Powder Technology 161 (2006) 169 174

www.elsevier.com/locate/powtec

Design principles for spray-roasted iron oxides for the

manufacturing of ferrites
Vassilios Zaspalis a,*, Marcel Kolenbrander b
a

Laboratory of Inorganic Materials, Chemical Process Engineering Research Institute, Center for Research and Development-Hellas,
P.O. Box 57001, Thessaloniki, Greece
b
Ferroxcube GmbH, Materials and Process Development, Essener Str. 4, 22419 Hamburg, Germany
Received 15 June 2004; received in revised form 24 August 2005
Available online 19 December 2005

Abstract
A range of high purity iron oxides are prepared by varying basic operation parameters of an industrial spray roasting process. These iron
oxides are investigated in relation to their morphology and subsequently evaluated as raw materials for MnZn ferrite preparation. It appears
that the most important morphological parameters for determining the reactivity (defined as firing shrinkage at equal compaction density) of
the high purity iron oxide, and consequently the final density and magnetic properties of the ferrite specimens, are the primary particle size
and the number of primary particles per aggregate. As found, the specific surface area of the iron oxide is of no predictive value for the
behavior of the iron oxide in a MnZn ferrite manufacturing process. A small primary particle size is important for a high reactivity;
however, when particles are packed together in large aggregates, they are not available for the prefiring reactions. As a result, reactive
sintering takes place leading to high porosity and bad microstructure. As found by the characterization methods employed in this article, the
optimum iron oxides for MnZn ferrite preparation should have a primary particle size between 0.45 and 0.55 Am with an aggregate size
below 1.60 Am.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Iron oxides; MnZn ferrite; Spray roasting

1. Introduction
In order to obtain either low magnetic power losses or
high magnetic permeabilities in MnZn ferrites for low or
high frequency applications, the iron oxide used for the
preparation should be of high chemical purity [1,2]. The
manufacturers of iron oxides have therefore increased their
efforts towards manufacturing of high purity material, very
often with remarkable success [3,4]. A process which is
widely used for the production of high purity iron oxide is
the so-called spray roasting or Ruthner process [5]. According to this method the iron source (usually steel industry side
or waste products) is diluted in low pH solutions followed
by pH-controlled precipitation of the impurities. The final
solution is spray-roasted in air at temperatures of 300 600

* Corresponding author. Tel.: +30 2310 498 114/115; fax: +30 2310 498 131.
E-mail address: [email protected] (V. Zaspalis).
0032-5910/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2005.10.004

-C resulting to the production of solid iron oxide (Fe2O3)

powder.
The impurity ions present in small amounts in the iron oxide
(e.g. Si, Ca, P etc.) had, however, positive effect on its
reactivity and consequently on the densification of the ferrites.
The most common mechanism was the formation of eutectic
mixtures during sintering [6,7]. Their removal causes a
reactivity loss that counteracts the advantages originating from
the high purity character.
In order to cope with this reactivity loss the iron oxide
should be morphologically upgraded. By means of a
morphology design, the negative effects introduced by the
absence of the impurities could be restored. In this work iron
oxides of different morphologies are produced by varying
basic operation parameters of the spray roasting process.
Subsequently, they are investigated as raw materials on the
reactivity and properties of MnZn ferrites. The results allow
the extraction of important conclusions concerning the
required morphological design of the spray-roasted iron oxide

170

V. Zaspalis, M. Kolenbrander / Powder Technology 161 (2006) 169 174

Table 1
Concentrations of most important contaminants in the spray-roasted iron oxide
samples, used as raw materials for ferrite preparation
Compound/Element

Concentration (wt.%)

