Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of the European Ceramic Society

journal homepage: www.elsevier.com/locate/jeurceramsoc

Original Article

The detailed microstructure of an alumina-zirconia-silica (AZS) fused cast

refractory material from the cast skin into the bulk analyzed using EBSD
⁎
Wolfgang Wisniewskia, , Christian Thiemeb, Christian Rüssela
a
Otto-Schott-Institut, Jena University, Fraunhoferstr. 6, 07743, Jena, Germany
b
Fraunhofer Institute for Microstructure of Materials and Systems IMWS, Walter-Huelse-Strasse 1, 06120, Halle, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: The interface of an industrially-produced block of an alumina-zirconia-silica (AZS) refractory material with the
ASZ refractories mold is extracted and prepared for analysis. The first 24 mm of crystal growth including the cast skin are
EBSD analyzed using X-ray diffraction (XRD) and scanning electron microscopy (SEM) including energy-dispersive X-
Crystallization hierarchy ray spectrometry (EDXS) and electron backscatter diffraction (EBSD).
Oriented growth
Apart from the bulk, two chemically different zones of crystallization occur near the interface to the mold.
Epitaxial
The growth model for this material begins with the primary crystallization of dendritic ZrO2 followed by sec-
ondary crystallization procedures in the respective crystallization zones. The layer adjacent to the mold crys-
tallizes to form alumina, secondary ZrO2 and mullite. An 001-oriented layer of mullite with the c-axis oriented
perpendicular to the mold material is formed in the second layer of crystallization while only alumina, zirconia
and residual glass occur further in the bulk.

1. Introduction
For a long time, fused cast refractories have been the state of the art
materials for high temperature applications in strongly corrosive con-
ditions such as glass melting furnaces where the materials are used in
direct contact with the glass melt. The first generation of these mate-
rials contains high amounts of around 50 wt% Al2O3 [1–7] while the
second generation primarily contains ZrO2 [5–15]. The main advantage
of fused cast refractories is their high thermal stability and corrosion
resistance which enable them to withstand the extreme conditions in a
glass tank for more than 10 years in the industrial practice. These fused
cast materials are produced via electric arc melting, subsequent casting
into molds and a long term thermal treatment. The cast block can weigh
up to 1000 kg, is lifted from the mold after a relatively short time and
transferred to a furnace which enables to control the further cooling to
room temperature which may take up to four weeks. Quite often these
blocks contain cracks and blowholes which dramatically decrease their
performance. A detailed study concerning the formation of such defects
[1] and the thermal stresses arising during such a production process
was published in 1994 [2].
The first generation is denoted as “alumina-zirconia-silica (AZS)
refractories” and generally shows wt% compositions of (43–50) Al2O3-
(33–42) ZrO2-(13–20) SiO2 [1,3,5–7] but can also contain small
amounts of e.g. B2O3 [1], Na2O [1,4], Fe2O3 [1] or TiO2 [1]. Al2O3 is
proposed to play a significant role during the corrosion of the AZS-
materials by glass melts: Al2O3 dissolved from the refractory material
forms an Al2O3-enriched interface layer between the melt and the re-
fractory material, enhancing the corrosion resistance [4]. Low amounts
of Al2O3 prevent this mechanism so that e.g. ZrO2 contents above 50 wt
% were reported to reduce the corrosion resistance in such environ-
ments [4]. Apart from alumina and ZrO2, mullite has been reported to
crystallize in the AZS-refractories [3,4]. It has been noted that the mi-
crostructure in AZS-refractories containing 33 wt% ZrO2 varies between
different locations in a block [3].
The necessity of producing large amounts of high-purity glasses, e.g.
used in LCD devices, led to the development of the second generation
denoted as “high zirconia (HZ) refractories” containing 83–94 wt%
ZrO2 [5–15]. The 3D-microstructure of the HZ-refractories has been
visualized by X-ray tomography [7,9]. They are primarily composed of
large ZrO2-dendrites [7–14] with residual glass filling the interdendritic
spaces. The long range connectivity of the crystal lattices leads to a high
creep resistance well above the glass transition temperature of the
amorphous matrix filling the interdendritic spaces [9]. Nevertheless,
the stresses in these materials also induce cracking, even after they have
been cooled to room temperature [10].
The Young’s moduli of an AZS and several HZ refractories have been
analyzed in the temperature range from 20 to 1550 °C [6,11,13] and
work concerning the thermo mechanical behavior of these materials has

⁎
Corresponding author.
E-mail address: wolfgang.w@uni-jena.de (W. Wisniewski).

https://doi.org/10.1016/j.jeurceramsoc.2019.01.051
Received 3 December 2018; Received in revised form 30 January 2019; Accepted 30 January 2019
0955-2219/ © 2019 Elsevier Ltd. All rights reserved.

