Acta Materialia 53 (2005) 927–937

www.actamat-journals.com

Polysilsesquioxane derived ceramic foams with gradient porosity

Jürgen Zeschky a,*, Thomas Höfner a, Claudia Arnold a, Rudolf Weißmann a,
Djamila Bahloul-Hourlier b, Michael Scheffler a, Peter Greil a
a
University of Erlangen-Nuernberg, Department of Materials Science, Glass and Ceramics, Martensstrasse 5, 91058 Erlangen, Germany
b
University of Limoges, SPCTS, 123 Avenue Albert Thomas, 87060 Limoges, France

Received 24 August 2004; received in revised form 27 October 2004; accepted 28 October 2004
Available online 30 November 2004

Abstract

Ceramic foams with a gradient porosity were manufactured using a self foaming process of a Si/SiC filled silicone resin in the
temperature range of 220–270
C. Subsequent pyrolysis at 1000
C resulted in Si–O–C micro composite foams with a pronounced
structural anisotropy. The porosity gradient in the foam was controlled by adjusting the polymer melt viscosity during foaming via
temperature. The porosity varied from 40% to 90% within 35 mm. The gradient of the porosity caused variations of the connectivity
density, structure model index and surface-to-volume ratio and influenced the gas permeability and fracture strength.
 2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Ceramic foam; Preceramic polymer; Graded material

1. Introduction [9], filters, membranes and catalyst carriers, supports in

Ceramic foams can be characterized as linked net-
works of irregularly shaped open or closed cell polyhe-
drons, with a low fractional density ranging from 0.05
to 0.50 [1]. The foam formation, however, is a complex
process which involves bubble nucleation, bubble growth
and pore coalescence [2]. The cellular foam structure of-
fers a high mechanical stability and non catastrophic
localized fracture behaviour at low density [3]. Fabrica-
tion of ceramic foams with an isotropic cell structure
has been well established for more than four decades.
Taking advantage of their high gas and liquid permeabil-
ity [4,5] and high thermal resistance [6,7], ceramic foams
are currently applied in various fields including metal
melt filters [8], heating elements in porous gas burners
*
Corresponding author. Tel.: +49 913 185 27561; fax: +49 913 185
28311.
E-mail address: zeschky@ww.uni-erlangen.de (J. Zeschky).
bio reactors [10–14] and sensor elements. Porous ceram-
ics are fabricated most commonly by replica techniques
which involve dip coating of a poly urethane foam or nat-
ural sponge, removing the excessive slurry and subse-
quent burn-out of the preform [15]. Other techniques
include space holder methods where a salt compact is
infiltrated, dried and where the salt is finally removed
by dissolution [16,17]. Gel casting [18–20], powder pro-
cessing [21] and gas bubble formation techniques [22]
are other commonly used manufacturing routes. Re-
cently, polysilsesquioxanes and poly silanes have at-
tracted increased interest for the fabrication of ceramic
foams [23–25]. Poly silsesquioxane was foamed with the
help of poly urethane acting as foaming agent. The foam
cell size and overall porosity were adjusted by adding sur-
factants to the silicone resin [26]. Average pore sizes ran-
ged from 200 to 400 lm and open porosities of 80–90% at
densities of 0.1–0.3 g/cm3 were obtained.
For special applications, however, ceramics with a
graded pore structure are desired. Pore gradient

