Freezing/thawing curves and characterization of porous network by a

capacitive sensor based apparatus

Antonin Fabbri, Teddy Fen-Chong & Olivier Coussy

Institut Navier LMSGC, France

ABSTRACT: A capacitive sensor-based apparatus is used to study the ice-water phase change in consolidated
porous media subject to freezing and thawing. This technique relies on the dielectric properties of water, ice,
air, and mineral substrate in the radio-frequency range. A semi-empirical method based on the Lichtenecker
model and on comparison between drying and freezing tests, provides an accurate estimation of the temperature-
dependent liquid water amount in frozen porous materials. This is then analyzed through thermoporometry
concepts to characterize pore size distribution and specific surface area. The results obtained are roughly in the
same order range than those obtained from gravimetric sorption/desorption isotherms.

1 INTRODUCTION

Damage induced by frost action upon concrete struc-

tures is a source of main concern in cold climates
(Pigeon 1984; Dash, Fu, and Wettlaufer 1995). The
mechanical response of a porous material is cred-
ited to result from the combination of the liquid-solid
expansion, the transport of unfrozen liquid water
through the porous network, and the presence of air
voids (Wang, Monteiro, Rubinsky, and Arav 1996).
A poromechanics-based approach has been recently
worked out to understand and quantify the phenomena
both at the pore scale (Coussy and Fen-Chong 2005)
and at the material scale (Coussy 2005). Whatever the
approach carried out, the freezing/thawing curve, that Figure 1. Schematic diagram of the capacitive sensorbased
is the saturation degree of unfrozen water versus tem- apparatus.
perature, turns out to be the key curve governing the
mechanical behaviour of porous materials upon frost
action. Freezing of cement based materials is usually similar materials have been described in (Fen-Chong,
studied with low temperature calorimetry (Bager and Fabbri, Guilbaud, and Coussy 2004). A preliminary
Sellevold 1986). However, this device only allows to work on freezing and thawing study of cement paste,
test very small specimens (characteristic length round stones, and fused glass beads, has been presented in
the millimeter) which are often crushed. Hence, tests (Fen-Chong, Fabbri, and Gaulard 2005).
on non-crushed heterogeneous materials like mortar The capacitive method is based on the contrast
and concrete (aggregate characteristic length round the between water (about 80) and ice (about 3) dielec-
centimeter) should be difficult to set up. tric constants in the radio-frequency range. A sample
Alternatively, it is possible to use a dielectric is inserted between two plane and circular stainless
method, which is less expensive and allows to test big- steel electrodes connected to a cryostat as sketched in
ger and heavier specimens. This paper presents the use figure 1. These are connected in parallel to a 30 MHz
of a spectroscopic-like dielectric capacitive method 50 MHz oscillator circuit, which forms an oscillat-
for studying the ice/water phase change in cohesive ing circuit. The experimental apparatus measures the
porous media such as cement paste, mortar and con- reduced frequency ( f ), which depends on the sample
crete. The experimental set-up, calibration, and use capacitance. Because a plane capacitance is propor-
for studying the drying (air/water phase change) of tionate to the real dielectric constant of the sample,

