Open Ceramics 9 (2022) 100240
Available online 23 February 2022
2666-5395/© 2022 The Authors. Published by Elsevier Ltd on behalf of European Ceramic Society. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Fabrication and characterization of porous mullite ceramics derived from
fluoride-assisted Metakaolin-Al(OH)
3
annealing for filtration applications
Amanmyrat Abdullayev
a
, Celal Avcioglu
a
, Tobias Fey
b
,
c
, Andr´
e Hilger
d
, Markus Osenberg
d
,
Ingo Manke
d
, Laura M. Henning
a
, Aleksander Gurlo
a
, Maged F. Bekheet
a
,
*
a
Fachgebiet Keramische Werkstoffe / Chair of Advanced Ceramic Materials, Institute of Materials Science and Technology, Technische Universit¨
at Berlin, 10623, Berlin,
Germany
b
Department of Materials Science and Engineering, Institute of Glass and Ceramics, University of Erlangen-Nürnberg, Erlangen, Germany
c
Frontier Research Institute for Materials Science, Nagoya Institute of Technology, Nagoya, Japan
d
Institute of Applied Materials, Helmholtz Centre for Materials and Energy, Berlin, Germany
ARTICLE INFO
Keywords:
Mullite whiskers
Ceramic membranes
Water filtration
In situ synchrotron X-ray diffraction
Synchrotron X-ray tomography (
μ
CT)
ABSTRACT
In this work, polycrystalline mullite whiskers are synthesized by fluoride-assisted method from metakaolin and
several aluminum-containing compounds such as γ-Al(OH)
3
, AlF
3
⋅3H
2
O, and
α
-Al
2
O
3
(corundum). The mullite
formation and crystallization are assessed both in ex situ and in situ synchrotron X-ray diffraction experiments
under synthesis conditions. Polycrystalline mullite starts to form from metakaolin, Al(OH)
3
, and AlF
3
⋅3H
2
O
reactants at 680
◦
C, whereas mullite does not form even at 1000
◦
C when corundum is used. Porous mullite
ceracmics are fabricated at sintering temperatures between 1000 and 1700
◦
C and tested for water permeance.
Scanning Electron Microscopy (SEM) and synchrotron X-ray tomography (
μ
CT) reveal that ceramics are
comprised of pore channels with an interlocked network of mullite whiskers. With competitive porosity (up to 63
%), compressive strength (up to 20 MPa), and pure water flux (up to 579 L/m
2
⋅h at 1 bar), fabricated mullite
ceramics are promising candidates for water filtration and purification.
1. Introduction
Water pollution, particularly contamination of clean water resources
and inadequate wastewater treatment, poses a significant risk to public
health and the ecosystem [1]. Hence, clean water and water sanitation
became one of the 17 sustainable development goals of the United Na-
tions (UN). One of the effective methods for wastewater treatment is the
use of ceramic membranes [2]. Commercially available ceramic mem-
branes are typically made from Al
2
O
3
, TiO
2
, SiC, and ZrO
2
[3]. However,
costly raw materials and high-temperature processing requirements
prevent their large-scale implementation in wastewater treatment.
Therefore, the fabrication of ceramic membranes from low-cost mate-
rials using relatively low-temperature manufacturing processes would
minimize their ecological footprint and encourage widespread applica-
tion, particularly in the Global South countries. In this direction, the past
few decades have witnessed a dramatic growth of research interest in
exploiting cost-effective ceramic membrane production methods [4,5].
Among various structural engineering materials, porous mullite
(3Al
2
O
3
⋅2SiO
2
) ceramics have attracted much attention regarding their
outstanding merits, such as high mechanical strength, good stability in
harsh conditions, good antifouling properties, thermal shock, creep, and
pressure resistances [6]. Besides, various affordable minerals and in-
dustrial wastes, such as aluminum sludge, coal fly ash, rice husk silica,
coal gangue, sago waste, topaz sand, diatomite, bauxite, ball clay, and
kaolin, have been employed as starting materials for the production of
mullite ceramics [5]. Nevertheless, its synthesis from clay minerals re-
mains the most viable option in terms of cost-efficiency. For instance,
mullite could be obtained by high-temperature treatment of kaolin,
which is the most prominent member of the clay family and costs only
1 $/kg. Kaolin transforms first into metakaolin (eq. (1)) in the temper-
ature range of 400–700 ◦C, before decomposing into crystalline mullite
and silica (SiO
2
) at temperatures above 1000
◦
C (eq. (2)) [7]. However,
the formation of the residual silica phase along with mullite is detri-
mental to the thermochemical properties. Accordingly, an additional
alumina source has been applied during the synthesis to react with the
excess silica in metakaolin to produce phase pure mullite ceramics [7,8].
In this regard, pure corundum (
α
-Al
2
O
3
) is the most preferred additional
alumina source due to its high purity [9]. However, difficulties arise in
* Corresponding author.
E-mail address: [email protected] (M.F. Bekheet).
Contents lists available at ScienceDirect
Open Ceramics
journal homepage: www.sciencedirect.com/journal/open-ceramics
https://doi.org/10.1016/j.oceram.2022.100240
Received 7 December 2021; Received in revised form 17 February 2022; Accepted 21 February 2022
Open Ceramics 9 (2022) 100240
2
reacting the excess silica when corundum (
α
-Al
2
O
3
) is mechanically
mixed to form mullite, because this reaction requires relatively high
temperatures (>1400
◦
C). The formation temperature of mullite could
be reduced to below 1400
◦
C when liquid phase forming additives, e. g.,
TiO
2
[10], Y
2
O
3
[11], Fe
2
O
3
[12] were added during the synthesis
process. Yet, even with the assistance of these additives, complete
mullite formation from kaolin and corundum mixture requires temper-
atures higher than 1100
◦
C. Besides, these liquid phase forming agents
may form residual glassy silicate phases, adversely affecting mullite
properties.
