Agglomerate bestimmen. F
ur SAXS konnte die dritte Synchroton Generation (ERSF, Grenobel)
ben
utzt werden, da diese gen
ugend Energie liefert um Messungen mit einer Volumenfraktion von weniger als 106 durchzuf
uhren.

Summary

The control of nanoparticle characteristics during flame synthesis is crucial since properties of the final product made from these particles depend on their size distribution,
morphology, extent of aggregation as well as chemical and phase composition. Flame
spray pyrolysis (FSP) for synthesis of functional oxide systems can be used for novel catalysts, stable quantum dots and fuel cells to mention some of their potential applications.
The influence of metal precursor and solvent composition on the morphology of SiO2 ,
Bi2 O3 and other oxide particles made by flame spray pyrolysis (FSP) was investigated in
the first chapter. Silica precursors with different boiling points were dissolved in xylene
and different solvents were used to dissolve tetraethyl-orthosilicate (TEOS). For Bi2 O3 ,
non-volatile bismuth nitrate pentahydrate was dissolved in different alcohols with different
boiling points. From these data and from the literature of FSP synthesis it is inferred
that hollow/inhomogeneous particles were formed at low combustion enthalpy densities
and when the solvent boiling point was comparable or smaller than the precursor melting
or decomposition point.
In the second chapter the investigation of homogeneity and chemical composition
was extended to a multiple system like yttria stabilized zirconia (YSZ). While in chapter one the production rate was relatively low (20 g/h), here, the production rate was
increased up to 350 g/h. The effect of liquid precursor composition on product particle
morphology, composition and crystallinity was investigated. Flame-made YSZ nanopar-

XI

XII
ticles of homogeneous composition and morphology were formed when using either only
organometallic zirconium and yttrium precursors or 2-ethylhexanoic acid as solvent and
inexpensive zirconium carbonate and yttrium nitrate hexahydrate as precursors. In contrast, and consistent with the literature, hollow or inhomogeneous YSZ particles were
made when organometallic zirconium and yttrium nitrate precursors of high water content were employed, especially at high production rate. The ratio of XRD-determined
small to large sizes for inhomogeneous crystalline particles is an effective quantitative
measure of their degree of inhomogeneity.
In the third and fourth chapter the finding of chapter one and two will be applied
by making mixed oxide for catalytic applications. First, thermostability of zirconia based
materials were tested by sintering them in an oven at different temperatures and second
the effect of doped WO3 /TiO2 powders with vanadia and silica on their specific surface
area, crystallinity, thermostability and catalytic behavior.
At the moment, most of the analysis of product particles was performed after the
final product has been formed. The combination and implementation of existing on-line
spray probing techniques has the potential to yield valuable local information on particle
size and morphology evolution without disturbing the process. This will be used to better
understand the FSP process itself that would lead to optimal process design and control
and finally tailor-made-products by FSP. Hence FTIR will be used for temperature measurement, Phase Doppler Anemometry (PDA) will allow to estimate the droplet lifetime
in the spray by measuring droplet velocity and droplet size evolution, thermophoretic
sampling (TS) to measure primary particle growth, while small angle X-ray scattering
(SAXS) and light scattering (LS) will enable the on-line monitoring of the nanoparticle
morphology and aggregate size evolution. Except from TS, these techniques are all in-situ
on-line methods.
Therefore, droplet and particle dynamics were studied in chapter five in situ
and non-intrusively in a particle-laden spray flame producing 15 g/h zirconia. Droplet
velocities were measured by 2D Phase Doppler Anemometry (PDA) and droplets smaller

XIII
than 4 m were used to estimate gas velocity. Gas temperature was measured by Fourier
Transform Infrared (FTIR) emission/transmission spectroscopy. In situ small angle
x-ray scattering (SAXS) was used to measure simultaneously the evolution of primaryparticle diameter, mass-fractal dimension, geometric standard deviation, particle volume
fraction, particle number concentration, agglomerate size and number of primary particles
per agglomerate of zirconia in the spray flame. For SAXS, a third-generation synchrotron
source is used where nano to micro scale measurement were possible.

CHAPTER

ONE

Criteria for flame spray synthesis of hollow, shell-like


or inhomogeneous oxides

Abstract

The influence of metal precursor and solvent composition on the morphology of SiO2 ,
Bi2 O3 and other oxide particles made by flame spray pyrolysis (FSP) was investigated.
Silica precursors with boiling points Tbp = 299-548 K dissolved in xylene were used as well
as different solvents (Tbp = 308-557 K) with tetraethyl-orthosilicate (TEOS) as the silica
precursor. For Bi2 O3 , non-volatile bismuth nitrate pentahydrate was dissolved in solvents
with Tbp = 338-468 K. Product powders were characterized by nitrogen adsorption, X-ray
diffraction and scanning and transmission electron microscopy. From these data as well
as from the literature of FSP synthesis of Bi2 O3 , CeO2 , MgO, ZnO, Fe2 O3 , Y2 O3 , Al2 O3
and Mg-Al spinel it is inferred that hollow/inhomogeneous particles are formed at low
combustion enthalpy densities and when the solvent boiling point is comparable or smaller
than the precursor melting or decomposition point.
1

1.1

Introduction

Silica and titania nanoparticles are produced in mega ton quantities annually by flame
technology, a manufacturing route that is investigated currently for synthesis of other
commodity and novel functional metal and ceramic nanoparticles [1, 2]. While volatile
metal-chlorides are fed as gases in a flame resulting in their metal oxides, liquids are
sprayed during FSP [3]. Liquid precursors considerably broaden the range of accessible
products from simple and mixed oxides for lasing materials [4] to alumina-supported
platinum catalysts [5] for enantioselective hydrogenation of ethyl pyruvate in synthesis of
chiral pharmaceuticals to name just a couple of examples.
Recent research activities have improved the understanding of particle formation and
growth during FSP for control of product particle size, crystallinity and even scale-up [4-8].
For example, Madler et al. [7] found that the specific surface area of silica made by FSP of
hexamethyldisilane (HMDSO) at constant liquid volumetric flow rate was decreased using
methanol, ethanol or iso-octane as solvents in this order. As the combustion enthalpy per
unit volume of these solvents was increased, higher flame temperatures enhanced sintering
resulting in larger particles. As result, this process has been scaled up for synthesis up
to 1 kg/h of silica [9] and ZrO2 [10] nanoparticles. Limaya and Helble [11] made zirconia
particles by FSP of zirconium(IV) n-butoxide in butanol and they postulated liquidand vapor-phase pathways of the final particle formation depending on the applied flame
temperature.
In some cases, however, FSP has produced inhomogeneous powders. Suyama and
Kato [12] made mostly hollow Mg-Al spinel by spray pyrolysis using magnesium and
aluminum nitrate co-dissolved in ethanol. Madler and Pratsinis [13] obtained hollow
bismuth oxide particles by FSP of bismuth nitrate dissolved in nitric acid/ethanol. By
replacing ethanol with acetic acid, they produced homogeneous, non-hollow, solid Bi2 O3
particles. Similarly, solid but inhomogeneous ceria particles were made by FSP using
cerium acetate dissolved in acetic acid [8]. Adding an iso-octane/2-butanol mixture to
the precursor solution resulted in rather homogeneous powders of 4-8 nm primary particle

3
diameter. Tani et al. [14] produced thinly-shelled hollow (eggshell-like) or solid -Al2 O3
particles by FSP of aqueous emulsion of aluminum nitrate when using air or oxygen,
respectively, as dispersion/oxidant gas. They also produced hollow TiO2 , ZrO2 , Y2 O3
or solid ZnO, Fe2 O3 , CeO2 and MgO particles by FSP of aqueous emulsion in air of
the corresponding nitrate or chloride precursors [14]. Some of the solid product powders
(CeO2 , ZnO and MgO) were quite inhomogeneous with a broad, nearly bimodal size
distribution.
Even though hollow or porous particles can be useful for insulation, catalyst supports
or encapsulation matrices [15], they are generally an undesired by-product. However,
systematic studies of the effect of process variables on FSP-made particle morphology
are still missing. It is therefore of interest to identify FSP conditions and criteria that
prevent synthesis of inhomogeneous powders. In this study the effect of liquid precursor
on product morphology is investigated during FSP synthesis of silica and bismuth oxide.
Design and operation criteria for synthesis of homogeneous particles are proposed and
compared with the present data and pertinent ones in the literature.

1.2
1.2.1

Experimental
Powder synthesis

Figure 1.1 shows a schematic of the experimental set-up of the small flame spray (FSP)
[7, 8]. Silica and bismuth oxide powders were produced by this FSP using air or oxygen (>
99.95% purity, Pan Gas, Luzern, Switzerland) as oxidant / dispersion gas. A constant 1.5
bar gas pressure drop at the burner nozzle tip was maintained by adjusting the nozzle-gap
width. The gas flow rates were controlled by mass flow controllers (Bronkhorst, Ruurlo,
The Netherlands). The liquid precursor solution was fed through the nozzle by a syringe
pump (IER-232, Inotech, Oberwinterthur, Switzerland). The resulting spray was ignited
by a ring of 18 supporting premixed flamelets sustained by 0.5 l/min CH4 and 2 l/min
O2 [8]. An additional oxygen sheath gas flow (5 l/min) was supplied through a sintered

4
metal plate ring 8 mm wide with an inner diameter of 9 mm surrounding the supporting
flamelets [8] .The particles were collected on a glassfiber filter (diameter of 150 mm;
Wathman GF/A) using a vacuum pump (Vaccubrand, Type RZ 16, Germany).
The effect of precursor composition on flame-made silica characteristics was investigated by spraying 1.5 ml/min of various silicon precursors (Table 1.1a, 0.88 M Si in
xylene (> 98%, Sigma-Aldrich, Fluka Chemie GmbH, Buchs, Switzerland)) to the flame.
The combustion enthalpy density was defined as the ratio of the fed liquid precursor combustion enthalpy (kJ/min) over the total gas flow (ggas /min) within the spray. The effect
of solvent composition (Table 1.1b) on silica characteristics was studied using tetraethylorthosilicate (TEOS) as Si-precursor. Since the combustion enthalpy per mass of alkane
solvents is similar but their densities vary from 0.62 (pentane) to 0.77 g/cm3 (hexadecane), the silica precursor concentration was adjusted between 0.64 M and 0.79 M. All
solvent experiments were performed at combustion enthalpy density of 6.4 0.1 kJ/ggas
and silica production rate of 4.65 g/h, silica with 3 l/min of air as dispersion gas using
liquid feed rates from 1.65 (hexadecane) to 2 ml/min (pentane).
Bismuth trinitrate pentahydrate (Bi(NO3 )3 5H2 O, > 98%; Sigma-Aldrich, Fluka
Chemie GmbH, Buchs, Switzerland) was dissolved in HNO3 / alcohol mixtures (15 vol% of
nitric acid (65 %, Sigma-Aldrich, Fluka Chemie GmbH, Buchs, Switzerland) and 85 vol%
of the corresponding alcohol, Table 1.1c). Since the density and combustion enthalpy per
unit mass of these alcohols vary, the feed rate and bismuth concentration were adjusted
(Table 1.1c) to keep production rate at 11.5 g/h Bi2 O3 and total combustion enthalpy
density of 2.2 0.03 kJ/ggas . The total combustion enthalpy density was increased to 4.0
0.09 and 4.7 0.2 kJ/ggas by increasing the precursor feed rates by a factor of 2 and
2.65, respectively. For Bi2 O3 3 l/min oxygen was used as dispersion/oxidant gas and 1.58
l/min CH4 and 1.38 l/min O2 for the supporting flamelets.

Figure 1.1: Sketch of a typical flame spray pyrolysis unit. The precursor mixture is
rapidly dispersed by oxygen or air and fed into the premixed flame methane / oxygen
stream.

Table 1.1: Process conditions and properties of solvents and precursors employed in the FSP synthesis of SiO2 and Bi2 O3 .
Solvent

boiling feed rate


point K ml/min

conc.
mol/l

sourcea
purity (%)

(a) Si-Precursor properties in xylene (Tbp = 410-416 K)


TMS, Si(CH3 )4
299-300 1.5
0.88
>99.5
HMDSO, C6 H18 OSi2
374
1.5
0.88
>99
TEMOS, Si(CH3 O)4
393-341 1.5
0.88
>99
TEOS, Si(CH3 CH2 O)4
436-440 1.5
0.88
>98
TEPOS, Si(CH3 CH2 CH2 O)4
548
1.5
0.88
>97
(b) Solvents for tertaethyl-orthosilcate (TEOS, Tbp = 436-440 K)
Pentane, C5 H12
Hexane, C6 H14
iso-Octane, C8 H18
Octane fraction
Decane fraction
Dodecane, C12 H26
Tetradecane, C14 H30
Hexadecane, C16 H34
(c) Solvent properties employed in
Methanol (MeOH)
Ethanol (EtOH)
Methoxy-2-propanol
Ethoxy-ethanol
Porpylene glycol propylether
Diethylene glycol-monoethylether
a

308-309 2.0
342
1.89
372
1.81
397-401 1.8
441-451 1.73
486-489 1.70
523-526 1.67
556-559 1.65
the FSP synthesis of
338
351
391
408-410
413-423
468

1.31
1.00
1.03
0.80
0.89
0.90

0.64
0.68
0.71
0.72
0.75
0.77
0.78
0.79
Bi2 O3
0.31
0.41
0.39
0.51
0.46
0.45

>99
>99.5
>99.5

90-95
>97
>98
>99.8
>99.9
>98
>99
>98.8
>98

Simga-Adrich, Chemie GmbH, Buchs, Switzerland. HMDSO, hexamethyldisilane; TEOS, tetraethyl-orthosilicate;TMS, tetramethylsilane; TEPOS,
tetrapropyl-orthosilica; TEMOS, tetramethyl-orthoslicate

1.2.2

Powder characterization

The powder specific surface area (SSA) was measured by nitrogen adsorption at 77 K
(BET-method, Micrometrics Inc., Gemini 2375, The Netherlands) after degassing samples
for at least 1 h at 150 C under nitrogen. The average BET-equivalent primary particle
diameter is dBET = 6/(SSA p ), where p is the density of SiO2 (2.2 g/cm3 ) or -Bi2 O3
(8.9 g/cm3 ). The crystallinity of bismuth oxide was measured by X-ray diffraction (40
kV, 40 mA) using a step size of 0.05 and a scan speed of 0.25 /min (Burker Advanced
D8, Karlsruhe, Germany). The XRD spectra were analyzed using the Topas 2.0 software
(Bruker AXS, 2000). Measured XRD patterns were regressed with the crystalline data
of -Bi2 O3 (PDF #78-1793 [16]). The morphology of product powders was analyzed by
scanning electron microscopy (SEM) operated at 30 kV (Hitachi Model S900). Both,
secondary-electron (SE) and back-scattered electron (BS) SEM images were taken to
distinguish from hollow and dense particles [13]. Further analysis was performed by
transmission electron microscopy (TEM, Zeiss 912 Omega; Jena, Germany operated at
100 kV using a slow CCD camera and the Proscan software). TEM pictures were prepared
by dipping the TEM grid into the product powder.

1.3

Results and Discussion

1.3.1

SiO2 Particle Size and Morphology

The effect of silica precursor/solvent composition on product primary particle size, dBET ,
was investigated at constant metal concentration, gas flow rate and combustion enthalpy
density. Using different silica precursors and solvents, the relative volatility of the sprayed
liquid was varied systematically. The relative volatility, TR , is the ratio of the boiling point
of the precursor over that of the solvent.
Figure 1.2 shows the dBET of silica as a function of TR for various precursors (Table
1.1a) dissolved in xylene (triangles) and for TEOS (circles) dissolved in various alkane

Figure 1.2: Primary particle BET diameter of silica using TEOS/alkanes (circles)
and xylene/Si-precursors (triangles) as a function of the relative boiling point (Tbp (Siprecursor)/Tbp (solvent)) at constant enthalpy density of 6.2 kJ/ggas . The volatility of
either precursor or droplet has no effect on particle size and morphology as homogeneous
particles were formed at all conditions.

9
solvents (Table 1.1b). Error bars represent twice the standard deviation of reproduced
experiments. The total combustion enthalpy density was kept constant at 6.2 kJ/ggas .
Increasing Si-precursor Tbp from 299 to 548 K (increases TR from 0.75 to 1.37) had no
effect on the dBET (13.8 1 nm). Likewise, increasing the solvent boiling point (Tbp ) from
308 to 559 K (decreasing TR from 1.42 to 0.79) had also no significant influence on dBET
indicating that the employed solvent/precursor compositions were of minor importance
in flame spray synthesis of silica at these conditions. Figure 1.3 shows TEM images of
homogeneous silica agglomerates made using TEOS/pentane (a, TR =1.42), TMS/xylene
(b, TR =0.75), TEOS/hexadecane (c, TR =0.79) and TEPOS/xylene (d, TR =1.37). No
distinct size differences or hollow particles were observed in any silica products. This
morphology is similar to fumed silica made by feeding gaseous Si-precursor (e.g. SiCl4 or
HMDSO) to flames resulting in fumed silica [17].
Apparently, silica particle formation takes place in the gas phase as the precursor
fully evaporates and forms product particles by chemical reaction, coagulation and sintering [7]. If particle formation would have taken place at or within the droplet, mass transfer
to the gas phase would be the rate-limiting step and spherical or broken shells may be
formed [15]. This was not observed and therefore dBET was not affected by the precursor
and solvent composition (Figure 1.2). Even though the difference in vapor pressure of
TEPOS (Tbp = 548 K) and xylene (Tbp = 410-416 K) is about 5 orders of magnitude at
room temperature, no significant difference in dBET was observed as the total combustion
enthalpy density was rather constant [7]. The droplet evaporation history for volatile
silica precursors has no significant effect on particle formation (size and morphology) so
major process parameters are combustion enthalpy and metal concentration in the flame.
This is in agreement with Briesen et al. [18] who found that the combustion enthalpy or
adiabatic flame temperature determines the product silica primary particle size made in
a vapor-fed premixed flame reactor.

10

Figure 1.3: Transmission electron microscope images of FSP-made silica at constant


combustion enthalpy density and production rate from a) TEOS/pentane, b) TMS/xylene,
c) TEOS/hexadecane and d) TEPOS/xylene. At all conditions homogenous particles were
formed.

11

1.3.2

Bi2 O3 Particle Size and Morphology

Since typically non-volatile metal precursors are employed in FSP, the effect of solvent
composition was investigated for synthesis of Bi2 O3 from non-volatile bismuth nitrate
dissolved in different alcohols (Table 1.1c). Experiments were carried out at three combustion enthalpy densities but constant metal concentration and gas flow rates.
Figure 1.4 shows the dBET of Bi2 O3 as a function of the employed alcohol (solvent)
boiling point, Tbp , at combustion enthalpy density of 2.2 (diamonds), 4.0 (circles) and
4.7 kJ/ggas (triangles). The primary particle size is not affected by the alcohol boiling
point (Tbp ) as with silica and remains constant at 14 1 nm, 17 1 nm and 23 1 nm
for combustion enthalpy densities of 2.2, 4.0 and 4.7 kJ/ggas , respectively. The increase in
particle size from 14 to 23 nm by increasing the combustion enthalpy density is consistent
with the current understanding of the effect of adiabatic flame temperature and product
particle size [18]. With increasing combustion enthalpy density from 2.2 to 4.7 kJ/ggas ,
the flame height increased from 55 to 75 mm similar to vapor-fed flame synthesis of
silica [18]. The increased residence time at high temperature in the flame leads to longer
residence time for particle sintering resulting in larger primary particles and lower specific
surface area [1, 7] For the lower combustion enthalpy densities, 2.2 and 4.0 kJ/ggas and
for the lowest boiling point solvent (MeOH), a slightly larger dBET was measured at each
combustion enthalpy. The latter Bi2 O3 powders, however, are inhomogeneous and contain
larger particles that decrease the SSA and increase the dBET as it will be shown shortly
by microscopy and X-ray diffraction.
Figure 1.5 shows TEM images of bismuth oxide particles made with a) EtOH as solvent at a feed rate of 3.5 ml/min resulting in combustion enthalpy density of 4.7 kJ/ggas ;
b) methoxy-2-propanol as solvent at a feed rate of either 1.80 ml/min resulting in combustion enthalpy density of 4.0 kJ/ggas or c) 0.9 ml/min resulting in combustion enthalpy
density of 2.0 kJ/ggas and d) ethoxy-ethanol as solvent at a feed rate of 1.6 ml/min resulting in combustion enthalpy density of 4.0 kJ/ggas . In all cases (Figure 1.5a-d), only
homogeneous particles were formed. Hollow or large particles were not found at high

12

Figure 1.4: Primary particle BET diameter of Bi2 O3 made by FSP from bismuth nitrate
pentahydrate/HNO3 /alcohol solutions as a function of the boiling point of the employed
alcohol solvent for three combustion enthalpy densities. At constant combustion enthalpy
density, the solvent composition has little influence on product primary particle size.
Higher production rate corresponding to higher combustion enthalpy density and longer
flames prolong the particle residence time at high temperature resulting larger particles.

13
combustion enthalpy density ( 4.7 kJ/ggas ) or with solvents having boiling points larger
than 391 K (Table 1.1c).
In contrast, Figure 1.6 shows images of inhomogeneous Bi2 O3 powder made using
EtOH as solvent at a feed rate of 1 ml/min resulting in a combustion enthalpy density
of 2.2 kJ/ggas . Figure 1.6a shows a TEM picture of a shell-like particle together with
fine particles ( 10 nm). Figure 1.6b shows a secondary electron (SE)-SEM image and
Figure 1.6c the corresponding backscattered electron (BS)-SEM image of larger, fine and
shell-like particles. Dense, large particles can be observed (white structures in Figure
1.6c) with particle diameters varying from 250 to 500 nm whereas the hollow ones appear
gray and the fine particles are barely visible. Similar results were obtained with MeOH
as solvent.
X-ray diffraction was used to further investigate particle homogeneity. Figure 1.7
shows XRD spectra of the Bi2 O3 powders in Figure 1.5a (broken line) and Figure 1.6
(solid line). This figure shows a clear distinction between a smooth (broken line) and
a hump-containing (solid line) XRD corresponding to homogeneous and inhomogeneous,
respectively, powders made at 4.7 and 2.2 kJ/ggas combustion enthalpy density. For the
inhomogeneous powder (Figure 1.6), a bimodal crystal size distribution can be derived
from XRD [16] resulting in average crystal sizes of 108 nm (15 % by mass) and 3 nm (85%
by mass) for each size mode. This gives a quantitative measure of crystalline powder
homogeneity and is used throughout this study.
The above results can be placed in the FSP parameter space of combustion enthalpy
density and solvent boiling point, Tbp : Figure 1.8 shows that hollow, shell-like or inhomogeneous powders were produced at low combustion enthalpy densities (< 4.7 kJ/ggas )
and with solvents having low boiling points (open symbols). In all other cases, solid
homogeneous, nanostructured Bi2 O3 powders were observed (filled symbols). Similarly,
Madler and Pratsinis [13] used acetic acid (AcOH, Tbp = 391 K) as solvent and produced
homogeneous powders. The particle morphology was independent of the combustion enthalpy density since they chose a high boiling point solvent (Figure 1.8, filled triangles).

14

Figure 1.5: Transmission electron microscope images of solid, nanostructured Bi2 O3


particles made by FSP from a bismuth nitrate pentahydrate/ HNO3 solution in a) EtOH
at a combustion enthalpy density of 4.7 kJ/ggas , b) methoxy-2-propanol at 4.0 or c) at
2.0 kJ/ggas and d) ethoxy-ethanol at 4.0 kJ/ggas .

15

Figure 1.6: Images of hollow, large and fine solid Bi2 O3 made by FSP from a bismuth nitrate pentahydrate/HNO3 /EtOH solution at a combustion enthalpy density of 2.0 kJ/ggas :
a) TEM of a hollow Bi2 O3 particle surrounded by very fine ones, b) secondary-electron
(SE) image and corresponding c) backscattered-electron (BS) images. The latter shows
solid particles (white contrast) of about 100-250 nm and a hollow broken shell of 2 m in
diameter (gray).

16

Figure 1.7: XRD spectra of Bi2 O3 powder made by FSP from bismuth nitrate dissolved
in EtOH for a combustion enthalpy density of 2.0 (solid line) and of 4.7 (broken line)
kJ/ggas . There is a distinct difference (hump) between inhomogeneous (solid line) and
homogeneous (broken line) Bi2 O3 .

17
When using, however, more than 20% EtOH with the balance of AcOH as solvent, inhomogeneous particles were formed [13]. By calculating the boiling point of AcOH/EtOH
mixture with the Antoine equation, the latter data are in excellent agreement with the
present study (Figure 1.8, open triangles). The melting point of Bi(NO3 )3 5H2 O is 349 K
[19] and in this study, solvents with Tbp = 338-468 K were used. Hollow particles were
formed if a concentration gradient is formed during solvent evaporation [20] and if the
molten oxide precursor decomposes within this layer. If the salt shells are impermeable to
the solvent, the pressure within the particle increases substantially and particles fragment
or explode [21] resulting in broken shells. From Figure 1.8 it may be inferred that bismuth
nitrate precipitates on the droplet surface forming a shell that can trap remaining solvents (MeOH and EtOH) which evaporate within that shell. The resulting high pressure
inside that hollow sphere disrupts it forming shell-like fragments. For high boiling point
solvents (methoxy-2-propanol to diethylene glycol-monoethylether, Table 1.1c) no shells
were formed. The bismuth nitrate decomposed in the hot solvent and solid nuclei are
formed throughout the droplet. Upon solvent evaporation, these nuclei form extremely
fine oxide particles.

