PERGAMON

Carbon 39 (2001) 857867

Planar fibre reinforced carbon aerogels for application in PEM

fuel cells
M. Glora, J. Fricke*
R. Petricevic,
Wurzburg

Physikalisches Institut der Universitat

and Bavarian Center for Applied Energy Research ( ZAE), Am Hubland,

D-97074 Wurzburg
, Germany
Received 10 December 1999; accepted 22 July 2000

Abstract
Organic precursor aerogels have been prepared from resorcinol-formaldehyde (RF) with a very high molar ratio of
resorcinol / catalyst of 1500, (RF / total) mass ratios between 30 and 25 wt. % (target density 300 kg m 23 ) and an
F /R-ratio between 1.5 and 3.0. These compositions show minimal shrinkage during ambient pressure drying. Subsequently,
the precursors were pyrolysed at 10508C to form carbon aerogels. The use of fibre additives from ceramic, carbon or
polymeric fibres (novoloid, PAN, aramide) increase the stability of the carbon aerogel matrix drastically and allow to prepare
thin planar sheets with thicknesses between 50 and 300 mm. By using untreated cellulose as the additive, a bimodal pore
structure is generated. During pyrolysis the fibreaerogel compounds were transformed into carbon aerogels exhibiting the
morphology and planarity of the RF precursors. The carbon aerogels were investigated with respect to their structure, using
scanning electron microscopy (SEM), small angle X-ray scattering (SAXS) and gas permeability measurements. The
prepared aerogel sheets show a rather smooth surface and a somewhat coarse interior structure, which is advantageous for
application in fuel cells. For the fibre reinforced gels, the structural changes due to a variation of the composition parameters
is less distinct than for the non-reinforced gels. The structure of the reinforced gel depends strongly on the time, the sol was
allowed to age prior to impregnation. In addition, it was found that the surface of certain fibre materials (e.g. cellulose)
influence the gelation catalytically, while novoloid rather slows the gelation down. The catalytic effect can be described by
a semi-empirical model, involving a catalytic penetration depth. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: A. Carbon aerogel, Electrodes; B. Pyrolysis; C. X-ray scattering; D. Diffusion

1. Introduction
Large efforts have been made in the past ten years to
develop proton exchange membrane (PEM) fuel cells.
Several automobile companies have shown that low-temperature fuel cells can be used as very efficient energy
converters in vehicles. A remaining problem is the high
production costs for the fuel cell aggregates. Compared to
internal combustion engines, which cost less than 50 US-$
per available kW, PEM fuel cells nowadays are prized at
about 5000 US-$ / kW. In order to yield a public acceptance
of fuel cell driven vehicles, their price must not be much
higher than for vehicles driven by combustion engines.
This dramatic cost reduction can be achieved only if
*Corresponding author. Tel.: 149-931-888-5740; fax: 149931-88-5158.
E-mail address: petricevic@zae.uni-wuerzburg.de (J. Fricke).

cheaper proton exchange membranes and efficient diffusion electrodes with very low platinum content become
available.
Carbonaceous gas diffusion electrodes for PEM fuel
cells are required to have a highly porous structure
combined with a high permeability with respect to oxygen,
hydrogen or methanol. In addition, the electrodes should
be available in thin sheets with high mechanical stability.
The oxygen electrode should be hydrophobic in order to
allow for a sufficiently fast release of water.
Carbon aerogels are highly porous, electrically conducting and permeable materials which show the required
properties and thus can be used in fuel cells in principle.
They have been studied for several years by the research
groups at Lawrence Livermore National Laboratory [13]

and the University of Wurzburg

[47]. Fibre-reinforced
carbon aerogels have been investigated and optimized for
fuel cells for the last four years at the University of

0008-6223 / 01 / $ see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0008-6223( 00 )00190-1