Al2O3⋅ 2SiO2⋅2H2O→
400−700 ◦CAl2O3⋅2SiO2+2H2O(1)
3(Al2O3⋅ 2SiO2)→
>1000 ◦C3Al2O3⋅ 2SiO2+4SiO2(2)
Recent studies have shown that gas-phase forming additives, such as
AlF
3
, are more beneficial than liquid phase forming agents to lower the
mullitization temperature as well as obtain phase-pure mullite [9,
13–16]. The challenge to reduce the mullitization temperature lower
than 1000 ◦C still remains when corundum is used. However, recently, a
synthesis of phase-pure mullite was achieved at temperatures as low as
800 ◦C by utilizing Al
2
(SO
4
)
3
instead of corundum as an alumina source
under the fluoride-assisted synthesis conditions [17]. But, since
Al
2
(SO
4
)
3
decomposes to toxic sulfur oxides, this compound is not
favorable for environmental reasons.
Accordingly, in the present work, we use natural mineral metakaolin
as a primary raw material and either gibbsite (γ-Al(OH)
3
) or corundum
(
α
-Al
2
O
3
) as an additional aluminum source to synthesize mullite using
the fluoride-assisted method. Although Chen et al. could synthesize
mullite from natural mineral kaolin combined with additives Al(OH)
3
and AlF
3
as alumina sources above 1300 ◦C, the exact amounts of used
alumina additives were not given in this work [18]. However, the for-
mation of mullite from different clay materials and alumina sources
might require a high amount of AlF
3
additive (i.e., 40 wt% of total
powder mixture) [9]. Thus, in this work, guided by in situ synchrotron
XRD experiments, we could successfully synthesize the mullite at much
lower temperatures (~700 ◦C) using low content of AlF
3
that does not
exceed 10 wt% of total powder. Moreover, the feasibility of porous
mullite membrane support fabrication is investigated using a mixture of
metakaolin, Al(OH)
3,
and AlF
3
⋅3H
2
O powders. By altering annealing
conditions, mullite ceramics with varying morphologies, pore sizes, and
mechanical strengths are obtained, then their performances are sys-
tematically evaluated.
2. Materials and methods
2.1. Materials
Metakaolin (Metamax, BASF, hereafter denoted as MK), which is a
calcined form of kaolin clay, was used as the main raw material.
Aluminum hydroxide (γ-Al(OH)
3
, >99%, Merck) corundum (
α
-Al
2
O
3
,
99.99%, AKP-50, Sumitomo), aluminum fluoride trihydrate (AlF
3
⋅3H
2
O,
≥97%, Ventron) were used as an additional alumina source and
mineralizer, respectively.
2.2. The synthesis of mullite powders
Crystalline mullites were synthesized by fluoride-assisted solid-state
route as described below. In order to ascertain the molar ratio of
alumina to silica, the silica and alumina content of the clay (MK) was
analyzed by XRF, revealing a ratio of 1:2 (Table 1). Thus, 13.7 g γ-Al
(OH)
3
and 2.7 g AlF
3
⋅3H
2
O were added as additional alumina sources to
11.0 g MK to obtain the P-1 precursor for synthesizing 3Al
2
O
3
⋅2SiO
2
(3:2
mullite). To study the influence of aluminum source on the crystalliza-
tion of mullite, a stoichiometric amount of
α
-Al
2
O
3
was added instead of
γ-Al(OH)
3
to AlF
3
⋅3H
2
O and MK to obtain the P-2 precursor. The powder
mixtures were mixed in an agate mortar and pestle for 10 min before
placing them in closed alumina crucibles. The crucibles were heated at
1000
◦
C for 30 min (heating and cooling rates of 5 K min
−1
) in a resistive
furnace (Nabertherm).
2.3. The fabrication of porous mullite ceramics
Cylindrical samples with 6 mm diameter and 12 mm height were
prepared for porosity, compressive strength,
μ
CT, and mercury intrusion
tests. Disc-shaped samples with 25 mm diameter and ~3 mm thickness
were prepared for the water permeance test as follows. The M-1 powder
mixture was first prepared from MK, Al(OH)
3,
and AlF
3
⋅3H
2
O, as
described above. Then, 3.85 wt% of 5 wt% PVA solution were added to
the powder mixtures as a binder for pressing. In the next step, the ground
powder mixtures were pelletized at 50 MPa for 5 min by uniaxial
compression (Paul-Otto Weber, Germany). After compression, the ob-
tained green bodies were dried at 24
◦
C for 24 h before heating them in
covered alumina crucibles. Alumina paste was applied between the
crucible and lid to minimize the escape of reactive gases from the system
at high temperatures. Then the closed crucibles were heated first at
1000
◦
C in a resistive furnace (Nabertherm) with a heating rate of 2.5 K
min
−1
, a holding time of 60 min, and a cooling rate of 5 K min
−1
to
obtain porous mullite bodies, this specimen is denoted as M-10. Subse-
quently, samples were sintered at several temperatures (1400
◦
C, 1500
◦
C, 1600
◦
C, and 1700
◦
C) with a heating rate of 5 K min
−1
, holding time
of 4 h and cooling rate of 5 K min
−1
. For convenience, the obtained
samples were denoted as M-14, M-15, M-16, and M-17 according to their
sintering temperatures.
2.4. Characterizations
2.4.1. Powder X-ray diffraction
The ex situ powder X-ray diffraction (PXRD) measurements were
performed at room temperature (RT) in a D8 Advance (Brucker, Ger-
many) using CuK
α
radiation in the 2θ range of 10–70◦with a step size of
0.02◦and step time of 8 s.
The in situ high-temperature synchrotron XRD experiments were
performed at the beamline 12.2.2, Advanced Light Source (Lawrence
Berkeley National Laboratory, California, USA) using monochromatic
synchrotron radiation with λ =0.495 Å (25 keV/30 mm spot size) in the
angle-dispersive transmission mode using a Pilatus detector. About 1 mg
of the powder sample was heated in a 0.7 mm outer diameter quartz
capillary (Hilgenberg GmbH, Germany) under quasi-flowing conditions
(O
2
:N
2
=1:4). The gases were injected through a 0.5 mm outer diameter
tungsten tube. The capillary was heated at 20 K min
−1
from RT to 950
◦
C
in an infrared heated SiC tube furnace as described elsewhere [19,20].