567
Copyright 2005 Taylor & Francis Group plc, London, UK
any liquid water variation inside the porous network
of the material results in resonant frequency variation.
2 LIQUID SATURATION DEGREE AND
DIELECTRIC CONSTANT
The experimental apparatus gives the f T relation
that can be expressed as a T relation, with T the
electrodes temperature and the dielectric constant
of the sample. To get the freezing curve, a rela-
tion between and the liquid water saturation ratio
Sw must be established. To do so, after a review of
some classical multi-scale schemes (Zakri, Laurent,
and Vauclin 1998; Fen-Chong, Fabbri, Guilbaud, and
Coussy 2004), the Lichtenecker model was chosen.
This model comes from the Effective Medium The-
ory in which the self-consistent scheme is adopted.
It assumes that the geometrical shape of ellipsoidal
inclusions follows a probability density beta function,
characterized by a parameter () ranging between 1
and 1. This finally gives (the proof is summarized in
reference (Zakri, Laurent, and Vauclin 1998):
When = 1 the pores (void inclusions) are parallel to
the external field, while for = 1 they are perpendic-
ular to the external field. The previous expression does
not apply for = 0 for which no particular shape nor
orientation is privileged with regard to the direction
of the external field. For such a completely disor-
dered medium the Lichtenecker model provides in
the form:
Let consider the pore volume filled both by liquid
water, with dielectric constant w , and by a non wet-
ting component (ice or air), with dielectric constant
nw to be an effective medium. If no particular pore
shape nor orientation is assumed to be privileged, the
overall dielectric constant constant w,nw of the porous
material can be expressed from (1) and (2) as:
So as to evaluate , drying test at 20 C were performed
on sandstones, fused glass beads and W/C = 0.5 ordi-
nary Portland cement paste. As sketched on figure 2,
a linear relation is found between and Sw , which
is well accounted for by letting = in (3), thus
we get:
Copyright 2005 Taylor & Francis Group plc, London, UK
Figure 2. Comparison between measured and calculated
dielectric constants of a non-saturated porous medium. The
calculation is made using the three-phase Lichtenecker model
with = in (3), thus leading to (4).
3 TESTED MATERIALS
Cement paste with water-cement ratio of 0.5 are pre-
pared with a 5-liter mortar mixer and cast in cylinder
moulds with a 40 mm diameter and 100 mm high.
Ordinary Portland Cement CEM I CPA 52.5 N CP2
and distilled water is used. One day after casting,
specimens are removed from their mould and stored
in moist condition (relative humidity rh = 95% 5%)
during 6 months. Then they are cut in about 15 mm
slices before being put in water until testing. Speci-
mens were saturated with degassed distilled water at
3 kPa air pressure and tested. Before saturation, some
specimens (index d) were dried in an oven at 70 C
3.1 Freezing-thawing curves
The equation (4) was obtained at = 20 C at which
the matrix and pore water dielectric constants, m and
w , are both known. Actually, the in-situ dielectric con-
stant of in-pore liquid water varies with ionic strength
(Barthel, Hetzenauer, and Buchner 1992) and adsorp-
tion (Mercury, Vieillard, and Tardy 2001), which in
turn depends on temperature and saturation. Further-
more the change in the dielectric constant of non-
evaporable water with temperature will significantly
affect the matrix dielectric constant. As a consequence
a further relation is needed between w,nw, Sw and .
The in-pore water of cement paste in the range of low
liquid saturation degrees Sw < 0.5 was experimentally
found not to freeze down to 45 C. This was previ-
ously observed by (Kaufmann 1999). Assuming that
no pore water evaporation takes place during cooling/
heating cycle on such low saturated samples in the
temperature range [0 C; 45 C], it becomes possi-
ble to experimentally determine the missing relation
between the (initial) liquid water degree saturation, the
temperature, and the sample dielectric constant w,a ,
where index a refers to air.
568
3.2 Dielectric constants of unsaturated cement
pastes
The changes in the dielectric constant w,a with respect
to temperature were first determined at liquid satura-
tion degree Sw held constant. Samples with different
liquid saturation degrees Sw were prepared by con-
serving initially dried specimens in an oven at 70 C
during 6 months, within atmospheres at various con-
trolled humidity rh = 12%, 33%, 44%, 85%, and 100%
by using salt solutions (Raoof 1998). This procedure
provided samples with liquid water saturation degrees
Sw = 0.12, 0.17, 0.18, 0.49 and 1 respectively. The
sample dielectric constant w,a was then determined
from 20 C to 5 C for Sw = 100%, and from 20 C
to 40 C for Sw = 0.12, 0.17, 0.18, 0.49. It must be
again stressed out that these temperature ranges were
allowed since primary freezing occurs only at 5 C for
saturated cement pastes, and no freezing is observed Figure 3. Evolution of the apparent dielectric constant and
for unsaturated cement pastes with Sw < 0.5 before its derivation with temperature at a fixed saturation ratio for
45 C. Whatever the porous material tested here, the cement paste P5-1d.
relation between the dielectric constant w,a and the
temperature was found to be linear (see figure 3A for
cement paste P5-1d), so that the corresponding slope
a(Sw ) is a function of liquid saturation only. Further-
more, as reported in figure 3B, a(Sw ) curve was found
to be a line. We finally write:

with a  0.225 C1 and b  0.002 C1 .

