Salt frost scaling of concrete – new insights regarding the damage mechanism

. To achieve the important goal of clinker reduction for concretes in cold and temperate climates, it is important to understand the underlying mechanisms for salt frost scaling. Two conflicting damage theories dominate the current State of the Art – the glue spall theory and the cryogenic suction theory. In this study, important aspects of both theories were evaluated experimentally using salt frost scaling tests on concrete and low temperature differential scanning calorimetry. It was found, that the generation of scaling in a salt frost attack could be better explained by the cryogenic suction theory, which is based on the uptake of highly concentrated brine at sub-zero temperatures. However, this theory cannot account for the pessimal effect of low de-icing salt concentrations in salt frost scaling tests, as it does not consider the moisture uptake during the thawing phase of a freeze-thaw cycle. By expanding the cryogenic suction theory by that aspect, a comprehensive theory for salt frost damage was obtained. In that theory the cryogenic suction of highly concentrated brine during frost is responsible for the generation of scaling. The moisture uptake during the thawing phase balances the de-icing salt concentration in the concrete and accounts for the pessimum effect.


Introduction
In the early 20th century, the practice to apply de-icing salts on pavements to enhance the removal of snow and ice was widely adopted.It then soon became obvious that this treatment could cause an increased surface scaling of concrete [1].In one of the first studies on the influence of the de-icer concentration Arnfelt [2] discovered already in 1943, that the most severe damage occurs at relatively low concentrations.Several studies (e.g.[3][4][5]) later confirmed this effect and located the pessimum concentration in the range of about 3 wt.-%for CaCl2 and NaCl.
That concentration became generally acknowledged and consequently the basis for many salt frost scaling test procedures, e.g. in [6].However, the cause for this pessimum phenomenon is still not fully understood.Existing theories for salt frost scaling attack thus struggle to account for it.
Two conflicting damage theories dominate the current State of the Art -the glue spall theory and the cryogenic suction theory.The present study addresses the ability of these theories to account for the pessimum effect.

Eutectic behaviour of salt solutions
When a salt solution freezes, pure ice crystals begin to form, as the solute does not enter the crystal lattice.The salt concentration in the remaining unfrozen solution increases and its freezing point decreases.As long as the temperature stays above the specific eutectic temperature (e.g.-21.3 °C for NaCl solutions), pockets of non-frozen brine are contained in the ice [7].As typical minimum temperatures in salt frost scaling tests are higher than the eutectic temperature, salt solutions usually do not freeze completely in these tests.The eutectic behaviour of salt solutions forms the basis for the two following damage theories.

Glue spall
The glue spall theory from Valenza and Scherer [7,8] attributes surface scaling to the interaction of an external ice layer with the concrete surface.After the initial freezing of water on concrete, a further drop in temperature causes the ice and the concrete to shrink.As the contraction of ice is about five times higher than that of concrete, stresses occur at the ice-concrete interface.Salt frost scaling is induced, when the ice layer cracks and these cracks propagate into the concrete (see Figure 1).This should only happen at moderate de-icer concentrations, when brine pockets weaken the ice layer.Ice formed from pure water does not contain brine pockets, and for this reason, it possesses a high tensile strength and doesn't crack.At high concentrations, the ice is too soft to damage the concrete.The glue spall theory thus uses the mechanical properties of saline ice to explain the pessimum concentration of 3 wt.-%NaCl in salt frost scaling tests.At -20 °C no scaling should occur outside the range of approximately 2-5 wt.-% NaCl [8].
The glue spall theory also predicts a dependency of the scaling intensity on the ice layer thickness, which is in contact with the concrete during frost.This dependency was confirmed in several studies [9][10][11].However, alternative approaches for explaining this effect exist.

Cryogenic suction of brine
The cryogenic suction theory by Lindmark [4] combines the classic theory of micro ice body growth [12] with aspects of frost heaving of soils [13], thus considering the uptake of de-icer solution in concrete at sub-zero temperatures.During frost, ice lenses form in coarse pores inside the concrete microstructure.Pore solution in very small pores remains unfrozen because of surface forces and growth constraints [12].
The proportion of frozen pore solution depends on the concentration of solutes and on the pore size distribution in the concrete.As unfrozen pore solution possesses a higher free energy than ice, moisture is drawn towards ice lenses from the surrounding hardened cement paste, until a thermodynamic equilibrium is reached, whereupon the growth of the ice lenses stops [4].This process is known as cryogenic suction or cryosuction.