Diffraction patterns were recorded every 60 s during the heating cycle.
2.4.2. SEM
The microstructure of samples was investigated using Scanning
Electron Microscopy (SEM) LEO 1530 (Carl Zeiss, Jena, Germany). The
specimens for SEM characterization were cut from the middle of cylin-
drical samples using a diamond disc and then washed in an ultrasonic
Table 1
Chemical composition of metakaolin as determined by XRF.
Compound Weight fraction, % Molar fraction, %
SiO
2
53.75 66.35
Al
2
O
3
43.82 31.82
Fe
2
O
3
0.45 0.21
CaO 0.16 0.21
K
2
O 0.18 0.14
Na
2
O 0.26 0.31
SO
3
0.02 0.02
P
2
O
5
0.43 0.14
TiO
2
0.86 0.79
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
3
bath to remove the loose dust particles before sputtering with a carbon
layer.
2.4.3. Synchrotron X-ray tomography (
μ
CT)
Synchrotron X-ray tomography was carried out at the BAMline at
BESSY II of Helmholtz-Centre Berlin, Germany. The synchrotron beam
was monochromatized to 20 keV using a double multilayer mono-
chromator with an energy resolution of about 1.5 %. The detector sys-
tem is comprised a CdWO
4
scintillator, a microscopic optic, and a pco.
edge camera with a 2560 ×2160 pixel
2
sCMOS chip. The pixel size was
0.72
μ
m, and the corresponding field of view was 1.8 ×1.6 mm
2
(length
×height). For the tomographic reconstruction, an implementation of a
filtered back-projection algorithm, the IDL-based library called “grid-
rec” was used [21].
The geodesic tortuosity is defined as a ratio of the lengths of the
shortest transportation paths (geodesic lengths) to euclidean lengths.
Tortuosities were calculated of a volume of 1500 ×1500 ×1500 voxel
using the geodesic distance transformation [22].
2.4.4. Porosity characterization
The open porosity of sintered mullite ceramics was determined by
the Archimedes method, according to the ASTM C-373-18 [23] as fol-
lows. The dried samples were first weighed before the Archimedes ex-
periments to determine W
d
, and then the samples were submerged into
the water until the open porosities were saturated with water. While the
samples were in the water, their weights were recorded to determine the
suspended weights (W
s
). Subsequently, the samples were taken out from
the water, and the surfaces were gently dried using a microfiber cloth,
before weighing the samples to determine their saturated weights (W
m
).
The open porosity (φ) was calculated from dry (W
d
), suspended (W
s
) and
saturated (W
m
) weights according to Eq. (3).
φ=Wm−Wd
Wm−Ws
*100 (%)(3)
Pore size distribution was analyzed with a mercury intrusion
porosimeter (Pascal 140, Porotec, Germany). It should be emphasized
that the mercury porosimetry relies on the Washburn equation, which is
applicable for cylindrical pores [24]. In this work, pores were assumed
to be cylindrical, and the pore size distributions were presented for
comparison purposes.
The diameter of disc-shaped samples before and after heat treatment
was measured using a digital micrometer (Micromar 40 EWR, Mahr,
Germany) with a precision of 0.001 mm. From the change in diameter of
the samples, shrinkage was calculated according to Eq. (4):
shrinkage =d0−df
d0
*100 (%)(4)
where d
0
and d
f
represent the diameter of samples before and after heat
treatment, respectively.
2.4.5. Mechanical stability
The mechanical stability of sintered cylindrical mullite ceramics, i.e.,
4.7–6.1 mm in diameter and 10–13 mm in height, was evaluated by
compressive testing in a RetroLine testing machine using testXpert
v.11 software (Zwick/Roell, Ulm, Germany). Briefly, cylindrical samples
were compressed with a testing speed of 0.5 mm min
−1
until fracture
occurred or deformation exceeded 10%. The compressive strength was
obtained from the quotient of maximum force and cross sectional area of
the test piece. At least four replicas were tested for each sample.
2.4.6. Pure water permeance
A lab-built dead-end filtration setup was used for the water per-
meance tests as described elsewhere [25]. Briefly, the ceramic mem-
brane support was first placed on the bottom of the stainless-steel
container and fixed with O-rings to prevent leakage or bypass of water.
Then, the container was filled with distilled water and pressurized with
nitrogen gas; the volume of water passed through the membrane was
recorded. Water flux was evaluated at different water pressures ac-
cording to Eq. (5):
Jw=Q
AΔt(5)
where J
w
is the water flux (L/m
2
⋅h), Q is the volume of permeated water
(L), A is the effective surface area of membrane (m
2
), and Δt is the time
spent for the permeation of water (h).
Fig. 1. XRD patterns of mullite prepared using
different sources of additional aluminum to
consume excess silica in metakaolin. P-1 is obtained
from the mixture of metakaolin, γ-Al(OH)
3
and
AlF
3
⋅3H
2
O; P-2 is obtained from the mixture of
metakaolin,
α
-Al
2
O
3
and AlF
3
⋅3H
2
O. Additionally,
XRD patterns of metakaolin plus AlF
3
⋅3H
2
O, meta-
kaolin plus γ-Al
2
O
3
, and metakaolin plus Al(OH)
3
are presented as controls, which prove that only P-1
precursor is capable of producing mullite at that at
1000 ◦C. Calculated patterns of
α
-Al
2
O
3
and mullite
are presented as a reference at the bottom of the
image.