1359-6454/$30.00  2004 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2004.10.039
928 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937
ceramics offer an interesting potential for being used
as biocompatible implants [27], preforms for liquid
metal infiltration [28–30], thermal barrier coatings
[31] and thermal shock resistant structures [32]. The
fabrication of porous ceramics with graded porosity
is more sophisticated as compared to isotropic ceramic
foams and usually requires multiple manufacturing
steps [33]. Graded ceramic foams were produced by
compacting ceramic powders with varying amounts
of porosity generating additives [33] such as polymers
or carbon [34], using 3D printing techniques [28], elec-
trophoretic deposition [30], stacking of tapes with dif-
ferent porosity [29] or by packing ceramic powders,
fibres or woven preforms to the desired structure
and partially sintering of the scaffolds [34]. Recently,
a method for the fabrication of ceramic foams with
graded porosity by foaming a silicone resin was
reported [35].
A novel manufacturing approach facilitates the fabri-
cation of ceramic foams with both isotropic and/or
graded porosity in one fabrication step [36]. A phenyl
methyl poly (silsesquioxane) containing small amounts
of ethoxy- and hydroxy-groups can be foamed in situ
without requiring additional foaming agent. When
heated to 220–300
C gaseous condensation products
evolve that are formed by the poly condensation reac-
tions of the polymer:
BSi–OH þ HO–SiB ! BSi–O–SiB þ H2 O " ð1Þ
BSi–OC2 H5 þ HO–SiB ! BSi–O–SiB þ C2 H5 OH "
ð2Þ
H2O and C2H5OH act as intrinsic foaming agents and
the foam structure will be stabilized by the concurrent
viscosity increase with proceeding cure of the polymer
melt. The microstructure can be tailored to optimize
the mechanical, magnetic and electrical properties by
loading the polymer with reactive or inert filler particles
[37]. The foaming temperature and thus the curing time
and the polymer melt viscosity can be varied to control
the sedimentation kinetics of the filler powder particles
and the rise of gas bubbles in the polymer melt. After
machining, the thermoset foam has to be pyrolysed to
convert the polymer into a ceramic foam. In the case
of polysilsequioxanes based polymers an amorphous
or partially crystalline silicon oxycarbide network
(Si–O–C) is formed which may contain additional filler
phases [38].
In this work, Si/SiC filled ceramic foams with a gradi-
ent porosity were manufactured by regulating the cross-
linking time and temperature. The porosity structure
was characterized by X-ray micro computer tomogra-
phy and correlated to the permeability behaviour and
fracture strength.
2. Experimental
2.1. Foam preparation
A methyl phenyl poly (silsesquioxane) [(C6H5)0.62-
(CH3)0.31R0.07SiO1.5]n with n
20, R = [–OH] and
[–OC2H5] (Silres H44, Wacker Chemie, Burghausen,
Germany) was used as preceramic polymer. The poly-
mer is solid at room temperature, has a melting interval
of 60–120
C and exhibits a density of 1.1 g/cm3. A
blend of Si (Silgrain, d50 = 7.6 lm, Elkem, Meerbusch,
Germany) and SiC (F500B, d50 = 16 lm, H.C. Starck,
Selb, Germany) was added to the polymer in a poly-
mer:Si:SiC weight ratio of 35:12:53 which corresponds
to a volume ratio of 59:7:34. Si and SiC act as inert fill-
ers when fired below 1000
C and were selected to tailor
the foam morphology, to adjust the desired porosity and
to increase the strength and structure stability of the
foam [38].
The polymer/filler blend was dry mixed in a ball mill
for 2 h. The polymer/filler mixture was filled in an alu-
minium tray and foamed at 245
C for 2 h in air. The
thermoset foam was cut into cylinders (r = 15 mm,
h = 35 mm) and bars (l = 120 mm, w = 20 mm, h = 35
mm). The foams were fired in nitrogen atmosphere
applying a heating rate of 3
C/min to 500
C and a hold
for 2 h. Subsequently, the foams were further heated by
5
C/min to 1000
C with a holding time of 4 h. The
cooling rate to room temperature was set to 5
C/min.
2.2. Viscosity and sedimentation measurements
Thin discs (r = 10 mm, h = 1 mm) were uniaxially
pressed from each 1.5 g of the polymer/Si/SiC blend.
The discs were used for shear viscosity measurements
at constant temperatures of 200–300
C and with
increasing temperature, from the melting temperature
of the polymer to 320
C. A heating rate of 5
C/min
was applied. A rotational rheometer setup with a 1 Hz
oscillating punch was used (Rheometer UDS 200, Phy-
sica Messtechnik GmbH, Ostfildern, Germany). The
complex shear viscosity g was calculated from the stor-
age modulus G 0 and loss modulus G00 .
Sedimentation of the filler powder particles above the
polymer melting temperature was examined using an
optical analyser (Turbiscan Lab Expert, Formulaction,
LÕUnion, France) that quantifies the transmittance and
the backscattering intensity of an introduced laser beam
with a wave length of 880 nm. Using evaluation methods
based on the Mie scattering theory a mean sedimenta-
tion velocity vs of the filler particles in the polymer melt
was calculated. As the maximum measurement tempera-
ture of Turbiscan is 60
C, a 1:1 wt% mixture of Silres
H44 and of the methyltriethoxysilane (MTES)
C7H18O3Si, which acts as a solvent was prepared. At
60
C the mixture has the same viscosity as plain Silres
 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937
H44 at 220
C. A H44:MTES:SiC mixture with a weight
ratio of 30:30:40 was stirred firmly and the velocity of
sedimentation vs of the filler powder particles was mea-
sured for a period of 104 s.
2.3. Structure characterization
The cellular microstructure of the foam was exam-
ined using an X-ray microtomograph (lCT40, Scanco
Medical AG, Bassersdorf, Switzerland). The X-ray tube
was equipped with a tungsten target and operated at
50 kV and 80 lA at a wave length of 0.024 nm. The
specimens were scanned in a near isotropic resolution
of 37 · 37 · 38 lm3. A CCD line array was used to
detect the transmitted intensity through the sample.
X-ray source and detector were covered with slit colli-
mators. The object was rotated through 360
with one
step per degree and the raw data were recorded as sinu-
grams. Generally, various approaches for the 2D recon-
struction of CT data are known that are mathematically
identical but differ in image quality [39]. In this case,
two-dimensional reconstructions of the slices were cal-
culated from the sinugrams by Radon back transforma-
tion using a Compaq DC10 Alpha workstation. From
the 2D images the fractional density qf was calculated
as the ratio of transmitted material in the measured slice
to the complete slice. The subsequent image analysis of
multiple slices facilitated the determination of a porosity
gradient in the examined volume of interest (VOI).
The morphometric structure model index (SMI) was
calculated from the three-dimensional tomography data.
The SMI was introduced for quantification of bone micro
architecture and can be used for foam structure analysis to
characterize the shape and anisotropy of cellular materi-
als as well [40]. The SMI was calculated according to
V  ðdS=drÞ
SMI ¼ 6  ; ð3Þ
S2
where S is the strut surface in a volume V and dS/dr is
the surface area derivative. The foam surface area is ren-
dered by triangulating the surface of a specific VOI with
the Marching Cubes method and the volume is defined
by setting up polyhedrons inside the VOI that match
the bounding surface triangles [41]. A SMI value of 4 de-
scribes spherical pores, any deviation of the spherical
cell geometry results in lower values. A rod cell structure
is characterized by a SMI value of 3 and plate pores
have a SMI of 0. Negative SMI values result from cells
with concave surface areas which is common when coa-
lescence of pores takes place.
The surface/volume ratio, the mean cell size and strut
thickness were calculated using the distance transforma-
tion (DT) method as described in [42]. The connectivity
density CD was computed from CT data using the
Conn-Eulor principle as explained in detail in [43]:
reconstructed 2D CT data images of two neighboured
929
slices are compared by the Boolean Exclusive-Or opera-
tor. The result is superimposed to the original pictures
and analyzed. All new bridges B (new connections),
holes (H) and islands (I) are counted and CD is
calculated:
P P P
Hþ I B
CD ¼  ; ð4Þ
2h  A
where h is the distance of the two slices (h = 38 lm) and
A the examined image area. In the case of determining
the CD of the pores, B and I denote cell windows and
cells, respectively. H, which is isolated strut material
completely surrounded by pores, is unlikely to occur.
The value is identical if CD is calculated for the strut
material, where B denotes new struts, H equal cells
and I isolated struts.
2.4. Gas permeability
The gas permeability of 5 mm thick slices (r = 15 mm)
was measured using a self made gas flow meter device.
The slices were sealed in an Al crucible that allows only
one directional gas flow along the cylindrical axis of the
specimen. The air mass flow q = Dm/Dt through the sam-
ple was adjusted by a digital mass flow controller
(Bronkhorst MFC-F201-AC-AAB-33-V, Bronkhorst
Hi-Tec, AK Ruurlo, Netherlands). The pressure drop
Dp through the sample was measured with a differential
gas pressure sensor (Bronkhorst DP-P506-AAB-33-V,
Bronkhorst Hi-Tec, AK Ruurlo, Netherlands). Accord-
ing to DarcyÕs law
gL q
j¼  ; ð5Þ
A Dp
the permeability j was determined from the q/Dp ratio. L
is the distance of the forced gas flow which corresponds
to the sample thickness (5 mm), A is the cross-section of
the foam (706 mm2) and g denotes the viscosity (air,
25
C, g = 1.83 · 105 Pa s). j, which has the unit m2,
is mostly specified in Darcy (1 D = 0.97 · 1015 m2).
The mass flow rate was set to 2.0 m3/h to ensure a lami-
nar Darcian flow through the foam and to minimize gas
slippage and Klinkenberg effects [44].
2.5. Compression and bending strength
The compression strength was measured at room
temperature for at least 10 cylindrical samples (r = 10
mm). Specimens with graded porosity were cut from
the top coplanarly to a sample height h ranging from
10 to 35 mm, respectively. The foams were loaded in a
universal testing machine (Instron 4202, Instron Corp.,
Canton, MA, USA). A constant crosshead speed of
0.5 mm/min was applied. From the maximum stress
after elastic deformation the crushing strength of the
foam was derived.
930 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937
The bending strength was measured in a four-point
setup (support roller distance: 100 mm, load roller dis-
tance: 50 mm) using the same testing device at a con-
stant crosshead speed of 0.5 mm/min at room
temperature. Bars with a geometry of 120 · 20 · 30
mm3 (length, width, height) were tested. The samples
were loaded as fired with the low fractional density part
of the specimens on the bottom (tensile stress) as well as
with the high fractional density part on the bottom side.
3. Results and discussion
3.1. Foaming
Foam formation by in situ blowing reactions requires
control of viscosity behaviour of the polymer melt. The
dependence of the viscosity of a polymer melt generally
follows an Arrhenius type relation [45]