1.3.3

Morphology mapping of FSP-made oxides

The limit between the region of hollow and homogeneous particles can be found close to
the melting point of the bismuth nitrate (Tmp = 349 K), thus corroborating the outlined
mechanism. The decomposition/melting point of bismuth nitrate (Td/mp = 349 K) can be
used to relate the present results to other FSP-studies of particle morphology. Figure 1.9
maps the powder morphology in the FSP parameter space of combustion enthalpy density
as function of the ratio Tbp /Td/mp . Filled symbols represent homogeneous powders while
open represent hollow or inhomogeneous powders.
Circles represent the Bi2 O3 data from this study (Figure 1.8). Hollow particles
and inhomogeneous Bi2 O3 powders were only found at Tbp /Td/mp < 1 and at combustion
enthalpy densities < 4.7 kJ/ggas . The triangles refer to Madler and Pratsinis [13] who

18

Figure 1.8: Morphology mapping of solid (filled symbols) and hollow/shell-like (open
symbols) particles as a function of solvent boiling point and combustion enthalpy density
during FSP synthesis of Bi2 O3 from bismuth nitrate pentahydrate.

19
prepared Bi2 O3 by FSP as discussed before (Figure 1.8). Inhomogeneous silica powders
(filled squares) were not made here since the employed Tbp /Td/mp was > 1.4 and the
combustion enthalpy densities were about 6.2 kJ/ggas . The inverse triangles show data of
ceria [8] made by FSP from cerium acetate (Td/mp = 573 K) dissolved in acetic acid (Tbp
= 391 K) resulting in Tbp /Tmp = 0.68. By adding a mixture of iso-octane/2-butanol, the
combustion enthalpy density was increased from 1.2 to 1.8 kJ/ggas at a liquid precursor
feed rate of 1 ml/min. A mixture of large and small particles (open inverse triangles)
was produced at combustion enthalpy densities of 1.2, 1.8 and 3.2 kJ/ggas (2 ml/min
feed rate). When increasing, however, the liquid feed rate to 4 ml/min resulting in a
combustion enthalpy density of 5.9 kJ/ggas , solid and homogeneous ceria particles were
formed (filled inverse triangles).
Tani et al. [14] made nanoparticles by FSP of aqueous emulsions of precursor metal
nitrates mixed with kerosene and a surfactant that was sprayed in a flame reactor. Here,
the Tbp /Td/mp was calculated using the boiling point of water (Tbp = 373 K) and the
melting point of the nitrates. The diamonds represent Al2 O3 powders with Tbp /Td/mp
= 0.92. Using oxygen as oxidant/dispersion gas, homogeneous alumina powder (filled
diamonds) was formed as the combustion enthalpy density was 5.4 kJ/ggas . Using air as
oxidant/dispersion gas, inhomogeneous alumina was formed (open diamonds) as the combustion enthalpy density was decreased to 4.2 kJ/ggas . Homogeneous iron oxide powder
(filled butterfly) was produced even though air was used as dispersion gas (Tbp /Td/mp =
1.16, 4.5 kJ/ggas ). Inhomogeneous powders were observed when making ceria (Tbp /Td/mp
= 0.88, 4.6 kJ/ggas , circles containing cross), zinc oxide (Tbp /Tmp = 0.92, 4.6 kJ/ggas ,
square containing cross), yttria (Tbp /Tmp = 1.00, 4.5 kJ/ggas , triangle containing dot)
and magnesium oxide (Tbp /Tmp = 1.01, 4.6 kJ/ggas , triangle containing cross, Figure 1.9).
Suyama and Kato [12] (square containing dot) prepared Mg-Al spinel by spray pyrolysis of Mg(NO3 )3 6H2 O and Al(NO3 )3 9H2 O dissolved in EtOH. The melting point of the
aluminum nitrate is 347 K, while the magnesium nitrate precursor decomposes at 371
K. They made hollow particles at Tbp /Tmp = 1.01 (Figur 1.9) and combustion enthalpy

20

Figure 1.9: Morphology mapping of various solid (filled symbols) and hollow/shell-like
(open symbols) ceramic oxide particles in the FSP parameter space of the ratio of the
solvent boiling point over the precursor decomposition or melting point (Tbp /Td/mp ) and
combustion enthalpy density. Homogeneous particles were made for Tbp /Td/mp >1.05 and
for combustion enthalpy densities > 4.7 kJ/ggas .

21
density of 1.2 kJ/ggas . Figure 1.9 shows that homogeneous powders were only found at
high combustion enthalpy densities (> 4.7 kJ/ggas ) and at Tbp /Td/mp > 1.05 for all these
oxides.

1.4

Conclusions

Formation of hollow/inhomogeneous ceramic oxide particles by flame spray pyrolysis


(FSP) was examined since FSP is one of the promising techniques for synthesis of a
wide spectrum of nanostructured oxides. By analyzing Bi2 O3 and silica powders made
with various precursors and solvents, low process temperatures and low boiling point solvents favor formation of hollow or inhomogeneous powders. These criteria were found to
match data with flame-spray-made Bi2 O3 , SiO2 , CeO2 , MgO, ZnO, Fe2 O3 , Y2 O3 , Al2 O3
or Mg-Al spinel. In particular, inhomogeneous particles were formed at low combustion
enthalpy densities (< 4.7 kJ/ggas ) and when the solvent boiling point is smaller than the
melting or decomposition point of the metal precursor (Tbp /Td/mp < 1.05).

References
[1] S. E. Pratsinis. Flame aerosol synthesis of ceramic powders. Prog. Energy Combust.
Sci., 24(3):197219, 1998.
[2] W. J. Stark and S. E. Pratsinis. Aerosol flame reactors for manufacture of nanoparticles. Powder Technol., 126(2):103108, 2002.
[3] M. Sokolowski, A. Sokolowska, A. Michalski, and B. Gokjeli. The in-flame-reaction
methode for Al2 O3 aerosol formation. J. Aerosol Sci., 8:219230, 1977.
[4] R. M. Laine, C. R. Bickmore, D. R. Treadwell, and F. Waldner. Ultrafine metals
oxide powders by flame spray pyrolysis. Patent #5614596, Sep. 28 1999.
[5] R. Strobel, W. J. Stark, L. Madler, S. E. Pratsinis, and A. Baiker. Flame-made platinum/alumina: Structural properties and catalyticlal behaviour in enantioselective
hydrogenation. J. Catal., 213:296304, 2003.

22
[6] C. R. Bickmore, K. F. Waldner, R. Baranwal, T. Hinklin, D. R. Treadwell, and R. M.
Laine. Ultrafine titania by flame spray pyrolysis of a titanatrane complex. J. Eur.
Ceram. Soc., 18(4):287297, 1998.
[7] L. Madler, H. K. Kammler, R. Mueller, and S. E. Pratsinis. Controlled synthesis
of nanostructured particles by flame spray pyrolysis. J. Aerosol Sci., 33(2):369389,
2002.
[8] L. Madler, W. J. Stark, and S. E. Pratsinis. Flame-made ceria nanoparticles. J.
Mater. Res., 17(6):13561362, 2002.
[9] R. Mueller, L. Madler, and S. E Pratsinis. Nanoparticle synthesis at high prodcution
rates by flame spray pyrolysis. Chem. Eng. Sci., 58(10):19691976, 2003.
[10] R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, and M. K. Akthar. Zirocnia
nanoparticles made in spray falmes at high production rates. J. Am. Ceram. Soc.,
50(12):30853094, 2004.
[11] A. U. Limaye and J. J. Helble. Morphological control of zirconia nanoparticles
through combustion aerosol synthesis. J. Am. Ceram. Soc., 85(5):11271132, 2002.
[12] Y. Suyama and A. Kato. Characterization and sintering of Mg-Al spinel prepared
by spray-pyrolysis technique. Cheram. Inter., 8(1):1721, 1982.
[13] L. Madler and S. E. Pratsinis. Bismuth nanopartilces by flame spray pyrolysis. J.
Am. Ceram. Soc., 85(7):17131718, 2002.
[14] T. Tani, N. Watanabe, K. Takatori, and S.E. Pratsinis. Morphology of oxide particles
made by the emuslsion combustion methode. J. Am. Ceram. Soc., 86(6):898904,
2003.
[15] G. L. Messing, S.-C. Zhang, and G. V. Jayanthi. Ceramics powder synthesis by spray
pyrolysis. J. Am. Ceram. Soc., 76(11):27072726, 1993.
[16] S. K. Blower and C. Greaves. The structure of -Bi2 O3 from powder neutrondiffraction data. Acta Cryst., 44:587589, 1988.
[17] H. K. Kammler and S. E. Pratsinis. Scaling-up the produciton of nanosize SiO2
partilces in a double difusion flame aerosol reactor. J. Nanoparticle Res., 1(4):467
477, 1999.
[18] H. Briesen, A. Fuhrmann, and S. E. Pratsinis. The effect of precursor in flame
synthesis of SiO2 . Chem. Eng. Sci., 53(24):41054112, 1998.

23
[19] D. R. Lide, editor. CRC Handbook of Chemistry and Physics. CRC Press, INC, Boca
Raton, Florida, 3rd electronic edition, 2000.
[20] I. W. Lenggoro, T. Hata, F. Iskandar, M. M. Lunden, and K. Okuyama. An experimental and modeling investigation of particle production by spray pyrolysis using a
laminar flow aerosol reactor. J. Mater. Res., 15(3):733743, 2000.
[21] A. Gurav, T. Kodas, T. Pluym, and Y. Xiong. Aerosol processing of materials.
Aerosol Sci. Tech., 19:411452, 1993.

24

CHAPTER

TWO

Morphology and composition of spray-flame-made


yttria-stabilized zirconia nanoparticles

Abstract
Homogeneous yttria stabilized zirconia (YSZ) of 8-31 nm of average crystallite and particle
diameter containing 3-10 mol% yttria are made by flame spray pyrolysis (FSP) of various
yttrium and zirconium precursors at production rates up to 350 g/h. Product particles are
characterized by N2 adsorption (BET), transmission electron microscopy (TEM), energydispersive X-ray spectroscopy (EDS) and X-ray diffraction (XRD). The effect of liquid
precursor composition on product particle morphology, composition and crystallinity is
investigated. The yttria content does not affect the product primary particle and crystal sizes of homogeneous YSZ. These are determined, in turn, by the process enthalpy
content and overall metal concentration. Flame-made YSZ nanoparticles of homogeneous
composition and morphology are formed when using either only organometallic zirconium
and yttrium precursors or 2-ethylhexanoic acid as solvent and inexpensive zirconium carbonate and yttrium nitrate hexahydrate as precursors. In contrast, and consistent with
the literature, hollow or inhomogeneous YSZ particles are made when organometallic zirconium and yttrium nitrate precursors of high water content are employed, especially at
high production rate. The ratio of XRD-determined small to large sizes for inhomogeneous
25

26
crystalline particles is an effective quantitative measure of their degree of inhomogeneity.
For such inhomogeneous particles nitrogen adsorption is not a reliable technique for the
average grain size as it relies on integral properties of the particle size distribution.

2.1

Introduction

Stabilized zirconia is important for its outstanding thermal stability, chemical resistance,
mechanical characteristics and ionic conductivity [1]. The high temperature cubic ZrO2
phase can be stabilized down to room temperature by the addition of yttria (Y2 O3 ), magnesia, calcium oxide or rare earth oxides [2]. Cubic ZrO2 is detected at room temperature
with as little as 1.5 mol% Y2 O3 . This is the limiting Y2 O3 concentration for solid solution
in monoclinic zirconia while at Y2 O3 concentrations higher than 7.8 mol% only cubic
ZrO2 is present. A yttria content of 17 to 40 mol% leads to the formation of cubic ZrO2
and Y4 Zr3 O12 while above that leads to pure Y4 Zr3 O12 [3]. Cubic ZrO2 has the highest
ion conductivity and yttrium-stabilized zirconia (YSZ) has been widely used in oxygen
sensors, solid oxide fuel cells (SOFC) [4] or as supporting material for platinum catalyst
for nitrogen oxide sensors [5]. Monodispersed nanoscale zirconia powder is favored for
preparation of advanced ceramics with uniform nanostructure and properties [6].
Yttria-stabilized zirconia (YSZ) is made by reaction sintering of the constituent oxides but this time-consuming method may result in a product with undesirable particle
size distribution and purity [1]. Zhang et al. [7] made solid, spherical, uniformly dispersed
4 mol% YSZ by spray pyrolysis of zirconyl hydroxyl chloride and yttrium nitrate. The
product average BET-particle diameter was controlled from 16 nm to about 1m. Pebler
[1] ultrasonically atomized a nitrate solution of zirconium and yttrium into a multipass
quartz reactor designed to extend the residence time up to 15 s. He reported the formation of pure cubic (10 mol% YSZ) solid spherical particles with an average diameter of
0.5 m. Shukla et al. [8] synthesized YSZ particles by a rapid-combustion route where
a saturated solution containing zirconyl nitrate, yttrium nitrate and carbohydrazine is

27
introduced into a muffle furnace at 620 K. The YSZ product consisted of cubic-fluorite
when made with more than 8 mol% of Y2 O3 . Karthikeyan et al. [9] made pure ZrO2 and
4.5 mol% YSZ powders by spraying solutions of 2.5 wt% zirconium butoxide/n-butanol
and 2 wt% zirconium/yttrium acetate/acid acetic/water, respectively, into a hydrogen
flame. Mostly tetragonal or monoclinic crystal structures were formed with sizes of 12
or 21 nm (zirconia) and 17 or 30 nm (YSZ), respectively. Yuan et al. [10] made pure
and yttria-stabilized zirconia by flame-assisted ultrasonic spray pyrolysis. They used
zirconium n-propoxide and yttrium nitrate hexahydrate dissolved in ethanol/HNO3 as
precursor at a Zr/Y ratio of 84:16 (8.5 mol% YSZ) with a total precursor concentration
(Zr+Y) of 0.1 and 0.2 M to make micron- and submicron-sized hollow or porous particles.
Xie [11] made well-dispersed YSZ nanoparticles with crystal sizes of around 12 nm by
sol-gel reactions of zirconyl chloride and yttrium chloride. Here, the effect of precursor
composition on product particle morphology, crystallinity and size is investigated during continuous, dry synthesis of YSZ at relatively high production rate. A focus is on
identifying conditions that would allow controlled synthesis of homogeneous, single or
polycrystalline, nanostructured YSZ from inexpensive precursors.

2.2
2.2.1

Experimental
Apparatus

Figure 2.1 shows the experimental set-up of the large flame spray pyrolysis (FSP) consists of an external-mixing gas-assisted stainless-steel nozzle (Schlick-D
use, Gustav Schlick
GmbH + Co, 970/4-S32) that is made of a capillary tube of 0.5 mm ID and an annular
gap that can be adjusted to keep a constant pressure drop (1 bar) across the nozzle [12].
Liquid precursor (13.5 to 81.1 ml/min) is fed through the capillary by a pulsation-free
precision piston pump (Isco Inc., 1000D). That liquid is dispersed (atomized) by 50 l/min
O2 (Pan Gas, Switzerland, 99.95%) unless otherwise noted. The resulting spray is ignited
by a supporting CH4 /O2 diffusion flame (CH4 = 2 l/min (Pan Gas, Switzerland, 99.5%),

28
O2 = 4.5 l/min) surrounding the nozzle [12]. Additional sheath O2 (15 l/min) is supplied
through an outer sintered metal plate ring to assure complete precursor conversion. The
dispersion O2 flow rate is controlled by a mass flow controller (Bronkhorst) while that
of sheath O2 by a calibrated rotameter (Vogtlin Instruments AG). Product powders are
collected with a Jet filter (FRR 4/1.2, Friedli AG, Switzerland) containing four baghouse
filters which are cleaned periodically by sequential air pressure shocks. Additionally, a
check valve is used to avoid disturbance of the spray flame by these shocks during particle collection. Small samples of the product powder (1 g) are collected on-line using
a bypass connected to the inlet pipe of the baghouse filter upstream of the check valve
[12]. For this, a vacuum pump (Vacubrand RE 16) is used and particles are collected on
a glass fiber filter (Whatman GF/A), 150 mm in diameter, that is located in a stainless
steel holder.

2.2.2

Precursor solution selection and preparation

The total metal concentration (Zr +Y) in the liquid precursor solution is kept constant at
0.5 M in all experiments. Typically a molar ratio of Zr/Y = 4.5 is used corresponding to
10 mol% YSZ in the product powder. Yttrium nitrate hydrate (Y(NO3 )3 xH2 O = YNx,
99.9%, Aldrich (x=6) and ChemPur (x=0.5)) is dissolved ultrasonically in EtOH (Alcosuisse, 99.9%, denatured with 2% methylethylketone) and then zirconium n-propoxide
(Zr(OC3 H7 )4 = ZP, 70 wt% in n-propanol, ChemPur) is added resulting in a clear solution.
Precursor solutions using YN6 and ZP at such ratios that correspond to 3, 5, 7, 8 and 9
mol% YSZ product powders are prepared as well. YSZ is made also by removing the H2 O
from YN6 with acetic anhydride (C4 H6 O3 = AcAn, Riedel de Haen, 99-100%). The AcAn
is added slowly to YN6 at room temperature under N2 and the resulting NOx is bubbled
through a NaOH solution. The yttrium precursor solution is then mixed with ZP and
EtOH to form the YSZ precursor solution. In addition, yttrium butoxide (Y(C4 H9 O)3,
YB, Aldrich, 0.5 M in toluene) is mixed with ZP and toluene (C7 H8 , Fluka, >99.5%) to
make another YSZ precursor mixture to be investigated here. Inexpensive precursors are

29

Figure 2.1: YSZ (Schematic of the flame spray pyrolysis (FSP) reactor for synthesis of
yttrium stabilized zirconia at high production rates using a commercially available nozzle.

30
examined also such as zirconium carbonate hydroxide oxide (Zr(OH)2 (CO3 )2 ZrO2 = ZC,
ZrO2 content 44.4 wt%, LU United Intl Inc.) that is first dissolved in acetic acid (Fluka,
99.8% AcOH) at 50 C and then 2-ethylhexanoic acid (2-EHA, Fluka, 99%) is added. This
solution is distilled at 160 C for 6 hours. For yttrium, YN6 is dissolved first in EtOH
and then 2-EHA is added and distilled at 120 C until a clear solution is formed. The
YSZ precursor is prepared by mixing these two precursor solutions using EtOH as an
additional solvent. Such precursor solutions are prepared containing yttria of 3, 5, 8 and
10 mol% in YSZ.

2.2.3

Powder characterization

The specific surface area (SSA) is determined from a five-point N2 adsorption isotherm at
77 K (Gemini III 2375 and Tristar 3000, Micromeretics Instruments Corp.) after degassing
the samples with N2 for 1 hour at 150 C. Assuming monodisperse spherical particles,
the average BET-equivalent primary particle diameter, dBET , is calculated by dBET =6/(
SSA)

where is the density of cubic 10 mol% YSZ (5.8 g/cm3 [13]). The densities of the

3 to 9 mol% YSZ powders are calculated by a linear interpolation from pure tetragonal
zirconia (6.1 g/cm3 [13]) to 10 mol% YSZ. The crystallinity of YSZ is measured by X-ray
diffraction (Bruker, D8, 40 kV, 40 mA, Karlsruhe, Germany) over a 2 range of 20 to 70 ,
step size 0.03 and scan speed 0.6 /min. The crystalline characteristics are obtained from
the XRD spectra using Topas 2.0 software (Bruker AXS, 2000) by the Rietveld method
[14] in which the effect of the equipment (e.g. X-ray source, slits) are incorporated. The
crystal size, dXRD , is calculated from the full width half maximum (FWHM) of the (111)
peak using Scherrers equation [14, ]: dXRD =0.9/ cos ), where is the wavelength of the
X-ray (0.154186 nm) and and represent the measured FWHM and the diffraction angle,
respectively. Measured XRD patterns are regressed with the crystalline data of cubic
YSZ structure (PDF #77-2288 [15]) while for powders with bimodal size distribution
the Rietveld method is used to calculated both crystal sizes (dXRD,l , dXRD,s ) [16]. The
morphology of the product powder is obtained by transmission electron microscopy (TEM)

31
with a Hitachi H600 or a JEOL 2000FX II electron microscopes operating at 100 or 200
kV, respectively, using magnification ranging from 50 to 800 k. The energy-dispersive Xray spectroscopy (EDS) analysis are made with an EDAX Genesis system with a resolution
of 130 eV to analyze the yttria content in the solid YSZ solution. During imaging special
attention is given to morphology and yttria distribution.

2.3

Results and Discussion

2.3.1

Effect of precursor on particle morphology

Figure 2.2 shows TEM images of 10 mol% YSZ powders made at production rates of a) 43
and b) 342 g/h using YN0.5/ZP/EtOH and at production rates of c) 54 and d) 324 g/h
using YN6/ZP/EtOH. Using YN0.5 fine solid, agglomerated YSZ nanoparticles with a
few large ones are made at 54 g/h (Fig. 1a). Increasing the production rate (that means
increased particle concentration and process enthalpy content) increases the size of all
particles. Using YN6 at a low production rate (54 g/h), large, hollow, shell-like and very
fine (only few nanometer in size) YSZ particles are made (Fig. 2.2c). At high production
rate (324 g/h) inhomogeneous powder is made also but without hollow particles (Fig.
2.2d, 250 nm particles and very fine ones). This is in agreement with Yuan et al. [10] who
reported also the formation of hollow or porous YSZ particles by hot-wall spray pyrolysis
of YN6/ZP/EtOH.
Here, it is shown (Fig. 2.2a, c) that the water content of YNx affects distinctly
the product powder morphology. Two mechanisms seem dominant for the formation of
inhomogeneous powder: Larger particles are formed directly form precursor droplets that
do not completely evaporate in the flame [17] while smaller ones are formed in the gas
phase. Possibly when YN6 is dissolved in EtOH and then mixed with ZP, one of the
four propoxides in ZP reacts with water forming an OH group [18]. The decomposition
point of the partially hydrolyzed ZP (Tdp = 500 C to ZrO2 for Zr(OH)4 [13]) and the YN
(Tdp = 600 C [6]) are much higher than the boiling point of EtOH (Tbp = 79 C). Hollow

32

Figure 2.2: Yttria stabilized zirconia (10 mol% Y2 O3 ) powders made by FSP of solutions
of yttrium nitrate (a, b) 0.5 hydrate (YN0.5) or (c, d) hexahydrate (YN6) and zirconium
n-propoxide (ZP) in EtOH made at a) 43, b) 342, c) 54, and d) 324 g/h.

particles are formed if a solute concentration gradient is created within a precursor droplet
during solvent evaporation so the molten precursor decomposes within this layer. The
partially hydrolyzed ZP precipitates first near the more supersaturated droplet surface
and forms a crust resulting in hollow or shell-like particles [17] consistent with Figure 2.2c.
These hollow or shell-like particles solidify at higher production rates as the combustion

33
enthalpy density within the spray is increased from 4.4 to 10.3 kJ/ggas (corresponding
to increasing the production rate from 54 to 324 g/h) and the flame height increases
from 9 to 40 cm. The increased particle residence time at high temperature results in
densification that leads to inhomogeneous particle size distribution (Fig. 2.2d) consistent
with Tani et al. [19] who monitored the evolution of hollow particles made by FSP. When
using YN0.5, the hydrate water concentration is rather low so that formation of partially
hydrolyzed zirconium is less favorable and therefore more homogeneous powder is made
(Fig. 2.2a, b). As the combustion enthalpy density increases from 3.7 to 10.3 kJ/ggas
(corresponding to increasing the production rate from 43 to 342 g/h) larger particles
are obtained. As the FSP gas-to-liquid mass ratio (GLMR) between dispersion gas and
liquid feed rate decreases from 7.1 to 0.9 for that increase in powder production rate,
larger droplets are formed [12, 20] leading to larger particles as expected for the increased
solids concentration and enthalpy content. The overall yttria content in the particles made
from YN6 is consistent (10 mol% measured by EDS) with the precursor inlet Zr/Y ratio.
These particles have, however, inhomogeneous composition as the average Y content in
the large and small particles is 11.2 and 2.9 mol%, respectively. EDS analysis showed
that the Y-content was constant within the large particle fraction of the inhomogeneous
powders. The large particles are formed directly from the droplets. This indicates also the
simultaneous co-precipitation of yttrium and zirconium precursors. The high H2 O content
of YN6 favors formation of Zr(OH)4 and simultaneous co-precipitation with YN resulting
in a slightly enhanced Y-content in the precipitate near the surface of the precursor
droplets that would formed the large particles. The smaller ones are formed in the gas
phase from unreacted zirconium propoxide and yttrium nitrate which reacts also in the
gas phase. Only small amounts of the yttrium nitrate can go into the gas phase to form
a solid solution with the zirconia. This may explain the low yttria content in the small
particles. Using YN0.5 there is less H2 O to react with ZP and form precipitated Zr(OH)4 .
As a result, both precursors evaporate and form small and homogeneous particles (Fig.
2.2a, b). Here an average yttria content of 8.3 mol% is obtained that is consistent for all

34
particles in contrast to YN6-made particles. The slightly smaller yttria content in YSZ
made from YN0.5 compared to the theoretical 10 mol% results from incomplete dissolution
of yttrium nitrate in EtOH. The solution is filtered before filling into the piston pump
and therefore some of the metered yttrium nitrate is removed from solution. As a matter
of fact, YN0.5 crystals were observed at the bottom of the container and on the filter.
In contrast, YN6 dissolves better than YN0.5 in EtOH so all the yttrium added into the
precursor solution ends up in the particles as discussed above since no precipitates in that
precursor solution were observed.
Figure 2.3 shows TEM images of a) pure zirconia [21] as well as YSZ containing b)
3, c) 5 and d) 8 mol% yttria using YN6/ZP/EtOH at a production rate of 200 g/h. The
YSZ morphology changes from homogeneous (3 mol%) to inhomogeneous powder with
increasing yttria content. Increased water amount from YN6 in the precursor solution
leads to formation of Zr(OH)4 and hence to inhomogeneous product particles. This is in
agreement with Yuan et al. [10] who reported the formation of spherical and dense, pure
ZrO2 particles while for 10 mol% YSZ they found hollow and porous particles.
Figure 2.4 shows TEM images of YSZ particles using YN6/AcAn/ZP/ EtOH solutions. Although the hydrate is completely removed from the precursor solution by the
AcAn, inhomogeneous product powder is made. Here the AcAn reacts with the YNhydrate to form AcOH that reacts with yttrium nitrate to form yttrium acetate (YA).
The remaining AcOH reacts then with ZP to form zirconium acetate (ZA) [22]. The ZA
(decomposition point = 320 C, [7]) and YA precipitate at the supersaturated surface of
the shrinking droplet (EtOH boiling point = 79 C) and form a crust resulting in hollow or
shell-like particles [17] that solidify within the flame. This is supported by EDS showing
that the yttria content of the large particles is higher than that of the small ones that are
made by ZP that evaporated from the droplets. Therefore inhomogeneous particles with
a bimodal size distribution are formed as it is shown on TEM images (Fig. 2.4). This is
consistent with Karthikeyan et al. [9] who made YSZ with a broad size distribution using
yttrium and zirconium acetate as precursors.

