Beruflich Dokumente
Kultur Dokumente
n e t
ABSTRACT
A great deal of information has been collected over the years about the phenomenology of salt scaling of concrete, and no single mechanism has
been identified that can account for all of the observations. We review the essential facts that any complete theory of scaling must explain, and
suggest two mechanisms that seem to be of primary importance. Experiments indicate that damage is worst when the water on the surface of the
concrete contains a small amount of solute, regardless of the type of solute. We argue that the role of solute is to create brine pockets in the ice that
act as mechanical flaws; when the weakened ice cracks, the crack runs to the surface of the concrete and does superficial damage. An analysis of
the stresses created in this way shows that they are more than sufficient to initiate cracks. A second phenomenon, which is present when chloride
salts are used, is expansion resulting from precipitation of Friedel's salt. The crystallization pressure from growth of these crystals causes significant
expansion of the paste, and results in high stresses on a local scale near the ice/paste boundary. This can substantially weaken the material and
facilitate scaling.
1359-5997 9 2004 RILEM. All rights reserved.
R/~SUMI~
Beaueoup d'informations ont Ot~ rassembldes au cours des dernikres annkes au sujet de la phknomOnologie d'Ocaillage du b~ton liOe aux sels,
mais aucun mOcanisme pouvant expliquer toutes ces observations n 'a Ot~ identifiO. Nous passons en revue les fairs essentiels que n 'importe quelle
thOorie complbte d'~eaillage doive expliquer, et suggkrons deux mOcanismes qui semblent ~tre d'importanee primaire. Les experiences indiquent
que les dommages sont les plus mauvais quand 1'eau sur la surface du bOton contient un peu de corps dissous, indOpendamment du type de corps
dissous. Nous arguons du fa# que le rrle du corps dissous est de erOer les poches de saumure dans la glaee qui agissent en tant que dOfauts
m~eaniques. Quand la glaee affaiblie fend, des dommages superficiels sont provoquOs par la propagation des fentes vers la surface du bOton. Une
analyse des pressions cr~Oes de cette faqon prouve qu 'elles sont au moins suffisantes pour initier des fissures. Un deuxiOme phOnombne, qui est
prOsent olz des sels de ehlorure sont employOs, est l 'expansion r~sultant de la precipitation du sel de Friedel. La pression de cristallisation de la
eroissance de ces cristaux cause l'expansion significative de la pdte, et provoque des pressions klevOes sur une ~chelle locale pros de l'interface
glaee/pdte. Ceci peut sensiblement affaiblir le matOriel et faciliter 1'~caillage.
1. INTRODUCTION nature of the solute (e.g., salts, alcohol, and urea show
similar behavior) [5]
When a pool of liquid freezes on a concrete surface, iii) the damage consists of small flakes of material removed
superficial damage called salt scaling (see Fig. 1) often occurs. from the surface [6]
The phenomenon has been described in detail in several iv) no scaling occurs without free liquid on the surface o f the
excellent reviews [1-3]. The principal observations are as sample (i.e., saturated surface-dry samples do not scale)
follows: [5, 7]
i) the damage is worst when the water contains a moderate v) damage is worse when the minimum temperature in the
amount (the so-called pessimum concentration) of solute cycle is lower [3, 7, 8]
[3-5] vi) the salt concentration in the exterior liquid is more
ii) the pessimum concentration is nearly independent of the important than that o f the pore liquid [1, 3]
vii) entrained air reduces the damage [1, 5]
Editorial Note
Prof. George W. Scherer is a RILEM Senior Member.