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
4
3. Results and discussion
3.1. Synthesis conditions of polycrystalline mullite powders assisted by in
situ synchrotron XRD experiments
Fig. 1 shows ex situ XRD patterns of the samples obtained by heating
P-1 and P-2 precursors at 1000
◦
C. The heat treatment of the precursor P-
1 results in the formation of phase-pure mullite, indicating the successful
reaction between Al(OH)
3
, AlF
3,
and MK at 1000
◦
C. In contrast, when P-
2 is used, the corundum (
α
-Al
2
O
3
) is still the primary phase, and only a
small amount of mullite phase is formed, which is the product of the
reaction of AlF
3
with silica [26]. This finding is also confirmed by
reacting only MK and AlF
3
⋅3H
2
O with a weight ratio of 1:1, which leads
to the formation of phase-pure mullite (Fig. 1). However, using such a
high amount of AlF
3
⋅3H
2
O is neither environmentally nor economically
feasible. These results suggest that
α
-Al
2
O
3
does not react with MK at
Fig. 2. In situ heating XRD patterns of precursor P-1 (a) and precursor P-2 (b) from room temperature to 950
◦C. Calculated patterns of phases are presented at the
bottom as reference.
Fig. 3. XRD patterns of mullite prepared from the mixture of metakaolin, γ-Al(OH)
3
and AlF
3
⋅3H
2
O at 1000 ◦C and sintered at 1000 ◦C (M-10) followed by another
sintering step at 1400 ◦C (M-14), 1500 ◦C (M-15), 1600 ◦C (M-16), and 1700 ◦C(M-17).
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
5
these experimental conditions. Moreover, several other control experi-
ments using different starting powder compositions, e.g. MK/Al(OH)
3
or
MK/γ-Al
2
O
3
, shows that mullite can not be obtained under these cir-
cumstances without AlF
3
, see Fig. 1. This finding is in agreement with
the literature, which shows the importance of fluoride for the
low-temperature formation of mullite [15–17]. It should be noted that
not only fluoride is important in mullite formation, but also Al
2
O
3
phase
is also crucial. For instance, in our previous work, we observed that in
the presence of AlF
3,
mullite was formed only if Al
2
(SO
4
)
3
was used as an
alumina source, while no mullite was formed when
α
-Al
2
O
3
or γ-Al
2
O
3
was used [17].
To get better insight into the formation mechanism of polycrystalline
mullite, the transformation of precursors P-1 and P-2 into poly-
crystalline mullite was followed by in situ synchrotron XRD experiments
during heating from room temperature up to 950
◦
C in air. Fig. 2 a and b
display the in situ XRD patterns collected during the heating of both
precursors under the air atmosphere. As shown in Fig. 2a, the precursor
M-1 contains crystalline gibbsite Al(OH)
3
and AlF
3
⋅3H
2
O in addition to
the amorphous MK phase.
The in situ XRD data reveal that the heating of the precursor P-1
leads to several consecutive changes with temperature. First, the
AlF
3
⋅3H
2
O phase is transformed into anhydrous rhombohedral AlF
3
at
200
◦
C by losing its chemically bonded water, which is consistent with
previous reports [17,27]. At 250
◦
C, gibbsite γ−Al(OH)
3
starts to
transform into the boehmite γ−AlOOH phase. The later phase trans-
formation continues until 330
◦
C, which agrees with previously reported
works [28]. At 540
◦
C, the boehmite phase starts to transform into an
amorphous alumina phase [28,29] that reacts with AlF
3
and amorphous
MK at 685
◦
C to form mullite phase. Surprisingly, no fluorotopaz
(Al
2
SiO
4
F
2
) phase is observed in the in situ XRD experiments. Previous
works reported that the fluorotopaz phase is formed from clay and Al
2
O
3
in the presence of silicon tetrafluoride (SiF
4
) gas under similar condi-
tions at temperatures below 900
◦
C, before decomposing endothermi-
cally to form mullite and SiF
4
above 1100
◦
C [30–32]. The results of the
in situ XRD experiments suggest that mullite can form directly from the
mixture of MK, γ-Al(OH)
3,
and AlF
3
⋅3H
2
O without forming fluorotopaz
as an intermediate phase, which is in agreement with several recent
works [33,34].
For the precursor P-2,
α
-Al
2
O
3
and AlF
3
⋅3H
2
O were found at room
temperature in addition to the amorphous MK phase. Similar to the
precursor P-1, AlF
3
⋅3H
2
O transforms into anhydrous AlF
3
at 200
◦
C
before the formation of a small amount of mullite phase at 685
◦
C.
However, the corundum
α
-Al
2
O
3
phase was stable during the entire
experiment, implying that corundum is not involved in the formation of
the mullite phase. This small amount of mullite formed is due to the
reaction between MK and AlF
3
. These results are in good agreement with
previous works that suggest temperatures higher than 1100
◦
C are
required for the formation of mullite from corundum and silica phases
[9,35,36].
3.2. The phase composition of mullite ceramics
The phase composition of mullite ceramics obtained by calcination
and sintering of P-1 precursors at various temperatures was investigated
using ex situ XRD characterization. As shown in Fig. 3, all samples
contain only a polycrystalline mullite phase.
The XRD reflections become sharper and narrower with increasing
the sintering temperatures, suggesting the increase in the crystallinity
and crystallite size of the mullite phase. However, other than that, no
significant microstructural change was observed, implying that obtained
mullite does not undergo any considerable structural change even for
high-temperature treatment at 1700
◦
C.
3.3. The microstructure of mullite ceramics
SEM images of the porous mullite ceramics, which were prepared at
several different temperatures, are presented in Fig. 4. Under fluoride-
assisted conditions, mullite forms in a solid-gas reaction, and due to
fewer constraints, mullite crystals grow faster along the c direction [14].
Consequently, these mullite whiskers with various lengths produce a
highly porous interlocked network of mullite, which is consistent with
previous works [9,37,38].
Surprisingly, there is not much difference observed in the micro-
structure when increasing the sintering temperature from 1000
◦
C to
Fig. 4. SEM images of cross-section of porous mullite membranes consisting of an interlocked network of mullite whiskers. Samples were prepared at different
sintering temperatures (M-10 at 1000
◦C, M-14 at 1400 ◦C, M-15 at 1500 ◦C, and M-16 at 1600 ◦C).