35

Figure 2.3: a) Pure zirconia [21] and YSZ powders with b) 3, c) 5, and d) 8 mol%
Y2 O3 made by FSP of a solution of yttrium nitrate hexahydrate (YN6) and zirconium
n-propoxide (ZP) in EtOH at a production rate of 200 g/h.

Figure 2.5 shows homogeneous YSZ powder containing 10 mol% of yttria made by
spraying a solution of YN6/EtOH/ZC/AcOH/2-EHA at a) 54 and b) 342 g/h powder
production rates. The 2-EHA solvent reacts with the Y- or Zr-precursor to form Y- and
Zr-octoate (2-ethylhexnoate). These octoates have lower melting point than the solvent
boiling point preventing solid precipitation in the droplet [23]. Using precursor with car-

36

Figure 2.4: a) YSZ powders containing 10 mol% Y2 O3 made by FSP of a solution of


yttrium nitrate hexahydrate (YN6) in acetic anhydride and zirconium n-propoxide (ZP)
in EtOH at production rates of a) 54 and b) 342 g/h.
boxylate ligands (2-ethylhexanoate in this work) facilitates evaporation and subsequent
particle nucleation in the gas phase followed by coagulation, sintering and agglomeration
[24]. Here, the zirconium and yttria precursors have similar volatility and decomposition
pathways. This leads to a uniform metal distribution in the gas phase so yttria and zirconia can be formed together resulting in YSZ solid solution. The EDS analysis shows
homogeneous yttria distribution at any measured location within these particles. Stark et
al, [25] produced by FSP homogeneous mixed ceria/zirconia nanocrystals with narrow size
distribution using cerium (III) acetate hydrate and zirconium tetraacetylacetonate dissolved in lauric/acetic acid mixture. Schulz et al. [26] also produced homogeneous silicaand alumina-doped ceria/zirconia nanoparticles of high crystallinity using zirconium carbonate, cerium (III) acetate hydrate, aluminum 2-ethylhexanoate and tetraethoxysilane
as precursor in 2-EHA and toluene.
Figure 2.6 shows 10 mol% YSZ powders made by spraying YB/ZP/ toluene at production rates of a) 108 and b) 216 g/h. Here, also homogeneous YSZ powder is produced

37

Figure 2.5: YSZ containing 10 mol% Y2 O3 made by FSP of a solution of yttrium nitrate
hexahydrate (YN6) and zirconium carbonate hydroxide oxide (ZC) in acetic acid and
2-ethylhexanoic acid at production rates of a) 54 and b) 342 g/h.

and the morphology is comparable to that made using YN6/EtOH/ZC/AcOH/2-EHA


(Fig. 2.5. The melting point of ZP (-70 C) and even the boiling point of YB (109 C)
are lower than the boiling point of the solvent toluene (110 C) resulting in homogeneous
particles. At these conditions no precipitation can take place on the droplet surface during its evaporation as there is no concentration gradient in the droplet. The precursor
remains in solution as the droplet evaporates. This is in agreement with Ishizawa et al.
[27] who prepared homogeneous YSZ powder by spray pyrolysis of zirconium n-butoxide
and yttrium isopropoxide in EtOH and never observed hollow or inhomogeneous product.
Although alkoxide precursors result in a homogeneous product, these are too expensive
to be used in YSZ manufacturing. If the ratio of the solvent boiling point to the precursor melting/decomposition point, TR , is smaller than one, inhomogeneous powders
are expected [23]. Here, in all cases TR < 1 when inhomogeneous powders are formed.
Jossen et al. [23] reported the formation of homogeneous powders when increasing the
combustion enthalpy density above 4.7 kJ/ggas . In our study the combustion enthalpy is

38

Figure 2.6: YSZ (10 mol% Y2 O3 ) made by FSP of a solution of yttrium butoxide (YB)
and zirconium n-propoxide (ZP) in toluene at production rates of a) 108 and b) 216 g/h.
about 10 kJ/ggas for the highest production rate and still inhomogeneous powders were
formed. Madler et al. [28] reported for ceria a TR = 0.68 and homogeneous powders were
not formed until a combustion enthalpy density of 5.6 kJ/ggas is used. Here the TR for
EtOH/ZA is 0.59 and for EtOH/Zr(OH)4 it is 0.46. This indicates that for low TR , the
combustion enthalpy density has to be larger than 4.7 kJ/ggas homogeneous particles to
be formed.

2.3.2

YSZ crystal and primary particle sizes

Figure 2.7 shows XRD patterns of YSZ powder made from a) YN6/ZP/EtOH at 54
g/h (Fig. 2.2c), b) YN6/AcAn/ZP/EtOH at 54 g/h (Fig. 2.5a), c) YN0.5/ZP/EtOH
(Fig. 2.2a) at 43 g/h and d) YN6/EtOH/ZC/AcOH/2-EHA at 54 g/h (Fig. 2.6a). All
patterns show the formation of a solid YSZ solution regardless of its morphology. The
XRD of inhomogeneous YSZ (Fig. 6a,b) shows very sharp peaks (full width half maximum
(FWHM) < 0.2) corresponding to large particles. The peak shift in Figure 2.7b may come

39

Figure 2.7:

XRD spectra of YSZ powders containing 10 mol% Y2 O3 made

from a) YN6/ZP/EtOH, b) YN6/AcAn/ZP/EtOH, c) YN0.5/ZP/EtOH and d)


YN6/EtOH/ZC/AcOH/2-EHA. In all cases the stable cubic structure was identified.

from the inhomogeneous yttria distribution in the YSZ powder. All XRD patterns from
inhomogeneous powders as those made from YN6/ZP/EtOH and YN6/AcAn/ZO/EtOH
showed always peak shifts (not shown here). The shifts come from the irregular yttria
distribution into the zirconia. Stark et al. [29] showed that increasing the ceria fraction
in Cex Zrx-1 O2 shifted the peak position towards higher diffraction angles.
In contrast, the XRD of homogeneous (Fig. 2.7d) and the mostly homogeneous

40

Figure 2.8: Effect of production rate on the a) dBET or dXRD and b) large YSZ particle
mass fraction made from YN6/ZP/EtOH, YN6/AcAn/ZP/EtOH or YN0.5/ZP/EtOH
solutions. Increasing the production rate increases the solids concentration and process
enthalpy density as liquid feed rate increases while the dispersion gas flow rate is constant.

(Fig. 2.7c) YSZ show broad peaks (FWHM > 0.6) that are typical for nanosized crystals.
Ishizawa et al. [27] showed that cubic and only minor amounts of tetragonal structures
are formed by spray pyrolysis when using 6 mol% Y2 O3 . Yuan et al. [10] reported also
the formation of cubic 10 mol% YSZ as well as Shukla et al. [8] for 8 and 10 mol% YSZ.
For the inhomogeneous (Fig. 2.7a,b) and for the mostly homogeneous powders (Fig. 2.7c)
two average cubic crystal sizes can be derived from XRD [28].
Figure 2.8a shows the dXRD of the small (filled symbols) and large (open symbols)
size fractions of YSZ made from YN6/ZP/EtOH (squares), YN6/AcAn/ZP/EtOH (triangles), and YN0.5/ZP/EtOH (diamonds) as a function of production rate. Figure 2.8b
shows the mass fraction of the large crystal fraction as a function of production rate. Using YN6/AcAn/ZP/EtOH and YN6/ZP/EtOH, large particles constitute about 60 and
85 wt%, respectively, of the XRD mass fraction of powders made at low production rate
while the large particle mass fraction increases to 85 and 99 wt%, respectively, at high

41
production rate. Increased production rates are achieved by increased supply of liquid
precursor resulting in larger droplets amplifying, thus, the inhomogeneity of the product. For YSZ made at 54 g/h using YN0.5/ZP/EtOH (Fig. 2.8a) also a bimodal crystal
size distribution was obtained that may appear inconsistent with TEM (Fig. 1a). The
corresponding XRD mass fraction of large particles is 17 wt% (Fig. 2.8a) which means
that only few larger particles exist by count that is typical for TEM. Increasing the production rate (e.g. 342 g/h, Fig. 2.8b) increases the fraction of larger particles to about
85 wt% consistent with TEM (Fig. 2.2b). Anyway, the dXRD for the particles made at
high production rate is about 60 nm and that for the small ones is 40 nm. This difference in dXRD is much less than that from YSZ powder made using YN6/ZP/EtOH
(120 and 6 nm, squares). As a result YSZ made from YN0.5 will tend to appear visually more homogeneous than that made from YN6. Taking the XRD diameter ratio,
RXRD = dXRD,s /dXRD,l , of small to that of large particles shows that inhomogeneous
(made from YN6/EtOH and YN6/AcAn) powders have a very small such ratio (RXRD <
0.4) while for mostly homogeneous powder the RXRD is between 0.9 and 0.5. For truly
homogeneous powders (RXRD = 1) only one crystal size was found (Fig. 2.8b, circles,
butterflies). Thus, XRD can be used to distinguish between inhomogeneous and homogeneous crystalline powders. Comparing RXRD with the EDS analysis (yttria distribution)
shows that when RXRD approaches 1 the yttria distribution is becoming homogeneous.
Figure 2.9a shows the dXRD for cubic YSZ and dBET as function of the yttria content from
two precursor solutions. For more than 5 mol% Y2 O3 -containing powder (Fig. 2.2c,d), a
bimodal crystal size distribution is obtained from XRD. Figure 2.9b shows the increase
of the large particle mass fraction (squares) with increasing Y2 O3 -content. It has to be
pointed out that at 3 mol% yttria only one crystal size (29 nm) was detected which indicates homogeneous morphology. Also the dBET remains constant at 20 nm comparable
to pure zirconia. Adding more than 3 mol% yttria when using YN6 increases the hydrate
water content and enhances the formation of Zr(OH)4 resulting in inhomogeneous powders. The dXRD increases from 37 to 116 nm (squares) for the large particle size fraction

42

Figure 2.9: :Effect of Y2 O3 -content on the a) dBET or dXRD and b) larger particle size mass fraction of YSZ made at 200 g/h by FSP of YN6/ZP/EtOH or
YN6/EtOH/ZC/AcOH/2-EHA solutions.

and decreases from 32 to 7 nm (butterflies) for the small size mass fraction for 5 to 10
mol% Y2 O3 -content, respectively.
Figure 2.9a shows also the dBET and dXRD from the homogeneous powders using
YN6/EtOH/ZC/AcOH/2-EHA (circles, triangles). The yttria content has no influence on
dBET (circles) or dXRD (triangles) that remain constant at about 20 nm. This indicates
that the yttria content does not affect particle or crystal size for homogeneous powders.
The similarity of dBET and dXRD indicates also the presence of softly agglomerated, single
crystals from 0 to 10 mol% YSZ. The influence of higher yttria content (> 10 mol%) on
particle size and morphology is not shown here. Adding more than 17 mol% yttria to
zirconia leads to the formation of Y4 Zr3 O12 which is not of interest for sensor or fuel cell
applications. For the mostly inhomogeneous YSZ particles made from YN6/ZP/EtOH
(Fig. 2.3b, c, d), the dBET (diamonds in Fig. 2.9a) increases from 20 nm for pure zirconia
[21] to 97 nm for 10 mol% YSZ. For small amounts of yttria (3 mol%) the dBET remains
constant at 20 nm while above that it increases to 24 (5 mol%), 40 (8 mol%) and finally

43

Figure 2.10: Average BET-equivalent and XRD (diamonds) diameter of homogeneous,


pure (open symbols) or yttria-stabilized zirconia (filled symbols) made from ZP/EtOH
[21] or YN6/EtOH/ZC/AcOH/2-EHA respectively, as a function of powder production
rate or process enthalpy and particle concentration) at O2 dispersion gas flow rate of 25
and 50 l/min.

to 97 nm (10 mol%) which is inconsistent with TEM and XRD. This indicates that the
dBET is not a reliable measure of particle size for inhomogeneous powders.
Figure 2.10 shows the dBET for 10 mol% YSZ powder (filled symbols) made from

44
YN6/ETOH/ZC/AcOH/2-EHA and pure zirconia (open symbols [21]) made from ZP/
EtOH for 25 (triangles) and 50 (squares) l/min O2 dispersion gas. The dBET of homogeneous YSZ and ZrO2 are identical when made at the same production rate. For example,
at a production rate of 324 g/h the combustion enthalpy density is 10.6 kJ/ggas and the
flame height is 391 cm for both FSP processes. This indicates that combustion enthalpy
density and precursor concentration are the main process parameters to control particle
size for homogeneous powders as is with vapor-fed flame reactors [30]. This similarity in
particle size of pure zirconia and YSZ indicates that the yttria content has little influence on particle size of homogeneous powders as also can be seen in Figure 2.9a (circles,
triangles). Using 25 l/min O2 , the dBET increases from 12 to 29 nm when increasing
the production rate from 54 to 216 g/h. At O2 dispersion gas flow rate of 50 l/min the
dBET is controlled from 8 to 31 nm when increasing the production rate from 54 to 324
g/h. At higher O2 flow rates mixing is intensified and combustion is accelerated shortening the flame height [31]. When increasing the O2 dispersion gas flow rate from 25
to 50 l/min, the flame height decreases, for example, from 40 to 30 cm at a production
rate of 217 g/h. The increase of the O2 dispersion gas flow rate decreases the droplet
concentration of the spray and the flame enthalpy content, thus, particle concentration
and particle residence time at high temperature decreases. This leads to faster quenching
and therefore smaller particles are formed. The corresponding dXRD (circles, diamonds)
closely follow the dBET at low production rate (up to 150 g/h) indicating single crystals of
YSZ while for production rates higher that 200 g/h dXRD is smaller than dBET indicating
polycrystallinity or increased degree of agglomeration. The dXRD of the inhomogeneous
product powder increases from 75 to 120 nm (YN6/ZP/EtOH) and from 75 to 110 nm
(YN6/AcAn/ZP/EtOH) as shown in Figure 2.8a. For the inhomogeneous YSZ powders,
large particles are formed directly from droplets as the solids precipitate at the droplet
surface. The average droplet diameter is 14 m at low production rate (54 g/h, 13.5 ml/min
liquid feed rate) while it increase to 22 m at the highest production rate (354 g/h, 81.1 liquid feed rate) [12, Appendix A1.1]. The large particles are directly related to the droplet

45
size. The above droplet size ratio at high to low production rate is 1.57 which is in good
agreement with the dXRD ratio of the large particles at high to low production rates, 1.6
(Fig. 2.8a: YN6/ZP/EtOH squares). At higher production rate also the residence time
in the hot temperature zone is increased which has a minor influence on large particles
growth. For the homogeneous particles (Fig. 2.10), particle formation takes place in the
gas phase from evaporated precursors resulting in smaller particles. For the latter, the
sintering time is smaller than for large particles ( d4p , [32]) resulting in faster particle
growth. By increasing the liquid flow rate (increasing the powder production rate) particle concentration and high temperature residence time are increasing leading the larger
particles. This may explain the increase of dXRD of homogeneous FSP-made YSZ powders
by a factor of 2-4 in Figure 2.10.

2.4

Conclusions

A systematic investigation of flame spray synthesis of nanostructured yttria-stabilized


zirconia (YSZ) was carried out at production rates up to 350 g/h. The precursor composition affects the morphology of YSZ powders made at identical combustion enthalpy
density, precursor concentration and constant process conditions. Inhomogeneous YSZ
powders exhibit a bimodal crystal size distribution. The diameter ratio between the two
modes can be used to estimate the degree of inhomogeneity. The average grain diameter
as determined by nitrogen adsorption of inhomogeneous YSZ powders does not represent
well the product particle characteristics.
The homogeneity of YSZ particles was improved by reducing the water content
in the precursor solutions. Flame-made YSZ particles have homogeneous composition
and morphology when using either organometallic precursors or 2-ethylhexanoic acid as
solvent and inexpensive zirconium carbonate as precursor. Increasing the production rate
increased drastically the dBET and dXRD of YSZ as both enthalpy content and metal
concentration increased resulting in larger particles and crystals. The yttria content (up

46
to 10 mol%) has little influence on homogeneous YSZ particle size and morphology.

References
[1] A. R. Pebler. Preparation of small particle stabilized zirconia by aerosol pyrolysis.
J. Mater. Res., 5(4):680682, 1990.
[2] R Nielsen, T. Wah, and A. Albany. Ullmanns encyclopedia of industrial chemistry,
volume 28 A. Wile-VCH Veralg GmbH, 1996.
[3] C. Pascual and P. Duran. Subsolidus phase-equilibria and ordering in the system
ZrO2 -Y2 O3 . J. Am. Cem. Soc., 66(1):2327, 1983.
[4] G. S. Pang, E. Sominska, H. Colfen, Y. Mastai, S. Avivi, Y. Koltypin, and
A. Gedanken. Preparing a stable colloidal solution of hydrous YSZ by sonication.
Langmuir, 17(11):32233226, 2001.
[5] S. Benard, L. Retailleau, F. Gaillard, P. Vernoux, and A. Giroir-Fendler. Supported
platinum catalysts for nitrogen oxide sensors. Appl. Catal. B-Env., 55(1):1121, 2005.
[6] M. Gaudon, E. Djurado, and N. H. Menzler. Morphology and sintering behaviour
of yttria stabilised zirconia (8-YSZ) powders synthesised by spray pyrolysis. Ceram.
Inter., 30(8):22952303, 2004.
[7] S. C. Zhang, G. L. Messing, and M. Borden. Synthesis of solid, spherical zirconia
particles by spray pyrolysis. J. Am. Ceram. Soc., 73(1):6167, 1990.
[8] A.K. Shukla, N.A. Dhas, and K.C. Patil. Oxides ion coduction of calcia and yttria
stabilized zirconia prepared by rapid combustion rout. Mater. Sci. Eng., B40:153
157, 1996.
[9] J. Karthikeyan, C. C. Berndt, J. Tikkanen, J. Y. Wang, A. H. King, and H. Herman.
Nanomaterial powders and deposits prepared by flame spray processing of liquid
precursors. Nanostruct. Mater., 8(1):6174, 1997.
[10] F. L. Yuan, C. H. Chen, E. M. Kelder, and J. Schoonman. Preparation of zirconia
and yttria-stabilized zirconia (YSZ) fine powders by flame-assisted ultrasonic spray
pyrolysis (FAUSP). Solid State Ion., 109(1-2):119123, 1998.
[11] Y. Q. Xie. Preparation of ultrafine zirconia particles. J. Am. Ceram. Soc., 82(3):768
770, 1999.

47
[12] R. Mueller, L. Madler, and S. E. Pratsinis. Nanoparticle synthesis at high production
rates by flame spray pyrolysis. Chem. Eng. Sci., 58(10):19691976, 2003.
[13] D. R. Lide. CRC Handbook of Chemistry and Physics. CRC Press, Inc., Boca Rota,
Florida, 3rd electorn edtion edition, 2000.
[14] R.C Rau. Routine crystallite-size determination by X-ray diffraction broadening.
Adv.X-Ray Anal., 5:105, 1962.
[15] M. Morinaga, J. B. Cohen, and J. Faber. X-ray-diffraction study of Zr(Ca,Y)O2x .
1. Average structure. Acta Crystallogr. Sect. A, 35:789795, 1979.
[16] R. W. Cheary and A. A. Coelho. Axial divergence in a conventional X-ray powder
diffractometer: I. Theoretical foundations. J. Appl. Crystal., 31:851861, 1998.
[17] A. Gurav, T. Kodas, T. Pluym, and Y. Xiong. Aerosol processing of materials. Aeros.
Sci. Technol., 19(4):411452, 1993.
[18] M. Z. C. Hu, J. T. Zielke, C. H. Byers, J. S. Lin, and M. T. Harris. Probing the
early-stage/rapid processes in hydrolysis and condensation of metal alkoxides. J.
Mater. Sci., 35(8):19571971, 2000.
[19] T. Tani, K. Takatori, and S. E. Pratsinis. Dynamics of hollow and solid alumina
particle formation in spray flames. J. Am. Ceram. Soc., 87(3):523525, 2004.
[20] R. Mueller. Characterization and synthesis of nanoparticles made in vapor and spray
flames. Ph.D. Thesis (ETH No. 15147), Department of Mechanical and Process
Engineering, 2003.
[21] R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, and M. K. Akhtar. Zirconia
nanoparticles made in spray flames at high production rates. J. Am. Ceram. Soc,
87(2):197202, 2003.
[22] J.R Lacher, J. Schwarz, and J.D Park. Nitration with uranium nitrate-nitrogen
tetraoxide-water complexe in the precense of acetic anhydride. J. Org. Chem.,
26:25362537, 1960.
[23] R. Jossen, S. E. Pratsinis, W. J. Stark, and L. Madler. Criteria for flame-spray synthesis of hollow, shell-like or inhomogenous oxides. J. Am. Ceram. Soc., 88(6):1388
1393, 2005.
[24] S. E. Pratsinis. Flame aerosol synthesis of ceramic powders. Prog. Energy Combus.
Sci., 24(3):197219, 1998.

48
[25] W. J. Stark, L. Madler, M. Maciejewski, S. E. Pratsinis, and A. Baiker. Flame
synthesis of nanocrystalline ceria-zirconia: effect of carrier liquid. Chem. Commun.,
5:588589, 2003.
[26] H. Schulz, W. J. Stark, Maciejewski M., S. E. Pratsinis, and A. Baiker. Flame-made
nanocrystalline ceria/zirconia doped with alumina or silica: Structure properties and
enhanced oxygene exchange capacity. J. Mater. Chem., 13(12):29792984, 2003.
[27] H. Ishizawa, O. Sakurai, N. Mizutani, and M. Kato. Homogeneous Y2 O3 -stabilized
ZrO2 powder by spray pyrolysis method. Am. Ceram. Soc. Bull., 65(10):13991404,
1986.
[28] L. Madler, W. J. Stark, and S. E. Pratsinis. Flame-made ceria nanoparticles. J.Mater.
Res., 17(6):13561362, 2002.
[29] W. J. Stark, M. Maciejewski, L. Madler, S. E. Pratsinis, and A. Baiker. Flame-made
nanocrystalline ceria/zirconia: structural properties and dynamic oxygen exchange
capacity. J. Catal., 220(1):3543, 2003.
[30] H. Briesen, A. Fuhrmann, and S. E. Pratsinis. The effect of precursor in flame
synthesis of SiO2 . Chem. Eng. Sci., 53(24):41054112, 1998.
[31] I Glassman. Combustion. Academic Press, San Diego, 1996.
[32] R. Mueller, R. Jossen, H. K. Kammler, and S. E. Pratsinis. Growth of zirconia
particles made by flame spray pyrolysis. AIChE J., 50(12):30853094, 2004.

CHAPTER

THREE

Thermal stability of flame-made zirconia-based mixed


oxides

Abstract
The thermal stability of pure and doped zirconia made by flame spray pyrolysis (FSP)
was studied by calcination at various temperatures and residence times. Powders were
analyzed by X-ray diffraction (XRD), nitrogen adsorption (BET), energy-dispersive X-ray
spectroscopy (EDS) and transmission electron micrograph (TEM). The metastable tetragonal structure of flame-made ZrO2 transformed into monoclinic after sintering. Doping
zirconia with yttria, ceria, lanthanum oxide, alumina or silica increased its thermal stability and also hindered that phase transformation. For all doped powders no monoclinic
phase was formed during calcination. The combination of 25 wt% ceria, 10 wt% lanthanum oxide and 65 wt% zirconia had the highest thermal stability.

3.1

Introduction

Tetragonal and cubic zirconia are of major importance in the engineering of advanced
materials for catalysis, solid oxide fuel cells, oxygen sensors and optics [1-3]. Tetragonal
zirconia is used as a structural material and as catalyst like in the hydrogenation of benzoic
49

50
acid to benzaldehydes [4] as well as in the isomerization of alkanes [5]. Cubic zirconia
doped with yttria or scandia is commonly used as electrolyte in solid oxide fuel cells. Three
different crystal phases exist for pure zirconia. The stable polymorph at room temperature
is monoclinic (m-ZrO2 ) that transforms to tetragonal (t-ZrO2 ) at 1170 C, and then to
fluorite-type cubic (c-ZrO2 ) at 2370 C [1, 6]. The most desirable t-ZrO2 is metastable
at room temperature and changes into m-ZrO2 during calcination at 700 C. This phase
transformation reduces the thermal stability and decreases the activity of zirconia as
catalyst [7]. As most applications for tetragonal and c-ZrO2 are below 1170 C (e.g. solid
fuel cell) there is a strong need to stabilize these high temperature phases. This can be
done by reducing the particle and crystal size to the nanometer regime [1, 6, 8]. Sunresh
et al. [9] showed that the transformation from tetragonal to monoclinic depends only on
crystal size. For pure zirconia they proposed a crystal size of 22.6 nm and for 1.5 mol%
yttria-doped zirconia this critical crystal size increases to 93.8 nm. Doping zirconia with
earth metals like Mg2+ , Ca2+ or Y3+ and Sc3+ is another possibility to stabilize that phase
and to avoid sintering [10]. Li et al. [11] investigated the transformation of metastable
t-ZrO2 to m-ZrO2 from sintered Zr(OH)4 by UV Raman spectroscopy and suggested that
the t-m transformation initially starts at the crystal surface of the calcined samples before
propagating into the interior of the particle. This is consistent with Li and Li [12] who
showed that impregnating zirconia with sulfate hinders the phase transformation from tZrO2 to m-ZrO2 and stabilizes the tetragonal phase even up to 800 C. Aozasa [13] showed
that zirconia doped with 25 wt% ceria and 10 wt% lanthanum oxide shows high thermal
stability compared to alumina-, lanthanum oxide- or ceria-doped zirconia: After sintering
at 900 C for 6 hours 82 m2 /g is conserved.
Mixed oxides with up to three components were made here in a single process step
using flame spray pyrolysis (FSP) at production rates up to 140 g/h. The thermal stability
of zirconia doped with yttria, alumina, lanthanum oxide, ceria and silica was investigated.

