Ordinary Portland cement mortar was used for all specimens in this research. Sand with a grain size of 0-2 mm, cement type CEM I 52.5 and water were mixed with a respective mass ratio of 6:2:1 to obtain a mortar with a water-to-cement ratio of 0.5.

The healing mechanism in this research requires a set of brittle capsules filled with a healing agent to be distributed in the cementitious matrix. The capsules were cut to a length of 50 mm from long borosilicate glass tubes with an internal diameter of 3 ± 0.05 mm and a wall thickness of 0.175 ± 0.03 mm. After cutting the capsules to the desired length, they were filled with a healing agent and the ends were sealed with a methyl methacrylate (MMA) based glue. Two different polyurethane (PU) precursors were used as healing agents in this study. The first PU precursor is a non-commercial product with a viscosity of 6700 mPas at 25°C. It was developed in the framework of another project (SHEcon) and has been used in previous research as healing agent for artificially created cracks (35, 36). The second PU precursor is a commercially available one with a viscosity of 200 mPas at a temperature of 25°C. Also this second precursor has already been used in previous research for self-healing of cementitious materials (33, 36, 37). Further details about both polyurethane precursors can be found in (36). Since the viscosity of the first healing agent is more than 30 times higher than the second one, the first PU precursor is denoted as PU_HV (High Viscosity) and the second one is denoted as PU_SLV (Super Low Viscosity) in this paper.

Reinforced mortar prisms with dimensions of 40 mm x 40 mm x 160 mm were cast in wooden formworks. Two screw threaded reinforcement bars with a diameter of 3 mm were placed in the molds in order to be able to create a controlled crack in the mortar prisms. The bars were positioned at 10 mm from the bottom of the specimens and at 10 mm from the sides. The specimens with the autonomous healing mechanism contained two capsules filled with a PU precursor. The capsules were fixed in the center of the specimens by gluing them on thin wires between the reinforcement bars. Figure 1 shows the preparation of the molds for the specimens with self-healing properties. When the molds were prepared, they were half-filled with mortar. This first layer of mortar was then compacted on a vibration table for 30 seconds. Subsequently the molds were completely filled with mortar and vibrated again for another 30 seconds. After that, the specimens were put in an air-conditioned room with a temperature of 20 ± 2°C and a relative humidity of at least 95%. 24 hours later, the mortar specimens were demolded and stored in the same environment for 27 days (curing period).

Cracks were introduced in the mortar prisms by means of a three-point bending test. Two different crack widths were targeted during the bending test to evaluate the healing efficiency for different crack widths. In order to register the width of the created crack during the bending test, a linear variable differential transformer (LVDT) was attached at the bottom of the sample using a small metal block and a thin metal plate. For some specimens a target crack width of 100 µm was desired. Therefore, the specimens were loaded at a rate of 3 µm/s until a crack width of 400 µm was obtained. Subsequently, the specimens were unloaded, causing the crack width to decrease to a value of around 100 µm due to the presence of the reinforcement bars. For some other specimens a target crack width of 300 µm was desired. These specimens were loaded at the same rate as previously mentioned, but until a crack width of 900 µm was reached. During unloading the crack width then decreased to a value of maximum 300 µm.

For the specimens with the autonomous healing mechanism, the capsules broke due to the creation of the crack. Breakage of the capsules could be noticed during the bending test by popping sounds. Moreover, the capsule breakage was noticed by a sudden small drop in the force during loading of the specimens. After crack creation, all specimens were stored in a climate room at a temperature of 20 ± 2°C and a relative humidity of 60% for 48 hours. During this period, the polyurethane that was released in the crack of specimens with the autonomous crack healing mechanism was given the time to harden inside the crack.

A total of 35 mortar specimens were prepared. An overview of the test series with their designation and the number of specimens per series is given in Table 1.

Due to the fact that the crack width reduced during unloading and the fact that the LVDT was still a few millimeters away from the bottom of the specimen, the final crack width could not be accurately detected from the last reading of the LVDT during the bending test. Therefore, the width of all cracks was measured microscopically using a stereomicroscope. Seven microscopic images were taken along the length of the crack with an intermediate distance of 5 mm. The crack width was measured three times on each image using the program ImageJ. Finally, the average width was calculated to represent the crack width of the crack in each sample.