Eg
gðT Þ ¼ g0  exp ; ð6Þ
RT
where g0 is an empirical constant and Eg is the activation
energy for viscous flow. R is the gas constant and T the
absolute temperature. As shown in Fig. 1, the viscosity
decreases with rising temperature. Heating by 5
C/
min, the polymer is molten completely at 120
C. The
influence of the crosslinking becomes dominant at tem-
peratures above 250
C. In the temperature range of
120–200
C the curve of Fig. 1 can be used to describe
the temperature dependence of the viscosity of the Silres
H44 melt. The numerical fit of Eq. (6) to the measured
data results in


32 kJ=mol
gðT Þ ¼ 0:01  exp : ð7Þ
RT
The activation energy of viscous flow of 32 kJ/mol is
comparable to that of other silicone melts which
100000
10000 Heating rate 5 ˚C/min
Viscosity (Pas)
1000
100
10
100 150 200 250
Temperature (˚C)
Fig. 1. Absolute value of the complex shear viscosity of Si/SiC filled
H44 during heating at 5
C/min from the melting point to 350
C.
depending on the molecular structure typically show
activation energies of 10–100 kJ/mol in the temperature
range below 300
C [46].
At temperatures above 200
C, a dependence on time
is superimposed to Eq. (7) due to the curing of the poly-
mer melt, Fig. 2. The reaction rate of the curing reac-
tions of the poly silsesquioxane during heating was
expressed by the general approach of Borchardt and
Daniel [47]:
da
¼ k  ðT Þ  ð1  aÞn : ð8Þ
dt
Here, da/dt is the curing reaction rate with a denoting
the degree of curing which may attain values between
0 and 1. The degree of curing was derived from experi-
mental data based on a thermo-barometrical analysis
of the gaseous condensation products which evolve dur-
ing the curing reactions [48]. The exponent n describes
the reaction order and k*(T) the specific conversion rate
at the temperature T which follows an Arrhenius type
equation


 Ea
k ðT Þ ¼ k 0  exp : ð9Þ
RT
Here, k0 is a material dependent constant and Ea is the
activation energy of the curing reaction. For the curing
reaction rate the following parameters were calculated
according to Eq. (8): n = 1.78, Ea = 32.7 kJ/mol and
k0 = 0.246. Previous results obtained from Silres H44
showed a reaction order ranging between n = 1 and 2.5
in the temperature interval of 200–300
C [49]. Si/SiC
filled Silres H44 has shown a very similar curing behav-
iour as compared to the unfilled poly silsesquioxane.
Combining the time dependence and temperature
dependence of the polymer melt viscosity was carried
out by applying a modification of MooneyÕs functional
analysis which is an extension of EinsteinÕs viscosity
equation for an infinitely dilute suspension [50]:
100000
10000
Viscosity (Pas)
1000
T = 270 ˚C
100 T = 220 ˚C
10
1
300 350 1 10 100 1000 10000
Time (s)
Fig. 2. Transient shear viscosity of Si/SiC filled H44 during foaming at
220 and 270
C.
 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937 931