51

3.2

Experimental

3.2.1

Precursor preparation and thermal stability characterization

Tetraethoxisilane (TEOS, Aldrich), Lanthanum (III) acetylacetonate hydrate (Strem Chemicals, 99.9%), yttrium nitrate hexahydrate (Aldrich, 99.9%), ceria octoate (Soctech S.A)
and aluminum ethylhexanoate basic (Strem Chemicals) were used as dopant precursors
while zirconium (IV) n-propoxide (Chempur, 70wt% in propanol) was used as zirconia
precursor. Lanthanum (III) acetylacetonate hydrate was dissolved in acetic anhydride
and 2-ethylhexanoic acid (2-EHA), yttrium nitrate hexahydrate in EtOH and 2-EHA [14]
and aluminum ethylhexanoate basic in 2-EHA. All solutions were distilled under nitrogen
to remove acetic acid, acetylacetone, water, EtOH and NO [14] until clear solutions were
formed. All solutions were diluted with toluene resulting in a total metal concentration
of 0.5 M.

3.2.2

Apparatus and processing

The experimental set-up of the pilot-FSP consists of a commercially available externalmixing stainless-steel gas-assisted atomizer (Chapter 2, Fig. 2.1) [15]. Oxygen is used as
dispersion gas. A 1-liter precision piston pump (Isco, Inc., 1000D) is used for pulsationfree supply of the precursor solution through the capillary tube. The resulting spray
is ignited by a supporting methane/oxygen diffusion flame (CH4 = 2 l/min, O2 = 4.5
l/min) surrounding the nozzle. The specific surface area of the powders was controlled by
increasing the liquid feed rate of the precursor solution, e.g. from 27.1 to 81.1 ml/min,
resulting in production rates of 100 to 300 g/h zirconia. The effect of scaling-up on particle
size [15], droplet size [16] and flame temperature [17] is given elsewhere.
Product powders are collected in a commercial Jet filter (Friedli AG, FRR 12/2.4)
consisting of twelve PTFE (polytetrafluoroethylene, Teflon) coated Nomex baghouse fil-

52
ters (total surface area: 10.2 m2 ) which are cleaned periodically by air pressure shocks.
Small samples ( 1 g) of product particles are collected with the aid of a vacuum pump
(Vaccubrand RE 16) on a glass fiber filter (Whatman GF/A) 150 mm in diameter located
in a stainless steel holder by a bypass connected to the Jet filter inlet pipe. Powders collected on the small glass fiber filter were identical (same specific surface area, crystallinity
and morphology; Appendix D) to the ones collected with the baghouse filters [15].
The thermal stability of the as-prepared powder samples was studied by placing
them in a furnace under air applying a heating and cooling rate of 10 K/min for 6 hours
at 900 C. The thermal stability of pure zirconia was also measured as function of time at
800 and 1000 C. For each experimental point a new sample was used.

3.2.3

Powder characterization

The specific surface area (SSA) is determined from a five-point N2 adsorption isotherm
in the relative pressure range of 0.05 to 0.25 at 77.3 K (BET analysis) using a Tristar
2000 (Micromeritics Instruments Corp.). Before N2 adsorption, all samples are degassed
(Flow prep 060, Micromeritics Instruments Corp.) under N2 at 150 C for 1 hour, to
remove water bound to the particle surface from air moisture. X-ray diffraction (XRD)
spectra are recorded with a Bruker D8 advanced diffractometer from 20 to 80 , step
0.03 and a scan speed 0.6 /min. Crystalline characteristics and average crystal size,
dXRD , are obtained from the XRD spectra using the Topas 2.0 software (Bruker AXS,
2000) on the basis of the fundamental parameter approach (Rietveld method) [18, 19]
in which the effects of the equipment (e.g. X-ray source, slits, etc.) are incorporated.
The morphology of the product powder is obtained by transmission electron microscopy
(TEM) performed on a Hitachi H600 and on a JEOL 2000FX II electron microscopes
operating at 100 kV and 200 kV, respectively, using magnification ranging from 50 to
800k. The energy-dispersive X-ray spectroscopy (EDS) analyses are collected with an
EDAX Genesis system with a resolution of 130 eV to analyze the doped material content
in the stabilized zirconia solution. During the imaging process particular attention is

53
given to characterizing particle morphology and doping distribution. Images are collected
with Gatan BioScan and MultiScan CCD cameras. Mixed oxides are made of ZrO2 (Z),
Y2 O3 (Y), CeO2 (C), SiO2 (S), La2 O3 (L) and Al2 O3 (A). All these are labelled as xByDZ,
where x and y are the weight fractions (%) of B and D oxide, respectively, and the balance
is the weight fraction (%) of ZrO2 . For example 5L25CZ means 5 wt% La2 O3 , 25 wt%
CeO2 and 70 wt% ZrO2 .

3.3

Results and Discussion

3.3.1

Pure ZrO2

Figure 3.1 shows the specific surface area (SSA) of pure ZrO2 made at 100 g/h (triangles,
initial SSA = 62 m2 /g) and 300 g/h (circles: initial SSA = 36 m2 /g) by FSP as a function
of calcination time at 800 C (filled symbols) and 1000 C (open symbols), respectively. The
SSA decreased from 62 to 27 m2 /g and from 36 to 22 m2 /g after 6 hours calcination at
800 C. When calcining zirconia at 1000 C for 6 hours the SSA for both powders decreased
to 12 m2 /g within the first two hours and then remained constant. It took only about 30
minutes for the initial difference in the higher SSA of 35 m2 /g to be levelled off (Fig. 3.1,
open circles, triangles).
Figure 3.2 shows the XRD patterns of pure zirconia for a) as-prepared and b) calcined zirconia at 900 C for 6 hours. The as-prepared zirconia contains 87 wt% of tetragonal phase while during calcination 96 % of monoclinic zirconia is formed with the balance
being tetragonal phase. This is in agreement with Mercera et al. [7] who investigated the
thermal stability of pure zirconia prepared by precipitation from zirconyl chloride at pH
= 10. They calcined the particles in air for up to 850 C and showed that the monoclinic
crystallites exhibit a more marked growth during calcination than the tetragonal ones. It
has been documented that the transformation takes place at 1200 C from monoclinic to
tetragonal, while the reverse takes place between 1150 and 1100 C. Here, the crystal size
for the as-prepared tetragonal zirconia is 12.6 nm and that for the sintered monoclinic

54

Figure 3.1:

SSA of zirconia nanoparticles as a function of calcination time at 800 C

(filled symbols) and 1000 C (open symbols) for an specific surface area of 62 (triangles)
and 36 m2 /g (circles)
.

phase it increases to 73.7 nm consistent with Sunresh et al. [9]. Mueller et al. [20] showed
that flame-made zirconia contains tetragonal phase also at particle sizes smaller 25 nm.
The thermal stability of ZrO2 also increased by doping with other metal oxides as also
shown by Stark et al.[21].

55

Figure 3.2: X-ray diffraction (XRD) patterns of a) as-prepared and b) calcined zirconia
at 900 C for 6 hours. The XRD of the as-prepared powders show mainly tetragonal phase
while after calcination mostly monoclinic phase has formed.

3.3.2

Yttria-doped zirconia

Figure 3.3 shows the dBET (circles) and dXRD (squares) of yttria-doped zirconia which
is commonly called yttria-stabilized zirconia (xYZ, where x is the wt% of yttria) of asprepared (open symbols) and calcined (filled symbols) for 6 h at 900 C as function of
the yttria content. Here, the Y2 O3 -content does not affect the as-prepared particle (open

56

Figure 3.3: dXRD (squares) and dBET (circles) as function of the yttria content for asprepared (filled symbols) and calcined xYZ (open symbols). The yttria content has no
influence on crystal size even when calcined the xYZ at 900 C for 6 hours.

circles) and crystal size (open squares) that remain constant at 18 nm (55 m2 /g) for all
cases. The similarity of dBET (open circles) and dXRD (open squares) indicates also that
as-prepared pure and yttria-doped zirconia particles are single crystals. After calcination
of pure zirconia its dBET increased from 18 to 71 nm while the crystal size increased only
from 18 to 39 nm, indicating that the resulting particles are polycrystalline. By increasing

57
the Y2 O3 -content to 14 wt% the thermal stability is improved further. For the 5, 9 and
14YZ (3, 5 and 8 mol% YZ) only tetragonal and cubic crystals were detected by XRD. For
all Y2 O3 -doped samples even after calcination no monoclinic phase was found regardless
of the Y2 O3 -content. After calcination the dBET decreases from 71 for pure ZrO2 to 34,
30 and 26 nm for 5, 9 and 14YZ, respectively. The dXRD of the calcined xYZ particles
decreases from 39 for pure ZrO2 to about 22 nm regardless of the Y2 O3 -content. For the
14YZ after calcination only cubic phase was formed while for the 9 and 5YZ the cubic
phase content was reduced to 98 and 96 wt%, respectively, with the balance being the
tetragonal phase. Khollam et al. [22] reported that the cubic phase of 18YZ (10 mol%
YZ) is maintained up to 800 C for 4 hours while at higher temperature the cubic phase
is destabilized into the monoclinic phase in contrast to our work where after sintering
at 900 C for 6 hours no monoclinic phase was found. Xu et al. [23] produced 3.5YZ (2
mol% YZ) containing 35 wt% monoclinic phase by a two step urea-based hydrothermal
synthesis. After calcining up to 1100 C no monoclinic phase was detected while at 900 C
the particle size increased from 10 to 25 nm (from TEM) which is comparable to this
work. The present study is consistent with Sunresh et al. [9] who pointed out that for
2.5YZ (1.5 mol% YZ) monoclinic ZrO2 will appear for particles above 93.8 nm, which is
above dBET of the present as-prepared and calcined powder.

3.3.3

Ceria-doped zirconia

Figure 3.4a depicts the as-prepared 35CZ powders that consist of well-crystalline, sharpedged nanoparticels of 5 to 10 nm in size. Figure 3.4b shows the same powder after
calcination: Regularly shaped particles with high crystallinity are obtained consistent
with Stark et al. [21].
Figure 3.5a shows the dXRD of as-prepared (open) and sintered CZ particles as
function of their ceria content (filled symbols). Pure zirconia has a crystal size of 12 nm
that decreases to 8.5 nm for 42CZ and increases again to 11 nm for pure ceria. After
calcination, the dXRD increases of pure zirconia to 38 nm while for 42CZ the dXRD has a

58

Figure 3.4: Transmission electron micrographs of a) as-prepared 35CZ, and b) corresponding samples after sintering at 900 C for 6 hours in air. Flame-made ceria/zirconia
forms highly crystalline nanoparticles of narrow particle-size distributions. Sintering affords well-defined, regular crystallites of about 15 nm in diameter. Surfaces are wellstructured and flat; no defects or twinned crystals appear.

minimum at 9.5 nm and for pure ceria it increases again to 17.5 nm. Comparing the dXRD
and the SSA of the as-prepared and the calcined powders show that 35CZ is most stable.
Figure 3.5b shows the SSA of as-prepared (open diamonds) and calcined (filled squares)
ceria-doped zirconia particles as function of the ceria content. Powders with 95 m2 /g
(35CZ) are made which is higher in terms of SSA than both pure zirconia (74 m2 /g) and
ceria (85 m2 /g). It can be seen from Figure 3.5b that mixed oxides are more stable than
either pure zirconia or ceria. An optimum was found for 35CZ also as 60 of the initial
95 m2 /g were conserved after calcination. This is in agreement with Stark et al. [21]
who made ceria/zirconia by FSP of cerium acetate and zirconium tetraacetylacetonate in
lauric/acetic acid solution. Their as-prepared powders had SSA up to 170 m2 /g by using
molar concentration of 0.15 M and a production rate ranging from 3.3 (pure zirconia)

59

Figure 3.5: a) dXRD and b) SSA of mixed ceria-zirconia oxide samples as-prepared (open
diamonds) and calcined (filled diamonds) ranging from pure zirconia (left) to pure ceria
(right).

to 4.6 g/h (pure ceria). Here, a 0.5 M solution was used at a liquid feed rate of 27.1
ml/min which corresponds to a production of 100 to 140 g/h depending on the Ce/Zr
ratio. The present higher particle concentrations and enthalpy content of the process
contribute to formation of particles with smaller SSA than Stark et al. [21]. However,
even at high production rates, the obtained SSA of the as-prepared and even the calcined
powders are larger than that of commercial powders. More specifically, the SSA of any
calcined powder mixture is larger than that of the Rhodia catalyst [24] (56 m2 /g) or the
precipitated made 25CZ from the corresponding nitrates by Bozo et al. [25] (42 m2 /g) or
Sutorik and Baliat [26] who achieved a SSA up to 26 m2 /g after calcination at 900 C for
2 h.
Figure 3.6 shows the effect of Al2 O3 or La2 O3 -doping on the dXRD and dBET of
a) pure zirconia and b) the optimal 35CZ (ZrO2 -content was kept constant). Adding
alumina to pure zirconia decreases the dBET of the as-prepared powders from 14 (74
m2 /g, pure zirconia) to 12.6 nm (86 m2 /g, 10AZ). This is in agreement with Lange and

60

Figure 3.6: dXRD and dBET of mixed oxide samples as-prepared (open symbols) and
after calcinations (filled symbols) of a) alumina doped-zirconia and b) lanthanum oxide
or alumina doped zirconia-ceria at the optimal condition. 10L25CZ shows higher thermal
stability as the optimal zirconia-ceria mixtures where 71 m2 /g of the initial specific surface
of 95 m2 /g are conserved.
Hirlinger [27] who showed that adding alumina to cubic zirconia suppresses grain growth
during sintering by colloidal processing. The thermal stability of alumina-doped zirconia
is similar to that of ceria- or yttria-doped zirconia. The dBET of the 10AZ increases from
12.6 (86 m2 /g) to 17.3 nm (63 m2 /g) which is similar to the optimal 35CZ after calcination
at 900 C for 6 hours. Even higher thermal stabilities can be obtained by doping 35CZ with
lanthanum oxide or alumina (Fig. 3.6b). After sintering the dBET is increased to about
13.5 nm (70 m2 /g) for all FSP-made lanthanum oxide-doped 35CZ powders (Fig. 3.6b).
The tetragonal phase content decreases with increasing alumina content from 92 (5AZ)
to 71 wt% (10AZ) with the balance being cubic phase. For the lanthanum oxide-doped
35CZ the cubic phase is dominant. Using 5 and 10 wt% lanthanum oxide 87 (5L30CZ)
and 97 wt% (10L25CZ) of cubic phase is formed for the mixed CeO2 /ZrO2 while doping
with 5 and 10 wt% alumina only 80 (5A30CZ) and 67 wt% (10A25C Z) of cubic phase

61

Figure 3.7: Transmission electron micrographs of a) as-prepared 10AZ, and b) corresponding samples after sintering. Flame-made 10AZ forms highly crystalline nanoparticles of narrow particle-size distributions. In contrast to 35CZ the particles are shaped
more irregularly with a tendency to form hexagons upon calcination.

is formed with the balance being tetragonal. For all doped zirconia powders the phase
content remains constant during calcination. The highest thermal stability was found
with 10L25CZ where 71 (13.5 nm) of the initial 80 m2 /g (11.7 nm) are conserved after
calcination, consistent with Aozasa [13] who showed that after calcining 10L25CZ at 900
for 6 hours its SSA decreases from the initial 102 to 82 m2 /g.
Figure 3.7 shows TEM-images of a) as-prepared (particle size about 10 nm) and
b) sintered 10AZ (particle size about 18 nm). The sizes obtained by TEM-images analysis agree well with the BET measurements (Fig. 3.6b). In contrast to 35CZ particles
(Fig. 3.4) the 10AZ ones have irregular shapes with tendencies to form hexagons upon
calcination.

62

Figure 3.8: Transmission electron micrographs of a) as-prepared 2SZ, and b) after calcination at 900 C for 6 hours in air. The powder consists of spherical primary particles.
Lattice planes are discernible indicating that particles are crystalline. Particles are coated
with a thin amorphous silica layer as determined by EDX.

3.3.4

Silica-doped zirconia

Figure 3.8 shows representative TEM-images of 2SZ of a) as-prepared and b) sintered at


900 C. The powder consists of spherical primary particles. Lattice planes are discernible
indicating that particles are crystalline. Particles are largely covered by thin amorphous
silica layers as determined by EDS. Here, for 2, 5 and 7SZ the SSA decreased from 83 to
60, 62 and 74 m2 /g, respectively, compared to 14.2 m2 /g for pure zirconia after calcination
at 900 C for 6 hours. Del Monte et al. [28] reported a critical particle size of 38.8 nm
for 2 wt% (4 mol%) silica which is larger than that of the FSP made particles even after
calcination (dXRD = 20 nm).
Figure 3.9 shows XRD patterns of 2SZ for a) as-prepared and b) calcined samples
as well as for 5 wt% silica-doped zirconia (c: as-prepared; d: calcined). The as-prepared
powders are mostly tetragonal structure (> 99% tetragonal) and do not change essentially

63

Figure 3.9: Fig. 9: X-ray diffraction patterns of 2 (a,b) and 5 wt% (c,d) silica doped zirconia of (a,c) as-prepared and (b,d) calcined at 900 C for 6 hours. For the as-prepared 2SZ
and 5SZ only tetragonal structure was formed while after calcining 1 wt% of monoclinic
phase was formed in both cases.
after sintering (>98% tetragonal). For the 7SZ even only tetragonal structure was found
after calcining. This is consistent with Wang et al. [29] who used wet chemical synthesis
to produce ZrO2 -SiO2 powders and found also only t-ZrO2 after sintering at 900 C for 2
hours in air. Del Monte et al. [28] produced silica-doped zirconia by wet phase chemistry
also and reported that for 2 mol% silica doped zirconia, the tetragonal crystal phase
remains stable up to 1000 C.

64

3.4

Conclusions

Flame-made doped zirconia nanoparticles produced at relative large production rates, up


to 140 g/h, have excellent thermal and crystallite stability that are superior to commercial powders and comparable to literature data. For pure zirconia the SSA decreases
drastically while a phase transition takes place from tetragonal to monoclinic. Adding
2 to 7 wt% of SiO2 to ZrO2 the tetragonal structure is stabilized during calcination and
the thermal stability is increased significantly. Yttria- or alumina-doped zirconia show an
enhanced thermal stability that increases for increasing doping levels while the formation
of monoclinic phase is prevented. With yttria, mainly the cubic phase is formed while
using alumina tetragonal phase is dominant. Adding only 13 wt% (10 mol%) ceria increases drastically the thermal stability of zirconia and with 35 wt% the highest specific
surface area is obtained after calcination. The highest thermal stability was found when
replacing 10 wt% of ceria with lanthanum oxide to that optimal ceria-zirconia and when
adding up to 7 wt% silica. Here, mostly the cubic phase is formed that is preserved during
sintering.

References
[1] T. Chraska, A. H. King, and C. C. Berndt. On the size-dependent phase transformation in nanoparticulate zirconia. Mater. Sci. Eng. A-Struct. Mater. Prop. Microstr.
Process., 286(1):169178, 2000.
[2] M. Gell. Application opportunities for nanostructured materials and coatings. Mater.
Sci. Eng. A-Struct. Mater. Prop. Microstr. Process., 204(1-2):246251, 1995.
[3] M. J. Mayo, J. R. Seidensticker, D. C. Hague, and A. H. Carim. Surface chemistry effects on the processing and superplastic properties of nanocrystalline oxide ceramics.
Nanostruct. Mater., 11(2):271282, 1999.
[4] K. Tanabe and T. Yamaguchi. Acid-base bifunctional catalysis by ZrO2 and its mixed
oxides. Catal. Today, 20(2), 1994.
[5] M. Hino and K. Arata. Synthesis of a highly-active superacid of platinum-supported
zirconia for reaction of butane. J. Chem. Soc.-Chem. Commun., (7):789790, 1995.

65
[6] R. C. Garvie. The occurrence of metastable tetragonal zirconia as a crystallite size
effect. J. Phy. Chem., 69(4):12381243, 1965.
[7] P. D. L. Mercera, J. G. Vanommen, E. B. M. Doesburg, A. J. Burggraaf, and J. R. H.
Ross. Zirconia as a support for catalysts - evolution of the texture and structure on
calcination in air. Appl. Catal., 57(1):127148, 1990.
[8] R. Ramamoorthy, D. Sundararaman, and S. Ramasamy. X-ray diffraction study of
phase transformation in hydrolyzed zirconia nanoparticles. J. European Ceram. Soc.,
19(10):18271833, 1999.
[9] A. Sunresh, M. J. Mayo, W. D. Porter, and C. J. Rawn. Crystallite and grainsize-dependent phase transformations in yttria-doped zirconia. J. Am. Ceram. Soc.,
86(2):360362, 2003.
[10] Y. W. Zhang, Y. Yang, S. Jin, C. S. Liao, and C. H. Yan. Long time annealing
effects on the microstructures of the sol- gel-derived nanocrystalline thin films of rare
earth-stabilized zirconia. J. Mater. Chem., 11(8):20672071, 2001.
[11] M. J. Li, Z. H. Feng, G. Xiong, P. L. Ying, Q. Xin, and C. Li. Phase transformation
in the surface region of zirconia detected by uv raman spectroscopy. J. Phys. Chem.
B, 105(34), 2001.
[12] C. Li and M. J. Li. Uv raman spectroscopic study on the phase transformation of
ZrO2 , Y2 O3 -ZrO2 and SO4
2 -/ZrO2 . J. Raman Spectrosc., 33(5):301308, 2002.
[13] S. Aozasa.
Zirconium-cerium-verbundoxid,
cokatalysator zur reinigung von abgasen, 2003.

verfahren zur herstellun und

[14] R. Jossen, R Mueller, S. E. Pratsinis, M. Watson, and K. M. Akhtar. Morphology


and composition of spray-flame-made yttria-stabilized zirconia nanoparticles. Nanotechnoloby, 26:S609917, 2005.
[15] R. Mueller, L. Madler, and S. E. Pratsinis. Nanoparticle synthesis at high production
rates by flame spray pyrolysis. Chem. Eng. Sci., 58(10):19691976, 2003.
[16] M. C. Heine and S. E. Pratsinis. Droplet and particle dynamics during flame spray
synthesis of nanoparticles. Ind. Eng. Chem. Res., 44(16):62226232, 2005.
[17] R. Mueller, R. Jossen, H. K. Kammler, S. E. Pratsinis, and M. K. Akhtar. Growth
of zirconia particles made by flame spray pyrolysis. AIChE J., 87(2):197202, 2004.

66
[18] R. W. Cheary and A. Coelho. A fundamental parameters approach to X-ray lineprofile fitting. J. Applied Crystallogr., 25:109121, 1992.
[19] R. W. Cheary and A. Coelho. Axial divergence in a conventional X-ray powder
diffractometer. I. theoretical foundations. J. Applied Crystallogr., 31:851861, 1998.
[20] R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, and M. K. Akhtar. Zirconia
nanoparticles made in spray flames at high production rates. J. Am. Ceram. Soc,
87(2):197202, 2004.
[21] W. J. Stark, L. Madler, M. Maciejewski, S. E. Pratsinis, and A. Baiker. Flame
synthesis of nanocrystalline ceria-zirconia: effect of carrier liquid. Chem. Commun.,
(5):588589, 2003.
[22] Y. B. Khollam, A. S. Deshpande, A. J. Patil, H. S. Potdar, S. B. Deshpande, and
S. K. Date. Synthesis of yttria stabilized cubic zirconia (ysz) powders by microwavehydrothermal route. Mater. Chem. Phys., 71(3):235241, 2001.
[23] H. R. Xu, L. Gao, H. C. Gu, J. K. Guo, and D. S. Yan. Synthesis of solid, spherical
CeO2 particles prepared by the spray hydrolysis reaction method. J. Am. Ceram.
Soc., 85(1):139144, 2003.
[24] Rhodia. www.rhodia-ec.com/site ec us/catalysis/page automiative.htm. 2002.
[25] C. Bozo, F. Gaillard, and N. Guilhaume. Characterisation of ceria-zirconia solid
solutions after hydrothermal ageing. Appl. Catal. a-Gen., 220(1-2):6977, 2001.
[26] A. C. Sutorik and M. S. Baliat. Solid solution behavior of Cex Zr1x O2 nanopowders
prepared by flame spray pyrolysis of solvent-borne precursors. Mater. Sci. Forum,
386-3:371376, 2003.
[27] F. F. Lange and M. M. Hirlinger. Grain-growth in 2-phase ceramics - Al2 O3 inclusions
in ZrO2 . J. Am. Ceram. Soc., 70(11):827830, 1987.
[28] F. Del Monte, W. Larsen, and J. D. Mackenzie. Stabilization of tetragonal ZrO2 in
ZrO2 -SiO2 binary oxides. J. Am. Ceram. Soc., 83(3):628634, 2000.
[29] S. W. Wang, X. X. Huang, and J. K. Guo. Wet chemical synthesis of ZrO2 -SiO2
composite powders. J. European Ceram. Soc., 16(10):10571061, 1996.

