Influence of synthesis parameters on properties and characteristics of poly (urea-formaldehyde) microcapsules for self-healing applications

on


Introduction
Some biological components such as bone or DNA present self-healing ability (Fratzl 2007). Materials that possess the ability to mimic biological components are known as self-healing materials. The main characteristic of these materials is their capacity to recover their proprieties after damage such as cracks occur (Blaiszik et al. 2010). Materials are normally susceptible to cracks during their service time, which may lead to degradation of mechanical properties and ultimately to failure. Thus, crack damage often results in high maintenance costs associated with shutdowns for repairs or even component replacement. One way to avoid such problems would be to mimic biological materials by producing materials with self-repairing ability (White 2001, Wool 2008, Hager et al. 2010, Zhang and Rong 2012, Binder 2013. In polymers and polymer-matrix composites, selfhealing agents can be incorporated to the matrix by two distinct ways according to the storage mode: (i) microencapsulated systems and (ii) vascular systems (Wu et al. 2008). The most studied self-healing system uses microencapsulation technology. This system was proposed by White et al. (2001), and comprises poly (urea-formaldehyde) microcapsules filled with dicyclopentadiene (DCPD) as healing agent embedded in an epoxy matrix. Another important component of this self-healing system is a solid catalyst (Grubbs' catalyst first generation), also embedded in the matrix. The catalyst is responsible for initiating self-healing reaction upon damage. The self-healing system proposed by White et al. (2001) is an autonomous process. In this case, damage works as a trigger in breaking microcapsules and releasing DCPD. The healing agent flows by capillarity to the damage area and undergoes ring opening metathesis polymerisation (ROMP) once in contact with Grubbs' catalyst. The resulting polymer, poly(DCPD) bonds the crack planes. Due to excellent compatibility between the matrix and poly(DCPD), the initial properties of the damaged material such as toughness, strain and stiffness can be recovered (White et al. 2001).
Microcapsules used in self-healing systems must be synthesised taking into account some main characteristics. Microcapsules need to possess a rough surface to ensure good adhesion between the polymeric matrix and polymeric microcapsules with no or low permeability to enclose the healing agent (in order to avoid diffusion), and have appropriate size and amount of filler material. Another important characteristic of microcapsules is a suitable wall thickness which is hard enough to endure matrix processing, yet fragile enough to be broken by a crack (Ollier et al. 2013). Finally, once embedded in a matrix, microcapsules must remain for a considerable period of time, being immune to healing agent leakage and diffusion until damage occurs (Brown et al. 2003).
Although widely described in the literature, no conclusive information has been presented on the importance of synthesis parameters for the processing and final characteristics of poly(urea-formaldehyde) microcapsules. Synthesis parameters (e.g. temperature, pH of the solution, agitation rate, emulsifier type, among others) are known to affect microcapsule characteristics and properties (Fan and Zhou 2010, Fan et al. 2013, Nesterova et al. 2012). An example of the influence of synthesis parameters on microcapsule characteristics is the variation of agitation rate which can produce microcapsules with different sizes and shell thickness. On the other hand, size and shell thickness may also be influenced by the pH of the synthesis.
Therefore, this study investigated the effects of pH of the solution and agitation rate on poly(urea-formaldehyde) microcapsule size and shell thickness. DCPD was used as healing agent for the produced microcapsules and the design of experiment (DOE) statistical method was used to investigate the influence of the synthesis parameters on the microcapsule characteristics. By manipulating multiple inputs at the same time, DOE can identify important interactions that may be missed when experimenting with one factor at a time. Microcapsules could then be produced using synthesis optimised conditions for self-healing applications through this statistical tool. Samples with microcapsules embedded were subsequently produced and damaged to evaluate the self-healing effect.