Please cite this article as: Wisniewski, W., Journal of the European Ceramic Society, https://doi.org/10.1016/j.jeurceramsoc.2019.01.051
W. Wisniewski et al.
been presented [7,13].
As ZrO2 plays a major role in these materials, it should be re-
membered that the cubic modification of pure ZrO2 (c-ZrO2) is ther-
modynamically stable at high temperatures but transforms to tetragonal
ZrO2 (t-ZrO2) below 2300 °C which in turn transforms to monoclinic
ZrO2 (m-ZrO2) below 1150 °C. The t-ZrO2 to m-ZrO2 transition is not
displacive and runs parallel to a change of the Zr coordination number
from eight to seven. The high temperature phases can be stabilized
down to room temperature by the incorporation of oxides of divalent
ions (e.g. Ca2+), [16] trivalent ions (e.g. Sc3+, Y3+), [17,18] rare earth
ions [17,18] or by crystal sizes considerably smaller than 30–100 nm
[19,20]. Increasing quantities of di- or trivalent cation oxides with
appropriate ionic radii increase the size below which this transition to
the monoclinic phase occurs. In the context of AZS and HZ refractory
materials, the martensitic phase transition to m-ZrO2 [12–14,21] and its
effect on micro cracking in the ZrO2-dendrites has been studied in detail
[12]. CaO [8] or Y2O3 [12] have been used to prevent this phase
transition to m-ZrO2 in HZ-refractories. While a crystallographic texture
of the parent cubic ZrO2-dendrites in relation to the thermal gradient
during cooling was initially not detected, [14] subsequent analysis by
neutron scattering enabled to measure a much larger number of den-
drites and showed an {001} fiber texture related to the thermal gradient
for m-ZrO2 near the surface (periphery) of the produced HZ-blocks
[15]. It was proposed that the driving force for this texture is stress
minimization during the phase transition to m-ZrO2 [15]. A similar
stress driven texture formation has been proposed to occur during the
surface crystallization of fused silica [22] where the transition from
cubic high-cristobalite to tetragonal low-cristobalite is also accom-
panied by a strong volume effect.
Although the microstructures of these fused cast materials show
many parallels to glass-ceramic materials, they are produced by crys-
tallizing the melt during initial cooling, i.e. the primary nucleation
occurs at high temperatures where the crystal growth velocity is com-
parably high. As a consequence, some of the crystals grow to fairly large
sizes. This is in contrast to the classical preparation of glass-ceramics
where the melt is first cooled to room temperature, then re-heated to a
temperature where the nucleation rate is high and finally to a tem-
perature attributed to fairly high crystal growth velocity. However,
applying a strong thermal gradient to a high temperature melt has
enabled to produce a glass-ceramic containing an oriented layer of
Bi4Ti3O12 which should not form via the classical preparation route
[23]. The method of electrochemically induced nucleation also enables
to produce glass-ceramic materials containing e.g. oriented fresnoite
[24] via the crystallization of high temperature melts.
While the microstructure of HZ-refractories has been intensely
analyzed as noted above, [6–15] we found comparably few reports
concerning details of the microstructure in AZS refractories [3,4,6,7].
Although variations of the microstructure within a block have been
noted, [3] a systematic growth model illustrating the hierarchy of phase
formation and crystal growth has not been formulated to the best of our
knowledge.
In this article, the microstructure of an industrially-produced AZS
fused cast refractory containing around 40 ma% ZrO2 is analyzed using
X-ray diffraction (XRD) and scanning electron microscopy (SEM) in-
cluding electron backscatter diffraction (EBSD). A growth model illus-
trating the phase formation and crystal growth near the mold material
is proposed.
2. Experimental section
The analyzed material was extracted from an industrially produced
AZS-block, a segment of which is shown in Fig. 1. During production,
the melt with a temperature between 1950 °C and 2000 °C is poured
into a mold of either water-cooled metal or resin-bound sand, subse-
quently lifted out after a maximum time of 1 h and transferred to a
cooling furnace where the blocks are cooled for times ranging from 2 to
2
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
Fig. 1. Photograph of a segment of the industrially-produced AZS-block with
the discernible cast skin at the bottom.
6 weeks. If the blocks are lifted from the mold too early, they are known
to break and release molten material from their core. After cooling,
these blocks have a cast skin which is removed before application. The
mol% chemical composition ca. 20 mm into the bulk of this block was
measured to be approximately 42 Al2O3∙22 ZrO2∙32 SiO2∙2 Na2O (i.e. ca.
47 Al2O3∙30 ZrO2∙21 SiO2∙2 Na2O ma%) using EDXS. The nominal
composition of the material was 46 Al2O3∙39.5 ZrO2∙13 SiO2∙1.1 Na2O
ma%. Further impurities probably occur due to the raw materials being
of technical quality.
X-ray diffraction (XRD) was performed on planes cut parallel to the
initial surface with CuKα–radiation in a Rigaku MiniFlex 300 dif-
fractometer using the θ-2θ arrangement in the 2θ-range from 10 to 60°.
In order to perform SEM studies, cross sections were manually polished
with abrasive slurries down to a diamond paste of 0.75 μm grain size
and a final finish of at least 30 min using colloidal silica (Logitech Syton
Typ SF1 (pH = 13.3, grain size 32 nm)) was applied. All samples were
contacted with Ag-paste and coated with a thin layer of carbon at about
10−3 Pa to avoid surface charging.
SEM analyses were performed using a Jeol JSM 7001 F SEM
equipped with an EDAX Trident analyzing system containing a
Digiview 3 EBSD-camera. Energy-dispersive X-ray spectrometry (EDXS)
was performed without a standard using acceleration voltages of 10 or
20 kV. EBSD-scans were performed using a voltage of 20 kV and a
current of ca. 2.40 nA. The scans were captured and evaluated using the
software TSL OIM Data Collection 5.31 and TSL OIM Analysis 6.2. All
data sets featured in this manuscript were cleaned using a 90° rotation
around the (001) plane normal of mullite as well as a 180° rotation
around the (101) plane normal of m-ZrO2 before applying the grain
Confidence Index (CI) standardization. Unreliable data points were then
removed by applying a CI filter of 0.1.
3. Results and discussion
A cross section of the interface near region extending ca. 24 mm into
the AZS-fused cast material from the bottom of a block was prepared for
analysis by SEM and EBSD. A photograph of the cross section prepared
for analysis is presented in Fig. 2 where the porous cast skin of the AZS-
material (right, at the interface to the mold material) is slightly yellow
at first, followed by a white layer which becomes increasingly yellow as
the distance from the interface increases. The marked crystallization
zones 1–3 will subsequently be described in detail.
In order to obtain a first insight into the phases and possible textures
occurring in this material, both compact and powdered samples pre-
pared from the interface near region or the bulk material were analyzed
by XRD. The acquired XRD-patterns are presented in Fig. 3 and mainly
show m-ZrO2 (baddeleyite space group P21/c) and trigonal α-Al2O3
(corundum, space group R-3c). The patterns A additionally indicate the
presence of mullite and show a very weak peak at 2Θ≈22° which could
be attributed to SiO2 in the form of cristobalite. Na-containing alumina
compounds (“β-Al2O3”) are not indicated by the presented XRD-pat-
terns. Further peaks highlighted by “?” cannot be attributed to any of
W. Wisniewski et al. Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Fig. 4. EBSD-patterns obtained from the polished cross section with a binning
of 2 × 2 indexable as: a) trigonal Al2O3, b) monoclinic ZrO2 and c) orthor-
hombic mullite.
Fig. 2. Photograph of the prepared cross section (cut perpendicular to the in-
terface with the mold) roughly segmented into the crystallization zones 1–3.
The Ag-paint was added to support the conductivity of the applied C-layer. The Table 1
locations “pore 1” and “Ag tip” are highlighted for later reference. Selected indexing parameters of the EBSD-patterns a), b) and c) in Fig. 4 in-
dexed using an IAT of 3°.
pattern phase Votes Fit [°] CI

a) alumina 6 2.29 0.000

mullite 12 2.09 0.025
monoclinic ZrO2 120 0.28 0.808
cubic ZrO2 5 2.27 0.008
cristobalite 10 1.99 0.008
b) alumina 53 0.93 0.292
mullite 8 2.09 0.008
monoclinic ZrO2 23 1.66 0.117
cubic ZrO2 23 1.89 0.133
cristobalite 10 2.38 0.025
c) alumina 4 2.21 0.000
mullite 40 1.20 0.175
monoclinic ZrO2 10 2.14 0.017
cubic ZrO2 5 2.28 0.008
cristobalite 7 2.29 0.008