CHAPTER

FOUR

Thermal Stability and Catalytic Activity of


Flame-made Silica-Vanadia-Tungsten oxide-Titania

Abstract

Vanadia (0.9 or 2 wt%) and silica (0 - 5 wt%) doping of flame-made tungsten oxidetitania nanostructure powders (anatase, 100 m2 /g, 10 wt% WO3 ) is investigated. The
effect of dopants on structural and chemical properties of these nanosized powders were
analyzed by nitrogen adsorption, X-ray diffraction (XRD), temperature programmed reduction (TPR), transmission electron microscopy (TEM) and Raman spectroscopy. After
calcination for 20 hours at 700 C in air, the thermally most stable composite powder
conserved its specific surface area (SSA) to 90 m2 /g and its anatase content to 96 wt%.
Tungsten oxide and vanadia form thin polymeric layers (1 nm) on the surface of the
titania support. Catalytic properties are strongly correlated to thermal stability of the
catalyst. Adding silica improves the thermal and crystal stability of the catalysts even at
higher reactor temperatures and as a result NO conversion by selective catalytic reduction
(SCR) with NH3 and turnover frequency were increased.
67

68

4.1

Introduction

Vanadia on ceramic supports is a well-known catalyst for selective oxidation of o-xylene


or 1-3 butadiene, ammoxidation of aromatic hydrocarbons as well as for selective catalytic
reduction (SCR) of NOx with NH3 at low temperature [1]. Activity and selectivity are
sensitive to the supporting material composition. Alumina, ceria or titania are examples
of such materials [2]. In particular for titania, catalytic performance strongly depends
on its crystallinity as for example anatase is the favorable phase for selective catalytic
oxidation [3-5]. Unfortunately, the thermal stability and anatase content become rather
poor at high temperatures. To avoid thermal degrading dopants like alumina, silica,
zirconia or tungsten oxide can be added [2]. The anatase phase can be stabilized up to
1273 K by adding ceria [6], silica [7] or zirconia [8]. Tungsten oxide additionally enhances
the NOx removal activity of vanadia-titania catalysts and widens the window of operation
temperatures for selective catalytic reduction [9].
Commonly vanadia-titania is prepared by wet processes: first a vanadia precursor
is dispersed onto the support that is filtered, dried and calcined. The dispersion step can
be made by impregnation [10], grafting (adsorption from solution) [11] or co-precipitation
[2]. In contrast to wet methods, Miquel et al. [12] used a counterflow diffusion flame
burner to make mixed vanadia-titania powders in a single step process: Pre-evaporated
VOCl3 and TiCl4 mixed with argon, reacted with hydrogen and oxygen to form particles
with a SSA of 45 m2 /g that were mostly rutile. Stark et al. [13] produced vanadia-titania
nanosized powders by combustion of titanium-tetra-isopropoxide and vanadium-oxo-triisopropoxide vapors in a methane/oxygen co-flow diffusion flame. Mostly anatase particles
with SSAs between 23 and 120 m2 /g were made at 13 g/h. X-ray diffraction and HRTEM
indicated that amorphous vanadia was dispersed on the surface of the anatase titania
spheres. Increasing the vanadia content from 2 to 5 wt% resulted in 30 times higher
activity while the activation energy remained constant for SCR of NOx with NH3 . This
process was scaled up to produce up to 200 g/h of powder while the specific surface area
was controlled from 25 to 100 m2 /g [14]. The catalytic activity for NO removal was twice

69
as that of conventionally prepared catalyst (vanadia-impregnated Degussa P25 titania).
Reiche et al. [11] made vanadia-tungsten oxide-titania particles by simultaneous
grafting steps of vanadia and tungsten oxide on titania. Increasing the tungsten oxide
content increased the SSA and inhibited the anatase to rutile transformation during calcinations. Djerad et al. [15] increased the thermal and crystalline stability of sol-gel-made
vanadia-tungsten oxide-titania by adding up to 9 wt% tungsten oxide. After calcination,
15 m2 /g of SSA remained while for lower tungsten oxide content, the SSA was too low to
be measured.
Here titania and tungsten oxide-titania doped with vanadia and/or silica are made
by the so-called flame spray pyrolysis (FSP) process [16]. The chemical and structural
properties of these materials as-prepared and calcined are investigated by nitrogen adsorption, X-ray diffraction (XRD), temperature programmed reduction (TPR), transmission
electron microscope (TEM) and Raman spectroscopy. The catalytic performance of these
materials was evaluated by analyzing turnover frequency and NO conversion by SCR with
NH3 in a fixed bed reactor.

4.2
4.2.1

Experimental
Apparatus

Precursor solutions resulting in powders of 10 wt% tungsten oxide, 0-2 wt% vanadia and
0-5 wt% silica and the balance TiO2 were prepared by dissolving appropriate amounts
of titanium isopropoxide (Sigma-Aldrich, Chemie GmbH, 95%), tungsten ethoxide (Alfa
Aesar, 95 %), vanadium isopropoxide (Strem, 99%) and tetraethyl-orthosilicate (SigmaAldrich, Chemie GmbH, 99%) in toluene (Sigma-Aldrich, Chemie GmbH). The total metal
concentration in solution was kept at 0.5 M and fed (5 ml/min) through a capillary by
a syringe pump (Inotech R232) and dispersed by 5 l/min oxygen (Pan Gas, > 99.95%)
forming a fine spray described in detail elsewhere ([16], Chapter 1, Fig. 1.1). The pressure
drop at the capillary tip was kept constant at 1.5 bar by adjusting the orifice gap at the

70
nozzle.
The spray was surrounded and ignited by a smaller premixed methane / oxygen
supporting flame ring issuing from an annular gap (0.15 mm spacing, at a radius of 6
mm, total flow rate of 4.7 l/min with fuel/oxygen ratio = 0.92). A sintered metal
plate ring (8 mm wide, starting at a radius of 8 mm) provided an additional oxygen
sheath flow (5 l/min) surrounding the supporting flame. Calibrated mass flow controllers
(Bronkhorst) were used to monitor all gas flows.
Product particles were collected on a glass fibre filter (Whatman GF/A, 15 cm in
diameter) with the aid of a vacuum pump (Vaccubrand). The thermal stability of the
as-prepared powder samples was studied in a furnace under air applying a heating and
cooling rate of 10 K/min for 20 hours at 700 C. For each experimental point a new sample
was used.
The SSA of the collected powders was analyzed by nitrogen adsorption at 77 K
using the BET method (Micrometrics, Tristar 3000). X-ray diffraction (XRD) patterns
of product powders were measured with a Bruker D8 advance diffractometer operating
with Cu(Ka) radiation. Crystallite size and phase composition were obtained using the
Rietveld method and the fundamental parameter approach [16].

4.2.2

Material Characterization

The morphology of the product powder was investigated by transmission electron microscopy (TEM) with a JEOL 2000FX II electron microscope operating at 100 or 200
kV using magnification ranging from 50 to 800k. Energy-dispersive X-ray spectra (EDS)
were obtained (EDAX Genesis) with a resolution of 130 eV to analyze the tungsten and
vanadium content. During imaging special attention was given to powder morphology
and tungsten and vanadium distributions.
Raman spectroscopy (Renishaw, inVia Raman microscope) operated with a HPNIR
laser ( = 785 nm) was used to characterize the vanadia and tungsten oxide species
between 500 to 1500 cm1 using a 1200 grooves/mm grating. The laser beam (5 mW)

71
was focused by a confocal microscope of 50 times magnification (laser spot size: 5m) and
was detected by a CCD camera. No laser-induced damage of the sample was observed.
The reducibility of vanadia with hydrogen was tested with temperature programmed
reduction (TPR) by flowing 20 ml/min of 5 vol% of hydrogen in argon in a tubular reactor
containing 40 mg of as-prepared samples. Samples were pretreated at 573 K for 30 min in
pure oxygen flow to assure complete oxidation of vanadia. The temperature was increased
linearly from 300 to 1223K at 10 K/min while H2 consumption was recorded continuously.
NO conversion is determined in a fixed bed reactor using the flame-made catalyst
powders. The composition of the reactor feed is 300 ppm NO, 360 ppm NH3 , 3% O2 ,
10% H2 O, and balance N2 . Space velocity is 83,000/hr. Conversion is measured at 220,
270 and 320 C by raising the reactor and preheater temperature to these levels, allowing
steady state to be established and then recording NO, NH3 and water concentrations with
a Nicolet FTIR. The reactor is bypassed at each temperature to measure concentrations
in the absence of reaction so the conversion is determined by difference.

4.3

Results and discussion

Table 4.1 summarizes the characteristics of all flame-made samples that are labeled as
xSiyVzWTi, where x, y and z are the weight fractions (%) of Si (silica), V (vanadia) and
W (tungsten oxide), respectively, and the balance is the weight fraction (%) of TiO2 . For
example 5Si0.9V10WTi means 5 wt% SiO2 , 0.9 wt% V2 O5 , 10 wt% tungsten oxide and
84.1 wt% TiO2

4.3.1

Powder characterization

Figure 4.1 shows the thermal stability of product powders calcined at 700 C for 20 hours
as function of their Si-content. As-prepared powders have a SSA of about 100 m2 /g (Table
4.1), regardless of their tungsten oxide, vanadia or silica content. Pure TiO2 (circle) has
poor thermal stability as its SSA decreases to 26 m2 /g upon calcination. Adding 10 wt%

72

Anatase, wt%
as-prepared calcined
18.8
18.1
18.7
17.9
19.1
17.5
18.9
18.9
17.1
15.3

41.8
35.5
50.1
115 (R)
28.4
23.3
20.0
82.9
29.8
19.8

dXRD , nm
as-prepared calcined

100.0
99.0
101.2
101.8
101.6
100.8
100.5
102.7
99.5
110.6

25.6
58.4
29.3
14.3
75.8
79.7
91.8
19.7
63.4
90.3

531
480
547
535
532
533
538
452

0.96
1
0.88
0.92
0.99
0.81
0.84
1

SSA, m2 /g
TPR of as-prepared samples
as-prepared calcined Tmax , C relative H2 consumption

Table 4.1: Characteristics of FSP-made samples of as-prepared and after calcination at 700 C for 20 hours.
Sample ID
88
94
90
93
90
95
96
95
97
99

18
85
80
0
90
91
95
55
90
96

Pure TiO2
10WTi
0.9V10WTi
2V10WTi
1Si0.9V10WTi
2Si0.9V10WTi
5Si0.9V10WTi
1Si2V10WTi
2Si2V10WTi
5Si2V10WTi
R: Rutile

73

Figure 4.1: The specific surface area of FSP-made pure and doped TiO2 powders as a
function of their silica content after calcination at 700 C for 20 h in air.

tungsten oxide increases the thermal stability of the resulting mixed oxide, retaining a SSA
of 58 m2 /g after calcination. This is in agreement with wet-made WO3 /TiO2 containing
6 wt% WO3 calcined at 700 C and retained a SSA of 45 m2 /g compared to 20 m2 /g of
pure TiO2 [17]. Adding 0.9 or 2 wt% vanadia to that 10WTi reduces thermal stability
of the mixed oxide resulting in a SSA of 29 (0.9V10WTi) and 14 m2 /g (2V10WTi) after
calcination (Fig. 4.1). This is consistent with Madia et al. [18] who investigated the
thermal stability of tungsten oxide-titania catalysts containing 1 to 3 wt% vanadia.
Adding 0-5 wt% SiO2 to V2 O5 -WO3 -TiO2 , however, improves its thermal stability
(Fig. 4.1): 1 wt% silica increases the SSA from 29 to 75 m2 /g (1Si0.9V10WTi, triangles)

74
and from 14 to 19 m2 /g (1Si2V10WTi, squares). When adding 5 wt% SiO2 , the resulting
SSA is 90 m2 /g, thus, retaining about 90 % of the SSA of the as-prepared powders,
regardless of vanadia content. This is in agreement with Reiche et al. [19] who grafted
vanadia on sol-gel made titania-silica catalysts.
Figure 4.2 shows the XRD patterns of as-prepared (solid line) and calcined (dotted
line) TiO2 and 10WTi. The as-prepared TiO2 contains mostly anatase that converts to
mostly rutile after calcination. The XRD diameter (dXRD ) of these anatase crystals increases from 18.8 to 41.8 nm (Tab. 4.1). For the 10WTi, however, anatase is conserved
during calcination and dXRD increased from 18.1 to 35.5 nm (Tab. 4.1). No crystalline
tungsten oxide was detected by XRD for both as-prepared and calcined powders, indicating amorphous tungsten oxide structure. Here, tungsten oxide suppresses the formation
of rutile at 700 C that leads also to an increased thermal stability as a SSA of 58 m2 /g
is retained [1, 17]. This is in agreement with Reiche et al. [11] who showed that adding
5 wt% tungsten oxide inhibited the transformation of anatase to rutile as well as with
Djerad et al. [15] who showed also that increasing the tungsten oxide loading of titania
to 9 wt% inhibited that transformation.
The XRD patterns in Figure 4.3 show the influence of silica on the anatase to rutile
transformation for a) xSi0.9V10WTi and b) xSi2V10WTi during calcination at 700 C for
20 hours in air (x = 0, 1 and 5). All as-prepared powders contain mostly anatase (Tab.
4.1). For the 0.9 wt% vanadia loading (Fig. 4.3a), the rutile content increases during
calcination in the absence of silica. This leads to an increase in dXRD form 18.7 to 50.1
nm (Table 4.1). Furthermore, monoclinic tungsten oxide is formed with peaks at 2 =
23.1 , 23.6 and 24.4 [20]. Comparing 10WTi (Fig. 4.2) with 0.9V10WTi and 2V10WTi
(Fig. 4.3) shows that vanadia promotes the formation of crystalline tungsten oxide. This
segregation takes place just before the anatase to rutile transition [17]. Adding silica
to 0.9V10WTi conserves anatase during calcination and no crystalline tungsten oxide is
detected. For increasing silica content, the dXRD of anatase after calcination decreases
from 28.4 (1 wt% silica) to 23.3 (2 wt % silica) and 20 nm (5 wt% silica).

75

Figure 4.2: XRD patterns of TiO2 and 10WTi powders as-prepared (solid line) and
calcined (dotted line) at 700 C for 20 h in air.

The same behavior was observed for 2 wt% vanadia loading (Fig. 4.3b). The
2V10WTi powder crystallinity is transformed completely to rutile resulting in dXRD of
115 nm upon calcination. Additionally, monoclinic tungsten oxide is detected when adding
1 wt% silica as tungsten oxide segregates and the transition from anatase to rutile takes
place. Increasing the silica content to 2 and 5 wt% impeded the segregation of the tungsten
oxide during calcination preserving the anatase phase. The dXRD after calcination is 82.9,
29.8 and 19.8 nm for 1, 2 and 5 wt% silica, respectively. From figures 4.2 and 4.3 it is
shown that vanadia promotes the anatase to rutile transition as well as crystallite growth.
Djerad et al. [15] showed also that the vanadia affects the anatase to rutile transition

76

Figure 4.3: XRD patterns of a) 0.9V10WTi and b) 2V10WTi powders both with 0 to 5
wt% silica loading as-prepared (solid line) and calcined at 700 C for 20 h in air (dotted
line).

which is faster at higher vanadia loading as shown here. Vanadia phases were not detected
by XRD as even 2 wt% vanadia was below its detection limit.
Figure 4.4 shows the evolution of the anatase content for as-prepared (open symbols) and calcined (filled symbols), 0.9V10WTi (triangles) and 2V10WTi (squares), TiO2
(circle) and 10WTi (diamonds) powders as a function of their silica content. For pure
TiO2 only 18 wt% of anatase remains after calcination but adding 10 wt% tungsten
oxide (10WTi) increases it to 85 wt%. Adding, however, 0.9 wt% vanadia to 10WTi
decreases the remaining anatase content to 80 wt% after calcination and even to 0 wt%
when adding 2 wt% vanadia. Adding progressively silica to 0.9V10WTi (triangles) and
2V10WTi (squares) powder increases drastically the remaining anatase content after calcination: for the former 90 (1 wt% silica) and 96 wt% (5 wt% silica) anatase remain after
calcination while the same trend is followed for the 2V10WTi powder (Fig. 4.4). For 0.9

77

Figure 4.4: Anatase content for as-prepared (open symbols) and calcined at 700 C for 20
h in air (filled symbols) TiO2 , 10WTi, 0.9V10WTi and 2V10WTi powders as a function
of their silica content.

and 2 wt% vanadia loading, silica suppresses the formation of rutile. It is widely accepted
that the presence of silica can retard the transformation of anatase to rutile phase by
interstitial defects [21] and keep the SSA through Ti-O-Si interaction [22, 23].
Crystalline tungsten oxide was only found after calcination at anatase contents less
than 80 wt%. The 10WTi and 2V10WTi were also calcined at 550 C for 100 h. The
anatase content decreased from 93 (Tab. 4.1) to 88 wt% and dXRD increased from 18
to 25 nm. Here, no crystallite tungsten oxide was detected. At lower calcination temperatures (550 C) the vanadia has less influence on crystallite transformation. Here the

78
20 nm

5 nm

Figure 4.5: TEM micrographs of FSP made 0.9V10WTi powder. The surface material
appears to be primarily tungsten oxide and vanadia.

transformation temperature is higher than 550 C and therefore at low temperature the
powder is more thermally stable.
Figure 4.5 shows TEM images of 0.9V10WTi. Here the crystalline particles are
between 5 and 20 nm. The surface layer appears to be primarily tungsten oxide that is
generally below 1 nm. In this sample the vanadia could not be detected by EDS as the
detection limit is < 1 wt% vanadium due to its overlapping signal with titanium. In the
sample 2V10WTi (not shown) the surface material appears to be primarily tungsten oxide
and vanadia with a similar layer thickness. The vanadia content of that sample detected
by EDS was 2-3 wt% consistent with the precursor composition.
Figure 4.6 shows Raman spectra of TiO2 and 10WTi as-prepared (solid line) and
calcined (dotted line). Raman band frequencies for anatase are expected at 144, 197, 399,

79

Figure 4.6: Raman spectra of TiO2 and 10WTi powders as-prepared (solid line) and
calcined (dotted line) at 700 C for 20 h in air.
519, 639 and 796 cm1 and for rutile at 143, 235, 300-380, 612 and 826 cm1 [24]. All
samples except for calcined pure TiO2 show only anatase Raman bands which is consistent
with XRD (Tab. 4.1). Calcined TiO2 shows an additional Raman band at 612 cm1
that corresponds to rutile, consistent with XRD. The as-prepared and calcined 10WTi
show also Raman bands at 960 and 984 cm1 , respectively, corresponding to polymeric
tungsta. The shift from 960 to 984 cm1 indicates an increasing degree of coordination
for the polymeric tungsta during calcination.
Figure 4.7 shows Raman spectra of a) xSi0.9V10WTi and b) xSi2V10WTi with x =
0, 1 and 5 wt% silica. For increasing silica content, the slope of the xSi0.9V10WTi spectra

80

Figure 4.7: Raman spectra of a) 0.9V10WTi and b) 2V10WTi powders both with 0 to
5 wt% silica loading as-prepared (solid line) and calcined (dotted line) a at 700 C for 20
h in air.
between 700 and 1000 cm1 increases resulting from the fluorescence of silica [25]. This
overlaps with the Raman band between 940 and 1000 cm1 that includes contributions
from polymeric metavanadate species, both tetrahedally and octahedrally co-coordinated
polymeric WOx groups [18]. No monomeric (1020 cm1 [26]), crystalline or bulk vanadia
(997 cm1 [26]) was detected by Raman. In the case of low vanadia loading (0.9V10WTi),
crystalline tungsten oxide is formed during calcination (Fig. 7a, x=0, Raman bands at
715 and 807 cm1 [1]). Also, mostly anatase is detected by Raman spectroscopy. There
is a small shoulder at about 600 cm1 that may be from the rutile Raman band which has
its maximum at 612 cm1 . This is in agreement with XRD where only 18 wt% of rutile
was found (Figs. 4.3 and 4.4). This shows further that vanadia promotes the formation of
crystalline tungsten oxide and rutile. When adding more than 1 wt% silica no crystalline
tungsten oxide and rutile is formed. Silica inhibits the formation of rutile as well of
crystalline tungsten oxide in agreement again with XRD (Fig. 4.3a).

81
Increasing the vanadia content to 2 wt% (Fig. 4.7b) results in Raman bands at
about 980 cm1 that correspond to polymeric vanadia and tungsten oxide independent
of silica content. There was no silica fluorescence, probably due to increased vanadia
content. After calcination, crystalline tungsten oxide and rutile are formed at low silica
loadings ( 1 wt%). At higher silica content ( 2 wt%) only anatase and no crystalline
tungsten oxide was detected. The Raman bands at 980 cm1 (as-prepared) are shifted
to higher wavelength after calcination. The band with a shoulder at about 1060 cm1 in
the absence of silica decreases to 1040 cm1 for 5 wt% SiO2 loading. This may indicate
the formation of monomeric vanadia. This was not observed for xSi0.9V10WTi (Fig.
4.7a). For flame-made powder monomeric vanadia is only formed directly at high vanadia
loadings consistent with Stark et al. [13] who only obtained monomeric vanadia at above
5 wt% vanadia.
Relative hydrogen consumption and temperature of maximal hydrogen consumption rate, Tmax , obtained by TPR are given in Table 4.1. In the V2 O5 -TiO2 system
different vanadia polymorphs such as monomeric, polymeric or crystalline VOx can coexist on the surface of the particles. Different Tmax do not necessarily represent different
stage of reduction but represent the reduction of different species. For all samples only
one reduction peak was found indicating that only one vanadia species corresponds to
polymeric vanadia. For 0.9V10WTi and 2V10WTi the Tmax are 531 and 480 C, respectively. For both samples up to 96% of hydrogen was consumed. Adding silica (up to 2
wt%) increases Tmax for 0.9 and 2 wt% vanadia loading to 547 and 538 C, respectively.
However the hydrogen consumption decreases to 88 and 81% for 0.9 and 2 wt% vanadia
loading, respectively, and 1 wt% silica. The silica may cover some vanadia on the surface
of the titania, therefore less vanadia is accessible for reduction. Reiche et al. [19] showed
that vanadia grafted on commercial titania-silica aerogels increases Tmax above that for
pure TiO2 support. Adding 5 wt% silica, for the 0.9 wt% vanadia-loaded powder does
not change the Tmax compared to lower silica loadings but the hydrogen consumption increases again to almost 100%. For the 2 wt% vanadia loading, the Tmax decreases to 452

82

C and the hydrogen consumption increases again to 100%. Sorrentino et al. [25] showed

that preparing vanadia-titania powders with and without silica Tmax depends only on the
preparation method, grafting or impregnation.

4.3.2

Catalytic performance

Figure 4.8 shows the NO conversion ((NOin -NOout )/NOin 100%) as a function of reactor
temperature for WO3 -TiO2 samples with different vanadia and silica content. For all
samples conversion increases with increasing reactor temperature. Sample 0Si0.9V10WTi
shows the lowest NO conversion at any reactor temperature that increases from 18 to
79% when increasing the reactor temperature from 230 to 320 C. When increasing the
vanadia content to 2 wt% the conversion is increased to 70 and 93% at 230 to 320 C,
respectively. Adding 1 wt% silica the increases again NO conversion to 29 and 86% for x
= 5 wt% silica it reacts 36 and 90% for these reactor temperatures. In contrast with 2 wt%
vanadia loading the NO conversion is decreased at 230 C reactor temperature from 70 to
54% when adding 1 wt% and to 50% when adding 2 wt% silica. At a reactor temperature
of 320 C, however, the conversion increases again to 92 and 97% for 1 and 2 wt% silica,
respectively. At 2 wt% vanadia loading, silica might deactivate the catalyst at low reactor
temperature by covering of the vanadia species on the surface. RAMAN analysis shows
as well the formation of monomeric vanadia that may deactivate the catalyst. Stark et
al. [13] reported that adding silica to the DeNOx catalyst deactivated the catalyst.
However, at 320 C and 2 wt% vanadia loading the conversion increases with increasing silica content as the silica improves the thermal stability of the catalyst. The NH3
selectivity ((NOin -NOout )/(NH3in -NH3out )100%) for the low (xSi0.9V10WTi) and high
vanadia loading (xSi2V10WTi) is almost 100%, independent of silica loading and even at
320 C .
Figure 4.9 shows the turnover frequency (TOF, mol NO sec1 /mol vanadia) as a
function of silica content at 270 C. Increasing the vanadia loading from 0.9 to 2 wt%
decreases the TOF from 0.61 to 0.56 s1 . In contrast Amiridis et al. [27] and Stark et

83

Figure 4.8: NO conversion of 0.9 (filled symbols) and 2 wt% (open symbols) vanadia
loading and different silica loadings as a function of the reactor temperature.
al. [13] showed that at low vanadia loading the TOF first increases to a maximum and
decreases again. Increasing the silica content from 0 to 5 wt% increases continuously
the TOF from 0.61 to 1.0 s1 for 0.9 wt% vanadia and from 0.56 to 0.79 s1 for 2 wt%.
This increase can been attributed to a higher specific activity of the polymeric vanadate
species and the accompanying increase of Brnsted acid sites with vanadia surface coverage
[28]. NO conversion strongly depends on the vanadia content as shown for flame- [13] or
wet- made [29] vanadia-titania catalyst. From this and TOF data it can be concluded
that the catalyst will be thermally deactivated when phase transformation, grain growth
and crystallization of vanadia and tungsten oxide take place. Silica can overcome this

84

Figure 4.9: CInfluence of silica loading on catalytic activity (turnover frequency) of 0.9
(open symbols) and 2 wt% (filled symbols) vanadia loading 10WTi.
deactivation by stabilizing the textural structure and surface area of the catalysts. These
are in agreement with Albinetti et al. [20] who studied the effect of silica on thermal
deactivation of a commercial DeNOx catalyst.

