Recovery of electro-mechanical properties inside self-healing composites through microencapsulation of carbon nanotubes

We report the successful microencapsulation of multi-walled carbon nanotubes suspended in a 5-ethylidene-2-norbornene (5E2N) self-healing monomer, into poly melamine urea formaldehyde shells through in situ polymerization. The average size of the microcapsules, their size-distribution, shell wall structural integrity and thickness are characterized by optical and scanning electron microscopy. The presence of carbon nanotubes (CNTs) inside the core liquid content, as well as their release after breaking is confirmed by microscopy and spectroscopy analyses. A small amount of CNTs inside the microcapsules is found to have no significant impact on the thermal stability of the system, as determined by thermogravimetric analysis and differential scanning calorimetry. Both the mechanical and the electrical properties of CNT-based self-healing materials can be restored up to 80% when CNT/5E2N microcapsules are incorporated into polymer composites, thus making them highly suitable for applications in aerospace.


Conceptual Route of Self-Healing with CNT/5E2N Microcapsules
The successful microencapsulation of CNTs with self-healing monomer, such as 5E2N, can open multiple opportunities for self-healing applications in space environment. In addition to making polymer composites more resistant mechanically [40][41][42][43][44][45] , the use of CNTs can improve the electrical conductivity of the composites 38,40,41,46 , avoid sparking in fuel filter housings 47,48 , act as sensors for monitoring damages in high performance batteries [49][50][51] , as well as contribute to the development of innovative electronic devices, such as circuits printed on flexible sheets 52 , anticorrosive coatings and conductive adhesives 11 . Recently, CNTs have been increasingly tested in various components of future generation aircrafts in which polymer composites are already used to make a majority of their components and structures 2 . As some of commercial passenger airplanes (Boeing 787, Airbus A350 and Bombardier Cseries models) presently employ polymer composites for up to 50% of their total weight 2 , the microencapsulation of CNT/5E2N suspensions inside these materials could give them self-healing abilities that would improve both their reliability and durability.
Caruso et al. 13 successfully synthesized microcapsules containing single wall carbon nanotubes (SWCNTs) mixed with organic solvents inside poly urea-formaldehyde (PUF) vessels to restore the electrical properties of various polymeric matrices. However, the SWCNTs used in these suspensions did not disperse efficiently within the solvents, leading to aggregation 13 . Similarly, Pastine et al. 53 encapsulated 1 wt. % of multiwall carbon nanotubes (MWCNTs) suspended in toluene inside polyamide shells by interfacial polymerization, which is intended to produce remotely triggerable microcapsules. In these reports 13,50,53 , however, the majority of core materials embedded within these microcapsules were solvents, which cannot undergo polymerization reaction for self-healing of cracks or damages of composites.
Regarding the healing agent, Bailey et al. 54 were able to encapsulate epoxy with high concentration (up to 20 wt. %) of CNTs using ethyl-phenyl acetate (EPA) solvent for conductive coating applications. A conductivity recovery of 64% (±23%) and a structural recovery of 81% (±39%) were measured by in situ electro-tensile tests conducted on coatings and substrate systems. Nevertheless, the epoxy/CNT core material of the microcapsules was dissolved in a solvent that is not suitable for self-healing applications below 0 °C 54 . Figure 1 shows the conceptual route of the self-healing process originally introduced by White et al. 8 , using CNT/5E2N suspension as microencapsulated healing agents. After encapsulation of the CNT/5E2N suspension into spherical PMUF shells, the microcapsules along with Grubbs' catalyst were uniformly dispersed inside the polymer matrix.
When cracks form inside the composite due to ageing or damage events, they propagate towards the microcapsules, so that the shell walls of these MC break and release the CNTs with the liquid monomer into the crack through capillary action 8 . Once the CNT/5E2N suspension comes into contact with the dispersed Grubbs catalyst, the latter locally triggers the ROMP reaction. The CNT/poly (5E2N) nanocomposite forms at the crack site, which partially restores the nominal functionality of the material.

Materials and Methods
Materials. EPON 828 (a bisphenol-A based epoxide resin) and epicure 3046 (an amidoamine room temperature curing agent) were purchased from Miller Stephenson Chemical Co.