2. EXPERIMENTAL PROCEDURE
3. RESULTS
Consequently, a pressure gradient forms in the pores 3.3 Warping with NaCI brine
each time the temperature drops, and the suction in the
pores near the lower surface causes that side to contract and Fig. 6 shows the deflection of a plate of cement paste
produce an upward deflection of the plate. The pore suction similar to that used in Fig. 4, except that the liquid contains
draws liquid into the plate from the pool, so the gradient 7~ NaC1. Nucleation occurs at about 4~ but there is no
gradually decreases and the plate becomes flat. The time oscillation of the deflection as there was when pure ice formed.
required for the relaxation is about 15 minutes, which is in We attribute this to the weakness of the ice, which contains a
good agreement with the hydraulic relaxation time found on large proportion of unfrozen brine. As shown in Fig. 7, even at
the same sample using the beam-bending method. the lowest temperature of this experiment (-10~ there is still
After nucleation of ice occurs, the temperature of the --45 vol% unfrozen brine in the layer on top of the cement
water (Top thermocouple) remains at 0~ because of the paste [21], so the layer does not have enough stiffness to
release of heat of fusion, as freezing progress; meanwhile, deflect the plate. However there is a continuously increasing
the temperature of the box (Bottom thermocouple)
decreases. By the time that freezing is complete, the
temperature of the surrounding box is about -5~ so the
sample coots quickly, resulting in a deflection (concave
upward) o f - 5 0 / a m . This deflection results from the
thermal contraction of the ice, whose linear thermal
expansion coefficient is eq = 5.1 x 10S~ -1 [16, p. 114],
whereas that of the cement paste is ~q ~ 1.0 x 105~ -1 [17].
The radius of curvature, R, of a bimaterial plate by thermal
expansion mismatch was analyzed by Timoshenko [18].
The result can be written as [19]
silanols [24, 25]; chloride adsorbs, largely irreversibly [26], is suggested by the observation that cements containing
and simulations indicate that the primary adsorption site is higher C3A contents are more susceptible to scaling [1].
the surface of calcium hydroxide crystals [27]. After To see whether noticeable damage results from salt
equilibration o f a 3 mm-thick plate of paste in 21 wt% NaC1 exposure, we performed a beam-bending measurement on a
for 24 h, we dried it and performed X-ray powder 3 mm plate of paste, then soaked it in 21 wt% NaC1
diffraction (XRD). Comparing the pattern to a sample that solution for - 4 0 h, and repeated the measurement with the
had not been exposed to the salt reveals that the only sample immersed in the salt solution. As shown in Fig. 11
differences are peaks corresponding to Friedel's salt [28], (Sample B), Young's modulus dropped by - 1 6 % (from
as shown in Fig. 10. This suggests that part of the expansion E = 17.1 to 14.6 GPa) and the permeability rose by - 2 5 %
may result from crystallization pressure upon precipitation (from 1.6 x 10 "21 m 2 to 2.0 x 10 -2~ m2). The sample was then
of Friedel's salt, which results from reaction of chloride placed in limewater (i.e., water saturated with calcium
with calcium aluminates. The expansion observed in hydroxide) for 5 days and remeasured while immersed in
NaNO3 indicates that sodium adsorption also contributes to limewater; the properties showed a small additional
the strain, since XRD performed on that sample indicated deterioration: E = 14.0 GPa and D = 2.1 x 10 21 m 2. These
no new phases.
4. DISCUSSION
changes are modest, but they are well outside the typical The bimaterial effect seems to account for most of the
sample-to-sample variation for our bending measurements phenomenology of scaling. Damage is seen only when there
(standard deviation ~ 10% of the mean [12]), where the is free liquid on top, because a certain thickness of ice is
maximum stress on the sample is -0.8 MPa. This point is needed to generate the stress. The pessimum results from
demonstrated by Sample A in Fig. 11, where the the mechanical properties of brine: pure ice is too strong to
measurement was repeated after each change of pore fluid. break, but impure water freezes into ice containing brine
From the evident change in properties, we conclude that pockets that constitute mechanical flaws. This explains why
some microcracking has resulted from the salt exposure. damage diminishes at high solute levels, where the ice is
However, it does not appear that a single exposure to a too weak to exert appreciable stress, and why the nature of
concentrated salt solution causes extensive microcracking the solute does not matter. Cracking of the ice creates
of cement paste, so the stress from crystallization pressure tensile stresses where the ice joins the surface of the paste
must be much less than the 6 MPa estimated above. The [29], and damage may therefore result from fatigue after
damage might be worse for mortar, however, because the many cycles. On the other hand, salt-induced swelling
paste will expand away from the aggregate, possibly causes cycles of stress that create some microcracking,
resulting in high local stresses. thereby weakening the surface, and may cause buckling of
The freezing experiments revealed another effect that may thin layers of paste over aggregate particles.