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
6
1600
◦
C. Only at 1600
◦
C, better densification of porous mullite matrix
can be seen, particularly on the wall or strut of pores, due to the increase
in the sinterability of mullite at such a high temperature. This is also
supported by porosity and shrinkage findings. Nevertheless, the nest-
like structure of interlocked mullite whiskers network remains present
in both samples.
μ
CT characterization also revealed that samples M-10 and M-16
display similar microstructure (Fig. 5). However, it should be noted that
the pixel size of
μ
CT equipment is 0.72
μ
m/voxel, which means smaller
pores were not detectable with this technique. Nevertheless, the skele-
tonized images show that both samples exhibit a connected network of
pores (see Fig. 5). Moreover, tortuosity of pores was calculated using
reconstructed
μ
CT tomograms, and it was found that M-10 and M-16
possess tortuosity factor of 1.21 and 1.29, respectively, being in this
Fig. 5. Partial volumes of reconstructed
μ
CT images for samples M-10 and M-16. a): The images show the network that contains nanopores. The main network was
rendered with a white surface, while sections without connection to the outside were rendered in red. b): Here, the now transparent main network contains spheres
with the pore equivalent diameter (colored accordingly). The spheres have been connected with each other based on the underlying network structure. (For
interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
Fig. 6. Pore size distribution of samples M-10 and M-16.
Fig. 7. Porosity, compressive strength, and shrinkage of porous mullite ceramics, which were prepared at several different temperatures: M-10 at 1000
◦C, M-14 at
1400 ◦C, etc.
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
7
respect very close to an ideal tortuosity which is equal to unity. These
values are comparable to those 1.12 and 1.16, which were recently re-
ported for whisker-based mullite membranes [9,39]. These findings
revealed that obtained mullite ceramics have high porosity and possess
excellent connectivity of pores; both have paramount importance for the
separation and filtration applications (i.e. permeance of membranes).
The pore size distribution of samples M-10 and M-16 are presented in
Fig. 6. Sample M-10 exhibits a broader range of pores with a median
diameter of 0.28
μ
m and bimodal pore size distribution. Also, the dis-
tribution of pore diameter skewed to the left, which means there are
pores with pore diameter as small as 50 nm. This is because there are two
possibilities for pores to form: i) interparticle pores are formed during
the whisker formation; ii) replicate pores are formed after the decom-
position of pore former/template. Here, Al(OH)
3
acts as a pore former
because it undergoes 60 % volume contraction due to decomposition at
high temperatures, which will produce extra porosity [40]. On the other
hand, sample M-16 has a quite narrow pore size range with a larger
median pore diameter of 0.78
μ
m compared to M-10. The larger and
more uniform pore size distribution of M-16 can be attributed to pore
widening and grain growth due to the coalescence of tiny particles into
larger ones. At the same time, their number decreases during heating to
high temperatures [40,41]. Hence, sample M-16 has larger pores with
lower porosity relative to M-10.
3.4. The porosity and mechanical stability of mullite ceramics
Open porosities, diametral shrinkage, and compressive strengths of
the sintered mullite ceramics are presented in Fig. 7. The samples M-10,
M-14, and M-15, possess high porosities of about 63 ±1 %, which can be
explained by the evolution of fluoride-based gases, O
2,
and water vapor
during the synthesis process. Besides, as depicted in SEM images (Fig. 4),
mullite whiskers grow as needle-like crystals, whereas their interlocked
network produces a nest-like structure with high porosity [9]. At sin-
tering temperatures higher than 1600
◦
C, the mullite approaches its
melting point of about 1800
◦
C, and the densification of mullite is
enhanced [42]. Thus, the porosity is remarkably decreased from 51 % to
11 % in the samples M-16 and M-17 sintered at 1600
◦
C and 1700
◦
C,
respectively.
As known, the densification of porous bodies is accompanied by a
decrease in their porosity and an increase in their mechanical strength
[7]. Indeed, as displayed in Fig. 7, the shrinkage and compressive
strengths of the samples are increased while their porosities are
decreased. The similar porosity and mechanical stability of M-10, M-14,
and M-15 suggest that the pore channels in mullite ceramics are stable
up to 1500
◦
C, which might be promising for high-temperature appli-
cations, such as hot-gas filtration or thermal insulation. Since samples
M-14 and M-15 do not show better mechanical strength than sample
M-10, the latter has been selected for permeance tests for economic
reasons because it is prepared at a lower temperature. However, the
sample M-16 still has adequate porosity and better mechanical stability
than the sample M-10, which makes it very promising for filtration ap-
plications, where higher porosity with good mechanical strength is
desired. In contrast, the sample M-17 has very low porosity despite its
quite good mechanical stability, so it is not appropriate for filtration
applications. Hence samples M-10 and M-16 were further investigated
for their potential applications.
3.5. Water permeance test
As noted above, samples M-10 and M-16 have been chosen in this
work for the water permeance tests. Pure water flux of samples was
measured as described in the experimental section. Additionally, using
all measured parameters porosity, pore size, and tortuosity, the pure
water flux can be calculated according to the widely used Hagen-
Poiseuille (H–P) equation [43]:
Jw=φ⋅r2⋅ΔP
8⋅
μ
⋅
τ
⋅Δx (6)
where J
w
is pure water flux (m/s or 1000 L/m
2
⋅h), φ is the porosity, r is
the average pore radius (m), ΔP is the transmembrane pressure differ-
ence (Pa),
μ
is the viscosity of water (Pa⋅s),
τ
is the tortuosity of the
membrane, and Δx is the thickness of the membrane (m). Fig. 8 displays
the experimental water flux (data points) along with the values calcu-
lated (lines, a grey area considering the standard deviation) according to
the eq. (6) from the available data (i.e. porosity, pore size distribution,
tortuosity, membrane thickness, and pressure drop).