MWCNTs were purchased from Bayer Material Science. These MWCNTs feature diameters in the 2-20 nm range, and lengths between 1-10 µm with 95% purity. The self-healing 5E2N monomer was made from a mixture of endo and exo, 99%, containing 100-200 ppm Butylated hydroxytoluene (BHT). The 2 nd generation Hoveyda-Grubbs (HG2) catalyst, the encapsulating shell materials melamine, formaldehyde, urea, as well as the emulsifying agent polyvinyl alcohol (PVA) and the surfactant sodium lauryl sulphate (SLS) were purchased from Millipore Sigma, Canada. All chemicals were used as received. We also used a conductive commercial silver/ epoxy ECCOBOND 56 C (Henkel Loctite Ablestik 56 C)/Catalyst 9, purchased from Ellsworth Adhesives, as electrodes for electrical conductivity measurements. characterization. The size and surface morphology of the microcapsules implemented in this work were investigated at various magnifications using an Optical Microscope (OM), a field emission JEOL JSM7600F electron microscope (FESEM) at an acceleration voltage of 5 kV, and a JEOL JEM-2100F transmission electron microscope (TEM), operating at 200 kV, respectively. The measurement of the microcapsules average diameter was obtained using the image analysis freeware: ImageJ.
Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) analysis were performed to determine the thermal stability of the microcapsules. For the TGA analysis, the microcapsule samples were heated from 25 to 700 °C under helium environment, using a heating rate of 10 °C/min thanks to a TGA Q500 instrument. For the DSC analysis, the samples were heated from 0 to 100 °C (under nitrogen environment) at a heating rate of 10 °C/min using a Mettler Toledo DSC system. All the data acquired from TGA and DSC testing were analyzed using the 'TA Universal Analysis' software. A Leica Optical Systems Series DM LM Raman spectrometer was employed to monitor in situ the ROMP reaction of 5E2N initiated by the HG2 catalyst. This apparatus was equipped with a thermoelectrically cooled charge coupled device (CCD) detector and a 600 lines/mm diffraction grating to collect data over a spectral range of 300-3300 cm −1 . A 785 nm laser line was used for optical excitation whose intensity of 0.5 mW was set to prevent sample damaging during the measurements. The Raman characterizations were performed using a confocal Invia Renishaw RM 3000 spectrometer, equipped with a digital camera and a 50× objective lens of 0.75 numerical aperture (NA) to study and compare the spectral signatures of microcapsules with and without CNTs.
Recovery of electrical and mechanical properties. The mechanical healing efficiency of the polymer composites are determined by measuring the recovery of fracture toughness [6][7][8][9]11,12 of the damaged samples according to the test protocol established by the research group of White et al. 8 . Cured epoxy samples of tapered double-cantiliver beam (TDCB) fracture geometry, incorporated into the medium with CNT/5E2N microcapsules and HG2 catalysts are prepared and tested under mode I fracture protocol 21 . The four studied samples were manufactured with 10 wt.% of microcapsules of 70 μm average diameter and 1 wt.% of HG2 catalyst. With this amount of microcapsules, the weight percent of CNT in the whole epoxy matrix would be about 0.008 wt.% (See the calculation detailed in the supplementary information section). Reference neat TDCB epoxy samples, containing no microcapsules and catalysts, as well as epoxy samples incorporated with catalyst and microcapsules containing 5E2N only were also manufactured and tested for comparison purpose. All the studied materials were first loaded until their complete failure, to be then allowed to heal for 24 hours at RT, and reloaded for evaluating the healing efficiency.
During these experiments, the mechanical healing efficiency, η h m , , is determined as 21 is the fracture toughness of the healed specimen, and P C is the corresponding critical failure loads of the virgin and healed specimens. The geometry of the samples are detailed in the supplementary information section.
The self-healing ability of microcapsules with and without CNTs to restore the electrical conductivity of silver/epoxy polymer containing HG2 catalyst were also fabricated and tested. Detailed information regarding the preparation of these samples, as well as the experimental tests and analyses that have been conducted are described in the supplementary information section.