p
aðt; T Þ erned by the time and temperature dependence of bub-
gðt; T Þ ¼ gðT Þ  1  ; ð10Þ
a0 ble nucleation, bubble growth and rise and concurring
crosslinking reactions. While bubble formation requires
where a0 is the degree of curing where no more viscous
a low viscosity of the polymer melt, stabilization of the
flow occurs and was set to 0.8. Applying Eq. (8) and
foam structure is caused by viscosity increase. In the
numerically fitting the experimental data of Fig. 2, p
presence of filler particles, heterogeneous nucleation
was calculated to 2.2 ± 0.4 which is in good agreement
is assumed to be the governing bubble formation
with the Einstein coefficient p = 2.5 [51]. Eq. (10) is valid
mechanism [52].
for viscosities up to 106 Pas. Fig. 3(a) shows the calcu-
Once bubbles are nucleated, increase of bubble size
lated values of the viscosity as a function of time and
due to cell growth will occur. Various theoretical ap-
temperature in a smoothed 3D plot. Values of
proaches have been carried out to estimate the bubble
g(T, t) > 106 Pas were set to 106 Pas. In Fig. 3(b), the re-
growth dynamics [53–55]. Generally, bubble growth in
gion of controlled foaming is displayed in the contour
viscoelastic liquids is based on diffusion mechanisms
plot of Fig. 3(a).
that do not follow FickÕs law. Complex models using
Heating the polymer to temperatures above 220
C
several approximations (IDS, infinitely dilute solute;
caused foam bubbles to be generated. Water, ethanol
TBL, thin layer boundary) have been studied. A detailed
and traces of benzene were released upon condensation
description is given in [54]. However, experimental data
reactions of the hydroxy and ethoxy groups of the poly
from Han and Yoo [56] on gas charged molten polymers
phenyl methyl (silsesquioxane). Foam formation is gov-
could be well described by an empirical approach.
Neglecting the solubility of individual gas molecules in
(a) the polymer melt, the bubble radius r tends to be a func-
tion of time t at a temperature T according to [54]:
6.0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5.5 rðT ; tÞ ¼ b  DðT ; tÞ  t; ð11Þ
5.0
4.5
where b is a dimensionless empirical constant. D(T, t) is
as])

4.0 the effective diffusion coefficient of the gaseous species

Viscosity (log[P

3.5 through the polymer melt into the growing bubble that
3.0 is related to the viscosity by the Stokes–Einstein
2.5 relation:
2.0
1.5 kB  T
1.0 DðT ; tÞ ¼ : ð12Þ
5 6  p  gðT ; tÞ  rg
0.5
0.0 4
Here, kB is BoltzmannÕs constant and rg is the radius of
200 3
)
s]

220 the gas molecules (for water and ethanol rg = 1.39 and
g[

240
lo

2 2.89 Å, respectively). Inserting Eqs. (10) and (12) in

e(

Foam 260
m

ing T 280
Ti

emp
eratu 300 1 Eq. (11) and numerically integrating over the time final-
re(˚C ly resulted in the calculation of the maximum bubble
)
size as a function of time, Fig. 4. The calculations were
(b) 5.0 performed for 220 and 270
C using the numerical solver
5.5 Region of gradient foaming of Maple V 7 (Maplesoft Inc., Waterloo, ON, CA). A
4.5 3.7 4.3 5.0
constant water/ethanol ratio of the gases diffusing into
1.8 2.4 the bubbles and a constant gas pressure within the bub-
4.0
1.1
ble were assumed. A critical bubble nucleus size of 0.2
3.5 Region of lm was taken in accordance to literature data [54].
Time (log[s])

0.50 homogenous
3.0 foaming The maximum bubble size radius rmax(a = 1) used for
0.10
normalization was determined experimentally from
2.5
lCT analysis. Values for rmax(a = 1) of 3.7 and 3.1
2.0 mm were derived when foaming was carried out at 220
and 270
C, respectively. When a degree of curing of
1.5
a = 0.8 and thus the end of viscous flow is reached, the
1.0 bubble growth rate is supposed to significantly slow
200 220 240 260 280 300 down as indicated by the kink of the 270
C curve.
Foaming Temperature (˚C) A schematic illustration of the various foaming stages
Fig. 3. Calculated viscosity of the polymer melt as a function of time is given in Fig. 5. A filler containing foam thermoset with
and temperature: (a) displays g(lg(t), T); (b) shows the contour plot of a homogeneous structure is obtained when heated at 270
the viscosity.
C. The rapid viscosity increase up to 1000 Pas/s of the
932 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937

1.2
60
Relative Maximum Bubble Size

1.0 Foaming temperature 245 ˚C

T = 270 ˚C 50

Fractional Density (%)

0.8
rmax(t)/rmax(t→∞)

40

0.6 T = 220 ˚C
30

0.4
20
Foaming temperature 260 ˚C
0.2
10

0.0
0
1 10 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0
Time t (s) Location y/h
Fig. 4. Experimental data of bubble radii as a function of time at 220 Fig. 6. Fractional density of foams with 245 and 260
C curing

C normalized to the maximum bubble radius (t ! 1, a ! 1). The temperature as a function of the sample height; y/h = 0 (top of the
full lines for 220 and 270
C were calculated according to Eq. (11). sample), y/h = 1.0 (bottom).

