Evolution of the water content and samples geometry
Figure 1 shows the water contents in groups A and B. The kinetics of water loss is higher for group A (Fig. 1.-a): the samples are supposed dry after around 15 days, while samples from group B (Fig. 1.-b) are supposed dry after 70 days (beginning of the « summer » curing parameters). Specimens from group A incorporate a small amount of water during the cold period (10 °C – 80 %RH) and specimens from group B experience a severe drying when going from cold to warm conditions.
For group A (Fig. 2), there is a good link between water content and height: specimens lose height (around 1 mm) at the same time as water (during the 35 °C – 20 %RH curing parameters, Fig. 3.-a) and seem to gain height and water during the cold curing parameters (not enough measurements were made on height to be sure about its gain during the cold conditions).
Group B (Fig. 3.-b) behaviour is slightly different: during the cold period (the first 60 days), the specimens lose water (Fig. 1.-b) but their height remains constant (around 60.3 mm). Moreover, during the 35 °C – 20 %RH curing process, what was left of water in the samples is lost in about 7 days, while the shrinkage happens during more than a month (from about 71 to 108 curing days). Then, the height does not seem to change anymore. This could be explained by the rheological behaviour of the binder and the mix moisture during curing. The volume variations could mainly be influenced by the binder movements on the aggregate (probably by wetting/dewetting). At 10 °C and 80 %RH the binder is less soft (around 107 Pa at 10 °C and 10 Hz) and there is still a high amount of water in the samples. This could prevent the mix shrinkage. After transition from 10 °C to 80 %RH to 35 °C and 20 %RH conditions, the binder is softer (around 106 Pa at 35 °C and 10 Hz) than at 10 °C so a shrinkage would be easier. The time difference between total loss of water (around 7 days) and the mix shrinkage (more than a month) can simply be explained by the high viscosity of the material resulting in a height loss inertia after water departure.
The height variations are also correlated to the void geometry that would change during curing and influence the modulus of the material. As a matter of fact, the void percentage of every sample was measured at the end of curing and the void content with curing time was deduced from the variations of geometry of each sample (Fig. 4), with the hypothesis that no material had been lost during the whole curing and that the void content is proportional to the height. At the end of the 4 months of curing, group A lost between 0,13 and 0,24 % voids and group B between 0,08 and 0,11 %, which would imply a stiffening of both groups.
Rheological behaviour of the mixture
The stiffness modulus of the samples was assessed with an oedometer device [9]. Usually, other types of mechanical tests (indirect tensile or two-point bending for example) are used for the characterisation of asphalt mixes as they are already very stiff at an early age. Asphalt cold mixes take time to cure and become more cohesive to be tested with these devices, they are very weak at young age. Therefore the oedometer test appears to be a suitable testing technique to measure their mechanical behaviour.
Figure 5 displays the mean evolutions of oedometric modulus versus time for the samples from groups A and B (3 samples for group A and 2 for group B).
The kinetics of stiffness evolution are very different from one another. Indeed group A evolves rapidly and its modulus reaches more than 3000 MPa after 60 days, when the one from group B reaches a value of about 2300 MPa for the same period. The second curing parameters applied result in different trends. The modulus of samples from group A remains constant while the modulus of group B increases. For both conditioning procedures, stiffness increases slowly during the cold period and more quickly during a warm period. Finally, the modulus values for both groups are similar at the end of the curing process, which shows that seasons sequence does not influence the curing of the material after about 130 days throughout the modulus.
Moreover, these evolutions can be compared to the water content in the samples (Fig. 6). Actually, in a general way, the modulus is lower when the water content is higher, and vice versa, when the water content is low, the modulus is higher. The same trends have already been examined for cold recycled bitumen emulsion mixtures containing cement [10] and for foamed asphalt cold recycled mixtures [11]. Besides the important stiffening at the beginning of curing for group A happens at the same time as the fast water loss of the samples (Fig. 1-a). In addition, the sudden increase of stiffness at the very beginning of the warm season for group B (around 70 days) can be correlated to the sharp decrease of water content at the same time (Fig. 1-b). But the presence of water does not explain the whole behaviour of the material as it continues to evolve even when the water content is stable [12].