SiO2
P2O5
V2O5
PbO
Na2O
K2O
SO3
Cl

0.008 0.010
0.002 0.003
0.001
0.001
0.001 0.002
0.001
0.001 0.005
0.06 0.1

between agglomerates and aggregates that take place during

compaction, the equivalent particle size, in general, depends
on (i.e. decreases with) the compaction force up to a force
above which a constant value is obtained since only
compaction and no fragmentation takes place. This constant
value is called the final equivalent particle size (in the
following to be abbreviated as fESP).
Scanning Electron Microscopy (SEM) is used to obtain
visual morphological impressions of the iron oxides. The spinel
to hematite ratio after prefiring is estimated with X-ray
diffraction (XRD) using a Siemens D-500 Kristalloflex X-ray
powder diffractometer with CuKa-radiation. The basic reflections used for the spinel and hematite structure were those at 2h
angles of 41- and 41.59- respectively.
MnZn ferrites having the chemical formula (Mn0.79Zn0.17
3+
Fe2+
0.04)Fe2 O4 are prepared by the mixed oxide method
according to which a mixture of weighted proportions of the
raw materials (Fe2O3: 70 wt.%, MnO: 24 wt.%, ZnO: 6 wt.%)
is dry mixed prefired to 800 -C and milled in ball-mills. The
milled powder is subsequently roll-granulated with 5 wt.%
polivinylalcohol binder and compacted towards toroidally
shaped specimens at pressures of 500 kg/cm2, resulting to a
compact density of 2.85 g cm 3. The final firing took place at a
peak temperature of 1300 -C with subsequent cooling under
equilibrium partial pressures of oxygen. Details of the ferrite
synthesis process are published elsewhere [9].
The quality of the iron oxide for ferrite preparation is
judged using the linear firing shrinkage (i.e. final specimen
density) and the magnetic power losses at a frequency of 100
kHz, a magnetic induction level of 200 mT and a temperature
of 100 -C; this is also the temperature where the power losses
exhibit their minimum value. The specimen density is
determined by the Archimedes method using water as the
immersion liquid. The linear firing shrinkage is determined
geometrically by measuring the dimensions of the specimen
before and after firing. The electromagnetic properties are
measured with an impedance-phase gain analyzer system
(Agilent 4284A LCR meter) equipped with oscilloscope
(Tektronix TDS 714L), power amplifiers (AG Series Amplifier
and Crown micro-tech 600) and frequency generator (Agilent
33120A), on specimens winded with copper wires, in order to

for the manufacturing of MnZn ferrites. Although all the

experiments reported in this work refer to spray-roasted high
purity iron oxides, it is believed that the results are of general
value not necessarily restricted to a particular iron oxide
manufacturing method.
2. Experimental
Ten morphologically different iron oxides are produced in
an industrial spray roasting process by affecting basic unit
operation process parameters like precipitation pH or spray
roasting temperature. Product chemical analysis is performed
with ICP-emission spectroscopy using a Perkin Elmer-40
spectrometer equipped with D-6000-AT + Cetac ultrasonic
nebulizer. The BET area is determined by nitrogen absorption
using a Quanta-Chrome Autosorb-1 analyzer. Particle size
analysis is performed on identically ultrasonically treated
aqueous dispersions of the oxides by laser scattering, using a
Malvern Mastersizer equipment. Equivalent spherical particle
size measurements are performed with a gas permeability
method developed in the laboratory. Detailed description of
the theoretical background and the experimental setup is
published elsewhere [8]. According to this method, the
resistance to the flow of a gas through a compact made of
the investigated powder is determined as a function of the
compaction force (i.e. compact density). Through basic
principles of gas flow dynamics through porous media, this
resistance can be converted to a so-called average equivalent
particle size. Because of the various fragmentation processes

19,5

3,6

19,4

3,4

19,3

S4

(%) firing shrinkage

BET surface area (m /gr)

4
3,8

3,2
3
2,8
2,6
2,4

S5
S10

19,2

S3

19,1

S1

19

S2

S8
S7

18,9

S9

18,8

2,2

S6

18,7

2
0

Iron oxide sample number

10

11

18,6
2,1

2,3

2,5

2,7

2,9

3,1

3,3

3,5
2

3,7

3,9

4,1

-1

BET surface area m g

Fig. 1. BET area of the iron oxide samples (left), and ferrite firing shrinkage as a function of the BET area (right; the numbers indicate the iron oxide sample number).

V. Zaspalis, M. Kolenbrander / Powder Technology 161 (2006) 169 174

fESP

2,6

0,8

2,4

0,7

2,2

0,6

2
0,5

1,8

fESP (m)

d50 (m)

average size(d50)
d50

0,4

1,6
1,4

0,3
0

10

11

Iron oxide sample number

Fig. 2. Average sizes d 50 and fESP of the iron oxide samples.

form inductors, from the alternating voltage across the

winding and the in-phase component of the current before
and after the specimen.
3. Results and discussion
The concentrations of typical contaminants in all iron oxide
samples are shown in Table 1. As seen, values are such that no
chemically induced differences in the reactivity can be present.
The somewhat high concentration of Cl ions is found to be
significantly reduced (down to 10 3 wt.%) after prefiring and
is therefore considered not to be of substantial importance. In
Fig. 1 (left) the BET surface area of the iron oxide samples is
shown while in Fig. 1 (right) the ferrite firing shrinkage is
shown as a function of the BET area of the iron oxide. The
expected relationship, the higher the BET area the higher the
shrinkage, is not found. As shown in Fig. 1 low shrinkages (i.e.
densities) are obtainable either at low (e.g. sample 1, Fig. 1) or
at high (e.g. sample 6, Fig. 1) BET values. As it will become
apparent in the following text the reason is the specific
relationship between iron oxide morphology and prefiring.
A powder is considered to consist of three elementary units
namely I) agglomerates, II) aggregates and III) primary
crystallites [10,11]. At a preliminary stage of the investigations,
it has been found that agglomerates that are weakly connected
groups of particles and aggregates (usually held together by
van der Waals or capillary forces) are being already destroyed
during the dry mixing process and thus play no role in the
subsequent process steps. The investigations are therefore
focused on the coexistence and the interactions of primary
crystallites and aggregates (i.e. groups of primary particles held