858

et al. / Carbon 39 (2001) 857 867

R. Petricevic

Wurzburg.
The precursor is made by condensing resorcinol
(R) and formaldehyde (F) in a molar ratio of F /R in water
with sodium carbonate as a catalyst (C) in a molar R /C
ratio. The concentration of RF is given by the mass ratio
m RF /m total . The process of making RF and carbon aerogels
has been described elsewhere [15]. It was shown that
their structure can be tailored precisely for their use in fuel
cells [48]. The solgel polymerization mechanism of RF
is described in Ref [3]. As shown by Bock [4] and Lu [6],
the size of particles and aggregates within a 5 to 200 nm
range can be tailored by adjusting the R /C and the mass
ratio. Even structures in the micron range can be created
this way [7,9]. The paper deals with the structural modification of planar fibre reinforced carbon aerogels by the
variation of certain synthesis parameters. The objective is
to tailor an asymmetric gas diffusion electrode possessing
a coarse backside and a smooth nanoporous frontside.
While the coarse backside promotes the gas supply to the
catalytic layer, a smooth frontside is necessary in order to
enable a highly dispersed deposition of platinum in contact
with both, electrode and membrane and for a better ohmic
contact with the proton exchange membrane.

2. Experimental
For all samples a RF solution with a R /C ratio of 1500
was used. Structural changes were investigated by varying
the RF mass ratio and the F /R ratio, respectively. At first
the RF mass ratio was varied between 25 and 30% by
keeping the F /R ratio at the standard value of 2.0. The
corresponding samples will be denoted as C25 . . . C30. In
addition a variation of the F /R ratio was performed
between 1.5 and 3.0 by keeping the RF mass ratio at 30 %.
These samples will be denoted as F /R51.5 . . . 3.0. Different fibre sheets were soaked with the liquid sol and
squeezed between two glass plates. This sandwich was
sealed and cured for one day at room temperature, for one
day at 508C and for an additional day at 908C. Deviations
from this standard procedures will be mentioned explicitly.
After aging and an optional exchange of the remaining
pore liquid with acetone, the resulting RF gels were dried
at 508C and ambient pressure. No shrinkage was observed
upon drying. The samples were pyrolysed for 4 hours at
10508C in an argon atmosphere. The linear shrinkage and
mass loss upon pyrolysis were monitored for all samples.
Scanning electron microscopy (SEM) was used to
visualize the macro and large-scale meso structure. The
permeability of the carbon aerogels was determined, as
previously reported [10], by measuring the steady state
flow of nitrogen and helium, respectively, with a fixed
pressure drop of 1000 Pa across the sample for different
mean pressures. The electrodes were glued with epoxy
onto the sample holder rim. The leakage through the rim
was negligible compared to the gasflow through the sample
itself. The permeability measurements give an estimate of

the size of the pores relevant for the gas transport.

Assuming in a zero order approach the pores to be
cylindrical capillaries with a constant diameter d c , the
permeability P can be written as the sum of a viscous and
molecular flow [11], approximated by
]]
f pav
2 2R 0 T
P ] ]d 2c 1 ] ]]d c .
t 32h
3 p Mm

(1)

R 0 is the gas constant, Mm the molar mass, h the viscosity,

pav the average pressure and T the temperature. The ratio
of porosity / tortuosity (f /t ) and the capillary diameter d c
are obtained by a two-parameter fit of the measured
permeability P versus the average pressure pav .
SAXS experiments were carried out at the beamline
JUSIFA [12] at HASYLAB, DESY, Hamburg. A q-range
between 0.1 nm 21 and 8 nm 21 could be covered by using
two different detectorsample distance geometries (3635
mm and 935 mm, respectively) and a corresponding
energy tune for each geometry, namely 13 keV for the
short distance and 7 keV for the long distance. For
correction purposes, empty beam, dark current, reference
and sensitivity measurements were monitored. Thus the
scattering cross section could be represented in absolute
units (cm 2 g 21 srad 21 ).
The scattering cross section was evaluated by applying
the two-phase media model (TPM) [13]. In this model l p
and l s are the mean chord lengths for the pore phase and
for the solid phase, respectively, derived from Porods
constant Km and the invariant Q 5 2p 2 C 2 ( rpart 2 r )
[13,14]. Assuming spherical shapes, l p and l s correspond to
4 / 3 times the mean pore and particle diameter, respectively. (S /m) m is the smooth meso- / macroscopic specific
surface area determined from Porods law, namely ds /
dV 5 2p 2 r 2part (S /m)q 24 5 Km q 24 . For light elements C 5
(NA /Mn )Zr e 5 8.504 ? 10 11 m kg 21 is a constant, including
the classical electron radius r e , the atomic number Z, the
molar weight Mn , and Avogadros constant NA . In case of
carbon aerogels, rpart represents the density of the microporous carbon particles which is assumed to be 1400
kg m 23 [9]. In a first order approximation of the TPM, the
observed q 23.8 -slope was replaced by a q 24 -slope. In order
to separate the meso- and macroscopic contribution (ds /
dV ) m of the secondary particle structure from the primary
micro structure part (ds / dV ) m , the following simple
superposition was used to fit the measured cross section
ds / dV 5 (ds / dV ) m 1 (ds / dV ) m
5 Km q 24 1 Nm Sm (q,r),