(a) T = 270 ˚C (b) T = 220 ˚C

foam bubbles (1.0–1.5 mm in diameter) rise by several
I I mm/h at a viscosity of less than 100 Pa. Curing tempera-
tures between 220 and 260
C lead to graded porosity in
Heterogeneous
bubble
the foam thermoset. At 220
C, a linear increase of the
nucleation fractional density ranging from 0.1 at the top to 0.6 at
t = 0 s, = < 10 Pas t = 0 s, < 10 Pas the bottom of the foam thermoset was obtained, Fig. 6.
Bubble nucleation, bubble growth and bubble rise take
Concurrent
II place simultaneously during the initial stages of foaming
bubble rise,
Rapid bubble growth growth and (step II in Fig. 5(b)) leading to an accumulation of pores
during fast viscosity nucleation, no
increase, nearly no significant at the foam surface, step III.
bubble migration
II
viscosity
increase
Polymer derived ceramic foams with a gradient pore
t = 300 s, < 10 Pas structure were also obtained by Colombo and Hellmann
[35] by blowing a mixture of a poly siloxane and a liquid
physical blowing agent. During foaming the viscosity re-
III
Bubble rise mained constant which gave rise for accelerated move-
t = 60 s, = 10 kPas and growth,
increasing ment of large pores to the top. Thus, a pronounced
viscosity
pore size gradient with large pores at the top and signif-
Stiff thermoset with icantly smaller pores at the bottom developed. In con-
t = 1200 s, 10 Pas < < 100 Pas
homogeneous pore
structure
trast, the foaming process in our work is characterized
by a distinctive increase of viscosity over time due to
III IV concurrent curing reactions taking place during bubble
Stiff thermoset
with graded formation and bubble ascent.
porosity
When the bubbles accumulate at the top coalescence
by strut rupture might occur resulting in reduction of
t = 600 s, > 1000 kPas t = 6000 s, > 1000 kPas
bubble number and increase of mean bubble size [57].
Fig. 5. Illustration of the different stages and kinetical effects during For coalescence to take place, however, a strut thickness
foaming at 270
C (homogeneous porosity) and 220
C (gradient below 0.1 lm and a significantly lower viscosity as given
porosity). in the poly silsesquioxane system were reported to be
necessary [58,59]. Hence, it may well be assumed that
polymer at temperatures above 260
C prevents the rise bubble coalescence should not affect accumulation of
of the gas bubbles and the fractional density remains small pores at the top which remain stable in size during
nearly constant over the cross-section of the foam. Cur- further annealing. With prolonged annealing, a pro-
ing temperatures of 270–300
C lead to foams with isotro- nounced viscosity increase by more than three orders
pic structure. At 260
C, a slight gradient is noticeable, of magnitude was observed. As a result, the ratio of bub-
Fig. 6. However, if the foaming temperature is set below ble ascent velocity (StokesÕ law) to bubble growth rate
260
C, curing reactions proceed significantly slower and (Eq. (11)) changes from >1 at low viscosity to 1 at
the initial low viscosity makes bubble rise evident. The high viscosity, Fig. 7. Estimation of bubble size after
 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937 933

10

0.5
dr/dt 1/η
dr/dt, dy/dt (relative units)

0.1

dy/dt 1/η

0 1 2 3
10 10 10 10
Viscosity (Pas)

Fig. 7. Bubble growth rate dr/dt and bubble migration velocity dy/dt
as a function of the polymer melt viscosity.

curing at 200
C for the growth period indicates the bub-

ble size at the bottom being three times larger as com-
pared to the bubble size at the top due to the non-
linear viscosity change as shown in Fig. 2.

3.2. Structure gradient

The fractional density qf(y) as a function of porosity

gradient g(y) can be expressed by
Fig. 8. 3D computer tomography reconstruction of a sample foamed
qf ðgÞ ¼ qf ð0Þ þ ðqf ðhÞ  qf ð0ÞÞ  gðyÞ ð13Þ at 245
C and pyrolysed at 1000
C.

with
 y m Fractional Density (%)
gðyÞ ¼ : ð14Þ 10 15 20 25 30 35 40 45 50
h
Here, y denotes the direction of the gradient and h is the 1.6 Mean cell diameter
sample height [60]. Fig. 8 shows a 3D reconstruction of a 1.4
micro computer tomography of ceramic foam blown at
245
C. g(y) is defined to range from 0 at the top (y = 0) 1.2
of the foam to 1 (y = 35 mm) at the bottom. qf(g(y)) for
Size (mm)

1.0
foaming temperatures 245 and 260
C is given in Fig. 6.
g(y) is linear for both temperatures and thus m = 1. 0.8
However, qf(0) and qf(1) show a dependence on the tem-
0.6
perature. Foaming at 245
C resulted in qf(0) = 0.52 and
qf(1) = 0.07 whereas at 260
C corresponding fractional 0.4
Mean strut thickness
densities of 0.30 and 0.24 as the foam borders were
0.2
reached. From Eq. (13), the overall porosity VP in the
foam can be calculated by 0.0 0.2 0.4 0.6 0.8 1.0
Z h Location y/h
1
VP ¼1  q ðyÞdy: ð15Þ Fig. 9. Mean cell diameter and mean strut thickness as a function of
h 0 f
the location y/h (equal to g(y)).
Here, VP equals 71.5% when foaming was carried out at
260
C and 70.5% for 245
C, respectively. The porosity
gradient is controlled by adjusting the foaming temper- dependence of the mean strut thickness and of the mean
ature while keeping a constant mean fractional density. cell diameter on the fractional density. The mean strut
The connectivity density CD, the surface/volume ra- thickness shows an exponential rise with increasing frac-
tio, the SMI and the mean strut thickness and cell diam- tional density. At g(y) = 0, the struts were 0.2 mm in
eter were correlated to g(y) and qf. Fig. 9 shows the diameter and increased to 1.0 mm at a fractional density
934 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937

of 0.5. The cell diameter increased linearly from 1.0 mm software package), the curve was fitted numerically by
at qf = 0.1 to 1.6 mm at qf = 0.35. In addition, pore coa- an exponential decay function. The dependence of the
lescence took place resulting in larger pores. This is con- surface/volume ratio is expressed by
firmed by the SMI, Fig. 10. At g(y) = 0 a SMI of 3 was S  q 
obtained which describes nearly ideal spherically shaped ¼ 1:53 þ 22:5  exp  f ðmm1 Þ: ð16Þ
V 0:16
cells. The SMI decreased linearly to 1 at qf = 0.45
which denotes a significant deviation of the spherical Fig. 11 shows the dependence of the connectivity density
shape of foam cells. The irregularly shaped foam cells CD on the fractional density qf that can be expressed by
 q 
were confirmed by microscopic and tomographic analy- CD ¼ 0:09 þ 6:81  exp  f ðmm3 Þ: ð17Þ
sis, Fig. 6. The deviation of the cell shape from the the- 0:10
oretically predicted sphere morphology is caused by a CD showed a similar dependence on the fractional den-
combination of processes including pore coalescence sity as the surface/volume ratio. As the theoretical max-
and cell collapse [1]. imum fractional density of fully interconnected open cell
Fig. 11 shows the surface/volume ratio as a function foams is qf = 0.26 [61], it is obvious that the lower part
of the fractional density and of the location g(y) ranging of the samples, which was characterized by
from 16 mm1 at g(y) = 0 to 2 mm1 at g(y) = 1. Using a 0.3 < qf < 0.5, contained isolated closed cells whereas
least squares algorithm (gnufit 1.2, part of gnuplot 3.7 the upper part of the foam showed increasing amounts
of cell windows in the structure. According to PengÕs
relation (Eq. (18)), the diameter d of the cell window
Fractional Density (%)
Pore which connects two foam cells of the same diameter D
10 15 20 25 30 35 40 45 50 Shape
correlates to the fractional density qf [61]:
3 2 0 13 3
p 6 3 3B 1 C7
Structure Model Index SMI