Design of experiment (DOE)
A factorial design was used to evaluate the influence of the solution pH and agitation rate (input variables) on microcapsules size and shell thickness (output variables). The design of experiments was composed of 2 2 experiments plus three central points. A total of seven experiments were conducted. Variable levels were selected based on previous synthesis and data found in the literature (White et al. 2001, Cosco et al. 2007, Blaiszik et al. 2008, Fan and Zhou 2011, Fan et al. 2013, Ahangari et al. 2014. The 2 2 factorial design is presented in Table 1, with level values and their respective coded levels for high, low and centre points. The output variables, mean shell thickness and mean microcapsule sizes were obtained by scanning electron microscopy (SEM). Size distribution was calculated based on data sets of 70 measurements and shell thickness was calculated based on data sets of 10 measurements in approximately eight different regions for each sample. The size of the area measured was approximately 5.4 cm (width) Â 4.5 cm (height). SEM analyses were performed in a Hitachi Tabletop TM-3000 scanning electron microscope. Microcapsules were placed onto a carbon conductive adhesive tape. ImageJ (version 1.51n, Bethesda, MD, USA) freeware was used to calculate microcapsule size distribution.
The statistical analysis of the data was carried out by ANOVA (analyses of variance), using Protimiza Experimental Design (https://experimental-design.protimiza.com.br; Protimiza Software 2017) online software. The significance of the input variables effect on the output variables was determined at a p < 0.10 level of significance.

Synthesis procedure
Microcapsule syntheses followed the procedure described by Brown et al. (2003) with a few modifications such as: 50 ml of 1.0 wt% aqueous solution of EMA, 20 ml of core material and agitation rate and pH of the solution were used according to the factorial design (Table 1). The 1:1.9 molar ratio of formaldehyde to urea proportion was maintained.
A solution was prepared using 150 ml of deionised water, 2.5 g of urea, 0.25 g of resorcine, 0.25 g of ammonium chloride and 50 ml of 1.0 wt% aqueous solution of EMA mixed in a 600 ml beaker under magnetic agitation. The pH of the solution was adjusted using NaOH (10%) and HCl (10%) according to the factorial design (Table 1). DCPD (20 ml) was added, followed by two drops of 1-octanol. The agitation rates are shown in Table 1. Next, 5.76 ml of formaldehyde were added while temperature was increased to 55 C. The solution was maintained under continuous agitation at 55 C for 4 h.
Microcapsules were filtered using a vacuum pump at the end of each synthesis. Quantitative filter paper with pore size of 25 mm was used to filter microcapsules. The microcapsules were then washed using deionised water and ethanol, and air dried at room temperature for 48 h. Finally, the microcapsules were collected and characterised by SEM.

Sample production and self-healing test
Microencapsulated dicyclopentadiene (DCPD) (2 wt%) and Grubbs' catalyst (0.2 wt%) were embedded in the epoxy matrix. Samples of neat epoxy and epoxy containing only microcapsules (2 wt%) without catalyst were also prepared. Samples were subsequently cracked using a guillotine-type equipment. The specimens are clamped at the ends and indented at the centre with a blade, thus inducing micro cracking.
Crack formation was observed by optical microscopy (OM). Since DCPD does not polymerise at room temperature, cracks were also observed after samples had been placed during 1 h in an oven at 50 C. This conditioning process was necessary to ensure that DCPD released from microcapsules was in liquid state to perform self-healing. Finally, samples were analysed inside the crack region using a Fourier-Transform infra-red (FTIR) microscope iN10 equipment from Thermo Scientific, with 100 mm resolution, DTGS detector and ATR mode.
The existence of peaks related to groups present in the poly (DCPD) in substitution to the peaks referring to groups related to DCPD in the FTIR analysis can indicate that there is self-healing in the damaged area of the sample. The presence of poly (DCPD) indicates the occurrence of reaction between DCPD and Grubbs catalyst after microcapsule cracking and heating of the sample.