7m

171

strongly together) and their influence in determining the

reactivity of the iron oxide.
In Fig. 2 the average particle size determined by laser
scattering (noted as d 50) and the fESP are shown for all iron
oxide samples. Since the oxides are initially dispersed in water
and subsequently subjected to ultrasonic treatment, the d 50 is
considered to represent the size of the aggregates; the size of
those units that are able to resist the forces exerted by the
sample preparation method, which are the aggregates. On the
other hand, the fESP value, which corresponds to the particle
size determined by gas flow dynamic principles, that is
independent of the compaction force, is considered to represent
the size of the primary crystallites. Those are the only powder
elements able to resist the very high compaction forces (ca.
4000 Nt). Two basic groups of iron oxide morphologies can be
recognized from the data shown in Fig. 2. Samples 1 to 5
belong to group A and are characterized by relatively small
aggregates and relatively large primary crystallites. In contrast,
samples 6 to 10 belong to group B and are characterized by
relatively large aggregate size and relatively small crystallite
size. An idealized schematic presentation of the modeled
morphological differences of the spray-roasted iron oxides is
shown on top of Fig. 2. The number of primary crystallites per
aggregate is significantly different between the two groups and
can be roughly estimated to be proportional to the figure of
merit (d 50/fESP)3. For example, the data shown in Fig. 2
indicate that there is a difference of a factor two in the previous
figure of merit between sample 5 ([d 50/fESP]3 32) and sample
6 ([d 50/fESP]3 64).
Representative SEM pictures of iron oxide samples 1, 3 and
6 are shown in Fig. 3. As can be observed there is a tendency
towards decreasing primary crystallite size, in agreement with
the results shown in Fig. 2. The average primary crystallite
sizes detected from gas flow permeability measurements are in
good quantitative agreement with those estimated from the
SEM images. However, no information about the aggregate
size can be extracted from the SEM images since the provided
information is of no statistically representative value. It can be
visually recognized from Fig. 3 that the particles of sample 6
are held together stronger than those of samples 1 and 3 with
enhanced neck formation areas. This may be taken as an
indication that the average aggregate size of sample 6 is larger
than those of samples 1 and 3, in qualitative agreement with the
earlier described morphological model.
In Fig. 4 (left) the spinel to hematite peak area ratio,
determined by XRD, is shown as a function of the fESP, for

7m

7m

Fig. 3. Scanning microscope pictures of iron oxide samples 1 (left), 3 (middle) and 6 (right).

Spinel/Hematite ratio after prefiring

V. Zaspalis, M. Kolenbrander / Powder Technology 161 (2006) 169 174

Spinel/Hematite ratio after prefiring

172

1,1

S1
1

S3
S6
S7

0,9

S2
S4
S5

S9
S8

S10

0,8
0,4

0,5

0,6

0,7

1,1

S1
1

S2
S6

S4
0,9

S5

S7

S9
S10

S3
S8
0,8
0

0,8

20

40

60

80

100

120

140

Crystallite size (fESP, m)

(d50 /fESP)

Fig. 4. The spinel to hematite peak area ratio of the prefired powders as a function: of the primary crystallite size of the iron oxide (left) and of the factor (d 50/fESP)3
indicative of the number of particles per aggregate in the iron oxide (right).