Fig. 2. Growth of ice lens due to cryogenic suction of brine
In the cryogenic suction theory, unfrozen brine from the outer ice layer acts as an additional moisture reservoir.The brine is drawn from the outside towards surface near ice lenses in the concrete, which now can grow further.This process should be considered as a specific kind of cryogenic suction, the cryogenic suction of brine, that takes place at the interphase of concrete and (frozen) salt solutions (see Figure 2).
The resulting stresses and increased saturation of the surface layer cause scaling according to the theory.The cryogenic suction process was verified in several experimental setups [4,14,15].In [15] it was also found, that the extent of additional ice formation due to brine uptake is influenced mainly by the porosity and by the chloride binding capacity of the hardened cement paste.
The destructive potential of the brine uptake at subzero temperatures was also displayed by gradually cooling HCP discs in brine (22.4 wt.-% NaCl) to -19 °C.The damage strongly resembled experiences with macroscopic ice lens formation (see. Figure 3) [16].Fig. 3. HCP damaged by cryogenic suction of brine [16] However, as the core of the cryogenic suction theory is the uptake of unfrozen, highly concentrated brine, scaling should be intensified with increasing de-icer concentrations.Thus, the diminishing effect of high salt concentrations in salt frost scaling tests cannot be fully explained by this theory.This indicates that additional processes for liquid uptake must be considered.
Lindmark [4] assumed, that during the thawing phase in the freeze-thaw cycle, when the test solution is completely liquid, diffusion processes balance the salt concentration in the concrete with the outer salt concentration in the test solution.If the salt concentration in the test solution is only moderate, the diffusion would dilute the salt concentration in the surface region of the concrete.
There are two drawbacks to this explanation.First, diffusion is rather slow process and might not be sufficient to account for the necessary change of the salt concentration in the surface near pore solution of the concrete.Second, it is a well known fact, that liquid uptake during freeze-thaw exposure also occurs, when pure water is used as test solution.This water uptake is typically considered to take place during the thawing phase in a freeze-thaw cycle, e.g. as described in the micro ice lens theory by Setzer [17].

Liquid uptake in thawing concrete
In freezing concrete, unfrozen solution from smaller pores is drawn towards ice lenses in coarser pores to satisfy the conditions of thermodynamic equilibrium (cryogenic suction, see Figure 4).When water is used as a test liquid, no uptake of external liquid is possible at temperatures below 0 °C, as external water is completely frozen [17].Cryogenic suction of brine thus cannot take place.Only very weak scaling occurs.During heating thermodynamics require the process to be reversed, but gel pore solution remains "trapped" at ice lenses up to temperatures near the melting point of bulk ice.Instead, the ice at the concrete surface melts first and the water is sucked up by the concrete.
Though the process starts at the concrete surface, moisture is transported several centimetres into the concrete, as the moisture flow follows the receding ice front [17].This intensive saturation of the concrete in subsequent freeze-thaw cycles causes internal damage in non-resistant concretes.

Experimental
In this study, important aspects of the different salt frost scaling theories were evaluated experimentally.Salt frost scaling tests were carried out on concrete, to investigate whether the glue spall theory can correctly reflect the influence of different salt concentrations on the scaling intensity.
Low temperature differential scanning calorimetry (LTDSC) was used to investigate, how different levels of internal salt concentration in hardened cement paste affects its ability to form additional ice due to the cryogenic suction of brine.In that way the impact of an additional uptake of salt solution due to ice lens pump effect was "simulated".The LTDSC measurements were carried out on hardened cement paste samples.

Concrete
Air entrained concrete was used for the salt frost scaling tests.The target fresh air content was 5.0 ± 0.5 vol.-%.The identical concrete composition was used to produce three specimen series for the salt frost scaling tests.As a certain variation was permitted for the fresh air content, the individual concrete series can differ in salt frost scaling resistance to some extent.The concrete composition is give in Table 1.After one day in mould the concrete was cured in water at 20 °C until an age of seven days.This was followed by a 21 day period in standard climate at 20 °C and 65 % relative humidity.The 28 day compressive strength of the concrete was about 50 MPa.