Indexing was performed using an interplanar angle tolerance (IAT) of 3°

Fig. 3. XRD-patterns from compact and powdered samples of the cast skin (A)
and in the bulk (B). The reference patterns of m- ZrO2 (JCPDS no. 01-088-
2390), trigonal Al2O3 (JCPDS no. 01-071-1123), SiO2 (low cristobalite, ICSD
no. 34,933) and a mullite phase of the composition Al4.64Si1.36O9.68 (JCPDS no.
01-079-1453) are presented for comparison. Peaks not attributable to any of
these phases are highlighted by “?”.
the reference patterns. They could e.g. be explained by the presence of
silimanite (Al2SiO5), however, none of the crystals subsequently ana-
lyzed by EDXS showed this composition. Hence silimanite or other
phases may only occur in small amounts. A notable difference between
the peak intensities acquired from compact and powdered samples is
not observed, i.e. textures are not indicated to occur in the areas ana-
lyzed by XRD.
After applying the appropriate polishing procedure to the sample
presented in Fig. 2, the high-quality EBSD-patterns presented in Fig. 4
could be acquired from this cross section. Selected indexing parameters
are stated in Table 1. As a rule of thumb, an EBSD-pattern can be
considered as reliably indexed by the applied software if it receives at
least 30 votes, a fit factor of less than 1° and a CI value larger than 0.1.
so as to be closer to the conditions supplied during the subsequently
performed EBSD-scans which were all recorded using a higher binning
of 4 × 4 to optimize the time required for the performed measurements.
Table 1 shows that these patterns may be reliably indexed as a) m-ZrO2,
b) trigonal Al2O3 (alumina) and c) orthorhombic mullite in agreement
with the dominant phases indicated to occur in this sample by the XRD-
measurements presented in Fig. 3. As the XRD-patterns acquired from
the surface also indicated the possible presence of cristobalite, a ma-
terial file of low-cristobalite was additionally supplied during all sub-
sequent EBSD-scans. In order to see whether any cubic/tetragonal ZrO2
could be detected in the material, i.e. stabilized by a very small grain
size, a material file of c-ZrO2 was supplied during all EBSD-scans. A
tetragonal file was not supplied because the lattice distortion of the unit
cell is too small to reliably separate c-ZrO2 and t-ZrO2 in the standard
EBSD- measurements performed here [25].
Chemical compositions measured by EDXS spot measurements at
various locations of interest are stated in Table 2 for later reference.
They are presented here for an easy comparison but will be discussed in
detail later on. It should be noted that the composition of mullite is not
stoichiometric but can be described as Al4+2xSi2-2xO10-x with x = 0.2 to
x = 0.9 which may also include other metal oxides [26].
Fig. 5 features an SEM-micrograph of the microstructure in crys-
tallization zone 1 adjacent to the interface of the AZS-material with the
mold. The superimposed phase + image quality (IQ)-map of an EBSD-
scan performed with a step size of 200 nm shows that the bright crystals
are attributed to m-ZrO2 while the areas of lower material contrast are
either attributed to alumina or do not provide indexable EBSD-patterns,
i.e. are probably composed of residual glass. The circled area contains a
small amount of mullite as well as a single grain indexed as c-ZrO2.
Apart from the large m-ZrO2 crystals in this microstructure, a high
amount of very small growth structures occur.
The large ZrO2 crystals are surrounded by a matrix of lower material
contrast, i.e. alumina or residual glass (black in the phase + IQ-map).

3
W. Wisniewski et al. Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Table 2 of Na detected in these separate pockets of glass indicates that the

Chemical compositions measured via EDXS spot measurements (spot size ˜1 μm composition of the residual glass is not homogeneous throughout the
in diameter) at various locations which will be noted in the subsequently pre- material, i.e. it depends on which phases crystallized from the melt.
sented results. The measurements were performed without a standard so the The inverse pole figure (IPF)+IQ-maps of b) alumina and c) m-ZrO2
margin of error is ca. ± 1 at%. The theoretical benchmark compositions for
are also presented in Fig. 5. They show that the alumina grains in the
mullite with x = 0.2 and x = 0.9 are stated for comparison.
scan do not have any orientation relationship to each other in the
at% Zr Al Si Na O scanned area while the apparent homogeneity of the m-ZrO2 grains in
the SEM-micrographs is misleading: whereas the alumina grains gen-
alumina small 5.1 35.9 0.0 0.1 58.9
mullite x = 0.2 0.0 27.8 10.1 0.0 62.0
erally only show one crystal orientation per grain, the m-ZrO2 grains
mullite x = 0.9 0.0 38.4 1.3 0.0 60.3 show multiple orientation domains which can be attributed to mono-
mullite 1 9.5 27.9 5.1 0.6 56.9 clinic twinning. The occurrence of these twins is in agreement with the
mullite 2 3.7 30.2 6.6 0.6 58.9 literature [21,27]. A method for reconstructing the high-temperature
mullite 3 0.1 38.1 6.8 0.5 54.5
cubic or tetragonal orientations from the monoclinic lattice has been
presented [21]. While only three orientation relations result from the
mol% ZrO2 Al2O3 SiO2 Na2O
cubic-to-tetragonal transformation, six orientation relations can result
res. glass1 0.9 19.9 78.3 0.9 from the tetragonal-to-monoclinic phase transition [21]. As dendrites
res. glass2 1.6 16.5 79.4 2.5 usually grow with a single crystal lattice and the 010-pole figure (PF) of
res. glass3 4.3 35.9 58.7 1.1
the ZrO2 dendrite circled in red indicates only six poles, it is very likely
res. glass4 0.6 12.8 84.5 2.1
that these large ZrO2-dendrites grew in the tetragonal modification.
Otherwise more orientations should appear as up to 3 × 6 = 18 poles
The chemical compositions measured for two such pockets of residual would be expected if the parent dendrite initially grew in the cubic
glass via the EXDS-spot measurements “residual glass 1” and “residual modification. While this is in agreement with the pouring temperature
glass 2” are presented in Table 2. The high amount of almost 80 mol% of less than 2000 °C, it is in contrast to the cubic ZrO2-dendrites formed
SiO2 at these locations is not surprising as SiO2 cannot be incorporated in HZ-refractories [14,21].
into the neighboring alumina or ZrO2 crystals. While the varying The element maps visualizing the distribution of Si and Na in the
amounts of Al and Zr could originate from varying contributions to the framed area in Fig. 5 a) show that both Si and Na are accumulated in
EDXS-signal by neighboring crystals, the significantly different amount the residual glass between the alumina and ZrO2 crystals which is in

Fig. 5. a) Phase + IQ-map of an EBSD-scan performed on the area

superimposed on an SEM-micrograph of the microstructure in
crystallization zone 1. EDXS-spot measurements were performed
at the marked locations residual glass 1 and 2, the results are
stated in Table 2. The 010-PF of the ZrO2-dendrite circled in red is
presented containing poles of both hemispheres. IPF + IQ-maps of
b) alumina and c) m-ZrO2 are also presented. EDXS element maps
visualize the relative distribution of d) Si and e) Na in the framed
area. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article).

4
W. Wisniewski et al.
Fig. 6. SEM-micrograph detailing the crystal circled in Fig. 5 a). The EBSD-patterns 1 and 2 were acquired at the locations 1 and 2. The phase map and EDXS-element
maps of Zr, Na, Al, Si and O extracted from a combined EBSD + EDXS scan performed on the framed area with a step size of 100 nm are also presented.
agreement with the spot measurements described above.
The area containing the particle indexed as c-ZrO2 is presented in
greater detail in Fig. 6. Three substructures are discernible in the pre-
sented SEM-micrograph in the areas attributed to c-ZrO2. In the pre-
sented phase-map, they neither have a bright material contrast com-
parable to m-ZrO2 nor are they enriched in Zr. Instead, the grain
predominantly attributed to mullite mainly contains elevated amounts
of Al and O as well as some Si in agreement with the expected com-
position of mullite. The EBSD-patterns 1 and 2 show many similarities,
but while pattern 1 is reliably indexed as mullite, pattern 2 can be in-
dexed as c-ZrO2. As EBSD alone is only sensitive to the crystallography
causing the diffraction pattern, these domains could e.g. be composed
of a cubic alumosilicate not detected by XRD, see Fig. 3. This result
illustrates that further phases may occur in this material, however, their
minimal amounts and tiny sizes make them difficult to identify and
probably irrelevant to the general properties of this material. Ulti-
mately, the information indicated by the XRD-patterns A in Fig. 3 is in
agreement with the EBSD-results acquired from the crystallization in
zone 1.
5
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
While the Figs. 5 and 6 illustrate the detailed microstructure in
crystallization zone 1, the performed EBSD-scans do not contain enough
grains to perform any representative texture analyses. Fig. 7 features a
larger area adjacent to the initial surface (right). Please note that the
dendritic grains of ZrO2 are smaller near the edge to the interface and
none are observed in the immediate 10–20 μm next to it. This indicates
that a thermal gradient affected crystal growth in this area so that the
dendritic grains had less time to grow than in the rest of zone 1. The
very fine crystallization in this area raises the possibility that a thin
layer of quenched glass (less than 50 μm thick) was formed as the melt
was poured into the mold. Frame 1 highlights the area featured in
Fig. 5a). The superimposed phase + IQ-map of an EBSD-scan performed
with a step size of 750 nm confirms that zone 1 primarily contains m-
ZrO2 and alumina and is roughly 200–400 μm thick before mullite
dominates the phase map, marking the boundary to crystallization zone
2. The EDXS element maps of the area in frame 2 spanning this
boundary are presented below: the boundary is most clearly discernible
in the element maps of Si and Na which are both distributed much more
homogeneously in zone 2 than in zone 1 where they form the discrete
W. Wisniewski et al. Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Fig. 7. a) Phase + IQ-map of an EBSD-scan superimposed on an SEM-micrograph overview of the microstructure adjacent to the interface (right). Areas where Ag-
paint covers the edge are shaded. The area in frame 1 was presented in greater detail in Fig. 5 a). The element maps of Zr, Al, Si, O and Na covering the boundary
between the crystallization zones 1 and 2 in frame 2 are presented below, the color coding is identical to that in Fig. 6. PFs visualizing the orientation distributions of
m-ZrO2, alumina and mullite in the entire scanned area are also presented.