Fabrication of CNT/5E2N microcapsules. Microcapsules containing MWCNTs suspended in 5E2N
were successfully synthesized by in situ polymerization following the process developed in ref. 27 . Controlled dispersion-emulsification-polymerization-recuperation steps were carried out to enable the microencapsulation of CNT/5E2N suspension in PMUF shells. The MWCNTs were first dispersed into the 5E2N monomer using sonication, where the power and duration were the main controlling factors in obtaining a stable CNT/5E2N suspension. Figure A1 in the appendix shows that the CNT/5E2N suspension remains stable even after one week. During emulsification, shell materials (melamine, formaldehyde and urea), core material (CNT/5E2N suspension), emulsifier (PVA) and surfactant (SLS) were mixed in an aqueous medium. Both the stirring speed and the concentration of emulsifier and surfactant were found to be critical to produce oil-in-water emulsions with the steady presence of micro droplets of the core material. During the polymerization step, the emulsion was heated to the reaction temperature of 86 °C at a heating rate of 2 °C/min, according to the process presented in ref. 27 .
The temperature was found to strongly influence the formation of urea formaldehyde (UF) micro particles, as well as the thickness, roughness and porosity of the microcapsule shells 55 . The reaction was then kept running at this temperature for about five hours. During the collection of the microcapsules, the emulsion was first cooled down, rinsed several times with deionized water and acetone, and filtered. The synthesized microcapsules were then dried in air and stored in glass vials before and between each characterization. The combination of the main fabrication parameters that resulted in the successful microencapsulation of CNT/5E2N is presented in Table 1. Figure 2a shows a typical optical microscopy image of the individually separated spherical free-flowing microcapsules. In Fig. 2b, the white arrows point CNTs, which have been released within the liquid core upon the crushing of few milligrams of microcapsules. Reference microcapsules that contain only 5E2N (i.e. without CNTs) do not show such features. Further, the liquid core content of the microcapsules is isolated and the released CNTs are seen under SEM, as shown in the inset of Fig. 2b. Figure 3(a-d) shows representative SEM micrographs obtained at various magnifications: x50, x30000, x2300 and x100000, respectively. Individually separated spherical microcapsules are shown in Fig. 3a, with core-shell structure displayed in Fig. 3b,c. The average shell thickness is found to be around 550 nm. This shell thickness of the microcapsules is sufficiently robust to survive handling and manufacturing of self-healing polymer composites 17 . The outer shells of the microcapsule walls are observed to be textured, as shown in Fig. 3c, with nanoparticles deposited onto its surface during the in situ polymerization process 17 . The white structures observed in Fig. 3d and pointed by arrows are CNTs, whose metallic properties make them brighter than the background contrast under electron beam exposure 56 . Figure 4 displays a typical size-distribution histogram of the microcapsules produced at 700 rpm stirring speed. Their diameters vary between 40 and 120 µm, around an average value of 70 µm. Both the dimensions and the size-distribution of the microcapsules strongly depend on the stirring speed 17 that was selected during the synthesis process.

Microscopic investigations.
The uneven sizes of the microcapsules may affect the reproducibilty of the self-healing efficiency especially at the nanoscale, because the core material could be not uniformly released. To minimize this effect, microcapsules of desired size-range can be sort using micro-sieves and their concentration in the composite can be increased. www.nature.com/scientificreports www.nature.com/scientificreports/ For TEM observations, we extracted the liquid released from few milligrams of crushed microcapsules and dissolved it into acetone and methanol. A tiny Cu grid with carbon film was then immersed into the solution and dried before placing it into the TEM. The TEM micrographs presented in Fig. 5 show the presence of CNTs in the liquid core extracted from the crushed microcapsules. These images thus confirm that the processing steps of the synthesis of microcapsules did not cause severe breakage to the CNTs and maintain their original high aspect ratios. Such features are very important for maximizing the benefits of incorporating CNTs/5E2N into the core of the microcapsules, especially in terms of their electrical properties. thermal stability analysis. TGA was carried out separately for reference microcapsules containing only 5E2N and microcapsules containing CNT/5E2N suspension. During these thermogravimetric experiments, the weight-loss of the samples was measured while the temperature was increased, as reported on Fig. 6a.     www.nature.com/scientificreports www.nature.com/scientificreports/ with the sudden burst of the microcapsule shell. This rupture is caused by the buildup of the vapor pressure inside the microcapsules. The tearing of the microcapsule shells releases the vaporized core materials in the TGA cell, thus resulting in rapid weight loss. For CNT/5E2N microcapsules, this phenomenon also occurs at lower temperatures, around ~220 °C.