Hardened cement paste (HCP)
Hardened cement paste was used as model material for concrete in measurements with low temperature differential scanning calorimetry (LTDSC).Fresh cement paste was produced with w/c-ratio of 0.50 using the same OPC as for the concrete.More information regarding the OPC properties is given in [15,16].The fresh paste was filled in a 50 ml test tube, which was sealed airtight and then slowly rotated for four hours to avoid sedimentation.Then it was stored in water saturated atmosphere at 20°C.
At an age of four weeks, discs of HCP with thickness of 1.0 to 1.5 mm were cut.The discs were cured in standard climate at 20 °C and 65% r.h. for one week.Then they were immersed in different NaCl solutions in the range of 0-18 wt.-% NaCl for two weeks.The intermediate drying for one week in standard climate was supposed to allow a quick uptake of the salt solutions.This was confirmed by measuring the depression of the freezing and melting point of the capillary pore solution in these samples [16].

Salt frost scaling test
The salt frost scaling resistance was investigated with the CDF test (Capillary suction of de-icing solution and freeze thaw test), which is standardized in CEN/TS 12390-9 [6].The scaled material was collected in paper filters, dried and weighted after the appropriate numbers of freeze-thaw cycles.The liquid uptake during presaturation and freeze-thaw test was determined by weighing the specimens.For the calculation of the liquid uptake the specimen weight was corrected for the amount of scaled material as described in [6].
Two modifications to the CDF test were applied in this study.First, different test solutions in the range between 0 and 18 wt.-%NaCl were used for the pre-saturation and the freeze-thaw-cycling, instead of 3 wt.-%NaCl.Second, with increasing NaCl concentration a higher amount of salt is trapped in the paper filters, which are used to collect the scaled material.This effect distorts the measured scaling intensity, especially for NaCl concentrations above 6 wt.-%.The average amount of salt, trapped in the paper filters, was thus determined in a preliminary investigation.The measured weights of scaled material were corrected by these values.
Three runs of the CDF test were carried out.All specimens for each run were produced from a single concrete batch to ensure a uniform concrete properties.Three specimens were used for each NaCl-concentration.

LTDSC
The effect of highly concentrated NaCl brine on the ice formation in HCP was investigated using LTDSC in a specific setup, which allowed the cyclic freezing and thawing of hardened cement paste in contact with brine.22.4 wt.-% NaCl (also referred to as brine) was used as test solution.This concentration corresponds to the liquidus concentration of NaCl solutions at -20 °C.
All experiments involving brine were conducted with a target minimum temperature of -20 °C.Consequently, the brine should have remained in liquid state during these experiments.
The experiments were carried out at a specimen age of ~28 days with a Netzsch Differential Scanning Calorimeter (DSC 214 Polyma) in combination with an intercooler, which can realize a minimum temperature of -70 °C.A small sample of approximately 3 x 3 mm² was obtained from each HCP disc.The HCP sample was surface dried with a paper towel and inserted into an aluminium crucible.Then a small drop of brine was applied on the sample, see Figure 5.The lid was then cramped tightly onto the crucible (cold welding).During the procedure the mass of sample, brine and crucible with lid was recorded with an accuracy of 0.01 mg.The whole procedure was carried out in less than one minute to avoid evaporation from the samples.The ratio of HCP to brine was ~20:1.Furthermore, samples without brine were also prepared for the initial measurement of the ice formation in the HCP without brine addition.

Fig. 5. LTDSC setup -HCP with brine on top (schematic)
The calorimetric measurements were started immediately after sample preparation.The temperature cycles ranged from +20 °C to -20 °C.The intention behind this test setup was to initiate freezing processes in the hardened cement paste while keeping the brine on the sample in liquid state, hence allowing the uptake of the brine by the cryogenic suction process.
Each cycle began with a cooling phase from +20 °C to -20 °C at 2 K/min.The minimum temperature of -20 °C was kept constant for 30 min to allow the cryogenic suction process to take place and to equilibrate the sample.Thereafter the sample was heated to +20 °C at 2 K/min.After an isothermal phase of 1 min the next cycle was started.Only one cycle was used for the initial measurement without brine and four subsequent cycles for the samples with brine on top.
The necessary heat of fusion for melting the frozen pore solution during the heating phase was used to assess the mass of ice in the samples, which equates the amount of water in the sample, which is freezable in the temperature range to -20 °C.A detailed description of the calculation process is given in [15].6 shows the results of the salt frost scaling tests as accumulated scaling after 28 freeze-thaw cycles for the different NaCl concentrations.Figure 7 shows the corrected accumulated scaling.By comparing these two figures it can be observed that in the range above 9 wt.-%NaCl the amount of scaling is overestimated, when the values are not corrected for the amount of salt, which is collected together with the scaled material.The following interpretation thus focusses on the corrected results

Fig. 6. Results of the CDF tests; Scaling (not corrected)
The most intense scaling occurs at 0.5 wt.-% NaCl (see Figure 7), which thus represents the pessimum concentration in the CDF tests.This is in accordance with recent results [18,19], but seems to contradict earlier findings [3][4][5].However, in earlier studies the NaCl concentrations below 2-3 wt.-% NaCl were not investigated.Consequently, the pessimum was located at 3 wt.-%.