(2)

with the form factor Sm of a primary inhomogeneity and a

constant factor Nm proportional to their number density
within a secondary particle. To be discussed is the mesoand macroscopic structure. That is why the primary
structure region (ds / dV ) m was replaced by a continuation
of the q 24 -slope of (ds / dV ) m . Graphically this means to

 et al. / Carbon 39 (2001) 857 867

R. Petricevic

put a smooth envelope around the secondary particles and

treat them as a homogeneous material with an effective
density rpart . Thus, the including micropores will not
contribute to the pore volume and surface, respectively.

3. Results

3.1. Compatible fibres

Since the shrinkage during the preparation (i.e. gelation
and drying) of the RF-aerogel is negligible, this stage of
the preparation of a fibre-reinforced compound is not
critical. To be discussed is the shrinking behaviour and the
shape stability of the compound during pyrolysis, which
determine the compatibility of a fibre with the aerogel
matrix. For a reinforcement of the aerogel matrix only few
types of fibres are useful. In order to prepare homogeneous
electrodes, fleeces, sheets, fabrics or tissues of the fibres
were employed. The fibres showing the best results are
given in Table 1. In addition, the mass loss of the fibres
upon pyrolysis, the linear in-plane shrinkage and the
density of the compound are given. Since there is no
quantitative measurement of the compatibility, the results
of different aerogelfibre compounds with regard to
preparation and stability will be given qualitatively. The
most important criterion is the macrosopic planarity and
homogeneity of the resulting electrode sheets.

3.2. Qualitative comparison of the fibres

For the reinforcement various inorganic and organic
fibres were used. The inorganic ones were ceramic and
carbon fibres exhibiting a high tensile strength, temperature stability and chemical inertness. Because of their
shape stability under pyrolysis conditions the aerogel
matrix is hindered effectively from shrinking. Saffil turns
out to be the best investigated material. Saffil is an
Al 2 O 3 ceramic fibre which is normally used as an insulation material for high temperature applications. The best

859

planarity is obtained with approximately 0.01 kgm 22

Saffil for a thickness of 200 mm. This corresponds to a
fibre fraction of 17 mass %. Saffil adheres strongly to the
carbon aerogel matrix. No other commercial ceramic fibre
has been found, which forms an as superior compound
with carbon aerogels. With regard to stability, planarity and
homogeneity, the only non-organic material showing comparable results is the carbon fibre fleece (SGL Carbon).
Besides its reinforcement property, this carbon fleece also
increases the electrical conductivity of the compounds
approximately by a factor of two.
Glass fibres are not suitable for the reinforcement of
carbon aerogels due to the low melting point of glass and
the reactivity with carbon. In addition, the derived compounds had high densities undesired in fuel cell stacks.
With regard to shrinkage and mass loss organic fibres
show similar properties as RF aerogels, because during
pyrolysis they are transformed into carbon without depolymerization as well. Thereby, fleeces from cellulose
(Royalin ), novoloid (Kynol), PAN (Freudenberg) and
aramide (TFP) showed the best results with regard to
planarity and stability of the corresponding carbon aerogel
compounds. Compounds with PAN and aramide seem to
be similar to the inorganic ones. Compounds with novoloid
fleece possess an increased elasticity, indicating a different
structure. The wettability with the RF sol depends on the
hydrophilic character of the fleece. Cellulose is very
hydrophilic and thus shows good wettability which simplifies the handling during preparation. The mechanical
stability of the aerogel compounds with cellulose is not as
high as for polymeric fibres but still sufficient for the use
as fuel cell electrodes.

