Design and optimization parameters of cement-based spherical macrocapsules for self-healing cement applications

. Encapsulated healing agents is a promising solution for extending the service life of critical infrastructure, providing long-term healing efficiency. This research focuses on the shell properties of cement-based spherical macro-capsules, aiming to achieve increased survivability during mixing of mortar mixtures and efficient triggering upon crack propagation. In this framework, the pan coating technique was examined for the production of capsules with a cementitious shell, developed for the protection of powder healing agents. The main properties that were studied included the crushing load as a function of capsules size and the shell hydration facilitated by different setting accelerators, and their consequent effect on the survivability and the triggering efficiency of the capsules. The results show that the use of setting accelerators allows the rapid densification of the shell microstructure and improves the crushing load of capsules, resulting in high survivability during mixing process. The enhanced compatibility of capsules with the matrix allowed the efficient triggering of capsules during crack propagation, initiating the autonomous healing process.


Introduction
Microcracking of cement and concrete infrastructure is an inevitable phenomenon that although initially does not affect the integrity of the structure, it allows the entrance of harmful substances that diminish its performance and lifetime. Self-healing of cementitious materials have gained significant attention as an effective solution to face the microcrack formation, without the need for external intervention, and thus additional maintenance and repair works [1].
To this direction, the development of encapsulated healing agents provides many advantages compared to other self-healing technologies (i.e. the vascular systems and the use of bacteria), such as tunable size and shell properties, multiple choices for healing agents and ease of incorporation in cement mixtures. Moreover, a crucial benefit is that capsules can be easily used in large scale applications, without the need of specific expertise and modification of the standard casting procedure. At the same time, the use of capsules minimizes the waste of healing agent, providing specific amounts of reactive components homogeneously distributed in the matrix [2,3]. However, several design parameters and performance requirements need to be fulfilled in order to ensure the efficient function of the capsules and optimize their performance and healing efficiency. For instance, the shell of the capsules should remain intact during mixing process of cement with water and aggregates, while at the same time it has to be easily damaged during crack propagation to activate the * Corresponding author: s.papaioannou@inn.demokritos.gr healing mechanism. Moreover, the shell should have the appropriate chemical and morphological characteristics in order to withstand the high pH values of cement and block the entrance of water inside the core of the capsules, enabling the long-term protection of the healing agent.
In this framework, several national and European research projects have been funded (SARCOS, SMARTINCS, RESHEALIENCE, RM4L, M4L, AKEISTHAI and HEALCON) bringing together researchers, engineers and scholars, working on different approaches. The research so far, shows the key role of the encapsulation methodology followed for the development of the capsules, since it controls their basic characteristics. Many chemical, physical and physicochemical techniques have been studied, such as polymerization, sol gel reaction, complex coacervation, 3D printing, extrusion and pan coating [4,5]. From the above methods, pan coating is considered very promising since it provides the ability to use materials of high chemical compatibility with the matrix [6], as well as to adjust the capsules properties in order to optimize their performance [7].
The present work aims at the study of the pan coating technique for macro-scale healing additives, focusing on the optimization of the mechanical behaviour and shell microstructure of cement-based capsules made by the pan coating technique. In this context, the effect of the type and concentration of the setting accelerator used for the shell formation was studied. The microstructure and hydration rate of the shell were examined in cross sections of capsules using Scanning Electron microscopy (SEM). Finally, the crushing load and survivability during mixing of the different types of capsules were studied and related to their morphological characteristics (circularity, shell thickness).