pockets of residual glass discernible in Fig. 5a). Please note that the
small grains of m-ZrO2 fail to be resolved in detail but this is not sur-
prising taking into account their size outlined in Fig. 5c) and the step
size of 750 nm applied to cover such a large area with a reasonable
number of data points.
The grain morphology of m-ZrO2 implies that the parent t-ZrO2
nucleated in the bulk and grew via the dendritic growth mechanism.
The weak, possibly stress-induced {001}-texture detected in HZ-re-
fractories via neutron scattering [15] is neither confirmed in the studied
AZS-material by the XRD-results presented in Fig. 3 nor by the EBSD-
results presented here: the 001-and 111-pole figures (PFs) of m-ZrO2 in
the scanned area presented in Fig. 7 do not indicate a clear orientation
preference. Perhaps the volume of material analyzed by EBSD is simply
too small to obtain results comparable with neutron scattering. After
all, stress-induced cracking has been reported in both the AZS- and HZ-
materials. The 0001-PF of alumina also fails to indicate an overall
orientation preference while the 001-PF of mullite shows that this phase
clearly exhibits an orientation preference within the scanned area.
However, most of the mullite is found more than 200 μm below the
initial surface, i.e. it did not contribute significantly to the XRD-pattern
A (compact) in Fig. 3.
The orientation preference of mullite as well as the distribution of Si
and Na in zone 2 outlined in Fig. 7 justify a more detailed analysis of
the microstructure in this layer of crystallization. Fig. 8a) shows the
detailed microstructure of zone 2 just beyond the border to zone 1
marked by the dashed line. An EBSD-scan was performed on the framed
area and the phase + IQ-map presented in Fig. 8b) shows that, as in
zone 1, the large bright crystals are composed of m-ZrO2. The matrix,
however, is generally attributed to mullite although the SEM-micro-
graph clearly shows that it is composed of nm-scale bright structures
surrounded by a dark matrix. The IPF + IQ-map of the scan presented
in Fig. 8c) confirms the twinning already described for m-ZrO2 but also

6
W. Wisniewski et al. Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Fig. 8. a) SEM-micrograph illustrating the microstructure in crystallization zone 2 just after the boundary to zone 1 illustrated by the dashed line. Locations of the
EDXS-spot measurements mullite 1 and 2 stated in Table 2 are highlighted. An EBSD-scan was performed on the framed area and data from this scan is presented
below: b) phase + IQ-map, c) IPF + IQ-map (the legend for m-ZrO2 is presented in Fig. 5), d) orientation + IQ-map containing only data points attributed to mullite,
e) EDXS element map of Si (same color coding as in Fig. 6) and f) 001-PF of mullite in the scanned area.

shows that only two orientation domains of mullite occur within the
scanned area. The orientation + IQ-map in Fig. 8d) proves that mullite
shows orientation deviations of up to 22° from the main orientation
domain in the scan.
Please note that the described results explain why these areas only
deliver EBSD-patterns indexable as mullite: the bright, nm-scale struc-
tures are most probably m-ZrO2, which, in addition to being very small,
may also show many twinning domains. Hence, the regions of a
homogeneous crystal lattice are probably too small to enable the for-
mation of an EBSD-pattern from this lattice. By contrast, the mullite
crystal lattice shows a relatively homogeneous orientation which en-
ables the formation of EBSD-patterns. Here the maximum information
depth of EBSD [28] which can reach more than 80 nm when using a
voltage of 20 kV [29] should be taken into account.
The element distribution of Si presented in Fig. 8e) shows that, as
with the residual glass described in zone 1, Si is accumulated in the
vicinity of the large m-ZrO2 crystals. However, these areas are generally
attributed to mullite in Fig. 8b). While the spot “mullite 1” is positioned
in an area highly permeated by the bright, fine grained structures, the
spot “mullite 2” is positioned in the dark area close to a large m-ZrO2
crystal. The corresponding chemical compositions stated in Table 2
confirm that both compositions match those possible for mullite with
some Zr included due to the relatively large information volume of
EDXS. The higher amount of Zr detected at the location mullite 1
supports the idea that the bright, fine grained structures are indeed
composed of some ZrO2 modification.
The 001-PF of mullite presented in Fig. 8f) visualizes that mullite
grew with its c-axis parallel to the prepared cut plane and almost per-
pendicular to the interface with the mold. Hence the question arises
whether the mullite in zone 2 forms a textured layer during further
growth into the bulk.
Fig. 9a) features a much larger scan performed further into the bulk,
i.e. ca. 1200 μm above the interface to the mold. The presented 001-PF
of mullite clearly shows an orientation preference with the c-axes
aligned perpendicular to the mold interface. While the orientation map
presented in Fig. 8d) shows orientation changes of 22° over a distance
of ca. 20 μm, the mullite grains in Fig. 9a) show orientation deviations
of less than 10° over distances of 100 μm or more. The greater detail in