As shown in Fig. 8, the experimental water flux of the sample M-10 is
within the calculated area (all dots are lying within the grey area), while
the calculated and experimental water flux values of the sample M-16
are differing a lot (dots are lying out of the grey area). These results can
be attributed to two reasons: First, the Hagen-Poiseuille equation, as
well as the pore size distribution measurements using mercury poros-
imetry, assume the perfectly cylindrical shape of the pores. However,
the porous mullite ceramics do not exhibit perfect cylindrical pores;
instead, they are composed of an interlocked network of mullite whis-
kers, as shown in Fig. 4. This deviation from the cylindrical shape of the
pores might cause this mismatch between the calculated and experi-
mental water flux values. Secondly, not all of the open porosity deter-
mined by the Archimedes method might originate from through pores.
Rather, also blind pores might contribute to the open porosity. However,
as they are not of service for water permeation through the membrane, it
can be assumed that the overall open porosity useful for permeation is
lower than what was determined by the Archimedes method.
Fig. 8. Pure water flux of membranes M-10 (a) and M-16 (b) as a function of pressure. Dots represent experimental values, and the line/grey area show the values
calculated according to the Hagen-Poiseuille (H–P) equation, eq. (6) (see text for the details). The mean and median pore radius were used to calculate the solid and
dotted lines, respectively.
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
8
Consequently, that led to a discrepancy between calculated and exper-
imental flux values.
Besides, a higher flux (both experimental and calculated) is observed
for the sample M-16 despite its lower porosity, i.e., the flux of the sample
M-10 at 1 bar is 479 L/m
2
⋅h, with 63 % porosity, but the flux of the
sample M-16 at 1 bar is 576 L/m
2
⋅h, with 51 % porosity. These results
can also be due to the deviation from the cylindrical pore shapes and
overestimation of porosity by Archimedes method, as explained above.
Additionally, it appears that pore radius has a more significant impact
on flux than porosity. Indeed, sample M-16 has larger pores with a
median diameter of 0.78
μ
m, while sample M-10 has smaller pores with
a median diameter of 0.28
μ
m.
The Hagen-Poiseuille equation appears to serve as a helpful instru-
ment to obtain a first estimation of the pure water flux even for the
presented whisker-based mullite ceramics. For more precise permeance
calculations for fibrous porous media, other numerical approaches –
such as the Carman-Kozeny model – have to be applied [44–46]. How-
ever, this would go beyond the scope of this study.
In Table 2, the main properties of ceramics fabricated in this work
are compared with other membranes with similar compositions previ-
ously reported in the literature. Apparently, samples M-10 and M-16
show competitive porosity, pore size and permeance values.
Overall, experimental pure water flux results are quite promising,
especially considering facile preparation, low-cost material metakaolin,
and low-temperature processes (at least for M-10). The sample M-10 has
high porosity with acceptable mechanical stability (withstanding
against pressures as high as 2 bar in the water permeance experiments),
making it suitable for applications where high mechanical strength is
not required. Moreover, due to its high thermal stability and low thermal
conduction of mullite, the sample M-10 can also be used directly in
thermal insulation applications. However, if high mechanical strength is
desired for water or hot-gas filtration applications, the sample M-16
sintered at 1600
◦
C could be promising.
4. Conclusions
In this work, the influence of aluminum sources, such as γ-Al(OH)
3
,
α
-Al
2
O
3
, and AlF
3
⋅3H
2
O, on the crystallization temperature of mullite
from metakaolin under fluoride-assisted condition has been investi-
gated. The results of ex situ XRD, and in situ synchrotron XRD experi-
ments revealed that γ-Al(OH)
3
led to the formation of phase pure mullite
at temperature as low as 680
◦
C, while corundum
α
-Al
2
O
3
did not react
even at 1000
◦
C. SEM characterization showed that the mullite grew as
needle-like whiskers. Moreover, the decomposition of Al(OH)
3
at high
temperatures produces extra porosity, making this process attractive for
membrane fabrication. Porous mullite ceramics were fabricated by
preparing at several temperatures from 1000 ◦C to 1700
◦
C and tested for
their applicability for water filtration. The results showed that inter-
locked mullite whiskers led to highly porous ceramics that presented
competitive properties as a membrane with 0.78
μ
m median pore
diameter, 51 % porosity, 20 MPa compressive strength, and 576 L/m
2
⋅h
at 1 bar pure water flux.
Conflicts of interest
The authors declare no conflict of interest.
CRediT authorship contribution statement
Amanmyrat Abdullayev: Conceptualization, Methodology, Inves-
tigation, Data curation, Writing – original draft, Preparation. Celal
Avcioglu: Methodology, Investigation, Writing – review & editing.
Tobias Fey: Investigation, Data curation. Andr´
e Hilger: Investigation,
Data curation. Markus Osenberg: Investigation, Data curation. Ingo
Manke: Investigation, Data curation. Laura M. Henning: Methodology,
Investigation, Formal analysis, Writing – review & editing. Aleksander
Gurlo: Conceptualization, Supervision, Writing – review & editing,
Project administration, Funding acquisition, Resources. Maged F.
Bekheet: Conceptualization, Methodology, Investigation, Data cura-
tion, Writing – review & editing, Project administration, Supervision, All
authors have read and agreed to the published version of the
manuscript.
Acknowledgments
We would like to thank Fabian Zemke and Franz Kamutzki for the
SEM measurements, both from Technische Universit¨
at Berlin. A.
Abdullayev expresses his gratitude to the German Academic Exchange
Service (DAAD) for scholarship support (grant number 91611173). The
authors further thank the Advanced Light Source (which is supported by
the Director, Office of Science, Office of Basic Energy Sciences, of the U.
S. Department of Energy under Contract No. DE-AC02-05CH11231),
where in situ XRD measurements were conducted at beamline 12.2.2 in
the framework of GUP proposal (ALS-10533).
References
[1] R.P. Schwarzenbach, T. Egli, T.B. Hofstetter, U. von Gunten, B. Wehrli, Global
water pollution and human health, Annu. Rev. Environ. Resour. 35 (2010)
109–136.
[2] Z. He, Z. Lyu, Q. Gu, L. Zhang, J. Wang, Ceramic-based membranes for water and
wastewater treatment, Colloids Surf. A Physicochem. Eng. Asp. 578 (2019)
123513.