Results and discussion
Figure 1(a) shows SEM images of microcapsules produced in synthesis 1 and the size distribution of microcapsules. This synthesis used a 1700 rpm agitation rate and pH of the solution at 4.5. The microcapsules were agglomerated and were not spherical in shape. The average shell thickness was about 2 mm, and it was possible to observe small particles and residues deposited on the microcapsule shell surfaces, in accordance to Brown et al. (2003), Then et al. (2011) and Fan et al. (2013). Free residue and free particles in the solution made the filtration process difficult. Figure 1(b) shows microcapsules produced in synthesis 2. In this case, the pH of the solution was set to 4.5 and the agitation rate was 1000 rpm. Microcapsules were agglomerated and were not spherical. Free particles were observed in solution and on the shell surface. The filtration process was difficult, probably because of those free particles and the solution residues which clogged the extractor filter paper. Microcapsules with average shell thickness of 1.77 mm were obtained. Moreover, a large amount of broken microcapsules were observed by SEM. Microcapsules probably broke during filtration and separation processes due to a thin shell wall.
The parameters used in synthesis 3 (Figure 1(c)) were a pH of the solution at 2.5 and 1700 rpm agitation rate, while synthesis 4 (Figure 1(d)) were a pH of 2.5 and 1000 rpm agitation rate. Although the effect of synthesis parameters time was not evaluated in this work, Brown et al. (2003) and Chuanjie et al. (2013) used 4 h as reaction time of urea-formaldehyde microencapsulation. Besides this, Cosco et al. (2006Cosco et al. ( , 2007 suggested that increasing reaction time from 4 to 6 h does not affect encapsulation yield. Microcapsules with a spherical shape were obtained for both syntheses. In addition, the microcapsules were not agglomerated. However, it was not possible to separate the microcapsules from the formed residue. The small amount of free particles in the solution resulted in a simple and efficient filtration process, as evidenced by Then et al. (2011). Thus, the processing parameters with a pH of 2.5 favoured microcapsule shell production instead of free particles. Also, small particles were deposited on the microcapsule shells, which increased shell thickness to approximately 4 mm.
All central points (syntheses 5, 6 and 7 with a pH of 3.5 and 1350 rpm agitation rate, Figure 1(e), (f) and (g), respectively) produced similar results. The microcapsules produced using this condition showed the best results in terms of filtration and amount of microcapsules produced. The pH at 3.5 resulted in the lowest amount of residue, which facilitated the filtration process. Microcapsules were disagglomerated, easily separated and had a smooth shell surface with few small particles deposited on top of it.
The design of experiments matrix with the coded input variables for high and low levels and centre points is shown in Table 1. The output variables are mean microcapsule size (microcapsule diameter) and mean shell thickness. Figure 2 illustrates how the measurements of microcapsule size (data sets of 70 measurements) and shell thickness (data sets of 10 measurements) were carried out from SEM analyses.
Analyses of residual sum of squares (residual SS) were performed to evaluate the influence and significance of pH of the solution and agitation variation on output variables. As polymerisation reactions present high fluctuation, 90% confidence level was used and p values inferior to 10% (<0.1) were considered significant, as recommended by Rodrigues and Iemma (2014).
Statistical evaluation for the output variable: mean microcapsule size (mm) Figure 3(a) shows the Pareto chart for mean microcapsule size. A statistical analysis based on residual sum of squares at 90% confidence was used. According to the chart, agitation rate was the only input variable with significant effect over the output variable. The chart shows that increasing agitation rate might cause a decrease in mean microcapsule sizes. The interaction effect between pH and agitation rate on microcapsule size was not significant. Thus, this output variable can be predicted by the model described in the following equation: Mean microcapsule size ¼ 79:33 À 28:67ÃAgitation rate (1) Table 2 shows the analysis of variance (ANOVA) for mean microcapsules size model based on residual SS.
The analysis of variance (ANOVA) showed that the model found for mean microcapsule sizes was not statically significant. The calculated F value (7.37) is a bit higher than the F value from the ANOVA table, at 90%. Also, the determination coefficient (R 2 ) is very low (0.5957) for the process. It was possible to conclude that the agitation rate significantly affects mean microcapsule size, even though the model was not effective in representing the process. Thus, agitation rate is one of the main variables responsible for microcapsule size control during synthesis, and the microcapsule size decreases when the agitation rate increases. Figure 3(b) shows the contour surface of mean microcapsule size. The influence of the agitation rate can be observed in the chart.
Statistical evaluation for the output variable: mean shell thickness (mm) The Pareto chart of the mean shell thickness output variable is presented in Figure 3(c). A statistical  analysis based on residual sum of squares at 90% confidence was used for this output variable. In this case, the effect of the input variable studied, agitation rate, on mean shell thickness was not significant. Hence, the model found can be described by the following equation: Mean shell thickness ¼ 3:72 À 1:53ÃpH þ0:56ÃpHÃAgitation (2) Table 3 shows the analysis of variance (ANOVA) for mean shell thickness model based on residual SS at 90% confidence level. According to ANOVA, the model found for mean shell thickness is valid (Equation (4)). The F calculated (16.35) value is higher than F from the ANOVA table (4.32). In addition, this model has a very high R 2 (0.8916). The chart of predicted and observed values is shown in Figure 3(d).
According to the obtained model (Equation (2)), the pH of the solution is the variable that most influences microcapsule shell thickness, followed by the interaction between pH and agitation rate. Lower pH values resulted in thicker microcapsule shells. The pH influencing phenomenon can be observed on the contour surface chart in Figure 3(e). Synthesis-adjusted conditions of pH of the solution and agitation rate were selected according to DOE. In this case, central points were found as the parameters with which the best microcapsules were produced. In accordance to Nesterova et al. (2011), microcapsules prepared in this synthesis condition could be easily filtered, separated and could be manipulated without being broken. Microcapsules also had size distribution and shell thickness that could resist processing conditions of thermoset polymers and polymeric matrix composite materials. Therefore, microcapsules produced at central point levels were chosen to be used in self-healing tests embedded in an epoxy matrix, along with Grubbs' first-generation catalyst. Figure 4 shows the epoxy sample containing microcapsules and Grubbs' catalyst before and after heating. Grubbs' catalyst and agglomerated microcapsules can  Figure 4. MO images before and after self-healing process (100Â). Arrows indicate where self-healing process were more evident.