Materials and methods
Spherical cementitious capsules were produced according to the methodology described elsewhere [7].
The core particles were produced by pelletisation, where a mixture of fine white OPC powder (CEM II 52.5 R, Aalborg White) with 15 wt. % expansive admixture (DENKA CSA #20) is agglomerated into spherical particles by uniform water spraying. The particles between 2 to 4 mm were selected by sieving for coating.
The coating process was based on the pan coating technique. The core particles were added in the rotating drum and sprayed with tap water, while grey OPC powder (CEM I 42.5 N, TITAN S.A.) was simultaneously added, creating an external layer (shell). The thickness of the shell was increased by repeating this process two more times. Depending on the batch, the liquid used in the latter step was: tap water (W_01, W_02), sodium silicate (Na2SiO3, Sigma Aldrich) water solution (10 wt. %), alkaline setting accelerator water solution (12 wt. %) (DOMOGUNIT L, DOMYLCO Ltd.) or alkali free setting accelerator water solution (20 wt. %) (DOMOGUNIT AF, DOMYLCO Ltd.). The above concentrations were selected according to manufacturer's suggestions, after initial trials. The mineralogical composition of both cements used for the formation of the capsules can be seen in Fig. 1, while the different water solutions, their concentration and the name of the different batches of capsules are presented in Table 1. The core particles of the capsules were prepared using white OPC sprayed with tap water as described in [7]. In this case, white OPC was used for the core particles in order to facilitate the visual separation between the core and the shell of the capsules. The morphology of representative core particles and capsules produced in the laboratory can be seen in images obtained on a Leica S6D optical microscope as presented in Fig. 2.  Representative core particles made of white OPC and cement-based capsules observed in stereomicroscope.
The particle size distributions of the core particles and the capsules were determined by analysing images of the optical microscope, using an opensource image analysis software (ImageJ). For increased representativeness of results, at least 300 capsules, randomly selected from each batch, were analysed. Although sieves of mesh sizes 2 and 4 mm were used for the selection of the core particles, their actual size according to the image analysis ranged between 2 to 5.5 mm (Fig. 3). The deviation between the two methods is mainly attributed to the geometry of the capsules, which is not completely spherical, allowing the passage of larger ellipsoidal capsules through the sieve openings of 4 mm. After shell formation, the size of the core particles increased, and higher fractions were obtained in the larger diameters, as shown in Fig. 3.   Fig. 3. Modification of core particles size distribution after shell formation, showing the increase of the capsules' percentage (ratio %) in the higher diameters, due to the development of the external shell around the core particles.

Examination of the shell microstructure
The microstructure of the shell was examined by SEM in cross sections of representative capsules of each batch. Before inserted in SEM, the capsules were dried in an oven at 40 o C for 24 h and then cut, mounted on SEM stubs and carbon-coated. SEM analysis was performed using a FEI, Quanta Inspect equipped with an EDAX ultra-thin 211 window analyser and Genesys analysis software package under a 25 kV accelerating voltage.

Mechanical performance of capsules
Crushing load of capsules is one of the main properties that affect the survivability of capsules during their incorporation in mortar mixtures, as well as the triggering efficiency of capsules during crack propagation. Thus, the measurement of crushing load allows the modification of the manufacturing process in order to adjust crushing strength towards optimum survivability and mechanical triggering. The crushing load of the capsules was measured according to an inhouse developed setup described in our previous work [7], as seen in Fig. 4. The tests were conducted using a 100 kN Instron loading frame implemented under displacement control conditions, with a constant rate of 100 μm/min. At least 20 capsules with diameters in the range of 3.8-4.2 mm of each type of capsules were tested, aiming to minimize the effect of size on the resultant values and collect comparable data. The mean value and standard deviation of each batch was calculated and compared.
For the purpose of studying the effect of capsules diameter on the mechanical behaviour of capsules, AlkAcc_02 capsules were categorized into different sizes. The crushing load and standard deviation of each size were measured. The results were interpreted in conjunction to the corresponded shell thickness values of the capsules and circularity index. The shell thickness was determined by examination of the damaged capsules under the optical microscope, while the circularity index was expressed according to Eq. 1: where A is the projected area of the particle and Prough is the perimeter of the projection.