3.3. Structure of the compound matrix

The influence of the fibres on the aerogel matrix can be
understood by studying the Saffil aerogel compounds for
example. Fig. 1 shows SEM images of a carbon aerogel
bulk material, a thin carbon aerogel film and a Saffil
aerogel compound film, each with a RF mass ratio of 25 %

Table 1
Characterization of fibres, which gave the best results in combination with an RC1500 / 30% (F /R52) aerogel a
Fibre material

Producer
(type)

Linear
shrinkage

Mass loss
(only fleece)

Thickness
(mm)

Bulk density
(kg m 23 )

Al 2 O 3
Carbon
Aramide
PAN
Novoloid
Cellulose
Cellulose
Cellulose
Cellulose

ICI (Saffil )
SGL
TFP
Freudenberg
Kynol
Royalin
Kleenex
S&S
S&S (RC60)

0%
0%
12%
15%
18%
26%
27%
22%
25%

0%
14%
76%
71%
42%
92%
74%
78%
82%

100
300
100
200
300
200
200
300
100

350
350
350
566
370
494
400
500
450

The uncertainties for shrinkage and mass loss are 62 %, for the thickness and density 620 %.

860

et al. / Carbon 39 (2001) 857 867

R. Petricevic

Fig. 1. SEM images of carbon aerogels as bulk, film and Saffil -reinforced film (S) with two different RF mass ratios. In the C25 film
image, part of the surface skin is visible and indicated by white arrows.

(C25) and 30 % (C30). Upon fibre reinforcement a finer

particle and pore structure of the gel network is generated.
In addition (upper right corner of the non-reinforced C25
film image), all aerogels possess a thin, rather dense
surface skin resembling the smooth surface of the glass
mould, in which the gelation took place. Part of the skin
was removed in order to visualize the inner structure by
SEM.
The mass-specific SAXS cross sections (ds / dV ) vs.
scattering vector q for these samples are plotted in Fig. 2
in a double log plot. For scattering vectors q , 1 nm 21 , a
slope of approximately 2 4 is observed. The compound
aerogel shows a higher scattering intensity in this region,
which is due to a smaller mean particle and pore structure.
A comparable scattering behaviour is observed for the
carbon aerogel film and the Saffilaerogel compound with
nearly the same thickness. The surface skin, which is
present in both samples, seems to determine the scattering
behaviour of the material. This conclusion only holds for
equal sample thicknesses which correspond to similar skin /
sample mass fractions.

Tests of the Saffil aerogel compounds in a 3 cm 2

hydrogen / air fuel cell (at 1 bar / 1 bar, 808C, 0.5
cm 3 min 21 ) show an abrupt voltage drop at current

Fig. 2. SAXS patterns of carbon aerogels as bulk (C30 bulk), film

(C30 film) and Saffil -reinforced film C30 (S).

 et al. / Carbon 39 (2001) 857 867

R. Petricevic
Table 2
Data from SAXS (two-phase media model) and permeability
measurements (capillary model) for the non-reinforced aerogel
films (C30 film, C25 film) and the Saffil -reinforced carbon
aerogels (C30 (S), C25 (S))

l s [nm]
l p [nm]
(S /m) part [m 2 g 21 ]
f /t
d c [mm]
a

C30 film

C25 film

C30 (S)a

C25 (S)

3263
107619
90612

4164
138624
7069

2362
145622
12662
0.2560.1
0.4560.1

2362
201614
127614
0.5360.2
0.5060.1

Standard recipe.

densities of approximately 200 mA cm 22 due to an insufficient water removal in the oxygen electrode (cathode).
In order to enhance the water, it was found that the pores
have to be enlarged. In a first approach, a decrease of the
RF mass ratio was expected to give larger pores, partially
due to sedimentation, according to Ref. [7]. However, such
a sedimentation could not be observed for compounds or
thin films. As Fig. 1 depicts, the particle size within the
inner parts of the C30 film grows from 100 nm to 2 mm for
the C25 film. Thus the particles are slightly larger than for
the corresponding bulk aerogels. In contrast to the nonreinforced films no significant change in particle size can
be observed for the Saffil aerogel compound upon a RF
mass ratio variation (C30 (S) and C25 (S)). In Table 2 the
data derived from the TPM model are given for these
samples. They represent average values of the inner bulk
structure and the surface skin. For the non-reinforced films,
both, l s and l p increase towards lower RF mass ratio, while