The dispersion of CNTs within the liquid 5E2N was found to affect the thermal stability of the medium. This effect is confirmed by the DSC analysis presented in Fig. 6b, where drops of liquid 5E2N and 0.1 wt. % CNT/5E2N suspension were heated separately from 0 to 100 °C. The difference in the slopes of the two lines (Fig. 6b) indicates the difference in the thermal properties for the 5E2N/CNTs, which possibly causes a quicker build-up of the vapor pressure inside the microcapsules. The presence of 0.1 wt. % CNTs inside the 5E2N liquid slightly decreases the maximum temperature to which the microcapsules can be heated before breaking. Moreover, even if this limit is somehow lower than that of the microcapsules containing only 5E2N, the thermal stability of about 220 °C makes the CNT/5E2N microcapsules still suitable for use in epoxy polymer composites, which may require curing at elevated temperatures (up to 180 °C) for some applications 37 . On the other hand, the operation temperatures in satellites and other space vehicles do not usually exceed 125 °C [58][59][60] . This suggests that the 5E2N/CNT microcapsules fabricated in our work are stable and operational in the thermal environment of an orbiting spacecraft. Figure 7 shows the Raman spectrum of the core liquid extracted from the reference microcapsules and the CNT/5E2N microcapsules, before (Fig. 7a) and after the polymerization process (Fig. 7b). The Raman peak observed around 1350 cm −1 (Fig. 7) corresponds to the spectral signature of the disorder-induced mode 'D-band' of CNTs. This mode is related to the presence of disorder in sp 2 -hybridized carbon systems, whose out-of-plane vibration is more prominent for MWCNTs 61 . The phonon peak at 1582 cm −1 refers to the 'G-band' , which is connected to tangential in-plane vibrations of sp 2 -bonded carbon atoms [61][62][63][64][65][66] .

Raman analysis.
These two representative Raman signatures confirm the presence of MWCNTs inside the CNT/5E2N microcapsules. None of the D band and G band peaks have been observed in the reference microcapsules, where the Raman phonon at 1568 cm −1 is attributed to the C=C stretching vibration of the norbornene ring of 5E2N 65,67 . Due to the addition of the catalyst to the core liquid of the microcapsules, another Raman peak appears around 1665.3 cm −1 (as indicated within the ellipse in Fig. 7b). As previously reported, this peak confirms the formation of poly (5E2N) after ROMP polymerization reaction for both types of microcapsules 10 .

Recovery of fracture toughness. The representative load-displacement curves of TDCB epoxy samples
where pure 5E2N and CNT/5E2N microcapsules have been incorporated inside the medium with HG2 catalysts are shown in Fig. 8 for mode I fracture tests.
Our measurements show that during the first loading, often named as 'virgin loading' in the literature [16][17][18][19][20][21][22] , the TDCB sample undertakes increasing load as the displacement is increased. As the load approaches to the peak load (around ~100 N in Fig. 8), the mode I crack is generated in the sample and propagates along the centreline (Fig. A2 in supplementary section) through the entire length of the sample leading to its complete failure. At this stage, the TDCB sample is separated into two parts and its load carrying capacity drops to zero. The two halves of the damaged TDCB sample are then put into contact for 24 hours during which the self-healing reaction occurs. After this period, the two halves of the TDCB sample are observed to be glued together and the reassembled sample is loaded for a second time (labelled as 'healed' in Fig. 8) similarly to the first loading, in order to determine its load carrying capacity after healing. It is to be noted that, the reference neat TDCB epoxy samples that do not contain any microcapsules and catalysts, could not carry any further load after their complete failure during the first loading. As expected for these materials, the two halves of the TDCB reference samples remained separated after 24 h and could not even be reloaded as no self-healing occurred.