A change in the void content could also be correlated to the height variations and therefore influence the modulus: a decrease in void content would lead to a decrease in height thus an increase in modulus; whereas an increase in void content would result in an increase in height and a decrease in modulus. For example such as seen in the 10 °C – 80 %RH period of group A where the change in height would counteract the stiffening than happens normally over time.
Additionally, the void content results mentioned above could only explain part of the stiffening as the final and initial states of samples from both groups A and B are similar while the void content variations are higher for group A than for group B.
The phase angle represents the delay between the stress and strain signals. The initial and final states of the phase angle is similar for both groups. This means that, as for the oedometric modulus, the seasons sequence does not impact the final properties of the material, at our time scale. Two trends can be observed on Fig. 7. First of all, the phase angle decreases with time during a warm simulated season (35 °C – 20 %RH). A decrease of phase angle during curing has been shown on cold asphalt mixes with RAP and with and without cement [13]. For group A, it starts around 27 ° and decreases to 23 ° at the end of the two first months of curing. The trend of group B is trickier to assess as the angle also seems to decreases at 35 °C – 20 %RH (3rd and 4th months) but each point remains in the standard deviation range of each other. Cold curing conditions (10 °C – 80 %RH) do not influence the phase angle. Indeed group A reaches 23 ° and 22 ° at the beginning and at the end of the simulated season respectively, while group B attains 27 ° and 26 ° respectively.
Moreover this data is in the same scale as can be found in the literature for cold mix asphalt [6].
Extracted binder evolution
As mentioned above, the comprehension of the physicochemical evolution of the binder could be extremely useful to grasp the whole mix behaviour and better understand the mechanisms at play during the curing process. The rheology, glass temperature and oxidation bring information on the chemical variations and ageing of the binder.
Rheological behaviour
Shear modulus and phase angle
Figure 8 highlights the differences in stiffening kinetics induced by the two different curing parameters. The literature shows that the shear modulus of a binder extracted from on-site cold mix asphalt increases with time [8], which is consistent with our results since the modulus of both groups has increased in four months. The standard deviation between the two furthest initial (t = 0) points is used to decide whether the following points are significantly different or not. For each mean value, this standard deviation is subtracted or added in order to get an upper and lower limit. If the modulus value of a point at curing time tn is included between the two limits of the previous point tn−1 then the studied point tn is considered having a similar modulus value than the tn−1 point.
Group A exhibits a fast increase of the complex shear modulus during the 60 first days at 35 °C – 20 %RH and its stagnation during the next 60 days at 10 °C – 80 %RH as the modulus values at these curing parameters are considered similar.
As for group B, the shear modulus slightly decreases with time during curing at 10 °C and 80 %RH as the first and last points of these curing parameters are considered significantly different. During curing at 35 °C – 20 %RH, the modulus then rises and reaches a value of about 7 MPa after 90 days in a similar way than the one for group A at the same curing parameters. This evolution would undoubtedly lead to the same final state for both groups.
The significance of deviation between the phase angle points has been assessed the same way as previously on the shear modulus. The phase angle of the extracted binders of both groups does not highly evolve as it stays around 47–52° for group A and between 52 and 59° for group B (Fig. 9). Curing conditions at 35 °C and 20 %RH generally cause a decrease in the phase angle on both groups, regardless of their position in the four months of curing. The influence of the 10 °C – 80 %RH curing conditions is trickier to examine: for group A (last two months of curing) there is no significant variation during those curing conditions. Phase angle values of group B (two first months of curing) slightly increase with time.
Thus the warm curing conditions (35 °C – 20 %RH) give the same response for the mix and the binder: a decrease of phase angle. This was expected since the viscous behaviour of the mix is caused in majority, if only, by the binder behaviour. Besides, the phase angle of the binder is higher than of the mix. This is due to the presence of aggregates in the mix that give a more elastic behaviour to the mix (lower phase angle) than the binder.
Representation in the Black space
The Black diagram of group A is illustrated in Fig. 10. Points 1, 2, 3 and 4 concerning binders which curing was stopped at 0, 15, 30 and 62 days respectively, during the warm period, show a stiffening and a decrease of phase angle with time. A stiffening is consistent with the literature and [14] showed that this progression is consistent with a physical hardening. The last two points (5 and 6, 91 and 120 days of curing, 10 °C – 80 %RH) indicate a stagnation of the binder.