4.4

Conclusions

The characteristics and thermal stability of silica-vanadia-tungsten oxide-titania powder


depend strongly on its composition. Pure TiO2 has very low thermal stability and during
calcination most anatase converts to rutile with large loss of its SSA. Addition of 10 wt%

85
tungsten oxide slightly improves its thermal and crystalline stability. In contrast, adding
even up to 2 wt% vanadia deteriorates both while adding silica improves both drastically:
90% of the initial SSA remained after calcination at 700 C for 20 hours upon adding 5
wt% silica to the 2V10WTi. Additionally the anatase content remains above 95 wt%.
Microscopic analysis shows that the vanadia and tungsten oxide are well distributed
on the surface of the titania support. Raman spectroscopy indicates polymeric vanadia
and tungsten oxide in all cases, which is believed to be the most active reaction sites
for NO reduction by NH3 . After calcination crystalline tungsten oxide was only formed
in the presence of vanadia and at low silica contents ( 1 wt%). Adding silica prevents
the crystallization of tungsten oxide. For all samples before and after calcination no
crystalline vanadia is found.
TPR analysis shows that in all cases the vanadia has a very good accessibility as
between 80 and 100% of the V5+ is reduced. It is shown that by adding silica (1 to 2 wt%)
the reducibility of vanadia is decreased while when adding 5 wt% silica the reducibility
is again increased. The 5Si2V10WTi powder has the lowest reduction peak as it is most
easily reduced. Adding silica can prevent this thermal deactivation and can again increase
the conversion as well as the TOF also at higher reactor temperature.

References
[1] J. Engweiler, J. Harf, and A. Baiker. WOx /TiO2 catalysts prepared by grafting of
tungsten alkoxides: Morphological properties and catalytic behavior in the selective
reduction of NO and NH3 . J. Catal., 159:259269, 1996.
[2] B. M. Reddy, I. Ganesh, and B. Chowdhury. Design of stable and reactive vanadium
oxide catalysts supported on binary oxides. Catal. Today, 49(1-3):115121, 1999.

[3] V. A. Nikolov and A. I. Anastasov. Pretreatment of a vanadia titania catalyst for


partial oxidation of ortho-xylene under industrial conditions. Ind. Eng. Chem. Res.,
31(1):8088, 1992.

86
[4] I. Nova, L. D. Acqua, L. Lietti, E. Giamello, and P. Forzatti. Study of thermal
deactivation of a De-NOx commercial catalyst. Appl. Catal. B-Envir., 35(1):3142,
2001.
[5] R. Y. Saleh, I. E. Wachs, S. S. Chan, and C. C. Chersich. The interaction of V2 O5
with TiO2 (anatase) - Catalyst evolution with calcination temperature and o-xylene
oxidation. J. Catal., 98(1):102114, 1986.
[6] B. M. Reddy, A. Khan, Y. Yamada, T. Kobayashi, S. Loridant, and J. C. Volta.
Structural characterization of CeO2 -TiO2 and V2 O5 /CeO2 -TiO2 catalysts by raman
and XPS techniques. J. Phys. Chem. B, 107(22):51625167, 2003.
[7] B. M. Reddy, A. Khan, P. Lakshmanan, M. Aouine, S. Loridant, and J. C. Volta.
Structural characterization of nanosized CeO2 -SiO2 , CeO2 -TiO2 , and CeO2 -ZrO2 catalysts by XRD, RAMAN, and HREM techniques. J. Phys. Chem. B, 109(8):3355
3363, 2005.
[8] M. Hirano, C. Nakahara, K. Ota, and M. Inagaki. Direct formation of zirconia-doped
titania with stable anatase-type structure by thermal hydrolysis. J. Am. Ceram. Soc.,
85(5):13331335, 2002.
[9] E. Y. Choi, I. S. Nam, and Y. G. Kim. TPD study of mordenite-type zeolites for
selective catalytic reduction of NO by NH3 . J. Catal., 161(2):597604, 1996.
[10] L. J. Alemany, L. Lietti, N. Ferlazzo, P. Forzatti, G. Busca, E. Giamello, and F. Bregani. Reactivity and physicochemical characterization of V2 O5 -WO3 /TiO2 DeNox
catalysts. J. Catal., 155(1):117130, 1995.
[11] M. A. Reiche, T. Burgi, A. Baiker, A. Scholz, B. Schnyder, and A. Wokaun. Vanadia
and tungsta grafted on TiO2 : influence of the grafting sequence on structural and
chemical properties. Appl. Catal. A-Gen., 198(1-2):155169, 2000.
[12] P. F. Miquel, Hung C.-H., and Katz J. L. Romation of V2 O5 -based mixed oxides in
flames. J. Mater. Res., 8(9):24042413, 1993.
[13] W. J. Stark, K. Wegner, S. E. Pratsinis, and A. Baiker. Flame aerosol synthesis
of vanadia-titania nanoparticles: Structural and catalytic properties in the selective
catalytic reduction of NO by NH3 . J. Catal., 197(1):182191, 2001.
[14] W. J. Stark, A. Baiker, and S. E. Pratsinis. Nanoparticle opportunities: Pilot-scale
flame synthesis of vanadia/titania catalysts. Part. Part. Sys. Charact., 19(5):306
311, 2002.

87
[15] S. Djerad, L. Tifouti, M Corocoll, and W Weisweiler. Effect of vanadia and tungsta
loadings on the physical and chemical characteristics of V2 O5 -WO3 /TiO2 catalysts.
J. Molec. Cat. A: Chem., 208:257265, 2004.
[16] L. Madler, H. K. Kammler, R. Mueller, and S. E. Pratsinis. Controlled synthesis
of nanostructured particles by flame spray pyrolysis. J. Aerosol Sci., 33(2):369389,
2002.
[17] G. Ramis, G. Busca, C Cristiani, L. Lietti, P. Forzatti, and F. Bergani. Characterization of tungsta-titania catalyst. Langmuir, 8:17441749, 1992.
[18] G. Madia, M. Elsener, M. Koebel, F. Raimondi, and A. Wokaun. Thermal stability
of vanadia-tungsta-titania catalysts in the SCR process. Appl. Catal. B-Enviro.,
39(2):181190, 2002.
[19] M.A. Reiche, A. Ortelli, and A. Baiker. Vanadia grafted on TiO2 -SiO2 , TiO2 and
SiO2 aerogels structural properties and catalytic behaviour in selective reduction of
NO by NH3 . Appl. Catal. B-Envir., 23:187203, 1999.
[20] S. Albonetti, S. Blasioli, M. Bufani, C. Lehaut-Burnouf, S. Augustine, E. Roncari,
and F. Trifiro. Effect of silica additive on the thermal stability of catalyst for NOx
abatement. Enviro. chem. Lett., 1:197200, 2003.
[21] M. K. Akhtar, S. E. Pratsinis, and S. V. R. Mastrangelo. Dopants in vapor-phase
synthesis of titania powders. J. Am. Ceram. Soc., 75(12):34083416, 1992.
[22] S. R. Kumar, C. Suresh, A. K. Vasudevan, N. R. Suja, P. Mukundan, and K. G. K.
Warrier. Phase transformation in sol-gel titania containing silica. Mater. Lett.,
38(3):161166, 1999.
[23] E. Pabon, J. Retuert, R. Quijada, and A. Zarate. TiO2 -SiO2 mixed oxides prepared
by a combined sol-gel and polymer inclusion method. Microporous Mesoporous Mat.,
67(2-3):195203, 2004.
[24] U. Balachandran and N.G. Eror. RAMAN spectra of titanium dioxide. J. Solid State
Chem., 42, 1982.
[25] A. Sorrentino, S. Rega, D. Sannino, A. Magliano, P. Ciambelli, and E. Santacesaria.
Performances of V2 O5 -based catalysts obtained by grafting vanadyl tri-isopropoxide
on TiO2 -SiO2 in SCR. Appl. Catal. A-Gen., 209(1-2):4557, 2001.
[26] G.T. Went, S.T. Oyama, and A.T. Bell. Laser raman spectroscopy of supported
vandaium oxide catalysts. J. Phys. Chem., 94:42404246, 1990.

88
[27] M. D. Amiridis, I. E. Wachs, G. Deo, Jehng J.-M., and D.U. Kim. Reactivity of V2 O5
catalysts for the selective catalytic reduction of NO by NH3 : Influence of vanadia
loading, H2 O and SO2 . J. Catal., 161:247253, 1996.
[28] J. Engweiler and A. Baiker. Vanadia suported on titania aerogels - morphological
properties and catalyitc behavior in the selective reduction of nitric-oxide by ammonia. Appl. Catal. A, 120:187, 1994.
[29] I. E. Wachs, G. Deo, B. M. Weckhuysen, A. Andreini, M. A. Vuurman, M. deBoer,
and M. D. Amiridis. Selective catalytic reduction of NO with NH3 over supported
vanadia catalysts. J. Catal., 161(1):211221, 1996.

CHAPTER

FIVE

Non-intrusive droplet and particle dynamics during


spray flame synthesis of nano ZrO2

Abstract

Droplet and nano particle dynamics are studied in situ and non-intrusively in a particleladen spray flame producing 15 g/h ZrO2 particles. In situ small angle X-ray scattering
(SAXS) is used to measure simultaneously the evolution of primary-particle diameter,
mass-fractal dimension, geometric standard deviation, particle volume fraction, particle
number concentration, agglomerate size and number of primary particles per agglomerate
in the spray flame. Droplet velocities are measured by 2D-Phase Doppler Anemometry
(PDA) and gas temperature is measured by Fourier transform infrared (FTIR) emission/transmission spectroscopy. A nucleation event is revealed early in the flame that
is accompanied by broad size distributions of nanoparticles that quickly approach the
asymptotic self-preserving size distribution by coagulation until the onset of agglomeration when their distribution narrows further by sintering. Wide angle X-ray scattering
(WAXS) indicates that early in the flame zirconia particles could be amorphous although
when collected on the filter have metastable tetragonal crystallinity.
89

90

5.1

Introduction

Flame spray pyrolysis (FSP) is a one step technique for synthesis of a broad spectrum of
inorganic nanoparticles from titania [1] to iron oxide [2], CeO2 [3] or ZrO2 , [4] or mixed
oxides like CeO2 /ZrO2 [5], MgAl2 O4 [6], Ta2 O5 /SiO2 for dental nanocomposites [7] or
yttria-stabilized zirconia [8] and even metal/ceramics such as Pt/Al2 O3 [9] for catalysis
to name a few. In FSP, metallic precursors are dissolved in an organic fuel that dispersed
through a nozzle forming a spray that is ignited. Droplet evaporation is followed by
combustion, particle formation and growth by condensation, coagulation, sintering and
eventual agglomerate formation [10].
Compared to vapor-fed flames [11], liquid-fed spray flames have typically higher gas
velocity [12] while maximum temperatures are similar. Therefore, particles are made in
spray flames with much higher cooling rates and much shorter high temperature residence
time than in vapor-fed flames that may result inhomogeneous FSP-made particle morphology [8] and composition [5]. Typically, the turbulence during FSP is quite high as
nozzle Reynolds numbers (Re) range between 10,000 and 23,000 [13]. For a better understanding of particle growth and formation of non-agglomerate or agglomerate particles by
FSP, flame temperature and gas velocity have to be known. Gas flame temperatures have
been measured in particle-laden spray flames [13, 14] using FTIR emission/transmission
spectroscopy while gas velocities were measured by Phase Doppler Anemometry (PDA)
using unburned precursor droplets as tracer particles [12]. The primary particle and agglomerate evolution can be studied by thermophoretic sampling (TS) and transmission
electron microscopy (TEM)-analysis [15]. Though, this technique could depict reliably
particle growth, it is intrusive and time consuming while sampling near the nozzle is not
possible as burning droplets impinge and destroy the TEM grid [13].
Several non-intrusive and in situ optical techniques have been developed to monitor combustion-made particle dynamics, including the 2-dimensional RAYLIX technique.
This is a combination of Rayleigh scattering and laser induced incandescence were the
integral extinction is detected from a single laser pulse measuring the particle volume frac-

91
tion, number density and average size [16]. A common method to study particle formation
is dynamic light scattering (DLS) that defines particle size through particle motion. The
temperature dependence of particle diffusion and DLSs difficulty in resolving agglomerate
structures limit DLS alone for in situ measurements in flame reactors [17]. Static light
scattering (SLS) is often used to measure in situ large agglomerate size and mass-fractal
dimension [18]. Here, the interpretation of the scattered intensity is difficult due to the
complexity of the Mie theory for small particles other than monodisperse particles of simple shape [19]. The relative large wavelength of visible light leads also to size limitations
[20]. In one of the best applications of light scattering, Rosner and his colleagues had
interfaced it with TEM to extract dynamics of flame-made alumina [21].
Small angle X-ray scattering (SAXS) and especially ultra-small angle X-ray scattering (USAXS) can overcome some of these limitations and unravel structures from nanoto microscale. The lower size detection limit depends on the highest probed scattering
vector. The third-generation synchrotron X-ray sources (ERSF in Grenoble or APS in
Argonne, Chicago) yields sufficient photon flux to probe time resolved scattering signal
in particle-laden flames even at low volume fraction (106 ) with short acquisition time
(< 1 second). Thus, Beaucage et al. [11] measured for the first time nanostructure inorganic particle evolution in a diffusion flame [22] using synchrotron X-ray scattering and
found that the silica particle size follows simple diffusion-controlled growth laws. At 5
mm above the burner they saw a burst in particle number density as there was sufficient
silica concentration to produce silica nuclei. Particles then start to rapidly coalesce at
flame temperatures of about 2300 K. Kammler et al. [23] also used a third-generation
synchrotron technique in a premixed silica producing flame [24] and found rather high
agglomerate fractal dimension of Df = 2.4 that might have formed mainly by monomercluster growth instead of cluster-cluster collisions.
Here, particle and agglomeration dynamics are studied by SAXS for the first time in
situ in a spray flame reactor [25]. Results cover the whole size range from primary particles
of about 2 nm to agglomerates up to 300 nm in diameter and even microsize droplets that

92
compared also to ex-situ product powder properties. SAXS measurements are coupled
with in situ gas temperature (FTIR emission/transmission) and velocity measurements
(PDA) to get a consistent data set for analysis of particle growth and agglomerate evolution during flame spray synthesis.

5.2
5.2.1

Experimental
Apparatus

Zirconia powder is produced by FSP at 1.5 bar gas pressure drop at the nozzle tip by
adjusting the nozzle-gap width as has been described in detail elsewhere (Fig. 5.1 [3]).
The gas flow rates are controlled by mass flow controllers (Bronkhorst, Ruurlo, The
Netherlands) and the liquid precursor solution is fed through the nozzle by a syringe pump
(IER-232, Inotech, Oberwinterthur, Switzerland). Zirconium n-propoxide (Zr(C3 H7 O)4 ,
70% in n-propanol, ZP, ChemPur, Germany) is dissolved in ethanol (EtOH, > 99.8%,
Alcosuisse) resulting in a 0.5 M precursor solution fed at 4 (flame A) or 8 ml/min (flame
B). The corresponding ZrO2 production rates are 15 or 30 g/h, respectively. The precursor
solution was fed through the center nozzle capillary and is dispersed into a spray by
flowing 5 l/min O2 through the surrounding annulus. The resulting spray is ignited by
a ring-shaped premixed flame (0.5 l/min CH4 and 2 l/min O2 ) surrounding the spray
[3]. An additional oxygen sheath gas flow (5 l/min) was supplied through a sintered
metal plate ring (8 mm wide with an inner diameter of 9 mm) surrounding this premixed
flame. Product powder was collected on a water cooled glass fiber filter (Whatman,
GF/A) 150 mm in diameter located at 20 cm above the nozzle tip. The specific surface
area (SSA) of product powders was measured by a five-point N2 adsorption isotherm in
the relative pressure range of 0.05 to 0.25 at 77 K (BET analysis) using a Tristar 3000
(Micromeritics Instruments Corp.). Prior to this, samples are degassed under N2 at 150 C
for 1h. Assuming monodisperse spherical primary particles, the BET-equivalent average
primary particle diameter, dBET , is calculated by dBET = 6000/(SSA), where = 6.1

93

Figure 5.1: Experimental schematic of the diagnostic technique. FTIR was used for
temperature measurement, PDA for gas velocity measurements, thermophoretic sampling
to monitoring particle growth and morphology and small angle X-ray scattering for monitoring in-situ particle structure growth. The optical image shows the installation of the
FSP at the ERSF in Grenoble at beam line ID2.

94
g/cm3 is the density of tetragonal ZrO2 . Crystallinity was measured by X-ray diffraction
(XRD) with Bruker D8 advanced diffractometer. Crystalline characteristics were obtained
from the XRD spectra using the Topas 2.0 software (Bruker AXS, 2000).

5.2.2

Spray flame characterization

Droplet velocities and diameters were measured using a 2D-Phase Doppler Anemometer
(2D-PDA, TSI. Inc., USA) operating with a water cooled 5W argon-ion laser (Innova
70, LA-70-5) generating a green and blue beam with wavelength of 514.4 and 488 nm,
respectively. The fiber optic based system is configured for first order refracted light with
the receiver located at 30 of the transmitter axis. The PDA was positioned by a 3D
traversing system (isel, 9450-XYZ500) controlled by a step motor (isel, C 142-4.1). The
droplet detection range is 1 to 100 m with 0.5 mm3 spatial resolution while at each point
about 40000 droplets were recorded. The reproducibility of the average droplet size and
velocity was within 5%. Axial droplet velocity measurements were performed by single
point measurements in 1 mm steps up to 45 and 60 mm height above the nozzle (HAN)
for flame A and B, respectively. Above these HAN the effective data rate was below
0.1 kHz compared to 16 kHz in the dense spray regions. Axial velocity measurements
were preformed radially at 8, 13, 18 and 25 mm HAN with the beams traversing from
right to left of the flame edges counting a total of about 200000 droplets at each HAN.
From the constant traverse speed, the radial position of droplets was reconstructed and
averaged over 1 mm steps also. The flame spray gas velocity was assumed to be the
velocity of droplets smaller than 4 m [12]. The nozzle Reynolds numbers based only on
the dispersion gas at the nozzle tip and were calculated by [26] :

Re =

l d0 (ug ul )
,
l

(5.1)

where d0 is the orifice equivalent diameter, l is the liquid viscosity and (ug - ul ) is the
difference of the gas, ug , and liquid, ul , velocity at the nozzle exit. The d0 is calculated
from the opening area of the dispersion gap which is 0.48 mm2 [25] corresponding to an

95
equivalent orifice diameter of 0.78 mm. The dispersion gas has approximately sonic speed
[26] and ul is 0.48 and 0.96 m/s for flame A and B, respectively.
Spray flame temperatures were measured by emission/transmission spectroscopy
[27, 28] using an FTIR spectrometer (Bomem, MB155S) operating in the range of 6500500 cm1 with 4 cm1 resolution and IR beam of 4 mm. The normalized radiance, Rn ,
calculated from the actual radiance is compared to a best fit using the blackbody Planck
function [29, 30]. Particle were collected thermophoretically [15, 24] and particle images
were obtained by transmission electron microscopy (HRTEM; Philips Tecnai F30, field
emission cathode, 300 kV).
In situ particle growth during FSP was measured using small-angle X-ray scattering
(SAXS) at the European Synchrotron Radiation Facility (ESRF, beam line ID02) in
Grenoble (France) [31]. A highly collimated monochromatic X-ray beam ( = 0.1 nm,
200 x 200 m2 penetrated the flame while the exposure time of the CCD camera was 0.8
s. Data were corrected by a procedure reported elsewhere [32] including subtraction of
background measurements made on a similar flame in the absence of the zirconia precursor
(pure ethanol). Particles in the range of 0.1 to 500 nm can be obtained form the SAXS
scattering spectra using the unified fit described in detail elsewhere [33, 34].

5.3

Theory

The structure of an object like a fractal agglomerate or just an single particle can be

determined by the scattering intensity I as function of the scattering vector


q . The

magnitude of
q yields to be
2
q=
sin

 

,
2

(5.2)

where is the scattering angle and the wavelength of the incident photons. For large q
the intensity deceases with the power law:
I(q) = BP orod q 4 ,

(5.3)

96

BP orod = 2re2 N ()2 S,

(5.4)

BP orod is the Porod prefactor where re2 is the electron cross section, N the particle number
density, the difference of electron density between the scatterer and the background
and S is the surface area of the particle. The Porod law is followed by the Guinier
law which has the shape of a kneelike decay in intensity. The Guinier law reflects the
structural size of a primary particle (Rg ). This kneelike decay in intensity follows the
Guinier law.
q 2 Rg2
I(q) = G exp
3



,

(5.5)

where
G = re2 N ()2 V 2 .

(5.6)

N is the number density and V is the volume of an primary particle. This Guinier regime
is called the first regime as it reflects only the primary particles.
At higher q range a second Power law can be observed with a weaker decay. The
magnitude of exponent is less than 3 and reflects the mass fractal dimension of an agglomerate. (eq. (5.3)):

I(q) = Bf q Df ,

(5.7)

At lower q the second power law is followed by a second Guinier law which represents
the radius of gyration of the agglomerates.
For example, an agglomerate has the structure of an branched aggregate which can
be described with a minimal path p across the agglomerate. p can be described as follows:

p=

RG2
RG1

Dmin
(5.8)

where Dmin is the minimum dimension, RG2 and RG1 are the agglomerate and primary
particle radii, respectively. Dmin = 1 when a linear path is formed in the agglomerate from

97
one to the other side. The relationship between Df and Dmin is called the connectivity
dimension.

c=

Df
Dmin

(5.9)

The number fraction of branches, br , can be calculated by the following equation:



br = 1

RG2
RG1

Dmin Df
(5.10)

and Dmin is defined [34, eq. 24]:


D

Dmin

Bf R f
Df
= Df G2
=
c
( 2 )G

(5.11)

Kammler et al. [35] determined from USAXS measurements of agglomerated and


non-agglomerated particles the particle volume to surface ratio that is compared to those
obtained by nitrogen adsorption measurements. The primary particle diameter was derived from

dV /S = 6

Q
V
=6 ,
BP orod
S

(5.12)

where Q is the Porod invariant. It is the integral of the scattering curve concerning the
primary particles.

Z
Q=

q 2 I(q)dq = 2 2 re2 N ()2 V

(5.13)

Beaucage et al. [36] presented a general index of polydispersity for symmetric particles. If the Porod and the Guinier (eq. (5.3) and (5.5)) of the primary particles show
an intersection, this indicates that the primary particles are polydisperse. The index of
polydispersity is given by following equation:
P DI =

BRg4
.
1.62G

(5.14)

The ratio BRg4 /G has a minimum of 1.62 for monodispered spheres. For monodisperse
spheres, PDI is 1, while a PDI of 4.93 or 5.56 is obtained for the self-preserving limit.

98
The PDI is directly related the geometric standard deviation:
r
g = exp () = exp

ln(P DI)
12

(5.15)

The number density of particles within the flame can be calculated by the scattering
invariant Q and the Guinier prefactor G.
N=

Q2
4[re2 ()2 ] 4 G

(5.16)

The contrast factor [re2 ()2 ] is calculated from


re2 ()2

re2

Na Ne
M

2

= 1.0232 1023 cm4

(5.17)

for zirconia in vacuum, where NA is the Avogadros number Ne the number of electrons
in a zirconia molecule and M the molar weight of zirconia. The electron cross section re2
is 7.49 1026 cm2 and the zirconia density is 5.8 cm2 /g for tetragonal crystalline structure.
The volume fraction V of zirconia in the flame is given by
V =

Q
2 2 re2 ()2

(5.18)

Agglomerates can be described by the radius of gyration, RG2 . RG2 reflects a high order
of moment, it is indicative of the largest agglomerates [34]. An other important number
to describe agglomerates is the number of primary particles in an agglomerate, z, and is
based on the weight average,
z=

N2 n2e,2
G2
(N1 /z) (ne,1 z)2
=
=
.
G1
N1 n2e,1
N1 n2e,1

(5.19)

G1 and G2 are the Guinier prefactor of the primary particles and of the agglomerates,
respectively, ne is the difference in electrons of the particle or agglomerates and the air,
(ne = V ).

99

5.4

Results and discussion

5.4.1

Spray flame

Figure 5.2a shows the axial gas velocity along the centerline for flame A (solid line)
and flame B (broken line). The gas is accelerated by droplet combustion until 14 mm
HAN [13] where the maximum gas velocity of 170 and 150 m/s is reached for flames A
and B, respectively. Then these velocities decay to 59 and 52 m/s at HAN 45 and 60
mm, respectively. There droplet data rates are below 0.1 kHz indicating nearly complete
droplet evaporation. The velocity profiles are extrapolated by an asymptotic decay of
1/HAN, as this is typically observed for turbulent jet [37]. The gas velocity for flame A is
higher than that of flame B as the momentum of the dispersion gas disperses more liquid
precursor with the latter flame. However, flame A has less combustible liquid so that the
visible flame height is reduced and the gas velocity decay is steeper [12].
Figure 5.2b shows the axial gas velocity as a function of the radial position, r, in
flame A at 8 (circles), 13 (triangles), 18 (diamonds) and 25 mm HAN (squares). The
radial measurements show good symmetry of the spray flame. The reconstructed radial
velocity data and single point radial measurements agree within 8 % due to lower statistics
compare to single point measurements. Within 1 mm radius the gas velocity decreases
by only 10 to 15 % compared to the maximum gas velocity on the centerline for all HAN.
For r > 2 mm the gas velocity decreases radially very fast reaching 20 m/s at 5 mm
where droplet data rate drops below 0.1 kHz compared to about 16 kHz on the center
line indicating the outer board of the spray flame. Both flames do not expand radially in
contrast to a pilot spray flame [38] where 27.1 and 81.1 ml/min is fed through a nozzle
and dispersed by 50 l/min O2 .
For different spray flames Table 5.1 shows the nozzle Re of the dispersion gas.
Flames 1 and 2 [13] have similar nozzle Re of about 19000 and do not depend on liquid
feed velocity and that is negligible at the nozzle compared to that of the dispersion gas.
For the large FSP unit [39], the nozzle Re is about twice as high as here as the equivalent

100

Figure 5.2: Axial gas velocity profiles of spray flames measured by Phase Doppler
Anemometry (PDA) for a) the centerline of flames A (4 ml/min, solid line) and B (8
ml/min, broken line) and b) at different radial positions of flame A at HAN 8 (circles),
13 (triangles), 18 (diamonds) and 25 mm (squares).

orifice diameter had been doubled [13] at similar dispersion gas velocity. Compared to
Madler et al. [25], the Re is similar as same nozzle geometry, flows and solvents were
used. Schulz et al. [14] used xylene as solvent which has a viscosity half of that of EtOH
resulting in twice as large Re as here.
Increasing the liquid feed rate from 4 to 8 ml/min almost doubles the visible spray
flame height from 7 to 13 cm. Figure 5.3 shows the integral (line-of-sight) centerline
temperature profiles of flame A (solid line) and B (broken line) and flame 1 and 2 [13] as
a function of residence time. The maximum gas temperature of flame A is 2620 K and
2800 K for flame B, both at 0.05 ms (3 mm HAN). This is consistent with the maximum
gas temperatures of 2500 and 2750 in flame 1 and 2 and with the peak temperature of
2650 K measured in a silica-laden / ethanol spray flame [25] and that of a titania-laden
flame using xylene as solvent where the maximum temperature was about 2500 K [14]. In

101

Table 5.1: Spray flame characteristics.