A capillary water absorption test was performed in a climate room at 20 ± 2°C and 60% relative humidity based on the European standard NBN EN 13057 (38). Before the start of the test, the specimens were preconditioned in an oven at 40 ± 2°C until a constant mass was reached. A mass change of less than 0.1% over a period of 24 hours was defined as constant mass. After that, the specimens were put in a climate room at 20 ± 2°C and 60% relative humidity for 24 hours before the start of the water absorption test. During that period, the lateral faces and the bottom (test) face of the specimens were partly covered with aluminum butyl tape (schematically illustrated in Figure 2). In this way, a confined region of 10 mm x 40 mm around the crack (or in the middle of the bottom surface in case of the uncracked specimens) was exposed to water during the absorption test.

Before starting the water absorption test, the initial mass of all specimens was registered by weighing them on a scale with an accuracy of 0.01 g. The specimens were then placed in a plastic container with water on two line supports so that they were immersed up to a depth of 2 ± 1 mm. The mass of the specimens was then recorded every half hour during the first eight hours of the test and after this first period the specimens were weighed again after 24, 48, 72 and 96 hours of water exposure.

By subtracting the initial mass of a sample from the recorded mass, the cumulative water uptake was calculated. Subsequently, the cumulative water uptake was divided by the surface area exposed to water during the absorption test. The cumulative water absorption per unit surface area (g/cm²) was then plotted for each specimen separately in function of the square root of immersion time (h^0.5). The gradient of the line from the intercept of the water absorption curve to the cumulative mass of the water uptake per unit area recorded at 24 hours was then defined as the sorption coefficient S (kg/m²/h^0.5).

After performance of the capillary absorption test, the specimens with a healed crack were split at the location of the crack in order to evaluate the polyurethane coverage on the crack. This was done in order to have an indication to what extent the PU precursors were able to fill up the crack. Those results could then be linked to the performance of a specimen during the capillary absorption test.

In order to quantitatively determine the surface area of the crack faces that was covered by polyurethane, the hardened PU on the crack surfaces was colored red and a photo of both surfaces was taken. Using the program GIMP, the total area that was colored red on the crack faces and the total cross sectional area of the specimens was determined. In this way, the percentage of the cross section covered by polyurethane was calculated.

During the three-point bending test it was seen that the polyurethane precursor was released in the crack almost immediately after breakage of the capsules. The precursor then reacted with moisture and solidified inside the crack. A few examples of the crack healing are given in Figure 3. In case the crack is healed with the high viscosity polyurethane (PU_HV) the healing of both relatively small cracks (crack width < 100 µm) and relatively large cracks (crack width > 200 µm) is quite similar (Figure 3a and b). A foaming reaction occurred during hardening of the precursor. Due to this foaming reaction and the high viscosity of the PU precursor, the crack was not always completely healed along the whole crack length. Mostly the cracks were found to be healed quite well in the middle part of the specimen, but sometimes remained unhealed near the edges of the crack (see Figure 3a and b).

Figure 3c shows an example of crack healing with PU_SLV for a limited crack width (< 100 µm). It can be seen that a part of the mortar surface next to the crack was covered with polyurethane. This leakage of polyurethane out of the crack was already noticed during the three-point bending test. It can be attributed to the fact that the PU_SLV precursor flows into the crack quite easily due to its very low viscosity and the crack volume is relatively small. Consequently, a part of the PU flowed out of the crack.

In case a larger crack (> 200 µm) was healed with PU_SLV (Figure 3d), much less leakage of polyurethane at the bottom of the sample was noticed, because the crack volume was larger. It should be mentioned that even in the case of a crack with a width of 300 µm, the crack volume remained lower than the total amount of PU precursor in the two capsules. This was done, because from previous research it was noticed that not all of the PU precursor flows out of the capsules when a crack is created (39). Usually only the PU precursor in the central part of the capsules, where the crack causes the rupture, flows into the crack. However, even though not all of the PU precursor flows into the crack, a good crack sealing was noticed visually along the whole length of the crack for both relatively small (< 100 µm) and large (> 200 µm) crack widths.

In order to make a quantitative evaluation of the self-healing performance, the self-healing efficiency (SHE) was determined according to the definition given in Van Belleghem et al. (3). The self-healing efficiency of a mortar specimen [%] was defined as [1]:

Where S_CR,M is the mean sorption coefficient of the cracked specimens [kg/m²/h^0.5], S_HEAL,i is the sorption coefficient of an individual healed specimen [kg/m²/h^0.5] and S_UNCR,M is the mean sorption coefficient of the uncracked specimens [kg/m²/h^0.5].