Figure 11-a shows a light flattening between the curves at 0 and 93 days of curing (loss of phase angle and gain in modulus). This phenomenon has already been shown in the literature [4] and is consistent with an ageing of the material. This is not visible on group A, which shows that group B has started to age but not group A. This difference between the curves is still small, the studied time scales are definitely too short to see a stronger ageing.
The series of the first 4 points (0, 14, 29 and 59 curing days at 10 °C – 80 %RH) on Fig. 11 show an even more significant trend to the one seen on Fig. 10 for the same simulated season: during the 2 months at 10 °C and 80 %RH, group B has lost stiffness and gained phase angle where A has not. This trend for group B (a gain in phase angle in addition to a loss of modulus) has never been noticed in the literature, to the author’s knowledge. This can be explained by the fact that the extracted binder of a cold mix cured in cold conditions (10 °C – 80 %RH for instance) has never, once again to the author’s knowledge, been studied. This difference in kinetics between A and B during cold conditions may be related to their initial states : group A has already been through a warmer period which may have started physicochemical processes that B has not seen yet, which would explain the inertia of softening seen on Fig. 10 between points 4 and 5. Then, after the transition to 35 °C – 20 %RH for group B (point 5), the binder gets even stiffer and elastic than its initial state, which implies a certain reversibility in the mechanisms which occur during a cold period.
Generally, the binder properties vary less when the cold curing conditions happen in second. A cold season in the first 2 months of laboratory curing allows variations in modulus and phase angle, whereas in the 2 last months there is no significant change in the binder rheology.
It is important to keep in mind that here the curing is only of 4 months in a climatic chamber this is why the data changes are very low. What would be interesting would be to model or test longer curing times, where the changes would be far more significant, and see how the trends would evolve.
Glass temperature, crystallised fractions and oxidation levels
The glass transition and the crystallised fractions (Fig. 12) do not significantly change during the whole curing process. Group A glass transition starts at -24.9 °C, then reaches − 23.1 °C at the end of the first two months (35 °C – 20 %HR) and decreases to -24.3 °C at the end of curing. An amplitude of less than a degree is not significant given the binder extraction procedure and the measuring process that could result in slight uncertainties. Moreover, the maximum amplitude (seen between 15 and 91 days of curing with respectively − 26.2 and − 22.6 °C of glass transition) is quite close to the difference between the two groups at initial state (-24.9 °C for A and − 27.3 °C for B), even though they should be considered similar as their composition and mixing protocol are identical. Concerning the glass transition evolution of group B, it starts at -27.3 °C (initial state), reaches − 27.8 °C after 59 days of curing and ends at -27.4 °C, which is also not considered significant. The mean maximum amplitude, reached between 29 and 59 days of curing is about 2.6 °C, which is also close to the difference between the two groups at initial state. Thus this is considered not significant. The evolutions of both groups are then considered as stable.
Concerning the crystallised fractions, their content stays around 3 % all along the curing process for the two groups. This shows that these fractions do not vary during the four months of curing and that the binders have still not aged significantly. Indeed, literature shows that these fractions increase with bitumen ageing [15].
Infra-red oxidation levels were measured on the extracted binders of the samples from groups A and B. Figure 13 shows the carbonyl (Fig. 13-a) and sulfoxide (Fig. 13-b) oxidation indexes of these specimens at different curing times.
The initial and final oxidation levels of both groups A and B are equivalent. Moreover the kinetics of oxidation for both groups at warm period are similar as well as the ones during cold period. During warm period (35 °C – 20 %RH, first two months of group A and last two months of group B), the carbonyl and sulfoxide oxidation indexes significantly increase with time. During cold period (10 °C – 80 %RH, last two months of group A and first two months of group B) they tend to be constant which shows that there is no oxidation mechanism during a cold weather. These results show the effect of a temperature of 35 °C, that accelerates oxidation.
The increase of stiffness and oxidation during curing, especially during 35 °C – 20 %RH shows an oxidation of the binder. Colder conditions (10 °C – 80 %RH) seem to slow down this evolution.