Flames

Reynolds Number Maximum


Temperature, K

Enthalpy density,
kJ/ggas

Flame A
Flame B
Flame 1 [13]
Flame 2 [13]
Madler et al. [25]
Schulz et al. [14]

19,070
19,040
46,100
45,500
19,030
34,541

9.5
14.4
7.2
14.5
5.1
15.9

2620
2800
2500
2750
2650
2500

the latter, the temperature profile forms a plateau up to 50 mm HAN and then decreases
to 1100 K at 120 mm. The difference in the temperature profile might be caused by
slower evaporation and combustion of xylene (boiling point: 144 C) compared to ethanol
(boiling point: 78 C). The maximum temperatures can be explained with the enthalpy
density in the flame that increases for higher liquid feed rates, leading to an increased
maximum flame temperatures (Tab. 5.1).
At 0.2 ms cooling rates are 7106 and 3106 K/s for flame A and B, respectively, (18
and 25 mm HAN, respectively) and there after decrease about one order of magnitude
to 6105 K/s and 6105 K/s for residence time of 0.4 to 1 ms. These cooling rates are
significantly higher than those of 3105 K/s close to the nozzle of a pilot FSP unit [13].
This difference in cooling rates can be explained by the 20 % slower gas velocity and the
higher surface area of the smaller spray jet per unit precursor volume fed. As a result
heat transfer is more pronounced in flames A/B than in flames 1/2. Further downstream,
cooling rates are similar for both flame types decreasing from 8 to 2105 K/s after 0.25
ms.
As gas volume (m3 ) increases linear with increasing temperature using ideal gas
law, the gas velocity (m/s) should be proportional to T 1/3 . No correlation could be find
between gas velocity and temperature as gas velocity depends also on liquid feed and
dispersion gas flow rates.

102

Figure 5.3: Centerline flame temperature of spray flames at liquid feed rates of 4 (A)
and 8 (B) ml/min dispersed with 5 l/min of oxygen and from Mueller et al. [13] at liquid
feed rates of 27.1 (flame 1) and 81.1 ml/min (flame 2) dispersed with 50 l/min along the
centerline. All temperatures are measured by in situ FTIR/transmission spectroscopy.
The residence time is calculated using the velocity data form Figure 5.2a and from Heine
and Pratsinis [38] for flame 1 and 2.

103

Figure 5.4: Scattering intensity as a function of the scattering vector, q, for flame A.
At t < 0.25 ms droplets are detected in the low q range (< 0.02
A1 ) while in the high q
range (> 0.04
A1 ) the nanosized particles are detected. For t > 0.25 ms no droplets are
detected.

5.4.2

Droplet evaporation by SAXS

Figure 5.4 shows the intensity of the SAXS measurements as a function of q vector (q =
4/ sin(/2)) for flame A. From 0.16 (1 mm HAN, butterflies) to 0.25 ms (7 mm HAN,
1 ) and particle (q > 0.04
circles), both droplets (q < 0.02 A
A1 ) structures can be
seen. Droplets are in the range of micrometers and therefore present at low q while the
nanosized particles are present at high q values. Comparing scattering spectra of a pure
EtOH spray flame at 0.16 and 1.85 ms shows no difference in intensity at high q that
demonstrating that no droplets are present at high q. The -4 slope at low q corresponds
to Porods power-law [40] that indicates the presence of micrometer size species in the

104
spray corresponding to droplets (Fig. 5.4). Droplets can be seen by SAXS until 0.25 ms
(7 mm HAN). At higher residence times (t > 0.25 ms, 7 mm HAN) the -4 slope is not
observed, indicating that from here on the droplet concentration is below the detection
limit of SAXS. In contrast to this, PDA measurements detected droplets present up to 45
mm HAN (Fig. 5.2a) as PDA can count single droplets. Also the intensity of the power
law is proportional to the contrast factor. For EtOH the contrast factor is 2 orders of
magnitude smaller than that of zirconia particles, therefore scattering from droplets can
be neglected at higher HAN.

5.4.3

Primary particle growth by SAXS

Figure 5.5 shows the evolution of average primary particle diameter along the centerline
of spray flame A determined from the SAXS volume to surface ratio, dV /S (circles), that
is equivalent to the Sauter, d3,2 , or BET-diameter, dBET [35]. First measurements are
made at 0.04 ms (0.4 mm) where dV /S is 6 nm, then dV /S decreases to 2.5 nm at 0.22
ms (3 mm HAN) before it increases again to about 10 nm at 0.68 ms (50 mm HAN).
The particle size of 2.5 nm is the smallest ever measured particle diameter in a flame by
SAXS. At the beginning of the process (0.04 ms), the supersaturation is low resulting
in formation of relative big critical clusters byF nucleation. As temperature increases
and more condensible species are formed supersaturation is increased and nucleation is
accelerated. Therefore, smaller critical clusters are formed, resulting in a smaller average
particle size [41]. This decrease in particle size at the early stage of the flame was also
seen by Kammler et al. [23] in premixed vapor-fed flames. At 0.21 ms particle growth
by coagulation becomes significant, leading to a continuous increase of dV /S . After 0.27
ms the particle growth rate decreases as the particle number concentration in the flame
is decreased as will be shown later on.
After 0.68 ms dV /S remains nearly constant till 1.85 ms (100 mm HAN, dV /S = 10.8
nm, circles). Primary particle growth stopped at 0.68 ms (50 mm HAN) corresponding to
a gas temperature of about 1620 K. This is consistent with Mueller et al. [13] who showed

105

Figure 5.5: Evolution of primary particle size, dV /S , obtained form in situ SAXS technique as a function of the flame residence time, t (circles). dV /S is calculated from the
volume to surface ratio. The filter particle size obtained with BET (triangle), SAXS
(circles) and TEM-image analysis are also shown (square).

with thermophoretic sampling and TEM-image analysis the growth of ZrO2 particles of
similar size and high temperature residence time stopped at 1600 K. From this point on the
primary particle distribution (Fig. 5.6) is essentially frozen resulting in soft agglomerates
that consist of smaller hard fractal-like agglomerates [42] and even physically bounded
single primary particles. Figure 5.5 shows also that FSP-made particles collected on the

106
filter have a dV /S and dBET (triangle) of 11.7 nm and 10 nm, respectively, that are in very
good agreement as it has been shown by Kammler et al. [35] and Beaucage et al. [36]
for other metal oxide powders made by vapor-fed flames. Furthermore, the Sauter mean
diameter from counting TEM images of particles collected on the filter is 10.6 nm (Fig.
5.5, square).
In situ wide angle X-ray scattering (WAXS) is also used to study the evolution of
crystal structure [43]. No signal is detected by the WAXS detector at any HAN during
ZrO2 synthesis by FSP indicating that a crystalline structure may not have been developed within the 1.85 ms. Particles collected on the filter from flames A and B show the
crystalline metastable tetragonal structure as measured by WAXS and XRD [39]. Thermophoretic sampling was not possible at the HAN where SAXS signal was strong enough
for particle detection as the presence of droplets ruined the TEM grids [39]. TEM analysis
of filter powder from flame A, however, was in good agreement with SAXS and BET data.
Figure 5.6 shows the evolution of geometric number-based standard deviation, g
(circles), of the primary particle size distribution (PPSD) assuming spherical particles and
a lognormal distribution [34]. At 0.04 ms (0.4 mm HAN) g is initially 1.49 and increases to
1.71 at 0.17 ms (1 mm HAN) and then g decreases continuously to 1.48 at 0.59 ms which
is closed to the self-preserving limit, for the free molecular regime it is 1.45 [44]. At this
position only single particles are present in the spray flame (Fig. 5.7). At 0.68 ms (50 mm
HAN), where the first agglomerates are formed, g decreases to 1.35 which is significantly
below that self-preserving limit. At this point two levels are used to fit the scattering
curve with the unified fit [33, 34]. The first level describe the primary particles while
the second the agglomerates. When (hard-) agglomerates are formed the size distribution
becomes narrower possibly by sintering of the constituent primary particles [45] during
agglomeration. At the end of the flame (1.85 ms; 100 mm HAN) g = 1.31 while for the
filter powder a higher value of g = 1.47 is detected. The increased g of the filter powder
results from radial differences in the flame where smaller primary particles are formed at
the flame edges and are mixed with other streamlines [46]. The higher g close to the

107

Figure 5.6: Evolution of the geometric standard deviation, g (circles), and zirconia
volume fraction (triangles) as a function of the residence time (circles) compared to the
filter powder (square).

nozzle tip indicate high polydispersity probably caused by continuous formation of new
particles [12]. The g has rather leveled off at t > 0.68 ms, indicating that the narrowing
of the PPSD by sintering has ended. The evolution of FSP-made ZrO2 g by SAXS here
is consistent with that measured by TS-TEM in larger FSP units [13].
Figure 5.6 shows also the total zirconia volume fraction (triangles) as a function of
the residence time along the burner centerline. The volume fraction increases rapidly to
about 1.7106 at 0.19 ms (15 mm HAN) and then rather to 3.22 106 at 0.68 ms (50
mm HAN). Finally the volume fraction decreases again to 2.4 106 at 1.85 ms (HAN 100
mm) by the radial expansion of the flame-spray jet by air entrainment [38].

108
Figure 5.7 shows the evolution of Df (triangles), Df m (diamonds), c (squares) and
BR (circles) above the burner. At t < 0.68 ms (50 mm HAN) no mass-fractal zirconia
is detected as in this high temperature zone non-agglomerated zirconia is formed and
particles fully coalesce. This is in agreement with Mueller et al. [13] who reported the
formation of single particles in the first part of the flame for both low (100 g/h) and high
(300 g/h) zirconia production rates. After their particles left the luminous region of the
spray flame agglomeration started. Here, the formation of agglomerates starts at t > 0.68
ms where primary particle growth stops and the gas temperature drops below 1600 K
(Fig. 2, 4). At 0.68 ms, Df = 1.2 that increases at 1.85 ms (100 mm HAN) to 1.47 and
finally increases to 1.53 at the filter. Between 0.68 and 1 ms, Df and Df m are similar
so c is close to one and BR close to zero. These indicate that first linear structures are
formed after 1 ms when branching increases as c and BR increases.
Figure 5.8 shows the number density of zirconia primary particles (triangles) in the
spray flame as a function of the residence time above the nozzle. A sharp increase in
particle density is observed and has its maximum at 0.22 ms (3 mm HAN) indicating
fast particle nucleation. The maximum value of g is observed at 0.22 ms (1 mm HAN,
Fig. 5) where the particle size of dV /S = 2.5 nm is smallest, indicating very broad size
distribution. Till this point nucleation is dominant, afterwards particle growth by coagulation and sintering is dominant. Nucleation can be driven by supersaturation of the
vapor or by chemical nucleation [47]. Nucleation leads to a minimum critical in particle
size, a maximum in g and a rapid increase in particle number concentration [41]. Here,
a critical volume fraction is found to be 1106 . At higher residence times the number
concentration decreases while the particle size increases (Fig. 5.5). The particle number
concentration at 0.68 ms (50 mm HAN) remains almost constant as particle growth stops
(Fig. 5.5). However, a small decrease is observed at the end of the spray that may result
from dilution by air entrainment.
Figure 5.8 shows also the evolution of the radius of gyration of the agglomerates
RG (circles). For t < 0.68 ms no data are shown as only single primary particles have

109

Figure 5.7: Evolution of the mass-fractal dimension, Df (triangles), Df m (diamonds)


connectivity dimension c (squares) and branch fraction (filled circles) as a function of
residence time above the nozzle obtained by in situ SAXS measurements along the centerline. Data from the filter powder is also shown. Below 0.68 ms residence time no data
are shown as particles are non-agglomerated.

been detected. At t > 0.68 ms formation of agglomerates starts as the gas temperature is
too low for full particle coalescence. The radius of gyration of the agglomerates is about
30 nm and stays constant till the end of the spray flame. The radius of gyration of the
agglomerates of the filter powder is about 140 nm. This is about 5 times larger than that
measured at 1.85 ms (100 mm HAN). At the filter, particles are collected that are formed
along different streamlines of the spray flame. The radius of gyration reflects a high
order moment of the agglomerates size, so it is an indicative of the largest agglomerates.

110

Figure 5.8: Evolution of the radius of gyration of the agglomerates, Rg (circles), particles
per agglomerate, z (squares), and primary particle number concentration (triangles) as
a function of the residence time obtained form in situ SAXS measurements along the
centerline (circles). Agglomerates are only formed after 0.68 ms.

Larger agglomerates may have formed at the edges of the flame where gas temperature
are significantly lower than that on the centerline [23].
Figure 5.8 shows also the evolution of the average number of primary particles per
agglomerate, z (squares). The agglomerates at 0.68 ms (50 mm HAB) consist of about 4
particles each that increase to 14 at 1.85 ms (100 mm HAN). As the radius of gyration
does not change considerably, this indicates that agglomerates form first a linear structure
and then start branching as discussed in Figure 5.7. The product powder has about 64

111
particles per agglomerates that result form mixing with other streamlines as has been
discussed already.

5.5

Conclusions

A series of non-intrusive diagnostics were used to obtain an understanding of particle formation and growth as well as agglomerate evolution, for the first time, during nanoparticle
synthesis by the highly versatile FSP process. Maximum temperatures up to 2800 K and
cooling rate of 7 106 K/s were measured during synthesis of nanostructured ZrO2 .
Droplets and particles could be distinguished by SAXS simultaneously as the former
were about 2 orders of magnitude larger than nanoparticles and the number concentration
is about 8 orders of magnitude smaller than that of the particles. The SAXS measurements show the evolution of primary-particle diameter, mass-fractal structures, geometric
standard deviation, zirconia volume fraction, zirconia number concentration, agglomerate
size and number of primary particles per agglomerate. SAXS data indicate a nucleation
burst at about 0.22 ms followed by primary particle growth that stops at about 1600 K,
where formation of agglomerates starts (t > 0.68 ms). Initially the geometric standard
deviation of the primary particles rapidly increases reflecting new particle formation, then
approaching 1.48, indicative the self-preserving limit of coagulation in the free molecular
regime, and at t 0.68 ms drops below that limit indicate of sintering. This in situ technique allows information on particle dynamics to be obtained in a single measurement.
WAXS showed that early in the flame no crystalline structures may exist while the filter
powder showed metastable tetragonal ZrO2 .

References
[1] C. R. Bickmore, K. F. Waldner, R. Baranwal, T. Hinklin, D. R. Treadwell, and R. M.
Laine. Ultrafine titania by flame spray pyrolysis of a titanatrane complex. J. Euro.
Ceram. Soc., 18(4):287297, 1998.

112
[2] S. Grimm, M. Schultz, and R. M
uller. Flame pyrolysis- a prepartion rout for ultrafine
pure gamma Fe2 O3 powders and control of their particle size. J. Mater. Sci., 32:1083
1092, 1997.
[3] L. Madler, W. J. Stark, and S. E. Pratsinis. Flame-made ceria nanoparticles. J.
Mater. Res., 17(6), 2002.
[4] J. J. Helble. Combustion aerosol synthesis of nanoscale ceramic powders. J. Aerosol
Sci., 29(5-6):721736, 1998.
[5] W. J. Stark, L. Madler, M. Maciejewski, S. E. Pratsinis, and A. Baiker. Flame
synthesis of nanocrystalline ceria-zirconia: effect of carrier liquid. Chem. Commun.,
(5):588589, 2003.
[6] C. R. Bickmore, K. F. Waldner, D. R. Treadwell, and R. M. Laine. Ultrafine spinel
powders by flame spray pyrolysis of a magnesium aluminum double alkoxide. J. Am.
Ceram. Soc., 79(5):14191423, 1996.
[7] H. Schulz, L. Madler, S. E. Pratsinis, P. Burtscher, and N. Moszner. Transparent
nanocomposites of radiopaque, flame-made Ta2 O5 /SiO2 particles in an acrylic matrix.
Adv. Funct. Mater., 15(5):830837, 2005.
[8] R. Jossen, R. Mueller, S. E. Pratsinis, M. Watson, and M. K. Akhtar. Morphology
and composition of spray-flame-made yttria-stabilized zirconia nanoparticles. Nanotechnology, 16(7):S609S617, 2005.
[9] R. Strobel, W. J. Stark, L. Madler, S. E. Pratsinis, and A. Baiker. Flame-made
platinum/alumina: structural properties and catalytic behaviour in enantioselective
hydrogenation. J. Catal., 213(2):296304, 2003.
[10] T.T. Kodas and M. J. Hampden-Smith. Aerosol Processing of Materials. Wiley-VCH,
New York, 1999.
[11] G. Beaucage, H. K. Kammler, R. Mueller, R. Strobel, N. Agashe, S. E. Pratsinis,
and T. Narayanan. Probing the dynamics of nanoparticle growth in a flame using
synchrotron radiation. Nature Mater., 3(6):031401, 2004.
[12] M. C. Heine and S. E. Pratsinis. Droplet and particle dynamics during flame spray
synthesis of nanoparticles. Ind. Eng. Chem. Res., 44(16):62226232, 2005.
[13] R. Mueller, R. Jossen, H. K. Kammler, and S. E. Pratsinis. Growth of zirconia
particles made by flame spray pyrolysis. AIChE J, 50(12):30853094, 2004.

113
[14] H. Schulz, L. Madler, R. Strobel, R. Jossen, and S. E. Pratsinis. Independent control
of metal cluster and ceramic particle characteristics during one-step synthesis of
Pt/TiO2 . J. Mater. Res., 20(9):25682577, 2005.
[15] R. A. Dobbins and C. M. Megaridis. Morphology of flame-generated soot as determined by thermophoretic sampling. Langmuir, 3:254259, 1987.
[16] H. Bockhorn, H. Geitlinger, B. Jungfleisch, T. Lehre, A. Schon, T. Streibel, and
R. Suntz. Progress in characterization of soot formation by optical methods. Phys.
Chem. Chem. Phys., 4(15):37803793, 2002.
[17] C. Oh and C. M. Sorensen. Light scattering study of fractal cluster aggregation near
the free molecular regime. J. Aerosol Sci., 28(6):937957, 1997.
[18] C. M. Sorensen, W. B. Hageman, T. J. Rush, H. Huang, and C. Oh. Aerogelation in
a flame soot aerosol. Phys. Rev. Let., 80(8):17821785, 1997.
[19] C. M. Sorensen. Light scattering by fractal aggregates: A review. J. Aerosol Sci.,
35:648, 2001.
[20] A. J. Hurd and W. L. Flower. In situ growth and structure of fractal silica
aggregates in a flame. J. Colloid Interface Sci., 122(1):178192, 1988.
[21] Y. C. Xing, U. O. Koylu, and D. E. Rosner. In situ light-scattering measurements
of morphologically evolving flame-synthesized oxide nanoaggregates. Applied Optics,
38(12):26862697, 1999.
[22] R. Mueller, H. K. Kammler, S. E. Pratsinis, A. Vital, G. Beaucage, and P. Burtscher.
Non-agglomerated dry silica nanoparticles. Powder Technol., 140(1-2):4048, 2004c.
[23] H. K. Kammler, G. Beaucage, D. J. Kohls, N. Agashe, and J. Ilavsky. Monitoring
simultaneously the growth of nanoparticles and aggregates by in situ ultra-smallangle X-ray scattering. J. Appl. Phys., 97(5), 2005.
[24] H. K. Kammler, R. Jossen, P. W. Morrison, S. E. Pratsinis, and G. Beaucage. The
effect of external electric fields during flame synthesis of titania. Powder Technol.,
135:310320, 2003.
[25] L. Madler, H. K. Kammler, R. Mueller, and S. E. Pratsinis. Controlled synthesis
of nanostructured particles by flame spray pyrolysis. J. Aerosol Sci., 33(2):369389,
2002.
[26] A.H. Lefebvre. Atomization and Sprays. Taylor & Francis, London, 1989.

114
[27] P. E. Best, R. M. Carangelo, J. R. Markham, and P. R. Solomon. Extension of
emission-transmissiontechnique to particulate samples using FT-IR. Combust. Flame,
66(1):4766, 1986.
[28] H. K. Kammler, S. E. Pratsinis, P. W. Morrison, and B. Hemmerling. Flame, temperature measurements during electrically assisted aerosol synthesis of nanoparticles.
Combust. Flame, 128, 2002.
[29] P. R. Griffiths and J. A. Haseth. Fourier transform infrared sprectroscopy. John
Wiley and Sons, New York, 1986.
[30] P. W. Morrison, R. Raghavan, A. J. Timpone, C. P. Artelt, and S. E. Pratsinis.
In-situ Fourier transform infrared characterization of the effect of electrical fields on
the flame synthesis of TiO2 particles. Chem. Mater., 9(12):27022708, 1997.
[31] T. Narayanan, O. Diat, and P. Bosecke. SAXS and USAXS on the high brilliance
beamline at the ESRF. Nucl. Instrum. Meth. A, 467:10051009, 2001.
[32] D. Pontoni, T. Narayanan, and A. R. Rennie. Time-resolved saxs study of nucleation
and growth of silica colloids. Langmuir, 18(1):5659, 2002.
[33] G. Beaucage. Approximations leading to a unified exponential power-law approach
to small-angle scattering. J.Appl. Crystal., 28:717728, 1995.
[34] G. Beaucage. Determination of branch fraction and minimum dimension of massfractal aggregates. Phys. Rev. E, 70(3):370374, 2004.
[35] H. K. Kammler, G. Beaucage, R. Mueller, and S. E. Pratsinis. Structure of
flame-made silica nanoparticles by ultra-small-angle x-ray scattering. Langmuir,
20(5):19151921, 2004.
[36] G. Beaucage, H. K. Kammler, and S. E. Pratsinis. Particle size distributions from
small-angle scattering using global scattering functions. J.Appl. Crystal., 37:523535,
2004.
[37] A. Bejan. Convection Heat Transfer. Wiley, New York, 1984.
[38] M. C. Heine, L. Madler, R. Jossen, and S. E. Pratsinis. Direct measurement of air
entrainment during nanoparticles synthesis in spray flames. Combust. Flame, 2005.
[39] R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, and M. K. Akhtar. Zirconia
nanoparticles made in spray flames at high production rates. J. Amer. Ceram. Soc.,
87(2):197202, 2004.

115
[40] A. Guinier and G. Fournet. Small-Angle Scatteirng of X-Rays. John Wiley and Sons,
New York, 1955.
[41] S. E. Pratsinis, T. T. Kodas, M. P. Dudukovic, and S. K. Friedlander. Aerosol reactor design - effect of reactor geometry on powder production and vapor-deposition.
Powder Technology, 47(1):1723, 1986.
[42] S. Tsantilis and S. E. Pratsinis. Soft- and hard-agglomerate aerosols made at high
temperatures. Langmuir, 20(14):59335939, 2004.
[43] D. O. de Zarate, C. Boissiere, D. Grosso, P. A. Albouy, H. Amenitsch, P. Amoros,
and C. Sanchez. Preparation of multi-nanocrystalline transition metal oxide (TiO2 NiTiO3 ) mesoporous thin films. New J. Chem., 29(1):141144, 2005.
[44] J. D. Landgrebe and S. E. Pratsinis. A discrete-sectional model for particulate production by gas-phase chemical-reaction and aerosol coagulation in the free-molecular
regime. J. Colloid Interface Sci., 139(1):6386, 1990.
[45] S. Tsantilis and S. E. Pratsinis. Evolution of primary and aggregate particle size
distributions during coagulation and sintering. AIChE J, 46:407415, 2000.
[46] T. Johannessen, S. E. Pratsinis, and H. Livbjerg. Computational fluid-particles
danymics for the flame synthesis of almunina particles. Chem. Eng. Sci., 55:177191,
2000.
[47] M. Frenklach, L. Ting, H. Wang, and M. J. Rabinowitz. Silicon particle formation
in pyrolysis of silane and disilane. Isr. J. of Chem., 36(3):293303, 1996.

116

CHAPTER

SIX

Research Recommendations
It is well known that particles as filler materials are more effective in reinforcing when
the particles are agglomerated. [1]. The characterization of those particles were made
by several people [2, 3, 4, 5, 6] using ultra small-angle x-ray scattering(USAXS), neutron
scattering (SANS), x-ray diffraction (XRD) or light scattering. Such materials consist
of primary particles, hard soft agglomerates in the range of a few nanometers to 100
micrometer. Figure 6.1 show a schematic of the formation of particles made by flame
process. Particle nucleate early in the flame and grow due to particle collision in hot
the temperature zone. Later in the flame particles form agglomerates with robust sinter
bridges so-called hard agglomerates. As gas temperature decreases the hard agglomerates
are clustered into dens agglomerates so-called soft-agglomerates with much less robust
bonding that can be disrupted by shear.
Schaefer et al. [3] produced silica structure by wet phase method using XRD, USAXS, SAXS and light scattering they get structural information of the wet phase made
silica from q values of 3 107 to 4
A. Structure form primary particle (few nanometer) to
soft agglomerates (100 m) could be determined. Sorensen at al. [7] show aerogelation
in a flame of soot aerosol using light scattering. Light scattering indicate the formation
the presence of two different soot structures. One the typical mass fractal soot (Df =
1.8) in the submicron range and second with non fractal dimension in the range of 5 m.
117

118

Figure 6.1: The basic steps of particle formation and growth by gas-to particle conversion.