2 qf ¼  pffiffiffi 4 2
 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA 5: ð18Þ
2 1  ðd=DÞ 5 2
1  ðd=DÞ
1

At the top, d/D equalled 0.4 and the cell window diam-
0
eter d  400–500 lm. At qf = 0.2, the window diameter
-1 decreased to d  300 lm, Fig. 12. At qf > 0.26, d/D < 0
and the formation of closed cells was inevitable.
-2
3.3. Properties
-3
0.0 0.2 0.4 0.6 0.8 1.0
Location y/h 3.3.1. Gas permeability
The permeability of porous ceramics is determined by
Fig. 10. Structure model index (SMI) of the pyrolysed foam as a the manufacturing process and pore structure. Ceramic
function of the location. The SMI value at y = 33 mm was not included
in the linear fit due to intense pore coalescence at the surface.
Fractional Density (%)
10 15 20 25 30 35 40 45 50
Fractional Density (%) 120 Calculated Window Diameterd (µm)
10 20 30 40 50 400
3.0 18
100
Gas Permeability κ (D)
Surface/Volume S/V (mm )

2.5 15 300
80
Connectivity Density CD (mm )
-3

2.0 12
60
200
1.5 9
40
100
1.0 6
20
-1

0.5 3
0 0
0.0 0.2 0.4 0.6 0.8
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 Location y/h
Location y/h
Fig. 12. Air permeability and cell window diameter as a function of
Fig. 11. Surface to volume ratio S/V and connectivity density CD of the fractional density. The permeability measurement was carried out
the foam cells in dependence of the location. Points denote the at 25
C at a constant mass flow of 2 m3/h. The cell window diameter
measured values, the curves shows the exponential fit. was calculated.
 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937 935

foams produced by reticulation processes are character- 10

ized by a permeability which is three to four times higher
than the permeability of foams of the same fractional
8
density fabricated by bubble generating techniques.

Bending Stress (MPa)

The relation of porosity and permeability of porous
materials ranging from low porous lime stone to high 6
porous ceramic foams has been studied experimentally
[62,63]. A linear dependence of the permeability on the 4
porosity was noted for low porous salt compacts having
a fully interconnected pore cell network [64]. However,
highly porous materials may show a significant devia- 2
tion of the linear dependence. Percolation Theory sug-
gests a power law relation between the permeability j 0
and the fractional density qf [65]: 0.0 0.1 0.2 0.3 0.4 0.5
Strain ε (%)
j ¼ c  ðqf;cr  qf Þl ; ð19Þ
Fig. 13. Typical stress–strain diagrams for graded porous ceramics
where c and l are empirical constants and qf,cr is the per- loaded in bending.
colation threshold which is the maximum fractional den-
sity that still shows percolation of the pores. Assuming
monomodal spherical pores, the percolation threshold
lies at qf,cr = 0.26 [61]. The permeability as a function
(a) 10
of g(y) was determined by measuring the slices that were Compaction
cut out of the cylindrical sample. The permeability of
each slice represented the corresponding value of the 8
Compression Stress (MPa)

foam part where it was cut out. The permeability

Collapse
k(g(y)) fitted well to Eq. (19) for qf < 0.26 by 6
kðgðyÞÞ ¼ 206  ½0:26  qf ðgðyÞÞ0:42 ; ð20Þ
confirming the dependence of the gas permeability on 4
the cell percolation. The Darcian air flow through the
polymer derived ceramic foam depended on the cell win- 2
dow size: similar to the cell window diameter d the per-
meability j of the open cell foam part decreased
exponentially with increasing fractional density, Fig. 0
0 5 10 15 20 25 30
12. At qf > 0.26, the permeability approached j = 0 as
Strain (%)
no percolation of open cell structure existed any more.
Minimum Fractional Density (%)
3.3.2. Strength 10 20 30 40 50
The bending strength of the porosity graded material (b) 12
depended on the fractional density. Typical stress–strain Equation 23
diagrams are given in Fig. 13. When the low fractional 10
Compression Strength (MPa)