Irrespective to the previous experiments the
changes in the dielectric constant w,a with respect to
liquid saturation Sw were then determined at temper-
ature held constant. By varying the relative humidity
rh, drying tests were performed at different tempera-
tures (6 drying tests from 20 C to 5 C). As found
in (Fen-Chong, Fabbri, Guilbaud, and Coussy 2004)
and reported in figure 4A for cement paste P5-1d, all
tests provide a linear relation between the dielectric
Figure 4. Evolution of the apparent dielectric constant and
constant w,a and the liquid saturation degree Sw . As
its derivation with saturation at a fixed temperature for
a consequence, the corresponding slope b is a func- cement paste P5-1d.
tion of temperature only. Furthermore, as reported
in figure 4B, b () is found to be a line. We finally
write: where coefficients a, b, c, d are specific to the tested
porous material. Their determination requires two tests
only: cooling of a fully saturated sample (Sw = 1) down
to the primary freezing, and a cooling test of a dried
with a  0.227 C1 and c  20.158. sample (Sw = 0) down to 45 C.
2 w,a
The Maxwell symmetry-like relation Sw
= Then combining (4) and (7) the freezing/thawing
2 behaviour of porous media can be described by:
Sw
a  a = 0.23 so obtained confirms the
w,a

existence of a material relation linking w,nw , Sw and

only. This was also observed for stone and fused bead
glass samples. For the tested porous materials, we can
where A = a b(i 1); B = bi ; C = c d(i 1)
eventually write:
and D = di . Once the changes in w,i with
respect to experimentally determined temperature,