The self-healing efficiency was evaluated for two crack width ranges. A ‘low crack width range’ and a ‘high crack width range’ were chosen so that at least three specimens of each series (CR, PU_HV and PU_SLV) were in each range. Specimens with crack widths from 62 to 83 µm were assigned to the ‘low crack width range’, specimens with crack widths from 214 to 295 µm were assigned to the ‘high crack width range’. The two crack width ranges are indicated by the vertical dashed lines in Figure 4.

As mentioned before, the mean value of the sorption coefficient of the uncracked mortar specimens was 1.89 kg/m²/h^0.5. For the cracked specimens in the ‘low crack width range’, the mean sorption coefficient was 6.22 kg/m²/h^0.5. With these two values, the self-healing efficiency of the autonomously healed specimens in the ‘low crack width range’ could be calculated using Equation 1. For three out of four specimens where crack healing occurred with PU_HV, the self-healing efficiency was between 63 and 78%. However, one specimen performed much worse. A self-healing efficiency of only 18% was found for this sample. It was also visually observed for this last specimen that only half of the crack at the surface of the specimen was filled with polyurethane. As mentioned previously, crack healing with PU_SLV resulted in a sorption coefficient that was lower than the one obtained for uncracked mortar. Consequently, a self-healing efficiency of more than 100% was found. For the three specimens in the ‘low crack width range’, the self-healing efficiency ranged between 116 and 124%.

In the ‘high crack width range’ the mean sorption coefficient of the cracked specimens amounted to 9.23 kg/m²/h^0.5. One of the specimens healed with PU_HV showed a very high self-healing efficiency of 108%. The other two specimens in this crack width range performed worse, but the self-healing efficiency still amounted to 53 and 70%. Clearly, a complete crack healing is possible with the high viscosity polyurethane, but it is certainly not always the case. Crack healing of cracks in the ‘high crack width range’ with PU_SLV on the other hand does always show a consistent (almost) complete crack healing. The self-healing efficiency of the five specimens in the ‘high crack width range’ ranged between 95 and 97%.

So far only the sorption coefficients have been used to evaluate the efficiency of the self-healing mechanism. This was a good parameter to compare the different test series, but it does not give complete information about the water absorption process in time. In this section the mean cumulative water absorption curves of the specimens in the ‘low crack width range’ and the ‘high crack width range’ are discussed. These curves are presented in Figure 5. From both graphs the major influence of a crack on the ingress of water is very clear. It can be seen that water enters very rapidly into the cracks at the beginning of the absorption test. At the first measurement, after half an hour of exposure to water, the mean total water absorption of specimens with a low crack width is already more than seven times higher than for uncracked specimens. For specimens with a high crack width this is even more than nine times higher. After 24 hours the water absorption rate in the cracked specimens starts to decrease as the specimens get faster to the saturated state. At the end of the test (at 96 hours of exposure) the total water absorption for specimens with a low crack width is nearly three times higher than the water absorption in uncracked specimens and four times higher in case of specimens with a high crack width.

Crack healing of specimens in the ‘low crack width range’ with PU_HV resulted in a capillary water absorption that was in between the water absorption of uncracked and cracked specimens (Figure 5a). The mean self-healing efficiency of this polyurethane for low crack widths is for all time steps between 51 and 58%. The self-healing efficiency was calculated at each time step according to Equation 1 by taking the values of the water absorption (g/cm²) instead of the capillary sorption coefficient. The large standard error of this absorption curve is attributed to the fact that a bad healing performance was found for one of the tested specimens, as discussed in the previous section.

Crack healing of small cracks with PU_SLV resulted in complete crack healing. The cumulative absorption curve is located completely under the curve of uncracked samples (self-healing efficiency ranging from 108 to 120%). During the first eight hours of the absorption test the water absorption rate of the healed specimens is lower than for the uncracked specimens, while in the period from 24 to 96 hours the sorption rate of healed and uncracked specimens is almost equal. As mentioned before, this can mainly be attributed to the fact that some polyurethane hardened on the mortar surface (see Figure 3c). The partial coverage of the test surface caused a very low sorption rate of the healed specimens in the beginning of the test.

For cracks in the ‘high crack width range’, crack healing with PU_HV resulted in a better healing performance than for low crack widths. The mean self-healing efficiency at all time steps was between 75 and 78%. Nevertheless, Figure 5b shows a very large standard error on the test results of the specimens healed with PU_HV. The large variability in healing performance was already noticed before by the large scatter on the sorption coefficients at high crack width (Figure 4).