Tsantilis et al. [8] developed a model for silica made by flame synthesis using SiCl4 as
precursor to distinguish regions for synthesis of soft agglomerates, hard agglomerates and
non-agglomerates. The process conditions (precursors initial volume fraction, maximum
temperature, residence time, cooling rate) promoting or suppressing the onset of hard or
soft agglomerates. Since now, there is no detail experimental studies of the formation of
soft agglomerates in the flame.

119

Figure 6.2: Schematic of the light scattering measuring system.

Figure 6.2 shows the schematic of a light scattering set-up for in-situ measurements
in flames. A water cooled 5W argon-ion laser (Innova 70, LA-70-5), generating a green
laser beam with wavelength of 514.4 nm is used at light source. The used power was about
500 mW. The scattered light by particles and agglomerates in the flame was collected by
a lens (L1) with 5 cm in diameter and focal point. This lens parallelize the scattered
light and send it through a narrow bandpass filter (F1) ( = 514nm). Only light of that
wavelength pass the filter . A seconde filter (F2) was installed to decreased the intensity.
The seconde lens (L2, 5 cm in diameter and focal point) focus again the scattered light. A
CCD camera detect the scattered light and was installed that scattered light of an angle
of 20 could detected. The CCD camera together with the lenses and the filters can be
turned to get scattering signals to an angle of 140 or q values between 0.6 and 10 m.
Preliminary experiments were done with a premixed flame [2] that produced agglomerated
silica at production rate of 14 g/h.
Figure 6.3 show the combination of USAXS and light scattering data of that premixed flame [2] at 28 and 40 mm above the burner.
The USAXS data form Kammler et al. [2] shows at 28 mm a dV /S of 10.7 nm, an
agglomeration radius of 128 nm and a fractal dimension of 2.4. At 40 the dV /S remains
constant, the agglomeration radius decrease to 99 nm and the fractal dimension to 2.3.
The light scattering data show a -4 slop for the third level at low q (2.5 - 6
A). The

120

Figure 6.3: In-situ scattering profiles of flame made silica at 28 (open symbols) and 40
mm (filled symbols). Data were accumulated by USAXS (diamonds) and light scattering
(circles).

third Guinier regime shows an agglomerate radius of about 600 nm. The combination of
the USAXS data and the light scattering data shows that in that premixed silica-laden
flame a third structure is formed, the so-called soft agglomerates. These agglomerates are
weakly bonded with a surface fractal dimension of 2 (smooth particles) and are not of
mass fractal structure.
It was shown that soft agglomerates are formed during flame process and can be
detected by scattering technique as shown by Sorensen at al. [7]. The formation of soft
agglomerates have to be compared with maximal flame temperature, cooling rates and

121
particle concentration within the flame to get better understanding of the formation of
flame process especially the formation of soft agglomerates. These information are also
helpful for further modelling work.
Scattering technique can also be used to detect droplet in a spray flame (Chap. 1).
The PDA reach its limitation to measured evaporation rates as the mass balance can not
be closed. The scattering technique can overcome this limitation as it measure in-situ the
volume to surface ratio of the droplets in the flame as shown for particles in chapter 5.
Therefore, evolution of volume to surface ratio in the spray flame can be used together
with velocity and temperature data droplet evaporation rates can be measured. This data
can be used to improve numerical model investigation.

References
[1] T. A. Witte, M. Rubenstein, and R. H Colby. J. Phys. II (France), 3:367372, 1993.
[2] H. K. Kammler, G. Beaucage, D. J. Kohls, N. Agashe, and J. Ilavsky. Monitoring
simultaneously the growth of nanoparticles and aggregates by in situ ultra-small-angle
x-ray scattering. J. Appl. Phys., 97(5), 2005.
[3] D. W. Schaefer, Ricker T., M. Agamalian, J. S. Lin, S. Sukumaran, c. Chen,
G. Beaucage, C. Herd, and Ivie J. Multilevel structure of reinforcing silica and carbon.
J. Appl. Crystl., 33:587591, 1998.
[4] G. Beaucage. Determination of branch fraction and minimum dimension of massfractal aggregates. Phys. Rev. E, 70(3):370374, 2004.
[5] M. Choi, J. Cho, J. Lee, and H.W. Kim. Measurements of silica aggregate particles
growth using light scatteringand thermophoretic sampling in a coflow diffusion flame.
J. Nanoparticles Res., 1:169183, 1999.
[6] C. M. Sorensen. Light scattering by fractal aggregates: A review. J. Aerosol Sci.,
35:648, 2001.
[7] C. M. Sorensen, W. B. Hageman, T. J. Rush, H. Huang, and C. Oh. Aerogelation in
a flame soot aerosol. Phys. Rev. Let., 80(8):17821785, 1997.
[8] S. Tsantilis and S. E. Pratsinis. Soft- and hard-agglomerate aerosols made at high
temperatures. Langmuir, 20(14):59335939, 2004.

122

APPENDIX

Reproducibility of nanoparticle production and color


investigation on titania particle made by pilot scale
FSP
The reproducibility of the large scale flame spray pyrolyse rector in respect to particle
size, morphology and composition was tested for titania production. 19 samples were
taken and compared for specific surface area (SSA) measured by N2 adsorption (BET),
crystallinity by x-ray diffraction (XRD) and carbon content by temperature programmed
oxidation (TPO). Results were statistically analyzed.
Titania made by the pilot scale FSP showed a yellowish coloration. The intensity of
the coloration was investigated by changing process parameter (liquid feed rate, titania
concentration or dispersion gas flow rate).

A.1

Experimental

The experimental set-up consists of an external-mixing gas-assisted stainless-steel nozzle


(Schlick-D
use, Gustav Schlick GmbH + Co, 970/4-S32) that is made of a capillary tube
of 0.5 mm ID and an annular gap that can be adjusted to keep a constant pressure drop
(1 bar) across the nozzle [1]. Liquid precursor (35 ml/min) is fed through the capillary
by a pulsation-free precision piston pump (Isco Inc., 1000D). That liquid is dispersed
123

124
(atomized) by 50 l/min O2 (Pan Gas, Switzerland, 99.95%). The resulting spray is ignited
by a supporting CH4 /O2 diffusion flame (CH4 = 2 l/min (Pan Gas, Switzerland, 99.5%),
O2 = 4.5 l/min) surrounding the nozzle [1]. The dispersion O2 flow rate is controlled by a
mass flow controller (Bronkhorst). Product powders are collected with a Jet filter (FRR
12/2.4, Friedli AG, Switzerland) containing twelve baghouse filters. Small samples of the
product powder (1g) are collected on-line using a bypass connected to the inlet pipe of
the baghouse filter upstream of the check valve [1]. For this, a vacuum pump (Vacubrand
RE 16) is used and particles are collected on a glass fiber filter (Whatmann GF/A), 150
mm in diameter, that is located in a stainless steel holder.

A.1.1

Reproducibility of pilot scale FSP

Titanium tetraisopropoxide (TTIP, Simga-Adrich, Chemie GmbH, Buchs) was dissolved


in toluene to give a 0.5 M solution. During TiO2 production 19 product samples were
taken collected on the by-pass filter for about 30 minutes, each. Those samples were
analyzed by nitrogen adsorption at 77 K (BET-method, Micrometrics Inc., Tristar 3000,
The Netherlands) to determine specifice surface area after degassing them for at least 1 h
at 150 C under nitrogen. X-ray diffraction (Burker Advanced D8, Karlsruhe, Germany)
was use to determined crystallinity (rutile and anatase) and crystal size. Carbon content
was determined using temperature programmed oxidation (TPO).

A.1.2

Investigation on the yellowish color of titania made by


pilot scale FSP

Titania was produced at different liquid flow rates (17.5 - 70 ml/min), O2 dispersion gas
flow rates (0 - 70 l/min) and molar concentrations (0.25 - 1.0 M) from TTIP in toluene.
The production rate was kept constant at 84 g/h. Particle were collected form the bypass filter and from the upper part of three filter candles in the baghouse filter. The
candles were cleaned with a vacuum cleaner after each experiments. All samples were

125

Figure A.1: Reproducibility of titania samples made at production rate of 84 g/h un


terms of dBET , dXRD and the anatase weight percent.
characterized by BET, XRD and temperature programmed oxidation (TPO).

A.2
A.2.1

Results
Reproducibility of pilot scale FSP

Figure A.1 shows, the BET-equivalent particle diameter, dBET (circles), the anatase
crystal size, dXRD (triangles) and the anatase weight percent (squares) of the 19 samples
collected on the by-pass filter. The dBET varied from 16.3 to 19.1 nm with an average
of 17.9 nm and a standard deviation of 0.8 nm. The anatase dXRD varied from 22.9 to

126
26.1 nm with an average of 24.9 nm and a standard deviation of 0.8 nm. The anatase
weight fraction varied form 85.1 to 86.0 wt% and had an average of 85.6 wt% resulting in
a standard deviation of 0.6 wt%. These results show a very good reproducibility of the
average particle size and crystallinity in agreement with other statistic from pilot scale
FSP during silica and zirconia production [1, 2].

A.2.2

Investigation on the yellowish color of titania made by


pilot scale FSP

All titania powders made on the pilot scale set-up had a yellowish coloration. The intensity
in color, however, decreases with decreasing titanium isopropoxide concentration in the
precursor solution. The dispersion gas flow rate had little influence on powder coloration.
The whitest powder was made at 35 ml/min liquid feed of 0.5 M solution and dispersed by
50 l/min O2 . The yellowish color of powders made from 0.25 to 0.5 M solution disappeared
after about one week in ambient atmosphere while the powder made from the 1 M solution
retained its yellow color within this time period. In contrast to this all laboratory scale
made titania powders were perfectly white. All powders collected from the by-pass and
baghouse filter of the pilot scale set-up had the same dBET and dXRD within 5%. The
carbon content determined by TPO was less than 0.2 wt% for all powders regardless of
solution molarity.

APPENDIX

Flame spray scale-up investigation


The scale up of zirconia production by flame spray pyrolysis (FSP) is investigated on a
lab-scale [3] and a pilot scale [1] set-up with production rate of 10 to 25 g/h and 50 to 600
g/h, respectively. 0.5 M (lab scale) and 0.5 or 1.0 M (pilot scale) precursor solution were
used. Liquid rates were varied between 3 to 8 ml/min (lab scale) and 13 to 81 ml/min
(pilot scale) at O2 dispersion gas flow rates of 5 and 25 or 50 l/min. Particle size made
on the pilot scale and on the lab-scale FSP are compared by using the gas to liquid mass
ratio (GLRM) defined as is the ratio of dispersion gas mass flow [kg/min] to the liquid
feed mass rate [kg/min].

B.1

Results

Figure B.1 shows the BET-equivalent particle diameter, dBET , as function of the gas to
liquid mass ratio (GLMR). As the combustion enthalpy density hardly depends on the
precursor concentration for the zirconium precursor the dBET can be normalized by the
following equation:
dBET cor = dBET (1/Cprec )x ,

(B.1)

where Cprec is the zirconium concentration in the precursor solution and dBET cor the
normalized particle diameter. The exponent x = 0.27 has been determined from agglom127

128

Figure B.1: ZrO2 primary particle diameter as a function of the gas to liquid mass ration
(GLMR).

eration and sintering calculations [4] by varying the zirconium concentration. This value
differs from the x = 2/5 dependency of the mass balance during coagulation-coalescence
dominated growth in the free molecular regime [5].Reason might be the effect of droplet
evaporation on particle size [6] and the actuel temperature and velocity profile in the spray
flame. Comparing to the zirconia made on the lab scale FSP (open squares) the dBET
shows much smaller values as that of the dBET from the pilot scale FSP. By multiplying
the concentration corrected dBET by a factor of 1.8 the labor scale FSP data match those
of the pilot scale FSP made zirconia (gray squares). The factor of 1.8 is an empirical

129
value and that incudes the effect of the particle residence time in the hot temperature
zone of the flame as well as particle number concentration and the specific design of the
set-up. The residence time of the zirconia made in the labor scale FSP is about 3 times
longer compare to the pilot scale made FSP zirconia for the same GLMR [6, 1].

130

APPENDIX

Systematical study of silica contaminated YSZ

C.1
C.1.1

Experimental
Precursor preparation

Zirconium carbonate (Zr(OH)2 (CO3 )2 ZrO2 , ZC, LU United Intl Inc) is dissolved in
AcOH and 2-ethylhexanoic acid (2-EHA) while yttrium nitrate hexahydrate (Y(NO3 )3
6H2 O) is dissolved in EtOH and 2-ethyl hexanoic acid. Those two precursor solutions
are mixed together to form a 10 mol% YSZ precursor solution using EtOH or toluene as
additional solvent. The YSZ powders are made on a lab scale set-up 1 at a liquid feed
rate of 5.4 ml/min (20 g/h) and 5 l/min O2 dispersion gas flow rate. Different flasks are
used to investigate systematically the silica contamination of YSZ powders. Therefore,
the Zr- and/or Y-precursor solutions are prepared in normal glass flasks and in silicon
free PFA (Perfluoralkoxy) flasks according to Table C.1.

C.1.2

Analysis

The samples are prepared by dispersing the provided dry powders directly onto holey
carbon-coated copper TEM grids. The grids were then viewed in the TEM at magnifications ranging from 50 to 400,000X. The samples are analyized by using a JEOL 2000FX II
131

132

Table C.1: Systematical preparation of YSZ precursor solutions using glass or silicon
free PFA flasks.
YSZ precursor solution
1

Zr- precursor solution

PFA

PFA

Glass

Glass

Y- precursor solution

Glass

PFA

PFA

Glass

TEM operated at 200kV. The energy-dispersive X-ray spectroscopy (EDS) are conducted
with an EDAX Phoenix EDS system with a measured resolution of 130eV. During the
imaging process particular attention is given to characterizing particle size and chemistry.
Images are collected with a Gatan MultiScan CCD camera.

C.2

Results

Table C.2 shows the composition found by EDS analysis. In samples 1 and 4 composition
of 1-3 wt% SiO2 , 9-13 wt% Y2 O3 and 85-90 wt% ZrO2 . Silica coating is significant and
also fills interstitial areas while in sample 2 and 3 the silica content is found to be < 0.5
wt%. The detection limit of silica is 0.5 wt%. Silica is only found in the product powder
when using normal glass flask for precursor preparation. Using PFA flask no silica is
detected by EDS.
The silica contamination of YSZ product powder may result from the precursor
preparation technique. By the preparation of the yttrium precursor, acid nitric (HNO3 )
is formed which may be responsible that silicon is leached from the glass flask. This
problem is well known in literature. Silicon is only leached from the glass flask preparing
the yttrium precursor while for the zirconia precursor preparation no silica is leached from
the glass flask. The pka of the formed (HNO3 ) is of 5 order of magnitude small than that
of the 2-ethyl hexanoic acid.

133

Table C.2:

EDS analysis collected from YSZ powder prepared accoring Tab.

C.1agglomerations
Sample

SiO2 , wt%

Y2 O2 , wt%

ZrO2 , wt%

3.0

10

87.0

0.5

10.9

88.7

3.1

10.3

86.6

2.4

10.7

86.9

3.0

10.4

86.6

2.4

10.8

86.8

<0.5

17.2

82.7

<0.5

17.1

82.9

<0.5

17.1

82.9

<0.5

20.1

79.9

<0.5

18.7

81.3

<0.5

19.6

80.4

134

APPENDIX

Thermophoretice sampling in spray flames

D.1

Experimental

Zirconia powder is produced by flame spray pyrolysis (FSP) at 1.5 bar gas pressure drop at
the nozzle tip by adjusting the nozzle-gap width as has been described in detail elsewhere
(Fig. 5.1). A constant 1.5 bar gas pressure drop at the burner nozzle tip was maintained by
adjusting the nozzle-gap width. The gas flow rates were controlled by mass flow controllers
(Bronkhorst, Ruurlo, The Netherlands). The liquid precursor solution was fed through the
nozzle by a syringe pump (IER-232, Inotech, Oberwinterthur, Switzerland). Zirconium
n-propoxide (Zr(C3 H7 O)4 , 70% in n-propanol, ZP, ChemPur, Germany) was dissolved in
ethanol (EtOH, >99.8%, Alcosuisse) resulting in a 0.5 M precursor solution fed at 4 that
resulted in ZrO2 production rat of 15. Particle where collected thermophoretically [7] (TSTEM) as described by Kammler et al. [8] (Fig. 5.1). Particle images were obtained by
transmission electron microscopy (HRTEM; Philips Tecnai F30, field emission cathode,300
kV).
135

136

D.2

Results

Figure D.1 shows images from particles thermophoretically collected from flame A (Chap
5) at 0.51 (40 mm HAN), 0.68 (50 mm HAN), 1.19 (75 mm HAN), 1.85 ms (100 mm HAN)
and from the filter. Monitoring the evolution of particle growth in lower stages of the
spray flame was not possible. Particles are between 3 - 20 nm and 6 - 33 nm at 0.51 and
1.85 ms. Particle diameters from the filter are 4 - 16 nm in contrast to the last sampling
position at 1.85 ms. The dV /S (Fig. D.2, circles) from SAXS measurements (Chap. 5) is
always smaller than Sauter mean diamter from particles on the TEM images (Fig. D.2)
taken by thermophoretic sampling. Particle counting show that the Sauter mean particle
diameter increases from 11 - 22 nm with increased HAN (Fig. D.2, squares) compare
to dV /S from SAXS where particle growth stop at about 10 nm (circles). A possible
reason for this difference may by the residence time of the TEM-grid in the spray flame
that is typical 40 ms. During TS the residence time (40 ms, [8]) of the grid in the hot
flame (> 1600 K) is 60 times higher compared to the residence time of 0.68 ms in the
hot temperature zone. Comparing the TEM-images from the filter and from the latest
taken position in-situ in the flame it can be seen that the TS-particles at 1.85 ms are
larger (22 nm) than the filter particles (10 nm) which is in agreement with corresponding
SAXS analysis (Fig. D.2). This may indicate that particle growth does not stop during
thermophoretic sampling. Still droplets are present even at 100 mm HAN who are deposit
on the grid, droplet combustion take place and partially destroyed the grid (Fig. D.1).
Therefore, TS is not applicable as long as droplets still exist in the flame. In diffusion
flames TS can be done as the gas velocity is by a factor of 10 to 20 slower [8, 9] and no
droplets are presented.

137

Figure D.1: TEMs from thermophoretic sampling during spray flame pyrolysis for zirconia nanoparticles production and a destroyed TEM grid.

138

Figure D.2: Comparison particle growth measured by SAXS (Chap. 5) and form TEMimages analysis.

139

Appendix: References
[1] R. Mueller, L. Madler, and S. E. Pratsinis. Nanoparticle synthesis at high production
rates by flame spray pyrolysis. Chem. Eng. Sci., 58(10):19691976, 2003.
[2] R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, and M. K. Akhtar. Zirconia
nanoparticles made in spray flames at high production rates. J. Am. Ceram. Soc,
87(2):197202, 2004.
[3] L. Madler, H. K. Kammler, R. Mueller, and S. E. Pratsinis. Controlled synthesis of
nanostructured particles by flame spray pyrolysis. J Aerosol Sci., 33(2):369389, 2002.
[4] R. Mueller, R. Jossen, H. K. Kammler, S. E. Pratsinis, and M. K. Akhtar. Growth of
zirconia particles made by flame srpay pyrolysis. AIChE J., 50(12):30853094, 2004.
[5] W. Koch and S.K. Friedlander. Particle growth by coalesence and agglomeration.
Part. Part. Syst. Charact., 8(1):8689, 1991.
[6] M. C. Heine and S. E. Pratsinis. Droplet and particle dynamics during flame spray
synthesis of nanoparticles. Ind. Eng. Chem. Res., 44(16):6222, 2005.
[7] R. A. Dobbins and C. M. Megaridis. Morphology of flame-generated soot as determined by thermophoretic sampling. Langmuir, 3:254259, 1987.
[8] H. K. Kammler, R. Jossen, P. W. Morrison, S. E. Pratsinis, and G. Beaucage. The
effect of external electric fields during flame synthesis of titania. Powder Technol.,
135:310320, 2003.
[9] G. Beaucage, H. K. Kammler, R. Mueller, R. Strobel, N. Agashe, S. E. Pratsinis,
and T. Narayanan. Probing the dynamics of nanoparticle growth in a flame using
synchrotron radiation. Nature Mater., 3(6):031401, 2004.

Curriculum Vitae
Rainer Jossen
August 19th 1974 born in Brig (VS), Switzerland
citizen of Naters (VS), Switzerland
09/198107/1987 Primarschule and Orentierungsschule Naters (VS)
08/199007/1995 Kollegium Spiritus Sanctus Brig (VS)
Matura, Typus C
07/199510/1995 Swiss army
10/199510/1999 Swiss Federal Institute of Technology Lausanne (EPFL)
10/199910/2001 Swiss Federal Institute of Technology (ETH)
academic degree: Dipl. Chem. Eng. ETH
12/2001 12.2005 Doctoral Thesis under the supervision of Prof. Dr. S. E. Pratsinis
(Particle Technology Laboratory) at the Swiss Federal Institute of
Technology.

Zurich, 2005

141

Refereed Publications
1. H. K. Kammler, R. Jossen, P. W. Morrison, Jr., S. E. Pratsinis, and G. Beaucage,
The effect of external electric fields during flame synthesis of titania, Powder
Technol. 135-136, 310-320 (2003).
2. R. Mueller, R. Jossen, S. E. Pratsinis, M. Watson, M. K. Akhtar, Zirconia nanoparticles made in spray flames at high production rates, J. Am Ceram. Soc., 87(2),
197-202 (2004).
3. R. Mueller, R. Jossen, H. K. Kammler S. E. Pratsinis, M. K. Akhtar, Growth
of zirconia particles made by flame spray pyrolysis, AIChE J., 50(12), 3085-3094
(2004).
4. R. Jossen, S. E. Pratsinis, W.J. Stark, L. Madler, Criteria for flame-spray synthesis
of hollow, shell-like or inhomogeneous oxides, J. Am Ceram. Soc., 88(6) 1388-1393.
(2005).
5. R. Jossen, R. Mueller, S.E. Pratsinis, M. Watson, K.M. Akhtar, Morphology
and composition of spray-flame-made yttria-stabilized zirconia nanoparticles, Nanotechnology, 16 S609-S617 (2005).
6. A. Teleki, S.E. Pratsinis, K. Wegner, R. Jossen, F. Krumeich, Flame-coating of
titania particles with silica, J. Mater. Res., 20(5) 1336-1347 (2005).
7. H. Schulz, L. Madler, R. Strobel, R. Jossen, S.E. Pratsinis, T. Johannessen, Independent control of metal cluster and ceramic particle characteristics during one-step
synthesis on Pt/TiO2 , J. Mater. Res., 20(9) 2568-2577 (2005).
8. M. C. Heine, L. Madler, R. Jossen and S. E. Pratsinis Direct measurement of
entrainment in flame sprays during nanoparticle synthesis, Combust. Flame, (2005)
in press.
9. R. Jossen, Heine, M.C., Pratsinis, S.E. and K. M. Akhtar, Thermal stability of
flame-made zirconia-based mixed oxides, CVD Journal (2006) accepted.

Presentations
1. R. Jossen, S. E. Pratsinis, W.J. Stark, L. Madler, Morphology, composition and
thermal stability of flame-made zirconia-base mixed oxides, (Poster) AIChE, Cincinnati, Ohio, USA, November, 2005.

142
2. R. Jossen, M. C. Heine, G. Beaucage, T. Narayanan, S. E. Pratsinis, Non-intrusive
dynamics during spray flame synthesis of ZrO2 , (Presentation) Millennium Inorganic Chemicals Inc., Baltimore, MD, USA, October, 2005.
3. R. Jossen, S. E. Pratsinis, W. J. Stark, L. Madler, Criteria for flame-spray synthesis
of hollow, shell-like or inhomogeneous oxides, (Presentation) University of Florida.,
Gainesville, FL, USA, January, 2005.
4. R. Jossen, M. C. Heine, S. E. Pratsinis, Synthesis of zirconia and titania based
materials by flame spray pyrolysis, (Presentation) Millennium Inorganic Chemicals
Inc., Baltimore, MD, USA, November, 2004.
5. R. Jossen, S. E. Pratsinis, W.J. Stark, L. Madler, Criteria for flame-spray synthesis
of hollow, shell-like or inhomogeneous oxides, (Presentation) AIChE, Austin, TX,
USA, November, 2004.
6. R. Jossen, R. Mueller, S.E. Pratsinis, M. Watson, K.M Akhtar, Morphology and
composition of spray-flame-made yttria-stabilized zirconia nanoparticles, (Presentation) AIChE, Austin, TX, USA, November, 2004.
7. A. Teleki, S.E. Pratsinis, K. Wegner, R. Jossen, F. Krumeich, Flame-coating of
titania particles with silica, (Poster) AIChE, Austin, TX, USA, November, 2004.
8. R. Jossen, S.E. Pratsinis, W. J. Stark, L. Madler, Criteria for flame-spray synthesis
of hollow, shell-like or inhomogeneous oxides, (Presentation) 19th ILASS Europe,
Nottingham, UK, September, 2004.
9. R. Jossen, S. E. Pratsinis, W.J. Stark, L. Madler, Criteria for flame-spray synthesis
of hollow, shell-like or inhomogeneous oxides, (Poster) PARTEC2004 conference,
Nuremberg, Germany , March, 2004.
10. R. Jossen, S. E. Pratsinis, W.J. Stark, L. Madler, Criteria for flame-spray synthesis of hollow, shell-like or inhomogeneous oxides, (Presentation) GVC conference,
Mannheim, Germany , March, 2003.
11. H. K. Kammler, R. Jossen, P.W. Morrision, JR, S.E. Partsinis, and G. Beaucage,
The effect of external electirc fields during flame synthesis of titania, AIChE,
Indianaplolis, IN, USA, November 2002

You might also like