density part of the specimen was loaded in tension and

the high fractional density part in compression, a bend- 8
ing strength of 7.2 ± 1.6 MPa and a failure strain of
0.21 ± 0.06% was obtained. When loaded bottom up, a 6
slightly higher failure stress of 8.5 ± 3.5 MPa and failure
strain of 0.34 ± 0.12% were measured. 4
In the case of compression loading, the specimens al-
ways failed in the foam top region (g(y) = 00.2). The 2
compression strength did not depend on the pore gradient
direction, i.e. if the sample was mounted in foaming direc- 0
tion or bottom up. The specimens always failed in the re- 100 80 60 40 20 0
gion with the lowest fractional density but did not show Fractional sample height (%)
catastrophic fracture behaviour, Fig. 14(a). During load- Fig. 14. Typical stress–strain diagrams for graded porous ceramics
ing the foam sample collapsed in a periodic manner with loaded in compression (a) and compression strength of the porosity
an approximate period of 5%. With increasing strain the graded ceramic foam as a function of the minimum fractional density (b).
936 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937
YoungÕs modulus raised as indicated by the systematic
variation of the slope of the regression line. This is in
accordance to the theoretical prediction that the YoungÕs
modulus of the foam E* is expected to increase with
increasing fractional density qf [66]:
E
¼ C  qnf ; ð21Þ
Es
where C is a geometrical constant set to 1 and n = 1.4 as
in agreement to literature data [67]. Es, the YoungÕs
modulus of the dense strut material is a material-depen-
dent constant. Fig. 14(b) shows the crushing strength rcr
of samples that were ground from the top to different
heights (g(y) = 0.23, 0.34, 0.46, 0.57, 0.68, 0.80 and
0.91) and therefore obtained different minimum frac-
tional densities whereas the maximum fractional density
remained constant at 0.52 for all specimens. Foams with
a minimum fractional density of qf = 0.1 obtained
rcr = 0.9 ± 0.3 MPa. The compression strength in-
creased with increasing minimum fractional density.
At half-height samples g(y) = 0.5, rcr = 4.3 ± 1.5 MPa
at a fractional density of 0.27. This is in agreement with
data from homogeneous ceramic foams of the same
composition and pyrolysis temperature [68]. At 8 mm
high samples with a qf = 0.4, a crushing strength of
8.6 ± 1.3 MPa was achieved.
The fractional strength of a brittle porous material
r*/rs, where r* is the failure strength of the ceramic
foam and of a porous ceramic and rs of the dense strut
material can be expressed as a function of the fractional
density [66]:
r 3=2
¼ C  U2  qf þ C 0 ð1  UÞ  qf ; ð22Þ
rs
U is the fraction of the material located in the struts
whereas 1  U is the material located in the cell faces
of closed cells. U depends on g(y). At g(y) = 0, the foam
is fully open celled and U = 1. C and C 0 are empirical
constants and were set to 0.2 and 0.18, respectively,
according to former results [69]. Eq. (22) can be ex-
pressed as a function of g(y) and qf(y):

[7]
2 3=2
r ðgðyÞÞ ¼ rs  0:2  UðgðyÞÞ  qf ðyÞ
 1992;71:1674.
þ0:18  ð1  UðgðyÞÞÞ  qf ðyÞ ð23Þ
As g(y) and qf(y) depend on T and the foam height,
r*(y) can be controlled by setting the temperature T
and adjusting the height h. The modulus of rupture rs
of the strut material was calculated to be 146 ± 32
MPa at g(y) = 0. At g(y) = 1, rs equals 103 ± 33 MPa.
This is in good agreement to the fracture strength of
dense silicon oxycarbide glass (153 MPa) [70] and of
the modulus of rupture of the strut material of polymer
derived ceramic foam filled with Al2O3 (90 ± 45 MPa)
[71]. The theoretical prediction of Eq. (23) as a function
of the fractional density qf is plotted in Fig. 14(b) and
was confirmed by the experimental data. The structure
independent modulus of rupture rs of the strut material
remains constant at 125 ± 20 MPa.
4. Conclusions
Si–O–C gradient foams were fabricated in a single
step manufacturing process using in situ foaming of por-
ous Si/SiC loaded poly silsesquioxane. A porosity gradi-
ent was generated by variation of the foaming
temperature. While the cell shape and the fractional den-
sity vary linearly with the foam height, the surface/vol-
ume ratio, the connectivity density and the mean strut
thickness scale exponentially. The crushing and the
bending strength are determined by the lowest fractional
density of the material.
Acknowledgements
The authors thank Rebecca Voigt for assistance at
rheology measurement and the Deutsche Forschungs-
gemeinschaft (DFG) under SCHE-628/1–2 for financial
support. The Institute of Science and Technology of
Metals of the University of Erlangen-Nuernberg is
gratefully acknowledged for providing gratuitous access
to their micro computer tomography. The authors wish
to thank the MAE, France, and the DAAD, Germany,
for travel support under PROCOPE 312-D/1022890.
References
[1] Weaire D. Adv Eng Mater 2002;4:723.
[2] Arefmanesh A, Advani SG. Polym Eng Sci 1995;35:252.
[3] Brezny R, Green DJ. J Am Ceram Soc 1989;72:1145.
[4] Innocentini MDM, Salvini VR, Pandolfelli VC, Coury JR. Am
Ceram Soc Bull 1999;77:78.
[5] Lopes RA, Segadães AM. Mater Sci Eng A 1996;209:149.
[6] Orenstein RM, Green DJ. J Am Ceram Soc 1992;75:1899.
She J, Ohji T, Deng Z-Y. J Am Ceram Soc 2002;85:2125.
[8] Woyansky JS, Scott CE, Minnear WP. Am Ceram Soci Bull
[9] Durst F, Weclas M. P I Mech Eng D-J Aut 2001;215:63.
[10] Binner JGP, Reichert J. J Mater Sci 1996;31:5717.
[11] Zimmerlin B, Müllner I, Leibold H, Seifert H. In: Heinrich JG,
Aldinger F, editors. Nano-scaled ceramic membranes for the
filtration of fine and sticky dust. Weinheim (Germany): Wiley–
VCH; 2001. p. 79.
[12] Carty WM, Lednor PW. Curr Opin Solid State Mater Sci
1996;1:88.
[13] Ambrogio M, Saracco G, Specchia V. Chem Eng Sci
2001;56:1613.
[14] Engin NÖ, Tas AC. J Eur Ceram Soc 1999;19:2569.
[15] Schwartzwalder K, Somers AV. Method for making porous
ceramic articles, US Patent No. 3.090.094 (May 21, 1963).
[16] Fitzgerald TJ, Michaud VJ, Mortensen A. J Mater Sci
1995;30:1037.
 J. Zeschky et al. / Acta Materialia 53 (2005) 927–937 937