MnZn ferrite mixtures incorporating the different iron oxide

samples, after prefiring at 800 -C. The highest reaction
progress towards the formation of the spinel structure is shown
by sample 1 which, as mentioned previously, is a sample
among those with large primary crystallite size. In accordance
to that, the lowest reaction progress is shown by sample 8,
which is a sample among those with the smallest primary
crystallite size. Although this result does not seem logical, it
can be explained on basis of the morphological considerations
of the iron oxides. The larger primary particles are packed in
relatively small aggregates (Fig. 2) and are accessible to
Mn3O4 and ZnO for reaction in the green mixture. On the
contrary, the very small particles are packed in large
aggregates; consequently a large part of them is not available
for reaction towards spinel formation. By taking the number of
particles per aggregate (d 50/fESP)3, calculated from the data
shown in Fig. 2, as a measure of the free available Fe2O3
particle reaction interfaces, the reaction progress results
become as expected (Fig. 4, right) and the irregularity is
restored.
In Fig. 5 the magnetic power losses of the sintered ferrites
and in Fig. 6 the final linear shrinkage during firing, are
shown as a function of the iron oxide raw material used. The
same basic groups A and B can be recognized in both figures;
in group A (experiments 1 5) there is a systematic dependency of the power losses on the morphology of the iron oxide
used. By decreasing the primary particle size at more or less
constant aggregate size (Fig. 2) the powder reactivity is
improved. This is reflected to higher firing shrinkage (thus

higher final density) and consequently lower power losses.

The development of the microstructure proceeds smoothly
since the higher homogeneity in the green mixture results to
higher chemical homogeneity in the prefired mixture (Fig. 4).
In group B (experiments 6 10) the magnetic power losses
remain at a relatively high level and do not exhibit a clear
trend on the iron oxide morphology. The shrinkage, although
lower than this corresponding to group A, shows a systematic
dependency on the morphology of the iron oxide. The
following explanation is put forward: in this group, the iron
oxides are characterized by an increasing aggregate size at a
more or less constant primary particle size. Unlike group A,
the aggregate size determines now the extent to which
prefiring has taken place. Prefiring in experiments 6 and 7
for example, is expected to have proceeded more than in
experiments 8 and 9. Consequently the shrinkage observed in
experiments 6 and 7 is lower than of experiments 8 and 9. The
high power loss level is related to the bad microstructure
development. As shown in Fig. 4, the samples of this group
exhibit a small reaction progress during prefiring, which is
becoming less as the aggregate size increases. As a consequence the uncompleted chemical reactions during prefiring
must take place during sintering which is becoming reactive.
Reactive sintering in ferrites produces pore growth that results
to limited densification and grain growth [12,13]. In Fig. 7
SEM microscope pictures are shown of cross sections of fired
MnZn ferrite specimens representative of the two previously
described groups. The existence of large pores is clearly
distinguishable in the microstructures of specimens of group
19,5

Firing shrinkage (%)

Power losses (mW/cm )

340

330

320

310

100kHz, 200mT, 100 C

300

19,3
19,1
18,9
18,7
18,5

10

11

Iron oxide sample number

Fig. 5. Magnetic power losses of MnZn ferrite specimens as a function of the

iron oxide raw material.

10

11

Iron oxide sample number

Fig. 6. Linear firing shrinkage of MnZn ferrite specimens as a function of the

iron oxide raw material.

V. Zaspalis, M. Kolenbrander / Powder Technology 161 (2006) 169 174

200 m

173

200 m

Fig. 7. SEM microscope images of microstructures of MnZn ferrites made with iron oxide sample number 5 (left; representative of all microstructures received from
iron oxides of group A) and number 8 (right; representative of all microstructures received from iron oxides of group B).

B. In Fig. 8 the average grain sizes of the fired specimens are

shown. As seen, experiments of group B have a lower average
grain size than experiments of group A. This is also a result of
the reactive sintering that limits densification and thus grain
growth.
In the previous paragraphs in the reactivity, that has been
defined as linear firing shrinkage (or final density), difference
between experiments of group A and B is explained on the
basis of the differences in the chemical homogeneity during
prefiring which, in turn, is a result of the different morphology
of the iron oxide raw material. Some thoughts are also
possible however towards explaining the reactivity differences
within each group. In Fig. 9 the characteristic particle sizes of
the prefired and milled powders are shown. In the experiments
of group A, the iron oxides with the smallest primary
crystallite size result also to the smallest particle size after
milling. This explains the increasing reactivity, within this
group, when going from iron oxide 1 to 5, which is also
associated with lower magnetic losses (Fig. 5). In group B
there is no trend in the d 90 particle size after milling, which is
in agreement with the more or less constant particle size of
those iron oxides (Fig. 2). Prefiring is less effective for the
larger aggregates and more reactivity is preserved for the final
firing process. Reactive sintering dominates the microstructure
development.