861

for the Saffil -reinforced samples only l p increases and l s

remains constant, in accordance with the SEM images.
The values of l s derived for the Saffil aerogel compounds are smaller than for the non-reinforced films, while
the pore size l p is larger. The latter is expected because the
reduced shrinkage of the matrix due to the fibre interaction
leads to a higher porosity. According to the TPM model,
the volume occupied by the fibres has to be treated as pore
volume too, when the scattering of the fibres is either
subtracted appropriately or negligible as in our case. This
gives an additional contribution to l p .
The nitrogen and helium permeabilities P* for the fibrereinforced aerogels and the bulk references, extrapolated to
standard pressure are shown in Fig. 3. The results for f /t
and d c derived from the capillary model are given in Table
2. The permeability of a commonly used E-Tek fuel cell
electrode (Fig. 3) will be used as a standard for a rough
performance classification of the aerogel electrodes. Compared to the reference bulk, the permeability of the
Saffil aerogel compounds is one order of magnitude
lower, in accordance with the previous results. For lower
RF mass ratios the slight increase of the permeability is the
result of an increase of f /t, d c and l p , respectively. The
porosity of the reinforced aerogel series varies around
84 % within the error range of 2 %. Correspondingly, f /t
is mainly determined by the change in tortuosity. Although
the mean pore size seems to be increased upon variation of
the mass ratio, the structure of the corresponding bulk
could not be reproduced with the Saffil -reinforced compounds. Their permeability is also far below the one of
commercial E-Tek electrode material with a porosity of
approximately 60 %. In order to yield a permeability

Fig. 3. Permeabilities at a pressure of 0.1 MPa for the carbon aerogel bulk materials and the Saffil -reinforced carbon aerogels (S) derived
from the RF mass ratio series, and for a commercial carbon paper loaded with platinum (E-Tek, Pt-loading: 0.4 mg cm 22 ).

 et al. / Carbon 39 (2001) 857 867

R. Petricevic

862

Fig. 4. SEM images of Saffil -reinforced carbon aerogels (S) with RC1500 / 30% and different F /R ratios.

comparable to commercial electrode materials, the pore

size of the aerogel matrix has to be increased by a factor
between three and four, according to Eq. (1). An effective
method to do so, has been found by investigating a series,
in which originally the influence of the formaldehyde /
resorcinol (F /R) ratio was to be studied.
Fig. 4 depicts the images of Saffil aerogel compounds
of the second series with different F /R ratios. The aerogel
particles forming around the fibres obviously increase with
higher F /R ratio: the sample F /R53.0 (S) of this series
shows considerably bigger particles than the sample F /R5
2.0 (S).
Another possibility to get larger particles is the aging of
the sol. The effect of such a procedure will be discussed
along obtained SAXS and permeability data for the sample
C30 (S) in Table 2 and the sample F /R52.0 (S) in Table
3. Both samples were prepared according to the same

recipe (indicated by a label), even fibre content and

thickness are comparable. The only difference was the
moment of impregnation. While the fibres of the first series
(variation of the RF mass ratio) were impregnated immediately after the sol was mixed (before the room temperature
step), the fibres of the second series (variation of the F /R
ratio) were soaked after the room temperature step (before
the 508C step). For the sample F /R52.0 (S) l p and l s are
larger by a factor of 3.5 and 3.3, respectively, and d c by a
factor of approximately 6 compared to C30 (S). Thus a
parameter, which was previously believed to be uncritical
proved to provide an effective method to generate large
scale structures in fibre-reinforced carbon aerogel compounds. The nitrogen and helium permeabilities P * of the
fibre-reinforced F /R series in comparison to the nonreinforced bulk series are given in Fig. 5. The permeabilty
is in the same order of magnitude as for the bulk material

Table 3
Data from SAXS (two phase media model) and permeability measurements (capillary model) for the Saffil -reinforced carbon aerogels (S)
with RC1500 / 30%, however different F /R ratios and additionally aged 1 day before RF impregnation

l s [nm]
l p [nm]
(S /m) part [m 2 g 21 ]
f /t
d c [mm]
a

Standard recipe.

F /R51.5 (S)

F /R52.0 (S)a

F /R52.5 (S)

F /R53.0 (S)

6065
343651
4865

7767
503678
3764
0.1260.05
2.8460.57

8467
576691
3464

8167
6416107
3564
0.3860.15
3.6260.72

 et al. / Carbon 39 (2001) 857 867

R. Petricevic

863

Fig. 5. Helium and nitrogen permeabilities at a pressure of 0.1 MPa for the carbon aerogel bulk materials and the Saffil -reinforced carbon
aerogels (S) derived from the F /R series (the latter is additionally aged 1 day before RF impregnation), and for a commercial carbon paper
loaded with platinum (E-Tek, Pt-loading: 0.4 mg cm 22 ).

without fibres. Upon the F /R variation an additional

increase of the particle and pore size is induced, which
makes the permeability performance of the Saffil aerogel
compound comparable and even superior to commercially
available E-Tek electrodes.