[17] Fitzgerald TJ, Mortensen A. J Mater Sci 1995;30:1025. [47] Borchardt HJ, Daniels F. J Am Chem Soc 1956:41.
[18] Sepulveda P. Am Ceram Soci Bull 1997;76:61. [48] Wild MJ, Buhler P. J Mater Sci 1998;33:5441.
[19] Fujiu T, Messing GL, Huebner W. J Am Ceram Soc 1990;73:85. [49] Scheffler M, Roisman TG, Takahashi T, Kaschta J, Buhler P,
[20] Ortega FS, Sepulveda P, Pandolfelli VC. J Eur Ceram Soc Greil P. [to be published].
2002;22:1395. [50] Mooney M. J Colloid Sci 1951;6:162.
[21] Monanaro L, Jorand Y, Fantozzi G, Negro A. J Eur Ceram Soc [51] Laven J, Stein HN. J Rheol 1991;35:1523.
1998;18:1339. [52] Baldwin DF, Park CB, Suh NP. In: Kumar A, Sealer VA, editors.
[22] Langguth K. Ceram Int 1995;21:237. A processing system for the extrusion of microcellular polymer
[23] Nangrejo MR, Bao X, Edirisinghe MJ. J Eur Ceram Soc sheets: shaping and cell growth control. New York: ASME;
2000;20:1777. 1994. p. 85.
[24] Bao X, Nangrejo MR, Edirisinghe MJ. J Mater Sci 1999;34:2495. [53] Arefmanesh A, Advani SG, Michaelides EE. Polym Eng Sci
[25] Colombo P, Modesti M. J Am Ceram Soc 1999;82:573. 1990;30:1330.
[26] Colombo P, Griffoni M, Modesti M. J Sol–Gel Sci Technol [54] Venerus DC, Yala N, Bernstein B. J Non-Newt Fluid Mech
1998;13:195. 1998;75:55.
[27] Tampieri A, Celotti G, Sprio S, Delcogliano A, Franzese S. [55] Amon M, Denson CD. Polym Eng Sci 1984;24:1026.
Biomaterials 2001;22:1365. [56] Han CD, Yoo HJ. Polym Eng Sci 1981;21:518.
[28] Moon J, Caballero AC, Hozer L, Chiang Y-M, Cima MJ. Mater [57] Jashnani IL, Lemlich R. Ind Eng Chem Fundam 1974;14:131.
Sci Eng A 2001;298:110. [58] deVries AJ. Mededel Rubber-Sticht 1957;321:91.
[29] Corbin SF, Zhao-jie X, Henein H, Apte PS. Mater Sci Eng A [59] Brady AP, Ross S. J Am Chem Soc 1944;66:1348.
1999;262:192. [60] Bao G, Cai H. Acta Mater 1997;45:1055.
[30] Put S, Vleugels J, Briest OVd. Acta Mater 2001;45:1139. [61] Peng HX, Fan Z, Evans JRG, Busfield JJC. J Eur Ceram Soc
[31] Steffens H-D, Babiaka Z, Gramlicha M. J Therm Spray Technol 2000;20:807.
1999;8:517. [62] Dullien FAL. Porous media: fluid transport and pore struc-
[32] Satyamurthy K, Hasselman DPH. J Am Ceram Soc 1979;62:431. ture. San Diego: Academic Press; 1992.
[33] Suresh S, Mortensen A. Fundamentals of functionally graded [63] Scheidegger AE. The physics of flow through porous
materials. London: The Institute of Materials; 1998. media. Toronto: University of Toronto Press; 1957.
[34] Maca K, Dobsak P, Boccaccini AR. Ceram Int 2001;577. [64] Cinar Y, Pusch G, Reitenbach V. Analysis of the pore structure of
[35] Colombo P, Hellmann JR. Mater Res Innov 2002;6:260. compacted salt material based on the measurement of hydraulic
[36] Gambaryan-Roisman T, Scheffler M, Buhler P, Greil P. Ceram properties. Orlando (FL, USA): Solution Mining Research Insti-
Trans 2000;108:121. tute; 2001.
[37] Greil P. J Am Ceram Soc 1995;78:835. [65] Sahimi M. Applications of percolation theory. London: Taylor
[38] Zeschky J, Höfner T, Lo JSH, Scheffler M, Greil P. In: Kainer & Francis; 1994.
KU, editor. High strength Si–O–C ceramic foams for the [66] Gibson LJ, Ashby MF. In: Clarke DR, Suresh S, Ward FRS,
reinforcement of Mg-alloys. Wolfsburg (Germany): Wiley– editors. Cellular solids: structure and properties. Cam-
VCH; 2003. p. 378. bridge: Cambridge University Press; 1997.
[39] Kulkarni A, Sampath S, Goland A, Herman H, Dowd B. Scripta [67] Colombo P, Hellmann JR, Shelleman DL. J Am Ceram Soc
Mater 2000;43:471. 2001;84:2245.
[40] Hildebrand T, Rüegsegger R. Comput Meth Biomech Eng [68] Höfner T, Zeschky J, Scheffler M, Greil P, Light weight ceramic
1997;15. sandwich structures from preceramic polymers. Cocoa Beach (FL,
[41] Guilak F. J Microsc 1994;173:245. USA): The American Ceramic Society [p. to be published].
[42] Hildebrand T, Rüegsegger R. J Microsc 1997;185:67. [69] Zeschky J, Lo JSH, Scheffler M, Hoeppel HW, Arnold M, Greil
[43] Gundersen HJG, Boyce RW, Nyengaard JR, Odgaard A. Bone P. Z Metallkd 2002;93:812.
1993;14:217. [70] Renlund GM, Prochazka S, Doremus RH. J Mater Res
[44] Innocentini MDM, Pandolfelli VC. J Am Ceram Soc 2001;84:941. 1991;6:2723.
[45] Rochow EG. In: Rochow EG, editor. Silicon and silicones. Ber- [71] Zeschky J, Scheffler M, Colombo P, Greil P. Cellular oxide ceramics
lin: Springer; 1987. from filler loaded silicone resins. In: Lin H-T, Singh M, editors.
[46] Noll W. In: Noll W, Hecht G, editors. Chemistry and technology Ceramic engineering and science proceedings, vol. 23. Cocoa
of silicones. Weinheim: Verlag Chemie (in German); 1968. Beach (USA): The American Ceramic Society; 2002. p. 285.