PU_SLV was able to reduce the water absorption through cracks with a relatively high crack width up to the level of uncracked mortar during the first eight hours of the absorption test (self-healing efficiency of 99 – 100%). After the first eight hours however, the water absorption rate of the healed specimens started to be higher than for the uncracked specimens. The specimens healed with PU_SLV do not behave as good as uncracked specimens anymore, but the self-healing efficiency still amounts to 88% after 96 hours of exposure to water. A possible cause for this increase in sorption rate could be that only the bottom part of the crack (below the level of the capsules) was healed. Water could then have penetrated into the mortar matrix next to the crack and reached the upper part that was not well healed. However, the evaluation of the crack surfaces in the next section provides proof that there was good filling of the crack in both the part below and above the capsules. Therefore, the increase in water absorption was not caused by incomplete crack filling. Another possible reason is that crack creation might have induced some damage to the bond between the mortar and the reinforcement bars. After eight hours, the water front in the matrix next to the crack most probably reached the position of the reinforcement bars. Some damage at the interface between the rebars and the mortar in the crack area could have caused an increase in the water absorption rate. However, this last hypothesis remains uncertain and could be validated in future research through visualization of the water uptake with X-ray or neutron radiography experiments.

Figure 6 shows a few examples of the crack surfaces of healed specimens in the ‘low’ and ‘high’ crack width ranges after splitting. As mentioned before, the hardened polyurethane on the crack faces was colored red in order to be able to distinguish it better from the mortar matrix. For the representative specimens in Figure 6 the original crack width is noted along with the total percentage of the cross section that was covered with polyurethane. An overview of the coverage percentage of the cross section of all specimens is given in Table 2. The total coverage percentage (CP) was taken as the sum of the coverage of both halves. This is justified since the fracture always occurred at the contact surface between the polyurethane and the mortar, so part of the polyurethane stuck to one half and another part to the other half after splitting the specimen.

All coverage percentages mentioned in Table 2 are relative to the total cross section of the prism. However, for both small and large crack width ranges the cracks did not run through the whole cross section. This is most clearly visible for the cracks in the ‘low crack width range’ (Figure 6a and c): no polyurethane was found at the top half of the cross section. This is simply because the crack did not reach the top of the specimen. Consequently, it was not possible to find a coverage percentage of 100%, but it is possible however that there was complete crack filling.

In case of crack healing in the ‘low crack width range’ with PU_HV, three out of four specimens had a CP of only 29%. However for two out of these three specimens a sorption coefficient of around 3.5 kg/m²/h^0.5 was found, while a much higher sorption coefficient of nearly 5.5 kg/m²/h^0.5 was found for the third one. The main reason for the bad healing behavior of this last specimen is that one side of the crack was not healed. No polyurethane coverage was found on one side of the cross section. This shows that not only the CP of the cross section is important, but also a complete filling of the crack over the whole width cross section is necessary to obtain a good healing performance.

When the crack healing occurred with the PU_SLV in the ‘low crack width range’ very similar values of the CP were obtained: between 46 and 52%. These values are also higher than the values found for PU_HV. When comparing Figures 6a and 6c it is also visible that healing with PU_SLV results in a more complete filling of the crack over the whole width of the cross section. This is mainly the result of the very low viscosity of the polyurethane which allows it to flow more easily in the crack.

Healing of cracks with PU_HV in the ‘high crack width range’ resulted in a large scatter of the sorption coefficient. For the specimen with the smallest crack width in this range (231 µm) the largest sorption coefficient was found. This can be attributed to the small CP of 36% that was found on the cross section (Table 2). The other two specimens had a similar CP (56-57%), but the sorption coefficient of one of them was three times smaller than the sorption coefficient of the other. Evaluation of the cross section again showed that for the specimen that performed worse, no complete healing over the whole width of the cross section was found. Most probably that was the main reason why the sorption coefficient of that specimen was much higher.

Figure 6d shows that when crack healing occurred with PU_SLV for cracks in the ‘high crack width range’ there was a very complete filling of the crack over the whole width of the cross section. Coverage percentages between 73 and 83% were found for all five specimens in this series. These high values for the CP agree well with the low sorption coefficients that were found for these specimens. The increase of water absorption rate after eight hours of water exposure of these specimens mentioned in section 4.4 is thus certainly not caused by an incomplete filling of the upper part of the crack.