3.4. Polymeric fibres

In the following the influence of the various fibre
materials on the structure of the carbon aerogel matrix has
been studied, while the other production parameters are
fixed. Deviations from the standard preparation will be
mentioned explicitly. The investigations were performed
on aerogel sheets reinforced with fibre fleeces from carbon,
aramide, PAN and novoloid. Fig. 6 shows an overview of
the rather dense (not resolvable) surfaces with lower
magnification (top) and the inner bulk structure resolvable
with higher magnification (bottom) after removal of the
surface skin. The rather dense surface skin appears to have
defects. The number of defects seems to be higher if the
supporting structures are coarser. Below the surface skin
carbon fibre, PAN and aramide-reinforced aerogels possess
a structure similar to the previously investigated Saffil reinforced samples while extremely coarse inner structures
appear for the novoloidaerogel compound. For all reinforced samples the aerogel particles are preferentially
bound close to the fibres.
The SAXS data are shown in Fig. 7. The total cross
section is the sum of the scattering contribution of both,
the carbonized fibre and the carbon aerogel matrix. The

SAXS patterns for the separately carbonized fibres are

shown in Fig. 8. Due the small fibre content of about
20 . . . 30 wt.%, the fibre contribution is negligible compared to the scattering cross section of the aerogel. The
carbon fibre, PAN and aramideaerogel compounds
appear to have a scattering behaviour similar to the
Saffil aerogel compound. The quantitative results given
by the TPM model are summarized in Table 4. The
calculated structural parameters of the compounds with
carbon, PAN and aramide fibers are in the same order of
magnitude as for the Saffil -reinforced sample C30 (S)
Table 2). It has to be mentioned that the aerogel reinforced
with carbon fibres was cured for two days at room
temperature (after impregnation!) instead of one. The rise
of the structure is not by far as large as expected for a
room temperature cure before impregnation, but it explains
the slight particle increase compared to C30 (S). The pore
and particle size for the novoloid-reinforced aerogel is
clearly larger than for the other compounds. The corresponding scattering pattern is typical for structures being
on a mm-scale. As for the F /R series the novoloid fibres
were impregnated after the ambient temperature cure.
Comparable is the sample F /R52 (S) (Table 3). The
altered curing explains part of the increase in particle size,
however the main difference to the F /R52 (S) sample is
attributed to an additional fibre influence.
The scattering cross section of the separately pyrolysed
novoloid fibres shows a similar scattering pattern as the
pyrolysed novoloidaerogel compound (Figs. 7 and 8).
Only the porod regime is shifted by a constant factor. This

864

et al. / Carbon 39 (2001) 857 867

R. Petricevic

Fig. 6. SEM images of carbon aerogels (RC1500 / 30%) reinforced with carbon, PAN and novoloid fibres.

implies that the SAXS contribution of the aerogel network

has a similar morphology as the fibres (fibre diameter 13
mm), but only on a smaller scale (l s 3 mm). In contrast to
the other polymers, the use of novoloid as fibre reinforcement enlarges the structures of the aerogel network,
which offers an opportunity to create even larger structures
than with the corresponding non-reinforced bulk material.

3.5. Cellulose

Fig. 7. SAXS patterns of carbon aerogels (RC1500 / 30%, F /R5

2) reinforced with polymeric fiber fleeces, cellulose and Saffil .

Fig. 8. SAXS patterns of the separately carbonized reinforcement

fibres.

A different way to control the structure of the aerogel

matrix is the use of cellulose as reinforcement. Unlike the
case with the previous investigated fibres, the structure of
the compound is not determined by the aerogel composition. The aerogel resembles the cellulose structure by

 et al. / Carbon 39 (2001) 857 867

R. Petricevic
Table 4
Results from the (SAXS) two phase media model for carbon
aerogels (RC1500 / 30%) reinforced with different fibre fleeces

l s [nm]
l p [nm]
(S /m) part [m 2 g 21 ]
a
b

Carbon a

Aramide

PAN

Novoloid b

4664
201628
6367

1862
5567
155619

4766
70610
6169

26286292
70556975
1.160.1

Two days cure at room temperature.