Triggering efficiency of capsules
Sufficient adhesion between the shell and the matrix in conjunction with appropriate mechanical properties of the capsules are required for ensuring the efficiency of the triggering mechanism. For this purpose, mortar mixtures with 20% of capsules (sand replacement by volume) were prepared, using sand:cement ratio of 1:3 and water:cement ratio of 1:2. Prismatic specimens of 2x2x8 cm 3 were prepared and left for curing for 28 days.
After the curing period of the specimens, they were left at room conditions for 24 h and loaded by threepoint bending at a constant rate of 100 μm/min until fracture. After loading, the specimens were manually split and samples were extracted, exposing the crack surfaces for SEM examination. SEM analysis was performed using a FEI, Quanta Inspect electron microscope (SEM-EDS) equipped with an EDAX ultrathin 211 window analyser and Genesys analysis software package under a 25 kV accelerating voltage.  Table 1 affected the shell microstructure, due to the different quantities of water and reactive compounds (sodium silicate and setting accelerators) sprayed.
The examination of the reference capsules W_01, revealed the formation of a porous shell, indicating that Core particles Capsules the quantity of water was not adequate for the hydration of the cement grains of the shell. As the water concentration increased, resulting in a water:OPC ratio equal to 3:10 (W_02), the hydration rate increased, however voids are presented between the hydrated cement grains. Replacement of water with sodium silicate did not improve in the porosity of the shell when low concentration was applied (SS_01), indicating the demand of higher quantity of water solution, and consequently sodium silicate to achieve improvement of microstructure. Indeed, a dense layer of hydrated cement was created on the external surface of the SS_02 capsules as a result of the reaction of cement particles and sodium silicate solution. Sodium silicate reacts with calcium hydroxide (CH) in the presence of water to form calcium silicate hydrate (C-S-H), the main product of cement hydration (Eq. 2). Below this external dense layer (approximately 20 μm thick), a relatively porous matrix of cement is observed. This indicates that the drying stage of the manufacturing process enabled the removal of the moisture from the internal of the shell, and resulted in the presence of an increased concentration of reactive cement phases in the internal layer of the ' shell. The homogeneity and porosity of the shell was not affected by the size of the capsules, since in all capsules the shell is formed layer by layer under the same conditions. 2Na + + SiO3 2-+ Ca 2+ + 2OH -+ Xh2O → CaO·SiO2·Xh2O + 2NaOH (2) Moreover, the effect of two types of setting accelerators (alkaline and alkali-free) on the shell microstructure was studied.
As shown in Fig. 5, the use of the alkaline accelerator resulted in the formation of a dense external layer on the capsules shell (AlkAcc_02), more than 40 μm thick. The alkaline accelerator is composed mainly of sodium aluminate, providing Na + and [Al(OH)4] − ions in the cement particles of the shell. During shell formation, [Al(OH)4] − ions react with Ca 2+ and SO4 2− ions present in the liquid phase of the cement, forming AFt (ettringite -C6AŜ3H32) and AFm (calcium monosulfoaluminate-C4AŜH12) phases, which reduce the setting time of the cement and increase the density of the hydrated products [8]. Further tricalcium aluminate (C3A) hydration proceeds with limited sulphate content and C-A-H phases might also be formed.
Using an alkali-free accelerator (AlkFreeAcc_02), elevated hydration of cement in the external layer of the shell was observed. However, compared to the alkaline accelerator, a less dense microstructure has been created. These differences were attributed to the chemistry of the setting accelerators. The main component present in alkali-free accelerators is aluminum sulphate, stabilized in an aqueous solution by the addition of an inorganic or organic acid. When this type of admixture is used in cement matrices, the mixing water is enriched with Al 3+ , SO4 2and H + ions. Due to the elevated pH in the paste, Al 3+ ions are converted into [Al(OH)4] -, which then reacts with calcium and sulphate ions present in solution. As this admixture also contributes to the increase of sulphate concentration in the liquid phase, ettringite is the main product formed in the matrix. Formation and growth of ettringite nanocrystals are the main processes that reduce setting times and increase the rate of strength development.