[3] E. Drioli, L. Giorno, E. Fontananova (Eds.), Comprehensive Membrane Science and
Engineering, Elsevier, Amsterdam [the Netherlands], 2017.
[4] L. Sun, Z. Wang, B. Gao, Ceramic membranes originated from cost effective and
abundant natural minerals and industrial wastes for broad applications - a review,
DWT 201 (2020) 121–138.
[5] A. Abdullayev, M.F. Bekheet, D.A.H. Hanaor, A. Gurlo, Materials and applications
for low-cost ceramic membranes, Membranes 9 (2019).
[6] H. Schneider, S. Komarneni, Mullite, Wiley-VCH, Chichester John Wiley
[distributor], Weinheim, 2005.
[7] S. Komarneni, H. Schneider, K. Okada, in: H. Schneider, S. Komarneni (Eds.),
Mullite Synthesis and Processing, Mullite, Wiley-VCH Verlag GmbH & Co. KGaA,
Weinheim, FRG, 2005, pp. 251–348.
[8] M. Salleh Choo, Matori Kok, A review on synthesis of mullite ceramics from
industrial wastes, Recycling 4 (2019) 39.
[9] M. Rashad, G. Logesh, U. Sabu, M. Balasubramanian, A novel monolithic mullite
microfiltration membrane for oil-in-water emulsion separation, J. Membr. Sci.
(2020), 118857.
Table 2
Benchmarking of main properties of membranes with similar compositions.
Raw material AlF
3
content, wt% Processing temperature, ◦C Porosity, % Median Pore size,
μ
m Pure water flux at 1 bar, L/m
2
⋅h Reference
Kaolin,
α
-Al
2
O
3
and AlF
3
40 1400 64 0.3 1031 [9]
Kaolin,
α
-Al
2
O
3
, Al(OH)
3
, AlF
3
5 1500 46 1.3 4000 [36]
Kaolin, Al(OH)
3
, AlF
3
NP
a
1550 39 1.16 NA
a
[18]
α
-Al
2
O
3
-coated diaomite +kaolin NA
a
1200 NA
a
0.12 200 [47]
α
-Al
2
O
3
NA
a
1350 38 0.15 110 [48]
Kaolin,
α
-Al
2
O
3
NA
a
1350 56 NA
a
582 [49]
Kaolin,
α
-Al
2
O
3
NA
a
1450 44 1.3 500 [50]
Metakaolin, Al(OH)
3
, AlF
3
10 1000 63 0.28 479 This work
Metakaolin, Al(OH)
3
, AlF
3
10 1600 51 0.78 576 This work
a
NA – not applicable, NP- used but not presented.
A. Abdullayev et al.
Open Ceramics 9 (2022) 100240
9
[10] T.S. Zhang, L.B. Kong, Z.H. Du, J. Ma, S. Li, Tailoring the microstructure of
mechanoactivated Al
2
O
3
and SiO
2
mixtures with TiO
2
addition, J. Alloys Compd.
506 (2010) 777–783.
[11] J. She, P. Mechnich, M. Schmücker, H. Schneider, Low-temperature reaction-
sintering of mullite ceramics with an Y
2
O
3
addition, Ceram. Int. 27 (2001)
847–852.
[12] H. Ye, Y. Li, J. Sun, Y. Sun, X. Wu, M. Yan, Novel iron-rich mullite solid solution
synthesis using fused-silica and
α
-Al
2
O
3
powders, Ceram. Int. 45 (2019)
4680–4684.
[13] B. L′′OCSEI, Mullite formation in the aluminium fluoride–silica system (AlF
3
–SiO
2
),
Nature 190 (1961) 907–908.
[14] K. Okada, N. Otuska, Synthesis of mullite whiskers and their application in
composites, J. Am. Ceram. Soc. 74 (1991) 2414–2418.
[15] M. Rashad, M. Balasubramanian, Characteristics of porous mullite developed from
clay and AlF
3
⋅3H
2
O, J. Eur. Ceram. Soc. 38 (2018) 3673–3680.
[16] X. Deng, Y. Wu, T. Wei, S. Ran, L. Huang, H. Zhang, F. Li, L. Han, S. Ge, S. Zhang,
Preparation of elongated mullite self-reinforced porous ceramics, Ceram. Int. 44
(2018) 7500–7508.
[17] A. Abdullayev, F. Zemke, A. Gurlo, M.F. Bekheet, Low-temperature fluoride-
assisted synthesis of mullite whiskers, RSC Adv. 10 (2020) 31180–31186.
[18] G. Chen, H. Qi, W. Xing, N. Xu, Direct preparation of macroporous mullite supports
for membranes by in situ reaction sintering, J. Membr. Sci. 318 (2008) 38–44.
[19] A. Doran, L. Schlicker, C.M. Beavers, S. Bhat, M.F. Bekheet, A. Gurlo, Compact low
power infrared tube furnace for in situ X-ray powder diffraction, Rev. Sci. Instrum.
88 (2017) 13903.
[20] L. Schlicker, A. Doran, P. Schneppmüller, A. Gili, M. Czasny, S. Penner, A. Gurlo,
Transmission in situ and operando high temperature X-ray powder diffraction in
variable gaseous environments, Rev. Sci. Instrum. 89 (2018) 33904.
[21] B.A. Dowd, G.H. Campbell, R.B. Marr, V.V. Nagarkar, S.V. Tipnis, L. Axe, D.
P. Siddons, Developments in synchrotron X-ray computed microtomography at the
national synchrotron Light source, in: U. Bonse (Ed.), Developments in X-Ray
Tomography II, SPIE, 1999, pp. 224–236.
[22] D. Legland, I. Arganda-Carreras, P. Andrey, MorphoLibJ: integrated library and
plugins for mathematical morphology with ImageJ, Bioinformatics 32 (2016)
3532–3534.
[23] C21 Committee, Test Methods for Determination of Water Absorption and
Associated Properties by Vacuum Method for Pressed Ceramic Tiles and Glass Tiles
and Boil Method for Extruded Ceramic Tiles and Non-tile Fired Ceramic Whiteware
Products, ASTM International, West Conshohocken, PA.