be the most important factor in scaling damage: a layer of pure The beneficial effect of air entrainment is more difficult
ice causes severe warping of the sample owing to the to explain on the basis of either of these mechanisms. It
mismatch in thermal expansion in the bimaterial composite of may be that air entrainment, by reducing bleeding, leads to
ice and paste. The stress produced from this mechanism can greater strength in the superficial layer of paste and thereby
amount to a few MPa, and increases as the temperature reduces damage. It is also possible that the air voids provide
decreases. Damage could result from shearing of asperities a site where crystals of Friedel's salt precipitated upon
from the surface at points where the ice and cement paste are exposure to NaC1 can grow without generating stress.
interlocked. However, a more likely mechanism is that the However, this remains to be proven, so the effectiveness of
frozen layer cracks, and the crack continues a short distance air entrainment presents an important challenge to the
into the concrete, resulting in superficial damage. A bimaterial or salt-swelling mechanism.
mechanism of this kind is used to produce a decorative surface
on glass [29]: a layer of epoxy is spread on a plate of glass,
which is then cooled until the thermal expansion mismatch 5. CONCLUSIONS
causes the epoxy to crack; the cracks penetrate into the glass
and remove a thin layer, resulting in a scalloped surface. The The warping experiment introduced in this study reveals
lower the temperature to which the body is cooled, the greater two important mechanisms that contribute to stress during
the probability that the layer will crack, so this mechanism is freezing of saline solutions on the surface of cement paste.
consistent with experiments which indicate that scaling First, there is a large mismatch in the thermal expansion
damage is worse when the minimum temperature of cooling is coefficients of the materials, so that the frozen layer
lower [3, 30]. develops high tension during cooling. We find that pure
A layer of pure ice does not crack, even when held at - water ice does not crack when cooled to -18~ on the
18~ for several hours. This may explain why scaling is not surface of cement paste. However, when the water contains
serious when pure water is used in the surface layer. solute, the unfrozen liquid creates brine pockets that
However, when any amount of salt is present, then the ice weaken the ice and promote cracking; as in the glue-spall
contains pockets or channels of brine that constitute process, the crack is expected to intersect the surface of the
mechanical flaws, and encourage fracture of the layer. This paste and cause damage. This mechanism accounts for most
could account for the pessimum: pure ice does not crack in of the phenomenology of salt scaling: (i) the pessimum
the temperature range used in scaling tests (typically down occurs because pure ice doesn't crack and high solute
to -20~ and high salt concentrations yield ice with so concentrations make the ice too weak to exert stress on the
much brine that they cannot impose significant stress at - paste; (ii) the pessimum is independent of the type of
20~ however, intermediate salt contents permit solute, because any solute creates brine pockets in ice; (iii)
development of substantial stresses and contain flaws that damage is in the form of flakes, as expected from the glue-
will promote cracking when the stress becomes high. This spall mechanism; (iv) damage requires free liquid, because
will lead to damage from low to moderate salt cracks can only form when a layer of ice is present; (v)
concentrations, but less damage at high concentrations. damage is worse when the minimum temperature is lower,
The occurrence of cracking in a layer of ice made from a because the stress in the ice increases and more cracks
3% solution of NaC1 is demonstrated by the photos that develop from smaller and smaller flaws; (vi) the exterior
were obtained by introducing a video camera lens through salt concentration is more important than the composition
the top of the box, as indicated in Fig. 2. The first cracks are of the pore liquid, because the role of the solute is to create
observed [22] when the temperature reaches -7.5~ brine pockets in the external ice layer.
weakening of the ice by the cracks results in upward The solute may play an additional role by contributing to
deflection because of the salt-induced swelling. Numerous weakening of the surface of the paste. As the saline solution
cracks are observed in the ice at the end of the experiment. freezes, it produces pure water ice plus a brine whose
J.J. Valenza, G. W. Scherer / Materials and Structures (2005) 479-488 487
concentration increases as temperature decreases; for NaC1 [13] Scherer, G.W., 'Measuring permeability of rigid materials by a
solutions, the brine approaches the eutectic concentration o f beam-bending method: I. Theory', J. Am. Ceram. Soc. 83 (9)
21 wt% at -21~ Exposure to sodium chloride solutions (2000) 2231-2239.