The comparison between the load-displacement curves of the virgin and healed samples reveals that the healed samples containing CNT/5E2N microcapsules are tougher, so that they can withstand more displacement before their final failures. This feature is consistent with highly crosslinked epoxy, which is naturally rigid and www.nature.com/scientificreports www.nature.com/scientificreports/ brittle 68 . When a crack is once again initiated inside the healed samples, its propagation is less uniform than within virgin epoxy samples due to the presence of healed CNT/poly (5E2N) nanocomposites inside the medium. Consequently, the load-displacement curves of the healed CNT/5E2N samples shown in Fig. 8 do not increase continuously with the displacements. This also favours the occurrence of a stick-slip fracture behavior 69 , characterized by a repetitive unstable propagation of the crack followed by a period of crack stoppage through the healed materials. These Stick-slip fracture behaviours in adhesives under mode I fracture conditions have already been observed in ref. 70 , where the substrate roughness is believed to play an important role at the microscopic level. In the adhesive layer, the bondline discontinuities 71 , even defects 72 result in the fluctuation of fracture energy that can lead to stick-slip fractures. According to ref. 73 , the non-linear behaviour of the load-displacement curve at the end of the rising stage originates from the crack formation process and changes in the shape of the crack front. At the microscale, these cracks were found to propagate in a zigzag way through the self-healed adhesive 74 , which can be regarded as low-grade crack deviations that also helps increase the fracture loads.
Last but not least, the inclusion of CNT fillers into the cracks may also affect its propagation 75 . To further verify this effect, the fracture surface of the TDCB samples has been examined by SEM. A representative SEM micrograph of the fracture surface of the TDCB samples is presented in Fig. 9, showing distinct poly (5E2N) film and broken microcapsules that have released their core content CNTs (shown in the inset).
In addition to confirming that the self-healing ROMP reaction is activated at the fracture surface where poly (5E2N) films are produced to glue and heal the crack, this figure also shows the presence of tails behind the CNT aggregates, as pointed by arrows at many different locations inside the poly (5E2N) layer. These features of crack are attributed to CNT aggregates, whose effect on the composite toughening can be associated with a crack pinning mechanism 20,68 . Such a phenomenon is caused by the bifurcation of the crack propagation path in the presence of the filler surface, followed by the subsequent meeting of the previously separated crack fronts 68 .
The dispersion of CNTs in monomer followed by in situ polymerization favours the formation of CNT/ Polymer nanocomposites. This results in a stronger and more active interface between the CNTs and polymer, which is the key for improving the structural performance of these composites 76 . According to ref. 25 , the incorporation of CNT fillers into 5E2N monomer may act as a cross-linking network in poly (5E2N) 25 , leading to a significant increase of its mechanical properties such as hardness, strength and stiffness.
Although the wt. % of CNTs in the whole epoxy matrix is only about 0.008 wt. %, the microcapsules enable the release of a well-dispersed suspension of CNT/5E2N into the crack sites. This ensures that a sufficient amount of CNTs come into contact with 5E2N. Their subsequent in situ polymerization produces strong poly (5E2N)/nanotube interfaces as evidenced in Fig. 9. If CNTs are pre-dispersed into the epoxy matrix and 5E2N-only microcapsules are used instead, they might not be liberated at the crack site and they are less likely to form an effective dispersion with the released 5E2N resulting in weaker poly(5E2N)/nanotube interfaces.
The average healing efficiencies achieved by the three types of tested samples, as calculated from Eq. (1), are compared in Fig. 10, showing that, in absence of any microcapsule and catalyst, the reference neat epoxy samples could not recover any fracture toughness. On the other hand, the samples containing catalyst and 5E2N-only microcapsules recover about 47% of their nominal fracture toughness, and the samples containing catalyst and CNT/5E2N microcapsules, about 80%. Such results prove indeniably that the presence of CNTs inside 5E2N significanlty improves the restoration efficiency of the structural integrity of self-healing composites, in agreement with the data reported in ref. 25 .