569
Copyright 2005 Taylor & Francis Group plc, London, UK

Figure 5. Freezing (upper graph) and thawing (lower graph)
of W /C = 0.5 cement paste submitted to different curing
conditions.
semi-empirical relation (8) gives the means to deter-
mine the thermodynamic state function linking Sw to .
3.3 Results
Figure 5 shows the estimated liquid water saturation
ratio variation with temperature during the cooling
and heating stages of cement paste samples. Satura-
tion varies almost linearly with temperature, except
around some particular temperatures where the slope
changes significantly. The linear saturation variation is
related to the progressive freezing (resp. thawing) of
the water (resp. ice) contained in the connected pores
which are sufficiently big (resp. small) to freeze (resp.
thaw). The saturation slope change is due to an impor-
tant water/ice phase change. For P5-1d, an important
freezing peak is observed around 8 C and a minor
second one around 43 C. The presence of these two
peaks is in good agreement with results reported for
the same kind of paste by (Kaufmann 1999; Bjaoui,
Revertegat, and Bournazel 2002). On P5-2, the first
freezing peak around 8 C is almost absent while two
others (around 28 C and 43 C) are much more
pronounced. Similar results were obtained on virgin
pastes by (Sellevold and Bager 1980). Our data are then
compared with low temperature calorimetry ones on
the same kind of specimen. Comparison on ice content
at each freezing peak between our data and (Sellevold
and Bager 1980) ones leads to measured relative dif-
ferences smaller than 20%. Hence, a good agreement
is found.
570
Copyright 2005 Taylor & Francis Group plc, London, UK
4 POROUS NETWORK CHARACTERIZATION
4.1 Relation between freezing temperature and
pore distribution
Both LMSGC home-made gravimetric sorption/
desorption apparatus (SD) and thawing curve analy-
sis via thermoporometry using capacitive sensor based
device (CS) allow to reach the pore distribution from
20 to 200 . For SD, the pore radius distribution was
obtained from a sorption Sw pc (rh) capillary curve
of a 30 mg crushed sample using the BJH method,
detailed in (Barrett, Joyner, and Halenda 1951). For
CS, it was obtained from a thawing test using thermo-
dynamical and mechanical equilibrium between ice
crystal and liquid water. Thawing test is preferred
to freezing one in this study because metastabil-
ity phenomena occurs in the latter (Scherer 1993).
Thermodynamical equilibrium induces the equality of
their chemical potentials. Neglecting the variation of
ionic concentration of the liquid solution, compressive
heat influence, variation of heat capacity and assum-
ing density of ice and water be equal and constant
(i = w = 0 ) basic thermodynamics gives:
with T0 the bulk melting temperature, pc the capillary
pressure defined as the difference of pressure from the
less wettable phase to the more wettable (pc = pi pw ),
Sf = 0 (sw0 si0 ) the entropy of fusion per unit of crys-
tal volume Cf = 0 (cp,w cp,i ) the heat capacity differ-
ence between water and ice per unit of crystal volume,
and g the function which links univocally pc to T .
From now, pores will be supposed to be cylindrical.
In this case, ice crystal thaws from cylindrical interface
(Setzer 1997). Thus, mechanical equilibrium of the
ice/water interface (Young-Laplace law) gives:
where is the water/ice interface energy, rp the pores
radii and e the unfreezable water layer length. Combi-
nation of (9) and (10) finally gives the relation linking
the thawing temperature and pore radius in which
melting occurs:
As Sw is linked to the volume of frozen pores Vi by the
relation Sw = 1 Vpores
Vi
, (11) allows to easily express
the pore radius distribution curve (rp ) = dr
dvi
p
(rp ) from
the Sw T curve. Samples used for comparison are of
 3 Table 1. Comparison between surface area
Capacitive device
Sorption device
2.5
2 Paste
dVi /drp [mm3/A/g]
1.5
0.5
0
0 20 40 60 80 100 120 140
Pore radius [A]
Figure 6. Comparison between pore radius distribution
from thermoporometry using CS and from sorption isotherm
using SD for W /C = 0.5 cement paste.
the same batch and conserved in the same condition
(P5-1ds conditions). Numerical applications are made
with: Sf = 1, 2 MJ/(Kg.K), Cf 2, 1 MJ/(Kg.K),
e = 9 (Brun, Lallemand, Quinson, and Eyraud
1977) T0 = 270 K (estimated from Raoult Law) and
= 36 + 0.25(T T0 ) (Zuber and Marchand 2004).
Results obtained by SD or CS are shown in figure 6.
From experimental results, our dielectric device seems
to be precise enough to characterize the fine porous
distribution of a cementitious material.
4.2 Determination of freezable porous network
surface area
The specific surface area of an elementary cylindri-
cal pore with radius rp can be expressed as (Quinson,
Astier, and Brun 1987):
with
 the dry density of the porous medium and
d wi the ice/water surface per unit of porous medium
volume expressed from Laplace Law as:
Note that (13) assumes that the solid matrix is unde-
formable. Combining (12) and (13), the explored
specific surface area can be evaluated by the relation:
Measurable specific area of all freezable porous
network (A in m2 /g) will be the value of for
Copyright 2005 Taylor & Francis Group plc, London, UK
values from capacitive sensor based apparatus
(CS), MIP and sorption/desorption test (SD) in
m2 /g.
MIP CS SD
P5-1d 16.81 82.38 145.46
Sw Sw, min , with Sw,min the volumic proportion
of unfreezable water. Thus, for cylindrical assumed
pores, using (14):
160 180 200
Specific area values from capacitive sensor based
apparatus (CS) experimental data using (15) are com-
pared with MIP and sorption/desorption test ones for
an W/C = 0.5 cement paste. The used specimens are
from the same batch and are cured in the same way.
MIP values are obtained using the Rootare and Pren-