causes expansion o f cement paste, and some microcracking [14] Vichit-Vadakan, W. and Scherer, G.W., 'Measuring
permeability of rigid materials by a beam-bending method: II.
seems to result. One indication that other solutes, such as Porous Vycor', J. Am. Ceram. Soc. 83 (9) (2000) 2240-2245.
alcohol, can damage the surface is the observation [15] that [15] Vichit-Vadakan, W. and Soberer, G.W., 'Measuring
immersion in pure isopropanol causes a hundred-fold Permeability of Rigid Materials by a Beam-Bending Method: III.
increase in the permeability o f cement paste. We are Cement Paste', J. Am. Ceram. Soc. 85 (6) (2002) 1537-44.
currently examining the effect o f exposure to other solutes [16] Pounder, E.R., 'Physics of Ice' (Pergamon Press, Oxford, 1965).
on the properties o f cement paste. [17] Valenza, J.J. and Scherer, G.W., 'Evidence of anomalous
thermal expansion of water in cement paste', Cem. Concr. Res.
35 (2005) 57-66.
ACKNOWLEDGEMENTS [18] Timoshenko, S., 'Analysis of bi-metal thermostats', J. Opt.
Soe. America 11 (1925) 233-255.
[19] Scherer, G.W., 'Relaxation in Glass and Composites' (Wiley,
This work was supported by National Science
New York, 1986; reprinted by Krieger, Malabar, FL, 1992) Ch.
Foundation Grant CMS-0200440.
16.
[20] Valenza II, J.J. and Scherer, G.W., 'Mechanism for salt scaling
of a cementitious surface', to be published in Proceedings of
REFERENCES RILEM conference, Evanston, IL (2004).
[21] Weast, R.C. and Astle, M.J. (eds.), 'CRC Handbook of
[1] Marchand, J., Pleau, R. and Gagn6, R., 'Deterioration of concrete Chemistry and Physics', 62nd ed., (CRC Press, Boca Raton, FL,
due to freezing and thawing', in 'Materials Science of Concrete 1981) D-232-233.
IV', eds. J. Skalny and S. Mindess (Am. Ceram. Soc., [22] Scherer, G.W. and Valenza II, J.J., 'Mechanisms of frost damage'
Westerville, OH, 1995) 283-354. to be published in 'Materials Science of Concrete', Vol. VII, eds.
[2] Pigeon, M. and Pleau, R., 'Durability of Concrete in Cold J. Skalny and F. Young (American Ceramic Society, 2005).
Climates' (E & FN Spon, London, 1995). [23] Stark, J. and Stth'mer, S., 'Investigation of compatibility of
[3] Lindmark, S., 'Mechanisms of Salt Frost Scaling of Portland cements with several salts', paper 4iv031 in Proc. 10th Int. Cong.
Cement-bound Materials: Studies and Hypothesis', Ph.D. thesis Chemistry Cement, Vol. 4 (S1NTEF, Trondheim, Norway, 1997)
(Report TVBN 1017), Lund Inst. Tech., Lund, Sweden, 1998. [24] Hung, S.Y and Glasser, F.P., 'Alkali binding in cement pastes.
[4] Marchand, J., Pigeon, M., Bager, D. and Talbot, C., 'Influence of Part I. The C-S-H phase', Cem. Concr. Res. 29 (1999) 1893-
chloride solution concentration of salt scaling deterioration of 1903.
concrete', A CI Mater. J. (July-Aug 1999) 429-435. [25] Viallis, H., Faucon, P., Petit, J.C. and Nonat, A., 'Interaction
[5] Verbeck, G.J. and Klieger, P., 'Studies of "Salt' scaling of between salts (NaC1, CsC1) and calcium silicate hydrates (C-S-
concrete', Highway Research Board Bull. 150 (1957) 1-17 H)', J. Phys. Chem. B 103 (1999) 5212-5219.