4. Summary and conclusions

By varying main operational parameters of the spray
roasting process a wide spectrum of morphologically different
iron oxides can be produced.
The reactivity of spray-roasted iron oxides for the manufacturing of ferrites (and the associated magnetic properties)
can be predicted by the combinative evaluation of two
morphological parameters, namely the primary particle size
and the primary particles per aggregate.
As is evident from the results of this study for optimum final
densities and magnetic properties the spray-roasted iron oxides
should have a primary particle size between 0.45 and 0.55 Am
and an aggregate size not larger than 1.60 Am (as those two
parameters are defined and measured in this article). Large
primary particles (i.e. > 0.55 Am) result in poor reactivity, low
final density and high magnetic losses. Small primary particles
(i.e. < 0.55 Am) result in high reactivity. If simultaneously
aggregates are small (i.e.< 1.60 Am) the prefired powder has
good homogeneity resulting therefore to high sinter density,
smooth microstructure development and good magnetic properties. If the aggregates are large, thus many particles per
aggregate, the prefired mixture will be inhomogeneous.
Reactive sintering will take place during firing having as a
d50

d90

15

1,9

4,9

1,8

4,8

1,7

4,7

1,6

4,6

13

12

1,5

11
0

10

11

Iron oxide sample number

Fig. 8. Average grain size (the lines indicate the average value of each series of
five experiments) as a function of the iron oxide raw material used.

d90 (m)

14

d50 (m)

Average grain size (m)

16

4,5
0

10

11

Iron oxide sample number

Fig. 9. Characteristic particle sizes after prefiring and milling of MnZn ferrite
powders as a function of the iron oxide raw material.

174

V. Zaspalis, M. Kolenbrander / Powder Technology 161 (2006) 169 174

result microstructures with large pores and limited density or

grain size.
The classical BET measurements as a predictive tool for the
reactivity of spray-roasted iron oxides in ferrite preparation
have a rather limited value. In the experiments reported in this
paper a relation between final firing shrinkage and BET is
absent. It is believed that such a relation is being destroyed by
the interference of the intermediate prefiring step.
Acknowledgements
Dr. A. Essing and dr. A. Rusch from Thyssen Krupp Stahl in
Dortmund (Germany) are acknowledged for kindly performing
the various spray roasting experiments that generated the iron
oxide samples used in this study.
References
[1] A. Goldman, Handbook of modern ferromagnetic materials, Ferrite
Processing, Kluwer Academic publishers, USA, 1999, pp. 305 308.
[2] M.F. Yan, D.W. Johnson Jr., Impurity induced exaggerated grain growth in
MnZn ferrites, J. Am. Ceram. Soc. 61 (7 8) (1978) 342 349.
[3] A. Chiba, O. Kimura, Production of highly pure iron oxide using a solvent
extraction method, Proceedings of the 5th Intern. Conf. On Ferrites (ICF5), India, 1989, pp. 55 60.

[4] N. Hirai, T. Murase, K. Okutani, T. Mori, Method for producing

powdered iron oxide, U.S. Patent 5032367, Chemirite Ltd. and TDK
Corporation, 1991.
[5] A. Rush, Iron oxides for hard and soft ferrites, Proc. of the Gorham/Intertech Conf. on Iron Oxides for Hard and Soft Magnets, Berlin-Germany,
2000, pp. 47 52.
[6] K. Hirota, O. Inoue, Sodium-doped MnZn ferrites; microstructure and
properties, J. Am. Ceram. Soc. 66 (12) (1987) 1755 1788.
[7] C.S. Liu, C.J. Chen, M.J. Tung, C.D. Wu, Effect of PbO B2O3 SiO2
glass on preparation of low loss MnZn ferrites, MRL Bull. Res. Dev. 5
(2) (1991) 23 26.
[8] V.T. Zaspalis, E. Kikkinides, M. Kolenbrander, R. Mauczok, Method for
the morphological characterization of powder raw materials for the
manufacturing of ceramics, J. Mater. Process. Technol. 142 (2003)
267 274.
[9] V.T. Zaspalis, E. Antoniadis, E. Papazoglou, V. Tsakaloudi, L. Nalbantian,
C.A. Sikalidis, The effect of Nb2O5 dopant on the structural and magnetic
properties of MnZn ferrites, J. Magn. Magn. Mater. 250 (2002) 98 109.
[10] W.F.M. Groot Zevert, A.J.A. Winnubst, A.J. Burggraaf, J. Mater. Sci. 25
(1990) 3449 3455.
[11] M.A.C.G.v.d. Graaf, J.H.H. ter Maat, A.J. Burggraaf, J. Mater. Sci. 408
(20) (1985) 1407.
[12] F. Toolenaar, M. Verhees, Reactive sintering of zinc ferrite, J. Mater. Sci.
23 (1988) 856 861.
[13] T. Yamaguchi, T. Nomura, Characterization and sintering of ferrite raw
material powders, Proceedings of the Intern. Conf. on Ferrites Japan,
1980, pp. 46 49.