Effect of type and concentration of setting accelerators on the mechanical performance of capsules
Crushing load is a crucial design and performance parameter of capsules, since it affects their survivability during the mixing process and triggering efficiency during crack propagation. Crushing load values of the different types of cement-based capsules are presented in Fig. 6. As shown in the figure, capsules prepared with the lower solution:cement ratio (1:10 by weight), presented low crushing load values, of approximately 5 N. The lower SS:cement ratio used during the formation of SS_01 capsules resulted in a similar crushing load to the W_01 capsules, indicating the demand of higher quantity of water solution, and consequently sodium silicate, to achieve improvement of mechanical properties. When solution:cement concentration increased from 1:10 to 3:10 (by weight), a significant improvement of crushing load was observed for both types of capsules studied, those that were sprayed with plain water and those that were sprayed with sodium silicate solution. More specifically, comparing crushing load of SS_01 and SS_02 capsules, it is evident that the extra spraying of the capsules after the drying, significantly improved their performance. SS_02 capsules exhibited almost five times higher crushing load values (approximately 23 N) in comparison to SS_01 capsules, where the maximum load was calculated at almost 5 N. The higher sodium silicate content in SS_02 capsules had a beneficial effect on the enhancement of the shell's strength, due to the elevated amount of C-S-H produced by the reaction of silicate ions from sodium silicate with portlandite. This was also evident in Fig. 3 where a denser microstructure of SS_02 shell can be seen, compared to SS _01.
The effect of the different setting accelerators on the crushing load of the capsules was revealed with the increase of the concentration of the setting accelerator. The beneficial role of sodium silicate was verified by the increased crushing load of SS_02 capsules by ~130%, compared to the reference capsules (W_02) produced by spraying of plain tap water. Since the same methodology was followed, the higher crushing load is attributed to the reaction of sodium silicate with the cement phases presented in the shell. The use of the alkaline accelerator further increased the crushing load of capsules that reached the average value of 25.5 N, presenting an increase of ~150% compared to the reference capsules (W_02). Similarly, the use of the alkali free accelerator resulted in an increase of the crushing load of the reference capsules (W_02), however a lower value was calculated compared to the capsules sprayed by sodium silicate and alkaline accelerator. Aiming to examine the effect of setting accelerator on the capsules crushing load, a specific and narrow range of diameters of capsules was used from each batch. In order to further examine the effect of capsules diameter on their crushing load values, the crushing load of AlkAcc_02 capsules with different diameters was measured and the results are presented in Fig. 7. It is shown that as the capsules size increases from 2.5 to 5.5 mm, the crushing load of the capsules presents an increase of approximately 480% and exceeds 30 N. This is mainly attributed to the shell thickness of the capsules, which as can be seen in Fig. 8, increases from 100 μm to ~550 μm. In contrast to the core of the capsules that is composed of sparsely stacked raw cement grains, the shell is composed of a hydrated layer of cement which is responsible for the strength development of the capsules. Thus, the thickness of this layer and its hydration rate are the main parameters that determine the strength of the capsules.  Additionally, it is observed that as the size of capsules increases, higher standard deviation values were presented in the crushing load values, as a result of the lower uniformity in the geometry of the capsules. As can be seen in Fig. 9, the circularity index of the capsules decreases as their size increases, denoting that larger capsules present a less spherical geometry with more irregularities that act as weak areas. These areas are more susceptible and act as crack initiation points during loading. Thus, capsules of higher diameters present a higher variation on their crushing load values.
Survivability of capsules was evaluated by conducting a washing test after the mixing process, in order to collect the intact capsules that survived during mixing and calculate their survivability ratio, as demonstrated in Fig. 10. The survivability values depend on the mechanical properties of the capsules and according to the results, they agreed with the crushing load values, since as the capsules' strength increased, more capsules survived the mixing process. Fig. 9. Circularity index as a function of capsules' diameter. Capsules prepared using the lower solution:cement ratio (1:10 by weight), presented lower survivability during mixing compared to the corresponded capsules prepared using a solution:cement ratio of 3:10. For both types of capsules, made by water and sodium silicate solution (W_01 and SS_01), 63% of capsules survived during mixing, while when the concentration of the spraying solution increased, the survivability increased to 66 and 75% respectively, illustrating the beneficial effect of crushing load enhancement. The replacement of sodium silicate solution with an alkaline setting accelerator solution resulted in a further increase of survivability to 77%, leading to a total increase of 17%, compared to the reference capsules W_02 made by spraying the same concentration of water. Similarly, to the alkaline accelerator, the use of the alkali free accelerator resulted in an increase of survivability of the reference capsules (W_02) equal to 12%, however due to the lower crushing load, lower survivability was calculated compared to the capsules sprayed by sodium silicate and alkaline accelerator solution.

Triggering efficiency of capsules
Towards studying the triggering mechanism of cementbased capsules, fractured surfaces of the specimens containing AlkAcc_02 capsules were examined in SEM. All the capsules found in the examined sections demonstrated successful triggering during crack propagation, as the shell was mechanically damaged, providing access of water to the healing agent. A representative example of an effectively activated capsule can be seen in Fig. 11a, where the different components, the core and the shell of the capsule, the binder and the aggregates can be distinguished. Also, some pores and secondary cracks can be noticed in this image. As shown in Fig. 11b, good interfacial bonding has been developed between the cement-based shell and the matrix due to their chemical affinity and hydrophilicity of the shell. The balance between adhesion strength and mechanical properties of the cement-based capsules resulted in efficient mechanical triggering, eliminating delamination of capsules and crack deflection phenomena observed in other types of capsules [9][10][11]. Fig. 11. Examination of the triggering mechanism of AlkAcc_02 capsules in fractured surfaces of cement mortar specimens in SEM. The images show an efficiently activated capsule in BSE mode (a) and the interface between the shell and the matrix in SE mode (b), indicating their good adhesion and interfacial bonding.

Conclusions
1. Pan coating is an easy, low cost and environmentally friendly technique that can be used for the industrial production of encapsulated healing agents, providing the advantage of adjusting their properties to each specific application. 2. The spraying solution of the shell and its concentration highly affect the hydration rate, the microstructure and the strength development of the shell. 3. Using an alkaline setting accelerator for the stabilization of the shell, the porosity of the shell is reduced, forming a dense external layer, more than 40 μm. 4. The optimization of the production parameters of the capsules resulted in the increase of their crushing load, followed by a high survivability ratio, up to 77%. 5. The mechanical properties of the capsules are highly affected by their size, shell thickness and circularity. 6. Due to their good adhesion with the matrix, capsules were efficiently triggered during crack propagation activating the healing mechanism.