[6] Jacobsen, S., 'Scaling and Cracking in Unsealed Freeze/Thaw [26] Wiens, U. and Schiessl, P., 'Chloride binding of cement paste
Testing of Portland Cement and Silica Fume Concretes', Thesis containing fly ash', paper 4iv016 in Proc. 10th Int. Cong.
report 1995: 101, Norwegian Inst. Tech., Trondheim, 1995 Chemistry Cement, Vol. 4 (S1NTEF, Trondheim, Norway, 1997).
[7] Sellevold, E.J. and Farstad, T., 'Frost/salt testing of concrete: [27] Kalinichev, A.G. and Kirkpatrick, R.J., 'Molecular dynamics
effect of test parameters and concrete moisture history', Nordic modeling of chloride binding to the surfaces of calcium
Concr. Res. (10) (1991) 121-138. hydroxide, hydrated calcium aluminate, and calcium silicate
[8] Studer, W., 'Internal comparative tests on frost-deicing-salt phases', Chem. Mater. 14 (2002) 3539-3549
resistance', in Int. Workshop on Resistance of Concrete to [28] Gon, S. and Guerrero, A., "Accelerated carbonation of Friedel's
Scaling due to Freezing in the Presence of Deicing Salts, Centre salt in calcium aluminate cement paste', Cem. Concr. Res. 33
Rech. Interuniv. B6ton, Univ. Sherbrooke-Univ. Laval, Quebec, (2003) 21-26.
August, 1993, 175-187. [29] Gulati, S.T. and Hagy, H., 'Analysis and measurement of glue-
[9] Stark, J. and Ludwig, H.M., 'Freeze-thaw and freeze-deicing salt chip stresses', J. Am. Ceram. Soc. 65 (1) (1982) 1-6; correction
resistance of concretes containing cement rich in granulated 65 (6) (1982) 320.
blast-furnace slag', paper 4iv035 in Proc. 10th Int. Cong. [30] Hammer, T.A. and Sellevold, E.J., 'Frost resistance of high-
Chemistry Cement, Vol. 4 (SINTEF, Trondheim, Norway, strength concrete', pp. 457-487 in 2no Symp. High Strength
1997). Concrete, SP- 121 (American Concrete Inst., Detroit, 1990).
[10] Litvan, G.G., 'Phase transitions of adsorbates: VI, Effect of [31] Vichit-Vadakan, W. and Scherer, G.W., 'Measuring
deicing agents on the freezing of cement paste', J. Am. Ceram. permeability and stress relaxation of young cement paste by
Soc. 58 (1-2) (1958) 26-30. beam-bending', Cement Concrete Res. 33 (2003) 1925-1932.
[11] Fagerlund, G., 'Studies of the destruction mechanism at freezing [32] Hildebrand, F.B., 'Advanced Calculus for Applications'
of porous materials', Contributions to Fondation Frangaise (Prentice-Hall, Englewood Cliffs, New Jersey, 1962).
d'l~tudes Nordiques, VIe Congres Int. : Les probl6mes pos6s par [33] Wolfram Research, 'Numerical inversion performed using the
la g61ifraction. Recherches fondamentales et appliqu6es (French package Numerical Inversion, #2000 by Amaud Mallet',
Foundation for Nordic Studies, 6th Int. Cong. on problems raised available at the Wolfram website.
by freezing. Fundamental and applied research), Le Havre, 23-25
April 1975.
[12] Valenza II, J.J. and Scherer, G.W., 'Measuring permeability of
rigid materials by a beam-bending method: V. Isotropic
rectangular plates of cement paste', J. Am. Ceram. Soc. 87 (10)
(2004) 1927-1931.
488 J.J. Valenza, G.W. Scherer / Materials and Structures (2005) 479-488
APPENDIX.
E X T R A P O L A T I O N OF CREEP CURVE
For a linearly viscoelastic material, the creep function can 1.3oi ........................................
be calculated from the stress relaxation function, so the data
from the beam-bending experiment can be used to predict the
creep rate of the paste. Of course, the calculation is accurate
only over the time scale over which the measurements were
made, but the logarithmic nature of the function permits a 1_0.9
relatively long extrapolation without serious error. The
mathematical details are discussed in [31]. The stress x