Self-healing tests
be observed on the sample surface ( Figure 4). The image suggests a significant reduction of the crack in the epoxy matrix. SEM images of the crack region in Figure 5(a,b) suggests that starting point was closed after the selfhealing promoted by heating. Figure 5(c) shows that the crack length was about 3 mm and some microcapsules agglomerated over a specific area (indicated by an arrow) away from the crack. This region is detailed in Figure 5(d). Figure 5(e,f) shows the material with some microcapsules ruptured by the crack while others surrounding the crack remained intact. From these images, it can be concluded that microcapsules produced at central point conditions hold adequate shell thickness in order to resist matrix processing and to be ruptured by a crack (Haiyan et al. 2012).
Another important consideration is related to microcapsule size. A microcapsule size of around 60 mm was three times larger than the crack face separation. Thus, when intercepted and ruptured by crack propagation, these microcapsules can carry enough healing agent to entirely fill the crack.

FTIR
According to optical microscopy (OM) analysis, selfhealing could not be entirely confirmed. However, FTIR analyses were performed to verify the presence of poly (DCPD) inside the cracks' planes in the samples. The FTIR spectrum of samples (neat epoxy, epoxy containing microcapsules and epoxy containing microcapsules plus Grubbs' catalyst) is shown in Figure 6. Knowing the characteristic peaks of DCPD and PUF wall shell material, it is possible to observe that the spectrum shows the main peaks attributed to the characteristic functional groups such as amide I, II and C¼O at 1650-1550 cm À1 . The spectrum also shows peaks of CH 2 OH, CH 3 and CN groups at 1400-1360 cm À1 (Zorba et al. 2008. The spectrum of microcapsules not only defines the characteristic peaks of PUF, but also shows the characteristic peaks of DCPD, which indicates that DCPD is microencapsulated with PUF. The spectrum of DCPD reveals C-H stretching vibration peak at about 2923 cm À1 and 2852 cm À1 . Peaks at about 1670 cm À1 and 1736 cm À1 also indicate the presence of DCPD. Along with the poly (DCPD) curing, the DCPD concentration was reduced and the structure of cyclic olefin of ¼C-H bonds changed into non-cyclic olefin structure, which should have been detected by IR spectrum at these peaks. In comparing this with the DCPD spectrum, the reduction in these peaks' intensity of poly (DCPD) was in fact a result of the cyclopentene double bond opening, and formed a cross-linked structure of poly (DCPD) (Sun et al. 2017).
Some characteristic peaks of the composite did not show considerable changes with respect to peaks of samples containing only microcapsules due to the wide crack face separation, low amount of microcapsules on the crack site (caused by poor dispersion of microcapsules in the sample), and low amount of catalyst in the matrix. These were the main reasons for the limited self-healing ability. Based on these results, new strategies for manufacturing samples with the objective of increasing dispersion of microcapsules and catalysts are proposed.

Conclusions
Design of experiment results showed that the microcapsule size tends to increase when the agitation rate decreases. In the other hand, when pH increases there is a reduction in the mean shell thickness. The agitation rate of 1350 RPM and pH of the solution at 3.5 led to microcapsules with adequate mean size ($60 mm) and the mean shell thickness of 4 mm.
Microcapsules produced under these conditions were tested in epoxy samples and SEM and FTIR analyses indicated significant reduction of the damage in the epoxy matrix, as well as the presence of PDCPD inside the crack planes, suggesting the occurrence of self-healing.