[24] E.W. Washburn, The dynamics of capillary flow, Phys. Rev. 17 (1921) 273–283.
[25] R.V. Kumar, A.K. Basumatary, A.K. Ghoshal, G. Pugazhenthi, Performance
assessment of an analcime-C zeolite–ceramic composite membrane by removal of
Cr(vi) from aqueous solution, RSC Adv. 5 (2015) 6246–6254.
[26] M. Rashad, M. Balasubramanian, A quantitative analysis of in-situ gases on the
properties of porous mullite developed from clay and AlF
3
⋅3H
2
O, Ceram. Int. 45
(2019) 1420–1423.
[27] X. Delong, L. Yongqin, J. Ying, Z. Longbao, G. Wenkui, Thermal behavior of
aluminum fluoride trihydrate, Thermochim. Acta 352–353 (2000) 47–52.
[28] P.S. Santos, H.S. Santos, S.P. Toledo, Standard transition aluminas. Electron
microscopy studies, Math. Res. 3 (2000) 104–114.
[29] I. Levin, D. Brandon, Metastable Alumina polymorphs: crystal structures and
transition sequences, J. Am. Ceram. Soc. 81 (1998) 1995–2012.
[30] B.R. Cho, J.B. Kang, Synthesis of porous mullite by vapor phase reaction, MSF
449–452 (2004) 209–212.
[31] J.R. Moyer, N.N. Hughes, A catalytic process for mullite whiskers, J. Am. Ceram.
Soc. 77 (1994) 1083–1086.
[32] J.R. Moyer, P.R. Rudolf, Stoichiometry of fluorotopaz and of mullite made from
fluorotopaz, J. Am. Ceram. Soc. 77 (1994) 1087–1089.
[33] D. XIE, L. Yang, L. Li, W. Wang, J. Liu, H. DU, X. Hu, The formation mechanism of
mullite whisker in the mullite fiber network, J. Ceram. Soc. Japan 126 (2018)
529–535.
[34] M. Rashad, U. Sabu, G. Logesh, C. Srishilan, M. Lodhe, A. Joy,
M. Balasubramanian, Mullite phase evolution in clay with hydrated or anhydrous
AlF3, J. Eur. Ceram. Soc. 40 (2020) 5974–5983.
[35] M. Rashad, U. Sabu, G. Logesh, M. Balasubramanian, Development of porous
mullite and mullite-Al2O3 composite for microfiltration membrane applications,
Separ. Purif. Technol. 219 (2019) 74–81.
[36] G. Chen, X. Ge, Y. Wang, W. Xing, Y. Guo, Design and preparation of high
permeability porous mullite support for membranes by in-situ reaction, Ceram. Int.
41 (2015) 8282–8287.
[37] X. Deng, S. Ran, L. Han, H. Zhang, S. Ge, S. Zhang, Foam-gelcasting preparation of
high-strength self-reinforced porous mullite ceramics, J. Eur. Ceram. Soc. 37
(2017) 4059–4066.
[38] A.J. Pyzik, C.S. Todd, C. Han, Formation mechanism and microstructure
development in acicular mullite ceramics fabricated by controlled decomposition
of fluorotopaz, J. Eur. Ceram. Soc. 28 (2008) 383–391.
[39] Li Zhu, Y. Dong, S. Hampshire, S. Cerneaux, L. Winnubst, Waste-to-resource
preparation of a porous ceramic membrane support featuring elongated mullite
whiskers with enhanced porosity and permeance, J. Eur. Ceram. Soc. 35 (2015)
711–721.
[40] R. Ahmad, J.-H. Ha, I.-H. Song, Synthesis of open-cell particle-stabilized Al
2
O
3
foam using Al(OH)
3
particles, Scripta Mater. 76 (2014) 85–88.
[41] J.A. Varela, O.J. Whittemore, E. Longo, Pore size evolution during sintering of
ceramic oxides, Ceram. Int. 16 (1990) 177–189.
[42] H. Schneider, J. Schreuer, B. Hildmann, Structure and properties of mullite—a
review, J. Eur. Ceram. Soc. 28 (2008) 329–344.
[43] M. Mulder, in: Basic Principles of Membrane Technology, second ed., Kluwer Acad.
Publ, Dordrecht, 1996 [u.a.].
[44] B. Xiao, W. Wang, X. Zhang, G. Long, J. Fan, H. Chen, L. Deng, A novel fractal
solution for permeability and Kozeny-Carman constant of fibrous porous media
made up of solid particles and porous fibers, Powder Technol. 349 (2019) 92–98.
[45] T.J. Donohue, C.M. Wensrich, Improving permeability prediction for fibrous
materials through a numerical investigation into pore size and pore connectivity,
Powder Technol. 195 (2009) 57–62.
[46] J. Cai, L. Luo, R.A. Ye, X. Zeng, X. Hu, Recent advances on fractal modeling of
permeability for fibrous porous media, Fractals 23 (2015), 1540006.
[47] J.-H. Ha, S.Z. Abbas Bukhari, J. Lee, I.-H. Song, C. Park, Preparation processes and
characterizations of alumina-coated alumina support layers and alumina-coated
natural material-based support layers for microfiltration, Ceram. Int. 42 (2016)
13796–13804.
[48] N. Das, S. Bandyopadhyay, D. Chattopadhyay, H.S. Maiti, Tape-cast ceramic
membranes for microfiltration application, J. Mater. Sci. 31 (1996) 5221–5225.
[49] M. Abbasi, M. Mirfendereski, M. Nikbakht, M. Golshenas, T. Mohammadi,
Performance study of mullite and mullite–alumina ceramic MF membranes for oily
wastewaters treatment, Desalination 259 (2010) 169–178.
[50] S. Sarkar, S. Bandyopadhyay, A. Larbot, S. Cerneaux, New clay–alumina porous
capillary supports for filtration application, J. Membr. Sci. 392–393 (2012)
130–136.
A. Abdullayev et al.