Fig. 7. Results of the CDF tests; Scaling (corrected)
It is also important to observe, that even at 6 wt.-% NaCl the scaling intensity is several times higher than in water.Both the location of the pessimum at 0.5 wt.-% and the scaling intensity at 6 wt.-% NaCl disagree with the glue spall theory, which predicts scaling at -20°C only in the range of about 2 to 5 wt.-%NaCl [8].This indicates that the glue spall theory cannot account for all aspects of salt frost scaling.
Figure 8 shows the accumulated liquid uptake of the concrete during the freeze-thaw cycling.The liquid uptake that occurred during the pre-saturation period is not included here.The step from pure water to 0.5 wt.-% NaCl causes a noticeable increase in liquid uptake.The range from 0.5 to 6.0 wt.-% is marked by uniform high values, before the liquid uptake decreases gradually with increasing NaCl concentrations.

Fig. 8. Results of the CDF tests; Liquid uptake during the freezethaw cycles (corrected)
These results show that the salt concentration has an influence on internal processes of the concrete during the freeze-thaw cycles.This indicates that the salt frost attack is not completely independent from internal processes.The cryogenic suction theory might thus be better suited to describe the salt frost scaling mechanism.

Brine uptake and ice formation
The impact of the salt concentration in the HCP on the ability to form ice at -20 °C was examined with LTDSC. Figure 9 shows the melting curves the HCP sample, which was pre-saturated with water.The initial cycle was carried out without brine and of the four succeeding cycles with brine on top of the specimen.The curves are displayed with a vertical offset as a function of the temperature.The total ice mass equals the amount of freezable water in the temperature range of the measurements (≥ -20°C).With subsequent freeze-thaw cycles the shape and position of the melting peak change and the amount of freezable water increases.Both effects indicate that brine is sucked into the HCP during the cycles.Thus, the freezing and melting temperatures of the original pore solution decrease, whereas they increase for absorbed brine due to the blending of both.The increase of the amount of freezable water indicates a binding of chlorides from the absorbed brine by the HCP.When the crucible was opened after the measurement it was revealed, that the brine was completely absorbed by the sample.
Thus, the uptake of highly concentrated brine due to the cryogenic suction process can indeed increase ice formation in HCP, despite the high amount of salt in the brine.However, it also shows that in order to explain the pessimum effect, an additional saturation process should be considered -the micro ice lens pump [17].
The increased salt concentrations in the HCP, which would result from the saturation by the micro ice lens pump were recreated by saturating the samples in salt solutions before the LTDSC measurement.
Figure 10 shows the LTDSC melting curves for the sample that was stored in 15 wt.-%NaCl solution before the measurement.Due to the increased salt content in the pore solution of the HCP, the melting peak of the initial cycle is shifted toward lower temperatures.A part of the water, that was freezable at -20 °C, now freezes in the temperature range below -20 °C.Consequently, the amount of freezable water at -20 °C is reduced, when compared to the sample that was pre-stored in water.
Nevertheless, the HCP takes up brine during the subsequent freeze-thaw cycles.This brine uptake causes a further reduction of the freezable water content, as the ability of the HCP to bind chloride ions from the brine is diminished due to the inherent high chloride content in the HCP.
Figure 11 gives an overview on the impact of the salt concentration during the pre-storage on the content of freezable water (-20 °C) in HCP.The values for the initial cycle without brine and after 4 cycles with brine are displayed.
There is a general trend towards a reduction of the initial content of freezable water with increasing salt concentration, with only one exception (at 0.5 % NaCl).Furthermore, the pre-storage in salt solution reduces the capacity of the HCP to form ice due to the brine uptake.Three ranges can be distinguished: Range 1 (Pre-storage in 0.0-3.0wt.-% NaCl): Here the maximum amount of freezable water occurs in the initial measurement.The freeze-thaw cycles with brine cause a strong increase in the freezable water content.
Range 2 (Pre-storage in 6.0-9.0 wt.-% NaCl): The increasing salt concentrations during pre-storage cause a clear reduction of the initial freezable water content.The brine uptake during freeze thaw cycles still causes an increase in the freezable water content, though much lower than in range 1.
The experiment thus reflects the experience from salt frost scaling tests, where increasing salt concentrations beyond the pessimum cause a reduction in scaling.This can be explained by the reduced capability of the HCP to form ice by cryogenic suction of brine.