7
W. Wisniewski et al.
Fig. 9b) enables to discern that some of the bright structures inside the
mullite grains are indexed as m-ZrO2, supporting the hypothesis that
the bright, fine grained structures discernible in Fig. 8a) are also ZrO2.
Furthermore, the chemical composition measured at the location
“mullite 3” and stated in Table 2 contains only 0.1 at% Zr but a much
larger amount of Si than detected in the areas providing EBSD-patterns
of mullite analyzed so far. Some areas fail to provide EBSD-patterns,
indicating the presence of a residual glass in this microstructure. The
composition measured at the location “residual glass 3” shows a much
higher content of Al2O3 than the residual glass detected in zone 1.
Fig. 9c) features the two large m-ZrO2 grains in greater detail so as to
show that some data points in these grains are again attributed to c-
ZrO2, by the indexing algorithm of the software. As these data points do
not cluster into a grain it is doubtful whether c/t-ZrO2 really occurs
here. Fig. 9d) visualizes the twinned microstructure of these m-ZrO2
grains already outlined above. As the bright, fine grained structures
follow the growth morphology of the mullite domains both in Fig. 8 and
Fig. 9, it is probable that they formed during (or after) the crystal-
lization of mullite.
Fig. 10. Phase + IQ-maps of performed EBSD-scans superimposed on an SEM-micrograph overview of the microstructure change from crystallization zone 2 to zone
3. PFs visualizing the orientation distributions of alumina and mullite in the scanned areas are presented.
8
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
Fig. 9. Crystallization zone 2 ca. 1200 μm
above the initial interface: a) Phase + IQ-map
of an EBSD-scan performed with a step size of
400 nm superimposed on an SEM-micrograph
of the area. The 001-PF of the mullite crystals
is presented. The framed area is presented in
greater detail: b) Phase + IQ-map of an EBSD-
scan performed with a step size of 100 nm su-
perimposed on an SEM-micrograph of the area.
Locations of the EDXS-spot measurements
“mullite 3” and “residual glass3” stated in
Table 2 are highlighted. The framed area is
presented in greater detail in c): phase + IQ-
map where some data points are attributed to
c-ZrO2. d) IPF + IQ-map of m-ZrO2 in the same
area.
Crystallization zone 2 grows to a thickness of ca. 1200 μm before
reaching the border to crystallization zone 3 which is outlined in Fig. 10
and occurs at a distance of ca. 1.5 mm from the initial interface. The
large “pore 1” is discernible and highlighted in Fig. 2 to serve as a
reference point. The phase maps show that zone 3 contains alumina in
addition to the m-ZrO2 observed throughout the material while mullite
is not detected. This is in agreement with the XRD-patterns B in Fig. 3.
While the detected 001-texture of mullite at the end of zone 2 is in
agreement with the results presented above, the 0001-PFs of alumina
do not indicate the random orientation distribution observed in crys-
tallization zone 1, see Fig. 7. Instead, the PFs indicate that multiple
orientation preferences occur in different regions.
As the low magnification in Fig. 10 is insufficient to visualize the
details of the microstructure occurring in crystallization zone 3, it is
presented in greater detail in Fig. 11. The phase + IQ-map of an EBSD-
scan performed with a step size of 300 nm is superimposed on an SEM-
micrograph of the area. The large, bright crystals are again composed of
m-ZrO2 while the large, grey domains attributed to alumina in the
phase maps of Fig. 10 show a microstructure very similar to that
W. Wisniewski et al.
described for crystallization zone 2. However, here the dark matrix is
composed of alumina instead of mullite. Large grains of alumina not
permeated by the bright phase also occur. Additionally, the inability to
acquire EBSD-patterns from large parts of the scanned area again in-
dicate the presence of a residual glassy phase. As the residual glass and
the ZrO2-free alumina grains are difficult to discern in the SEM-mi-
crograph, it must be questioned whether the 17% residual glassy phase
reported to occur in the AZS-material in Ref. [6] is in fact not similarly
composed of crystalline Al2O3 and residual glass. The EDXS-spot results
performed at the highlighted locations confirm that the spot “alumina
small” indicates a local composition containing ≈5 at% Zr. As some of
the larger bright structures are attributed to m-ZrO2 in the phase + IQ-
map, it is probable that the fine growth structures are also composed of
m-ZrO2. Alumina crystals free of ZrO2 as well as those encasing ZrO2
domains have also been reported for an AZS-material of similar com-
position [3]. As significant amounts of Na are only detected in the re-
sidual glass but not in the crystal phases, both types of alumina crystals
are probably composed of the same phase and not e.g. α- and β-alumina
as proposed in Ref. [3].
The clearly different morphology of the ZrO2 grains inside the
9
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
Fig. 11. Phase + IQ-map of an EBSD-scan superimposed on a
detailed SEM-micrograph of the microstructure in zone 3. The
0001-PF of alumina and 111-PF of m-ZrO2 illustrate orienta-
tions within the area framed in white. Locations of the EDXS-
spot measurements “alumina small” and “residual glass 4”
stated in Table 2 are highlighted.
Fig. 12. a) Phase + IQ-map of an EBSD-scan superimposed on
the SEM-micrograph of the transition from zone 2 (right) to
zone 3 (left). b) IPF + IQ-map of the same area featuring only
data points attributed to alumina, the color legend is the same
as presented in Fig. 5 b). Crystal orientations at selected lo-
cations are visualized by the wire frames of alumina unit cells.
alumina domains raises the question whether they show an orientation
relationship with each other or the encasing alumina lattice. The area
framed in Fig. 11 is well suited to address this question as it contains
only one domain of alumina but multiple grains attributed to ZrO2. As
expected, the presented 0001-PF of alumina in this area indicates a
single orientation in the featured area. Similarly, the 111-PF of m-ZrO2
indicates the presence of only two main orientations (each contributing
two poles in opposing quadrants). Furthermore, the orientations circled
in red show a [111]-plane basically parallel to a [0001]-plane of the
alumina matrix, pointing towards an epitaxial relationship between
these grains, similar to those observed for tips of Y2O3-stabilized ZrO2
growing onto ε-Y2Si2O7-dendrites [30] or the continuation of a
Bi4−xNdxTi3O12 lattice in the form of Bi2Si1−yTiyO5 due to a mod-
ification of the melt chemistry during crystal growth [31]. This or-
ientation relationship was detected in all analyzed alumina grains with
ZrO2 inclusions. As the parent alumina grains must have grown before
their ZrO2-inclusions, these grains prove a second step of crystallization
after the primary growth of the large ZrO2 dendrites in this area.
The chemical composition measured for the residual glass in zone 3
contains more than 84% SiO2, hence the entire SiO2 of the initial
W. Wisniewski et al. Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

composition probably occurs as residual glass in zone 3 which re-

presents the bulk of the AZS-block. In agreement with the XRD-patterns
B in Fig. 3, cristobalite was not reliably detected by EBSD in any of the
measurements presented here.
After clarifying the detailed microstructure in zone 3, the non-
random orientations of alumina in zone 3 must be explained. Hence the
area of the border from zone 2 to zone 3 in Fig. 10 is presented in
greater detail in Fig. 12. Fig. 12a) shows that, as everywhere else in this Fig. 14. Photograph of the analyzed sample with the locations of performed
material, the crystals of bright contrast are m-ZrO2. The grey structures EDXS-area measurements 1–3c highlighted by black frames.
in zone 2 (right) are composed of m-ZrO2 permeated mullite while the
darkest structures are solely composed of mullite. The areas of residual
low and its crystal growth velocity is high. It is probable that such long-
glass described in Fig. 9 are comparably rare. The dark structures in
range crystal lattices in the microstructure have a similar effect on the
zone 3 are both alumina and residual glass which have a very similar
creep resistance as the large ZrO2-structures observed in HZ-materials
contrast and are hence difficult to discern at this magnification.
[9]. The increased size of the alumina domains from the border to zone
Fig. 12b) presents the IPF + IQ-map of the data points attributed to
2 towards the bulk, discernible by comparing the scans at the border
alumina in this EBSD-scan. Local orientations at specific locations are
(Figs. 10 and 12), near pore 1 (Fig. 10) and further in the bulk (Fig. 13),
visualized by the wire frames of unit cells. While most of the alumina is
indicate that a slight thermal gradient may affect the crystallization
oriented with the c-axes more or less parallel to the current cut plane,
within the analyzed region. By contrast, the grain size of the large ZrO2
some (red in the IPF + IQ-map) are oriented with their c-axes almost
crystals is comparable throughout the analyzed sample except for the
perpendicular to it. Additionally, all the crystals seem to originate from
first ca. 100 μm adjacent to the mold.
the same region. It may be concluded that this EBSD-scan covers a cut
The color changes in the photograph of Fig. 2 and the complex
plane through an alumina spherulite of several hundred μm in dia-
phase formations including multiple zones of crystallization raise the
meter. Hence this alumina formed via bulk nucleation which explains
question whether the analyzed material shows a homogeneous che-
the lack of an overall texture comparable to that of mullite in zone 2. At
mical composition from the bottom to the maximum distance from the
the same time, the formation of such large scale spherulites indicates
interface analyzed here, i.e. about 24 mm into the bulk. EDXS-area
that the nucleation rate of alumina was much lower here in zone 3 than
measurements covering several hundred μm2 each were performed at
the nucleation rate of alumina in zone 1 where this phase only forms
the locations highlighted in Fig. 14. The chemical compositions calcu-
grains originating from individual nucleation.
lated from the acquired spectra are stated in Table 3.
A significant size increase is discernible between the alumina
The amount of ZrO2 decreases from zone 1 to zone 3 in which it
spherulite at the border to zone 2 and the structures observed near pore
remains constant considering the margin of error. The maximum Al2O3
1. Even larger orientation domains of alumina occur further into the
and minimum SiO2 average concentrations of the three zones are
bulk as illustrated by Fig. 13 which presents the microstructure near the
measured in zone 2. The highest concentration of Na2O is measured in
location “Ag tip” highlighted in Fig. 2. The presented phase + IQ-maps
zone 1, the lowest in zone 2 and then it remains constant throughout
illustrate alumina domains more than 1 mm long while the 0001-PF of
zone 3. The most probable reason for the apparent chemical variations
the data points attributed to alumina confirm that there is no overall
are dynamical conditions when the arc melting process is started, i.e.
texture for alumina in this area. Such huge orientation domains can
inhomogeneous melt compositions are poured into the mold until the
only form at high temperatures where the nucleation rate of alumina is
process reaches stable operating conditions. This assumption is sup-
ported by the thickness of zone 2: it is difficult to imagine that a layer of
≈1 mm formed via diffusion after the crystallization of ZrO2 but before
the subsequent crystallization processes. Furthermore, diffusion leads
to concentration gradients at the boundaries, as e.g. described during
the crystallization of mullite in a melt with another composition, [32]
not the homogeneously different element distribution e.g. implied for
Na at the boundary from zone 1 to zone 2 in Fig. 7. As none of these
compositions match the nominal composition which contains ≈10%
more ZrO2 and ≈10% less SiO2, it must be assumed that further
changes of the composition occur even further into the bulk.
In summary, it may be stated that the observed chemical variations
in addition to the occurring thermal gradient lead to the clearly dis-
cernible cast skin. The pores in the cast skin are relatively few and
rather large, i.e. it is unlikely that they were formed by gasses released
from the melt, e.g. by reboiling. Instead they were probably introduced
to the melt physically when it was poured into the mold. Furthermore,