zlow equation (Rootare and Prenzlow 1967). Sorption/
desorption test ones are from BET theory analy-
sis (Brunauer, Emmett, and Teller 1938). Numerical
applications are made with = 1.62 g/cm3 values
given in 4.1 . Results are reported in table 1.
The CS values appear to be strongly higher than
MIP ones. This large gap can be explained by the dam-
age caused by the high pressure needed for the MIP
experiment and the inability of the latter to scan finer
porosity. Actually the mercury-injection porosimetry
technique is known to underestimate the specific area
(Raoof 1998). Furthermore, the lower values provided
by the CS experiment by comparison to the SD exper-
iment can be attributed to the screening effect of
unfreezable water with regard to the roughness of the
internal walls of the porous network. In addition, the
CS cooling test is restricted to 45 C whereas ice
formation is still observed down to 80 C (Jehng,
Sprague, and Halperin 1996).
5 CONCLUSION
A spectroscopic-like dielectric capacitive apparatus
was set-up in order to experimentally determine the
thermodynamic state relation relying the ice content
on the current temperature. This quantitative method
has been successfully applied to centimetric cement
paste samples by combining the results provided
by drying tests and freezing tests. The experimen-
tal results agree well with the data existing in the
literature obtained independently on the same kind
of specimens by low temperature calorimetry. Pore
571
radii distribution and specific surface area are found
to be consistent with those obtained on similar sam-
ples from sorption/desorption isotherms by means of
a homemade apparatus.
REFERENCES
Bager, D. and E. Sellevold (1986). Ice formation in hardened
cement paste - part II : Drying and resaturation on room
temperature cured pastes. Cement and Concrete Research
16, 835844.
Barrett, E., L. Joyner, and P. Halenda (1951). The determi-
nation of pore volume and area distributions in porous
substances. i : Computations from nitrogen isotherms.
Journal of the American Chemical Society 73, 373380.
Barthel, J., H. Hetzenauer, and R. Buchner (1992). Dielec-
tric relaxation of aqueous electrolyte solutions. i-solvent
relaxation of 1:2, 2:1 and 2:2 electrolyte solutions.
Berichte der Bunsen-Gesellschaft Physical Chemistry
Chemical Physics 96, 988997.
Bjaoui, S., E. Revertegat, and J.-P. Bournazel (2002).
Mcanismes de formation de la glace au sein des ptes de
ciment et betons. Revue franaise de Gnie Civil 6(78).
Brun, M., A. Lallemand, J. Quinson, and C. Eyraud (1977).
A new method for the simultaneous determination of
the size and the shape of pores : The thermoporometry.
Thermophysica Acta 21, 5988.
Brunauer, S., P. Emmett, and E. Teller (1938). Adsorption
of gases in multimolecular layers. Journal of American
Chemical Society 60(2), 309319.
Coussy, O. (2005). Poromechanics of freezing materials.
Journal of the Mechanics and Physics of Solids
Submitted to publication.
Coussy, O. and T. Fen-Chong (2005). Crystallization, pore
relaxation and cryosuction. Comptes Rendus Mecanique,
submitted to publication.
Dash, J. G., H.-Y. Fu, and J. S. Wettlaufer (1995). The
premelting of ice and its environmental consequences.
Reports on Progress in Physics 58, 115167.
Fen-Chong, T., A. Fabbri, and F. Gaulard (2005). Freez-
ing and thawing porous media : Experimental study
with a dielectric capacitive method. Comptes-Rendus
Mecanique, Submitted to publication.
Fen-Chong, T., A. Fabbri, J.-P. Guilbaud, and O. Coussy
(2004). Determination of liquid water content and
572
Copyright 2005 Taylor & Francis Group plc, London, UK
dielectric constant in porous media by the capacitive
method. Comptes- Rendus Mecanique 332, 639645.
Jehng, J., D. Sprague, and W. Halperin (1996). Pore struc-
ture of hydrating cement paste by magnetic resonance
relaxation analysing and freezing. Magnetic Resonance
Imaging 14(7/8), 785791.
Kaufmann, J. (1999). Experimental identification of damage
mechanisms in cementitious porous materials on phase
transition of pore solution under frost deicing salt attack.
Ph. D. thesis, cole Polytechnique Fdrale de Lausanne,
Ecublens, CH-1015 Lausanne.
Mercury, L., P. Vieillard, and Y. Tardy (2001). Thermody-
namics of ice polymorphs and icelike water in hydrates
and hydroxydes. Applied Geochemistry.
Pigeon, M. (1984). Microstructure et rsistance au gel des
ciments et btons. Ph. D. thesis, Universit Pierre et Marie
Curie, Paris 6, France.
Quinson, J. F., M. Astier, and M. Brun (1987). Determination
of surface areas by thermoporometry. Applied Catalysis
30, 123130.
Raoof, A. (1998). Adsorption distribution et dynamique de
leau dans les milieux poreux. Laboratoire Central des
Ponts et Chausses, Paris.
Rootare, H. and C. Prenzlow (1967). Surface areas from
mercury porosimeter measurements. Journal of Physical
Chemistry 71(8), 27332736.
Scherer, G. W. (1993). Freezing gels. Journal of Non Crys-
talline Solids 155, 125.
Sellevold, E. J. and D. H. Bager (1980). Low temperature
calorimetry as a pore structure probe. 7TH Interna-
tional Congress on the Chemistry of Cement, Paris 4,
394399.
Setzer, M. J. (1997). Action of frost and deicing chemicals:
Basic phenomena and testing. In J. Marchand, M. Pigeon,
and M. J. Setzer (Eds.), RILEM Proceedings 30, Freeze-
Thaw Durability of Concrete. E & FN SPON.
Wang, K., P. J. M. Monteiro, B. Rubinsky, and A. Arav (1996).
Microscopic study of ice propagation in concrete. ACI
Materials Journal, 370377.
Zakri, T., J.-P. Laurent, and M. Vauclin (1998). Theoretical
evidence for lichteneckers mixture formulae based on the
effective medium theory. Journal of Physics D: Applied
Physics 31, 15891594.
Zuber, B. and J. Marchand (2004). Predicting the volume
instability of hydrated cement systems upon freezing
using poro-mechanics and local phase equilibra. Concrete
Science Engineering 37(268), 257270.