Rule et al. 16 , who studied the effects of size and wt.% of microcapsules containing DCPD on the healing performance of epoxy showed that the healing efficiency can be optimized by choosing different combinations of microcapsule size and concentration for given sizes of cracks to be healed. According to the authors 16 , a high concentration (e.g. 20 wt%) of large (up to 400 μm) microcapsules can deliver more healing agents into the cracks for better healing. However, high concentration of larger microcapsules might negatively affect the virgin properties of the epoxy 20 , as well as raise some manufacturing issues related to the increase of the resin viscosity, and difficulties in achieving uniform dispersion. From our results, we recommend to increase the concentration of CNTs in the microcapsule core to release more CNTs into the crack and produce more poly (5E2N)/CNT interfaces, so that the healed material will be toughened through the crack-pinning mechanism occuring in more sites, as shown in Fig. 9  www.nature.com/scientificreports www.nature.com/scientificreports/  The Fig. 11 demonstrates two types of electrically conductive samples whose conductive paths were first disrupted by introducing a damage/cut (cutting with razor blade) to the conductive path so that no electrical current could pass through it. The samples were then left untouched for 24 hours to allow for self-healing. The samples which did not contain any microcapsules or contain 5E2N-only microcapsules, could not recover the conductivity. The samples that contain the CNT/5E2N microcapsules are found to be able to recover their electrical conductivity, up to 82% of its nominal value as shown in Fig. 11. Despite possible variations due to the presence of microcapsules non-uniform in size, no significant change in electrical properties is recorded in the studied samples after healing. This may result from the very low percolation threshold for electrical conductivity inside the CNT/polymer composites 76 .
This makes the incorporation of CNTs into 5E2N monomer relevant to restore the electrical conductivity and/ or connections inside damaged electronic circuits. The microcapsules containing such CNTs can thus permit a better repairing/restoring of key electronic components and/or optoelectronic devices. Aïssa et al. 77 demonstrated that the performance of a flexible fluidic antenna incorporated with conductive nanocomposite materials are greatly influenced by the electrical conductivity of the nanocomposites.
In addition, the mechanical properties of the nanocomposite materials are found to contribute significantly to the overall elasticity of the antenna structure, which can be deformed and/or bent repeatedly without losing its performance. The CNT/5E2N microcapsules offer the possibility of restoring significantly both the electrical conductivity and the structural integrity of systems that may degrade due to aging effects and damaging events occurring in harsh space environments. Since the 5E2N healing agent remains active below 0 °C 10 and the presence of CNTs inside encapsulated 5E2N does not usually affect the kinetics of the ROMP reaction, as reported in the Fig. A5 of the ESI section for CNT/5E2N with Grubbs catalyst, we infer that the improvement of self-healing abilities due to the presence of CNTs is relevant for applications in space environment. This provides an alternative solution to extend both their lifetime and their reliability, even for operation at low temperatures and/or temperature gradients similar to space environment conditions.

conclusions and perspectives
MWCNTs suspended in 5E2N self-healing monomer capable of undergoing ROMP reaction were successfully encapsulated using an in situ polymerization technique. An experimental trial and error procedure was used to identify suitable process parameters including the wt. % of CNT, both the power and duration of sonication for dispersion, as well as the concentration of emulsifying agent and surfactant. After successful encapsulation of 5E2N/CNT suspensions, the average dimension, size-distribution, shell-wall topology and thickness of the microcapsules were characterized by microscopic observations. Optical, SEM and TEM images show the formation of microcapsules, as well as the presence of CNTs inside their core and the release of their content upon crushing. As confirmed by Raman analysis, the ROMP reaction is observed for pure 5E2N and mixed CNT/5E2N microcapsules. The structural integrity of these CNT/5E2N microcapsules was investigated up to 600 °C, showing that such system can resist to external temperatures of 220-250 °C, which makes them suitable to be used in advanced polymer composites for aerospace. After testing, 80% of the fracture toughness and 82% of the electrical conductivity are found to be autonomically restored inside the polymer composite containing 5E2N/CNT self-healing microcapsules. These new 5E2N/CNT PMUF shells offer unique self-healing solutions to devices/structures that require restoration of either or both electrical and mechanical properties with efficient functioning, even at low temperatures. Their ease of being integrated into a great number of composite materials www.nature.com/scientificreports www.nature.com/scientificreports/ makes them suitable to protect and extend the lifetime of telecommunication equipment and sensors exposed to accidental events and collisions with small debris, as well as critical ageing effects that can cause their irreversible damage and failure. The use of encapsulated 5E2N/CNT could provide a significant added-value to commercial products and devices requiring long operation periods inside facilities or environments where they cannot be repair and/or replaced by new ones. Example of such devices/structures are flexible fluidic antennas, where both electrical conductivity and mechanical properties ensure their efficient functioning, to be used in wireless sensing or monitoring radio systems, switches, radio frequency identification (RFID) tags, conformal circuits for health monitoring or in other military and space applications.