Table 3
Chemical compositions calculated from the EDXS-area measurements outlined
in Fig. 14 without a standard. The margin of error is ca. ± 1%.
location ZrO2 Al2O3 SiO2 Na2O
[mol%] [mol%] [mol%] [mol%]

1 26.8 44.3 25.2 3.7

Fig. 13. Phase + IQ-maps of EBSD-scans performed with a step size of 3500 nm 2 25.0 50.2 23.2 1.6
superimposed on an SEM-micrograph of the area near the location highlighted 3a 23.2 44.0 30.1 2.7
3b 21.6 41.8 34.0 2.6
as “Ag tip” in Fig. 2. The 0001-PF of the combined alumina data of the scans is
3c 23.7 41.2 32.6 2.5
presented.

10
W. Wisniewski et al.
Fig. 15. Ternary phase diagram Al2O3/ZrO2/SiO2 redrawn based on Ref. [34],
the nominal composition (neglecting Na2O) is marked by the red dot. The red
arrow indicates the shift caused by the subtraction of ZrO2 from the melt via
crystallization (For interpretation of the references to colour in this figure le-
gend, the reader is referred to the web version of this article).
it is acceptable to assume that the entire layer of analyzed material
crystallized during the time the block stayed in the mold, i.e. within
60 min: it is hard to imagine how a layer of less than 30 mm thickness
can withstand the pressure resulting from ≈1000 kg of (partially
crystallized) melt at high temperatures when the block is lifted from the
mold.
With the exception of the ca. 20 μm of material immediately ad-
jacent to the mold, all three described crystallization zones contain
large ZrO2 dendrites. Hence the crystallization of the analyzed AZS-
refractory material began with the primary bulk nucleation of t-ZrO2
throughout the entire analyzed area, independent of any local varia-
tions of the chemical composition within the latter. The reduced size of
these dendrites close to the mold indicates that a thermal gradient af-
fected their growth, i.e. they had less time to grow because the melt was
cooled faster in this area.
Neglecting the Na2O content, the liquidus temperature of the melt is
roughly 1900 °C as illustrated by the red dot in Fig. 15. Removing ZrO2
from this composition via crystallization lowers the liquidus tempera-
ture, possibly as low as 1700 °C (see red arrow in Fig. 15), while the
crystallization enthalpy of ZrO2 is simultaneously released into the
melt. Both effects serve to suppress the nucleation of secondary phases
for some time, allowing the primary ZrO2 dendrites to grow to lengths
of more than 100 μm in all three zones of crystallization except at the
immediate interface to the mold where the melt is quenched to a glassy
state. As a change of the coarse dendritic morphology towards finer
growth is not observed in the primary dendrites, e.g. well documented
during the crystallization of Sr-fresnoite, [33] it is improbable that the
growth velocity changed significantly due to a decreasing temperature
while they were formed. Hence these primary dendrites probably
stopped growing due to the formation of a ZrO2-depleted zone around
them.
Measured values concerning the temperature in the melt after
pouring or during cooling are not available but could clarify this matter
in future experiments. Sadly, the means for acquiring such data are
currently unavailable to the authors.
As the subtraction of ZrO2 can hardly lead to large shifts in the SiO2,
Al2O3 and Na2O concentrations, the chemical differences between the
zones 1–3 and the nominal composition were probably present before
11
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
Fig. 16. Illustration of the probable development of the thermal gradient in the
analyzed crystallization zones 1–3 as time progresses from time 1 to time 4.
Boundary conditions set by the casting temperature Tcast, primary and sec-
condary liquidus temperatures TL and the glass-transition temperature Tg are
outlined.
any secondary crystallization occurred and the zone formation is not
directly comparable to that e.g. observed due to a thermal gradient
[23]. Hence these three zones contain secondary melts of different,
unknown compositions after the primary ZrO2 crystallization. However,
the oriented mullite layer in zone 2 and the alumina spherulites in zone
3 indicate that the ZrO2 dendrites did not serve as nucleation sites for
the crystallization of other phases.
An illustration of the development of the thermal gradient in the
analyzed area over time is presented in Fig. 16 for reference during the
subsequent discussion concerning the secondary crystallization in the
respective zones 1–3. While the glass transition temperature (Tg) of
neither the primary melt nor the residual melts is known, it is accep-
table to assume that the Tg of the secondary melts are lower than the Tg
of the primary melt due to their lower ZrO2 content. As the melting
points of ZrO2 and alumina are above the casting temperature Tcast,
both may nucleate as soon as the temperature drops below the liquidus
temperature (TL) of the respective melt. Mullite, however, can only
nucleate if TL is below ≈1830 °C.
Just after pouring the melt, i.e. time 1 in Fig. 16, the outer 10–20 μm
of the melt are quenched below Tg of the primary melt due to the
contact with the mold. At this time, the primary growth of dendritic t-
ZrO2 begins in the zones 1 to 3 and probably further in the bulk which
hence must be below the primary liquidus temperature (TL(primary)). As
a result, the secondary local liquidus temperatures of the residual liquid
phases (TL(sec.)) in the respective zones are decreased and the nucleation
of further phases is prevented for some time. As the thermal energy is
removed from the block, the thermal gradient flattens and shifts into
the melt.
3.1. Secondary crystallization in zone 1
At time 2 in Fig. 16, the temperature of the mold and the quenched
area adjacent to it has increased above the local Tg due to heat flow.
The resulting high nucleation rates allow the formation of the fine-
grained area less than 50 μm thick immediately adjacent to the mold in
Fig. 7.
Because of the somewhat flattened thermal gradient, both alumina
and mullite may nucleate in zone 1 while the zones 2 and 3 are still
above their respective secondary liquidus temperatures. The steep
thermal gradient in zone 1 at this time means that mullite would pre-
ferably nucleate near the mold while alumina predominantly nucleates
near the boundary to zone 2. This is supported by the phase distribution
in Fig. 7: mullite is only detected closer to the mold but not near the