processes on pessimum concentration
By considering cryogenic suction of brine during frost in combination with liquid uptake during thawing, the pessimum effect in salt frost scaling tests can be explained.
Cryogenic suction of brine can take place, when ice formation in the concrete has begun and when a minimum amount of salt is present in the test solution.The suction process should be strongest at the minimum temperature in a freeze-thaw cycle (see Figure 12).
The salt concentration in the unfrozen brine mainly depends on the type of solute and on the minimum temperature.The concentration of NaCl in the liquid brine is 22.4 wt.-% at a minimum temperature of -20 °C.

Fig 12. Cryogenic suction of brine at -20 °C
The initial concentration of the de-icer solution only influences the amount of unfrozen brine, but not its concentration, once the de-icer solution has started to freeze.The cryogenic suction of brine is probably limited to the immediate surface layer of the concrete (<1 mm), which will be removed from concrete by scaling, to some extent.
Parallel to the cryogenic suction of brine an internal distribution of moisture occurs inside the pore structure of the concrete, as the pore solution from the unfrozen gel pores is drawn towards lenses of frozen pore solution in the capillaries.The hardened cement paste shrinks.
When the thawing phase starts, a heating gradient causes the outer test solution to melt before the ice in the concrete can melt.Due to the rise in temperature the hardened cement paste starts to expand and the gel pores strive to regain their pore solution.However, as that pore solution is still trapped in the frozen capillaries, the internal redistribution of moisture is inhibited.
Instead, the concrete starts to take up the melted test solution.As the moisture flow follows the receding ice front into the concrete an intensive saturation occurs (see Figure 13), which exceeds the uptake by cryogenic suction of brine.
As the melted test solution has resumed its initial concentration (e.g. 3 wt.-%NaCl), this process dictates how much salt is taken up by the concrete.

Conclusions
Two conflicting damage theories dominate the current State of the Art for salt frost scaling-the glue spall theory and the cryogenic suction theory.In this study, important aspects of both theories were reviewed experimentally using salt frost scaling tests on concrete and low temperature differential scanning calorimetry.
The glue spall theory attributes the surface scaling to the interaction of an external ice layer with the concrete surface.The salt frost scaling tests revealed, that severe scaling can also occur at salt concentrations, where the ice is either too "strong" or too "soft" to allow scaling.Possibly the glue spall effect might amplify the damage at low salt concentrations.However, the theory cannot account for all aspects of a salt frost attack.
The cryogenic suction theory considers the uptake of highly concentrated brine at sub-zero temperatures as the cause for scaling.The ability of this process to increase ice formation in hardened cement paste was shown in this study and also elsewhere.However, this theory cannot account for the pessimal effect of low de-icing salt concentrations in salt frost scaling tests, as it does not consider the moisture uptake during the thawing phase of a freeze-thaw cycle.
By expanding the cryogenic suction theory with that aspect, a comprehensive description for salt frost attack was obtained.In there the cryogenic suction of highly concentrated brine during frost is responsible for the generation of scaling.The moisture uptake during the thawing phase balances the de-icing salt concentration in the concrete and thus reduces the severity of the attack with increasing salt concentration.The combination of both processes can account for the pessimum effect.
The research was supported by the Deutsche Forschungsgemeinschaft (DFG) under grant LU1652/29-1.The authors would like to thank Professor Marianne T. Hasholt of DTU for the valuable discussion on the manuscript.

Fig. 1 .
Fig. 1.Crack formation at the concrete surface due to the glue spall effect (schematic)

Fig. 4 .
Fig. 4. Diffusion of unfrozen pore solution from gel pores to an ice lens in a capillary pore, where it can freeze (schematic)

Fig 11 .
Fig 11.Change in freezable water (-20 °C) due to cryogenic suction of brine, depending on NaCl concentration during pre-storage

Fig 13 .
Fig 13.Liquid uptake with proceeding thawing progressAssuming, that the major part of liquid uptake occurs during thawing, suction by the micro ice lens pump should account for reduced scaling intensity, when test solutions exceed a certain salt concentration.