Fibre impregnation after room temperature step.

building a rather dense and homogenous coating on its

inner surface. The SEM image in Fig. 9 depicts aerogel
compounds with an untreated, and thus rather hydrophilic
cellulose and a hydrophobically treated cellulose. Besides
the coating, certain cellulose fibre fleeces (e.g. cellulose
filter papers) yield porous aerogel structures in the mm
regime obtained in the novoloidaerogel compounds. After
subjecting the cellulose to a hydrophobic treatment before
applying the RF impregnation, the rather dense surface
coating resembles an aerogel network obtained e.g. in
Saffil aerogel compounds.
With this kind of solgel deposition, the size of the
cellulose structure can be subsequently adjusted towards
smaller pores. Upon pyrolysis some cellulose materials
show a mass loss of more than 90 %. As a result, merely a
rather dense and crack-free carbon aerogel shell around the
depolymerized cellulose is left. Thus a method was found
to further increase the porosity of carbon aerogel sheets
rather than their mechanical stability. The SAXS curve of a
celluloseaerogel compound is also shown in Fig. 7.
Besides the shoulder at large scattering vectors caused by

865

the primary inhomogeneities, another shoulder appears for

small scattering vectors, which is not present for the
separately pyrolysed cellulose (Fig. 8). This shoulder is
caused by the deposition of the aerogel onto the cellulose
and corresponds to structures on a length scale of several
nanometers.

4. Discussion
An increase of the RF particle sizes upon dilution of the
starting solution is expected for a sol being near the
gelation limit which is the case here and was discussed
previously for non-reinforced RF aerogels [7]. A dilution
of the RF sol can be achieved either directly by lowering
the RF mass ratio, i.e. by adding water, or indirectly by an
increase of the F /R-ratio. The latter can be explained as
follows: The stoichiometric balance of the polycondensation reaction is reached at F /R 2. Thus a further increase
of the F /R-ratio results in a dilution via an excess of
formaldehyde.
The structural changes upon integration of fibres as well
as the occurrence of a surface skin can only be explained
by a catalytic effect due to a locally increase of polycondensation sites (e.g. OH surface groups) provided by
the corresponding interfaces. The influence of the fibres
depends on the respective material. If the RF sol is aged
for one day at ambient temperature before the impregnation is performed, the fibre influence on the aerogel
structure can be more or less suppressed while the visible structure of the surface skin remains unchanged.
In order to understand the correlation between dilution

Fig. 9. SEM images of aerogel compounds with an untreated (hydrophilic) and a hydrophobically treated cellulose.

866

et al. / Carbon 39 (2001) 857 867

R. Petricevic

and fibre-induced effects, a semi empirical model was

developed: In a first approximation, the surface of the glass
mould (in which the gelation took place) as well as the
surface of the reinforcement fibres will be treated as a
generalized surface S surrounding the volume V of the sol.
The polarity of this surface plays an important role. A
hydrophilic surface is equivalent to a locally increased
catalyst concentration (providing more sites for the polycondensation reaction). Accordingly, an intensified nanoscopic surface deposition of RF reactants takes place. We
observed that not only the glass mould behaves like a
hydrophilic surface, but also novoloid and cellulose whereas Saffil , carbon, PAN and aramide appear to be hydrophobic and thus are less favorable with respect to surface
deposition.
In Fig. 10 the observed effects are sketched for different
generalized surface / volume ratios (S /V ). S /V corresponds
to a reciprocal mean distance between two adjacent
surfaces. At very low S /V ratios (S /V 0) the influence of
the surface is negligible. This holds for the case of the
non-reinforced bulk aerogel. At a certain intermediate S /V
ratio given by S /V l 21 the influence of a hydrophilic
surface becomes considerable, where l is a characteristic
catalytic penetration depth. In this case part of the RF
reactants will be consumed via surface deposition. The
remaining diluted RF sol forms large particles according to
the aforementioned dilution effects, consequently gelation
is delayed. This case holds for the novoloidaerogel

compounds and the thin, non-reinforced films. At very

high S /V ratios (S /V 4 l 21 ), the catalytic influence of the
surface is present within the whole volume V of the sol. In
the hydrophilic case the RF groups are consumed completely via surface deposition. This holds for the untreated
cellulose fibres. For a hydrophobic surface a typical
aerogel network is built rather than a deposition. The more
filigree structure compared to S /V 0 indicates a remaining catalytic effect of the corresponding surface. This holds
for the aerogel compounds reinforced with Saffil , carbon,
PAN, aramide and hydrophobically treated cellulose. For
fibre diameters d f < fibre length, the following estimation
of the S /V-ratio can be derived:
S
4 rA
]] ]
,
V d f rf d