W. Wisniewski et al.
boundary to zone 2.
It must be noted that the alumina crystals expel SiO2 during growth
while mullite can incorporate SiO2 into its crystal lattice. An enrich-
ment of SiO2 in the vicinity of the alumina grains will increase the local
viscosity, hindering their growth. This effect will be more significant at
lower temperatures, i.e. closer to the mold. The SiO2 not incorporated
into the mullite crystals forms clear pockets of residual glass enriched in
SiO2 during this secondary crystallization and negligible amounts of
further phases, e.g. indexed as c-ZrO2 in Fig. 6, may occur.
3.2. Secondary crystallization in zone 2
At time 3 in Fig. 16, the secondary melt in zone 2 near the boundary
to zone 1 drops below the local secondary liquidus temperature TL(sec.2)
of the residual liquid phase in zone 2 which must be lower than the
melting point of mullite at 1830 °C A localized nucleation of mullite is
triggered at this interface while alumina nucleation should be possible
throughout this zone at this time. However, the alumina nuclei may be
prevented from significant growth because they expel SiO2 as discussed
above and hence fail to be detected. Mullite is not hindered by this
effect and may thus grow through the partially crystallized melt to-
wards zone 3 which is confirmed by the orientation selection of mullite
along this direction. The preferred growth of mullite along the [001]-
direction in a melt is in agreement with previous observations, [32,35]
so that the detected 001 texture is in agreement with the result expected
from a kinetic selection. It also proves that relatively few mullite nuclei
are formed and the crystal growth velocity is fairly high.
As mullite may incorporate only minor Zr-concentrations of less
than 0.8%, [26] the ZrO2 remaining in the secondary melt is expelled
and crystallized into the bright, fine grained structures observed in the
areas providing EBSD-patterns indexable as mullite. It seems likely that
this ZrO2 grows epitaxially onto the already present mullite crystals.
3.3. Secondary crystallization in zone 3
As all the necessary elements for the crystallization of mullite occur
in zone 3, the secondary liquidus temperature of the residual liquid in
zone 3 TL(sec.3) is probably above the melting point of mullite, pre-
venting its formation. At time 4 in Fig. 16, the temperature in zone 3
drops below TL(sec.3), allowing the formation of the alumina spherulites
and the epitaxial, secondary (probably tetragonal) ZrO2 domains en-
cased within them. At this high temperature the above described layer
enriched in SiO2 around the alumina particles is not an effective dif-
fusion barrier. A small thermal gradient may occur at the boundary of
zone 3, leading to smaller spherulites in this region. As zone 3 crys-
tallized at the highest temperature, it seems probable that the bulk of
the block shows the same crystallization process. If so, the bulk prob-
ably contains a higher amount of dendritic ZrO2 and even larger crys-
tals.
Ultimately, the microstructure formation in this material is com-
pleted by the ZrO2 transformation to the m-ZrO2 detected in the mea-
surements presented above during the final cooling stage. As the ana-
lyzed 24 mm of the AZS-fusion cast material contain three
distinguishably crystallized layers, none of which definitely represent
the bulk of the material, it must be questioned whether single SEM-
micrographs can be considered to represent the microstructure of a
given refractory material [6,7]. The significantly varying grain sizes of
primary and secondary ZrO2 outlined above provide an alternative
explanation for the idea of “dendrite dissolution” due to a heat treat-
ment of 2 h at 1100 °C [3]: considering the high thermal stability of
ZrO2, it is possible that Ref. [3] simply presents micrographs of primary
and secondary ZrO2 from different locations.
The intermingled structures of secondary ZrO2 described in the
zones 2 and 3 have been denoted as “eutectic”, but considering that
both ZrO2-free grains of mullite and alumina occur in the analyzed
material, this mode of eutectic crystallization is improbable.
12
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
4. Conclusion
The fusion cast process used to produce the analyzed AZS-material
leads to the formation of a cast skin at the melt/mold interface con-
taining three layers with compositions differing from that observed
further in the bulk. While the outer parts of this cast skin are usually
removed before application, it is important to consider their existence
when analyzing this class of materials.
The analyzed layer of material containes three zones of different
chemical composition, probably originating from dynamic conditions
during the early stages of the arc melting process. The melt is only
quenched to a glassy state in the 10–20 μm immediately adjacent to the
mold. The crystallization of the analyzed material began with the pri-
mary formation of ZrO2-dendrites throughout the analyzed area. This
subtraction of ZrO2 from the melt leads to the formation of three dif-
ferent secondary melts of unknown chemical composition in the re-
spective areas.
The secondary crystallization in these zones leads to the formation
of a highly oriented layer of mullite in zone 2 as well as a secondary
crystallization of tetragonal ZrO2. In zone 3, the secondary ZrO2 fre-
quently occurs in an epitaxial relationship with large alumina domains.
In the bulk of the analyzed material, SiO2-containing phases do not
crystallize and hence SiO2 is highly enriched in the residual glass. The
crystallized t-ZrO2 transforms to m-ZrO2 during cooling which shows a
heavily twinned structure. Neither the crystallization of ZrSiO4 nor that
of cristobalite is observed in any case.
A fragmentation of ZrO2 dendrites due to annealing at 1100 °C is
unlikely and the morphological differences presented in the literature
are better explained by the two-step formation of ZrO2. It is very im-
portant to explicitly state the origin of an analyzed sample when ana-
lyzing this class of materials.
Conflict of interest statement
There are no conflicts of interest.
Acknowledgments
This work was supported by Deutsche Forschungsgemeinschaft
(DFG) in Bonn Bad Godesberg (Germany) via project nr. RU 417/14-1.
References
[1] S.L. Cockcroft, J.K. Brimacombe, D.G. Walrod, T.A. Myles, Thermal stress analysis
of fused-cast AZS refractories during production: part I, industrial study, J. Am.
Ceram. Soc. 77 (1994) 1505–1511.
[2] S.L. Cockcroft, J.K. Brimacombe, D.G. Walrod, T.A. Myles, Thermal stress analysis
of fused-cast AZS refractories during production: part II, development of thermo-
elastic stress model, J. Am. Ceram. Soc. 77 (1994) 1512–1521.
[3] T. Asokan, Microstructural featuring of fusion cast Al2O3-ZrO2-SiO2 refractories, J.
Mater. Sci. Lett. 13 (1994) 343–345.
[4] L.J. Manfredo, R.N. McNally, The corrosion resistance of high ZrO2 fusion-cast
AI2O3-ZrO2-SiO2 glass refractories in soda lime glass, J. Mater. Sci. 19 (1984)
1272–1276.
[5] A. Ito, S. Endo, R.W. Brown, K.H. Sandmeyer, Fusion-cast zirconia refractories for
use with problem glasses, Ceram. Bull. 59 (1980) 746–747.
[6] E.Y. Fogaing, M. Huger, C. Gault, Elastic properties and microstructure: study of
two fused cast refractory materials, J. Europ. Ceram. Soc. 27 (2007) 1843–1848.
[7] M. Boussuge, Investigation of the thermomechanical properties of industrial re-
fractories: the French programm PROMETHEREF, J. Mater. Sci. 43 (2008)
4069–4078.
[8] V.A. Sokolov, Fusion-cast refractories in the high-zirconia region of the ZrO2 –
SiO2– CaO system, Refract. Ind. Ceram. 46 (2005) 37–40.
[9] K. Madi, S. Forest, M. Boussuge, S. Gailliègue, E. Lataste, J.-Y. Buffière, D. Bernard,
D. Jeulin, Finite element simulations of the deformation of fused-cast refractories
based on X-ray computed tomography, Comput. Mater. Sci. 39 (2007) 224–229.
[10] E. Lataste, J.P. Erauw, C. Olagnon, G. Fantozzi, Microstructural and mechanical
consequences of thermal cycles on a high zirconia fuse-cast refractory, J. Europ.
Ceram. Soc. 29 (2009) 587–594.
[11] C. Patapy, C. Gault, M. Huger, T. Chotard, Acoustic characterization and micro-
structure of high zirconia electrofused refractories, J. Eur. Ceram. Soc. 29 (2009)
3355–3362.
W. Wisniewski et al.
[12] A. Sibil, T. Douillard, C. Cayron, N. Godin, M. R’mili, G. Fantozzi, Microcracking of
high zirconia refractories after t→m phase transitions during cooling: an EBSD
study, J. Eur. Ceram. Soc. 31 (2011) 1525–1531.
[13] C. Patapy, N. Gey, A. Hazotte, M. Humbert, D. Chateigner, R. Guinebretiere,
M. Huger, T. Chotard, Mechanical behaviour characterization of high zirconia
fused-cast refractories at high temperature: influence of the cooling stage on mi-
crostructural changes, J. Eur. Ceram. Soc. 32 (2012) 3929–3939.
[14] C. Patapy, M. Huger, R. Guinebretière, N. Gey, M. Humbert, A. Hazotte, T. Chotard,
Solidification structure in pure zirconia liquid molten phase, J. Eur. Ceram. Soc. 33
(2013) 259–268.
[15] C. Patapy, F. Gouraud, M. Huger, R. Guinebretière, B. Ouladiaff, D. Chateigner,
T. Chotard, Investigation by neutron diffraction of texture induced by the cooling
process of zirconia refractories, J. Eur. Ceram. Soc. 34 (2014) 4043–4052.
[16] J.R. Hellmann, V.S. Stubican, Stable and metastable phase relations in the system
ZrO2‐CaO, J. Am. Ceram. Soc. 66 (1983) 260–264.
[17] C. Pascual, P. Duran, Subsolidus phase equilibria and ordering in the system
ZrO2‐Y2O3, J. Am. Ceram. Soc. 66 (1983) 23–27.
[18] M.R. Thornber, D.J.M. Brevan, E. Summerville, Mixed oxides of the type MO2
(Fluorite)-M2O3. V. Phase studies in the systems ZrO2-M2O3(M= Sc, Yb, Er, Dy), J.
Solid State Chem. 1 (1970) 545–553.
[19] R. Srinivasan, L. Rice, B.H. Davis, Critical particle size and phase transformation in
zirconia: transmission Electron microscopy and X‐ray diffraction studies, J. Am.
Ceram. Soc. 73 (1990) 3528–3530.
[20] G. Baldinozzi, D. Simeone, D. Gosset, M. Dutheil, Neutron diffraction study of the
size-induced tetragonal to monoclinic phase transition in zirconia nanocrystals,
Phys. Rev. Lett. 90 (2003) 216103.
[21] M. Humbert, N. Gey, C. Patapy, E. Joussein, M. Huger, R. Guinebretière, T. Chotard,
A. Hazotte, Identification and orientation determination of parent cubic domains
from electron backscattered diffraction maps of monoclinic pure zirconia, Scripta
Mater. 63 (2010) 411–414.
[22] W. Wisniewski, S. Berndt, M. Müller, C. Rüssel, Stress induced texture formation in
surface crystallized SiO2 glass, CrystEngComm 15 (2013) 2392–2400.
[23] W. Wisniewski, S. Slavov, C. Rüssel, Y. Dimitriev, Growing oriented layers of
13
Journal of the European Ceramic Society xxx (xxxx) xxx–xxx
Bi4Ti3O12 in Bi2O3/TiO2/SiO2/Nd2O3/Al2O3 glass-ceramics by melt quenching, Sci.
Rep. U. K. 8 (2018) 8639.
[24] W. Wisniewski, K. Thieme, C. Rüssel, Fresnoite glass-ceramics – a review, Progr.
Mater. Sci. 98 (2018) 68–107.
[25] S. Martin, H. Berek, C.G. Aneziris, U. Martin, D. Rafaja, Pitfalls of local and
quantitative phase analysis in partially stabilized zirconia, J. Appl. Cryst. 45 (2012)
1136–1144.
[26] H. Schneider, J. Schreuer, B. Hildmann, Structure and properties of mullite—a re-
view, J. Europ. Ceram. Soc. 28 (2008) 329–344.
[27] T. Höche, M. Deckwerth, C. Rüssel, Partial stabilization of tetragonal zirconia in
oxynitride glass-ceramics, J. Am. Ceram. Soc. 81 (1998) 2029–2036.
[28] W. Wisniewski, C. Rüssel, An experimental viewpoint on the information depth of
EBSD, Scanning 38 (2016) 164–171.
[29] W. Wisniewski, S. Saager, A. Böbenroth, C. Rüssel, Experimental evidence con-
cerning the significant information depth of Electron backscatter diffraction
(EBSD), Ultramicroscopy 173 (2017) 1–9.
[30] W. Wisniewski, S. Seidel, C. Patzig, T. Höche, C. Rüssel, Surface crystallization of a
MgO/Y2O3/SiO2/Al2O3/ZrO2 glass: growth of an oriented β-Y2Si2O7 layer and
epitaxial ZrO2, Sci. Rep. UK 7 (2017) 44144.
[31] W. Wisniewski, S. Slavov, C. Rüssel, Y. Dimitriev, Phase formation, crystal or-
ientations and epitaxy in Bi2O3/TiO2/SiO2 (/Nd2O3) glass-ceramics, CrystEngComm
19 (2017) 2775–2785.
[32] W. Wisniewski, R. Carl, G. Völksch, C. Rüssel, Mullite needles grown from a MgO/
Al2O3/TiO2/SiO2/B2O3/CaO glass melt: orientation and diffusion barriers, Cryst.
Growth Des. 11 (2011) 784–790.
[33] W. Wisniewski, M. Patschger, C. Rüssel, Viscous fingering and dendritic growth of
surface crystallized Sr2TiSi2O8 Fresnoite, Sci. Rep.-U. K. 3 (2013) 3558.
[34] M.C. Greca, J.V. Emiliano, A.M. Segadaes, Revised phase equilibrium relationships
in the system Al2O3 ZrO2 SiO2, J. Eur. Ceram. Soc. 9 (1992) 271–283.
[35] S. Berndt, A. Gawronski, C. Patzig, W. Wisniewski, T. Höche, C. Rüssel, Oriented
crystallization of a β-Quartz solid solution from a MgO/Al2O3/SiO2 glass in contact
with tetragonal ZrO2 ceramics, RSC Adv. 5 (2015) 15164–15171.