(3)

where rf is the bulk density of the fibre, rA weight per unit

area of the fibre fleece and d the thickness of the fibre
aerogel compound. For some of the used fibre materials we
could estimate the S /V-ratios according to this equation,
which are 0.02 mm 21 for novoloid, 0.03 mm 21 for carbon
fibres and 0.05 mm 21 for Saffil . From these data we
conclude that for novoloid l is about 50 mm (using S /V
l 21 ), while for carbon and Saffil l < 20 m m (using
S /V 4 l 21 ) holds. Thus the characteristic catalytical
penetration depths seem to be in the range of approximately 1 mm up to 50 mm.

Fig. 10. Model of a generalized surface (5contact surface between the sol and the boundary1fibers). S /V is the surface / volume ratio of the
system. S /V 0: the surface has no influence on the gel structure: S /V l 21 : the sol is diluted upon the build-up of a rather dense surface
coating, thus between the coated surfaces larger particles form; l corresponds to a characteristic catalytic penetration depth. S /V 4 l 21 : the
catalytic influence of the surface is present within the whole volume of the sol. A strong surface polarity (hydrophilic) causes a deposition
exhibiting inhomogenities in the nm-range rather than a build-up of a typical gel network; the latter is formed for a weak surface polarity
(hydrophobic).

 et al. / Carbon 39 (2001) 857 867

R. Petricevic

5. Conclusion
Planar fibre reinforced carbon aerogels can be tailored
for an optimal structure suitable for asymmetric gas
diffusion electrodes in fuel cells. Especially the observed
fine porous skin above a coarse diffusion layer is advantageous, such a structure allows a finer deposition of
platinum and improves the contact to the proton exchange
membrane (PEM). In addition, the rather coarse structure
behind the skin guarantees a fast release of water in the
cathode. Large progress has been made in explaining the
effects leading to the desired structure of carbon aerogel
electrodes.

Acknowledgements
We are grateful to Drs. M. Lipinski and R. Leuschner /
Siemens AG for cooperation and helpful discussions. Part
of this work was supported by the BMBF (German
Ministry for Research / Bonn).

References
[1] Pekala RW. J Mater Sci 1989;24(9):322131.
[2] Pekala RW, Kong FM. Polym Prepr 1989;30(1):22132.

867

[3] Pekala RW, Alviso CT. Mat Res Soc Symp Proc
1992;270:313.
[4] Bock V, Emmerling A, Saliger R, Fricke J. J Por Materials
1997;4(4):28794.
R, Tillotson T, Geis S, Fricke
[5] Saliger R, Bock V, Petricevic
J. J Non-Cryst Solids 1997;221(2-3):14450.
[6] Lu X, Caps R, Fricke J, Alviso CT, Pekala RW. J Non-Cryst
Solids 1995;188(3):22634.
R, Reichenauer G, Bock V, Emmerling A, Fricke
[7] Petricevic
J. J Non-Cryst Solids 1998;225(1):415.
R, Fricke J. WO99 /
[8] Leuschner R, Lipinski M, Petricevic
27597, 1999.
[9] Saliger R, Fischer U, Herta C, Fricke J. J Non-Cryst Solids
1998;225(1):815.
[10] Reichenauer G, Fricke J. Mat Res Soc Symp Proc
1997;464:34550.
[11] Reichenauer G, Stumpf C, Fricke J. J Non-Cryst Solids
1995;186(1):33441.
[12] Haubold HG, Gruenhagen K, Wagner M, Jungbluth H, Heer
H. Rev Sci Instrum 1989;60(7):19436.
[13] Emmerling
A,
Fricke
J.
J
Non-Cryst
Solids
1992;145(1):11320.
R, Wang P, Beck A, Fricke J. J
[14] Emmerling A, Petricevic
Non-Cryst Solids 1995;185(3):2408.